Health demand

Chapter 3

The Demand for Healthcare

1. Introduction Even if the health care sector in most countries is heavily regulated, so that it will only make limited sense to think of a demand after its goods and services in the usual way, it still is a fruitful approach to begin our investigation of the economics of the health care sector by a consideration of consumer behavior. Irrespectively of concrete institutional details there is still a need for services to be satisfied in this sector, and most important for our economic point of view, they will have to be balanced against many other needs directed agains commodities and services from other sectors.

The fact that “health”, whatever it may be, certainly cannot be bought directly in the marketplace does of course not exclude that an economic theory of consumer choices is established in relation to “health”. There are many other cases form economic life where rather abstract services are demanded, which cannot be bought directly (such as for example benefits derived from environment and biodeiversity). It can even with some right be argued that it is the rule rather than the exceptioon that the good which is wanted by the consumer is actually available in the market. The consumer doesn’t want carbon hydrates, fat and stabilizer which is what she gets according to what is written on the package, rather the goal is to get social intercourse, excitement, status, etc. On her way to achieve these goails the consumer will have to carry out concrete purchases of commodities, and it is mainly these acts which are the object of the economic analysis.

What shall interest us in the present context is the trade-off between ordinary consumption on one side and the consumption of commodities and services from the health care sector on the other. This means that we are not aiming at deriving an aggregate demand for “health”, something which was not to expected in view of the ambiguous nature of the concept of “health”. The models to be considered are rather directed at enhanding our understanding of a few aspects of the problem which may be considered as particularly interesting and promising for our overall understanding of health behavior; there is no general theory in this field, not yet at least.

The different aspects of the concept of health, which may have a role to play in describing consumption choices and other economic behavior, may be ability to use available time according to own wishes (rather than being ill), as is the case in the Grossman model to be treated below, or it may be the cost in time or money of being ill (as in Phelps and

1

Newhouse). In both situations we consider the consequences for consumption, so that what is actually studied, is the rational choice of a consumer in cases where some of the choices are related to health. The assumption of rationel and consistent behavior in such choice situations gives rise to a certain pattern of behavior, which can be made subject to an analysis of comparative statics.

In a health care system with full third-part financing there is not much left over for choice, since in this case it is rational to choose as much as is made available. This solution may rightly be expected to be inefficient, since the individual will not balance the demand for health care with the cost of providing it. However, not only the system of paying for services is specific in the health care sector, more often than not the services themselves have characteristics which make them somewhat different from ordinary commodities and services, and then the system of third-party-payment may not be improper.

In the following two sections, we consider the Grossman health capital model, where the choice of consumption over a lifetime is seen as an investment problem, so that the consumer can choose between investing in her own health or investing in the money market. The fact that these two alternatives are open to the consumer means that the choice of investment in health, and consequently the current state of health, will depend on aggregate economic variables; we investigate this dependence. Following this, we consider some other models of consumer choice, most of them much simpler in there structure since they do not incorporate intertemporal choices, and we review some of the empirical evidence in the field, including the Rand experiment in health insurance and state of health.

2. Choice of health as an intertemporal consumption problem (1) In the Lancaster model of consumption characteristics outlined in the previous chapter, the characteristics, and among those the characteristics pertaining to health, were something produced and consumed alongside with ordinary commodiities. It might however be questioned whether this formulation is quite satisfactory, in particular since there is a long-run aspect of health which cannot reasonably be assumed away. When people quit smoking, it is usually not by fear of having a lung cancer at some instant in the evening, but some time in next 20 – 30 years, and the other way around: Eating a healthy diet will not necessary give a better feeling of health the same day (usually rather to the contrary) but it will in the course of some time. Certainly, much consumption related to health – indeed most of what is not treatment here and now, have a long-term aspect, and correspondingly the choice of consumption should be studied in the long perspective.

This is exactly what is done in one of the by now classical models in the field, the consumption choice model of Grossman (1972). Before looking at the model in full detail, what we shall certainly do, it may be useful the outline its logical structure and main results, and this is what will be done in the present section.

The basic idea of the model, considerably simplified, is that health is considered as a personalized capital good, which can be built up by the consumer through investment in her own health, something which is certainly costly in terms of money and time; one may think of investment as healthy diet, recreational travels in the Caribbean, frequent visits in gyms and at the doctor (whereby it is understood that the consumer pays herself for most of 2

this). The health capital is subject to a certain reduction being worn out by use (this rate of decrease may reasonably be assumed to be small in the consumers early life but increasing with age); its payoff is ability to work, measured in possible working ours – and with known money wages, directly in money. In our first (actually also in the more detailed) approach we may abstract from other payoffs such as the sheer pleasure of being healthy; here it is good to be healthy just because you can work more (or, slightly more sophisticated, because you have more time to be used either for work or for leisure, not wasted by illness).

