Journal of Personal Finance Volume 13 Issue 1

Volume 13 Issue 1 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 IARFC

National Financial Plan Competition Featuring

William L. Moore CLU, ChFC, FIC, RFC® Sagemen

Jon M. Rogers Ph.D., ChFC, CLU, RFC® Rogers Financial Group

NEW YORK, NEW YORK • LAS VEGAS HOTEL & CASINO • MAY 1, 2014

Peter J. D’Arruda CTC, RFC® Financial Safari

The Educational Process Begins with You! J oin the 2014 Financial Plan Competition.

• Receive recognition as a National Judge of the Competition Finals.

• Learn from industry expert, (guest speaker to be announced.)

• Attend our Pre-Event, April 30 – Impact Branding workshop Discover and build your own personal brand.

Les Anderson MBA, RFC® Cetera Advisors

• Network during the welcome reception. Mentor future financial planning professionals.

• RFC Continuing Education Credits available – Impact Branding (3), Competition Process (3), Feature Presentation (1).

Sponsored by

Ed Morrow CLU, ChFC, RFC® IARFC Chairman & CEO

Visit www.IARFC.org/ FinancialPlanCompetition

Volume 13, Issue 1

1

Journal of Personal Finance

Volume 13, Issue 1 2014

The Official Journal of the International Association of Registered Financial Consultants

2

Journal of Personal Finance

CONTENTS Editor’s Notes…………………………………………………………………………………………………………………..7 How Prior Outcomes Affect Individual Investors’ Subsequent Risk Taking……………………………………………...8 Keith Jacks Gamble, Assistant Professor of Finance, DePaul University Bjorn Johnson, Assistant Professor of Finance, DePaul University We present empirical evidence of how prior outcomes affect individual investors’ subsequent risk taking. Investors who experience big gains or big losses are likely to exit the stock market; however, investors remaining in the market increase their portfolio risk taking following losses. They replace stocks sold with new positions of higher (lower) value following recent losses (gains), thereby leading to increased (decreased) stock portfolio risk taking. Our results are consistent with predictions of consumption habit formation interacting with the disposition effect. Our results cannot be explained by information, simple rebalancing, the house money effect, or the break even effect.

Does Visually Displaying Probability Outcomes Change Stock Selection?………………………………………………38 Tim S. Griesdorn, Ph.D., CFP®, AFC®, Human Development and Family Studies, Iowa State University Hyrum L. Smith, Ph.D., CFP®, CPA®, Personal Financial Planning, Woodbury School of Business, Utah Valley University Prospect theory assumes individuals are more risk averse with gains than to equivalent losses and indicates how an investment question is framed influences individual decision making. However, limited research has been conducted on the influence of visually displaying the probability of different outcomes, versus in written form, on investment decision making. Hypothetical investment scenarios tested the effect of visually displaying probability information.

Results suggest that the visual display of probability shifts investment preferences to the stock with the greatest probability of a gain. Thus, it appears investors are willing to take more financial risks when probability information is displayed visually. Financial planners could incorporate visual displays when communicating with clients to help match the clients’ risk capacity with risk tolerance.

Consideration of Retirement Income Stages in Planning for Retirement……………………………………………..…52 Kyoung Tae Kim, Ph.D. Candidate, The Ohio State University Sherman D. Hanna, Ph.D., The Ohio State University Samuel Cheng-Chung Chen, Ph.D., Suntrust Mortgage Previous retirement adequacy studies have ignored expected retirement income stages. Ignoring retirement income stages results in biased estimations of retirement adequacy. This study analyzes retirement income stage theoretically and then empirically. Based on the 1995 to 2007 Survey of Consumer Finances (SCF) datasets, about 73% of working households with the head and/or spouse/partner age 35-70 and working full-time will have more than one retirement income stage. When income stages are taken into account, the proportion of households with retirement adequacy ranges from 44% in 1995 to 58% in 2007. Ignoring retirement income stages results in adequacy proportions being 23 to 28 percentage points higher. Financial planners and researchers evaluating retirement adequacy should take retirement income stages into account.

©2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1

Post-Recession, Post-Legislation Credit Use: Insights from an Online Survey……………………………….………….65 Barbara O’Neill, Ph.D., Extension Specialist in Financial Resource Management, Rutgers Cooperative Extension Jing Jian Xiao, Ph.D., Professor, Department of Human Development and Family Studies, University of Rhode Island This article describes findings from data collected in 2009-2013 from a sample of 1,081 U.S. respondents to a 20question online self-assessment tool called the Wise Credit Management Quiz. The quiz incorporates frequently cited expert recommendations for credit use. Key provisions of recent federal credit legislation are also described. This study was conducted to investigate post-recession, post-legislation credit management practices of Americans in an effort to inform financial education and counseling efforts. The average quiz score for checking credit scores was higher than that for checking credit reports even though free credit scores are not required by law and free credit reports are. The least frequently performed practices were comparing at least three lenders before applying for credit and taking advantage of “float” time between when purchases are made and when credit card payments are due. The highest credit management quiz scores were associated with respondents who were older and those who had higher incomes and educational levels.

3

4

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 manuscripts in topics related to household financial decision making. The journal is committed to providing high quality article reviews in a single-reviewer format within 45 days of submission. JFP encourages submission of manuscripts that advance the emerging literature in personal finance on topics that 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

-

EDITORIAL BOARD The journal is also seeking editorial board members. Please send a current CV and sample review to the editor. JPF is committed to providing timely, high quality reviews in a single reviewer format.

CONTACT Michael Finke, Editor Email: jpfeditor@gmail.com www.JournalofPersonalFinance.com

Š2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1

5

JOURNAL OF PERSONAL FINANCE VOLUME 13, ISSUE 1 2014 EDITOR Michael S. Finke, Texas Tech University

ASSOCIATE EDITOR Wade Pfau, National Graduate Institute for Policy Studies (GRIPS)

EDITORIAL ASSISTANT Carey Yeary, Texas Tech University

EDITORIAL BOARD Steve Bailey, HB Financial Resources Joyce Cantrell, Kansas State University Dale Domian, York University Monroe Friedman, Eastern Michigan University Joseph Goetz, University of Georgia Clinton Gudmunson, Iowa State University Sherman Hanna, The Ohio State University George Haynes, Montana State University Douglas Hershey, Oklahoma State University Karen Eilers Lahey, University of Akron Doug Lambin, University of Maryland, Baltimore County Rich Landsberg, Advanced Consulting Group Jean Lown, Utah State University Angela Lyons, University of Illinois Ruth Lytton, Virginia Tech University Lewis Mandell, University of Washington and Aspen Institute Yoko Mimura, University of Georgia Robert Moreschi, Virginia Military Institute Edwin P. Morrow, Financial Planning Consultants David Nanigian, The American College Barbara O’Neill, Rutgers Cooperative Extension Cliff Rob, University of Alabama Jing Xiao, University of Rhode Island Rui Yao, University of Missouri Tansel Yilmazer, University of Missouri Yoonkyung Yuh, Ewha Womans University

Mailing Address: IARFC Journal of Personal Finance The Financial Planning Building 2507 North Verity Parkway Middletown, OH 45042-0506

© Copyright 2014. International Association of Registered Financial Consultants. (ISSN 1540-6717)

6

Journal of Personal Finance

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 Editor.

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

Š2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1

7

EDITOR’S NOTES

This issue begins with two novel studies that investigate behavioral determinants of investment decisions.

The first tests some common theories of suboptimal investor behavior and finds that some do a better job of explaining investor behavior than others. Investors who do very well or very badly tend to leave the market. Those who stay in the market take more risks if they’ve done poorly and fewer risks if they’ve done well.

One theory of investor behavior is the so-called house money effect. Once you’ve won at the slots, you may be more likely to blow your winnings on the craps table. Easy come, easy go. Are stock investors also prone to take greater risks after a big win? As it turns out, they don’t. Winners are more likely to back away from the table or take smaller bets. The first paper by Keith Jacks Gamble and Bjorn Johnson is among the first to providing convincing evidence that the house money effect doesn’t explain trading behavior among stock investors.

Many of the behavioral theories we use to explain investor preferences come from experimental studies that ask subjects to compare two different risky choices. The second paper by Tim Griesdorn is the first to show that how we present risk-related information can have a significant impact on which choice a subject prefers. Visual representation of risk can give people a much more accurate idea of the tradeoffs that exist for each outcome.

Griesdorn finds that subjects are more likely to take risks when probabilities are displayed visually than when they are shown numerically. This is an important finding when we consider how to present information to less sophisticated investors in a way that helps them make decisions that best reflect preferences for risk – especially if they may have limited quantitative skills.

I should also note that this will be my last issue as editor of the Journal of Personal Finance. In 2014, my current assistant editor Wade Pfau and Joe Tomlinson will begin serving as the new editors. I am extremely indebted to the IARFC for their commitment to providing members with a high quality academic journal, and to the staff and publisher for their assistance over the last three years. I am also grateful to new editors for agreeing to chart the Journal’s course in the future, and look forward to watching the Journal and the field of personal finance grow.

~Michael Finke

8

Journal of Personal Finance

HOW PRIOR OUTCOMES AFFECT INDIVIDUAL INVESTORS’ SUBSEQUENT RISK TAKING

Keith Jacks Gamble, Assistant Professor of Finance Kellstadt Graduate School of Business, DePaul University Bjorn Johnson, Assistant Professor of Finance Kellstadt Graduate School of Business, DePaul University

We present empirical evidence of how prior outcomes affect individual investors’ subsequent risk taking. Investors who experience big gains or big losses are likely to exit the stock market; however, investors remaining in the market increase their portfolio risk taking following losses. They replace stocks sold with new positions of higher (lower) value following recent losses (gains), thereby leading to increased (decreased) stock portfolio risk taking. Our results are consistent with predictions of consumption habit formation interacting with the disposition effect. Our results cannot be explained by information, simple rebalancing, the house money effect, or the break even effect.

Introduction There is an extensive experimental literature documenting how prior outcomes systematically affect subsequent risk taking among student subjects with small stakes in the laboratory. Thaler and Johnson (1990) find increased risk taking following gains, which they dub the "house money effect," and increased risk taking following losses when there is a chance to break even. Post et al.

(2008) document similar behavior using high-stakes field data provided by contestants' risk taking behavior on the internationally popular television game show "Deal or No Deal.'' Unlike the risky choices in a laboratory, the risky propositions present in a financial market are among the most common and important high stakes financial decisions people face. An extensive empirical literature in finance supports the relevance of these behavioral effects of prior outcomes on subsequent risk taking even among financial professionals.1 A wide variety of behavioral biases plague 1

See Coval and Shumway (2005) for bond futures traders at the

Chicago Board of Trade; Liu et al. (2010) for Taiwanese market makers in TAIEX index options; Frino et al. (2008) for professional

individual investors (Barber and Odean (2011)), but in this study we document a notable exception to this extensive literature. Individual investors do not exhibit these behavioral effects of prior outcomes on their subsequent risk taking behavior. Instead, our results are consistent with an interplay between two effects, one operating on the portfolio level (such as the consumption habit formation of Constantinides (1990) and Campbell and Cochrane (1999)) and one operating on the stock level (such as the disposition effect of Shefrin and Statman (1985) and Odean (1998)).

Using a large database of account information on individual investors from a major discount brokerage firm in the U.S. combined with stock return data, we measure investors’ portfolio returns and their subsequent risk taking.

For the full sample, investors who experience extreme returns, either large losses or large gains, in the first six traders on the Sydney Future Exchange; O’Connell and Teo (2009) for institutional currency traders; Savor and Gamboa-­‐Cavazos (2011) for short sellers; and Beshears and Milkman (2011) for stock analysts.

©2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1 months of the year decrease the risk of their stock portfolio over the final six months of the year. The larger the magnitude of the initial return, the more investors lower their subsequent risk taking. This finding is inconsistent with the house money effect and the break even effect driving individual investors' behavior.

We examine the underlying causes of the decrease in risk taking following large gains and losses, and find it to be mainly related to net selling of stocks, which is driven by the investors who are exiting the stock market. Extreme past returns induce an increase in total portfolio turnover, of both sales and purchases of stocks. We find that sales dominate among investors who experience large gains or large losses.

However, for the subset of households who stay invested in the stock market, we instead find that they increase their risk taking slightly following losses, while lowering their risk somewhat after gains. Thus, even among this subsample, the house money effect does not drive their subsequent risk taking behavior. Only among this subsample do we find any evidence that is consistent with the break even effect impacting individual investor behavior.

We further investigate which stocks investors trade in response to past returns. In the case of large prior portfolio gains, the investors tend to sell stocks that have done well in the past, consistent with the well-established disposition effect; however, we show that they do not fully replace those holdings with other stock positions. Thus, the disposition effect has a large impact on portfolio risk. Overall, they end up with a net sale in dollar terms. In the case of large prior losses at the portfolio level, the investors again tend to act in accordance with the disposition effect: while investors with large prior losses are inclined to sell some stocks, they are reluctant to sell stocks that have done poorly. Instead, they sell the stocks that have done relatively well. However, they do not buy new stocks, nor increase their holdings of existing stock positions by the same amount. Thus, again they end up with a net sale in dollar terms. In sum, the end result of these behaviors is that the overall risk taking in the portfolio is reduced after both large gains and losses.

Our results cannot be explained by information or overconfidence. First, we may rule out trading based on any useful information, simply because the most frequent traders in the sample perform the worst (Odean (1999), Barber and Odean (2000)), and the stocks sold by investors on average outperform the stocks that they purchase (Odean (1998)).

Another hypothesis is that investors trade as a result of overconfidence in their abilities or in the information they possess. This hypothesis has been tested by Statman et al.

(2006), who find trading to be positively related to past

9 returns. Related to this is the finding by Choi et al. (2009) that 401(k) investors chase returns in their retirement portfolios; investors who have personally experienced gains increase their contribution rates more than their coworkers who have experienced losses, which the authors attribute to naive learning on the part of the investors. However, none of these hypotheses are consistent with our results since we find that both prior losses and gains make the investor cut down on their risk taking. Further, our findings cannot be due simply to rebalancing, since any standard type of portfolio rebalancing would increase stock holdings after losses.

Although it is difficult to attribute our results to any single explanation, our observations are consistent with an interplay between consumption habit formation and the disposition effect. Which one of these effects that will dominate in any given situation would be determined by the magnitude and sign of the past return. To see how this would work, consider the following. In the habit formation model, which is essentially a model that takes into account changing risk aversion in response to asset returns, the investor has established a minimum sustenance level based on past consumption, which she will be very reluctant to fall below.

As a consequence, the investor becomes extremely risk averse after a large negative shock to her portfolio threatens to drive her consumption below her sustenance level. In response to the rapidly increasing risk aversion, the investor will take flight from her stock positions and effectively close out of the market if the returns are low enough. Thus, after large losses, habit formation should be a dominating force driving the decrease in risk taking on the portfolio level we observe for the full sample.

On the other hand, after large gains, the investor will find herself further away from the habit level, and would hence be willing to take on more risk. However, due to the concavity of the utility function, the further away the investor gets from her habit level, the slower the risk tolerance grows, so the rebalancing motives may now fade into the background as it becomes less costly in utility terms to deviate from optimal portfolio holdings, and other effects could well dominate. In fact, as the literature on portfolio choice in the presence of transaction costs shows (see e.g.

Magill and Constantinides (1976) and Constantinides (1986)), investors are rather insensitive to deviations from their optimal portfolio weights in the standard portfolio choice setting of Merton (1973) (or in the case of Campbell and Cochrane (1999), as long as they are sufficiently far away from their consumption habit level).

10

Journal of Personal Finance

In this case of large past gains, a stock-level effect such as the disposition effect may gain importance as an explanation for our observation of selling after large gains for the full sample. The disposition effect is largely an empirical phenomenon, but recently there has been some work to attempt to find its theoretical underpinnings (see e.g.

Barberis and Xiong (2009)), where it is argued that the investor sells an asset at a gain in order to lock in a profit relative to the reference price of the asset, something the investor particularly enjoys under some alternative preference specifications, such as prospect theory (Kahneman and Tversky (1979)). This effect may help explain our finding that the investors in our sample are net sellers of stock after recent gains. Taken together, these two effects--consumption habit formation and the disposition effect--could potentially shed light on our puzzling finding that investors tend to decrease risk taking after both large gains and losses.

In a related study, Calvet et al. (2009) find that Swedish households actively rebalance about one half of the passive variation in their proportion of wealth invested in risky assets. Households whose portfolios rise in value tend to be net sellers of risky assets, and households whose portfolios fall in value tend to be net purchasers of risky assets. While our findings at first might seem to be at odds with their results, Calvet et al. (2009) also examine the households' decision to exit the stock market, and find that undiversified households and households with an extremely high initial risky share tend to be more likely to exit. This could potentially reconcile our results, since in their rebalancing analysis, Calvet et al. (2009) require the households to maintain their participation in the stock market throughout the measurement period. Another potential explanation for the difference in the results may be that while Calvet et al.

(2009) predominantly use the risky share (defined as the proportion of total wealth invested in stocks and mutual funds) as their measure of risk taking, we use trades-adjusted portfolio volatility, which takes into account both changes in dollar exposure and stock portfolio volatility.

Our findings are important not only for understanding individual investor behavior and the potential welfare consequences of the observed decrease in risk taking following large gains or losses, but also have wider implications: if the trades of individual investors are correlated, aggregate gains and losses can have an impact on aggregate trading by individual investors and hence drive asset prices and returns. Research in this direction has been conducted by Kumar and Lee (2006) and Barber et al.

(2009), among others. For market-wide implications, also see Griffin et al. (2007). As markets become more integrated, the gains and losses of individual investors may also become more correlated, increasing the possibility of systemic effects arising from the investor behavior documented in our study.

The remainder of the paper is structured as follows.

Section 2 describes the data and construction of measures.

Section 3 contains the methodology and results, and Section 4 concludes.

Data and Measures Data Sources and Definitions We use a large dataset of the trades and monthly holdings of individual investors at a major discount brokerage firm in the United States. Barber and Odean (2000), the first to analyze these data, provide a detailed description of the dataset, which we summarize here. It includes information on 78,000 households from January 1991 to December 1996. Many of these households have multiple accounts with the brokerage firm, such as individual retirement accounts and custodial accounts. We aggregate the holdings of each household for our analysis.

The vast majority of households (85%) hold a common stock in at least one month. Investments in common stocks comprise about 60% of the value of all investments held at the firm. As in Barber and Odean (2000), we focus our analysis on the common stock holdings of each investor. Our main results also hold when including mutual funds and other traded securities in the investors’ portfolios, such as closed-end funds and REITs. We examine how prior outcomes affect future risk taking among stock investors. To calculate the prior outcome for an investor, we measure that investor’s stock portfolio return over the first six months of the year and then proceed to measure the change in risk taking during the final six months of the year. In the analysis we use both halves of the year to measure returns and subsequent risk taking. In other words, in addition to measuring returns over the first six months of the year and change in risk taking over the final six months of the year, we also measure returns over the last six months of the year, and the change in risk taking over the first six months of the following year. This is consistent with investors having a slowly updating reference point when they measure gains and losses. In what follows, when we refer to the first six months of the year (or the first half of

Š2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1 the year), this is not to be interpreted as the first six months in the calendar year, but rather as the first half of a one year time-frame. However, our results also hold when restricting the sample to only calendar year time-frames.

We begin by calculating daily portfolio return using the holdings as of the end of the previous six month period, and end with the return on the last day of trading in the final month of the six month period. We use the monthly portfolio snapshots of the data taken at the end of each month for each investor. Following Barber and Odean (2000), we assume that investors hold these stocks for the following month.

Thus, we ignore any intra-month round-trip trades, and we effectively assume that investors only trade on the final day of the month. These assumptions greatly simplify the computation involved in calculating each return and risk measure, while not biasing our results in any systematic direction. Since the monthly positions data do not provide the portfolio positions for the beginning of January 1991, we begin our analysis with the end of June 1991 holdings.

To be included in our main sample in a given year, the household must hold at least one stock at the beginning of the first six months of the year and a minimum of one stock for at least one month in the second half year. For both our return and risk measures we need daily return data for each common stock, which we obtain from the Center for Research in Security Prices (CRSP). We also obtain daily return data for the risk free rate in the United States. If the return data does not include a return for a stock on a day that an investor held the stock, we exclude that stock from the investor’s portfolio for that day.

Construction of Measures We compound an investor’s daily portfolio return over the first half of the year to measure their prior outcome. We use a net return of zero as the reference level for gains and losses, which is perhaps the most natural choice. Other reasonable reference points could be the risk free rate, as suggested by Barberis and Xiong (2009), or the return on a market index. We find no difference in our results when using these benchmark returns as the reference point from where to measure gains and losses. Because investors hold highly undiversified portfolios, there is significant crosssectional variation in return across investors within the same six-month period. We exploit this variation to examine how return over the first six months of the year affects risk taking in the following six months.

We use the percentage volatility of an investor’s daily stock portfolio return over the second half of the year to measure how much risk the investor is taking in her stock

11 portfolio. Although it would be natural to measure the change in risk taking during the second half of the year relative to the volatility during the first half of the year, this measure is problematic because we cannot be sure that the first six months of return is driving the change in risk rather than vice versa. For example, consider an investor who increases her portfolio volatility dramatically during the middle of the first half of the year for reasons unrelated to prior outcomes and suppose this increase in volatility during the middle of the first half of the year increases the cumulative return for the first six months. Even if this investor makes no changes during the second half of the year, her change in risk taking is positive since she holds a less risky portfolio for the first three months of the year. In this example, the increase in volatility created the strong first six month return, not the direction of the effect we want to analyze. Another related problem is that extreme outcomes in the returns variable are positively related to volatility during the same period, and hence a negative difference between the volatility during the second and the first half of the year may simply be due to the spike in volatility during the first half of the year, generated by the extreme return realization, tapering off.

Therefore, we instead measure the baseline level of risk taking as the daily return volatility of the investor’s holdings as of the end of the first six months of the year held passively for the remainder of the year. Thus, if an investor makes no changes to her portfolio during the second half of the year, her change in risk taking relative to this passive measure will be zero. On the other hand, if the investor actively changes her holdings in the portfolio during the second half of the year, this change will impact the actual portfolio volatility measured over this period, and thus we will observe a difference between this actual (active) measure of volatility and the passive baseline volatility. This method breaks the effect of a change in volatility on the measure of the prior outcome, allowing us to analyze the direction of the effect we are interested in; how the prior outcome affects subsequent risk taking. This method is similar in spirit to that of Huang et al. (2011) who examine the risk shifting behavior by mutual fund managers.

More specifically, the portfolio return used to calculate the active measure of volatility is the actual return realized by the investor, taking any rebalancing observed in the monthly snapshots into account. The active return for household h over the period t to T is calculated as

12

Journal of Personal Finance !

đ?‘&#x;!,!,! = Â

(1 + !!!

đ?‘&#x;!,! đ?‘¤!,!,! )

(1)

!

where đ?‘&#x;!,! is the return on stock i during day t, and đ?‘¤!,!,! is the weight of stock i in the portfolio of household h at the beginning of day t. In contrast, the returns used to compute the passive volatility measure take the portfolio weights as of end of the first six-month period and assume that the investor passively holds the same portfolio for the remainder of the year. Thus, the passive returns over the period are calculated as the buy-and hold returns of the portfolio over the second half of the year: !

đ?‘&#x;!,!,! = Â

[�!,!,!!! !

(1 + đ?‘&#x;!,! )]

(2)

!!!

In order to properly assess the volatility of the portfolio, we need to assign suitable weights to the stocks in the portfolio. This is not trivial, as we want to account properly for the sale and purchases of stocks during the six month period. For example, if an investor sells a stock, it is likely that the volatility increases as measured within the stock portfolio, since the portfolio now contains one less stock and thus is less diversified. However, intuitively we would expect the sale to express a preference by the investor to reduce the risk of her overall portfolio, as she effectively shifts funds from a risky asset (the stock) to a risk free asset (cash). Unfortunately, the dataset does not provide a direct assessment of cash balance in the account, so we calculate a cash balance based on the transaction data. If a sale of stock is not directly offset by a purchase of another stock, the proceeds are presumably either held in cash or consumed.

There is the possibility that the investor withdraws the proceeds from the stock sale and reinvests the funds in the stock market, either directly or indirectly, through another broker. We ignore this possibility in what follows, noting that there is no reason to believe that this behavior would be more pronounced in certain sections of the past return spectrum, so it should not have an impact on our results.

Similarly, if an investor purchases a stock in the middle of the six-month period, without having sold any stocks during any of the previous months in the period, we can infer that the investor previously had these funds accessible in cash.

The funds may also have been accumulated from labor income, which in a permanent income sense is as good as cash. Thus, we can keep track of the cash balance of the investor as implied by the sales and purchases we observe over the period and work backwards to obtain the implied cash balance at the beginning of the period. Armed with the

implied cash balance, we now define the weight of stock i at the beginning of day t in the portfolio belonging to household h as �!,!,! =

đ??¸đ?‘„!,!,!!! ! đ??¸đ?‘„!,!,!!! +  đ??śđ?‘Žđ?‘ â„Ž!,!!!

(3)

where đ??¸đ?‘„!,!,!!! is the equity value of stock i at the end of the previous day, and đ??śđ?‘Žđ?‘ â„Ž!,!!! is the implied cash balance at the end of the previous day. The cash balance is assumed to grow at the risk free rate.

As a result of this definition of the portfolio weights, the active volatility is the standard deviation of daily total portfolio returns, while taking the changing cash balance into account over the half year period, and with monthly rebalancing of the holdings to reflect the actual trades made by the investor. Passive volatility, on the other hand, is the standard deviation of the daily returns of the investor’s passive holdings, taking the positions at the beginning of the six month period and assuming the investors hold this portfolio for the remaining six months, while only taking the beginning of period cash balances into account. As mentioned before, only if the investor does not change the composition of her portfolio over the entire period are the two measures of volatility the same.

