MIT LIBRARIES 3 9080 02618 3563 llfll HMH HI ' il §llfl$$ii ; Digitized by the Internet Archive in 2011 with funding from Boston Library Consortium Member Libraries http://www.archive.org/details/inequalitysocial00farh2 OBwev B31 M415 Technology Department of Economics Working Paper Series Massachusetts Institute of INEQUALITY, SOCIAL DISCOUNTING AND ESTATE TAXATION Emmanuel Farhi Ivan Werning Working Paper 05-1 April 27, 2005 Rev. May 23, 2005 Room E52-251 50 Memorial Drive Cambridge, MA 02142 This paper can be downloaded without charge from the Social Science Research Network Paper Collection 2803 http://ssrn.com/abstract=71 at Inequality, Social Discounting and Estate Taxation* Emmanuel MIT Ivan Werning MIT, NBER and UTDT efarhi@mit.edu iwerning@mit.edu Farhi June 2004 First Draft: This Version: April 2005 Abstract To what degree should societies allow inequality to be inherited? What taxation play in shaping the intergenerational transmission of welfare? tions by modeling altruistically-linked individuals productivity shocks. uals where Our positive economy efficiency requires immiseration: is who role should estate We explore these ques- experience privately observed taste or identical to models with inequality grows without infinite- lived individ- bound and everyone's consumption converges to zero. However, under an intergenerational interpretation, previous work only characterizes a particular set of Pareto-efficient allocations: those that value only We where the social welfare on their welfare so that the effective social discount rate is lower than the private one. For any such difference in social and private discounting we find that consumption exhibits mean-reversion and that a steadystate, cross-sectional distribution for consumption and welfare exists, where no one is trapped at misery. The optimal allocation can then be implemented by a combination of income and the initial generation's welfare. study other efficient allocations criterion values future generations directly, placing a positive weight estate taxation. We find that the optimal estate tax is progressive: fortunate parents face higher average marginal tax rates on their bequests. 1 Introduction Societies inevitably choose the inheritability of inequality. tunity for newborns and incentives for altruistic parents act plays out to determine long-run inequality Some balance between is struck. We explore equality of oppor- how and draw some novel implications for this balancing optimal estate taxation. we thank Daron Acemoglu, Fernando Alvarez, George-Marios Angeletos, Blanchard, Ricardo Caballero, Dean Corbae, Mikhail Golosov, Bengt Holrn- *For useful discussions and comments Abhijit Banerjee, Gary Becker, Olivier strom, Narayana Kocherlakota, Robert Lucas, Casey Mulligan, Roger Myerson, Chris Phelan, Gilles Saint-Paul, Nancy Stokey. Jean Tirole and seminar and conference participants at Chicago, Minnesota, MIT and the Texas Monetary Conference held at the University of Austin in honor of the late Scott Freeman. This work begun motivated by a seminar presentation of Chris Phelan at MIT in May 2004. We also have gained significant insight from a manuscript by Scott Freeman and Michael Sadler we thank Dean Corbae for bringing it to our attention. — Existing normative models of inequality reach an extreme conclusion: inequality should be perfectly inheritable and rise steadily a vanishing lucky fraction to tions without bound, with everyone converging to absolute misery and This immiseration result bliss. on preferences (Phelan, 1998); and obtains invariably Thomas and or productivity (Aiyagari and Alvarez, effort or 1995). weak assump- in partial equilibrium (Green, 1987, Worrall, 1990), in general equilibrium (Atkeson and Lucas, 1992), ments with moral-hazard regarding work We robust; requires very is and across environ- with private information regarding preferences 1 depart minimally from these contributions, adopting the same positive economic models, but a slightly different normative work with criterion. In a generational context, previous lived agents characterizes the instance where future generations are not considered On only indirectly through the altruism of earlier ones. (2005) proposes a social planner with equal weights on infinite- directly, but the opposite side of the spectrum, Phelan all Our future generations. interest here is in exploring a class of Pareto-efficient allocations that take into account the current population along We with unborn future generations. and vanishing Pareto weight on the expected place a positive utility of future generations, this leads effectively to a social discount rate that is lower than the private one. This relatively small change produces a drastically different bounded, a steady-state, cross-sectional distribution exists bility is possible lower and everyone avoids misery. Indeed, welfare bound that between mation We social is is strictly better long-run inequality remains consumption and welfare, social mo- typically remains above an endogenous than misery. This outcome holds however small the difference and private discounting, and regardless of whether the source of asymmetric infor- privately observed preferences or productivity shocks. begin by modeling a positive economy that Each generation by Atkeson and Lucas (1992). live for for result: one period and are endowment of the only altruistic is is identical to the taste-shock setup developed composed of a continuum towards a single descendant. There consumption good in each period. is of individuals who a constant aggregate Individuals are ex-ante identical, but experience idiosyncratic shocks to preferences that are only privately observed —thus ruling out first- best allocations. Feasible allocations must be incentive compatible and must satisfy the aggregate resource constraint in When all periods. only the welfare of the to that of an economy with first generation is considered, the planning problem infinite-lived individuals. Intuitively, is equivalent immiseration then results from the desire to smooth the dynastic consumption path: rewards and punishments, required for incentives, are best delivered component that leads cross-sectional inequality to of the distribution 1 Many permanently. As a result, the consumption process inherits a random-walk is grow endlessly without bound. consistent with a constant aggregate find the immiseration result perplexing and some even endowment only find it if Infinite spreading everyone's consump- morally questionable, but it is also incon- venient from a practical standpoint. Long-run steady-states often provide a natural benchmark to study dynamic economies, but such long-run analyses are not possible for private-information economies without a steady-state distribution with positive consumption. This has impaired the study of long-run implications of optimal taxation, so common in the Ramsey taxation literature. tion eventually converges to zero. Note, that as a consequence, no steady-state, cross-sectional distribution with positive consumption exists. Across generations, this arrangement requires a lock-step link between the welfare of parent and child. Of course, the perfect intergenerational transmission of welfare improves parental incentives but at the expense of exposing newborns to the risk of their parent's luck. Individuals would value being insured against the uncertainty of their family's fortune the biggest risks one faces in By contrast, it life is regarding the family one — born is often recognized that one of it is into. remains optimal to link the fortunes of parents and children in our model, but no longer in lock-step. Rewards and punishments are distributed over in a front-loaded random walk manner. This creates a mean-reverting tendency —that is all consumption in strong enough to bound long-run inequality. future descendants, but The result is —instead of a a steady-state consumption and welfare, with no fraction of the population stuck cross- sectional distribution for at misery. We also study a repeated Mirrleesian version of our economy and derive implications for optimal estate taxation. In this model, individuals have identical preferences with regard to consumption and work but are heterogenous in the productivity of their work effort, productivity and work effort private is —only the resulting output is effort. Information about publicly observable. We show that the analysis from the taste-shock model carries over to this setup, virtually without change. In particular, a very similar Bellman equation characterizes the solution to the social planning prob- lem: consumption exhibits mean-reversion and has a steady-state cross-sectional distribution. This outcome highlights the fact that our results More importantly, the do not require any particular asymmetry of information. Mirrleesian model offers locations can be implemented new insights into estate taxation. by combining income and estate taxes. Specifically, we Feasible al- find that a progressive estate tax, which imposes a higher average marginal tax rate on the bequests of fortu- nate parents, is optimal. This result reflects the mean-reversion of consumption: more fortunate must have a declining consumption dynasties, with relatively high levels of current consumption, path induced by higher estate tax rates that lower the net rates of return across generations. Finally, an important methodological contribution of planning problem recursively. In doing so, (1987) to situations where private and we extend paper this ideas introduced by Spear state variable, and Srivastava we are able to reduce the and our analysis and results heavily exploit social preferences differ. dynamic program to a one-dimensional to reformulate the social is Indeed, the resulting Bellman equation. Related Literature. Our paper social planning most closely related to problem with no discounting of the planning problem exists then problem have is it must solve a strictly positive inequality static and future. He shows that if a steady state for the maximization problem, and that solutions to social mobility. of a steady-state distribution for the planning Phelan (2005), who considered a problem discounting. Unlike the case with no discounting, there steady-state distributions, and as a result, the methods is for Our paper establishes the existence any difference no associated in social static planning we develop here this and private problem are very different. for In overlapping-generation models without altruistic links, efficient place positive direct out that market equilibria that are Pareto all weight on future generations. Bernheim (1989) was the many dynastic extension of these models with altruism, in the Kaplow are not attainable by the market. The estate tax is negative — to point efficient allocations (1995) argued that these Pareto efficient allocations and that they can be implemented by market are natural social objectives taxation policy. Pareto first it is a subsidy — so equilibria with estate as to internalize the externality of giving on future generations. contributes to a large literature on dynamic economies with asymmetric information. Our work In addition to the tion work mentioned above, this includes recent research on dynamic optimal taxa- Golosov, Kocherlakota and Tsyvinski, 2003; Albanesi and Sleet, 2004; and Kocherlakota, (e.g., 2004). This application has been handicapped by the immiseration result of a steady-state distribution with positive consumption, clusions for optimal taxation. Our work Our results provide also indirectly related to Sleet is making it and by the non-existence difficult to draw long-run con- an encouraging way to overcome generation. In this environment, as in Phelan's, Sleet and Yeltekin derive for the current a foregone conclusion that immiseration will to solve for the best subgame-perfect equilibria. from a Lagrangian and use these to numerically turns out that the best allocation in their no-commitment asymptotically equivalent to the optimal one with commitment but featuring a more is patient welfare criterion. way it it is how first-order conditions simulate the solution. Interestingly, environment is problem. and Yeltekin (2004), who study an Atkeson-Lucas environment with a utilitarian planner, who lacks commitment and cares only not obtain, so that the interesting question this of characterizing the Thus, our own approach and results provide an indirect, but effective no-commitment problem and of formally establishing that a steady-state distribution with no one at misery exists. The and rest of the sets up the paper is organized as follows. Section 2 introduces the economic environment social planning problem. In Section 3, we develop a recursive version of the planning The problem and draw its then put to use Section 4 to characterize the solution to the social planning problem. Here derive our main consumption. some in connection to our original formulation. results We on mean-reversion and on the existence discuss these results in Section 5 related problems and reformulations. In Section setup with productivity shocks and focus on its we turn Bellman equation is we of a steady-state distribution for and develop 6, resulting intuition for them by studying to the canonical optimal-taxation implications for estate taxes. Section 7 offers some conclusions from the analysis. All proofs omitted in the main text are contained in the Appendix. 