Public Debt and Optimal Taxes without Commitment

Begoña Domínguez¤y CREB Parc Cientí…c de Barcelona, and IDEA Universitat Autònoma de Barcelona April, 2002

Abstract This paper studies the optimal time-consistent tax policy for an economy with public debt. In the absence of bonds, the optimal time-consistent tax policy is very di¤erent from that under full-commitment. In particular, it has been shown that the optimal time-consistent tax rate on capital income is di¤erent from zero at the steady state. The present paper shows that this result hinges on the balanced budget constraint. In the presence of debt, we extend the Chamley-Judd result to economies without commitment showing that the optimal time-consistent tax rate on capital income converges to zero. Keywords: Fiscal Policy; Optimal Taxation; Time-Consistency JEL classi…cation: E61, E62, H21, H63
¤

I want to express my gratitude and appreciation to Jordi Caballé for his very useful comments

and kind guidance. I have bene…ted greatly from comments of Andrés Erosa. I also wish to thank Juan Carlos Conesa, Ariadna Dumitrescu, Hugo Rodríguez-Mendizábal and workshop participants at UAB and CREB for their comments. Financial support from the Spanish Ministry of Education through grant AP96-07972685 and from the Spanish DGES under project PB98-0870 are also acknowledged. All the remaining errors are mine.
y

Corresponding address: CREB, Parc Cientí…c de Barcelona, C/ Baldiri Reixac, 4-6, 3rd ‡oor,

B2, 08028 Barcelona, Spain.

1

1

Introduction

This paper studies the properties of the optimal time-consistent taxes for an economy with public debt. The previous literature on optimal taxation without commitment has focused only on economies without public debt with the main result of optimal time-consistent capital taxes di¤erent from zero at the steady state. In particular, Benhabib and Rustichini (1997) obtained that the optimal steady state policy is characterized by subsidies to capital. The intuition for this result is that capital subsidies encourage the accumulation of capital, which could become high enough to act as a commitment device against deviation from the announced policy. In this paper we explore whether the previous result holds when a market for public debt is present. To this end, we allow governments to issue debt. The implications of this new environment are twofold. On the one hand, the time-inconsistency problem could be worsened by the possibility of defaulting on debt payments. On the other hand, governments have a new instrument, debt, to a¤ect the bene…ts and costs of deviating from the announced policy. Given that the government is benevolent and that the choice of not issuing bonds is still a sustainable outcome, the optimal management of public debt alleviates the consequences of the time-inconsistency problem. Our main result is that the optimal time-consistent capital tax turns out to be zero at the steady state. Thus, once governments have the possibility of issuing debt, public debt becomes the central commitment device against deviation. One of the main concerns in the theory and practice of …scal policy is how much capital income should be taxed. Currently, the capital income is taxed heavily in both the U.S. and the E.U. economies. However, the theory prescribes the opposite, that is, capital income tax rates should be set close to zero. This result was …rst shown by Chamley (1986) and Judd (1985), who proved the optimality of a zero capital income tax rate at the steady state for economies with identical and heterogeneous agents, respectively. Later on, Lucas (1990) and Chari, Christiano and Kehoe (1994) extended this result to economies with endogenous growth and to stochastic economies, respectively.1 How to reconcile the observed taxes and the
1

Atkenson, Chari and Kehoe (1999) uni…ed di¤erent extensions of the Chamley-Judd result.

2

theory of capital taxation? An enlightening work by Aiyagari (1995) showed that for economies with incomplete insurance markets and borrowing constraints, the optimal tax rate on capital income is positive, even in the long-run. When capital investment yields pure pro…ts, Jones, Manuelli and Rossi (1997) obtained that the asymptotic capital tax rate is not longer zero. For life-cycle economies, Garriga (2000) and Erosa and Gervais (2002) showed that, if the government has no access to age-dependent taxes, the optimal capital tax is di¤erent from zero both during the transition and at the steady state. Recently, Yakadina (2001) emphasized the role of public debt in the properties of optimal taxation for stochastic economies. She found that, when the amount of debt that a government can issue is restricted by lower and upper exogenous limits, the expected long-run capital taxes are di¤erent from zero. She argued that this result is due to the lack of complete insurance against future uncertainty. A recent line of research has focused on the time-inconsistency problem of optimal taxation to account for actual capital tax rates. A limitation that previous studies share is the assumption that governments can commit to follow a prescribed policy plan. However, it is generally recognized that actual governments have no such a perfect commitment technology. Then, without commitment, and under a policy plan that sets capital taxes to zero and labor taxes to some positive distorsionary amount, the incentives to revise the policy plan and tax the pure rents generated by the capital input could be high.2 In this sense, positive capital taxes could accrue from a zero capital tax announcement that is not credible. Thus, time-inconsistency problems could explain actual capital tax rates. Following this reasoning, we …nd the papers of Benhabib and Rustichini (1997) and Klein and Ríos-Rull (2000). Benhabib and Rustichini (1997) characterized the optimal time-consistent capital taxes at the steady state. They …rst set the optimal taxation problem for an economy with debt. However, for simplicity, the analytical and the numerical characterization
2

In a …nite-horizon economy, the incentives to revise the plan and tax heavily the long-run last

period capital are very high (e.g., Fisher (1980)). However, this is not clear in an in…nite-horizon setup, in which there is no last period and reputation forces come into play. There, the bene…ts from higher than announced capital tax rates could be outweighed by the loss of reputation.

3

of capital taxes are carried out in a framework where governments cannot issue bonds.3 To solve for the optimal time-consistent taxes, they introduced an incentive compatibility constraint for each period into the government optimization problem. This constraint says that the welfare value of continuing with the announced policy must be at least as large as the welfare value of deviating from it. The deviation value is endogenous and depends on the consequences of deviating, which are speci…ed as follows: once the government deviates, individuals expect capital income to be taxed at a maximal rate (equal to one) at all future periods and, therefore, they decide not to save. The solution of this problem yields a steady state. The steady state is called incentive-constrained, if the incentive compatibility constraints bind at the steady state. For utilities that are linear in consumption, Benhabib and Rustichini (1997) showed analytically that, when the steady state is incentive-constrained, the optimal time-consistent capital taxes are di¤erent from zero. In order to study concave utilities and determine the sign of the capital taxes, they calibrated the model and obtained that the optimal steady state policy is a subsidy to capital. Following a di¤erent approach, Klein and Ríos-Rull (2000) studied the optimal policy for an economy without commitment. They focused on the properties of Markov perfect equilibria, which are time-consistent by construction. Throughout the paper, they imposed a balance budget so that governments cannot issue debt. This assumption is made to overcome theoretical and numerical problems associated with allowing issuing debt. They calibrated the model and their results can account for the magnitude of the empirical capital income tax rates. Summing up, from Benhabib and Rustichini (1997) and Klein and Ríos-Rull (2000), we know that time-inconsistency problems may explain the optimality of long-run capital taxes di¤erent from zero for economies without public debt. However, would this result hold in the presence of debt? In other words, is the timeinconsistency or the absence of a market for public debt the basis for this result? The present paper explores the role of public debt and time-inconsistency problems in relation to the steady state optimal taxation. Our model follows the approach
3

We show that the assumption on governments’ inability to issue debt cannot be made for

simplicity purposes since it turns out to be crucial for the results.

4

of Benhabib and Rustichini (1997). 5 . This equilibrium speci…es the consequences of deviating. First. respectively. We show that the optimal management of debt results in steady state levels of capital and debt. Moreover. Therefore. We propose a deviation that implies reversion to a Markov equilibrium. Thus. which will be similar to those of Benhabib and Rustichini (1997). hence. on the transition towards the steady state. The intuition for this result is that the incentives to deviate from the announced policy at the steady state depend only on the steady state level of capital. public debt can make the announcement of a zero capital tax at the steady state credible. If the incentive compatibility constraints are not satis…ed for that level of capital. Section 6 concludes. Second. Our analysis is carried out both for utilities that are linear in consumption and for utilities that are concave in consumption. governments face a constraint that embeds the trade-o¤ between continuing with and deviating from the announced policy. which in turn depend on the transition. which are consistent with a zero capital tax rate. we allow governments to issue debt. Section 4 and 5 characterize the optimal time-consistent taxes for an economy without and with public debt. There is a steady state level of capital for which the optimal steady state capital tax is zero. When choosing the optimal policy. the optimal time-consistent capital tax is di¤erent from zero so as to have no incentives to deviate. The explanation for this …nding is as follows. we consider an economy without a market for public debt. For economies with public debt. Consequently. Section 2 presents the model. capital taxes are di¤erent from zero. The Appendices provide a general characterization of the time-consistent taxes and the policy after a deviation and contain the proofs of the results. We obtain that. Section 3 analyzes the steady state. if the steady state is incentive-constrained. for economies without public debt. the steady state depends on the steady state levels of both capital and debt. We show that the optimal time-consistent capital tax rate is zero at the steady state. the incentives and costs of deviating from the announced policy at the steady state depend on both steady state levels of capital and debt. then the capital taxes can optimally adjust the steady state level of capital so as to satisfy the incentive compatibility constraints. The plan of the paper is the following. the steady state capital and debt depend on the initial conditions and.