Note how the cash balance mutes the volatility of the portfolio. In the absence of any cash balance, the portfolio volatility would simply be the volatility of the return on the stock holdings. However, when introducing the cash balance, the total volatility will be dragged towards zero by the portion of the portfolio allocated to the implied cash account. Thus, a sale will have a negative impact on portfolio volatility going forward, while a purchase will have a positive impact on portfolio volatility going forward. This happens since in the case of a purchase, the portfolio prior to the purchase was dragged down by the cash balance implied by the observed purchase. Note also how sales, by themselves, will not have any impact on the calculation of passive volatility, as the passive volatility only take beginning cash balances into account. In contrast, any purchases not financed by prior sales during the period will lower the passive volatility as these purchases are assumed to be financed out of the implied cash account held at the beginning of the period.

Differences in the active volatility measure and the passive one must stem from one the following two channels: a change in money invested in the stock portfolio, i.e. net purchases or sales over the period; or a change in the stock

Š2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1 portfolio composition so that more or less risky or correlated assets are shifted in and out of the portfolio. In contrast, any changes in the volatilities of individual stocks, or the correlations among the stocks held, are absorbed by the passive volatility benchmark and hence do not impact the difference between the two volatility measures. This is important since stocks with spikes in return and volatility in one period are likely to slowly revert back to the mean in subsequent periods. These stocks would also be more likely to populate the portfolios of investors with extreme portfolio returns, which are the ones that are the focus of our study.

Sample Summary Table I presents summary statistics for our measure of an investor’s prior outcome and various measures of stock portfolio volatility and portfolio characteristics. To ensure that outliers do not drive our results, we drop all observations with volatility and total equity values in the top and bottom half-percentile for each time period of the sample. The table is sorted into deciles based on the prior outcome (past return). There is a large dispersion in past return, with investors in the lowest decile earning on average a negative 25% semi-annual return, while the investors in the top decile earn on average close to 50% returns. The mean return for the full sample is around 7.2% on a semiannual basis, which is somewhat lower than for the market, which yielded a return of around 7 to 9% semi-annually over the same period, depending on the exact market index used.

The second column in the table displays the current return, which is measured during the second half of the year.

Moving on to the volatility measures, we see that past volatility is much higher for the deciles with the highest and lowest past returns, as we would expect, since an extreme return for the portfolio is generally associated with a spike in volatility over the same period. More interestingly, investors who experienced an extreme return in the past have persistently high volatility in their returns, so the active and passive measures of volatility for the following period are high as well. This is partly because the investors experiencing extreme returns hold riskier portfolios than the average investor, but may also be due to a persistence in volatility on the stock level. As already mentioned, we control for this latter channel by defining the change in risk taking as the difference between active and passive volatility over the same time period. Although the passive and active measures of volatility exhibit a largely similar pattern across past return deciles, the difference between these measures shows a distinct trend for the full sample, as we shall see in the next section.

13 Some final observations from Table I are that the total equity and number of stocks in an investor’s portfolio as of the end of the first half of the year are significantly lower for the deciles with the most extreme past outcomes. This is as expected, since more concentrated portfolios contain more idiosyncratic risk and hence are more likely to exhibit extreme outcomes. Another reason for the lower portfolio value in the lowest past return decile is simply that this portfolio has lost a significant part of its value over the past half year by construction.

Table II reproduces the statistics from Table I, but for the surviving sample of investors only. Since these investors represent the vast majority of the total observations (89%, as can be seen by comparing the last rows of Table I and II), most of the statistics in Table II are similar to those of Table I. However, one important difference, as will become clear below, is the trend in the active and passive volatility measures across deciles.

Method and Results Prior Returns and Changes in Risk Taking To examine the effect of prior outcomes on future risk taking, we first examine how risk taking changes after sorting observations by return deciles. Table III presents the change in risk taking by return decile for the full sample of investors. We sort households by the prior period’s return on the household portfolios into deciles. Then, we consider two measures of changes in risk; the average magnitude of the change in risk taking for each decile, and the proportions of investors either increasing or decreasing their risk taking.

As a measure of the magnitude, we use average Delta Vol, the mean percentage difference between the volatility of an investor’s actual stock holdings over period final six months of the year and the passive volatility of the investor’s holdings as of the end of the previous six-month period, assuming a buy-and-hold position in the stocks. This percentage is slightly negative on average and becomes even more negative in the extreme return deciles, indicating that investors tend to decrease their risk taking after large gains or losses. Thus, in aggregate, it is the absolute magnitude of the returns that seems to matter, rather than the direction of the returns. This is however not the case when only considering the investors who stay in the sample, as presented in Table IV. Here, we find the Delta Vol measure to be positive on average, and exhibiting a slight downward trend across return deciles, so investors tend to increase volatility relative to the passive benchmark after losses, and increase it less after gains.

14

Journal of Personal Finance

The alternative measure we use only accounts for the direction of investor’s changes in risk, and not the magnitude. It is therefore more robust to outliers and tells us how many of the investors in the sample change their risk taking in response to prior outcomes. This measure is separated into three parts. Vol Up is the percentage of observations in which investors increase their risk taking in the final six months of the year. For the full sample in Table III, this percentage exhibits a hump-shape across the return deciles. Vol Down is the percentage of observations in which investors decrease their risk taking in the final six months of the year. This measure is lower on average for the lower return deciles than among the higher return deciles, but stays relatively flat overall for the full sample. Up / Down is the ratio between the number of observations in which investors increase their risk and the number of observations in which investors decrease their risk. This measure exhibits a humpshape across the return deciles. Thus, for the full sample, this pattern indicates that more extreme magnitudes of prior outcomes decrease investors’ propensity to take on risk.

Looking at the Up / Down measure for the surviving sample only, in Table IV, we find that this group of investors tend to increase their risk on average, and that this effect is less pronounced after large gains.

To more rigorously examine the potentially different impact of positive returns and negative returns we observe in the decile sorts, we now estimate an ordinary least-squares regression where we separate returns into those observations of positive return and those of negative return. The regression model used in the analysis !!"#$%!!"#!!! = !+!! !"#!!!!! ! !!!!"#!!!!! ! !!!"#$!"#!!!!! ! !!!!!!! ! !!!!

(4)

where RET is the return on investor h’s stock portfolio over the first six months of the year. DUMNRET is a dummy that takes value one if the return on an investor’s portfolio is negative over the first six months of the year, and zero otherwise. NRET is the return on an investor’s stock portfolio over the first six months of the year, whose value is the product of RET and DUMNRET, and is thus only different from zero in the negative domain of returns. We look for the incremental effect of returns on changes in risk taking by estimating a regression of Delta Vol on RET, the return on an investor’s stock portfolio over the first six months of the year, NRET, the return on an investor’s stock

portfolio over the first six months of the year when that value is negative and zero otherwise, the negative returns dummy DUMNRET, and !, which is a set of control variables. The control variables include the number of stocks in the portfolio, the dollar value of the equity positions, as well as Vol Reversion, which is a variable measuring volatility reversion and is based on past changes in total volatility of the equity portfolio. More specifically, the Vol Reversion variable is defined as the difference between lagged volatility and twice lagged volatility. It attempts to capture the reversion to mean volatility (or preferred risk habitat in the language of Dorn et al. (2009)) after a spike in volatility during the first six months of the year due to extreme returns. We also include time fixed effects for each half-year to account for time trends in risk taking and to control for market effects. All regressions employ robust standard errors as in White (1980).

Table V reports the results of these regressions for a number of specifications for the full sample, while Table VI reports the results for the surviving sample only. Starting with the results for the full sample in Table V, the loadings on the return variables are all statistically significant. In model 1, which does not include any controls, the loading on RET is -0.025, and the loading on NRET is 0.057. To understand these loadings, let us examine the marginal effect of a 10% increase and decrease, respectively, in past returns on the next period risk taking. We find that a 10% increase in return during the first half of the year results in a 0.25% decrease in risk taking during the second half of the year, while each 10% decrease in return results in a 0.32% decrease in risk taking (the overall effect in the negative domain is obtained by adding the RET coefficient to the NRET coefficient). To consider the economic effect of these findings, note that in our decile sorts for the full sample (see Table I), we observe an absolute magnitude of returns in the region of 30% for the top and bottom deciles, and an average volatility in the mid 30%. Thus, our regression results tell us that investors on average are expected to decrease their risk taking by close to one incremental percentage point after both gains and losses in the most extreme decile, which translates into close to 3% of the level of volatility. This is also consistent with the pattern found in Table III for the full sample, after taking out the mean change in volatility of the middle deciles. After adding controls for portfolio size, number of stocks and changes in past volatility, the results are quite similar. When considering the surviving sample in Table VI, we find the same downward-sloping pattern as shown in Table IV, meaning that these investors tend to

©2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1 increase their risk taking after losses, and decrease it after gains.

Figures 1 and 2 repeats the regressions in Tables V and VI, except that we replace the continuous returns variables with dummy variables with specific cut-offs to better estimate the non-linear relationship of positive and negative returns on future risk taking. We consider the following cutoffs for past returns: less than -20%, -20% to -10%, -10% to -5%, -5% to 5%, 5% to 10%, 10% to 20% and greater than 20%. We use -5% to 5% as the baseline in our regressions.

The results in Figures 1 and 2 are qualitatively similar to those in Table V and VI respectively.

Prior Returns and Investor Trading Next, we examine the determinants of the change in volatility after large gains or losses. In the previous section, we found that investors in the full sample tend to decrease their overall risk in their stock portfolio after large absolute levels of return, but that investors who stay in invested actually increase their risk after losses while decreasing it after gains. However, it is still not clear what actions investors take to change the volatility of their portfolios in response to past outcomes. Given the measure of change in risk taking that we employ, there are only two ways in which the investor can change her risk taking: a change in money invested in the stock portfolio, i.e. net purchases or sales over the period; or a change in the stock portfolio composition so that more or less risky or correlated assets are held. In this section we investigate the trading behavior of the investor to see whether overall turnover and net sales of stocks can explain the change in risk taking following prior gains or losses.

We will first address the results for the full sample, and then the results for the surviving sample. Table VII shows how investors’ trading behavior in the full sample differs across deciles when sorted on past returns. We find that overall dollar turnover in the final six months of the year is lower for the losing investors, while the turnover is relatively flat for the deciles with medium and high returns.

The same pattern holds for the dollar value of the net sales, which is defined as the total value of all stock sales minus the total value of all stock purchases over the period. This measure includes trades of partial positions as well as changes in positions over months within the six-month period, but does not include intra-month round-trip trades.

However, we know from Table I that the investors in the most extreme deciles also hold the smallest portfolios in terms of value and number of stocks, and this is especially true on the loser side, since these portfolios by construction

15 will have declined in value over the past six months.

Therefore, a more appropriate measure of net sales is the proportion of turnover that are comprised of net sales, or more precisely, the dollar value of net sales divided by the dollar value of overall turnover. This measure explains the decrease in risk taking we observe for the past losers and past winners, as we find that for the full sample, the more extreme deciles have a higher proportion of net sales compared to the middle deciles. This finding is supported by another measure of net sales, which is the dollar value of net sales as a percentage of the value of the portfolio at the end of the first six months of the year. A graphical representation of this pattern can be found in Figure 3, which displays total sales and purchases as a percentage of portfolio value. In the figure, we note that although both sales and purchases exhibit a U-shaped pattern, the increase in sales is more pronounced than the increase in purchases in the extreme return deciles, leading to the U-shaped pattern in net sales observed in Table VII. Finally, note how the overall dollar turnover as a percentage of the portfolio value displays a U-shaped pattern (again, see Table VII and Figure 3), meaning that extreme past returns tend to induce investors to trade. This is consistent with the Barber and Odean (2008) finding that extreme returns during the past day drive trading on an individual stock level, which they attribute to attention.

Other explanations, such as rebalancing, may also be applicable, although the rebalancing explanation is difficult to reconcile with the net sales on the losing end. If anything, an investor is expected to increase the holdings of the risky asset following a downturn in the stock market, as the investor’s risky share of her overall portfolio now will be below her preferred level of overall risk taking (see e.g.

Merton (1973)).

To better take into account the interplay between past returns, trading statistics, and portfolio characteristics, we repeat the analysis in a regression framework analogous to the one employed in the previous section. The results for the full sample, presented in Table IX, display a U-shaped relation between past return and future trading and net sales, as evident by the positive and significant loading on RET and the negative and significant loading on NRET across all specifications, combined with the fact that the sum of the loadings on RET and NRET is consistently negative. Thus, the regression estimates, which are measured in dollar terms, are consistent with the ratio measures from the preceding analysis.

In order to get a sense of the relative magnitudes of the impact on overall portfolio volatility from net sales versus

16

Journal of Personal Finance

changes in portfolio volatility among the stocks in the portfolio, let us consider the following simple decomposition. Recall that across all deciles of past return, the total change in portfolio volatility, Delta Vol, is on average -1.0 percent points, with the benchmark passive volatility being on average 26.0 percent (see Tables I and III). We know from Table VII that the average net sales as a proportion of the value of the investor’s portfolio is 5.8 percent. For the portion of the portfolio that is sold, the volatility goes to zero as the proceeds from any sales that are not directly offset by any purchases translate into cash.

However, the net sale of stocks may occur at any time over the second half of the year, with sales that take place early during the period having a larger impact on the overall change in portfolio volatility for the period. Assuming that on average, the sales occur during the middle of the sixmonth period, we find that a reduction of stocks in the portfolio by 5.8 percent means that the overall portfolio volatility goes down by 26.0 * 0.058 / 2 = 0.75 percentage points. In other words, under this assumption, 75 percent of the total magnitude of the average Delta Vol is explained by net sales, leaving 25 percent to be explained by changes in volatility within the stock portfolio. Note that this case assumes that the volatility of the stocks net sold is the same as the average portfolio volatility at the beginning of the period. If the stocks net sold have a higher than average volatility, which is to be expected, the impact from net sales would be even higher. While this discussion only addresses the average effects in the sample, the pattern is similar for the various deciles of portfolios based on past return.

The discussion above relates to the full sample. On the other hand, investors who remain in the sample (survivors) exhibits markedly different trading behavior. These investors have negative net sales on average (meaning that they are net buyers of stocks), and as a percentage of the value of their stock portfolios, this net buying is highest after past large losses. These results are presented in Table VIII (sorts) and Table X (regressions), and in Figure 4.

Prior Returns and Stock-Level Trading In this section we further investigate the underpinnings of the main observations we have made so far about aggregate portfolio changes in response to prior outcomes.

To do so, we set out to answer the question of which stocks the investors trade after prior gains or losses. Although we may a priori expect the trading to be focused on poorly performing individual stocks in the portfolio, it is not necessarily the case. For example, the pervasive empirical

phenomenon dubbed the disposition effect (see Shefrin and Statman (1985) and Odean (1998)) tells us that on an individual stock level, investors tend to hold on to losers and sell winners. In the domain of losses, then, it is reasonable to expect investors to engage in trading stocks other than the ones that perform especially badly.

Our analysis proceeds by examining how past stock returns impact the trading in each stock in the portfolio at a time. Thus, in contrast to the earlier analysis, which was conducted on a household and date level, we now use each stock in the household’s portfolio at each date as one observation. We break up the past portfolio returns into the return attributable to a specific stock i, called RET, and the return attributable to the rest of the portfolio (i.e. the complement of stock i), called RETX. These returns are measured over the first six months of the year, as before.

Although the various stocks composing the portfolio exhibit a positive correlation most of the time, there is still significant idiosyncratic variation in the stock returns within each time period, which we exploit in a pooled regression framework where we include both of the past return variables simultaneously. The dependent variable in the regressions is the change in the dollar value of the holdings of stock i over the final six months of the year, adjusting for the return of that stock over the period. This measure is similar to, but distinct from, the net sales measure employed in the trading analysis in the previous section. The distinction is that the changes of holdings measure do not take any round-trip trades occurring within the six month period into account, which is done in order to focus on the changes in risk taking on an individual stock level exhibited by the investor, rather than the trading activity in that stock.

Since the analysis is carried out on the individual stock level, our measure of changes in holdings is more appropriate than using a modified version of our Delta Vol measure, which is only feasible to use on the portfolio level. The explanatory variables include the two return variables RET and RETX, negative return dummies NRET and NRETX and negative return intercept terms DUMNRET and DUMNRETX, for stock i and the rest of the portfolio, respectively, using the dummy approach described in Section 3.1, along with the dollar size of the portfolio and the number of stocks held as controls.

The results from the regressions are presented in Table XI for the full sample and Table XII for the surviving sample, respectively. Since the effects are stable across specifications in the tables, and are similar for both the full and surviving sample, we focus our discussion on the full

Š2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1 model (model 3), where both return variables are included, which means that any effects of past returns, from either stock i or the rest of the portfolio, on changes in holdings of stock i are marginal effects, controlling for the other source of return. We find that the coefficient of RET is negative and significant, meaning that a large positive past return in the stock leads to a subsequent reduction in the holding of that stock. As we can see from the loading on NRET, the marginal effect of a negative return on the stock is positive, implying that investors reduce their position in the stock after a large decline in the price of that stock. However, as we need to add the coefficients of RET and NRET to obtain the overall impact of the stock return on changes in holdings of that stock, we find the overall effect to have a somewhat negative slope, meaning that investors tend to increase their positions slightly in the stock after it performed poorly. On the other hand, the impact of the past return of the rest of the portfolio (excluding stock i) on the change in holdings of stock i has a pattern that is essentially the opposite of what we found for stock i’s own return. Here, we find that RETX has an insignificant impact on changes in the holdings of stock i, implying that the past returns on other stocks in the portfolio have little impact on the investor’s decision to sell or buy stock i. In contrast, we see a large and positive coefficient on the NRETX variable, meaning that a large negative return on the rest of the portfolio is associated with a reduction in the holding of stock i. These results are graphically represented in Figures 5 and 6, which employs the same type of multiple dummy regression technique as outlined in Section 3.1 to produce Figures 1 and 2.

To understand these results, consider how an investor would act if she starts out fully invested in an equalweighted portfolio of stocks, and intends to keep each stock at an equal weighting by rebalancing the portfolio periodically. In this case, after one of the stocks has had a large positive return, the investor will find that this stock is now over-weighted relative to the other stocks in the portfolio, and so the investor will reduce her holdings of this stock and invest the proceeds into the other stocks in her portfolio. Conversely, if the stock instead had a large negative return over the period, the investor will now be under-weighted in this stock and thus will act to increase her holdings in the stock by selling some of the other stocks in the portfolio and use the proceeds to finance the purchases of the underperforming stock. In sum, we should expect to see a negative relationship between past returns and changes in holdings for the same stock over both gains and losses.

Now let us instead consider what the investor will do to keep the portfolio equal-weighted after all the stocks but one do well in a period. This single stock will now be under-

17 weighted in the portfolio, and so the investor will increase her holdings of this stock and decrease her holdings of the other stocks. The opposite is true if all the stocks but one do poorly; this single stock will now comprise a too large fraction of the portfolio and the investor will reduce her holdings in this stock and increase her holdings of the other stocks. Hence, we should expect to see a positive relation between changes in holdings of a specific stock and past returns on the rest of the portfolio, or in other words, the opposite relationship compared to the one between the stock’s own return and changes in holdings.

Returning to our results, we see that this is partly what is happening, but only to a certain extent. In particular, we find that the investors tend to sell their winners, as they would in a rebalancing scenario, but are somewhat reluctant to increase their holdings in a particular stock after losses.

They do however sell the stock after the rest of the portfolio has experienced large losses, which is again rebalancing at work, but they fail to purchase more of the stock after the rest of the portfolio has experienced large positive returns.

Thus, it seems as if the investors only complete one leg of the necessary rebalancing act of selling the assets that have become over-weighted and buying the assets that have become under-weighted. For the full sample, the net result is that they reduce their positions in one stock by more than they increase their position in other stocks, and thus cut down on their overall portfolio exposure to risky assets after both large prior gains and losses. We know that this holds true in the aggregate, partly from the concave shape of the curves in Figure 5, and the slope coefficients in Table XI, but it is also confirmed by our result in Section 3.2, where we found that the investors on average sell more stocks than they purchase after a large prior gain or loss. However, for the surviving sample, the investors make more purchases than they make sales, so even though they exhibit a similar pattern in their stock-level trading, they purchase more than enough new stocks after prior losses to outweigh the sales, resulting in an overall increase in risk taking after losses.

Conclusion We find that those investors who experience big gains or losses in the first six months of the year are more likely to decrease the risk in their stock portfolio over the final six months of the year. The bigger the initial gain or loss, the more investors lower their subsequent risk taking. This result is inconsistent with the house money effect and break even effect driving individual investors' behavior. This main result is driven by investors leaving the stock market. In contrast, for the sample of investors remaining in the market, we find that investors increase their risk taking following

18

Journal of Personal Finance

losses, while lowering their risk after gains. Thus, we find only limited support for the break even effect, and no support for the house money effect, even in this subsample.

Overall, our results best match the predictions of consumption habit formation interacting with the disposition effect.

A large literature documents the effect of behavioral biases on individual investors' behavior. Odean (1999) and Barber and Odean (2000) show that individual investors trade too much, which is related to overconfidence (Barber

and Odean (2001)) and sensation seeking (Grinblatt and Keloharju (2009)). They buy attention grabbing stocks (Barber and Odean (2008)). They sell their winners too soon and hold on to their losers too long (Odean (1998)). Our finding that individual investors overall risk-taking behavior is not driven by the house money effect or break even effect is particularly remarkable given this wide variety of behavioral biases that do plague individual investors.

Š2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1

19

References Barber, B. & Odean, T. (2000), Trading is hazardous to your wealth: The common stock investment performance of individual investors, Journal of Finance 55(2), 773-806.

Barber, B. & Odean, T. (2001), Boys will be boys: Gender, overconfidence, and common stock investment, Quarterly Journal of Economics 116(1), 261-292.

Barber, B. & Odean, T. (2008), All that glitters: The effect of attention and news on the buying behavior of individual and institutional investors, Review of Financial Studies 21(2), 785-818.

Barber, B. & Odean, T. (2011), The behavior of individual investors, in G. Constantinides, M. Harris, & R. Stulz (Eds.), Handbook of the Economics of Finance (in press).

Amsterdam: Elsevier.

Barber, B., Odean, T. & Zhu, N. (2009), Systematic noise, Journal of Financial Markets 12, 547-569.

Barberis, N. & Xiong, W. (2009), What drives the disposition effect? An analysis of a long-standing preference-based explanation, Journal of Finance 64(2), 751-784.

Beshears, J. & Milkman, K. (2011), Do Sell-Side Stock Analysts Exhibit Escalation of Commitment?, Journal of Economic Behavior and Organization 77(3), 304-317.

Calvet, L., Campbell, J. & Sodini, P. (2009), Fight or flight?

Portfolio rebalancing by individual investors, Quarterly Journal of Economics 124(1), 301-348.

Campbell, J. & Cochrane, J. (1999), By force of habit: A consumption-based explanation of aggregate stock market behavior, Journal of Political Economy 107(2), 205-251.

Choi, J., Laibson, D., Madrian, B. & Metrick, A. (2009), Reinforcement learning and savings behavior, Journal of Finance 64(6), 2515-2534.

Constantinides, G. (1986), Capital market equilibrium with transaction costs, Journal of Political Economy 94, 842862.

Constantinides, G. (1990), Habit formation: A resolution of the equity premium puzzle, Journal of Political Economy 98(3), 519-543.

Coval, J. & Shumway, T. (2005), Do behavioral biases affect prices?, Journal of Finance 60(1), 1-34.

Dorn, D., Huberman, G., Han, B. & Kumar, A. (2009), Preferred risk habitat of individual investors, Journal of Financial Economics, 97(1), 155-173.

Frino, A., Grant, J. & Johnstone, D. (2008), The house money effect and local traders on the Sydney Futures Exchange, Pacific-Basin Finance Journal 16(1-2), 8-25.

Griffin, J., Nardari, F. & Stulz, R. (2007), Do investors trade more when stocks have performed well? Evidence from 46 countries, Review of Financial Studies 20(3), 905-951.

Grinblatt, M. & Keloharju, M. (2009), Sensation seeking, overconfidence, and trading activity, Journal of Finance 64(2), 549-578.

Huang, J., Sialm, C. & Zhang, H. (2011), Risk shifting and mutual fund performance, Review of Financial Studies 24(8), 2575-2616.

Kahneman, D. & Tversky, A. (1979), Prospect theory: An analysis of decision under risk, Econometrica 47, 263-291.

Kumar, A. & Lee, C. (2006), Retail investor sentiment and return comovements, Journal of Finance 61(5), 2451-2486.

Liu, Y., Tsai, C., Wang, M. & Zhu, N. (2010), Prior consequences and subsequent risk taking: New evidence from the Taiwan Futures Exchange, Management Science 56(4), 606-620.

Magill, M. & Constantinides, G. (1976), Portfolio selection with transaction costs, Journal of Economic Theory 13(2), 245-263.

Merton, R. (1973), An intertemporal capital asset pricing model, Econometrica 41(5), 867-887.

OConnell, P. & Teo, M. (2009), Institutional investors, past performance, and dynamic loss aversion, Journal of Financial and Quantitative Analysis 44(01), 155-188.

Odean, T. (1998), Are investors reluctant to realize their losses?, Journal of Finance 53, 1775-1798.

Odean, T. (1999), Do investors trade too much?, American Economic Review 89(5), 1279-1298.

Post, T., Van den Assem, M., Baltussen, G. & Thaler, R. (2008), Deal or no deal? Decision making under risk in a large-payoff game show, American Economic Review 98(1), 38-71.

Savor, P. & Gamboa-Cavazos, M. (2011), Holding on to your shorts: When do short sellers retreat?, working paper.

