The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Catching up with the Keynesians Lars Ljungqvist and Harald Uhlig Working Papers Series Macroeconomic Issues Research Departm ent Federal Reserve Bank of Chicago October (W P -96-15) ■...» M m Hn FEDERAL RESERVE B A N K OF CHICAGO Catching u p with the Keynesians* Laxs Ljungqvist Re s e a r c h D e p a r t m e n t Federal Reserve B a n k of C h i c a g o Chicago, IL 6 0 6 9 0 USA ljimg@frbchi.org and H a r a l d Uhlig C e n t E R for E c o n o m i c R e s e a r c h Tilburg University Postbus 90153 5 0 0 0 L E Tilburg THE NETHERLANDS e-mail: uhlig@kub.nl September 1996 *The views expressed herein are those of the authors and not necessarily those of the Federal Reserve Bank of Chicago or the Federal Reserve System. A bstract This paper examines the role for tax policies in productivity-shock driven economies with “catching-up-with-the-Joneses” utility func tions. The optimal tax policy is shown to affect the economy countercyclically via procyclical taxes, i.e., “cooling down” the economy with higher taxes when it is “overheating” in booms and “stimulating” the economy with lower taxes in recessions to keep consumption up. Thus, models with catching-up-with-the-Joneses utility functions call for traditional Keynesian demand management policies. Parameter values from Campbell and Cochrane (1995) are also used to illustrate that the necessary labor taxes can be very high, in the order of 50 percent. However, Campbell and Cochrane’s nonlinear version of the aspiration level in the catching-up-with-the-Joneses preferences has the additional implication that consumption bunching can be welfare enhancing. 1 Introduction Envy is one important motive of human behavior. In macroeconomics, the ories built on envy have been used in trying to explain the equity premium puzzle as described by Mehra and Prescott (1985). Abel (1990) and most re cently Campbell and Cochrane (1995) postulate utilityfunctions exhibiting a desire to “catch up with the Joneses”,i.e.,ifothers consume more today, you yourself will experience a higher marginal utility from an additional unit of consumption in the future.1 In some ways, the idea of “catching up with the Joneses” is a variation of the theme of “habit formation”,see Constantinides (1990). The key difference is that “catching up with the Joneses” postulates a consumption externality since agents who increase their consumption do not take into account their effect on the aggregate desire by other agents to “catch up”. Thus, this externality allows room for beneficial government intervention. The optimal tax policy would induce agents in the competitive equilibrium to behave in a first-best manner, which is given by the solution to a social planner’s problem with habit formation. While “catching up with the Joneses” has been the focus of quite some research in the asset pricing literature, its implications with respect to pol icy making have rarely been explored. The purpose of this paper is to do exactly that. In particular, we examine economies driven by productivity shocks where agents care about consumption as well as leisure, and there is a “catching-up” term in the consumption part of the utility function. For simplicity, the model abstracts from capital formation.23 In this framework, we examine the role for taxing labor income. The optimal tax policy turns 1Gall (1994) explores an alternative assumption where agents’ preferences depend on current instead of lagged per capita consumption ( “keeping up with the Joneses” as com pared to “catching up with the Joneses”). 3As noted by Lettau and Uhlig (1995), the inclusion of capital formation in models based on catching-up-with-the-Joneses utility functions have the implication that con sumption becomes excessively smooth. For a similar observation and a possible remedy in models with habit formation, see Boldrin, Christiano and Fisher (1995). 1 out to affect the economy countercyclically via procyclical taxes, i.e., “cool ing” down the economy with higher taxes when it is “overheating” due to a positive productivity shock. The explanation is that agents would other wise end up consuming too much in boom times since they are not taking into account the “addiction effect” of a higher consumption level. In re cessions, the effect goes the other way around and taxes should be lowered to “stimulate” the economy by bolstering consumption. Thus, models with catching-up-with-the-Joneses utility functions call for traditional Keynesian demand management policies. W e also use parameter values from Camp bell and Cochrane (1995) to illustrate that the necessary labor taxes can be very high, in the order of 50 percent. However, Campbell and Cochrane’s nonlinear