WP\4I jCpq3 POLICY RESEARCH WORKING PAPER 1943 The Implications There is strong empirical evidence that people discount of Hyperbolic Discounting the future hyperbolically, for Project Evaluation applying larger annual discount rates to near-term returns than to returns in the Maureen Cropper distant future. The problem David Laibson with hyperbolic discounting is that it leads to time- inconsistent plans - a person who discounts the future hyperbolically will not carry out the consumption plans he makes today. Hyperbolic discounting provides a rationale for lowering the required rate of return on investment projects but does not justify treating environmental projects differently from other investment projects. The World Bank Development Research Group Environment and Infrastructure July 1998 I POLICY RESEARCH WORKING PAPER 1943 Summary findings The neoclassical theory of project evaluation is based on preferences, plans made in 1998 would not be followecl models in which agents discount the future at a constant - because the low discount rate applied to returns in, exponential rate. But there is strong empirical evidence say, 2020, will become a high discount rate as the year that people discount the future hyperbolically, applying 2020 approaches. larger annual discount rates to near-term returns than to Since it makes sense to analyze only plans that will returns in the distant future. actually be followed, Cropper and Laibson characterize This has led some policymakers to argue that, in the equilibrium of an intertemporal game played by an evaluating programs with benefits spread over decades individual who discounts the future hyperbolically. (such as subway systems and abatement of greenhouse Along an equilibrium consumption path, the individual gases), a low long-term discount rate should be used. In will behave as tlhough he were discounting the future at a fact, some economists have suggested that higher constant exponential rate. The individual's consumption discount rates be applied in the present and lower rates path is, however, Pareto inferior: He would be better off in the future. if he could force himself to consume less and save more. Cropper and Laibson demonstrate that this is This provides a rationale for government subsidization incorrect. of interest rates or, equivalently, lowering the required The problem with hyperbolic discounting is that it rate of return on investment projects. leads to time-inconsistent plans - a person who Although hyperbolic discounting provides a rationale discounts the future hyperbolically will not carry out the for lowering the required rate of return on investment consumption plans he makes today. projects, it does not provide justification for those who Cropper and Laibson note that if social decisionmakers seek to treat environmental projects differently from were to use people's 1998 hyperbolic rates of time other investment projects. This paper - a product of Environment and Infrastructure, Development Research Group - is part of a larger effort in the group to develop benefit-cost methods for environmental decisionmaking. Copies of the paper are available free from the World Bank, 1818 H Street NW, Washington DC 20433. Please contact Anna Maranon, room MC2-533, telephone 202-473-9074, fax 202-522-3230, Internet address amaranon@worldbank.org. Maureen Cropper may be contacted at mcropper@worldbank.org. July 1998. (13 pages) The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the countries they represent. Produced by the Policy Research Dissemination Center The Implications of Hyperbolic Discounting for Project Evaluation MVaureen Cropper David Laibson THE IMPLICATIONS OF HYPERBOLIC DISCOUNTING FOR PROJECT EVALUATION Maureen Crolpper and David Laibson The neoclassical theory of project evaluation (Arrow and Kurz, 1970) is based on models in which agents discount the future at a constant exponential rate. There is, however, strong empirical evidence that people discount the iuture hyperbolically, applying larger annual discount rates to near-term returns than to returns in the distant future (Ainslie, 1992; Cropper, Portney and Aydede, 1994). In this paper we trace out the implications of hyperbolic preferences for private investment choices and public policy The immediate problem posed by hyperbolic discounting is that it leads to time- inconsistent plans: A person who discounts the future hyperbolically will not carry out the consumption plans he makes today. From today's perspective the