1). labor. u0 (ct ) = ¯rt+1 u0 (ct+1 ) . The utility functions u(¢) and v(¢) are u(ct ) = c1¡¾ l1+" t .2 The Model Our economy mimics that of Benhabib and Rustichini (1997). where f is increasing. the representative individual chooses consumption. The representative individual has an instantaneous utility function that depends on both consumption ct and labor supply lt so that his welfare takes the following form: 1 X ¯ t [u (ct ) ¡ v (lt )] . (4) (5) where u0 (ct ) and v0 (lt ) denote the partial derivatives of the instantaneous utility functions (2) with respect to consumption and labor at date t. and assets at+1 so as to maximize his welfare (1) subject to the budget constraint rt at + wt lt ¸ at+1 + ct . concave. t=0 (1) with ¯ 2 (0. 1¡¾ 1+" (2) with ¾ ¸ 0 and " > 0. (3) where rt is the after-tax return on assets at date t and wt is the wage rate net of labor taxes at date t. respectively. The …rst-order conditions for this optimization problem are wt u0 (ct ) = v0 (lt ) . continuous and 6 . a representative competitive …rm and a benevolent government. Taking the government policy as given. and v(lt ) = t . lt ). The production function is f (kt . We consider an in…nite-horizon economy with three agents: a representative consumer. The individual’s assets take the form of capital kt and debt holdings bt so that the following market equilibrium condition holds: at = kt + bt : (6) The representative competitive …rm produces output using capital and labor.

the government could choose an initial capital tax rate so high that the resulting tax revenues could …nance all future government expenditures. Taking factor prices as given. lt ) = wt . we assume that the initial after-tax return on assets satis…es r0 ¸ rmin .homogenous of degree one. the government budget constraint can be written as bt+1 + (fk (kt . In particular. 7 . The resulting …rst-order conditions are ¢ 1 ¡ ¿ k fk (kt . Next. In order to ensure distorsionary taxation. lt ) with respect to capital and labor are denoted fk (kt . a competitive equilibrium for this economy is de…ned as follows: 4 (10) (11) We assume non-negative government bonds in order to help identify a Markov equilibrium to revert to once a government deviates from the announced policy. The resource constraint is kt+1 + ct + G · f (kt . lt ) . the …rst period after-tax return on assets is restricted by a lower bound. the …rm chooses non-negative amounts of capital and labor so as to maximize pro…ts. Hence. lt ) = rt . lt ) ¡ rt ) (kt + bt ) + (fl (kt . lt ) and fl (kt . where rmin ¸ 0 is su¢ciently close to zero. Note that this lower bound can be viewed as an upper bound on the initial capital tax rate. (9) The government can only sell bonds.4 Those bonds are issued with the after-tax return rate rt . In the absence of that upper bound. t ¡ ¢ 1 ¡ ¿ l fl (kt . lt ) ¡ wt ) lt ¸ G + rt bt . The partial derivatives of f (kt . bt+1 ¸ 0. t t The government …nances debt payments and an exogenous public spending G per capita through taxes on labor and capital income and the issue of one-period bonds. respectively. respectively. that is. t ¡ (7) (8) where ¿ k and ¿ l are the capital and labor income tax rates at date t. lt ). We consider an upper bound smaller or equal to one but su¢ciently close to one.

the government takes explicitly into account an incentive compatibility constraint. wt . we specify a general functional form for the deviation value that depends on the current levels of capital and bonds. the government chooses among the time-consistent policies so as to maximize the welfare of the representative individual. :::. For the time being. the budget constraint (3). This deviation value depends on the consequences associated with the deviation from the announced policy. After a deviation. ¿ l t=0 .1 will provide a speci…c form. the exogenous public spendt t sequence fct . for t = 1. and the initial assets b0 and k0 . (ii) factors are paid their marginal products according to equations (7) and (8). rt . bt ) is the value of welfare after a deviation at date t: We assume that V D (kt . and (iii) all markets clear (equations (6) and (11) hold. bt+1 . this optimization problem also satis…es the government budget 8 . bt+1 . which makes positive savings unattractive. 2. The government chooses the sequences fct . kt+1 . 2. the incentive compatibility constraint (12) and the non-negativity constraint bt+1 ¸ 0. bt ) is di¤erentiable in both variables. the lower bound on the initial after-tax return on assets (10). the economy reverts to some bad equilibrium. lt . kt+1 .1 The Government Optimization Problem Without commitment. (12) where V D (kt . individuals expect maximal capital taxes. wt g1 such that: (i) the representative individual maxt=0 ing G per period. The incentive compatibility constraints link current and future governments as follows: 1 X s=t ¯ s¡t (u (cs ) ¡ v (ls )) ¸ V D (kt . which provides a welfare called the deviation value. lt . where after a deviation. a competitive equilibrium is a imizes his welfare (1) subject to the budget constraint (3). In order to do so.© ª1 De…nition 1 Given a sequence of tax rates ¿ k . which says that the welfare value of continuing with the announced policy must be higher than the welfare value of deviating from it. the latter with equality). the resource constraint (11) . bt ) .5 The Lagrangian for this 5 Given equations (3) and (11). Section 5. rt g1 so as to maxit=0 mize the welfare of the representative individual (1) subject to the …rst-order conditions (4) and (5).

debt. t=1 s=t where ¯ t » t . ¯ t ´t . (11). ´t fkt ¡ ¯ ¡1 ´t¡1 + » t (fk (kt . and (10). with £ ¤ 1 + ¯ ¡t (° ¤ ¯)t = 1 + ° 1 ¯ ¡1 + ° 2 ¯ ¡2 + ::: + ° t¡1 ¯ ¡(t¡1) + ¯ ¡t ° t . (12) . 6 These equations are the …rst-order conditions when the optimal debt is non-negative. (14) £ ¤ 00 1 + ¯ ¡t (° ¤ ¯)t v 0 (lt ) ¡ » t (fl (kt . lt ) ´ t . bt ) . bt ) . ¹t u0 (ct ) ¡ » t lt = 0. ¸t¡1 u0 (ct ) ¡ » t (kt + bt ) = 0. (4). these conditions take a di¤erent form (see Appendix 7:1). bt ) . labor. ¯ t ¸t . and ·0 are the Lagrange multipliers associated with constraints (3). return rate. capital. ¯ t ¹t . Moreover. lt ) ¡ wt ) + ¹t v (lt ) = fl (kt . lt ) ¡ rt ) = ° t ¯ ¡t VkD (kt . lt ) ¡ G ¡ kt+1 ) ¡ rt (kt + bt ) ¡ wt lt )] + ·0 (r0 ¡ rmin ) Ã 1 ! 1 X X (13) + °t ¯ s¡t (u (cs ) ¡ v (ls )) ¡ V D (kt . » t ¯rt = » t¡1 ¡ ¯° t ¯ ¡t VbD (kt . respectively. ° t .optimization problem is L´ 1 X ¯ t [u (ct ) ¡ v (lt ) + ¸t (¯rt+1 u0 (ct+1 ) ¡ u0 (ct )) t=0 + ¹t (wt u0 (ct ) ¡ v0 (lt )) + ´ t (f (kt . (5). The …rst-order conditions for consumption. (15) (16) (17) (18) (19) constraint (9). at the initial date 0. and wage rate are respectively as follows:6 £ ¤ 1 + ¯ ¡t (° ¤ ¯)t u0 (ct ) + u00 (ct ) (¸t¡1 rt ¡ ¸t + ¹t wt ) = ´ t . 9 . lt ) ¡ kt+1 ¡ ct ¡ G) + » t (kt+1 + bt+1 + (f (kt .

3 The Steady State From now on. Moreover. taking into account this distinction. since capital and debt are non-negative. the second-order conditions are not clearly satis…ed because they involve second and third derivatives of the utility function. kt . (3). (11) and (12) can be written respectively as follows: wu0 (c) = v 0 (l) . rt ) to the government optimization problem for some initial conditions k0 and b0 such that ct . Taking into account conditions (7) and (8). First.Notice that. a steady state is de…ned. (20) ¯r = 1. This means a very high capital tax at the initial date 0. at a steady state. the individual’s total savings satisfy at+1 ¸ 0 in equilibrium. bt . (5) . we assume that an optimal interior solution exists. and rt are constant. the …rst-order condition for the after-tax return implies r0 = rmin . As usual in this literature. we will focus on the equilibrium at the steady state. wt . we provide three lemmas that will help us to characterize the optimal taxes at the steady state. Next. lt . the following transversality condition holds: lim ¯ t u0 (ct ) rt (kt + bt ) = 0: t!1 For an initial condition a0 > 0. taking into account the speci…c utility functions (2). the capital and labor tax rates ¿ k and t ¿ l are also constant at a steady state. constraints t (4). The remaining tax rates are obtained from the competitive equilibrium conditions (7) and (8). lt . wt . Therefore. we distinguish between steady states where the incentive compatibility constraint (12) binds and those where this constraint does not bind. Finally. Consequently. We de…ne a steady state as follows: De…nition 2 A steady state equilibrium is an optimal solution x = (ct . 10 (21) . kt . bt .

l) ¡ = ° t ¯ ¡t VkD (k. rt is de…ned analogously and is determined by ~ ~ l ~ ~ ~ equations (20) ¡ (27) for b = 0. 1¡¾ 1+" (24) Moreover. wt . Otherwise. This de…nition for incentive-constrained and unconstrained steady states and the analysis of the system (20) ¡ (28) will allow us to present di¤erent lemmas. we give the following de…nition: De…nition 3 An incentive-constrained steady state is a steady state such that the incentive compatibility constraint (24) is binding. The incentive compatibility constraint (24) could be binding or not at the steady state. ³ ´ a steady state x = ct . c £ ¤ 1 + ¯ ¡t (° ¤ ¯)t v0 (l) ¡ » t (fl (k. l) ´ t . Consequently. b) : (27) (28) Constraints (20) ¡ (24) and the …rst-order conditions (25) ¡ (28) yield a steady state for an economy with debt. ¯ k + c + G = f (k. ~t . Those 11 . l) ¡ w) + » t "w = fl (k. l) . b) . (25) (26) µ ¶ ¢ 1 1¡ (´t + » t ) fk (k. kt . plugging equations (18) and (19) into (14)¡(17) . b) + ¡´t + ´ t¡1 . For an economy without a market for public debt. the following optimality conditions must be satis…ed at a steady state: µ ¶ £ ¤ 0 £ ¤ k+b ¡t 1 + ¯ (° ¤ ¯)t u (c) ¡ ¾» t + ¾ » t+1 ¡ » t = ´t.µ ¶ 1 ¡ 1 (k + b) + wl = c. ¯ ¯ » t = » t¡1 ¡ ¯° t ¯ ¡t VbD (k. the steady state is incentive-unconstrained. The properties of the optimal taxes without commitment will depend crucially on this distinction. (22) (23) µ 1 1¡¯ ¶· ¸ c1¡¾ l1+" ¡ ¸ V D (k.

