The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
clevelandfed.org/research/workpaper/index.cfm Working Paper 9409 THE WELFARE EFFECTS OF TAX SIMPLIFICATION: A GENERALEQUILIBRIUM ANALYSIS by Jang-Ting Guo and Kevin J. Lansing Jang-Ting Guo is a professor of economics at the University of California, Riverside, and Kevin J. Lansing is an economist at the Federal Reserve Bank of Cleveland. Working papers of the Federal Reserve Bank of Cleveland are preliminary materials circulated to stimulate discussion and critical comment. The views stated herein are those of the authors and not necessarily those of the Federal Reserve Bank of Cleveland or of the Board of Governors of the Federal Reserve System. September 1994 clevelandfed.org/research/workpaper/index.cfm ABSTRACT This paper employs a dynamic general-equilibrium model to analyze various schemes for simplifying the U.S. tax system, such as a uniform tax levied on all types of income regardless of source, and the elimination of specific tax breaks like the depreciation allowance. Under each scheme, the government selects a balanced-budget fiscal policy (consisting of tax rates and the level of public expenditures) which maximizes household welfare given the constraints imposed by the particular tax system. We find that a uniform tax system does almost as well as a system with separate taxes on labor and capital incomes, provided that a depreciation allowance is maintained. Without the depreciation allowance, a uniform tax system significantly reduces household welfare, even though marginal tax rates are lower under this scheme. The welfare differences between the various distortionary tax systems are much smaller than the potential welfare gains from switching to a system of nondistortionary, lump-sum taxes. The various tax systems are also shown to display very different behavior for the movement of tax rates and aggregate economic variables over the business cycle. clevelandfed.org/research/workpaper/index.cfm 1. Introduction Every year, Congress passes a bill that tinkers in some way with our tax system. During the 1980s however, two major tax bills were enacted that fundamentally altered the structure of the federal income tax: the Economic Recovery Tax Act of 1981 (ERTA) and the Tax Reform Act of 1986 (TRA). Although many changes to the tax code have been made since ERTA and TRA, the current U.S.tax system derives most of its basic structure from these two laws. In this paper, we first review the basic features of these tax laws and then construct a dynamic general-equilibrium model to analyze an issue which is closely linked to tax reform, namely, tax simplification. Our primary finding is that a simplified system which involves a uniform tax levied on all types of income does almost as well as a more complicated system with separate tax rates on labor and capital income, provided that a depreciation allowance is maintained. In 1981, ERTA imposed a dramatic 23 percent, across-the-board cut in all marginal tax rates, and reduced the top marginal rate for individual income from 70 to 50 percent. Statutory marginal rates were scaled back to levels similar to those that prevailed in 1965. In addition, ERTA introduced new incentives for investment and saving, such as an increased investment tax credit, a generous accelerated depreciation schedule, and an extension to the eligibility rules for Individual Retirement Accounts (IRAs).' ERTA also helped to eliminate "bracket creep" by indexing tax brackets, personal exemptions, and the standard deduction for inflation. In 1986, TRA brought about the most significant overhaul of the federal tax system since its inception in 1913. The act lowered marginal tax rates for individuals and corporations, dramatically reduced the number of tax brackets, broadened the tax base by eliminating or reducing many tax breaks, and helped to "level the playing field by reducing the dispersion of marginal tax rates across alternative income-producing activities. Also, the lowering of marginal rates reduced the attractiveness of tax evasion and tax avoidance activities. Because of these features, TRA was viewed as taking a significant step toward the goal of achieving a simpler, more efficient federal tax system. An important result of 'some of the investment incentives in ERTA, such as the generous accelerated depreciation schedule, were scaled back somewhat by the Tax Equity and Fiscal Responsibility Act of 1982 (TEFRA). See Economic Report of the President, 1987 and 1989. clevelandfed.org/research/workpaper/index.cfm TRA, and one which we will focus on here, was that average marginal tax rates on labor and capital income were brought closer together. The figures in table 1, taken from Economic Report of the President, illustrate this point.' Table 1: Average Marginal Tax Rates before and after TRA 1986 Source of Income Before TRA After TRA Labor Income Capital Income Source: Economic Report of the President, 1987, Table 2-6, page 91. TRA also reduced the variation in marginal tax rates within each category of income. In the labor income category, the number of individual income brackets was reduced to only two, set at 15 and 28 p e r ~ e n t . In ~ the capital income category, TRA eliminated the investment tax credit (which, under ERTA, had applied to equipment but not structures), eliminated the capital gains preference by taxing gains as ordinary income, decelerated depreciation allowances on real estate, imposed limitations on passive business and real estate losses, and phased out the deductibility of nonmortgage consumer interest. Furthermore, the lowering of personal marginal rates and the top corporate tax rate reduced the attractiveness of tax shelters. By imposing a more uniform tax on alternative sources of income, TRA attempted to eliminate incentives in the tax code that had directed resources to less productive activities offering high after-tax returns. These perverse incentives result in a loss of output and a reduction in economic welfare. In addition, a simpler, more efficient tax system could be expected to contribute to increased compliance and lower administration costs. In this paper, we examine the issue of tax simplification using a dynamic general-equilibrium model which is calibrated to the U.S. economy. Based on the results of TRA, tax simplification is %or additional description and analysis of TRA 1986, see Slemrod (1991) and the two symposia in The Journal of Economic Perspectives, Summer 1987 and Winter 1992. 