version of the aspiration level in the catching-up-with-the-Joneses preferences has the additional implication that consumption bunching can be welfare enhancing. As an example, we show how welfare can be improved upon in their framework by inducing business cycles in an otherwise station ary environment. The paper is organized as follows. In section 2, we examine a simple oneshot model as well as an infinite horizon version, where agents care about “keeping up with the Joneses”. The assumption being that contemporaneous average consumption across all agents enters the utility function. In that case, it turns out that there is a constant tax rate on labor, which delivers the first best outcome independent of the productivity shock. In section 3, we allow the agents’aspiration level to be a geometric average of p a s t per-capita consumption, i.e., specifying a utility function which exhibits “catching up with the Joneses”. This framework has Keynesian-style countercyclical policy implications. In section 4, we examine the utility function used by Campbell and Cochrane (1995), adding another layer of complexity. Here, we are in particular interested in the quantitative tax implications of their parameter values. Section 5 concludes. 2 2 K eeping up w ith th e Joneses W e imagine an economy with many consumers, each with the same utility function 1 - 7 where c > 0 is the individual’s consumption, C > 0 is average consump tion across all agents and n > 0 is labor supplied by the individual. The parameters oc G [0,1), 7 > 0 and A > 0 determine the relative importance of average consumption, the curvature of the consumption term and the relative importance of leisure. This utility function captures the notion of “keeping-up-with-the-Joneses”,i.e., average consumption decreases an indi vidual’s level of utility and increases his marginal utility of an additional unit of consumption. This specification is different from the formulation in Abel (1990), who uses ratios rather than differences to aggregate consumption, but is in line with the catching-up formulation in Campbell and Cochrane (1995). No “keeping-up” is imposed on the leisure part of the utility func tion. In other words, we assume that agents are competing in, say, having the biggest car or the biggest house rather than having the most amount of leisure. The utility in leisure is also assumed to be linear. This assumption is partly done for convenience, but can also be motivated by indivisibilities in the labor market and is an often used assumption in the real-business cycle literature, see e.g. Hansen (1985) and the explanations therein. W e imagine that the production function takes the form c = On, where 6 is a productivity parameter. Thus, there is no capital, and output is simply linear in labor. The government levies a flat tax r on all labor income and the tax rev enues are then handed back to the agents in a lump-sum fashion. Let v be the lump-sum transfer to each agent. Since all agents are identical, the 3 government’s budget constraint can be written as = rd n v . A competitive equilibrium is calculated by having an agent maximize the utilityfunction above with respect to c and n subject to his budget constraint, c = (1 — A r)6 n + v. consumer’s optimal consumption is then found to be c= aC+^(l-r)^ where average consumption C , (1 ) is taken as given by the individual agent. However, in an equilibrium it must be true that c = C , so the equilibrium consumption level is 1 c = C = 1 — a The government’s optimal choice of r can be deduced from the solution to the social planner’s problem. The social planner would take the exter nality into account be setting C in the utility function above, and then = c maximize with respect to consumption and labor subject to the technology constraint. The first-best outcome is then given by c* = C Comparing the social planner’s solution to the competitive equilibrium, we find: Proposition 1 (“Keeping up with Joneses”) T h e fir s t - b e s t c o n s u m p t io n a llo c a t io n c a n be a c h ie v e d w ith a ta x ra t e t = 4 a . This result is quite intuitive. A fraction a of any increase in the representative agent’s consumption does not contribute to his utility since itisoffset through the consumption externality. It is therefore socially optimal to tax away a fraction or of any labor income so that the agent faces the correct utility tradeoff between leisure and consumption. It can also be noted that the optimal tax is independent of the productivity parameter 8. While the tax can potentially be high depending on the value of a, it does not react to current economic conditions. In particular, we do not get any Keynesian effects in the sense of setting taxes procyclically. Given