and converging to zero. As Lemma 1 states. we analyze the system of constraints and …rst-order conditions (20) ¡ (28). the problem (13) and the Ramsey problem yield di¤erent transitions and di¤erent steady states. the optimal capital tax rate is zero at incentive-unconstrained steady states. which are necessary conditions for optimality. the Lagrange multipliers f°g1 associated with i the incentive compatibility constraints (12) are positive. If the steady state is incentive-constrained. problem (13) de…nes implicitly a reputation mechanism through equation (24). we need to de…ne 1 + g and 1 + j as follows: 2 3 1 VkD (k. if the incentive constraints (12) are binding during the transition. we must clarify that a solution to problem (13) that exhibits an incentive-unconstrained steady state is not equivalent to the solution of the Ramsey problem. l) ¡ ¯ 7 (29) Chamley (1986) discussed the dependence of steady state values on initial conditions when government bonds are present. However. Therefore. However. 12 . with a di¤erent history. In order to determine whether the steady state is incentive-constrained. A Lagrangian for the Ramsey problem would be that of (13) without the incentive compatibility constraints (12). b) ¡ ¯ ¡ (1 + D) fk (k. The …rst result is the following: Lemma 1 If x is incentive-unconstrained. then the solutions of the Ramsey problem and the problem (13) would coincide.lemmas will become an important device in order to characterize the optimal timeconsistent taxes for economies with public debt. whereas the Ramsey solution is time-inconsistent. See Appendix 8. b) ¡ ¯ h ³ ´i 5 . we obtain the same long-run result as Chamley (1986) for his Ramsey problem. We …nd an optimality condition for being at incentive-constrained steady states. Before stating that condition. if the incentive compatibility constraints (12) are not binding during the transition. Under this condition. 1+g ´4 1 1 D Vk (k. then ¿ k = 0: Proof. then. summable. the properties of the steady state taxes could be quite di¤erent. which makes the solution of problem (13) time-consistent. On the one hand.7 On the other hand.

l) 5 c D D ¡¾Vb (k. (32) Lemma 2 shows some optimality conditions for incentive-constrained steady states. then 1 + j 2 0. 7 k+b fl (k. some incentive-constrained steady states will satisfy the Chamley-Judd result. l) ¡ w) u l 3 (30) where We can then present the next lemma: # u0 (c) (fl (k. From Lemma 3. See Appendix 8. whether the steady state is incentive-constrained or unconstrained. then 1 + g 2 0. l) c (31) Lemma 2 An incentive-constrained steady state x satis…es the following properties: ´ ³ 1 (i) If ¾ = 0. b) 0 (c) (f (k. l) ¡ w) ¡ ¢ D´ : 1 (1 + ") w ¡ (1 ¡ ¾) fl (k. 1). ¯ : Proof. some incentive-constrained steady states will have the property of an optimal zero capital tax: Lemma 3 For ¾ > 0. if x is incentive-constrained and 1 + j 2 (0. then ¿ k = 0: Proof. These conditions help us to identify an important feature of the steady state. See Appendix 8. we know that capital taxes are optimally zero at incentive-unconstrained steady states. From Lemma 1. 1 VbD (k. Are capital taxes di¤erent from zero at incentive-constrained steady states? As the next lemma shows. b) ¡ D à ¡ ¢ ! 7. 13 . l) + ¯ ¾ k+b fl (k. namely. ¯ : Moreover. we can conclude that being at an incentive-constrained steady state is not a su¢cient condition for optimal capital taxes di¤erent from zero. b) = ¡ D: ¯ ´ ³ 1 (ii) If ¾ > 0.and 6 6 1+j ´6 6 4 " 2 7 7 ¡¯VbD (k. In fact.

those lemmas only hold for economies with public debt. the possibility of issuing debt is also a major determinant of the taxation scheme. is important for the properties of the optimal taxes both during the transition and at the steady state. 8 We cannot apply Lemmas 1. and 3 on the basis of the …rst-order condition for debt (28) : Thus. the existence of a market for public debt and the speci…c form of the utility do matter for the properties of optimal taxation. 2. statement (ii) and condition (32) of statement (i) hold. First. namely. through the issue of bonds governments seek to smooth consumption over time. This di¤erent behavior a¤ects the properties of the optimal time-consistent taxes. The proofs of the properties of the optimal taxes are contained in the Appendices. whether the utility is linear in consumption and whether governments can issue debt. From these lemmas. Appendix 7:1 provides a general characterization regardless of the presence of debt and Appendix 8 includes the proofs of the speci…c taxation properties for economies without debt and for economies with debt. which will be crucial to characterize the steady state optimal taxation. Regarding Lemma 2. 14 . For economies without debt. they care about the deviations from a consumption pattern. two features turn out to be important. individuals are indi¤erent about the timing of consumption. ´ h 1 For ¾ = 0. First. Lemma 3 cannot be established. 2 and 3 for economies without a market for public debt. Then. Lemma 1 holds for utilities that are linear in consumption. when the utility is strictly concave.Therefore. We will next proceed to characterize the optimal time-consistent taxes. Moreover. an incentive-constrained steady state always satis…es that 1 + j 2 1. we consider an economy with public debt. Second. Hence. and 3. the current issues of debt will a¤ect not only the present but also the future incentives to deviate. However. the steady state values depend on the initial conditions when government bonds are present. the presence of debt a¤ects the intertemporal dynamics of the economy. the type of utility.8 In addition. but the rest of Lemma 2 does not apply. but it does not need to hold for concave utilities. we cannot apply Lemma 3 for utilities that are linear in consumption. When the utility is linear in consumption. 2. We have established Lemmas 1. First. individuals prefer to smooth consumption over time and. we examine an economy without debt. either linear or strictly concave. thus. Given these intertemporal linkages. ¯ : In this section we have developed a theoretical basis through Lemmas 1.

the optimal time-consistent taxes at a steady state x satisfy the following properties: ¡ " ¢ (i) ¿ l 2 0. The optimal allocation and policy are determined by equations (20)-(27) setting b = 0. b) > ¯ . governments run a balanced budget each date. we should reconsider the balanced budget constraint. b) < ¯ . Note that the bene…ts and costs of deviating at the steady state depend only on the steady level of capital. See Appendix 8. At incentive-constrained steady states. a capital tax di¤erent from zero provides the appropriate incentives to change the level of capital so as to satisfy the incentive compatibility constraint (24). ~ 1 (ii) If x is incentive-constrained and VkD (k. For strictly concave utilities. then the optimal capital tax rate is di¤erent from zero. min ¾+" . For an economy without bonds. 1+" : then ¿ k < 0: Proof. the optimal time-consistent taxes at a steady state x satisfy the following properties: ¡ £ ¤¢ (i) ¿ l 2 0. For utilities that are linear in consumption. Propositions 1 and 2 capture the results on capital taxes of Benhabib and Rustichini (1997). the steady state is characterized by a positive and bounded tax rate on labor income. if the incentive compatibility constraint (24) is binding. See Appendix 8. 1 : 1+" Proof. Notice 15 1 1 (ii) If x is incentive-constrained and VkD (k. However. the optimal time-consistent capital tax is di¤erent from zero for economies without public debt. then ¿ k > 0: If VkD (k. Thus. b) > ¯ u0 (ct ).4 An Economy without Public Debt In an economy without bonds. we state the following: ~ Proposition 2 For ¾ > 0. then ¿ k > 0: ~ . Moreover. the steady state optimal taxation is characterized as follows: ~ Proposition 1 For ¾ = 0.

the government will choose the issues of debt taking into account how the bene…ts of deviating are a¤ected. 5 An Economy with Public Debt In this section we consider an economy with a market for government bonds. l) satis…es an elasticity of substitution between capital and labor greater than one. capital is not essential for positive production. This means that. 1) . l) fkl (k. Notice that. The optimal allocation and policy are now characterized by equations (20)-(28). B > 0. Thus. when governments can issue debt. Hence. in order to …nance the public spending. the speci…c form of the deviation plays now a role that was absent without a market for debt. e ¸ 0.9 9 (33) 1 ¯ ¡ ' with ' continuous strictly increasing function. As a result. 16 . and A (e) = Cobb-Douglas and a linear function.that public debt is an instrument that links past and future variables and makes the steady state dependent on the transition. under this assumption. which takes the following form: f (k. l))] > 1 since f (k. the …rst-order condition for debt (28) must hold. the government chooses tax rates on labor and capital income and the issue of debt. and ' (0) = 0: This production function is a weighted average of a The linear component of the speci…c production function (33) can be viewed as home produc- tion. This condition links past and future e¤orts to satisfy the budget constraint through the current incentives to deviate. where ® 2 (0. l) = A (e) k + Bl + ek ® l1¡® . that is. l) fk (k. To this end. A speci…c example for this type of production functions is that of Benhabib and Rustichini (1997). l) is homogenous of degree one. we assume that the production function f (k. l)) /(f (k. we then need to specify a particular deviation form. which can be written as [(fl (k. the presence of debt could seriously a¤ect the previous results. Before characterizing the optimal time-consistent taxes.