3 ~ ohigh-income r individuals, the 15 percent tax bracket and the personal exemption were phased out, creating an implicit third bracket, at 33 percent. Before TRA, there were fourteen tax brackets ranging from 11 to 50 percent. In 1990, the Omnibus Budget Reconciliation Act (OBRA90) created a new statutory bracket at 31 percent. OBRA93, enacted in August 1993, added two statutory brackets for high-income individuals. With OBRA93, there are now five statutory brackets at 15, 28, 31, 36 and 39.6 percent. See Economic Report of the President, 1994, Table 1-4, p. 34. clevelandfed.org/research/workpaper/index.cfm modeled as a system which imposes a uniform tax on all types of income, regardless of source. The more complicated system in our model is one in which capital and labor income are taxed at different rates. A basic principle underlying TRA was to broaden the tax base by eliminating tax breaks, thus allowing revenue requirements to be met with lower marginal rates. We capture this principle by eliminating the model's only tax break, the depreciation allowance. We compare the uniform tax system, with and without the depreciation allowance, to the more complicated tax system. We also compare all three systems to one that employs nondistortionary, lump-sum taxes. The various tax systems are evaluated in terms of economic welfare (as measured by steady-state household utility), output (as measured by steady-state GNP), and business-cycle characteristics (as measured by the relative standard deviations and cross-correlations of aggregate economic variables). Under each scheme, the government chooses a balanced-budget fiscal policy that maximizes household welfare, given the constraints imposed by the particular tax system. In this way, we endogenize the choice of fiscal policy, which consists of a set of stationary decision rules for tax rates and the level of public expenditures. Our approach differs from that of Cooley and Hansen (1992), who evaluate the welfare effects of various combinations of exogenous tax rates. From the perspective of choosing an optimal fiscal policy, a uniform tax system imposes an additional constraint on the government's decision problem, namely, that the tax rate on capital income must be equal to the tax rate on labor income. Because a policy of equal tax rates is available to the government, but not chosen, under the complicated system, we know that the additional constraint is binding and thus results in a lower level of household utility than in the unconstrained case. Our aim is to quantify this welfare effect and to evaluate the additional impact of the depreciation allowance. We find that the difference in welfare between the complicated system and the simplified tax system with a depreciation allowance is less than 0.1 percent of GNP, or $23 per person per year in 1993. The steady-state output loss associated with this version of the simplified tax system is slightly less than 1 percent of GNP. These results suggest that a uniform tax system could actually be welfareimproving if sufficient cost saving were realized in the areas of compliance and administration. Without the depreciation allowance, a uniform tax system significantly reduces household welfare and steadystate output, as compared to both the complicated system and a uniform tax system with a depreciation clevelandfed.org/research/workpaper/index.cfm allowance. Household welfare is reduced by approximately 2 percent of GNP or $523 per person per year in 1993. The output loss is over 4 percent of GNP. Eliminating the depreciation allowance reduces household welfare because this tax break operates as an implicit subsidy to capital accumulation, and partially offsets the income tax distortion. This is true even though the marginal tax rate on capital income is lower under a system with no depreciation allowance. Because of the government's desire to tax profits as much as possible, a high tax rate on capital combined with a full depreciation allowance is superior to a low tax rate on capital combined with no depreciation allowance. Our result points out the important role played by the depreciation allowance in encouraging capital accumulation. Thus, the inflation-induced erosion of unused nominal depreciation allowances, which are carried over from year to year in historical cost terms, may impose a significant welfare cost on the U.S. e ~ o n o m y . ~ Moreover, we find that the welfare differences between the various distortionary tax systems are much smaller than the potential welfare gains from switching to a system of nondistortionary, lump-sum taxes. Our calculations indicate that distortionary taxes impose a welfare cost exceeding 14 percent of GNP. Because of distributional issues and concerns over the perceived fairness of the tax code, we do not view the lump-sum tax system as a realistic policy option. However, it does provide a useful benchmark for our analysis. Finally, we show that the various tax systems have very different implications for the comovement and relative variability of tax rates and aggregate economic variables over the business cycle. Under the complicated tax system, the capital tax moves countercyclically and displays high variability because it is used to absorb shocks to the government's budget constraint. In the model, budget shocks are caused by changes in the size of the tax base due to business-cycle fluctuations. Under the simplified tax system, government expenditures are much more variable because a separate ~ instrument for absorbing budget shocks is not available. Also, for a given variance of the technology shock, we find that output, consumption, and hours worked are more variable under a system of lump- i 4~udd(1989) emphasizes this point by treating the inflation erosion of unused depreciation allowances as an effective increase in the tax rate on capital. See Altig and Carlstrom (1991) for a business-cycle model that incorporates inflation's effect on the nominal taxation of capital income. Pecorino (1993) examines the growth effects of the depreciation allowance in an endogenous growth model. clevelandfed.org/research/workpaper/index.cfm sum taxes than under any of the distortionary systems. This occurs because movements in distortionary tax rates tend to offset the effects of technology shocks, thus resulting in lower output variability. The remainder of this paper is organized in the following manner: Sections 2 and 3 describe the model and the solution method. The choice of parameter values is discussed in section 4. Section 5 presents quantitative welfare comparisons based on steady-state analysis. In section 6, we examine the business-cycle characteristics of the various tax schemes. Section 7 provides concluding remarks. 2. The Model The model economy consists of many identical, infinitely lived households, identical private firms, and the government. Households derive direct utility from government-provided public goods which are financed by taxes on households and firms. Following Benhabib and Farmer (1994), we assume that firms exert some degree of monopoly power over the production of intermediate goods so that they realize positive economic profits even though the market for final goods is perfectly competitive. The profits are equal to the difference between the value of output and the payments to labor and capital inputs. The purpose of introducing profits is to obtain a positive optimal tax rate on capital, consistent with U.S. observation^.