Journal of Personal Finance Vol 13 Issue 2 by IARFC

Volume 13 Issue 2 2014 www.journalofpersonalfinance.com

Journal of Personal Finance

Techniques, Strategies and Research for Consumers, Educators and Professional Financial Consultants

IARFC INTERNATIONAL ASSOCIATION OF REGISTERED FINANCIAL CONSULTANTS

2014 National Financial Plan Competition

and the Winner Is... Bryant University

Bryan t L-R Ja Universitymie P epin, presents “f Laure i n Fay ctional” B r ne an d Kyle ewster fam ily pre Creed on senta ti

on:

First Place — Bryant Universitty Smithfield RI, team members Kyle Creedon, Lauren Fayne, Jamie Pepin and their instructor Mara Derderian

The results are in and the winner is Bryant University! So ended the 2014 Financial Plan Competition 2 day event in Las Vegas. On hand were the student finalists, their advisors, family, IARFC members as judges and professional presenters. The previous day’s activities of the Impact Branding Workshop and Reception gave the students an opportunity to network and get to know the financial professionals that were part of the event. “Coach Pete” D’Arruda, Les Anderson and Ed Morrow highlighted why branding is such an important part of the Financial Advisor’s public persona and how to build it.

The Results First place – Bryant University Smithfield RI, team members Kyle Creedon, Lauren Fayne, Jamie Pepin and their instructor Mara Derderian. Second Place – Shepherd University Shepherdstown WV, team members Bonnie Baily, Akhtar Khan, Avery Mendzela and their professor Nicolas Pologeorgis. Third Place – Bowling Green University Bowling Green OH, team members Dana Kaufman, Elizabeth Kevorkian, Becca Smoody and their professor Matthew Garrow. Contact: 800.532.9060

info@iarfc.org

www.IARFC.org

Volume 13, Issue 2

Journal of Personal Finance

Volume 13, Issue 2 2014 The Official Journal of the International Association of Registered Financial Consultants

ÂŠ2014, IARFC. All rights of reproduction in any form reserved.

Journal of Personal Finance

CONTENTS Editor’s Notes.........................................................................................................................................................................................8 Portfolio Size Matters.........................................................................................................................................................................9 Gordon Irlam, Independent Software Developer In contrast to target date funds that vary asset allocation by age alone, it is important to take into account both the client’s age and the client’s portfolio size relative to spending goals when determining an optimal asset allocation. Stochastic Dynamic Programming (SDP) is a mathematical optimization technique that can be used to determine optimal dynamic adjustments to asset allocation in response to evolving portfolio wealth and time horizons. Using SDP, portfolio size appears at least as important to asset allocation decisions as age. For a 25 year old, asset allocation using SDP requires approximately 20-34% less resources than traditional target date asset allocation approaches. Financial planners can provide much value to their clients by incorporating portfolio size into a dynamic asset allocation strategy.

Parental and Early Influences on Expectations of Financial Planning for Retirement.........................................17 Janet L. Koposko, M.S., Doctoral Candidate, Oklahoma State University Douglas A. Hershey, Ph.D., Professor of Psychology, Oklahoma State University This investigation was designed to test a theoretically-grounded model of the psychomotivational dimensions that underlie retirement planning. In developing the hypothesized model, special consideration was given to positive early influences on development that could potentially impact other dimensions known to predict successful planning practices. Participants were 722 college students who reported the extent of childhood personal finance lessons learned, their retirement goal clarity and knowledge of financial planning, and expectations of future planning and anticipated satisfaction with life in retirement. As hypothesized, two measures of early financial influences were predictive of other variables known to underlie the retirement planning decision-making process, and one’s vision of satisfaction in retirement. Results and implications are discussed in terms of the way in which motivational forces, particularly those that occur early in life, contribute to perceptions of future planning efforts.

The Role of Trans-Generational Financial Knowledge and Self-Reported Financial Literacy on Borrowing Practices and Debt Accumulation of College Students................................................................................28 Chad Smith, Ph.D., Associate Professor, Clarion University of Pennsylvania Gustavo Barboza, Ph.D., Professor, Clarion University of Pennsylvania This paper studies the effects of trans-generational financial knowledge, self-reported financial knowledge, academic performance, and overall financial literacy on financial management practices, using a sample of 380 college students. Exploratory estimates using a series of ordered Probit models indicate that academic status and self-reported overconfidence on financial knowledge relate positively to the amount of debt a student carries. More interesting, our estimates provide robust support to the hypothesis that a transgenerational financial knowledge effect from parents to students plays a major role in reducing the amount of financial burden students assume, both in the form of student loans and credit card balance. In addition, students that maintain a good credit card history, as reflected by high repayment rates, are more likely to hold lower debt amounts than otherwise. A high debt level implies that students are living beyond their means and consequently developing unhealthy financial management practices. Our paper provides evidence in favor to the hypothesis that early financial education is a means to reduce or maintain low levels of indebtedness. Our empirical estimates also point out to the presence of a strong overconfidence effect, as reflected by unrealistically high self-reported financial knowledge, leading students to an incorrect decision making process in favor to holding more debt. We argue that lack of personal financial literacy is at the core of high debt accumulation.

Volume 13, Issue 2

A Better Systematic Withdrawal Strategy--The Actuarial Approach ..........................................................................51 Ken Steiner, Fellow of the Society of Actuaries, Retired Retirees generally have at least two potentially conflicting financial goals: (i) spend enough each year to maintain a certain standard of living throughout retirement and (ii) not spend so much that accumulated savings run out prior to death. Corollaries to these two primary goals include: (i) having relatively predictable and stable inflation-adjusted spendable income from year to year (ii) having spending flexibility to meet unforeseen expenses, (iii) maximizing the general level of spendable income and (iv) not leaving too much unspent at death. An optimal retirement spending strategy should address each of these goals to some degree, depending on the preferences of the individual retiree.

A Rule-of-Thumb Approximation for Time Value of Money Calculations..........................................................57 David N. Swingler, Ph.D., Professor of Engineering, St. Mary’s University, Halifax, N.S., Canada A simple rule-of-thumb is presented for the classic financial calculations centered on the present value of a series of equal future payments. It is demonstrated that it is a useful addition to the armamentaria of engineering students engaged in engineering economics and/or finance courses.

Race, Trust, and Retirement Decisions.........................................................................................................................62 Terrance K. Martin Jr., Ph.D., Assistant Professor, University of Texas-Pan American Michael Finke, Ph.D., Professor and Director of Retirement Planning and Living, Texas Tech University Philip Gibson, Ph.D., Assistant Professor, Winthrop University Using the 2008 National Longitudinal Survey of Youth, this study investigates whether racial differences in trust can explain the decision to consult a financial planner and the variation in accumulated retirement wealth. Blacks and Hispanics are more likely to report having low trust compared to non-black, non-Hispanic respondents. The results provide evidence that low trust has a greater impact among blacks than non-black, non-Hispanic respondents. Low trust has a weaker impact among Hispanics on the decision to consult a financial planner and on the accumulation of retirement wealth. There is no evidence that hiring a financial planner has a larger impact on retirement wealth among blacks or Hispanics than it does among non-black, non-Hispanic households.

Loss Aversion Under Cognitive Load............................................................................................................................72 Michael A. Guillemette, Ph.D., CFP®, Assistant Professor, University of Missouri Russell N. James III, Ph.D., J.D., CFP®, Professor and CH Foundation Chair, Texas Tech University Jeff Larsen, Ph.D., Associate Professor, University of Tennessee An experiment was conducted to explore whether loss aversion is altered when individuals are placed under a higher level of cognitive load. The coefficients of monetary loss aversion were measured for 30 participants under low and high cognitive load. Memorizations of differing spans of digits were used to manipulate cognitive load. Participants’ skin conductance was measured to quantify emotional responses to gains and losses. No statistically significant evidence was found that loss aversion, as measured by choice, is altered when individuals are placed under a higher level of cognitive load. Statistically significant evidence was found that a higher level of cognitive load temporarily reduces an individual’s emotional arousal to absolute and relative small dollar losses.

Journal of Personal Finance

CALL FOR PAPERS JOURNAL OF PERSONAL FINANCE (www.JournalofPersonalFinance.com)

OVERVIEW The new Journal of Personal Finance is seeking high quality submissions that add to the growing literature in personal finance. The editors are looking for original research that uncovers new insights—research that will have an impact on advice provided to individuals. The journal is committed to providing high quality article reviews in a single-reviewer format within 60 days of submission. Potential topics include: • Household portfolio choice • Retirement planning and income distribution • Individual financial decision-making • Household risk management • Life cycle consumption and asset allocation • Investment research relevant to individual portfolios • Household credit use • Professional financial advice and its regulation • Behavioral factors related to financial decisions • Financial education and literacy Please check the “Submission Guidelines” on the Journal’s website (www. JournalofPersonalFinance.com) for more details about submitting manuscripts for consideration.

CONTACT Wade Pfau and Joseph Tomlinson, Co-Editors Email: jpfeditor@gmail.com www.JournalofPersonalFinance.com

Volume 13, Issue 2

JOURNAL OF PERSONAL FINANCE VOLUME 13, ISSUE 2 2014 Co-Editors Wade Pfau, The American College Joseph Tomlinson, Tomlinson Financial Planning, LLC

Editorial Board Dale L. Domian, Ph.D., CFA, CFP™, York University Michael S. Finke, Ph.D., CFP™, RFC® Texas Tech Joseph W. Goetz, Ph.D., University of Georgia Clinton Gudmunson, Ph.D., Iowa State University Sherman Hanna, Ph.D., The Ohio State University George W. Haynes, Ph.D., Montana State University Douglas A. Hershey, Ph.D., Oklahoma State University Karen Eilers Lahey, Ph.D., The University of Akron Douglas Lamdin, Ph.D., University of Maryland Baltimore County Jean M. Lown, Ph.D., Utah State University Angela C. Lyons, Ph.D., University of Illinois Carolyn McClanahan, MD, CFP™, Life Planning Partners Yoko Mimura, Ph.D., California State University, Northridge Robert W. Moreschi, Ph.D., RFC®, Virginia Military Institute Ed Morrow, CLU, ChFC, RFC®, IARFC David Nanigian, Ph.D., The American College Barbara M. O’Neill, Ph.D., CFP™, CRPC, CHC, CFCS, AFCPE, Rutgers Rosilyn Overton, Ph.D., CFP™, RFC®, New Jersey City University Jing Jian Xioa, Ph.D., University of Rhode Island Rui Yao, Ph.D., CFP™, University of Missouri Tansel Yilmazer, Ph.D., CFP™, The Ohio State University Yoonkyung Yuh, Ewha Womans University Seoul, Korea Mailing Address:

IARFC Journal of Personal Finance The Financial Planning Building 2507 North Verity Parkway Middletown, OH 45042-0506 Postmaster: Send address changes to IARFC, Journal of Personal Finance, The Financial Planning Building, 2507 North Verity Parkway, Middletown, OH 45042-0506 Permissions: Requests for permission to make copies or to obtain copyright permissions should be directed to the Co-Editors. Certification Inquiries: Inquiries about or requests for information pertaining to the Registered Financial Consultant or Registered Financial Associate certifications should be made to IARFC, The Financial Planning Building, 2507 North Verity Parkway, Middletown, OH 45042-0506.

Disclaimer: The Journal of Personal Finance is intended to present timely, accurate, and authoritative information. The editorial staff of the Journal is not engaged in providing investment, legal, accounting, financial, retirement, or other financial planning advice or service. Before implementing any recommendation presented in this Journal readers are encouraged to consult with a competent professional. While the information, data analysis methodology, and author recommendations have been reviewed through a peer evaluation process, some material presented in the Journal may be affected by changes in tax laws, court findings, or future interpretations of rules and regulations. As such, the accuracy and completeness of information, data, and opinions provided in the Journal are in no way guaranteed. The Editor, Editorial Advisory Board, the Institute of Personal Financial Planning, and the Board of the International Association of Registered Financial Consultants specifically disclaim any personal, joint, or corporate (profit or nonprofit) liability for loss or risk incurred as a consequence of the content of the Journal.

General Editorial Policy: It is the editorial policy of this Journal to only publish content that is original, exclusive, and not previously copyrighted. Subscription requests should be addressed to: IARFC Journal of Personal Finance The Financial Planning Building 2507 North Verity Parkway Middletown, OH 45042 Info@iarfc.org 1-800-532-9060

Journal of Personal Finance

EDITORS’ NOTES

his issue begins with a paper by Gordon Irlam that applies the economists’ life-cycle finance approach to determining optimal asset allocations for retirement. The author demonstrates the inappropriateness of the common rule of thumb that stock allocations should be determined by age. He demonstrates that portfolio size also needs to be considered. Applying the life-cycle finance approach and the use of accompanying tools such as stochastic dynamic programming is gaining more attention as a research area, and it shows promise for developing practical applications. The second paper by Janet Koposko and Douglas Hershey deals with the impact of early life influences on planning for retirement many years later. The authors conduct a survey of college students who report the extent of childhood personal finance lessons learned, and the study relates this early experience to expectations of future planning and anticipated satisfaction with retirement. They find that early experiences are likely to carry even much later in life. The next paper by Chad Smith and Gustavo Barboza bears some similarity to the Koposko/Hershey paper, but focuses on the impact of early influences on how college students deal with current financial issues. They find that financial lessons imparted from parents to students can play a strong role in reducing the financial burdens students assume. They also find that overconfidence can play a role in leading students to take on too much debt. In the next paper, Ken Steiner proposes an actuarial approach to planning for taking withdrawals from savings to support retirement. His particular method bears similarities to the approach actuaries take in dealing with pension plans, and involves taking a fresh look at assets and liabilities each year,

and making changes to the spending plan as appropriate. He also suggests a smoothing technique to avoid too much disruption to spending plans. Next, we present a short paper by David Swingler that may appeal to those interested in financial math. He is an engineering professor, and he demonstrates the process he has gone through to develop a rule-of-thumb to apply to a common problem in finance math regarding the present value of a series of future payments. The paper by Terrance Martin, Michael Finke, and Philip Gibson deals with the important issue of how race and trust affect the decision to seek financial planning services and the accumulation of retirement wealth. The study reports differences between black and Hispanic households in terms of the impact of low trust on financial planning decisions. Finally, Michael Guillemette, Russell James, and Jeff Larsen provide us with a paper in the relatively new subject area of applying neuroscience to financial planning research. They report on experiments to test whether loss aversion is altered when subjects are placed under higher cognitive load, with more demands placed on mental processing. We will likely be seeing more neuroscience research in areas such as risk tolerance assessment. As new co-editors, we welcome the submission of research papers that uncover new insights in personal finance and show the potential to have an impact on the financial advice provided to individuals.

—Wade Pfau —Joe Tomlinson

Volume 13, Issue 2

Portfolio Size Matters Gordon Irlam, Independent Software Developer, Los Altos, California

In contrast to target date funds that vary asset allocation by age alone, it is important to take into account both the clientâ&#x20AC;&#x2122;s age and the clientâ&#x20AC;&#x2122;s portfolio size relative to spending goals when determining an optimal asset allocation. Stochastic Dynamic Programming (SDP) is a mathematical optimization technique that can be used to determine optimal dynamic adjustments to asset allocation in response to evolving portfolio wealth and time horizons. Using SDP, portfolio size appears at least as important to asset allocation decisions as age. For a 25 year old, asset allocation using SDP requires approximately 20-34% less resources than traditional target date asset allocation approaches. Financial planners can provide much value to their clients by incorporating portfolio size into a dynamic asset allocation strategy.

Journal of Personal Finance

Introduction Decisions about asset allocation—how much to allocate to risky investments like stocks versus safer investments like bonds—are arguably the most important decisions in financial planning. Those providing advice on asset allocation have recognized the importance of age as a key variable influencing allocation choices. However, portfolio size relative to spending goals can be shown to be equally important. I’ll briefly discuss various approaches that have been used in asset allocation research and then demonstrate an optimization approach, using stochastic dynamic programming to illustrate the importance of both age and portfolio size. Research on asset allocation has taken one of three main approaches. The first approach is based on the statistical properties of the asset classes such as means, variances, and covariances, and includes mean-variance optimization (MVO) (Markowitz (1952)), the Black-Litterman model (1992), and Sharpe’s gradient method (1987). Such techniques require specification of an investor’s tolerance for risk, which is difficult to judge. As well, because they are measured over single time periods, these techniques say little about the probabilities of portfolio success over long periods of times. Markowitz (1991) recognized this point, noting that his mean-variance optimization is meant for institutional investors, and in personal finance one must use multiple period simulations instead of single-period risk-adjusted wealth accumulation. The second approach involves simulating different static asset allocations over a longer period of time and then picking the best one. This approach was used in the Trinity study (Cooley, Hubbard, and Walz (1998)). These historical simulation or Monte Carlo simulation approaches only consider portfolios with a static stock/bond asset allocation, or they employ age dependent rules for asset allocation. Most existing works also only consider fixed life expectancies, which as we will see, limits the applicability of the results. And finally, at least for the Trinity study, only retirement portfolios are considered, not the full asset accumulation-depletion process over a lifetime. The final approach treats asset allocation as a mathematical optimization problem, and solves it analytically or numerically using stochastic programming or stochastic dynamic programming (SDP) techniques (Bellman (1953), Mulvey and Vladimirou (1989)). I take a numerical SDP approach as described in more detail later. My approach somewhat resembles the simulation approach. It performs a series of single time-step simulations. However, unlike the simulation approach, it does not perform these simulations with predetermined asset allocations regardless of the simulated circumstances. Rather, the approach works backward by determining optimal strategies at each stage of life so that earlier decisions will be optimized after accounting for what later optimal behaviors will be. An important point is that both MVO and Monte-Carlo simulation ignore portfolio size as a key variable. In the first

approach, investor tolerance for risk might to some extent provide a proxy for portfolio size, but this is very fuzzy. In the second approach, portfolio size is calculated, but then ignored when analyzing the results. Shiller (2005a) was the first to suggest that asset allocation might be portfolio size dependent. This suggestion was lent weight by the analysis of Basu and Drew (2009), who found that portfolio size effects exist during the accumulation phase, but they did not attempt to vary asset allocation as a function of the portfolio size. Blanchett and Kasten (2010) is a related study which argued that using dynamic asset allocation responding to the funded status of a retirement plan (present value of assets divided by present value of liabilities) can improve retiree outcomes. In this paper I explore asset allocation for fixed withdrawal rates schemes. Schemes such as the 4% rule are widely popular, and hence worthy of study. Despite their popularity however, we need to begin by following Scott, Sharpe, and Watson (2009) and sounding a note of caution against them. A scheme that does not respond to the performance of the portfolio is likely to underperform one that is responsive to portfolio performance.

Methods Asset allocator implementation I have developed an asset allocator called Optimal Portfolio Algorithm (OPAL). This asset allocator does not generate returns from a probability distribution. Instead, it makes use of returns drawn from the annual historical record. The asset allocator computes everything in terms of Relative Portfolio Size (RPS), which is the portfolio size relative to the annual retirement withdrawal amount. For those used to thinking about withdrawal rates, during retirement the RPS is simply 100 divided by the withdrawal rate. For example, if the portfolio size is $500,000 and the annual retirement withdrawal amount is $25,000, the withdrawal rate is 5% and the RPS is 20. For the present study we limit ourselves to two asset classes, stocks and bonds. We consider them in increments of 1%, namely 0/100, 1/99, through 100/0 stocks/bonds. At first, computing the optimal asset allocation might seem computationally intractable. With 101 asset allocation possibilities, over a period of 70 years, there are 10170 different asset allocation combinations to be evaluated. The key to breaking this intractability is the realization that if optimal asset allocations for time t+1 are known, then computing the optimal asset allocation for time t is as simple as trying all of the asset allocation choices for time t, followed by making use of the already known asset allocation decisions for time t+1. The asset allocator computes the optimal asset allocations in reverse time order by starting at the final year, and then working back to the initial year. For the final year, and for each given RPS value, computing the optimal asset allocation is simple. We try each asset allocation possibility, perform a set of single time step simulations corresponding to the annual return historical record,

Volume 13, Issue 2

compute the portfolio success probability, and decide on the best asset allocation for maximizing success. Then we move back to the prior year. Again for each given RPS value, and for each asset allocation possibility, we perform a set of single time step simulations, but this time we use the resulting RPS values from these single time step simulations to look up the portfolio success probabilities associated with future times. We then combine these success probabilities and pick the best asset allocation. In this way we build up the optimal asset allocation and success probability working backward for every age and every RPS value. It is computationally intractable to pre-compute the optimal asset allocations and success probabilities associated with all RPS values. Instead we bucketize RPS, computing the results just once for all RPS values lying within a given bucket. A bucket size of 0.001 RPS units gives good results for the problem at hand. To make our description more concrete suppose that we know the optimal asset allocations and success probabilities for all bucketized RPS values for age 50, and we are currently computing the optimal asset allocation and success probability for age 49, RPS 10.000. For each asset allocation possibility, such as 35/65 stocks/bonds, we perform a set of single time set simulations, based on the historical record. For instance in one year the (stock, bond) returns might have been (+22%, -3%) and in another (-15%, -8%) giving overall returns for the asset allocation being considered of 5.75% and -10.45%. If we are currently saving 0.5 RPS units at the end of each year towards retirement, this carries forward from age 49 to RPS values at age 50 of 11.075 and 9.455 respectively. Using these RPS values we retrieve the bucketized success probabilities for age 50 of, say, 99.3% and 98.4%. We combine these two success probabilities along with the success probabilities of all the other returns values in the historical record. This gives us the success probability for 35/65 at age 49. We repeat for all the other asset allocation choices, and pick the best asset allocation. We then repeat for all the other RPS values other than 10.000, and finally we move back to computing the optimal asset allocations and success probabilities for age 48. Measures of portfolio performance It is common to report on portfolio performance in terms of not running out of money before death. But there is a big difference between running out of money a year before you die and ten years. The magnitude of failure is more important than the probability of failure. We capture this idea through the use of time weighting, which measures the number of years in which a client remains solvent divided by the total number of years lived. We attempt to maximize and report performance as time weighted portfolio success. Variable length actuarial life expectancy Asset allocation for retirement planning has traditionally been performed using fixed life expectancies. Unfortunately, this isnâ&#x20AC;&#x2122;t very realistic. To get a handle on the impact of life expectancy on portfolio success, we follow the approach of Finke, Pfau, and Williams

(2012) and use actuarial data directly within the asset allocator. Death while remaining solvent is considered a portfolio success. The asset allocator combines the probabilities of death with the year by year probabilities of solvency to compute the overall probability of success. This weighted average is performed as a calculation without the use of randomly drawn ages of death or other Monte Carlo techniques.

Input data Stock and bond returns The primary stock and bond market returns data used by the asset allocator are provided by Shiller (2005b, updated). They are for US stock market returns and constant maturity 10-year US treasury notes for the period 1926-2011. The Shiller data set extends back to 1871, but it is only the later part we make use of, since the later part is more indicative of the present era. Stock market returns are adjusted to include dividends. Bond market returns are adjusted to include changes in market price. Returns data generation One of the problems when performing and analyzing asset allocations is the lack of historical returns input data. We have at best 140 years of reasonable quality annual data. This is sufficient for the OPAL asset allocator, which only cares about the returns distribution, not returns sequences. To evaluate how well the asset allocator would have performed in comparison to other asset allocation strategies, we will provide further comparisons based on realistic returns sequences. For this, we use the bootstrapping technique of picking fixed 20-year long blocks of data from the historical record at random and concatenating them together. No wrapping of data is performed, so the least weight is given to the returns data at the start and end of the historical period, which is also the case for historical simulation studies with overlapping data periods. Returns from the different asset classes for a given year are kept together. For each comparison we generate and analyze 100,000 synthetic returns sequences. Life expectancy data The US Government produces the US Life Tables (Arias (2011)) that report the probability of death by age. The asset allocation comparisons we perform are based on the US Life Tables for 2007 using the unisex combined male/female tables for all racial/ ethnic groups. To provide meaningful results near but below age 100, the life expectancy table is naively extended from age 100 to age 109 using the death rate given for age 99 (30% per year). Key simulation parameters All calculations and results are created in real, inflation-adjusted terms. A key parameter to be specified is the Relative Contribution Rate (RCR), which is the annual retirement savings contribution amount during the accumulation years divided by the annual withdrawal amount during retirement. It is a self-selected

Journal of Personal Finance

audience, but informal discussions on the Bogleheads Forum show people typically contributing twice as much or more, even after adjusting for inflation, than they were ten years earlier. This holds throughout their working years. The reasons for this vary, and examples to justify increased savings with age include promotions, kids leaving home, and paying off student debt and mortgages. To loosely capture this, we default to an RCR of 5% at age 25, increasing by 7% in real terms annually. For example, if the annual withdrawal amount during retirement is $50,000, this corresponds to saving $2,500/year at age 25 and $35,000/ year at age 64.

success probability map for OPAL with this scenario. As would be expected, the likelihood of success has a pyramid shape as assets are expected to be built up during the accumulation phase and depleted during the retirement phase. Figure 1 shows the success probabilities when using optimal asset allocations for each age and RPS level. Next, Figure 2 shows the optimal asset allocations computed by OPAL for each age and RPS level.

Readers may be more used to seeing savings rates as a percent of gross income, rather than our RCR measure, but speaking in terms of savings rates would add complications by needing to also define the lifetime income path and the withdrawal amount as a percentage of pre-retirement income. Further extending our example, if the client’s salary was $30,000 at age 25 and $100,000 at age 64, then the savings rate was 8.3% at age 25 and 35% at age 64. To avoid having to provide additional complications related to the client’s lifetime income path, this analysis is conducted solely in terms of the RCR measure. All simulation runs start at age 25. Accumulation normally lasts for 40 years, with retirement beginning at age 65. SDP can handle multiple asset classes, but the SDP analysis performed here makes use of just two asset classes, stocks and bonds, rebalanced annually to the target asset allocation. Taxes and portfolio management fees have not been included at this stage.

Results Optimal portfolio for a fixed life expectancy First, before providing analysis with survival probabilities for each age, let’s consider a fixed 70 year portfolio horizon, with 40 years until retirement, an initial RCR of 5% increasing by 7% annually, and 30 years in retirement. The OPAL program will optimize and report the time spent solvent. Figure 1 shows the

Figure 1. Success probabilities for OPAL using optimal asset allocations with a fixed life expectancy.

Figure 2. Optimal asset allocation computed by OPAL for a fixed life expectancy. Optimal asset allocation displays strong horizontal and vertical phenomena. In other words, asset allocation is dependent both on RPS and on age. This is important. Traditionally, asset allocation has been viewed as a function of age alone, as seen, for instance, with lifecycle or target date mutual funds. Very roughly, for an RPS below 5 (again, in retirement, this would imply a withdrawal rate above 20%), stocks dominate. This is easy to understand. A client might not have a very high chance of portfolio success, but the best chance is to gamble everything on stocks, for stocks produce a higher return and thus a better expected outcome than any of the alternatives. For example, consider an individual with $50,000, 10 years to live, and needing $20,000 a year to survive. Betting on either stocks or bonds the client is likely to fail, but betting on stocks on average will provide the client with a longer period during which the portfolio meets the client’s income needs. In the real world, for such circumstances, a better strategy than betting everything on stocks might be to advise the client to reduce their expenditures or gain additional income. Above an RPS of 15 (or below a withdrawal rate of 6.7%), a balanced bond-dominant portfolio makes the most sense. Success is highly likely, and it becomes more likely the more we can reduce the variance of returns in order to prevent a catastrophic outcome. Bonds should dominate, and for later ages if all that matters is the success probability, a 100% bond portfolio appears appropriate. Again, in such circumstances, an alternative strategy

Volume 13, Issue 2

might be to understand clearly about the clientâ&#x20AC;&#x2122;s spending and legacy goals, and then potentially advising to fund retirement spending with a bond ladder and devote the remainder to stocks. Most of the interesting action occurs between an RPS of 5 and 15. This includes the yellow region representing balanced portfolios. The case for a bond dominant portfolio for the asset rich is far more equivocal than the case for a stock dominant portfolio for the asset poor because the difference between the success probabilities for the asset rich as asset allocation is varied is far smaller than it is for the asset poor.

The need for personalized asset allocation that depends on assets, future income streams and saving potential, age, and retirement goals implies an important role and responsibility for the financial advisor.

Optimal portfolio for a variable length actuarial life expectancy Figure 3 shows the optimal asset allocation for a variable length actuarial life expectancy for an individual matching the US longevity distribution and maximizing time weighted success as the optimization goal. This moves the analysis away from the fixed age of death shown in Figure 2.

Figure 4. Sample asset allocation paths computed by OPAL for a variable length actuarial life expectancy. Optimal asset allocation involves frequent asset allocation changes. On a yearly basis, a change of 5-6% is typical, while over 5 year periods the allocation changes are typically in the range of 12-16%. These changes in asset allocation policy are significant, but not beyond the bounds of what a typical investor would likely feel comfortable making.

Figure 3. Optimal asset allocation for an individual with a variable length actuarial life expectancy.

Leaving an inheritance Up until this point our sole concern has been with avoiding portfolio failure. However, leaving a legacy in the form of an inheritance is often an important secondary concern. It is straightforward to mathematically include a term expressing the value of any inheritance provided. The key issue to be decided though

With variable life expectancies, we do not observe much difference (compared to Figure 2) in optimal asset allocations during the pre-retirement years as clients still have low mortality rates and long time horizons. However, we can observe significant changes in asset allocation for the post retirement period. At later ages, the region for balanced portfolios moves to higher RPS levels between 5 and 10, and the width of the balanced portfolio region becomes far wider than before. Knowing the precise age of death leads to a sharp, precisely calculated portfolio of all stocks or all bonds, and that sharpness is dulled with an unclear terminal date. The need for rebalancing Figure 4 shows sample paths for the optimal asset allocation over time using different simulation paths for portfolio returns. The frequent change in asset allocation implies the need for periodic portfolio checkups, particularly after significant market events.

Figure 5. Optimal asset allocation for an individual with a variable life expectancy and an interest in leaving an inheritance.

Journal of Personal Finance

is the weight to be placed upon the inheritance by the recipient(s) and also the importance to the client for providing the inheritance. We treat the inheritance as if it is to be divided by 5 parties who each value it the same as one year’s worth of client spending, or equivalently it goes to one party who values it over 5 years. A second parameter is how important leaving an inheritance is relative to the client’s own welfare. For this we use a value of 20%, indicating a genuine but not exceptionally strong interest in leaving an inheritance. Figures 5 show the optimal asset allocations for an individual with a variable life expectancy in the presence of an inheritance. The presence of an inheritance motive gives rise to two common investing maxims: stocks for the asset rich and stocks for the young. In these cases we can observe more aggressive allocations than seen in Figure 3. Without considering any inheritance motive, bonds for the asset rich rules the day since bonds are less volatile. Once an inheritance motive enters the picture, avoiding the very small chance of portfolio failure by using bonds takes a back seat to maximizing the value of any inheritance by including stocks. For the young, the risk taken by gambling on stocks is more than offset by the long time over which a successful gamble can pay off in the form of allowing for a bequest to be made. Sensitivity to scenario changes Additional testing was performed to explore the sensitivity of the results presented here to changing scenario assumptions. Table II summarizes the sensitivity of the optimal asset allocation map to various alterations in our variable life expectancy baseline case. The life expectancy of a couple is calculated as that of the last-surviving spouse, and there is no drop in income required when the first spouse dies. Several of these results may appear counter-intuitive, but remember that the asset allocation map shows the optimal action to take for a particular RPS, not what to do if it looks likely the client will fail to reach that RPS. And remember we are concerned not just with the client’s portfolio success, but also with the client leaving an inheritance. Pre-retirement, for “reduce retirement age” the client is likely to save less money in the future than they otherwise would have done. This means providing for the client’s own income security is more important than maximizing the value of any inheritance. This can be done by switching from a stock heavy allocation toward bonds. This appears as the broad thrust of the change in scenario, although it should be noted that there is a small part of the map where the RPS is not too small and not too large and where reducing the retirement age causes stock holdings to increase. Similar considerations apply to “reduce contribution rate” and “reduce contribution rate growth”. Evaluating some simple asset allocation rules Table III shows the improvement provided by SDP relative to a variety of different asset allocation rules at ages 25 and 65. The dataset for which the optimal asset allocation is computed and

change in scenario

change in optimal asset allocation

change in success probability

reduced equity returns

increase bonds

decreases

less frequent rebalancing

increase bonds

virtually unchanged

individual to couple

virtually unchanged

reduce retirement age

pre-retirement: primarily increase bonds

decreases

reduce contribution rate

pre-retirement: primarily increase bonds

decreases

reduce contribution rate growth

pre-retirement: primarily increase bonds

decreases

decreased value of leaving an inheritance

early years: increase bonds

virtually unchanged

share an inheritance more narrowly

early years: increase bonds

virtually unchanged

Table II. Sensitivity of the optimal asset allocation map to various changes in scenario. the rules are evaluated covers the years 1926-2011 with average real returns to equity scaled down to 5% per annum in order to match more conservative projections for the future. The initial RCR grows by 7% per annum. The metric under which the rules are evaluated includes a desire to leave an inheritance with the same parameter values as before. The simplified Vanguard Target Retirement and simplified Fidelity Freedom asset allocations are for the age-based target date funds of their respective companies using the reported asset allocations as of December 2010 and March 2011, respectively. They have been simplified by treating all stock holdings as US stocks, and all cash and bond holdings as 10 year notes. As such they are intended to be indicative of typical glide path rules, not an assessment of the precise performance of their respective funds. strategy

age 25

age 65

age in bonds

34%

15%

age minus 10 in bonds

26%

11%

simplified Vanguard Target Retirement glide path

20%

12%

simplified Fidelity Freedom glide path

23%

12%

Table III. Reduction in asset requirements for SDP compared to alternative asset allocation strategies.