2 A Social Insurance The backbone of our Problem model requires a tradeoff between insurance and incentives. This tradeoff can be due to private information regarding either productivity or preferences. For purposes of comparison, we 6, first adopt the Atkeson-Lucas taste-shock specification. In Section we adapt our arguments to a repeated Mirrleesian model with privately observed productivity shocks. Similar arguments could be applied to moral-hazard situations with unobservable Our An positive economy identical to that of Atkeson-Lucas is —the differences are only normative. can be interpreted as a dynasty of individuals who have infinite-lived agent Under altruistically linked. effort choices. this interpretation, finite lives but are Atkeson-Lucas and others focus on a particular 2 set of efficient allocations: those that only directly consider the welfare of the initial generation. In contrast, our interest here future generations. with lies This approach infinitely-lived social planner who more patient than by a continuum of individuals who have and point in time, our economy identical preferences, live for one period, is populated and are replaced single descendant in the next. Parents born in period t are altruistic towards their only child their utility v t satisfies vt where ct > is uous derivative = Ei_i [9 t u(c t ) + fivt+i } own consumption and P £ (0, 1) is the altruistic weight placed on the The utility function u(c) is assumed continuous and concave, with a contin- the parent's descendant's utility v t +\. The all individuals. Demography, Preferences and Technology. At any by a that directly weigh the welfare of asymptotically equivalent to postulating preferences for an is is efficient allocations for all c taste shock 9 e 6 > is satisfying the Inada conditions lim^o u'(c) = oo and lim^oo u'(c) = 0. distributed identically and independently across individuals and time. This specification of altruism is consistent with individuals having a preference over the entire future consumption of their dynasty given by oo vt = YP J 5 ^t-i[e t+ su{ct+s )]. (1) s=0 In each period, a resource constraint limits aggregate consumption to be no greater than constant aggregate endowment e precisely those adopted Define U = > 0. set of all possible utility values. be unbounded so that the extremes u is These specifications choices preferences and technology are by Atkeson-Lucas. u(M+) to be the cost function c(u) some defined on U = u(0) and u = lim^oo u(c) Note that we allow may be as the inverse of the utility function c = utility to finite or infinite. u~ l . The For simplicity, we = < 6^ = 6. We denote the density < 9<i < by p(6) and adopt the normalization that E[0] = Yl n=i @nP{9 n = 1- The level of dynastic utility v then always belongs to the set V = w(M + )/(l — (3) with extremes v = u/(\ — P) and v = u/{\ — /?). assume contains a finite number of shocks 9 6-y ) Social Welfare. We t depart from Atkeson-Lucas by assuming that the social welfare criterion can be represented by preferences given by the utility function j^E-itM*)], (2) t=o 2 The final paragraph in Atkeson Lucas (1992) discusses the possible importance of relaxing this assumption. with J3 > P. Thus, social preferences are identical to the individual preferences given by (1), except for the discount factor. Future generations are This setup puts weight on the welfare of future generations directly. already indirectly valued through the altruism of the current generation. also directly included in the welfare function the social discount factor If, in addition, they are must be higher than To p. see this, consider the utilitarian welfare criterion oo oo ^a*E_ 1 Ut = ^^E_ 4=0 where S t and = P + P t_1 a + + Pa l • • • more social preferences are = rion (3) approaches (2) with 1 '1 + a Then 1 [^«(c t )] (3) > the discount factor satisfies . In the limit 8t+i/5 t patient. max{/?, a}. — > max{/3, a}, so the welfare crite- 3 4 ' Atkeson-Lucas' analysis applies to the case with weight 1 t=0 P = $, so we focus on the case where enough placed on future generations to ensure that the long-run social discount factor remains is strictly higher than the private one, rest of the paper, it is $> p. Although we adopt the preference in (2) directly for the straightforward to adapt the arguments to the welfare criterion specifications are slightly different for any finite The two (3). horizon but are identical for the long-run, which is our primary concern. Information and Incentives. Taste shock their descendants. The realizations are privately observed by individuals and mechanisms revelation principle then allows us to restrict our attention to that rely on truthful reports of these shocks. Thus, each dynasty faces a sequence of consumption functions {q}, where ct (0 (6o, 8i, that a 1 . . ,8 t ). maps £ : . A t ) = dynasty's reporting strategy a {a t t histories of shocks 9 into a current report -* G*. We use ct* t } is a sequence of functions a t Any . strategy Given an allocation {c t }, l ) = t for all 9 the utility obtained from any reporting strategy a l e — > : a induces a history to denote the truth- telling strategy with a* (9 t i+1 = t represents an individual's consumption after reporting the history 9 of reports 0*. is oo [/({cj.a;/?)^ J2 t 3 = /?M*(*W)) PK?)- 0<£0t+l Bernheim (1989) performs similar intergenerational discount factor calculations Caplin and Leahy (2005) argue that these ideas deterministic dynastic saving model. discounting within a lifetime, leading to a social discount factor that is in his welfare analysis of a also apply to intra-personal greater than the private one not only across generations, but within generations as well. 4 Onc can also adopt the more general welfare {at}. In particular, the sequence a = (1 criterion J^^L -j3)/(l - (i) and at = a E_iut t ao/3* for for t > some sequence of positive Pareto weights 1 delivers S t+ i/5 t = ft for all t = 0, 1, . . An allocation {c t } is incentive compatible if truth-telling is optimal: U({ct },a*;P)>U({ct },a;P). (4) Social Planning Problem. Following Atkeson-Lucas, we identify each dynasty with a number v, which we interpret as that all its initial dynasties with the entitlement to expected, discounted same entitlement v receive the same treatment. distribution of utilities v across the population of dynasties: ip(A) will receive An expected discounted allocation is AC the set utility in a sequence of functions that a dynasty with initial entitlement v gets at date For any given is feasible v to initial distribution of incentive compatible for (i) it is if: entitlements dynasties entitled to all initial exceed the fixed endowment v; and We v. then let ip denote a where 0^(0') represents the A (ii) it e, sequence of shocks we say that an delivers expected We optimum maximizes the average feasible allocations. ip That let e*(ip) allocation {cvt } utility of at least denote the lowest resource and an endowment level e to is relevant for is level e /3 = ip such —the j3. weighed by social welfare function (2), the social planning problem given an is, #*. average consumption in the population does not (iii) problem studied by Atkeson and Lucas (1992) which social who consumption that there exists a feasible allocation that delivers the distribution of utility entitlements efficiency assume the fraction of dynasties after reporting the dynasties; all e in all periods. t v, and resources ip is We = R. each {<:£} for utility, vq initial distribution of ip, over all entitlements maximize U({cvt },a*,P)diP(v) subject to v = U{{cvt },a*;P) > U{{cv },a; (3) t and for all v, /^cf(9')Pr(«')#)<e J Our social planning e problem is > and where J3 Steady States. on distributions to the planning utilities v t of that problem is is = 0,1,... (5) non-empty constraint well defined, with a are interested in situations with Our focus £ l (3 e > set, for all e > e*(ip); we e*(ip). of utility entitlements ip such that the solution features, in each period, a cross-sectional distribution of continuation also distributed according to consumption to replicate itself over time. these properties a steady state and denote ip. We also require the cross-sectional distribution term any them by utility constitutes a state variable that follows a We ip*. initial distribution of As we Markov shall process, entitlements with demonstrate below, continuation and steady states are then invariant distributions of this process. Note that in the Atkeson-Lucas case, with ft = j3, the non-existence of a steady state with a consequence of the immiseration result: starting from any non-trivial positive consumption is initial distribution ip and resources e*(ip) the sequence of distributions converges weakly to the 7 distribution having steady states A 3 ip* full mass at misery, with zero consumption that exhaust a strictly positive aggregate for everyone. endowment We seek non-trivial e in all periods. Bellman Equation In this section we study a relaxed version of the social planning problem whose solution coincides with that of the original problem at steady states. The relaxed problem has two important advan- problem can be solved by studying a First, the relaxed tages. dynasty with entitlement v subproblems set of —one —which avoids the need to keep track of the entire population. for each Second, each of these subproblems admits a simple recursive formulation, which can be characterized quite sharply. We believe that the general approach we develop may be here useful in other contexts. Consider the relaxed planning problem where the sequence of resource constraints by the replaced single intertemporal condition /oo oo ^Q for (5) is some t positive sequence {Qt} with J>^)Pr(^)^)<eX>, Ym^o Qt < °°- O ne can interpret (6) problem as representing this a small open economy facing intertemporal prices {Qt}- The relaxed and original versions of the planning problem are related in that any solution to the former which happens to satisfy the A resource constraints in (5) must also be a solution to the latter. the converse: there must exist some Lagrangian argument establishes positive sequence {Qt} such that the solution to the original planning problem also solves the relaxed problem. Most importantly, any steady-state solution to the relaxed problem Our is a steady-state solution to the original one. are only compatible with q forward. Q = focus on steady states leads naturally to = $, so Attaching a multiplier A the Lagrangian L = JL v dijj(v) we adopt > l q t for some q > this value for the relaxed 0. Indeed, steady states problem from to the intertemporal resource constraint (6), this point we can form where oo and study the optimization of L subject to v = U {{$} equivalent to the pointwise optimization, for each v = it satisfies U({tf}, a*;/3) Theorem > U{{c v t ), a; /?). Our first result v, of , a* ; /?) > U {{cv } t , a; the subproblem: k(v) (3) = for all v. sup characterizes this value function L v This is subject to and shows that a Bellman equation. 