^ As owners of the firms, households receive net profits in the form of dividends, but consider them to be outside their control, similar to wages and interest rates. It is assumed that profits are initially taxed at the firm level, then distributed as dividends and taxed again at the household level. This formulation is intended to capture the double taxation of corporate dividends in the U.S. economy. Furthermore, in the complicated tax system, we assume that the government can distinguish between labor and capital income, but cannot distinguish between the different categories of capital income, such as profits, dividends, and capital rental income. Therefore, the complicated system includes only two types of distortionary taxes: a labor tax and a capital tax. ones, Manuelli, and Rossi (1993b) show that the existence of profits and a restriction on the menu of available tax instruments (the absence of a separate profits tax) is one method of obtaining a positive optimal tax rate on capital in the steady state. Without profits, the optimal steady-state tax on capital is zero (see Judd [I9851 and Chamley [1986]). clevelandfed.org/research/workpaper/index.cfm 2.1 The Household's Problem Households maximize a discounted stream of within-period utility functions over consumption and leisure, subject to a sequence of budget constraints. The decision problem can be summarized as max .. c, h, k,,, {lnc, -Ah, + s i n g , ) O<p<l, A, BLO (1) t-O subject to kt+, = ( 1-6 )kt + x, 0 <6 < 1 , ko given. In the above equations, c, represents private consumption goods. Households are endowed with one unit of time each period and work h, hours during period t. Household preferences also include a separable term representing the utility provided by public consumption goods g,. Examples of public consumption goods that might affect household utility are national defense, police protection, and government provision of food and housing during natural disasters. Public goods are assumed to be noncongestable and free of specific user charges. Households maximize the utility function in (1) over c,, and h,, but view g, as outside their control. The logarithmic form of the within-period utility function has been chosen for tractability and for comparability with previous business-cycle literature. The separability in c, and g, implies that public consumption does not affect the marginal utility of private consumption, a specification supported by parameter estimates in McGrattan, Rogerson, and Wright (1993). The symbol E, is the expectation operator conditional on information available at time t, and P is the constant household discount factor. The fact that utility is linear in hours worked reflects "indivisible labor," as described by Rogerson (1988) and Hansen (1985). This means that all variation in economywide hours is due to variations in the number of employed workers, as opposed to variations in hours per worker. Real business cycle models with indivisible labor are better able to match some key characteristics of aggregate labor-market data. Specifically, U.S. data display a large volatility of hours worked relative clevelandfed.org/research/workpaper/index.cfm to labor productivity and a near-zero correlation between hours and produ~tivit~.~ Equation (2) represents the period budget constraint of the household. The terms xt and kt represent gross private investment and private capital, respectively. Households derive income by supplying labor and capital services to firms at rental rates wt and rt , and pay taxes on labor and capital income at rates z, and z,, respectively. The term Tt represents a lump-sum tax. An additional source of income is the firms' net profits fit, which are distributed to households as dividends and are taxed at the same rate as capital rental income rt k,. The term (Iz, 6kt represents the depreciation allowance, where the parameter (I can be set to either 1 or 0,depending on whether this tax break is maintained in the simplified tax system. Equation (3) is the law of motion for private capital, given a constant rate of depreciation 6. Households view tax rates, wages, interest rates and dividends as determined outside their control. 2.2 Household Optimality The Lagrangian for the households' problem is defined as: The household first-order conditions with respect to the indicated variables and the associated transversality conditions (TVC)are: 'see Hansen (1985), and Christiano and Eichenbaum (1992). 7 clevelandfed.org/research/workpaper/index.cfm TVC: limEopthtkt+,= 0. t +- The government's problem is solved by finding the set of utility-maximizing allocations ct , ht and k;,,, such that the household's first-order conditions (5) and budget constraint (2) are satisfied. Given these optimal allocations, the government uses the household equilibrium conditions to recover the appropriate tax rates z, and z, that will support these allocations in a decentralized economy? 2 3 The Firm's Problem The firm's problem is based on the model developed by Benhabib and Farmer (1994). A unique final good yt is produced out of a continuum of intermediate goods y,, i E [0,1], using the following constant-returns-to-scale technology: yt = [I,' yi,Xdillh, 0 < x <1. We assume that the final goods sector is perfectly competitive, but that intermediate goods producers each exert a degree of monopoly power that is captured by the parameter X. In the special case when x = 1, all intermediate goods are perfect substitutes in the production of the final good, and the intermediate sector becomes perfectly competitive. All intermediate goods are produced using the same technology with labor and capital as inputs: yit = exp(zt) kiP1hi:', Zt+l = P z Z t +&,+I where O < a i < l , a , + a , 2 1, O<pz<l, ~,-iid(0,0;), zo given. (6) (3 Notice that equation (6) allows for the possibility of increasing returns to scale in the production 7 ~ e Chari, e Christiano, and Kehoe (1993) for a more complete discussion of this equilibrium concept 8 clevelandfed.org/research/workpaper/index.cfm of intermediate goods8 The intermediate technology is subject to economywide exogenous shocks ( z ,), which are revealed to agents at the beginning of period t. These shocks generate equilibrium businesscycle fluctuations in the model. Due to their monopoly power, producers of intermediate goods earn an economic profit which is taxed at the rate z, . Under the assumption that factor markets are competitive, a symmetric equilibrium (k= kt and hit=ht for all i) implies the following expressions for the aggregate production function (y, = y, ), the rental rate of capital, and the real wage. where 8, = X a, e2 = X a,. and From (8), the aggregate technology can be rewritten as 1 y, = exp(nt)[k," h t e 2 ] ~ , where 1 (0, + €I2)- 2 1. X Depending on the value of X, the aggregate technology can demonstrate either constant or increasing returns to scale. The after-tax profits, distributed to households in the form of dividends, are 2.4 The Government's Problem The government chooses an optimal program of taxes and public expenditures in order to maximize the discounted utility of the household. The vector !Pt= { gt , z, , z, , Tt ) summarizes government policy implemented at time t. This is a dynamic version of the Rarnsey (1927) optimal tax problem, involving a Stackelberg game between the government and households. To avoid timeconsistency problems, we assume that the government can commit to a set of state-contingent, stationary policy rules announced at time zero. Also, to make the problem interesting, we rule out any time-zero 'see Benhabib and Farmer (1994) and Farmer and Guo (1994) for more details regarding this formulation. 