Volume 13, Issue 2

Look at the first line of the table. Consider the time weighted success probability for age in bonds. The same time weighted success probability can be achieved by SDP for an individual age 25 with an RPS of 0 with a starting RCR that is 34% lower than the default 5% starting RCR used for age in bonds. Similarly, for age 65, the SDP RPS value is 15% lower than the benchmark RPS value of 20 used by age in bonds, but both achieve the same success probability. At age 25, SDP performs 20-34% better than the traditional age in bonds and glide path rules. At age 65 with an RPS of 20, SDP performs 11-15% better than the alternatives. These savings in asset requirements represent a major advantage of SDP.

Discussion The importance of portfolio size Asset allocation must be based on portfolio size in addition to age. The importance of portfolio size is highlighted by the asset allocation graphs showing very strong vertical non-uniformity in addition to horizontal non-uniformity. This represents something of a challenge for target date funds which are predicated on horizontal non-uniformity, but assume vertical uniformity in asset allocation decisions. Investor psychology The biggest challenge to using SDP for asset allocation is investor psychology. There are two issues. The first has to do with the conventional assumption that the asset poor should not be taking risks on highly volatile assets such as stocks. First, the asset poor need to focus on reducing their retirement expenditures, but once these clients have reached their limits with spending reductions, investing in stocks is the best available alternative. This does not simply apply to minimizing the chance of portfolio failure, but also to minimizing the time spent insolvent. When investing in stocks, portfolios may lose value, but this is more than offset by the likelihood of improved successful outcomes. The second issue is what to do if a client’s RPS slips significantly following a stock market downtown. Having just experienced major market losses, the client is now being told to invest heavily in stocks. This is the opposite of what most people want to do in such circumstances. They want to become more “conservative” by investing in bonds. Investing in bonds in such circumstances will reduce portfolio volatility, but it also reduces the chance of portfolio success. This analysis does assume that clients can be convinced to stay with the recommended asset allocations and not cave in to fear (selling stocks after losses) or greed (increasing stock allocations after market gains). Periodic portfolio reviews As we have seen, the optimal asset allocation moves after significant market events. This means an investor needs to monitor and adjust their asset allocation after such events. They cannot simply

follow a glide path. At these portfolio reviews a number of things need to happen. Communicating to a client their current risk level assuming an optimal portfolio in easy to understand terms, such as “a 5% chance of portfolio failure with a mean length of 7 years”, is always appropriate. Obviously the optimal asset allocation map can also be used to determine the client’s appropriate asset allocation. The prospect of insolvency can be detected by looking at the odds of portfolio success, and in such cases it may be possible to cut back on portfolio expenditures earlier to avoid a subsequent catastrophic failure. Showing a client their performance with an asset allocation or success probability graph, and then making appropriate recommendations for adjusting asset allocation, contribution rates, or retirement spending plans, is a better approach than asking the client to specify an optimal asset allocation with meanvariance indifference curves. The approach described here can be applied to all phases of life.

Implementation A free website where readers can use SDP to calculate asset allocations for individualized client circumstances is online at www.aacalc.com.

Acknowledgments I am grateful for feedback from Ross Williams and Jaipal Tuttle. I appreciate the time Rinku Banerji put into verifying the mathematical theory underlying our work. William Souza did an amazing job converting OPAL from Python to Java.

References Arias, Elizabeth. 2011. United States Life Tables, 2007, National Vital Statistics Reports, 59, 9: 2011-1120. http://www.cdc.gov/nchs/products/ life_tables.htm Basu, Anup K., and Michael E. Drew. 2009. “Portfolio Size Effect in Retirement Accounts: What Does It Imply for Lifecycle Asset Allocation Funds?” The Journal of Portfolio Management 35, 3 (Spring): 61-72. Bellman, Richard. “Bottleneck problems and dynamic programming.” Proceedings of the National Academy of Sciences of the United States of America 39.9 (1953): 947. Blanchett, David M., and Gregory W. Kasten. 2010. “Improving the ‘Target’ in Target-Date Investing.” Journal of Pension Benefits 18, 2 (Winter): 11–18. Black, Fischer, and Robert Litterman. 1992. “Global Portfolio Optimization.” Financial Analysts Journal 48, 5 (September/October): 28–43.

Journal of Personal Finance

Cooley, Phillip L., Carl M. Hubbard, and Daniel T. Walz. 1998. “Retirement Savings: Choosing a Withdrawal Rate That Is Sustainable.”American Association of Individual Investors Journal 20, 2 (February): 16-21. Finke, Michael, Wade D. Pfau, and Duncan Williams. 2012. “Spending Flexibility and Safe Withdrawal Rates.”Journal of Financial Planning 25, 3 (March): 44-51. Markowitz, Harry M. 1952. “Portfolio Selection.”The Journal of Finance 7, 1 (March): 77-91. Markowitz, Harry M. 1991. “Individual versus Institutional Investing.” Financial Services Review 1, 1: 1-8. Merton, Robert C. 1969. “Lifetime Portfolio Selection Under Uncertainty: The Continuous Time Case.” Review of Economics and Statistics, 51, 3 (August): 247-257. Mulvey, John M., and Hercules Vladimirou. “Stochastic network optimization models for investment planning.” Annals of Operations Research 20.1 (1989): 187-217. Scott, J. S., Sharpe, W. F., & Watson, J. G. (2009). The 4% Rule--At What Price?. Journal of investment management, (3), 31. Sharpe, William F. 1987. “An Algorithm for Portfolio Improvement.”In Kenneth D. Lawrence, John B. Guerard, Jr. and Gary D. Reeves Advances in Mathematical Programming and Financial Planning, eds. Kenneth D. Lawrence, John B. Guerard, Jr. and Gary D. Reeves. (Vol. 1, p. 155-170). Greenwich: JAI Press. Shiller, Robert J. 2005a. “Lifecycle Portfolios as Government Policy.”The Economists’ Voice 2, 1: 1-8. Shiller, Robert J. 2005b, updated. Irrational Exuberance. Princeton: Princeton University Press. http://www.econ.yale.edu/~shiller/data.htm

Volume 13, Issue 2

Parental and Early Influences on Expectations of Financial Planning for Retirement Janet L. Koposko, M.S., Doctoral Candidate, Department of Psychology, Oklahoma State University Douglas A. Hershey, Ph.D., Professor, Department of Psychology, Oklahoma State University

Author Note: The authors are indebted to Celinda Reese-Melancon and Maureen Sullivan for critical comments on an earlier draft of the manuscript. Correspondence should be addressed to the first author at the Department of Psychology, Oklahoma State University, Stillwater, Oklahoma 74078 or via email at koposko@okstate.edu. This investigation was designed to test a theoretically-grounded model of the psychomotivational dimensions that underlie retirement planning. In developing the hypothesized model, special consideration was given to positive early influences on development that could potentially impact other dimensions known to predict successful planning practices. Participants were 722 college students who reported the extent of childhood personal finance lessons learned, their retirement goal clarity and knowledge of financial planning, and expectations of future planning and anticipated satisfaction with life in retirement. As hypothesized, two measures of early financial influences were predictive of other variables known to underlie the retirement planning decision-making process, and oneâ&#x20AC;&#x2122;s vision of satisfaction in retirement. Results and implications are discussed in terms of the way in which motivational forces, particularly those that occur early in life, contribute to perceptions of future planning efforts.

Journal of Personal Finance

Introduction

saving in adulthood. From an applied perspective, if early influences are found to play a role in retirement saving practices in adulthood, then it becomes important to focus attention on this dimension. This is because unlike some motivational forces that are relatively immutable (such as personality traits or income limitations), early financial learning experiences can be carefully cultivated through modeling and intervention efforts.

The process of how individuals go about making financial plans for retirement is not a simple one or one that is easy to explain, and evidence suggests that many Americans fail to adequately plan and save for the post-employment period (VanDerhei & Copeland, 2010; Wiener & Doescher, 2008). A survey by the Employee Benefit Research Institute (Helman, Copeland, Adams, & VanDerhi, 2013) found that 57 percent of employees have less than $25,000 saved for retirement, and only 21 to 28 percent felt confident that they would be able to save enough to live comfortably after leaving the workforce. A similar lack of personal retirement savings can be found throughout much of the western world, particularly in countries where individual workers shoulder the responsibility for a portion of their own retirement income (Hershey, Jacobs-Lawson, & Austin, 2013).

The conceptual goal of the present investigation is to test a psychomotivational model of financial planning for retirement. The hypothesized model (see Figure 1) includes variables previously shown to motivate financial planning activities (e.g., financial knowledge, retirement goal clarity, future time perspective), in addition to variables that tap positive early financial learning experiences believed to contribute to a pattern of planning success. To test the model, we examined the experiences, attitudes, perceptions, and beliefs of an important yet understudied segment of the population—undergraduate college students. Although most published studies on this topic focus attention on middle-aged and older working adults, we felt that it was important to examine the future financial planning and saving intentions of college-age adults, inasmuch as intentions have been shown to be one of the best predictors of future behavior (Ajzen, 1991). An additional rationale for focusing on younger adults is because a large majority of these individuals stand on the threshold of entering the workforce, where they will be required to make important programmatic retirement saving decisions, and most will set in place a pattern of saving practices that could extend decades into the future.

Saving opportunities may be restricted by factors such as a limited income, not having access to an employer-sponsored retirement plan, or having other major expenses (e.g., a child’s college tuition) that limit discretionary resources. However, even individuals who do not face these saving challenges are sometimes financially ill-prepared for old age. This could, in part, be due to certain motivational forces that predispose some, but not others, to plan and save for retirement (Hershey, 2004; Lunt & Livingstone, 1991). A number of key motivational dimensions that influence saving have been identified in previous investigations. However, few studies have focused on positive financial learning experiences that occur early in life, and how those experiences contribute to a pattern of effective

Non-Family Early Influences

Parental Influences on Saving

H10

H11

Future Time Perspective

Retirement Goal Clarity

Financial Knowledge

Conscientiousness

Expected Expectations H1 Satisfaction of Financial with Life in Planning for Retirement Retirement

H8 H2 H3 H4 Figure 1. Hypothesized model of influences on expectations of financial planning for retirement and expected satisfaction with life in retirement. All paths shown in the model are expected to have beta weights with positive valences.

Volume 13, Issue 2

Role of Motivational Forces Financial Literacy. As a motivational construct, financial literacy involves financial knowledge, behavior, and attitudes, and it is used to refer to the range of awareness, knowledge, and skills that help people to make good decisions when it comes to managing money (OECD INFE, 2011). Many individuals who live in Western societies tend to demonstrate low levels of financial literacy (Lusardi & Mitchell, 2011a; Lusardi & Mitchell, 2011b), and it has been argued that literacy levels among youth and young adults are insufficient to make reasonably informed financial decisions (Anderson, Zhan, & Scott, 2004; Mandell & Klein, 2007). The situation described above can be rectified by educating children and adolescents about personal finance so as to promote sound financial saving habits over the course of one’s life (OECD, 2005). In recent decades, a handful of private and government programs have been instituted that are designed to teach children about personal finance (Anderson et al., 2004; Jump$tart Coalition, 2012). Although the need for approaches to early financial education has been recognized (Anderson et al., 2004; Shobe & Sturm, 2007), the implementation of worthwhile programs often face barriers because they can be costly and time consuming to administer. Furthermore, controversy exists as to the most effective means of educating children and adolescents about finances, and how early intervention programs should be best evaluated (McCormick, 2009). Some researchers have suggested that because parents have the primary influence on their children’s development, it is them who should be responsible for serving as positive role models so as to help their children achieve a reasonable degree of financial literacy (Heckman & Grable, 2011), make sound economic decisions (Webley & Nyhus, 2006), and develop healthy financial behaviors and attitudes (Jorgensen, 2010; Lusardi, Mitchell, & Curto, 2010). In the present investigation, self-rated financial knowledge will be used as the indicator of financial literacy.1 Based on these considerations, it is hypothesized that financial knowledge will be not only positively related to expectations of financial planning for retirement (path H5; Adams & Rau, 2011; Hershey, Jacobs-Lawson, McArdle, & Hamagami, 2007; Van Rooij, Lusardi, & Alessie, 2011), but also to expected satisfaction with life in retirement (path H2; Elder & Rudolpha, 1999; Gutierrez & Hershey, 2014). Personality Factors. Personality represents a second motivational dimension that has been shown to be associated with retirement planning and saving. Two personality traits in particular have received a fair amount of attention in the literature. Conscientiousness refers to the extent to which one is mindful of planning and responsive to making preparations, and it has been shown to be related to aspirational motivations in retirement (Robinson, Demetre, & Corney, 2010). This trait has also been shown to be predictive of another personality trait, future time perspective (Hershey & Mowen, 2000), which itself has been tied to planning practices. As a trait, future time perspective characterizes the extent to which individuals enjoy thinking about events in the distant future. Persons who are more future oriented or who feel more connected to possible future events tend to be more effective at planning and saving for retirement than those who are

not (Knoll, Tamborini, & Whitman, 2012; Wiener & Doescher, 2008). Being future oriented has also been associated with the desire to think about and discuss retirement plans with others (Yang & Devaney, 2011), which, we believe, should help to refine and clarify individuals’ long-range financial goals. Based on these findings, it is predicted that conscientiousness will be positively related to future time perspective (path H11; Hershey & Mowen, 2000; Webley & Nyhus, 2006). It is also predicted that future time perspective will be positively linked to both retirement goal clarity (path H9; Hershey et al., 2007; Yang & Devaney, 2011) and to expected satisfaction with life in retirement (path H3; Gutierrez & Hershey, 2014). Goals. The clarity of individuals’ retirement goals represents a third important dimension that has been linked to planning and saving practices. Financial advisors would argue that it is beneficial to calculate one’s financial needs well in advance of retirement, as doing so not only allows one to set critical savings goals, but it also allows one to establish a metric against which savings efforts may be measured. Yet, many individuals fail to carry out a future needs analysis that will facilitate setting a concrete savings goal, because they do not consider the task worthwhile (Mayer, Zick, & Marsden, 2011). In one recent study by Petkoska and Earl (2009), financial goals were shown to be a significant predictor of engaging in activities designed to increase financial knowledge and preparedness. That same investigation demonstrated that being in possession of clear and meaningful retirement goals played an important adaptive role in other (non-financial) domains, such as health and leisure. In other work by Stawski, Hershey, and Jacobs-Lawson (2007), the clarity of individuals’ retirement goals was found to be positively related to financial planning activities, which in turn, was linked to regular savings contributions. Based on the evidence cited above, it is anticipated that retirement goal clarity will be positively related to financial knowledge (path H7; Hershey et al., 2010; Petkoska & Earl, 2009). Expected Satisfaction with Life in Retirement. The fourth motivational dimension that will be examined as a part of this study involves expectations of satisfaction with life in retirement. Financial security is one key component when it comes to experiencing a high quality of life in old age, and insufficient engagement in planning and saving activities over the course of one’s career is likely to hinder post-employment satisfaction (Couture, 2011; Elder & Rudolpha, 1999). Moreover, previous research (Quick & Moen, 1998) has demonstrated that differences in planning behaviors lead to different quality of life outcomes in old age. In light of these linkages between planning and anticipated future quality of life, in the present study we use the Gutierrez and Hershey (2014) Expected Satisfaction with Life in Retirement Scale (SWLRS), which is based on the well-known Diener, Emmons, Larson, and Griffin (1985) Satisfaction with Life Scale (SWLS). It is hypothesized that financial knowledge will be positively related to expectations of financial planning for retirement (path H5; Adams & Rau, 2011; Hershey, et al., 2007; Van Rooij et al., 2011). Furthermore, it is anticipated that expectations of financial planning for retirement will be positively related to expected satisfaction with life in retirement (path H1; Elder & Rudolpha, 1999; Quick & Moen, 1998).

Journal of Personal Finance

Role of Early Learning Experiences In addition to the motivational forces identified in the previous section, planning and saving practices may also be realistically influenced by positive early financial learning experiences. Shobe and Sturm (2007) have made a strong argument to suggest that a lack of financial literacy among children and adolescents is a serious problem, and financial learning opportunities should ideally be introduced as early in life as possible. Studies have shown that parental influences play a considerable role in how individuals go about forming their attitudes, beliefs, and behaviors, both in the area of finance (Jorgensen, 2010) and in other life domains (Webly & Nyhus, 2006). Early parental and social influences on retirement planning and saving have been found to have a significant effect on retirement goal clarity (Hershey, Henkens, & Van Dalen, 2010) and financial knowledge (Guiterrez & Hershey, 2014). Furthermore, having parents who planned for their own retirement has been found to be predictive of one’s income (Dan, 2004), and income, in turn, has been shown to predict savings contributions (Hira, Rock, & Loibl, 2009; Lunt & Livingstone, 1991). Whereas positive parental and family learning experiences can increase financial planning involvement, more formal financial education also has the potential to make a significant contribution (Bernheim, Garrett, & Maki, 2001). Some schools include personal finance components as part of the curriculum (Fox, Bartholomae, & Lee, 2005; Spielhofer, Kerr, & Gardiner, 2010), and focused education in personal economics and related areas have been shown to help increase overall levels of financial literacy (Van Rooij, et al., 2011). Therefore, in addition to the role of parental influences on planning and saving, exposure to non-family early influences, such as school-based educational programs, should help to improve lifespan financial planning. In the present investigation, two different measures of early learning (parental influences and non-family influences) will be employed to assess the extent to which early financial learning experiences influence expectations of not only future planning and saving, but also expectations of satisfaction with life in retirement. Indeed, one of the clear value added aspects of the present study involves the inclusion of early learning indicators in the theoretical model to be tested. Based on the considerations regarding early learning experiences in the preceding paragraphs, it is hypothesized that non-family early influences will be positively related to financial knowledge (path H6; Bernheim et al., 2001; Van Rooij et al., 2011). It is also hypothesized that parental influences on saving will be positively related not only to future time perspective (path H10; Hershey & Mowen, 2000), but to financial knowledge as well (path H8; Bernheim et al., 2001; Walker, 2012). Furthermore, it is anticipated that parental influences on saving will be positively related to expected satisfaction with life in retirement (path H4; Gutierrez & Hershey, 2014). Theoretical Framework Elements of the theoretical foundation for the current study draw upon the life course perspective (also known as life course theory) (Crosnoe & Elder, 2002; Elder, 1994; Elder, 1998a, 1998b;

Umberson, Crosnoe, & Reczek, 2010). The life course perspective is a broad, meta-theoretical view of adult development. One aspect of the model maintains that individuals’ decisions are influenced by past life events as well as future expectations. Following from this observation, positive early financial learning experiences are likely to influence the way individuals think about retirement at present, and those present viewpoints are posited to shape expectations of future planning and saving practices. Core propositions found in image theory (Beach, 1998; Beach & Mitchell, 1987) also serve to buttress the proposed theoretical framework. Image theory researchers maintain that individuals do not use a formal analytical process when making significant life decisions (Beach, 1998); but rather, they make decisions on the basis of three things: (i) how well an action plan (in this case, making savings contributions) is likely to achieve one’s goals, (ii) whether the action plan is consistent with one’s morals, values, and beliefs, and (iii) whether the types of tactics and strategies associated with the action plan are reasonable and effective. Furthermore, like the life course perspective, image theory holds that lifespan planning and decision making is colored by personal experiences, previous consequential life decisions, and other contextual and situational factors. This study was designed to contribute to the extant literature in four different ways. First, it will build upon existing investigations by testing a theoretical model that is designed to replicate and extend the field of forces that underlie retirement planning practices. Second, as mentioned above, by studying college students we will examine a large and important segment of the population that has received scant attention in the literature on retirement finances. Third, by taking individuals’ early financial influences into account, we seek to take existing theoretical models in a novel and profitable direction. Finally, the present study is unique in that it will test a theoretically-derived model that is conceptualized from a lifespan perspective (Baltes, 1987; Baltes, Staudinger, & Lindenberger, 1999). This is accomplished by examining the way in which early financial influences shape perceptions and beliefs, as well as the way in which perceptions and beliefs lead to expectations of future financial sufficiency and quality of life.

Method Participants All participants in the study (N = 722) were students attending a large, mid-western state university. Each respondent earned partial credit in a psychology course for their participation. The mean age of the sample was 19.51 years (SD = 2.83), and 64.0 percent of the sample was female. The majority of the participants self-identified as being White (80.5 percent) and non-Hispanic (91.1 percent). At the time of testing, the majority of respondents were unemployed (72.4 percent). Only 3.0 percent of participants held jobs where they worked more than 32 hours per week. Measures The present study utilized a number of different scales and measures, some of which were existing scales that had been used

Volume 13, Issue 2

in previous investigations and others that were developed for the purpose of this study. All but the last scale listed below used a 7-point Likert-type response format (1 = strongly disagree; 7 = strongly agree). Each scale is described in detail below. Future Time Perspective. This 5-item scale (M = 5.66; SD = 1.09) measures the extent to which individuals are prone to think about the future, specifically in the context of retirement planning. The measure used in this investigation is a modified version of the Hershey et al. (2007) scale.2 A sample item is, “I enjoy thinking about how I will live years from now in the future.” Psychometric evaluation of the measure revealed a single factor structure and a coefficient alpha level of .89. The future time perspective score for each participant is the mean of the five items, with higher scores indicating a greater tendency toward future-oriented thinking. Financial Knowledge. This 3-item scale (M = 3.62; SD = 1.56) measures self-reported knowledge of financial planning for retirement (Hershey et al., 2010). A sample item is, “I know more than most people about retirement planning.” Psychometric evaluation of the measure revealed a single factor structure and a coefficient alpha level of .92. The financial knowledge score for each participant is the mean of the three items, with higher scores indicating higher levels of perceived financial knowledge. Retirement Goal Clarity. This 5-item scale (M = 3.73; SD = 1.52) measures the extent to which individuals report thinking about and setting specific goals for retirement (Stawski, Hershey, & Jacobs-Lawson, 2007). A sample item is, “I have a clear vision of how life will be in retirement.” Psychometric evaluation of the measure revealed a single factor structure and a coefficient alpha level of .91. The retirement goal clarity score for each participant is the mean of the five items, with higher scores indicating a greater degree of retirement goal clarity. Conscientiousness. This 3-item scale (M = 5.45; SD = 1.19) measures the extent to which individuals are efficient and precise when engaged on a task (Hershey & Mowen, 2000; Mowen, 2000). A sample item is, “I am organized.” Psychometric evaluation of the measure revealed a single factor structure and a coefficient alpha level of .87. The conscientiousness score for each participant is the mean of the three items, with higher scores indicating higher levels of task-related conscientiousness. Expected Satisfaction with Life in Retirement Scale. This 4-item scale (M = 5.04; SD = 1.16) assesses expectations of satisfaction with retirement among individuals who are not yet retired (Gutierrez & Hershey, 2014). A sample item is, “I expect that in retirement my life will be close to ideal.” Psychometric evaluation of the measure revealed a single factor structure and a coefficient alpha level of .89. The retirement satisfaction with life score for each participant is the mean of the four items, with higher scores indicating expectations of greater satisfaction with life in retirement. Expectations of Financial Planning for Retirement. This 3-item scale (M = 5.23; SD = 1.09) is a new scale designed for the present study to assess participants’ expectations of how easy or difficult they anticipate finding the task of retirement planning.

A sample item is, “Success at financial planning for retirement will be something that will come easily to me.” Psychometric evaluation of the measure revealed a single factor structure and a coefficient alpha level of .84. The expectations of financial planning for retirement score for each participant is the mean of the three items, with higher scores indicating expectations of minimal difficulties in carrying out financial planning tasks. Parental Influences on Saving. This 4-item scale (M = 5.67; SD = 1.24) is a new measure designed for the present study to assess the effect one’s parents had on money management and saving. A sample item is, “My parents had a strong influence on my current opinions about saving.” Psychometric evaluation of the measure revealed a single factor structure and a coefficient alpha level of .86. The parental influences on saving score for each participant is the mean of the four items, with higher scores indicating a greater degree of positive parental influences on saving. Non-Family Early Learning Experiences. This 5-item scale (M = 0.45; SD = 0.09) is a new measure designed for the present study to asses financial knowledge derived during childhood or adolescence from sources beyond one’s family or parents. A sample item is, “In school I took a course on money management, investing, or personal finance.” The response format for each of the five items was dichotomous (0 = no; 1 = yes); therefore, the total score for each participant was the sum of the five dichotomous items. The degree of internal consistency (KR-20) is adequate at .67. Higher scores on this measure indicate more in the way of financial learning experiences in school or community-based settings. The last three (newly developed) scales listed above were created by identifying dimensions relevant to the scale topic, writing items that reflect those dimensions, and then pilot testing those items to determine whether they were suitable as part of the three measures. Procedure Participants completed an online questionnaire that was designed using the web-based software SurveyGizmo (Widgix, 2012). Most questions contained in the instrument used 7-point Likerttype scales; one measure (non-family influences) used dichotomous (yes/no) scoring. Each of the scales and measures contained in the questionnaire is described in Table 1; a complete list of scales and their corresponding items can be found in the Appendix. Following the completion of testing, all participants were thanked for their participation and given contact information for the investigators should they have any questions about the study. Analysis Plan. In terms of an analysis plan, a measurement model will first be tested to ensure that all scale items load on their respective constructs. Once the factor structure for the scales has been confirmed, the path model shown in Figure 1 will be tested. As part of that process, the statistical significance of slope parameters in the model will be evaluated, and the overall goodness of fit of the broader theoretical framework will be assessed.

Journal of Personal Finance

22 Non-Family

.16

Early Influences

.38 2

R = .22

Parental Influences on Saving

.32 2

R = .30

.35

Future Time Perspective

Retirement Goal Clarity

.25

.67 2

R = .59

Financial Knowledge

Conscientiousness

R = .40

R = .30

Expectations

Expected

Planning for Retirement

with Life in Retirement

.28 of Financial .30 Satisfaction

.30 .08

.14

.25 .30 Figure 2. Observed model of influences on expectations of financial planning for retirement and anticipated satisfaction with life in retirement. All path parameters shown are standardized beta weights, and all were found to be statistically significant at the .01 level.

Results The data were cleansed and examined for skew, kurtosis, outliers, and any other possible issues that may lead to either distributional distortions or violations of the assumptions of general linear model analyses. Prior to testing the model shown in Figure 1, a measurement model was created to ensure that the factor structure of the items were as hypothesized for each scale. One independent variable, the non-family early influences measure, was not included in the measurement model because it utilized a different type of response format. The measurement model was evaluated using the Analysis of Moments Structures (AMOS) software version 19 (Arbuckle, 2010). Model fit indices for both the measurement and path model were interpreted according to criteria established by Hu and Bentler (1999), as well as Hooper, Coughlan, & Mullen (2008). The measurement model was found to be a good fit to the data, χ2 (303) = 1221.24, p < .01, TLI = .92; CFI = .93; RMSEA = .07. No appreciable cross-loadings were observed and the model fit could not be improved by re-specifying paths to non-hypothesized constructs. In sum, the computation of this measurement model demonstrates empirical evidence that the items for the various scales loaded on their respective factors, which served to pave the way to compute the hypothesized path analysis model. The path model shown in Figure 1 was then analyzed in order to compute values for the eleven path parameters and establish metrics reflecting overall goodness-of-fit. Exogenous variables were allowed to correlate.3 As is often the case when using structural equation modeling software, the initial run of the model was found to have a less than adequate fit, χ2 (14) = 433.17, p < .01, TLI = .55, CFI = .78, RMSEA = .20. Modification indices

revealed that the fit could be improved by deleting the path from parental influences on saving to expected satisfaction with life in retirement (H4). Modification indices also suggested that fit could be improved by adding three new paths to the model: one from conscientiousness to expectations of financial planning for retirement, a second from non-family early influences to goal clarity, and a third from future time perspective to expectations of financial planning for retirement. It was decided that all three of these paths were theoretically plausible; therefore, each was incorporated into the revised model. Next, a revised path model was tested that contained all eight original variables, but now thirteen paths. In this model, exogenous variables were again allowed to correlate. The resulting specification was shown to be a good fit to the data, χ2 (12) = 68.74, p < .01, TLI = .93, CFI = .97, RMSEA = .08. Moreover, all thirteen path parameters were shown to be statistically significant at the .01 level. A graphic representation of the revised model, which contains R2 values for each endogenous variable and standardized beta weights for each path, is shown in Figure 2. As seen in the figure, this model did an excellent job in accounting for variance among the endogenous variables, capturing between 22 to 59 percent of the total variance operating for each construct.

Discussion The overarching goal of the present investigation was to test a theoretically driven, lifespan model of retirement planning. It was expected that the hypothesized paths shown in Figure 1 would reveal a number of important relationships between key retirement planning constructs, and those predicted relationships would account for appreciable amounts of variance among the

Volume 13, Issue 2

endogenous variables. The revised path model was found to meet those expectations. Indeed, the findings provide important insights into the way college students think about the retirement planning process. Two broad take-away messages are worth noting at the outset of the discussion. The first is that the field of forces that influence the anticipated retirement planning practices of young (mostly non-working) college students is quite similar to the motivational forces that shape the planning and saving behaviors of older, working adults. This is seen by the fact that many of the variables (and relationships between variables) identified as important in the present investigation have also been shown to play a role in studies carried out with members of middle-aged and older cohorts (Adams & Rau, 2011; Hira et al., 2009; Hershey et al., 2007; Hershey et al., 2010; Petkoska & Earl, 2009; Webley & Nyhus, 2006). The second broad finding is that early financial influences do indeed have an effect on individuals’ motives to save for retirement, which is a topic that has received scant attention in the extant literature on financial and retirement planning (Doyle, 2007; Jorgenson, 2010; Lusardi et al., 2010). Both findings suggest important theoretical and applied implications, which are discussed in the following paragraphs. Two different theoretical frameworks were used in order to position the present investigation within the existing literature. These frameworks were the life course perspective (Elder, 1998a) and image theory (Beach & Mitchell, 1987). The findings from the observed path model were consistent with both of these theories. One key proposition of the life course perspective is that individuals’ lives are embedded in social contexts (Elder, 1998a), and an individual’s family structure is one such context. Therefore, the fact that parental influences on saving was predictive of individuals’ future time perspective is consistent with life course theory. What this suggests is that for many of the college students involved in this study, forward-thinking attitudes were promoted in the social context of the home environment. The life course perspective also suggests that individuals have “linked lives,” and that each individual is influenced by significant others in his or her life sphere (Elder, 1998a). This premise was also supported by the data, in that individuals who reported having positive parental influences ultimately developed higher levels of financial knowledge (H8). However, the prediction that parental influences would be related to superior expectations of satisfaction with life in retirement (H4) was not supported by the data. Perhaps this non-significant hypothesized finding is due to the number of years that transpire between one’s early learning experiences and how they envision their quality of life decades into the future. Another key element of the life course perspective is human agency, or the idea that individuals shape their lives by choosing to engage (or choosing not to engage) in certain types of activities (Elder, 1994). Choosing to take part in non-family related financial learning activities during one’s formative years is consistent with the notion of human agency, and it appears that the nature of these experiences helps to shape individuals’ future behaviors when it comes to planning and saving. Both of these life course theory elements—linked lives and human agency—provide the-

oretical support for the observed relationships between parental influences on future time perspective (H10), and non-family early influences on financial knowledge (H6), respectively. The second theoretical framework used as a foundation for the present investigation was image theory (Beach, 1998; Beach, 1990; Beach & Mitchell, 1987). The “trajectory image” in image theory refers to a decision-maker’s goal state, or in other words, the state the individual desires to achieve in the future (cf., Austin & Vancouver, 1996). In this investigation, the extent to which one thinks about future goal states was represented by the measure of future time perspective, and this measure was predictive of not only retirement goal clarity (H9), but also expectations of satisfaction with life in retirement (H3). Taken together, this pair of findings provides empirical support for the closely aligned constructs of one’s orientation to time, the clarity of one’s goals, and one’s vision of the future. Beyond the trajectory image, Beach’s image theory posits that individuals make decisions in the context of two other images: the “strategic image” and the “value image” (Beach, 1990; Beach & Mitchell, 1987). The strategic image represents the plans and tactics individuals use to achieve their goals. In terms of the present investigation, financial knowledge could be thought to serve as a proxy indicator for the strategic image. The third of the three images, the value image, represents personal values, morals, and ethics held by the decision maker. In the present investigation, the early influence variables—parental influences on saving and non-family early influences—could be considered to reflect one’s personal financial values, inasmuch as they shape personal beliefs about the world that the individual carries forward into adulthood. In the observed model of retirement planning, both non-family early influences and parental influences on saving were predictive of financial knowledge (H6 and H8, respectively), which are reflective of the theoretical link between one’s value image and strategic image. Financial knowledge, in turn, was predictive of variables involving future expectations (i.e., expectations of financial planning for retirement [H5] and expected satisfaction with life in retirement [H2]), which is reflective of the theoretical link between one’s strategic image and trajectory image. In short, these observed empirical relationships are consistent with the flow of influences posited in Beach’s theoretical framework. Implications also exist in terms of the way in which personality traits influence individuals’ retirement planning decisions. Two personality variables—conscientiousness and future time perspective—were included in the hypothesized model. In previous investigations, conscientiousness has been shown to be associated with future time perspective and knowledge of financial planning (Hershey & Mowen, 2000; Webley & Nyhus, 2006). In the observed model, the first of these two findings (H11) was replicated. Furthermore, the observed model revealed that conscientiousness was predictive of expectations of financial planning for retirement, which is a non-hypothesized empirical link not previously demonstrated. Given the fact that one’s level of conscientiousness tends to remain developmentally stable over the life course (Gallagher, Fleeson, & Hoyle, 2010), low levels of this personality trait could represent a true barrier to envisioning oneself as being

Journal of Personal Finance

effective when it comes to planning and saving for the future. Like conscientiousness, the measure of future time perspective served to replicate and extend associations with other measures in the model. In support of H9, future time perspective was predictive of retirement goal clarity among these college student respondents, which is a relationship that has previously been established among a sample of older adults (Hershey et al., 2007). Furthermore, in support of H3, future time perspective was predictive of expectations of financial planning for retirement, which is an effect that has not previously been demonstrated. Knowledge of the linkages between personality traits and expectations of future life satisfaction might benefit intervention specialists, who not only face the challenge of getting their clients to plan and save, but also to envision a financially secure and worry-free old age. One final, broader theoretical implication has to do with the use of multivariate models to capture complex decision making processes. The goal of understanding complex thought has been the subject of increased attention in recent years (Bakken, 2008; Bargh, 2011; Klein, 2005; Qudrat-Ullah, 2008). In the present investigation, eight variables were analyzed in relation to one another, which resulted in a holistic picture of the forces that drive an individual to save for retirement. The results from this analytic effort serve to replicate and extend existing multivariate models of retirement planning (e.g., Adams & Rau, 2011; Gutierrez & Hershey, 2014; Hershey et al., 2007; Hershey et al., 2010; Hershey & Mowen, 2000; Webley & Nyhus, 2006). The complex and dynamic nature of the model tested brings into sharp focus the important role of human agency (Elder, 1994), which suggests that individuals make decisions within the context of multiple forms of opportunities and constraints. In terms of applied implications, the findings from this study should help retirement counselors and financial professionals develop more effective and efficient approaches to intervention. In the present investigation, early learning experiences were found to play a prominent role in shaping attitudes toward and knowledge of retirement planning. This would suggest that the scaffolding of youth financial education programs could help individuals acquire a solid level of financial literacy by their early twenties (Cowen, Blair, & Taylor, 2011). Indeed, we believe that early financial learning experiences can translate into positive attitudes toward money management, saving, and financial independence if they are introduced to children and adolescents at the

right time and in a meaningful manner. Whereas certain psychomotivational dimensions that shape planning practices (such as elemental personality traits) tend not to be malleable (Gallagher et al., 2010), it is encouraging to note that early financial learning experiences (which can take the form of parental modeling or formal interventions) may be a particularly effective means of nurturing individuals into becoming both interested in planning and competent when it comes to saving. The results of this study offer a number of valuable insights into the psychological mechanisms that underlie retirement planning. However, certain limitations should be acknowledged. These limitations include the fact that the scales used relied upon self-report, which may be subject to certain reporting biases; the investigation relied on the use of correlational data, which limits the ability to draw causal conclusions (Cliff, 1983); and the observed findings cannot be generalized to non-college age populations. To address the first limitation, in future investigations researchers might consider using objective measures (where applicable) in conjunction with self-reports. With regard to the second limitation, in future studies a true experimental design could be employed, in which one group of children is assigned to complete a financial literacy program (while the contrast group does not). A different experimental design might involve having one group of parents receive money management training, while training is withheld from a second (matched) group of parents, to observe the effects of the intervention on their children. And the third limitation cited above could be addressed by designing a study in which the model shown in Figure 2 is tested on populations other than college-aged students. The results of this study make both theoretical and applied contributions to the existing literature on the psychology of retirement planning. From a broad theoretical perspective, the findings suggest that we should take seriously the impact early learning experiences have on an individualâ&#x20AC;&#x2122;s development. It appears the long-terms effects of positive financial lessons learned in the home, the school system, or the community, not only extend oneâ&#x20AC;&#x2122;s view of the future, but they also help to clarify retirement goals and enhance levels of financial literacy. From an applied perspective, our findings provide educators, retirement counselors, and financial professionals excellent reasons and motives to promote forward-thinking youth intervention programs designed to foster appropriate levels of financial competence.