the solution above, one can easily examine a dynamic model, in which there are periods denoted by t = 0,1,2,... and agents have the utility function '(c . - q C,)1-1't= 0 where E q \ - - , 1-7 is the expectation operator conditioned upon information at time 0 and /? € (0,1) is a discount factor. The production function is the same as before, and so are the budget constraints of the government and the agents. There isnow also some stochastic process driving productivity 8t . Computing the competitive equilibrium and the social planner’s solution amounts to the same calculations as above, since this dynamic model simply breaks into a sequence of one-shot models. The first-best solution is again achieved at r = a, i.e., there are no cyclical consequences for the tax rate. Finally, itisworth pointing out that the tax analysis here is closely related to the literature on redistributive taxation when individual welfare depends on relative income. Given a social welfare function, Boskin and Sheshinski (1978) analyze how the standard results of optimal tax theory axe altered when individuals care about relative income, and they demonstrate that the scope for redistribution becomes much larger. Persson (1995) extends their argument by showing that high taxation can even constitute a Pareto improvement as long as individuals’ pre-tax incomes are not too different. In fact, his discussion of the special case of identical individuals corresponds 5 directly to our treatment of “keeping up with the Joneses”. 3 C atching up w ith th e Joneses 3.1 T he m od el W e now assume that the utility function does not depend on current aver age consumption as assumed above, but rather on some measure X t of p a s t average consumption, (c* t=0 \ x tr * * - 1 -7 In particular, we let the aspiration level X t - - . (2 ) be a geometric average of past per-capita consumption levels, Xt with 0 < ^ < 1 and 0 < a = a ( l - <j>)Ct-, + , (3) < 1. Otherwise, the production technology and the budget constraints of the consumers and the government are the same as before. In addition, we now need to.be more careful about the productivity process. W e postulate the following stochastic process, where xp 6 [0,1) and et is i.i.d, has mean zero and is bounded below by €t > — 1 .®3 3The stochastic process (4) is approximately the same as postulating an AR(1) process for the logarithm of 9t , log(tft) = (1 - V1)log(0) + xj>log(0<—i)+ c,. Thus, our exact analytical results below pertaining to the stochastic process (4) can also be interpreted as approximations to the corresponding formulas valid for the more commonly used AR(1) process for the logarithm of 9t . 6 For the competitive equilibrium in this model, one finds analogously to (1) that the agent will set consumption equal to Thus, given a first-best path for consumption c\ = C*, one can achieve this outcome with a sequence of taxes rt satisfying n = i - j ' ( c ; - x ly. (6) To characterize the optimal tax policy, we now turn to the social planner’s problem. 3.2 S o l v i n g t h e social p l a n n e r ’s p r o b l e m The social planner maximizes the utility function (2) subject to the produc tion technology and the constraint (3), taking as given the process for and the initial conditions X q and 0q . $t Since this maximization problem is a concave one, we can analyze it by using first-order conditions. Let Xt be the Lagrange multiplier for the constraint (3). The two first-order conditions with respect to C t and X t+ i can then be written as ( C .- X ,) " 1 = £ + «(l-«A,, (7) A, = W [ ( C , „ - A , t, r ] + « [ A , til- (8) The first equation contains the additional third term a(l — <^)Atas compared to the corresponding equation of the private agent’s optimization problem. Here, the social planner takes into account the “bad” effect on future utility of additional aggregate consumption today, since itraises the aspiration level X t+ i tomorrow and beyond. In particular, a fraction a ( l — <f>) of an increase in today’s per-capita consumption spills over to -X’i+i,and the shadow value of a higher X t + i is given by A*. Equation (8) shows in turn how the shadow 7 value A* is the sum of the expected effect on tomorrow’s discounted marginal utility of consumption and its impact on still future periods. The latter effect is captured by the discounted expected value of At+i multiplied by where $ <f>, is the fraction of the aspiration level that carries over between two consecutive periods. Using the two first-order conditions (7) and (8) as well as the constraint (3), the steady-state consumption level can be calculated to be C* = 1 1— a Comparing this expression to the agent’s consumption rule in equation (5) and noting that X — a C , we see that the first-best