1 The value function k(v) k(v) is continuous, concave, and satisfies the Bellman equation = maxE[6u(0) - \c(u(6)) + 0k(w(9))] (7) subject to = v 9u(9) E[9u(6) + f3w(9) > + f3w(9)] 9u(9') (8) + pw{9') for This recursive formulation imposes a promise-keeping constraint that telling the truth today necessary to satisfy the taken care of in full is optimal the truth if is 9' e 0. (9) and an incentive con- Of told in future periods. incentive-compatibility condition by evaluating the value function (7) (S) 9, out one-shot deviations from truth-telling, guaranteeing Intuitively, the latter rules straint (9). all (4). course, this Intuitively, the rest at the continuation utility: is for is implicitly any given continuation value w(9), envision the planner in the next period solving the remaining sequence problem by selecting an entire allocation that Then k(w(9)) incentive compatible from then on. is represents the value to the planner of this continuation allocation. Taken together, a pair u(9) and w{9) that yields satisfies (8)-(9) an allocation that pasted with the corresponding continuation allocations for each w(9), satisfies the full incentive-compatibility The (4). objective function in (7) then captures the relevant value of allocations constructed in this way. Among We tained. shows that the maximum in the Bellman equation (7) is atu w the policy functions g (9,v) and g (9,v) denote the unique solutions for u and w, other things, let For any respectively. Theorem 1 initial utility entitlement the policy functions (g u ,g w ) by setting uq{9q) inductively for t Our next in this is 1 by u t (0') = g {9 t ,v t {9 t = -1 )) result elucidates the connection way and Theorem > u Vq, an allocation {u t } can then be generated from u g (9 ,VQ) initially and defining t- 1 and Vt+^O*) = g w (9 u v t {9 )). ii t and v +i(9 (0') t t ) between allocations generated from the policy functions solutions to the planning problem. 2 (a) An allocation {u t } is optimal for the relaxed problem, given w u generated by the policy functions (g ,g lifetime utility of Vq; (b) ) vq, if starting at vq, is incentive compatible, an allocation {u t } generated by the policy functions u (g ,g and only if it and delivers a w starting at ), 4-1 = and delivers utility vq; (c) an allocation {u t } generated by the has lmij-.oo/?' E_i?; t (# ) u w policy functions {g ,g ), starting from v Q is incentive compatible if vq, , t t limsup E_ 1 p v (a t for all -1 t {9 -1 )) > reporting strategies a. Part (a) of Theorem 2 implies that either the solution to the relaxed planning generated by the policy functions of the Bellman equation, or there and part (c) of (c). the theorem show that the The can be verified a Theorem first case is guaranteed if is we can no solution many all. is Parts (b) verify the limit condition in latter is automatically satisfied for all utility functions that are in at problem bounded below and other cases of interest. 2 involves various applications of versions of the Principle of Optimality. For example, for any given policy functions [g u ,g w ) and an initial value vo, the individual dynasty faces a recursive dynamic programming The case with (3 = J3 can be studied by the same approach. Recall that the efficiency problem studied by Atkeson and Lucas (1992) minimizes resources e subject to the sequence of resource constraints (5) and v = U({cvt ), a*; > U {{cv } (3) t , a; (3) for all v. Consider the relaxed version of this problem that replaces the sequence of resource constraints with the single intertemporal constraint (G) for some sequence {Q constraints (5) is it t }- Then the solution to this problem satisfies the resource also a solution to the original problem. with constant relative risk aversion this case, if utility functions, Although no steady state exists in the relaxed problem with Q = characterize the original one, for an appropriately chosen value of qAL > 0, q AL t not necessarily equal to P. we can take the Since the constraint (6) binds, the left hand KAL (v) = inf side of this inequality. This minimization can then be is subject to (8) and > t /?). The each v let associated Bellman (10) > This problem can be thought of as the limiting version of the problem and is case as (3 not necessarily 0. Theorems 1 and 2 Bellman equation. its Optimal Inequality In this section We U{{c },a; for = mmE[c(u{d))+q AL KAL (w(8))] co and where the discount factor in the objective q AL also apply to this 4 (9). = U ({<*}, a*; P) > done pointwise: then KAL (v) — to v J2Zo <&&(&) subject equation for this problem A objective function for the relaxed problem as we exploit the connection between the characterize the solution and derive a key equation that dynamics of consumption. The main enough to imply the existence of result of the section exists. Mean Reversion We now are in a position to Theorem 2, 1 = to establish that these forces are strong Finally, study the Bellman equation's optimization problem. The value function k(v) main, with lim v ^ v k'(v) mean-reverting forces in the -co. is strictly If utility is we provide and ensure that a solution to the planning justify the use of first-order conditions with the following Lemma is illustrates an invariant distribution with no misery. sufficient conditions to verify part (c) of problem Bellman equation and the planning problem. To begin, we lemma: concave and differ~entiable on the interior of unbounded below, then lim v ^yk'(v) = 1. its do- Otherwise problem with state variable vt Conditions (8) and (9) then amount to guessing and verifying a solution to the Bellman equation of the agent's problem in particular, that the value function that satisfies the Bellman equation, with truth telling, is the identity function. However, one also needs to verify that this value function represents the true optimal value for the dynasty from the sequential problem. This verification is accomplished by part (c) of . — Theorem 2. 10 lim v ^y k'(v ) = Let A let jjl{9,9') = oo. k'(v) be the multiplier on the left-hand side of the promise-keeping constraint be the multipliers on the incentive constraints - with equality if u The (9) is interior. for k(v) derived in - \c' (u(9)))p(6) Lemma 1, 9Xp(9) The (9). must be solution for w(9) - +p$r p(9, ff)-PY, e' — Using the envelope condition k'(v) J k'{g w °') = Inada conditions °- e' A and adding up across Y M is 0, interior, given the satisfy the first-order condition 0Xp{9) < 9') 9^(9, and must pk' [w(9))p(9) and first-order condition for u(9) £ + ]T 9(i(9, 9') - (8) (9,v))p{9) = this 9, becomes ik\v). (11) P eeo This key equation can be represented in sequential notation as E where {v t } is t - l [k'{v t+l (9 t ))]=ik'(v generated by the policy function g w Regressive (hereafter: CLAR) Markov process. t t -1 (9 Thus, {k'(v t )} . (12) )) is a Conditional Linear Auto Note that we can translate anything about the process {k'(vt )} into implications for the process {vt }, since the derivative k'(v) u strictly decreasing. Likewise, using the policy function g (9,v), conclusions is continuous and about the process {v t } provide information about the process for consumption. The conditional expectation in (12) illustrates that for the process {k'(v t )} maximum at v* > toward zero. v with k'(v*) from misery. This feature is = Lemma 0, 1 1 creates a force for mean implies that the value function k(v) has an interior itself embodies an interesting form of social mobility. with mobility ensured between them. social hierarchies, Descendants of individuals with current welfare above v* will eventually fall below dynasties initially entitled to welfare below v* are guaranteed to access levels above fall the value function k(v). it is it. Similarly This rise important to stress the role played by the non-monotonicity of Although mean reversion stems from non-monotonicity of k(v) that ensures that reversion Atkeson-Lucas case a is it. of families illustrates a strong intergenerational mobility in the model. In deriving this result, case — away key to our results on the existence of invariant distributions. can divide the population into two and reversion so reversion occurs towards this interior utility level Economically, the mean-reversion equation We @/0 < CLAR 11 < 1 in not toward misery. equation (12), By it is contrast, in the may hold, but the value function in this misery. Indeed, the envelope and first-order equation similar to (12) monotone and reversion then occurs toward is ft/ft conditions for the Bellman equation (10) yield qAL e which when similar to condition (12) is KAL (v) the value function q AL > h Crucially, unlike the case with j3. strictly increasing, so that is KAL (v) > negative process, which implies by the Martingale Convergence almost surely utilities w g (9,v) follows, v t — > some to (a.s.) v and q — > Proposition mass unbounded 7(1 7 = — — a.s. » Immiseration then bounds are v) satisfies the (l - < 1 CLAR equation (11). In addition: - *V(0, v)) < 7 (1 - + 9 n ^ -E[9\9> = (/?//?) k'(v)) u(6) = u w(9) > v e 6. For values of v such that bound in (13) holds for 1 unbounded below, continuation is (l - (13) £) max {(1 + 6 n — E[6 < 6 n ])/9 n } and k'(v) > 1, we have = 0/P)k\v) k'(v) < 1, the lower bound in (13) holds; the upper sufficiently high v. illustrates a powerful a parent + 9n]/9n _ l }. k'(w(9)) Proposition guaranteeing the existence bounded below, then for low enough values of v such that is all 9 critical for below, then - fc») + (P/P) min {1 2<n<N how much t) (3. 6 G 9, where the constants are given by 7 (b) if utility for KAL (v at misery. The policy function g w (9, 1 if utility is all $ > evolution. These its of an invariant distribution with no for must converge it result pushes the characterization of reversion past the average behavior of the {k't } process by deriving bounds for (a) a non- is t )} here This highlights the importance of the non-monotonicity of the value a.s. function k(v) for our results in the case of Our next Theorem that /?, Since incentives must be provided using continuation finite value. w g (9,v), this rules out anything other than ^ {KAL {v Thus, 0. > ft utility g tendency away from misery. w (9, v) remains bounded even as v supposed to be punished, his child 6 With logarithmic utility qAL = /3 yields a and a < 1 the appropriate value of q Ah that , For example, with utility is — always somewhat spared. solution with constant average consumption. yields constant consumption, 12 -00. Thus, no matter is strictly With above (3. u(c.) = c l ~a /{I -a) Main Result: Existence of an Invariant Distribution with We now state the then main result of this section: if a solution to the relaxed planning problem exists, The proof admits an invariant distribution with no misery. it conditional-expectation equation (12) and the bounds in Proposition mean-reversion properties discussed in the previous subsection. Proposition 2 The existence of an invariant distribution for the Markov process {vt } implied by g is bounded above, or 7 Proposition < is guaranteed ip* of this result relies on the Thus, 1. it makes use exists if either: utility is Theorem part (a) of 2, leaves This situation contrasts strongly with the Atkeson-Lucas unbounded we show that a open only two possibilities: case, with /? = /?, (ii) illustrate this 0, also provides an efficient method (i) no solution where a solution solution to the planning problem can be guaranteed so that case Our Bellman equation We = below, utility but does not admit a steady state, and everyone ends up at misery. Towards the end of section lem. both with no mass at misery, ip*({v}) the relaxed problem admits a steady-state invariant distribution with no misery; or exists. of 7 1 when combined with 2, w No Misery for explicitly solving the (i) this holds. planning prob- with two examples, one analytical and another numerical. Example 1. Suppose utility is CRRA with a = 1/2, so that u(c) = 4c / 2 for c > and c(u) = u 2 /2 for u > 0. For f3 = (3 Atkeson-Lucas show that the optimum involves consumption inequality 1 growing without bound and leading to immiseration. Consider the relaxed problem where we ignore the non-negativity constraints on u and w, k(v) = maxE[0u(0) - \u(6) 2 + J3k(w(6))] u,w subject to (8) and (9). the value function is This is a linear-quadratic dynamic programming problem, so a quadratic and the policy functions are linear in u g (6,v) = 7 ?