To see by an intuitive argument what will come out of rational consumer choice in a world as the one sketched above, let us assume that the consumer wants to change the health capital stock at some time t be a small unit. The gain to be achieved by this additional unit of health capital is the marginal product of health capital at time t, denoted Gt (remember that this product is measured as time saved from illness, which multiplied by the wage rate w gives the increase in income in the next period. The cost is s (the ratio between capital and investment), multiplied by the cost per unit health investment Ct−1 (the investment cost is dated one period earlier). In optimum we have sCt−1 = wGt or

wGt .

Ct−1 This was the period-t-optimization condition. In the longer run, there is one more consideration to be made, namely that the investment shall match the built-in deterioration of health capital, which occurs at a rate d, as well as the time preference given by the discount rate of interest r. Consequently, on a steady-state path we must have that the rate of growth in capital s, must equal the rate of decrease, corrected by discounting, that is d + r. In total we therefore get wGt = d + r.

Ct−1 We see that the left hand side can be identified as the marginal efficiency of investment in health, the payoff per dollar invested in this particular type of capital. If we correct this by the rate of deterioration of the health capital, then it should equal the marginal efficiency of any other investment, in particular investment in the money market, which has the payoff r per dollar invested.

This rather simple version of the Grossman model may now be supplemented with further assumptions, mainly of a very conventional type. This will give us a possibility of doing comparative statics, which may give insigts into the reaction of the public to changes in either the health-related data of the situation or in the economic conditions for intertemporal choice. s=

3

marginal efficiency of capital

d+r r

Health capital

Optimum size

Figure 1. Decreasing marginal efficiency of health investment with the size of the health capital. When the rathe of depreciation increases, e.g. due to aging, the optimal size of the health capital is reduced.

Thus it sounds quite plausible that there should be a decreasing relationship between size of health capital and marginal efficiency of health investment, the payoff per dollar of additional investment in health.The more healthy one becomes, the less will be the payoff from having even better health; the relationship is illustrated in Figure 1, which also shows how the consumer adapts in her optimization to changes in the depreciation rate. If it is assumed that the rate of depreciation increases with age, then optimal choice will lead to a reduction over time of the size of the health capital, since it will not be optimal to keep up the previous state of health as it becomes too costly to replace all the capital which is worn out.

In its consequence, this means that health capital will decrease over the years and eventually hit the level of 0, presumably to be interpreted as death. It should be remembered, however, that the model is one of fixed horizon and full certainly, so that death does not come as a surprise, but as a planned final state (if terminal size of health capital does not matter for the utility of the consumer).

A more interesting consequence of the same type of analysis will emerge if we assume that the payoff of investment in health capital is changed, something that may happen as a result of new medical technology. When it becomes easier to acquire new units of health capital (Ct−1 decreases), then the value of Gt , which will equalize the left and right sides of the equation above will become smaller. Inserting this into the decreasing relationship between the size of the health capital and its marginal efficiency of investment will tell us that optimal size of health capital increases.

We can also follow the effects of a change in the wage rate w on the optimal stock of health capital. The wage rate appears explicitly in the numerator of the fundamental expression for optimal choice of health capital, so it would seem obvious that an increased 4

wage rate makes health investment more attractive and therefore shifts the optimal size of the health capital in the upwards direction. However, this effect will have to be modified due to the fact that the wage rate does enter also in the denominator, even if not with the same weight: An increase in the wage rate of 10% appears in the numerator with the full 10% , while the denominater increases only to the extent that the expenditure on health investment contains labor cost (of others such that medical professionals, personnel attending the machines in the gymnasium etc., and own time evaluated at opportunity cost, i.e. the wage to be earned in alternative occupation) rather than remuneration of other factors of production.

In an extreme case the labor content of health investment might be 100%, in which case changes in the wage rate would leave the optimal amount of health capital unchanged, but under ordinary circumstances it is to be expected that the value of the expression on the left hand side will be reduced, and in order to retain equality with the right hand side, the amount of health capital “installed� must be increased. Thus, a rise in wages entails an increase in the general state of health, perhaps not a very shocking new piece of insight, but at least an illustrative example of the way in which a precise model of optimizing behavior, in casu the Grossman model of long term choices of consumption and health, can be used to produce forecasts of aggregate behavior. marginal efficiency of capital

d+r

new optimum

optimum size

Health capital

Figure 2. The effects of an increase in the wage rate. Since the change affects both numerator and denominator, but the latter less than the first, the overall effect will be an increase in the marginal efficiency of health investment at any level of health capital, so that the curve shifts upwards. For fixed r and d the new optimum will be reached at a higher value of the health capital.

5

3. Choice of health as an intertemporal consumption problem (2) In the present section we discuss the Grossman model in somewhat greater detail than that of the previous section. As before, we consider a single consumer facing an intertemporal choice problem. There is a given planning horizon of n periods. This does not necessarily mean that the consumer lives with certainty in these n periods and then dies, since life and death may be considered as consequences of the size of the health capital and consequently are depending on the consumer’s own decisions.