steady-state allocation is supported by a tax of - - <t>) For example, ifthe aspiration level is simply a times the level of yesterday’s per-capita consumption (<f> = 0), we get f = a/3. This formula is rather intu itive compared to the simple model above of “keeping up with the Joneses”, where we got r = a. Since the consumption externality now enters the utility function with a one-period lag, the adverse future effect of being “addicted” to today’s consumption is discounted by /? so the optimal steady-state tax rate is also scaled down by /?. In order to characterize the optimal consumption and taxation outside of a steady state, we can actually solve the dynamic equations in closed form. The substitution of equation (7) into (8) yields a first-order difference equation in the shadow value At, which can be solved forward in the usual manner, OO At = p A E t j =0 where 8 1 With the law of motion for 0* in (4), one can then calculate X t to be A ‘ , / W ( l 8 )6 + i - 6$ U M (i - l\ (10) 0) ' After substituting this expression into the first-order condition (7), the opti mal consumption level is found to be Cf = Xt + ( a i- w \7 i - « + /j__ iu U e) - w r' i- 8 x i> ) (ii) The tax necessary to support this optimal consumption allocation is then given by equation (6). Rather than calculating the tax rate rt,it is more appealing to calculate the ratio of taxes to after-tax income. Using equations (6) and (7), we get <*(1 ~ rt 1- <f>) (12) O tX t. A rt With the productivity process in (4), A4 is given by (10) and the tax ratio can then be rewritten as in the following proposition. Proposition 2 (“Catching up with Joneses”) T h e ta x ra te Tt s u p p o r t in g the fir s t - b e s t c o n s u m p t io n a llo c a t io n c a n be s o lv e d fro m Tt = 1 — r< <*/?(! ~ 4>) ( , 1-8 $ V ,1 I- M 8 (13) 6 ) ' w ith a s t e a d y - s ta t e v a lu e o f - _ aft(i T 3.3 l- f i* 4) ‘ Tax p o licy im p lication s Corollary 1 (“Catching up with Joneses”) T h e o p t im a l ta x p o l ic y a ffe c ts th e e c o n o m y c o u n t e r c y c lic a lly v ia p r o c y c lic a l ta x e s. 9 This corollary follows directly from equation (13), the tax ratio (and thus the tax rate itself) varies positively with productivity 0t.4 Thus, we get Keynesian-style policy recommendations. A government that maximizes wel fare should “cool down” the economy during booms via higher taxes because agents would otherwise consume too much as compared to the first-best so lution. Likewise, the government should “stimulate” the economy during recessions by lowering taxes and thereby bolstering consumption. Of course, these optimal fiscal policies axe here driven by a rather unorthodox argu ment. Taxation is needed to offset the externalities associated with private consumption decisions. One individual’s consumption affects the welfare of others through agents’desire to “catch up with the Joneses”. To shed light on how different parameters affect the cyclical variations of optimal taxation, let w t be the relative deviation of the tax ratio r t/ ( l — r t) from its steady-state value. That is, u t tells us, how the ratio of taxes to after-tax income responds to productivity shocks relative to its steady-state value. PYom equation (13), we can calculate n ( f 1 — rt \1 — _ t 1-0 1— ) 8rj> -9 (14) $ Doing comparative statics on this expression, we see that the size of the cycli cal tax effect in absolute terms varies negatively with 0 and positively with a, /? and <f>. The intuition for this is straightforward by considering the tax response to a positive productivity shock. A higher 0, i.e., a more persistent productivity shock, means that future production and consumption oppor tunities are also expected to be better than average. The anticipation of the economy being able to sustain a higher consumption level for a prolonged period of time mitigates the adverse effects of making people “addicted” to 4It is worth noting that this result holds for a much larger class o f stochastic processes than given by equation (4). According to equations (9) and (12), the optim al tax rate goes up with 9t as long as E t [SJLo ^ &7+i + j ] decreases less than proportionally with the inverse o f 0 t - 10 higher consumption today. It is therefore socially optimal to take more ad vantage of a persistent productivity shock, so the optimal tax hike is lower with a higher ip. In contrast, preferences with a higher weight on yesterday’s consumption (a higher a ), a higher degree of persistence in the aspiration level (a higher <f>), or a higher emphasis on the future (a higher ft) give rise to a larger cyclical tax effect. The reason is, of course, that the consumption externality is more important for such preferences and the government