discount rate between two distant periods, t and t+1, is a long-term low discount rate. But, when period t arrives, the individual will apply a short term high discount rate to consumption in period t+1. Because it makes sense to discuss investment decisions only along consumption paths that will actually be carried out, one must begin by characterizing time-consistent plans for a consumer with hyperbolic preferences. This can be done by allowing the consumer's different temporal selves to play a game and to analyze the equilibrium of this game. In the case of a finite-lived consumer with quasi-hyperbolic preferences, the game has a unique subgame perfect equilibrium, which (as Arrow has conjectured) can be characterized by an Euler equation similar to that in the Ramsey model. The consumption path that characterizes the equilibrium of the hyperbolic consumer is thus observationally equivalent to the consumption path 1 of a consumer who discounts the future-exponentially. Moreover, the consumption rate of discount along this path should always equal the rate of return on capital. This suggests that one should discount future returns using the rate of return on capital, whether consumers have hyperbolic or exponential preferences. This is not, however, the end of the story. As Phelps and Pollak (1968) demonstrated many years ago, the equilibrium of the game played by quasi-hyperbolic consumers is Pareto- inefficient. Consumers in all years would be better off if they each saved more, but, absent a commitment mechanism, this will never occur. This implies that there is a role for government policy when preferences are hyperbolic. Specifically, the government can induce Pareto improvements by subsidizing the return on capital or, equivalently, by lowering the required rate of return on investment projects. Calibration of the hyperbolic model implies that the magnitude of this subsidy should be about two percentage points annually. We hasten to add that this conclusion does not favor environmental projects vis-a-vis other forms of investment. Hyperbolic preferences provide a motive for lowering the required return on all capital investment projects because of the under-saving that occurs along the hyperbolic equilibrium path, but do not favor one type of capital over another. The remainder of the paper is organized as follows. The next section introduces the notion of hyperbolic preferences. Section 2 describes the intertemporal game played by a consumer with quasi-hyperbolic preferences and characterizes the equilibrium of that game. Section 3 adds a government to the model of section 2 and section 4 concludes. 2 1. Hyperbolic Discounting and Its Consequences The neoclassical theory of optimal growth assumes that people have stationary time preferences: that the choice between two payoffs depends only on the absolute time interval separating them. There is, however, strong empirical evidence that people are more sensitive to a given time delay if it occurs closer to the present than if it occurs farther in the future (Ainslie, 1992; Cropper, Portney and Aydede, 1994). In other words the discount rate that applies to near- term consumption tradeoffs is higher than the discount rate that applies to long-term consumption tradeoffs. Loewenstein and Prelec (1992) present an axiomatic analysis of such preferences, which implies a generalized hyperbolic discount function, i.e., a function of the form (P(t) = (1+at)"', a,y > 0. (1) As a - 0, (p(t) approaches the exponential function. When a is very large, p(t) approximates a step function, implying that all periods after the first receive approximately equal weight. For a > 0, q(t) lies below the exponential function at low t and above it at high t. In what follows we approximate the hyperbolic function with a quasi-hyperbolic function, first proposed by Phelps and Pollak (1968) for intergenerational analysis and then applied by Laibson (1997) for intrapersonal analysis. 