Let the size of the health capital in each of the periods considered be H0 , H1 , . . . , Ht , . . . , Hn .

We assume now that there is a certain payoff structure connected with health capital, so that a stock n Ht of health capital at time t gives rise to a flow of health capital services ht at that time. The size of ht is an expression of the total benefits from being healthy which the consumer obtains in the given period.

Life and death enters into the model through the existence of a minimum size of the health capital, Hmin . If the stock gets beyond this minumum size, the consumer dies. Moreover, the existing stock of health capital is subject to a continuous devaluation or depreciation, so that the stock is adjusted from one period to the next by Ht+1 = Ht + It − dt Ht ,

(2)

where It is the health investment of the period and dt the rate of depreciation effective for period t; as mentioned earlier, it seems reasonable that this rate of depreciation changes over time due to aging.

The consumer is endowed with a utility function of the form U(h0 , . . . , hn , ξ0 , . . . , ξn ) which – at least at the formal level – depends on the current benefit derived from the health capital in each of the periods; the variables ξt denote the consumption at time t of other characteristics (in the Lancaster sense) than those related to health.

It is assumed that the consumer in each period produces both the health investments It and the bundle of other consumption characteristics ξt in a household technology, here given by production functions It = G H (mt , T tH ), ξt = G F (xt , T t ), assumed to be time independent. The variables entering as factors of production, are mt , the amount of health releated commodities and services purchased in period t, the bundle xt of ordinary consumption commodities bought in the market, whereas the two variables T tH and T t denote the time used by the consumer in producing health investments and characteristics bundle, respectively. If it is assumed that the household production satisfies constant returns to scale, we may replace the production function G H by a function of only one variable, It = mt gH (τtH ), τtH = 6

T tM mt

where gH (τtH ) = G H (1, τtH ) (production per unit of commodities inserted, a standard approach in growth theory). We then may express the marginal products in the production of health investments as dgH ∂It = H = g0H , H ∂T t dτt ∂It = gH − τtH g0H . ∂mt It remains only to add the budget constraints. Regarding the income of the consumer we assume that there is only a single budget restruction, which states that discounted future purchases of health related services and consumption goods should not exceed the discounted future incomes (obtained by selling labor power) together with possible initial purchasing power: n n X X [pt mt + vt · xt ](1 + r)−t = wt T tW (1 + r)−t + A0 , t=0

t=0

where wt is the wage rate at time t, and is the time expenditure in the labor market.

Since the use of time is explicitly taken care of in the model, we need a budget restriction also here, namely T tW + T tH + T t + T tL ≤ Ω, T tW

where the first three variables have already been introduced as use of time in the labor market and time expenditure in household production of health investment and consumer characteristics; the time spent in producing consumer characteristics – or some of it – may be interpreted as leisure time. The last of the four variables, T tL , denotes loss of time due to illness. We need to relate this variable also to the remaining variables, and this is done by assuming that it depends on the size of the health capital, T tL = L(Ht ).

Letting the time loss depend on health seems natural, as does the assumption that dL/dHt < 0, the more health, the less time is wasted due to illness. The constant Ω denotes the total time available in the period.

We may now solve the consumer’s problem which amounts to choosing investment and consumption over time in such a way that total utility is maximized, given the constraints determined by technology, budget, and time. We derive the first order conditions in the standard way, constructing the Lagrangian and setting its derivatives to zero. We assume that in the optimum considerd, all the constraints are satisfied as equalities, and therefore we can eliminate the time constraint by inserting T tW = Ω − (T tH + T t + T tL ) in the ordinary budget constraint.

Having done this, we obtain a Lagrangian of the form n X U(h0 , . . . , hn , ξ0 , . . . , ξn ) − λ[ [vt · xt + Ct + w(T t + T tL )](1 + r)−t ], t=0

where Ct = pt mt −wT tH is the health expenditure of the period, that is purchase of commodity inputs and use of time evaluated at offer cost (which is the wage rate); we have left all 7

constants out of the Lagrangian, which does not change the first order conditions where they will not show up anyway. Differentiating with respect to It−1 , health investment at time t − 1, we obtain Ut0

∂ht ∂Ht ∂Ht ∂It−1

+ · · · + Un0

∂hn ∂Hn ∂Hn ∂It−1

−λ[πt−1 (1 + r)−(t−1) + w +··· + w

∂T tL ∂Ht (1 + r)−t ∂Ht ∂It−1

∂T nL ∂Hn (1 + r)−n ] = 0. ∂Hn ∂It−1

There are three different types of members of this expression; the first type (indexed from t to n) emerges from differentiating the utility function w.r.t. the health components (U depends on ht which depends on health capital at time t, and the latter is influenced by It−1 ).