(0)t; g"(0 v) = 7» + 7oW t it follows that v: + 7 «(0) For taste shocks with sufficiently small amplitude we can guarantee, by continuity with the deterministic case = [vl,vh] with v L convex and original 7 0, > that 7™(0) 0. < 1 and 7™(0) > Moreover, g u (6,v) utility turns > for v implying a unique bounded ergodic set e problem with non- negativity constraints on When utility is for utility [vl,v h ]. Hence, since the planning problem is out to be strictly positive at the steady state, this solution does solves the bounded below, we u. either require that utility for a small dispersion of the shocks, as a simple It 0, way be bounded above, or that 7 of ensuring that the ergodic set seems very plausible, however, that these conditions could be dispensed with. 13 is < 1, which is bounded away from ensured misery. Figure Example 1: To 2. w illustrate the for the logarithmic case We iterated Figure c(v(l — = w against c(g (9,v)(l reveals a unique, low values of is j3 0.9, J3 = 0.975, e = = A" 1 = 0.6, 6 h for k(v) until convergence. we compute the 1.2, 0, = 0.75 solution and p = 0.5. 8 plots the policy function for continuation utility in consumption-equivalent units, consumption, c(v(l 1 with c Policy function g (9.w) 2: numerical value of our recursive formulation, on the Bellman equation 1 (3)) Figure Policy functions g (6,v) v. — /?)) — against g (3)), c while Figure 2 does the same for the policy function for Both policy functions are monotonic and smooth. Figure (9, v). bounded ergodic Note that both policy functions become nearly set for v. This illustrates the result, discussed immediately after Proposition 1, flat for that utility kept above some endogenous bound. Figure 3 displays the steady-state, cross-sectional distribution of dynastic utility measured in — consumption-equivalent units, c(v(l /?)) implied by the solution to the planning problem. 9 long-run distribution has a smooth bell-curve shape mean-reverting dynamics of the model, since distribution of taste shocks. 8 The incentive The figure also details of this numerical exercise and promise- keeping constraints. — a feature that must be due to the smooth, cannot be a direct consequence of our two-point shows the invariant distributions where as We it The follows: we for various values of solved for u (0) as a function of We then maximized over w(6). employed a grid w (9) using the for v defined in terms of equally spaced consumption-equivalent units c((l — 0)v) = {0.01, ,2}. Results with a grid size of 100 and 300 were similar; we report the latter. We used Matlab's splines package to interpolate the value function . and used fmincon.m as our optimization routine over w{6). Our iterations . . were initialized with the value function corresponding to the feasible plan that features constant consumption: ( ,(1 _ p) k Q (v) _ Ac(t; (i _ , 3)) 1-0 10~ 10 and verified that the policy functions had also converged. Note that g is well within the interior of [.01,2], so that the arbitrary upper and lower bounds from our grid choice were not found to be binding. We stopped the iterations when ||fc n (u) — fc n _i(v)|| < w (9,v) 9 The invariant distribution was approximated by generated a Monte Carlo simulation for the dynamics of the 1 {v t } process generated by g " with an arbitrary initial value of vq. Since this process converges to a unique invariant distribution ip* starting from any initial value of vo, the frequencies in a long time-series sample approach the , , frequencies of ip" . To create the figure we used Matlab's Wavelet Toolbox simulated Monte-Carlo sample. 14 to approximate the density from the 0.61 0.6 utility in Figure P. The degree intuitively 3: 0.62 consumption equivalent units Steady-State Distributions of Dynastic Utility of inequality appears to decrease with higher values of by the coefficient CLAR equation on the These simulations of the impulse response to shocks. we approach the Atkeson-Lucas (12) case, /3 — > /3, /?. This outcome and the discussion is in Section 5 suggested on features also support the natural conjecture that as the resulting sequence of invariant distributions blows up, since no steady state with positive consumption exists when $ = P- We now turn briefly to issues of uniqueness and stability for the invariant distribution guaranteed by Proposition 2. This question is of level vq, That is, if convergence toward the distribution then the fortunes of distant descendants of the individual's present condition. present, but its influence is Indeed, under some conditions it represents an even stronger notion (12) discussed in the previous occurs starting from any i/j* —the distribution of exerts some wiped we can guarantee To that the social optimum generated by g w converges weakly to , (12) ensures influence on the initial distribution ip Q ip* in our model does see this, suppose the ergodic set for the {A;^} compact. This is guaranteed, for example, by applying Proposition the policy function g w (9,v) is monotone in v, the invariant distribution ip* from any independent out. is in the sense that, starting is initial utility dies out over time, so that the advantages or disadvantages display this strong notion of social mobility. process — their welfare At the optimum, the past always bounded and of distant ancestors are eventually if interest because by the mean- reversion condition of social mobility than that implied subsection. economic , 1 when 7 < is 1. Then, unique and stable the sequence of distributions {ip t }, This follows since the conditional-expectation equation enough mixing to apply Hopenhayn-Prescott's Theorem. 10 The monotonicity of the 'See pg. 382-383 in Stokey and Lucas with Prescott (1989). 15 w policy functions for continuation utility 1 and seems intuitive and plausible, as illustrated by Examples 11 2. Another approach suggests uniqueness and convergence without relying on monotonicity. Grunwald, Hyndman, Tedesco and Tweedie (1999) show that one-dimensional, with the Feller property that are bounded below and satisfy a a unique and stable invariant distribution. the sense that initial distribution is fast, in ments of their theorem have been of irreducibility, which distribution. We do is hold if condition, such as (12), have Moreover, convergence to this distribution from any occurs at the geometric rate @/$. All the require- it verified already for our likely to CLAR Markov processes irreducible we were to model, except for the technical condition assume that the taste shock has a continuous not pursue this formally other than to note that the forces for reversion in (12) could be further exploited to establish uniqueness and convergence. Our focus on steady states, where the distribution of utility entitlements replicates itself over time, has exploited the fact that the relaxed and original planning problems we can do more and for the logarithmic utility case Proposition 3 there exists is an endowment level e*(ip) initial distribution of utility entitlements ip such that the solution to the original social planning problem generated by the policy functions (g u ,g w ) from the relaxed problem with monotone e* is a e*(i; )<e*(i; One increasing, in that coincide. However, characterize transitional dynamics. any // utility is logarithmic, then for must a if ip ' -< ip b Q = t ft in the sense of first- order stochastic . The function dominance then b ). interesting application of this result modified to select the best to the situation where the planning problem is initial distribution instead of taking one as given. ip, Then dynasties are treated identically and started with identical utility level v* solving k'(v*) is all initial = 0. The optimal allocation then evolves according to the dynamics implied by the policy function g w (6,v). The cross-sectional distribution of welfare will spread out from its initially egalitarian condition as dynasties experience varying luck in the realization of their shocks. By convergence to a unique invariant distribution ip* of the Markov implied by the policy function g w (9, v) takes on additional economic meaning. It implies applying Proposition process {v t } 3, the stability of the cross-sectional distributions of welfare and consumption in the population. That if is, the Markov process {v t } generated by g w welfare and consumption eventually settle As mentioned for any (ip,e) as in Section 3, for any down stable, then the cross-sectional distributions of is to the steady state. utility function specification the solution to a relaxed problem with some sequence of prices {Qt}, that are not necessarily exponential. Proposition 3 identifies the distributions More lead to exponential prices in the logarithmic case. pair (ip,e), one can characterize the solution we can show that Q = t ft + A/5* for and endowment pairs (ip, e) generally, with logarithmic utility for some constant A. The that any entire optimal allocation can then be characterized by the policy functions from a non-stationary Bellman equation. Since n Indeed, can be shown that g w (9,v) is strictly increasing in v. However, although we know of no counterexample, we have not found conditions that ensure the monotonicity of g w {9, v) for all 6^0. for the general case it 16 prices are asymptotically exponential, in that lim t _ 00 l P~ Qt = follows that long-run 1, it dynamics are always dominated by the policy functions (g u ,g w ) from the relaxed problem with exponential prices Q = t that /?* we have characterized. Sufficient Conditions for Verification We conclude this section by describing sufficient conditions to exist at the steady state we First, steps. ip* identified consumption Lemma is by the policy functions establish that allocations generated compatible by verifying the condition finite for in part (c) of under the invariant distribution in Proposition 2. This involves two by the policy functions are indeed incentive Theorem 2. Second, we verify that average ip*. 2 The allocation generated from the policy functions (g u ,g w ), starting from any guaranteed to be incentive compatible in the following cases: (a) is a solution to the planning problem bounded below; (c) utility is logarithmic; < or (d) 7 1 or 7 > utility is bounded above; vq, is (b) utility 0. We now find sufficient conditions that guarantee average consumption is finite under the invariant distribution is ip* . If the ergodic set for utility v bounded and average consumption is is bounded away from the extremes, then consumption trivially finite. Even when a bounded ergodic set for utility v cannot be ensured, finite average consumption can be guaranteed for a large class of utility functions. Lemma 3 Average consumption is finite under the invariant distribution ip* ^Tc(g u (8,v))p(9)diP*(v)<cx> e if either (a) the of ergodic set for v is bounded; or (b) such that d{u{c)) is a convex function c. Note that a bounded ergodic set is guaranteed by 7 with sufficiently small amplitude; and condition aversion utility functions with The a> < (b) holds, for 1, which and thus has full range. for ensured for taste shocks all constant relative risk 1. utility, A. For instance, in the case of average steady state consumption is a power function of In fact, in this case the entire solution for consumption of degree one in the value of the planning problem is example, for value of average consumption depends on the value of constant relative risk aversion 5 utility is endowment any endowment e. is A, homogenous This ensures a steady state solution to the social level. Discussion: Mean-Reversion This section develops an intuitive understanding of the key mean-reversion property discussed previously. We first derive the impulse response of consumption to a one-time taste shock. 