# Public investment, tax evasion, and the welfare effects of a tariff reform.

I. INTRODUCTION

The literature challenging the relevance of the first-best result that a small open economy gains from a tariff reform has focused on the revenue-neutral case wherein the loss of tariff revenue is fully neutralized by a coordinated increase in domestic taxes (Emran and Stiglitz, 2005; Munk, 2006). (1) While this is a useful benchmark, the experience of developing countries indicates that the trade reforms are rarely revenue neutral. An International Monetary Fund (IMF) staff review of various country experiences found that "nearly half of the low-income countries that cut their tariff rates over the past 20 yr, and suffered an associated revenue loss, recovered less than 70 percent of this lost revenue from other sources" (IMF, 2005; also see Lin, 2000, for evidence from the Chinese tariff reform). This finding is consistent with the stylized fact that trade taxes account for one-third of the total tax income in developing countries (Dean, Desai, and Reidel, 1994; Tanzi, 1992), while the amount of revenue government can collect from other sources is limited due to tax evasion and a large informal sector (Acharya, 1985; Bearse, Glomm, and Janeba, 2000; de Soto, 1989).

Due to the loss in revenue from the tariff reform, the first and major casualty on the expenditure side has been the public investment in physical and social infrastructure (Roubini and Sachs, 1989; World Bank, 1988). Roubini and Sachs note, "in periods of restrictive fiscal policies and fiscal consolidation capital expenditure are the first to be reduced (often drastically)." This is disturbing for the social return on infrastructure investment is typically much higher than the return on private investment as levels of infrastructure in developing countries are suboptimally low (Lin, 2000; Pohl and Mihaljek, 1992). As a result, the welfare losses from reduced infrastructure investment have been of the first order, not just second order small. These losses must be taken into account for correctly assessing the welfare outcome of the tariff reforms of past few decades.

The objectives of this paper were, therefore, the following: a quantitative assessment of the welfare effect of the tariff reforms of past few decades in a model that recognizes the link from the tariff cut [right arrow] revenue loss [right arrow] lower public investment that is missing in the current literature. For this purpose, an overlapping generations model is used in which, consistent with the overwhelming evidence from developing countries, there is tax evasion. There is a growing literature that recognizes the importance of tax evasion for policy analysis for developing countries (Arana, 2004; Chen, 2003; Emran and Stiglitz, 2005; Gupta, 2007). As discussed later, the presence of tax evasion rationalizes the government's inability or unwillingness to generate offsetting revenue from domestic taxes.

Economic theory only tells us that with multiple sources of distortions, as in our case, a tariff reform may lower welfare. However, to go beyond this and to assess the actual welfare outcome of the tariff reforms of past few decades, a quantitative analysis is necessary. When this is done, the results turn out to be much more pessimistic: there is a strong presumption that the effects have been negative as welfare falls in most of the scenarios considered in the paper; compared to a potential welfare gain of .339% of gross domestic product (GDP) for the revenue-neutral reform, the fall in welfare might have been as large as .869% of GDP. As each period in the model is 20-yr long, these gains and losses are large as they are percentages of the GDP for 20 yr.

The paper also suggests why a benevolent government may have been unwilling to recover the lost trade revenues through increased domestic taxation. If the government cannot effectively fight tax evasion, a coordinated domestic tax reform will only partially recover revenue lost due to the tariff reform. Since empirical evidence suggests that when government revenue falls, public investment is not only the major but also the first casualty (Roubini and Sachs, 1989), the partial recovery of lost revenue would fail to stem the fall in public investment and will only saddle the economy with additional distortionary losses due to larger tax evasion. The government would, therefore, avoid domestic tax increases despite a significant decrease in its revenues.

When the model is extended to include other relevant features of developing countries, additional losses arise that are quantitatively significant and therefore further tilt the balance against the desirability of a tariff reform. For example, with the inclusion of elastic labor supply, the potential gain for the revenue-neutral reform falls from .339% to .112% of GDP. The inclusion of the audit cost in the model also reduces the potential gain for the revenue-neutral reform by a similar amount (.339% vs. .128%). Furthermore, the calibrated model with audit cost also shows that it is not possible for the government to increase its revenues by simply raising the audit rate. This happens because the marginal cost of increasing the audit rate is higher than the marginal revenue raked in by the increased audit.

The remaining part of the paper is organized as follows. Section II outlines the model. Section III contains the details of calibration. The welfare analysis of the tariff reform is presented in Section IV. Section V extends the model to include elastic labor supply and audit cost and discusses policy implications of the paper. Section VI concludes.

II. THE MODEL

The paper considers a small open overlapping generations economy that uses labor and capital to produce a homogeneous good, which can be consumed or used as input for the production of capital. The capital is produced by combining the domestic good and an imported input in a fixed proportion. Since these countries usually import "critical" inputs and machines for which there is little scope for substitution from within the country (see Buffie, 2001), the assumption of fixed proportions is a reasonable approximation for the developing countries. The economy also imports a consumer good that is not produced domestically. The country cannot borrow from abroad, and hence, the current account is balanced in each period. Each generation in the economy lives for two periods. The population of each generation is constant and has measure 1. All agents in a generation are born identical. Each agent has measure 0 and is endowed with one unit of labor when young, which he supplies inelastically.

The choice of the model is dictated by following considerations. The assessment of welfare implications in presence of public investment needs a dynamic model as public investment affects the intertemporal trade-off faced by the agents in the economy. In addition, as the economy typically spends a significant time away from the steady state, in models with capital, a comparative static analysis in a static model would fail to capture the welfare changes occurring during the transition to the new steady state. The overlapping generations model has been used for policy analysis in the presence of tax evasion by Arana (2004), Chen (2003), and Gupta (2007), and it simplifies analysis. (2) Finally, as the paper focuses on the effects of the fall in public investment and the distortionary loss arising from tax evasion, in the baseline model, it is assumed that the labor is supplied inelastically. Elastic labor supply, which strengthens the results of the paper, is introduced later in Section V.

A. Preferences and Utility Maximization

The agents are modeled as having a time additive separable (von Neumann-Morgenstern) utility function where utility depends on consumption in each period. (3) Let [~.V](E, P, [~.P]) denote the per-period (indirect) utility function where P is the price of the domestic good, [~.P] is the price of the imported consumer good, and E is the consumption expenditure. [~.V](*) is strictly increasing, strictly concave, and twice continuously differentiable in E. It also satisfies Inada conditions in E. By choosing domestic good as numeraire, we define V(E, [~.P]) [equivalent to] [~.V] (E, 1, [~.P]).

The representative agent of generation t has labor income [w.sub.t] when young, which also equals the wage. When old, that is, in period t + 1, he derives income from the capital accumulated in period t. The government levies a tax at the rate [[tau].sub.[i,t]] on the labor income of period t, which the agent can evade. (4) On receiving his income, the agent decides his saving, [s.sub.t], and the fraction of labor income, [x.sub.t], on which to evade tax. (5) He cannot diversify away the risk of being caught while evading taxes, although at the time of choosing [x.sub.t], he knows the probability of his being caught and the penal tax rate, [[tau].sub.[i,t].sup.p].

The government audits a fraction, p, of the returns. On audit, the underreporting of labor income is detected with probability 1. While p represents both the audit rate and the probability of being caught, in what follows, one or the other interpretation will be highlighted according to what appears more natural in that context. An agent caught evading taxes in period t pays taxes from his saving at a higher penal tax rate, [[tau].sub.[i,t].sup.p] in the same period. After paying penal taxes, if any, the remaining amount is used to accumulate capital. Let the capital accumulated by the agent be [k.sub.[1,t+1]], if not caught, and [k.sub.[2,t+1]] otherwise.

The government levies tariff at rate [[tau].sub.[c,t]] on the quantity of the imported consumer good that is imported in period t. As world prices of all imported goods are normalized to 1, the domestic price of imported consumer good, [~.P], equals 1 + [[tau].sub.[c,t]]. The government also imposes tariff at rate [[tau].sub.[e,t]] on the imported capital input in period t. (6) With fixed proportions technology in capital production, the imported capital input bears a constant ratio to the total capital production, and hence, one can instead assume that a constant fraction, [gamma], of the capital stock is imported. Thus, the domestic price of capital in period t in terms of domestic good becomes 1 + [gamma][[tau].sub.[e,t]].

The representative agent of generation t, therefore, solves the following problem:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (1)

subject to

[E.sub.[y,t]] + [s.sub.t] [less than or equal to] [[x.sub.t] + (1 - [x.sub.t])(1 - [[tau].sub.[i,t]])][w.sub.t] + [j.sub.t], (2)

[E.sub.[0,t+1].sup.1] [less than or equal to] [[r.sub.[t+1]] + (1 + [gamma][[tau].sub.[e,t+1]])(1 - [[delta].sub.k])][k.sub.[1,t+1]], (3)

[E.sub.[0,t+1].sup.2] [less than or equal to] [[r.sub.[t+1]] + (1 + [gamma][[tau].sub.[e,t+1]])(1 - [[delta].sub.k])][k.sub.[2,t+1]], (4)

(5) [k.sub.[1,t+1]] = [[s.sub.t]/[(1 + [gamma][[tau].sub.[e,t]])]],

(6) [k.sub.[2,t+1]] = [[([s.sub.t] - [[tau].sub.[i,t].sup.p][x.sub.t][w.sub.t])]/[(1 + [gamma][[tau].sub.[e,t]])]], 0 [less than or equal to] [x.sub.t] [less than or equal to] 1.

where [beta] is the subjective discount factor; [E.sub.[y,t]] is the consumption expenditure of an agent of generation t when young; [E.sub.[0,t+1].sup.1] is his consumption expenditure when old (i.e., in period t + 1) if he is not caught evading taxes; [E.sub.[0,t+1].sup.2] is his consumption expenditure when old if he is caught evading taxes; [j.sub.t] is the lump-sum transfer from the government; [r.sub.[t+1]] is the capital rental from period t to t + 1; and [[delta].sub.k] [member of] [0, 1] is the rate of depreciation of private capital. As the utility function is strictly increasing in expenditure in each period, all budget constraints hold with equality in equilibrium.

B. Technology and Profit Maximization

On the production side, following Barro (1990), government spending augments the productivity of each firm. The specification follows Futagami, Morita, and Shibata (1993) as the stock of public capital (G), and not public spending, affects productivity. The firms are identical and have Cobb-Douglas production function, exhibiting constant returns to scale in private capital, K, and labor, L. Thus, one can assume that there is a single firm in the economy and its output is given by

(7) F(K, L; G) = [AG.sup.[theta]][K.sup.[alpha]][L.sup.[1-[alpha]] [equivalent to] Y,

where [alpha] > 0, [theta] > 0 and [alpha] + [theta] < 1, and G, K, L, and Y are economy-wide aggregates. Note that although there are external increasing returns to scale at the aggregate level, the production function is characterized by decreasing returns in the accumulable factors, and there is no long-run growth in the economy. (7) The output per person of the generation supplying labor is

(8) y = f(k; G) = [AG.sup.[theta]][(K/L).sup.[alpha]] = [AG.sup.[theta]][k.sup.[alpha]].

The firm's problem is straightforward. It chooses capital and labor to maximize profit in each period,

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (9)

C. The Government Budget Constraint

To finance public investment, the government has access to three sources of revenue. Let [X.sub.t] be the (average) fraction of income not reported by the agents of generation t and [W.sub.t] be their aggregate wage income in period t. Then, the government revenue from labor income tax is [[[tau].sub.[i,t]] (1 - [X.sub.t]) + p[[tau].sub.[i,t].sup.p] [W.sub.t]. Its revenues from the tariff on imported capital equals [[tau].sub.[e,t]][gamma][[K.sub.[t+1]] - (1 - [[delta].sub.k])[K.sub.t]].Let [[kappa].sub.t] [member of] [0, 1] be the expenditure share of the imported consumption good, then the amount spent on consumer imports is [[kappa].sub.t][[E.sub.[y,t]] + (1 - p)[E.sub.[0,t].sup.1] + p[E.sub.[0,t].sup.2], of which a fraction [[[tau].sub.[c,t]]/(1 + [[tau].sub.[c,t]])] is tariff revenue. Thus, the government revenue is

(10) [[bar.R].sub.t] = [[[tau].sub.[i,t]](1 - [X.sub.t]) + p[[tau].sub.[i,t].sup.p][X.sub.t]][W.sub.t] + [[tau].sub.[e,t]][gamma][[K.sub.[t+1]] - (1 - [[delta].sub.k])[K.sub.t]] + [[[tau].sub.[c,t]]/[1 + [[tau].sub.[c,t]]]][[kapp].sub.t][[E.sub.[y,t]] + (1 - p)[E.sub.[0,t].sup.1] + p[E.sub.[0,t].sup.2]].

The government revenue in excess of public investment is rebated to the current young in a lump-sum manner. Hence, the government's budget constraint for period t is simply

(11) [[bar.R].sub.t] = [G.sub.[t+1]] - (1 - [[delta].sub.G])[G.sub.t] + [J.sub.t],

where [[delta].sub.G] [member of] [0, 1] is the rate of depreciation of the public capital; [J.sub.t] is the transfer made to generation t in period t; [G.sub.[t+1]] is the stock of public capital that enters the production function of the firms in period t + 1.

D. The Competitive Equilibrium

The competitive equilibrium for this economy is defined as:

Definition. A competitive equilibrium for this economy is a sequence {[E.sub.[y,t]], [E.sub.[0,t+1].sup.1], [E.sub.[0,t+1].sup.2], [x.sub.t], [s.sub.t], [k.sub.[1,t]], [k.sub.[2,t]], [r.sub.t], [w.sub.t], [K.sub.t], [L.sub.t], [X.sub.t], [S.sub.t], [W.sub.t], [G.sub.t], [j.sub.t], [J.sub.t], [[tau].sub.[i,t]], [[tau].sub.[i,t].sup.p], [[tau].sub.[e,t]]} such that for every t

(1) given {[r.sub.[t+1]], [w.sub.t], [j.sub.t], [[tau].sub.[i,t]], [[tau].sub.[i,t].sup.p], [[tau].sub.[e,t]), {[x.sub.t], [s.sub.t], [E.sub.[y,t]], [E.sub.[0,t+1].sup.1], [E.sub.[0,t+1].sup.2]} solves the optimization problem (Equation 1) for generation t;

(2) given {[r.sub.t], [w.sub.t], [G.sub.t]}, {[K.sub.t], [L.sub.t]} maximizes profit of the firm as in Equation (9);

(3) given {[x.sub.t], [s.sub.t], [E.sub.[y,t]], [E.sub.[0,t].sup.1], [E.sub.[0,t].sup.2], [w.sub.t], [K.sub.t], [G.sub.t]}, government policy {[G.sub.[t+1]], [J.sub.t], [[tau].sub.[i,t]], [[tau].sub.[i,t].sup.p], [[tau].sub.[e,t]]} satisfies government's budget constraint (Equation 11);

(4) aggregate and individual level variables are consistent; and

(5) markets for capital, labor, and output clear (8)

[K.sub.[t+1]] = (1 - p)[k.sub.[1,t+1]] + p[k.sub.[2,t+1] = [S.sub.t] - p[[tau].sub.[i,t].sup.p][X.sub.t][W.sub.t], (12)

(13) [L.sub.t] = 1,

[Y.sub.t] = (1 - [[[[tau].sub.[c,t]][[kappa].sub.t]]/[1 + [tau]c, t]])([E.sub.[y,t]] + [E.sub.[0,t-1].sup.1] + [E.sub.[0,t-1].sup.2]) + ([K.sub.[t+1]] - (1 - [[delta].sub.k])[K.sub.t]) + ([G.sub.[t+1]] - (1 - [[delta].sub.G])[G.sub.t]). (14)

E. Solving for the Competitive Equilibrium

The firm's profit maximization yields following first-order conditions:

(15) [K.sub.t]: [r.sub.t] = [F.sub.k] ([K.sub.t], [L.sub.t]; [G.sub.t]),

(16) [L.sub.t]: [w.sub.t] = [F.sub.L]([K.sub.t], [L.sub.t]; [G.sub.t]).

The first-order conditions for interior solution for maximization of agent's utility are

[s.sub.t]: [V.sub.E]([E.sub.[y,t]], 1 + [[tau].sub.[c,t]] = [beta][[[[r.sub.[t+1]] + (1 + [gamma][[tau].sub.[e,[t+1]])(1 - [[delta].sub.k])]]/[1 + [gamma][[tau].sub.[e,t]]]] x [(1 - p)[V.sub.E]([E.sub.[0,t].sup.1], 1 + [[tau].sub.[c,t+1]]) + p[V.sub.E]([E.sub.[0,t].sup.2], 1 + [[tau].sub.[c,t+1]]) (17)

[x.sub.t]: [V.sub.E]([E.sub.[y,t]], 1 + [[tau].sub.[c,t]]) = [beta][p[[tau].sub.[i,t].sup.p]/[[tau].sub.[i,t]]] [[[[r.sub.[t+1]] + (1 + [gamma][[tau].sub.[e,t+1])(1 - [[delta].sub.k])]]/[1 + [gamma][[tau].sub.[e,t]]]] [V.sub.E]([E.sub.[0,t].sup.2], 1 + [[tau].sub.[c,t+1]]). (18)

The first-order condition for [s.sub.t] is standard. In the first-order condition for [x.sub.t], the left-hand side is the marginal benefit of evading taxes on labor income and right-hand side is the marginal cost. It can be seen from Equation (18) that an increase in [[tau].sub.[i,t].sup.p] increases the cost of tax evasion and hence reduces [x.sub.t]. It can also be shown that as this decrease in tax evasion reduces the need for precautionary saving, [s.sub.t] falls as well. (9)

The government policy specifies the tax rates and the fraction ([[zeta].sub.t]) of government revenue that is rebated to the agents as transfers so that

(19) [G.sub.[t+1]] - (1 - [[delta].sub.G])[G.sub.t] = (1 - [[zeta].sub.t])[[bar.R].sub.t].

Solving for the competitive equilibrium involves solving Equations (2-8), (10-13), and (15-19), details of which are relegated to the Appendix. In particular, for computing the steady state, these 16 equations can be used to find the steady-state values of s, x, [E.sub.y], [E.sub.0.sup.1], [E.sub.0.sup.2], y, [k.sub.1], [k.sub.2], r, w, Y, K, L, [bar.R], G, and J, given the government policy defined by {[[tau].sub.i], [[tau].sub.i.sup.p], [[tau].sub.e], [[tau].sub.c], [zeta]}

III. CALIBRATION OF THE MODEL

To quantify the welfare outcome of a tariff reform, it is necessary to turn to numerical simulation, which requires choosing the functional forms, and values for the parameters. It may be emphasized that while the numerical simulations require choosing particular values of every parameter in the model, there are only a few whose values affect the outcome one is usually interested in. We do sensitivity analysis for such parameters where data are lacking or show a wide variation.

The utility function is chosen to be constant-elasticity-of-substitution-and-constant-relative-risk-aversion (CES-CRRA) with indirect utility function given by

(20) V(E, [P.sub.c]) = [1/1 - [sigma]][[[E/[P.sub.c]]].sup.[1-[sigma]]],

where [P.sub.c] [equivalent to] [[a.sub.1] + (1 - [a.sub.1])[(1 + [[tau].sub.c]).sup.[1-[micro]]].sup.[1/[1-[micro]] is the exact consumption-based price index and [a.sub.1] > 0 is a preference parameter. It allows us to choose the values for intertemporal (1/[sigma]) and intratemporal ([mu]) elasticities of substitution in consumption that are in accordance with the empirical facts.

In this setup, [sigma] plays a dual role: as the coefficient of relative risk aversion and as the inverse of the elasticity of intertemporal substitution. Using estimates for low- and middle-income countries in Ogaki, Ostry, and Reinhart (1996), [sigma] is set at 2, which implies elasticity of intertemporal substitution of .5. The value of [sigma] = 2 is also consistent with empirical evidence if it is interpreted as the coefficient of relative risk aversion.

In highly aggregated demand systems with 5-11 goods, the estimated compensated own-price elasticities lie in the range .15-.6. To account for the fact that the model has only two goods, and hence, the scope of substitution is much less, [mu] is set at .3 (this yields compensated own-price elasticity of imported good of .255) and sensitivity analysis is done for [mu] = .4.

Durlauf and Johnson (1992) study convergence across national economies and find that the share of physical capital in income or output ([alpha]) varies between .3 and .4. The poor countries have a capital share of income of .3, whereas it is .4 for the countries with intermediate income. For the developed countries, they find this share to be .33. For Latin American economies, Elias (1992) estimates a value of .5. Consistent with Elias as well as to account for the fact that capital implicitly also includes the intermediate inputs in the model, [alpha] is given a value of .5 (Table 1). The (annual) return to private capital (or the real interest rate) of 8% for developing countries is taken from Buffie (2001). In steady state, the return to private capital equals the capital rental, r. The share of the imported capital in private capital varies considerably across the developing countries. The parameter [gamma] is set at .5, which is in the middle of the range of estimates (.35-.65) in Buffie (2001) that are consistent with Taylor's (1990) illustrative Social Accounting Matrix (SAM) and Dervis, de Melo, and Robinson (1982). The depreciation of private and public capital at 7% and 4% per year are typical. The difference reflects the facts that private capital implicitly also includes imported intermediates in the model and public capital consists primarily of physical infrastructure, which typically depreciates slowly compared to plant and machinery. With each period in the model lasting 20 yr, this yields [[delta].sub.k] = .766 and [[delta].sub.G] = .558.