9 clevelandfed.org/research/workpaper/index.cfm levies on private-sector assets that might be used to finance all future expenditures. In addition, we assume that the government adheres to a period-by-period, balanced-budget constraint? With these assumptions, the government's problem is subject to: (i) household first-order conditions and budget constraint, (ii) firm profit maximization conditions, (iv) 4 (v) T~~ = T (vi) T,, = 1, -- T,, Tt = 0, for the complicated tax system, = T ~ ,@ = { = 0, = Tt 1, 0 ] , g,, Tt = 0, for the simplified tax system, for the lump-sum tax system. Constraints (i) and (ii) summarize rational maximizing behavior on the part of private agents and constitute "irnplementability" constraints imposed on the government's choice of policy. Constraint (iii) is a general version of the government budget constraint, where the squared term on the right-hand side reflects the double taxation of firm dividends. Finally, (iv) through (vi) speclfy the constraints associated with each of various tax systems we intend to analyze. The summation of the household budget constraint (2) and the government budget constraint (iii) yields the following resource constraint for the economy: Because the resource constraint and the government budget constraint are not independent ' ~ d d i ngovernment ~ debt to the model introduces complications that we wish to avoid here. Specifically, equilibrium for a model with debt and capital imposes an ex ante arbitrage condition on the expected returns from government bonds and private capital. The steady-statelevel of debt is thus indeterminate (see Charnley [1985]).Furthermore, in a stochastic environment, the government can vary the ex post combination of the capital tax and the bond interest rate in many different ways to raise needed revenue, yet still satisfy ex ante arbitrage (see Zhu 119921 and Chari, Christiano, and Kehoe 119931).Excluding government debt allows us to pin down a unique time path for the optimal capital tax. clevelandfed.org/research/workpaper/index.cfm equations, equation (12) will be used in place of (iii) in solving the government's problem. 3. Solution of the Model The governrnent's problem under commitment can be solved using the recursive algorithm developed by Kydland and Prescott (1980). A recursive structure is obtained by defining the household lagged shadow price h,-, as a "pseudo-state variable." Including h,-,in the state vector provides a link to the past by which the policymaker at time t considers the fact that household decisions in earlier periods depend on current policy by means of expectations. This is the mechanism by which the commitment problem can be solved using dynamic programming. Appendix A describes the procedure for formulating the recursive version of (1 1) and numerically solving the dynamic programming problem. 4. Calibration of the Model To obtain quantitative welfare estimates from the model, as many parameters as possible are assigned values in advance on the basis of empirically observed features of postwar U.S. data. Parameter choices are also guided by the desire to obtain steady-state values for key model variables that are consistent with long-run averages in the U.S. economy. Table 2 summarizes the choice of parameter values and is followed by a brief description of how they were selected. Table 2: Parameter Set Parameters and Values Agent = 0.962 A = 2.50 B = 0.35 Firms 0,= 0.31 0, = 0.60 x = 0.91 6 = 0.07 Government $ = { I , 0) Households p, = 0.85 a, = 0.02 The time period in the model is taken to be one year. This is consistent with the time frame of most government fiscal decisions and the frequency of available data on average marginal tax rates. The discount factor of P=0.962 implies an annual rate of time preference equal to 4 percent. The household utility parameter A is chosen such that hours worked in the distortionary tax systems is close to 0.3. This is in line with time-use studies, such as Juster and Stafford (1991), which indicate that households spend clevelandfed.org/research/workpaper/index.cfm approximately one-third of their discretionary time in market work. The value of B is chosen to yield a steady-state value of g/GNP near 0.22, the average ratio for the U.S. economy from 1947 to 1992. The exponents in the Cobb-Douglas production function are chosen on the basis of two criteria. First, the chosen values of 8, and 8, are in the range of the estimated shares of GNP received by capital and labor in the U.S. economy (see Christian0 [1988]). Second, the model's share of GNP devoted to monopoly profits (= 1- 8, - 8,) is chosen to yield a reasonable value for the steady-state tax on capital ( T ~in) the complicated tax system. Because a separate profits tax is not available, the government uses the tax on private capital to recapture a portion of these profits. In the model, the steady-state ratio of profits to GNP is 0.09 and the resulting steady-state tax on capital is 0.30. This value of T~ coincides with the average effective corporate tax rate in the United States from 1947 to 1980, as estimated by x is chosen such that the aggregate scale. The value of x =0.91 yields Jorgenson and Sullivan (1981).1° The monopoly power parameter production technology demonstrates constant returns to yt = exp (2,) k:34 hp.66. We also experiment with smaller values of X, such that the aggregate technology is characterized by increasing returns to scale. The private capital depreciation rate of 6 = 0.07 is based on the value estimated in Braun and McGrattan (1993) and is consistent with values commonly used in the real business cycle literature. Together with the values of P and €4, this depreciation rate implies a steady-state ratio of private capital to GNP in the range of 2.27 to 3.1 1 for the various tax systems, and a ratio of private investment to GNP in the m g e of 0.16 to 0.22. The corresponding averages for the U.S. economy from 1947 to 1992 are 2.58 and 0.21. The process governing technology shocks was estimated using annual data from 1947 to 1992. The series for zt was constructed by computing the changes in output not accounted for by changes in the productive inputs." The estimated parameters, p, = 0.85 and o,= 0.02, represent values close to 1°~igherprofit levels imply a higher steady-state tax on capital in our model. When profits are zero (€I, +€I,= 1). the optimal steady-state tax on capital is zero. If a separate profits tax was available, the government would choose to tax profits at 100 percent and ordinary capital income (r, k,) at zero percent (see footnote 5). production function residual was measured as z, = InGNP, - 0.34hkI- 0.66hh,. The private capital stock k, is defined as fixed private capital + stock of consumer durables + residential capital from Fixed Reproducible Tangible Wealth in the United States. U.S. Department of Commerce (1993). Real GNP and the labor input (h,= LHOURS) are from Citibase. clevelandfed.org/research/workpaper/index.cfm those estimated by other studies using annual data, such as Benhabib and Jovanovic (1991). In the simulations, the estimated value of o, yields a standard deviation of output in the range of 1.74 to 2.96 percent. The U.S. average over this period is 2.46 percent. 