After this come the members derived from the constraint; first we have an expression which gives us the effects at t −1, when the investment is carried out; πt−1 denotes the marginal cost in production of health investment, and they can be found from the expressions w = πt−1 g0H (price of a factor equals the value of the marginal product of that factor) or alternatively pt−1 = πt−1 (gH − g0H ). Following this we have members showing the effects on the budget constraint of these investments in the years to follow; this effect is indirect and stems from the effect of the health capital stock on the time available to the consumer.

It remains to rewrite the expressions ∂I∂Ht−1τ for τ = t, . . . , n slightly. For this we use the relation (2) connection investment and capital. For τ = 1 the partial derivative equals 1, since all of the investment is carreid over as health capital in the next period. For the following periods the depreciation of health capital has to be taken into account, so that ∂Ht+1 ∂Hn = 1 − dt , = (1 − dt ) · · · (1 − dn−1 ). ∂It−1 ∂It−1 We insert these expressions in the first order condition as derived above, and move the members around, isolating the expression λπt−1 (1+r)−(t−1) on the left hand side, and inserting ∂hτ = −L0 (Hτ ) = γτ , τ = t, . . . , n. ∂Hτ Here L0 is the derivative of the function determining the dependence of healthiness on health capital. It is seen that it is also the derivative of health consumption w.r.t. health capital, and that could not have been derived from what has been assumed so far, but amounts to a new assumption, saying that healthiness is important because (and, to be sure, only because) it makes more time available. Whether this assumption is a plausible one, is another matter.

Be this as it may, we end up with the expression πt−1 (1 + r)−(t−1) =

wγt (1 + r)−t + (1 − dt )wγt+1 (1 + r)−(t+1) + . . . + (1 − dt ) · · · (1 − dn−1 )wγn (1 + r)−n U 0 γt U 0 γn + t + . . . + (1 − dt ) · · · (1 − dn−1 ) n . λ λ 8

We now have a formula for the discounted marginal cost of investment in health at time t − 1. To get on to something useful we reduce the expression by finding the similar expression for marginal cost one period later, that is πt (1 + r)−t , which of course looks much like the former one, namely as follows, πt (1 + r)−t =

wγt (1 + r)−(t+1) + . . . + (1 − dt+1 ) · · · (1 − dn−1 )wγn (1 + r)−n 0 Ut+1 γt+1 Un0 γn + + . . . + (1 − dt+1 ) · · · (1 − dn−1 ) . λ λ

Now we multiply the last equation by (1 − dt ) and subtract it from the first, which means that several of the members cancel each other; what is left is only πt−1 (1 + r)−(t−1) = wγt (1 + r)−t +

Ut0 γt + (1 − dt )πt (1 + r)−t . λ

Isolating the two members containing γt on the right hand side and multiplying by (1+r)t , this side of the equation becomes γt [w +

Ut0 (1 + r)t ]. λ

On the other side we are left with πt−1 (1 + r) − πt (1 − dt ) = πt−1 + πt−1 r − πt + πt dt , a somewhat messy expression, which however may be improved by some cosmetic changes: We introduce the quantity πt−1 − πt π˜ t−1 = πt−1 which is the percentual change in marginal cost from t−1 to t; from this we obtain πt−1 −πt = πt−1 π˜ t−1 . Next we cheat slightly and put πt dt = πt−1 dt , and we have a ready-made expression γt [w + Ut0 λ−1 (1 + r)t ] = r − π˜ t−1 + dt . πt−1 Comparing with the intuitive derivation of the fundamental relationship of the Grossman model presented in the previous section, we see that we have obtained basically the same formula; the added difficulties of the present approach has the advantage of giving a better understanding of the relationship. On the left hand side we still have marginal payoff of health investment (but now we have also the “pure” consumption effect of health in the formula), and on the right hand side we have the quantities which relate to the time aspect, whereby the correction term π˜ t−1 has been added; this term is not an inessential one since it expresses the effects of the structure over time of the investment and its effects over the periods to follow.

On the other hand, the fact that our derivation here gives us basically the same result as we used in our small exercises in comparative statics in the previous section is reassuring, as it shows that this comparative statics was fundamentally sound and indeed may be extended to several other relevant situations.

9

4. Other models of demand for services of the health care sector While the classic Grossman model focusses on the investment aspects of health behavior, other models of demand for health have emphasized the short run, where health expenditure has to do with actual illness and these expenditures compete with other consumption goods for the short term budget of the consumer. Nevertheless, some of the characteristic features of the Grossman model may be found also here, such as the idea that health is a consumption characteristic which must be produced by the household on the basis of commodities bought in the market as well as the time given up by the household.