TABLE 1

Parameter Values for the Calibrated Model

Preferences

[beta] = .1898; [sigma] = 2, [micro]=.3, [kappa] = .15

Production function

[alpha] = .5; [theta] = .2; [[delta].sub.k] = .766; [[delta].sub.G] = -558; [gamma] = .5

Government policy

[[tau].sub.i] = .3: [[tau].sub.i.sup.p] = .6; [[tau].sub.e] = .4; [[tau].sub.c] = .8

[chi] = 2; [zeta] = .9394

Other

p = .1163

Pohl and Mihaljek (1992) show that the public capital is highly productive in developing countries. Analyzing the rate of return on 1,015 World Bank projects implemented in developing countries, they find the median and the average (annual) rates of return to be 14% and 16%, respectively. While Pohl and Mihaljek (1992) suggest that the projects submitted by governments for the World Bank financing may primarily include projects with above-average rates of return, Easterly (1999) summarizes evidence showing that the return to public investment in developing countries (especially in physical infrastructure) may actually be even higher (19%-29%). We choose a very modest value of 12% for the return on public capital and do sensitivity analysis for 16%. There are very different estimates of the elasticity of national output with respect to public capital ([theta]) varying from close to 0 to .2 (see Ai and Cassou, 1995; Lynde and Richmond, 1993). In our simulations, [theta] is set at .2, which is the estimate obtained by Canning and Fay (1993) and Fay (2001) using large cross-country data sets. Once the return on public investment is chosen, the choice of [theta] merely determines the share of public investment in public expenditure (1 - [zeta]), and [theta] = .2 gives a more realistic estimate for latter. The results of the paper, however, only depend on the higher return to public capital compared to private capital.

Developing countries have an escalated structure of protection with higher tariffs on consumer goods and lower tariffs on imported capital and intermediates. For example, as Vernengo (2004) reports, the average tariff on capital and intermediates in Brazil over 1960-1980 was 50% of that on the consumption goods. Berlinski (2000) provides similar evidence for Argentina. Accordingly, and following Edwards (1995), [[tau].sub.e] and [[tau].sub.c] are set to .4 and .8, respectively, which are typical pre-reform values for the developing countries. The tax rate on labor income ([[tau].sub.i]) of .3 is set so as to yield the share of tariffs in government revenue that is consistent with the estimates in Tanzi (1992) and Dean, Desai, and Riedel (1994).

The second consumption good in the model is entirely imported, although in reality a large portion of its demand is met by domestic production of this good or its very close substitutes. The share of the imported component of this good in consumption is about 10%, whereas the share rises to 15%-25% when consumption of the portion that is produced domestically as well as the domestically produced close substitutes is included (see Buffie, 2001). Each of these numbers is a valid choice for the share of the imported good in consumption from two different perspectives. From the perspective of matching the contribution of tariffs to government revenue, the consumption share of the imported consumer good in the model should be set to 10%. However, the tariff on imported consumer good also raises the price of its close domestic substitute(s). Hence, to capture the distortionary effect of the tariffs, the consumption share of the imported consumer good in the model should be set slightly higher. To balance these conflicting considerations, the results are presented for two cases. The baseline case sets the consumption share of imported good (kappa) at .15 in the initial steady state. Then, sensitivity analysis is done for a lower value of .1. It should be noted that the value of [kappa] in the baseline case corresponds to a scenario with larger gains from tariff reduction.

The values of [beta], x, p, and [zeta] are estimated from the model to match the data on the ratio of government revenue to GDP ([bar.R]/Y), the penal tax rate ([[tau].sub.i.sup.p]), and the returns to the public and the private capital. The ratio of the government revenue to GDP has been ascertained from Summers and Heston (1991) (Mark 5.6a), which reveals considerable variation across countries. It ranges from 10% to 30% for the middle 90% of the countries, and the average is lower for the developed countries than the developing countries. The model is calibrated for [bar.R]/Y = .20. We set [[tau].sub.i.sup.p] = 2[[tau].sub.i] = .6, which besides being empirically reasonable also yields plausible estimates of p in the range .1-.2. This implies that less than 1% of returns are audited every year, which accords with audit rate in India. (10)

A. Political Constraints on Government Policy

The paper does not model the political process that determines the ability of the government to fight tax evasion. The penal tax rate may be already very high, and a higher rate may infeasible due to the widespread nature of tax evasion. The paper also does not model the process by which government decides the extent to which to offset the loss of revenue arising from tariff reform. There may be resistance to raising the domestic tax rates; in cases where statutory tax rate can be raised, the penal tax rate may be already very high. In the face of an inertial penal tax rate, increasing the statutory tax rate may only increase evasion and not enable the government to restore [bar.R]/Y to its initial level. These constraints on the government's policy choices are labeled as "political constraints" for want of a better term. (11)

While the constraints on government policy are exogenous, it is nonetheless possible to analyze their impact on the welfare consequences of a tariff reform. To this end, define [[epsilon].sub.r] to be the fraction of lost tariff revenues that is offset by a coordinated increase in domestic taxes. Thus, [[epsilon].sub.r] is a quantitative measure of the severity of the constraints faced by the government. A value of [[epsilon].sub.r] < 1 implies that the government can only partially offset the loss of revenue from the tariff reform--the tariff reform is not revenue neutral; in particular, [[epsilon].sub.r] = 0 implies completely passive domestic tax policy with no changes in the domestic tax rates. The constraints on the ability to make up lost tariff revenues have been significant; recall the findings of the IMF staff review that "nearly half of the low-income countries that cut their tariff rates ... recovered less than 70 percent of this lost revenue from other sources" (IMF, 2005). In case of China, Lin (2000) provides a similar evidence where the share of tariffs in government revenue declined from over 10% in 1985 to 3.4% in 1998.

To recover fraction [[epsilon].sub.r] of its lost revenue, the government has two potential instruments, [[tau].sub.i] and [[tau].sub.i.sup.p], at its disposal. (12) Hence, a rule that links the penal tax rate to the statutory tax rate is needed to uniquely determine the government policy. A general linear penal tax rate rule has the form

(21) [[tau].sub.i.sup.p] = [[bar.[tau]].sub.i.sup.p] + [chi][[tau].sub.i],

where [[bar.[tau]].sub.i.sup.p] > 0 and [chi] > 0 are parameters. Landskroner, Paroush, and Swary (1990) study tax evasion as a portfolio choice under such a rule where [[bar.[tau]].sub.i.sup.p] has the interpretation of a penalty on the evaded income and [chi] of the penalty on the evaded tax. A proportional penal tax rate rule (i.e., a rule with [[bar.[tau].sub.i.sup.p] = 0) minimizes tax evasion, and as shown by Yitzhaki (1974), it also eliminates the substitution effect as defined in Allingham and Sandmo (1972). Thus, government policy is considered constrained if the penal tax rate increases less than proportionally with the statutory tax rate (i.e., [[bar.[tau].sub.i.sup.p] > 0). To quantify this constraint on government policy, define [[epsilon].sub.p] as the elasticity of the penal tax rate with respect to the statutory tax rate. A constraint on public policy in this dimension implies that [[epsilon].sub.p] is less than 1. In an interesting observation, Yitzhaki (1974) notes that United States and Israel were the only countries in 1974 that had a proportional penal tax rate system.

If the government budget shrinks, the brunt of the resource crunch is borne by public investment. (13) As mentioned before, Roubini and Sachs (1989) note, "in periods of restrictive fiscal policies ... capital expenditures are the first to be reduced (often drastically)." This agrees with the findings in the World Development Report (World Bank, 1988) that in the face of fiscal tightening, the public investment fell far more sharply (35%) than other current expenditures such as wages (10%). Hicks (1991) comes up with corresponding estimates of 27.8% and 7.2%. (14) These findings are not hard to understand. The effects of reduction in public investment become visible only when the gradual deterioration of public roads and overcrowding of existing infrastructure impacts productivity. A reduction of transfers and the public sector wage bill has more immediate consequences for politicians. Political expediency results in a disproportionate reduction in public investment.

Rodrik (1996) analyzes of the role of incentives of policy makers in economy policy reforms. For our purpose of quantifying the welfare effect of reduced public investment in wake of tariff reforms, it is, however, not necessary to formally model these incentives. What is critical is to be able to capture two empirically relevant outcomes: (1) public investment is the first and major casualty in the face of declining government revenue and (2) the decline in public investment is three to four times greater (as in Hicks, 1991 and the World Bank, 1988) when the government is able to recover only part of its lost tariff revenues (as in IMF, 2005).

The first fact suggests that the relationship between the fall in public investment and government revenue is monotonic but highly nonlinear. Such nonlinearity can be captured parsimoniously by making the ratio of the post-reform public investment to the pre-reform public investment ([lambda]), an exponential function of the reduction in revenue ([[epsilon].sub.r]). A polynomial relationship, while a plausible alternative, will involve more parameters; and the parameters will have less intuitive interpretation. As what is relevant for our results is the actual empirical relationship to which the function is calibrated and not its analytical form, the following rule is used to link [lambda] and [[epsilon].sub.r]:

[lambda] [equivalent to] 1 - [[1 - [upsilon]]/[1 - exp[-[psi]]]][1 - exp[-[psi](1 - [[epsilon].sub.r])]], (22)

where v[member of] [0, 1] and [psi] > 0 are parameters.

This specification implies that public investment decreases by fraction 1 - v when the government is constrained to keep domestic rates of taxation unchanged at pre-reform levels. In addition, public investment (as a fraction of government revenue) is unaffected if the government can fully neutralize the loss of revenue from tariff reform by a coordinated increase in domestic taxes which occurs if [[epsilon].sub.r] = 1. Furthermore, a higher value of [psi] implies that the brunt of the initial fiscal crunch falls on public investment with a greater intensity. This can be seen from Figure 1, which represents this relationship for v = .7 and two values of [psi], 3 and 5. In either case, public investment falls by 30% if the government is constrained to keep the domestic rates of taxation unchanged at the pre-reform levels, but for the higher value of [psi] = 5, there is a proportionally larger reduction in public investment (the lower curve in Figure 1) when the fall in government revenue is smaller, that is, [[epsilon].sub.r] is higher.

[FIGURE 1 OMITTED]

To match the second fact mentioned above, we set [upsilon] = .7 and [psi] = 3 so that, for the 50% tariff reduction considered later, when the government is able to recover only 75% of its lost tariff revenues ([[epsilon].sub.r] = .75), public investment falls by 3-4 times more (16.7% vs. 5%).

B. The Steady State

The steady state for the calibrated model is presented in Table 2 and is representative of a typical developing country. The share of tariffs (T) in government revenue, T/[bar.R], is .3991. This value is somewhat high, but well within the range of the estimates reported in Tanzi (1992) and Dean, Desai, and Riedel (1994). The larger share results from the higher consumption share of the imported consumer good that is assumed so that the welfare gain arising from the reduction in consumption distortion as a result of the tariff reform can be accurately captured.

TABLE 2

Steady State of the Calibrated Model

Tax evasion

x = .2591

Consumption-output ratios

Ey/y = .4548; [[E.sub.0.sup.1]/Y] = .5686; [E.sub.0.sup.2] = .2063

Saving- and capital-output ratios

s/y = .1220; (1 + [gamma][[tau].sub.e])K/Y = 0.1130; (1 + [gamma][[tau].sub.e])[[k.sub.1]/y] = 0.1220; (1 + [gamma][[tau].sub.e])[[k.sub.2]/y] = 0.4425

Annualized private capital to output ratio = 2.256 Government

[[bar.R]/Y] = .2; J/Y = .1879; G/Y = .0217; G/K = .1924; T/[bar.R] = .3991

Other

r = 8%, return to public capital = 12%

The annualized capital output ratio at 2.256 is also within the range of empirical values presented in Buffie (2001). Agents evade tax on 25.91% of their income, which is at the lower end of the range of estimates in the literature. (15) In the data, tax evasion is higher because, with the annual filing of tax return, the risk arising from being caught at evading taxes is much smaller than in the model where tax is assessed only once during entire working life. More specifically, for the same degree of risk aversion, the risk is much higher in the two-period model as being caught at evasion results in the agent consuming significantly less during entire period of retirement.

The value of [zeta] at .9394 is high as it implies only 6.1 % of the government revenue is invested in public capital. For Latin America, Glomm and Rioja (2003) estimate about 15% of government revenue is spent on infrastructure. However, the results of the paper depend only on the fact that public capital is much more productive than private capital on the margin. The stock of public capital does not matter as the loss depends on the inefficiency generated by the initial divergence between public and private capital productivity and the proportion by which public investment falls. Irrespective of the level of the stock of public capital, the tariff reform reduces public investment in the same proportion.

One reason that the calibrated share of public investment in government revenue is smaller than what is reported in official government data is the widespread corruption in the governments in the developing countries. The calibrated share is based on the return to public capital in World Bank-funded projects. Since the corruption in World Bank projects is presumably smaller, this rate of return on public capital is closer to the actual return on public capital. When this return is used to compute the economy-wide stock of public capital and investment, only the fraction of public investment funds that actually gets invested is accounted for. In contrast, in the official government accounts, public funds that "leak" due to corruption or inefficiency also appear as public investment. In the model, they appropriately appears as transfers because there is no production of goods and services associated with them. That they appear as transfers to the current young is also reasonable as current young represent the working generation.

In view of rampant corruption in developing countries, it is not surprising that the fraction of public funds that actually gets invested may be no larger than one half (see Bearse, Glomm, and Janeba, 2000 and references cited therein). In fact, this reconciles the apparently contradictory evidence that although public capital is found to be more productive than private capital, the productivity in public sector is much lower than the private sector in developing countries (Bearse, Glomm, and Janeba, 2000) when calculations are based on official government accounts.

IV. THE WELFARE EFFECTS OF A TARIFF REFORM

Consider a 50% reduction in tariff on both the imported capital and the imported consumer good. This approximately corresponds to the change in tariff levels that occurred in many developing countries during 1980s and 1990s. (16) The welfare effect of this reduction is assessed by numerically solving the nonlinear model. The details of numerical computations can be found in the Appendix.

To quantify the welfare gains or losses arising from the tariff reform, the additional consumption needed or enjoyed by each generation when young is discounted to the time of the reform using the domestic interest rate. To this, the surplus consumption enjoyed by the current old is added. Adding the net present values of the gains or losses accruing to various generations gives a measure of the net welfare effect of the tariff reform. In the overlapping generations framework, government can potentially achieve Pareto improvement through intergenerational transfers (see Auerbach and Kotlikoff, 1987). To abstract from these effects and to focus on the effects of tax evasion and fall in public investment, all revenue in excess of the public investment continues to be rebated in lump sum to the current young.

A. Numerical Simulation of the Reform

Table 3 presents the welfare outcome of the tariff reform for the varying degree of constraints on government policy. Each entry is the welfare gain as percent of the current period GDP.

In absence of any constraints on government policy, the government sets both [[epsilon].sub.r] and [[epsilon].sub.p] at 1. In this case, the reform is revenue neutral. As can be seen, the overall welfare gain from tariff reform amount to .339% of the current period GDP as distortions due to tariffs fall. This is a large welfare gain as each period in the model corresponds to 20 yr. The gains are, however, not evenly distributed. The current old gain as the tariff reform shifts tax burden away from them by reducing the tax on imported consumer good. Current young and future generations, however, lose from the reform. The reform raises labor income tax rate from .3 to .369. This lowers the income of the current young making them worse off despite the fall in the price of the imported consumer good. Their loss amounts to .817% of the current period GDP.

Future generations lose as capital stock in the economy falls reducing their consumption possibilities. Across steady states, capital stock falls by 6.31% as the young respond to declining income by saving less as fraction of total GDP and by reducing tax evasion. The fall in tax evasion also reduces precautionary saving (see Kimball, 1990). (17) The reduction in saving reduces capital accumulation as the saving turns into capital except for the amount paid as penalties for tax evasion. The capital stock declines despite the reduction in tariff on the imported capital. For example, in the period of reform, with a proportional increase in penal tax rate from .6 to .738, x falls from .259 to .188 along with a fall in the saving in the period of reform by 10.02%. As a result, the capital stock declines by 1.84% in the next period.

B. Constrained Policy and Welfare Loss

The gains that occur in an unconstrained scenario, however, vanish for very reasonable constraints on government policy along either dimension--ability to raise revenue or ability to fight tax evasion. First, consider the case of revenue-nonneutral reforms in which the government is unable to fully offset its revenue loss from the tariff reform ([[epsilon].sub.r] < 1), although it can effectively fight tax evasion ([[epsilon].sub.p] = 1). For such reforms, the associated reduction in public investment causes a welfare loss. If the government can increase [[tau].sub.i] to offset only 50% of the shortfall in revenues ([[epsilon].sub.r] = .5), then the loss outweighs the usual gains from tariff reduction. The resulting loss in an overall welfare is .055% of the current period GDP. This scenario is not unrealistic; recall, many countries have been able recover less than 70% of lost tariff revenues through coordinated domestic tax increases.

If the government is able to recover its revenue loss ([[epsilon].sub.r] = 1) by increasing domestic taxes, but cannot effectively fight tax evasion ([[epsilon].sub.p] < 1), tax evasion increases and the resulting distortion reduces the usual gain from the tariff reform. As Table 3 also shows, this gain is almost wiped out when [[epsilon].sub.p] is .5, which is a very plausible value in light of observation in Yitzhaki (1974) cited earlier. In this case, capital accumulation is higher in physical terms but is accompanied by a larger distortion from tax evasion, and in numerical simulations, future generations still lose from the reform. (18) Thus, given the current system of taxation, a welfare gain is not ensured even for revenue-neutral reforms.

The gains from tariff reform disappear even with very modest constraints on public policy. When the government offsets 75% of its revenue loss by increasing the statutory tax rate and raising the penal tax rate with elasticity .75 ([[epsilon].sub.r] = .75, [[epsilon].sub.p] = .75), there is a welfare loss of .017% of GDP. The constraints are so mild that the government revenue decreases by less than a percentage point from 20% of GDP to 19.13% and less than one-quarter (23.3%) of this reduction in the government budget falls on public investment. On the tax evasion front, these constraints imply that the government increases [[tau].sub.i.sup.p] from 60% to 68.69%, which is only marginally lower than the unconstrained value of 71.8% for the corresponding increase in [[tau].sub.i] from 30% to 35.9%. The situation gets far worse with greater, yet plausible, constraints on government policy in the developing countries. The reform now yields a loss of a magnitude similar to the gain that results in the absence of constraints on government policy (see Table 3).

The source of loss from revenue-neutral reforms is very similar to those in the static models of tax evasion where there is an intratemporal trade-off between the distortionary loss due to tax evasion and the benefits of revenue collected from taxation. For the revenue-nonneutral reforms, the trade-off in the model is, however, inherently inter-termporal, as by reducing the stock of public capital, the reform affects the (intertemporal) saving decision of the agents. The reduced availability of public capital reduces the marginal product of private capital, which reduces the saving in the economy. Thus, the stock of private capital falls as well, further reducing the productive capacity of the economy.

How does the stylized two-period structure of the model affect the results? This, if anything, significantly understates the losses arising from reduced public investment as the stock of public capital in the period of the reform (which has a duration of 20 yr) is predetermined. In reality, the public capital will start deteriorating much earlier. With agents living for multiple periods, the tax evasion will also be higher because, as mentioned earlier, the risk associated with tax evasion will be much lower. In multiperiod models, it would also be possible to set the measure of retirees relative to the working generations to a realistic value of less than 1. The gains of retirees, therefore, will have a lower weight in computation of the overall gain from the reform. Thus, the basic result that modest constraints on government policy can wipe out the gains from tariff reforms is likely to be robust to the extension of model to include agents living more than two periods.

It is also clear from Table 3 that for a given [[epsilon].sub.r], the government will set [[epsilon].sub.p], as high as it can as it leads to highest welfare. In other words, the government would fight tax evasion, as effectively as it can. Similarly, if government can effectively fight tax evasion ([[epsilon].sub.p] = 1), it would offset the fall in its revenues to the extent possible. However, note from Table 3 that when the governmen=t cannot raise penal tax rate proportionally with the statutory tax rate ([[epsilon].sub.p] < 1), the welfare loss is not monotonic in [[epsilon].sub.r]. Hence, in presence of both a limited ability to fight tax evasion ([[epsilon].sub.p] < 1) and a limited flexibility in coordinated domestic tax increases ([[epsilon].sub.r] < 1), avoiding an increase in domestic taxation would be better, despite a significant decrease in government revenue. This interaction of the political constraints suggests that, in view of the losses arising from tax evasion, the government may be "unwilling" to raise domestic taxes even when it is "able" to do so.

The nonmonotonicity of losses in [[epsilon].sub.r], when government fails to fight tax evasion effectively ([[epsilon].sub.p] < 1), arises from the change in the relative magnitude of two conflicting effects. The increase in [[epsilon].sub.r] mitigates the fall in revenue, which helps contain the adverse impact of reduced public investment. On the other hand, it also leads to a loss from tax evasion as [[epsilon].sub.p] < 1. How these effects vary with [[epsilon].sub.r] is determined by different factors. For the positive public investment effect, it depends on what fraction of additional revenue is invested in public capital, while for the negative tax evasion effect, it is determined by the elasticity of the penal tax rate with respect to the statutory tax rate, that is, [[epsilon].sub.p].