'Specifically, we examine a representative consumer who lives T periods and whose period-t self receives utility from the consumption sequence (CO,C1 .....CT) according to T-t Ut(CO,C1,...,CT) = u(ct) + p X 8iu(ct,i), 0 < P,6 < 1. (2) i=1 When 0 < J < 1 the discount structure in (2) mimics the qualitative properties of the hyperbolic 3 function, while maintaining most of the analytical tractability of the exponential discount function. We shall refer to the discount factors ( 1, p5, p52, p63 ...... } as quasi-hyperbolic. Figure 1 graphs the exponential discount function for 6 = 0.97, the hyperbolic discounting function with a 105 and y = 5x103, and the quasi-hyperbolic discounting function with 3 = 0.6 and 5 = 0.99. To illustrate the time-inconsistency problems to which quasi-hyperbolic preferences give rise, consider what happens if self 0 chooses the consumption sequence (co,c1 .,cT) to maximize (2) subject to the constraints (3) and (4), O< ct p> O. (5) Consider now the equilibria of this game. As Laibson has shown (1996), when T is finite, the game played by different temporal selves has a unique subgame perfect equilibrium. Each self s equilibrium consumption strategy is a linear function of its inherited wealth, ct = tWt, and the consumption path is characterized by u'(ct) = R6u'(ct+i)[t+i (,-1) + 1], (6) where k+ = Oct,+ (Wt+1,T)/8Wt-+. As T - co, equation (6) converges to u'(ct) = RWu'(ct+l)[R* (P-1) + 1], (6') where X is the solution to the non-linear equation = I - (6R'-P)/P[P;*(P 1 ) + 1 ]l/P. (7) When P = 1, equation (6') is identical to the condition that characterizes the optimal consumption path in the Ramsey model. Even when 0i # 1, there is still an observational 5 equivalence result conjectured by Arrow, between the equilibrium of the game with quasi- hyperbolic preferences and the Ramsey model. Specifically, the consumption path corresponding to (6') is identical to the consumption path generated by a Ramsey model in which utility is discounted at the constant exponential rate 6 = 6[)* (f-1) + 1]. In contrast to the Ramsey model, however, the consumption path that characterizes the game with quasi-hyperbolic preferences is not Pareto-efficient. As Phelps and Pollak (1968) first pointed out, all selves would be better off if they all consumed less than the equilibrium consumption rate, but there is no mechanism to guarantee that this strategy will be followed.' The intuition behind this result is simple. While requiring self t to save more lowers self t's utility, requiring all other selves to save more raises self t's utility, and the second effect dominates the first. This suggests that it may be possible for the government to enact policies to increase saving that will be Pareto-improving. 3. Pareto-Improving Government Policies in a Quasi-Hyperbolic World In a world in which consumers can be made better off by lowering the fraction of wealth 'This can be seen as follows. Write self t's utility as a function of X:, the fraction of wealth consumed in the long run: Ut(k) = U($Wt) + p6U(p(1_ ?* )RWt) + p62U(X. (I- X ) 2R2Wt) + p63U(I(- X_ ) 3R3Wt)+. Phelps and Pollak showed that aUt(Q)/O* < 0. 6 that they consume, there are two ways that the government can improve welfare. One is to subsidize interest rates (raise R to R); the other is to penalize consumption. Laibson (1996) focuses on both strategies. Here we consider only interest rate subsidies. We model the government as a sequence of players {0,1,2.,T} who can tax consumers and use the proceeds to subsidize iinterest rates. An essential feature of each government is that it can implement policies only with a lag: due to delays in the budget process, government t picks the lump-sum tax in period t+l, Tt+l, and Rt+1-R, the interest rate subsidy in period t+l.2 The effect of this assumption is to give the government a commitment technology. The time t government cannot iinstantaneously overturn the policies of the time t-1 government. In this way, the government is able to overcome the self control problem that plagues consumers. The goal of the government at time t is to maximize the well-being of self t. The government's policy instrument influences marginal tradeoffs between periods t+1 and t+2. Between t+1 and t+2 the time-t government would ideally like the consumption path to be: u'(ct+i) = 6Ru'(ct+2), (8) where R is the unsubsidized marginal rate of transformation. Note that ,3 does not appear in this equation, as 6 is the relevant discount faLctor between period t+1 and t+2 from government t's perspective. Equation (8) implies 2Since the consumers in this economy are not liquidity-constrained, the timing of lump- sum taxes is irrelevant. We therefore focus cn the choice of interest rate subsidy. 