This deviation will be speci…ed as the reversion to a Markov equilibrium. the policy plan ¦ induces future histories by ht = (ht¡1 . bs+1 ). For an economy without commitment. we de…ne both sustainable and Markov equilibria. denoted ¦t (ht¡1 ). government and individuals take their corresponding decisions sequentially. ws . Then. In this section we will develop a speci…c deviation form. A continuation policy of ¦ is (¦t (ht¡1 ) . t ¡ 1). we …rst describe the economy without commitment. ls (hs )) ¡ rs (hs¡1 )) as (hs¡1 ) + (fl (ks (hs¡1 ) . Once the government decides the current policy. Consider …rst the government at date t. (34) subject to government budget constraint bs+1 (hs¡1 ) + (fk (ks (hs¡1 ) . Given some history ht¡1 and given that future allocations evolve according to F . where ¼ s = (rs . as+1 ). At the beginning of date t. In order to de…ne a sustainable equilibrium. Finally. the government chooses a current policy as a function of history ht¡1 = ( ¼ s j s = 0. we will next frame both a continuation policy and a continuation allocation into an optimization problem. An allocation at date s is denoted ds = (cs . Given a history ht and a policy plan ¦t . and a plan for future policies under all possible future histories. ¦t (ht¡1 )) and so on. Given a history ht¡1 . the representative individual chooses consumption. the government chooses a continuation policy that maximizes the welfare of the representative individual 1 X s=t ¯ s¡t [u (cs (hs )) ¡ v (ls (hs ))] . for which we have only assumed dependence and di¤erentiability with respect to the current levels of capital and public debt.5. ¦t+1 (ht¡1 . Ft+1 (ht . ls . :::. ¦t (ht¡1 )) . and a plan for future allocations. ¦t+1 (ht )) . a continuation allocation of F is (Ft (ht ) . In order to do so. ::::).1 A Deviation So far we have speci…ed a general functional form for the deviation value. :::). ls (hs )) ¡ ws (hs¡1 )) ls (hs ) ¸ G + rs (hs¡1 ) bs (hs¡1 ) . we propose a Markov equilibrium. denoted Ft (ht ). labor and assets holdings at date t as a function of history ht . 17 (35) . following Chari and Kehoe (1993).

the lower bound on the initial after-tax return on assets rt ¸ rmin . F ) : for all dates s ¸ t. ¦. and histories are induced by ¦ from ht¡1 . ¦. bt h0t¡1 . A sustainable equilibrium is utility-Markov if the past history in‡uences payo¤s only to the extent that it changes the current state variables. ¼ t . where V and W are de…ned in equations (34) and (36) . This de…nition can be formalized as De…nition 5 A sustainable equilibrium is said to be utility-Markov if for any pair of ¡ ¡ ¢ ¡ ¢¢ histories ht¡1 and h0t¡1 such that (kt (ht¡1 ) . F ) = V h0t¡1 . where future of welfare that is denoted V (ht¡1 . The solution of this problem provides a value Consider the representative individual at date t. bt (ht¡1 )) = kt h0t¡1 . ( ii ) given a policy plan ¦. for date t. the continuation allocation of F solves the consumer’s problem for every history ht . ¦. F ) = W h0t¡1 . Given some history ht and given that future policies evolve according to ¦. (36) subject to the individual budget constraint rt at (ht¡1 ) + wt lt (ht ) ¸ at+1 (ht ) + ct (ht ) . respectively. the non-negativity constraint bs+1 ¸ 0 for all dates s ¸ t. 18 . and rs (hs¡1 ) as (hs¡1 ) + ws (hs¡1 ) ls (hs ) ¸ as+1 (hs ) + cs (hs ) . F ) : Using these programs. ¦. ¦. ¼ t . (38) (37) for all dates s > t. we de…ne a sustainable equilibrium as follows: De…nition 4 A sustainable equilibrium is a pair (¦. ¦. F ) that satis…es the following conditions: ( i ) Given the allocation rule F . capital and public debt at date t. F . then ¡ ¢ ¡ ¢ (i) V (ht¡1 . the representative individual chooses a continuation allocation so as to maximize 1 X s=t ¯ s¡t [u (cs (hs )) ¡ v (ls (hs ))] . F . and (ii) W (ht¡1 . the continuation policy of ¦ solves the government’s problem for every history ht¡1 . where future histories are induced by ¦ from ht : We denote the welfare resulting from this optimization problem by W (ht . namely.

s=t which in turn yields the consumer allocation rule F m . the market the non-negativity constraint bs+2 ¸ 0 at date s ¸ t. (41) subject to the budget constraint (3) at date s ¸ t. (40) at date s ¸ t. br ) . the …rst-order conditions (4) and (5) at date s ¸ t. The second program de…nes the value function W (kt . a solution for them will be sustainable. Given the way these two optimization problems (39) and (41) have been constructed. the government budget constraint (9) is not required to hold at date t. and the lower bound on the initial after-tax return on assets rt ¸ rmin . The sequence fds . the …rst-order conditions (4) and (5) at date s ¸ t. This allocation rule is de…ned for all possible histories. including those in which the government deviates. as the next lemma shows: Lemma 4 The pair (¦m . ¼ s g1 that solves problem s=t ¯ s¡t [u (cs ) ¡ v (ls )] . ¼ t ) ´ max 1 X s=t to deviate from the announced policy. and kt and debt bt . the incentive constraint (40) at date s ¸ t. the market ¯ s¡r [u (cs ) ¡ v (ls )] ¸ V (kr . Here equation (40) ensures that the government will have no incentives (39) gives us a policy plan ¦m for the government.We will next follow Chari and Kehoe (1993) to propose a Markov equilibrium by means of solving two di¤erent programs. the incentive constraint 1 X s=r (9) at date s ¸ t. The …rst program de…nes the value function V (kt . To this end. See Appendix 8. the non-negativity constraint bs+1 ¸ 0 at date s ¸ t. (39) subject to the budget constraint (3) at date s ¸ t. bt ) = max 1 X s=t ¯ s¡t [u (cs ) ¡ v (ls )] . ¼s+1 g1 that solves the optimization problem (41). F m ) is a sustainable equilibrium. We choose fds . given the initial capital kt and debt bt . 19 . the government budget constraint clearing condition (6) at date s ¸ t. given ¼ t and the initial capital clearing condition (6) at date s ¸ t. Proof. the government budget constraint (9) at date s > t. bt .

By analyzing problem (39). both capital and debt become zero. namely. that is. thus. Moreover. F m ). ks = bs = 0 for all dates s ¸ t + 1. governments have incentives to tax them as high as possible.. this sustainable equilibrium is utility-Markov by construction. . our Markov equilibrium is such that defection satis…es ks = bs = 0 for all dates s ¸ t + 1: Appendices 7:2 and 8 contain a general characterization of the policy after a deviation and the proofs of the lemmas. Summing up. if a unique solution exists. as = 0 for all dates s ¸ t + 1: In particular. individuals prefer not to save. a government will choose a lower than the announced initial after-tax return on assets. In particular. if the value of the total assets at is strictly positive. 20 savings are never positive. we state the following lemma: Lemma 5 The pair (¦m . we can derive some properties of the Markov equilibrium (¦m . Thus. when total assets are strictly positive at date t. This minimum after-tax return can be viewed as a very high capital tax. t+2. F m ) satis…es as = 0 and. Proof. they expect very high capital taxes for all future dates. when savings are strictly positive. See Appendix 8. they should provide the same welfare. Consequently. if there is more than one solution. given the non-negativity constraints. choosing the policy plan at that current date. This result holds for all dates t+1. A deviation occurs at date t when a government chooses a policy plan from date t on di¤erent from the announced plan. the government plans for any future date must provide a welfare from that date on that is at least as large as the welfare value of re-optimizing. that is. rmin . if the after-tax return on savings is su¢ciently low. On the one hand. On the other hand. Under this incentive constraint. It is obvious that. Therefore. Since individuals are aware of those incentives to tax heavily the assets income. the …rst-order condition for assets holdings (5) dictates individuals not to save and. this is uniquely determined by the current capital and debt.. if rmin is su¢ciently close to zero. then it is optimal to set the initial after-tax return to its minimum value rmin . our . Given the speci…c production function.The programs (39) and (41) de…ne a sustainable equilibrium thanks to equation (40).

using Lemma 1. thus. the wider the 21 apply Lemmas 1.2 Optimal Time-Consistent Taxes Once we have found a Markov equilibrium to revert to after a deviation. thus. 1 ¯ 1 the announced after-tax return is such that r = ¯ . Hence. 2. For utilities that are linear in consumption.respectively. In the presence of public debt. if rmin · 1 ¯ ¡ 1. our deviation clearly extends the environment of Benhabib and Rustichini (1997) to economies with public debt. and the > rmin ¸ 0 is satis…ed because minimum after-tax return was assumed to satisfy rmin ¸ 0. 1+" : (ii) If rmin · 1 ¯ ¡ 1. We have found in Proposition 1 that the long-run capital tax could be di¤erent from zero for an economy without bonds. the optimal capital tax rate is zero. the optimal time-consistent taxes at a steady state x satisfy the following properties: ¡ " ¢ (i) ¿ l 2 0. individuals decide not to save. we turn to characterize the optimal taxes without commitment for an economy with government bonds. First. Second. we can incentive-constrained steady states of Lemma 2 does not hold and. Two remarks are in order here. 5. the existence of a market for debt and its relation with the time-inconsistency problem a¤ect clearly the properties of the optimal taxes. Moreover. we …nd that the smaller the steady state debt is. See Appendix 8. and 3. which yields the value of welfare after deviating. thanks to equation (21). In this section we have identi…ed a Markov equilibrium to revert to after a deviation. we obtain the following result: Proposition 3 For ¾ = 0. then x is incentive-unconstrained and. This means rmin = 0 at date t and ks = 0 for all dates s ¸ t + 1: Therefore. the numerical results of Benhabib and Rustichini (1997) are found for a deviation that speci…es complete default and. Thus. the steady state must be incentive-unconstrained. therefore. ¿ k = 0: Proof. This deviation value depends on the current stock of capital and debt and on the minimum after-tax return on assets rmin . then condition (32) for . We show that.