Figure 2 illustrates how these effects vary with [[epsilon].sub.r]. For a given [[epsilon].sub.p], as [[epsilon].sub.r] rises, the public investment effect becomes progressively stronger (solid curve in Figure 2) because a larger fraction of additional revenue collected goes to fund public investment. Note that this line asymptotes to x-axis as [[epsilon].sub.r] approaches 0: the positive effect rapidly vanishes with fall in [[epsilon].sub.r] as the increase in government revenue causes hardly any increase in public investment (see Figure 1). The loss from tax evasion effect also rises with [[epsilon].sub.r] as government's attempt to raise more revenue leads to larger tax evasion. But, as [[epsilon].sub.p] is unchanged, this effect does not rise as rapidly with [[epsilon].sub.r] as the public investment effect as suggested by the dotted lines (linearity is just for convenience of exposition) in Figure 2.

[FIGURE 2 OMITTED]

From Figure 2, it evident that given [[epsilon].sub.p] < 1, there is a threshold value of [[epsilon].sub.r], [[bar.[epsilon]].sub.r], such that, for [[epsilon].sub.r] < [[bar.[epsilon]].sub.r] the welfare outcome is worse than with [[epsilon].sub.r] = 0. Moreover, [[epsilon].sub.r] rises when [[epsilon].sub.p] falls: the adverse effect of tax evasion is greater when [[epsilon].sub.p] is smaller, and therefore, [[epsilon].sub.r] must rise more so that the positive effect of increase public investment is able to overcome the higher welfare loss arising from increased tax evasion for lower [[epsilon].sub.p]. When [[epsilon].sub.p] = .75, for example, it is not worthwhile to undertake domestic tax reform if the government can recover only 25% of its lost revenues. For [[epsilon].sub.p] = .5, increasing domestic taxes is not useful even if 50% of lost revenue can be recovered. Intuitively, when the government cannot effectively fight tax evasion, a coordinated increase in domestic taxation that only partially offsets its revenue loss just saddles the economy with distortionary losses due to tax evasion. It does not deliver a countervailing benefit by preventing the fall in public investment.

C. Sensitivity Analysis

Table 4 contains the results of sensitivity analysis for the parameters that affect the welfare outcome of the reform; and they are as expected. A higher return on public capital magnifies the loss arising from the reduction in government revenue. For [[epsilon].sub.r] = .75 and [[epsilon].sub.p] = 1, the reform delivers a welfare gain of .091% (Table 3) when the return on public capital is 12%. This turns into a loss of .190% with the increase in return on public capital to 16% (panel 1, Table 4); recall, this is the average rate of return on public capital found by Pohl and Mihaljek (1992). In the worst case, welfare loss is .869%. It may be noted that even a return of 16% on public investment is much lower than the estimates reported in Easterly (1999).

When the consumption share of the imported consumer good is smaller, the initial distortion from tariffs is also smaller, and so is the gain from the tariff reform. In the absence of any constraints of government policy, welfare gain reduces from .339% to .230% of GDP when [kappa] falls from .15 to .1 (Table 3 and panel 2, Table 4). Qualitatively, however, the welfare outcome is similar: mild constraints on government policy still cause a welfare loss.

As expected, a higher elasticity of substitution in consumption raises the gain from the reform due to the increased possibility of substitution towards the now cheaper imported consumer good, while the loss from the constraints on the government policy remains unchanged. The welfare outcome of the tariff reform, therefore, becomes more favorable, although reasonable constraints of government policy still cause gains from the reform to disappear (panel 3, Table 4).

V. SOME EXTENSIONS AND POLICY IMPLICATIONS

The result so far convincingly shows that the gains from tariff reform disappear for modest constraints on public policy, not only for the benchmark parameter values but also for the plausible alternative combinations of parameter values. In the same vein, we now investigate how the welfare calculus of the reform is affected if some additional features of the developing economies are included in the model. In particular, this section examines the implications of elastic labor supply and nontrivial cost of collection of domestic taxes. In the latter case, it also estimates the marginal cost of increasing the audit rate, p (which is also the probability of being caught in the model), and investigates how an increase in p in the presence of audit cost would affect the net revenue of the government. The section ends with some remarks on the policy implications.

A. Elastic Labor Supply

To allow for the labor supply to endogenously respond to policy changes, the utility function of the agents is augmented to include a term ([upsilon](*)) capturing the disutility from supplying labor ([l.sub.t]), which is equivalent to a preference for leisure.

The representative agent of generation t now solves the following problem:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (23)

subject to

[E.sub.[y,t]] + [s.sub.t] [less than or equal to] [[x.sub.t] + (1 + [x.sub.t])(1 - [[tau].sub.[i,t]])][w.sub.t][l.sub.t] + [j.sub.t], (24)

and Equations (3-6). The only change in the constraints for optimization is that now the labor income depends on the amount of labor supplied. The optimal choice of labor supplied depends on the trade-off between the disutility from labor and the utility from consumption made possible by the income earned by supplying labor. This trade-off is captured by the following first-order condition for [l.sub.t]:

[upsilon]'([l.sub.t]) = [V.sub.E]([E.sub.[y,t]], 1 + [[tau].sub.[c,t]])[[x.sub.t] + (1 - [x.sub.t]) (1 - [[tau].sub.[i,t]])][w.sub.t]. (25)

For numerical simulations, the disutility from labor is assumed to be given by

[upsilon](l) = -a[(b - l).sup.[xi]], (26)

which on substitution in Equation (25), along with use of Equation (20), gives

a[xi][(b - [l.sub.t]).sup.[[xi]-1]] = [([[E.sub.[y,t]]/[1 + [[tau].sub.[c,t]]]]).sup.-[sigma]]

[[x.sub.t] + (1 + [x.sub.t])(1 - [[tau].sub.[i,t]])] - [[w.sub.t]/[1 + [[tau].sub.[c,t]]]]. (27)

Equation (27) equates the marginal disutility from working (on left) to the marginal utility gain from additional income (on right). It is clear that, ceteris paribus, an increase in tax evasion raises labor supply as it increases the effective wage rate.

Before simulations, however, we need to calibrate the extended model for three new parameters, a, b, and [xi]. Three targets are used to find the values of these parameters. First target is the amount of labor is supplied by the agent in the initial steady state. This is set to 1 unit as in the baseline model. By ensuring that the initial equilibrium is same for both the extended model and the baseline model, this allows a comparison of the results of the two models. The second target is the fraction of the time agent spends working and is set at 30%, as is commonly assumed in the models with labor-leisure choice (e.g., see Atolia, Chatterjee, and Turnvosky, 2009). Last target sets the compensated wage elasticity of labor supply ([[xi].sub.w.sup.c]) at .27, which is well within the range of empirical estimates. (19)

The calibration yields the following values of the parameters: a = -.0044; b = 10/3, and [xi] = -11/3. For these parameter values, the Frisch wage elasticity of labor supply is .5, which also on the lower side of the range of values that Gourio and Noual (2006) experiment with. It may be mentioned that for the extended model, the Frisch elasticity of labor supply, [[epsilon].sub.w], with respect to (real) wage is

(28) [[epsilon].sub.w] = [b - l/l][1/1 - [xi]]

The welfare results for the tariff reform when labor supply is elastic are collected in Table 5. A comparison of the results in Table 5 and Table 3 shows that the inclusion of elastic labor supply strengthens the results of the paper; the welfare gain from tariff reform is uniformly lower with elastic labor supply. For example, the welfare gain from tariff reform in the best-case scenario with no constraints on government policy falls from .339% to .112% of GDP. For modest constraints with [[epsilon].sub.r] = .75 and [[epsilon].sub.p] = .75, the welfare loss jumps from .017% to .203% of GDP.

The reason is not very hard to understand. In the baseline model, labor supply is perfectly inelastic, and hence, the tariff reform constitutes a move from a distortionary to a nondistortionary source of raising government revenue. Whereas with elastic labor supply, the tariff reform is a move from one distortionary source to another. As a result, the losses from tariff reform are higher with elastic labor supply.

In other words, there is no tax burden associated with public investment after the implementation of tariff reform when the labor supply is inelastic. However, when labor supply is elastic, the provision of public investment via labor income taxation gives rise to a tax burden, which reduces the gain from the tariff reform. This tax burden can be measured as the difference between the welfare effects of tariff reform with inelastic and elastic labor supply, which allows us to answer an open question in public finance and expenditure literature: what is the tax burden associated with public investment when it is financed by distortionary taxation? (20), (21)

To provide an answer to this question, consider the unconstrained case where the level of public investment after the tariff reform is same as before the reform. In this case, with elastic labor supply, there is an additional welfare loss of .339 - .112 = .227% of GDP. Of this loss, 6.06%, that is, .0138% of GDP, is due to the increased financing of public investment by distortionary labor income tax. This follows from the fact the fraction of government revenue that is used for public investment is 1 - [zeta] = .0606. This loss is associated with the

increase in financing of public investment from distortionary labor income tax by .229% of GDP.

It is necessary to make one adjustment before we can estimate the tax burden per dollar of the revenue raised for public investment from distortionary taxation. The need for adjustment arises because the welfare loss computed above is measured in net present value terms, whereas the revenue is measured in flow terms. Given the per-period steady-state interest rate of 4.66, the net present value of the additional revenue raised for public investment from distortionary taxation amounts to (4.66/3.66) x .229 = .291% of GDP. Thus, the tax burden per dollar for the additional revenue raised from distortionary domestic taxation comes to .0138/.291 = 4.74 cents.

It is also instructive to look at the response of labor supply to the tariff reform. Figure 3 shows the time path and the new steady-state values of the labor supply for two cases. It also shows the level of the pre-reform labor supply. The labor supply falls on impact and across steady states when there are no constraints on government policy, and the government is able to completely recover the revenue lost due to the removal of tariffs by increasing domestic taxes. On the other hand, it rises when government policy is constrained. It is also worth noting that the labor supply is lower in short term than in the new steady state.

[FIGURE 3 OMITTED]

The difference in the response of labor supply between the two cases is tied to the wealth effects of the tariff reform. In the unconstrained case, there is a positive wealth effect. As leisure is normal good, the agent increases leisure at the expense of labor supply. The leisure falls for the other case as the wealth effect is negative.

There are two conflicting effects operating in the model that render the substitution effect weaker compared to the wealth or the income effect of the reform. For a given policy scenario, a higher rate of labor income taxation is also associated with a higher public investment, and hence, with a higher marginal product of labor, which tends to mitigate the negative effect of increased taxation.

B. Cost of Collecting Domestic Taxes

One of the classic arguments in favor of imposition of tariffs by the developing countries relies on the fact that the cost of collection of domestic taxes is very high for these countries, whereas the collection of tariffs costs very little. The analysis so far has ignored this cost of collection argument in favor of tariffs. This section includes the cost of collection into the model to assess the quantitative significance of the argument.

While in practice, there may be many components of the cost of collection, for simplicity and as in Cremer and Gahvari (1996, 2000) and Reinganum and Wilde (1985), we will interpret this cost as the audit cost. Accordingly, it is posited that the cost of collection (C) of the domestic labor income tax is an increasing function of the revenue government tries to collect ([[tau].sub.i]W) and the audit rate (p), that is,

(29) C([[tau].sub.i]W, p) = c(p)[[tau].sub.i]W.

The assumed linearity of the audit cost in tax revenue is a conservative assumption. It is quite likely that successive increases in labor income tax rate will motivate agents to try ever more harder to evade taxes. This would make the cost of detection a convex function of the tax revenue that government tries to collect, which would make the conclusion of this section stronger. The positive dependence of the audit cost on p follows from the fact that a higher p will increase the proportion of returns that are audited, and as was the case with higher tax rate, will also motivate agents to try harder to avoid being caught evading taxes. In what follows, the exact functional dependence of the audit cost on p will not be needed.

After inclusion of the audit cost, the government's budget constraint becomes

(30) [[bar.R].sub.t] = [G.sub.[t+1]] - (1 - [[delta].sub.G])[G.sub.t] + [J.sub.t] + [C.sub.t].

The government still spends fraction (1 - [zeta]) of its revenues on public investment but transfers adjust with the audit cost.

For the numerical simulation of tariff reform in the model with audit cost, the audit cost function needs to be calibrated. The World Development Report (World Bank, 1988, p. 85) states "The administrative costs of trade and excise taxes normally range from 1 to 3 percent of revenue collected ... for personal income taxes it can reach 10 percent." Assuming tariff collection to be costless without any loss of generality, the audit cost function is calibrated so that the audit cost (for domestic labor income tax) is 6% of the collected revenue. Thus, the difference in the cost of collection of tariffs and domestic income tax is well within the evidence in the World Development Report.

The results of the numerical simulation are shown in Table 6. As expected, the audit cost reduces the welfare gain. For example, for the unconstrained case almost two-thirds of the welfare gain disappears (.128 vs. .339). For [[epsilon].sub.r] = .5 and [[epsilon].sub.p] = .5, the presence of audit cost more than doubles the initial loss from the tariff reform (-.258 vs. -.522). More importantly, the simulations suggest that the cost of collection argument is quantitatively important and can tip the balance against the tariff reform, an argument also made by Munk (2006). Furthermore, it is easy to see that, if both elastic labor supply and audit cost are simultaneously included in the model, the tariff reform would clearly become an unattractive proposition even if the government policy is unconstrained.

So far, the audit probability has been treated as exogenous in the model. One might ask, what if the government also changed p, the audit rate, when it changed the labor income tax rate pursuant to the tariff reform? However, this question begets the following question: what prevents the government from increasing p prior to the implementation of tariff reform? If the reason is political opposition, the original question asked above cannot be answered in the context of this model as the paper takes the political economy considerations as exogenous. Therefore, the government is assumed to not increase p any further in the pre-reform situation because increasing p fails to increase net government revenue. In other words, the pre-reform audit rate is optimal.

The optimality of the pre-reform audit rate allows me to estimate the slope of audit cost function with respect to p at the calibrated value of p. In particular, it implies that one percentage point increase in p increases the audit cost by .175% of GDP. This follows from the fact that, in the pre-reform steady state, one percentage point increase in p raises the government revenue from labor income tax by .175% of GDP. Also, the audit cost function in (Equation 29) implies that the audit cost associated with one percentage point increase in p will increase proportionally with [[tau].sub.i].

Having calibrated the audit cost function, it is now possible to simulate the tariff reform and compare the increase in gross government revenue with the corresponding increase in audit cost in the new steady state. Table 7 shows the results for the different levels of policy constraints when p is increased by one percentage point--the increase in audit cost in parentheses. As the numbers in parentheses are always larger in each cell, for the calibrated model, an increase in p, pursuant to the implementation of tariff reform, results in a net loss of revenue for the government.

C. Some Policy Implications

The results both of the sensitivity analysis and of extending the model lead to a strong presumption that the tariff reforms undertaken over the past few decades in the developing countries might have reduced welfare via the tariff cut [right arrow] revenue loss [right arrow] lower public investment channel.

There are different ways to interpret this result from the policy perspective. One can argue that high administrative costs, pervasive tax evasion, and highly productive public investment are important features of the developing countries, and added together, their adverse welfare effects provide a very potent argument against the IMF and World Bank's advocacy to reduce tariffs.

A more positive vantage point to view the results of the paper is the following. The paper does not argue against the reduction of tariffs, but it argues in the favor of a proper sequencing of economic reforms. The developing countries must undertake reforms to ameliorate the constraints on government policy prior to the liberalization of tariffs. These reforms must empower governments so that they are able to fight tax evasion and neutralize the loss of revenue from future tariff reductions. Furthermore, any halfhearted attempts at domestic tax reforms will not suffice; they will only saddle the economy with distortionary effects of taxation without generating any offsetting benefits by preventing the fall in public investment. In light of this result, the apparent unwillingness of the governments to partially recover the lost tariff revenues may have been a rational response.

In today's world, the countries with high tariff barriers and heavy dependence on tariff revenues are mainly in sub-Saharan Africa. Given their level of economic development, this process of empowerment may take some time (see Munk, 2006). While this need for "carefully sequencing trade liberalization with domestic tax reforms" is slowly being recognized in policy circles (see IMF, 2005), the paper shows that the welfare outcome of the reform critically hinges on it. Accordingly, the future attempts at tariff reforms should therefore be undertaken within a broader program of economic reforms and would require planning and capacity building over a longer time horizon.

The package of reforms would also need to include complementary reforms on expenditure side to curb wasteful public expenditure. However, there are some important differences in the nature of the constraints faced by the developing country governments when choosing to reform their tax system and curbing wasteful public expenditure. While it faces political constraints in both situations, in view of the political clout of the public sector employees in the developing countries, such constraints may be much more stringent on the expenditure side. In contrast, although the political constraints may be less severe, but given the level of economic development of the tariff-dependent countries, the technological constraints are likely to be far more binding and important when it comes to reforming the tax system.

It is well known that while trade reforms may enhance the welfare of a country, they quite often also result in a significant redistribution. It is not uncommon that some groups may gain and others may lose and that the losses to individual groups may be much larger than the overall welfare gain to the country. Therefore, implementation of trade reforms involves important political economy considerations. This paper highlights the fact that a tariff reform may lead to a redistribution across generations: Recall, even in the best-case scenario, the current old gain from the reform while the current young and future generations lose.

The fact that future generations lose is important. They are not represented in the political process whose outcome decides whether such reforms are undertaken. While in the real world, current generations would take the interest of future generations into account to some extent, the intergenerational dimension of redistribution does raise some serious questions. Here is one: Being the representative of the current generations, did the governments of the developing countries fail to adequately resist the pressure from the IMF and the World Bank to reduce tariffs as a significant part of the cost was to be borne by the future generations?

VI. CONCLUSIONS

The literature starting with Emran and Stiglitz (2005) has highlighted the fact that it is plausible for a small open economy to lose from a tariff reform. They show that the value added tax (VAT) and World Bank's advocacy for the replacement of border taxes with a VAT can reduce welfare. Munk (2006) argues that the developing countries may not benefit from such a coordinated tariff-tax reform as the extra administrative costs of domestic taxation may exceed the allocational benefits of freer trade.

This paper shows that the tariff cut [right arrow] revenue loss [right arrow] lower public investment link leads to a strong presumption that the tariff reforms of the past few decades in the developing countries have reduced welfare. It also lends a strong quantitative support to the administrative cost argument of Munk (2006).

There are different ways to interpret this result from policy perspective. One can argue that high administrative costs, pervasive tax evasion, and scarcity of productive public investment provide a very potent argument against the IMF and World Bank's advocacy for reduction of tariffs. A more positive vantage point to view the results of the paper is the following. The paper does not argue against the reduction of tariffs. But it argues that this should be done within a broader program of economic reforms and that, given the level of economic development of tariff-dependent countries, it would require planning and capacity building over a longer time horizon.

The paper only considers public investment in physical infrastructure. There are many other forms of public investment besides physical infrastructure that are not being considered here. Their inclusion will only strengthen paper's results. The stylized nature of the two-period model also, if anything, significantly understates the loss from the tariff reform, as in a multiperiod model the effects of deterioration of public capital will be felt much earlier, and the tax evasion will be higher. Thus, the basic result that modest constraints on government policy can wipe out the gains from trade reforms will only get stronger in models in which agents live for more than two periods and in which there are other forms of public investment. However, future research must also take into account the pro-competitive gains of the freer trade in presence of imperfect competition and its impact on economic growth.

APPENDIX

This appendix provides the details needed for numerically computing the nonlinear solution for the transition dynamics of the baseline model of the paper. This solution is obtained from the equations that define the competitive equilibrium of the economy (see Section II of the paper). These equations come from solving the firm's profit maximization problem and the agent's utility maximization problem, the imposition of the budget constraints for the agents and the government, the specification of the government policy, and finally taking account of certain market clearing and aggregate consistency conditions.

Beginning with firm's profit maximization, the first-order conditions of their problem Equations 15 and 16) imply

(A1) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

(A2) [[w.sub.t]/[Y.sub.t]] = [[w.sub.t]/[y.sub.t]] = (1 - [alpha]).