7 (CtJ/Ct+1)P = SR. (9) The time t government can implement this path by choosing an interest subsidy k = kt+1 such that the generalized Euler equation (with subsidized interest rate) is consistent with the government's desired consumption path, (ct+Ct1)= R [k(R)(i-1) + 1], (10) where :(Ik) is the value of X: implied by equation (7) when R is replaced by k. Equations (9) and (10) jointly imply that the time t government picks IR such that AR = SR [*(I )(I-1) + 1]. (11) This argument holds for all governments (i.e., for all times t). So, in equilibrium R t+1 = R for all t ˇ 0. To solve explicitly for the interest rate subsidy as a function of model parameters, we rely on the fact that ct+, = X(k)Wt+,, for all s 2 1. Together with (4) this implies that Ct+s+i/ct+s= (1- x(R)) k. (12) Inserting (12) into (9) yields );(k) as a function of R and R, [ (1- ?:(Ik)) f]pj= 6R, (13) which, together with (11) yields the interest rate subsidy as a function of model parameters, R-R = [(I-P)(I-(5R'-P)/P]- R/J3. (14) 8 To illustrate the magnitude of the interest subsidy, we consider plausible values for the parameters of the quasi-hyperbolic discount:ing function, P and 6, the gross returns on capital, R, and the elasticity of marginal utility with respect to consumption, p. Suppose that p = 3 and R = exp(.04), i.e., the gross return on capital is 4%. The values 3 = 0.6 and 6 = 0.99 pictured in Figure I are roughly consistent with empiric:al evidence on intertemporal choices presented by Ainslie .(1992). Together, these parameters imply an interest rate subsidy of over two percentage points (R-R = 0.02 1). We note that this subsidy puts the economy on a Pareto-efficient path. The equilibrium path that arises in the game with government is identical to the consumption path that would be chosen by self 0 if self 0 could commit all[ future selves. Note that on the equilibrium path u'(ct) = 8Ru'(ct±1) for all t > 0. This is exactly the equilibrium path self 0 would like to implement. Hence, the equilibrium path is lPareto-efficient, as any perturbation to the path would make self 0 worse off. 4. Conclusions When agents discount future utility of consumption using a quasi-hyperbolic rather than an exponential function, the equilibrium consumption path in the economy is no longer Pareto efficient. All consumers would be better oif if they each saved more, but there is no way to coordinate this behavior. This suggests a role for government policy: By subsidizing interest rates (reducing the required return on investment projects) the government can help to overcome the self-control problem that results from hyperbolic preferences. 9 As we noted in the introduction, however, this is not a pro-environment result. It does not justify applying a lower discount rate to an environmental project (e.g., a reforestation project) than would be applied to the building of a steel mill. Are there assumptions that would justify such a practice? The answer is yes. In a world of quasi-hyperbolic preferences one can justify applying a lower discount rate to environmental projects under the following conditions: (1) The production of environmental services (YE) from an environmental capital stock (E), e.g., a forest, is a separate process from the production of private output (Y) from private capital (K) (separability in production). (2) The environmental consumption good (cE) and the private consumption good (c) are imperfect substitutes in the utility function (separability in consumption). (3) The government controls the rate of consumption of the environmental good and, as is assumed above, the government can act only with a lag--in period t it chooses the amount of the environmental good that will be consumed in period t+I (c5t+I).3 Under these conditions one can show that the government will choose to consume a lower 3Formally, suppose that: Yt = AtKta O < ct < Kt Kt = (1-d)Kt., + Yt - ct yEt = AtEa O < CEt < E-t E,t = (ldEt YE1td) - cE t T-t The preferences of self t and government t are given by v(cEt,ct) + 13 Y. 6iv(cEt+i,ct+i). i=l 10 fraction of the environmental capital stock than consumers will choose to consume out of private capital, and that the steady-state rate of return on environmental capital will lie below the return on private capital. Thus, a lower discount rate should be applied to environmental projects than to private investments. The intuition behind this result is as follows: As long as the government can act only with a lag, it is prevented from over-consuming the environmental good, as consumers are tempted to do in the case of a private good. Furthermore, because of the assumed lack of substitutability between the environmental good and other goods, both in production and in consumption, consumers cannot undo the government's choices. This result is, however, a fragile one: it will fail to hold if any one of the three assumptions listed above is violated. In particular, if there is substitutability in production or consumption between the environmental good and other goods, the same rate of return will apply to both environmental and non-environmental capital. This underscores the main point of this paper. While hyperbolic discounting provides a rationale for lowering the required rate of return on investment projects, it does not provide justification for those who seek to treat environmental projects differently from other investment projects. 