17 We then revisit the problem with an alternative Bellman equation that full useful as a source of intuition. is Impulse Response Consider a version of our model where only the period, there are is two possible values deterministic: = > 1 for t We 1. compare generation faces uncertainty. shock 6 G {0l,9h}, but this to the case In the thereafter the first economy with no uncertainty in the first To This allows us to trace out the consumption response to the taste shock over time. period. simplify, We 6t for the taste first we adopt logarithmic utility. begin by studying a subproblem of the deterministic planning problem from the second generation onward, that planning problem is for = t 1,2, — For a given, promised continuation utility the Vj, is oo fcdetfa) = maxV^'" 1 7T t=\ {a} (logQ - Ac*) v , ' subject to oo 4=1 The associated Bellman equation is = max kdet(v t ) Ct,Vt+l subject to vt = log Ct + pv t+ j. The (log(ct ) - Aq + (3kdet (vt+1 ) \ I first-order *4etK+l) ct and envelope conditions imply that = (14) !*£*(*)> = A-^l-fc^M). (15) Condition (14) shows that {k'det (v t )} reverts geometrically towards zero at the rate (3/$. This is In the logarithmic case, it a deterministic version of the conditional-expectation equation translates directly into back to a common consumption by the discounting coincide, so that level c t = c{y-y/{\ Turning to the Thus, consumption reverts first-order condition (15). steady state at the same rate; deviations from the steady-state level of consump- tion have a half-life of (log 2 (/3 P))~ new (12). — first P= l $, Note that in the Atkeson-Lucas case when social and private consumption remains perfectly constant /?)). generation at t maxE[0 o log = 0, the planning problem solves - (c o (0)) A o cq(0) + pk det co,vi ( Vl {8))} subject to 0l log (c (9 L )) + p Vl (9 L > ) 6 L log (c 18 (M) + Pv 1 (9 H ). after the shock at its where we omit the other incentive constraint since > convex. At the optimum, c {6 H ) is it is c (9 L ), Vi(8 H ) not binding at the optimum, and the problem < Vi(6 L ), and E[k'det (v 1 (6))]=0, implying that Vi(6 H ) and equal to A -1 < < v* = Vi(9 L ) where k'{v*) Note that average consumption 0. That constant in all periods. Figure 4 shows the consumption response to a taste shock in the periods. is is, we use and (14) on consumption from the shock (15) for t > 1 dies out over time first starting at v 1 (9 L ), v* period, for subsequent The and v±(6 H ). effect and consumption returns to a common steady- state level. Again, this illustrates that the influence of past fortunes eventually vanishes for distant We descendants. also plot the Atkeson-Lucas case with has a permanent impact on the consumption of Figure To provide 4: Consumption path incentives for the reporting a low taste shock. first is > = (3, consumption, for a where the luck of the first 1 in response to taste shock at t = generation, society rewards the descendants of an individual Rewards can take two forms and society makes use exploits differences in preferences: it of both. in present- value terms. tilting the given present value, in the direction preferred by individuals. 12 consumption profile Somc readers may recognize this when conceding their child's The second 12 is toward the present. Similarly, punishments involve method The allows an adjustment in the pattern of Since individuals are more impatient than the planner, this latter form of reward by generation descendants. standard and involves increased consumption spending, more subtle and is first for t all (3 delivered tilting the as the time-honored system of rewards and punishments used by snack or reducing their TV-time. In these instances, the child values some goods more than the parent wishes, and the parent uses them to provide incentives. parents last favorite 19 consumption path toward the future. intensively to provide incentives common returns to a affecting the both In cases, earlier —rewards and punishments are front-loaded. first Indeed, consumption steady-state level in the long-run regardless of the initial shock because consumption of very distant descendants incentives to the consumption dates are used more not an efficient is way for society to provide generation. Another Bellman Equation Here we develop another Bellman equation that holds This alternative formulation q = is useful, for any value of q not necessarily equal to J3. both as a source of intuition and to motivate our focus on P- Consider the following cost minimization problem oo *(«,«)= mm ^g^c^PrC* (C ' } ), gt t=0 subject to the incentive compatibility constraint 1 U ({ct} , a* ; (3) > U({ct},a; (5) and oo 4=0 oo t=0 that delivering utility v is, and v for the planner and individual, respectively. Then the value function must satisfy the Bellman equation K{v,v) = maxE\c(u(9))+qK(w(9),w{9))], u,w,u> -* subject to v = E[6u(6) + J3w(6)] v = E[6u(6) + 0w(6)] and 6u{9) + f3w{9) > 9u{6') This formulation could be used to derive equation formulation, however, The more convenient (7) is slightly is that it all + 0w{6') for all 9, 9' e 0. of our results, although the lower-dimensional for that purpose. The advantage of this cost-minimization lends itself naturally to economic interpretations. following story provides a useful reinterpretation and source for intuition. infinite-lived Bellman Consider an household with two members, husband and wife, and assume that consumption a public good —there is is no intra-period resource allocation problem. However, husband and wife 20 disagree on how Suppose the wife to discount the future. more patient, but only the husband problem characterizes the constrained Pareto problem for this is can observe and report taste-shock realizations. Then this cost-minimization The Pareto Pareto frontier between husband and wife. K = household, in the sense that the isocost curve K(v, v) K represents, given resources frontier is non-standard in that the , not it is everywhere decreasing and does not represent the usual transfer of private goods between two agents. Instead, it arises from differences in preferences that generate a disagreement about the optimal consumption path for the only public good. Since disagreement on preferences the Pareto frontier is non-monotone and the highest possible interior utility level for the the left The of this point must = husband, where Ki(v*,v*) utility for the wife attained for an Reductions in the husband's 0. utility to also decrease utility for the wife, for a given level of resources. first-order conditions can be rearranged to deliver PK (v,v)=qK 2 K K 2 1 (v,v) 2 (v,v) = " E p (w(9),w(9)) (16) Kx (w{e),w(0))' K (w(e),w(6)) (17) 2 Condition (16) can then be used to argue that a steady-state requires q {K2t } is bounded, is would increase without bound; likewise, q if > {K2t } $, then = $. Indeed, if q < J3, decreases toward zero. then Both situations clearly do not lend themselves to the existence of an invariant distribution for (v,v). On the other hand, distribution When K 2 q is = if q = (3 K (v t ,Vt) 2 is constant along the optimal path and an invariant possible. (3, the state (v t ,v t ) moves along a one-dimensional locus given by Intuitively, since (vq,vq). then no incentives are required for the wife, she the sense that the marginal cost of delivering welfare to her Figure 5 shows that the curve K 2 (v,v) = curves from below, and cuts Ki(v,v) perfect insurance for the = K 2 from above. 2 (v,v) = perfectly insured in held constant across time. is continuation (v ,i>o) for is K utilities cuts the isocost Intuitively, incentives require foregoing husband and accepting fluctuations Starting from (v* v*), rewards can be delivered in two ways. , in v as rewards and punishments. The optimum makes use of both forms of rewards, explaining the shape of the schedule for continuation utilities. The first form of rewarding involves increasing resources K, and can be seen as an upward movement along the diagonal Ki(v,v) allocation of these resources that is along the Pareto frontier, which at rewards and, as a above the more 0. However, the husband is also rewarded by allowing an to his liking, which can be represented as lateral (v*, v*) is horizontally flat. result, (v,v) travels initial isocost curve. = along K 2 (v,v) = Note that punishments section of the Pareto frontier. Thus, ex-ante efficiency 13 K 2 The (v ,v will ) movements solution combines both forms of to the right of K\{y,v) and push the agent on the upward sloping demands ex-post inefficiency. 13 Returning to the analogy in footnote 12: parents often complain that the punishments they choose to their children hurt them more than they do their kids. 21 = inflict, on Figure Condition (17) is the —Ki/K2 and eventually husband with Isocost curves of Here, it v) implies that the slope of the isocost curve (Pareto reverts geometrically toward 0. Thus, (vt ,vt ) moves along K2(v,v) = K2(vq,vo) reverts toward (v* v*). Intuitively, the solution deviates from (v*, v*) to provide the , incentives, but it is efficient to revert the wife: Patience ensures that the wife has her 6 K(v, analog of the conditional-expectation equation (12) obtained from the one-dimensional Bellman equation. frontier), 5: way back to this point of maximum efficiency for in the long-run. Estate Taxation We now turn to a repeated Mirrleesian economy and study optimal taxation. In this version of our model, individuals have identical preferences over consumption and work regarding their labor productivity, which distributed across generations is effort but are heterogenous privately observed by the individual and dynasties. We and independently continue to focus on the case where the social welfare criterion discounts the future at a lower rate than individuals. Unlike the taste-shock model, here even between parent and child, there would still if we were to restrict allocations to feature be a non-trivial planning problem. period the situation would then be identical to the static, always optimal to With link Indeed, in each nonlinear income tax problem originally studied by Mirrlees (1971). Moreover, in the absence of altruism, so that actually coincides with this static solution. no altruism, however, (3 we = 0, the social shall see optimum below that it is link welfare across generations within a dynasty to enhance incentives for parents. Despite differences between the Mirrleesian economy and our taste-shock model, our previous analysis can be adapted virtually without change. 22 In particular, a recursive representation can be derived, and the Bellman equation can be used to characterize the solution and to establish that a steady-state, invariant distribution exists. This highlights the fact that our model requires asymmetric information, but not any particular form of We some it. focus on an implementation of the allocation that uses income and estate taxes, and derive interesting results for the latter. We find that estate taxation should be progressive: fortunate parents should face a higher average marginal tax rate on their bequests. mean reflects the reversion in consumption explained in the previous section. A more This result higher estate tax ensures that the fortunate face a lower net rate of return across generations, and that consequently their consumption path decreases over time toward the mean. Repeated Mirrlees: Productivity Shocks Each period economy of this depends on the Utility generation t is identical to the canonical optimal taxation setup in Mirrlees (1971). level of consumption c and work effort n. t differ t t t in ], regarding their productivity in translating work effort into output. productivity w, exerting work effort n, produces output y is assume that individuals have identical preferences that satisfy V =E _ 1 [u(c )-h(n )+pVt+1 but We = wn. We An individual with assume that productivity w independently and identically distributed across dynasties and generations. Thus, the produc- tivity talents of parent assumption, if the and optimum child are unrelated skills are represents a assumed physical environment. 