An example of this kind of models is provided by Phelps og Newhouse [1974], where the health expenditure C of a consumer is supposed to consist of direct outlays to the amount of cP, where P is the price of health care services and c the part of this price left for the consumer (the rest is covered by an insurance or another scheme for health care financing), and a time use t which may be evaluated in money at the price W, which expresses the value of one hour for a consumer (as previously possibly corresponding the hourly wage which the consumer might receive if working). Altogether we have a budget constraint C = cP + Wt

(1)

for health expenditures. This is not a standard budget constraint, since the amount C may depend on the remaining choices of the consumer as well as on the income, but even though overly simplified the model may still be used to illustrate some peculiarities of the demand for health care services: First of all we see that given (1) it is conceivable that the demand for health care may depend only weakly on prices, since the consumer pays only the part c, and â&#x20AC;&#x201C; what is more important â&#x20AC;&#x201C; the substitution effect of a change in price may lead to a considerable shift in the time use and only a minor change in the demand for health care.

time

time

health care Figure 3. The Newhouse-Phelps model: In the figure at the left there is a usual budget line corresponding to the expression (1), which however is quite sharply sloped since

10

health care

the share c is low and time is considered as very important. In the figure to the right the budget constraint shown is one arising from a scheme for reimbursement of health expenditure where the percentage reimbursed increases after a certain level of expenditure has been reached.

Secondly, the share which is reimbursed by the health insurance scheme, will often vary with the size of the consumption. Typically, there is a certain initial use which is not refunded at all or where the reimbursement is very small, whereas further consumption is reimbursed at a higher rate (this corresponds to the recently â&#x20AC;&#x201C; March 2000 â&#x20AC;&#x201C; introduced scheme for reimbursement of outlays for medicin, where the first Dkr.500 must be paid by the individual, whereafter some reimbursement occurs). The budget constraint corresponding to (1) will no longer be a straight line but will display a kink at the level where the higher reimbursement rate comes into display. This is illustrated in Figure 3.

The kinked budget line gives rise to a particular pattern of behavior with the consumer, obviously depending on the shape of the indifference curves. If the indifference curves are smoothly curved, the consumer will never have her optimum in the kink, but rather at some distance to the right or to the left; it may even happen that there are two optima on the same indifference situated at each side of the kink. If the curvature of the indifference curves is small enough, the optimum will be rather far away from the kink point, meaning that consumers will either use very little health care or conversely use very much, but that there will be few or no consumers close to the consumption given by the kink point.

It may be debatable whether much can be learned from this model which was not known at the outset, but at least the model gives somw useful hints about the adjustments which will necessarily follow from the introduction of a particular rule for own payment versus reimbursement; it may as suggested by the model result in a polarization of behavior such that some consumers reduce their use of health care while other increase it.

In some cases, the reimbursement depends on consumption also in previous periods, negatively if a heavy use of the insurance means that the coverage will be reduced, or positively if the reimbursement comes into function only after a certain cumulated exoenditure.

In such situations a dynamic approach is needed: Increased expenditure today will entail either smaller or greater expenditures in future periods since todays expenditure is added to the cumulated use of the insurance. We shall not develop such models (which anyway are not very much developed in the literature; the basic reasoning, namely that consumption should be seen in close connection with the health plan of the consumer, is important, and we shall return to this again and again.

5. On the elasticity of demand for health care It may be argued that the demand for the services of the health care sector is determined directly by an underlying physical need for treatment rahter than by prices as discussed hitherto. The basis for such an argumentation is that a service often has a definite and very limited character; an individual suffering from a particular disease usually needs particular and well-defined treatments in a similarly well-defined quantity (visits to the doctor, treatment in hospital, medicin etc.). If the price of this well-defined treatment is raised or 11

lowered, it only means that the consumer will have to part with a greater or smaller part of her consumer surplus, possibly even all of it. In the worst case the price may be so high that it deters consumption, but as long as this has not happened, the same service will be delivered. This means that demand is inelastic up to a certain limit.

If this can be shown to be the case, it would mean that the classical argument against providing the services free of charge to the consumers would lose power. According to this argument, goods and services delivered free of charge will be demanded and used in surplus of what was economically rational (in terms of the consumersâ&#x20AC;&#x2122; own preferences) as illustrated in Figure 4. If the consumer has a decreasing demand curve and the price is 0, then demand will surpass the level where the corresponding price equals marginal cost. This in its turn means that services are delivered beyond the point where the cost for society of delivering them balances the value for the individual of receiving them, so that the allocation is inefficient.

Price

Price

quantity

quantity

Figure 4. Inelastic demand: In the figure to the left we have the typical inelastic demand for treatment of an acute disease. The quantity of treatment demanded is the same at all prices up to a certain level, where the patient can no longer afford to be treated.

In the figure to the right the demand of the patient consists of several distinct demands for the service, corresponding to all the diseases from which the patient may suffer at a given time. While each of these demands fits with the figure to the left, the total demand for service will be more elastic.

If demand is inelastic, so that the demand curve is horizontal, then this argument is clearly of no impact, and in this case the idea of providing services free of charge looks attractive. However, the argument is based on the assumption that the services provided by the health care sector are distinct and well-defined in relation to the need of the individual.