The constraints of the agent's problem (Equations 2-6) yield

[[E.sub.[y,t]]/[y.sub.t]] = [[x.sub.t] + (1 - [x.sub.t])(1 - [[tau].sub.[i,t]])](1 - [alpha]) + [[j.sub.t]/[y.sub.t]] - [[s.sub.t]/[y.sub.t]], (A3)

(A4) [[E.sub.[0,t+1].sup.1]/[y.sub.t]][[r.sub.[t+1]] + (1 + [gamma][[tau].sub.[e,t+1]])(1 - [[delta].sub.k])][[k.sub.[1,t+1]]/[y.sub.t]],

[[E.sub.[0,t+1].sup.2]/[y.sub.t]][[r.sub.[t+1]] + (1 + [gamma][[tau].sub.[e,t+1]])(1 - [[delta].sub.k])][[k.sub.[2,t+1]]/[y.sub.t]], (A5)

(A6) [[k.sub.[1,t+1]]/[y.sub.t]] = [1/[1 + [gamma][[tau].sub.[e,t+1]]]][[s.sub.t]/[y.sub.t]],

(A7) [[k.sub.[1,t+1]]/[y.sub.t]] = [1/[1 + [gamma][[tau].sub.[e,t+1]]]][[[s.sub.t]/[y.sub.t]] - [[tau].sub.[i,t].sup.p][x.sub.t](1 - [alpha])],

and from the first-order conditions for the agent's utility maximization problem (Equations 17 and 18), we get

(A8) [1/[[([E.sub.[y,t]]/yt).sup.[sigma]]]] = [beta]([[r.sub.[t+1]]/[1 + [gamma][[tau].sub.[e,t+1]]]] + (1 - [[delta].sub.k])) [[[1 - p]/[[([E.sub.[0,t+1].sup.1]/[y.sub.t]).sup.[sigma]]]] + [p/[[([E.sub.[0,t+1].sup.2]/[y.sub.t]).sup.[sigma]]]]], [[[tau].sub.[i,t]]/[[([E.sub.[y,t]]/[y.sub.t]).sup.[sigma]]]] = [beta]([[r.sub.[t+1]]/[1 + [gamma][[tau].sub.[e,t+1]]]] + (1 - [[delta].sub.k]))

(A9) [[P[[tau].sub.[i,t].sup.p]]/[[([E.sub.[0,t+1].sup.2]/[y.sub.t]).sup.[sigma]]]]

The government's revenue (Equation 10) as fraction of GDP is given by

[[[bar.R].sub.t]/[Y.sub.t]] = [[[tau].sub.[i,t]](1 - [x.sub.t]) + p[[tau].sub.[i,t].sup.p][x.sub.t]](1 - [alpha]) + [[tau].sub.[e,t]][gamma] [[K.sub.[t+1]]/[Y.sub.t] - (1 - [[delta].sub.k])[K.sub.t]/[Y.sub.t]] + [[[[tau].sub.[c,t]]/[1 + [[tau].sub.[c,t]]]] [[kappa].sub.t] [[[E.sub.[y,t]]/[Y.sub.t]] + (1 - p)[[E.sub.[0,t].sup.1]/[Y.sub.t]] + p[[E.sub.[0,t].sup.2]/[Y.sub.t]]], (A10)

and government's budget constraint (Equation 11) implies

[[[bar.R].sub.t]/[Y.sub.t]] = [[G.sub.[t+1]]/[Y.sub.t]] - (1 - [[delta].sub.G])[[G.sub.t]/[Y.sub.t]] + [[J.sub.t]/[Y.sub.t]] = [[G.sub.[t+1]]/[Y.sub.t]] - (1 - [[delta].sub.G])[[G.sub.t]/[Y.sub.t]] + [[j.sub.t]/[y.sub.t]]. (A11)

In addition, from Equation (7), the output of the economy is

(A12) [Y.sub.t] = [AG.sub.t.sup.[theta]][K.sub.t.sup.[alpha]],

which can also be used to obtain

[[Y.sub.[t+1]]/[Y.sub.t]] = [([[[[G.sub.[t+1]]/Yt]]/[[[G.sub.t]/[Y.sub.t]]]]).sup.[theta]] [([[[[K.sub.[t+1]]/Yt]]]/[[[K.sub.t]/[Y.sub.t]]]]).sup.[theta]]. (A13)

Finally, note that the aggregate consistency condition for the stock of capital can be written as

[[[K.sub.t] + 1]/[Y.sub.t]] = (1 - p)[[k.sub.[1,t+1]]/[Y.sub.t]] + p[[k.sub.[2,t+1]]/[Y.sub.t]] = (1 - p)[[k.sub.[1,t+1]]/[y.sub.t]] + p[[k.sub.[2,t+1]]/[y.sub.t]] (A14)

These equations can be recursively solved for the transition path of the economy. To show this, we begin by noting that, for t [greater than or equal to] 1, the government policy is given by

[[tau].sub.[c,t]] = .20; [[tau].sub.[e,t]] = .40, [[bar.R]/[Y.sub.t]] = .20, [[J.sub.t]/[[bar.R].sub.t]] = [zeta]; [[tau].sub.[i,t].sup.p] = [chi] [[tau].sub.[i,t]],

and the government varies [[tau].sub.[i,t]] to satisfy its budget constraint. Given the government policy, one can calculate [[kappa].sub.t] as follows:

(A15) [[kappa].sub.t] = [[(1 - [a.sub.1])[(1 + [[tau].sub.[c,t]]).sup.[1-[mu]]]]/[[a.sub.1] + (1 - [a.sub.1])(1 + [[tau].sub.[c,t]])).sup.[1-[mu]]]]].

At the beginning of each period t, [K.sub.t] and [G.sub.t] are known, and (A1-A14) are 14 equations that can be solved for [r.sub.[t+1]], [[w.sub.t]/[y.sub.t]], [[E.sub.[y,t]]/[y.sub.t]], [E.sub.[0,[t+1]].sup.1], [E.sub.[0,[t+1]].sup.2], [x.sub.t] [[s.sub.t]/[y.sub.t]], [[k.sub.[1,[t+1]]/[y.sub.t]], [[k.sub.[2,[t+1]]/[y.sub.t]], [[K.sub.[t+1]]/[Y.sub.t]], [[Y.sub.[t+1]]/[Y.sub.t]], [[G.sub.[t+1]]/[Y.sub.t]], [Y.sub.t], and [[tau].sub.[i,t]]. Thus, one can recursively solve for the transition dynamics of the economy.

REFERENCES

Acharya, S. Aspects of the Black Economy in India. New Delhi, India: Government of India, 1985.

Agenor, P.-R. Infrastructure Investment and Maintenance Expenditure: Optimal Allocation Rules in a Growing Economy. Centre for Growth and Business Cycle Research Discussion Paper Series 60. School of Economic Studies, University of Manchester, Manchester, UK, 2005.

Ai, C., and S. Cassou. "A Normative Analysis of Public Capital." Applied Economics, 27, 1995, 1201-9.

Alam, A., and M. Sundberg. "A Decade of Fiscal Transition." Policy Research Working Paper No. 2835, The World Bank, Washington, DC, 2002.

Allingham, M. G., and A. Sandmo. "Income Tax Evasion: A Theoretical Analysis." Journal of Public Economics, 1, 1972, 323-38.

Alm, J., R. Bahl, and M. N. Murray. "Tax Base Erosion in Developing Countries." Economic Development and Cultural Change, 39, 1991, 849-72.

Amin, G. "Globalization, Consumption Patterns, and Human Development in Egypt." Economic Research Forum Working Paper No. 9929, Cairo: Economic Research Forum, Cairo, Egypt 1999.

Arana, O. M. V. "Economic Growth and the Household Optimal Income Tax Evasion." Discussion Paper No. 275, Department of Economics, Universidad Nacionale de Colombia 2004.

Atolia, M. "Tax Evasion in an Overlapping Generations Model." Manuscript, Florida State University 2008.

Atolia, M., S. Chatterjee, and S. J. Turnovsky. "How Misleading is Linearization? Evaluating the Dynamics of the Neoclassical Growth Model." Manuscript, Florida State University 2009.

Auerbach, A. J., and L. J. Kotlikoff. Dynamic Fiscal Policy. Cambridge: Cambridge University Press 1987.

Barrett, C. B., S. M. Sherlund, and A. A. Adesina."Shadow Wages, Allocative Efficiency, and Labor Supply in Smallholder Agriculture." Agricultural Economics, 38(1), 2007, 21-34.

Barro, R. J. "Government Spending in a Simple Model of Endogenous Growth." Journal of Political Economy, 98(5), Part II 1990, S103-S125.

Bearse, P., G. Glomm, and E. Janeba. "Why Poor Countries Rely Mostly on Redistribution in-Kind." Journal of Public Economics, 75, 2000, 463-81.

Berlinski, J. "International Trade and Commercial Policies of Argentina (An Overview of Selected Evidence and Issues of the 20th Century)." Manuscript, Instituto and Universidad T. Di Tella, 2000.

Buffie, E. F. Trade Policy in Developing Countries. New York: Cambridge University Press 2001.

Canning, D., and M. Fay. "The Effect of Transportation Networks on Economic Growth." Working Paper, Columbia University 1993.

Chen, B. L. "Tax Evasion in a Model of Endogenous Growth." Review of Economic Dynamics, 6, 2003, 381-403.

Corsetti, G., and N. Roubini. "Optimal Government Spending and Taxation in Endogenous Growth Models." NBER Working Paper 5851, National Bureau of Economic Research 1996.

Cremer, H., and F. Gahvari. "Tax Evasion and the Optimum General Income Tax." Journal of Public Economics, 60, 1996, 235-49.

--. "Tax Evasion, Fiscal Competition and Economic Integration." European Economic Review, 44, 2000, 1633-57.

de Soto, H. The Other Path--The Invisible Revolution in the Third World. New York: Harper & Row Publishers 1989.

Dean, J. M., S. Desai, and J. Riedel. "Trade Policy in Developing Countries Since 1985: A Review of Evidence." World Bank Discussion Paper No. 267, The World Bank, Washington, DC 1994.

Dervis, K., J. de Melo, and S. Robinson. General Equilibrium Models for Development Policy. New York: Cambridge University Press 1982.

Diamond, P. A., and J. A. Mirrlees. "Optimal Taxation and Public Production: I-Production Efficiency." American Economic Review, 61(1), 1971, 8-27.

Dropsy, V., and N. Grand. "Exchange Rate and Inflation Targeting in Morocco and Tunisia." Manuscript, Institut de la Mediterran ee, Marseille 2004.

Durlauf, S. N., and P. A. Johnson. "Local Versus Global Convergence Across National Economies." NBER Working Paper No. 3996, National Bureau of Economic Research, 1992.

Easterly, W. "When Is Fiscal Adjustment an Illusion?" Economic Policy, 14(28), 1999, 55-86.

Edwards, S. Crisis and Reform in Latin America--From Despair to Hope. New York: Oxford University Press 1995.

Elias, V. Sources of Growth: A Study of Seven Latin American Economics. San Francisco: ICS Press 1992.

Emran, M. S., and J. E. Stiglitz. "On Selective Indirect Tax Reform in Developing Countries." Journal of Public Economics, 89, 2005, 599-623.

Fay, M. "Financing the Future: Infrastructure Need in Latin America, 2000-05." Working Paper No. 2545, The World Bank, Washington, DC 2001.

Feige, E. L. "How Big is the Irregular Economy?" Challenge, November 1979, 5-13.

Futagami, K., Y. Morita, and A. Shibata. "Dynamic Analysis of an Endogenous Growth Model with Public Capital." Scandinavian Journal of Economics, 95, 1993, 607-25.

Glomm, G., and F. Rioja. "Populist Budgets and Long Run Growth." Manuscript, Indiana University 2003.

Gourio, F., and P.-A. Noual. "The Marginal Worker and The Aggregate Elasticity of Labor Supply." Manuscript, Boston University 2006.

Gupta, R. "Tax Evasion and Financial Repression." Journal of Economics and Business, 60(6), 2007, 517-535.

Hatzipanayotou, P., M. S. Michael, and S. M. Miller. "Win-Win Indirect Tax Reform: A Modest Proposal." Economic Letters, 44, 1994, 147-51.

Hicks, N. L. "Expenditure Reductions in Developing Countries Revisited." Journal of International Development, 3, 1991, 29-37.

International Monetary Fund. "Doha Trade Talks: The Main Risk Is a 'Gutless' Agreement." IMF Survey, 34(9), 2005, 148-49.

Keen, M., and J. E. Ligthart. "Coordinating Tariff Reduction and Domestic Tax Reform." Journal of International Economics, 56, 2002, 489-507.

Kimball, M.S. "Precautionary Saving in the Small and in the Large." Econometrica, 58, 1990, 53-73.

Landskroner, Y., J. Paroush, and I. Swary. "Tax Evasion and Portfolio Decisions." Public Finance, 45, 1990, 132-57.

Lin, S. "The Decline of China's Budgetary Revenue: Reasons and Consequences." Contemporary Economic Policy, 18, 2000, 477-90.

Lynde, C., and J. Richmond. "Public Capital and Total Factor Productivity." International Economic Review, 34, 1993,401-44.

Munk, K. J. "Tax-tariff Reform with Costs of Tax Administration." Department of Economics Working Paper 2006-14, Department of Economics, University of Aarhus 2006.

Ogaki, M., J. Ostry, and C. Reinhart. "Savings Behavior in Low- and Middle-Income Developing Countries." IMF Staff Papers, 43, 1996, 38-71.

Pohl, G., and D. Mihaljek. "Project Evaluation and Uncertainty in Practice: A Statistical Analysis of Rate-of-Return Divergences of 1,015 World Bank Projects." World Bank Economic Review, 6, 1992, 255-77.

Reinganum, J. F., and L. L. Wilde. "Income Tax Compliance in a Principal-Agent Framework." Journal of Public Economics, 26, 1985, 1-18.

Rochjadi, A., and J. H. Leuthold. "The Effect of Taxation on Labor Supply in a Developing Country: Evidence from Cross-Sectional Data." Economic Development and Cultural Change, 42, 1994, 333-50.

Rodrik, D. "Understanding Economic Policy Reform." Journal of Economic Literature, 34, 1996, 9-41.

Roubini, I., and J. Sachs. "Government Spending and Budget Deficits in the Industrial Countries." Economic Policy, 4(8), 1989, 99-132.

Singh, I., L. Squire, and J. Strauss. "A Survey of Agricultural Household Models: Recent Findings and Policy Implications." World Bank Economic Review, 1(1), 1986, 149-79.

Summers, R., and A. Heston. "The Penn World Table (Mark 5): An Expanded Set of International Comparisons, 1950-1988." Quarterly Journal of Economics, 106, 1991, 327-68.

Tanzi, V. "Structural Factors and Tax Revenue in Developing Countries: A Decade of Evidence," in Open Economies: Structural Adjustment and Agriculture, edited by I. Goldin and L. A. Winters. Cambridge: Cambridge University Press 1992, 267-289.

Taylor, L. "Structuralist CGE Models." in Socially Relevant Policy Analysis: Structuralist Computable General Equilibrium Models for the Developing World, edited by L. Taylor. Cambridge, MA: MIT Press 1990, 1-70.

Vernengo, M. "External Liberalization, Stabilization, and the Labor Market in Brazil." Latin American Perspectives, 31(4), 2004, 62-75.

World Bank. "World Development Report." Washington, DC: The World Bank, 1988.

Yitzhaki, S. "Income Tax Evasion: A Theoretical Analysis." Journal of Public Economics, 3, 1974, 201-2.

(1.) For first-best case, see Diamond and Mirrlees (1971), Hatzipanayotou, Michael, and Miller (1994), and Keen and Ligthart (2002).

(2.) Infinite-horizon models with heterogeneous agents and idiosyncratic noninsurable shocks can only be solved in some special cases as the whole distribution of wealth across agents becomes the state variable for the dynamic economy.

(3.) The agents do not have any bequest motive.

(4.) We do not consider capital income tax as it is a small fraction of government revenue in developing countries. Furthermore, in the model, unlike the labor income tax which is nondistortionary, the capital income tax will be distortionary even in absence of tax evasion and increase in capital income tax pursuant to reform will, therefore, reduce the welfare gain (or increase welfare loss) from the reform. However, note that even the labor income tax is distortionary in the extension analyzed in Section V.A.

(5.) The informal sector is an important and large component of developing economies. Although this equivalence has not been made explicit, it is possible to interpret the share of the labor income on which tax evaded as the proportion of labor that the agent supplies in the informal sector. With this reinterpretation, the changes in degree of tax evasion in the model will correspond to the changes in the size of informal economy and the amount of labor employed therein.

(6.) A reader may question the rationale for tax on capital as, even in the second-best world, it is desirable to have production efficiency (see Diamond and Mirrlees, 1971). Since we are concerned with the welfare effects of actual tariff reforms in developing countries, we want to capture the extant practices in these countries. Indeed, these countries levied tariffs on both productive inputs and consumption goods albeit at different rates.

(7.) It may be mentioned that, in this paper in general, the economy-wide variables (aggregate or average) are denoted by capital letters, while those relating to the individual agents are denoted by (corresponding) small letters. However, note an exception: both aggregate and individual consumption expenditures will be denoted by E.

(8.) The aggregate resource constraint can be written in this form as the world prices of all goods are normalized to 1.

(9.) Atolia (2008) analyzes in detail the effect of tax policy on tax evasion, intertemporal resource allocation, and growth.

(10.) The calibration process adds three additional equations to the set of equations listed at the end of Section II. The extended system is then solved for the steady-state values of all endogenous variables and [beta], p, and [zeta]. The three equations that are added set the values of [bar.R]/Y and the return to public and private values to their target values.

(11.) The political nature of these constraints is, however, recognized in policy circles (see IMF, 2005).

(12.) For now, the audit rate (p) is kept unchanged as change in p implies a change in administrative costs, and these costs are not considered until Section V.B.

(13.) Since there is a sustained mismatch between revenue and expenditure, borrowing by government whether in domestic or international market is not feasible and the expenditure has to be curtailed and public investment is first to be axed. In addition, in many developing countries, domestic bond markets are nonexistent or very small.

(14.) For proportionally greater adverse effect of fiscal tightening on public investment in transition economies, see Alam and Sundberg (2002). Amin (1999) provides similar evidence for Egypt, and Dropsy and Grand (2004) do so for Morocco and Tunisia. Also see Easterly (1999) for a thorough discussion of this issue.

(15.) Estimating tax evasion is especially problematic as it is an illegal activity. However, estimates for many countries are as high as 50%; for example, see Acharya (1985) for India, Feige (1979) for Italy, Alm, Bahl, and Murray (1991) for Jamaica.

(16.) In some cases, particularly in Latin America, fall in tariffs has been larger. As larger tariff reduction yields progressively smaller gains from reduced consumption distortion and larger losses from increased tax evasion and reduced public investment, we are considering a conservative scenario.

(17.) For a detailed exposition of this outcome based on Equations (17 and 18), the reader is referred to Atolia (2008).

(18.) The value of the physical capital stock, however, falls as its price in terms of domestic good falls from 1.4 to 1.2.

(19.) Rochjadi and Leuthold (1994) in a study of effects of taxation on labor supply in Indonesia estimate a value of .50 for males and .59 for females. In another extensive study covering seven countries. Singh, Squire, and Strauss (1986) find estimates ranging from .11 to .45 when profits of agricultural households are allowed to vary. Recently, Barrett, Sherlund, and Adesina (2007) have estimated uncompensated wage elasticity of .12. The compensated wage elasticity is typically much higher than the uncompensated elasticity. For example, in Rochjadi and Leuthold (1994), the uncompensated elasticity is estimated to be 0, whereas the estimate of compensated elasticity is more than .5. Thus, the chosen value of [[epsilon].sub.w.sup.c] is empirically reasonable, and if anything, a conservative choice and will tend to understate the losses arising from increased labor income taxation.

(20.) The tax burden is associated not only with public investment but also with transfers as both are financed through distortionary taxes when labor supply is elastic.

(21.) Futagami, Morita, and Shibata (1993), Corsetti and Roubini (1996), and Agenor (2005) analyze the trade-off between increase in productivity due to public investment and the tax burden associated with raising revenue for such investment.

ABBREVIATIONS

IMF: International Monetary Fund

GDP: Gross Domestic Product

MANOJ ATOLIA *

* I am grateful to Ed Buffie and two anonymous referees who provided extensive feedback. Any errors remaining are my own.

Atolia: Department of Economics, Florida State University, Tallahassee, FL 32306. Phone 850-644-7088, Fax 1-850-644-4535, E-mail matolia@fsu.edu

doi: 10.1111/j.1465-7287.2009.00176.x

The literature challenging the relevance of the first-best result that a small open economy gains from a tariff reform has focused on the revenue-neutral case wherein the loss of tariff revenue is fully neutralized by a coordinated increase in domestic taxes (Emran and Stiglitz, 2005; Munk, 2006). (1) While this is a useful benchmark, the experience of developing countries indicates that the trade reforms are rarely revenue neutral. An International Monetary Fund (IMF) staff review of various country experiences found that "nearly half of the low-income countries that cut their tariff rates over the past 20 yr, and suffered an associated revenue loss, recovered less than 70 percent of this lost revenue from other sources" (IMF, 2005; also see Lin, 2000, for evidence from the Chinese tariff reform). This finding is consistent with the stylized fact that trade taxes account for one-third of the total tax income in developing countries (Dean, Desai, and Reidel, 1994; Tanzi, 1992), while the amount of revenue government can collect from other sources is limited due to tax evasion and a large informal sector (Acharya, 1985; Bearse, Glomm, and Janeba, 2000; de Soto, 1989).

Due to the loss in revenue from the tariff reform, the first and major casualty on the expenditure side has been the public investment in physical and social infrastructure (Roubini and Sachs, 1989; World Bank, 1988). Roubini and Sachs note, "in periods of restrictive fiscal policies and fiscal consolidation capital expenditure are the first to be reduced (often drastically)." This is disturbing for the social return on infrastructure investment is typically much higher than the return on private investment as levels of infrastructure in developing countries are suboptimally low (Lin, 2000; Pohl and Mihaljek, 1992). As a result, the welfare losses from reduced infrastructure investment have been of the first order, not just second order small. These losses must be taken into account for correctly assessing the welfare outcome of the tariff reforms of past few decades.