14 = n 1 /j so that, defining 9 = iy~ 7 , write total utility over consumption and output as being subject to taste shocks oo oo Y^pE^u^-eMyt)} t=0 The it this with incentives in this way, and not a mechanical result For convenience, we adopt the power disutility function h(n) we can assumed nonheritable. Given features intergenerational transmission of welfare, then social decision to provide altruistic parents originating from the —innate =^^E_ [A(c -i)-^% 1 i i )] -E_x[0 o %o)] t=l right-hand side of this equation leads to a convenient recursive representation of the planning problem in the continuation utility defined are abusing notation slightly by folding the The by v _1 /? t = S YlT=o0 E(-i[«(ct+ s -i) — 6t+sh(yt+ s )] (where we into the definition of the utility function u(c)). resource constraint requires total consumption not to exceed total output plus some fixed constant endowment f 14 As Y^ in the taste Albanesi and Sleet. cV et t( ) Pr (^) (tyiv) < f J2 v et t( ) Pr (# £ ) d ^( v ) + e i = 0, 1, . . = j3 leads to immiseration. This case has been studied by shock model, here the case with who impose an exogenous lower bound on dynastic welfare to circumvent immiseration. (2004), 23 where individuals are indexed by their initial utility entitlement v, with distribution in the ip population. We continue to assume social discounting problem is t is > lower than private discounting: t to choose an allocation {c^(9 ),y^(9 )} to maximize average (3. The planning social welfare subject to the incentive-compatibility constraints and the resource constraints. Using the the utility function and applying similar reasoning as in the taste- last expression for shock model yields the Bellman equation = max k(v) for the associated relaxed E[w_ - Xc(u_) - 8h{9) + Xy(h{8)) U-,h,w = v -8h{d) E[u_ - 8h(6) + (3w{6) > problem + J3k(w(8))] + (3w(6)} -9h{9') where the function y(h) represents the inverse of the + Pw(8'), disutility function, y = h~ l that justify the study of this Bellman equation, are similar to those that underlie in the context of the taste-shock model. previously, The . The arguments Theorems 1 and 2 results regarding steady states parallel those obtained and imply that an invariant distribution exists with no immiseration, as in Proposition 2. Implementation with Income and Estate Taxation Any allocation that is incentive compatible taxes on labor income and estates. Here and we can be implemented by a combination of feasible describe this implementation, and explore first some features of the optimal estate tax in the next subsection. For any incentive-compatible and feasible allocation {cvt {9 tation along the lines of Kocherlakota (2004). t i ), In each period, conditional on the history of their and any inherited wealth, individuals report dynasty's reports 6 we propose an implemen- y^{9 )} shock 9 their current t , produce, consume, pay taxes and bequeath wealth subject to the following set of budget constraints ct where Rt-i,t +b < t l y t (9 ) -T t (0<) + (1 - n^R^A-i i? t _ 1}4 6 t _i at rate 7^(0*). Given a tax policy {Tl'(9 {R t ,t+i}] an allocation for t t ) ) T (9 t ), (18) . = 0. Individuals are and a proportional tax on inherited r^fl*), yl{9 )} ) an equilibrium consists of a sequence of interest rates consumption, labor income and bequests {cv {9 strategy {cr^(9 )} such that: . ly t t 15 = 0, 1, the before-tax interest rate across generations, and initially 6_i is subject to two forms of taxation: a labor income tax wealth t (i) {c t , bt , cr t } maximize dynastic t ) ) ^(O 1 )}; and a reporting utility subject to (18), taking the In this formulation, taxes are a function of the entire history of reports, and labor income y t is mandated given However, if the labor income histories y ( 0* —» R* being implemented are invertible, then by the this history. : we can rewrite T and r as functions of this history of labor income and avoid having to mandate labor income. Under this arrangement, individuals do not make reports on their shocks, but instead simply choose taxation principle a budget-feasible allocation of consumption and labor income, taking as given prices and the tax system. 24 {R sequence of interest rates clears so that fE_i[b^(9 incentive compatible if, t )} t^ and the tax policy {Tt r t yt } t+ i} d(p(v) , = in addition, for all i it = 0, 1, induces truth For any feasible, incentive-compatible allocation . We . . , for any reporting strategy b^(9 t ) = satisfies cr, 1 is we construct an incentive-compatible ) = — 4 yt(# ) ct (9 t ) and v u'(c t {8 t {i? t _i it }. say that a competitive equilibrium {c", y^}, PH -i,t any sequence of interest rates the asset market (ii) telling. competitive equilibrium with no bequests by setting T^{9 for and as given; )) These choices work because the estate tax ensures that the resulting consumption allocation {c"(cr f (#'))} with no bequests the consumption Euler equation U'(c?(aW)-)=/?iWE u '(^ Ot+i The labor income tax is such that the budget constraints are satisfied with this consumption cation and no bequests. Thus, this no-bequest choice is optimal for the individual regardless of the reporting strategy followed. Since the resulting allocation follows that truth telling is optimal. then ensure that the asset market As noted above, in our The resource clears. is incentive compatible, economy without capital only the after-tax interest rate matters so the In the next subsection, set the interest rate to the reciprocal of the social discount factor, Rt-i,t it it constraints together with the budget constraints t natural because by hypothesis, 16 implementation allows any equilibrium before-tax interest rate {i2 _ 1)t }. we allo- — $~ l - This choice is represents the interest rate that would prevail at the steady state in a version of our economy with capital. Optimal Progressive Estate Taxation In this subsection we derive an important intertemporal condition that must be satisfied by the optimal allocation. This condition has interesting implications for the optimal estate tax, computed using (19) at the optimal allocation. Let A be the multiplier on the promise-keeping constraint and on the incentive constraints. Then the first-order conditions for let n(6, 9') represent the multipliers u_ and w(9) are c'(u_)-A-A = 0V(w(9))p(9)-p\p(e)-J2M6') + 'E ti0',O) = l 16 Since the consumption Euler equation holds with equality, the same estate tax can be used to implement alloca- tions with any other bequest plan with income taxes that are consistent with the budget constraints. 25 and the envelope condition is k' — (v) Y k'{w{e)) P {0) J Using c'(w_) = l/u'(c_) = A negative arrive at the Modified Inverse Euler equation first (3 > side tax r(^ ,6 t ) , is novel, since is the standard inverse Euler zero it is when /? = (3 and is strictly Xl (3. In our environment, the relevant past history f_1 V^ term on the right-hand side The second term on the right-hand when ik'(v), JY"W) left-hand side together with the equation. = p + k'(v) we u'(c_) The Together these imply A. is encoded in the continuation utility so the estate can actually be reexpressed as a function of t_1 t> t (0 ) and 8t Abusing notation we . then denote the estate tax by T t (v,9 t ). Since we focus on the steady-state, invariant distribution, we also drop the time subscripts The average and write t(v,9). estate tax rate f(v) 1 is then defined by - f(v) = £ (1 - t(v,6)) P (6) (21) e Using the modified inverse Euler equation (20) we obtain f(v) = -AV(c>))(J-l) In particular, this implies that the average estate tax rate However, are subsidized. implementation. is negative, f(v) recall that before-tax interest rates are not As a consequence, neither are the estate taxes < 0, so that bequests uniquely determined in our computed by (19). particular choice for the before-tax interest rate, however, the tax rates are pinned acquires a corrective, Pigouvian role. externalities down and Differences in discounting can be interpreted as a form of from future consumption, and the negative average tax can then be seen as a way of countering these externalities as prescribed by Pigou. In our setup without depends on the choice of the before-tax interest be a robust steady-state outcome In our With our model of past shocks it is more encoded rate. in a version of our in the promised continuation function of consumption, which, in turn, is progressive: the average tax on transfers for However, the negative tax on estates would economy with interesting to understand capital, this result how capital. the average tax varies with the history utility v. The average tax an increasing function of v. more fortunate parents higher. is is an increasing Thus, estate taxation is ^This equation can also be derived from an elementary variation argument. This is done in Farhi, Kocherlakota and Werning (2005), who also show that this equation, and its implications for estate taxation, generalize to an economy with capital and an arbitrary process for skills. 26 Proposition 4 In economy, the optimal allocation -can the repeated Mirrlees be implemented by a combination of income and estate taxes. At a steady-state, invariant distribution average estate tax f(v) defined by (19) and (21) The is ip* the optimal increasing in promised continuation utility v. progressivity of the estate tax reflects the mean-reversion in consumption. must face lower net , The fortunate consumption path decreases towards the mean. 18 rates of return so that their Conclusions 7 Should privately-felt parental altruism implications for inequality? affect the social contract? To address these questions, what are the long-run If so, we modeled a central tension in society: the tradeoff between ensuring equality of opportunity for newborns and providing incentives for altruistic parents. Our model's answer is that society should indeed exploit altruism to motivate parents, linking the welfare of children to that of their parents. However, of future generations directly, the inheritability of we also find that if we value the welfare good or bad fortune should be tempered. This produces a steady-state outcome in which welfare and consumption are mean-reverting, long-run inequality What of our is bounded, social mobility is possible and misery avoided by everyone. is instruments should society use to implement such allocations? For a Mirrleesian version model we progressive: find an important role for the estate tax. more fortunate parents should The optimal tax on inheritances is face a higher average marginal tax rate on their bequests. This result illustrates an interesting way in which the conflict between corrective and redistributive taxation is optimally resolved. Further examination of other situations with similar conflicts remains an interesting direction for future work. 19 Appendix Proof of Theorem Weak 1 concavity of the value function k{v) follows because the relaxed sequence problem has a concave objective and a convex constraint continuity over the interior of its also its domain. be established as follows. Define the k*(v) set. The weak concavity If utility is of the value function k{v) implies bounded, continuity at the extremes can first-best value function = max^T/^E-x^u^) - Ac(u t (0*)) ] 4=0 subject to v at any finite = X^o^'^-il^*^*)]- Then extremes v and 18 Farhi, Kocherlakota 19 Some v. Then k*(v) is continuous and k(v) < k*(v), continuity of k(v) at finite extremes follows. and Werning (2005) explore more general versions of this result Thus, k(v) and discuss other progress along these lines can be found in Amador, Angeletos and Werning (2005). 27 with equality is intuitions. continuous. The constraint (6) with q Toward a = $ and continuation implies that utility utility are well-defined. contradiction, suppose T ]T/?'