5. Steady-State Welfare Analysis Based on our choice of parameter values, table 3 shows the steady-state values of key model variables for each of the various tax systems. Of the three distortionary tax systems, the complicated system has the highest level of steady-state utility and the largest output, followed closely by the simplified system with a depreciation allowance. Table 4 indicates that the difference in utility between these two systems is less than 0.1 percent of GNP, which translates to a loss of $23 per person per year in 1993.12To put this number in perspective, Cooley and Hansen (1991) estimate the welfare cost of a 5 percent annual inflation to be 0.34 percent of GNP annually. The simplified tax system could actually be welfare-improving if sufficient cost savings were realized in the areas of compliance and administration. However, the necessary savings represent about 9 percent of the estimated total resource cost of administering the federal tax system.13Cost savings of this magnitude would appear to be very difficult to achieve. Also, the simplified system's lack of an explicit capital tax may be undesirable from the standpoint of redistributing resources in an economy where wealth is highly concentrated in the hands of rich taxpayers.14 When the depreciation allowance is eliminated from the simplified system, household welfare is reduced by over 2 percent of GNP, or $523 per person per year in 1993. Eliminating the depreciation allowance reduces household welfare because it operates as an implicit subsidy to capital accumulation, '%his figure is based on a nominal GNP of $6,510 billion and a total U.S. population of 258.2 million in 1993. 13~osen(1992, p. 351) estimates the total resource cost of administering the federal tax system to be $35.3 billion in 1982, or 1.11 percent of 1982 GNP. If we assume that the ratio of resource cost to GNP is approximately the same in 1993, and that the necessary cost savings to make the simplified system desirable amount to 0.1 percent of GNP, then the necessary savings represent 9 percent (=0.1/1.11) of the total resource cost. "?n the United States, the top 20 percent of households own about 80 percent of total wealth and earn about 42 percent of pretax income. See McDermed, Clark, and Allen (1989). figures 13.1 and 13.2, and Rosen (1992), table 8.1. See Lansing (1994b) for an optimal tax model with rich and poor households. clevelandfed.org/research/workpaper/index.cfm and partially offsets the income tax distortion. This is true even though the marginal tax rate on capital income (2,) is lower under this scheme. Because of the government's desire to tax profits as much as possible (see Jones, Manuelli and Rossi [1993b]), a high tax rate on capital income combined with a full depreciation allowance is superior to a low tax rate on capital income combined with no depreciation allowance.15 Table 5 indicates that the welfare effects of tax simplification are much smaller than the potential welfare gains from switching to a system of lump-sum taxation. Under the complicated system, the welfare loss from distortionary taxation translates to $3,533 per person per year in 1993. However, our representative household framework abstracts from any distributional role fuKied by a system of distortionary taxes. The lump-sum tax system should thus be viewed as a benchmark, but not as a realistic policy option. '%hen profits are zero (1-9, -9,=0) and the government can tax labor and capital income separately, eliminating the depreciation allowance has no effect whatsoever on the steady-state allocations. This can be seen from the government's recursive problem in appendix A, for the complicated tax system (T,#T,); note that the depreciation allowance parameter I$ only appears in the profit term of the (transformed) household budget constraint clevelandfed.org/research/workpaper/index.cfm Table 3: Steady-State Comparison of Tax Systems Variable Complicated Utility -2.764 Tax Rate zh= 0.22 2, = 0.30 W/ Simplified Depr Allow Simplified No Depr Allow Lump-sum U.S. Economya -2.766 -2.798 -2.547 n.c. z = 0.25 z = 0.20 T l y = 0.23 zh= 0.27, 0.23 zk = 0.30, 0.57 T l y =0.28 an.c. =not comparable. The U.S. values for k / y and g / y represent the 1947-92 averages, from Citibase. Sources for U.S. tax rates are as follows: The two values for T, are averages from Barro and Sahasakul(1986), 194783, and McGrattan, Rogerson, and Wright (1993), 1947-87, respectively. The two values for z, are averages from Jorgenson and Sullivan (1981, table l l ) , 1947-80, and McGrattan, Rogerson, and Wright (1993), 1947-87, respectively. The value of T / y is the average of total government receipts (federal, state, and local) as a percentage of GNP, 1947-92, from Economic Report of the Presides, 1994, table B-80, p. 363. Table 4: Effects of Tax Simplification Welfare and Output Loss versus Complicated Tax Systema Tax System W/ Welfare Loss = AUl(5y) Output Loss = Ay 1y Simplified Depr Allowance Simplified No Depr Allowance 'AU and Ay are normalized using the steady-state values of h and y from the complicated tax system, where h is the marginal utility of private consumption (to convert AU into consumption units) and y is GNP. Source: Authors' calculations. clevelandfed.org/research/workpaper/index.cfm Table 5: Effects of Distortionary Taxes Welfare and Output Loss versus Lump-sum Tax Systema Tax System Welfare Loss = AU/(hy) Output Loss = Ay l y Complicated W/ Simplified Depr Allowance Simplified No Depr Allowance 'AU and Ay are normalized using the steady-state values of h and y from the lump-sum tax system, where h is the marginal utility of private consumption (to convert AU into consumption units) and y is GNP. Source: Authors' calculations. The welfare costs in table 4 tend to be higher for an economy with increasing returns to scale. When x = 0.75, the aggregate production technology is y, = exp (2,) k0.41 hbgO.In this case, the welfare losses in the first column of table 4 become 0.19 percent and 2.69 percent of GNP, respectively, for the two simplified systems. 6. Dynamic Simulation of Business-Cycle Behavior Our simulation results (table 6 and figures 1 and 2) show that the various tax systems have very different implications for the comovement and relative variability of tax rates and aggregate economic variables over the business cycle. In the model, shocks to the government's budget constraint are caused by changes in the size of the tax base due to business-cycle fluctuations. Because we have imposed a balanced-budget restriction, government debt is not available to help smooth public expenditures when budget shocks occur. In the complicated system, however, the shock absorbing function is performed by a state-contingent capital tax, which moves countercyclically and displays high variability relative to the labor tax. For example, a positive technology shock generates more tax revenue because GNP and household incomes (the tax base) increase. This motivates a reduction in T~ because government spending requirements can be met using a lower tax rate. Absorbing shocks mainly by changes in T ~ , as opposed to changes is T ~is ,efficient because capital is completely inelastic within a given period, clevelandfed.org/research/workpaper/index.cfm unlike labor supply.16 The shock absorbing feature of z, reinforces the variability of household investment, but allows the government to