The objectives of this paper were, therefore, the following: a quantitative assessment of the welfare effect of the tariff reforms of past few decades in a model that recognizes the link from the tariff cut [right arrow] revenue loss [right arrow] lower public investment that is missing in the current literature. For this purpose, an overlapping generations model is used in which, consistent with the overwhelming evidence from developing countries, there is tax evasion. There is a growing literature that recognizes the importance of tax evasion for policy analysis for developing countries (Arana, 2004; Chen, 2003; Emran and Stiglitz, 2005; Gupta, 2007). As discussed later, the presence of tax evasion rationalizes the government's inability or unwillingness to generate offsetting revenue from domestic taxes.

Economic theory only tells us that with multiple sources of distortions, as in our case, a tariff reform may lower welfare. However, to go beyond this and to assess the actual welfare outcome of the tariff reforms of past few decades, a quantitative analysis is necessary. When this is done, the results turn out to be much more pessimistic: there is a strong presumption that the effects have been negative as welfare falls in most of the scenarios considered in the paper; compared to a potential welfare gain of .339% of gross domestic product (GDP) for the revenue-neutral reform, the fall in welfare might have been as large as .869% of GDP. As each period in the model is 20-yr long, these gains and losses are large as they are percentages of the GDP for 20 yr.

The paper also suggests why a benevolent government may have been unwilling to recover the lost trade revenues through increased domestic taxation. If the government cannot effectively fight tax evasion, a coordinated domestic tax reform will only partially recover revenue lost due to the tariff reform. Since empirical evidence suggests that when government revenue falls, public investment is not only the major but also the first casualty (Roubini and Sachs, 1989), the partial recovery of lost revenue would fail to stem the fall in public investment and will only saddle the economy with additional distortionary losses due to larger tax evasion. The government would, therefore, avoid domestic tax increases despite a significant decrease in its revenues.

When the model is extended to include other relevant features of developing countries, additional losses arise that are quantitatively significant and therefore further tilt the balance against the desirability of a tariff reform. For example, with the inclusion of elastic labor supply, the potential gain for the revenue-neutral reform falls from .339% to .112% of GDP. The inclusion of the audit cost in the model also reduces the potential gain for the revenue-neutral reform by a similar amount (.339% vs. .128%). Furthermore, the calibrated model with audit cost also shows that it is not possible for the government to increase its revenues by simply raising the audit rate. This happens because the marginal cost of increasing the audit rate is higher than the marginal revenue raked in by the increased audit.

The remaining part of the paper is organized as follows. Section II outlines the model. Section III contains the details of calibration. The welfare analysis of the tariff reform is presented in Section IV. Section V extends the model to include elastic labor supply and audit cost and discusses policy implications of the paper. Section VI concludes.

II. THE MODEL

The paper considers a small open overlapping generations economy that uses labor and capital to produce a homogeneous good, which can be consumed or used as input for the production of capital. The capital is produced by combining the domestic good and an imported input in a fixed proportion. Since these countries usually import "critical" inputs and machines for which there is little scope for substitution from within the country (see Buffie, 2001), the assumption of fixed proportions is a reasonable approximation for the developing countries. The economy also imports a consumer good that is not produced domestically. The country cannot borrow from abroad, and hence, the current account is balanced in each period. Each generation in the economy lives for two periods. The population of each generation is constant and has measure 1. All agents in a generation are born identical. Each agent has measure 0 and is endowed with one unit of labor when young, which he supplies inelastically.

The choice of the model is dictated by following considerations. The assessment of welfare implications in presence of public investment needs a dynamic model as public investment affects the intertemporal trade-off faced by the agents in the economy. In addition, as the economy typically spends a significant time away from the steady state, in models with capital, a comparative static analysis in a static model would fail to capture the welfare changes occurring during the transition to the new steady state. The overlapping generations model has been used for policy analysis in the presence of tax evasion by Arana (2004), Chen (2003), and Gupta (2007), and it simplifies analysis. (2) Finally, as the paper focuses on the effects of the fall in public investment and the distortionary loss arising from tax evasion, in the baseline model, it is assumed that the labor is supplied inelastically. Elastic labor supply, which strengthens the results of the paper, is introduced later in Section V.

A. Preferences and Utility Maximization

The agents are modeled as having a time additive separable (von Neumann-Morgenstern) utility function where utility depends on consumption in each period. (3) Let [~.V](E, P, [~.P]) denote the per-period (indirect) utility function where P is the price of the domestic good, [~.P] is the price of the imported consumer good, and E is the consumption expenditure. [~.V](*) is strictly increasing, strictly concave, and twice continuously differentiable in E. It also satisfies Inada conditions in E. By choosing domestic good as numeraire, we define V(E, [~.P]) [equivalent to] [~.V] (E, 1, [~.P]).

The representative agent of generation t has labor income [w.sub.t] when young, which also equals the wage. When old, that is, in period t + 1, he derives income from the capital accumulated in period t. The government levies a tax at the rate [[tau].sub.[i,t]] on the labor income of period t, which the agent can evade. (4) On receiving his income, the agent decides his saving, [s.sub.t], and the fraction of labor income, [x.sub.t], on which to evade tax. (5) He cannot diversify away the risk of being caught while evading taxes, although at the time of choosing [x.sub.t], he knows the probability of his being caught and the penal tax rate, [[tau].sub.[i,t].sup.p].

The government audits a fraction, p, of the returns. On audit, the underreporting of labor income is detected with probability 1. While p represents both the audit rate and the probability of being caught, in what follows, one or the other interpretation will be highlighted according to what appears more natural in that context. An agent caught evading taxes in period t pays taxes from his saving at a higher penal tax rate, [[tau].sub.[i,t].sup.p] in the same period. After paying penal taxes, if any, the remaining amount is used to accumulate capital. Let the capital accumulated by the agent be [k.sub.[1,t+1]], if not caught, and [k.sub.[2,t+1]] otherwise.

The government levies tariff at rate [[tau].sub.[c,t]] on the quantity of the imported consumer good that is imported in period t. As world prices of all imported goods are normalized to 1, the domestic price of imported consumer good, [~.P], equals 1 + [[tau].sub.[c,t]]. The government also imposes tariff at rate [[tau].sub.[e,t]] on the imported capital input in period t. (6) With fixed proportions technology in capital production, the imported capital input bears a constant ratio to the total capital production, and hence, one can instead assume that a constant fraction, [gamma], of the capital stock is imported. Thus, the domestic price of capital in period t in terms of domestic good becomes 1 + [gamma][[tau].sub.[e,t]].

The representative agent of generation t, therefore, solves the following problem:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (1)

subject to

[E.sub.[y,t]] + [s.sub.t] [less than or equal to] [[x.sub.t] + (1 - [x.sub.t])(1 - [[tau].sub.[i,t]])][w.sub.t] + [j.sub.t], (2)

[E.sub.[0,t+1].sup.1] [less than or equal to] [[r.sub.[t+1]] + (1 + [gamma][[tau].sub.[e,t+1]])(1 - [[delta].sub.k])][k.sub.[1,t+1]], (3)

[E.sub.[0,t+1].sup.2] [less than or equal to] [[r.sub.[t+1]] + (1 + [gamma][[tau].sub.[e,t+1]])(1 - [[delta].sub.k])][k.sub.[2,t+1]], (4)

(5) [k.sub.[1,t+1]] = [[s.sub.t]/[(1 + [gamma][[tau].sub.[e,t]])]],

(6) [k.sub.[2,t+1]] = [[([s.sub.t] - [[tau].sub.[i,t].sup.p][x.sub.t][w.sub.t])]/[(1 + [gamma][[tau].sub.[e,t]])]], 0 [less than or equal to] [x.sub.t] [less than or equal to] 1.

where [beta] is the subjective discount factor; [E.sub.[y,t]] is the consumption expenditure of an agent of generation t when young; [E.sub.[0,t+1].sup.1] is his consumption expenditure when old (i.e., in period t + 1) if he is not caught evading taxes; [E.sub.[0,t+1].sup.2] is his consumption expenditure when old if he is caught evading taxes; [j.sub.t] is the lump-sum transfer from the government; [r.sub.[t+1]] is the capital rental from period t to t + 1; and [[delta].sub.k] [member of] [0, 1] is the rate of depreciation of private capital. As the utility function is strictly increasing in expenditure in each period, all budget constraints hold with equality in equilibrium.

B. Technology and Profit Maximization

On the production side, following Barro (1990), government spending augments the productivity of each firm. The specification follows Futagami, Morita, and Shibata (1993) as the stock of public capital (G), and not public spending, affects productivity. The firms are identical and have Cobb-Douglas production function, exhibiting constant returns to scale in private capital, K, and labor, L. Thus, one can assume that there is a single firm in the economy and its output is given by

(7) F(K, L; G) = [AG.sup.[theta]][K.sup.[alpha]][L.sup.[1-[alpha]] [equivalent to] Y,

where [alpha] > 0, [theta] > 0 and [alpha] + [theta] < 1, and G, K, L, and Y are economy-wide aggregates. Note that although there are external increasing returns to scale at the aggregate level, the production function is characterized by decreasing returns in the accumulable factors, and there is no long-run growth in the economy. (7) The output per person of the generation supplying labor is

(8) y = f(k; G) = [AG.sup.[theta]][(K/L).sup.[alpha]] = [AG.sup.[theta]][k.sup.[alpha]].

The firm's problem is straightforward. It chooses capital and labor to maximize profit in each period,

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (9)

C. The Government Budget Constraint

To finance public investment, the government has access to three sources of revenue. Let [X.sub.t] be the (average) fraction of income not reported by the agents of generation t and [W.sub.t] be their aggregate wage income in period t. Then, the government revenue from labor income tax is [[[tau].sub.[i,t]] (1 - [X.sub.t]) + p[[tau].sub.[i,t].sup.p] [W.sub.t]. Its revenues from the tariff on imported capital equals [[tau].sub.[e,t]][gamma][[K.sub.[t+1]] - (1 - [[delta].sub.k])[K.sub.t]].Let [[kappa].sub.t] [member of] [0, 1] be the expenditure share of the imported consumption good, then the amount spent on consumer imports is [[kappa].sub.t][[E.sub.[y,t]] + (1 - p)[E.sub.[0,t].sup.1] + p[E.sub.[0,t].sup.2], of which a fraction [[[tau].sub.[c,t]]/(1 + [[tau].sub.[c,t]])] is tariff revenue. Thus, the government revenue is

(10) [[bar.R].sub.t] = [[[tau].sub.[i,t]](1 - [X.sub.t]) + p[[tau].sub.[i,t].sup.p][X.sub.t]][W.sub.t] + [[tau].sub.[e,t]][gamma][[K.sub.[t+1]] - (1 - [[delta].sub.k])[K.sub.t]] + [[[tau].sub.[c,t]]/[1 + [[tau].sub.[c,t]]]][[kapp].sub.t][[E.sub.[y,t]] + (1 - p)[E.sub.[0,t].sup.1] + p[E.sub.[0,t].sup.2]].

The government revenue in excess of public investment is rebated to the current young in a lump-sum manner. Hence, the government's budget constraint for period t is simply

(11) [[bar.R].sub.t] = [G.sub.[t+1]] - (1 - [[delta].sub.G])[G.sub.t] + [J.sub.t],

where [[delta].sub.G] [member of] [0, 1] is the rate of depreciation of the public capital; [J.sub.t] is the transfer made to generation t in period t; [G.sub.[t+1]] is the stock of public capital that enters the production function of the firms in period t + 1.

D. The Competitive Equilibrium

The competitive equilibrium for this economy is defined as:

Definition. A competitive equilibrium for this economy is a sequence {[E.sub.[y,t]], [E.sub.[0,t+1].sup.1], [E.sub.[0,t+1].sup.2], [x.sub.t], [s.sub.t], [k.sub.[1,t]], [k.sub.[2,t]], [r.sub.t], [w.sub.t], [K.sub.t], [L.sub.t], [X.sub.t], [S.sub.t], [W.sub.t], [G.sub.t], [j.sub.t], [J.sub.t], [[tau].sub.[i,t]], [[tau].sub.[i,t].sup.p], [[tau].sub.[e,t]]} such that for every t

(1) given {[r.sub.[t+1]], [w.sub.t], [j.sub.t], [[tau].sub.[i,t]], [[tau].sub.[i,t].sup.p], [[tau].sub.[e,t]), {[x.sub.t], [s.sub.t], [E.sub.[y,t]], [E.sub.[0,t+1].sup.1], [E.sub.[0,t+1].sup.2]} solves the optimization problem (Equation 1) for generation t;

(2) given {[r.sub.t], [w.sub.t], [G.sub.t]}, {[K.sub.t], [L.sub.t]} maximizes profit of the firm as in Equation (9);

(3) given {[x.sub.t], [s.sub.t], [E.sub.[y,t]], [E.sub.[0,t].sup.1], [E.sub.[0,t].sup.2], [w.sub.t], [K.sub.t], [G.sub.t]}, government policy {[G.sub.[t+1]], [J.sub.t], [[tau].sub.[i,t]], [[tau].sub.[i,t].sup.p], [[tau].sub.[e,t]]} satisfies government's budget constraint (Equation 11);

(4) aggregate and individual level variables are consistent; and

(5) markets for capital, labor, and output clear (8)

[K.sub.[t+1]] = (1 - p)[k.sub.[1,t+1]] + p[k.sub.[2,t+1] = [S.sub.t] - p[[tau].sub.[i,t].sup.p][X.sub.t][W.sub.t], (12)

(13) [L.sub.t] = 1,

[Y.sub.t] = (1 - [[[[tau].sub.[c,t]][[kappa].sub.t]]/[1 + [tau]c, t]])([E.sub.[y,t]] + [E.sub.[0,t-1].sup.1] + [E.sub.[0,t-1].sup.2]) + ([K.sub.[t+1]] - (1 - [[delta].sub.k])[K.sub.t]) + ([G.sub.[t+1]] - (1 - [[delta].sub.G])[G.sub.t]). (14)

E. Solving for the Competitive Equilibrium

The firm's profit maximization yields following first-order conditions:

(15) [K.sub.t]: [r.sub.t] = [F.sub.k] ([K.sub.t], [L.sub.t]; [G.sub.t]),

(16) [L.sub.t]: [w.sub.t] = [F.sub.L]([K.sub.t], [L.sub.t]; [G.sub.t]).

The first-order conditions for interior solution for maximization of agent's utility are

[s.sub.t]: [V.sub.E]([E.sub.[y,t]], 1 + [[tau].sub.[c,t]] = [beta][[[[r.sub.[t+1]] + (1 + [gamma][[tau].sub.[e,[t+1]])(1 - [[delta].sub.k])]]/[1 + [gamma][[tau].sub.[e,t]]]] x [(1 - p)[V.sub.E]([E.sub.[0,t].sup.1], 1 + [[tau].sub.[c,t+1]]) + p[V.sub.E]([E.sub.[0,t].sup.2], 1 + [[tau].sub.[c,t+1]]) (17)

[x.sub.t]: [V.sub.E]([E.sub.[y,t]], 1 + [[tau].sub.[c,t]]) = [beta][p[[tau].sub.[i,t].sup.p]/[[tau].sub.[i,t]]] [[[[r.sub.[t+1]] + (1 + [gamma][[tau].sub.[e,t+1])(1 - [[delta].sub.k])]]/[1 + [gamma][[tau].sub.[e,t]]]] [V.sub.E]([E.sub.[0,t].sup.2], 1 + [[tau].sub.[c,t+1]]). (18)

The first-order condition for [s.sub.t] is standard. In the first-order condition for [x.sub.t], the left-hand side is the marginal benefit of evading taxes on labor income and right-hand side is the marginal cost. It can be seen from Equation (18) that an increase in [[tau].sub.[i,t].sup.p] increases the cost of tax evasion and hence reduces [x.sub.t]. It can also be shown that as this decrease in tax evasion reduces the need for precautionary saving, [s.sub.t] falls as well. (9)

The government policy specifies the tax rates and the fraction ([[zeta].sub.t]) of government revenue that is rebated to the agents as transfers so that

(19) [G.sub.[t+1]] - (1 - [[delta].sub.G])[G.sub.t] = (1 - [[zeta].sub.t])[[bar.R].sub.t].

Solving for the competitive equilibrium involves solving Equations (2-8), (10-13), and (15-19), details of which are relegated to the Appendix. In particular, for computing the steady state, these 16 equations can be used to find the steady-state values of s, x, [E.sub.y], [E.sub.0.sup.1], [E.sub.0.sup.2], y, [k.sub.1], [k.sub.2], r, w, Y, K, L, [bar.R], G, and J, given the government policy defined by {[[tau].sub.i], [[tau].sub.i.sup.p], [[tau].sub.e], [[tau].sub.c], [zeta]}

III. CALIBRATION OF THE MODEL

To quantify the welfare outcome of a tariff reform, it is necessary to turn to numerical simulation, which requires choosing the functional forms, and values for the parameters. It may be emphasized that while the numerical simulations require choosing particular values of every parameter in the model, there are only a few whose values affect the outcome one is usually interested in. We do sensitivity analysis for such parameters where data are lacking or show a wide variation.

The utility function is chosen to be constant-elasticity-of-substitution-and-constant-relative-risk-aversion (CES-CRRA) with indirect utility function given by

(20) V(E, [P.sub.c]) = [1/1 - [sigma]][[[E/[P.sub.c]]].sup.[1-[sigma]]],

where [P.sub.c] [equivalent to] [[a.sub.1] + (1 - [a.sub.1])[(1 + [[tau].sub.c]).sup.[1-[micro]]].sup.[1/[1-[micro]] is the exact consumption-based price index and [a.sub.1] > 0 is a preference parameter. It allows us to choose the values for intertemporal (1/[sigma]) and intratemporal ([mu]) elasticities of substitution in consumption that are in accordance with the empirical facts.

In this setup, [sigma] plays a dual role: as the coefficient of relative risk aversion and as the inverse of the elasticity of intertemporal substitution. Using estimates for low- and middle-income countries in Ogaki, Ostry, and Reinhart (1996), [sigma] is set at 2, which implies elasticity of intertemporal substitution of .5. The value of [sigma] = 2 is also consistent with empirical evidence if it is interpreted as the coefficient of relative risk aversion.

In highly aggregated demand systems with 5-11 goods, the estimated compensated own-price elasticities lie in the range .15-.6. To account for the fact that the model has only two goods, and hence, the scope of substitution is much less, [mu] is set at .3 (this yields compensated own-price elasticity of imported good of .255) and sensitivity analysis is done for [mu] = .4.

Durlauf and Johnson (1992) study convergence across national economies and find that the share of physical capital in income or output ([alpha]) varies between .3 and .4. The poor countries have a capital share of income of .3, whereas it is .4 for the countries with intermediate income. For the developed countries, they find this share to be .33. For Latin American economies, Elias (1992) estimates a value of .5. Consistent with Elias as well as to account for the fact that capital implicitly also includes the intermediate inputs in the model, [alpha] is given a value of .5 (Table 1). The (annual) return to private capital (or the real interest rate) of 8% for developing countries is taken from Buffie (2001). In steady state, the return to private capital equals the capital rental, r. The share of the imported capital in private capital varies considerably across the developing countries. The parameter [gamma] is set at .5, which is in the middle of the range of estimates (.35-.65) in Buffie (2001) that are consistent with Taylor's (1990) illustrative Social Accounting Matrix (SAM) and Dervis, de Melo, and Robinson (1982). The depreciation of private and public capital at 7% and 4% per year are typical. The difference reflects the facts that private capital implicitly also includes imported intermediates in the model and public capital consists primarily of physical infrastructure, which typically depreciates slowly compared to plant and machinery. With each period in the model lasting 20 yr, this yields [[delta].sub.k] = .766 and [[delta].sub.G] = .558.

TABLE 1

Parameter Values for the Calibrated Model

Preferences

[beta] = .1898; [sigma] = 2, [micro]=.3, [kappa] = .15

Production function

[alpha] = .5; [theta] = .2; [[delta].sub.k] = .766; [[delta].sub.G] = -558; [gamma] = .5

Government policy

[[tau].sub.i] = .3: [[tau].sub.i.sup.p] = .6; [[tau].sub.e] = .4; [[tau].sub.c] = .8

[chi] = 2; [zeta] = .9394

Other

p = .1163

Pohl and Mihaljek (1992) show that the public capital is highly productive in developing countries. Analyzing the rate of return on 1,015 World Bank projects implemented in developing countries, they find the median and the average (annual) rates of return to be 14% and 16%, respectively. While Pohl and Mihaljek (1992) suggest that the projects submitted by governments for the World Bank financing may primarily include projects with above-average rates of return, Easterly (1999) summarizes evidence showing that the return to public investment in developing countries (especially in physical infrastructure) may actually be even higher (19%-29%). We choose a very modest value of 12% for the return on public capital and do sensitivity analysis for 16%. There are very different estimates of the elasticity of national output with respect to public capital ([theta]) varying from close to 0 to .2 (see Ai and Cassou, 1995; Lynde and Richmond, 1993). In our simulations, [theta] is set at .2, which is the estimate obtained by Canning and Fay (1993) and Fay (2001) using large cross-country data sets. Once the return on public investment is chosen, the choice of [theta] merely determines the share of public investment in public expenditure (1 - [zeta]), and [theta] = .2 gives a more realistic estimate for latter. The results of the paper, however, only depend on the higher return to public capital compared to private capital.