ÂŠ2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 2

References Adams, G. A., & Rau, B. L. (2011). Putting off tomorrow to do what you want today: Planning for retirement. American Psychologist, 66, 180192. doi: 10.1037/a0022131 Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior & Human Decision Processes, 50, 179. doi:10.1016/0749-5978(91)90020-T Anderson, S. G., Zhan, M., & Scott, J. (2004). Targeting financial management training at low-income audiences. Journal of Consumer Affairs, 38, 167–177. doi: 10.1111/j.1745-6606.2004.tb00470.x Arbuckle, J. L. (2010). IBM SPSS Amos 19 user’s guide. Chicago, IL: Amos Development Corporation. Austin, J. T., & Vancouver, J. B. (1996). Goal constructs in psychology: Structure, process, and content. Psychological Bulletin, 120, 338-375. doi: 10.1037/0033-2909.120.3.338 Bakken, B. E. (2008). On improving dynamic decision-making: Implications from multiple-process cognitive theory. Systems Research and Behavioral Science, 25, 493-501. doi: 10.1002/sres.906 Baltes, P. B. (1987). Theoretical propositions of life-span developmental psychology: On the dynamics between growth and decline. Developmental Psychology, 2.1, 611-626. doi:10.1037/0012-1649.23.5.611 Baltes, P. B., Staudinger, U. M., & Lindenberger, U. (1999). Lifespan psychology: Theory and application to intellectual functioning. Annual Review of Psychology, 50, 471-507. doc10.1146/annurev.psych.50.1.471 Bargh, J. (2011). Unconscious thought theory and its discontents: A critique of the critiques. Social Cognition, 29, 629-647. doi: 10.1521/ soco.2011.29.6.629 Beach, L. R. (1990). Image theory: Decision making in personal and organizational contexts. Chichester, UK: Wiley. Beach, L. R., & Mitchell, T. R. (1987). Image theory: Principles, goals, and plans in decision making. Acta Psychologica, 66, 201-220. doi: 10.1016/0001-6918(87)90034-5 Beach, L. R. (Ed.). (1998). Image theory: Theoretical and empirical foundations. Mahwah, NJ: Lawrence Erlbaum Associates. Bernheim, B. D., Garrett, D. M., & Maki, D. M. (2001). Education and saving: The long-term effects of high school financial curriculum mandates. Journal of Public Economics, 80, 435-465. doi: 10.1016/S00472727(00)00120-1 Cliff, N. (1983). Some cautions concerning the application of causal modeling methods. Multivariate Behavioral Research, 18, 115-126. Couture, L. (2011). Planning for retirement. Activities, Adaptation, and Aging, 35, 267-273. doi: 10.1080/01924788.2011.602927 Cowen, J., Blair, W., & Taylor, S. (2011). The use of scaffolding in the financial planning classroom: An Australian case study. Australian Accounting and Business Finance Journal, 5, 3-16. Retrieved from: http://ro.uow. edu.au/aabfj/vol5/iss3/2 Crosnoe, R., & Elder, G. H. Jr. (2002). Successful adaptation in the later years: A life course approach to aging. Social Psychology Quarterly, 65, 309-328. Dan, A. A. (2004). What are people doing to prepare for retirement? Structural, personal, work, and family predictors of planning. Doctoral dissertation, Case Western Reserve University. DAI-A 65/01, 251. Retrieved from: http:etd.ohiolink.edu/view.cgi/Dan%20 Amy%20Anne. pdf?case1079241195 Diener, E., Emmons, R. A., Larsen, R. J., & Griffin, S. (1985). The satisfaction with life scale. Journal of Personality Assessment, 49, 71-75. doi: 10.1207/s15327752jpa4901_13

Doyle, N. (2007). Children’s pensions: How to develop an accessible and understandable personal savings habit that produces an adequate retirement income. Pensions, 12, 213-216. doi: 10.1057/palgrave.pm.5950057 Elder, G. H., Jr. (1998a). The life course and human development. In R. M. Lerner (Ed.), Handbook of Child Psychology: Vol. 1. Theoretical models of human development, 5th ed., (pp. 939-991). New York: Wiley. Elder, G. H., Jr. (1998b). The life course as developmental theory. Child Development, 69, 1-12. doi: 10.1111/j.1467-8624.1998.tb06128.x Elder, G. H., Jr. (1994). Time, human agency, and social change: Perspectives on the life course. Social Psychology Quarterly, 57, 4-15. doi: 10.2307/2786971 Elder & Rudolpha (1999). Does retirement planning affect the level of retirement satisfaction? Financial Services Review, 8, 117-127. Fox, Bartholomae, & Lee (2005). Building the case for financial education. Journal of Consumer Affairs, 39, 195-213. doi: 10.1111/j.17456606.2005.00009.x Gallagher, P., Fleeson, W., & Hoyle, R. H. (2010). A self-regulatory mechanism for personality trait stability: Contra-trait effort. Social Psychological and Personality Science, 2, 335-342. doi: 10.1177/1948550610390701 Goldsmith, E., & Goldsmith, R. E. (1997). Gender differences in perceived and real knowledge of financial investments. Psychological Reports, 80, 236-238. doi: 10.2466/PR0.80.1.236-238 Goldsmith, R. E., Goldsmith, E. B., & Heaney, J. G. (1997). Sex differences in financial knowledge: A replication and extension. Psychological Reports, 81, 1169-1170. doi: 10.2466/pr0.1997.81.3f.1169 Gutierrez, H. C. & Hershey, D. A. (2014). Age differences in expected satisfaction with life in retirement. International Journal of Aging and Human Development, 78-116. doi: http://dx.doi.org/10.2190/AG.78.2.a Heckman, S. J., & Grable, J. E. (2011). Testing the role of parental debt attitudes, student income, dependency status, and financial knowledge have in shaping financial self-efficacy among college students. College Student Journal, 45, 51-64. Helman, R., Adams, N., Copeland, C., & VanDerhei, J. (2013). The 2013 Retirement Confidence Survey: Perceived Savings Needs Outpace Reality for Many (Issue Brief No. 384). Washington, D. C.: EBRI. Retrieved from http://www.ebri.org/publications/ib/index.cfm ?fa=ibDisp&content_id=5175 Hershey, D. A. (2004). Psychological influences on the retirement investor. CSA Journal, 22, 31-39. Hershey, D. A., Henkens, K., & Van Dalen, H. P. (2010). Aging and financial planning for retirement: Interdisciplinary influences viewed through a cross-cultural lens. International Journal of Aging and Human Development, 70, 1-38. doi: 10.2190/AG.70.1.a Hershey, D. A., Jacobs-Lawson, J.M., & Austin, J. T. (2013). Effective financial planning for retirement. In M. Wang (Ed.), The Oxford handbook of retirement (pp. 402-430). New York: Oxford University Press. Hershey, D. A., Jacobs-Lawson, J. M., McArdle, J. J., & Hamagami, F. (2007). Psychological foundations of financial planning for retirement. Journal of Adult Development, 14, 26-36. doi: 10.1007/s10804-0079028-1 Hershey, D. A. & Mowen, J. C. (2000). Psychological determinants of financial preparedness for retirement. The Gerontologist, 40, 687-697. doi: 10.1093/geront/40.6.687 Hira, T. K., Rock, W. L., & Loibl, C. (2009). Determinants of retirement planning behaviour and differences by age. International Journal of Consumer Studies, 33, 293-301. doi: 10.1111/j.1470-6431.2009.00742.x

Journal of Personal Finance

Hooper, D., Coughlan, J., & Mullen, M. (2008). Structural equation modeling: Guidelines for determining model fit. Journal of Business Research Methods, 6, 53-60. Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1-55. doi: 10.1080/10705519909540118 Jorgensen, B. L. (2010). Financial literacy of young adults: The importance of parental socialization. Family Relations, 59, 465-478. doi: 10.1111/j.1741-3729.2010.00616.x Jump$tart Coalition (2012). Jump$tart Coalition for Personal Financial Literacy. Retrieved from http://jumpstart.org/about-us.html Klein, J. (2005). The contributions of a decision support system to complex educational decisions. Educational Research and Evaluation, 11, 221234. doi: 10.1080/13803610500101140 Knoll, M. A. Z., Tamborini, C. R., & Whitman, K. (2012). I do…want to save: Marriage and retirement savings in young households. Journal of Marriage and Family, 74, 86-100. doi: 10.1111/j.17413737.2011.00877.x Lunt, P. K., & Livingstone, S. M. (1991). Psychological, social, and economic determinants of saving: Comparing recurrent and total savings. Journal of Economic Psychology, 12, 621-641. doi: 10.1016/01674870(91)90003-C Lusardi, A., & Mitchell, O. S. (2011a). Financial literacy around the world: An overview. Journal of Pension Economics and Finance, 10, 497-508. doi: 10.1017/S1474747211000448 Lusardi, A. & Mitchell, O.S. (2011b). Financial literacy and retirement planning in the United States. Journal of Pension Economics and Finance, 10, 509-525. doi: 10.1017/S147474721100045x Lusardi, A., Mitchell, O. S., & Curto, V. (2010). Financial literacy among the young. The Journal of Consumer Affairs, 44, 358-380. doi: 10.1111/j.1745-6606.2010.01173.x Mandell, L., & Klein, L. S. (2007). Motivation and financial literacy. Financial Services Review, 16, 105-116. Mayer, R. N., Zick, C. D., & Marsden, M. (2011). Does calculating retirement needs boost retirement savings? The Journal of Consumer Affairs, 45, 175-200. doi: 10.1111/j.1745-6606.2011.01199.x McCormick, M. H. (2009). The effectiveness of youth financial education: A review of the literature. Journal of Financial Counseling and Planning, 20, 70-79.

Robinson, O. C., Demetre, J. D. & Corney, R. (2010). Personality and retirement: Exploring the links between the Big Five personality traits, reasons for retirement and the experience of being retired. Personality and Individual Differences, 48, 792-797. doi: 10.1016/j.paid.2010.01.014 Shobe, M. A., & Sturm, S. L. (2007). Youth individual development accounts: Retirement planning initiatives. Children & Schools, 29, 172181. Spielhofer, T., Kerr, D., & Gardiner, C. (2010). Personal finance education: Effective practice guide for schools. Slough, UK: National Foundation for Educational Research. Stawski, R. S., Hershey, D. A., & Jacobs-Lawson, J. M. (2007). Goal clarity and financial planning activities as determinants of retirement savings contributions. International Journal of Aging and Human Development, 64, 13-32. doi: 10.2190/13GK-5H72-H324-16P2 Umberson, D. Crosnoe, R., & Reczek, C. (2010). Social relationships and health behavior across the life course. Annual Review of Sociology, 36, 139-157. doi:10.1146/annurev-soc-070308-120011 VanDerhei, J., & Copeland, C. (2010). The EBRI retirement readiness rating: Retirement income preparation and future prospects. (Issue Brief No. 344). Washington, D. C.: EBRI. Retrieved from http://ebri.org/publications/ib/index.cfm? fa=ib Disp&content_id=4593 Van Rooij, M. C. J., Lusardi, A., & Alessie, R. J. M. (2011). Financial literacy and retirement planning in the Netherlands. Journal of Economic Psychology, 32, 593-608. doi:10.1016/j.joep.2011.02.004 Walker, R. B. (2012). Sustainable personal retirement. McGraw-Hill Research Foundation: Retrieved from: www.mcgraw-hillresearchfoundation.org Webley, P., & Nyhus, E. K. (2006). Parents’ influence on children’s future orientation and saving. Journal of Economic Psychology, 27, 140-164. doi: 10.1016/j.joep.2005.06.016 Widgix (2012). SurveyGizmo [Computer Software] (2012). Boulder, CO: Widgix, LLC. Wiener, J., & Doescher, T. (2008). A framework for promoting retirement savings. The Journal of Consumer Affairs, 42, 137-164. doi: 10.1111/j.1745-6606.2008.00102.x Yang & Devaney (2011). Intrinsic rewards of work, future time perspective, the economy in the future, and retirement planning. Journal of Consumer Affairs, 45, 419-444. doi:10.1111/j.1745-6606.2011.01211.x

Mowen, J. C. (2000). The 3M Model of Personality: Theory and Empirical Applications to Consumer Behavior. Boston: Kluwer Academic Press. OECD INFE (2011). Measuring financial literacy: Questionnaire and guidance notes for conducting an internationally comparable survey of financial literacy. Paris: OECD. OECD (2005). Recommendation on principles and good practices for financial education and awareness. Retrieved from http://www.oecd.org/ dataoecd/7/17/35108560.pdf Petkoska & Earl (2009). Understanding the influence of demographic and psychological variables on retirement planning. Psychology and Aging, 24, 245-251. doi: 10.1037/a0014096 Qudrat-Ullah, H. (2008). Future directions on complex decision making using modeling and simulation decision support. In H. Quadrat-Ullah, J. M. Spector, & P. I. Davidsen (Eds), Complex Decision Making: Theory and Practice (pp. 323-337). Cambridge, MA: Springer. Quick H. E. & Moen, P. (1998). Gender, employment, and retirement quality: A life course approach to the differential experiences of men and women. Journal of Occupational Health Psychology, 3, 44-64. doi: 10.1037/1076-8998.3.1.44

Endnotes 1Use

of a self-report measure of financial knowledge was decided upon because it is more efficient to administer than an objective measure of knowledge, and both objective and self-reported financial knowledge have previously been shown to be positively correlated (Goldsmith & Goldsmith, 1997; Goldsmith, Goldsmith, & Heaney, 1997).

2The

original Hershey et al. (2007) future time perspective scale contained six items, four of which were reverse coded. In an effort to improve the level of internal consistency, the four reverse coded items were replaced with the following three new items: “I look forward to life in the distant future,” “It is important to take a long-term perspective on life,” and “My close friends would describe me as future oriented.”

3The

three correlations among exogenous measured variables (i.e., parental influences, conscientiousness, and non-family influences) were all quite small, and for that reason are not shown in Figure 1.

Volume 13, Issue 2

Appendix: List of Scales Used in the Study Expected Satisfaction with Life in Retirement Scale

Conscientiousness Scale

1. I expect that in retirement my life will be close to ideal.

1. I am organized.

2. Once I enter retirement, the conditions of my life will be excellent.

2. I am orderly.

3. After I retire, I will be satisfied with life.

3. I am efficient.

4. After I retire, I will have gotten the important things I wanted in life. Non-Family Early Influences Measure Expected Financial Planning for Retirement Scale 1. I expect to meet my financial goals in terms of planning and saving for the future. 2. I think I will do a good job of planning and saving for retirement. 3. Success at financial planning for retirement will be something that will come easily to me.

Self-rated Financial Knowledge Scale 1. I know a great deal about financial planning for retirement. 2. I have informed myself about financial preparation for retirement.

1. In school I took a course on money management, investing, or personal finance. 2. In the past, I have seen a guest speaker, educator, or other person talk about financial planning. 3. At some point during school, I studied the general structure of how social security and pension plans work. 4. When I learned about career planning and career exploration in school, I learned about typical retirement saving options that are offered to employees by their employer. 5. I had to do an assignment or class project in the past that involved making either a real or mock budget. This involved describing the types of things I would spend money on and how I could save money to get the things I need.

3. I know more than most people about retirement planning.

Goal Clarity Scale 1. I have set clear goals for gaining information about retirement. 2. I have thought a great deal about my quality of life in retirement. 3. I set specific goals for how much will need to be saved for retirement. 4. I have a clear vision of how life will be in retirement. 5. I have discussed retirement plans with a spouse, friend, or significant other.

Future Time Perspective Scale 1. I enjoy thinking about how I will live years from now in the future. 2. I like to reflect on what the future will hold. 3. I look forward to life in the distant future. 4. It is important to take a long-term perspective on life. 5. My close friends would describe me as future oriented.

Parental Influences on Saving Scale 1. Growing up, my parents helped me to imagine situations when I might need extra money to fall back on. 2. My parents made sure that I understood the value of money and that money is a limited resource. 3. Saving money for the future was an important lesson I learned as a child. 4. My parents suggested to me concrete ways to save money on my own.

Journal of Personal Finance

The Role of Trans-Generational Financial Knowledge and Self-Reported Financial Literacy on Borrowing Practices and Debt Accumulation of College Students Dr. Chad Smith, Associate Professor, Department of Management, Marketing and Human Resources, Clarion University of Pennsylvania Dr. Gustavo Barboza, Professor, Department of Management, Marketing and Human Resources, Clarion University of Pennsylvania

This paper studies the effects of trans-generational financial knowledge, self-reported financial knowledge, academic performance, and overall financial literacy on financial management practices, using a sample of 380 college students. Exploratory estimates using a series of ordered Probit models indicate that academic status and self-reported overconfidence on financial knowledge relate positively to the amount of debt a student carries. More interesting, our estimates provide robust support to the hypothesis that a trans-generational financial knowledge effect from parents to students plays a major role in reducing the amount of financial burden students assume, both in the form of student loans and credit card balance. In addition, students that maintain a good credit card history, as reflected by high repayment rates, are more likely to hold lower debt amounts than otherwise. A high debt level implies that students are living beyond their means and consequently developing unhealthy financial management practices. Our paper provides evidence in favor to the hypothesis that early financial education is a means to reduce or maintain low levels of indebtedness. Our empirical estimates also point out to the presence of a strong overconfidence effect, as reflected by unrealistically high selfreported financial knowledge, leading students to an incorrect decision making process in favor to holding more debt. We argue that lack of personal financial literacy is at the core of high debt accumulation.

ÂŠ2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 2

1. Introduction Issues relating to financial literacy, the role that parents and educators play in transmitting financial knowledge to the youth and the relationship to debt management remain as relatively unexplored topics in the field of financial education. These issues become relevant as considerations regarding effective and efficient personal financial management are of the outmost importance concerning individuals’ present and future financial wellbeing. Financial management revolves around a continuous process of consumption-savings decisions, and it is directly affected by the overall knowledge perception and actual behavior towards financial decision-making processes. In a fast changing financial environment, where financial innovation increases the complexity of every day financial decision, the evidence overwhelmingly indicates that students’ financial knowledge is not keeping up with these changes and therefore they are ill-equiped to perform basic financial decisions. The negative implications deriving from this financial knowledge gap are reasons for concern, and could potentially be great. If we assume that an effective financial decision-making process is fundamental to a healthy financial management, individuals would be better off if they were to learn and master basic financial concepts at a relatively young age. If this condition is met, one would also expect that by the time the youth become college students they would be able to demonstrate adequate capacities to make effective financial decisions. However, previous research on financial literacy of college students reveals the existence of a significant gap between the formal and informal means of personal financial education, manifesting itself in large amounts of debt holdings. For instance, Lewin (2011) indicates that on average graduating college students accumulate $25,250 of a burden in debt, up 5% from 2009, with a projection to reach $27,000 for 2011 and 2012. By the same token, Nellie Mae Corp., (2002) indicated that on average graduating college students have $20,402 of a burden in debt when leaving college. While we recognize that some of this debt is related to educational expenses, it is also relevant to note that a good amount of it relates to credit card debt. Kuchler (2013) notes that in 2010, 70% of US households have at least one credit card and an average credit card balance month-to-month of approximately $12,900. Manning (2000) notes that overconfident college students are fueled by easy access to consumer credit allowing for enhanced lifestyle activities, and peer “competitive consumption” pressure – like in keeping up with the Joneses behavior – on college campuses that heightens the social acceptance, thus increasing the level of personal debt. To illustrate the severity of the problem, Manning (2000) argues that the lack of personal finance education in schools has created a sense of complacency and acceptance among those students unaware of financial debt, allowing for irrational borrowing behavior practices to develop. One particular area that has received special attention is the rapid growth of credit card debt among the youth. In an effort to curb the rampant increase in credit card debt held by the youth, the Card Accountability

and Responsibility and Disclosure (CARD) Act of 2009 was approved, requiring that individuals 21 years of age or younger, must have a cosigner or submit financial information to guarantee repayment capabilities (See Debbaut, Ghent, & Kudlyak, 2013 for more details)1. An early inspection of the literature indicates that financial education – formal and/or informal – seems to be at the core of finding a solution to the observed problem of over-indebtedness by college students. In particular, the evidence provides support to the hypothesis that educating students on financial responsibility and financial literacy before entering, or while in college, could possibly lower the accumulated students’ debt by the time they leave college.2 In support of this, Lusardi, Mitchell, and Curto (2009a) indicate that financial education should take place as early as high school. This becomes more relevant, as recent developments in the field of financial and consumer behavior point to the differences between naïve and sophisticated individuals as a possible explanation for larger amounts of debt balances held by the naïve3 (See Kuchler, 2013). It is believed that financial education (formal and informal), could ameliorate the presence of naïve behavior and present bias preferences. In this regard, and given the perceived existent gap between actual and perceived financial literacy, we propose to study borrowing practices of college students – and parallel knowledge on debt management – to determine the roles of trans-generational financial knowledge, and whether overconfidence and/or lack of financial literacy and knowledge are at the core of student borrowing practices and debt accumulation. We organize this paper as follows. The next section reviews the most relevant literature on financial literacy and debt management. The third section describes the research design, the estimation model and the data. Because of the nature of our data, we then provide a thorough interpretation of the Ordered PROBIT model estimations, and conduct corresponding impact analysis. The last section presents some general conclusions and policy recommendations.

2. Literature Review Financial education matters because it provides individuals with adequate tools and places them in a suitable position to conduct effective financial decisions. The literature is clear (see Borden, Lee, Serido, and Collins, 2008; Chen, 2005; Fox, Bartholomae, & Lee, 2005; and Martin & Oliva, 2001; among others) to argue in Our research acknowledges the possible implications in terms of self-selection by individuals 21 or younger to apply for a credit card under the new Act, but does not aim at addressing the possible implications vis-à-vis those that now cannot apply for credit cards because of not meeting eligibility criteria. We leave those considerations for future research. 1

For instance, Kim, Sorhaindo, and Garman (2006) suggest that personal financial stress could affect work performance and Jennings (2007) reinforces the need for higher education role in providing financial management to students. 2

In general, naïve individuals are believed to have a present bias for consumption that prevents them from fully incorporating future discounts appropriately and therefore failing to maintain effective repayment rates on borrowed money. This is to say, they tend to hold positive debt balances for longer periods and eventually fail to repay in full. 3

Journal of Personal Finance

favor of reinforcing financial responsibility and increasing financial literacy through active teaching and education programs that could prove useful in moderating students’ borrowing behavior, and prevent the escalation of debt accumulation early on in their lives. In this context, Martin (2007) found that there is a clear connection between financial knowledge of individuals’ behavior and financial education problems. There exists consensus as well on the importance of financial education as a key determinant to financial knowledge as indicated by Bernheim and Garrett (2003), Shim, Barber, Card, Xiao, and Serido (2010), Goetz, Durband, Halley, and Davis (2011), Volpe and Mumaw (2010), and Brown, Klaauw, Wen, and Zafar (2013), among others. However, research conducted by Fannie Mae shows that three-quarter of consumer financial literacy programs were not offered until early 2000 (Bernheim & Garrett, 2003). Despite the stated importance of appropriate financial knowledge, Kim, Sorhaindo, and Garman (2006) note “the great majority of secondary schools and colleges do not offer students courses in personal finance” (p.474). It is relevant to note that the implications of a lack of financial literacy/knowledge spread beyond ill-prepared individuals to make effective financial decisions. In particular, studies by Ornelas and Kleiner (2003), Chen and Volpe (2002), and Kim, Garman, and Quach (2005) bring forth the hypotheses that absenteeism, health issues, job satisfaction, and reduced productivity can result from issues relating personal financial stress. Because of its importance, investigating the causes and implications of how financial decisions relate to financial stress, education, gender, and other factors is a relevant and timely research topic. Most notably, studies such as, Chen and Volpe (1998 & 2002), Lusardi and Tufano (2009b), Lusardi, Mitchell, and Curto (2009a), Lusardi (2008), Peng, Bartholomae, Fox, and Cravener (2007), Wells (2007), Allen, Edwards, Hayhoe, and Leach (2007), Fiona, Wai-Lap, and Yin-Kwan (2012), Ford, Devoto, Kent, Harrison (2007), provide strong evidence on the issues of gender differences, academic performance, and the important role that financial education have on an effective financial decision-making process and overall debt management. More importantly, these studies provide robust insights into the potential benefits of increased financial literacy as a mechanism to lower and possibly eliminate unnecessary debt. The literature consistently shows that financial literacy provides individuals with better judgment regarding their finances (Allen et al., 2007; Bernheim & Garrett, 2003; Bowen & Jones, 2006; Carlin & Robinson, 2012; Chen, 2005; Chen & Volpe, 1998 & 2002; Mojtaba & Taihyeup, 2011). For instance, Carlin and Robinson (2012) demonstrate (using data from a controlled test group), that financial literacy can be taught, but more importantly, the manner and content taught potentially could have a more significant role in improving financial decision making. Also the evidence indicates that financially informed subject groups saved four times more than other non-informed subject groups, and demonstrate a delay in immediate gratification when compared with their counterparts. This is to say, financial knowledge leads to a reduction of present bias behavior as individuals become less naïve.

In a related study, Chen and Volpe (1998) addressed the issues of financial literacy levels and financial decision-making skills involving a sample of 924 college students from 13 different campuses. Chen and Volpe (1998) reported that 89% of students were able to make good spending decisions in a hypothetical situation if he or she had higher levels of financial literacy. In comparison only 68% of the students were capable of making adequate choices when (s) he had lower levels of financial literacy. Their study also found that demographic variables such as age, gender, class ranking, and work experience are significant predictors of better decision-making process regarding spending habits. On a subsequent study, Chen and Volpe (2002) focused on gender differences as a determinant of personal financial literacy performance, finding evidence that on average women know less about personal finance than men. In addition, Chen and Volpe (2002) find – as expected – that college students with a business major were more likely to know about personal finance than non-business majors. As a side note, however, the participants of this study indicated that they acquired financial knowledge primarily through informal channels – such as parent(s). This is something that we take particular interest in our study. Perhaps, more troubling is the fact that the subjects indicate a significant amount of his/her financial knowledge is not derived from formal educational channels. This may point out to a selective bias in favor of the business majors not because of the major itself, but because business students know about financial topics due to other unrelated reasons. Further evidence supporting the role and importance of informal channels of financial knowledge transmission is provided by Allen et al. (2007). They surveyed 1,293 college students about their interactions with their parents in regards to money and credit and found a strong positive interaction when parents worked together to form a financial plan. In a similar context regarding the value of financial literacy among high school and college students, Peng et al. (2007) measured investment knowledge, savings, financial education, financial experience, income and inheritance, and demographics, and found evidence in support of the hypothesis that students taking a class on personal finance while attending college improve investment knowledge. Bernheim and Garrett (2003) studied the impact of state mandated financial classes on students’ financial performance, and found that for states participating on mandated financial education programs, saving and overall wealth have a higher rate after five years than states who did not enforce mandated classes. In addition, Tennyson and Nguyen (2001) found that financial knowledge is enhanced if the mandate is not in the subject of “generic” financial education but more on specific topics on personal finance. Further evidence by Hung, Parker, and Yoong (2009), state that financial literacy does have a positive impact on financial decision making processes; however, the notable concept is the difference between knowledge and the behavior aspect of financial literacy. 4

Other supporting evidence to note is from DesJardins, Ahlburg, and McCall (2002) focused on how changes in financial-aid packaging affect students’ departure decisions over time and Heckman and Grable (2011) study on the shaping of financial self-efficacy of college students. 4

Volume 13, Issue 2

Another issue receiving significant attention in the literature relates to students’ self perception on financial knowledge. In general, the evidence points out that students believe they are of the sophisticated type, when in fact their behavior resembles the naïve more closely (See Kuchler, 2013). For instance, Wells (2007) studied why students were not concerned about mounting personal debt and the effects of such debt in their future. Wells (2007) found that students’ optimism was greatest about debt as freshmen and sophomores (naïve behavior combined with present bias preferences); whereas an observed reversal on optimism takes place for graduate students, seniors, and juniors (sophisticated type, and self aware of possible present bias preferences), as they grow more concerned about their accumulated debt. Roberts and Jones (2001) investigated the attitudes associated with how access to cash versus access to credit cards relates to compulsive buying. Using a sample of 406 college students – with an approximately equal distribution between females and males – they found support to the common belief that credit cards have a significant role in the compulsive buying of college students and more relevantly in individuals consequently carrying a month-to-month balance; that is, students were unable or unwilling to repay credit card in full at the end of the billing cycle. Furthermore, they note the existence of a close relationship between attitudes (power, prestige, distrust, and anxiety) and compulsive buying. In this context, Borden et al. (2008) conducted a pilot study on the influence that an hour and a half of educational seminar had on changing college students’ financial knowledge, attitudes, and behaviors toward the use of credit cards. Their findings suggest that participation in the seminar not only effectively increased the financial knowledge from pre-testing to post-testing, but also indicated that males demonstrated better financial knowledge than females. In a more comprehensive study Lusardi and Tufano (2009b) expanded the implications of financial illiteracy to the set of financial experiences of individuals, beyond the traditional behavior of debt management. Their research shows that low levels of debt literacy are the norm for elderly, women, certain minorities, and people at the lower income distribution brackets. In general, Lusardi and Tufano (2009b) note that those with less financial knowledge paid a disproportionately larger percentage of fees and finance charges than otherwise. The results also point out that only about one-third of the population seems to fully comprehend how compounding interest is calculated or even how credit cards function. Kuchler (2013) provides strong evidence that naïve type of individuals have a difficult time staying on track with ex-ante repayment plans on credit card balance because of a significant present bias, whereas sophisticated individuals are able to more accurately smooth out consumption-savings decisions and consequently achieve higher repayment rates on credit card balances. Naïve individuals are therefore more likely to carry a month-tomonth balance on their credit cards. Further evidence from Lusardi, Mitchell, and Curto (2009a) notes that only 27% of those surveyed possessed knowledge in basic5 Lusardi et al (2009a) pp8-9, use what is known as the big three questions to assess basic financial literacy. These questions measure basic knowledge on inflation, compound interest and risk management in stock purchases. 5

financial concepts. They conclude that most young adults are not equipped to make financial decisions. An interesting aspect to the research is that young adults receive financial literacy through two main sources, parents and the educational system. In particular, Lusardi et al (2009a) note that young adults with college-educated parents tend to have a better understanding of financial concepts. Another interesting result reveals that women and minorities show even less of an understanding of financial literacy (See Chen & Volpe, 2002 for more on these results). Furthermore, other studies provide evidence on the role that proper types of financial education play as a significant factor in achieving financial literacy (Mandell, 1997), while others indicate that much of the financial education is being conducted through business and community organizations, and not through educational institutes (Fox, et al. 2005). The review of the literature provides strong support to the argumentation that individuals who are more financially literate are better equipped to perform efficient and effective financial decisions. In addition the evidence indicates that parental influences, accessibility to financial education, and the mode and content in which the literacy is being presented are all relevant determinants of efficient financial decision making. However, the evidence also provides robust support to the fact that there are large and persistent gaps and a generalized lack of financial knowledge. Furthermore, females and individuals with low levels of financial experiences are at a disadvantaged position to make effective financial decisions. Our study provides support to some of the existent literature but it also aims at exploring a direct correlation between trans-generational knowledge and the effects of financial knowledge on debt accumulation and borrowing behavior. We also propose to study the relationship between how students self-report their financial knowledge vs. actual financial literacy and the implications for debt management.