£a lim T—>oo t u(ct) t=o is not defined, for some utility is concave 9u(c) s > — 1. This implies that limx^oo ^] t=0 < Ac + B for B> some A, so it ^2(3 m&x{Es 9 u{c t t t ),0} ^ t=0 $* = E_iCf i A^(?E t ), = 0} oo. Since a ct +B 4=0 t=0 limit yields lini7'_ 00 l T < ^ ]P f? E s q + S < t=0 max{£'s 6 u(c follows that T T J Taking the 0, i /? oo. Since there are finitely many s histories S+1 £ X^t=o^*^-i c< = °°- ^ there is a non-zero measure of such agents this implies a contradiction of (6). Thus, for both the relaxed and unrelaxed problems utility and continuation this implies limj^oo utility are well defined given the other constraints on the problem. This is important for our recursive formulation below. We next prove two lemmas that imply the rest of the theorem. Consider the optimization problem on the right hand side of the Bellman equation: supE[0u(0) - \c(u{9)) + J3k(w{6))] (22) u,w v 6u Define m = max c > Oi e e (0) e(0w(c) = + pw(6) > — Ac) E[6u(6) Qu and k + pw(6') (0') (v) + 0w(6j\ = k (v) (23) e for all 0, 6' — m/(l — $) < (24) 0. The problem in (22) is equivalent to the following optimization with non-positive objective: supE[0u(0) -Xc(u(9)) -m + pk{w{9))} (25) subject to (23) and (24). Lemma Proof. A.l The supremum If utility is and compactness in (22), or equivalently (25), is attained. bounded the result follows of the constraint set. arguments apply when utility is immediately by continuity of the objective function So suppose utility is unbounded above and below only unbounded below or only unbounded above. We — similar first show that lim k(v) v—*oo and then use this result to restrict, maximum attained. is without = lim k(v) v—>— oo loss, = -co the optimization within a compact 28 (26) set, ensuring a To establish these limits, define oo = sup J2 P h(v; 0) l - E-i [QMS*) m - Ac(u(0*)) t=o = subject to v t E_i Ylt^oP ^tu(9 incentive constraints it t Since this corresponds to the same problem but without the )- < follows that k(v) h(v,J3). — then the desired limits (26) follow. Since 9u h(v,P) < h(v,0) — Xc(u) = lim,,-^ h(v, If < ra and < ft lim„__ 00 h(v, ft it = — oo, J3) follows that — 777 - XC(v,P) - v = J3) (27) where oo C(u,/?) = V^'E_ic(u(0')) inf i=0 = subject to v t t Yl'tLoft '^-i[6t'u(9 problem, with solution «(#') such that lim„__ 00 ^(v) = = Note that C(v,/3) )]1 and lim _ 00 7 i; C(v,P) Fix a = this defines a closed, bounded > we can Xc(u)) is this defines a closed, that we can in turn, imply the limits (26). interval for w(9), for each restrict the bounded restrict the is M < |if(0)| 9. VyW It with (9u — Xc(u)) interval for u(9), for each maximization to \u(9)\ < M v>u 9. continuous over this restricted Lemma set, the A. 2 The value function k(v) Proof. Suppose that for concave with the limits is when > maxE[0u(0) - Xc(u(9)) u — > Thus, there exists an set. < > k/p(9). v>w 00 m — 00 or u — M 00 such ViU < > —00, Since the objective function be attained. Bellman equation Xc(u{9)) 29 — either (7)-(9). some v k(v) M . maximum must satisfies the (26), . Hence, we can restrict the maximization in (25) to a compact is 00 be restrict the follows that there exists — —00 > we can non-positive, maximization over u{9) so that 9u(9) strictly concave, < the constraints (23) and (24) and let k k/(j3p(9)). Since k{w{9)) such that we can restrict the maximization to — Using the inequality (27) this establishes Then, since the objective maximization to w{6) such that k(w(9)) Since (9u (6 1 (v)))~ 1 the corresponding value of (25). Similarly, 1 /3) verifies and Then 00. = ±-E[cfe)1-/3 T Take any allocation that v. (v) first-best allocation positive multiplier j(v), increasing in v = —00 and lim^oo h(v, (3) = —00. = —00 and hm ,_ 00 h(v,$) = —00, which, so that lim^^.oo h(v, linv^-oo h(v,$) some , (c')~ (9 t y(v)) for a standard convex is + J3k(w(9))] where the maximization Then subject to (23) and (24). is k(v) > E[6u{9) - there exists A> such that + J3k(w(6))] + A Xc(u{9)) (u,w) that satisfy (23) and (24). But then by definition for all > Y^P^-ilOMO*) - k{w(6)) for all allocations u that yield w (9) Ac(«t(^))] and are incentive compatible. Substituting, we find that oo k(v) > J^p'E^leMe^-xc^tie 1 ))] a contradiction with the definition of k(v). for all incentive-compatible allocations that deliver v, Namely, that there should be a plan with value arbitrarily close to maxu w E[6u(9) - Xc(u(9)) , By and + J3k(w(9))] definition, for every v delivers v and e > subject to (23 +a ) and there exists a plan k(vo). We conclude that k(v) < (24). {u^Q 1 ; v, e)} that is incentive compatible with value oo ]T ft E_j [OMe v,e)-X 1 c(u 4 (^; v,e))] ; > k(v) - e. t=o Let (u*{6),w*(9)) E Sirgmax UiW E[9u(9) andut(0') = ut-i((0 u . . . - + J3k{w(9))]. Xc(u{9)) ,9 t );w*(6 ),e) for t > Consider the plan u (9Q ) = u*(9 ) Then 1. oo = E_xk^(0o)-Ac(u*(0o))+^^^E o [0 m ^ +1 (0 i+1 )-Ac(^ +1 (0 t+1 ))] t=0 > Since e (23) > and maxE[0u(fl)-Au(0)+/3/c(w(#))l was arbitrary follows that fc(u) > ma,x u<w E[9u(9) — Xc(u(9)) + J3k(w(9))} subject to (24). Finally, together to (23) it - /k and both inequalities imply k(v) = max U}W E[9u(9) — Xc(u(6)) + j3k(w(9))} subject (24). Proof of Theorem 2 Part (a). Suppose the allocation {u incentive compatible and t } is generated by the policy functions starting from v delivers lifetime utility v 30 . , is After repeated substitutions of the Bellman equation ( / ) we , arrive at T = Y,P ^-i{8Me )-\c{u t k(v Q ) t t t {d ))} + E- k(v T {e T )). (3 (28) 1 t=o Since k(v is ) bounded above this implies that oo k(v < ) YtP'E-iWMe 1 ) HMO 1 ))}, 4=0 so {ut } optimal, by definition of k(vo). is Conversely, suppose an allocation {u t } is optimal given v Then, by definition . it must be incentive compatible and deliver utility vq. Define the continuation utility implicit in the allocation w = J^p (9 Q ) t -1 E o [9 t u t (0 t )}- t=i and suppose that plan {u t } u either u ^g (9) (9; u incentive compatible, is v (9) ) w (9) ^ g w (9; v for some 9 £ 0. and w (9) satisfy (23) and (24). The or ), Since the original Bellman equation then implies that k(v E[g u (9;v )-Xc(g u (9;v ))+Pk(g w (9;v = ) > E[u - Xc(u (9) ))} + Pk(w Q (9))] (9)) oo > E_j [uo(6 ) - \c(u + X>* E-i [MO') ~ HMO'))} (9 ))] t=i The argument applies > 1. We Part if (7), while the second inequality follows the Thus, the allocation {u t } cannot be optimal, a contradiction. definition of k(wo(6)). t does not maximize inequality follows since u first the plan A similar not generated by the policy functions after some history 9 is l and conclude that an optimal allocation must be generated from the policy functions. (b). First, suppose an allocation {u t v t ] generated by the policy functions (g u , at Vq satisfies lim t _ oo t /? E_iu t t (0 ) = Then, 0. after repeated substitutions of (8), , g w starting ) we obtain T r = Y,P ^-A e tMO t t )\ + P T E^[vT (9 T (29) )}. t=0 Taking the limit we get v v Next, . Hm -,ooP t t we show ^~iVt{9 Suppose lim supj^oo t ) /5 E_iv '^ , -i[OtM0 t )] so that the allocation that for any allocation generated by (g ,g = (9 w {u t } delivers lifetime starting from finite v ), , utility we have 0. unbounded above and limsup^^ t t t Yl'tLoP u utility is ( = ) = oo. |t /?' E_it> t (6 Since the value function k(v) 31 is ) > 0. Then >P implies that non-constant, concave and reaches an interior maximum, we can bound t <alimsup 5 E_ limsup/J'E-ifc^fl*)) / (28) implies that k(v We finite values. Similarly, = ) liminf t _ 0O /3*E_ 1 'U f (<9 < ) with a b, < 0. Thus, + 6 = -oo u i (^) /3' E_iut (0') < 0. unbounded below and that liminf utility is = + -co, a contradiction since there are feasible plans that yield conclude that lim sup^,^ suppose 1 av t—too t—>oo and then < the value function so that k(v) < -co. Using k(v) + av b, with a > 0, i _ 00 /3 t < E_it't(^ ) < > 0. 0. Then Then after we conclude that = -oo liminf/^E.i/cfWfl')) t—>oo implying k(v The two Part = — oo, ) a contradiction. Thus, we must have liminf established inequalities imply lim^oo/? (c). t Suppose limsup 4 _^ 0O repeated substitutions of E_if /3 t (a'(^ 4 E_ |t > t )) u t (6 1 for ) = i t _too /? E_iU£ (^ t ) 0. every reporting strategy a. (9), T T v>Y,P' E -i [°^t (*W)] + f E-iVr {° {0 T )) t=o implying T v Therefore, {u t } Proof of Part (a) is > incentive compatible, since v Lemma is u, (^(fl*))] t • attainable with truth telling from part (b). 1 t t (Strict Concavity) Let {u t (9 ,Vo),vt (9 ,v )} be the plans generated from the policy no claim of incentive compatibility functions starting at vq (note: utility values v a and Vb, ^ with v a v?(9 2 part (b) implies that immediately implies that {u°(9 there exists a finite time T t Define the average v^. <(0') Theorem ]T 0- E_i [0 lim inf l ) = au = t {9\v a ) avti^Va) {«j(#\fa )} and + (\ is required). - a^O ,v 1 b) + (1-0^(9*;^) {«((#*, i> b a )} deliver = av a + va and v b respectively. This , (1 — a)v b such that t u l t (9 ; va )}y^J2 i=0 t=o 32 initial utilities )} delivers initial utility v ]T 0* E_i [6 Take two p* E_! [OMB*; v b )} . . It also implies that so that ut for some and history i+1 e 6>* t (9 ;v a )^u T Consider iterating . t (6 -v b ), t (30) times on the Bellman equations starting from v a Vf,: T = £]/?*£_! [0^(0*;^) -c(u k{va ) (0'; t Ua ))] +^ T E_ 1 fe(u T (0 T ; Ua )) 4=0 T = £)i&'E_ 1 pt w fc(v 6 ) t (e*;«i,)-c(Ti t (fi r :r ;^))]+ & E_ 1 A:(wr^ ;«6)), t ) t=o and averaging we obtain ak(v a ) + (l-a)k(v b ^^E.!^^ )- [ac(^(0 = ) 4 i ;t; Q )) + (l- a )c( Ut (^;^))]] 4=0 +£ T E_x [ak(v T (9 T ; v a )) + (1 - a )^ r (# r «>))] ; T J^'E^u?^) -Qc^f^*))] < + T E_ k(v$(9 T l )) t=0 where the that from the strict inequality follows we have the inequality (3(1), concavity of the cost function c strict and the weak concavity of the value function k. inequality follows from iterating on the Bellman equation for v satisfies the is Bellman equations constraints at every The , the fact last weak a a since the average plan (u ,v ) This proves that the value function k(v) step. strictly concave. (b) (Differentiability) Since the value function k(v) is, a (it) there is concave, it is sub-differentiable at least one sub-gradient at every point v. Differentiability can then is —that be established by the following variational envelope arguments. Suppose first that utility is unbounded below. Fix an interior value v for initial utility. For a neighborhood around vo define the test function W(v) Since W(v) is = E [9(g u (9, v + (v- (see v Q )) - Xc(g u (9, v ) + (v - v the value of a feasible allocation in the neighborhood of with equality at v Theorem ) . Since W'(vq) exists Theorem 4.10, in Stokey, k'(v Finally, since c'(u) > this by the upper bound k(v) < ) = + 0k (g™ (6, v i> it follows that W'(v Q ) =1- k'(v) + m/(l — < J3) 33 ))] W(v) < k(v), by application of the Benveniste-Scheinkman follows, Lucas and Prescott, 1989), that k'(v shows that h(v,0) it )) ) also exists and A E[c'(u*(9))}. 1. The limit lim^-oo (31) k'(v) = 1 is introduced in the proof of Theorem inherited 1, since Urn,,—.*, Jj;/l(u,/?) The it is = 1. = — oo limit lim„_^ c k'(v) inherited by the upper since. lim^oo = -^h(v,P) long as g u (9,v ) > bound Then + m/(l — $) h(v,P) 9 0, for all loss in generality utility function, so , interior t; t t any oo. = Otherwise 1, )) the plan {{v/va)u t } is incentive compatible as is required. set. > for all reporting strategies a solution {u t } to the planner's sequence problem 2, utility are possible here a different envelope argument bounded below, then limsup^^ E-iP v (a(9 Theorem < suppose that the G 0. However, corner solutions with g u (9,vo) t that, applying u if introduced in the proof of Theorem provide one that exploits the homogeneity of the constraint If utility is = — oo, the argument above establishes differentiability at a point v even with Inada assumption on the We < bounded below, and without is zero. is k(v) from lim^^ k(v) -oo. Next, suppose utility of zero consumption follows immediately is a so ensured. Then, for and attains value v for the agent. In addition the test function W(v) = J2~PtE 9t - f=0 satisfies W(v) < W(vq) k(v), Scheinkman Theorem, that The proof of lim^ s we show = = — oo and fc(i>o) k'(vo) exists k'(v) that lim^^^ k'(v) = t t {9 )-\c(-u v \v is differentiate. and equals W'(v the is -u same t {9 t ) follows It from the Benveniste- ). as in the case with utility unbounded below. Finally, problem oo. Consider the deterministic planning oo k(v) = meix2_]p{ u ~ t Ac(u t )) t=o subject to v = P tu t- Note that k(v) is plans are trivially incentive compatible, it Yl'tLo must have limv -, u k'(v) = CLAR with \im v -> u tf(v) follows that k(v) < = oo. Since deterministic k(v), with equality at v. Then we oo to avoid a contradiction. Proof of Proposition The differentiable 1 equation was shown in the main text, so we focus here on the bounds. Consider the program maxVpn {0 n un ii. in < -c(u n +(3k(w n )} ) n V = Y^Pn(9nUn + P^n) n 9 n un This problem and constraint on u. its But of neighboring agents + Pwn > 9 n u n+1 + Pw n+X for notation require some discussion. this notation allows us to consider is n = We 1, 2, . . , . K- 1, do not incorporate the monotonicity bunching in the following way. bunched, then we group these agents under a single index and 34 If any let set pn be Likewise the total probability of this group. this group, which is what is let 9 n represent the conditional average of within relevant for the promise-keeping constraint and the objective function. Let 9 n be the taste shock of the highest agent in the group. The incentive constraint must rule the highest agent in each group from deviating and taking the allocation of the group above him. Of course, every combination of bunched agents leads to a We program. different study all of them. The optimal allocation of our problem must solve one of these programs, although not necessarily the condition The one that yields the highest value, since is may not be feasible if the monotonicity violated. first-order conditions are Pn{6n - - Xc'{u n ) + Qn A0 n } Pn0k'(w n - PX} + = where, by the Envelope theorem, A Consider first case with utility tion holds with equality. n 1 ' = - n _ 1/u n _ 1 Mn-:) < = so that the first order condition for consump- the first-order conditions for consumption, we get =l-k'{v) \E[c'(u(6))} first-order conditions for - n k'(v). unbounded below, Summing \i /?