Developing countries have an escalated structure of protection with higher tariffs on consumer goods and lower tariffs on imported capital and intermediates. For example, as Vernengo (2004) reports, the average tariff on capital and intermediates in Brazil over 1960-1980 was 50% of that on the consumption goods. Berlinski (2000) provides similar evidence for Argentina. Accordingly, and following Edwards (1995), [[tau].sub.e] and [[tau].sub.c] are set to .4 and .8, respectively, which are typical pre-reform values for the developing countries. The tax rate on labor income ([[tau].sub.i]) of .3 is set so as to yield the share of tariffs in government revenue that is consistent with the estimates in Tanzi (1992) and Dean, Desai, and Riedel (1994).

The second consumption good in the model is entirely imported, although in reality a large portion of its demand is met by domestic production of this good or its very close substitutes. The share of the imported component of this good in consumption is about 10%, whereas the share rises to 15%-25% when consumption of the portion that is produced domestically as well as the domestically produced close substitutes is included (see Buffie, 2001). Each of these numbers is a valid choice for the share of the imported good in consumption from two different perspectives. From the perspective of matching the contribution of tariffs to government revenue, the consumption share of the imported consumer good in the model should be set to 10%. However, the tariff on imported consumer good also raises the price of its close domestic substitute(s). Hence, to capture the distortionary effect of the tariffs, the consumption share of the imported consumer good in the model should be set slightly higher. To balance these conflicting considerations, the results are presented for two cases. The baseline case sets the consumption share of imported good (kappa) at .15 in the initial steady state. Then, sensitivity analysis is done for a lower value of .1. It should be noted that the value of [kappa] in the baseline case corresponds to a scenario with larger gains from tariff reduction.

The values of [beta], x, p, and [zeta] are estimated from the model to match the data on the ratio of government revenue to GDP ([bar.R]/Y), the penal tax rate ([[tau].sub.i.sup.p]), and the returns to the public and the private capital. The ratio of the government revenue to GDP has been ascertained from Summers and Heston (1991) (Mark 5.6a), which reveals considerable variation across countries. It ranges from 10% to 30% for the middle 90% of the countries, and the average is lower for the developed countries than the developing countries. The model is calibrated for [bar.R]/Y = .20. We set [[tau].sub.i.sup.p] = 2[[tau].sub.i] = .6, which besides being empirically reasonable also yields plausible estimates of p in the range .1-.2. This implies that less than 1% of returns are audited every year, which accords with audit rate in India. (10)

A. Political Constraints on Government Policy

The paper does not model the political process that determines the ability of the government to fight tax evasion. The penal tax rate may be already very high, and a higher rate may infeasible due to the widespread nature of tax evasion. The paper also does not model the process by which government decides the extent to which to offset the loss of revenue arising from tariff reform. There may be resistance to raising the domestic tax rates; in cases where statutory tax rate can be raised, the penal tax rate may be already very high. In the face of an inertial penal tax rate, increasing the statutory tax rate may only increase evasion and not enable the government to restore [bar.R]/Y to its initial level. These constraints on the government's policy choices are labeled as "political constraints" for want of a better term. (11)

While the constraints on government policy are exogenous, it is nonetheless possible to analyze their impact on the welfare consequences of a tariff reform. To this end, define [[epsilon].sub.r] to be the fraction of lost tariff revenues that is offset by a coordinated increase in domestic taxes. Thus, [[epsilon].sub.r] is a quantitative measure of the severity of the constraints faced by the government. A value of [[epsilon].sub.r] < 1 implies that the government can only partially offset the loss of revenue from the tariff reform--the tariff reform is not revenue neutral; in particular, [[epsilon].sub.r] = 0 implies completely passive domestic tax policy with no changes in the domestic tax rates. The constraints on the ability to make up lost tariff revenues have been significant; recall the findings of the IMF staff review that "nearly half of the low-income countries that cut their tariff rates ... recovered less than 70 percent of this lost revenue from other sources" (IMF, 2005). In case of China, Lin (2000) provides a similar evidence where the share of tariffs in government revenue declined from over 10% in 1985 to 3.4% in 1998.

To recover fraction [[epsilon].sub.r] of its lost revenue, the government has two potential instruments, [[tau].sub.i] and [[tau].sub.i.sup.p], at its disposal. (12) Hence, a rule that links the penal tax rate to the statutory tax rate is needed to uniquely determine the government policy. A general linear penal tax rate rule has the form

(21) [[tau].sub.i.sup.p] = [[bar.[tau]].sub.i.sup.p] + [chi][[tau].sub.i],

where [[bar.[tau]].sub.i.sup.p] > 0 and [chi] > 0 are parameters. Landskroner, Paroush, and Swary (1990) study tax evasion as a portfolio choice under such a rule where [[bar.[tau]].sub.i.sup.p] has the interpretation of a penalty on the evaded income and [chi] of the penalty on the evaded tax. A proportional penal tax rate rule (i.e., a rule with [[bar.[tau].sub.i.sup.p] = 0) minimizes tax evasion, and as shown by Yitzhaki (1974), it also eliminates the substitution effect as defined in Allingham and Sandmo (1972). Thus, government policy is considered constrained if the penal tax rate increases less than proportionally with the statutory tax rate (i.e., [[bar.[tau].sub.i.sup.p] > 0). To quantify this constraint on government policy, define [[epsilon].sub.p] as the elasticity of the penal tax rate with respect to the statutory tax rate. A constraint on public policy in this dimension implies that [[epsilon].sub.p] is less than 1. In an interesting observation, Yitzhaki (1974) notes that United States and Israel were the only countries in 1974 that had a proportional penal tax rate system.

If the government budget shrinks, the brunt of the resource crunch is borne by public investment. (13) As mentioned before, Roubini and Sachs (1989) note, "in periods of restrictive fiscal policies ... capital expenditures are the first to be reduced (often drastically)." This agrees with the findings in the World Development Report (World Bank, 1988) that in the face of fiscal tightening, the public investment fell far more sharply (35%) than other current expenditures such as wages (10%). Hicks (1991) comes up with corresponding estimates of 27.8% and 7.2%. (14) These findings are not hard to understand. The effects of reduction in public investment become visible only when the gradual deterioration of public roads and overcrowding of existing infrastructure impacts productivity. A reduction of transfers and the public sector wage bill has more immediate consequences for politicians. Political expediency results in a disproportionate reduction in public investment.

Rodrik (1996) analyzes of the role of incentives of policy makers in economy policy reforms. For our purpose of quantifying the welfare effect of reduced public investment in wake of tariff reforms, it is, however, not necessary to formally model these incentives. What is critical is to be able to capture two empirically relevant outcomes: (1) public investment is the first and major casualty in the face of declining government revenue and (2) the decline in public investment is three to four times greater (as in Hicks, 1991 and the World Bank, 1988) when the government is able to recover only part of its lost tariff revenues (as in IMF, 2005).

The first fact suggests that the relationship between the fall in public investment and government revenue is monotonic but highly nonlinear. Such nonlinearity can be captured parsimoniously by making the ratio of the post-reform public investment to the pre-reform public investment ([lambda]), an exponential function of the reduction in revenue ([[epsilon].sub.r]). A polynomial relationship, while a plausible alternative, will involve more parameters; and the parameters will have less intuitive interpretation. As what is relevant for our results is the actual empirical relationship to which the function is calibrated and not its analytical form, the following rule is used to link [lambda] and [[epsilon].sub.r]:

[lambda] [equivalent to] 1 - [[1 - [upsilon]]/[1 - exp[-[psi]]]][1 - exp[-[psi](1 - [[epsilon].sub.r])]], (22)

where v[member of] [0, 1] and [psi] > 0 are parameters.

This specification implies that public investment decreases by fraction 1 - v when the government is constrained to keep domestic rates of taxation unchanged at pre-reform levels. In addition, public investment (as a fraction of government revenue) is unaffected if the government can fully neutralize the loss of revenue from tariff reform by a coordinated increase in domestic taxes which occurs if [[epsilon].sub.r] = 1. Furthermore, a higher value of [psi] implies that the brunt of the initial fiscal crunch falls on public investment with a greater intensity. This can be seen from Figure 1, which represents this relationship for v = .7 and two values of [psi], 3 and 5. In either case, public investment falls by 30% if the government is constrained to keep the domestic rates of taxation unchanged at the pre-reform levels, but for the higher value of [psi] = 5, there is a proportionally larger reduction in public investment (the lower curve in Figure 1) when the fall in government revenue is smaller, that is, [[epsilon].sub.r] is higher.

[FIGURE 1 OMITTED]

To match the second fact mentioned above, we set [upsilon] = .7 and [psi] = 3 so that, for the 50% tariff reduction considered later, when the government is able to recover only 75% of its lost tariff revenues ([[epsilon].sub.r] = .75), public investment falls by 3-4 times more (16.7% vs. 5%).

B. The Steady State

The steady state for the calibrated model is presented in Table 2 and is representative of a typical developing country. The share of tariffs (T) in government revenue, T/[bar.R], is .3991. This value is somewhat high, but well within the range of the estimates reported in Tanzi (1992) and Dean, Desai, and Riedel (1994). The larger share results from the higher consumption share of the imported consumer good that is assumed so that the welfare gain arising from the reduction in consumption distortion as a result of the tariff reform can be accurately captured.

TABLE 2

Steady State of the Calibrated Model

Tax evasion

x = .2591

Consumption-output ratios

Ey/y = .4548; [[E.sub.0.sup.1]/Y] = .5686; [E.sub.0.sup.2] = .2063

Saving- and capital-output ratios

s/y = .1220; (1 + [gamma][[tau].sub.e])K/Y = 0.1130; (1 + [gamma][[tau].sub.e])[[k.sub.1]/y] = 0.1220; (1 + [gamma][[tau].sub.e])[[k.sub.2]/y] = 0.4425

Annualized private capital to output ratio = 2.256 Government

[[bar.R]/Y] = .2; J/Y = .1879; G/Y = .0217; G/K = .1924; T/[bar.R] = .3991

Other

r = 8%, return to public capital = 12%

The annualized capital output ratio at 2.256 is also within the range of empirical values presented in Buffie (2001). Agents evade tax on 25.91% of their income, which is at the lower end of the range of estimates in the literature. (15) In the data, tax evasion is higher because, with the annual filing of tax return, the risk arising from being caught at evading taxes is much smaller than in the model where tax is assessed only once during entire working life. More specifically, for the same degree of risk aversion, the risk is much higher in the two-period model as being caught at evasion results in the agent consuming significantly less during entire period of retirement.

The value of [zeta] at .9394 is high as it implies only 6.1 % of the government revenue is invested in public capital. For Latin America, Glomm and Rioja (2003) estimate about 15% of government revenue is spent on infrastructure. However, the results of the paper depend only on the fact that public capital is much more productive than private capital on the margin. The stock of public capital does not matter as the loss depends on the inefficiency generated by the initial divergence between public and private capital productivity and the proportion by which public investment falls. Irrespective of the level of the stock of public capital, the tariff reform reduces public investment in the same proportion.

One reason that the calibrated share of public investment in government revenue is smaller than what is reported in official government data is the widespread corruption in the governments in the developing countries. The calibrated share is based on the return to public capital in World Bank-funded projects. Since the corruption in World Bank projects is presumably smaller, this rate of return on public capital is closer to the actual return on public capital. When this return is used to compute the economy-wide stock of public capital and investment, only the fraction of public investment funds that actually gets invested is accounted for. In contrast, in the official government accounts, public funds that "leak" due to corruption or inefficiency also appear as public investment. In the model, they appropriately appears as transfers because there is no production of goods and services associated with them. That they appear as transfers to the current young is also reasonable as current young represent the working generation.

In view of rampant corruption in developing countries, it is not surprising that the fraction of public funds that actually gets invested may be no larger than one half (see Bearse, Glomm, and Janeba, 2000 and references cited therein). In fact, this reconciles the apparently contradictory evidence that although public capital is found to be more productive than private capital, the productivity in public sector is much lower than the private sector in developing countries (Bearse, Glomm, and Janeba, 2000) when calculations are based on official government accounts.

IV. THE WELFARE EFFECTS OF A TARIFF REFORM

Consider a 50% reduction in tariff on both the imported capital and the imported consumer good. This approximately corresponds to the change in tariff levels that occurred in many developing countries during 1980s and 1990s. (16) The welfare effect of this reduction is assessed by numerically solving the nonlinear model. The details of numerical computations can be found in the Appendix.

To quantify the welfare gains or losses arising from the tariff reform, the additional consumption needed or enjoyed by each generation when young is discounted to the time of the reform using the domestic interest rate. To this, the surplus consumption enjoyed by the current old is added. Adding the net present values of the gains or losses accruing to various generations gives a measure of the net welfare effect of the tariff reform. In the overlapping generations framework, government can potentially achieve Pareto improvement through intergenerational transfers (see Auerbach and Kotlikoff, 1987). To abstract from these effects and to focus on the effects of tax evasion and fall in public investment, all revenue in excess of the public investment continues to be rebated in lump sum to the current young.

A. Numerical Simulation of the Reform

Table 3 presents the welfare outcome of the tariff reform for the varying degree of constraints on government policy. Each entry is the welfare gain as percent of the current period GDP.

TABLE 3 Welfare Gains and Losses under Alternative Government Policies [[epsilon].sub.r] [[epsilon].sub.p] 0 .25 .5 .75 1 1 -.171 -.133 -.055 .091 .339 .75 -.173 -.175 -.133 -.017 .206 .5 -.176 -.231 -.241 -.175 .006 .25 -.180 -.312 -.414 -.451 -.376

In absence of any constraints on government policy, the government sets both [[epsilon].sub.r] and [[epsilon].sub.p] at 1. In this case, the reform is revenue neutral. As can be seen, the overall welfare gain from tariff reform amount to .339% of the current period GDP as distortions due to tariffs fall. This is a large welfare gain as each period in the model corresponds to 20 yr. The gains are, however, not evenly distributed. The current old gain as the tariff reform shifts tax burden away from them by reducing the tax on imported consumer good. Current young and future generations, however, lose from the reform. The reform raises labor income tax rate from .3 to .369. This lowers the income of the current young making them worse off despite the fall in the price of the imported consumer good. Their loss amounts to .817% of the current period GDP.

Future generations lose as capital stock in the economy falls reducing their consumption possibilities. Across steady states, capital stock falls by 6.31% as the young respond to declining income by saving less as fraction of total GDP and by reducing tax evasion. The fall in tax evasion also reduces precautionary saving (see Kimball, 1990). (17) The reduction in saving reduces capital accumulation as the saving turns into capital except for the amount paid as penalties for tax evasion. The capital stock declines despite the reduction in tariff on the imported capital. For example, in the period of reform, with a proportional increase in penal tax rate from .6 to .738, x falls from .259 to .188 along with a fall in the saving in the period of reform by 10.02%. As a result, the capital stock declines by 1.84% in the next period.

B. Constrained Policy and Welfare Loss

The gains that occur in an unconstrained scenario, however, vanish for very reasonable constraints on government policy along either dimension--ability to raise revenue or ability to fight tax evasion. First, consider the case of revenue-nonneutral reforms in which the government is unable to fully offset its revenue loss from the tariff reform ([[epsilon].sub.r] < 1), although it can effectively fight tax evasion ([[epsilon].sub.p] = 1). For such reforms, the associated reduction in public investment causes a welfare loss. If the government can increase [[tau].sub.i] to offset only 50% of the shortfall in revenues ([[epsilon].sub.r] = .5), then the loss outweighs the usual gains from tariff reduction. The resulting loss in an overall welfare is .055% of the current period GDP. This scenario is not unrealistic; recall, many countries have been able recover less than 70% of lost tariff revenues through coordinated domestic tax increases.

If the government is able to recover its revenue loss ([[epsilon].sub.r] = 1) by increasing domestic taxes, but cannot effectively fight tax evasion ([[epsilon].sub.p] < 1), tax evasion increases and the resulting distortion reduces the usual gain from the tariff reform. As Table 3 also shows, this gain is almost wiped out when [[epsilon].sub.p] is .5, which is a very plausible value in light of observation in Yitzhaki (1974) cited earlier. In this case, capital accumulation is higher in physical terms but is accompanied by a larger distortion from tax evasion, and in numerical simulations, future generations still lose from the reform. (18) Thus, given the current system of taxation, a welfare gain is not ensured even for revenue-neutral reforms.

The gains from tariff reform disappear even with very modest constraints on public policy. When the government offsets 75% of its revenue loss by increasing the statutory tax rate and raising the penal tax rate with elasticity .75 ([[epsilon].sub.r] = .75, [[epsilon].sub.p] = .75), there is a welfare loss of .017% of GDP. The constraints are so mild that the government revenue decreases by less than a percentage point from 20% of GDP to 19.13% and less than one-quarter (23.3%) of this reduction in the government budget falls on public investment. On the tax evasion front, these constraints imply that the government increases [[tau].sub.i.sup.p] from 60% to 68.69%, which is only marginally lower than the unconstrained value of 71.8% for the corresponding increase in [[tau].sub.i] from 30% to 35.9%. The situation gets far worse with greater, yet plausible, constraints on government policy in the developing countries. The reform now yields a loss of a magnitude similar to the gain that results in the absence of constraints on government policy (see Table 3).

The source of loss from revenue-neutral reforms is very similar to those in the static models of tax evasion where there is an intratemporal trade-off between the distortionary loss due to tax evasion and the benefits of revenue collected from taxation. For the revenue-nonneutral reforms, the trade-off in the model is, however, inherently inter-termporal, as by reducing the stock of public capital, the reform affects the (intertemporal) saving decision of the agents. The reduced availability of public capital reduces the marginal product of private capital, which reduces the saving in the economy. Thus, the stock of private capital falls as well, further reducing the productive capacity of the economy.

How does the stylized two-period structure of the model affect the results? This, if anything, significantly understates the losses arising from reduced public investment as the stock of public capital in the period of the reform (which has a duration of 20 yr) is predetermined. In reality, the public capital will start deteriorating much earlier. With agents living for multiple periods, the tax evasion will also be higher because, as mentioned earlier, the risk associated with tax evasion will be much lower. In multiperiod models, it would also be possible to set the measure of retirees relative to the working generations to a realistic value of less than 1. The gains of retirees, therefore, will have a lower weight in computation of the overall gain from the reform. Thus, the basic result that modest constraints on government policy can wipe out the gains from tariff reforms is likely to be robust to the extension of model to include agents living more than two periods.

It is also clear from Table 3 that for a given [[epsilon].sub.r], the government will set [[epsilon].sub.p], as high as it can as it leads to highest welfare. In other words, the government would fight tax evasion, as effectively as it can. Similarly, if government can effectively fight tax evasion ([[epsilon].sub.p] = 1), it would offset the fall in its revenues to the extent possible. However, note from Table 3 that when the governmen=t cannot raise penal tax rate proportionally with the statutory tax rate ([[epsilon].sub.p] < 1), the welfare loss is not monotonic in [[epsilon].sub.r]. Hence, in presence of both a limited ability to fight tax evasion ([[epsilon].sub.p] < 1) and a limited flexibility in coordinated domestic tax increases ([[epsilon].sub.r] < 1), avoiding an increase in domestic taxation would be better, despite a significant decrease in government revenue. This interaction of the political constraints suggests that, in view of the losses arising from tax evasion, the government may be "unwilling" to raise domestic taxes even when it is "able" to do so.

The nonmonotonicity of losses in [[epsilon].sub.r], when government fails to fight tax evasion effectively ([[epsilon].sub.p] < 1), arises from the change in the relative magnitude of two conflicting effects. The increase in [[epsilon].sub.r] mitigates the fall in revenue, which helps contain the adverse impact of reduced public investment. On the other hand, it also leads to a loss from tax evasion as [[epsilon].sub.p] < 1. How these effects vary with [[epsilon].sub.r] is determined by different factors. For the positive public investment effect, it depends on what fraction of additional revenue is invested in public capital, while for the negative tax evasion effect, it is determined by the elasticity of the penal tax rate with respect to the statutory tax rate, that is, [[epsilon].sub.p].

Figure 2 illustrates how these effects vary with [[epsilon].sub.r]. For a given [[epsilon].sub.p], as [[epsilon].sub.r] rises, the public investment effect becomes progressively stronger (solid curve in Figure 2) because a larger fraction of additional revenue collected goes to fund public investment. Note that this line asymptotes to x-axis as [[epsilon].sub.r] approaches 0: the positive effect rapidly vanishes with fall in [[epsilon].sub.r] as the increase in government revenue causes hardly any increase in public investment (see Figure 1). The loss from tax evasion effect also rises with [[epsilon].sub.r] as government's attempt to raise more revenue leads to larger tax evasion. But, as [[epsilon].sub.p] is unchanged, this effect does not rise as rapidly with [[epsilon].sub.r] as the public investment effect as suggested by the dotted lines (linearity is just for convenience of exposition) in Figure 2.