must consequently be more resolute in moderating agents’consumption behavior. As a point of reference, the largest tax effect as defined by (14) is attained for transient one-period productivity shocks (ip = 0). The percentage devi ation of the tax ratio from its steady-state value responds then one-for-one to the percentage change in the productivity from its steady state. However, besides noting that the cyclical tax effect can be large relative to the mag nitude of the productivity shock, it is also important to keep in mind that most aggregate economic shocks are usually relatively small so the cyclical tax changes considered here are really examples of extreme “fine tuning” of taxes. Finally, Figure 1 illustrates the consumption dynamics in response to a productivity shock. After a one-percent initial shock to 9 t at time t = 0, the hump-shaped dashed line traces out the response of consumption from the steady state when taxes are adjusted optimally and the solid line displays the consumption response when the tax rate is not changed but kept constant at its steady-state value. As a parameterization, we used ip = 0.9, a = 0.8, (3 = 0.97 and varied 7 € {0.5,1.5}. Not surprisingly, the consumption response becomes muted with a higher 7 ,since a more rapidly diminishing marginal utility of consumption reduces the attractiveness of increasing consumption. It is interesting to note that for both values of 7 in Figure 1 the deviation of consumption from steady state is reduced by around 25 % under optimal tax adjustment as compared to keeping the tax rate constant at its steadystate value. The figure also contain the change in the tax ratio u>t needed to accomplish this “cooling down” of the economy. 11 PanalA: Gamma-0.5 Panel B: Gamma -1.5 Figure 1: Consumption dynamics in response to a one-percent productivity shock from the steady state. The dash-dotted line depicts the optimal re sponse in the tax ratio rt/(l— rt). The parameters are 7 € {0.5,1.5} (panel A and panel B, respectively), ^ = 0.9, a = 0.8, and 12 0 = 0.97. 4 T h e C a m p b e l l - C o c h r a n e utility f u n c t i o n W e now turn to the utility function proposed by Campbell and Cochrane (1995) extended with a linear disutility term for labor. These preferences are then also given by our expression (2), but the aspiration level X t is now a complex nonlinear function of current and past per-capita consumption as shown below.56 A useful concept when studying this model is the “surplus consumption ratio” defined for an individual as < h -X t st = ----- , Ct and the upper case letter S t will be used to denote the economy-wide value of s t. In an equilibrium, S t will of course be equal to st since all agents are identical. Campbell and Cochrane postulate an implicit law of motion for X t by A writing a law of motion for St- Let St = log S t be the logarithm of the economy-wide surplus consumption ratio, and likewise C t = log C t is the log arithm of average consumption across all agents. Abstracting from economic growth, it is then assumed that S t evolves according to st where S = (1 - <t>) log S + <f>St-t + A(5t_0 (C t - C t- i j , (15) is the steady-state value of S , and the function A (S') is given by A( S ) -{ S - 'y / l -2(5 — log 5) — 1, 0, S < S „ (16) S _ > S^max , 5Our notation differs from Campbell and Cochrane (1995) in order to stay consistent with the notation above. In particular, we use hats rather than sm all letters to denote logs, while sm all letters still denote the individual’s choice variables. We use instead of for the discount factor and we abstract from growth, i.e., their parameter g is here set equal to zero. 6 13 with Sm ax = log 5 + (l — S2) /2.6 Given equation (15), one can back out the implied law of motion for X t. imation shows that the log of X t Near the steady state, a log-linear approx is a moving average of past consumption in logs and it does not depend on contemporaneous consumption, x t = iog(i - s ) + (i - <f>) £; p e t - ;.! . i —0 For our purposes, the steady-value S can be thought of as a parameter in this model. By picking a value of S , we are effectively choosing a particular preference specification.7 Taking X t as given, the agent maximizes utility subject to the usual budget constraint. Analogously to the previous sections, the agent’s optimal consumption is found to be °t = X t + ^ ( 1 — rt)^ , which can also usefully be written as (^ “(T=W (17) Instead of solving the social planner’s problem, we now turn to a more modest question. In a steady state, we ask what tax rate is needed to support the best possible constant consumption level. (The word ‘constant’will soon be shown to be restrictive in terms of maximizing welfare). 