for zero steady state debt. the steady state incentives to deviate depend on the steady state level of capital. ¿ k = 0: Proof. which makes the steady state level of total assets higher. then x is incentive-unconstrained and. thus. This result can help us to understand that of Benhabib and Rustichini (1997). 1 : 1+" (ii) If rmin = 0. thus. min ¾+" . then the incentive compatibility constraint (12) is not binding at the steady state. See Appendix 8. then the steady state debt must be strictly positive. Let us now consider Corollary 1. Corollary 1 states that the capital tax rate is zero. In turn. if the optimal steady state debt is zero. Without a market for debt. Therefore. we show the following: Proposition 4 For ¾ > 0. ¿ k = 0: ³ ´³ ´ 1 1 (iii) If x is incentive-constrained and 0 < rmin · 2¡¯ ¡ 1 . When the incentive constraint (12) binds at the steady state. Suppose that the optimal steady state debt that solves the system (20) ¡ (28) is zero. we can say that. we get the following result: Corollary 1 For ¾ = 0. then Benhabib and Rustichini (1997) obtained that a subsidy to capital would promote a higher level of steady state capital in order to satisfy the incentive compatibility constraint (12). then we have a strictly positive level of debt. See Appendix 8. the optimal time-consistent taxes at a steady state x satisfy the following properties: ¡ £ ¤¢ (i) ¿ l 2 0. then x is incentive-unconstrained and. There is a steady state level of capital consistent with an optimal capital tax rate equal to zero. if b = 0. 22 . For utilities that are strictly concave in consumption. More precisely.range of rmin that su¢ces to have an optimal zero capital tax rate. then ¿ k = 0: ¯ Proof. if the incentive constraint (12) binds at the steady state. In that case. When the incentive constraint (12) is not satis…ed for that steady state capital. the optimal management of debt makes public debt become an endogenous commitment device. In particular.

the steady state values depend on the initial conditions and. if rmin = 0. the steady state is that incentive-constrained steady states satisfy 1+j 2 (0. hence. in the presence of debt. the government can choose a debt sequence so that the transition leads to a steady state that is incentive-unconstrained or where the incentive compatibility constraints are alleviated. the labor taxes will take a di¤erent value in these economies. For economies with public ³ ´ 1 debt. the optimal capital tax rate is zero. Regarding labor taxation. Thus. ¯ for incentiveconstrained steady states of Lemma 2 does not hold. for a range of rmin . All in all. Therefore. thus. we …rst show that. those conditions include rmin = 0: Therefore. the incentives and costs of deviating. Moreover. depend on the steady state capital and debt. Moreover. the optimal time-consistent policy for economies with public debt is very di¤erent from that for economies without debt. but it is zero for economies with debt. Note …rst that these conditions are su¢cient but not necessary. then condition 1 + j 2 0. However. we have found that the optimal long-run labor tax rate is positive and bounded both in economies without and with public debt. we obtain 23 . the optimal time-consistent capital tax rate is zero for an economy with debt. by Lemma 3. As a result. to minimize the costs of meeting the incentive constraints (12). First. Now. we obtain that the optimal long-run capital tax rate is di¤erent from zero for economies without debt. The intuition for these results is as follows. Second. the steady states that exhibit a capital tax rate di¤erent from zero in Benhabib and Rustichini (1997) and in Propositions 1 and 2 have an optimal zero capital tax once a market for debt is introduced. 1) and. note that. incentive-unconstrained and. on the transition towards the steady state. which build the incentive compatibility constraint (12).Proposition 2 obtained that the optimal time-consistent capital tax at the steady state is di¤erent from zero for economies without debt. which in turn depend on the past decisions. then the introduction of a market for debt and the optimal management of that debt will be welfare improving. That optimal zero capital tax rate is obtained under some conditions on the minimum after-tax return on assets rmin . since no issuing debt is still a sustainable outcome. thus. ¿ k = 0: Moreover.

the main commitment device against deviation is public debt. we have proved that the presence of a market for public debt is crucial for the properties of the optimal tax policy without commitment. For instance. In this paper we have identi…ed a particular Markov equilibrium. These results highlight the importance of debt in relation to time-inconsistency and optimal taxation. In this paper we have analyzed the best sustainable equilibrium. while the previous literature on optimal taxes without commitment had focused on economies without debt. In this framework the capital tax is the instrument to make the policy credible. In the present paper we have shown that the mere presence of a market for debt alleviates the long-run e¤ects of the time-inconsistency problem. for economies with public debt.6 Conclusions This paper has studied the optimal steady state taxes for an economy without commitment. we would obtain a di¤erent equilibria to revert to after a deviation. We have found that the properties of the optimal time-consistent capital taxes hinge on the assumption of a balanced budget. This equilibrium is non-Markov because governments take decisions based not only on the current state variables but also on the future reactions to its actions. Nevertheless. if a government deviates. However. Therefore. However. the optimal time-consistent capital tax rate could be di¤erent from zero at the steady state. the economy reverts to some speci…c bad equilibrium. we believe that our main result of an optimal zero 24 . we extend the Chamley-Judd result to economies without commitment showing that the optimal time-consistent tax rate on capital income is zero at the steady state. Moreover. In a seminal paper Lucas and Stokey (1983) showed that a rich composition of debt can help overcome the time-inconsistency problem of optimal …scal policy. the Markov equilibrium would be di¤erent from the one in this paper. The optimal management of debt can make a long-run zero capital tax announcement credible. if we had assumed a production function and a minimum after-tax return rate that allowed for positive savings. Then. once a market for debt is introduced. For economies without government bonds. under di¤erent assumptions.

The intuition for this conjecture is that we have established a theoretical basis for the optimality of zero capital taxes based on a general functional form for the deviation. 25 .capital tax at the steady state could still hold.

forthcoming. [8] Erosa. [7] Chari. 26 . and Constant Discounting. and Patrick J. 23. “Optimal Taxation of Capital Income in General Equilibrium with In…nite Lives.” Journal of Political Economy. and Patrick J. [6] Chari. [10] Garriga. Borrowing Constraints. [9] Fischer. Stanley (1980).” Journal of Economic Theory. Christiano. 103. “Optimal Capital Income Taxation with Incomplete Markets. 54.. [3] Atkinson. “Pigou. 119-128.” Journal of Economic Dynamics and Control. and Public Goods. S. 2.References [1] Aiyagari.” Econometrica. Kehoe (1994). Jess and Aldo Rustichini (1997). [4] Benhabib. “Optimal Fiscal Policy in Overlapping Generations Economies. “Sustainable Plans and Debt. Anthony and Nicholas Stern (1974).V. Carlos (2000). 230-61. 77.V. 1158-1175.” Journal of Political Economy. “Optimal Taxation in Life-Cycle Economies. Kehoe (1993). 93-107. Andrew. 61. Andrés and Martin Gervais (2002).” Federal Reserve Bank of Minneapolis Quarterly Review. 617-652. Cooperation.V. V . V. Lawrence J. 102. Taxation. 41. [2] Atkenson. Christophe (1986). and the Benevolent Dissembling Government. Kehoe (1999). Rao (1995). and Patrick J. 607-622.” CREB Working Paper 0105. 231-259. “Optimal Fiscal Policy in a Business Cycle Model.” Journal of Economic Theory. “Optimal Taxes Without Commitment.” Journal of Economic Theory. Chari. [5] Chamley.” Review of Economic Studies. “Taxing Capital Income: A Bad Idea. “Dynamic Inconsistency. V.

73.“Supply-Side Economics: An Analytical Review.” mimeo. Rodolfo E. and Nancy L. “Time-Consistent Optimal Fiscal Policy. “Optimal Fiscal and Monetary Policy in an Economy without Capital. Robert E. [14] Lucas. 59-83. Larry E. “Optimal Capital . [15] Lucas. (1985). Kenneth L.” Oxford Economic Papers. Irina (2001). (1990). “On the Optimal Taxation of Capital Income.Labor Taxes under Uncertainty and Limits on Debt.” Journal of Monetary Economics. Stokey (1983). 27 . Paul and José-Víctor Ríos-Rull (2000).” Journal of Economic Theory. “Redistributive taxation in a simple perfect foresight model.” mimeo. 42.[11] Klein. [13] Judd. 12. and Peter E. 55-93. [16] Yakadina. Manuelli. 93117. 293-316. Robert E. 28. Rossi (1997).” Journal of Public Economics.. [12] Jones.

7 Analytical Characterization In this appendix we …rst characterize the optimal time-consistent taxes and. then. To study the transition. bonds. labor. r0 ¸ rmin . we suppose that the economy is at the steady state and we do the following backward-looking exercise. (3). Therefore.1 The Optimal Time-Consistent Taxes In order to characterize the optimal time-consistent taxes at a steady state. all the variables attain their respective steady state values from date 1 onwards. and wage rate. 28 (46) . return rate. Next. 1 X t=1 (42) (43) (44) (45) ¯ t¡1 [u (ct ) ¡ v (lt )] ¸ V D (k1 . (12) . which can be written. we solve for the steady state values in conjunction with the conditions for the initial date. l0 ) . we …rst study the transition of the economy towards the steady state so as to determine the sign of the Lagrange multipliers. respectively: u0 (c0 ) + u00 (c0 ) (¹0 w0 ¡ ¸0 ) = ´ 0 . respectively. (11). 7. as follows: w0 u0 (c0 ) = v0 (l0 ) . u0 (c0 ) = ¯r1 u0 (c1 ) . b1 ) . (5) . The government chooses the next date initial debt such that the transition back to the steady state takes place in one date. the properties of the economy after a deviation. r0 (k0 + b0 ) + w0 l0 = k1 + b1 + c0 . and (10) . k1 + c0 + G = f (k0 . capital. and by the following …rst-order conditions for consumption. The solution at date 0 of Lagrangian (13) is characterized by constraints (4).

