The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Vol. 30, No. 4 ECONOMIC REVIEW 1994 Quarter 4 Tax Structure, Optimal Fiscal Policy, and the Business Cycle 2 by Jang-Ting Guo and Kevin J. Lansing Cross-Lender Variation in Home Mortgage Lending 15 by Robert B. Avery, Patricia E. Beeson, and Mark S. Sniderman The Efficiency and Welfare Effects of Tax Reform: Are Fewer Tax Brackets Better than More? by David Altig and Charles T. Carlstrom FEDERAL RESERVE BANK OF CLEVELAND 30 ■ ECONOMI C REVIEW 1994 Quarter 4 Vol. 30, No. 4 Tax Structure, Optimal Fiscal Policy, and the Business Cycle 2 by Jang-Ting Guo and Kevin J. Lansing The real business cycle (RBC) approach to the study of aggregate fluctua tions is now a well-established paradigm in macroeconomics. Most RBC models abstract from government fiscal policy altogether or treat it as some exogenous stochastic process. This article develops an RBC model in which government fiscal variables such as tax rates and public expendi tures are endogenous. The authors characterize the "optimal” behavior of fiscal policy over the business cycle for two different tax structures and re late this behavior to movements in private-sector variables like output, con sumption, labor hours, and Investment. As a benchmark, they also provide a comparison between the model and U.S. data. Cross-Lender Variation in Home Mortgage Lending 15 by Robert B. Avery, Patricia E. Beeson, and M arkS . Sniderman This study evaluates the feasibility of using information collected under the Home Mortgage Disclosure Act (HMDA) to form quantitative measures of lender activity for use in enforcement. By evaluating three firm-level meas ures— loan application, approval, and origination rates— the authors find that cross-lender variation in minority and low-income originations primar ily reflects differences in home mortgage application rates, not in approval rates. The authors also compare gross measures of lender performance with indices controlling for property location and loan applicant charac teristics and determine that they perform similarly. This suggests that most of the variation in lender behavior is idiosyncratic and cannot be attributed to variance in applicant characteristics reported in the HMDA data or to dif ferences in the geographic markets served by the lenders. The Efficiency and Welfare Effects of Tax Reform: Are Fewer Tax Brackets Better than More? 30 by David Altig and Charles T. Carlstrom On the wish list of many members of the new Congress is an income tax system characterized by constant marginal tax rates, typically referred to as a flat-tax system. In reality, what we are likely to see Is a continuation of the worldwide trend toward replacing systems with high marginal-rate progressive with those that have a smaller number of rates that are flat over relatively broad in come ranges. In this article, the authors compare a simple two-bracket tax code with an approximation to traditional structures that entail steeply rising marginal tax rates. Their conclusion — that the simpler rate structures are not necessarily more efficient than alternatives with numerous, highly progressive brackets — serves as a cautionary note to potential reformers. Economic Review is published quarterly by the Research Depart ment of the Federal Reserve Bank of Cleveland. Copies of the Review are available through our Corporate Communications and Community Affairs Department. Call 1-800-5433489, then immediately key in 1-5-3 on your touch-tone phone to reach the publication request option. If you prefer to fax your order, the number is 216-579-2477. Coordinating Economist: Jagadeesh Gokhale Advisory Board: Charles T. Carlstrom Joseph G. Haubrich Peter Rupert Editors: Tess Ferg Robin Ratliff Design: Michael Galka Typography: Liz Hanna Opinions stated in Economic Re view are those of the authors and not necessarily those of the Fed eral Reserve Bank of Cleveland or of the Board of Governors of the Federal Reserve System. Material may be reprinted pro vided that the source is credited. Please send copies of reprinted material to the editors. ISSN 0013-0281 2 Tax Structure, Optimal Fiscal Policy, and the Business Cycle by Jang-Ting Guo and Kevin J. Lansing Introduction The real business cycle (RBC) approach to the study of aggregate fluctuations is now a wellestablished paradigm in macroeconomics. The early groundbreaking articles in this area (Kydland and Prescott [1980a, 1982] and Long and Plosser 119831) completely abstracted from government behavior, yet were reasonably suc cessful in capturing the broad comovements and relative variabilities of the economic aggre gates that characterize the business cycle. More recently, researchers have introduced elements of government fiscal policy into these models to help match various business cycle facts. For example, Christiano and Eichenbaum (1992) include stochastic government spending in the household utility function to help explain the low observed correlation between labor hours and real wages (as measured by average labor productivity) in postwar U.S. data. This works in their model because shocks to gov ernment spending impact the marginal utility of private consumption and thereby induce shifts in the household labor supply. These interact with labor demand shifts (caused by technology shocks) to produce a low correlation between wages and hours. Braun (1994) and Jang-Ting guo is an assistant profes sor of economics at the