6The purpose of Campbell and Cochrane’s rather complicated preference specification is to assure that c* — X t> 0, and that the risk-free rate is constant when is a random walk with drift. 7To understand the correspondence between S and the preference specification, let us consider the model in Section 3 where S = (<?— X )/ C = (C —aC )/C = 1 — a. That is, S maps directly into a and is unaffected by the tax rate. (The only exception being a 100 % tax rate which would close down all economic activity, and the surplus consumption ratio would no longer be defined.) 14 Proposition 3 (Campbell-Cochrane) I n a s t e a d y s t a te w ith 7 > 1, th e re e x is ts a u n iq u e c o n s u m p t io n le v e l t h a t c a n n o t be im p r o v e d u p o n th ro u g h a o n c e - a n d - f o r - a ll c h a n g e to a n o t h e r c o n s u m p t io n le v e l. T h e s t e a d y - s ta t e t a x ra te s u p p o r t in g t h is best p o s s ib le c o n s t a n t c o n s u m p t io n le v e l is g iv e n by (l-S). i - M The derivation of this proposition is deferred to the appendix, and here we only note that the tax rate is the same the steady-state tax rate for the as usual linear version of the aspiration level in Proposition 2. To see this, we only have to use the observation in footnote 7 which is that the parameter a and the steady state surplus consumption ratio satisfy the relationship S = 1 — a. It is also interesting to take a look at the quantitative tax implication of Proposition 3. Campbell and Cochrane use the parameters <j>= 0.97 and S 0 = 0.973, = 0.049. For these parameters, we obtain f = 0.494, i.e., almost 50 percent of labor income should be taxed away in steady state in order to support the best possible constant consumption level. Taken seriously, this would indicate that current labor taxes are too low in the United States, but about right in, say, the Netherlands or Sweden. Finally, we have refrained from using the word ‘first-best outcome’sim ply since we have not presented the optimal solution to the social planner’s problem. The nonconcave character of this maximization problem makes it analytically intractable, so here we rather use an example to demonstrate that consumption bunching can improve upon a constant consumption allocation. In particular, Figure 2 explores the welfare consequences of a temporary oneperiod increase in consumption starting from a steady state with the best possible constant consumption level C , 15 as described in Proposition 3. The x-axis in Figure 2 shows the size of the one-period consumption deviation as percentage of C , and the y-axis depicts the life-time utility associated with that policy. It is clear from the figure that there are one-time consumption deviations that can increase life-time utility. The intuition is that a tempo rary consumption increase acts as an ‘investment’in the surplus consumption ratio S because of the persistence parameter <f> in equation (15). But let us first consider what happens in the first period when the consumption hike takes place. Equation (15) shows how the log of S increases by the logdeviation in consumption multiplied by the steady-state value of the A-function. Since the A-function is decreasing in the surplus consumption ratio, it follows that the negative impact on S is smaller in the next period when consump tion reverts back to C . In fact, there is no effect at all ifthe log of the surplus consumption ratio has reached S mas in equation (16) when the A-function be comes zero. (This critical point shows up as a kink on the curve in panel B of Figure 2.) In consecutive periods, welfare is positively affected by the slowly decaying surplus consumption ratio (while consumption is kept constant at C ). Concerning the parameterization in Figure 2, we have used Campbell and Cochrane’s values mentioned above and their parameter 7 = 2.372, and we have set A = $ = 1. Figure 2 suggests that a first-best outcome for the Campbell-Cochrane utility function will involve consumption cycles even in an otherwise station ary environment. The social planner would like to exploit the law of motion for the surplus consumption ratio in order to increase the well-being of in dividuals. The rationale for this is that the dynamics of the law of motion for the surplus consumption ratio can be said to exhibit increasing returns to scale. For a related argument on welfare-improving cycles in models with increasing-returns-to-scale production technologies, see Murphy, Shleifer and Vishny (1989). 16 PanelA Panel B Figure 2: Impact on life-time utility of a one-period consumption deviation from the steady state in Proposition 3. Panel B is a magnification of the left-hand portion of panel A. The parameters are /? = 0.973, S = 0.049, 7 = 2.372, and A = 9 = 1. 