We get expressions for ¿ k and ¿ l from the …rst-order conditions equations (25) ¡ (28) and (46) ¡ (51). ¤¢ ¡ £ We will next proceed to show that ¿ l 2 0. l0 ) ¡ w0 ) + ¹0 v (l0 ) = fl (k0 . r. combining conditions (25) and (26). (43) implies c0 = c1 = c. and (45) yield (7) and (8).v0 (l0 ) ¡ » 0 (fl (k0 . equations (20) ¡ (24) . any ¾ regardless of whether debt is present or not. (44) . the economy is already at the steady state at date 1. First. Atkinson and Stern (1974) showed that » t > 0 in a second-best framework. We solve for the initial and the steady state Lagrange multipliers using By the Kuhn-Tucker theorem. Using ¯r = 1. l. the Lagrange multipliers for constraints (12) and which does not converge to zero. conditions (20) and (21) (19) . ¯ 00 (47) (48) » 1 ¯r1 = » 0 ¡ ¯° 1 ¯ ¡1 VbD (k1 . (42). b. which does not converge to zero. that is. k. ·0 ¡ » 0 (k0 + b0 ) = 0. Note that ¡» t . c. therefore r1 = r. (10) satisfy respectively ° t ¸ 0 and ·0 > 0. condition fl0 . 1) for 1+" combining condition (28) and VbD (k. the Lagrange multiplier for the budget constraint (3). l0 ) ´ 0 . we obtain » t > 0. using equations (18) and £ (52) 10 excess burden of taxation. c 00 imply the unity upper bound on both tax rates. b1 ) + 1 (¡¯r1 ´ 1 + ´ 0 ) . l1 ) ¡ r1 ) = ° 1 ¯ ¡1 VkD (k1 . (´1 + » 1 ) (fk (k1 . w0 . it results that (´t + » t ) > 0. min ¾+" . b) · 0 (as proved later). we get µ ¶ ¤ k+b u (c) [¸t ¡ ¸t¡1 r + ¹t w] = ¾ » t+1 ¡ » t ¡ ¾» t · 0. r0 = rmin : Moreover. the initial after-tax return on assets is set optimally at its minimum value. Thus. Next. 1 and ¿ k 2 (¡1. wg. ¹0 u0 (c0 ) ¡ » 0 l0 = 0: (49) (50) (51) Given an initial condition for total assets a0 > 0. represents the marginal 29 . b1 ) .10 Conditions (18) and (19) imply ¸t > 0 and ¹t > 0: Moreover. Equation (50) implies » 0 > 0: Next.

Combining equations (52) and (27). if ¾ » t+1 ¡ » t ¸ 0 holds. that is. l) ¡ = °t¯ Vk (k. l) ¡ w) . l)] ¡ ¾ » t+1 ¡ » t c £ ¤ 1 + ¯ ¡t (° ¤ ¯)t u0 (c) (fl (k. 1+" . b) ¸ ¯ . if VkD ¡ ¯ u0 (c) + ° ¯ ¡t Z · 0. we get µ · ¸ ¶ 1 1 0 1 ¡t D (´ t + » t ) fk (k. we can write equation (53) as £ ¶ µ ¤ k+b fl (k. and ¸t < ¯ ¸t¡1 must hold in order to have a well-de…ned (53) obtain (1 + ") w ¡ (1 ¡ ¾) fl (k. The consequences of deviating have been speci…ed in 30 . it results £ ¤ that ¾ » t+1 ¡ » t ¸ 0 and Z · 0 always hold.2 The Economy After Deviation We …rst compute VkD (kt . ¯ ¯ ¯ (54) with £ ¤ Z = ¾ » t+1 ¡ » t µ k+b c ¶ ¡ ¾ » t ¡ » t¡1 £ ¤ ¶¶ µ µ k+b : 1+ c 1 From equation (54) . l) = » t [(1 + ") w ¡ (1 ¡ ¾) fl (k. b) ¡ ¯ u0 (c) ¸ 0 implies ´ ³ 1 1 ¿ k ¸ 0: Otherwise. Therefore. ¾ = 0. Therefore. £ ¤ whose RHS is positive. fl (k. then we > ¿ l: We will next derive conditions to determine the sign of the optimal capital taxes. if VkD (k. ¾+" 1+" Lagrangian. then. l) ¡ w) . if VkD (k. equation (25) becomes µ ¶ µ ¶ £ ¤ k+b » t "w ¡ ¾ » t+1 ¡ » t ¡ ¾» t fl (k.whose LHS is positive thanks to inequality (52). Second. l) > 0 and. in turn. Hence. it follows that. bt ) and VbD (kt . First. 7. b) · ¯ . then VkD (k. l) = c ££ ¤ ¤ 1 + ¯ ¡t (° ¤ ¯)t u0 (c) + » t (fl (k. l) ¡ w > 0 and. ¿ l 2 0. Using conditions (20) and (26). it results that. if Z · 0. b) ¡ u (c) ¡ Z. then ¿ k · 0: Notice that. for utilities that are linear in consumption. ¿ l > 0: Moreover. we study the optimality conditions after a deviation. bt ) and. then ¿ k · 0: t 1 then ¿ k ¸ 0: Otherwise. 1 since u00 (c) · 0. the optimal steady state ¡ " ¢ 1 taxes satisfy the following results. ¹t wt > 0. thus.

Hence. t d d wt+s lt+s = cd .Lemma 5 as follows: once a government deviates. bt ) = (59) To …nd these derivatives. bt ) ¡ d ¢¡¾ @cd ¡ d ¢" @lt t = ct . lt+s . t+s ¡ ¢ ¡ d¢ d wt u0 cd = v 0 lt . individuals expect a minimum return on assets after taxes at all future dates so that agents do not save. bt ) = ¡ ¡ 1¡¾ 1+" 1¡¯ 1¡¾ 1+" (58) which allows us to write VkD (kt . ¡ lt @kt @kt @kt d @V D (kt . once the government deviates at date t. the economy is characterized by equations (3) . t t 31 (60) 11 The variables with a superscript d are those after a deviation. we use the speci…c utility functions (2) and equations (55) ¡ (57) to form the following set of equations: ¡ ¢¾ ¡ d ¢1+" rmin (kt + bt ) + cd lt ¡ cd = 0. t+s at date t + s for all s > 0. bt ) = d @V D (kt . (4) and (11) . V D (kt . and ¡ ¢ d cd + G = f kt . t (55) (56) (57) at date t. the welfare after a deviation at date t is à ¡ ¢1¡¾ ! à ¡ ¢1+" ! · ¸ Ãà ¡ d ¢1¡¾ ! à ¡ d ¢1+" !! d lt ct+s lt+s cd ¯ t + . ¡ d ¢ cd + G = f 0. . bt ) ¡ d ¢¡¾ @cd ¡ d ¢" @lt t = ct : ¡ lt @bt @bt @bt VbD (kt . lt .11 Thus. t ¡ ¢ ¡ ¢ 0 d d wt+s u0 cd t+s = v lt+s . which take the following form: d d rmin (kt + bt ) + wt lt = cd .

t t 32 . labor. bt ) = ¡rmin 4 ¡ ¢ ¢5 : kt +bt d d d (1 + ") wt ¡ (1 ¡ ¾) fld kt . we obtain ·· µ ¶¸ ¸ ¡ d¢ ¡ ¢ ¡ ¢ kt + bt 0 D d d Vk (kt . lt ¡ wt + ¹d v lt = fld kt . return rate. t t t ·d ¡ » d (kt + bt ) = 0. lt + (1 ¡ ¾) + ¾rmin fk kt . b) = 0: For some rmin > 0. Thus. bt ). bt ) : If rmin = 0. From this problem. lt ¡ ¾rmin ktcd t fld kt . lt ´d . Notice that Lemma 5 provides a necessary condition for the continuation allocations from any history of the Markov equilibrium (¦m . lt t We will next study the optimality conditions after a deviation at date t. lt ¡ rmin cd t 2 3 ¡ d¢ ¡ ¡ ¢ ¢ 0 d d u ct fld kt . it turns out that the allocations that solve problem (39) solve the same problem but replacing constraint (40) by the conditions provided in Lemma 5. lt t (62) Proceeding in the same way for VkD (kt . t t t t (63) (64) (65) ¡ d¢ ¡ ¡ ¢ ¢ ¡ ¢ 00 ¡ d ¢ d d d v0 lt ¡ » d fld kt . ¡ 1 ¡ ¾ ct @bt @bt µ @cd t @bt ¶ ¡ ¢ d ¡ fld kt . and wage rate at date t: ¡ ¢ ¡ ¢ d u0 cd + u00 cd ¹d wt = ´ d . lt ¡ ¾rmin cd fld kt .Let us …rst compute VbD (kt . lt = 0: t (61) ¶ = 0. by implicit di¤erentiation. using equations (60) ¡ (61). we obtain the following system: µ ¶ µ d¶ h ¡ d ¢¾¡1 ¡ d ¢1+" i @cd ¡ d ¢¾ ¡ d ¢" @lt t lt + (1 + ") ct lt = ¡rmin . lt µ d @lt @bt ¡ ¢ d cd + G ¡ f kt . we obtain the following …rst-order conditions for consumption. F m ). bt ) = u ct fk kt . lt ¡ wt ³ ´ ¡ VbD (kt . the derivative (59) becomes @bt @bt 2 3 ¡ d¢ ¡ ¡ ¢ ¢ 0 d d u ct fld kt . then VbD (k. lt ¡ wt 4 ³ ´ ¡ ¡ ¢ ¢5 : +b d d d (1 + ") wt ¡ (1 ¡ ¾) fld kt . Solving for d @cd @lt and t .

t 1+" ¡ d kt . lt ¡ wt . (57) . t which implies (1 + d ") wt ¡ (1 ¡ ¾) fld Using the previous inequality. fld kt . lt ¡ wt > 0 and. condition (64) implies ´ d + » d > 0: Next. min ¾+" . and the lower bound on the initial after-tax return on assets. respectively. 1 : From equation (62).Here ¹d . ¿ l > 0: Next. the optimal labor tax rate after a deviation satis…es ¡ £ ¤¢ d that ¿ l 2 0. lt ¡ wt . t d ct ¡ ¢ d d d whose LHS is positive. which implies ·d > 0. we get µ µ ¶ ¶ ¡ ¢ ¡ ¢ kt + bt d d d d » t (1 + ") wt ¡ (1 ¡ ¾) fld kt . t t t t (56). The government sets rt = rmin . lt ¡ ¾rmin fld kt . » d . t ¡ ¢ d we get » t > 0: Using equation (66) . we obtain d » d "wt + » d ¾ t t d d ¢¡ ¡ ¢ ¢ wt lt d ¡ d d d (67) wt = ´ t + » d fld kt . ¡ ¢ d ¹d u0 cd ¡ » d lt = 0: t t t (66) plugging equation (63) into (67). ´d . (63) and (64). it results that VbD (kt . Therefore. t t combining equations (55) . and ·d are the Lagrange multipliers associated with constraints (4). lt = cd t ¡ ¢¡ ¡ ¢ ¢ d d u0 cd fld kt . thus. lt > 0: 33 . From condition (65). lt ¢ ¡ ¾rmin µ kt + bt cd t ¶ ¡ ¢ d fld kt . bt ) · 0.