University of California, Riverside, and Kevin J. Lansing is an economist at the Fed eral Reserve Bank of Cleveland. For helpful comments and suggestions, the authors thank David Altig, Charles T. Carlstrom, Jagadeesh Gokhale, Joseph G.Haubrich, Finn Kydland, Peter Rupert, and an anonymous referee. McGrattan (1994) show that a similar result can be obtained by introducing stochastic distortionary taxes to shift the labor supply curve. A common feature of these studies is that govern ment policy is viewed as exogenous. In this paper, we develop an RBC model in which government fiscal variables such as tax rates and public expenditures are endogenous. Our objective is to characterize the “optimal” behavior of these policy variables over the business cycle and to relate this behavior to movements in private-sector variables like out put, consumption, labor hours, and investment. As a benchmark, we also provide a comparison between the model and U.S. data. We build on the recent work of Chari, Chris tiano, and Kehoe (1994), who develop a com petitive RBC model in which a government policymaker chooses an optimal sequence of distortionary taxes on labor and capital income in a dynamic version of the Ramsey (1927) optimal tax problem. Our model differs from theirs in three main respects. First, we introduce monopoly profits into the production sector of the economy such that the optimal steady-state tax on capital is positive, consistent with U.S. observations. In a competitive model, this tax rate is zero (see Judd [19851 and Chamley [19861). Second, our model 3 incorporates the "indivisible labor” specification of Rogerson (1988) and Hansen (1985). In standard RBC models (which abstract from government), the indivisible labor specification serves to increase the variability of hours rela tive to the real wage to a value that is more in line with U.S. data. Third, we endogenize the time series of government spending by includ ing in household preferences a separable term that represents the utility provided by public goods. In the Chari, Christiano, and Kehoe model, government spending follows an exoge nous stochastic process. We compare simulations from our model to post-WWII, annual U.S. data and an otherwise similar model with nondistortionary lump-sum taxes. Our results can be summarized as follows: For a given stochastic process of the technology shock, we find that optimal distortionary taxes reduce the variability of output and labor hours, but increase the relative variability of household investment, compared to the model with lump sum taxes. This result can be traced to the behav ior of the optimal distortionary tax rates on labor and capital income. The optimal labor tax in the model is procyclical, which reduces the variability of hours (and output) by providing households with an implicit insurance mechanism against variations in their after-tax wage. The optimal capital tax in the model is countercyclical and dis plays a high standard deviation relative to the labor tax. This tends to increase the variability of household investment relative to output, but pro vides an efficient means of absorbing shocks to the government’s budget (which are caused by changes in the size of the tax base over the busi ness cycle). In addition, we find that the distortionary tax model underpredicts the variability of hours worked relative to the real wage in U.S. data, de spite our specification of indivisible labor. This result is due to the procyclical optimal labor tax, which tends to reduce the variability of hours worked in comparison to a standard RBC model with indivisible labor. Finally, both mcxlel versions capture the procyclical behavior of government spending in annual U.S. data, but underpredict its variability over the business cycle. We find that this comparison, as well as comparisons along some other dimensions, are substantially improved if we exclude U.S. data prior to 1954 to avoid the influence of the Korean War. However, a few comparisons, such as the correlation between gov ernment spending and output, become worse. The remainder of the paper is organized as follows: Sections I and II describe the model and the solution method. The choice of parameter values is discussed in section III. Section IV examines the business cycle characteristics of the two tax structures and compares them to U.S. data. Concluding remarks are presented in section V. I. The Model The model economy consists of three types of agents: households, firms, and the govern ment. Households obtain direct utility from government-provided public goods, which are financed by taxes on households and firms. Following Benhabib and Farmer (1994), we postulate that firms w hich produce intermedi ate goods exhibit some degree of monopoly power such that they realize positive economic profits even though the final-goods sector of the economy is perfectly competitive. The prof its are equal to the difference between the value of output and the payments made to in puts. The reason for introducing profits is to obtain a positive optimal tax rate on capital un der the distortionary tax structure, consistent with U.S. observations.1 As owners of the firms, households receive net profits in the form of dividends. It is as sumed 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. Further more, under the distortionary tax structure, we assume that the government can distinguish between labor and capital income, but cannot distinguish