[FIGURE 2 OMITTED]

From Figure 2, it evident that given [[epsilon].sub.p] < 1, there is a threshold value of [[epsilon].sub.r], [[bar.[epsilon]].sub.r], such that, for [[epsilon].sub.r] < [[bar.[epsilon]].sub.r] the welfare outcome is worse than with [[epsilon].sub.r] = 0. Moreover, [[epsilon].sub.r] rises when [[epsilon].sub.p] falls: the adverse effect of tax evasion is greater when [[epsilon].sub.p] is smaller, and therefore, [[epsilon].sub.r] must rise more so that the positive effect of increase public investment is able to overcome the higher welfare loss arising from increased tax evasion for lower [[epsilon].sub.p]. When [[epsilon].sub.p] = .75, for example, it is not worthwhile to undertake domestic tax reform if the government can recover only 25% of its lost revenues. For [[epsilon].sub.p] = .5, increasing domestic taxes is not useful even if 50% of lost revenue can be recovered. Intuitively, when the government cannot effectively fight tax evasion, a coordinated increase in domestic taxation that only partially offsets its revenue loss just saddles the economy with distortionary losses due to tax evasion. It does not deliver a countervailing benefit by preventing the fall in public investment.

C. Sensitivity Analysis

Table 4 contains the results of sensitivity analysis for the parameters that affect the welfare outcome of the reform; and they are as expected. A higher return on public capital magnifies the loss arising from the reduction in government revenue. For [[epsilon].sub.r] = .75 and [[epsilon].sub.p] = 1, the reform delivers a welfare gain of .091% (Table 3) when the return on public capital is 12%. This turns into a loss of .190% with the increase in return on public capital to 16% (panel 1, Table 4); recall, this is the average rate of return on public capital found by Pohl and Mihaljek (1992). In the worst case, welfare loss is .869%. It may be noted that even a return of 16% on public investment is much lower than the estimates reported in Easterly (1999).

TABLE 4 Sensitivity Analysis for Welfare Gains and Losses under Alternative Government Policies Return on public investment =16% [[epsilon].sub.r] [[epsilon].sub.p] 0 .25 .5 .75 1 1 -.741 -.660 -.498 -.190 .339 .75 -.745 -.704 -.578 -.300 .208 .5 -.749 -.763 -.690 -.460 .011 .25 -.754 -.849 -.869 -.741 -.368 [kappa] = .l [[epsilon].sub.r] [[epsilon].sub.p] 0 .25 .5 .75 1 1 -.257 -.220 -.148 -.008 .230 .75 -.258 -.248 -.199 -.081 .140 .5 -.259 -.284 -.268 -.182 .012 .25 -.261 -.333 -.370 -.339 -.199 [micro] = .4 [[epsilon].sub.r] [[epsilon].sub.p] 0 .25 .5 .75 1 1 -.028 .010 .080 .234 .481 .75 -.030 -.031 .011 .127 .351 .5 -.033 -.089 -.095 -.027 .155 .25 -.037 -.165 -.263 -.295 -.215

When the consumption share of the imported consumer good is smaller, the initial distortion from tariffs is also smaller, and so is the gain from the tariff reform. In the absence of any constraints of government policy, welfare gain reduces from .339% to .230% of GDP when [kappa] falls from .15 to .1 (Table 3 and panel 2, Table 4). Qualitatively, however, the welfare outcome is similar: mild constraints on government policy still cause a welfare loss.

As expected, a higher elasticity of substitution in consumption raises the gain from the reform due to the increased possibility of substitution towards the now cheaper imported consumer good, while the loss from the constraints on the government policy remains unchanged. The welfare outcome of the tariff reform, therefore, becomes more favorable, although reasonable constraints of government policy still cause gains from the reform to disappear (panel 3, Table 4).

V. SOME EXTENSIONS AND POLICY IMPLICATIONS

The result so far convincingly shows that the gains from tariff reform disappear for modest constraints on public policy, not only for the benchmark parameter values but also for the plausible alternative combinations of parameter values. In the same vein, we now investigate how the welfare calculus of the reform is affected if some additional features of the developing economies are included in the model. In particular, this section examines the implications of elastic labor supply and nontrivial cost of collection of domestic taxes. In the latter case, it also estimates the marginal cost of increasing the audit rate, p (which is also the probability of being caught in the model), and investigates how an increase in p in the presence of audit cost would affect the net revenue of the government. The section ends with some remarks on the policy implications.

A. Elastic Labor Supply

To allow for the labor supply to endogenously respond to policy changes, the utility function of the agents is augmented to include a term ([upsilon](*)) capturing the disutility from supplying labor ([l.sub.t]), which is equivalent to a preference for leisure.

The representative agent of generation t now solves the following problem:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (23)

subject to

[E.sub.[y,t]] + [s.sub.t] [less than or equal to] [[x.sub.t] + (1 + [x.sub.t])(1 - [[tau].sub.[i,t]])][w.sub.t][l.sub.t] + [j.sub.t], (24)

and Equations (3-6). The only change in the constraints for optimization is that now the labor income depends on the amount of labor supplied. The optimal choice of labor supplied depends on the trade-off between the disutility from labor and the utility from consumption made possible by the income earned by supplying labor. This trade-off is captured by the following first-order condition for [l.sub.t]:

[upsilon]'([l.sub.t]) = [V.sub.E]([E.sub.[y,t]], 1 + [[tau].sub.[c,t]])[[x.sub.t] + (1 - [x.sub.t]) (1 - [[tau].sub.[i,t]])][w.sub.t]. (25)

For numerical simulations, the disutility from labor is assumed to be given by

[upsilon](l) = -a[(b - l).sup.[xi]], (26)

which on substitution in Equation (25), along with use of Equation (20), gives

a[xi][(b - [l.sub.t]).sup.[[xi]-1]] = [([[E.sub.[y,t]]/[1 + [[tau].sub.[c,t]]]]).sup.-[sigma]]

[[x.sub.t] + (1 + [x.sub.t])(1 - [[tau].sub.[i,t]])] - [[w.sub.t]/[1 + [[tau].sub.[c,t]]]]. (27)

Equation (27) equates the marginal disutility from working (on left) to the marginal utility gain from additional income (on right). It is clear that, ceteris paribus, an increase in tax evasion raises labor supply as it increases the effective wage rate.

Before simulations, however, we need to calibrate the extended model for three new parameters, a, b, and [xi]. Three targets are used to find the values of these parameters. First target is the amount of labor is supplied by the agent in the initial steady state. This is set to 1 unit as in the baseline model. By ensuring that the initial equilibrium is same for both the extended model and the baseline model, this allows a comparison of the results of the two models. The second target is the fraction of the time agent spends working and is set at 30%, as is commonly assumed in the models with labor-leisure choice (e.g., see Atolia, Chatterjee, and Turnvosky, 2009). Last target sets the compensated wage elasticity of labor supply ([[xi].sub.w.sup.c]) at .27, which is well within the range of empirical estimates. (19)

The calibration yields the following values of the parameters: a = -.0044; b = 10/3, and [xi] = -11/3. For these parameter values, the Frisch wage elasticity of labor supply is .5, which also on the lower side of the range of values that Gourio and Noual (2006) experiment with. It may be mentioned that for the extended model, the Frisch elasticity of labor supply, [[epsilon].sub.w], with respect to (real) wage is

(28) [[epsilon].sub.w] = [b - l/l][1/1 - [xi]]

The welfare results for the tariff reform when labor supply is elastic are collected in Table 5. A comparison of the results in Table 5 and Table 3 shows that the inclusion of elastic labor supply strengthens the results of the paper; the welfare gain from tariff reform is uniformly lower with elastic labor supply. For example, the welfare gain from tariff reform in the best-case scenario with no constraints on government policy falls from .339% to .112% of GDP. For modest constraints with [[epsilon].sub.r] = .75 and [[epsilon].sub.p] = .75, the welfare loss jumps from .017% to .203% of GDP.

TABLE 5 Elastic Labor Supply and Welfare Gains and Losses under Alternative Government Policies [[epsilon].sub.r]] [[epsilon].sub.p]] 0 .25 .5 .75 1 1 -.286 -.297 -.259 -.133 .112 .75 -.288 -.328 -.313 -.203 .033 .5 -.290 -.369 -.390 -.309 -.091 .25 -.294 -.431 -.516 .502 -.344

The reason is not very hard to understand. In the baseline model, labor supply is perfectly inelastic, and hence, the tariff reform constitutes a move from a distortionary to a nondistortionary source of raising government revenue. Whereas with elastic labor supply, the tariff reform is a move from one distortionary source to another. As a result, the losses from tariff reform are higher with elastic labor supply.

In other words, there is no tax burden associated with public investment after the implementation of tariff reform when the labor supply is inelastic. However, when labor supply is elastic, the provision of public investment via labor income taxation gives rise to a tax burden, which reduces the gain from the tariff reform. This tax burden can be measured as the difference between the welfare effects of tariff reform with inelastic and elastic labor supply, which allows us to answer an open question in public finance and expenditure literature: what is the tax burden associated with public investment when it is financed by distortionary taxation? (20), (21)

To provide an answer to this question, consider the unconstrained case where the level of public investment after the tariff reform is same as before the reform. In this case, with elastic labor supply, there is an additional welfare loss of .339 - .112 = .227% of GDP. Of this loss, 6.06%, that is, .0138% of GDP, is due to the increased financing of public investment by distortionary labor income tax. This follows from the fact the fraction of government revenue that is used for public investment is 1 - [zeta] = .0606. This loss is associated with the

increase in financing of public investment from distortionary labor income tax by .229% of GDP.

It is necessary to make one adjustment before we can estimate the tax burden per dollar of the revenue raised for public investment from distortionary taxation. The need for adjustment arises because the welfare loss computed above is measured in net present value terms, whereas the revenue is measured in flow terms. Given the per-period steady-state interest rate of 4.66, the net present value of the additional revenue raised for public investment from distortionary taxation amounts to (4.66/3.66) x .229 = .291% of GDP. Thus, the tax burden per dollar for the additional revenue raised from distortionary domestic taxation comes to .0138/.291 = 4.74 cents.

It is also instructive to look at the response of labor supply to the tariff reform. Figure 3 shows the time path and the new steady-state values of the labor supply for two cases. It also shows the level of the pre-reform labor supply. The labor supply falls on impact and across steady states when there are no constraints on government policy, and the government is able to completely recover the revenue lost due to the removal of tariffs by increasing domestic taxes. On the other hand, it rises when government policy is constrained. It is also worth noting that the labor supply is lower in short term than in the new steady state.

[FIGURE 3 OMITTED]

The difference in the response of labor supply between the two cases is tied to the wealth effects of the tariff reform. In the unconstrained case, there is a positive wealth effect. As leisure is normal good, the agent increases leisure at the expense of labor supply. The leisure falls for the other case as the wealth effect is negative.

There are two conflicting effects operating in the model that render the substitution effect weaker compared to the wealth or the income effect of the reform. For a given policy scenario, a higher rate of labor income taxation is also associated with a higher public investment, and hence, with a higher marginal product of labor, which tends to mitigate the negative effect of increased taxation.

B. Cost of Collecting Domestic Taxes

One of the classic arguments in favor of imposition of tariffs by the developing countries relies on the fact that the cost of collection of domestic taxes is very high for these countries, whereas the collection of tariffs costs very little. The analysis so far has ignored this cost of collection argument in favor of tariffs. This section includes the cost of collection into the model to assess the quantitative significance of the argument.

While in practice, there may be many components of the cost of collection, for simplicity and as in Cremer and Gahvari (1996, 2000) and Reinganum and Wilde (1985), we will interpret this cost as the audit cost. Accordingly, it is posited that the cost of collection (C) of the domestic labor income tax is an increasing function of the revenue government tries to collect ([[tau].sub.i]W) and the audit rate (p), that is,

(29) C([[tau].sub.i]W, p) = c(p)[[tau].sub.i]W.

The assumed linearity of the audit cost in tax revenue is a conservative assumption. It is quite likely that successive increases in labor income tax rate will motivate agents to try ever more harder to evade taxes. This would make the cost of detection a convex function of the tax revenue that government tries to collect, which would make the conclusion of this section stronger. The positive dependence of the audit cost on p follows from the fact that a higher p will increase the proportion of returns that are audited, and as was the case with higher tax rate, will also motivate agents to try harder to avoid being caught evading taxes. In what follows, the exact functional dependence of the audit cost on p will not be needed.

After inclusion of the audit cost, the government's budget constraint becomes

(30) [[bar.R].sub.t] = [G.sub.[t+1]] - (1 - [[delta].sub.G])[G.sub.t] + [J.sub.t] + [C.sub.t].

The government still spends fraction (1 - [zeta]) of its revenues on public investment but transfers adjust with the audit cost.

For the numerical simulation of tariff reform in the model with audit cost, the audit cost function needs to be calibrated. The World Development Report (World Bank, 1988, p. 85) states "The administrative costs of trade and excise taxes normally range from 1 to 3 percent of revenue collected ... for personal income taxes it can reach 10 percent." Assuming tariff collection to be costless without any loss of generality, the audit cost function is calibrated so that the audit cost (for domestic labor income tax) is 6% of the collected revenue. Thus, the difference in the cost of collection of tariffs and domestic income tax is well within the evidence in the World Development Report.

The results of the numerical simulation are shown in Table 6. As expected, the audit cost reduces the welfare gain. For example, for the unconstrained case almost two-thirds of the welfare gain disappears (.128 vs. .339). For [[epsilon].sub.r] = .5 and [[epsilon].sub.p] = .5, the presence of audit cost more than doubles the initial loss from the tariff reform (-.258 vs. -.522). More importantly, the simulations suggest that the cost of collection argument is quantitatively important and can tip the balance against the tariff reform, an argument also made by Munk (2006). Furthermore, it is easy to see that, if both elastic labor supply and audit cost are simultaneously included in the model, the tariff reform would clearly become an unattractive proposition even if the government policy is unconstrained.

TABLE 6 Audit Cost and Welfare Gains and Losses under Alternative Government Policies [[epsilon].sub.r] [[epsilon].sub.p] 0 .25 .5 .75 1 1 -.374 -.368 -.306 -.148 .128 .75 -.377 -.418 -.396 -.273 -.023 .5 -.382 -.483 -.522 -.454 -.250 .25 -.387 -.577 -.720 -.766 -.673

So far, the audit probability has been treated as exogenous in the model. One might ask, what if the government also changed p, the audit rate, when it changed the labor income tax rate pursuant to the tariff reform? However, this question begets the following question: what prevents the government from increasing p prior to the implementation of tariff reform? If the reason is political opposition, the original question asked above cannot be answered in the context of this model as the paper takes the political economy considerations as exogenous. Therefore, the government is assumed to not increase p any further in the pre-reform situation because increasing p fails to increase net government revenue. In other words, the pre-reform audit rate is optimal.

The optimality of the pre-reform audit rate allows me to estimate the slope of audit cost function with respect to p at the calibrated value of p. In particular, it implies that one percentage point increase in p increases the audit cost by .175% of GDP. This follows from the fact that, in the pre-reform steady state, one percentage point increase in p raises the government revenue from labor income tax by .175% of GDP. Also, the audit cost function in (Equation 29) implies that the audit cost associated with one percentage point increase in p will increase proportionally with [[tau].sub.i].

Having calibrated the audit cost function, it is now possible to simulate the tariff reform and compare the increase in gross government revenue with the corresponding increase in audit cost in the new steady state. Table 7 shows the results for the different levels of policy constraints when p is increased by one percentage point--the increase in audit cost in parentheses. As the numbers in parentheses are always larger in each cell, for the calibrated model, an increase in p, pursuant to the implementation of tariff reform, results in a net loss of revenue for the government.

TABLE 7 Increase in Government Revenue and Audit Cost (in parentheses) for One Percentage Point Increase in the Audit Rate in the New Steady State [[epsilon].sub.r] [[epsilon].sub.p] 0 .25 .5 1 .163(.175) .162(.185) .162(.195) .75 .163(.175) .165(.186) .168(.197) .5 .163(.175) .170(.187) .176(.200) .25 .163(.175) .176(.190) .189(.205) [[epsilon].sub.r] [[epsilon].sub.p] .75 1 1 .161(.205) .159(.214) .75 .170(.208) .170(.218) .5 .182(.212) .186(.225) .25 .202(.221) .214(.237)

C. Some Policy Implications

The results both of the sensitivity analysis and of extending the model lead to a strong presumption that the tariff reforms undertaken over the past few decades in the developing countries might have reduced welfare via the tariff cut [right arrow] revenue loss [right arrow] lower public investment channel.

There are different ways to interpret this result from the policy perspective. One can argue that high administrative costs, pervasive tax evasion, and highly productive public investment are important features of the developing countries, and added together, their adverse welfare effects provide a very potent argument against the IMF and World Bank's advocacy to reduce tariffs.

A more positive vantage point to view the results of the paper is the following. The paper does not argue against the reduction of tariffs, but it argues in the favor of a proper sequencing of economic reforms. The developing countries must undertake reforms to ameliorate the constraints on government policy prior to the liberalization of tariffs. These reforms must empower governments so that they are able to fight tax evasion and neutralize the loss of revenue from future tariff reductions. Furthermore, any halfhearted attempts at domestic tax reforms will not suffice; they will only saddle the economy with distortionary effects of taxation without generating any offsetting benefits by preventing the fall in public investment. In light of this result, the apparent unwillingness of the governments to partially recover the lost tariff revenues may have been a rational response.

In today's world, the countries with high tariff barriers and heavy dependence on tariff revenues are mainly in sub-Saharan Africa. Given their level of economic development, this process of empowerment may take some time (see Munk, 2006). While this need for "carefully sequencing trade liberalization with domestic tax reforms" is slowly being recognized in policy circles (see IMF, 2005), the paper shows that the welfare outcome of the reform critically hinges on it. Accordingly, the future attempts at tariff reforms should therefore be undertaken within a broader program of economic reforms and would require planning and capacity building over a longer time horizon.

The package of reforms would also need to include complementary reforms on expenditure side to curb wasteful public expenditure. However, there are some important differences in the nature of the constraints faced by the developing country governments when choosing to reform their tax system and curbing wasteful public expenditure. While it faces political constraints in both situations, in view of the political clout of the public sector employees in the developing countries, such constraints may be much more stringent on the expenditure side. In contrast, although the political constraints may be less severe, but given the level of economic development of the tariff-dependent countries, the technological constraints are likely to be far more binding and important when it comes to reforming the tax system.

It is well known that while trade reforms may enhance the welfare of a country, they quite often also result in a significant redistribution. It is not uncommon that some groups may gain and others may lose and that the losses to individual groups may be much larger than the overall welfare gain to the country. Therefore, implementation of trade reforms involves important political economy considerations. This paper highlights the fact that a tariff reform may lead to a redistribution across generations: Recall, even in the best-case scenario, the current old gain from the reform while the current young and future generations lose.

The fact that future generations lose is important. They are not represented in the political process whose outcome decides whether such reforms are undertaken. While in the real world, current generations would take the interest of future generations into account to some extent, the intergenerational dimension of redistribution does raise some serious questions. Here is one: Being the representative of the current generations, did the governments of the developing countries fail to adequately resist the pressure from the IMF and the World Bank to reduce tariffs as a significant part of the cost was to be borne by the future generations?

VI. CONCLUSIONS

The literature starting with Emran and Stiglitz (2005) has highlighted the fact that it is plausible for a small open economy to lose from a tariff reform. They show that the value added tax (VAT) and World Bank's advocacy for the replacement of border taxes with a VAT can reduce welfare. Munk (2006) argues that the developing countries may not benefit from such a coordinated tariff-tax reform as the extra administrative costs of domestic taxation may exceed the allocational benefits of freer trade.

This paper shows that the tariff cut [right arrow] revenue loss [right arrow] lower public investment link leads to a strong presumption that the tariff reforms of the past few decades in the developing countries have reduced welfare. It also lends a strong quantitative support to the administrative cost argument of Munk (2006).

There are different ways to interpret this result from policy perspective. One can argue that high administrative costs, pervasive tax evasion, and scarcity of productive public investment provide a very potent argument against the IMF and World Bank's advocacy for reduction of tariffs. A more positive vantage point to view the results of the paper is the following. The paper does not argue against the reduction of tariffs. But it argues that this should be done within a broader program of economic reforms and that, given the level of economic development of tariff-dependent countries, it would require planning and capacity building over a longer time horizon.

The paper only considers public investment in physical infrastructure. There are many other forms of public investment besides physical infrastructure that are not being considered here. Their inclusion will only strengthen paper's results. The stylized nature of the two-period model also, if anything, significantly understates the loss from the tariff reform, as in a multiperiod model the effects of deterioration of public capital will be felt much earlier, and the tax evasion will be higher. Thus, the basic result that modest constraints on government policy can wipe out the gains from trade reforms will only get stronger in models in which agents live for more than two periods and in which there are other forms of public investment. However, future research must also take into account the pro-competitive gains of the freer trade in presence of imperfect competition and its impact on economic growth.

APPENDIX

This appendix provides the details needed for numerically computing the nonlinear solution for the transition dynamics of the baseline model of the paper. This solution is obtained from the equations that define the competitive equilibrium of the economy (see Section II of the paper). These equations come from solving the firm's profit maximization problem and the agent's utility maximization problem, the imposition of the budget constraints for the agents and the government, the specification of the government policy, and finally taking account of certain market clearing and aggregate consistency conditions.