Given equation (25) . Since we try to identify incentive-constrained steady states. l) ¡ w ¡t » t = 1 + ¯ (° ¤ ¯)t . where 1 + g takes the value de…ned in equation (29). ° t ¯ ¡t = (1 + g)t ° 0 . (1 + ") w ¡ fl (k. that is. we get ° t ¯ ¡t = (1 + g) ° t¡1 ¯ ¡(t¡1) . l) (69) 1 1 If VkD = ¯ . thus. (31) and (68) into (27). 34 . In addition. equations (20)¡(24) yield an incentive-unconstrained Plugging equations (25). ´ t = ´t¡1 . or equivalently. ¿ k = 0: ¥ (68) which can be written as · ¸ fl (k. if 1 + g were greater or equal to ¯ . Solving for » t in equation (53). when public debt is present. Second. Thus. Moreover. 1+g must be greater than 1 0 so that ° t is positive.8 Proofs of the Propositions and Lemmas Proof of Lemma 1 If the steady state is incentive-unconstrained. there exists a real number g such that ° t ¯ ¡t = (1 + g) ° t¡1 ¯ ¡(t¡1) . then the …rst-order condition for debt (28) implies » t = » t¡1 . ¾ = 0. condition (27) implies fk (k. ¯ : First. equations (28) and (69) imply that condition (32) must hold at incentive-constrained steady states. an incentive-constrained steady state of an economy with public debt must satisfy conditions (29) and (32). l) = Proof of Lemma 2 We will next show that. if the steady state is ´ ³ 1 incentive-constrained. Therefore. l) » t = » t¡1 + ° t ¯ ¡t 1 ¯ and. then fk = ¯ . we get µ µ ¶ ¶ £ ¤ 1 1 ¡t ¡t D = °t¯ Vk (k. b) ¡ : 1 + ¯ (° ¤ ¯)t [1 + D] fk (k. ´ t is also constant. l) ¡ w : (1 + ") w ¡ fl (k. then Lagrangian (13) would not be well-de…ned. then 1+g belongs to 0. l) ¡ ¯ ¯ (70) steady state. we obtain ¸ · £ ¤ fl (k. for utilities that are linear in consumption. Solving for ° t ¯ ¡t in equation (70) and rearranging terms. that is. we consider 1 VkD 6= ¯ . Thus. for ¾ = 0.

l) c 35 (73) . we require that · (72) " # u0 (c) (fl (k. (1 + ") w ¡ (1 ¡ ¾) fl (k. l) + ¯ ¾ k+b fl (k. ¯ : Notice next that. ¾ > 0. l) ¡ w) . F which becomes equation (30) : By arguments analogous to those for ¾ = 0. VbD (k. b) must be strictly ´ ³ 1 negative at incentive-constrained steady states: In particular. b) = 0. b) > ¡ E + ¡1 F : ¯ ¯ ¯ 1 More precisely. b) k+b fl (k. b) 1+j ´ . l) ¡ w) 1 ¡ ¢ > VbD (k. if the ³ ´ 1 steady state is incentive-constrained.We will next show that. l) ¡ ¢ ¡¾VbD (k. b) . ¡ 1 ¯ (1 + ") w ¡ (1 ¡ ¾) fl (k. where ¸ F ¡ E ¡ ¯VbD (k. » t can be written as follows: » t = (1 + ¯ ¡t (° ¤ ¯)t )E + ° t+1 ¯ ¡(t+1) F: Here E > 0 and F ¸ 0 take the following forms: E = u0 (c) (fl (k. if VbD (k. then condition (71) can be written as » t = » t¡1 + ° t ¯ ¡t E. for 1 + j 2 0. b) must satisfy · µ ¶ ¸ 1 1 1 D ¡ (E ¡ F ) > Vb (k. l) c F = : (1 + ") w ¡ (1 ¡ ¾) fl (71) Considering equation (71) at date t ¡ 1 and rearranging terms. in order to have 1 + j > 0 and 1 + j < ¯ . ¯ . for utilities that are concave in consumption. there exits a real number j such that ° t ¯ ¡t = (1 + j) ° t¡1 ¯ ¡(t¡1) : Using the necessary conditions (25) . then 1 + j 2 0. VbD (k. that is. (26) and (28) . we obtain ° t+1 ¯ ¡(t+1) = (1 + j) ° t ¯ ¡t . which implies ° t = 0 through equation (28) : Therefore.

the policies that solve problem (41) are ¼ t and. then ° t ¯ ¡t approaches zero and. ¯ (1 + ") w ¡ (1 ¡ ¾) fl (k. by the recursivity of (39) .and respectively. £ ¤ If 1 + j belongs to (0. they solve problem (36) : 36 . in turn. b) > ¡ 4 ¯ (1 + ") w ¡ (1 ¡ ¾) f (k. we obtain that both ³ ´ 1 » t and ´t become constant. l) ¡ 1 ¡ 1 ¾ ¡ k+b ¢ f (k. Moreover. ¥ u (c) (fl (k. bt . ¥ Proof of Lemma 4 First. ¯ ¯ where E and F are de…ned in equation (71). Moreover. we must show that given a policy plan ¦m . bt+1 ) from date t + 1 on. 1) for zero capital taxes at incentive-constrained steady states can be written as 1 1 ¡ (E ¡ F ) > VbD (k. thanks to condition (27). Thus. l) ¡ w) 1 5. l) ¡ w) D Vb (k. those generated by ¦m for date s > t: As a result. More precisely. we require condition (73) and · ¸ 1 u0 (c) (fl (k. the allocations generated from F m are the optimal response to that policy. l) converges to a positive constant. given that constraint (3) holds for all dates s ¸ t for problem (41). l) (75) respectively. the continuation allocation of F m solves the individual’s problem (36) for every history ht . b) > ¡ . Note that the solution of problem (41) for (kt . The condition 1 + j 2 (0. in order to have 1+j > 0 and 1 + j < 1. the policies that solve problem (41) are exactly the policies that the consumer faces when solving (36). Using conditions (25) and (28) . l) l l ¯ c 2 0 3 (74) Proof of Lemma 3 The proof of Lemma 2 shows that a real number j exists such that ° t ¯ ¡t = (1 + j)t ° 0 . 1 + ¯ ¡t (° ¤ ¯)t approaches zero and. then equations (37) and (38) are both satis…ed. Hence. Then. ³ ´ VbD (k. b) > ¡ E. thus. ¼ t ) at date t + j for j ¸ 1 coincides with the solution of (39) for (kt+1 . 1). ¿ k = 0 at the steady state. (´t + » t ) ¯ ¡ fk (k.

¾ = 0. ¥ (5) of the individual implies as = 0 for all dates s ¸ t + 1: This result holds because 37 . ¥ Proof of Lemma 5 In order to characterize the Markov equilibrium (¦m . F m ) satis…es as = 0 for all dates s ¸ t + 1: If the initial savings are strictly positive. When the economy is at date s. the optimality conditions imply that rt = rmin . given the allocation rule F m . then ¿ k > 0 at 1 incentive-constrained steady states. b) · ¯ . ¥ Proof of Proposition 1 For utilities that are linear in consumption. suppose that the government at date t chooses rmin at all future dates s ¸ t + 1. the pair (¦m . ¼ s g1 s=t+1 that solves the government problem at date We will show that the equilibrium (¦m . this implies ks = bs = 0 for all dates s ¸ t + 1. given the non-negativity constraints for capital and debt. equilibrium. 1+" holds. b) > ¯ . Second. Hence. If VkD (k. which gives rise to an optimal as > 0. F m ) satis…es as = 0 for all dates s ¸ t + 1: Moreover. F m ) is a sustainable t must solve the government problem at date t + 1. the continuation policy of ¦m solves the government’s problem (34) for every history ht¡1 . Appendix 7:1 showed the ¡ " ¢ 1 following results. the …rst-order condition the production function satis…es an elasticity of substitution between capital and labor greater than one. which violates the incentive constraint (40) for the government problem (39) at date t. this is the case since they satisfy constraint (40). we will focus on the government problem (39) yielding ¦m . that is. First. This means that. if individuals follow the allocation rule F m and the government policies from date t + 1 onwards are generated from ¦m . then there is no policy ¼ t at date t which satis…es the budget constraint (35) and improves welfare. By construction of ¦m and F m . F m ). the inequality is reversed. the government problem (39) yields a …rst-order condition for the initial after-tax return implying that rs = rmin . ¿ l 2 0. Since the pair (¦m . Second. It su¢ces to show that no deviation improves welfare. and at any arbitrary date s ¸ t. the sequence fds . If rmin is su¢ciently close to zero. we must prove that. if VkD (k. at > 0. First. for an arbitrary date t. suppose that the government problem at date s chooses some rs > rmin .Second. that is.