The important part of this is that the quantity demanded should be something determined alreade by medical considerations independent of any economic data. As long as these assumptions hold, the principle of free delivery is welfare superior to selling in a market, at least if welfare depends on equality in health, since too high a price may leave some people untreated.

12

However, som care must be taken; it may well be the case that the demand for treatment for a particular disease is inelastic, but treatment as a whole will involve several different types of treatment, corresponding to different diseases, and since the latter may be of greater and smaller importance for the everyday situation of the individual, a certain sensitivity to prices will result: If a leg is broken, not many options are left open with respect to treatment, but it is less clear whether we want to go to the dentist every time we have some unpleasant sensations in the teeth.

To this should be added the time and uncertainty aspects of demand. There is a marked difference between the demand which stems from a disease which is present on one side and the demand connected with a health insurance on the other side. The considerations related to insurance and health service of non-insurance type (as the Danish) are postponed to a more thorough treatment in Chapter 6, but it should be noticed that a discussion of the elasticity of demand for health with respect to price cannot be separated from the discussion of the methods for financing health expenditure.

6. Empirical investigations The dependence of demand for health care on price is a matter in which the public holds some interest, since the debate on privatization of the latter years has also touched upon health care. Consequently the rather few empirical investigations which have been made into thus topic have been commented at length in the literature. As it is often not simple to get a clear answer if a too detailed theory is taken as the starting point, most of the investigations have been restricted to observing the dependence of demand on a few key variables, typically price and income. To this should be added the important questions about the dependence on the health insurance system covering (part of) the expenses of the individual – questions which clearly can be answered only by observing behavior in countries where there is a possibility of choice between such schemes, such as in USA. The influence of income: There have been several investigations of the dependence of demand on income. The common feature of these investigations is that one finds income elasticities below 1 – when income grows by 1% , health care expenditures grow by less than 1%. This common result of the empirical investigation does however cover a considerable variation (from 0,17 to 0,85), so it is a question how much should be put into the coincidence of the findings with respect to income elasticities (Rosett og Huang [1973]). The differences in observed result is partly due to variances in the methods of registrating demand and in particular demand for health care – we are back in the old problem of measuring health and determining when an expenditure is health care and when it is ordinary consumption, since this “ordinary consumption” will have an effect on health as well. Also differences in method are behind the differences in results.

It is also a question whether the empirical findings should be interpreted in the direct sense, so that an average family is expected to use a steadily diminishing share of its income on health, or whether it is not just an expression of the not particularly surprising fact that the more wealthy families are less subject to illness due to more healthy conditions of living and working, and a more healthy life style in general. As we have seen, this is an aspect of reality which may well be captured by the Grossman model, where the long perspective in the 13

health related choices of the consumer may lead to large expenditures in some periods and considerably smaller expenditures in others, while the empirical investigations necessarily have to consider the expenditureover a shorter period, since it is usually too complicated and costly to follow the individual family over several years (an exception from this is the RAND experiment to be commented upon below).

A hint that the connection between income and health care expenditures in the family may be more complex than what is shown by the direct empirical investigations, is obtained from an investigation of expenditure on hospitals in France (Mizrahi og Mizrahi [1982]; there is an element of user payment in hospital admissions in the French system â&#x20AC;&#x201C; as in many other European countries though not in Denmark). Here it turns out that the expenditure decreases with family income, but at a closer sight a certain regularity is observed: There are not fewer admissions, but the number of hospital days is markedly smaller at each admission for the more wealthy than for the poor families. Again there are several possible explanantions: It is to be expected that the wealthier due to their life style will have another pattern of diseases, and in particular they will more often need short and quick treatments (by-pass operations) rather than the poor who may suffer from more chronic diseases (rheumatisms, tuberculosis etc.). However, the observed pattern may also be an effect of considerations of offer cost â&#x20AC;&#x201C; the loss for the wealthy of lying in a hospital bed rather than earning money outside is larger than that of the poor (a consideration to which we will return later).

It may be attempted to estimate income elasticities not only using national data, but also using cross sections of countries, and this has been done starting with Newhouse [1977].

When comparisons are made in this way, one does get income elasticities above 1, so that individuals in rich countries on the average use a larger percentage of their income on health than those of poor countries. This fits rather well with the first impression obtained from the the aggregate data on percentage of GNP spent on healthcare â&#x20AC;&#x201C; rich countries use a large part, poor countries a smaller part of their GNP on healthcare expenditure.

The influence of social security and price of healthcare: The formal models of demand for health have been based on the assumption that a price has to be paid by the individual for the healthcare received and for the preventive healthcare undertaken. This may however fit rather badly with the realities of most countries, a circumstance which must be taken into account in connection with the empirical verification and estimation of the relationships.