(//„ ) The one this imply 1 OiK Xd( Ul ) XE[c'(u e )] 1-A 0i pi h h e1 This implies ^<^-(i-a£, Using PPI p we get p fc'M > p -^"-4 8 x) Ok-i /%-i 6K Pk = + n = 01 01 K ^ (uk) , we > 0i Pk P "0i P A +£ J XE[d(u g )] 1-A K K K 1 ~ X > -(1-A); ~ Ok-i 35 N get This implies V-K-l ,// k'{v) 7T~ P Similarly, writing the first-order conditions for (1 1 l TK 7 K-1 — — 1 1 0i J Using k (w K ) we = —A + - — P Pk get k' (%) < _I^ + k'(y) e;-o; P _ A A) 7 &k-i we have Wi, Qx 1 + (1 @K-\ wk < w n < For any n, 1 A -r 1 ft +I PY 0i k'(wn ) < P P 1 ®K-\ (l-fc») + l-|P > l-k'(gw + 1 arrive at the expression in the text Turning to the bounded w we take the worst most unfavorable to each bound, noting that satisfied + K TK-\ @K-\ @K-1 (9,v)) when A > implies k'(w(9)) 1 with = utility case, /i n = = all eK Qk-\ & K-\ — case scenario: k'{y) > - k'(v)) +i-L we choose the subproblem 0. the first-order conditions and constraints are w (6) = P~ l v > v. (P/P)k'(v). Since the problem is strictly convex, this represents and u(9) = k'(0~ v) note that 1 1 (i P is 1 k'{v) Oi > P that Ok-i we obtain + P To eK 1 + 1 P After rearranging, P 6k-i 1 n < ex ~6k I k'(v) ®k-\ 'K-l for every K u and The first-order condition for the unique solution. Recall that in the arguments above establishing the lower bound involved no assumption on interior solutions did require u{9) >u for all 9, for u, so this holds for all v. which must be true for The upper bound, on the other hand, high enough v, i.e. for low enough k'(v). Proof of Proposition 2 Consider first the case with utility unbounded below. Since the derivative k'(v) strictly decreasing, we can x < 1 if utility is continuous and define the transition function Q(x,9) for all is = k'{g w ((k')- 1 {x),9)) unbounded below. For any probability distribution probability distribution defined by tq(v)(A) = / l {Q{Xt o )eA) 36 dn(x)dp(6) fj,, let Tq(/i) be the any Borel for set A. Define + 7g + ---+73 rq TQ n = ' For example, Tq with k'(v = ) the empirical average of {k'(v t )}?=1 over {5x ) is iTL The x. n following lemma histories of length all n starting establishes the existence of an invariant distribution by considering the limits of {Tq iTI }. Lemma A. 3 If utility is that converges weakly, Proof. The bounds unbounded below, then for each x < in distribution, to i.e. an invariant distribution on (—00, (13) derived in Proposition lim<2(z,0) = lim imply that 1 k'(g w {9,v)) {Tq^ u )(5 x )} there exists a subsequence 1 for all 9 = 0/J3 < 1) under Q. 6 1. l)x0-> (—00, 1) to a continuous transition function Q(x,9): (-00, 1] x0-t (— 00, 1), with Q(l,8) = P//3 and Q(x,6) = Q(x,8), for all x € (—00, 1). It follows that TU maps probability distributions over — 00, 1] to probability We first extend the continuous transition function Q(x, 6) (—00, : ( distributions over (—00, 1), and Tq(S x ) = Tq(5 x ), for all x € (—00, 1). We next show that the sequence {Tq u (S x )} is tight, in that for any e > set A e such that TU (5 X )(/LJ > 1 — e, for all n. The expected value of is simply E-Wivtie*' 1 ))] with x = k'(v ) < 1. Recall that E-i^Mfl* there exists a compact the distribution T?(6 X ) -1 ))] = 1 (P / J3) k' {v ) -» 0. This implies that min{0,fc'(^o)} < E.!^'^^*- 1 ))] I ?3(<y (-00, -A)(-A) + for all A> 0. Rearranging, 1 isw(-°o.-a)< -^ tight. is Tightness implies that there exists a subsequence T^ ^ weakly to some distribution n. Since Q(x,6) But the Tq(tt). linearity of Tq is and since 0(n) Recall that (-00, for 1). — > 00 we must have Tq{tv) Tq maps it continuous in ~ = x )-Tq(§ x ) tv = Tq (tt) follows that 7r that converges weakly, x, then Tq{Tq^^{5 x )) + TQMn)V x ^y- i.e. in dis- converges ) n. probability distributions over (-00, This implies that such distributions, n(<5 x ) implies that T^)+1 (5 TQ(TQ M n)(8*)) = °-'> 1 which implies that (T^(<5 X )}, and therefore {TQ n (5 x )}, tribution, to - T?(5x )(-oo, -A))l (1 1] to probability distributions over has no probability mass at {1}. Since Tq and Tq coincide = Tq(tt), so that under Q. 37 n is an invariant distribution on (-00, 1) The argument for the case with utility bounded below function Q(x,8) as above, but for is unbounded above but 7 < v. Define the utility level v \im v ^y g vl > bounded above then is w (0, = v) let argument is w is > w we must have that v continuous with Q(x,9) then a simple modification of the one above for playing the role of 1 If utility 00 for the ergodic set for minimum of the policy continuous over this compact is w of g over v e minimum v^ by the Finally, the transition function v. < Next, define vl to be the 1. Define the transition can take on any real value. k'(v) defined since g is In both cases, since g v. = v by k'(vo) function g w over v e [fo,^], which utility now since very similar. then there exists an upper bound v H 1 > i£E, all is [vq, v), this < which minimum k'(vi) < is is above misery: rest of the unbounded below, with utility If defined since The 00. set. k'iyLj (things are actually slightly simpler here, since no continuous extension of Q required). is Proof of Proposition 3 Consider indexing the relaxed planning problem by by setting A = e _1 , with associated value func- We first show that if an initial distribution ip satisfies the condition J k'(v; e) dip(v) = 0, the solution to the relaxed problem and original problem coincide. We then show that for any tion k(v; then e e). initial distribution Since utility is there exists a value for e that satisfies this condition. unbounded below, we have k'{v t e) ; =E t _i [l — v \c!(u t (#*))] . Applying the law of iterated expectations to (12) then yields t E_ 1 [l-Ac'«(^)] = ^0fc'( U ;e With logarithmic utility c'(u) for allt of 0, 1, Theorem Now / = . . 2, it . The = allocation follows that it k'(v; e) dtp(v) = 0. Note that — — 00. > must k'(v It — j^ = implies /E_i[q] dtp = A -1 = e Lemma 2 below, and applying part (c) k'(v;e)dip(v) by solve the original planning problem. The homogeneity = J incentive compatible is consider any initial distribution k(v; e) e c{u), so that — tp. We argue that + k (v find a value of A = e _1 such that problem implies that of the sequential -r log(e) we can - log(e); 1) is strictly increasing in e —— and log(e); 1 limits to 1 as e — 00, and to —00 as follows that J k'(v ]e )d^(v) defines a unique value of e* for any immediately by using the fact that = 1 J k (v - ^-log(e);l) #(u) = initial distribution k'(-;l) is ip. The monotonicity a strictly decreasing function. 38 of e*(ip) then follows Lemma Proof of (a) If utility is also 2 bounded below, then the unbounded below, but bounded above. Then k'(g w (9, •)) w w that k'(g (9,v)) = 1. It follows that ma,x k' lim^-oo -co such (g v that g w (9,v) (b) If utility is > With logarithmic bounded below, the w (9,v) - g 9*~ l for all pairs of histories w (9,v) and 9 < l~ l . u g (9,v) - (9 v)) 2 7 > > 1 ))] 1 ^*- )] = A. < of Theorem 2. one can show that: log (9/9). follows that It ^E-!^*- t 1 1, (c) 9' u g (9,v) we have lim _ 00 /5* E_ 1 [u limsup^^E-iMff'- (d) If implies attained, so there exists a vl is , 1 The v^ incentive constraint '1 1' 1 1 ) > v^O ) - tA Then 1 Theorem — (9/(3) log(0/0) /^[^(cr'- ^part (b) of utility is continuous and Proposition immediately from part result follows utility this implies that then implies that g So suppose (b). vl- d(u(6)) that is Using the first-order conditions from the proof of Proposition (c) From from part result follows > t (0' -1 )] = 1 )] -PtA. Since 0. lim^^ (3 tA = l 0, it follows 0. then the bound in (13) implies that k' (g w (9, > v)) 1 - (3/(3 and the result follows then we can define n = 1 — (1 — /3/$)/(l — 7), and define vh by k'iyn) = «• Then for all v < v # we have g w (9,v) < v. It follows that the unique ergodic set is bounded above v) > We can now apply the argument in (a) so there exists a, > —00 such that w immediately. If 7 < 1, by vh- vl Proof of Part (a) is (b), recall then all c Lemma g J k'(v) dip*(v) = under the invariant distribution solutions for consumption are interior. with g (9,v) = for all 9 in this for some compact If utility is 9 can only occur for low set. enough ip*. is bounded. For part If utility is unbounded below bounded below, then corner c levels of v, so that g (9,v) Recall that for interior solutions l-k'(v) = XE[c'{g u (9,v))]=XE[c'(u(c(g u (9,v)))) Applying Jensen's inequality we obtain c The u vl- 3 immediate since by continuity of the policy functions, consumption that (9, u E[c{g (9,v))] dP(v)X\ < JE[c'(u(c(g^9,v))))) result then follows since c'(u(c)) is an increasing function of 39 c. #» = 1. is solutions bounded, References [1] [2] [3] Aiyagari, Rao, and Fernando Alvarez (1995), "Stationary Private Information: A Tale of Kings and Slaves," mimeo. [5] Amador, Manuel, and George-Marios Angletos, and Ivan Werning Bernheim, B. Douglas (1989) mimeo. "Intergenerational Altruism, Dynastic Equilibria Studies, and Social v56, nl (Jan), pp. 119-128. Caplin, Andrew, and John Leahy (2004) "The Social Discount Rate," Journal of Political Economy, vll2, n6 (December): 1257-68. Emmanuel, Narayana Kocherlakota and Ivan Werning (2005) work [7] Farhi, [8] Green, Edward tual [9] (2005), "Correc- Atkeson, Andrew, and Robert E. Lucas Jr. (1992) "On Efficient Distribution with Private Information," Review of Economic Studies, v59, n3 (July): 427-53. Welfare," Review of Economic [6] with Albanesi, Stefania, and Christoper Sleet (2004), "Dynamic Optimal Taxation with Private Information," mimeo. tive vs. Redistributive Taxation, "work in progress, [4] Efficient Distributions J. in progress. (1987) "Lending and the Smoothing of Uninsurable Income, "in Contrac- Arrangements for Intertemporal Trade: 3-25. Grunwald, G.K., R.J. Hyndman, L.M. Tedesco, and R.L. Tweedie (2000) "NonGaussian Conditional Linear AR(1) Models," Australian and New Zealand Journal of Statistics, 42(4), 479-495. [10] [11] Golosov, Mikhail, Narayana Kocherlakota, and Aleh Tsyvinski (2003) "Optimal Indirect and Capital Taxation," Review of Economic Studies, v70, n3 (July): 569-87. Kaplow, Louis (1995) "A Note on Subsidizing Gifts," Journal of Public Economics, v58, n3 (Nov): 469-77. [12] [13] Kocherlakota, Narayana (2004) "Zero Expected Wealth Taxes: Dynamic Optimal Taxation," mimeo. Mirrlees, tion," [14] [15] [16] [17] [18] James (1971) "An Exploration in the Theory of A Mirrlees Approach to Optimum Income Taxa- Review of Economic Studies, v38, n2: 175-208. Phelan, Christopher (1998) "On the Long Run Implications ard," Journal of Economic Theory, v79, n2 (April 1998): 174-91. of Repeated Moral Haz- Phelan, Christopher (2005) "Opportunity and Social Mobility," Federal Reserve Bank Minneapolis, Research Department Staff Report 323 (March). Sleet, Christopher, and Sevin Yeltekin (2004) of "Credible Social Insurance," mimeo. Spear, Stephen E., and Sanjay Srivastava (1987) "On Repeated Moral Hazard with Discounting," Review of Economic Studies, v54, n4 (October): 599-617. Stokey, Nancy L., and Robert E. Lucas, Jr., with "Recursive Methods in Economic Dynamics," Harvard Press. 40 Edward C. Prescott (1989) [19] Thomas, Jonathan, and Tim Worrall (1990) "Income formation: An Example of a Fluctuation and Asymmetric InRepeated Principal- Agent Problem," Journal of Economic Theory ' Vol 51(2), pp 367-390. 41 U 8 3 G 025 Date Due Lib-26-67 MIT LIBRARIES 3 9080 02618 3563