Shefrin, H. & Statman, M. (1985), The disposition to sell winners too early and ride losers too long: Theory and evidence, Journal of Finance 40, 777-790.

Statman, M., Thorley, S. & Vorkink, K. (2006), Investor overconfidence and trading volume, Review of Financial Studies 19(4), 1531-1565.

Thaler, R. & Johnson, E. (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.

White, H. (1980), A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity, Econometrica 48(4), 817-838.

20

Journal of Personal Finance

Table I

Summary Statistics for Portfolios Sorted on Past Return: Full Sample This table presents averages of returns, volatility and investor portfolio characteristics when sorting the sample based on past returns. The sorts are based on the return during the first six months of the year, labeled past return in the table. Sorts are done within half-year cohorts. Current return is the average portfolio return of the investor during the final six months of the year. Past volatility is the annualized volatility of the investor's daily portfolio returns over the first six months of the year. Active volatility is the annualized volatility of the investor's daily portfolio returns over the final six months of the year. Passive volatility is the annualized volatility of the investor's daily portfolio returns over the final six months of the year, using the holdings as of the end of the first six months of the year. Total equity is the dollar value of the stock holdings in the average investor's portfolio at the end of the six first months of the year. Nu mber of stocks is the number of stocks in the average investor's portfolio at the end of the six first months of the year. All figures are equal-weighted averages. Standard errors appear in parentheses. The sample is truncated at the 0.5 and 99.5 percentiles of the volatility and total equity value measures for each time period of the sample.

Decile Sorted on Past Return Full Sample 1 (Lowest) 2 3 4 5 6 7 8 9 10 (Highest) No. Observations

Past Current Return (% ) Return (% ) 7.2 6.2 (0.04) (0.04) -24.6 8.3 (0.07) (0.16) -9.1 7.0 (0.03) (0.11) -3.2 7.3 (0.03) (0.10) 0.8 7.0 (0.03) (0.09) 4.2 6.9 (0.03) (0.09) 7.6 6.4 (0.03) (0.09) 11.2 5.3 (0.04) (0.09) 15.6 4.6 (0.04) (0.10) 22.2 4.7 (0.04) (0.11) 47.5 5.1 (0.17) (0.14) 328,210 328,210

Past Vol (% ) 26.5 (0.03) 43.5 (0.12) 28.1 (0.08) 23.9 (0.07) 21.7 (0.07) 20.7 (0.07) 20.0 (0.06) 20.6 (0.06) 22.2 (0.07) 26.0 (0.07) 38.5 (0.11) 328,210

Active Vol (% ) 25.0 (0.03) 39.4 (0.13) 26.6 (0.09) 22.8 (0.08) 21.0 (0.07) 20.3 (0.07) 19.6 (0.07) 20.0 (0.07) 21.7 (0.07) 25.0 (0.08) 33.4 (0.10) 328,210

Passive Vol (% ) 26.0 (0.03) 41.1 (0.14) 27.7 (0.09) 23.7 (0.08) 21.8 (0.07) 21.0 (0.07) 20.3 (0.07) 20.8 (0.07) 22.6 (0.07) 26.1 (0.08) 35.3 (0.10) 328,210

Š2014, IARFC. All rights of reproduction in any form reserved.

Total Equity ($) 54,303 (180) 28,147 (344) 40,964 (442) 50,903 (523) 60,472 (603) 67,641 (656) 71,182 (687) 68,037 (666) 60,812 (608) 50,341 (524) 44,578 (500) 328,210

Number of Stocks 5.3 (0.01) 3.3 (0.02) 4.5 (0.02) 5.3 (0.03) 6.1 (0.04) 6.8 (0.05) 7.0 (0.05) 6.5 (0.04) 5.6 (0.04) 4.5 (0.03) 3.4 (0.02) 328,210

Volume 13, Issue 1

21 Table II

Summary Statistics for Portfolios Sorted on Past Return: Survivors This table presents averages of returns, volatility and investor portfolio characteristics when sorting the sample based on past returns. The sorts are based on the return during the first six months of the year, labeled past return in the table. Sorts are done within half-year cohorts. Current return is the average portfolio return of the investor during the final six months of the year. Past volatility is the annualized volatility of the investor's daily portfolio returns over the first six months of the year. Active volatility is the annualized volatility of the investor's daily portfolio returns over the final six months of the year. Passive volatility is the annualized volatility of the investor's daily portfolio returns over the final six months of the year, using the holdings as of the end of the first six months of the year. Total equity is the dollar value of the stock holdings in the average investor's portfolio at the end of the six first months of the year. Nu mber of stocks is the number of stocks in the average investor's portfolio at the end of the six first months of the year. All figures are equal-weighted averages. Standard errors appear in parentheses. The sample is truncated at the 0.5 and 99.5 percentiles of the volatility and total equity value measures for each time period of the sample. Surviv iors are the investors who still own at least one stock at the end of the final six months of the year.

Decile Sorted on Past Return Full Sample 1 (Lowest) 2 3 4 5 6 7 8 9 10 (Highest) No. Observations

Past Current Return (% ) Return (% ) 7.1 6.4 (0.04) (0.04) -24.3 8.5 (0.07) (0.18) -9.1 7.1 (0.04) (0.12) -3.3 7.5 (0.04) (0.11) 0.8 7.3 (0.04) (0.10) 4.2 7.2 (0.04) (0.10) 7.5 6.6 (0.04) (0.10) 11.1 5.5 (0.04) (0.10) 15.5 4.6 (0.04) (0.11) 22.1 4.8 (0.05) (0.12) 46.8 5.0 (0.16) (0.15) 292,706 292,706

Past Vol (% ) 26.3 (0.03) 42.6 (0.12) 27.9 (0.08) 23.9 (0.07) 21.6 (0.07) 20.7 (0.07) 20.0 (0.07) 20.6 (0.07) 22.1 (0.07) 25.9 (0.08) 37.9 (0.11) 292,706

Active Vol (% ) 25.9 (0.03) 40.3 (0.13) 27.5 (0.09) 23.7 (0.08) 21.8 (0.07) 21.2 (0.07) 20.4 (0.07) 20.8 (0.07) 22.6 (0.07) 26.0 (0.08) 34.6 (0.10) 292,706

Passive Vol (% ) 25.5 (0.03) 39.7 (0.14) 27.1 (0.09) 23.2 (0.08) 21.4 (0.08) 20.7 (0.07) 20.0 (0.07) 20.4 (0.07) 22.2 (0.07) 25.6 (0.08) 34.4 (0.10) 292,706

Total Equity ($) 55,739 (196) 29,207 (377) 41,978 (480) 52,854 (575) 62,304 (657) 68,716 (709) 72,603 (742) 69,489 (720) 62,195 (656) 52,060 (579) 46,004 (543) 292,706

Number of Stocks 5.4 (0.01) 3.3 (0.02) 4.5 (0.03) 5.5 (0.03) 6.3 (0.04) 6.8 (0.05) 7.1 (0.05) 6.6 (0.05) 5.7 (0.04) 4.6 (0.03) 3.4 (0.02) 292,706

22

Journal of Personal Finance Table III

Changes in Risk Taking when Investors are Sorted on Past Return: Full Sample This table presents changes in risk taking when sorting the sample based on past returns. The sorts are based on the return during the first six months of the year. Sorts are done within half-year cohorts. Delta vol is the change in risk taking over the final six months of the year, defined as the difference between active volatility and passive volatility during the final six months of the year. Active volatility is the annualized volatility of the investor's daily portfolio returns over the final six months of the year. Passive volatility is the annualized volatility of the investor's daily portfolio returns over the final six months of the year, using the holdings as of the end of the first six months of the year. Vol Up is the proportion of investors who increase their portfolio volat ility during the final six months of the year, compared to the passive volatility benchmark (i.e. delta vol is positive). Vol Down is the proportion of investors who decrease their portfolio volatility during the final six months of the year, co mpared to the passive volatility benchmark (i.e. delta vol is negative). Vol Same is the proportion of investors who hold their portfolio volat ility during the final six months of the year constant compared to the passive volatility bench mark (i.e. delta vol is zero ). Up / Do wn is the rat io of the nu mber of investors who increase their risk to the nu mber o f investors who decrease their risk. All figures are equal-weighted averages. Standard errors appear in parentheses. The sample is truncated as described in Table I. Decile Sorted on Past Return Full Sample 1 (Lowest) 2 3 4 5 6 7 8 9 10 (Highest) No. Observations

Past Return (% ) 7.2 (0.04) -24.6 (0.07) -9.1 (0.03) -3.2 (0.03) 0.8 (0.03) 4.2 (0.03) 7.6 (0.03) 11.2 (0.04) 15.6 (0.04) 22.2 (0.04) 47.5 (0.17) 328,210

Delta Vol (% ) -1.0 (0.01) -1.7 (0.06) -1.1 (0.04) -0.8 (0.04) -0.8 (0.04) -0.7 (0.04) -0.7 (0.03) -0.8 (0.04) -0.9 (0.04) -1.1 (0.04) -1.9 (0.06) 328,210

Prop. Vol Up (% ) 46.7 (0.09) 41.1 (0.27) 45.5 (0.27) 47.6 (0.28) 48.9 (0.28) 50.2 (0.28) 50.6 (0.28) 49.9 (0.28) 47.3 (0.28) 45.0 (0.27) 40.7 (0.27) 328,210

Prop. Vol Down (% ) 44.8 (0.09) 44.5 (0.27) 44.7 (0.27) 44.3 (0.27) 44.8 (0.27) 44.1 (0.27) 44.2 (0.27) 43.8 (0.27) 45.1 (0.28) 45.7 (0.27) 47.1 (0.28) 328,210

Prop. Vol Same (% ) 8.5 (0.05) 14.4 (0.19) 9.8 (0.16) 8.1 (0.15) 6.4 (0.13) 5.8 (0.13) 5.2 (0.12) 6.3 (0.13) 7.6 (0.15) 9.4 (0.16) 12.2 (0.18) 328,210

Š2014, IARFC. All rights of reproduction in any form reserved.

Ratio Up / Down 1.04 0.92 1.02 1.07 1.09 1.14 1.15 1.14 1.05 0.98 0.86

Volume 13, Issue 1

23 Table IV

Changes in Risk Taking when Investors are Sorted on Past Return: Survivors This table presents changes in risk taking when sorting the sample based on past returns. The sorts are based on the return during the first six months of the year. Sorts are done within half-year cohorts. Delta vol is the change in risk taking over the final six months of the year, defined as the difference between active volatility and passive volatility during the final six months of the year. Active volatility is the annualized volatility of the investor's daily portfolio returns over the final six months of the year. Passive volatility is the annualized volatility of the investor's daily portfolio returns over the final six months of the year, using the holdings as of the end of the first six months of the year. Vol Up is the proportion of investors who increase their portfolio volat ility during the final six months of the year, compared to the passive volatility benchmark (i.e. delta vol is positive). Vol Down is the proportion of investors who decrease their portfolio volatility during the final six months of the year, co mpared to the passive volatility benchmark (i.e. delta vol is negative). Vol Same is the proportion of investors who hold their portfolio volat ility during the final six months of the year constant compared to the passive volatility bench mark (i.e. delta vol is zero ). Up / Do wn is the rat io of the nu mber of investors who increase their risk to the nu mber o f investors who decrease their risk. All figures are equal-weighted averages. Standard errors appear in parentheses. The sample is truncated as described in Table I. Surv iviors are the investors who still o wn at least one stock at the end of the final six months of the year.

Decile Sorted on Past Return Full Sample 1 (Lowest) 2 3 4 5 6 7 8 9 10 (Highest) No. Observations

Past Return (% ) 7.1 (0.04) -24.3 (0.07) -9.1 (0.04) -3.3 (0.04) 0.8 (0.04) 4.2 (0.04) 7.5 (0.04) 11.1 (0.04) 15.5 (0.04) 22.1 (0.05) 46.8 (0.16) 292,706

Delta Vol (% ) 0.4 (0.01) 0.6 (0.04) 0.4 (0.03) 0.5 (0.03) 0.4 (0.03) 0.4 (0.03) 0.4 (0.03) 0.4 (0.03) 0.4 (0.03) 0.4 (0.03) 0.2 (0.05) 292,706

Prop. Vol Up (% ) 51.4 (0.09) 45.7 (0.29) 50.0 (0.29) 52.7 (0.29) 53.6 (0.29) 54.9 (0.29) 55.4 (0.29) 54.7 (0.29) 52.3 (0.29) 49.9 (0.29) 45.3 (0.29) 292,706

Prop. Vol Down (% ) 39.2 (0.09) 38.3 (0.28) 39.0 (0.28) 38.9 (0.29) 39.4 (0.29) 38.9 (0.29) 38.9 (0.28) 38.3 (0.28) 39.4 (0.29) 39.9 (0.29) 41.1 (0.29) 292,706

Prop. Vol Same (% ) 9.3 (0.05) 16.0 (0.21) 11.0 (0.18) 8.5 (0.16) 7.0 (0.15) 6.2 (0.14) 5.7 (0.14) 6.9 (0.15) 8.4 (0.16) 10.2 (0.18) 13.6 (0.20) 292,706

Ratio Up / Down 1.31 1.19 1.28 1.36 1.36 1.41 1.42 1.43 1.33 1.25 1.10

24

Journal of Personal Finance Table V

Regressions of Changes in Risk Taking on Past Return: Full Sample This table reports the estimates for pooled OLS regressions of delta vol on past returns and investor portfolio characteristics. Delta vol is the change in risk taking over the final six months of the year, defined as the difference between active volat ility and passive volatility during the final six months of the year. Active volatility is the annualized volatility of the investor's daily portfolio returns over the final six months of the year. Passive volatility is the annualized volatility o f the investor's daily portfolio returns over the final six months of the year, using the holdings as of the end of the first six months of the year. RET is the return on an investor's portfolio over the first six months of the year, while NRET is the return on an investor's portfolio over the same period if this value is negative, and zero otherwise. DUM NRET is a dummy that takes on the value one if RET is negative and zero otherwise. Total equity is the dollar value of the stock holdings in the average investor's portfolio at the end of the six first months of the year. Nu mber of stocks is the number of stocks in the average investor's portfolio at the end of the six first months of the year. Vol reversion is the difference between lagged volatility and twice lagged volatility. The regressions are estimated with fixed effects for each half-year. The samp le is truncated as described in Table I. The t -statistics (in parentheses) are based on White's (1980) heteroskedasticity-robust standard errors.

Dependent Variable: Delta Vol Intercept RET NRET DUMNRET

(1) 0.005 (11.88) -0.025 (-16.38) 0.057 (16.19) -0.003 (-5.69)

Log (Total Equity)

(2) -0.047 (-36.21) -0.022 (-14.23) 0.039 (11.05) -0.002 (-4.55) 0.005 (43.90)

Log (No. Stocks)

Model (3) -0.004 (-6.81) -0.021 (-13.92) 0.046 (13.09) -0.002 (-4.93)

0.005 (29.06)

(4) -0.047 (-34.23) -0.022 (-14.31) 0.039 (11.14) -0.002 (-4.59) 0.005 (34.40) -0.001 (-2.05)

5.10% 328,210

5.58% 328,210

Vol Reversion Adjusted R-Square No. Observations

4.90% 328,210

5.58% 328,210

Š2014, IARFC. All rights of reproduction in any form reserved.

(5) -0.047 (-30.79) -0.024 (-13.09) 0.044 (11.03) -0.003 (-5.61) 0.005 (30.96) 0.000 (-1.00) -0.009 (-4.53) 5.81% 273,196

Volume 13, Issue 1

25 Table VI

Regressions of Changes in Risk Taking on Past Return: Survivors This table reports the estimates for pooled OLS regressions of delta vol on past returns and investor portfolio characteristics. Delta vol is the change in risk taking over the final six months of the year, defined as the difference between active volat ility and passive volatility during the final six months of the year. Active volatility is the annualized volatility of the investor's daily portfolio returns over the final six months of the year. Passive volatility is the annualized volatility o f the investor's daily portfolio returns over the final six months of the year, using the holdings as of the end of the first six months of the year. RET is the return on an investor's portfolio over the first six months of the year, while NRET is the return on an investor's portfolio over the same period if this value is negative, and zero otherwise. DUM NRET is a dummy that takes on the value one if RET is negative and zero otherwise. Total equity is the dollar value of the stock holdings in the average investor's portfolio at the end of the six first months of the year. Nu mber of stocks is the number of stocks in the average investor's portfolio at the end of the six first months of the year. Vol reversion is the difference between lagged volatility and twice lagged volatility. The regressions are estimated with fixed effects for each half-year. The samp le is truncated as described in Table I. The t -statistics (in parentheses) are based on White's (1980) heteroskedasticity-robust standard errors. Surviviors are the investors who still o wn at least one stock at the end of the final six months of the year.

Dependent Variable: Delta Vol Intercept RET NRET DUMNRET

(1) 0.002 (4.18) -0.005 (-4.18) 0.000 (-0.11) -0.001 (-1.61)

Log (Total Equity)

(2) -0.020 (-20.36) -0.004 (-3.08) -0.008 (-2.91) 0.000 (-1.04) 0.002 (25.55)

Log (No. Stocks)

Model (3) 0.003 (5.07) -0.005 (-4.44) 0.001 (0.28) -0.001 (-1.71)

-0.001 (-3.74)

(4) -0.026 (-24.09) -0.006 (-4.84) -0.004 (-1.35) -0.001 (-1.48) 0.003 (29.56) -0.004 (-21.58)

0.12% 292,706

0.54% 292,706

Vol Reversion Adjusted R-Square No. Observations

0.12% 292,706

0.36% 292,706

(5) -0.023 (-20.96) -0.005 (-3.30) -0.006 (-2.25) -0.001 (-2.17) 0.003 (25.77) -0.004 (-19.25) -0.008 (-5.35) 0.56% 240,800

26

Journal of Personal Finance Table VII

Investor Trading Behavior when Investors are Sorted on Past Return: Full Sample This table presents statistics on investor trading behavior when sorting the sample based on past returns. The sorts are based on the return during the first six months of the year. Sorts are done within half-year cohorts.

Portfolio turnover is the dollar value o f the overall portfo lio turnover during the final six months of the year.

Net sales volume is the dollar value of the net sales during the final six months of the year, defined as the value of the aggregate sales during the period minus the value of the aggregate purchases during the period.

Proportion net sales is the dollar value of net sales during the final six months of the year divided by the dollar value of the overall portfolio turnover. Turnover as % of value is the dollar value of the overall turnover during the final six months of the year divided by the total equity value of the portfolio at the end of the first six months of the year. Net sales as % of value is the dollar value of net sales during the final six months of the year divided by the total equity value of the portfolio at the end of the first six months of the year. All figures are equal-weighted averages. Standard errors appear in parentheses. The samp le is truncated as described in Table I. Decile Sorted on Past Return Full Sample 1 (Lowest) 2 3 4 5 6 7 8 9 10 (Highest) No. Observations

Past Return (% ) 7.2 (0.04) -24.6 (0.07) -9.1 (0.03) -3.2 (0.03) 0.8 (0.03) 4.2 (0.03) 7.6 (0.03) 11.2 (0.04) 15.6 (0.04) 22.2 (0.04) 47.5 (0.17) 328,210

Portfolio Turnover ($) 21,550 (139) 13,438 (368) 17,725 (389) 20,354 (432) 22,044 (459) 23,479 (432) 24,996 (465) 24,248 (436) 23,248 (467) 22,255 (436) 23,711 (497) 328,210

Net Sales Volume ($) 2,628 (64) 1,364 (147) 2,233 (168) 2,226 (175) 2,658 (191) 2,929 (207) 2,907 (236) 3,075 (206) 3,212 (222) 2,931 (195) 2,748 (253) 328,210

Prop. Net Sales (% ) 3.4 (0.12) 5.5 (0.39) 4.8 (0.39) 4.6 (0.39) 2.7 (0.39) 2.6 (0.39) 1.1 (0.39) 1.7 (0.39) 1.8 (0.39) 3.3 (0.39) 5.7 (0.39) 328,210

Turnover % of Value 37.6 (0.41) 39.6 (0.94) 37.6 (0.99) 36.1 (0.97) 33.8 (0.57) 35.4 (1.47) 34.2 (0.64) 34.7 (1.03) 37.0 (1.59) 39.0 (0.79) 48.1 (2.58) 328,210

Š2014, IARFC. All rights of reproduction in any form reserved.

Net Sales % of Value 5.8 (0.21) 5.6 (0.80) 5.6 (0.53) 5.3 (0.73) 5.9 (0.30) 4.7 (0.84) 5.1 (0.57) 5.8 (0.37) 6.0 (0.74) 6.5 (0.71) 7.2 (0.79) 328,210

Volume 13, Issue 1

27 Table VIII

Investor Trading Behavior when Investors are Sorted on Past Return: Survivors This table presents statistics on investor trading behavior when sorting the sample based on past returns. The sorts are based on the return during the first six months of the year. Sorts are done within half-year cohorts.

Portfolio turnover is the dollar value o f the overall portfo lio turnover during the final six months of the year.

Net sales volume is the dollar value of the net sales during the final six months of the year, defined as the value of the aggregate sales during the period minus the value of the aggregate purchases during the period.

Proportion net sales is the dollar value of net sales during the final six months of the year divided by the dollar value of the overall portfolio turnover. Turnover as % of value is the dollar value of the overall turnover during the final six months of the year divided by the total equity value of the portfolio at the end of the first six months of the year. Net sales as % of value is the dollar value of net sales during the final six months of the year divided by the total equity value of the portfolio at the end of the first six months of the year. All figures are equal-weighted averages. Standard errors appear in parentheses. The samp le is truncated as described in Table I. Surv iviors are the investors who still own at least one stock at the end of the final six months of the year.

Decile Sorted on Past Return Full Sample 1 (Lowest) 2 3 4 5 6 7 8 9 10 (Highest) No. Observations

Past Return (% ) 7.1 (0.04) -24.3 (0.07) -9.1 (0.04) -3.3 (0.04) 0.8 (0.04) 4.2 (0.04) 7.5 (0.04) 11.1 (0.04) 15.5 (0.04) 22.1 (0.05) 46.8 (0.16) 292,706

Portfolio Turnover ($) 20,335 (150) 13,040 (403) 16,832 (423) 19,626 (484) 20,832 (493) 21,864 (469) 22,991 (474) 22,332 (460) 21,407 (487) 21,430 (484) 22,992 (541) 292,706

Net Sales Volume ($) -551 (57) -533 (136) -278 (146) -517 (156) -473 (166) -751 (179) -940 (201) -680 (165) -569 (189) -338 (176) -429 (251) 292,706

Prop. Net Sales (% ) -7.4 (0.12) -6.4 (0.38) -6.1 (0.39) -5.9 (0.39) -7.3 (0.39) -7.3 (0.39) -8.9 (0.39) -8.8 (0.39) -9.2 (0.39) -7.9 (0.38) -6.4 (0.39) 292,706

Turnover % of Value 32.0 (0.40) 33.8 (1.04) 32.2 (1.10) 31.0 (1.04) 28.8 (0.59) 29.7 (0.91) 29.3 (0.69) 30.0 (1.16) 29.9 (0.87) 33.5 (0.84) 42.1 (2.89) 292,706

Net Sales % of Value -2.3 (0.21) -3.7 (0.88) -2.9 (0.61) -2.6 (0.76) -1.2 (0.29) -1.9 (0.70) -2.1 (0.62) -1.7 (0.39) -1.4 (0.28) -2.5 (0.75) -2.7 (0.87) 292,706

28

Journal of Personal Finance Table IX

Regressions of Investor Trading Behavior on Past Return: Full Sample This table reports the estimates for pooled OLS regressions of various trading variables on past returns and investor portfolio characteristics. In Panel A, the dependent variable is portfolio turnover, wh ich is the dollar value of the overall portfolio turnover during the final six months of the year. In Panel B, the dependent variable is net sales volume, which is the dollar value of the net sales during the final six months of the year, defined as the value of the aggregate sales during the period minus the value of the aggregate purchases during the period. RET is the return on an investor's portfolio over the first six months of the year, while NRET is the return on an investor's portfolio over the same period if this value is negative, and zero otherwise. DUM NRET is a du mmy that takes on the value one if RET is negative and zero otherwise. Total equity is the dollar value of the stock holdings in the average investor's portfolio at the end of the six first months of the year. Number of stocks is the number of stocks in the average investor's portfolio at the end of the six first months of the year. The regressions are estimated with fixed effects for each half-year. The sample is t runcated as described in Table I. The t-statistics (in parentheses) are based on White's (1980) heteroskedasticity-robust standard errors.

Intercept RET NRET DUMNRET Log (Total Equity) Log (No. Stocks) Adjusted R-Square No. Observations

Panel A: Turnover (1) (2) -211,123 -220,198 (-93.26) (-88.04) 14,094 10,693 (15.81) (12.07) -51,857 -45,328 (-29.10) (-25.68) -153 -404 (-0.40) (-1.05) 22,269 24,370 (101.16) (85.78) -7,240 (-20.10) 14.16% 14.40% 328,210 328,210

Panel B: Net Sales (1) (2) -20,471 -21,456 (-19.13) (-18.41) 1,675 1,306 (3.53) (2.79) -4,476 -3,767 (-5.26) (-4.53) -282 -309 (-1.58) (-1.74) 2,132 2,360 (20.46) (18.04) -786 (-4.45) 1.43% 1.44% 328,210 328,210

Š2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1

29 Table X

Regressions of Investor Trading Behavior on Past Return: Survivors This table reports the estimates for pooled OLS regressions of various trading variables on past returns and investor portfolio characteristics. In Panel A, the dependent variable is portfolio turnover, wh ich is the dollar value of the overall portfolio turnover during the final six months of the year. In Panel B, the dependent variable is net sales volume, which is the dollar value of the net sales during the final six months of the year, defined as the value of the aggregate sales during the period minus the value of the aggregate purchases during the period. RET is the return on an investor's portfolio over the first six months of the year, while NRET is the return on an investor's portfolio over the same period if this value is negative, and zero otherwise. DUM NRET is a du mmy that takes on the value one if RET is negative and zero otherwise. Total equity is the dollar value of the stock holdings in the average investor's portfolio at the end of the six first months of the year. Number of stocks is the number of stocks in the average investor's portfolio at the end of the six first months of the year. The regressions are estimated with fixed effects for each half-year. The sample is t runcated as described in Table I. The t-statistics (in parentheses) are based on White's (1980) heteroskedasticity-robust standard errors.