maintain a very smooth series for g,ly,, as compared to the other tax systems (see figure 2). Under the two simplified tax systems, the government prefers to absorb budget shocks mainly by increasing the variability of g, . Changes in g, do not affect household decisions due to the additively separable way in which g, enters the household utility function. The government maintains a low variability in z, to reduce fluctuations in household labor supply. Interestingly, the simplified tax systems do a better job of matching the relatively high standard deviation of public expenditures in U.S. data. In the complicated and lump-sum tax systems, the standard deviation of g, is much lower than the U.S. value. For these tax systems, it appears that an additional shock is needed to increase the variability of government spending.17 The labor tax (in the complicated system) and the uniform income tax (in both simplified systems) are procyclical (table 6b). The government uses these taxes to help smooth households' aftertax income from labor. For example, a positive technology shock is accompanied by increases in z, and 2,. This provides households with an implicit insurance mechanism against earnings variability, thus leading to lower standard deviations of household consumption and hours worked in the distortionary tax systems versus those in the lump-sum tax system (table 6a). The model's prediction that the capital tax should display more variability than the labor tax is consistent with the U.S. tax rate estimates we have chosen for comparison.18 The model disagrees with the data, however, in predicting a negative correlation between z, and z,. The U.S. tax rate estimates 1 6 ~ h optimality e of using a state-contingentcapital tax to absorb budget shocks has been shown previously by Judd (1989) and Chari, Christiano, and Kehoe (1993). Bohn (1988) shows how nominal public debt can be used to absorb budget shocks. 17~dding preference shocks to households' demand for public goods (to simulate wars, for example) would increase the variability of g, in these tax systems. See Lansing (1994a) for an optimal policy model with preference shocks. 18~igures3 and 4 display the tax rate series before detrending. For quantitative comparisons (table 6) detrending is necessary because the U.S.labor tax displays a distinct upward trend, while the U.S.capital tax displays a downward trend. These trends have no counterpart in the model. The trend in T, is possibly linked to the phenomenon of "bracket creep," which existed before tax schedules were indexed for inflation in 1985. Regarding the trend in T,, Auerbach and Poterba (1988) argue that the downward trend is due to increasingly generous investment tax credits and accelerated depreciation schedules. clevelandfed.org/research/workpaper/index.cfm display a positive correlation.19This suggests there may be rigidities in the U.S. tax code (not accounted for in the model) that loosely connect the movement of capital and labor taxes over the business cycle. Explicit modeling of such rigidities is a topic for our future research. Without a depreciation allowance, the simplified tax system displays wildly counterfactual behavior, predicting negative correlations with output for ct , ht , and 4 . This behavior can be traced to the higher variability of zt that results from the lack of a separate tax on capital. The variability of zt is such that movements in the tax rate can actually override the effects of technology shocks in determining the behavior of household consumption and inve~tment.~ For a given variance of the technology shock, output is more variable under the system of lumpsum taxes than under any of the distortionary systems (table 6a). This occurs because movements in the distortionary tax rates partially offset the effects of technology shocks, resulting in lower output variability. The lump-sum tax system does a reasonably good job of matching the relative standard deviations of hours and productivity, a feature typical of indivisible labor models (see Hansen 119851). In all three distortionary systems, however, the standard deviation of hours relative to productivity is too low, despite the presence of indivisible labor. The insurance features of the optimal tax rates are responsible for the reduced variability of hours in these systems. Similar to the case for government spending, our results suggest that an additional shock is needed in the distortionary tax models to increase the variability of hours worked. Such a specification would be consistent with the position of Aiyagari (1994), who uses a variance decomposition analysis to argue that the behavior of hours in U.S. data is driven by multiple shocks. 191n the U.S. data, the correlation coefficient between (detrended) z, and z, equals 0.41, using the z, series from Barro and Sahasakul (1986) and the z, series from Jorgenson and Sullivan (1981). The correlation is 0.52 using the z, and z, series from McGrattan, Rogerson, and Wright (1993). ''~xce~tfor the double taxation of dividends, the income tax with no depreciation allowance is equivalent to a production tax which takes the form j,= (l-z,)exp (z,) [k:' h P ] ' I X .A tax of this fonn can be modeled as a negative technology shock. See Abel and Blanchard (1983). clevelandfed.org/research/workpaper/index.cfm Table 6a: Business-Cvcle Statistics for the Various Tax Svstems Standard Deviation in Percent (all series detrended)a Series Complicated Simplified w/ Depr Allow GNP=y, 2.32 2.34 Tax Rates zh 4.12 z, 11.31 Simplified No Depr Allow U.S. ~conorn~~ Lump-sum 1.74. 2.96 2.46 5 5.65, 3.65 z 4.27 z 7.80 Tly 1.79 2, Tly 16.4,4.82 2.44 Table 6b: Business-Cvcle Statistics for the Various Tax Svstems (continued) Contemporaneous Correlation with Output (all series detrended)a Series Tax Rates Complicated z, 0.99 6.96 Simplified w/ Depr Allow z 0.99 Simplified No Depr Allow z 0.99 U.S. Lump-sum T l y 6.91 cono om^^ zh 0.37, 0.05 z, 0.10, 0.32 T l y 0.14 aModel statistics are means over 100 simulations, each 46 periods long, after dropping the first 50 periods. Parameter values are from table 2. All series were logged and detrended (except for tax rates which were only detrended) using the HodrickPrescott Nter (see Prescott [1986]). The filter smoothing parameter was set at A = 100 since all data are at annual frequency. b ~ . data ~ . cover the period 1947 to 1992. The following quarterly series from Citibase were annualized before computing statistics: y, = GNPQ, c, = GCDQ+GCNQ (nondurables and services), g, = GGEQ, h, = LHOURS (household survey), y, 1H , = GNPQLHOURS. The series for x,is fixed business investment + consumer durables expenditures +residential investment. The series for k, is fixed private capital + stock of consumer durables + residential capital. Both x, and kt are annual series from Fixed Reproducible Tangible Wealth in the United States, U.S. Department of Commerce (1993). Sources for U.S. tax rate data are described in the notes to table 3. All series were logged (except tax rates) and detrended as in the model. clevelandfed.org/research/workpaper/index.cfm 7. Concluding Remarks While our model is admittedly a very abstract and simplified representation of the vastly complex U.S. tax code, we believe that it can