3. The Data and Model Description Data for this research comes from a survey administered to a sample of 380 graduate and undergraduate business students attending a Pennsylvania Higher Education System University. For the purpose of this study, we define the dependent variable, as the amount of existing debt at the time of the survey. We use the following categorical classification, as the survey asked students to provide answers relating to the amount of debt they hold using the following ranges: 1=$0, 2=$1-$2,000, 3=$2,001-$5,000, 4=$5,001$10,000, 5=$10,001-$25,000, 6=$25,001-$50,000, and 7=greater than $50,000. We separate debt into three types: Student Loans, Credit Card Balance6, and Other Debts. It is also interesting to note that only 172 students hold a balance on credit cards7, while 299 In the case of Credit Cards, it is relevant to point out that the CARD Act of 2009, restricts access to credit cards for individuals younger than 21. A person younger than 21 can only apply for credit card if (s)he has a cosigner – like a parent – or can provide evidence of a steady flow of income, such as employment verification. In our data, any person younger than 21 years of age must fulfill any of these requirements, since all the information was collected between 2010-11. 6

The relatively low proportion of students holding credit cards could be the result of the CARD regulations and lack of cosigning and or verification of income status for those 21 years of age or younger. 7

Journal of Personal Finance

students have educational loans, and 295 students report holding other type of debt. We divide the list of independent variables in the following groups: demographics, behavioral, self-reported financial knowledge, financial experience, and financial literacy.8 Our list of independent variables aims at assessing the participants’ knowledge of debt management while trying to understand his or her current borrowing behavior, while controlling for standard demographic characteristics. The demographic variables serve as controls and provide the benchmark for our analysis. These variables allow us to account for possible differences in academic performance, academic status, age, gender, ethnicity, employment status, and academic major. Several of these variables have been identified in the literature review as having potentially important effects in the levels and differences on debt accumulation by students. We place particular emphasis on gender, academic status, and academic performance differences. Our intuitive expectation is that the estimated coefficient for academic performance will demonstrate the existence of a negative relationship with the level of debt. That is as student’s GPA decreases, (s)he would be more likely to hold higher levels of indebtedness. This would reflect a less than optimal understanding of the negative implications to holding debt. Regarding gender, the literature indicates that females are more likely to hold larger amounts of debt, particularly as it relates to credit card; therefore we expect to have a negative estimated coefficient for the gender variable, since we define male=1 and female =0. For the academic status variable, we expect a positive coefficient, indicating that levels of indebtedness increase as students advance in their college career. This should be particularly relevant for the student loan variable, where the increase in debt is a natural progression of academic progress. For the rest of the demographic variables, we do not have a predetermined expected sign for the coefficients, therefore we will argue as our null hypothesis that coefficients will not be different than zero; that is all individuals are affected across the border by issues relating to debt management. The behavioral category group includes two questions regarding individuals’ credit card frequency of use and corresponding repayment behavior. These variables serve as proxies to assess individuals’ time preferences (present bias issues), and potential issues relating to procrastination and lack of commitment, which may in turn be reflected in behavioral delays on credit card repayment. In the case of credit card use – as a proxy for actual decision-making – we use a categorical scale regarding frequency use of credits cards (Never=1, Seldom=2, Regularly=3, Often=4, Always=5), and regarding repayment behavior on credit card balance (Always late=1, Do not meet minimum payment=2, Minimum Payment=3, Payment in Full=4). In this regard we expect that as the frequency of repayment increases the level of debt will decline and it should be represented by a negative coefficient. Students that pay in full 8 A full

description of the variables is available in Table 1A in the appendix

would tend to suffer less of a present bias issue and therefore are more likely to balance their purchases to their disposable income, and consequently less likely to hold debt. This should be more prevalent in the case of credit card debt, than in the other types of debt. For the frequency of use of credit card, the null hypothesis would be that as frequency increases individuals are more likely to hold a larger amount of debt. That is the coefficient is expected to be positive. If the evidence were to reject the null hypothesis in favor of the alternative, then increased frequency of use of credit card would imply that individuals use the credit card, but make sure to also pay in full at the end of the billing cycle. In this case, an individual would be characterized as having perfect foresight and little to no present bias preferences. The underlining assumption is that individuals with demonstrated present biased preferences are more likely to hold a month-to-month balance in their credit card as they tend to value present consumption significantly more than future consumption. Under the self-reported financial knowledge and financial experience category we asked questions that allow us to assess students’ perceptions on these topics. We use either a Likert Scale where responses ranged from strongly disagree (1), disagree (2), neutral (3), agree (4), and strongly agree (5); or classified the variables as yes/no responses (Dummy variables). The self-reported financial questions are designed with the objective to assess how much students believe they know about personal finance issues, and how they achieve this knowledge. In this way we are able to measure the degree of experiential financial knowledge. We can then relate their perceived financial knowledge to their behavior as reflected by the levels of debt reported. Among the questions that students were asked in the survey we find: do you understand the cost of buying using a credit card? do you discuss financial issues with your parents? do you know the effects of only making minimum payments in your credit card? do you know what compound interest is? do you read economic articles on the newspaper? do you have a mortgage, car loan, or cash loans? among others. If students’ self-assessed financial knowledge is accurate, then the expected sign of the estimated coefficients is negative. That is, high levels of self-assessed financial knowledge would lead to lower levels of indebtedness, and low levels of financial knowledge to high levels. If students on the other hand, are overconfident and overestimate their financial knowledge, that is they are naïve, then the estimated coefficient would be positive; the more they report knowing, the higher the probability of holding large amounts of debt. Finally, the financial literacy questions assess actual financial knowledge making use of three questions9. The questions range from simple, to intermediate, to complex. These are the questions we use:

Lusardi et al 2009 use similar questions to assess financial literacy levels.

Volume 13, Issue 2

1) The difference between a bond and a stock is (Select the correct answer): a. Bonds are certificates of indebtedness and stocks of ownership b. Bonds are issued by firms and stocks by the government c. Stocks have fixed returns and bonds have variable returns d. Stocks provide yield for a fixed time period whereas bonds provide yields even after maturity.

2) If an individual invests $100 at an interest rate of 5% annually, how much would she have in ten years? a. $ 254.00 b. $ 1,050.00 c. $ 1,005.00 d. $ 163.00

Journal of Personal Finance

3) How much would an individual receive if she invests $100 per year for a period of 10 years at rate of 5% annually? a. $1,050.00 b. $1,025.00 c. $1,320.00 d. $1,005.00 The difference between financial literacy questions and the self-reported financial knowledge allows to assess actual gaps between what students believe they know vis-Ă -vis what they actually know. We then can compare the effects of both sets of variables, on financial management decisions as they relate to levels of indebtedness. Here, we expect a negative coefficient as our null hypothesis; the higher the financial literacy knowledge the lower the amount of debt. In other words, the more correct

answers a student achieves the higher her/his financial literacy level and therefore the lower the expected amount of debt (s)he would hold. Table 1A, in the appendix presents basic descriptive statistics and corresponding coding of variables10. Tables 1 and 1b below present the main statistical information by debt type and demographic variables, as well as the relationship between financial literacy, trans-generational knowledge and self-reported financial knowledge11. 10 A copy

of the survey is available from the authors upon written request.

11 We thank an anonymous referee for the insightful suggestion to expand Table 1 and to add Table 1b.

(continued)

ÂŠ2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 2

ned above, we code students’ responses on the outstanding debts (Student Loans, Credit

The model (4a) d Other debt) using a discrete categorical variable; therefore the most appropriate model As explained above, we code students’ responses on the outstand(4b) (Student Loans, Credit Card model. Balance,This and model Other debt) mations isingandebts ordered dependent variable allow us to compute the using a discrete categorical variable; therefore the most appromodel may to usehold in our estimations is anrange. ordered ilities thatpriate a student debt in a given Independent addition the use of an ordered (4c) variable model. This model allow us to compute the expected probabilities that a student mayof hold debt inofa independent given range. Invariables on the discrete will allow us to determine the effects a series addition the use of an ordered dependent model will allow us to In this regard, to obtain the marginal effects of changes in the determine the effects of a series of independent variablesto on obtain the In this regard, the marginalvariables effects we of require changes the explanatory variables we re explanatory theinstandard normal cumulative debt amounts. The basic model description has the following general specification: discrete ordered students’ debt amounts. The basic model descrip- distribution function evaluated at and respectively. According to tion has the where arefollowing assumedgeneral independent and identically distributed random variables as is usual, is probability the the marginal effects function, the of occurrence ofrespectively. one standard normal cumulative distribution function evaluated at matrixand A ere are assumed independent andspecification: identically distributed random variables as usual, the matrix particular assumed independent and identically distributed random variables as usual, is categorical the matrix response increases, assuming that the corwhere are assumed independent and identically distributed random variables as usual, is the matrix responding is (1) positive, as value of theto variable is unobserved. According of explanatory β is the vectortoofthe coefficients toeffects be (1) estimated and function, the βprobability ofthe occurrence ofexplanatory one particular categorical is unobserved. According to explanatory variables, variables, β is the vector of coefficients tomarginal be estimated and increases. In other words, as x increases then the probability of is unobserved. According to variables, β is the vector of coefficients to be estimated and where are assumed independent and identically distributed is values unobserved. According to In the reverse case, of explanatory variables, β is the vector of coefficients to be estimated andhigher achieving of y increases as well. Greene (2003) what one does observe is of explanatory increases, assuming that the corresponding β is positive, as the value of the explanatory variable random variables as observe usual, the matrix eene (2003) what one does is is when β<0 then as x increases the probability of achieving lower ) what one does observe isvector of coefficients to be estimated and variables, β is thewhat is values of y increases. Furthermore, in our case an increase in the Greene (2003) one does observe is unobserved. According to Greene (2003) does as observe x increases then thevalue probability of increase achieving values y increases as w In what otherone words, probability reflects an in higher the overall debtoflevel as is (2) (2) revealed by students’ responses. Therefore positive β coefficients (2) then (2) x increases the probability achieving reverse case, when β<0 as that indicate higher survey values (higher levels) lower increasevalues of y (2)of debt ey is available from the authors upon written request. ymous referee for the insightful suggestion to expand Table 1 and to add Table 1b.overall outstanding debt. Marginal effects for dummy variables as Work andprobability Gender are value computed using probabilities Furthermore, in our case ansuch increase in the reflects anthe increase in the overall deb that result when the variables take the two different values and . the restTherefore of the variables areβheld constant at their mean . revealed by students’ responses. positive coefficients indicate that value. higher survey valu . : In general, according to EViews-Manual, “the sign of βj shows : . the direction of the change in the probability of falling in the : : debt levels) increase overall outstanding debt. Marginal effects for dummy variables such as W end-point ranking (y=0 or y=1) when xij changes. changes in the opposite of theassociated sign of βj and areparameters unknown parameters to be estimated with .that It follows that the direction probabilities withchanges in the same where are unknown to be estimated with are . It computed follows the probabilities associated withwhen the ere using the probabilities that result variables take the two differe Gender direction as the sign of βj.” (p 653). unknown parameters to be estimated with . It follows that the probabilities associated with are unknownparameters parameters estimated . It follows that the probabilities associated with where where are unknown to to be be estimated withwith β. achieving event are givenassociated by ieving each eventeach are given by It follows that the probabilities with eachvariables are held constant 4. Results andmean Discussion and theachieving rest of the at their value. In general, according to h event are given by event are given by achieving each event are given by We perform three sets of estimations of the Ordered Probit Manual, “the sign of βj shows the direction of the change in the probability of falling in the Model, as reported (3a) 2, 4, and 6. The individual estima(3a) in Tables (3a) tions allow us to analyze the determinants of different types of (3a) (3a) changes in the opposite direction of the sign ranking (y=0 or y=1) when xdebt ij changes. (Student Loans Value, (3b) Credit Card Balance, and Other Debt) (3b) respectively.(3b) We separate debt by type based on the assumption (3b) (3b) 653). and responds changes in thethat same as the signaof βj.” (p purpose eachdirection type of debt serves different (3c) (3c) to different economic decision-making processes. In each of the (3c) (3c) estimations, we organize(3c) the results as follows. The first set of estimates (Model 1) includes only demographic variables. We Underthe thethat condition that that Under condition ,, then thenall allprobabiliprobabilities will Inofthe of an ordered thenbe proceed to include the additional three categories of varider the condition , then all probabilities will be positive. Inpositive. the case ancase ordered ties will be positive. In the case of an ordered probabilities model, dition that , then all probabilities will be positive. In the case of an ordered probabilities model, it is relevant to keep in mind that in order to provide a meaningful interpretation of the ables, behavioral, self-reported financial knowledge-experience, Under the condition that all probabilities will be positive. In the case of an ordered AND DISCUSSION 1. , then RESULTS it is relevant toneed keep in mind that inthe order to provide a meaningful and financial literacy, in the subsequent estimations. results, we first to transform estimated coefficients from equation (1) into marginal effects. babilitiesprobabilities model, it is relevant in mind thatininmind orderthat to provide interpretation of the model, ittois keep relevant to keep in order a to meaningful provide a meaningful interpretation of the of the results, first toto transform estimated odel, itConventional isinterpretation relevant to use keep mindwe that in need order providethe a ismeaningful interpretation of the of in Ordinary Least Square estimates not appropriate when using discrete effects. (ordered) data; results, we first need to transform the estimated coefficients from equation (1) into marginal coefficients from equation (1) into marginal effects. Conventional Before we proceed with the econometric we would like is the estimated coefficient are not regular coefficients in the sense ordinary estimations with Probitanalysis, We perform sets ofof of theeffects. Ordered Model, aspredomireported in Tables 2 ults, wethat first to use transform the estimated from equation (1)estimations into marginal Conventional of Ordinary Least Square estimates is notthree appropriate when using discrete (ordered) data; use ofneed Ordinary Least Square estimates is notcoefficients appropriate when to explore the data. First, Student Loans are the most st needcontinuous to transform the estimated coefficients from equation (1) into marginal effects. dependent variable. 6.regular The individual estimations allow analyze thehold, determinants of seen different types of debt (Stude that the estimated coefficient are coefficients sense ofusordinary estimations with usingisdiscrete (ordered) data; that is thenot estimated coefficient arein the nant type oftodebt students as it can be in Table 1. We Value,estimations Credit Balance, and Other Debt) respectively. Wetypes separate type based on th nventional of Ordinary Least is not Card appropriate when using that discrete (ordered) notuse regular coefficients invariable. theSquare sense ofestimates ordinary with continuous dependent also observe students have other of debtdebt and by in general use of Ordinary Least Square estimates is not appropriate discrete (ordered) continuous dependent variable. Incidentally, in the Orderedwhen Probitusingkeep balances greater than zero. While a balance greater than Model the marginal effects of x on the probabilities of occurrence zero is expected in student loans a; that is the estimated coefficient are not regular coefficients in the sense of ordinary estimations withas they are long term loans, a e estimated coefficient are not regular in the sense of ordinary estimations with of each of the events are not equalcoefficients to the coefficient estimates β as month-to-month balance in credit cards has different implications in the conventional LSIncidentally, estimations. To the marginal effects regarding individuals’ x on theas we will see later. Also, no tinuous dependent variable. in obtain the Ordered Probit Model the marginal effects ofbehavior we need toIncidentally, take the partial of theProbit probability x on theis observed between males and females in pendent variable. in derivative the Ordered Modelfunction the marginalstatistical effects of difference with respect to the vector of parameters x. Thus, the marginal analterms of general borrowing practices. When we separate the data babilities of occurrence of each of the events are not equal to the coefficient estimates β as in the yses of changes in the independent variable(s) on the probabilities into otherβdemographic f occurrence of each of the events are not equal to the coefficient estimates as in the classifications we observe marked differin the case of three categories are given by: ences. For instance, older people report holding larger balances in ventional LS estimations. To obtain the marginal effects we need to take the partial derivative of the S estimations. To obtain the marginal effects we need to take the partial derivative of the

bability function with respect to the vector of parameters x. Thus, the marginal analyses of changes in

Journal of Personal Finance

all debt categories, with the Age group 5 holding debt in excess of $25,000 for student loans and $10,000-$25,000 for other types of loans. However we also observe a mean reversal for the much older students (non-traditional in group 6 – older than 37) whom hold less student loans and overall less debt than some of their peers. This could be the result of non-traditional students already holding a permanent job, and consequently having less need to borrow and more capacity to repay. It may also mean older students are receiving other financial help through employers or government agents. Regarding academic status student loans balance increase with seniors and masters students having the highest level – high category 4 and high category 5 respectively. However, while controlling for academic status, this increased relationship does not replicate when looking at reported credit card balance and other type of debt. All students regardless of class standing hold about the same amount of debt – greater than zero – indicating that credit card use is extensive and generalized across all groups. Analysis by academic performance (GPA conditional) reveals a negative relationship with debt holding; i.e., students with higher GPA12 have on average less debt. While we cannot state with certainty, we can hypothesize in favor of a trans-generational effect, where students performing academically better have less need to borrow as a result of coming from more affluent families that have probably invested more in education (See Allen et al., 2007). In addition, one may expect that better academically prepared students are more likely to receive scholarships, and therefore have a lesser need for student loans. Furthermore, students with a higher GPA also hold less credit card balances, possibly indicating to a positive relationship with higher financial literacy13. Finally, if we take a look at the repayment behavior on credit cards, we observe that while the majority of those using Credit Cards do pay the balance in full every month14, there are approximately 25% that only pay minimum and a small number that do not meet minimum repayment requirements, as it is represented by a mean of 2.024 for the 18-20 years of age group. Now, if we look at selected independent variables in the subgroups Behavioral, Self-Reported Financial Knowledge, and Financial Literacy, we observe the following. First, the initial analysis supports the claim that students who only make minimum payments are more likely to carry higher amounts of debt.15 The second subgroup – Self-Reported Financial Knowledge – shows that students reporting to discuss financial issues with their Notice in Table 1, that are coding for GPA indicates that lower code values refer to higher GPA, that is as coding values increase, GPA decreases. 12

13 For a further and more elaborated discussion on the issue of actual financial literacy and student academic performance, please see …We have omitted this information for the review process to stay anonymous. 14 In this regard, it is relevant to notice that for an unspecified number of credit card holders (21 or younger), it is quite possible that payments on credit card balances are made by the cosigner, normally a parent. We cannot differentiate this behavior as our data does not allow for this classification separation. 15 Notice that data on repayment behavior is limited to the categories, minimum payment and payment in full, as the other categories only have one observation.

parents tend to hold lower amounts of debt than otherwise. On the other hand, those reporting to have higher levels of financial knowledge – as measured by the variables CINT, and HIST – tend to report having higher debt levels. The data indicates the negative effects that overconfidence may have on debt. For the Financial Literacy subgroup, the data indicates that lower levels of financial literacy result on average in higher levels of debt as expected. Table 1b further decomposes the data to better understand the relationship among financial trans-generation knowledge and selected variables in the Behavioral, Self-Reported Financial Knowledge, and Financial Literacy subgroups. When looking at the mean value, higher values indicate that students report discussing financial issues with their parents, and therefore the trans-generational financial effect is present. Mean data for the Behavioral category indicates that those reporting making only minimum payment on their credit cards hold a lower mean than those making payments in full. This is, lower repayment rates are associated with lower levels of interaction with parents, and consequently with less financial knowledge as reflected by larger month-to-month balances. For the self-reported financial knowledge subsample, the mean values indicate that those reporting to know how compound interest works (CINT value of 1), hold a lower value than otherwise. In other words, they have less interaction with their parents, but report to know more. This same result holds for the HIST variable, yet the difference in mean values is small. Finally, the financial literacy subsample mean scores indicate that those answering correctly also have a larger interaction with their parents regarding financial issues. This provides support to the hypothesis that a trans-generational financial knowledge effect improves financial literacy. Upon completion of the estimations, and as indicated earlier,we need to transform the β estimates into corresponding probabilities in order to obtain meaningful results. Coefficient estimates (β) reported in Table 2, 4 and 6 need to be interpreted with caution as they only indicate the direction of the effect and the relative significance across them; yet they are not directly meaningful in absolute terms in changes of probabilities. Given the cumulative , and with some significant amount of normal function computations, the corresponding probabilities evaluated at the average value of all explanatory variables (continuous) and using the corresponding estimated coefficients, along Equations (3a-c) are reported as noted in Table 3, 5, and 7. Operationally, we compute the marginal effects using the average value for each of the independent (continuous) variables. For simplicity, we select only the most comprehensive model in each of the estimations of the marginal probabilities, normally the last model reported in the estimations in Tables 2, 4, and 6. Thus, to illustrate how we perform the computation, let us take the case of Financial Literacy as reported by interaction with parent regarding financial issues (Trans-generational effect). Here, we use the average values for the described variables as reported in Table 1 (Descriptive Statistics). Using the corresponding β estimates from Model 4 in Table 2, the resulting =-0.6687

Volume 13, Issue 2

and the corresponding =-1.5623 for the first limit, and -1.4136, -0.5251, 0.1115, 1.1653 and 2.0921 for the rest of limits.16 We repeat this computation for each of the alternative models and each variable17. We compute values for the marginal effects using the normal cumulative probability function and report them in Tables 3, 5 and 7 respectively. In addition, in the case of dummy variables, we follow a similar procedure, yet we evaluate the cumulative function at the two values (0,1) and then take the difference between them to obtain the marginal effect of changes in the discrete variable, while holding the rest of the variables at the average level. Finally, notice that as a cross check of the accuracy of the estimated model and consequent probabilities, the sum of the marginal effects across all probability levels equals zero. This is the expected outcome because all limit values are increasing for the combined models, and therefore changes in the independent variable must result in probability changes that balance out.

16 Limits for Ordered Probit Models estimates are reported in Table 2A in the appendix. 17

The curious reader can perform alternative computations to validate our results.

Once the marginal effects are computed, we then use both the coefficients and marginal effects to conduct the empirical analysis in the next section. 4.1 Student Loans We study first the determinants of student loans (See Table 2). According to the information provided in the introduction, it is almost expected that a student will have to seek some form of student loan18 to finance her/his education. Because of this, our interest is to determine the probability of holding more or less debt as it relates to a set of demographic variables, two behavioral variables, self reported financial knowledge indicators, experience with financial instruments, and actual financial knowledge measures. 4.1.1. Controls Our exploratory estimates provide preliminary evidence indicating that academic status is positively related to studentsâ&#x20AC;&#x2122; loans and it is statistically significant at the 1% level. This is to say that 18

In our sample only 27% of the subject reported not having Student Loans.

Journal of Personal Finance

upper classmen have a positive probability to carry larger student loans as they spend more years in the education system (See Table 3 for estimated marginal effect on Probabilities). However, while the evidence confirms the expectations, students on average will hold about $10,000 in student loans at the end of the bachelorâ&#x20AC;&#x2122;s degree. MBA student will hold debt in the range of $10,000 to $25,000. Because most students do not hold currently a steady

job, we estimate that the projected debt amounts are high and most likely beyond what a student could currently manage effectively. For the remaining demographic variables, the estimates do not provide any statistical significance; that is students in our sample are affected basically in the same way. Our evidence does not provide support to the discussion provided in the literature review regarding a significant gender difference.

ÂŠ2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 2

4.1.2 Behavioral We now turn our attention to the effect that the behavioral measures (How does a person pay her/his credit card balance, and what is the frequency of use of credit cards) might have on student loans. In this case, it is interesting to note that both variables demonstrate a negative and statistically significant coefficient, confirming our hypotheses. That is, as the frequency of credit card repayment increases19 (from paying only minimum payments and finally to paying all balance in full every month), the probability of holding lower levels of student loans also increases. The coefficient for credit card repayment holds about twice the magnitude compared to credit card frequency of use (Table 2). In addition, the corresponding marginal effect of credit

19 It is relevant to remember from our discussion in the data section that the first two categories â&#x20AC;&#x201C; always late=1 and less than minimum payment=2 â&#x20AC;&#x201C; only received one response each. Therefore, we eliminated them from the analysis and technically the HPAY variable is treated like a dummy variable, where the only two forms of behavior are minimum payment, and payment in full.

card repayment holds about twice the size in relation to frequency of credit card use (Table 3). Intuitively, this could be interpreted as, when students are capable of making payments in full every month and therefore not hold a month-to-month balance, then the probability of holding larger amounts of student loan decreases more than by corresponding increases in the frequency of use. In other words, students that are capable to maintain low to no month-to-month balance in their credit cards, are also more likely to hold lower levels of student loan debt. While it is recognizable that having zero student loans is very difficult for the typical student, our evidence indicates that those students that are capable of having a high probability of repaying the credit card in full every month are also more likely to keep student loans at lower levels than their peers who might have present bias issues and suffer of lack of control (as reflected by month-to-month balance in their credit card). This result is important as it implies an estimated probability of holding lower level of student loans increasing by 5.5% with every improvement in repayment practices (minimum payment vs. payment in full) on credit cards. Incidentally, the

Journal of Personal Finance

probability of holding higher levels of student loans decreases by an accumulated 18% (Loan categories 5-6-7 combined). In addition, high frequency of credit card use has the same effect on student loan balance (as credit card payment behavior), with an estimated probability of half of the repayment behavior. This evidence indicates that on the positive side, students with better repayment history have a superior knowledge of debt management and therefore are more likely to borrow less and maintain a good credit record history. This is, they are characterized by having less present biased behavior and better self-control and display more dynamically consistent preferences. The tentative conclusion at this point is that as long as students maintain a high repayment rate of credit card balance (pay in full every month) they would maintain lower levels of overall debt. 20 4.1.3 Self Reported Financial Knowledge and Financial Literacy Let us now turn to study the effects of financial literacy on debt management. Here we approximate financial literacy in two ways: perceived (self-reported) and actual. In the optimal scenario, one would expect that there is a high correlation between the two types of measures, and therefore a reinforcing effect among them in regards to debt management. In the lesser than optimal scenario, perceived knowledge is higher than actual, and individuals behave as if they know more than they actually do. Finally, in the worst case scenario, when both perceived and actual are low, then students would be in an uncomfortable and unprepared situation to make adequate decisions regarding debt management. Recall that the survey asked students to self-assess their level of financial knowledge (see Table 1A, items D). Correspondingly, our estimates for self-reported financial knowledge provide some preliminary and interesting results. First, let us state that if self-reported financial knowledge is accurate (i.e. provides a true reflection of what students do know about managing financial information), the estimated coefficients should be negative. Alternatively self-reported knowledge would overestimate the actual knowledge. In this regard, our estimates in Table 2, for knowledge on how budgets work, and trans-generational financial knowledge – as reported by financial literacy derived from informal instruction with parents21 – provide the expected negative coefficients and are statistically significant, indicating the more students report knowing on each of these two categories the higher the probability to hold lower levels of student loans. We have particular interest on the results derived from the trans-generational knowledge as it provides evidence supporting previous findings in the literature.22 This result indicates that parents play an important and significant role in assisting students to manage their finances. This is a relevant result because it provides strong 20 It is relevant to keep in mind of the possibility that for some individuals in the 21 year or younger category, repayment on credit card balances could being made by a third party, namely a cosigner, such as parents. As noted, we cannot provide a full separation of effects given data restrictions. This issue is beyond the scope of this study. 21 See Chen and Volpe (2002) and Allen et al. (2007) on this topic. Also see Cavanaugh (2013), Jorgensen (2010), and Martin and Oliva (2001) on the issue of trans-generational financial knowledge.

evidence in favor of early financial education as a mean to reduce or maintain low levels of indebtedness. The estimated coefficient for the trans-generational effect is the largest of all coefficients (-0.8 and statistically significant at the 1%) and represents as well the largest marginal effect of all variables used in the estimations. As we will see later, this positive effect from parents onto children is also present and statistically significant as it relates to credit card debt. In particular, the estimated probability of holding low levels of debt is 9.5% per increment in the self-reported financial knowledge acquired from parents. As we look at some other variables capturing self assessed financial knowledge, we observe that students tend to display a naïve behavior through the overestimation of their actual level of financial knowledge. Students report having high levels of knowledge about financial/investment options and understanding on how compound interest works. However, these variables have positive and statistically significant coefficients, and consequently resulting in lower probabilities of holding low level of student loans. Clearly, the magnitude of the coefficient (and corresponding probabilities – see Table 2 and 3), and statistical significance, provide evidence to conclude that students tend to overestimate their knowledge and consequently are more likely to borrow more than they should. This behavior is based on a false perception of their actual financial knowledge. In particular, lack of knowledge of what compound interest23 is, with an estimated coefficient of 0.66 and marginal effect almost as large (but in the opposite direction) as the trans-generational knowledge effect, indicates a significant overestimation of financial knowledge. This overestimation results in a probability reduction of 7.7%, 2%, 13% and 3% to hold lower levels of student loans for the lowest four categories respectively; and consequently increases the higher end student loan levels, with marginal effects of 12%, 10% for the upper categories of 5 and 6. Students’ low levels of preparation on key financial concepts24 and the incorrect overconfidence demonstrated in actual knowledge leads to students borrowing higher levels than they should. This becomes more relevant, as student loans are long term loans, and therefore the added cost at the margin could amount to large sums of money. In other words, self-reported knowledge on how compound interest works, yields the unexpected sign, and reveals that students overestimate their knowledge in a key financial variable. As noted by Lusardi et al 2009a, lack of knowledge on compound interest puts individuals in a highly disadvantageous position to make effective and efficient financial decisions.

22 See Lusardi, Mitchell and Curto (2009a) for more detail in the parent-youth relationship regarding transmission of knowledge. 23 This lack of knowledge in regards to compound interest is also evident in the low level of correct responses observed in the questions designed to measure actual financial literacy. Average results on these two questions could be found in the summary statistics Table 1A in the appendix. 24 Lusardi et al 2009 find that a large segment of the population has severe issues understanding the concept of compound interest.

Volume 13, Issue 2

Estimates on any of the financial experience variables do not have a statistically significant effect on student loan levels. Finally, while our measures of financial literacy have the expected sign (higher financial literacy reduces overall debt holding), they are not statistically significant. Because of the importance stressed in the literature review section, this is an area in which further research is necessary. Intuitively, formal education on financial topics should result in higher levels of financial literacy, which

in turn should reduce individualsâ&#x20AC;&#x2122; propensity to over borrow, and lead them to learn how to control any possible present biased tendencies. This hypothesis is still subject to empirical verification. 4.2 Credit Cards Before we proceed with the empirical analysis in this section, let us remember that intuitively effective and efficient use of a credit card is a relatively simple concept: you use the card

Journal of Personal Finance

to complete a purchase with the expectation to repay in full at the end of the billing cycle; therefore when the billing cycle is up, you fulfill your expectations and pay in full the balance. By doing this, any individual restricts consumption to her/his budgetary constraint, and incurs no additional fees or interest charge. Any behavior that deviates from this practice, results in added costs to the ticket price of the item/service purchased, and consequently reduces individualâ&#x20AC;&#x2122;s utility and overall welfare levels. In addition, carrying a balance in a credit card reflects issues relating to lack of commitment, and present-bias preferences. The evidence indicates that a month-to-month

balance has unfortunately become one of the main problems that individuals face currently (See information presented in the introduction). We have particular interest in these estimates as credit card debt is a major problem affecting a large segment of the population. An early detection of credit card misbehavior use could have large positive effects on consumersâ&#x20AC;&#x2122; wellbeing. Therefore, our second set of estimates addresses the issue of credit card balances. Results from the estimations are in Table 4 and corresponding marginal effects on Table 5 respectively. Table 4 presents four alternative models. Corresponding estimated probabilities are given in Table 8.

ÂŠ2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 2

4.2.1 Controls In this section we want to place some emphasis on a few key findings out of all possible estimations. The first two models in Table 4 reveal a very interesting result as the coefficient for GPA is statistically significant and negatively related to low level of credit card balance holding. Specifically, this result implies that students with a lower GPA will hold larger credit card balances. The estimated coefficient have an approximate value of 0.23 and 0.17 in Models 1 & 2 (Table 4). Notice that as GPA declines, the marginal effects (Table 5) indicates that the probability of holding a balance of zero decreases, while the probability of carrying a month-to-month balance increases for all other levels of debt. This result is interesting per se and raises additional questions as to why lower GPA students have to use credit cards as a mean to finance expenses beyond their budgetary constraint.25 On the one hand, higher holding GPA students know better about the possible negative effects of credit card use and holding a month-to-month balance, as they have demonstrated a superior understating of key financial concepts â&#x20AC;&#x201C; such as how compound interest works â&#x20AC;&#x201C; over their peers. This is implicitly reflected by their higher GPA. On the other hand, student with lower GPA, may rely more heavily on credit card use (and consequently maintain larger month-tomonth balances) as a mean to finance their expenses as a result of lack of knowledge on the possible negative effects of this behavior. Students possessing these characteristics consequently incur in over-indebtedness and place themselves in an uncomfortable financial situation.26 These hypotheses are beyond the current scope of this study, but we believe that they are worth further exploration in future research. For now, what we do observe is a negative relationship where, lower GPA implies a larger monthto-month Credit Card balance. The rest of the control variables do not provide evidence of statistically significant differences across groups; that is all individuals may be affected in about the same way regarding the use of credit cards. This is very similar and consistent with the evidence provided under the Student Loans section. 4.2.2 Self-Reported Financial Knowledge and Literacy From the previous section we brought forth the hypothesis that the working mechanism linking GPA and higher month-to-month balance on credit card may be related to issues regarding financial illiteracy. We explore this hypothesis in further detail here. Our hypothesis is partially supported by the negative (expected) and statistically significant coefficient for Financial Education through parental conversations about financial issues. As it was the case for student loans, self-reported financial knowledge acquired through the trans-generational effects is negative and statistically significant at the 1%. The magnitude of the

25 Of course, we do not have enough data to test the alternative hypothesis of whether, students carrying a month-to-month balance also have savings accounts. This behavior has been observed in previous research, but it is beyond the scope of this study.