Beginning with firm's profit maximization, the first-order conditions of their problem Equations 15 and 16) imply

(A1) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

(A2) [[w.sub.t]/[Y.sub.t]] = [[w.sub.t]/[y.sub.t]] = (1 - [alpha]).

The constraints of the agent's problem (Equations 2-6) yield

[[E.sub.[y,t]]/[y.sub.t]] = [[x.sub.t] + (1 - [x.sub.t])(1 - [[tau].sub.[i,t]])](1 - [alpha]) + [[j.sub.t]/[y.sub.t]] - [[s.sub.t]/[y.sub.t]], (A3)

(A4) [[E.sub.[0,t+1].sup.1]/[y.sub.t]][[r.sub.[t+1]] + (1 + [gamma][[tau].sub.[e,t+1]])(1 - [[delta].sub.k])][[k.sub.[1,t+1]]/[y.sub.t]],

[[E.sub.[0,t+1].sup.2]/[y.sub.t]][[r.sub.[t+1]] + (1 + [gamma][[tau].sub.[e,t+1]])(1 - [[delta].sub.k])][[k.sub.[2,t+1]]/[y.sub.t]], (A5)

(A6) [[k.sub.[1,t+1]]/[y.sub.t]] = [1/[1 + [gamma][[tau].sub.[e,t+1]]]][[s.sub.t]/[y.sub.t]],

(A7) [[k.sub.[1,t+1]]/[y.sub.t]] = [1/[1 + [gamma][[tau].sub.[e,t+1]]]][[[s.sub.t]/[y.sub.t]] - [[tau].sub.[i,t].sup.p][x.sub.t](1 - [alpha])],

and from the first-order conditions for the agent's utility maximization problem (Equations 17 and 18), we get

(A8) [1/[[([E.sub.[y,t]]/yt).sup.[sigma]]]] = [beta]([[r.sub.[t+1]]/[1 + [gamma][[tau].sub.[e,t+1]]]] + (1 - [[delta].sub.k])) [[[1 - p]/[[([E.sub.[0,t+1].sup.1]/[y.sub.t]).sup.[sigma]]]] + [p/[[([E.sub.[0,t+1].sup.2]/[y.sub.t]).sup.[sigma]]]]], [[[tau].sub.[i,t]]/[[([E.sub.[y,t]]/[y.sub.t]).sup.[sigma]]]] = [beta]([[r.sub.[t+1]]/[1 + [gamma][[tau].sub.[e,t+1]]]] + (1 - [[delta].sub.k]))

(A9) [[P[[tau].sub.[i,t].sup.p]]/[[([E.sub.[0,t+1].sup.2]/[y.sub.t]).sup.[sigma]]]]

The government's revenue (Equation 10) as fraction of GDP is given by

[[[bar.R].sub.t]/[Y.sub.t]] = [[[tau].sub.[i,t]](1 - [x.sub.t]) + p[[tau].sub.[i,t].sup.p][x.sub.t]](1 - [alpha]) + [[tau].sub.[e,t]][gamma] [[K.sub.[t+1]]/[Y.sub.t] - (1 - [[delta].sub.k])[K.sub.t]/[Y.sub.t]] + [[[[tau].sub.[c,t]]/[1 + [[tau].sub.[c,t]]]] [[kappa].sub.t] [[[E.sub.[y,t]]/[Y.sub.t]] + (1 - p)[[E.sub.[0,t].sup.1]/[Y.sub.t]] + p[[E.sub.[0,t].sup.2]/[Y.sub.t]]], (A10)

and government's budget constraint (Equation 11) implies

[[[bar.R].sub.t]/[Y.sub.t]] = [[G.sub.[t+1]]/[Y.sub.t]] - (1 - [[delta].sub.G])[[G.sub.t]/[Y.sub.t]] + [[J.sub.t]/[Y.sub.t]] = [[G.sub.[t+1]]/[Y.sub.t]] - (1 - [[delta].sub.G])[[G.sub.t]/[Y.sub.t]] + [[j.sub.t]/[y.sub.t]]. (A11)

In addition, from Equation (7), the output of the economy is

(A12) [Y.sub.t] = [AG.sub.t.sup.[theta]][K.sub.t.sup.[alpha]],

which can also be used to obtain

[[Y.sub.[t+1]]/[Y.sub.t]] = [([[[[G.sub.[t+1]]/Yt]]/[[[G.sub.t]/[Y.sub.t]]]]).sup.[theta]] [([[[[K.sub.[t+1]]/Yt]]]/[[[K.sub.t]/[Y.sub.t]]]]).sup.[theta]]. (A13)

Finally, note that the aggregate consistency condition for the stock of capital can be written as

[[[K.sub.t] + 1]/[Y.sub.t]] = (1 - p)[[k.sub.[1,t+1]]/[Y.sub.t]] + p[[k.sub.[2,t+1]]/[Y.sub.t]] = (1 - p)[[k.sub.[1,t+1]]/[y.sub.t]] + p[[k.sub.[2,t+1]]/[y.sub.t]] (A14)

These equations can be recursively solved for the transition path of the economy. To show this, we begin by noting that, for t [greater than or equal to] 1, the government policy is given by

[[tau].sub.[c,t]] = .20; [[tau].sub.[e,t]] = .40, [[bar.R]/[Y.sub.t]] = .20, [[J.sub.t]/[[bar.R].sub.t]] = [zeta]; [[tau].sub.[i,t].sup.p] = [chi] [[tau].sub.[i,t]],

and the government varies [[tau].sub.[i,t]] to satisfy its budget constraint. Given the government policy, one can calculate [[kappa].sub.t] as follows:

(A15) [[kappa].sub.t] = [[(1 - [a.sub.1])[(1 + [[tau].sub.[c,t]]).sup.[1-[mu]]]]/[[a.sub.1] + (1 - [a.sub.1])(1 + [[tau].sub.[c,t]])).sup.[1-[mu]]]]].

At the beginning of each period t, [K.sub.t] and [G.sub.t] are known, and (A1-A14) are 14 equations that can be solved for [r.sub.[t+1]], [[w.sub.t]/[y.sub.t]], [[E.sub.[y,t]]/[y.sub.t]], [E.sub.[0,[t+1]].sup.1], [E.sub.[0,[t+1]].sup.2], [x.sub.t] [[s.sub.t]/[y.sub.t]], [[k.sub.[1,[t+1]]/[y.sub.t]], [[k.sub.[2,[t+1]]/[y.sub.t]], [[K.sub.[t+1]]/[Y.sub.t]], [[Y.sub.[t+1]]/[Y.sub.t]], [[G.sub.[t+1]]/[Y.sub.t]], [Y.sub.t], and [[tau].sub.[i,t]]. Thus, one can recursively solve for the transition dynamics of the economy.

REFERENCES

Acharya, S. Aspects of the Black Economy in India. New Delhi, India: Government of India, 1985.

Agenor, P.-R. Infrastructure Investment and Maintenance Expenditure: Optimal Allocation Rules in a Growing Economy. Centre for Growth and Business Cycle Research Discussion Paper Series 60. School of Economic Studies, University of Manchester, Manchester, UK, 2005.

Ai, C., and S. Cassou. "A Normative Analysis of Public Capital." Applied Economics, 27, 1995, 1201-9.

Alam, A., and M. Sundberg. "A Decade of Fiscal Transition." Policy Research Working Paper No. 2835, The World Bank, Washington, DC, 2002.

Allingham, M. G., and A. Sandmo. "Income Tax Evasion: A Theoretical Analysis." Journal of Public Economics, 1, 1972, 323-38.

Alm, J., R. Bahl, and M. N. Murray. "Tax Base Erosion in Developing Countries." Economic Development and Cultural Change, 39, 1991, 849-72.

Amin, G. "Globalization, Consumption Patterns, and Human Development in Egypt." Economic Research Forum Working Paper No. 9929, Cairo: Economic Research Forum, Cairo, Egypt 1999.

Arana, O. M. V. "Economic Growth and the Household Optimal Income Tax Evasion." Discussion Paper No. 275, Department of Economics, Universidad Nacionale de Colombia 2004.

Atolia, M. "Tax Evasion in an Overlapping Generations Model." Manuscript, Florida State University 2008.

Atolia, M., S. Chatterjee, and S. J. Turnovsky. "How Misleading is Linearization? Evaluating the Dynamics of the Neoclassical Growth Model." Manuscript, Florida State University 2009.

Auerbach, A. J., and L. J. Kotlikoff. Dynamic Fiscal Policy. Cambridge: Cambridge University Press 1987.

Barrett, C. B., S. M. Sherlund, and A. A. Adesina."Shadow Wages, Allocative Efficiency, and Labor Supply in Smallholder Agriculture." Agricultural Economics, 38(1), 2007, 21-34.

Barro, R. J. "Government Spending in a Simple Model of Endogenous Growth." Journal of Political Economy, 98(5), Part II 1990, S103-S125.

Bearse, P., G. Glomm, and E. Janeba. "Why Poor Countries Rely Mostly on Redistribution in-Kind." Journal of Public Economics, 75, 2000, 463-81.

Berlinski, J. "International Trade and Commercial Policies of Argentina (An Overview of Selected Evidence and Issues of the 20th Century)." Manuscript, Instituto and Universidad T. Di Tella, 2000.

Buffie, E. F. Trade Policy in Developing Countries. New York: Cambridge University Press 2001.

Canning, D., and M. Fay. "The Effect of Transportation Networks on Economic Growth." Working Paper, Columbia University 1993.

Chen, B. L. "Tax Evasion in a Model of Endogenous Growth." Review of Economic Dynamics, 6, 2003, 381-403.

Corsetti, G., and N. Roubini. "Optimal Government Spending and Taxation in Endogenous Growth Models." NBER Working Paper 5851, National Bureau of Economic Research 1996.

Cremer, H., and F. Gahvari. "Tax Evasion and the Optimum General Income Tax." Journal of Public Economics, 60, 1996, 235-49.

--. "Tax Evasion, Fiscal Competition and Economic Integration." European Economic Review, 44, 2000, 1633-57.

de Soto, H. The Other Path--The Invisible Revolution in the Third World. New York: Harper & Row Publishers 1989.

Dean, J. M., S. Desai, and J. Riedel. "Trade Policy in Developing Countries Since 1985: A Review of Evidence." World Bank Discussion Paper No. 267, The World Bank, Washington, DC 1994.

Dervis, K., J. de Melo, and S. Robinson. General Equilibrium Models for Development Policy. New York: Cambridge University Press 1982.

Diamond, P. A., and J. A. Mirrlees. "Optimal Taxation and Public Production: I-Production Efficiency." American Economic Review, 61(1), 1971, 8-27.

Dropsy, V., and N. Grand. "Exchange Rate and Inflation Targeting in Morocco and Tunisia." Manuscript, Institut de la Mediterran ee, Marseille 2004.

Durlauf, S. N., and P. A. Johnson. "Local Versus Global Convergence Across National Economies." NBER Working Paper No. 3996, National Bureau of Economic Research, 1992.

Easterly, W. "When Is Fiscal Adjustment an Illusion?" Economic Policy, 14(28), 1999, 55-86.

Edwards, S. Crisis and Reform in Latin America--From Despair to Hope. New York: Oxford University Press 1995.

Elias, V. Sources of Growth: A Study of Seven Latin American Economics. San Francisco: ICS Press 1992.

Emran, M. S., and J. E. Stiglitz. "On Selective Indirect Tax Reform in Developing Countries." Journal of Public Economics, 89, 2005, 599-623.

Fay, M. "Financing the Future: Infrastructure Need in Latin America, 2000-05." Working Paper No. 2545, The World Bank, Washington, DC 2001.

Feige, E. L. "How Big is the Irregular Economy?" Challenge, November 1979, 5-13.

Futagami, K., Y. Morita, and A. Shibata. "Dynamic Analysis of an Endogenous Growth Model with Public Capital." Scandinavian Journal of Economics, 95, 1993, 607-25.

Glomm, G., and F. Rioja. "Populist Budgets and Long Run Growth." Manuscript, Indiana University 2003.

Gourio, F., and P.-A. Noual. "The Marginal Worker and The Aggregate Elasticity of Labor Supply." Manuscript, Boston University 2006.

Gupta, R. "Tax Evasion and Financial Repression." Journal of Economics and Business, 60(6), 2007, 517-535.

Hatzipanayotou, P., M. S. Michael, and S. M. Miller. "Win-Win Indirect Tax Reform: A Modest Proposal." Economic Letters, 44, 1994, 147-51.

Hicks, N. L. "Expenditure Reductions in Developing Countries Revisited." Journal of International Development, 3, 1991, 29-37.

International Monetary Fund. "Doha Trade Talks: The Main Risk Is a 'Gutless' Agreement." IMF Survey, 34(9), 2005, 148-49.

Keen, M., and J. E. Ligthart. "Coordinating Tariff Reduction and Domestic Tax Reform." Journal of International Economics, 56, 2002, 489-507.

Kimball, M.S. "Precautionary Saving in the Small and in the Large." Econometrica, 58, 1990, 53-73.

Landskroner, Y., J. Paroush, and I. Swary. "Tax Evasion and Portfolio Decisions." Public Finance, 45, 1990, 132-57.

Lin, S. "The Decline of China's Budgetary Revenue: Reasons and Consequences." Contemporary Economic Policy, 18, 2000, 477-90.

Lynde, C., and J. Richmond. "Public Capital and Total Factor Productivity." International Economic Review, 34, 1993,401-44.

Munk, K. J. "Tax-tariff Reform with Costs of Tax Administration." Department of Economics Working Paper 2006-14, Department of Economics, University of Aarhus 2006.

Ogaki, M., J. Ostry, and C. Reinhart. "Savings Behavior in Low- and Middle-Income Developing Countries." IMF Staff Papers, 43, 1996, 38-71.

Pohl, G., and D. Mihaljek. "Project Evaluation and Uncertainty in Practice: A Statistical Analysis of Rate-of-Return Divergences of 1,015 World Bank Projects." World Bank Economic Review, 6, 1992, 255-77.

Reinganum, J. F., and L. L. Wilde. "Income Tax Compliance in a Principal-Agent Framework." Journal of Public Economics, 26, 1985, 1-18.

Rochjadi, A., and J. H. Leuthold. "The Effect of Taxation on Labor Supply in a Developing Country: Evidence from Cross-Sectional Data." Economic Development and Cultural Change, 42, 1994, 333-50.

Rodrik, D. "Understanding Economic Policy Reform." Journal of Economic Literature, 34, 1996, 9-41.

Roubini, I., and J. Sachs. "Government Spending and Budget Deficits in the Industrial Countries." Economic Policy, 4(8), 1989, 99-132.

Singh, I., L. Squire, and J. Strauss. "A Survey of Agricultural Household Models: Recent Findings and Policy Implications." World Bank Economic Review, 1(1), 1986, 149-79.

Summers, R., and A. Heston. "The Penn World Table (Mark 5): An Expanded Set of International Comparisons, 1950-1988." Quarterly Journal of Economics, 106, 1991, 327-68.

Tanzi, V. "Structural Factors and Tax Revenue in Developing Countries: A Decade of Evidence," in Open Economies: Structural Adjustment and Agriculture, edited by I. Goldin and L. A. Winters. Cambridge: Cambridge University Press 1992, 267-289.

Taylor, L. "Structuralist CGE Models." in Socially Relevant Policy Analysis: Structuralist Computable General Equilibrium Models for the Developing World, edited by L. Taylor. Cambridge, MA: MIT Press 1990, 1-70.

Vernengo, M. "External Liberalization, Stabilization, and the Labor Market in Brazil." Latin American Perspectives, 31(4), 2004, 62-75.

World Bank. "World Development Report." Washington, DC: The World Bank, 1988.

Yitzhaki, S. "Income Tax Evasion: A Theoretical Analysis." Journal of Public Economics, 3, 1974, 201-2.

(1.) For first-best case, see Diamond and Mirrlees (1971), Hatzipanayotou, Michael, and Miller (1994), and Keen and Ligthart (2002).

(2.) Infinite-horizon models with heterogeneous agents and idiosyncratic noninsurable shocks can only be solved in some special cases as the whole distribution of wealth across agents becomes the state variable for the dynamic economy.

(3.) The agents do not have any bequest motive.

(4.) We do not consider capital income tax as it is a small fraction of government revenue in developing countries. Furthermore, in the model, unlike the labor income tax which is nondistortionary, the capital income tax will be distortionary even in absence of tax evasion and increase in capital income tax pursuant to reform will, therefore, reduce the welfare gain (or increase welfare loss) from the reform. However, note that even the labor income tax is distortionary in the extension analyzed in Section V.A.

(5.) The informal sector is an important and large component of developing economies. Although this equivalence has not been made explicit, it is possible to interpret the share of the labor income on which tax evaded as the proportion of labor that the agent supplies in the informal sector. With this reinterpretation, the changes in degree of tax evasion in the model will correspond to the changes in the size of informal economy and the amount of labor employed therein.

(6.) A reader may question the rationale for tax on capital as, even in the second-best world, it is desirable to have production efficiency (see Diamond and Mirrlees, 1971). Since we are concerned with the welfare effects of actual tariff reforms in developing countries, we want to capture the extant practices in these countries. Indeed, these countries levied tariffs on both productive inputs and consumption goods albeit at different rates.

(7.) It may be mentioned that, in this paper in general, the economy-wide variables (aggregate or average) are denoted by capital letters, while those relating to the individual agents are denoted by (corresponding) small letters. However, note an exception: both aggregate and individual consumption expenditures will be denoted by E.

(8.) The aggregate resource constraint can be written in this form as the world prices of all goods are normalized to 1.

(9.) Atolia (2008) analyzes in detail the effect of tax policy on tax evasion, intertemporal resource allocation, and growth.

(10.) The calibration process adds three additional equations to the set of equations listed at the end of Section II. The extended system is then solved for the steady-state values of all endogenous variables and [beta], p, and [zeta]. The three equations that are added set the values of [bar.R]/Y and the return to public and private values to their target values.

(11.) The political nature of these constraints is, however, recognized in policy circles (see IMF, 2005).

(12.) For now, the audit rate (p) is kept unchanged as change in p implies a change in administrative costs, and these costs are not considered until Section V.B.

(13.) Since there is a sustained mismatch between revenue and expenditure, borrowing by government whether in domestic or international market is not feasible and the expenditure has to be curtailed and public investment is first to be axed. In addition, in many developing countries, domestic bond markets are nonexistent or very small.

(14.) For proportionally greater adverse effect of fiscal tightening on public investment in transition economies, see Alam and Sundberg (2002). Amin (1999) provides similar evidence for Egypt, and Dropsy and Grand (2004) do so for Morocco and Tunisia. Also see Easterly (1999) for a thorough discussion of this issue.

(15.) Estimating tax evasion is especially problematic as it is an illegal activity. However, estimates for many countries are as high as 50%; for example, see Acharya (1985) for India, Feige (1979) for Italy, Alm, Bahl, and Murray (1991) for Jamaica.

(16.) In some cases, particularly in Latin America, fall in tariffs has been larger. As larger tariff reduction yields progressively smaller gains from reduced consumption distortion and larger losses from increased tax evasion and reduced public investment, we are considering a conservative scenario.

(17.) For a detailed exposition of this outcome based on Equations (17 and 18), the reader is referred to Atolia (2008).

(18.) The value of the physical capital stock, however, falls as its price in terms of domestic good falls from 1.4 to 1.2.

(19.) Rochjadi and Leuthold (1994) in a study of effects of taxation on labor supply in Indonesia estimate a value of .50 for males and .59 for females. In another extensive study covering seven countries. Singh, Squire, and Strauss (1986) find estimates ranging from .11 to .45 when profits of agricultural households are allowed to vary. Recently, Barrett, Sherlund, and Adesina (2007) have estimated uncompensated wage elasticity of .12. The compensated wage elasticity is typically much higher than the uncompensated elasticity. For example, in Rochjadi and Leuthold (1994), the uncompensated elasticity is estimated to be 0, whereas the estimate of compensated elasticity is more than .5. Thus, the chosen value of [[epsilon].sub.w.sup.c] is empirically reasonable, and if anything, a conservative choice and will tend to understate the losses arising from increased labor income taxation.

(20.) The tax burden is associated not only with public investment but also with transfers as both are financed through distortionary taxes when labor supply is elastic.

(21.) Futagami, Morita, and Shibata (1993), Corsetti and Roubini (1996), and Agenor (2005) analyze the trade-off between increase in productivity due to public investment and the tax burden associated with raising revenue for such investment.

ABBREVIATIONS

IMF: International Monetary Fund

GDP: Gross Domestic Product

MANOJ ATOLIA *

* I am grateful to Ed Buffie and two anonymous referees who provided extensive feedback. Any errors remaining are my own.

Atolia: Department of Economics, Florida State University, Tallahassee, FL 32306. Phone 850-644-7088, Fax 1-850-644-4535, E-mail matolia@fsu.edu

doi: 10.1111/j.1465-7287.2009.00176.x

Printer friendly Cite/link Email Feedback | |

Author: | Atolia, Manoj |
---|---|

Publication: | Contemporary Economic Policy |

Geographic Code: | 0DEVE |

Date: | Apr 1, 2010 |

Words: | 15790 |

Previous Article: | Democratic aid and the democratization of recipients. |

Next Article: | Should a mandatory subordinated debt policy be introduced in the United Kingdom? Evidence from the issuance activity of banks and building societies. |

Topics: |