Surviviors are the investors who still own at least one stock at the end of the final six months of the year.

Intercept RET NRET DUMNRET Log (Total Equity) Log (No. Stocks) Adjusted R-Square No. Observations

Panel A: Turnover (1) (2) -206,104 -215,396 (-83.34) (-78.57) 14,791 11,179 (14.44) (10.92) -51,936 -45,065 (-26.27) (-22.98) 259 -2 (0.61) (-0.01) 21,760 23,894 (90.31) (76.62) -7,298 (-18.82) 12.84% 13.09% 292,706 292,706

Panel B: Net Sales (1) (2) 6,814 7,049 (7.13) (6.63) 376 467 (0.79) (0.99) 1,559 1,385 (1.92) (1.75) 285 291 (1.77) (1.81) -503 -557 (-5.36) (-4.56) 184 (1.17) 0.16% 0.16% 292,706 292,706

30

Journal of Personal Finance Table XI

Regressions of Changes in Individual Stock Holdings on Past Return: Full Sample This table reports the estimates for pooled OLS regressions of changes in individual stock holdings on the past return of that particular stock and the past return on the rest of the portfo lio, as well as a set of control variables. The dependent variable, change in holdings of stock i, is defined as the change in the dollar value of stock i over the final six months of the year, adjusted for the stock's return over the period. RET is the return of stock i over the first six months of the year, wh ile NRET is the return of that stock over the same period if this value is negative, and zero otherwise. DUM NRET is a dummy that takes on the value one if RET is negative and zero otherwise. RETX is the return of the rest of the portfolio (excluding stock i) over the first six months of the year, while NRETX is the return of the rest of the portfolio over the same period if this value is negative, and zero otherwise. DUMNRETX is a du mmy that takes on the value one if RETX is negative and zero otherwise. Total equity is the dollar value of the stock holdings in the average investor's portfolio at the end of the six first months of the year. Number of stocks is the number of stocks in the average investor's portfolio at the end of the six first months of the year. The regressions are estimated with fixed effects for each half-year. The samp le is truncated as described in Table I. The tstatistics (in parentheses) are based on White's (1980) heteroskedasticity-robust standard errors.

Dependent Variable: Change in Holdings of Stock i Intercept RET NRET DUMNRET

(1) 18,873 (69.62) -961 (-12.12) 726 (5.71) 373 (9.89)

RETX NRETX DUMNRETX Log (Total Equity) Log (No. Stocks) Adjusted R-Square No. Observations

-2,744 (-75.29) 3,671 (63.92) 7.40% 1,013,819

Model (2) 19,399 (70.55)

-187 (-1.34) 3,886 (14.04) 24 (0.58) -2,777 (-76.15) 3,636 (64.01) 7.35% 1,013,798

Š2014, IARFC. All rights of reproduction in any form reserved.

(3) 19,313 (69.37) -950 (-12.01) 601 (4.73) 380 (10.09) -83 (-0.60) 3,807 (13.76) -7 (-0.16) -2,768 (-75.69) 3,626 (63.75) 7.43% 1,013,798

Volume 13, Issue 1

31 Table XII

Regressions of Changes in Individual Stock Holdings on Past Return: Survivors This table reports the estimates for pooled OLS regressions of changes in individual stock holdings on the past return of that particular stock and the past return on the rest of the portfo lio, as well as a set of control variables. The dependent variable, change in holdings of stock i, is defined as the change in the dollar value of stock i over the final six months of the year, adjusted for the stock's return over the period. RET is the return of stock i over the first six months of the year, wh ile NRET is the return of that stock over the same period if this value is negative, and zero otherwise. DUM NRET is a dummy that takes on the value one if RET is negative and zero otherwise. RETX is the return of the rest of the portfolio (excluding stock i) over the first six months of the year, while NRETX is the return of the rest of the portfolio over the same period if this value is negative, and zero otherwise. DUMNRETX is a du mmy that takes on the value one if RETX is negative and zero otherwise. Total equity is the dollar value of the stock holdings in the average investor's portfolio at the end of the six first months of the year. Number of stocks is the number of stocks in the average investor's portfolio at the end of the six first months of the year. The regressions are estimated with fixed effects for each half-year. The samp le is truncated as described in Table I. The tstatistics (in parentheses) are based on White's (1980) heteroskedasticity-robust standard errors. Surviviors are the investors who still own at least one stock at the end of the final six months of the year.

Dependent Variable: Change in Holdings of Stock i Intercept RET NRET DUMNRET

(1) 15,291 (57.64) -832 (-10.43) 906 (7.21) 315 (8.60)

RETX NRETX DUMNRETX Log (Total Equity) Log (No. Stocks) Adjusted R-Square No. Observations

-2,068 (-58.22) 2,689 (47.78) 7.51% 923,080

Model (2) 15,616 (58.13)

-61 (-0.44) 2,832 (10.66) 30 (0.75) -2,089 (-58.81) 2,665 (47.77) 7.48% 923,061

(3) 15,595 (57.29) -827 (-10.39) 818 (6.51) 322 (8.78) 37 (0.27) 2,735 (10.29) 6 (0.15) -2,085 (-58.49) 2,660 (47.60) 7.53% 923,061

32

Journal of Personal Finance

Figure 1

0.4 0.2

De lta Vol (%)

0 (-INF,-20]

(-20,-10]

(-10,-5]

(-5,5)

[5,10)

[10,20)

[20,INF)

-0.2 -0.4

-0.6 -0.8

-1

Past Re turn (%)

Changes in Risk Taking and Past Return: Full Sample This figure displays the estimates from a regression of change in volatility on a series of dummies for various regions of past return and a set of control variables. The dependent variable, delta vol, is the change in risk taking over the final six months of the year, defined as the difference between active volatility and passive volatility during the final six months of the year. Active volatility is the annualized volat ility of the investor's daily portfolio returns over the final six months of the year. Passive volatility is the annualized volatility of the investor's daily portfolio returns over the final six months of the year, using the holdings as of the end of the first six months of the year. Past return is the return on an investor's portfolio over the first six months of the year. The past return is separated into a series of dummy variab les, divided into regions with fixed breakpoints as indicated in the figure. The past return dummies take the value one if the past return variable is inside a part icular reg ion, and zero otherwise. The controls are the continuous variables log of total equity, log of number of stocks in the portfolio, and vol reversion. Total equity is the dollar value of the stock holdings in the average investor's portfolio at the end of the six first months of the year.

Nu mber of stocks is the number of stocks in the average investor's portfolio at the end of the six first months of the year. Vol reversion is the difference between lagged volatility and twice lagged volatility. The regressions are estimated with fixed effects for each half-year.

The delta vol associated with the middle return dummy is normalized to zero. The samp le is truncated as described in Table I. The figure displays error bars of a 95% confidence interval based on White's (1980) heteroskedasticity-robust standard errors.

Š2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1

33

Figure 2

0.5 0.4

De lta Vol (%)

0.3

0.2 0.1 0

(-INF,-20]

(-20,-10]

(-10,-5]

(-5,5)

[5,10)

[10,20)

[20,INF)

-0.1 -0.2 -0.3

Past Re turn (%)

Changes in Risk Taking and Past Return: Survivors This figure displays the estimates from a regression of change in volatility on a series of dummies for various regions of past return and a set of control variables. The dependent variable, delta vol, is the change in risk taking over the final six months of the year, defined as the difference between active volatility and passive volatility during the final six months of the year. Active volatility is the annualized volat ility of the investor's daily portfolio returns over the final six months of the year. Passive volatility is the annualized volatility of the investor's daily portfolio returns over the final six months of the year, using the holdings as of the end of the first six months of the year. Past return is the return on an investor's portfolio over the first six months of the year. The past return is separated into a series of dummy variab les, divided into regions with fixed breakpoints as indicated in the figure. The past return dummies take the value one if the past return variable is inside a part icular reg ion, and zero otherwise. The controls are the continuous variables log of total equity, log of number of stocks in the portfolio, and vol reversion. Total equity is the dollar value of the stock holdings in the average investor's portfolio at the end of the six first months of the year.

Nu mber of stocks is the number of stocks in the average investor's portfolio at the end of the six first months of the year. Vol reversion is the difference between lagged volatility and twice lagged volatility. The regressions are estimated with fixed effects for each half-year.

The delta vol associated with the middle return dummy is normalized to zero. The samp le is truncated as described in Table I. The figure displays error bars of a 95% confidence interval based on White's (1980) heteroskedasticity-robust standard errors. Surv iviors are the investors who still own at least one stock at the end of the final six months of the year.

34

Journal of Personal Finance

Figure 3

30

% of Portfolio Value

25 20 15 10 5 0 1

2

3

4

5

6

7

8

9

10

Deciles Sorted on Past Return Sales

Purchases

Stock Trades when Investors are Sorted on Past Return: Full Sample This figure displays average investor stock sales and purchases when sorting the sample based on past returns. The sorts are based on the return during the first six months of the year. Sorts are done within half-year cohorts. Sales are the dollar value of the aggregate sales during the final six months of the year d ivided by the total equity value of the portfolio at the end of the first six months of the year. Purchases are the dollar value of the aggregate purchases during the final six months of the year divided by the total equity value of the portfolio at the end of the first six months of the year. The values of the bars are the equalweighted averages of each decile. The sample is truncated as described in Table I.

Š2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1

35

Figure 4

25

% of Portfolio Value

20

15

10

5

0 1

2

3

4

5

6

7

8

9

10

Deciles Sorted on Past Return Sales

Purchases

Stock Trades when Investors are Sorted on Past Return: Survivors This figure displays average investor stock sales and purchases when sorting the sample based on past returns. The sorts are based on the return during the first six months of the year. Sorts are done within half-year cohorts. Sales are the dollar value of the aggregate sales during the final six months of the year d ivided by the total equity value of the portfolio at the end of the first six months of the year. Purchases are the dollar value of the aggregate purchases during the final six months of the year divided by the total equity value of the portfolio at the end of the first six months of the year. The values of the bars are the equalweighted averages of each decile . The samp le is truncated as described in Tab le I. Surv iviors are the investors who still own at least one stock at the end of the final six months of the year.

36

Journal of Personal Finance

Figure 5

600 Change in Holdings of Stock i ($)

400 200 0 -200

(-INF,-20]

(-20,-10]

(-10,-5]

(-5,5)

[5,10)

[10,20)

[20,INF)

-400 -600 -800 -1000 -1200 -1400 Past Return (%) Stock i (RET)

Rest of Portfolio (RETX)

Changes in Individual Stock Holdings and Past Return: Full Sample This figure displays the estimates fro m a regression of change in individual stock hold ings on a series of dummies for various regions of the past return of that particular stock and the past return on the rest of the portfolio, as well as a set of control variables. The dependent variable, change in holdings of stock i, is defined as the change in the dollar value of stock i over the final six months of the year, adjusted for the stock's return over the period. Past return is the return over the first six months of the year measured on either stock i, called RET, or on the rest of the portfolio (excluding stock i), called RETX. Each of these past return variables are separated into a series of dummy variables, divided into regions with fixed breakpoints as indicated in the figure. The past return dummies take the value one if the past return variable is inside a particular region, and zero otherwise. The controls are the continuous variables log of total equity and log of nu mber of stocks in portfolio. Total equity is the dollar value of the stock holdings in the average investor's portfolio at the end of the six first months of the year. Nu mber of stocks is the number of stocks in the average investor's portfolio at the end of the six first months of the year. The regressions are estimated with fixed effects for each half-year. The change in holdings associated with the middle return du mmy is normalized to zero. The samp le is truncated as described in Tab le I. The figure d isplays error bars of a 95% confidence interval based on White's (1980) heteroskedasticity-robust standard errors.

Š2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1

37

Figure 6

Change in Holdings of Stock i ($)

400 200 0 (-INF,-20]

(-20,-10]

(-10,-5]

(-5,5)

[5,10)

[10,20)

[20,INF)

-200 -400 -600 -800 -1000 -1200 Past Return (%) Stock i (RET)

Rest of Portfolio (RETX)

Changes in Individual Stock Holdings and Past Return: Survivors This figure displays the estimates fro m a regression of change in individual stock hold ings on a series of dummies for various regions of the past return of that particular stock and the past return on the rest of the portfolio, as well as a set of control variables. The dependent variable, change in holdings of stock i, is defined as the change in the dollar value of stock i over the final six months of the year, adjusted for the stock's return over the period. Past return is the return over the first six months of the year measured on either stock i, called RET, or on the rest of the portfolio (excluding stock i), called RETX. Each of these past return variables are separated into a series of dummy variables, divided into regions with fixed breakpoints as indicated in the figure. The past return dummies take the value one if the past return variable is inside a particular region, and zero otherwise. The controls are the continuous variables log of total equity and log of nu mber of stocks in portfolio. Total equity is the dollar value of the stock holdings in the average investor's portfolio at the end of the six first months of the year. Nu mber of stocks is the number of stocks in the average investor's portfolio at the end of the six first months of the year. The regressions are estimated with fixed effects for each half-year. The change in holdings associated with the middle return du mmy is normalized to zero. The samp le is truncated as described in Tab le I. The figure d isplays error bars of a 95% confidence interval based on White's (1980) heteroskedasticity-robust standard errors. Surviviors are the investors who still o wn at least one stock at the end of the final six months of the year.

38

Journal of Personal Finance

DOES VISUALLY DISPLAYING PROBABILITY OUTCOMES CHANGE STOCK SELECTION? Tim S. Griesdorn, Ph.D., CFP®, AFC® Human Development and Family Studies, Iowa State University Hyrum L. Smith, Ph.D., CFP®, CPA® Personal Financial Planning, Woodbury School of Business, Utah Valley University

Prospect theory assumes individuals are more risk averse with gains than to equivalent losses and indicates how an investment question is framed influences individual decision making. However, limited research has been conducted on the influence of visually displaying the probability of different outcomes, versus in written form, on investment decision making. Hypothetical investment scenarios tested the effect of visually displaying probability information. Results suggest that the visual display of probability shifts investment preferences to the stock with the greatest probability of a gain.

Thus, it appears investors are willing to take more financial risks when probability information is displayed visually.

Financial planners could incorporate visual displays when communicating with clients to help match the clients’ risk capacity with risk tolerance.

Introduction Little is known about how visual displays of probabilities of different outcomes change perceived financial risk and an individual's decision-making process.

Prospect theory provides some insight into how risk is perceived. Prospect theory explains that investors tend to be more loss averse with gains than equivalent losses; they feel the psychological pain of losses twice as much as they feel corresponding gains (Kahneman & Tversky, 1979). Prospect theory has been tested using hypothetical investment scenarios presented in a word problem format (Kahneman & Tversky, 1979; Levy & Levy, 2002b). When word problem questions are framed in terms of a potential loss, respondents tend to exhibit risk seeking behavior, and when word problem questions framed in terms of a potential gain, respondents become loss averse. This study, the first of its kind, explores subjects’ responses to a visual display of the probability scenario of different investment outcomes and obtains qualitative feedback about rationale for selections.

The questions posed in the original prospect theory research were one-sided, meaning either in the form of a direct gain or loss scenario. This new research tests mixed outcome potentials, exploring possibilities of gain or loss

possibilities in one investment scenario. In addition, the scenarios are provided both in written and visual form describing the probabilities of different investment outcomes and participants were asked to indicate why they selected a particular investment. Finally, by asking questions with the same expected value but with varying loss probability and dollar amounts, this research explores which information, dollar amounts or probability, has the greatest influence on the decision-making process. The purpose of this study is to determine if an individual’s risk perception can be manipulated by the introduction of a visual display of probability.

Literature Review The use of visual displays of information and how they influence consumer decision making has been most widely studied in the filed of medical decision making. As Lipkus and Hollands (1999) point out, most of the research conducted has been atheoretical, without an understanding of why information displayed visually would induce greater risk aversion. Lipkus and Hollands note in their research pictorial representations along with verbal descriptions help

©2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1 improve comprehension. Other researchers have confirmed the finding that visual display of information leads to increased comprehension (Brase, 2009; Chua, Yates, & Shah, 2006; Fuller, Dudley, & Blacktop, 2001; Gigerenzer & Edwards, 2003; Stone, Yates, & Parker, 1997). It should also be noted that pictoral representations are not the only way to influence comprehension, but how numbers are presented can influence a person’s probability in estimating a correct response. Studies have shown when conditional probabilities are used, respondents’ likelihood of a correct response diminishes. However, when natural frequencies are given respondents’ likelihood for a correct response increases (Chapman & Liu, 2009; Gigerenzer, 1991; Hoffrage, Gigerenzer, Krauss, & Martignon, 2002).

Similarly, Bruin, Fischholff, Millstein and Halpern-Felsher (2000) find answers to probability questions depends upon how the question is phrased, in either a singular or distributional mode.

The pictures or graphical representations used to display the probability can also influence the decision making process. Brunette (2012) summarizes research in various forms of representation of risk, such as risk circles, risk ladders, stick figures, and array of dots and conducts an experiment utilizing all forms of risk reprepresentation to conclude risk communication method can influence biases.

Brunette finds the insensitivity to scope and gain/loss asymmetry biases can influenced based upon the method used to represent risk. Lipkus and Hollands (1999) find stick figures, facial displays, asterisks and histograms can increase risk aversion. These findings are similar to the framing effect discussed by Tversky and Kahneman (1986) which demonstrated people are more risk averse when scenarios are framed in terms of mortality rates when compared with survival rates.

Studies involving the visual depiction of probability with financial behavior are very limited. Hanna, Gutter, and Fan (2001) developed a measure of financial risk tolerance similar to the Barksy, Juster, Kimball, and Shapiro (1997) measure that is included in the Health and Retirement Study, a longitudinal panel survey of U.S. residents age 50 and older. The Hanna, Gutter, and Fan measure of financial risk tolerance used a lifetime income gamble in which a small sample of college students had to choose between a potential decrease in lifetime income versus a possible increase in lifetime income using a variety of scenarios with potential gains and losses expressed in a word problem format. Hanna and Lindamood (2004) modified the Hanna, Gutter, and Fan measure by adding a graphical depiction of the probability to clarify the impact of the different income scenarios. In the

39 Hanna, Gutter, and Fan study, relative risk aversion was 6.6; and in the Hanna and Lindamood study, the relative risk aversion was 4.4. Respodents who viewed the graphical depiction of the probability have lower realative risk aversion scores, or stated another way, increased financial risk tolerance. This paper expands the literature by asking uncertain investment related questions to a broader sample in both word and graphical formats.

The field of visual analytics attempts to understand how analytical reasoning can be assisted with visual interactive features; however, little research has been done in the field of personal financial planning. Schuchardt, Hanna, Hira, Lyons, Palmer, and Xiao (2009) note the need to understand why consumers misunderstand the financial risks they face when investing and suggest this be a future research priority.

Rudolph, Savikhin, and Ebert (2009) found that when consumers are given a tool that displays return, risk, and correlation information of potential portfolio selections, consumers make more optimal risk/return decisions when compared with wealth-time plot information. Rudolph, Savikhin, and Ebert used computer software to create the visual representations and the software has an interactive feature that allows the consumer to modify assumptions.

Their research showed visually displaying financial information increased understanding of the financial information being presented, improved decision making, and increased confidence in decisions. Bhandari, Hassanein, and Deaves (2008) found that decision support aids like graphs and feedback prompts could improve investment decision making by mitigating common cognitive biases of diversification heuristic, representative heuristic, and investment default selection which help simplify decision making but are likely to lead to sub-optimal results.

Cognitive limitations and information overload can lead to inferior decision making. Agnew and Szykman (2005) found that limiting the number of choices in a defined contribution plan helps to reduce the information overload experienced by plan participants with above average financial knowledge; however, those with below-average skills still found the investment decision to be overwhelming regardless of how the information was presented. Iyengar, Huberman, and Jiang (2004) found that while more mutual fund choices in a 401(k) plan presented more opportunities to investors it also led to information overload and reduced overall plan participation rates.

Further, humans have limited cognitive resources and time in which to make decisions. This leads to experiencing bounded rationality, or being limited in their ability to make more rational or optimal financial decisions, and the

40

Journal of Personal Finance

development of heuristics or shortcuts in decision making to deal with these limitations. Lang (2000) present evidence that the manner in which television messages are communicated, such as in more cuts from one scene to another, can significantly influence an individual's ability to process and retrieve information. In a similar fashion, this study investigates whether the manner in which invetment probabilities are displayed influences an individual's ability to process information and select stocks.

Theoretical Framework Kahneman and Tversky (1979) termed the idea of additional psychological pain experienced with a financial loss as prospect theory. Prospect theory assumes investors are loss averse and will become risk seeking to avoid the psychological pain of experiencing a loss. This pattern of behavior leads to an S-shaped utility curve where the gain portion of the curve is concave, meaning investors are more likely to be risk averse with gains, taking less risk with potential gain scenarios. The loss portion of the S-shaped utility curve is convex with investors being more willing to take additional risks to avoid the pain of realizing a loss.

With this decision making pattern, investors are quick to cash in gains and hold on to losses in the hopes the losses will be reduced in the future.

In addition to inconsistently dealing with probable gains and losses, humans have limited cognitive resources and time in which to make decisions. This leads to experiencing bounded rationality, or being constrained in the ability to make the most optimal financial decision. One theory that describes this effect is the limited capacity model of motivated mediated message processing (LC4MP). The limited capacity model of motivated mediated message processing assumes humans have a finite capacity for cognitive processing of information (Lang, 2000). The theory suggests people may not have the resources available to fully process a question because they are unfamiliar with a topic, may simply lack the motivation to devote the energy needed to fully understand the question, and that visual elements tend to utilize more cognitive resources. Therefore, it is possible that people don’t encode probability information as effectively when presented in a word problem format versus in a visual display where there is an increased likelihood of the data being included in the mental calculation. The additional encoding of the visual display of probability may assist people in understanding and differentiating the alternative scenarios. The visual display and increased cognition should make the probability

information easier to recall, thus participants being more likely to select the stock from each visual display question that presents the lowest probability of a loss when compared to the word format question.

Hypotheses Prospect theory predicts that people are loss averse and feel the psychological pain of loss about twice as much as gains of the same dollar value, therefore people should have a greater preference for stock scenarios with no chance of loss versus similar scenarios with a chance of loss and similar dollar value gains. The visual display of probability facilitates encoding of information, thus the probability aspect of the problem should be easier to recall, and therefore respondents should have a preference for the selection with lowest probability of a loss.

H1: Respondents will be more likely to select the stock from each word format question that presents the lowest dollar value of a loss.

H2: Respondents will be more likely to select the stock from each visual display question that presents the lowest probability of a loss when compared to the word format question.

Method This study used a survey to present hypothetical investment scenarios and then presented the probabilities of different outcomes so subjects had a clear frame of reference regarding the probability of the event. Each participant saw both the word problem format and the visual display format of the hypothetical investment selections and then made a selection. The results of each selection were compared to determine if the respondent had changed their investment selection or not when the probability information was displayed differently. This study replicated two hypothetical mixed outcome investment scenarios from Levy and Levy (2002b) and are termed a small-dollar-return and a largedollar-return. These questions were chosen because they had the same statistical expected value, but one option dominates another in terms of Stochastic Dominance. Prospect Stochastic Dominance occurs when an investor has an sshaped utility function, and one gamble has greater expected utility than another (Levy & Levy, 2002b). In the largedollar-return scenario, stock G (25% chance of $3,000 potential loss) dominates stock F (50% chance of $1,500 potential loss) in terms of Prospect Stochastic Dominance (Levy & Levy, 2002b); thus loss averse investors should

Š2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1 choose stock G. Second-degree Stochastic Dominance occurs when one gamble is superior to another when the investor has a concave utility function, or in other words, is risk averse. In the small-dollar-return scenario, stock G dominates stock F in terms of Second-degree Stochastic Dominance (SSD) (Levy & Levy, 2001); thus any risk averse investor should choose stock G. Sampling Procedures A 37-question survey was created which included investment related questions based on the Levy and Levy (2001; 2002a; 2002b) studies but displayed in both visual and written form as well as other demographic information.

Personal face-to-face interviews were the primary survey mode. Survey design allowed for randomized survey questions with half of the sample viewing the word problem first and half viewing the visual display question displayed first.

A cluster sample of five census tracts within a western Texas City suburb were chosen as the sample frame.

Because this population had a higher percentage of middleto-upper-income households, they were more likely to own stocks and have experience with hypothetical investment scenarios. While age, race, and education profile of this sample may not be generalized to the U.S. population, the random ordering of numeric and visual displays helped ensure the validity of the design of the study within the sample assuming any reporting biases are not systematic. A total of 34 blocks were randomly selected within these tracts; these blocks contained 601 households. The number of completed surveys was 200, for a response rate of 33% collected over seven weeks: October 2 through November 21, 2010. However, the sample could also be viewed as a 400 observation panel (visual display or not) with individual fixed effects for unobservable characteristics. Each household who participated was eligible to receive a $2 bill as an incentive payment for participation. The key questions of the survey replicated two hypothetical mixed outcome investment scenarios from Levy and Levy (2002b) and are referred to as small-dollar-return and large-dollar-return questions and two questions based on prospect theory provided by Kahneman and Tversky (1979).

Small-dollar-return scenario.