be useful for examining key questions about the institutional structure of our tax system. In our view, the strength of the model lies in the ability to capture the general-equilibrium effects of endogenous fiscal policy. Using this approach, we find that a uniform tax on all types of income yields almost the same welfare as a more complicated system with separate tax rates on labor and capital income. We also find that the depreciation allowance plays an important role in encouraging capital accumulation. In a related paper (see Guo and Lansing [1994]), we extend our examination of various tax structures to include an analysis of single versus double taxation of firm dividends. In general, we find that the assumed structure of the tax system can have profound effects on the characteristics of optimal fiscal policy and the resulting behavior of aggregate economic variables over the business cycle. Although our model abstracted from economic growth, we expect that tax structure would continue to have a significant impact in a model where growth is endogenous. Jones, Manuelli, and Rossi (1993a) employ an endogenous growth model to derive optimal fiscal policy with separate taxes on labor and capital, and no depreciation allowance. They also examine a case where tax rates on labor and human capital are restricted to be the same, but physical capital is taxed separately. In our current research, we are performing tax simplification experiments similar to those done here in the context of an endogenous growth model (see also Pecorino [I9931 and Stokey and Rebelo [1993]). Our work thus far agrees with the viewpoint of Slemrod (1990), who argues that selecting an optimal tax structure can often be more important than the setting of optimal tax rates. clevelandfed.org/research/workpaper/index.cfm APPENDIX A A.l Recursive Formulation of the Government's Problem To formulate the recursive version of the government's problem specified in equation (1 I), we first substitute the household first-order conditions (5) into the household budget constraint (2), the resource constraint (12), and the utility function U(.)to eliminate z,, z,, and c, .' The resource constraint (12) can then be used to substitute out g,. For the lump-sum tax system, we use the household first-order conditions to eliminate r,, w,, and c,; the government budget constraint to eliminate T, ; and the resource constraint to eliminate g, .2 The vector of state variables for the government's problem in period t is s,={z,, kt, h,,]. In the transformed problem, the government's decision variables are h , , h, , and k+,. Using primes (') to denote next-period quantities, the recursive version of the government's problem is shown in (A.1). The Bellman equation in (A.1) summarizes the recursive nature of the problem The first two constraints represent two versions of the household budget (after substituting in the first-order conditions), depending on whether distortionary or lump-sum taxes are used. The next constraint imposes the condition of equal tax rates on labor and capital for the simplified tax system. The symbols u,, and u, represent composite error terms that arise due to the presence of E, in the first-order condition for kt+,.The remaining constraints define the production technology, the factor prices r, and w, , and the law of motion for the exogenous technology shock. The dynamic programming problem applies for all t > 0. The problem at t = 0 must be considered separately, as shown by Kydland and Prescott (1980), Lucas and Stokey (1983), and Charnley (1986). At t = 0, the stock of private capital is fixed. Optimal policy thus implies a high initial tax on capital to take full advantage of this nondistortionary source of revenue. We assume that this form of lump-sum taxation is insufficient to finance the entire stream of future expenditures. The analysis here will focus ., ,+, 'Due to the presence of the expectation operator in the first-order conditions for k the substitution is accomplished using = f, ( - ) - u,, where f ,(.) is a function of random variables and ,u, is the forecast error. The assumption the expression E f of rational expectations implies E ,.,u, = 0. ,(a) 2 ~ u to e the presence of monopoly profits in our model, the competitive equilibrium is not Pareto optimal. Therefore, the government's problem with lump-sum taxes must be solved in the same manner as the problem with distortionary taxes. clevelandfed.org/research/workpaper/index.cfm on policy in stationary stochastic equilibrium, i.e., when t approaches infinity. The linear-quadratic approximation method used to solve (A.l) is accurate only in the neighborhood of the deterministic steady-state. Consequently, we do not solve the t=O problem or compute the transition path to the stationary equilibrium. v(s) = s = { z , k, h-l } U(.) = In(l/h) - A h where m a . E,{ U ( .) subject to A h + -kA-1 - h + r-$6 m wh Bln - 1 -kt+u, = O (l-01-02)y A h + I h - ' -1+6]k A + 2 flh -x- fl [v(st)I s ] } + k', h . 5 r-$6 x (for the distortionary tax systems) - (0,+0,)y + u l = 0 (fortax the lump-sum system) u Z = O (for the simplified tax system only) Equilibrium is defined as a value function V(s) and an associated set of stationary decision rules that satisfy (A.l). The decision rules dictate a set of household allocations and prices at time t that can clevelandfed.org/research/workpaper/index.cfm be implemented by means of the government's chosen policy. The government's explicit policy rules for tax rates and public expenditures are recovered by substituting the implementable allocations and prices into the applicable first-order conditions and budget constraints. A.2 Computation Procedure The dynamic programming problem in (A. 1) is solved numerically using a variant of the linearquadratic approximation technique first used by Kydland and Prescott (1982). An approximate version of (A.l) is obtained by first substituting as many nonlinear constraints as possible into the return function U(.).For constraints that cannot be substituted in, a Lagrange multiplier is used to incorporate the constraint into a redefined return function, and the multiplier is treated as an additional decision variable. A quadratic approximation of the return function is then computed in terms of the logarithms of all variable^.^ The solution algorithm exploits the certainty equivalence property of linear-quadratic control problems. The optimal decision rules for the approximated economy can be obtained by solving the deterministic version of the model4 An initial guess Vo is made for the optimal value function V(s) in the quadratic version of (A.l). Sequential candidate value functions Vi are then computed by successively iterating on the Bellman equation until the value function has converged. We verified that the algorithm always converged to the same value function regardless of the initial guess for Vo. Once the process has converged, log-linear decision rules that dictate household equilibrium allocations are computed. Log-linear policy rules for z, , z, (or 7,) and g, can then be computed using the household first-order conditions and the budget constraints, log-linearized around the steady state. 3 ~ h log-linear e version of the Kydland-Prescott method is described in Chrktiano (1988). 