Most citizens of the countries, where empirical research has been carried out, have some kind of health insurance arrangement, and though the insurance schemes can vary rather much, a common element will be that a considerable part of healthcare expenditure is not paid by the patient herself. This means that direct measurements of price elasticities cannot readily be maid, and in a larger perspective one needs to include that demand depends on the particular type of health insurance scheme that covers the individual investigated.

A classical investigation of the importance of the insurance scheme was carried through in California by Helms e.a. [1978], who recorded the effects of the introduction of a user cost for the poorest part of the population to the amound of $ 1 per visit to the doctor. It turned out that the number of such visits decreased by 8% , while hospitalizations (which were not included in the new rules about user payments) increased by 17% . It cannot automatically be concluded that the fall in the use of services which were covered by user payments induced the increase in the visits to hospitals and emergency wards, but at least it is a plausible 14

hypothesis. Another conclusion of the investigation was that the new arrangement had led to an increase in the total expenditure on healthcare, since the costly hospital treatments had increased more than the not so costly treatments by the doctor had fallen. Independently of how user payments are viewed in a larger perspective, it is clear that they should at least be constructed in such a way that they do not induce unwanted substitutions.

In countries with healthcare systems that allow for a choice between different levels of coverage, there is a possibility of studying the variation of health expenditure with coverage, and as expected on finds that the expenditure is considerably higher for schemes with high levels of coverage than for those with lower level (in France the yearly expenditure for the schemes with highest coverage will be around 3 times as high, cf. Labourdette [1988]).

Again there is no easy conclusion to make from this, since the choice of coverage is in itself conditioned by the general situation of the family (young and healthy choosing low coverage, elderly choosing higher coverage), so that what is observed may be a result of the choices rather than an economic reaction on the given insurance scheme. It may be added that the French data also have shown that higher use of healthcare is not uniform over all types of medical treatments; for example is the average consumption of medical drugs not greater for people with 100% coverage than in the other groups (Jeunemaitre og Dumez [1986]).

The RAND health insurance experiment. A particularly thorough investigation of the influence of the insurance scheme on healthcare expenditure was carried out in the US over the years 1974 – 82. A group of individuals, around 7000, who were not covered by the public healthcare insurance scheme Medicare for persons aged above 62, were randomly distributed among 5 specified health insurance schemes, and these persons were then followed over 3 – 5 years, with information collected about their health and their use of the scheme.

The five schemes were as follows: Table 1. Health insurance schemes in the RAND experiment Scheme

Deductible

Full coverage from:

Ceiling over user payment

1 2 3 4 5

none (free healthcare) 25% 50% 95% 95% of outpatient 0% of hospitalization

5–15% 5–15% 5–15% none

$ 1000 $ 1000 $ 1000 $ 150 per person or $ 450 per family

Over the period, there was a certain dropout, so that the final sample has a unequal representation of the five groups (most in 1, fewest in 3).

The main results of the investigation are summarized in Table 2. As a whole the tendency is rather clear, pointing to a positive correlation between higher coverage and more intensive use of the services provided. There is however some variation within the categories (which hare are shown only in aggregated form and after a selection). A rather important extra information for the assessment of the experiment is whether one could find a change in health status as a result of being covered by one or another scheme, but this did not show up in the data which was obtained from the experiment.

15

Table 2. Main results from the RAND health insurance experiment, as percentage of those of scheme 1.

Ydelsestype

1 (absolute numbers)

1. Outpatient treatm.

4.55 cost $ 340 2. Hospitalization 0,128 cost $ 409 3. Prescription medicin 5,43 cost $ 60,1 4. Visits to emergency ward 0,304 cost $ 32

1

2

3

4

5

100 73 67 60 60 100 76 66 60 69 100 82 72 77 90 100 91 110 77 91 100 82 80 67 79 100 76 60 57 73 100 79 82 65 80 100 86 91 70 82

Kilde: Søgaard [1990]

Care should be taken that the results are not unduly over-interpreted, even though at the face of it they show a considerable price sensitivity in the demand for healthcare. There is nothing wrong with the result as such, but it must be used in the proper context rather than as a general message to the effect that user payments have a direct effect on the demand for and the cost of healthcare. As we shall see later – in connection with the discussion of health insurance – there may well be an effect but it is more sophisticated, working through the choice of insurance scheme combining insurance premium and user payments. We shall return to this topic at several instance later.

The influence of time: We have already in connection with the Newhouse-Phelps model discussed the importance of time in the consumer’s optimization problem. This aspect has been made the object of an empirical investigation by Coffey [1983], who compared number of visits to the doctor with assessment of time, the latter measured by opportunity cost in the form of hourly wage (for employed; for unemployed he used an estimated reservation wage). It turned out that this assessment of time had a negative influence on the probability of a visit (10% increase in hourly wages resulted inl 15$ reduction of visits). On the other hand it had no effect on the total number of visits given that a treatment was initiated. Also the investigation showed a markedly weaker reaction for publicly employed (where it was down at 6%).