In the small-dollar-return scenario the participants saw the following word format and visual display versions of the hypothetical investment scenario: Word problem version:

41 Suppose that you decided to invest $10,000 either in stock F or in stock G. Which would you choose, F or G, when it is given that the dollar gain or loss one year from now will be as follows: Stock F

Stock G

Outcome

Probability

Outcome

Probability

-$500 Loss

1 out of 4 or 25% 1 out of 4 or 25% 1 out of 4 or 25% 1 out of 4 or 25%

$0 Gain

1 out of 2 or 50% 1 out of 2 or 50%

$500 Gain $1,000 Gain $2,000 Gain

$1,500 Gain

Please circle your preference below: Stock F

Stock G

Visual display version: Suppose that you decided to invest $10,000 either in stock F or in stock G. Which would you choose, F or G, when it is given that the dollar gain or loss one year from now will be determined as follows: Stock F dollar gain or loss will by calculated by the following: Twelve balls are placed in a bag, 6 red and 6 black. One ball will be pulled out of the bag. If a red ball is pulled out of the bag, your investment will lose $0, and if a black ball is pulled out of the bag, your investment will gain $1,500.

Stock G dollar gain or loss will by calculated by the following: Twelve balls are placed in a bag, 3 red, 3 black, 3 yellow and 3 green. One ball will be pulled out of the bag. If a red ball is pulled out of the bag, your investment will lose $500, and if a black ball is pulled out of the bag, your

42

Journal of Personal Finance

investment will gain $500. If a yellow ball is pulled out of the bag, your investment will gain $1,000, and if a green ball is pulled out of the bag, your investment will gain $2,000.

Please circle your preference below: Stock F

Stock G

Visual display version: Suppose that you decided to invest $10,000 either in stock F or in stock G. Which would you choose, F or G, when it is given that the dollar gain or loss one year from now will be determined as follows: Stock F dollar gain or loss will by calculated by the following: Which investment would you prefer? Please circle your preference below: Stock F

Stock G

Twelve balls in a bag, 6 red and 6 black. One ball will be pulled from the bag. If a red ball is pulled out of the bag, your investment will lose $1,500, and if a black ball is pulled out of the bag, your investment will gain $4,500.

Briefly describe your decision making process for this stock selection (e.g. picked the one with the biggest potential gain, smallest possible loss, smallest chance of a loss, etc.): To prevent potential priming effects (e.g. answering stock F for all scenarios) the answer to this question was reverse coded. Each participant was handed a laminated card with a single scenario at a time and asked to indicate their preference. The preference was recorded, and then the participant was given the next scenario. When analyzing the data the reverse coding of this question was corrected, and the analysis that follows is shown as if this question was not reverse coded.

Large-dollar-return scenario.

Stock G dollar gain or loss will by calculated by the following: Twelve balls in a bag, 3 red and 9 black. One ball will be pulled from the bag. If a red ball is pulled out of the bag, your investment will lose $3,000, and if a black ball is pulled out of the bag, your investment will gain $3,000.

The second hypothetical investment question or largedollar-return scenario question was shown as follows: Word problem version: Suppose that you decided to invest $10,000 either in stock F or in stock G. Which would you choose, F or G, when it is given that the dollar gain or loss one year from now will be as follows: Stock F Outcome -$1,500 Loss $4,500 Gain

Probability 1 out of 2 or 50% 1 out of 2 or 50%

Stock G Outcome -$3,000 Loss $3,000 Gain

Probability 1 out of 4 or 25% 3 out of 4 or 75%

Which investment would you prefer? Please circle your preference below: Stock F

Stock G

Š2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1

43

Briefly describe your decision making process for this stock selection (e.g. picked the one with the biggest potential gain, smallest possible loss, smallest chance of a loss, etc.): Other investment scenario questions.

In addition to these two mixed outcome hypothetical investment scenario questions, two single sided investment scenarios coming directly from Kahneman and Tversky (1979) were included in their original word format but also visually displayed. The first hypothetical investment scenario had respondents chose between a sure gain of $3,000 and an 80% chance of a $4,000 gain with a 20% chance of $0 gain. The second hypothetical investment scenario had respondents choose between a sure loss of $3,000 and an 80% chance of a $4,000 loss with a 20% chance of a $0 loss.

Empirical Model The key dependent variable in this study was whether the respondent selected a different stock when shown a visual display of probability versus the original Levy and Levy (2002b) word problem format. The following potentially confounding variables were controlled for in the study: age, education, financial risk tolerance, gender, net worth, and numeracy. These control variables were included to consider the possibility that different sub-population groups may be more or less reliant on visual displays. To estimate whether stock selection was related to graphical representation of the probability of gains and losses in the scenario, logistic regression analysis was conducted. For the financial risk tolerance measure, the study uses a 13-item scale developed by Grable and Lytton (1999). A factor analysis of this scale indicates it is useful in measuring investment risk, risk comfort and experience, and speculative risk. The reliability of the scale as measured by Cronbach’s alpha has ranged from .70 to .85 (Grable & Roszkowski, 2008; Yang, 2004). The study uses an eightitem subjective numeracy scale, which measures mathematical skill, developed by Fagerlin et al. (2007).

Demographic questions regarding age, gender and education of the respondent were included as well as financial questions regarding income, net worth, and asset allocation.

Since the dependent variable is binary (coded 1 if individual selected different stock under visual display than word format; otherwise coded 0), binomial logistic regression was utilized. Due to the small sample size the analysis was conducted using a parsimonious model of just the most significant variables. The parsimonious model is included to avoid possible over-fitting of the model which Hosmer and Lemeshow (2000) indicates as a common problem with logistic regression analyses using small samples. The regression model can be expressed as follows: log (pi /( 1-pi )) = β 0 + β 1 Decision Rationale Probe i + β 2 Financial Risk Tolerance i +

β 3 Net Worth i + β 4 Education i + ε

In this model,

pi = the probability that the household i

selected a different stock when shown the visual display versus the word format display of the question. Decision Rational Probei, and Educationi were represented as dummy variables; Financial Risk Tolerancei, and Net Worthi, were treated as continuous variables related to different household factors and concepts described in the Theoretical Framework section of this paper.

Results Descriptive Statistics Table 1 includes the descriptive statistics for the sample.

The sample was more highly educated and had higher net worth than the average for the region. Fifty percent of respondents changed stock selection after seeing the visual display of probability in the small-dollar-return scenario.

44

Journal of Personal Finance

Table 1.

Descriptive Statistics of Total Sample (n=200)

Age Subjective Numeracy Scale (8-48) Risk Tolerance Scale (13-48) Investment Allocation % of Investments Currently in Cash % of Investments Currently in Bonds % of Investments Currently in Stocks % of Investments Currently in Other Resp. Who Saw Word Problem Format First % Who Changed Stock Selection Large $ question Small $ question Income Category Under $50k $50k to $75k $75 to $100k $100 to $125k $125 to $150k Over $150k Refused/Did Not Answer Net Worth Category $0 or less Up to $25k $25k to $50k $50k to $75k $75k to $100k $100k to $150k $150k to $200k $200k to $250k $250k to $300k $300k to $400k $400k to $500k Over $500k Refused/DK/Did Not Answer Educational Attainment High School Graduate or Less Some College Associates Degree Bachelor’s Degree Some Graduate School Graduate Degree

Mean

Std. Dev.

52 39 25

17.12 4.84 4.92

40% 18% 39% 3%

36% 23% 31% 11% Percentage 49% 35% 50% 18% 27% 20% 16% 4% 16% 2% 5% 12% 7% 8% 7% 6% 3% 7% 5% 9% 6% 24% 4% 7% 23% 8% 26% 8% 30%

©2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1

45

As shown in Figure 1, after answering the word problem format for the small-dollar-return scenario, 40% of respondents preferred stock F ($500 loss potential and highest possible gain of $2,000) whereas 60% preferred stock G (no loss potential and highest possible gain of $1,500). On the other hand, after seeing the visual display of probability information in the small-dollar-return scenario, 65% of respondents preferred stock F ($500 loss potential and highest possible gain of $2,000), whereas 35% preferred stock G (no loss potential and highest possible gain of $1,500).

Figure 1.

80% 60% 20%

60% 40%

65% 35%

Word Problem

Visual

Figure 2.

Large $ Return Investment Scenario

80% 60% 40% 20% 0%

Small $ Return Investment Scenario

40%

the large-dollar-return scenario scenario respondents had a clear preference for stock G ($3,000 loss potential and highest possible gain of $3,000) over stock F ($1,500 loss potential and highest possible gain of $4,500). 63% (125 respondents) preferred stock G prior to the visual display of probability information and 68% (134 respondents) preferred stock G after the probability was visually displayed.

Stock F

63% 37%

68% 32%

Word Problem

Visual

Type of Stock Senario Ques3on Answered

Fewer respondents changed their preference in this scenario. Thirty-nine respondents or 19.5% of the sample changed their stock preference from stock F ($1,500 loss potential and highest possible gain of $4,500) to stock G ($3,000 loss potential and highest possible gain of $3,000), whereas 15% (30 respondents) changed their preference from stock G to stock F, and 65% (129 respondents) did not change their preference. Table 2 shows the descriptive statistics for the each of the two mixed outcome investment scenarios.

Type of Stock Senario Ques3on Answered

Table 2 highlights that seventy-five percent of respondents who changed stock selection in the small dollar scenario switched from stock G (no loss potential and highest possible gain of $1,500) to stock F ($500 loss potential and highest possible gain of $1,500). Figure 2 shows the results for the the large-dollar return scenario. In

Table 2.

Dependent Variable Distribution Word Problem

Large $ Scenario Stock F Stock G # who changed stock selection Changed from F to G Changed from G to F

Stock G

Stock G

0%

Small $ Scenario Stock F Stock G # who changed stock selection Changed from F to G Changed from G to F

Stock F

77 121

73 125

Visual Display

# Meeting Condition

% Meeting Condition

99 24 75

50% 25% 75%

69 39 30

35% 56% 44%

128 70

64 134

46

Journal of Personal Finance

A 2 X 2 ANOVA analysis was conducted to determine if there was any order effect to seeing the word problem displayed first. The results of the ANOVA indicate the order of presentation was non-significant as well as the interaction between order of presentation and the stock selected, F(4,196)=1.38, p=.24. There was a highly significant effect, F(4,196)=28.26, p<.01, on the type of question (word vs visual) as previously discussed.

After the visual stock scenarios were shown and selections indicated, respondents were asked, “Briefly describe your decision making process for this stock selection.” Responses were recorded and then later grouped into five categories of similar responses: 75% chance of gain, highest dollar potential gain, smaller chance of a loss, smallest dollar loss potential, and other (no response or other response). In the hypothetical investment scenario, respondents could focus on the dollar amount or the probability of the outcome happening, and they could talk in terms of gains or losses. Therefore, four response choices were developed and any response falling outside of one of these four choices was included in a category called “other response” and are displayed in Table 3.

The qualitative data of stock selection rationale of the small-dollar-return scenario suggests 46% of the respondents indicated the decision of stock selection had to do with the potential gains offered, whereas 34% of the sample indicated potential losses were the reason behind their selection (refer to Table 3).

In the small-dollar-return investment scenario, those who chose stock F (25% chance of a $500 loss, $500 gain, $1,000 gain, or $2,000 gain) tended to focus on the potential gain, whereas those who chose stock G (50% chance of

either $0 loss, or $1,500 gain potential) tended to focus on the no loss outcome. As Table 3 shows, when combining categories, 44.0% of the respondents spoke in terms of percentage or probability of gain or loss, while 36% mentioned the absolute dollar value of gain or loss.

The qualitative data of the large-dollar-return scenario as displayed in Table 4 shows that 56% of the respondents indicated the rationale used to make the decision of stock selection had to do with the potential gains offered by the selection, whereas only 32% of the sample indicated potential losses were the reason behind their selection.

In the large-dollar-return investment scenario, those who chose stock F ($1,500 loss, $4,500 gain potential with a 50% probability of each) tended to focus on the potential loss, whereas those who chose stock G (25% chance of $3,000 loss, 75% chance of $3,000 gain) tended to focus on the probability of a gain outcome. As Table 4 shows, when combining categories, the majority of respondents (66%) spoke in terms of percentage or probability of gain or loss, while 22% mentioned the absolute dollar value of gain or loss.

Regression Analysis In addition to the ANOVA 2 x 2 analysis, logistic regression results for the small-dollar and large-dollar returns are presented to control for the possibility that the visual displays may influence variables or demographic groups in different ways. Table 5 displays the logistic regression results for the small-dollar-return investment scenario. In the regression, people who changed from stock G (50% chance of either $0 loss, or $1,500 gain potential) to stock F (25% chance of a $500 loss, $500 gain, $1,000 gain, or $2,000 gain) were compared with the rest of the sample.

Table 3.

Table 4.

Answers to Follow-up Question on Small Dollar Return Investment Scenario

Answers to Follow-up Question on Large Dollar Return Investment Scenario

N Potential Gains 75% Chance of Gain Highest $ Potential Gain Potential Losses Smaller % Chance of a Loss Smallest $ Loss Potential Other No Response or Refused Other Response Total

Percent

66 26

33% 13%

22 45

11% 23%

29 12 200

15% 6% 100%

Potential Gains 75% Chance of Gain Highest $ Potential Gain Potential Losses Smaller Chance of a Smallest $ Loss Loss Other Potential No Response or Refused Other Response Total

©2014, IARFC. All rights of reproduction in any form reserved.

N

Percent

89 21

45% 11%

41 21

21% 11%

22 6 200

11% 3% 100%

Volume 13, Issue 1

47

Seventy four respondents, or 37% of the sample, changed their investment preference from stock G to stock F. The logistic regression results indicated the significant variables were the probe question responses and financial risk tolerance scale. Those respondents who had lower financial risk tolerance scale scores were more likely to switch from stock G ($0 loss potential) to stock F ($500 loss potential).

In addition, those respondents who focused on gains, either dollar gain potential or percentage chance of a gain, were more likely to change from stock G (50% chance of gain) to stock F (75% chance of gain), with those highlighting 75% chance of gain as the major reason for changing being

almost eight times more likely to change from stock G to stock F. Table 6 displays the logistic regression results for the large-dollar-return investment scenario. In the large-dollarreturn investment scenario, 39 respondents, or 19.5% of the sample, changed their investment preference from stock F (50% chance of $1,500 loss, 50% chance of $4,500 gain) to stock G (25% chance of $3,000 loss, 75% chance of $3,000 gain). Of the respondents who changed their selection, 28 individuals or 72%, indicated the reason for making this choice was the greater percentage chance of experiencing a

Table 5.

Logistic Regression for Small $ Return Stock G to F Change Risk Tolerance Scale Probe Questions (vs Small $ Loss) 75% Chance of Gain Small % Chance Loss Other Answer Large $ Gain Potential Net Worth Education (vs High School or Less) Some College Associateâ&#x20AC;&#x2122;s Degree Bachelorâ&#x20AC;&#x2122;s Degree Some Graduate School Graduate Degree N Pseudo R2 ***p<.001, ** ***p<.001, **p<.01, p<.01,**p<.05 p<.05

Coefficient -.114

Odds Ratio .893

2.072 1.041 -.031 1.168 -.055

7.942 2.832 .969 3.215 .947

-.382 -.122 -.213 -.920 -1.075

.682 .885 .808 .398 .341

200 0.269

p-value .006** .000*** .077 .971 .037* .247 .603 .888 .769 .339 .145

48

Journal of Personal Finance

Table 6.

Logistic Regression for Large $ Return Stock F to G Change Risk Tolerance Scale Probe Questions (vs 75% Gain) Small % Chance Loss Other Answer Small $ Loss Potential Large $ Gain Potential Net Worth Education (vs Graduate Degree) High School or Less Some College Associateâ&#x20AC;&#x2122;s Degree Bachelorâ&#x20AC;&#x2122;s Degree Some Graduate School N Pseudo R2 ***p<.001, ** p<.01, * p<.05

Coefficient .028

Odds 1.028 Ratio

Sig. .540

-.559 .142 -20.282 -20.089 .040

.572 1.152 .000 .000 1.041

.261 .879 .998 .998 .487

-.655 -.122 -.191 .445 -19.388

.520 .886 .826 1.560 .000

.578 .832 .832 .371 .998

200 0.256

gain on their investment although the logistic regression results indicated no significant variables.

The visual display of probability information did not change the propensity for respondents to prefer the certain outcome (stock G) or the risky choice (stock F) in the prospect theory questions from Kahneman and Tversky (1979). Eighty-two percent of the sample preferred stock G (a sure gain of $3,000), whereas 75% of the sample preferred stock F (a 20% chance of no loss and an 80% chance of a $4,000 loss).

Discussion Two hypotheses were proposed in this study. The first hypothesis, that participants would be more likely to select the stock from each word format question that presents the lowest dollar value of a loss, is only partially supported in this research. In the word format small-dollar-returninvestment scenario, 60% of respondents preferred stock G (no chance of loss) after reading the word problem version of the investment scenario. On the other hand, in the word format large-dollar-return-investment scenario, 63% preferred stock G (25% chance of $3,000 loss) over stock F (50% chance of $1,500 loss). It is possible the larger dollar amount involved made respondents more sensitive to the chance of loss. Another possibility is the small-dollar-return question had a no loss option and respondents chose stock G due to a certainty effect.

The second hypothesis, that participants will be more likely to select the stock that presents the lowest probability of loss after seeing the visual display of probability, is partially supported in this research. In both the visual display of the small and large dollar-return investment scenarios, respondents had a clear preference for the stock with the least percentage chance of a loss, with 65% preferring stock F ($500 loss, $500 gain, $1,000 gain, $2,000 gain potential with a 25% probability of each) in the small-dollar-returninvestment scenario, and 68% preferring stock G (25% chance of $3,000 loss) in the large-dollar-return-investment scenario. In both of these scenarios the chance for experiencing a loss was 25%. However, when respondents were asked to explain their investment decision making process, the respondents indicated percentage chance of gain, not loss was their primary rationale. Rather than choosing the stock with the least chance of loss, participants consistently chose the stock with the highest chance of a gain. For example, in the large-dollar-return investment scenario, 45% of respondents indicated they preferred this stock due to the high chance of investment gains, whereas 21% of respondents spoke in terms of small chance of experiencing a loss.

Similarly, in the small-dollar-return investment scenario, 33% chose stock F ($500 loss, $500 gain, $1,000 gain, $2,000 gain potential with a 25% probability of each) indicated the rationale for their decision was based upon the

Š2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1 potential gain, whereas 11% chose stock G (50% chance of $0 loss, 50% chance of $1,500 gain) said they preferred it due to the no loss outcome. In this small-dollar-return scenario, regardless of the order in which questions were presented, more of the respondents (65%) preferred stock F when the probability information is visually displayed. This result suggests that visual display of probability of different investment outcomes (versus in word format) may mitigate the behavioral tendency of individuals to be more risk averse in the domain of gains than in the domain of losses as proposed by prospect theory. Since each outcome has the same expected value statistically speaking, prospect theory would have suggested a preference for stock G since it has no loss potential. However, respondents preferred stock F over stock G by almost a two to one margin. This could be explained by the fact that visual elements tend to draw more upon cognitive resources as suggested by the theory of limited capacity model of motivated message processing (LC4MP), thus increasing the likelihood that data is being more effectively processed and that households are more consistent in their aversion to gains and losses.

Respondents were also far more likely to change their stock selection in the small-dollar-return investment scenario with 50% changing their response from their first indicated preference. Of the respondents who changed their preference, 75% changed from stock G to stock F, and only 25% changed from stock F to stock G. Those who switched from stock G to stock F were more likely to have lower risk tolerance scale scores, indicating participants with a low risk tolerance score were willing to accept the risk of a $500 loss for an increased likelihood of a gain outcome. The respondents who changed their mind from stock G to F indicated the reason was due to the greater probability of gain associated with stock G, whereas those who changed their preference to stock F from G indicated the small dollar amount of loss as the reason for their decision. A $500 loss on a $10,000 investment over a one year period of time would equate to 5% loss, and this value may be close enough to zero for respondents to accept the risk of loss.

Anecdotally, several of the respondents indicated a $500 loss as being small enough to absorb for a 75% chance of a gain; however, it is impossible to determine the number of respondents who felt that way based on the data available.

The results have some stochastic dominance significance as well. Remember, for the large-dollar investment scenario, stock G dominates stock F in terms of Prospect Stochastic Dominace, which means investors who are loss averse should prefer stock G. The majority of investors (63%-68%) do prefer stock G regardless of how

49 the question was presented. This result was in contrast to the Levy and Levy (2002b) study in which 76% of the sample of practitioners, MBA students, and professors preferred stock F. In the small-dollar investment scenario, investors who are risk averse should prefer stock G, because stock G dominates stock F in terms of second-degrees stochastic dominance. In the word-problem version the majority of investors preferred stock G and would be considered risk averse. However, when shown the visual depiction of the same scenario, the majority of investors selected stock F, thus no longer remaining risk averse.

Implications Overall, participants tended to prefer the stock with the greatest percentage chance of gain when the probability was visually depicted. This tendency was more prevelant in the scenario with returns closer to zero. In a small-dollar-return scenario, when respondents had a choice between stock F with a 50% chance of no loss or gain and a 50% chance of a $1,500 gain or stock G with a 25% of a $500 loss, a 25% chance of a $500 gain, a 25% chance of a $1,000 gain, and a 25% chance of a $2,000 gain, respondents were strongly influenced by the addition of a visual display of probability to the question. In the word problem question format, 60% of respondents were selecting the option with no loss potential as suggested by prospect theory. When the probability was visually displayed, only 35% of respondents preferred this option. The research with the small-dollarreturn investment scenario suggests that the visual display of probabilities and outcomes in investment scenarios shift the preference towards the outcome which has the greater probability for a gain. Sixty-five percent of respondents were willing to take a small risk of $500 for a greater likelihood of a gain in their investment. The change in preference could be due the increased cognitive attention given to the probability information. The process of encoding the question into memory might have been facilitated by the use of a visual representation of the probability. Financial advisors and planners should consider adding or ensuring visual displays of information are provided in addition to written form since visual information utilizes different cognitive processes and can lead to reduction in cognitive biases.

This study therefore suggests that a financial planner simply not administer a financial risk tolerance profile questionnaire and assume the results from this profile will be able to help them predict the clientâ&#x20AC;&#x2122;s investment preferences.

This study indicates some investors with similar financial risk tolerance profile scores would be willing to take some risk of loss for potential investment gains, while others

50

Journal of Personal Finance

would prefer to have no loss exposure. Clarification of these types of investment scenarios is needed so the expectations between client and financial planner can be clearly established. Planners should be aware that some visual displays of investing scenarios can influence investors towards preferring a riskier stock selection.

recent events or events that are easier to recall have a stronger influence on decision-making than events that are harder to recall. This bias could have influenced the respondentâ&#x20AC;&#x2122;s willingness to take risk in hypothetical investing scenarios versus previous studies utilizing the same questions.

Limitations

It is possible some interviewer bias and social desirability bias have occurred with the study. In particular, the social desirability of being good with math for the subjective numeracy scale is a concern. The scale has a potential range of 8 to 48, but the sample results range was 25 to 48 with a mean score of 39.4. Perhaps the use of a numeracy scale that asked respondents to calculate answers to mathematical problems would have been a more accurate measure of actual numeracy skills.

A number of limitations with regard to the study need to be acknowledged. The sample size is small and fairly homogeneous in nature. Future studies would benefit from a larger sample size from a survey that includes respondents randomly chosen from more than one metropolitan area or a more focused field study. Using a large-sample survey might better support an investigation of a few key variables, such as visual or numeric display on investment selection, while a field study, although perhaps more complex or qualitative in nature, might support a broader investigation of how various characteristics of individuals within a focused subpopulation group (e.g., individuals with no stock ownership) relate to or moderate decision making (e.g., investment selection) (Schwab, 2004).

It is impossible to determine what affect the recent stock market returns and the economic recession has made on respondent decision-making. The suvey was conducted in October and November, 2010 a short time after the 2009 great recession in which the Dow dropped by roughly 50% to 7000 but had then rebounded back to 10,000. The availability bias (Thaler & Sunstein, 2008) indicates more

When respondents were given a visual representation of probability they tended to talk in terms of probability. Future studies could be improved by asking the same probing questions after the word problem format questions as well. A review of literature reveals a lack of qualitative studies on hypothetical investment scenarios. Researchers ask quantitative hypothetical questions and then go on to infer the respondent motivation that corresponds with their answers. A better way to understand respondent choice and decision-making may be to ask them to explain their decision-making process directly after the answer selection.

Š2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1

51

References Agnew, J. R., & Szykman, L. R. (2005). Asset allocation and information overload: The influence of information display, asset choice, and investor experience. The Journal of Behavioral Finance, 6 (2), 57-70.

Barsky, R. B., Juster, F. T., Kimball, M. S., & Shapiro, M. D. (1997). Preference parameters and behavioral heterogeneity: An experimental approach in the health and retirement study. Quarterly Journal of Economics, 112 (2), 537-579.

Benartzi, S., & Thaler, R. H. (2001). Naive diversification strategies in defined contribution saving plans. The American Economic Review, 91 (1), 79-98.

Bhandari, G., Hassanein, K., & Deaves, R. (2008). Debiasing investors with decision support systems: An experimental investigation. Decision Support Systems, 46, 399-410.

Fagerlin, A., Zikmund-Fisher, B. J., Ubel, P. A., Jankovic, A. J., Derry, H. A., & Smith, D. M. (2007). Measuring numeracy without a math test: Development of the subjective numeracy scale. Medical Decision Making, 27 (5), 672680.

Grable, J. E., & Lytton, R. H. (1999). Financial risk tolerance revisited: The development of a risk assessment instrument. Financial Services Review, 8 (3), 163-181.

Grable, J. E., & Roszkowski, M. J. (2008). The influence of mood on the willingness to take financial risks. Journal of Risk Research, 11 (7), 905-923.