While we do not have supporting evidence, one would expect that lower GPA holding students will also end up having relatively lower paying jobs, further complicating their financial position should they over borrow while studying. Of course, this is just a hypothesis in need of empirical verification. 26

trans-generational effect holds about the same in comparison to the student loans estimates. The estimated coefficients in models 2 and 3 (Table 4) are -0.793 and -0.637 respectively. We interpret this as an indication of the robustness of parental education regarding financial issues. More interesting is the evidence from the marginal effects analysis (Table 5) which indicates that an increase in self-reported Financial Knowledge through the trans-generational effect has the second27 largest positive effect to achieve a balance of zero in credit cards. The interaction between students and their parents increases the probability of holding a zero dollar month-to-month balance in their credit card by an impressive 15.2%. This result is very relevant as students that learn from parents about financial issues are significantly less likely to carry credit card balances and most likely make full payments on balances on a regular basis. As noted, Financial Education through interaction with parents plays a major role to effectively learn how to manage borrowing and consequently keep lower credit card balances. In fact the positive marginal effect of trans-generational financial education is larger for credit cards than for the student loans. Because of the significant emphasis placed in the literature, we take interest in the measure of financial literacy. Our measure of financial literacy (see questions discussed earlier) yield the expected negative sign, indicating that students possessing higher levels of financial literacy are more likely to hold no monthto-month balance in their credit card. The estimated coefficient in Model 4 Table 4, is -0.223, and represent a possible 5.3% marginal effect. However, this relationship is not statistically significant at conventional levels (only marginally significant at the 15%). Nevertheless, we believe that the lack of significance is due to the low level of correct answers provided by the students. In this regard, the evidence indicates that very few students are capable to answering all financial literacy questions correctly. In fact, only 10% of all surveyed students were capable of correctly answering the three questions. This in itself is a worrisome fact and speaks of the need to increase formal instruction regarding basic financial concepts. For the most part, however, we believe that the evidence supports our hypothesis that financially literacy reduces credit card balance holdings. Notice however, that when we introduce the financial literacy question, GPA is no longer statistically significant, despite maintaining the expected sign. Intuitively, students capable of answering correctly more financial literacy questions can overcome any possible deficiency as indicated by a lower GPA. The rest of the self-reported financial knowledge variables are not statistically significant at any conventional levels, irrespectively of whether they have the expected sign or not.28 The only variable holding a statistically significant coefficient is the variable 27 The highest positive effect comes from an increase in repayment in credit cards at 18.8%.

The inclusion of additional financial experience variables such as Car Loan (DCAR), Mortgage (MRTG), and Cash Loans (CASHL), results in a significant reduction in the number of observations. None of these variables are significant. The rest of the variables hold about the same level of economic and statistical significance.

Journal of Personal Finance

measuring knowledge regarding the effects on credit history (HIST) of only making minimum payments in the credit card.29 In this case, studentsâ&#x20AC;&#x2122; responses result in an unexpected positive and statistically significant coefficient, indicating that students have overconfidence in their knowledge and therefore are more likely to hold a positive month-to-month balance in their credit cards. The size of the coefficient is impressively large at 1.351 with a corresponding marginal effect of -32% against holding no month-to-month balance, and an 18.6% increase in holding a balance in the range of $1-$2000 (Table 4 Model 4 and Table 5). This result is counterintuitive as students reporting to know the negative effects of only making minimum payments in the credit card, tend to carry a month-to-month balance. If indeed a subject fully comprehends to possible negative implications of only making a minimum payment, (s)he should pay in full every month. 4.2.3 Behavioral Furthermore, our estimates find economic and statistically significant relationships between repayment frequency and credit card balance, as expected. Particularly, and this should not come as a surprise, high frequency of repayment (HPAY) has the single largest effect on reducing credit card balance (-0.788 in Model 4-Table 4), followed very closely by self-reported financial literacy through informal education with parents. While this concept may seem common knowledge, the facts expressed in our statistical analysis indicate that not all students follow these simple rules/suggestions. In particular, the marginal effects analysis in Table 5 indicates that an increase in the repayment behavior from minimum payment to payment in full will increase the probability of holding zero balance in the credit card by 18.8%. The moral of the story is simple, if you want to eliminate your credit card debt, pay it in full. It is surprising how simple this concept is yet it is not understood and practiced by all. Repayment differences correspond to behavioral differences between naĂŻve and sophisticated type of individuals as explained earlier. Therefore, those holding a month-to-month Credit Card balance in excess of $0 (Zero), are living beyond their means and incurring in lesser than optimal financial decisions. This is particular true if we consider that most students in our sample are still very young to maintain balances on credit card. As noted by Lusardi and Tufano (2009b), credit card payments, other than payment in full, result in accumulated fees and finance charges, consequently increasing the cost of purchases incurred through credit cards. 4.3 Other Debts The last set of estimates look at the other debt category. Our preliminary results continue to provide robust evidence in support to the argument that frequency of payment on credit card balances is a major determinant of debt management. In this regard, as frequency of payment increases the probability to hold lower amounts of other debt increases. We continue to explore this relevant result, because it provides further evidence in support to

The actual question is: Do you understand the effect of just making minimum payments on your credit card, which can lead to a negative payment history and compound interest payments far exceeding the original balance? Answer code: Yes=1, No=0.

the hypothesis that individuals that maintain realistic expectations regarding their time preferences in terms of debt management, indeed are capable to maintain overall lower levels of debt, and therefore behave more responsibly regarding their finances. We propose that those with a high repayment rate are more likely to have dynamically consistent preferences, have less present bias problems, and consequently do not suffer from lack of commitment and self control issues. In addition, we also find that our trans-generational variable has a negative effect on holding larger amounts of other debt, this is to say that receiving financial knowledge from parents allows students to have a higher probability to accumulate lower amounts of debt. The trans-generational effect proves to be robust and consistent across all three types of debt analysis that we perform in this study. However, the size of the coefficient (-0.56) and corresponding marginal effect 0.05 of the trans-generational variables are the lowest of the three types of debt (Model 3 in Tables 6 and 7). Furthermore, we find that those students that report holding a job have a higher probability to accumulate more debt. In addition, AGE is related positively and statistically significant to the amount of debt in other categories. However, the statistical significance of these variables is dependent on the model specification. In general, the results of this third set of estimations are less robust than the previous two. A possible explanation for this lack of consistent results could be that the survey includes too many types of debt (Car, Mortgage, other) in one collapsed category. Further research regarding the other types of debt is necessary, our research at this point cannot provide any further information, leaving us with more questions than answers at this time. 4.4 Projected Probabilities by Type of Debt We conclude our statistical analysis by looking at Table 8 regarding the predicted probabilities for each debt category. Results in this table are computed using the most comprehensive of each models as originally reported in Tables 2, 4, and 6, and therefore use the corresponding marginal effects from Tables 3, 5, and 7. Based on our exploratory results, we are able to predict that students have a 58% probability of holding student loan balances between $5,000 and $25,000, and a 33% probability to have debt in excess of $10,000. In addition, there is a 10% probability that students will have student loans amounting to $25,000-$50,000, and a very small probability (1.8%) to fall in the $50,000+ range. Less positive, there is also a very small probability of about 8% that students will hold student loan in the lowest bracket of $0-$2,000. This is to say, that based in our empirical evidence, students will finish their studies already holding a significant amount of debt in student loans. The estimated probabilities are in line with the amounts reported initially in the introduction, confirming the validity of our estimates and corresponding projections. This is a concern given that most of the students will enter the work force already facing a significant challenge (See Martin & Oliva, 2001 for similar results). This is more relevant if we consider that the main determinants of student loan management are related to their performance in school, and financial literacy obtained from their parents (See Jorgensen & Savla,

ÂŠ2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 2

2010; Serido, Shim, Mishra, & Tang, 2010; and Shim et al., 2010 for a related issue on Financial Socialization). With regards to credit card balance, our empirical results indicate the existence of an estimated probability of 15.5% that no month-to-month balance will be carried out to the next period, i.e., students will pay in full every month. This proportion of

the students will be characterized as possessing near perfect foresight regarding their financial management, and not having present-bias issues leading to procrastination in credit card repayment, and therefore being able to limit consumption to the available disposable income. However, there is an overwhelming 79.6% probability that students will have a month-to-month credit card balance in the range of $1-$2,000. Also, there exists

Journal of Personal Finance

ÂŠ2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 2

a smaller projected probability of 3.6% that the balance will be between $2,001 and $5,000. Finally, with regards to other debt types, it appears that about half of the students will not carry any other debt, while the other half will have debts in the range of $1-$2,000 (Prob=12.7%) and more importantly, 35.9% in the range of $5,001-$10,000. Results for other type of debt are less conclusive than the previous two. Now, when we combine the estimated probabilities, we observe that students have a significant high chance to hold large amounts of debt presently. As we observe an overestimation on the amount of self- reported financial literacy students possess, the concern is that students may believe – wrongly – that they are making the correct financial decisions. The reality presented by the data is indeed very different. It becomes almost obvious that students are not fully understanding the implications of acquiring financial obligations so early in their lives and the consequences that interest compounding may have on their ability to successfully manage their financial future. The main conclusion from our analysis is that low levels of financial literacy, high self-reported financial knowledge, and lack of interaction with parents regarding financial issues, results in excessive borrowing and debt accumulation early on in students’ lives. As simple as 1+1=2, results derived from these probability computations indicate that students have a higher chance of accumulating large amounts of debt. On itself, this is a very concerning result as it indicates that most students will enter the professional job market already holding debt amounts beyond their financial capacity. We, however, recognize that for most students, there may be very little options to finance their education if it is not through student loans. The consistent use of credit cards, as reflected by the outstanding balances, complicate the financial standing of students as they tend to leverage themselves beyond their financial capacity.

5. Conclusions The results presented in this paper allow some tentative conclusions. Empirical estimates of this study provide robust statistical support to the hypothesis in favor of an existence trans-generational effect regarding financial knowledge, where students that report discussing financial issues with their parents are significantly less likely to hold larger amounts of debt. In particular, our study sheds light to understand the important role that financial education has on preventing the youth from engaging in over-borrowing, and therefore accumulating large amounts of debt too early in their lives. In addition, our empirical evidence allows us to conclude that students that maintain a healthy management of credit cards, (paying in full the balance every month) are more likely to hold lower debt amounts in credit cards, but also in student loans. While it still remains unclear, whether the trans-generational effects leads to higher levels of financial literacy, the evidence drawn from this study is strong enough/robust to move forward the argument, that lower levels of indebtedness could effectively be achieved by promoting a direct interaction between the youth

and their parents. In general we can also conclude that students (from our sample) tend to self-report higher levels of financial knowledge. These self-reported levels of financial knowledge are not aligned with the actual measures of financial literacy. The existence of this discrepancy raises the question and concern as to what the effect of a more accurate self-assessment on financial knowledge could have on students’ debt management behavior. This question is however, beyond the scope of this research. Clearly, discussing financial issue with parents prevents students from holding larger debt amounts, yet it does not necessarily translate into higher levels of financial literacy, but higher levels of conventional/common financial practical knowledge. This difference, while subtle may actually have larger and longer-term implications, as the youth would be able to prevent over-borrowing, yet they might not be in a desirable position to actually conduct effective and efficient financial decisions. Finally, we argue that teaching financial knowledge to all college students may create the necessary synergy to provide better financial management knowledge as they enter the workforce. If this is correct, formal financial education would supplement the trans-generational effect mentioned earlier. Nevertheless, our results are exploratory in nature and therefore subject for further research. Several questions remain unanswered; such as how can the relationship between financial literacy and students-parents interaction be assessed to provide students with a more adequate overall financial knowledge to effectively make financial decisions?

References Allen, W. M., Edwards, R., Hayhoe R. C., & Leach, L. (2007). Imagined interactions, family money management patterns and coalitions, and attitudes toward money and credit. Journal of Family Economic Issues. 28, 3 -22. Bernheim, B. D., & Garrett, D. (2003). The effects of financial education in the workplace: Evidence from a survey of households. Journal of Public Economics. 87 (7/8), 1487-1519. Borden, M. L., Lee, U., Serido, J., & Collins, D. (2008). Changing college students’ financial knowledge, attitudes, and behavior through seminar participation. Journal of Family and Economic Issues. 29, 23-41. Bowen, F. C., & Jones, M. H. (2006). Empowering young adults to control their financial future. Journal of Family and Consumer Sciences. 98, 33-40. Brown, M., Klaauw, W., Wen, J., & Zafar, B. (2013) Financial education and the debt behavior of the young. Federal Reserve Bank of New York Staff Reports. No. 634. Carlin, B., & Robinson, D. T. (2012). What does financial literacy training teach us? The Journal of Economic Education, 43(3), 235-247. CAVANAUGH, A. R. (2013). Rich dad vs. poor dad: Why leaving financial education to parents breeds financial inequality & economic instability. Selected Works, 1-49. Retrieved from http://works.bepress.com/cgi/viewcontent.cgi?article=1002&context=afton_cavanaugh Chen, H., & Volpe P. R. (2002). Gender differences in personal financial literacy among college students. Financial Services Review. 11, 289 – 307. Chen, H., & Volpe, R. P. (1998). An analysis of personal financial literacy among college students. Financial Services Review, 7(2), 107-128.

Journal of Personal Finance

Chen, M. (2005). Ethics: An urgent competency in financial education. Journal of American Academy of Business, Cambridge. 6, 74-80.

Lusardi, A., & Tufano, P. (2009b). Debt literacy, financial experiences, and overindebtedness. NBER Working Paper Series. JEL No. D91, 1-35.

Debbaut, P., Ghent, A. & Kudlyak, M. (2013) Are young borrowers bad borrowers? Evidence form the Credit CARD Act of 2009, Working Papers Series, The Federal Reserve Bank of Richmond. WP 13-09.

Manning, R. (2000). Credit cards on campus. The social consequences of student credit dependency. New York: Basic Books, 159- 193.

DesJardins, L. S., Ahlburg, A. D., & McCall, P. B. (2002). Simulation the longitudinal effects of changes in financial aid on student departure from college. The Journal of Human Resources. 37, 653-679. EViews 5.1 User Guide. Quantitative Micro Software. Fiona, C., Wai-Lap, C., & Yin-Kwan K. (2012). Financial knowledge and aptitudes: Impacts on college students’ financial well-being. College Student Journal. 46,114-132. Ford, W. M., Devoto, S., Kent, W. D., & Harrison, T. (2007). Threat, intimidation, and student financial knowledge: An empirical study. Journal of Education for Business. 82, 131-140. Fox, J., Bartholomae, S., & Lee, J. (2005). Building the case for financial education. The Journal of Consumer Affairs. 39, 195-214. Goetz, J. W., Durband, D. B., Halley, R. E., & Davis, K. (2011). A peerbased financial planning & education service program: An innovative pedagogic approach. Journal of College Teaching & Learning, 8(4), 7-14. Greene, William. (2003). Econometric analysis, 5th ed., Upper Saddle River, NJ: Prentice Hall. Heckman, S., & Grable J. (2011). Testing the role of parental debt attitudes, student income, dependency status, and financial knowledge have in shaping financial self-efficacy among college students. College Student Journal. 45, 51-64. Hung, A., Parker, A., & Yoong, J. (2009). Defining and measuring financial literacy. RANDCorporation Publications Department, Working Papers: 708, 28 pages. Jennings R. J. (October 2007). Higher education must fill the void in student financial management. Diverse Issues in Higher Education, 24, 46. Jorgensen, B., & Salva, J., (2010). Financial literacy of young adults: The importance of parental socialization. Family Relations, (4), 465-478. Kim, J., Garman T. E., & Quach, A. (2005). Workplace financial education participation and retirement savings by employees and their spouses. Journal of Personal Finance 4, 92 -108. Kim, J., Sorhaindo, B, & Garman, T. E. (2006). Relationship between financial stress and workplace absenteeism of credit counseling clients. Journal of Family Economic Issues 27, 458-478. Kuchler, T. (2013). Sticking to your plan: Hyperbolic discounting and credit card debt paydown. Stanford Institute of Economic Research (SIEPR) Discussion Paper 12-025.

Mandell, L (1997). Personal financial survey of high school seniors. Jump Start Coalition for Personal Financial Literacy, April. Washington, DC. Martin, M. (2007). A literature review on the effectiveness of financial education. Federal Reserve Bank of Richmond WP 07-03. Martin A., & Oliva, J. C. (2001). Teaching children about money: Applications of social learning and cognitive learning developmental theories. Journal of Family & Consumer Sciences, 93(2), 2 29. Nellie Mae (2002, April). Undergraduate students and credit cards: An analysis of usage rates and trends. Retrieved from http://www.westmont. edu/~phunter/ma5/eg/credit.pdf Mojtaba, S. & Taihyeup D. (2011). Improving financial literacy of college students: A cross sectional analysis. College student Journal. 45, 1. Ornelas, S. & Kleiner H. B. (2003). New developments in management job related stress. Equal Opportunities International. 22, 64 – 70. Peng, M. T., Bartholomae, S., Fox, J. J., & Cravener, G. (2007). The impact of personal finance education delivered in high school and college courses. Journal Family Economic Issues. 28, 265-284. Roberts, A. J., & Jones, E. (2001). Money attitudes, credit card use, and compulsive buying among american college students. The Journal of Consumer Affairs. 35, 213-240. Serido, J., Shim, S., Mishra, A., & Tang, C. (2010.) Financial parenting, financial coping behaviors, and well-being of emerging adults. Family Relations, (4), 453-464. Shim, S., Barber, B., Card, N., Xiao, J., & Serido, J. (2010). Financial socialization of first-year college students: The role of parents, work, and education. Journal of Youth and Adolescents, 39(12), 1457-1470. Tennyson, S. & Nguyen, C. (2001). State curriculum mandates and student knowledge of personal finance, Journal of Consumer Affairs. (Winter) 35, 241 -262. Volpe, C. & Mumaw, K. (2010). Mortgage meltdown reveals importance of financial literacy education. Research and Theory, 9, 61- 77. Volpe, R.P., Chen, H., & Pavlicko, J.J. (1996). Personal investment literacy among college students: A survey. Financial Practice and Education. 6, 68-94. Wells, C. (2007). Optimism, inter-temporal choice, and college student debt. Journal of personal Finance, 5(4), 44-66.

Lewin, T. (2011, November 2). College graduates’ debt burden grew, yet again, in 2010. The new york times. Retrieved from http://www.nytimes. com/2011/11/03/education/average-student-loan-debt-grew by-5-percent-in-2010.html Lusardi, Anna Maria. (2008). Financial literacy: An essential tool for informed consumer choice? NBER Working Paper Series. WP 14084 Lusardi, A., Mitchell, S. O., & Curto, V. (2009a). Financial literacy among the young: Evidence and implications for consumer policy. PRC Working Paper Pension Research Council. WP2009-09, 1-35.

Volume 13, Issue 2

Appendix

Journal of Personal Finance

ÂŠ2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 2

A Better Systematic Withdrawal Strategyâ&#x20AC;&#x201D; The Actuarial Approach Ken Steiner, Fellow, Society of Actuaries, Retired

Retirees generally have at least two potentially conflicting financial goals: (i) spend enough each year to maintain a certain standard of living throughout retirement and (ii) not spend so much that accumulated savings run out prior to death. Corollaries to these two primary goals include: (i) having relatively predictable and stable inflation-adjusted spendable income from year to year (ii) having spending flexibility to meet unforeseen expenses, (iii) maximizing the general level of spendable income and (iv) not leaving too much unspent at death. An optimal retirement spending strategy should address each of these goals to some degree, depending on the preferences of the individual retiree.

Journal of Personal Finance

Introduction There are four general sources of retirement income in the U.S: (i) Social Security, (ii) life annuity income from pension plans or insurance company products (iii) systematic withdrawals from accumulated savings and (iv) other income, including income from employment in retirement. Recent research has shown that delaying commencement of Social Security, possibly working longer and combining life annuity products with systematic withdrawals from accumulated assets may all be components of an optimal retirement spending strategy1. This paper focuses on alternative systematic withdrawal approaches and concludes that “The Actuarial Approach” is a better approach than alternative systematic withdrawal approaches that are commonly used. It can coordinate income from the other retirement income sources and has the potential to do a better job of balancing the potentially conflicting goals of most retirees. The following sections describe The Actuarial Approach, provide an example of its use and briefly discuss why The Actuarial Approach is a better systematic withdrawal strategy.

The Actuarial Approach The five steps involved in The Actuarial Approach are briefly described below. It is anticipated that this process will be followed at least once a year in order to determine the retiree’s total spendable amount (or “spending budget”) for the year. The author recommends that the retiree’s spending budget for a given calendar year be determined at or near the beginning of such year. Step 1: Gather Data. In the first year of using The Actuarial Approach, the following data may be required: 1) the amount of the retiree’s accumulated savings2, 2) how those assets are or will be invested, 3) the retiree’s health status (and that of the retiree’s spouse or significant other), 4) the amount of immediate lifetime income payable from sources other than Social Security, 5) amounts payable from deferred income annuities and year(s) of commencement of such payments, and 6) the amount of accumulated assets desired to be left to heirs. In subsequent years, the data above will need to be supplemented with results from steps 3 and 4 below for the previous year and the increase in the Consumer Price Index for the year just ended. If the retiree is determining withdrawals from accumulated savings in order to “bridge” to a deferred Social Security commencement age, the retiree will need an estimate of the deferred Social Security benefit and number of years until its expected commencement. Step 2: Make Relevant Assumptions. Next, assumptions need to be made about the expected rate of future return on accumulated assets, the expected payout period and the expected rate of future inflation. The expected rate of future return assumption should generally be coordinated with the assumption for future inflation, as discussed below.

Some actuaries and economists believe that assumed investment return should approximate a risk-free interest rate, as the higher expected returns associated with investment in riskier assets, such as equities, also carry a higher risk of volatility, meaning the returns might vary significantly over time. Therefore, assuming a risk-free interest rate for all asset classes is more conservative and automatically adjusts for the extra risk inherent in investing in riskier assets with higher expected returns. Assumed Investment Return Recommendation: The investment return assumption should generally reflect investment return assumptions inherent in immediate annuity contracts with perhaps some small adjustment for greater expected returns on equities. At the current time, the author recommends use of an investment return assumption of 5% per annum (nominal, not real), or slightly lower if the retiree plans to invest mostly in bonds. This recommended assumption is consistent with nominal interest rates currently used to price immediate life annuities of approximately 4% per annum.3 An assumption must also be made about longevity. How long will the retiree(and/or the significant other) need these payments? Life expectancies are based on average mortality experience, but people can and do outlive their life expectancy. In fact, if you experience average mortality, you have about a 50% probability of outliving your life expectancy based on standard mortality tables. If life expectancy is used for the expected payout period, experience losses will occur each year that the retiree survives, and future actuarially determined withdrawals will decrease, all things being equal. There are many sites on the Internet that can provide life expectancy estimates. The author recommends the Society of Actuaries Simple Life Expectancy calculator found at http://www.soa.org/research/software-tools/ research-simple-life-calculator.aspx Expected Payout Period Recommendation: In order to avoid having decreasing withdrawals from year to year that result from living an additional year, the author recommends using an assumed expected payout period equal to 95 minus the retiree’s current age, or the retiree’s life expectancy if greater. Generally, retirees will want to make sure that their retirement income keeps pace with rising inflation. The assumption about future inflation should be consistent with the assumption made for future investment return. Assumed Inflation Recommendation: The author’s current recommendation for an inflation assumption is 3% per annum, or the investment return assumption minus 2% if the investment return assumption is lower than 5% per annum. Note that some retirees may want to “front-load” their real annual spendable income to some degree. For example, they may not be concerned about later year expenses, such as long-term care, and wish to travel more in their early years of retirement. In this case, they may wish to consider using an annual desired increase in withdrawal assumption that is less than expected inflation. Other

Volume 13, Issue 2

retirees may wish to be more conservative in selection of assumptions used in order to avoid decreasing spendable amounts. Of course, the retiree can also be conservative by spending less than the spending budget for the year.

corridor, the Total Spendable Amount for the year is equal to the result of Calculation #2. If the result of Calculation #2 falls outside the 10% corridor, the Total Spendable Amount for the year will equal the applicable corridor upper or lower limit.

Step 3: Perform Calculations to Determine Preliminary Spendable Amount4. The Preliminary Spendable Amount is the answer to the mathematic problem, “What total spendable amount (from accumulated savings and annuities) may be spent in the current year, to be increased each subsequent year by a constant percentage so that accumulated assets will exactly equal the amount desired to be left to heirs at the end of the expected payout period?” The author’s website provides a simple spreadsheet tool for this calculation (“Excluding Social Security 2.0”). More sophisticated tools for this purpose might be available on the Internet or elsewhere. However, the specific tool the retiree uses for this calculation is generally less important than making reasonable assumptions (as described in Step 2) and diligently following the five step process which constitutes The Actuarial Approach each year.

Step 5: Store the Results for Next Year. A critical component of “The Actuarial Approach” is to periodically adjust results to reflect actual experience and changes in assumptions, if necessary. As discussed above, the author recommends that retirees revisit their spending budget at the beginning of each year. Therefore, at a minimum, the Preliminary Spendable Amount (the result of Step 3) and Total Spendable Amount (the result of Step 4) for the current year should be placed in a file to be used in the budget determination for the next year. It may also be desirable to print out the input tab from the website spreadsheet as well to see what assumptions and data were used in the previous year.

Note that this simple tool coordinates income provided by fixed immediate annuities, fixed deferred annuities and withdrawals from the retiree’s accumulated assets. It does not coordinate income from Social Security (hence the name “Excluding Social Security”) or inflation-indexed annuities, as these sources of retirement income are already assumed to be inflation-adjusted and simply need to be added to the result from the spreadsheet to obtain the retiree’s theoretical total spendable amount for the year. The two “Run-Out” tabs of this spreadsheet show year by year withdrawals from accumulated savings, total spendable amounts (Excluding Social Security) and accumulated assets in nominal and inflation-adjusted dollars assuming all assumptions made in the input tab are exactly realized each future year and amounts withdrawn from accumulated savings equal the budgeted amount. However, the user should note that it is likely, for many reasons, that the Preliminary Spendable amount will be different from the amount actually spent by the retiree during the year.

The Actuarial Approach as described above may seem to be more complicated than it actually is. If the data required in Step 1 are relatively easy to pull together, the annual budgeting process should take no more than 10 minutes to complete.

Step 4: Apply Smoothing Algorithm to Develop the Spending Budget For The Year. In order to provide a more stable and predictable spending budget pattern, it is advisable to smooth experience gains and losses and changes in assumptions. However, this desire to have spendable income stability needs to be balanced against the conflicting goals of not running out of money and not under spending. Recommended Smoothing Algorithm: Calculation #1: Multiply the Preliminary Spendable Amount by both 90% and 110% to develop a 10% corridor around the Preliminary Spendable Amount for the year. Calculation #2: Increase last year’s Total Spendable Amount (the spending budget for the previous year) by the increase in CPI during the previous year and add any previously deferred annuity amount that commences in the current year. Note that it is the spending budget from the previous year that is used in this calculation and not the actual amount spent by the retiree. If the result of Calculation #2 falls inside the 10%

Example

Here is an example: Richard retired on January 1, 2013 at age 65. At that time, he used about 20% of his accumulated savings to buy an immediate life annuity that pays him $15,000 per year. At the beginning of 2013, he had $800,000 left after his annuity purchase. He inputted the author’s recommended assumptions (5% interest, 3% inflation, 30 years expected payout period (95-65)) and $10,000 as the desired amount of assets at death into the spreadsheet in the author’s website, to determine a total spendable amount (excluding Social Security) for 2013 of $45,179 ($30,179 from accumulated savings and $15,000 from the annuity). He deposited $30,179 in his non-interest bearing spending account and decided to invest half of the remaining assets ($769,821) in equities and the other half in a variety of fixed income investments. During 2013, Richard spent exactly the amount in his spendable account plus the $15,000 from the annuity. Easy Steps to Determine Richard’s Spending Budget for 2014 The first step in the process is gather asset data as of the end of 2013. Richard’s equity investments yielded almost 29% during 2013 and his fixed income investments yielded about 1%, so his end-of-year assets are $884,909 (compared with expected end-ofyear assets from the previous year’s calculation of $808,312, or an asset gain for 2013 of $76,597). Richard determined that the Consumer Price Index has increased by 1.3% during 2013. He pulls out his file containing his beginning-of-year 2013 calculations. The second step is to review the assumptions used for 2013 and see if they are still appropriate for 2014. Richard decides to use the same assumptions for 2014 as 2013.

Journal of Personal Finance

The third step is to determine a preliminary spending value for 2014 by inputting new data amounts into the Excluding Social Security spreadsheet on the author’s website . If the same assumptions and life annuity and bequest amounts are input as for 2013, $884,909 is inputted for accumulated savings and 29 years for the expected payout period, Richard’s preliminary 2014 spendable amount is $49,947 ($34,947 from accumulated savings + $15,000 from the annuity). The fourth step in the process is to apply the recommended smoothing algorithm to the preliminary spending value. Richard determines his 2014 total spendable amount as last year’s total spendable amount ($45,179) increased by 1.3% ($45,766), but not less than 90% of the preliminary 2014 total spendable amount of $49,947 (.9 X $49,947 = $44,952)and not more than 110% of $49,947 (or $54,942). Since last year’s budget amount increased with inflation for the previous year falls inside the 10% corridor, Richard’s total spendable amount for 2014 is $45,766 ($30,766 from accumulated savings and $15,000 from the annuity). Richard then places the results of his 2014 calculations into his retirement file to be used next year when he determines his 2015 spending budget.

Comparison of The Actuarial Approach with Three Commonly Used Systematic Withdrawal Strategies In September, 2013 the Stanford Center on Longevity, in collaboration with the Society of Actuaries Committee on PostRetirement Needs and Risks, released “The Next Evolution in Defined Contribution Plan Design.”5 In that paper, three common systematic withdrawal strategies were examined: The 4% Rule, the constant 4% withdrawal approach and the Required Minimum Distribution (RMD) rule established by the Internal Revenue Service. None of these three strategies attempts to coordinate spendable income from accumulated savings with other annuity income that the retiree may have or may expect to receive in the future. This, in and of itself, is a significant deficiency, as research has shown that it is generally financially prudent to manage risks in retirement by diversifying sources of retirement income. Further, none of these strategies can be used if a retiree desires to use some or all of his accumulated savings to help defer commencement of Social Security benefits. In addition, none of these strategies anticipates payment of a specific bequest motive. Even if the retiree has no annuity income and does not plan to defer commencement of Social Security benefits or leave specific amounts to heirs, each of these three strategic withdrawal strategies has additional shortcomings when compared with The Actuarial Approach. These shortcomings are briefly discussed below.

his accumulated savings in the first year of retirement and then increases this amount by accumulated inflation in subsequent years. This “set and forget” withdrawal strategy does provide a very stable and predictable withdrawal pattern. However, it does this by ignoring the impact of actual experience. Further, this approach anticipates that the retiree will withdraw exactly the amount dictated by this strategy each year, and therefore offers no flexibility in actual withdrawals. If experience (including actual spending vs. budgeted spending) is relatively favorable, there is no adjustment in withdrawals and not enough money is spent. If experience is relatively unfavorable, too much money may be spent. Since it is generally designed to be “safe”, there is theoretically more likelihood of the former occurring rather than the later. Also, each year’s spendable amount depends entirely on the amount of accumulated assets in the first year of retirement, and therefore may be highly dependent on when the individual retires. Some argue that this method is preferable because it is “simple.” However, adjustments are supposed to be made under this approach for different expected payout periods and investment mixes and frequently proponents of this approach indicate that specified or unspecified adjustments should be made to reflect actual experience. After factoring in all these “adjustments”, the author does not find this approach to be any more simple than The Actuarial Approach. Under the Constant 4% Withdrawal Rule, only 4% of the participant’s accumulated assets are withdrawn each year. While this approach does adjust for actual experience, there is no smoothing of year to year results and while a 4% withdrawal rate may be appropriate at age 65, it is far too low at older ages, so too much money is likely to remain at death. In addition, this rule does not consider the effects of inflation on retiree purchasing power. Under The RMD rule, specific percentages based on expected longevity are applied to accumulated savings. The Stanford/SoA paper suggested 3.5% for ages prior to 70. This approach also adjusts for actual experience, but like the Constant 4% Rule, there is no smoothing of year to year results and it too tends to understate withdrawals, particularly in the early years of retirement. Also, like the 4% constant withdrawal approach, this rule does not consider the effects of inflation on retiree purchasing power. The following graphs show withdrawal patterns and remaining assets under these three approaches compared with The Actuarial Approach for the period 1998 to 2014 for a person retiring at age 65 with $500,000 of assets in 19986. For this period, the 4% Rule produces a ruler-flat inflation adjusted withdrawal pattern, but it fails to maximize the retiree’s desire to maximize spending. The other two approaches also failed to maximize spending and their withdrawal patterns were much less stable from year to year than under The Actuarial Approach.