4 See Sargent (1987, p. 36). Specifically, the stochastic terms E , , u,, and u, are set equal to their unconditional means (zero) in the numerical algorithm. With a quadratic objective, the first-order conditions are linear in all variables. This allows the expectation operator in (A.l) to be passed through the expressions, dropping out any stochastic terms. clevelandfed.org/research/workpaper/index.cfm FIG 1: SIMULATED SERIES FOR TAX RATES 0.70 0.65 - lncorne Tow Rate (simplified system, w/ depr ollow) 7 0.60 - I 0.55 - 0.50 - 0.45 - +T , , *- 7 - Labor Tow Rote (complicated system) - lncorne Tow Rate (simplified system, no depr ollow) 0 [L x 0.25 \ b-a 0.10 0 4 8 12 16 20 24 28 32 36 40 . 44 48 Year FIG 2: U.S. & SIMULATED SERIES FOR GOVT EXPENDITURES (fraction of GNP) 0.30 I I I - g/y g/y g/y I I (lump-sum system) (complicoted system) (simplified system. w/ depr allow) ha' 0.10 0 4 8 12 16 20 24 Year 28 32 36 40 44 48 clevelandfed.org/research/workpaper/index.cfm FIG 3: U.S. SERIES FOR CAPITAL TAX RATES - 7k -7, - 1952 1957 1962 1967 Capital Tax Rate from McGrattan et 01. (1993) Copital Tax Rate from Jorgenson & Sullivan (1981) 1972 1977 1982 - 1987 1992 1987 1992 Year FIG 4: 1947 1952 1957 U.S. SERIES FOR AVERAGE AND LABOR TAX RATES 1962 1967 1972 Year 1977 1982 clevelandfed.org/research/workpaper/index.cfm REFERENCES Abel, Andrew B. and Olivier Blanchard. (1983). "An Intertemporal Model of Savings and Investment," Econometrica, 5 1,675-692. Aiyagari, S. Rao. (1994). "On the Contribution of Technology Shocks to Business Cycles," Federal Reserve Bank of M i ~ e a p ~ l iQuarterly s, Review, Winter, 22-34. Altig, David and Charles T. Carlstrom. (1991). "Inflation, Personal Taxes, and Real Output: A Dynamic Analysis," Journal of Money, Credit and Banking, 23, 545-571. Auerbach, Alan J. and James M. Poterba. (1988). "Why Have Corporate Tax Revenues Declined?" in E. Helpman, A. Razin, and E. Sadka, eds., Economic Effects of the Government Budget. Cambridge, MA: MIT Press. Barro, Robert J. and Chaipat Sahasakul. (1986). "Average Marginal Tax Rates from Social Security and the Individual Income Tax," Journal of Business, 59, 555-566. Benhabib, Jess and Roger E.A. Farmer. (1994). "Indeterminacy and Increasing Returns," Journal of Economic Theory, 63. 19-41. Benhabib, Jess and Boyan Jovanovic. (1991). "Externalities and Growth Accounting," American Economic Review, 81, 82-113. Bohn, Henning. (1988). "Why Do We Have Nominal Government Debt?" Journal of Monetary Economics, 21,127-140. Braun, R. Anton and EUen R. McGrattan. (1993). "The Macroeconomics of War and Peace," in Macroeconomics Annual, National Bureau of Economic Research, forthcoming. Chamley, Christophe P. (1986). "Optimal Taxation of Income in General-Equilibrium with Infinite Lives," Econometrica, 54, 607-622. Chari, V.V., Lawrence J. Christiano, and Patrick J. Kehoe. (1993). "Optimal Fiscal Policy in a Business Cycle Model," National Bureau of Economic Research, Working Paper No. 4490. Christiano, Lawrence J. (1988). "Why Does Inventory Investment Fluctuate So Much?" Journal of Monetary Economics, 21, 247-280. Christiano, Lawrence J. and Martin Eichenbaum. (1992). "Current Real-Business-Cycle Theories and Aggregate Labor-Market Fluctuations," American Economic Review, 82, 430-450. Cooley, Thomas F. and Gary D. Hansen. (1991). "The Welfare Costs of Moderate Inflation," Journal of Money, Credit, and Banking, 23,483-497. . (1992). "Tax Distortions in a Neoclassical Monetary Economy," Journal of Economic Theory, 58, 290-3 16. Farmer, Roger E.A. and Jang-Ting Guo. (1994). "Real Business Cycles and the Animal Spirits Hypothesis," Journal of Economic Theory, 63,42-72. Guo, Jang-Ting and Kevin J. Lansing. (1994). "Tax Structure, Welfare, and the Stability of Equilibrium in a Model of Dynamic Optimal Fiscal Policy, "Federal Reserve Bank of Cleveland, Working Paper 9410. Hansen, Gary D. (1985). "Indivisible Labor and the Business Cycle," Journal of Monetary Economics, 16, 309-327. Jones, Larry E., Rodolfo ManueUi, and Peter E. Rossi. (1993a). "Optimal Taxation in Models of Endogenous Growth," Journal of Political Economy, 101,485-517. .(1993b). "On the Optimal Taxation of Capital Income," National Bureau of Economic Research, Working Paper No. 4525. clevelandfed.org/research/workpaper/index.cfm Jorgenson, Dale and Martin A. Sullivan. (1981). "Inflation and Corporate Capital Recovery," in C.R. Hulten, ed., Depreciation, Inflation, and the Taxation of Income from Capital. Washington, D.C.: Urban Institute Press. Judd, Kenneth L. (1985). "Redistributive Taxation in a Simple Perfect Foresight Model," Journal of Public Economics, 28, 59-83. . (1989). "Optimal Taxation in Dynamic Stochastic Economies: Theory and Evidence," Hoover Institution, reproduced. Juster, F. Thomas and Frank P. Stafford. (1991). "The Allocation of Time: Empirical Findings, Behavioral Models, and Problems of Measurement," Journal of Economic Literature, 29, 471-522. Kydland, Finn E. and Edward C. Prescott. (1980). "Dynamic Optimal Taxation, Rational Expectations, and Optimal Control," Journal of Economic Dynamics and Control, 2, 79-91. . (1982). "Time to Build and Aggregate Fluctuations," Econometrica, 50, 1345-1370. Lansing, Kevin J. (1994a). "Optimal Fiscal Policy When Public Capital is Productive: A Business Cycle Perspective," Federal Reserve Bank of Cleveland, Working Paper 9406. . (1994b). "OptimalFiscal and Monetary Policy in a Business Cycle Model with Heterogenous Agents," Federal Reserve Bank of Cleveland, Working Paper (forthcoming). Lucas, Robert E., Jr. and Nancy L. Stokey. (1983). "Optimal Fiscal and Monetary Policy in an Economy without Capital," Journal of Monetary Economics, 12, 55-94. McDermed, Ann A., Robert L. Clark, and Steven G. Allen. (1989). "Pension Wealth, Age-Wealth Profiles, and the Distribution of Net Worth," in R.E. Lipsey and H. Stone Tice, eds., The Measurement of Savings, Investment, and Wealth. Chicago: University of Chicago Press. McGrattan, Ellen R., Richard Rogerson, and Randall Wright. (1993). "Household Production and Taxation in the Stochastic Growth Model," Federal Reserve Bank of Minneapolis, Staff Report No. 166. Pecorino, Paul. (1993). "Tax Structure and Growth in a Model with Human Capital," Journal of Public Economics, 52, 251-271. Prescott, Edward C. (1986). "Theory ahead of Business Cycle Measurement," Federal Reserve Bank of Minneapolis, Quarterly Review, Fall, 9-22. Ramsey, F.P. (1927). "A Contribution to the Theory of Taxation," Economic Journal, 37,47-61. Rogerson, Richard. (1988). "Indivisible Labor, Lotteries, and Equilibrium," Journal of Monetary Economics, 21, 3-16. Rosen, Harvey S. (1992). Public Finance. Boston: Irwin, Inc. Sargent, Thomas J. (1987). Dynamic Macroeconomic Theory. Cambridge, MA: Harvard University Press. Slemrod, Joel. (1990). "Optimal Taxation and Optimal Tax Systems," Journal of Economic Perspectives, 4, 157-178. . (1991). Do Taxes Matter? The Impact of the Tax Reform Act of 1986. Cambridge, MA: MIT Press. Stokey, Nancy L. and Sergio Rebelo. (1993). "Growth Effects of Flat-Rate Taxes," National Bureau of Economic Research, Working Paper No. 4426. Zhu, Xiaodong. (1992). "Optimal Fiscal Policy in a Stochastic Growth Model," Journal of Economic Theory, 58, 250-289.