Hanna, S. D., & Lindamood, S. (2004). An improved measure of risk aversion. Financial Counseling and Planning, 15 (2), 27-38.

Hanna, S. D., Gutter, M. S., & Fan, J. X. (2001). A measure of risk tolerance based on economic theory. Financial Counseling and Planning, 12 (2), 53-60.

Hosmer, D. W., & Lemeshow, S. (2000). Applied logistic regression. Hoboken, NJ: John Wiley & Sons Inc.

Â

Iyengar, S. S., Huberman, G., & Jiang, W. (2004). How much choice is too much? Contributions to 401(k) retirement plans. In Mitchell, Olivia, Utkus, & S. (Eds.), Pension Design and Structure: New Lessons from Behavioral Finance (pp. 83-96). Oxford, UK: Oxford University Press.

Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47 (2), 263291.

Lang, A. (2000). The information processing of mediated messages: A framework for communication research.

Journal of Communication, 50, 46-70.

Levy, H., & Levy, M. (2002a). Experimental test of the prospect theory value function: A stochastic dominance approach. Organizational Behavior and Human Decision Processes, 89 (2), 1058-1081.

Levy, M., & Levy, H. (2002b). Propsect theory: Much ado about nothing? Management Science, 48 (10), 1334-1349.

Levy, M., & Levy, H. (2001). Testing for risk aversion: A stochastic dominance approach. Economics Letters, 71 (2), 233-240.

Rudolph, S., Savikhin, A., & Ebert, D. S. (2009). FinVis: Applied visual analytics for personal financial planning.

IEEE Symposium on Visual Analytics Science and Technology (pp. 195-202). Atlantic City: IEEE. Schuchardt, J., Hanna, S. D., Hira, T. K., Lyons, A. C., Palmer, L., & Xiao, J. J. (2009). Financial literacy and educational research priorities. Journal of Financial Counseling and Planning, 20 (1), 84-94.

Thaler, R. H., & Sunstein, C. R. (2008). Nudge: Improving decisions about health, wealth, and happiness. New York: Penguin Books.

Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. Science, 185 (4157), 1124-1131.

Yang, Y. (2004). Characteristics of risk preferences: Revelations from Grable & Lytton's 13-item questionnaire.

Journal of Personal Finance, 3 (3), 20-40.

52

Journal of Personal Finance

CONSIDERATION OF RETIREMENT INCOME STAGES IN PLANNING FOR RETIREMENT

Kyoung Tae Kim, Ph.D. Candidate The Ohio State University Sherman D. Hanna, Ph.D. The Ohio State University Samuel Cheng-Chung Chen, Ph.D. Suntrust Mortgage

Previous retirement adequacy studies have ignored expected retirement income stages. Ignoring retirement income stages results in biased estimations of retirement adequacy. This study analyzes retirement income stage theoretically and then empirically. Based on the 1995 to 2007 Survey of Consumer Finances (SCF) datasets, about 73% of working households with the head and/or spouse/partner age 35-70 and working full-time will have more than one retirement income stage. When income stages are taken into account, the proportion of households with retirement adequacy ranges from 44% in 1995 to 58% in 2007. Ignoring retirement income stages results in adequacy proportions being 23 to 28 percentage points higher. Financial planners and researchers evaluating retirement adequacy should take retirement income stages into account.

Introduction The retirement of the baby boom generation, with Social Security eligibility starting in 2008, will increase stress on the Social Security system (Butrica, Iams & Smith, 2003). Although defined benefit pension plans provide guaranteed retirement income, increasingly employers have switched retirement plans to defined contribution pension plans, such as 401Ks and IRAs. Beyond the increasing responsibility of managing retirement accounts, retirement planning is becoming increasingly challenging due to increasing life expectancy (Kim, 2005). Extended life expectancy requires more financial resources for retirement.

Furthermore, Van Solinge and Henkens (2010) concluded that an individualâ&#x20AC;&#x2122;s subjective life expectancy influences the retirement decision making process. Because of the complexity of retirement planning, how to prepare for an

affordable retirement is an important contemporary issue for workers in the U.S. Over the last two decades, retirement adequacy has been an important issue for researchers. Hanna and Chen (2008) found that research on retirement adequacy of working households has produced a wide range of estimates, with estimates of the proportion of workers on track for an adequate retirement ranging from 31% to 80%. These studies have explored the development of retirement adequacy by analyzing adequacy indicators, such as replacement ratio (Palmer, 1992, 1994), prescribed savings rate (Moore & Mitchell, 1997; Mitchell & Moore, 1998), the ratio of retirement wealth to needs (Yuh, Hanna & Montalto, 1998), and financial failure rate (Ameriks, 2000, 2001). However, previous retirement adequacy studies have ignored expected retirement income stages. Studies that ignore retirement

Š2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1

53

income stages result in biased estimations of retirement adequacy. In this study, the concept of retirement income stages is analyzed theoretically and then empirically. A retirement income stage is defined as a period in which the projected number of retirement income sources is constant.

Retirement adequacy is defined as being able to maintain preretirement spending, which is estimated by using a version of Palmer’s (1992, 1994) required retirement ratio concept. The main purpose of this study is to explore how many retirement income stages are held by households and to analyze the effect of taking stages into account at retirement on projected retirement adequacy.

Literature Review Life Cycle Saving Model. A number of different consumption theories have been developed over the decades, and the most widely accepted is the life cycle hypothesis (Ando & Modigliani, 1963). This theory implies that households should plan to smooth consumption over the rest of life despite fluctuations in current income. Moreover, it posits that consumption and saving reflect an individual’s stage in the life cycle, which is generally proxied by age. For example, young households are expected to spend more than income since they typically have relatively low earnings and incur other higher expenses regarding education and housing. Households in the middle period of the life cycle begin to save for retirement and pay their debts. After retirement, dis-saving is expected to occur again. The plausibility of this assumption depends partly on whether households can approximate the complex calculations for retirement and other changes in income. Some households face even greater complexity, because retirement represents more than one stage in terms of inflation-adjusted income flows, but previous researchers have assumed only one stage and have not differentiated between households with one retirement stage and those expecting multiple stages.

Retirement Consumption. Several studies have analyzed retirement consumption based on actual spending patterns.

Bernheim, Skinner and Weinberg (2001) used changes in food expenditure to estimate retirement consumption from the Consumer Expenditure Survey (CES). The findings suggested that the average consumption dropped by 14% at retirement. Based on the 2001 Consumption and Activities Mail Survey (CAMS), Hurd and Rohwedder (2003) concluded that non-retired singles’ spending declined about 17%, while among couples the reduction average was about 12% between 1992 and 2000. Butrica, Goldwyn and Johnson (2005) found that the spending percentage of income increased with age. The median level of expenditures dropped by 21% between the youngest (53-64)

and oldest age groups (75 and over), whereas the median value of income before taxes declined by 49%. Aguiar and Hurst (2009) found that food expenditures fall during the retirement period. Between the ages of 60 and 68, consumption on total food, clothing, and transportation declines by 7%, 18%, and 15% respectively from the 19842003 Consumer Expenditure Survey (CES).

On the other hand, previous retirement adequacy studies have attempted to measure retirement consumption on the basis of pre-retirement income. Palmer (1992, 1994) suggested that the required replacement ratio is a proxy of retirement needs as long as the retirement income can cover needs. Yuh, Montalto, and Hanna (1998), Yuh, Hanna, et al.

(1998), Chen (2007), Munnell, Webb, and Golub-Sass (2007) have assessed retirement consumption by using the replacement ratio approach. Furthermore, most financial planning textbooks have followed the replacement ratio approach to project retirement consumption. For example, Tacchino and Littell (1999) assumed that 60-80% of current salary is appropriate for retirement needs projection, and Dalton et al. (2005) and Grable, Klock, and Lytton (2012) assumed 70 to 80% of current salary for retirement needs.

However, none of these authors mentioned any justification of the default proportion for retirement needs.

Retirement Adequacy. To determine retirement adequacy, Palmer (1992) and Palmer (1994) focused on required replacement ratios, while Moore and Mitchell (1997) and Mitchell and Moore (1998), focused on prescribed saving rates. The required replacement ratio assumes pre-retirement spending is a proxy for optimistic post-retirement spending, and post-retirement income should be able to maintain postretirement spending. The prescribed savings rate is a specific saving rate under which a household can maintain a preretirement living standard when they retire. Moore and Mitchell’s approach is to determine the prescribed rate of saving to meet the required replacement ratio as determined

by Palmer. Moore and Mitchell (1997) concluded only 31% of households were saving enough, assuming retirement at age 62, while 40% were saving enough if they retired at 65.

Yuh, Montalto, et al. (1998) defined retirement adequacy as when individuals are able to retire at their planned retirement age and can maintain their pre-retirement consumption level. They concluded that about 52% would have enough assets and income for retirement, assuming investment assets earned historical mean returns. However, based on pessimistic projection of investment returns, only 42% would be on track for retirement adequacy. Another indicator of retirement adequacy beyond the replacement rate, the prescribed savings rate and the net worth/needs

54

Journal of Personal Finance

ratio, is the failure rate of financial assets (i.e., running out of financial assets). Ameriks (2000) used the 1992 and 1995 SCF datasets and financial planning software to project the failure rate of financial assets, and found about 56% would have enough for an adequate retirement.

Scholz, Seshadri, and Khitatrakun (2006) proposed an augmented model of life cycle saving hypothesis, which is based on the dynamic optimal programming method. They constructed this model to find the optimal net worth in retirement by using the Health and Retirement Study (HRS) dataset waves 1-4, in which households were assumed to maximize their utility and were subject to restricted resources. They then compared the optimal net worth with the real net worth from the HRS, and found that about 80% of American households reach this optimal new worth level, and were defined as well-prepared for retirement.

Court, Farrell, and Forsyth (2007) analyzed retirement adequacy of baby boomers (born from 1946 to 1964). After they formally retired, 60% of boomers will need to work (following formal retirement) just to maintain 80% of their current consumption, and more than 40% (29 million) will be working at age 65. Munnell, et al. (2007) used the 2004 SCF and concluded that 43% of households will not be able to maintain their standard of living in retirement even if they retire at age 65.

Hurd and Rohwedder (2011) performed 100 simulations of consumption and wealth paths of a sample of 66-69 yearolds by using data from the Health and Retirement Study (HRS) and data from the 2001-2007 Consumption and Activities Mail Survey (CAMS). They concluded that 71% of persons in the target age group were adequately prepared for retirement, but there is substantial variation by observable characteristics: 80% of married persons were adequately prepared compared with just 55% of single persons.

Retirement income stage. A few studies have discussed retirement stages, although none have taken retirement stages into account in projecting retirement adequacy of current workers. Everett and Anthony (2002) discussed classifying retirement into stages for planning calculations, but did not attempt to empirically assess adequacy. No previous research on retirement adequacy except for Chen and Hanna (2005) and Chen (2007) has attempted to address the technical question of calculating retirement adequacy by accounting for retirement income stages.

We also reviewed six financial planning and personal finance textbooks (Tacchino & Littell, 1999; Dalton, 2004; Dalton et al., 2005; Mittra, Sahu, & Crane, 2007; Garman & Forgue 2012, Grable et al., 2012) and eight online retirement calculators provided by AARP, Fidelity, T. Rowe Price, SSCalc.net, FINRA, CNN money AOL and Vanguard (Appendix 1). Most financial planning textbooks and online calculators have used three common retirement resources, which are Social Security, pension plans, and personal saving. However, all of the textbooks and online calculators assumed constant retirement income over the retirement period.

Methodology Data and sample selection. In this study, datasets from the 1995, 1998, 2001, 2004 and 2007 SCF were used for empirical analyses. The most recent dataset, 2010 SCF is not included because this study focuses on retirement stage analysis during normal times. The 2010 survey covered an extended period of a severe economic recession, which was very different from the 1995 to 2007 period. Bricker, Kennickell, Moore, and Sabelhaus (2012) reported that both median and mean net worth fell dramatically over the period 2007 to 2010; 38.8% and 14.7%, respectively.

The sample includes households with the head or spouse/partner who is age 35 to 70, currently working full time and indicating planned retirement ages. The reason for including only those households with either a head or spouse currently working is that our purpose is to project future retirement adequacy, and respondents who are younger than 35 are more likely to have major changes in jobs or marital status, which would make projection of retirement adequacy less accurate. Therefore, the sample includes only those 35 years or older. Social Security benefits do not increase after age 70, and workers must withdraw minimum amounts from traditional Individual Retirement Accounts starting after age 70.5, so it makes sense to restrict our analyses to those under 70. Our sample restrictions are similar to Yuh, Montalto, et al. (1998). About 16% of the sample households responded that they will never retire. For those households, we assumed that their planned retirement age equals 70. Also, households who answered â&#x20AC;&#x2DC;never retiredâ&#x20AC;&#x2122; from part-time job were assumed to be retired from that job at age 75. There were also some technical restrictions related to plausible projections of life expectancy, resulting in a final total sample size of 8,435 (Table 1).

Š2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1 Retirement income stage. A retirement income stage is defined as a period in which the real income remains constant. However, some events might cause small income changes, for example, changes in income tax rules. In order to have a more meaningful and technically feasible definition, only projected changes in the number of income sources are considered. Therefore, the income stage is formally defined as a period in which the projected number of income sources is constant. Whenever the projected number of income sources changes, one new stage is created. The income stage starts with the planned retirement age and ends with death of the individual or couple. The maximum stage number for a married household is eight, based on the stage drivers of Social Security retirement benefits, Defined Benefit pensions, and part-time job wages.

Retirement Income. Our calculation of resources during retirement was somewhat similar to the methods reported by Chen (2007), with calculation of retirement income from projected retirement assets combined with estimated income from Social Security pensions, defined benefit pensions, and part-time wages. Investment asset accumulation was based on projecting current balances and the assumption that current contributions to retirement investments would remain constant in real terms until retirement (Moore & Mitchell, 1997; Mitchell & Moore 1998; Yuh, Hanna, et al, 1998; Yuh, Montalto, et al., 1998). The current allocation of retirement investments were computed based on stock, bond, and cash equivalent categories, along with investment real estate. Following Yuh, Hanna, et al. (1998), lognormal projection was used to estimate the future value of the balances and of the contributions based on the long term inflation-adjusted mean and variance of each investment category at the time of the survey (e.g., Ibbotson Associates, 2008). Total income was estimated for each stage by adding continuing income such as Social Security pensions to annuity income from the projected future balances of retirement investments. To calculate feasible income for multiple stages an iterative process was used, and to simplify the estimation for households with more than two stages, we combined all stages but the last stage into one new stage (for details, see Chen, 2007).

The Survey of Consumer Finances datasets do not include total spending, so we estimated spending for each household using methods similar to those used by Chen (2007) and Palmer (1992; 1994). We estimated spending as a percent of pretax income averages for 12 income categories from the Bureau of Labor Statistics 2007 Consumer Expenditure Survey published results (U. S. Department of Labor, 2009). The highest income category in

55 the published tables was $150,000 and above, so to estimate spending for households with income above $150,000 we projected amounts above the published income categories using a power function estimation from the lower income categories.

Results Stage Effect. About 27% of households have one stage, 35% have two stages, and almost 39% have more than two stages (Table 2). Quite a few households (5%) have more than 5 stages. Single heads planning to retire at age 62 or later, and couple households where both plan to retire at age 62 or later in the same calendar year and have only one source of retirement income other than pensions, would typically have only one retirement income stage. Our sample characteristics of households are similar to the result reported by Kim, Chen, and Hanna (2012), as 33% plan to retire before age 62 and almost 64% plan to have a part-time job after retirement from full-time work. Both of these patterns would result in multiple retirement income stages.

Over 20% of households have defined benefit pension plans.

These findings confirm that many households have multiple retirement income stage drivers: planned retirement before age 62, defined benefit pensions, planning for part-time jobs, and couples with planned retirement ages in different calendar years.

Table 3 shows the median length of household stages by the total number of stages. The median length of New Stage 1 is the sum of the lengths of the first N-1 old stages. The length of New Stage 2 is the difference between remaining life expectancy and the length of New Stage 1. The total length of New Stage 1 and New Stage 2 is equal to the remaining life expectancy at retirement. For example, consider a household with three stages. The length of Stage 1 is 3 years, Stage 2 is 4 years and life expectancy is 30 years. To make our calculations feasible for statistical analyses, the Old Stage 1 and Old Stage 2 are combined into New Stage 1. The Old Stage 3 would be New Stage 2.

Therefore, the length of New Stage 1 is 7 years (3+4 years).

The length of New Stage 2 is 23 years (30-7years).

The first retirement age is defined as headâ&#x20AC;&#x2122;s age when the first retirement happens. For couples, first retirement is calculated by comparing headâ&#x20AC;&#x2122;s planned retirement age and spouseâ&#x20AC;&#x2122;s planned retirement age. The earlier the age of first retirement, the more likely a household is to have multiple stages, as shown in Table 3. For example, households with four stages have a median first retirement age of 56, compared to the median first retirement age of 65 for households with only one stage. In general, if the first

56

Journal of Personal Finance

retirement age is before 62 there will be more than one retirement stage because the household will have to wait until the head turns 62 to start receiving the Social Security pension. The last stage length decreases as the households have more stages. For example, households with three stages have a last stage length of 15.7 years, while households with seven stages have a last stage length of 10.3 years. The reason for this phenomenon is because first N-1 stages already count part of the life expectancy, so that the last stage length decreases.

The length of the first stage is generally longer than that of the middle stages. Households with two stages have a median first retirement age of 65, suggesting that a second stage is not generated only by retirement before the minimum Social Security age of 62. It is often generated by having a defined benefit pension and a part-time job. In contrast, households with more than two stages have a median first retirement age before age 62 and it is likely that the first stage results from retirement before the minimum Social Security age of 62. This can be verified by adding the length of first stage to the first retirement age which results in an age close to 62. For example, in households with four stages, the length of first stage is 7 years, and the first retirement age is 56. The sum of 7 and 56 is 63, which is very close to the minimum Social Security age of 62. Since the first stage usually happens before age 62, and middle stages usually happen between 62 and 65, the length of the first stage is generally greater than that of the middle stages.

Replacement Ratio. Category 1 includes households with two New Stages and with projected retirement assets high enough to allow for equal spending in New Stage 1 income and New Stage 2. Category 2 includes households with two New Stages but lacking sufficient projected retirement assets to have equal spending in the two stages.

Category 3 includes households with only one stage. In this study, we assess the projected retirement adequacy of U.S households.

As Table 4 shows, the median replacement ratio ranges from 53% to 120% across categories and new stages. If stage partitions are ignored, the replacement ratio ranges from 53% to 265%. As expected, the replacement ratio ignoring income retirement stage partitions is much higher than with stage partition, because the method ignoring stage partitions recognizes all retirement income at the first planned retirement age. In contrast, the method with income stage partitions recognizes retirement income when income really occurs. The overestimation of the replacement ratio is

substantial for households in Category 1. Comparing the replacement ratio in Category 1 between stage partitions and ignoring the stage partition, the average overestimation of replacement is around 130%. Within each survey year, the replacement ratio is highest in Category 1, but is lowest in Category 3. This is because households in Category 3 are less likely to have defined benefit pensions and part-time jobs.

Table 5 shows mean retirement adequacy proportions, in other words, the proportions of households estimated to be able to maintain preretirement spending. The mean retirement adequacy proportion ranges from 26% to 83% across categories and new stages. Ignoring stage partitions, the proportion ranges from 26% to 98%. It is as expected that the retirement adequacy proportion ignoring stage partition is higher than that with stage partition because the median replacement ratio of ignoring stage partition is higher. The across categories combined results are presented in Table 6. Under the stage partition method, the average adequacy proportion steadily increased from 1995 to 2004, with 2007 about the same as 2004. Ignoring stage partitions, the adequacy proportion has a pattern similar to that with stage partitions, except that the proportions are overestimated by 23% to 28%.

Conclusion and Implication The main focus of this study is the effect of taking income stages into account at retirement. About 73% of households have more than one stage, and only 27% of households have one stage. Planned first retirement age is negatively related to the number of stages. For example, when the first retirement age is before 62, there will be more than one retirement stage because the household will have to wait until the head turns 62 to start receiving the Social Security retirement benefit. The length of last stage is the longest, since it is the period between the age of income leveling off and the age of death. Furthermore, the last stage length decreases as the households have more stages. About 26% of households have a defined contribution pension and about 20% of households have a defined benefit pension.

About 40% of households plan to retire between the ages of 62 and 65. About 64% of households plan to have a parttime job after retirement from full-time employment.

By comparing mean income replacement ratios with benchmark ratios, mean retirement adequacy is calculated.

The overall adequacy, taking retirement income stages into account, steadily increased in each survey period from 1995

Š2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1 to 2007. The combined stage adequacy proportion ranges from 44% to 58%. When ignoring stage partitions, the adequacy proportion ranges from 71% to 82%. Ignoring stage partition overestimates the adequacy proportion by 28 percentage points in 1995 and 1998, 23 percentage points in 2001 and 2007, and 25 percentage points in 2004. The result that accounting for retirement income stages gives much lower levels of retirement adequacy than the result that ignored retirement income stages is an extremely important result, and all future researchers on retirement adequacy need to carefully consider retirement income stages.

This study has implications for both financial planners and researchers. The research confirms that retirement income stages should be considered in householdsâ&#x20AC;&#x2122; planning for retirement. Retirement income stages represent multiple income cash flows. In other words, a household with more retirement income stages has more income cash flows. Due to the complexity of cash flow management, discrepancies may exist in forecasting the timing and amount of future cash flow. Such discrepancies could result in financial planning failures. Therefore, better understanding of retirement income stage analysis by financial planners is necessary to reduce the risk of such financial planning failures for their clients. Furthermore, it may encourage households to create a financial portfolio that will generate greater retirement income.

57 The method we used in calculating retirement adequacy was simplified in order to analyze large national datasets (1995-2007 SCF datasets), but the process that financial planners can use is similar to the usual capital needs analysis, except for consideration of retirement income stages. Appendix 2 shows the process. The basic calculations assume that a life annuity would be purchased, and therefore the remaining life expectancy needs to be estimated or, for a couple, the joint life expectancy.

However, if the household has more than one retirement income stage the remaining life expectancy should be estimated as of the beginning of the last stage of retirement.

Our analysis has an important implication for developing retirement adequacy measurement. Previous retirement adequacy studies have used different measurements of adequacy and datasets as discussed above, with estimates of the proportion of workers on track to achieve retirement adequacy ranging from 39% to 80%. We discussed the effect of having more than one planned retirement income stage, first analyzed theoretically and then empirically. Since accounting for retirement income stages gives much lower levels of retirement adequacy than estimates ignoring these stages, future retirement adequacy researcher needs to consider their possible impact on household preparedness for retirement.

58

Journal of Personal Finance

References Aguiar, M., Hurst, E. (2009). Deconstructing lifecycle expenditure, University of Rochester, manuscript Ameriks, J. (2000). Using retirement planning software to assess Americans' preparedness for retirement: An Update. Benefits Quarterly, Fourth Quarter, 37-51.

Ameriks, J. (2001). Assessing retirement preparedness with planning software: 1998 update. Benefit Quarterly, Fourth Quarter, 44-53.

Ando, A., & Modigliani, F. (1963). The life cycle hypothesis of saving: Aggregate implications and tests. American Economic Review, vol. 53, 55-84.

Bernheim, D., Skinner, J. & Weinberg, S. (2001). What Accounts for the Variation in Retirement Wealth Among U.S. Households? American Economic Review, 91(4), 832 – 857.

Butrica, B. A., Iams, H. M., & Smith, K. E. (2003). It’s all relative: understanding the retirement prospect of babyboomers. Working Paper 2003-21 of Center for Retirement Research at Boston College (CRRP).

Butrica, B. A., Goldwyn, J. H., & Johnson, R. W. (2005).

Understanding expenditure patterns in retirement, The Urban Institute.

Bricker, J., Kennickell, A. B., Moore, K. B., & Sabelhaus, John. (2012). Changes in U.S. Family Finances from 2007 to 2010: Evidence from the Survey of Consumer Finances. Federal Reserve Bulletin 98(2): 1-80.

Chen, C. C. & Hanna, S. D. (2005). Change in retirement adequacy, 1995-2001: Accounting for stages of retirement. Consumer Interests Annual, 52.

Chen, C. C. (2007). Changes in retirement adequacy, 19952004: Accounting for retirement stages. Dissertation, The Ohio State University.

Court, D., Farrell, D., & Forsyth, J.E. (2007). Serving aging Baby Boomers. The McKinsey Quarterly. 4, 103-113.

Dalton, M. (2004). Chapter 2: Introduction to Retirement Funding. Retirement Planning and Employee Benefit (pp. 9-45). Kenner, LA: Your Money Education Resource.

Dalton, M., Dalton J., Gangelosi R., Guttery R., & Wasserman, S. (2005). Chapter 15: Introduction to Retirement Planning. Personal Financial Planning: Theory and Practice (pp. 667-694). St. Rose, LA: Kaplan.

Everett, M. D. & Anthony, M. S. (2002). A stages model for planning retirement income distribution. Journal of Financial Planning, 6, 42-51.

Garman, E. T. & Forgue, R. E. (2012). Chapter 17: Retirement Planning. Personal Finance (pp. 510545). Mason, OH: Southwestern Cengage Learning.

Grable, J. E., Klock, D. D., & Lytton, R. H., (2012). Chapter 12: Retirement Planning. The Case Approach to Financial Planning (pp. 565-629). Erlanger, KY: National Underwriter Academic Series.

Hanna, S. D. & Chen, C. C. (2008). Retirement savings. in J. Xiao, Handbook of Consumer Finance Research, Springer Publishing, 35-46.

Hurd, M. D. & Rohwedder, S. (2011). Economic preparation for retirement. NBER, Working Paper 17203.