Volume 13, Issue 2

55 $35,000 $35,000 $30,000 $30,000 $25,000 $25,000 $20,000

Constant 4%

$20,000 $15,000

RMD Constant 4% Steiner RMD Actuarial Approach

$15,000 $10,000

4% RuleActuarial Approach Steiner 4% Rule

$10,000 $5,000

1998 1998 1999 1999 2000 2000 2001 2001 2002 2002 2003 2003 2004 2004 2005 2005 2006 2006 2007 2007 2008 2008 2009 2009 2010 2010 2011 2011 2012 2012 2013 2013 2014 2014

$5,000 $0

Annual Withdrawals Under Four Alternative Systematic Withdrawal Strategies 1998-2014 Remaining Assets Under Four Alternative (in 1998 Dollars) Systematic Withdrawal Strategies Remaining Assets Under Four Alternative Systematic Withdrawal Strategies (in Nominal Dollars) (in Nominal Dollars)

$900,000 $900,000 $800,000 $800,000 $700,000 $700,000 $600,000 $600,000 $500,000

Constant 4% RMD Constant 4% Steiner RMD Actuarial Approach

$500,000 $400,000 $400,000 $300,000

4% RuleActuarial Approach Steiner

$300,000 $200,000

4% Rule

$200,000 $100,000

1998 1998 1999 1999 2000 2000 2001 2001 2002 2002 2003 2003 2004 2004 2005 2005 2006 2006 2007 2007 2008 2008 2009 2009 2010 2010 2011 2011 2012 2012 2013 2013 2014 2014

$100,000 $0

Remaining Assets Under Four Alternative Systematic Withdrawal Strategies (in Nominal Dollars)

Conclusion It is not unreasonable for retirees to diversify their sources of retirement income. Research has shown combining life insurance annuity products with strategic withdrawals from accumulated assets to be an efficient strategy for managing risks in retirement7. An optimal systematic withdrawal strategy is one that can coordinate the various sources of retirement income to meet a retiree’s financial goals in retirement. It is the author’s belief that

The Actuarial Approach is a better withdrawal strategy that can result in a better overall retirement spending strategy, provided reasonably conservative assumptions and methods are selected and the five step actuarial process is diligently followed. For more discussion of the Actuarial Approach, a brief biography of the author and a lengthy disclaimer regarding The Actuarial Approach please visit the author’s blog at http://howmuchcaniaffordtospendinretirement.blogspot.com/

Journal of Personal Finance

Footnotes: 1 Research

on deferring commencement of Social Security benefits: “Efficient Retirement Design--Combining Private Assets and Social Security to Maximize Retirement Resources”, John B Shoven and Sita N. Slavov, Stanford Institute for Economic Policy Research. Research on Efficient Frontier for Retirement Income: Dr. Wade D. Pfau, “An Efficient Frontier for Retirement Income”, Journal of Financial Planning, February, 2013 and follow-up article, “Why Retirees Should Choose DIAs over SPIAs”, Advisor Perspectives, September 24, 2013

2Accumulated

assets may include an estimate of home equity if the retiree expects to downsize, apply for a reverse mortgage or otherwise tap into home equity to fund retirement spending.

drawals of accumulated savings and annuity payments) are designed to increase by an inputted percentage each year. Adjustments would need to be made to accommodate other desired spending patterns. The author’s website does contain a separate spreadsheet (Social Security Bridge) for the purpose of coordinating total spending with a decision to delay commencement of Social Security. 5”The

Next Evolution in Defined Contribution Plan Design”, by Steve Vernon, FSA, Consulting Research Scholar, Stanford Center on Longevity in collaboration with the Society of Actuaries’ Committee on Post-Retirement Needs and Risks.

on an annuity purchase rate for a male age 65 of approximately $600 per month for a premium of $100,000 as shown in Incomesolutions. com and assuming life expectancy of approximately 22 years from Society of Actuaries Annuity-2000 table with 1% mortality improvement, the author derived an interest rate of about 4.6% using the life expectancy annuity certain approximation method. An investment return assumption of 5% per annum would be a compounded investment return, not an arithmetic return.

based on author’s calculations. Consistent with interest rates inherent in immediate annuity purchase rates, 1998 initial assumptions for The Actuarial Approach were 7% investment return and 4% inflation changing to 6% investment return and 4% inflation in 2004 and 5% investment return and 3% inflation in 2009. Desired amount of bequest at death was $10,000. Investment experience based on investment mix of 25% large cap equities, 25% mid-cap equities, 25% mid-term bonds and 25% short-term bonds with annual rebalancing of investments. Source of investment returns by assets class: Investor Cookbooks Historical Asset Class Return Charts. Source of CPI data from Tab 8A Society of Actuaries--Statistics For Employee Benefits Actuaries.

4The

7See

3Based

simple spreadsheet tool referred to in this section (Excluding Social Security 2.0) does not anticipate the retiree having more than one deferred annuity. Adjustments would have to be made to accommodate multiple deferred annuity starting dates. Total theoretical withdraws (from with-

6Graphs

footnote 1. The author makes no claim regarding the optimal mix of immediate annuities, deferred annuities and self managed assets to be included in a retirement portfolio. He leaves this project up to more well qualified retirement researchers.

There are many risks associated with self-insuring your own retirement. The general process described in the article is made available as a self-help tool for independent use and is not intended to provide investment or financial advice. As with all planning tools, the reasonableness of the results (in this case, the “annual spendable amount”) is a function of the accuracy of the data and assumptions.

Volume 13, Issue 2

A Rule-of-Thumb Approximation for Time Value of Money Calculations David N. Swingler, Ph.D., Professor, Engineering Division, Saint Maryâ&#x20AC;&#x2122;s University, Halifax, Nova Scotia, Canada

A simple rule-of-thumb is presented for the classic financial calculations centered on the present value of a series of equal future payments. It is demonstrated that it is a useful addition to the armamentaria of engineering students engaged in engineering economics and/or finance courses.

Journal of Personal Finance

Introduction

Discussion

Engineers have long been associated with approximations, often

codified as “rules-of-thumb” which help them deal with otherwise complex mathematical situations. These commonly find application not only at the outset of projects where ballpark estimates are useful, but also during projects as guidance while undertaking a more complete exposition of the situation, and even at the end for final “sanity checking”. The sanity-checking mechanism is also at play while reading third party reports, for instance, just to make sure that what is written makes on-going sense to the reader. The most useful, and gratifying, of these rules-of-thumb are perhaps those that are simple enough to be handled via mental arithmetic Engineering students are encouraged to use the same philosophy: nothing is more humiliating than having a nice piece of work undone by a careless use of a spreadsheet formula or mis-pressing of the buttons on a sophisticated calculator, producing a significant error which makes all the ensuing results completely useless. The proverb “spoiling the ship for a ha’porth of tar” springs to mind here.

We begin by simply asserting that a useful approximation to b is given by

, i≤0.2

(1A)

This is felt to encompass most situations of “everyday” practical , i≤0.2 . use interest but for completeness for ni > 3 we

, i≤0.2

If we compute the ,percentage error between , i≤0.2 (1B)

and

, i≤0.2 IfEquation we compute percentage errorand between A justification for (1A) isthe given in the appendix it might be remarked that the derivation therein requires only the mathematical skills of a first- or second-year engineering student. In passing, remark thattoin this financial $ bˆ is the Thewesolution is language, given by solving (approximate) value of any one of a stream of n equal and equi, i≤0.2 approximation of (1A) it becomes simply: spaced future payments with a present value of P = $1.

One field where rules-of-thumb play little role is that of finance The solution to this isˆ given by solving (with the exception of the ubiquitous rule-of-72). This is almost If we compute the percentage error between b and If we compute the percentage error between and viab via simply: approximation of (1A) it becomes certainly due to the notion that monetary amounts must be accu=618 . rate to the penny, which is true at the time contracts are signed but may not be true at every point in the process, as outlined above. Additionally, university courses in finance and economics =618 . are relatively new additions to the field of engineering education. And finally bankers are just not engineers and do not think like forval-i, with n The solutionthentothethis is givenresults by ofsolving representative Figure 1 ensue for various them. ues of i from 1 to 20% and n from 1 to 121 periods. We see from approximation of (1A) it becomes simply: = The purpose of this brief note is to suggest that there is a useful Figure 1 that the peak errors are not much more than very roughly “rule-of-thumb” for the classic and most common time value of ± 6% which is deemed entirely adequate for our rule-of-thumb money problem which involves dealing with the present value, P, usage, especially as the underlying = for b seems quite =618expression . of series of n equi-spaced future payments, each of value, A, with subtle over this wide range of n and i. Further we might observe an interest rate of i, via the formula that the peak negative errors of about -6% occur in the vicinity of ni=3 which makes their occurrence easily predictable (and largely A = positive = 250,000 correctable if so desired). The peak errors of about +5% /100 = occur in the vicinity of ni=1 (roughly), again useful knowledge. where the factor b is given by Zero error occurs near ni=2 (roughly). Finally we note that at = 250,000 /100 A= ni=3 where the approxima= .two parts of the piecewise continuous = 24000/ (2.4/21) =$2 . tion abut, there no discontinuity. This is aexpression felicitous effect. Weis start with the correct for given by

The first payment is at the end of the first period. As an aside, we note that Hawawini and Vora [1, p. (v)] have indicated that the , i≤0.2 and problem of calculating i from known A, P and n is non-trivial , i≤0.2 has a very long history.

, i≤0.2

We start with the correct expression for A=

= 250,000 and let

/100 = 250,000 so that the above becomes

and let so that the above becomes , i≤0.2We start with the correct expression for given by

If we compute the percentage error between compute the percentage error between and via

and

via

and let

so that the above becomes

given b

/100

Volume 13, Issue 2

, i≤0.2

If we compute the percentage error between

and

via

Figure 1. Errors in the approximation of Equation (1).

Examples

, i≤0.2 The solution to this is given by solving approximation of (1A) it becomes simply:

for i, with n=9. With our

We take our second example from a worked example in the We now present simple examples of the utility of the rule-ofcourse text Contemporary Engineering Economics by Park et al thumb of Equation (1A). It might be of interest to note that these , , i≤0.2 =618 . [2, p.78]. Paraphrasing it states: A lottery winner is expecting 21 were literally the first examples that came to the hand of the annual payments of $24,000 (after tax) which he wants to use author as he thought about this topic. We start with one ofIfthe we compute the percentage error between and via against a bank loan at 10% interest. How much can he borrow earliest examples of interest calculations, taken from [1, p. 1] from the bank? Here we have A =$24k, n=21, i=0.1 and we use A 16th century Italian university was loaned 2814 ducats and = = 24000/ (2.4/21) =$210,000 repaid it in nine annual installments of 618 ducats beginning at the end of the first year. What was the effective interest rate? The solution to this is given by solving for i, with n=9. With our (Recall that interest rate calculations, accurate ones anyway, are which is also (just about) manageable by mental arithmetic alone. approximation of (1A) it becomes simply: regarded as mathematically non-trivial.) The correct answer is $207,569 but the book’s authors offer the interesting caveat that the whether the bank would loan that =618 . /100 = 250,000 The solution to this is given by solving 2814b = 618 for i, with amount A = depends = 250,000 /100 on the applicant’s creditworthiness, suggestn=9. With our approximation of (1A) it becomes simply: ing that our $210k figure would nicely suffice for this ballpark estimate. 1 2  2814 + i  =618 . We start with the correct expression for given by Our last problem is taken from another worked example in the 9 3  same text=[2, p.76]. Here a company has borrowed $250,000 for = 24000/ (2.4/21) =$210,000 equipment. The loan carries an interest rate of 8% and is to be It is convenient to mentally rearrange this as (1 + 6i) = 9 * 618 / 2814 repaid in annual installments over the next six years. What is the and if we observe that the right-hand side is almost exactly two amount annual installment? Here we have then an approximate answer immediately falls out as i =16⅔%by and let so that of thethe above becomes mental arithmetic only. (Given that the numbers involved strongly suggest that the amount repaid was exactly twice that A= = 250,000 /100 = 250,000 /100 loaned, one wonders whether there is an ancient typographical error in the data as given here). With the data as actually given, or A = 22% of $250,000 = $55,000, again via mental arithmetic our approximation yields i =16.3% which is the correct answer We start with the correct expression for given by only. Note the efficacy of turning bˆ into a percentage figure. to three significant digits. The fact that the rule-of-thumb is this The correct answer is $54,079 but requires access to (and correct accurate lies in the fortuitous positioning of this problem on the application of) significant computational resources. left in Figure 1 in the vicinity of n=9, i=16% where the approximation error is small. and let

so that the above becomes

Journal of Personal Finance

Students will of course complain that Equation (1A) is just another formula to memorize. To this end is useful to refer to it via the mnemonic “the one ’n two-thirds formula” and to indicate that, as it deals with loan repayments, it is natural for the quantity to reduce with n, hence n appears as a denominator, but increase with i so i appears as a numerator term. It should also be stressed this is not a replacement for the “correct” results, but simply an auxiliary tool.

= 250,000

/100 = 250,0

Appendix I : Derivation of the approximation We start with the correct expression for given by

We start with the correct expression for b given by

and let

so that the above becomes

and let ni = p so that the above becomes

Escalating payment streams It should be observed that approximation of equation (1A) can be readily extended to a stream of escalating payments. Rather than digress here, this material is adumbrated in Appendix II.

We begin the by lettingn be nbylarge. beletting large. a cla We approximation begin the approximation nFollowing be alarge. Follo We begin the approximation by letting Following We begin the approximation by letting n be large. Fo L’hopital’s rule for the bracketed term in the denominator, we arrive a We begin the approximation by letting n be large. Following a classic application of L’hopital’s rule for the bracketed term in the denominator, classic application of L’hopital’s rule for the bracketed term in the L’hopital’s rule for the bracketed term in the denominato We begin the approximation by letting n be large. Following a classic application of L’hopital’s rule for the bracketed term in the denominator, arrive at the approximation denominator, we arrive at thewe approximation Conclusion L’hopital’s rule for the bracketed term in the denominator, we arrive at the approximation . . We have presented a useful rule-of-thumb for time value of . money problems, of benefit to students in engineering econom. ics and finance classes and perhaps even to their professor, idly . musing, while exam invigilating, what the payments might be It transpires,It not altogether that is reasonably linea transpires, notintuitively, altogether intuitively, that is reaso on the $300,000 mortgage she’d need to take out on that new It transpires, not altogether intuitively, that isthat reasonIt transpires, not altogether intuitively, is rea is reasonably linear for and transpires, not altogether intuitively, that house she is interested in, It with, say, an amortization of 25 years ably for and it is this unitymight whereuse p=0. this use is reasonably linear for andFor It transpires, not altogether that answer: (100/25 and an interest rate of 6%. intuitively, [Her approximate + unity Swingler paper it is where p=0. For this reason we ≈reaan itlinear is unity where p=0. reason we For might 2/3 * 6)% of $300,000 = 8% of $300,000 = $24,000pa or about itmight is unity where p=0. For this reason we might use ≈ and with a little it is unity where p=0. For this reasonsonwewemight use use and with a little numerical $2000pm]. regression it is unity where p=0. For this reason we might use ≈ numerical and with a little analysis on the interval 0  p  3 , m is regression analysis on the interval 0 ≤ p ≤ 3, m is found to be immediately leads to the final result about 0.67. This immediately leads to the final result

≈

= or or or for ni ≤ 3. or

or for ni ≤ 3.

≈

≈=

p / n 1  ≈2  1 2 = ˆ =b  ≈ 1  p  =  i = b 1  e p n  3  n 3

for ni ≤ 3.

for ni ≤ 3. 1 2  3. ni ≤ 3. bˆ    i  for ni ≤ for n 3 

Notwithstanding the constraints in this analysis, the main body of Notwithstanding thethat constraints in this analysis, the main body of this note demonstrates the approximation works well down approximation works downfortothe as special low ascase n=1offor i<0.2 (altho to as low as n=1 for i<0.2well (although n=1, . where where just1+i, 1+i,ofof course). completeness, ni>3,wei<0.2, we bb isisjust course). For For completeness, ni>3,. i<0.2, . simply. use

bˆˆ=i i.

Appendix II: Escalating Payment Streams

Escalating payment streams can be readily handled by simple m In particular, if the payment amounts increase by e% each period we have and where the nth is where n(i-e)≤2, i≤0.25, e≤i i≤0.25, Again, then gives (approximately) the n(i-e)≤2, e≤i Again gives (approxi

where i≤0.25, e≤i Again gives where n(i-e)≤2, i≤0.25, e≤i asymptotically Again givesn(i-e)≤2, (approximately) the first payment. It is (appro correct as (i-e)→0 has utility, for utility, example, asymptotically correct asand (i-e)→0 and has forinex 1 2 where n(i-e)≤2, i≤0.25, e≤iasymptotically Again gives (approximately) the first payment. It is asymptotically correct asin(i-e)→0 and the has utility, for paymentand ofpayment ahas stream escalating retirement from a princi correct as (i-e)→0 utility, example, first ofofafor stream escalating retirement payouts fro bˆ  of  i estimating  e  . payouts n 3 payment of a stream of escalating retirement payouts asymptotically correct as (i-e)→0 and has utility, for example, in estimating the first portfolio yield is i and the inflation rate is e. It has a sweet spot nf payment of a stream of escalating retirement principal of Prate where portfoliopayouts yield isfrom i anda the inflation is e. the It has a sw portfolio yield is i and the inflation rate is e. It has a payment of a stream of escalating payouts from amakes principal ofIt Pfor where theforspot itisideal retirement income streams. Forstreams. instanc portfolio retirement yield is i and the which inflation rate e.makes has aideal sweet n≈30, (i-e)≤0.6 which itsuch suchnear retirement income ˆmakes gives (approximately) the first payment which is Here which it ideal for such retirement income stream A Pn≈30, bform portfolio yield is i and thewhich inflation rateitisideal e. Itfor hasrights aprincipal sweet spot near (i-e)≤0.6 ofprincipal $500,000 and his portfolio has a yield of i=7% and ©2014, IARFC. All ofretirement reproduction inany reserved. makes such income streams. For instance if retiree has a of $500,000 and his portfolio has a yield of s principal of $500,000 and his portfolio has a yield which makes it ideal for such retirement income streams. instance ifaerror awith retiree aover period. The distribution is of now but there peak with e=3% over ahas period years then theareapprox principal of $500,000 and inflation his For portfolio has yield of i=7% hecomplex, wants match inflation e=3% a n=30 period of to n=30 years theno

. Volume 13, Issue 2

It transpires, not altogether≈intuitively, that =

= is reasonably linear for

Escalating Payment Streams it is unityAppendix where p=0.II:For this reason we might use or

Escalating payment streams can forbenireadily ≤ 3. handled by simple modification(s) to equation (1A). In particular, if the payment amounts increase by e% each period such that the first payment is A and the nth is A*(1 + e) ^ (n – 1), then we have

≈

. or

for ni ≤ 3.

Here A = P bˆ gives (approximately) the first payment which is subsequently indexed by e% per period. The error distribution is now complex, but there are peak errors of about ±10% for ni≤3, i≤0.1, e≤i so again it covers a wide range of “everyday” situations. The errors generally are worst as (i-e)→0.

≈

and

and with a Acknowledgement little The author would like to acknowledge the helpful comments of Prof. Moshe Milevsky, York University, Toronto on an early draft of this document.

References [1] G. A. Hawawini and A. Vora, The History of Interest Approximations, Arno Press, USA, 1980, ISBN 0-405-13480-0 [2] C. S. Park, R. Pelot, K. C. Porteous, M. J. Zuo, Contemporary Engineering Economics, Addison Wesley Longman, Toronto, 2001, ISBN 0-201-61390-5

. To avoid this problem perhaps a more useful approximation is for gives (approximately) the first payment. It is where n(i-e)≤2, i≤0.25, e≤i Again an escalating immediate payment stream similar to the above but asymptotically (i-e)→0 and of hastheutility, for Inexample, in estimating the first where the firstcorrect paymentas is at the beginning first period. this case have peak of aboutretirement ±10% for thispayouts formula: from a principal of P where the payment of we a stream oferrors escalating portfolio yield is i and the inflation rate is e. It has a sweet spot near n≈30, (i-e)≤0.6 which makes it ideal for such retirement income streams. For instance if a retiree has a principal of $500,000 and his portfolio has a yield of i=7% and he wants to match inflation with e=3% over a period of n=30 years then the approximate first annual payment isn(i-e)≤2, abouti≤0.25, $26,700. The true isgives about $27,400k. The difference is a wholly where n(i-e)≤2, e≤iAgain Again (approximately) the first payment. It is where i≤0.25, e≤i (approxiA =value P bˆ gives acceptable mately)-2.6%. the first payment. is asymptotically correct as (i-e)→0 asymptotically correct as It(i-e)→0 and has utility, for example, in estimating the first and has utility, for example, in estimating the first payment of from a principal of P where the payment of a stream of escalating retirement payouts a stream of escalating retirement payouts from a principal of P portfolio yield is i and the inflation rate is e. It has a sweet spot near n≈30, (i-e)≤0.6 where the portfolio yield is i and the inflation rate is e. It has a which makes it ideal such which retirement For instance if a retiree has a sweet spot near n≈30,for (i-e)≤0.6 makes itincome ideal forstreams. such principal of $500,000 and his portfolio has a yield of i=7% and he wants to match retirement income streams. For instance if a retiree has a prininflation e=3% a period n=30 years then the approximate first annual cipal ofwith $500,000 and over his portfolio has a of yield of i=7% and he wants is to match withThe e=3% overvalue a period of n=30$27,400k. years payment aboutinflation $26,700. true is about The difference is a wholly then the approximate first annual payment is about $26,700. The acceptable -2.6%. true value is about $27,400k. The difference is a wholly acceptable -2.6%.

Journal of Personal Finance

Race, Trust, and Retirement Decisions Terrance K. Martin Jr., Ph.D., Assistant Professor, Department of Economics and Finance, University of Texas-Pan American Michael Finke, Ph.D., Professor and Director of Retirement Planning and Living, Department of Personal Financial Planning, Texas Tech University Philip Gibson, Ph.D., Assistant Professor, Department of Accounting Finance and Economics, Winthrop University

Using the 2008 National Longitudinal Survey of Youth, this study investigates whether racial differences in trust can explain the decision to consult a financial planner and the variation in accumulated retirement wealth. Blacks and Hispanics are more likely to report having low trust compared to non-black, non-Hispanic respondents. The results provide evidence that low trust has a greater impact among blacks than non-black, non-Hispanic respondents. Low trust has a weaker impact among Hispanics on the decision to consult a financial planner and on the accumulation of retirement wealth. There is no evidence that hiring a financial planner has a larger impact on retirement wealth among blacks or Hispanics than it does among non-black, non-Hispanic households.

ÂŠ2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 2

Introduction Employee responsibility for retirement funding makes retirement savings and retirement planning decisions critical to future financial wellbeing. Unfortunately, the low levels of savings among many households threatens their financial security in retirement. Minority households, on average, have lower retirement savings (Ariel Education Initiative & Aon Hewitt, 2012) and lower levels of retirement planning (Honig, 1996). In 2007, more than 50% of black and Hispanic workers reported retirement savings of less than $10,000, while only 25% reported attempting to calculate retirement income needs (Helman, et al., 2007). Planning for retirement is complex for all households. A professional financial planner may help an individual improve their retirement savings behavior by providing access to expert financial knowledge (Collins, 2012), by expanding their comprehension of the consequences of low savings, and by reducing psychic costs of making complex choices. Yet, compared to non-black, non-Hispanic households, both black and Hispanic households have a lower likelihood of seeking professional financial advice (Helman et al., 2007). The main purpose of this study is to investigate whether racial differences in trust affect an individual’s decision to both seek financial advice for retirement and to save for retirement. Previous studies have explored behavioral and life cycle explanations for the wide variation in retirement savings outcomes (Bernheim et al., 2001, Bernartzi and Thaler, 2004). Financial literacy may also impact awareness of saving needs (Lusardi and Mitchell, 2009). This study contributes to the existing literature by demonstrating that trust can also be used to explain the variation in household retirement decisions and retirement savings. Prior studies have identified the importance of trust as a possible predictor of financial market participation. For example, Guiso,Sapienza and Zingales (2008) assess the role of trust in stock market participation, and show a positive relation between trust and rates of stock ownership. Bennett and Robson (2004) examine the effect of trust on the supply of advice and find that trust alone may not maximise the satistfaction of clients. In a recent study of defined contribution plan participation, Agnew, Szykman, Utkus, & Young (2012) evaluate trust’s role in voluntary and involuntary retirement plans. In this study we take a micro-level approach by evaluating the influence of trust among racial groups on the decision to save for retirement and obtain help from an advisor. Life experiences, values, communities and cultural backgrounds make race an important potential predictor of one’s ability to trust (Uslaner, 2002; Smith, 2010). To test our hypotheses, we draw on data from the National Longitudinal Survey of Youth (NLSY), which offers financial and demographic information as well as a commonly used measure of trust. The results of this study suggest that black and Hispanic respondents are less trusting than non-black, non-Hispanic respondents. However, the impact of low trust appears to be more relevant

among blacks than any other racial group. Among Hispanics, trust has a small impact on the decisions to consult a financial planner and to save for retirement. Low trust, more than cultural socialization, among blacks appears to affect the decision to consult a financial planner and to save for retirement relative to non-black, non-Hispanic respondents. The article proceeds as follows. First we provide a survey of the related literature demonstrating evidence of lower retirement wealth among blacks and Hispanics; here we also explore potential explanations for lower trust of financial services among minorities. Second, we draw on agency theory to inform the conceptual model and to test the hypotheses. Third, we describe our data and methods. Fourth, we present our results regarding the impact that trust has on retirement savings and on using a financial planner. Finally, we provide recommendations for future studies and present our conclusions.

Literature Review Use of a Financial Planner: Although it is documented that many households make poor financial decisions (Campbell, 2006), only 25% seek the assistance of a financial planner1 (Hanna, 2011). Financial planners often provide their clients with many services such as investment management, tax planning, estate planning, insurance planning, and retirement planning. There is evidence that those who employ the services of a financial professional are more likely to engage in positive financial behaviors. Marsden, Zick and Mayer (2011) find that clients who work with a financial professional benefit from improved goal setting, calculation of retirement needs, retirement account diversification and are more likely to have an emergency fund. Impact of trust: Trust involves a relationship with another person in which there exists both uncertainty and interdependence (Williams, 1993; Rousseau, Sitkin, Burt, & Camerer 1998). Trust can be divided into two distinct types: generalized trust and personalized trust. Generalized trust is a belief about a random individual among a group of distinguishable groups of individuals. Personalized trust is a belief about a specific indiviudal borne out of frequent interactions (Bottazzi, Rin, & Hellmann, 2011). Bjørnskov and Nannsted (2008) note factors that affect socio-economic differences within a country increase or decrease trust within that country. For example, income inequality is a robust predictor of generalized trust (Knack and Keefer, 1997; Ulsaner, 2002). Uslaner (2002) finds that blacks are less likely to report generalized trust than whites. Life experiences, values, communities, and cultural backgrounds contribute to one’s ability to trust (Nannsted, 2008; Smith, 2010; Uslaner, 2002).

Respondents are asked whether they use a financial planner without additional details such as whether the planner provides comprehensive financial advice, how they are compensated and trained, or whether they provide a fiduciary standard of care. 1

Journal of Personal Finance

Most commercial transactions include some element of trust Most commercial transactions include some element of trust (Arrow, 1972). A household will choose to rent the services of a financial planner only if the expected benefits exceed the expected costs. The transactions which occur between a financial planner and his/her clients can be viewed as a principle-agent relationship (client-financial planner) (Finke, Huston, & Waller, 2009). The principal (client) assumes that the agent (planner) will not make self-serving recommendations. Trust affects expectations within agency relationships (Tyler and Stanley, 2007). A worker with less general trust will be less willing to pay for professional financial advice. They will expect to spend more time monitoring advisor recommendations and will also assume that benefits will be reduced by self-serving recommendations. According to Tyler and Stanley (2007), trust is integral to service relationships because such services may be credence goods. Purchase of a credence good, such as financial advice, will always involve some degree of trust since quality is difficult to verify even after advice is received (Dulleck and Kerschbamer, 2006). Race, trust, and a financial planner: The use of a professional financial planner may improve retirement savings behavior by highlighting the consequences of inadequate savings (Campbell, 2006), improving portfolio diversification (Kramer, 2012), lowering portfolio turnover (Kramer and Lensik, 2012), improving investment performance, reducing mistakes (Horn et al., 2009), reducing cognitive load (Engelmann, et al., 2009), and by discouraging inertia (Chalmers and Reuters, 2012). However, the 2007 Minority Retirement Confidence Survey finds that black and Hispanic workers who save for retirement are more likely to seek the help of friends or family than any other source (Helman, et al., 2007). As an explanation of the difference in findings, Brehm and Rahn (1997) argue that experiences with discrimination may explain the pervasiveness of lower trust among blacks. Discriminatory practices from agents within the financial and capital markets, combined with a history of restricted access to these markets, may lead to a negative perception of all agents. Mistrust in health care expressed by blacks may, for example, be explained by differences in access to quality health care services which leads to racial differences in the trust of healthcare agents such as doctors (Boulware, Cooper, Ratner, LaVeist, & Powe, 2003; Brandon, Isaac, & LaVeist, 2005). Accordingly, racial differences in trust can hinder a minority householdâ&#x20AC;&#x2122;s willingness to consult a financial planner for retirement planning advice. In finance, there is evidence of racial discrimination in credit and consumer markets. In credit markets, Oliver and Shapiro (1997) report that black households pay on average 0.5% higher interest rates on home mortgages than white households even after controlling for creditworthiness. Furthermore, blacks and Hispanics were far more likely to be rejected for a mortgage (Munnell, Tootell, Browne, & McEneaney, 1996) at a comparable level of creditworthiness. Black households are also more likely to be offered mortgages from subprime lenders who charge higher credit-adjusted fees and interest rates (Williams et al. 2005). Black and Hispanic

households paid on average $339 more in automobile financing charges than white households of comparable credit risk profiles. Weller (2009) finds evidence that blacks face higher credit costs and are more likely than whites to be denied loans. Negative financial market experience may explain the resistance of blacks and Hispanics to seek professional financial advice for retirement. Race, trust, and retirement wealth: The persistent disparity in retirement plan participation and contribution rates may lead to a decline in retirement wellbeing for the two largest minority groups in the United States. Black and Hispanic households in a sample of seven large employers are less likely than nonblack, non-Hispanic households to opt into employer sponsored retirement plans, and less likely to change their default savings rate (Pagliaro & Utkus, 2011). Among those who do participate in retirement plans, blacks and Hispanics are less likely to contribute at a rate higher than white workers and are more likely to remain in the default investment option (Munnell & Sullivan, 2009; Pagliaro & Utkus, 2011). The Ariel Education Initiative and Aon Hewitt (2012) report significantly lower 401(k) savings among black and Hispanic households than white households. In 2010, among workers making less than $30,000, average retirement savings among black, Hispanic and white workers were $7,557, $8,949, and $14,563, respectively. A 2012 Vanguard study (Pagiliaro & Utkus, 2012) using individual account data from six large defined contribution plans finds that, on average, retirement wealth among black and Hispanic households was $17,367 and $9,784 lower than white households (Pagiliaro & Utkus, 2012). Brown (2007) observes a lower likelihood of black and Hispanic participation in financial markets. Brown (2007) also finds that low-income whites are more likely to invest in financial markets than higher income black and Hispanic households. Among those that do participate in financial markets, blacks had lower net growth in their retirement accounts than Hispanics and whites during the great recession (Kochlar, Fry, & Taylor, 2011). Within the same salary quantiles, the percentage change in mean retirement savings between 2007 and 2010 among black workers was less than 1% while white workers realized an average increase of 5.25% (Ariel Education Initiative and Aon Hewitt, 2012)

Theoretical framework Households with limited financial knowledge should seek the services of a professional financial advisor if the expected benefit of improved choice exceeds the costs of advising services. A greater level of distrust increases both the perceived agency costs of hiring an agent and decreases the perceived benefit of following the advice2 (Zingales, 2011). An agency relationship exists whenever principals hire agents, to perform some service and delegate some decision making authority to the agents (Jensen & Meckling, 1976). 2

The presence of high monetary costs and the absence of high bonding costs.

ÂŠ2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 2

Conflicts of interest potentially exist in any transaction in which decision making is delegated to a more knowledgeable agent who has an incentive to act in his/her own interest. In a relationship where trust is present, there is a reduction in the perceived costs of the principle–agent exchange. This is due to the reduced need to monitor performance by evaluating the quality of advice and identifying self-serving behavior (Moore, 1999). Conversely, trust may also make rent extraction easier, leading to lower-quality client recommendations. For this reason, trust is viewed as a client’s perception of the agency costs associated with consulting a financial planner for retirement. Since there is evidence that financial service professionals are more likely to engage in self-serving behavior in transactions with minority customers3, we hypothesize blacks and Hispanics with lower levels of general trust will be significantly less likely to rely on expert financial advice and accumulate retirement wealth.

Method Data: This paper uses data from the 2008 administration of the National Longitudinal Survey of Youth (NLSY79). The survey consists of a sample of 12,686 randomly selected men and women born between 1957 and 1965. The respondents in this survey are ideal for this study since they are in their peak life cycle saving years. The sample includes those who were interviewed for the NLSY79 retirement module in 2008. Sixty percent of the respondents were asked a series of retirement planning questions. The sample was further reduced to include only those who answered whether they consulted a financial planner and completed the trust instrument. The final sample consists of 7,550 respondents. The racial composition of the final sample includes 2,346 blacks, 1,449 Hispanics and 3,755 non-black, non-Hispanics. The trust distribution of the sample includes 2,590 low trust respondents, 2,190 moderate trust respondents, and 2,770 high trust respondents. The NLSY79 oversamples black and Hispanic households, making it an ideal data source for studying racial differences.