11 References Ainslie, George W. (1992) Picoeconomics, Cambridge: Cambridge University Press. Arrow, Kenneth J. and Mordecai Kurz. (1970) Public Investment, The Rate of Return and Optimal Fiscal Policy, Baltimore: Johns Hopkins University Press. Cropper, Maureen L., Sema K. Aydede and Paul R. Portney. (1994). " Preferences for Life Saving Programs: How the Public Discounts Time and Age. " Journal of Risk and Uncertainty, 8, 243-265. Laibson, David 1. (1996) "Hyperbolic Discount Functions, Undersaving, and Savings Policy," NBER Working Paper 5635, June 1996. Laibson, David I. (1997) "Golden Eggs and Hyperbolic Discounting." Quarterly Journal of Economics, 112,443-477. Loewenstein, George, and Drazen Prelec. (1992) "Anomalies in Intertemporal Choice: Evidence and an Interpretation." Quarterly Journal of Economics 57, 573-598. Phelps, E.S., and R.A. Pollak (1968) "On Second-Best National Saving and Game- Equilibrium Growth." Review of Economic Studies, 35, 185-199. Strotz, Robert H. (1955) "Myopia and Inconsistency in Dynamic Utility Maximization." Review of Economic Studies, 23, 165-180. 12 Figure 1: Three Discount Functions 1 \ ' ' ' I 0.9 0.8 0.7 Exponential 0.6 - Quasi-hyperbolic U- pe~~~~~ c 0 0.5 - - 3 n.^ Hyperbolic '0 0.4 - >0.3 - 0.2 0.1 ol 0 5 10 15 20 25 30 35 40 45 50 Year Policy Research Working Paper Series Contact Title Author Date for paper WPS1925 Half a Century of Development Jean Waelbroeck May 1998 J. Sweeney Economics: A Review Based on 31021 the Handbook of Development Economics WPS1926 Do Budgets Really Matter? Emmanuel Ablo June 1998 K. Rivera Evidence from Public Spending Ritva Reinikka 34141 on Education and Health in Uganda WPS1927 Revenue-productive Income Tax Fareed M. A. Hassan June 1998 A. Panton Structures and Tax Reforms in 85433 Emerging Market Economies: Evidence from Bulgaria WPS1928 Combining Census and Survey Data Jesko Hentschel June 1998 P. Lanjouw to Study Spatial Dimensions Jean Olson Lanjouw 34529 of Poverty Peter Lanjouw Javier Poggi WPS1929 A Database of World Infrastructure David Canning June 1998 A. Abuzid Stocks, 1950-95 33348 WPS1 930 The Main Determinants of Inflation in Ilker Domac June 1998 F. Lewis Albania Carlos Elbrit 82979 WPS1931 The Cost and Performance of Paid Ariel Dinar June 1998 F. Toppin Agricultural Extenion Services: The Gabriel Keynan 30450 Case of Agricultural Technology Transfer in Nicaragua WPS1932 Air Pollution and Health Effects: Bart D. Ostro June 1998 C Bernardo A Study of Respiratory Illness GL nnar S. Eskeland 31148 Among Children in Santiago, Chile Tarhan Feyzioglu Jose Miguel Sanchez WPS1933 The 1997 Pension Reform in Mexico Gloria Grandolini June 1998 C. Zappala Luis Cerda 87945 WPS1934 WTO Accession for Countries Constantine Michalopoulos June 1998 L. Tabada in Transition 36896 WPS1935 Explaining the Increase in Inequality Branko Milanovic June 1998 G. Evans during the Transition 85734 WPS1936 Determinants of Transient and Jyotsna Jalan June 1998 P. Sader Chronic Poverty: Evidence from Martin Ravallion 33902 Rural China WPS1937 Aid, the Incentive Regime, and Craig Burnside June 1998 E. Khine Poverty Reduction David Dollar 37471 Policy Research Working Paper Series Contact Title Author Date for paper WPS1938 What Explains the Success David Dollar June 1998 E. Khine or Failure of Structural Adjustment Jakob Svensson 37471 Programs? WP51939 Second Thoughts on Second Arturo J. Galindo June 1998 M. Cervantes Moments: Panel Evidence on William F. Maloney 37794 Asset-Based Models of Currency Crises %APS1940 The Structure of Labor Markets in William F. Maloney June 1998 M. Cervantes Developing Countries: Time Series 37794 Evidence on Competing Views W/VPS1941 Are Labor Markets in Developing William F. Maloney June 1998 M. Cervantes Countries Dualistic? 37794 W'vPS1942 Poverty Correlates and Indicator- Christiaan Grootaert July 1998 G. Ochieng Based Targeting in Eastern Europe Jeanine Braithwaite 31123 and the Former Soviet Union