Hurd, M. D. & Rohwedder, S. (2003). The retirementconsumption puzzle: anticipated and actual declines in spending at retirement. NBER, Working Paper 9586.

Ibbotson Associates. (2008). Stocks, bonds, bills, and inflation yearbook. Chicago, IL: Ibbotson Associates.

Kim, H., & DeVaney, S. A. (2005). The selection of partial or full retirement by older workers. Journal of Family and Economic Issues, 26, 371–394.

Kim, K. T., Chen, C. C. & Hanna, S. D. (2012). Does greater complexity reduce retirement adequacy? evidence from the Survey of Consumer Finances, 1995-2007, Consumer Interests Annual, 58.

Mitchell, O. S., & Moore, J. F. (1998). Can Americans afford to retire? New evidence on retirement saving adequacy. The Journal of Risk and Insurance, 65(3), 371-400.

Mittra, S., Sahu, A. P., & Crane, R. A. (2007). Chapter 1011. Practicing Financial Planning for Professionals.

Holly, MI: Rochester Hills Publishing.

Moore, J. F. & Mitchell, O. S. (1997). Projected retirement wealth and savings adequacy in the Health and Retirement Study. NBER Working Paper 6240.

Munnell, A. H., Webb, A. & Golub-Sass. F. (2007). Is there really a retirement saving crisis? An NRRI Analysis.

Center for Retirement Research, Boston College, Chestnut Hill, MA. Palmer, B. A. (1992). Establishing retirement income objectives: the 1991 retire project. Benefits Quarterly, Third Quarter, 6-15.

Palmer, B. A.(1994). Retirement income replacement ratios: An update. Benefits Quarterly, Second Quarter, 59-75.

Scholz, J., Seshadri, A., & Khitatrakun, S. (2006). Are Americans saving “optimally” for retirement? Journal of Political Economy, 114, 607–643.

Tacchino, K. B., & Littell. D. A. (1999). Chapter 19-23.

Planning for Retirement Needs (pp. 331-437). Bryn Mawr, PA: The American College.

U.S. Department of Labor (2009). Consumer Expenditure Report: Consumer expenditures in 2007. Retrieved from http://www.bls.gov/cex/csxann07.pdf.

©2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1 Van Solinge, H., & Henkens, K. (2010). Living longer, working longer? The impact of subjective life expectancy on retirement intentions and behaviour.

European Journal of Public Health, 20(1), 47-51.

59 Yuh, Y., Hanna, S. D., & Montalto, C. P. (1998). Mean and pessimistic projections of retirement adequacy.

Financial Services Review, 9(3), 175- 193.

Yuh, Y., Montalto, C. P., & Hanna S. D. (1998). Are Americans prepared for retirement? Financial Counseling and Planning, 9(1), 1-12.

60

Journal of Personal Finance

Table 1 - Sample Size by Survey Year SCF Survey Year 1995

Sample size with restriction

Sample size without restriction

1,453

4,299

1998

1,605

4,305

2001

1,675

4,442

2004

1,903

4,519

2007

1,799

4,418

Total

8,435

21,983

Note: Restrictions are described in the Methodology Section, and include head being 35 or older, but no more than 70 and being in the labor force.

Table 2 - Distribution of Number of Retirement Income Stages Stages

Percentage

Cumulative Percentage

1

26.52

26.52

2

34.63

61.15

3

21.65

82.80

4

11.87

94.67

5

4.71

99.38

6

0.60

99.98

7

0.02

100.00

Table 3 â&#x20AC;&#x201C; Median Period Length of Old Stage and New Stage Median Period Length subgroup by stages (SCF 1995-2007) p1

p2

p3

p4

p5

Median New Stage Length p6

1 stage

21.71

2 stages

6

12.87

3 stages

9

2

15.68

4 stages

7

5

2

14.16

5 stages

6

3

4

2

12.25

6 stages

4

2

3

2

2

12.01

7 stages

2

2

1

2

1

2

P7

10.26

First Retirement Age

newN1

newN2

24.32

N/A

65

6

12.87

65

11

15.68

60

16

14.16

56

19

12.25

55

19

12.01

55

16

10.26

56

Note: This table is a summary of aggregate data result

Š2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1

61

Table 4 - Median Income Replacement Ratios by SCF 1995-2007

With Stage Partition Category1 Category2 Category3 No Stage Partition

1995

1995

1998

1998

2001

2001

2004

2004

2007 New Stage 1

2007 New Stage 2

New Stage 1

New Stage 2

New Stage 1

New Stage 2

New Stage 1

New Stage 2

New Stage 1

New Stage 2

109% 78%

109% 75%

111% 93%

112% 81%

118% 93%

120% 86%

119% 85%

119% 80%

115% 88%

116% 82%

67%

N/A

53%

N/A

66%

N/A

64%

N/A

62%

N/A

1995

1998

2001

2004

2007

Category1

230%

240%

265%

248%

238%

Category2

181%

192%

195%

196%

200%

Category3

67%

53%

66%

64%

62%

Category 1: Households have two New Stages. New stage 1 spending and New Stage 2 spending can be equalized by accumulated retirement assets.

Category 2: Households have two New Stages. New stage 1 spending and New Stage 2 spending cannot be equalized by accumulated retirement assets.

Category 3: Households have only one stage.

62

Journal of Personal Finance

Table 5 - Mean Retirement Adequacy Proportion by SCF 1995-2007 (Compared to Benchmark Replacement Ratio) 1995

1995

1998

1998

New Stage 2

71%

2001 New Stage 1 79%

With Stage Partition

New Stage 1

New Stage 2

New Stage 1

New Stage 2

Category1

67%

67%

71%

Category2

29%

35%

Category3 Overall proportion, accounting for stages No Stage Partition

32%

N/A

2001

New Stage 2

79%

2004 New Stage 1 82%

37%

33%

26%

N/A

2004

82%

2007 New Stage 1 83%

2007 New Stage 2 83%

46%

43%

44%

43%

50%

42%

33%

N/A

35%

N/A

38%

N/A

44%

47%

55%

57%

58%

1995

1998

2001

2004

2007

Category1

96%

96%

98%

96%

97%

Category2

83%

89%

93%

94%

96%

Category3 Overall proportion, ignoring stages

32%

26%

33%

35%

38%

71%

76%

78%

82%

81%

Category 1: Households have two New Stages. New stage 1 spending and New Stage 2 spending can be equalized by accumulated retirement assets.

Category 2: Households have two New Stages. New stage 1 spending and New Stage 2 spending cannot be equalized by accumulated retirement assets.

Category 3: Households have only one stage.

Table 6 - Retirement Adequacy Difference between Stage Partition Method and Non-Stage Partition Method 1995

1998

2001

2004

2007

Stage 1

43%

48%

55%

57%

59%

Stage 2

45%

47%

54%

57%

56%

Average of S1 & S2

44%

47%

55%

57%

58%

Non-stage partition Difference between adequacy rates ignoring stage partition and counting stage partition

71%

76%

78%

82%

81%

28%

28%

23%

25%

23%

Š2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1 Appendix 1

63 Online Retirement Planning Programs

To select retirement planning calculators, we used search engines such as Google and Yahoo. For each of the eight retirement calculators listed below, we attempted to identify the programâ&#x20AC;&#x2122;s assumptions about retirement income stages based on different scenarios such as different retirement income sources, different retirement ages of a couple household and so forth.

For the purpose of this study, we did not attempt to assess assumptions in the retirement calculators other than the three parameter values.

1. AARP Retirement Planning Calculator http://www.aarp.org/work/retirement-planning/retirement_calculator/ 2. Fidelity Retirement Quick Check http://personal.fidelity.com/retirement/retirement_frame. shtml.cvsr 3. T. Rowe Price Retirement Income Calculator http://www3.troweprice.com/ric/ric/public/ric.do 4. SSCalc.net Social Security Calculator http://www.sscalc.net/sscalc%20c.php5 5. FINRA (Financial Industry Regulatory Authority) Retirement Calculator http://apps.finra.org/investor_Information/Calculators/1/RetirementCalc.aspx 6. CNN money. http://cgi.money.cnn.com/tools/retirementplanner/retirementplanner.jsp 7. AOL Retirement Planning Calculator http://www.walletpop.com/calculators/retirement 8. Vanguard Retirement Income Calculator https://retirementplans.vanguard.com/VGApp/pe/pubeducation/calculators/RetirementIncomeCalc.jsf

64

Journal of Personal Finance

Appendix 2

Calculating the Capital Needs for Retirement with Multiple Retirement Income Stages

A retirement income stage is defined as a period when non-investment income does not substantially change in inflationadjusted terms.

General example with two stages: Y1 = non-investment income in Stage 1, with length = N1 Y2 = non-investment income in Stage 2, with length = N2 C=desired spending per year, assumed to be constant over retirement.

Annual gap in Stage 1 = G1 = C - Y1 In this example, we will ignore income taxes on amounts withdrawn from retirement investments, but this should be considered in calculations.

Assume that an inflation-adjusted interest rate is used – this should be the aftertax inflation-adjusted return for annuities if an annuity is purchased. We calculate present value with annual payments at the beginning of each year.

Capital needs to fund Stage 1 = PV(R, N1, G1). In the example below, annual payments at the beginning of each year are assumed, but this is arbitrary.

Capital needs to fund Stage 2, as of the beginning of Stage 2 = PV(R, N2, G2 ).

Capital needs to fund Stage 2, as of the beginning of Stage 1 = PV(R, N2, G2 ) / [(1+R) N1 ] Example: A worker retires at age 55, has no non-investment income until 7 years later, when he starts to collect a Social Security pension. His desired spending per year is $50,000 in constant dollars. His Social Security pension will be $20,000 per year.

(Assuming real rate of return of 3%, remaining life expectancy = 25 years, ignoring income taxes) Retirement Income Stage 1(55-61, N1 = 7 years): Annual gap = desired spending – aftertax non-investment income = 50,000 - 0 = 50,000 Capital needs to fund Stage 1 = PV (3%, 7, 50000) = $320,860 Retirement Income Stage 2(62-86, N2 = 25 years): Annual gap = desired spending – aftertax non-investment income = 50,000 – 20,000 = 30,000 Capital needs to fund Stage 2 (as of the beginning of Stage 2) = PV(3%, 25, 30000) = $538,066 Capital needs to fund Stage 2 (as of the beginning of Stage 1) = PV(3%, 25, 30000) / [(1+0.03)7] = $437,497 Therefore, total capital needs to fund Stage 1&2 = $758,357 Note that if retirement income stages are ignored, we would need to either assume that the Social Security pension starts at age 55, which would underestimate the capital needs, or somehow address the needs to fund Stage 1.

©2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1

65

POST-RECESSION, POST-LEGISLATION CREDIT USE: INSIGHTS FROM AN ONLINE SURVEY Barbara O’Neill, Ph.D., Extension Specialist in Financial Resource Management, Rutgers Cooperative Extension Jing Jian Xiao, Ph.D., Professor, Department of Human Development and Family Studies, University of Rhode Island

This article describes findings from data collected in 2009-2013 from a sample of 1,081 U.S. respondents to a 20question online self-assessment tool called the Wise Credit Management Quiz. The quiz incorporates frequently cited expert recommendations for credit use. Key provisions of recent federal credit legislation are also described. This study was conducted to investigate post-recession, post-legislation credit management practices of Americans in an effort to inform financial education and counseling efforts. The average quiz score for checking credit scores was higher than that for checking credit reports even though free credit scores are not required by law and free credit reports are. The least frequently performed practices were comparing at least three lenders before applying for credit and taking advantage of “float” time between when purchases are made and when credit card payments are due. The highest credit management quiz scores were associated with respondents who were older and those who had higher incomes and educational levels.

Introduction It is fair to say that much has happened during the past decade to affect the way that people handle credit. In addition to passage and implementation of two major federal credit reform laws, the Fair and Accurate Credit Transactions Act (FACTA) and the Credit Card Accountability and Disclosure (CARD) Act, the country experienced a severe financial crisis. The financial crisis led to tightened credit standards (e.g., higher FICO credit scores required for the best credit terms), high unemployment, stagnant incomes, decreased home values, an increased number of foreclosures and “underwater” homeowners with mortgages greater than the value of their homes, and the Great Recession that officially ran from December 2007 to June 2009 (U.S. Business Cycle, n.d.). Consumer Reports, noting the severity of the Great Recession, predicted “The recession’s effects will linger as consumers continue to get squeezed in a credit crunch more severe than we’ve seen in years” (The New Credit Crunch, 2009, p. 15).

With all these recent major changes, it is important to assess how consumers are handling credit in the postrecession, post-legislation 2010s decade. Knowledge of the strengths and weaknesses of household financial management can inform research, public policy, financial education programs and publications, and credit counseling

interventions. For example, it is helpful to know if people are requesting free credit reports, since there are no cost barriers (except, perhaps, the cost of a postage stamp for mailed credit report requests) due to the FACT Act, and what percentage of credit users avoid “borrowing themselves out of debt” with high-cost predatory loans and credit card advances. In addition, are there reported signs of financial distress such high consumer debt-to-income ratios and late payments, or is there evidence of wise credit management practices such as shopping around for credit, avoiding highcost credit card insurance, and paying credit card bills in full to avoid interest charges?

This article describes findings of a study of credit and debt management practices using data collected from a sample of 1,081 U.S. respondents to a 20-question online self-assessment tool called the Wise Credit Management Quiz. The study was conducted to investigate the postrecession, post-legislation credit practices of Americans.

Data were collected after the Great Recession ended between November 5, 2009 (inception of the online quiz) and June 30, 2013, a 44-month post-recessionary period during a fragile and prolonged U.S. economic recovery.

Thus, this study adds to the growing body of literature about how households manage their finances and use credit in tough economic times.

66

Journal of Personal Finance

While the results are not widely generalizable, they nevertheless provide valuable insights into the strengths and weaknesses of Americans’ debt portfolios and whether or not they are following expert advice such as keeping credit card debt utilization ratios low and avoiding co-signing (Bucci, 2011). Some quiz items are behaviors that people can initiate themselves (e.g., requesting a free credit report and checking their credit score). Others may have been beyond their control as a result of the Great Recession (e.g., being “upside down” on a loan such as a mortgage).

Review of Literature This literature review examines the aftermath of financial crises, the financial status and practices of U.S. households in recent years, and studies about credit card knowledge and behavior. Most of the research studies reported below collected data during the same approximate time period as the current study.

Aftermath of Financial Crises No recession since the Great Depression was deeper or longer than the most recent one (Cronin, 2011). In June 2009, the Great Recession of 2007-2009 officially ended but its effects continued to linger. A post-financial crisis economic meltdown is nothing new, however. Reinhart and Rogoff (2009) studied the aftermath of past financial crises, including the Great Depression and 1973 oil crisis, and found they are typically marked by four events: deep and prolonged asset market collapses (e.g., housing market), profound declines in economic growth, high unemployment in both the public and private sector, and an explosion in government debt as tax revenues decline (Mian, Sufi, & Trebbi, 2013; Ritholtz, 2009).

In most cases, a decade ensues where unemployment remains higher and home prices remain lower than they were at pre-crisis levels. Large numbers of people also repay their debt, a process called deleveraging (Reinhart & Rogoff, 2009). Unfortunately when consumers, businesses, and government all deleverage at the same time and incomes are stagnant, the inevitable outcome is slow economic growth and higher unemployment as the only way to cut debt is through reduced spending (Collingwood, 2012). While precautionary savings and debt reduction are often excellent strategies for individual households, they can worsen a recession when done on a large scale because they can cause collective harm. This phenomenon has been referred to as the “paradox of thrift” (Paradox of Thrift, 2013).

Financial Status and Practices of U.S. Households Several recent studies have investigated the impact of the financial crisis and Great Recession on the financial management practices of U.S. households. One of the most comprehensive is a study of financial crisis impacts between January 2008 and June 2010 by Pew Research Center with 2,967 respondents (Taylor et al., 2010). The study found that 62% of Americans reduced their spending since the recession began in December 2007 and 48% said they were in worse financial shape than before the recession. Four in ten respondents with savings or a retirement account made withdrawals to pay their bills and 71% said they have bought less expensive brands to help make ends meet.

Another study (Financial Capability, 2009), conducted after the financial crisis began by the FINRA Investor Education Foundation found that nearly half of 1,488 respondents reported difficulty paying monthly expenses and bills and a majority did not have funds set aside for emergencies or savings for predictable life events (e.g., retirement). Only 49% of respondents reported that they had set aside funds sufficient to cover expenses for three months.

Four in ten (41%) paid their credit card bills in full and nearly one quarter (23%) of the sample engaged in high-cost “alternative” forms of borrowing (e.g., payday loans, tax refund loans, pawn shops, car title loans, and rent-to-own stores).

In 2012, the FINRA National Financial Capability Study (Financial Capability, 2013) was replicated with a national sample of 25,509 adults. There were two slight improvements over 2009: more respondents reported having an emergency fund and fewer reported difficulty making ends meet. Still, more than half (56%) of respondents did not have a three-month emergency reserve and 42% and 16%, respectively, found it somewhat difficult and very difficult to pay all their bills. Almost half of respondents (49%) paid their credit card bills in full and nearly a third (30%) used at least one alternative borrowing method.

The 2009 MetLife Study of the American Dream (2009) was also conducted during the height of the financial crisis and included 2,243 online responses. Respondents reported paring spending (e.g., restaurant meals and vacations) and half of the sample said they could only meet their financial obligations for one month if they lost their job. In a survey conducted two years later (the 2011 MetLife Study), 42% of 2,420 respondents said they could not cover their financial obligations for more than a month if they were unemployed.

©2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1 A marked finding from this study was that Americans were replacing traditional notions of the “American Dream” such as material possessions, homeownership, and a high net worth, with metrics based on their personal values.

Another analysis of the effects of the financial crisis was a 2009 Federal Reserve Board study (Bricker, Bucks, Kennickell, Mach, & Moore, 2011) of families that participated in the 2007 Survey of Consumer Finances. The 2007 data were collected just as the economy started to turn down and 2009 re-interviews took place between July 2009 and January 2010. Information from the two years provides a unique measure of how families were affected by the financial crisis. The study found that families took on more debt and made less money. Median total debt rose from $70,300 to $75,600 and median household income dropped from $49,800 to $50,100 (Bricker et al.). Most families (63%) experienced wealth losses and the median percentage change was 18% of 2007 wealth.

In a follow-up paper that included data from both the 2007 and 2010 Survey of Consumer Finances (Bricker, Kennickell, Moore, & Sabelhous, 2012), Federal Reserve researchers found that decreases in family income were substantially smaller than declines in median and mean net worth (which fell 38.8% and 14.7% respectively) between the two surveys, driven most strongly by decreased housing prices. Median U.S. family net worth was $126,400 in 2007 and $77,300 in 2010. The overall value of families’ debt decreased slightly between the two waves of SCF data collection but their leverage ratio (ratio of the sum of debt to the sum of assets) rose from 14.8% in 2007 to 16.4% in 2010 due to asset value declines.

O’Neill and Xiao (2012) examined performance of 20 recommended financial practices using a six-year online survey data set through December 2010. Among the 12 significant time-period differences in a positive direction was more frequent payment of credit card bills in full to avoid interest charges. One possible explanation is that people were charging less per month in the post-recession period making it easier to quickly repay a small balance versus a larger one. Another recent study by the John J. Heldrich Center for Workforce Development at Rutgers University (Szeltner, Van Horn, & Zukin, 2013; Van Horn, 2013), with a nationally representative sample of 1,090 American workers, found that 73% of Americans either lost a job themselves or knew someone who had in the past four years. As a result, a majority (56%) of Americans reported having less money in savings than before the recession began. Nearly half withdrew funds from savings to make

67 ends meet and three in ten borrowed money from family or friends.

Credit Card Knowledge and Behavior Since the current study examines credit management practices in the aftermath of the Great Recession and recent legislation, it is instructive to examine recent studies about credit knowledge and practices. Lyons, Rachlis, and Scherpf (2007) studied consumers’ credit knowledge and concluded that most consumers generally understand the basics of credit reporting and their right to dispute errors. They also identified knowledge gaps, such as how to dispute errors and behaviors that impact credit scores, and recommended educational programs to address them, especially for individuals with low credit knowledge. The study employed a 58-question instrument with 23 credit knowledge questions and was administered to a random sample of 1,578 U.S. consumers. Not surprisingly, higher levels of income and education were found to significantly increase respondents’ credit knowledge score. Some, but not all prior credit experiences (e.g., exceptions included experiencing identity theft and living in a state with free credit report laws prior to the implentation of the FACT Act) were also found to significantly influence credit knowledge.

Financial knowledge may be associated with credit behavior. Chen and Volpe (1998) found that higher financial knowledge was linked to positive financial decisions among college students. Robb (2011) likewise found that students with higher scores on a measure of personal financial knowledge were more likely to engage in more responsible credit card behavior. Xiao, Serido, and Shim (2011) found that financial knowledge reduced risky credit behaviors among first year college students. A subsequent study that used the same data set found that subjective knowledge had a more significant impact than objective knowledge on reducing risky credit behavior among college students (Xiao, Tang, Serido, & Shim, 2011).

Delpechitre and DeVaney (2006) studied differences in credit card usage among various demographic groups using data from the 2001 Survey of Consumer Finances. Their results showed that African-American and Hispanic households were less likely than White households to be convenience users of credit cards and older household heads were also more likely to be convenience users. In addition, compared with those with a college degree, those with a high school education or less and those with some college were less likely to be convenience users.

Allgood and Walstad (2011) used probit analysis with a large national sample of 28,146 adults from the inaugural

68

Journal of Personal Finance

(2009) FINRA National Financial Capability Study to investigate financial knowledge and credit card behavior.

This study showed that a combination of actual and perceived financial knowledge significantly influence credit card behavior. It also found that about 4 in 10 (41.8%) survey respondents said they paid their credit card bills in full. The study investigated five credit card behaviors paying credit card balances on time, carrying over a balance and paying interest on it, making only a minimum payment on a credit card, being charged a late fee, and going over a credit card limit. In the 2012 FINRA National Financial Capability Study (Financial Capability, 2013), nearly three out of every five credit card holders (59%) engaged in at least one behavior resulting in either interest or fees and over a third (35%) engaged in two or more such behaviors.

Almost half (49%) carried over a balance and paid interest in some months, 34% sometimes paid only the minimum payment, 16% were sometimes charged a late fee, 11% sometimes used credit cards for a cash advance, and 8% were sometimes charged an over-the-limit fee for exceeding their credit line. In the 2012 Consumer Financial Literacy Survey (2012), with a national random sample of 1,007 adults, nearly two in five Americans (39%) carried credit card debt from month to month, down from 41% in 2010 and 40% in 2011. Most adults have reviewed neither their credit score (55%) nor their credit report (62%).

Credit Legislation During the 2000s Decade As noted above, two major federal credit reform laws, the Fair and Accurate Credit Transactions Act (FACTA) and the Credit Card Accountability and Disclosure (CARD) Act were implemented within the past decade. Since this study investigated credit management practices in years immediately following the passage of these laws, it adds to literature about whether or not legislation may have affected credit card practices. A brief description of each piece of legislation is provided. The credit laws are mentioned because several items on the online quiz relate specifically to credit management practices affected by these laws (e.g., payment of credit card late fees and availability of free credit reports to consumers nationwide upon request).

The FACT Act became law in November 2003 and was fully implemented in September 2005. It is enforced by the Federal Trade Commission (FTC) and the new Consumer Financial Protection Bureau or CFPB (Bucci, 2011). Since FACTA’s full implementation, consumers can request a free copy of their credit report annually from each of the “big three” credit reporting agencies or CRAs (Experian, Equifax,

and TransUnion). The law directed the CRAs to establish a centralized source for free credit report requests and to provide consumers with three methods of access: a toll-free number (877-322-8228), a website, and a postal address (Annual Credit Report Request Service, P.O. Box 105281, Atlanta, GA 30348-5281). The website www.annualcreditreport.com was developed for the purpose of filling orders for free annual credit reports. The CARD Act also contained provisions to help reduce identity theft and to assist identity theft victims (Bucci, 2011). A “Plain English” explanation of the FACT Act is available at http://www.privacyrights.org/fs/fs6a-facta.htm.

The CARD Act became law in May 2009 and was fully implemented in August 2010. This law provides protection against credit card interest rate increases and restricts various fees, such as over-the-limit fees that now require consumers to opt-in (give their permission) to exceed their credit limit and pay a fee of $25 or $35 if the credit limit is exceeded twice in six months (Susswein, 2010-2011). Late fees were also cut to $25 from the previous $35 or $39 charged by many creditors prior to the CARD Act taking effect. Like over-the-limit fees, card issuers are allowed to charge $35 in late fees to repeat offenders within a sixmonth period (Saha-Bubna, 2010). In addition, the CARD Act eliminated costly two-cycle average daily balance calculations that result in higher interest charges on the credit cards of cardholders who revolve a balance. Creditors are also required to provide 45 days’ advance notice of interest rate changes and are prohibited from increasing interest rates based on debt balances held with another creditor (i.e., so-called universal default clauses). A summary of major provisions of the CARD Act is available at http://banking.senate.gov/public/_files/051909_ CreditCardSummaryFinalPassage.pdf.

The Wise Credit Management Quiz The Wise Credit Management Quiz includes 20 items that are based on expert recommendations about how to manage credit wisely (e.g., paying credit card bills in full and having all debt balances listed as “current” or “paid as agreed” in a credit report) and on practices and debt danger signals to avoid (e.g., using predatory loans and taking out new loans or credit cards to pay off existing debt balances).