µ ¶ 1 1 we get that 1 + k > ¯ : Thus. Therefore. from equation (22) . then the necessary condition (32) for incentive-constrained steady states does not hold. min ¾+" . l) ¡ w) ¡ ¢ Q= . ¿ l 2 0. First. Lemma 1 implies ¿ k = 0: Using equations (8). As a result. Second. Appendix 7:2 showed that ¿ l 2 0. P = ¾ k fl (k. if VkD (k. which is increasing in time. Q > 0. l) c " # u0 (c) (fl (k. b) ¡ ¯ u0 (c) ¸ 0. if rmin · 1 ¯ ¡ 1. then ld ¸ l. l) : ¯ ¯ 38 ¡ 1. equation ¡ £ ¤¢ (76) implies (1 + ") w ¡ (1 ¡ ¾) fl (k. (78) . l) > 0 and. we will show that. " ¡ " ¢ For ¾ = 0. l) + ¾ k fl (k. a solution £ ¤ 1£ ¤ for Lagrangian (13) requires » t+1 ¡ » t · ¯ » t ¡ » t¡1 . we obtain » t+1 = P t [» 0 ¡ Rt Q] : Here we have (76) converges to a negative number. we know that rmin < ¯ : Second. thus. and Rt > 0. 1+" . then » t+1 # ¡ ¢ (1 + ") w ¡ (1 ¡ ¾) fl (k. Solving for » t+1 in equation (53) . We will next show that. l) c £ ¤ £ ¤ £ ¤ Rt = P ¡1 1 + ¯ ¡t (° ¤ ¯)0 + P 1 1 + ¯ ¡t (° ¤ ¯)t + ::: + P ¡t 1 + ¯ ¡t (° ¤ ¯)t . we can write condition (32) as à ! µ ¶ d ¿l 1 ¿l = rmin : ¯ " ¡ (1 + ") ¿ l " ¡ (1 + ") ¿ ld 1 First. If P · 1. given that » t+1 ¡ » t ¸ 0. l) c ¡ ¢ . where Z is de…ned in equation (54). Z · 0 holds. it results that. then ¿ k ¸ 0: ¥ Proof of Proposition 3 with P > 0. we obtain µ ¶ 1 1 k+ ¡ 1 b + l"+1 + G = f (k. if rmin · d (77) 1 ¯ which implies ¿ l < ¿ l : Combining equations (22) and (23). (31) and (62) . P > 1 must be satis…ed: Next. (c) 1 From Appendix 7:1. 1 : 1+" ¤ £ We next characterize the capital tax. ¾ k fl (k.Proof of Proposition 2 £ ¤ We show …rst that ¾ » t+1 ¡ » t ¸ 0.

1 ): 1+" The proof of Lemma 2 showed that incentive-constrained steady states require VbD (k. Proof of Proposition 4 £ ¤ £ ¤ Equation (28) implies that ¾ » t+1 ¡ » t ¸ 0 holds and. min ¾+" . it follows ¡ 1. then incentive-constrained ¯ 1 steady states satisfy 1 + j 2 (0. ld : (79) From inspection of equations (78) and (79) . ¿ l 2 (0. by Lemma 1. then equation (80) is satis…ed for all rmin : ¥ · ¿ l and. which implies ¿ l · ¿ l through equation (8). ¿ k = 0: ¥ Proof of Corollary 1 1 Since rmin < ¯ . thus. it results that. thus. if b = 0. ¿ k = 0. we …nd ¡ ¢"+1 ¡ ¢ rmin (k + b) + ld + G = f k. ³ ´³ ´ 1¡¯ 1 We proceed next to show that. if rmin · then ¿ ld d (80) 1 ¯ that the steady state must be incentive-unconstrained and. the steady state is incentive-unconstrained and. 1). if 2¡¯ ¸ rmin . ¯ (1 + ") (1 ¡ ¿ l ) ¡ (1 ¡ ¾) ¯ (1 + ") 1 ¡ ¿ ld ¡ (1 ¡ ¾) where ³ ´ 3 d ¡ d¢ # 2 (1 + ") 1 ¡ ¿ l ¡ (1 ¡ ¾) rmin u ct 4 ³ ´5 : Y = ¡ ¢ 1 0 kt +bt u (c) ld ¡ (1 ¡ ¾) ¡ ¾r (1 + ") 1 ¡ ¿ ¯ min d " 0 ct ³ kt +bt cd t ´A : (81) 39 . If rmin = 0. Using equations (62) and (8). As a result. condition (75) for 1 + j < 1 at incentive-constrained steady states becomes 0 ¡ ¢ d µ ¶ 0 l u0 cd ¿ l u (c) ¿ 1 t > rmin @ ¯ (" + ¾) ¡ (1 + ") ¿ l (" + ¾) ¡ (1 + ") ¿ ld ¡ ¾r min Inequality (81) can be written as à ! µ ¶ d 1 u0 (c) ¿ l 1 u0 (c) ¿ l ¡ ¢ > ¤ Y. b) be strictly negative and belong to the range speci…ed by (72).Using constraints (60) and (61). b) = 0. therefore. ¯ ¯ then ld ¸ l. Hence. the system (60) ¡ (61) implies that VbD (k. if µ ¶ 1 1 rmin (k + b) · k + ¡ 1 b. condition (77) does not hold. As a result.

we prove by total di¤erentiation 1 ¯ d sustainable consumption to the consumption after a deviation at date t. l) l ¤ ¤ ¾ 4c + " ¡ w4l . l) 4¤ > ¡fl (k. c l µ ¶ wl 1¡¾ 4¤ ¡ (1 + ") w4¤ · 0: c l c If 1 ¡ ¾ (84) (85) wl ¸ 0. l) l c fl (k. we can write < 0. l) : t that is.In order to show that inequality (81) holds at incentive-constrained steady states. we get t c (82) Since rmin · 1 ¯ ¡ 1. From 1 ¡ ¾ < 0. in turn. c wl c¾ l" d . l) with the following sign: sign d¿ ¡ l Here ¡ wl labor is greater than one. we proceed in two steps. then inequalities (84) and (85) imply l · ld and c < cd and. l) 4¤ > 0. l) (86) < 1 because the elasticity of substitution between capital and (87) . Since ¿ l = 1 ¡ ¿ l > ¿ l : Consider next 1 ¡ ¾ c fl (k. we show that Y < 1 holds ³ ´³ ´ 1¡¯ 1 whenever 2¡¯ ¸ rmin . First. l) l fl (k. equations (22) and (60) allow us to write ¡ ¢¾ ¡ d ¢1+" cd ¡ cd lt · ct ¡ c¾ l1+" : t t (83) Equations (82) and (83) imply the following two inequalities: 4¤ ¡ fl (k. 1). Let us denote by 4¤ the change in c from the best c ¯ that. if ¡ d¢ cd ¡ f k. lt > c ¡ f (k. then c < cd and ¿ l > ¿ l . Thus. l) ¶ ¢ µ µ ¶¶ ¶ µ fll (k. we can write " # ¾ wl ¡ 1 c fl (k. 1+j 2 (0. Combining equations (23) and (61) . c l c (1 + ") w 40 µ fll (k. ¡ 1 ¸ rmin . 4¤ > 0 means c < cd . l) l wl wdl : = ¡sign ¾ dc + " ¡ c fl (k. l) µ µ µ ¸ ¶· ¶¶ 1 wl ¤ fll (k. 4¿ l = ¡ fl (k. it follows that c " # ¾ wl ¡ 1 c 4¤ > ¡ 4¤ : l c (1 + ") w Combining equations (84) and (87) . Second. l) 4¤ > ¡4¤ .

which implies µ 1 (1 + ") w ¶· µ ¶ ¸ wl fl (k. Y < 1: We can write condition Y < 1 as # ¡ ¢ 1 µ ¶ " ³ ´ i rmin u0 cd ¡ ¯ u0 (c) h kt + b t d t ¾rmin < (1 + ") 1 ¡ ¿ l ¡ (1 ¡ ¾) : 1 0 cd u (c) t ¯ Using equation (74). l) ¾ wl ¡ 1 c then c < cd and ¿ l > ¿ : ¢ ¡ d If 4¤ < 0. l) ¾ ¡ 1 + (1 + ") w 4¤ > 0: c c d The previous inequality means that 4¤ > 0. ¥ < 1. thus. then sign ¡4¤ l > 0 and. there c Consider l > ld . we can write l ¶¶ µ µ µ ¶ wl ¤ fll (k. (1 + ") 1 ¡ ¿ ¡ (1 ¡ ¾) > ¾ ¡1 ¯ c as ¿ l > ¿ l : Combining equations (88) and (89) . l · ld or l > ld . that is. ¡ rmin c ¯ u0 (c) which is satis…ed whenever rmin Since ³c´ 1¡¯ < ¯ · ¸ "³ ³ c´ + (1 ¡ ¯) cd ´³ 1¡¯ ¯ d h 1 2¡¯ ih 1¡¯ ¯ i ¸ rmin implies that (88) (89) à ¡ ¢ !#¡1 u0 cd : u0 (c) ¸ rmin is a su¢cient condition for cd Y < 1 and. for zero capital taxes at incentive-constrained steady states. c < cd : Concerning labor. namely. We will next show that. 4¤ < 0: Using equations (86) and (87) . l) l ¤ ¤ w4l ¸ sign (¡4¿ l ) = sign ¾ 4c + " ¡ c fl (k. l) l c + "¡ sign w4¤ : ¡ (1 + ") l fl (k. again ¿ l > ¿ l : Hence. if l ¿ ld 1 ¯ are two possibilities. we get · ¸µ ¶ ´ i h ³ 1 kt + bt ld . given previous results. thus. it results that 1 2¡¯ ´ 41 . then it is clear that ¿ l > ¿ l : ¡ 1 ¸ rmin . If l · ld . that is. l) ! µ Ã" à ! µ ¶¶# ¾ wl fll (k. we obtain " à ¡ ¢ !# µ d¶ u0 cd 1 c rmin < [1 ¡ ¯] .