7. Supplier-induced demand A recurrent theme in the debate about healthcare provision is that connected with demand created by the provider, the supplier-induced demand. This is based on the idea that the medial profession has a particularly easy access to the creation of demand for their own products. At the outset, there are several other professions who have this option, typically those engaged in consultancy of some sort, but in the public debate it has mainly been connected to the medical profession. It is not easy to establish empirical foundation of a theory of supplier-induced demand, even though many attempts have been made over time.

What is investigated is typically the connection between density of practitioning doctors 16

and demand for consultations, but such investigations remain rather inconclusive since the causation may well have the opposite direction.

There have also been given many theoretical explanations of the way in which healthcare providers, in particularly doctors and hospitals, can create new demand. One of the most frequent refers to the principal-agent relationship between patient and doctor, where the patient must leave the judgement about character and seriousness of the disease from which she is suffering – and as a consequence the amount of treatment necessary – to the doctor. We shall have more to say about principal-agent relationships later, and indeed a discussion of supplier-induced demand falls more naturally under a supplier behavior than under demand. Instead we take a brief look at a classical model which based on a rather simple considerations derive a kind of supplier-oriented demand. A a closer view, it is perhaps rather a demand based on the prices of the suppliers, depending on the latter in an entirely nonstandard way. The model was proposed by Feldstein [1981].

The point of departure is the US healthcare system with hospitals that are private but not profit maximizing; typically, hospital i will charge a price pi for its treatment, which may not cover all the cost ci of this treatment at hospital i; the remaining part of the cost is covered by a subsidy di from the “owners”, who may be a university of a philantropic institution.

We use the drastically simplifying assumption that the hospital has only one treatment with one associated price (which anyway will be needed when the model is used for empirical estimation). The hospital direction are responsible for the cost which adds up to the amount ci , and they are considered as consisting of remuneration of the input li of labor and ki of materials together with overheads.

The main point is now, that in a non-profit-maximizing hospital the direction will try to maximize some indicator of quality ui (l, k), depending on the input of labor and other productive factors. The general economic constraints ci = pi + di , namely that cost must be equal to the income that the hospital can get either from patients or from owners, becomes a constraint in the quality optimization problem of the direction (where pi is a quantity controlled by the direction); the outcome of the optimization is a choice which depends on external factors (namely wages w prices v on other inputs, together with the subsidy di ). Out of this one gets a price pi (w, v, di ) as well as an optimal allocation of productive factors. produktionsfaktorer.

Turning next to the users of the hospitals, then each of these have a conception about the way in which quality is produced, formalized in the same way as that of the hospital direction, namely as a subjective indicator w(k, l) of quality, and therefore the user will direct her attention to the hospital which according to the available information gets closest to maximizing w. However, the inputs cannot be directly observed, but wich knowledge of w, v and all the di the user can find her own p∗i (w, v, di ) for each hospital, and if this p∗i looks like the actual price (which is the result of the optimization by the i-direction), this coincidence is taken as a signal that the hospital is a good choice.

The consequence is that the prices charged by the hospitals becomes an informational signal, which in its turn triggers demand, since price has become an indicator of quality. In particular this means that demand will vary with price in a way totally different from what 17

is observed for ordinary consumer goods; a higher price can be read as a signal of higher quality by the users. The main message of the model is that there may be rational behavior on both sides of the market behind what at a first sight is seen as supplier-induced demand; it is not necessarily caused by greedy providers. kvalitet. In particular, this means that demand varies with price in a way which differs much from what is known for ordinary consumption goods. A higher price may be read as a signal of superior quality from the point of view of the user. We shall not pursue this idea any further at this point (and Feldstein introduced the model only as a step towards empirical analysis of price and demand). The main message of the model is that there may be a rational core in what seems to be only an idea in search for a theory (supplier-induced demand), but possibly it has to be found in another context and point to other policy measures.

8. References Feldstein,M.S. [1981], Hospital costs and health insurance, Harvard University Press, Cambridge, Mass.

Grossman, M. [1972], On the concept of health capital and the demand for health, Journal of Political Economy 80.

Jeumaitre, A., H. Dumez [1986], Controverses autour d’une d´er`eglementation des prix des m´edicaments, Revue francaise des Affaires sociales, 176 – 190.

Mizrahi,A., A. Mizrahi [1982], Consommation m´edicale et travail f´eminin, Consommation, 1.

Newhouse, J.P., C.E. Phelps [1974], Coinsurance, the price of demand and the demand for medical services, Revue of Economics and Statistics 56.

Rosett, R.W., L. Huang [1972], The effect of health insurance on the demand for medical care, Journal of Political Economy 81.

Søgaard, J. [1990], Sygeforsikringseksperiment i USA: Efterspørgselsreaktioner og sundhedskonsekvenser, i Andersen,P., T. Christiansen (red.): Brugerbetaling i sundhedssektoren, teori, viden, holdninger, Odense Universitetsforlag, Odense, 92–121.

18