17 <f> = 0.97 and 5 Conclusions This paper examined the role for tax policies in simple productivity-shock driven economies with “catching-up-with-the-Joneses” utilityfunctions. These utility functions give rise to consumption externalities, but taxation can be used to get back to the first-best solution. The optimal tax policy turns out to affect the economy countercyclically via procyclical taxes. When the economy is “overheating” due to a positive productivity shock, a welfaremaximizing government should raise taxes to “cool down” the economy. Like wise, taxes should be cut in recessions to “stimulate” the economy by bol stering consumption. Thus, models with catching-up-with-the-Joneses utility functions call for traditional Keynesian demand management policies. W e also used parameter values from Campbell and Cochrane (1995) to illustrate that the necessary labor taxes can be very high, in the order of 50 percent. However, Campbell and Cochrane’s nonlinear version of the aspiration level in the catching-up-with-the-Joneses preferences has the additional implica tion that consumption bunching can be welfare enhancing. An example was used to illustrate how welfare can be improved upon in their framework by inducing business cycles in an otherwise stationary environment. 18 Appendix P r o o f o f P r o p o s itio n 3 Let us start our argument from an arbitrary initial steady state with consumption 0 (and a surplus consumption ratio of S ) . At time t consider an alternative future consumption allocation of C t = Co = 0, we for all t > 0. According to equation (15), the associated sequence of surplus consumption ratios can be expressed in log form as So = log S + A(Co — log C ), St = (1 - <j>) log5 + (j>St-i = (1 - ^ ) log5 + tfS o , fort > 1 , where A = A(log S ) = 5 -1 — 1. That is, the sequence of surplus consumption ratios is given by for t > 0. The life-time utility associated with such an alternative consumption alloca tion is For 7 > 1, this expression is concave in C o and the first-order condition with respect to C q is (18) 19 For any initial steady state C , equation (18) can be used to solve for the best C o when we are constrained to only consider once-and-for-all changes in the consumption level. To find the unique constant consumption level that cannot be improved upon in this way, we solve for C in equation (18) such that C o = 0 . The best possible constant consumption level is then found to be (19) To support this consumption allocation in a competitive equilibrium, we first solve for the tax rate in the agent’s first-order condition in equation (17), _ A ( c t s t)7 with a steady-state value of After substituting equation (19) and A = 5 -1 — 1 into this expression, we arrive at the steady-state tax rate in Proposition 3. 20 References [1] Abel, Andrew B. (1990), “Asset Prices under Habit Formation and Catching up with the Joneses,” A m e r i c a n E c o n o m i c R e v ie w 8 0 , 2, 38-42. [2] Boldrin, Michele, Lawrence J. Christiano and Jonas D. M. Fisher (1995), “Asset Pricing Lessons for Modeling Business Cycles,” Federal Reserve Bank of Minneapolis, Working Paper 560. [3] Boskin, Michael J. and Eytan Sheshinski (1978), “Optimal Redistribu tive Taxation when Individual Welfare depends upon Relative Income,” Q u a r t e r ly J o u r n a l o f E c o n o m i c s 92, 589-601. [4] Campbell, John Y. and John H. Cochrane (1995), ”By Force of Habit: A Consumption-Based Explanation of Aggregate Stock Market Behavior,” National Bureau of Economic Research, Working Paper 4995. [5] Constantinides, George M. (1990), “Habit Formation: A Resolution of the Equity Premium Puzzle,” J o u r n a l o f P o lit ic a l E c o n o m y 98, 3, 519- 543. [6] Gali, Jordi (1994), “Keeping Up with the Joneses: Consumption Exter nalities, Portfolio Choice, and Asset Prices,” a n d B a n k in g J o u r n a l o f M o n e y , C r e d it , 26, 1, 1-8. [7] Hansen, Gary D. (1985), "Indivisible Labor and the Business Cycle,” J o u r n a l o f M o n e t a r y E c o n o m ic s 16, 309-327. [8] Lettau, Martin and Harald Uhlig (1995), “Can Habit Formation be Rec onciled with Business Cycle Facts?”, Tilburg University, CentER Dis cussion Paper 9554. [9] Mehra, Rajnish and Edward Prescott (1985), “The Equity Premium Puzzle,” J o u r n a l o f M o n e t a r y E c o n o m ic s 21 15, 145-161. [10] Murphy, Kevin M., Andrei Shleifer and Robert W. Vishny (1989), “Building Blocks of Market Clearing Business Cycle Models,” in N B E R M a c r o e c o n o m ic s A n n u a l 1 9 8 9 ,eds. Olivier Jean Blanchaxd and Stanley Fischer, Cambridge, MA: M IT Press. [11] Persson, Mats (1995), “W h y are Taxes so High in Egalitarian Societies?” S c a n d in a v ia n J o u r n a l o f E c o n o m ic s 22 97, 569-580.