Empirical Models Trust Model: In this study we focus on the impact of generalized trust. Trust is investigated first as an effect (dependent variable), second as a cause (independent variable), and last as a moderating variable. In 2008, NLSY administrators included a question that measures the respondent’s level of generalized trust. Respondents were asked, “Generally speaking, how often can you trust other people? (Always, most of the time, about half the time, once in a while, or never).” Sapienza, Toldra-Simats, and 3 It is also possible that the characteristics of financial planners may differ by race if, for example, those who cater to minority clients tend to be less highly trained, provide less comprehensive services, or are more likely to receive transaction-based compensation.

Zingales (2013) and Ulsaner (2011) confirm the effectiveness of this trust instrument as a measure of economic outcomes. We create three dichotomous trust categories: low trust, moderate trust, and high trust. If respondents chose option 5 (never), the respondents are classified as ‘low trust’. Response options 4 (once in a while) and 3 (half of the time) are classified as moderate trust. If respondents answered 2 or 1, the respondents are classified as high trust. Using an ordered logistic regression model, trust is modelled as a function of factors found to be empirically related to generalized trust. Variables include race, religion, education, marriage, divorce, death of a spouse, as well as age, gender, wage income, and wealth (see Bjørnskov, 2007; Nannsted, 2008). Planner use model: In the 2008 retirement model, respondents are asked “Have you [or] [Spouse/partner’s name] ...consulted a financial planner about how to plan your finances after retirement?” 17.35% of the sample report having consulted a financial planner. There is no further information about the type of financial planner used by the household, so the planner may not necessarily provide extensive retirement planning advice and may include professionals using the term planner from various financial services business models. This article largely adapts the empirical model used by Hanna (2011) with the addition of this article’s hypothesis variables trust and an interaction of race and trust. We use a binary logistic regression to model the consultation of a financial planner for retirement. The consultation of a financial planner is modeled as a function of trust, race, trust and race interaction term, age, marital status, education level, home ownership, income, net worth, and risk preference. The full sample is used to compute the analysis. We separately estimate the impact of race and trust on retirement savings, and then both race and trust are included in the full model as well as an interaction term. Accumulated retirement wealth model: Retirement wealth accumulated over the life cycle is modeled as a function of factors affecting household savings decisions. The retirement wealth variable is a linear combination of household assets held within qualified retirement accounts as well as assets held in tax advantaged accounts including individual retirement accounts (IRAs). Control variables including financial planner, financial planner interacted with race, education, marital status, home ownership, business ownership, age, gender, family size, income, net worth and risk preferences. Due to a large percentage of the sample having $0 retirement wealth, a Type I Tobit regression model is used at a left censored minimum of $0 (Tobin, 1958). We report the marginal effects for the unconditional expected value of the accumulated retirement wealth, E (retirement wealth*), where retirement wealth* = max (0, min (retirement wealth, upper bound max)). Four specifications of the model are used: race and financial planner only; race, financial planner, and trust only; race, financial planner, and the interaction of race and financial planner only; and lastly race, financial planner, and the interaction of race and financial planner and trust.

Journal of Personal Finance

Results Descriptive Results: Table 1 presents summary frequencies. Blacks are more likely to report low trust and less likely to report high trust relative to the other race groups. 47% of black respondents identify themselves as having low trust while only 21% consider themselves high trusting. Hispanics show a higher propensity to report low trust compared to non-black, non-Hispanics. Non-black, non-Hispanic respondents show a higher percentage of respondents reporting high trust than other groups, and the lowest percentage identifying themselves as low trust (49% and 23%, respectively). Across all races, respondents are more likely to have high trust as income and net worth increase. For example, among Hispanics in the lowest income quintile 59% have low trust, while only 19% report high trust. Among blacks in the lowest net worth quintile 53% have low trust, while only 14% report high trust. Table Table 11 Descriptive Descriptive Statistics: Statistics: Frequencies Frequencies byby race race group group and and trust trust level. level. Hispanic Hispanic Low Moderate Moderate High High Low Trust Trust Trust Trust Trust Trust Sample Sample 4242 2727 3131 Marital Marital Status Status 5050 2727 2323 Never Never Married Married 3636 2525 3636 Married Married 5353 3131 1616 Separated Separated 4848 2222 2929 Divorced Divorced 5252 2828 2020 Widowed Widowed Education Education Level Level 4747 2828 2626 High High School School oror Less Less 1212 2222 6666 Some Some College College 2222 3131 4747 College College 3838 2424 3838 Graduate Graduate Permanent Permanent Income Income Lowest LowestIncome Income quintile quintile quintile quintile 22 quintile quintile 33 quintile quintile 44 Highest Highest Income Income quintile quintile Net Net Worth Worth Lowest Lowest NW NW quintile quintile quintile quintile 22 quintile quintile 33 quintile quintile 44 Highest Highest NW NW quintile quintile

Black Black Low Moderate Moderate High High Low Trust Trust Trust Trust Trust Trust 4747 3232 2121

White White Low Moderate Moderate Low Trust Trust Trust Trust 2323 2828

High High Trust Trust 4949

5252 4242 4545 5151 5353

3131 3333 3434 3030 3535

1717 2525 2121 1818 1212

2727 1919 4343 3131 3131

2929 2727 2929 2929 3333

4444 5353 2929 4040 3737

5151

3131

1818

2929

3030

4141

2222 3131 4949

3838 3838 3030

4040 3131 2222

99 1414 2020

2121 2525 3030

7070 6161 5151

5959 4545 4141 2828 2828

2222 3232 2626 3030 2525

1919 2323 3232 4242 4747

5757 5050 4343 3434 3333

3030 2828 3333 3838 3535

1313 2222 2424 2828 3232

4242 2929 2525 2121 1313

2828 3030 2929 2727 2525

2929 4242 4646 5252 6262

5757 5050 3838 3232 2727

2424 2525 3131 3030 2525

1919 2424 3131 3838 4848

5353 4949 4848 3636 3737

3232 3232 2929 3535 3131

1414 1919 2323 2929 3232

3535 3434 2828 1919 1414

3030 3131 3030 2626 2525

3434 3535 4242 5454 6161

Source Source National National Longitudinal Longitudinal Survey Survey ofof Youth Youth 1979, 1979, 2008 2008 Administration Administration

2222

ÂŠ2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 2

In Table 2, frequencies are reported for those consulting a financial planner by race and level of trust. Approximately 17.4% of the respondents reported consulting a financial planner for retirement. For Hispanics and non-black non-Hispanics, the likelihood of consulting a financial planner increases with trust. However, among blacks 37% report low trust, 34% report moderate trust, and 30% report high trust. Trust is much higher among non-Hispanic, non-blacks among whom 14% report low trust, 26% moderate trust, and 60% high trust. Table 2 Descriptive Statistics: Frequencies of the use of a financial planner by trust and race group. Hispanic Black Low Moderate High Low Moderate Trust Trust Trust Trust Trust Consulting Financial Planner 30 26 44 37 34 Marital Status Never Married 36 36 28 48 33 Married 28 23 50 29 35 Separated 0 100 0 54 15 Divorced 38 25 38 37 39 Widowed 0 100 0 100 0 Education Level High School or Less 35 31 34 40 37 Some College 9 18 73 30 37 College 26 30 44 30 30 Graduate 33 17 50 47 24 Permanent Income Lowest Income quintile 43 43 14 64 23 quintile 2 50 33 17 49 36 quintile 3 38 23 38 34 35 quintile 4 25 22 53 25 31 Highest Income quintile 25 25 49 28 38 Net Worth Lowest NW quintile 25 50 25 38 34 quintile 2 56 25 19 37 35 quintile 3 38 24 38 41 36 quintile 4 27 40 33 31 33 Highest NW quintile 23 14 63 38 30

High Trust

Low Trust

White Moderate Trust

High Trust

18 37 31 24 0

17 13 0 21 10

25 25 43 31 30

58 62 57 49 60

23 33 40 29

20 7 10 15

30 20 25 30

50 72 65 55

14 15 31 44 34

32 16 18 18 10

26 35 22 25 27

42 49 60 56 64

28 28 24 36 32

16 21 23 16 11

30 33 33 22 26

54 46 44 61 64

Source National Longitudinal Survey of Youth 1979, 2008 Administration

Table 3 shows mean values of accumulated retirement wealth 23 by race and trust level for all households and then by those consulting a financial planner. Regardless of race, respondents reporting low trust have accumulated less retirement wealth. In general, blacks report less retirement wealth among all households and also

amongst respondents consulting a financial planner. The mean retirement wealth of low trusting blacks is $23,128 compared to $67,116 for the high trusting black respondents. Low trusting Hispanics on average have only accumulated 32% of the retirement wealth of high trusting Hispanics - $27,676 versus $87,836.

Table 3 Table showing the mean retirement wealth by race and trust for all households and by those using a financial planner

Low Trust

Hispanic

Black

White

Moderate Trust

High Trust

Retirement Wealth

27,677

$ 61,286

Retirement Wealth

56,042

$ 97,841

$ 202,775

87,836

Low Trust

All Households

23,128

$ 71,894

Low Trust

$ 33,560

$ 67,116

$ 91,762

By use of a financial planner

Source National Longitudinal Survey of Youth 1979, 2008 Administration

High Trust

59,258

59,167

99,105

High Trust

99,865

152,956

$ 199,677

240,218

Journal of Personal Finance

Multivariate Analysis Table 4 provides the proportional odds estimates obtained using an ordered logistic regression model. Recall that trust has three levels: low, average, and high. Blacks and Hispanics are 50% less likely to have high trust versus lower trust categories than nonblack, non-Hispanics. Education, income, and net worth positively affect the odds of being in the high trust group compared Table 4 Binary logistic regression: Dependent variable is trust level. Parameter Race (ref. non-Black, non-Hispanic) Black Hispanic Education (ref. High School or less) Some College College Graduate School Marital Status (ref Never married) Married Separated Divorced Widowed Income (ref. Lowest quintile) quintile 2 quintile 3 quintile 4 Highest Income quintile Net Worth (ref Lowest quintile) quintile 2 quintile 3 quintile 4 Highest NW quintile Religious Affiliation (ref. None) Protestant Baptist Episcopalian Lutheran Methodist Presbyterian Catholic Jewish Other Risk Preference (ref. Conservative) Aggressive Moderate Other Factor Age Pseudo R square

0.499 0.509

*** ***

2.663 1.697 1.245

*** *** ***

1.174 0.995 0.975 0.898

1.246 1.282 1.402 1.436

*** *** *** ***

1.092 1.201 1.601 1.779

** *** ***

1.014 0.945 1.597 1.33 1.139 1.124 1.248 1.21 1.027 1.129 1.204 1.006 0.1674

***,**,* indicates significance at the 0.01, 0.05, and 0.1 levels, respectively Source National Longitudinal Survey of Youth 1979, 2008 Administration

** **

* ***

to the low trust group. As income increases, the odds of being in the high trust group increases compared to the lowest income quintile. Religiosity increases the odds of being in the high trust group. Episcopalian and Lutheran have 1.59 times and 1.33 times higher odds of being in the high trust group versus the lower trust groups compared to respondents with no religious affiliation.

Odds of the likelihood of consulting a financial planner are presented in Table 5. The first model only controls for race and all other control variables with the exception of trust. The second controls for trust level all other control Tablemodel 5 Binary logistic regression: Use ofand a financial planner is thevariables outcome. with the exception of race. The third controls for both race and trust. In the first model, there is no significant difference between Race Only Trust Only black, non-black and non-Hispanic respondents on the decision Parameter to(ref. consult a financial planner. Hispanic respondents are 29% less Race non-Black, nonlikely than non-black, non-Hispanic respondents to consult a Hispanic) financial planner for retirement. In the second model in which we Black 1.12

Hispanic 0.71 *** Trust (ref. Moderate Table 5 Binary logistic regression: Use of a financial planner is the outcome. Trust) Low Trust 0.82 ** Trust Race Only Trust Only Race and Parameter High Trust 1.08 Race (ref. non-Black, nonEducation (ref. High Hispanic) School or less) Black 1.12 1.18 * Some College 2.600.73****** Hispanic 0.71 ***2.69 *** College 2.07 *** 2.05 *** Trust (ref. Moderate Trust) Graduate School 1.62 *** 1.59 *** Low Trust 0.82 ** 0.82 ** Marital Status (ref Never High Trust 1.08 1.10 married)(ref. High Education Married 0.99 0.96 School or less) Some College 2.69 ***0.55 **2.60 *** Separated 0.562.57** *** College 2.07 *** 2.05 *** Divorced 1.01 0.982.01 *** Graduate School 1.62 *** 1.59 *** 1.60 *** Widowed 1.09 1.10 Marital Status (ref Never Income (ref. Lowest married) Married 0.99 0.96 0.99 quintile) Separated 0.55 ** 0.56 ** 1.00 0.970.55 ** quintile 2 Divorced 1.01 0.98 1.01 1.41 *** 1.371.10** quintile 3 Widowed 1.09 1.10 1.41 *** 1.38 *** Income (ref. 4 Lowest quintile quintile) 1.91 *** 1.87 *** Highest Income quintile 1.00 0.97 0.99 quintile 2 Net Worth (ref Lowest quintile) 1.41 *** 1.37 ** 1.40 *** quintile 3 0.931.39 *** quintile 1.41 ***0.93 1.38 *** quintile 4 2 1.91 ***1.20 1.87 *** 1.181.88 *** Highest Income quintile 3 quintile Netquintile Worth (ref Lowest quintile) 2.16 *** 2.06 *** 4 0.93 0.93 quintile 2 3.22 *** 3.010.93 *** Highest NW quintile 1.20 1.18 1.19 quintile 3 Risk Preference (ref. Conservative) 2.16 *** 2.06 *** 2.10 *** quintile 4 3.22 ***1.19 * 3.01 *** Aggressive 1.193.14* *** Highest NW quintile Risk Preference (ref. Conservative) Moderate 1.22 ** 1.22 ** Aggressive 1.19 * 1.19 * 1.18 * Other Factors Moderate 1.22 ** 1.22 ** 1.21 ** Family Size 1.00 0.99 Other Factors Family 1.00 0.99 Age Size 1.01 1.011.00 Age 1.01 1.01 Home owner 1.07 1.081.01 Home owner 1.07 1.08 1.07 Business Owner 1.86 *** 1.861.85****** Business Owner 1.86 *** 1.86 *** Pseudo R square 0.1961 Pseudo R square 0.1998 0.1976 0.1976 0.1961 NN 7550 7550 7550 75507550 ***, indicates significance at the 0.01, and 0.10.05, levels,and respectively. ***,**,* **,* indicates significance at 0.05, the 0.01, 0.1 levels, Source National Longitudinal Survey of Youth 1979, 2008 Administration

respectively. Source National Longitudinal Survey of Youth 1979, 2008 Administration

ÂŠ2014, IARFC. All rights of reproduction in any form reserved.

Race

1 0

0 1

2 2 1

0 0 1 1

0 1 1 1

0 1 2 3

1 1

1 1 1 1 0.19 75

Volume 13, Issue 2

control for trust but not race, the results indicate that respondents with low trust are 18% less likely to consult a financial planner. However when we control for general trust, black respondents are more likely to choose a financial planner while Hispanic respondents are still less likely to see a planner. Education, income, net worth, and risk preference are all positively associated with the decision to consult a financial planner. The marginal effects for the unconditional expected value of accumulated retirement wealth obtained from Tobit regression analysis are reported in Table 6. In the first model, there is a significant difference in retirement wealth between black and Hispanic respondents compared to non-black, non-Hispanic respondents. On average, black respondents report $9,023 less retirement wealth than non-black, non-Hispanic. Likewise, Hispanic respondents report $11,851 less retirement wealth than non-black, non-Hispanic. Respondents using a financial planner on average report accumulating $25,742 more retirement wealth than those not using a financial planner for retirement advice. In the second model, trust level is added. The effect of being black on the outcome variable is no longer significantly different from non-black, non-Hispanic respondents, and the effect

of being Hispanic remains significant but to a lower degree. On average, Hispanics record $9,625 lower retirement savings than non-black, non-Hispanic respondents. Respondents with low trust on average report $11,703 less retirement wealth than those with moderate trust. In the third model, two interaction terms are added (black*Financial Planner and Hispanic*Financial Planner). The results suggest that there is no evidence of a statistically significant difference in the effect of having a financial planner. The partial effect of being black loses statistical significance at the 5% alpha level but not at a 10% level. In the final model, all control variables are included. The effect of blacks once again disappears when trust level is controlled for in the model. Likewise, the effect of Hispanic loses statistical significance at the 5% level. The use of a financial planner has the largest positive impact on retirement savings ($27,414), and those with lower trust have saved $7,130 less. Although blacks have saved less for retirement, this may be explained by their lower trust and then by their less frequent use of a financial planner. Hispanics, on the other hand, still have significantly less saved for retirement even after accounting for control variables including trust and financial advice.

Table 6 Tobit regression: dependent variable is retirement wealth. Parameter Race (ref. non-Black, non-Hispanic) Black Hispanic Financial Planner Interaction Terms Black*Financial Planner Hispanic*Financial Planner Trust (ref. Moderate Trust) Low Trust High Trust Education (ref. High School or less) Some College College Graduate School Marital Status (ref Never married) Married Separated Divorced Widowed Log of Income Log of Net Worth Risk Preference (ref. Conservative) Aggressive Moderate Other Factors Family Size Age Male Home owner Business Owner

Race and FP

-$9,023.20 -$11,851.03 $25,742.48

** ** ***

Race, FP and Trust

-$5,455.20 -$9,652.28 $24,791.00

** ***

-$11,703.25 $7,506.51

** *

Race, FP , Interaction

Race, FP Interaction, and Trust

-$7,885.93 -$10,575.56 $28,409.30 -$5,430.72 -$7,256.31

-$4,313.77 -$8,381.17 $27,414.17 -$5,385.11 -$7,130.27

* ** ***

* ***

-$11,694.98 $7,511.77

** *

$69,245.85 $49,950.48 $16,000.74

*** *** **

$63,850.85 $46,937.80 $14,951.60

*** *** **

$69,125.90 $49,919.16 $15,985.48

*** *** **

$63,735.59 $46,909.10 $14,936.83

*** *** **

$38,635.05 $768.71 -$25,951.91 $2,552.96 $4,203.28 $11,609.76

***

*** ***

$38,058.92 $730.47 -$25,118.17 $2,956.90 $4,170.68 $11,409.91

***

*** ***

$38,729.31 $617.31 -$25,917.77 $2,611.52 $4,216.11 $11,604.12

***

*** ***

$37,964.79 $881.09 -$25,152.43 $2,898.48 $4,157.94 $11,415.60

$26,253.79 $23,030.66

*** ***

$25,519.35 $22,157.71

*** ***

$26,249.59 $23,026.47

*** ***

$25,514.43 $22,153.56

*** ***

* *** **

$1,064.87 $963.25 -$6,630.65 $19,268.11 -$18,123.24

* *** ***

$1,161.84 $1,007.35 -$6,980.06 $19,550.38 -$17,913.72

* *** **

$1,039.02 $963.02 -$6,643.63 $19,362.73 -$18,311.61

* *** ***

$1,188.40 $1,007.55 -$6,966.67 $19,454.71 -$17,720.21

***, **,* indicates significance at the 0.01, 0.05, and 0.1 levels, respectively. Source National Longitudinal Survey of Youth 1979, 2008 Administration

* *** ***

Journal of Personal Finance

Conclusion and Implication Trust is the expectancy of positive outcomes one may receive based on the expected action of another party in an interaction characterized by uncertainty (Bhattacharya, et al., 1998). Trust can be further categorized as generalized or personalized. Using the NLSY79 data, we estimate the relation between generalized trust and both consulting a financial planner and accumulated retirement wealth. Multivariate results suggest that blacks and Hispanics are more likely to report low generalized trust. Race is a strong predictor of the willingness to report trusting others. Black and Hispanic respondents are 50% less likely to report having higher trust compared to non-black, non-Hispanic respondents. The racial difference observed in the data is consistent with previous findings of low trust among minority groups (Pager and Shepherd, 2008; Smith, 2010). Race and trust are also related to the use of a financial planner for retirement concerns. It is worth noting that when only race is controlled in the model there is no statistical difference between blacks and non-black, non-Hispanic respondents in the likelihood of consulting a financial planner. However, Hispanics are 29% less likely than non-black, non-Hispanic respondents to consult a financial planner. Respondents with low trust are 18% less likely than moderate trust respondents to seek retirement advice from a financial planner. When race and trust are controlled in the same model, black respondents become statistically more likely (18%) to see a financial planner. This suggests that the use of financial planners would be slightly higher if blacks had higher general levels of trust, but the lower use of financial planners among blacks can be primarily attributed to demographic characteristics other than trust. The marginal impact of financial planner use among Hispanics is not increased when we include general trust. This suggests that trust is not the primary barrier to the use of financial planners by Hispanics. Race and trust are also related to the accumulation of retirement wealth. There is strong evidence of racial differences in the accumulated retirement wealth. Even when controlling for the use of a financial planner, blacks and Hispanics on average report $9,023 and $11,851 less retirement wealth than non-black, non-Hispanic respondents. When trust is added to the model, the difference between blacks and non-blacks, non-Hispanics decreases to $4,314 and is no longer statistically significant. This result suggests that low trust may be inhibiting blacksâ&#x20AC;&#x2122; willingness and/ or ability to save for retirement. Also, once trust is controlled the retirement savings disparity among Hispanics relative to non-black, non-Hispanic respondents is reduced by $2,200, but remains statistically lower. The strongest predictors of retirement saving are education, risk tolerance, home ownership and marital status. Differences in these household characteristics among racial groups explain most of the disparity in retirement savings among these groups. Cultural mistrust is defined as the lack of trust that a minority group may have for whites (Terrell & Terrell, 1981). In

discussing the black community, Terrell and Terrell (1981) point out that because blacks have a long history of financial market discrimination, they have developed mistrust of investing in these markets. This general lack of trust among black and Hispanic households may produce an affinity to work with black or Hispanic personal financial advisors. Affinity refers to a preference to work with someone based on common characteristics that are specific to a particular group (Perri & Brody, 2012). However, compared to whites, the black and Hispanic financial advisors are underrepresented within the advising community. According to the 2013 Bureau of Labor Statistics, only 6.5% of personal financial advisors identified as being black or African American and 5.2% identified themselves as Hispanic and Latino4 Due to the limited availability, black and Hispanic households might be reluctant to seek financial advice because they are unable to find someone who shares a similar background.

References Agnew, J. R., Szykman, L. R., Utkus, S. P., & Young, J. A. (2012). Trust, plan knowledge and 401(k) savings behavior*. Journal of Pension Economics and Finance, 11(1), 1. Arrow, K. J. (1972). Gifts and exchanges. Philosophy & Public Affairs, 343362. Bennett, R. J., & Robson, P. J. (2004). The role of trust and contract in the supply of business advice. Cambridge Journal of Economics, 28(4), 471-488. Bhattacharya, R., Devinney, T. M., & Pillutla, M. M. (1998). A formal model of trust based on outcomes. Academy of management review, 23(3), 459-472. Bottazzi, L., Da Rin, M., & Hellmann, T. F. (2011). The importance of trust for investment: Evidence from venture capital (No. w16923). National Bureau of Economic Research. Boulware, L. E., Cooper, L. A., Ratner, L. E., LaVeist, T. A., & Powe, N. R. (2003). Race and trust in the health care system. Public health reports, 118(4), 358. Brandon, D. T., Isaac, L. A., & LaVeist, T. A. (2005). The legacy of Tuskegee and trust in medical care: is Tuskegee responsible for race differences in mistrust of medical care? Journal of the National Medical Association, 97(7), 951. Brehm, J., & Rahn, W. (1997). Individual-level evidence for the causes and consequences of social capital. American journal of political science, 999-1023. Brown, D. A. (2007). Pensions and Risk Aversion: The Influence or Race, Ethnicity, and Class on Investor Behavior. Lewis & Clark L. Rev., 11, 385. Campbell, J. (2006). Household Finance. Journal of Finance, 1553-1604. Dulleck, U., Kerschbamer, R., & Sutter, M. (2009). The economics of credence goods: On the role of liability, verifiability, reputation and competition. Guiso, L., Sapienza, P., & Zingales, L. (2008). Trusting the stock market. The Journal of Finance, 63(6), 2557-2600.

http://www.bls.gov/cps/cpsaat11.pdf.

ÂŠ2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 2

Hanna, S. D. (2011). The Demand for Financial Planning Services. Journal of Personal Finance, 10(1).

Pagliaro, C. A., & Utkus, S. P. (2011). Diversity and defined contribution plans: The role of automatic plan features. Vanguard Research.

Helman, R., Copeland, C., & VanDerhei, J. (2012). The 2012 Retirement Confidence Survey: Job insecurity, debt weighs on retirement confidence, savings. EBRI Issue Brief, (369).

Perri, F. S., & Brody, R. G. (2012). The optics of fraud: affiliations that enhance offender credibility. Journal of Financial Crime, 19(3), 305-320.

Helman, R., VanDerher, J., & Copeland, C. (2007). Minority workers remain confident about retirement, despite lagging preparations and false expectations. Washington D.C: Employee Benefit Research Institute Honig, M. (1996). Retirement Expectations: Differences by Race, Ethnicity, and Gender. The Gerontologist, 36(3). Kochlar, R., Fry, R., & Taylor, P. (2011). Wealth Gaps Rise to Record Highs between Whites, Blacks and Hispanics. Pew Research Center. Marsden, M., Zick, C. D., & Mayer, R. N. (2011). The value of seeking financial advice. Journal of family and economic issues, 32(4), 625-643. Moore, M. (1999). Truth, trust and market transactions: What do we know? The Journal of Development Studies, 36(1), 74â&#x20AC;&#x201C;88. Munnell, A. H., Tootell, G. M., Browne, L. E., & McEneaney, J. (1996). Mortgage lending in Boston: Interpreting HMDA data. The American Economic Review, 25-53. Oliver, M., & Shapiro, T. (2006). Black Wealth/White Wealth: A New Perspective on Racial Inequality 2E. Routledge.

Rousseau, D. M., Sitkin, S. B., Burt, R. S., & Camerer, C. (1998). Not so different after all: A cross-discipline view of trust. Academy of management review, 23(3), 393-404. Sapienza, P., Toldra, Simats, A., & Zingales, L. (2013). Understanding trust. The Economic Journal. Smith, S. S. (2010). Race and trust. Annual Review of Sociology, 36, 453475. Smith, T. W. (1997). Factors relating to misanthropy in contemporary American society. Social Science Research, 26(2), 170-196. Terrell, F., & Terrell, S. L. (1981). An inventory to measure cultural mistrust among Blacks. The Western Journal of Black Studies. Uslaner, E. M. (2002). The moral foundations of trust. Cambridge University Press. Weller, C. E. (2009). Credit access, the costs of credit and credit market discrimination. The Review of Black Political Economy, 36(1), 7-28. Zingales, L. (2011). The role of trust in the 2008 financial crisis. The Review of Austrian Economics, 24(3), 235-249.

Journal of Personal Finance

Loss Aversion under Cognitive Load Michael A. Guillemette, Ph.D., CFP®, Assistant Professor, Department of Personal Financial Planning, University of Missouri Russell N. James III., Ph.D., J.D., CFP®, Professor and CH Foundation Chair in Personal Financial Planning, Director of Graduate Studies in Charitable Planning, Department of Personal Financial Planning, Texas Tech University Jeff Larsen, Ph.D. Associate Professor, Department of Psychology, University of Tennessee

An experiment was conducted to explore whether loss aversion is altered when individuals are placed under a higher level of cognitive load. The coefficients of monetary loss aversion were measured for 30 participants under low and high cognitive load. Memorizations of differing spans of digits were used to manipulate cognitive load. Participants’ skin conductance was measured to quantify emotional responses to gains and losses. No statistically significant evidence was found that loss aversion, as measured by choice, is altered when individuals are placed under a higher level of cognitive load. Statistically significant evidence was found that a higher level of cognitive load temporarily reduces an individual’s emotional arousal to absolute and relative small dollar losses.

Volume 13, Issue 2

Introduction The Great Recession tested the ability of financial planners to keep their clients invested in the stock market when prices fell dramatically between 2008 and 2009. Ideally, comprehensive financial planners would like clients to focus on their longterm goals when they are making financial decisions. However, misguided expectations and cognitive biases often cause clients to focus on irrelevant short-term market returns. One of the most difficult tasks a financial planner has is to help clients understand the emotional triggers that may cause them to make unsatisfactory financial decisions. If a client experiences persistent cognitive biases there may be an increased probability of sub-optimal financial behaviors, such as myopia (short-term thinking) and the tendency to be more sensitive to losses relative to gains (loss aversion). Strategies that distract clients from short-term gains and losses may help them avoid these cognitive biases. Myopia and loss aversion are two cognitive biases that cause people to make suboptimal choices. Individuals who exhibit myopic thinking are more likely to favor investment outcomes that are more present oriented (Finke & Huston, 2004). When individuals were shown monthly stock returns versus annualized stock returns, they invested a higher percentage of their hypothetical portfolio in bond funds (Thaler, Tversky, Kahneman & Schwartz, 1997). A strong positive correlation has also been found between stock market returns and risk tolerance, and individuals are less risk tolerant when equity valuations are more favorable (Guillemette & Finke, 2014). People tend to take less risk and pull money out of equity mutual funds following a decline in stock prices when fear and uncertainty are high. This is consistent with the finding that prior losses result in greater levels of loss aversion (Barberis, Huang & Santos, 2001). Individual equity investors lost 1.56 percent annually in dollar-weighted returns between 1991 and 2004 due to poor market timing (Friesen & Sapp, 2007). The purpose of this paper is to examine the association between cognitive load and response towards loss. If a client experiences a monetary loss, but is distracted (i.e. under higher cognitive load), they may have less of an emotional response to the loss. This is because the distraction may help reduce myopic behavior, thereby reducing loss aversion (Benartzi & Thaler, 1995). This study aims to answer two questions. The first question is whether loss aversion is reduced when people are distracted and choose between monetary gains and losses. The second question is whether distractions reduce physiological response toward monetary gains and losses. Physiological response, which can also be thought of as emotional arousal, is quantified based on the amount of sweat-induced moisture on the skin, as sweat is controlled by the central nervous system.

Review of Literature Theoretically risk preferences can be described using either expected utility theory or prospect theory. The word “risk

tolerance” is commonly used by financial planners to describe the level of risk a client is willing to accept. Risk tolerance, from an expected utility theory standpoint, is measured by the willingness to accept variation in monetary outcomes, or consumption (Hanna, Fan & Chang, 1995). Clients who are more willing to accept variation in monetary outcomes are more risk tolerant, or less risk averse, compared to people who prefer greater certainty. The word “loss aversion” is also used by financial planners to describe clients’ willingness to accept risk. Prospect theory modified the expected utility theory concept of risk tolerance by overweighting losses from an arbitrary reference point (Kahneman & Tversky, 1979). Expected utility theory treats gains and losses the same, whereas evidence suggests that the dissatisfaction people experience from losses is approximately two times more than the satisfaction they receive from gains (Tversky & Kahneman, 1992). Therefore, loss aversion may better explain clients’ willingness to accept risk within the loss domain. Recent research provides evidence that loss aversion can be altered. The perspective people take when making decisions between certain monetary amounts and gambles has been found to alter their aversion to losses (Sokol-Hessner et al., 2009). When people perceived their monetary choices as part of a holistic process, instead of in isolation, they became less loss averse. Sokol-Hessner et al. (2009) defined the holistic approach as emphasizing choices in their greater context, such as creating a financial portfolio. An understanding of the two brain systems that guide decision making is important in determining whether loss aversion is altered when individuals are placed under greater cognitive load. Cognitive-experiential self-theory (CEST) characterizes a dual system that competes for control of our actions and inferences. CEST emphasizes two distinct but interactive systems of information processing which are defined as rational and experiential processing (Epstein, 1994). The rational system is logical, analytic and involves a conscious appraisal of events. The experiential system involves preconscious, impulsive processing. Processing in the rational system is slower, and action is more delayed, compared to the experiential system. The rational system is unique to humans and uses the working memory system of the prefrontal cortex (Baddeley, 2000). The CEST dual-system model is similar to a wide variety of other dual-system models proposed for consumer and financial decision making such as the dual-self model (Fudenburg & Levine, 2006), the planner-doer model (Thaler & Shefrin, 1981), the affective and deliberative systems (Loewenstein & O’Donoghue, 2004), the Type 1 and 2 model popularized by Daniel Kahneman (Kahneman, 2011) and even the passions-spectator model proposed by Adam Smith in the 18th century (Ashraf, Camerer & Lowenstein, 2005). There is growing evidence from the field of neuroeconomics that the amygdala, which is part of the experiential system, plays a significant role in individuals’ aversion to losses. The amygdala is an almond-shaped collection of neurons located deep within the medial temporal lobe of the brain. The amygdala is involved in processing emotion and learning, particularly for

Journal of Personal Finance

negative outcomes, and it plays a key role in the processing of losses (Bechara, Damasio, Damasio & Lee, 1999; Breiter, Aharon, Kahneman, Dale & Shizgal, 2001). Two individuals with focal bilateral amygdala lesions were studied under a series of monetary gain and loss gambles. The two individuals showed a dramatic reduction in loss aversion compared to the control participants (Martino, Camerer & Adolphs, 2010). Loss aversion has also been found to correlate with amygdala activity in response to losses relative to gains under mixed (gain or loss) gambles (Sokol-Hessner, Camerer &Phelps, 2012). When individuals are placed under greater cognitive load it activates the working memory system of the prefrontal cortex.

Figure 1. Experimental design example

Slide 1 Remember this number sequence in the order it is displayed: 7872319 Press the SPACE BAR to continue.