Some items are similar, but not identical, to those found in the studies cited above. Seven of the quiz items are answered with a “yes” or “no” response and the remainder utilize a Likert-type scale with responses ranging from “I never do this” to “I always do this.” “Yes” and “I always do this”

©2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1 responses are worth 5 points. Thus, the higher the quiz score (close to 100), the more frequently a person is practicing recommended credit management practices such as keeping their consumer debt-to-income ratio less than 20%, checking their credit report annually for errors, comparison shopping for loans, paying credit card bills in full to avoid interest charges, checking credit scores, keeping a list of credit accounts and creditor contact information, and avoiding predatory loans, late payments, upside down loans, high-cost credit insurance, and co-signing loans for others. Quiz users receive a score and an interpretation of their score upon completion of the quiz as well as a list of recommended action steps to improve their use of credit.

In the years since the financial crisis began and implementation of the CARD Act eliminated or restricted previously lucrative practices, creditors have been wooing high spenders with rewards programs to capture the fees paid by merchants that they do business with (Simon, 2012).

Thus, there is a quiz item that explores whether or not survey respondents are receiving credit card “perks” (e.g., cash back points) and are “convenience users” (i.e., credit card users who pay their bills in full to avoid interest charges). The question is asked in this way because interest and/or fees would probably offset any rewards earned by credit card users who revolve a high balance (needed to earn valuable rewards) from month to month.

Several quiz items refer to practices that are frequently recommended to raise a person’s credit score. Among these practices are - always paying as agreed (i.e., making at least the minimum payment by the due date), keeping debt balances low compared to available credit limits, and using a variety of types of credit (Importance of Credit History, 2012). Another practice, taking advantage of “float” time between the date of purchases and the date payments are due, provides consumers with borrowed money for almost two months interest-free. Two keys to using this strategy successfully are being a convenience user and knowing the date that a credit card’s billing cycle closes and making purchases right after that date.

69 Methodology and Sample The online Wise Credit Management Quiz has two primary purposes: to provide users with feedback on their credit management practices and to generate data to support ongoing empirical research about respondents’ financial behavior. It can be accessed from the website http://njaes.rutgers.edu/money/wise-credit/. As noted previously, total quiz scores can range from 20 to 100 with higher scores indicating more frequent performance of the 20 recommended credit management practices. Data were drawn from a convenience sample of 1,106 online respondents during a 44-month period between November 5, 2009 (inception of the online quiz) and June 30, 2013.

Respondents accessed the quiz through online searches and publicity by eXtension, Cooperative Extension agents, financial professionals, media reports, and social media (e.g. placing the quiz link in Tweets).

After excluding non-U.S. residents from the original sample of 1,106 online quiz respondents, the sample size used in this study was 1,081. Demographic characteristics are shown in Table 1. The sample was 74.2% female and 83.2% white compared to 50.7% and 79.6%, respectively, in the U.S. population (People Quick Facts, 2010). It also skewed toward younger ages. About 4 in 10 respondents (41.1%) were under age 35, 42.8% were age 35 to 54, and the remaining 16.1% of respondents were age 55 and older.

Respondents were almost equally divided between single and married persons and came from 48 states and the District of Columbia. Their demographic characteristics are shown in Table 1.

70

Journal of Personal Finance

Table 1 Descriptive Statistics About Online Survey Sample (N = 1,081) Demographic Characteristic % Gender Male 25.8 Female 74.2 Age Under 25 16.6 25-34 25.5 35-44 20.4 45-54 22.4 55 or older 16.1 Education High school or lower 20.5 Some college or associate degree 29.4 Bachelor’s degree or higher 50.1 Income Less than $25,000 15.4 $25,000-$49,999 31.4 $50,000-$74,999 25.7 $75,000-$99,999 13.7 $100,000 or higher 13.8 Race White 83.2 Non-white 16.8 Marital status Single, no minor children 40.0 Single, with minor children 8.3 Married, no minor children 24.7 Married, with minor children 27.0

Findings As noted above, the purpose of this study was to examine the credit and debt management practices of a convenience sample of Americans in the aftermath of implementation of two major federal credit laws and the Great Recession. The study presents a “snapshot” of the strengths and weaknesses of how people manage credit and whether they are following expert recommendations such as keeping a low consumer debt-to-income ratio and checking their credit report annually for errors and evidence of identity theft. The study also adds to existing literature on credit management behavior differences with respect to several demographic variables (gender, age, education, income, race, and marital status).

Among the 20 credit management behavior questions, the first 7 are yes or no questions. The answer of yes is coded as 5 and the answer of no is coded as 1. The remaining 13 questions are multiple choice questions and are coded as follows: 5 = I always do this; 4 = I usually (almost

always) do this; 3 = I do this about 50% of the time; 2 = I seldom or rarely (every once in a while) do this; and 1 = I never do this. The average total score and scores for each of the behavior questions are presented in the Table 2.

Demographic Differences in Credit Management Practices ANOVA analyses were conducted to examine credit management behavior differences with respect to gender, age, education, income, race, and marital status variables.

Not surprisingly, the highest credit management quiz scores were associated with respondents who were older and those who had higher incomes and educational levels. These respondents undoubtedly had the benefit of more years of credit management experience and more financial resources with which to repay their debt. Those who were married without children, White, and male also had higher online quiz scores.

These findings support those of the 2009 and 2012 FINRA Investor Education Foundation studies that found that women and those with low education, AfricanAmericans, and Hispanics were less likely to correctly answer financial literacy questions (Financial Capability, 2012, 2009). While those studies measured financial knowledge and this one measured credit management practices, people make financial decisions based upon what they know or don’t know and their level of financial literacy affects their financial behavior (Hilgert, Hogarth, & Beverly, 2003). Table 3 provides a summary of the findings with respect to demographic differences with * signs indicating a significance level of 5%. The following are some interesting findings: Gender Differences- The average total credit quiz score for males was 78.9 and for females was 73.3. The difference was statistically significant. In addition, 12 out of the 20 credit management behaviors showed gender differences, among which males performed better than females.

Age Differences- For the total credit quiz score, the age differences showed an approximately positive association.

The score of the group aged 24 or younger was 70.2, the group aged 25-34 was 75.5, the group aged 35-44 was 73.0, the group aged 45-54 was 77.1, and the age group 55 or older was 77.1. In addition, 17 individual behaviors also showed age differences. These differences had four patterns, among which four behaviors showed positive associations (perform more frequently when older), one behavior showed

©2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1 a negative association (perform less frequently when older), six behaviors showed a U pattern (perform the worst in a

71 middle age group), and six behaviors showed a reverse U pattern (perform the best in a middle age group).

Table 2 Scores for Credit Management Behaviors (N =1,081) Credit Management Behaviors Total score

Average score 74.8

1.

My latest credit report lists all my credit accounts as "current" or "paid as agreed".

3.6

2.

My current ratio of monthly consumer debt payments to monthly net income is less than 20%.

3.3

3.

I have checked my credit score within the last two years.

3.9

4.

I have NEVER had any of the following: collection accounts, judgments, liens, repossession, wage garnishment, foreclosure, or bankruptcy.

3.6

5.

The last time that I got a loan or credit card, I compared at least three different lenders before applying.

2.8

6.

I am NOT "upside down" on any of my loans.

3.7

7.

I have a list of my credit card account numbers and creditor contact information to refer to in the event of loss or theft.

3.2

8.

I pay my credit card bills in full every month to avoid interest charges.

3.4

9.

I check my credit report annually for errors and/or evidence of identity theft.

3.4

10. I AVOID making late payments on credit card bills and incurring late fees.

4.4

11. I charge no more than half (50%) of the maximum limit available on my credit cards.

3.6

12. I keep a running list of my outstanding credit card expenses as I use my credit card(s) so I am aware of my current total debt level.

3.7

13. It is my personal policy NOT to co-sign a loan for anyone.

4.3

14. I use at least one credit card that has some type of "perk" such as product or service discounts and/or cashback rewards AND am a convenience user.

3.4

15. I AVOID using predatory payday loans, car title loans, check-cashing stores, rent-to-own stores, and pawnshops to obtain cash and or merchandise.

4.5

16. I use or have used several different types of credit (e.g., personal loan, car loan, credit cards, student loan, and mortgage).

4.0

17. I AVOID buying high-cost credit life insurance, credit disability insurance, and credit unemployment insurance when securing a loan and/or credit card.

4.5

18. I AVOID taking out new loans or credit cards to pay off existing debt balances.

4.3

19. I AVOID taking credit card cash advances.

4.5

20. I make credit card purchases at the beginning of my monthly billing cycle to take advantage of maximum "float" time between when purchases are made and when payments are due.

2.7

72

Journal of Personal Finance

Education Differences- For the total quiz score, education showed a positive association. For consumers with high school or less education, the quiz score was 69.6, for those with some college, trade/vocational school, or an associate degree, the score was 70.6, and for those with a bachelors degree or higher, the score was 79.3. In addition, 18 individual behaviors showed education differences.

Among them, 12 behaviors showed positive associations (higher education is associated with better credit management behaviors) and six showed a U pattern (the mid education group shows the lowest behavior score).

Income Differences- For the total quiz score, income showed a positive association. From the lowest to the highest income groups, the scores were 68.4, 71.3, 75.4, 80.9, and 82.3, respectively. 19 behaviors also showed income differences. Among these behaviors, one showed a reverse U pattern (a middle income group performs the best), six showed a U pattern (a middle income group

performed the worst), and the remaining 12 behaviors showed a positive pattern (the higher the income, the better the behavior).

Racial Differences- The average total credit quiz score of whites was significantly higher than nonwhites, 76.0 vs.

68.6. In addition, whites showed significantly higher scores than nonwhites on 14 individual behaviors.

Marital Status Differences- Among the four marital status categories, singles with children had the lowest total score of 68.8 while married respondents without children had the highest score of 78.6. The average scores for singles without children and for married persons with children were 73.4 and 75.0, respectively. A total of 16 behaviors also showed marital status differences. Among these behaviors, married with children had the lowest score on one behavior, singles without children on two behaviors and singles with children had the lowest scores on 13.

Š2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1

73

Table 3 Significant Differences in Credit Management Practices by Demographic Characteristics Behavior Total score

Gender m>f*

Q1 Q2

m>f*

Q3 Q4

m>f*

Q5

m>f*

Q6

m>f*

Q7

Age positive*

Education positive*

Income positive*

Race w>nw*

Marital status single with children*

positive*

positive*

positive*

w>nw*

single with children*

U*

U*

single with children* single no children*

reverse U*

positive*

positive*

U*

U*

U*

positive*

positive*

U*

w>nw*

single with children*

U*

w>nw*

single with children*

U*

U*

w>nw*

single with children*

U*

positive*

positive*

positive*

positive*

w>nw*

single with children*

U*

positive*

U*

w>nw*

single with children*

positive*

reverse U*

w>nw*

married with children*

reverse U*

positive*

positive*

w>nw*

single with children*

U* U*

Q8

m>f *

Q9

m>f *

Q10 Q11

m>f *

Q12

m>f *

Q13

U*

Q14

m>f *

positive*

U*

positive*

w>nw*

single with children*

Q15

m>f *

reverse U*

positive*

positive*

w>nw*

single with children*

Q16

positive*

positive*

positive*

w>nw*

single no children*

Q17

reverse U*

positive*

positive*

w>nw*

single with children*

reverse U*

positive*

positive*

w>nw*

single with children*

reverse U*

positive*

positive*

w>nw*

single with children*

negative*

U*

U*

Q18

m>f *

Q19 Q20 *p<.05

m>f *

74

Journal of Personal Finance Discussion

Data were collected at least four, and up to eight, years following full implementation of the FACT Act which requires free credit reports upon request. Yet the average quiz score for checking a credit report annually was the fifth (tied) lowest (3.4). Interestingly, and similar to findings from the 2012 Consumer Financial Literacy Survey, the average score for checking credit scores was higher (3.9) than for checking credit reports even though free credit scores are not required by law and free credit reports are. Perhaps respondents obtained free credit scores automatically from potential lenders or used free non-FICO credit scoring websites such as CreditKarma. Another possibility is that a three-digit number is simply easier to understand than a multi-page report, which makes obtaining credit scores more attractive to consumers.

Another quiz item with a slightly lower average score (3.3) than that for checking credit reports is the question pertaining to having a current consumer debt-to-income ratio (i.e., monthly nonmortgage debt payments divided by takehome or net pay) lower than 20%. This means that a number of respondents were at or above what is considered a “danger zone” for overextension. A ratio of 15% or less is desirable and a ratio of 16% or above is considered problematic because a household is making high debt payments and could quickly experience financial difficulty if a disruption in income were to occur (Garman & Forgue, 2008). A 20% debt-to-income ratio is like working for five days and getting paid for four because a full day’s income is already “spoken for.” The Great Recession and its aftermath took a toll on many people who experienced a loss or reduction in income. When that occurs, their consumer debt-to-income ratio subsequently increases because consumer debt payments are divided by a smaller disposable income. This could be what some survey respondents had experienced by the time they took the online quiz.

Two even lower-scoring items are comparing at least three different lenders before applying for credit (the average score is 2.8) and taking advantage of the maximum “float” time between when credit card purchases are made and when payments are due (2.7). Both of these practices require proactive planning and time to implement and are obviously not being performed very frequently. The third lowest scoring item, with an average score of 3.2, was having a list of credit card account numbers and creditor contact information to refer to in the event of loss or theft. Again, this practice requires some time to research and pull together data into one document. Perhaps a widely marketed, user-friendly

downloadable “fill in the blanks” tool to help do this would increase compliance with this frequently recommended practice.

On the plus side, respondents infrequently reported engaging in high-cost or high-risk practices. These scores included 4.5 for avoiding credit-based insurance (e.g., credit life insurance), predatory payday loans, and cash advances and4.3 for avoiding new loans or credit cards to pay off existing debt balances and for not co-signing a loan for anyone. Many also practiced behaviors that result in a high credit score such as avoiding late payments on credit cards (score of 4.4), using different types of credit (score of 4.0), and, to a lesser degree, charging no more than half of the maximum limit available on credit cards (score of 3.6).

Finally, the demographic differences in credit management practices that were found in this study were not surprising. The “usual suspects” of demographic groups that consistently score lowest on financial literacy tests (Financial Capability, 2012, 2009) and in reports about quartiles of household income and wealth (Bricker et al., 2011) were found to have the lowest credit management quiz scores. This includes females, single parents, young adults, minorities, and those with lower incomes and educational levels.

Implications Following are three implications from this study for financial practitioners: § There is a need to make credit management easier, perhaps with a collection of “fillable forms” that consumers can download to give themselves a periodic credit “check up.” People are not performing many recommended practices such as consolidating their credit data and keeping a running total of credit card expenses. Perhaps these tools can even be developed easily as phone apps. § Despite having the ability to request free credit reports for over eight years, many people still do not do so. Financial practitioners might consider having consumers fill out a credit report request form when they are together in a face-to-face meeting or at financial education programs, even supplying a stamped, self-addressed envelope, if necessary, to encourage compliance. § Outreach to vulnerable populations who score lowest on assessments of financial literacy and financial behavior continues to be warranted. Increased financial knowledge can result in increased financial behavior (Hilgert, Hogarth, & Beverly, 2003) so a good way to improve credit management practices is to start with financial education that addresses the knowledge gaps and wise credit use barriers of specific target audiences.

©2014, IARFC. All rights of reproduction in any form reserved.

Volume 13, Issue 1

75 References

Allgood, S. & Walstad, W.B. (2011, July). The effects of perceived and actual financial knowledge on credit card behavior. Networks Financial Institute at Indiana State University, Working Paper 2011-WP-15. Retrieved July 10, 2013, from http://papers.ssrn.com/sol3/ papers.cfm?abstract_id=1896365.

Bricker, J., Kennickell, A.B., Moore, K.B., & Sabelhaus, J. (2012, June). Changes in U.S. finances from 2007 to 2010: Evidence from the survey of consumer finances.

Washington DC: Federal Reseve Bulletin, 98(2).

Retrieved July 11, 2013, from http://www.federalreserve.gov/Pubs/Bulletin/2012/ articles/scf/scf.htm.

Bricker, J., Bucks, B., Kennickell, A., Mach, T., & Moore, K. (2011). Surveying the aftermath of the Storm: Changes in Family Finances from 2007 to 2009.

Washington, DC: Federal Reserve Board. Retrieved March 25, 2011, from http://www.federalreserve.gov/ pubs/feds/2011/201117/201117pap.pdf.

Bucci, S. (2011). Credit management kit for dummies.

Hoboken, NJ: John Wiley & Sons.

Chen, H., & Volpe, R. P. (1998). An analysis of personal financial literacy among college students. Financial Services Review, 7(2), 107-128.

Collingwood, H. (2012). The recovery’s silent assassin: How debt deleveraging killed the economy. The Atlantic. Retrieved April 24, 2012, from http://www.theatlantic.com/business/archive/2011/10/ the-recoverys-silent-assassin-how-debt-deleveragingkilled-the-economy/246463/.

Cronin, B. (2011, October 31). Slow recovery feels like a recession. The Wall Street Journal, p. A5.

Delpechitre, D. & DeVaney, S.A. (2006). Credit card usage among White, African American and Hispanic Households. Consumer Interests Annual, 52, 448-454.

Financial capability in the United States: Findings from the 2012 national financial capability study (2013).

Washington, DC: FINRA Investor Education Foundation. Retrieved July 10, 2013, from http://www.usfinancialcapability.org/downloads/NFCS_ 2012_Report_Natl_Findings.pdf.

Financial capability in the United States (2009).

Washington, DC: FINRA Investor Education Foundation. Retrieved May 6, 2012, from http://www.finrafoundation.org/web/groups/foundation/ @foundation/documents/foundation/p120535.pdf.

Garman, E.T. & Forgue, R. E. (2008). Personal finance (9th edition). Boston: Houghton Mifflin Company.

Hilgert, M. A., Hogarth, J.M., & Beverly, S. G. (2003, July).

Household financial management: The connection between knowledge and behavior. Federal Reserve Bulletin. Retrieved May 7, 2012, from http://www.federalreserve.gov/pubs/bulletin/ 2003/0703lead.pdf.

Importance of credit history and successful saving (2012).

Washington, DC: America Saves and Experian.

Retrieved May 1, 2012, from http://www.americasaves. org/images/newsletters/creditscore.pdf.

Lyons, A.C., Rachlis, M., & Scherpf, E. (2007, Winter).

What’s in a score? Differences in consumers’ credit knowledge using OLS and quantile regressions. Journal of Consumer Affairs, 41(2), 223-249 Mian, A. Sufi, A., & Trebbi, F. (2013). Resolving debt overhang: Political constraints in the aftermath of Financial Crises. Retrieved July 11, 2013, from http://faculty.arts.ubc.ca/ftrebbi/research/mst4.pdf.

O’Neill, B. & Xiao, J.J. (2012). Financial behaviors before and after the financial crisis: Evidence from an online survey. Journal of Financial Counseling and Planning, 23(1), 3-16.

Paradox of thrift (2013). Investopedia. Retrieved July 10, 2013, from http://www.investopedia.com/ terms/p/paradox-of-thrift.asp.

People quick facts (2010). Washington DC: U.S. Census Bureau. Retrieved August 16, 2010, from http://quickfacts.census.gov/qfd/states/00000.html.

Reinhart, C.M. & Rogoff, K.S. (2009). The aftermath of financial crises. American Economic Review, 99(2): 466–72.

Ritholtz, B. (2009, January 24). The aftermath of financial crises. The Big Picture. Retrieved July 11, 2013, from http://www.ritholtz.com/blog/2009/01/the-aftermath-offinancial-crises/.

Robb, C. A. (2011). Financial knowledge and credit card behavior of college students. Journal of family and economic issues, 32(4), 690-698.

Saha-Bubna, S. (2010, August 21-22). Credit firms gird for cut in late fees. The Wall Street Journal, p. B3.

Simon, R. (2012, March 24-25). Credit-card rewards get richer. The Wall Street Journal, P. B7, B10.

Susswein, R, (2010-2011, Winter). Over limit? It’s up to you. Consumer Action News, p. 2.

Szaltner, M., Van Horn, C., Zukin, C. (2013). Diminished lives and futures: A portrait of America in the greatrecession era. New Brunswick, NJ: John J. Heldrich Center for Workforce Development. Retrieved July 11, 2013, from http://www.heldrich.rutgers.edu/sites/ default/files/content/Work_Trends_February_2013.pdf.

Taylor, P., Morin, R., Kochhar, R., Parker, K., Cohn, D., Lopez, M.H., Fry, R., Wang, W., Velasco, G., Dockterman, D., Hinze-Pifer, R., & Espinoza, S. (2010).

A balance sheet at 30 months: How the great recession has changed life in America. Washington, DC: Pew Research Center, Social and Demographic Trends Project. Retrieved August 11, 2010, from http://pewsocialtrends.org/assets/pdf/759-recession.pdf.

The 2012 consumer financial literacy survey (2012).

Washington, DC: National Foundation for Credit Counseling. Retrieved July 11, 2013, from

76

Journal of Personal Finance

http://www.nfcc.org/newsroom/FinancialLiteracy/ files2012/FLS2012FINALREPORT0402late.pdf.

The 2011 MetLife study of the American dream: The do-it yourself dream (2011). New York: Metropolitan Life Insurance Company. Retrieved July 11, 2013, from https://www.metlife.com/assets/cao/gbms/studies/metlif e-2011-american-dream-report.pdf.

The 2009 MetLife study of the American dream: Rebooting the American dream. shifted. altered. not deleted (2009). New York: Metropolitan Life Insurance Company. Retrieved February 16, 2010, from http://www.metlife.com/assets/cao/gbms/studies/ 09010229_09AmDreamStudy_WEB.pdf.

The new credit crunch (2009, July). Consumer Reports, 74(7), 15-17.

U.S. business cycle expansions and contractions (n.d.).

Retrieved April 24, 2012, from http://www.nber.org/cycles.html.

Van Horn, C. (2013). Working scared (or not at all): The lost decade, great recession, and restoring the shattered American dream. Lanham, MD: Rowman & Littlefield Publishers.

Xiao, J. J., Serido, J., & Shim, S. (2011). Financial education, financial knowledge, and risky credit behaviour of college students. In D. Lamdin (ed.).

Financial decisions across the lifespan: Problems, programs, and prospects (pp113-128). New York: Springer.

Xiao, J. J., Tang, C., Serido, J., & Shim, S. (2011).

Antecedents and consequences of risky credit behavior among college students: Application and extension of the Theory of Planned Behavior. Journal of Public Policy & Marketing, 30(2), 239-245.

Š2014, IARFC. All rights of reproduction in any form reserved.

IARFC 2014 National Financial Plan Competition Registration New York, New York Las Vegas Hotel & Casino

Step 1

Step 2

Check your events

Price*

Schedule of Events

Impact Branding Workshop

Please check appropriate box:

Welcome Reception

Impact Branding Workshop 4/30/14

National Financial Plan Competition

6:30pm – 7:30pm

Impact Branding & Reception

$225

Workshop, Reception & Competition

$275 $105

Reception & Competition

National Financial Plan Competition 5/1/14

$225 $20 $95

Special Packages

2pm – 5pm

Welcome Reception 4/30/14

Figure your pricing. Don’t forget to take advantage of Special Packages for more than one event.

*Travel and Lodging are not included in the price. The IARFC has negotiated discounted room rates starting at $60 per night at New

7:30am – 3pm

York, New York Las Vegas Hotel & Casino, registration details will be available November 15, 2013.

For complete agenda and details visit: www.IARFC.org or call 800.532.9060

*Registration Deadline for Events: March 31, 2014

Step 3

REGISTER NOW Please print or type information below.

Mr.

Put total event price on the Total Payment line, fill out form below and and return to the IARFC.

Total Payment: _____________________________

Ms. Mrs.

_________________________________________________________________ First Name Last Name _________________________________________________________________ Firm or B/D

Payment Information:

Credit Card Check made payable to IARFC

_________________________________________________________________ Street Address

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

_________________________________________________________________ City State Zip Code

________________________________________________________ Signature

________________________________________________________________ Phone _________________________________________________________________ E-mail

7 RFC® approved CE credits Continuing Education Credits available - Impact Branding (3), Competition Process (3), Feature Presentation (1)

Registration Deadline and Cancellation Policy: Workshop attendees may register or cancel up to 30 days prior to the event date. Cancellations will receive a full refund, less a $20 processing fee.

See New York, New York Las Vegas Casino & Hotel cancellation policy, which is separate from IARFC. Note: Program Subject to Change

PLEASE RETURN THIS COMPLETED FORM TO IARFC

IARFC INTERNATIONAL ASSOCIATION OF REGISTERED FINANCIAL CONSULTANTS

International Association of Registered Financial Consultants Financial Planning Building 2507 North Verity Parkway P.O. Box 42506 Middletown, OH 45042-0506

Fax: 513.424.5752 Questions: 800.532.9060 www.IARFC.org info@iarfc.org

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

2014 Sponsors National Financial Plan Competition Corporate Sponsors

Individual Sponsors

Gold: Cetera Advisors

Silver: Harvey Charbonneau Michael Coy Vicki Diggles Enyi P. Enyi Victor Ferreira Christopher Hill Eileen Kohler Brian Low Patrick Lyman Burnett Marus Calvin Mendez Fred Ostermeyer Jeffrey Rattiner Jeffrey Reitzel Paul Schwabe Mark J. Sheahan Thomas Stark Winston Williams

Royal Alliance

Visit www.IARFC.org/ FinancialPlanCompetition

Gold: Jay Bailey Michelle Blair Douglas Brown Isabel Cooper Frank Eberhart Donald Fortney Andrew Kwittken Robert Meints, II Rick Mueller, II Chris Roberts Nick Royer Richard Villers

Platinum: Les Anderson Steve Bailey Albert Coletti Ed Morrow Rosilyn Overton Gary Padussis Jon Rogers Micah Shilanski

Diamond: Pete Dâ&#x20AC;&#x2122;Arruda Lloyd Lowe Sr. Edward Skelly

Join in the Sponsorship and have your name listed here and on IARFC.org. Thank you for Supporting the Future of Financial Planning. We have Individual Sponsorship Opportunities starting at $50.

For more information call: 800.532.9060