LOSS AVERSION UNDER COGNITIVE LOAD

Physiological response to loss is typically measured using skin conductance response (SCR). SCR, also referred to as electrodermal response or galvanic skin response, measures emotional arousal. SCR is measured in units called microsiemens (µS). Latency is the time between the onset of the stimulus and the beginning of the SCR. The latency period is typically one to three seconds in duration (Figner & Murphy, 2011). The time between the onset of the SCR and its peak amplitude is referred to as rise time and is typically one to three seconds in duration (Figner & Murphy, 2011). The difference between the onset (baseline) of the SCR and the peak is referred to as the amplitude and is one of the most common SCR measures (Figner & Murphy, 2011).

Slide 2 Using Usingthe thekeyboard, keyboard,select selectwhich whichoption optionyou youprefer. prefer.

$12 $0 -$8

A = certain

Experimental Design

-$8

B = 50/50 Slide 3

Thirty two individuals participated in this study. Two participants’ -$8 Slide 3 results were not used in the final analysis. One participant did not understand the experiment and the other did not finish. Ages Slide 4 -$8 ranged from 27 to 56 for the 15 male and 15 female participants. The mean age and standard deviation of the participants was 39 the last number sequence you were asked to remember in the order it was display Type and 5, respectively. Participants’ education levels ranged from Press the SPACE BARSlide to proceed to the next screen. acquiring less than a high school diploma to earning an advanced 4 degree. The average participant had a college degree. Participants were recruited from a southern U.S. state through the use of an Type the last number sequence you were asked to remember in online advertisement that can be found in the appendix. The the order it was displayed. advertisement was circulated on Craigslist.com. On the first screen participants were asked to memorize either a two digit (low cogn Press the SPACE BAR to proceed to the next screen. E-Prime 2.0 was used to program and run the experiment on digitfor(high cognitive load) number sequence in the order it was displayed, wh a desktop computer. E-Prime 2.0 is software seven that is used computerized experiment design, data collection, and analysis. Each participant was endowed with $30 priorreferred to the experiment to as forward digit span task. Forward digit span tasks have been widely u and asked to put the money in their pocket or purse. Evidence On the first screen participants were asked to memorize either a has shown that people place more value on an item that they two cognitive digit (low cognitive load) or seven&digit (high cognitive literature to manipulate resources (Jensen Figueroa, 1975; load) Gilbert, Gie physically possess compared to an equivalent item they do not number sequence in the order it was displayed, which is referred possess (Kahneman, Knetsch & Thaler, 1991). Participants were to as forward digit span task. Forward digit span tasks have been & Alfieri, 1997) theliterature average hascognitive been found to be able to r informed that their $30 could go up or down 1995; during Trope the experiwidely usedand in the to person manipulate resources ment and that they could potentially lose their entire endowment. (Jensen & Figueroa, 1975; Gilbert, Giesler & Morris, 1995; Figure 1 provides an example of the experimental design.of seven digits maximum accurately (Ericsson & Chase, 1982). Thefound digittostring inp Trope & Alfieri, 1997) and the average person has been be able to recall a maximum of seven digits accurately (Ericsson & Chase, digitand string occurred after all financial after all financial choices had1982). beenThe made allinput financial outcomes had been exper choices had been made and all financial outcomes had been experienced for the particular set of choices and outcomes made under particular set of choices and outcomes made under each digit set. Thus, the inputtin each digit set. Thus, the inputting of digits occurs outside of the ©2014, IARFC. All rights of reproduction in any form reserved.

occurs outside of the measured time span and, consequently, the differences betwee

Volume 13, Issue 2

measured time span and, consequently, the differences between typing or speaking would not be expected to affect the measured responses. On the second screen participants were asked to choose between a certain or uncertain amount of money. The uncertain choice was between two different monetary amounts. Each uncertain amount was assigned a 50% probability weight. After the selection was made the monetary outcome was displayed on the third screen for eight seconds. This was to allow adequate time to measure participantsâ&#x20AC;&#x2122; SCR to monetary outcomes. Recall accuracy has been found to be reduced when digits are held in memory over time (Baddeley & Hitch, 1974). Therefore, we would expect the eight second response delay to reduce digit recall accuracy. Participants were then asked to recall the number sequence on the forth screen. Participants wore a Q Sensor 2.0 wrist band, developed by Affectiva, which captured SCR response every 125 milliseconds. Substantial work has been completed on the validation of the sensor (Yeykelis, Cummings & Reeves, 2014) and various academic studies have used the sensor to measure SCR (Henriques, Paiva & Antunes, 2013; De Looff, Kuijpers & Nijman, 2014). SCRs were measured to make sure participants were responding emotionally to the monetary gains and losses. SCRs were also measured to determine whether a higher level of cognitive load alters an individualâ&#x20AC;&#x2122;s physiological response to gains and losses. Participants wore the sensor on their left wrist and were asked to type on the computer keyboard using their right hand. One participant did not register SCR so they were excluded from the physiological analysis. In order to measure loss aversion each participant answered 312 questions under gain-only, loss-only and mixed choices. As more questions are asked, it reduces uncertainty about the unknown parameter, loss aversion. 10 trials of digit span of two were conducted during each of which no more than 16 financial risk choices were made, totaling 156 choices. Similarly, 10 trials of digit span of seven were conducted during each of which no more than 16 financial risk choices were made, totaling 156 choices. Question and load order were the same for every participant. The sample size was 30 and the total number of data points was 9,360. Figure 1, Slide 2 is an example of a mixed choice question as there is a chance to gain or lose money. For gain-only questions there was zero chance of a loss and for loss-only questions there was zero chance of a positive monetary outcome. Monetary values ranged from -$28 to $28 for each question. Participants were not informed of their number sequence recall accuracy or the sum of their outcomes until after the experiment was completed. They were paid the initial $30 endowment, plus or minus half the sum of their outcomes, since every question was asked twice. Due to institutional review board guidelines participants were informed prior to the experiment that they could not owe the experimenter money. Participants were paid up to an additional $20 for digit recall accuracy in order to provide participants with an incentive for remembering the digit sequences. There was very little variation in the performance of participants

who provided accurate responses on the forward digit span task. The mean digit recall accuracy was 96.52 percent. The average recall for the two digit and seven digit loads were 98.14% and 94.9%, respectively. A $10 participation fee was paid when the experiment was completed.

Methodology Equation 1 displays the model used to derive the coefficient of loss aversion under high and low cognitive load for each participant. Whether the participant accepted or rejected the uncertain monetary choice (AR) is coded as one or zero, respectively, and is the dependent variable. In Figure 1, Slide 2, the uncertain monetary choice is option B. Analysis of variance (ANOVA) and analysis of covariance (ANCOVA) assume the error terms are normality distributed so these methods of analysis were not selected. A logistic regression model is appropriate to examine the relation between a binary dependent variable and a number of independent variables (Davidson & MacKinnon, 2004). The regression coefficients are estimated using maximum likelihood estimation. The continuous independent variables in Equation 1 include the gain of the uncertain monetary choice, the loss of the uncertain monetary choice, the certain outcome, the previous monetary outcome (PMO) and the time at which each question was answered. In order for a risk averse participant to select the uncertain monetary choice its expected value should be greater than the certain monetary choice (Arrow, 1971; Pratt, 1964). An example of the gain of the uncertain monetary choice would be the $12 in Figure 1, Slide 2. The hypothesized direction of effect would be positive as participants should be more likely to accept the uncertain monetary choice as the gain of the uncertain monetary choice increases. An example of the loss of the uncertain monetary choice would be the -$8 in Figure 1, Slide 2. As the loss of the uncertain monetary choice becomes less negative we would expect participants to be more likely to accept the gamble. The zero dollar choice in Figure 1, Slide 2 would be an example of a certain outcome. The hypothesized direction of effect for the certain monetary outcome would be negative. As the certain monetary outcome increases participants should be less likely to accept the uncertain monetary choice. House money and break-even effects1 were controlled for by including the PMO the participant observed in the prior question. For example, if the participant observed a value of $10 prior to observing the outcome of -$8 in Figure 1, Slide 3, then $10 would be included as the PMO. The PMO should be positively associated with choosing the uncertain monetary choice as prior gains have been found to increase risk-taking (Thaler and Johnson, 1990). A cumulative lag was not selected for the PMO because recent gains and losses are more salient than outcomes that occurred during a past period. Participants were not shown a cumulative total of their 1

Refer to Thaler and Johnson (1990) for an explanation of these effects.

Journal of Personal Finance

Q-Q plots of the AMP distributions before and after the squaremonetary outcomes until the experiment had concluded as this may have biased their risky choices. Monetary outcomes beyond root transformation. A log transformation was not selected due to the prior outcome may still have influenced subsequent risky the presence of zero values. Guillemette/James/Larsen paper their risky choices. choices Monetarywere outcomes beyond the choice. In order to limit this, biased the uncertain monetary priorsystematically outcome may still have influenced choice.value In order to limit theto determine whether participants were responding physIn this, order ordered so no moresubsequent than threerisky expected uncertain monetary choices were systematically ordered so no more than three expected value iologically to monetary gains and losses dummy variables were gain outcomes or three expected value loss outcomes appeared gain outcomes or three expected value loss outcomes appeared consecutively. The time at created forwhich absolute gains and losses. If a participant observed a consecutively. The time at which each question was answered each question was answered was included in the model to control for the order and interval in non-negative outcome the absolute gain variable (AG) was coded was included in the model to control for the order and interval in which each question was asked. as one and the absolute loss variable (AL) was coded as zero. which each question was asked. Prospect theory states that individuals place a greater emphasis Equation 1. Loss aversion Equation 1. Lossmodel aversion model on relative gains and losses compared to absolute gains and losses (Kahneman & Tversky, 1979). To account for relative gains and AR = gain*β1 + loss* β 2 + certainty* β 3 + PMO* β 4 + time* β 5 + e AR = gain*β1 + loss* β 2 + certainty* β 3 + PMO* β 4 + time* losses the certain outcome was subtracted from the observed outβ5+e come to create a relative gain/loss variable. If the relative gain/ loss was greater than zero the relative gain variable (RG) was TheThe coefficient ofofloss wasderived derivedbybytaking taking coefficient lossaversion aversion ( ) was thethe negative beta for the loss of coded as one, otherwise the relative loss variable (RL) was coded negative beta for the loss of the uncertain monetary choice and as one. For example, suppose the certain outcome was $5 and the the uncertain choice by the beta formonetary the gain of the uncertain monetary dividing monetary it by the beta forand thedividing gain ofitthe uncertain uncertain outcomes were $15 and -$5. If the participant chose the choice for each participant. Equation 2 shows how the coefficient uncertain outcome and observed $15 the RG would be $10, since choice for each participant. Equation 2 shows how the coefficient of loss aversion was derived of loss aversion was derived from Equation 1. We hypothesize the opportunity cost was $5. the beta for losses greater than the beta fromthat Equation 1. coefficient We hypothesize that thewill betabe coefficient for losses will be greater than the beta coefficient for gains, on average, based on prior experimental Variables were created to account for both absolute and relative research & Kahneman, 1992; & Traub, 2002). coefficient for(Tversky gains, on average, based on priorSchmidt experimental research (Tverskygains & Kahneman, and losses. If a participant was exposed to an absolute loss If the beta coefficient for losses is greater than the beta coefficient and a relative loss the absolute loss-relative loss variable (ALRL) 1992; Traub,that 2002). If the beta is coefficient for losses is greater than the beta forSchmidt gains it&means a participant loss averse. A Shapiro-Wilk was coded as one. If a participant experienced an absolute test was run on the loss aversion coefficients and the null hypoth- gain and a relative gain the absolute gain-relative gain variable coefficient for gains it means that a participant is loss averse. A Shapiro-Wilk test was run on esis of a normal distribution was rejected at an alpha level of (AGRG) was coded as one. If an absolute gain and a relative loss 0.05. Therefore, we use a Wilcoxon signed-rank test to determine were observed the loss aversion coefficients and the null hypothesis of a normal distribution was rejected at an the absolute gain-relative loss variable (AGRL) whether the two-digit load control group is statistically different was coded as one. No outcome was both an absolute loss and from loadwe group. Wilcoxon signed-rank is a awhether alpha levelthe of seven-digit 0.05. Therefore, use aAWilcoxon signed-rank test test to determine relativethe gain. Choosing the certain outcome was used as the non-parametric test that is used to compare non-normally distribreference group. Based on prior experimental findings (Pennings two-digit load control group is different the seven-digit load group. A uted matched samples tostatistically assess whether theirfrom average ranks differ & Smidts, 2003; Booij, Van Praag & Van De Kuilen, 2010) (Siegel, 1956). we hypothesize that the SCR to ALRL will be greater than the Wilcoxon signed-rank test is a non-parametric test that is used to compare non-normally response to AGRG. If the participant experiences an AGRL we Equation 2. Derivation of the coefficient of loss aversion that they will be conflicted and not consistently distributed matched samples to assess whether their average ranks differ (Siegel,hypothesize 1956). register a SCR reading.

λ = -β loss / β gain

Equation 3 displays the modelof used to analyze Equation 2. Derivation of the coefficient loss aversion

participant’s physiological arousal to absolute and relative gains and losses. This model is important because it tells us whether participants were responding emotionally to the gains and losses and also whether they responded to the losses more than the gains. The SCR amplitude (AMP) variable is the dependent variable and is created by taking the SCR at the onset of the stimulus and subtracting it from the maximum SCR value up to 6000 milliseconds later. In Figure 1, AMP is measured from the time at which Slide 3 (-$8) appears on the computer screen until six seconds later. All AMP values are non-negative numbers. The distribution of the AMP variable contains a high number of zero values as 28.82% of questions resulted in a µS reading of zero. When the dependent variable contains a large number of zero values the use of an ordinary least squares model is not appropriate as regression coefficients will be biased (Maddala, 1987). A Tobit model is used as this type of model produces data similar to the AMP variable data. The AMP variable was square-root transformed to reduce skewness so that the model produced data similar to the data that is produced by a Tobit model. The appendix contains the

Equation 3. Model to determine physiological response to absolute and relative gains and losses AMP = ALRL* β 1 + AGRG* β 2 + AGRL* β 3 + e In order to determine the effect cognitive load has on absolute and relative gains and losses a Tobit model was created that includes interaction effects. The model is displayed in Equation 4. The high cognitive load variable (HCL) is interacted with ALRL, AGRG and AGRL. Due to the variable inclusion principle lower-order terms, ALRL, AGRG, AGRL and HCL, should be included in the model. When lower-order terms are added to the regression model with higher-order multiplicative terms, the standard errors of the higher-order terms are unaffected (Aiken & West, 1991). We hypothesize that when absolute and relative gains and losses are interacted with HCL they will be negatively associated with the AMP variable. This is because the distraction of higher cognitive load should lessen participants’ emotional responses to monetary gains and losses.

Volume 13, Issue 2

Equation 4: Physiological response to absolute and relative gains and losses under high cognitive load

Table 1. Coefficients of loss aversion estimates under high and low cognitive load

AMP = ALRL* β 1 + AGRG* β 2 + AGRL* β 3 + HCL* β 4 + ALRL*HCL* β 5 + AGRG*HCL* β 6 + AGRL*HCL* β 7+ e

Results The mean and standard deviation for the coefficients of loss aversion under HCL were -1.3283 and 0.9436, respectively. The mean and standard deviation for the coefficients of loss aversion under low cognitive load was -1.2084 and 0.7171, respectively. The coefficients of loss aversion under high and low cognitive load are reported for each participant in Table 1.

Participant 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

High cognitive load -0.7414 -1.3678 -3.0347 -1.1328 -0.7339 -5.0785 -2.2220 -0.9940 -0.5090 -1.3120 -1.2952 -0.9247 -1.0867 -2.1087 -2.4519 -1.3729 -1.5397 -0.3226 -0.8768 -0.7017 -0.4315 -0.8253 -0.6477 -0.3361 -1.6127 -1.2475 -1.2678 -0.9933 -1.5221 -1.1565

Low cognitive load -0.7857 -1.7557 -2.4550 -1.0392 -0.6430 -3.5409 -1.9824 -1.3472 -0.4584 -1.2389 -2.2222 -0.9576 -0.9300 -1.8603 -0.6005 -1.0441 -1.3567 -0.2317 -0.9540 -0.9698 -0.3046 -0.9054 -0.6428 -0.1106 -1.5000 -1.3439 -1.1008 -1.0567 -1.7084 -1.2050

The results for the Wilcoxon signed-rank test comparing the mean rank of the coefficients of loss aversion under low and high cognitive load are displayed in Table 2. The Z-value was -1.3678, which is not statistically significant.

Z-Value -1.3678

Table 2. Wilcoxon Table Wilcoxon signed-rank signed-rank test test comparing and low cognitive load Z-Value -1.3678 p-Value 0.1707

under high high and low co under

Journal of Personal Finance

Physiological response to gains and losses are reported in Table 3. The ALRL and AGRG variables are both statistically significant which means that the participants responded emotionally to the gains and losses. The coefficient for the ALRL variable is approximately two times greater than the coefficient for the AGRG variable. This means that physiologically, SCR AMP was twice as strong for losses compared to gains. This is consistent with the finding that the dissatisfaction experienced from losses is approximately two times greater than the satisfaction derived from gains (Tversky & Kahneman, 1992). The AGRL variable is not statistically significant. The Akaike information criterion (AIC) for the Tobit model displayed in Table 3 is 1991. The AIC provides a measure of the relative quality of the model. The AIC incorporates the model’s goodness of fit as well as a penalty when the number of independent variables increases.

Table 3. Physiological response to absolute and relative gains and losses Variable Intercept ALRL AGRG AGRL

Estimate (µS) 0.0928*** 0.0200*** 0.0104** 0.0002

Standard Error 0.0031 0.0055 0.0052 0.0129

t-Value 29.85 3.63 2.00 0.01

*p<.10; **p<.05; ***p<.01 Table 4 displays the Tobit model results for the interactions between the HCL variable and the ALRL, AGRG and AGRL variables. Again, the dependent variable is SCR AMP. There is statistically significant evidence that physiological response to absolute and relative losses is reduced when individuals are placed under HCL. The AIC for the model displayed in Table 4 is 1994.

Table 4. Physiological response to absolute and relative gains and losses under high cognitive load Variable Intercept ALRL AGRG AGRL HCL ALRL*HCL AGRG*HCL AGRL*HCL *p<.10; **p<.05; ***p<.01

Estimate (µS) 0.0936*** 0.0297*** 0.0106 0.0023 -0.0017 -0.0189* -0.0003 -0.0041

Standard Error 0.0044 0.0078 0.0073 0.0183 0.0061 0.0110 0.0104 0.0257

t-Value 21.49 3.79 1.44 0.12 -0.27 -1.71 -0.03 -0.16

Volume 13, Issue 2

Conclusions and Limitations We find no statistically significant evidence that loss aversion, as measured by choice, is altered when individuals are placed under HCL. The findings provide statistically significant evidence that HCL temporarily reduces physiological responses to absolute and relative small dollar losses during the period that HCL is experienced. The long-term duration of the affective dampening created by temporary cognitive load is unknown. However, at least in the short term, the results suggest that presenting loss results in the context of other cognitively complex information should work to dampen the normal affective response to losses. For financial planners seeking to temporarily dampen the emotional response to losses, presenting such information within the context of other cognitively complex factors may be, at least momentarily, effective. Limitations of this study include the use of small monetary amounts, a small sample size, and the use of house money. The use of small amounts of money is a limitation as participants risk preferences might have been different had larger amounts been used. A larger sample size would reduce uncertainty about the unknown parameters in the models. The use of house money, as opposed to using the participants’ money is another limitation as a mental accounting effect has been found in the prior literature (Thaler & Johnson, 1990). The house money effect would predict that loss aversion would be lower than average in this study since participants were informed prior to the experiment that they could not owe the experimenter money.

Ideas for Future Research This paper focused on the effect cognitive load has on loss aversion. However, there are a variety of other factors which may alter loss aversion within the domain of personal finance. The Certified Financial Planner™ designation (CFP®) is a signal of specific human capital in the area of investments and it imposes a fiduciary responsibility on advisors who provide financial planning services to clients. Brain scans have indicated that clients who know their advisors have the CFP® designation are less likely to second guess their advisors’ recommendations in an experimental stock market game (James, 2013). It would be valuable to go one step further and investigate whether investment advice from a CFP® (or Chartered Financial Analyst) actually reduces loss aversion. Most financial planners do not distinguish between labor income and retirement income when assessing client risk tolerance. However, the marginal propensity to consume has been found to alter risk preferences (Kimball, 1990) and is directly related to mental accounting. It would be important for financial planners to understand whether mental accounting alters loss aversion. If loss aversion is different when clients are asked about pre-retirement versus post-retirement assets, financial planners would need to account for this when assessing client risk preferences.

Finally, the tendency for individuals to make investment decisions on the basis of information provided by trades of other market participants (also known as “herding”) has been found to influence investor behavior in both the U.S. and international stock markets (Sias, 2004; Wylie, 2005; Tan, Chiang, Mason & Nelling, 2008). Wermers (1999) found that stocks that herds purchased outperformed stocks that herds sold during the following six months from 1975-1994. It may be beneficial to study what effect, if any, herding has on loss aversion.

References Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Sage. Arrow, K. J. (1971). Essays in the theory of risk-bearing (Vol. 1). Chicago: Markham Publishing Company. Ashraf, N., Camerer, C. F., & Loewenstein, G. (2005). Adam Smith, behavioral economist. Journal of Economic Perspectives, 131-145. Baddeley, A. (2000). The episodic buffer: a new component of working memory? Trends in Cognitive Sciences, 4(11), 417-423. Baddeley, A. D., & Hitch, G. (1974). Working memory. Psychology of learning and motivation, 8, 47-89. Barberis, N., Huang, M., & Santos, T. (2001). Prospect theory and asset prices. Quarterly Journal of Economics, 116(1), 1-53. Bechara, A., Damasio, H., Damasio, A. R., & Lee, G. P. (1999). Different contributions of the human amygdala and ventromedial prefrontal cortex to decision-making. Journal of Neuroscience, 19(13), 5473-5481. Benartzi, S. & Thaler, R. (1995). Myopic Loss Aversion and the Equity Premium Puzzle. Quarterly Journal of Economics. 110(1), 73-92. DOI: 10.2307/2118511. Booij, A. S., Van Praag, B. M., & Van de Kuilen, G. (2010). A parametric analysis of prospect theory’s functionals for the general population. Theory and Decision, 68(1-2), 115-148. Breiter, H. C., Aharon, I., Kahneman, D., Dale, A., & Shizgal, P. (2001). Functional imaging of neural responses to expectancy and experience of monetary gains and losses. Neuron, 30(2), 619-639. Davidson, R., & MacKinnon, J. G. (2004). Econometric theory and methods (p. 623). New York: Oxford University Press. De Looff, P. C., Kuijpers, E., & Nijman, H. L. I. (2014). Stress levels of psychiatric nursing staff. Journal of Nursing Education and Practice, 4(7), 1. Epstein, S. (1994). Integration of the cognitive and the psychodynamic unconscious. American Psychologist, 49(8), 709. DOI: 10.1037/0003066X.49.8.709. Ericsson, K. A., & Chase, W. G. (1982). Exceptional Memory: Extraordinary feats of memory can be matched or surpassed by people with average memories that have been improved by training. American Scientist, 70(6), 607-615. Figner, B., & Murphy, R. O. (2011). Using skin conductance in judgment and decision making research. A handbook of process tracing methods for decision research, 163-184. Finke, M. S., & Huston, S. J. (2004). Risk and myopic financial decisions. Journal of Personal Finance, 3(3), 99-112.

Journal of Personal Finance

Friesen, G. C., & Sapp, T. R. (2007). Mutual fund flows and investor returns: An empirical examination of fund investor timing ability. Journal of Banking & Finance, 31(9), 2796-2816. DOI: 10.1016/j.jbankfin.2007.01.024. Fudenberg, D., & Levine, D. K. (2006). A dual-self model of impulse control. The American Economic Review, 1449-1476. Gilbert, D. T., Giesler, R. B., & Morris, K. A. (1995). When comparisons arise. Journal of Personality and Social Psychology, 69(2), 227. DOI: 10.1037/0022-3514.69.2.227. Guillemette, M. A. & Finke, M. S. (2014). Do Large Swings in Equity Values Change Risk Tolerance. Journal of Financial Planning, 27(6), 44-50. Hanna, S., Fan, J. X., & Chang, Y. R. (1995). Optimal life cycle savings. Financial Counseling and Planning, 6, 1-15. Henriques, R., Paiva, A., & Antunes, C. (2013). On the need of new methods to mine electrodermal activity in emotion-centered studies. In Agents and Data Mining Interaction (pp. 203-215). Springer Berlin Heidelberg. James III, R. N. (2013). Brain Activity Suggests Planning Designation Helps Calm Investors. Journal of Financial Planning, 26(2), 52-59. Jensen, A. R., & Figueroa, R. A. (1975). Forward and backward digit span interaction with race and IQ: Predictions from Jensen’s theory. Journal of Educational Psychology, 67(6), 882. Kahneman, D. (2011). Thinking, fast and slow. New York, NY: Farrar, Straus and Giroux. Kahneman, D., Knetsch, J. L., & Thaler, R. H. (1991). Anomalies: The endowment effect, loss aversion, and status quo bias. Journal of Economic Perspectives, 5(1), 193-206.

Tan, L., Chiang, T. C., Mason, J. R., & Nelling, E. (2008). Herding behavior in Chinese stock markets: An examination of A and B shares. PacificBasin Finance Journal, 16(1), 61-77. Thaler, R. H., & Johnson, E. J. (1990). Gambling with the house money and trying to break even: The effects of prior outcomes on risky choice. Management Science, 36(6), 643-660. Thaler, R. H., & Shefrin, H. M. (1981). An economic theory of self-control. The Journal of Political Economy, 392-406. Thaler, R., Tversky, A., Kahneman, D. & Schwartz, A. (1997). The effect of myopia and loss aversion on risk-taking: an experimental test. Quarterly Journal of Economics, 112, 647−661. DOI: 10.1162/003355397555226. Trope, Y., & Alfieri, T. (1997). Effortfulness and flexibility of dispositional judgment processes. Journal of Personality and Social Psychology, 73(4), 662. DOI: 10.1037/0022-3514.73.4.662. Tversky, A., & Kahneman, D. (1992). Advances in prospect theory: Cumulative representation of uncertainty. Journal of Risk and Uncertainty, 5(4), 297-323. DOI: 10.1007/BF00122574. Wermers, R. (1999). Mutual fund herding and the impact on stock prices. The Journal of Finance, 54(2), 581-622. Wylie, S. (2005). Fund manager herding: A test of the accuracy of empirical results using UK data*. The Journal of Business, 78(1), 381-403. Yeykelis, L., Cummings, J. J., & Reeves, B. (2014). Multitasking on a Single Device: Arousal and the Frequency, Anticipation, and Prediction of Switching Between Media Content on a Computer. Journal of Communication, 64(1), 167-192.

Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica: Journal of the Econometric Society, 263-291. Kimball, M. S. (1990). Precautionary saving and the marginal propensity to consume (No. w3403). National Bureau of Economic Research. Loewenstein, G., & O’Donoghue, T. (2004). Animal spirits: Affective and deliberative processes in economic behavior. Available at SSRN 539843. Maddala, G. S. (1987). Limited dependent variable models using panel data. Journal of Human Resources, 307-338. Martino, B., Camerer, C. F., & Adolphs, R. (2010). Amygdala damage eliminates monetary loss aversion. Proceedings of the National Academy of Sciences, 107(8), 3788-3792. DOI: 10.1073/pnas.0910230107. Pennings, J. M., & Smidts, A. (2003). The shape of utility functions and organizational behavior. Management Science, 49(9), 1251-1263. Pratt, J. W. (1964). Risk aversion in the small and in the large. Econometrica, 122-136. Schmidt, U., & Traub, S. (2002). An experimental test of loss aversion. Journal of Risk and Uncertainty, 25(3), 233-249. Sias, R. W. (2004). Institutional herding. Review of financial studies, 17(1), 165-206. Siegel, S. (1956). Nonparametric statistics for the behavioral sciences. New York: McGraw-Hill Book Company. Sokol-Hessner, P., Camerer, C. F., & Phelps, E. A. (2012). Emotion regulation reduces loss aversion and decreases amygdala responses to losses. Social Cognitive and Affective Neuroscience. DOI: 10.1093/scan/ nss002. Sokol-Hessner, P., Hsu, M., Curley, N. G., Delgado, M. R., Camerer, C. F., & Phelps, E. A. (2009). Thinking like a trader selectively reduces individuals’ loss aversion. Proceedings of the National Academy of Sciences, 106(13), 5035-5040. DOI: 10.1073/pnas.0806761106.

Volume 13, Issue 2

AVERSION UNDER COGNITIVE LOAD 24

Appendix

Appendix Q-Q plot of the AMP distribution before square root transformation

Q-Q plot of the AMP distribution after square root transformation

Recruitment Advertisement Date The Department of Personal Financial Planning is conducting a risk assessment study. The experiment is expected to last approximately 1 hour and you could earn between $10-$200. If you are interested in participating or have any questions regarding this study, please e-mail XXX at XXX. Please put “Risk Assessment Experiment” in the subject line if you are interested in participating. Individuals studying personal financial planning are not eligible to participate. All information will be kept confidential. Thank you for your consideration in helping us with this study.

Sincerely, XXX Principal Investigator

IARFC National Financial Plan Competition April 29-30, 2015 — Charlotte, NC

focusing on the future Sponsorship The IARFC has extended its invitation to students to participate in the 2015 National Financial Plan Competition. The finalists and their faculty advisor will present their comprehensive financial plans to a live audience. We invite the IARFC members to join the competition and become part of the judging process. Expenses incurred by participating in the judging of the Financial Plan Competition are not the responsibility of the IARFC.

Professionals who love what they do for a living feel fortunate and look for a way to pay it forward – especially to the next generation. For that reason, RFC Ed Skelly of Sterling Financial Partners in Ashburn, Virginia traveled to Las Vegas and judged the the 2014 National Plan Competition. As a Diamond Sponsor, Ed is encouraged by what the students are learning in school. “Today the presentations by the three teams were absolutely phenomenal,” commented Skelly. “I think the nine students who presented did a wonderful job.”

2014 Winning Team — Bryant Universitty Smithfield RI, team members Kyle Creedon, Lauren Fayne, Jamie Pepin and their instructor Mara Derderian (far L and R, Les Anderson and Ed Morrow)

Become a 2015 Sponsor Sponsorship levels

Silver

Gold

Platinum

Diamond

Individual

$50

$100

$250

$500

Corporate

$1,000

$5,000

$7,500

$10,000

Unable to sponsor but interested in mentoring. Send more information on Corporate Sponsorship. Interested in judging the Financial Plan Competition. Please print or type information below. Mr.

Ms.

Mrs.

_________________________________________________________________ First Name Last Name _________________________________________________________________ Firm or B/D _________________________________________________________________ Street Address _________________________________________________________________ City State Zip Code ________________________________________________________________ Phone _________________________________________________________________ E-mail

Payment Information: Credit Card

Check made payable to IARFC

___________________________________________ _____________ Credit Card # (Discover, MasterCard, Visa, AmEx) Exp. date ________________________________________________________ Signature

PLEASE MAIL OR FAX THIS COMPLETED FORM TO IARFC

IARFC INTERNATIONAL ASSOCIATION OF REGISTERED FINANCIAL CONSULTANTS

International Association of Registered Financial Consultants P.O. Box 42506 Middletown, OH 45042-0506

Fax: 513.424.5752 Phone: 800.532.9060 Email: info@iarfc.org

focusing on the future

IARFC INTERNATIONAL ASSOCIATION OF REGISTERED FINANCIAL CONSULTANTS

National Financial Plan Competition April 29-30, 2015 â&#x20AC;&#x201D; Charlotte, NC

Bringing public and industry recognition to IARFC members and incoming young professionals.

Give Back

Your Practice

Participation as a sponsor for the National Plan Competition is mutually beneficial and allows for various levels of interaction with the students and members.

Attend a Business Owner Consulting Workshop. Business owners are an underserved market that are looking for solutions to many issues. Address important items such as loan cancellation strategies, buy-outs, succession, retention and incentive plans plus much more.

Getting Involved

Raising Awareness

Finalists and their faculty advisor will present their comprehensive financial plans to an audience. We invite you to the competition and become part of the judging process.

National Financial Plan Competition is a way to get hands-on development of young people by teaching them the skills necessary to provide the services we currently give to the next generations.

learn more

Top: Nick and Jerry Royer co-host their weekly radio show every Saturday, go to www.AskJerryandNick.com for showtimes; Bottom: Nick answers a listenerâ&#x20AC;&#x2122;s question.

www.iarfc.org/FinancialPlanCompetition

The Register | September-October 2014

Page 83

Journal of Personal Finance International Association of Registered Financial Consultants - IARFC 2507 North Verity Parkway P.O. Box 42506 Middletown, Ohio 45042

www.journalofpersonalfinance.com

Sponsors National Financial Plan Competition 2015 Individual Sponsors

Silver Anonymous Donation Patrick Astre George Atkinson Andrew Bearre Bill Brice George Brown Judy Clark James Davidson Elisabeth Dawson George Flack James Flanagan Jessica Harris Chris Hill Gerard Iacangelo Rene Isuk Norman Johnson Darren Josephson

David King Charles McKinney Timothy Morgan Chris Nelson Yukiko Ogino-Christman Frederick Ostermeyer Gary Padussis Bryan Philpott Barbara Pietrangelo John Repass Bob Shlesinger Stuart Spivak Charles Reed Terry Mary Jo Walker Brenda Whitman Kirk Barr Young Salvatore Zimbardi

Gold

Platinum

Diamond

Jan Belyeu Ahmed Edris Frank Eberhart Patrick Lyman Joseph Lolli

Michell Blair David Howard

Jon Rogers

learn more: www.IARFC.org/FinancialPlanCompetition

Be listed here as these 2014 Corporate Sponsors are!