between the various categories of capital income, such as profits, dividends, bond interest, and capital rental income. Therefore, this version of the model includes only two types of distortionary taxes: a labor tax and a capital tax. ■ 1 Jones, Manuelli, and Rossi (1993) show that the existence of profits and a restriction on the menu of available tax instruments (the ab sence of a separate profits tax) is one method of obtaining a positive opti mal tax rate on capital in the steady state. Without profits, the optimal steady-state tax on capital is zero. The Household’s Problem There is a continuum of identical, infinitely lived households, each of which maximizes a stream of discounted utilities over sequences of consumption and leisure: (1) max •fi’o X ^ cr br -h,+■ t= o ^nct ~ A bt +B\ngt) 0 < |3< 1, A, B > 0 . (3) In this utility function, (3 is the household dis count factor and ct represents private consump tion goods. The symbol Et is the expectation operator conditional on information available at time t. Each household is endowed with one unit of time each period and works ht hours dur ing period t. The fact that utility is linear in hours worked draws on the formulation of indivisible labor described by Rogerson (1988) and Hansen (1985). This means that all fluctuations in labor hours are due to changes in the number of work ers employed, as opposed to variations in hours per worker.2 Household preferences also include a term representing the utility provided by aggre gate public consumption goods gt. The separabil ity in ct and gt implies that public consumption does not affect the marginal utility of private con sumption, a specification supported by parameter estimates in McGrattan, Rogerson, and Wright (1993). Households view gt as outside their con trol. Examples of public consumption goods that might affect household utility are national de fense, police protection, and government provi sion of food and shelter during natural disasters. Public goods are assumed to be noncongestable and free of specific user charges. The representative household faces the fol lowing within-period budget constraint: (2) kt+1 = (1 - 5) kt+ x t , 0 < 8 <1 . Households view tax rates, wages, interest rates, and dividends as determined outside their control. Household Optimality The household first-order conditions with respect to the indicated variables and the associated transversality conditions (TVC) are (4a) c ■X = — ' c, (4b) b,: (4 0 kt+, : X,(\-\b,)W = A x, = p £ , x (+1 [(1 - T 41+ i ) ( ri+1 - 8) + 1] (4d) b X =M BE u t + i •■ 'W ^ t ^X t + 1 ~ 1 k t+ 1 ) r b t+ 1 + (4e) TVC: lim E 0 t— kt+1 = 0, lim E 0 $ fX tbt+1 = 0 , t— c t+ x t + b t+1< ( l -\ht)w t ht + (1-T kt)(rtk,+ the firm’s net profits, Kt (which are distributed to households as dividends), and the interest earned on government bonds, rhtbt. Dividends and interest are taxed at the same rate as capital rental income, rtkt. The term Tktdkt represents the depreciation allowance built into the U.S. tax code, and Tt is a lump-sum tax. The following equation describes the law of motion for the capital stock, given a constant rate of depreciation 8: Kt + rbtbt) + Tktdkt + bt- Tt, where X, is the Lagrange multiplier associated with the budget constraint (2) in period t. The interpretation of X t is that it represents the k 0 ,b 0 given, w'here xt is investment, kt is the stock of physi cal capital, and bt+l represents one-period, real government bonds carried into period t + 1 by the household. Households derive income by supplying labor and capital services to firms at rental rates u\ and rr and pay taxes on labor and capital income at rates xht and xkt, respectively. Two additional sources of household income are ■ 2 The linearity of (1) in h t implies that the effective labor-supply elas ticity of the representative household is infinite. In a decentralized economy, both Rogerson and Hansen show that this utility function can be supported by a lottery that randomly assigns workers to employment or unemployment each period, with the firm providing full unemployment insurance. Wage contracts call for households to be paid based on their expected (rather than actual) number of hours worked. RBC models with indivisible labor are better able to match some key characteristics of aggregate labor market data. Spe cifically, U.S. data display a large variability of hours worked relative to the real wage, and a weak correlation between hours and the real wage (see Christiano and Eichenbaum [1992]). 5 marginal utility of an additional unit of after-tax in come received in period t. The transversality con ditions ensure that the household’s within-period budget constraint (2) can be transformed into an infinite-horizon, present-value budget constraint. The Firm’s Problem This section closely follows the model devel oped by Benhabib and Farmer (1994). Suppose there exists a continuum of intermediate goods yit, i e [0,11 and a unique final good yt that is produced using the following constant-returnsto-scale technology: (8c) Due to their monopoly power, intermediategoods producers earn an economic profit that is taxed at rate xkt. The firm’s after-tax profits, distributed to households in the form of divi dends, are (9a) nt= ( l - x b )(yt- rt k t - w t ht). (9b) 7t,= (1 - xkt) (1 - 01 - 02 )yt. The Government’s Problem (5 ) yt = y ^di We assume that the final-goods sector is per fectly competitive, but that intermediate-goods producers exert a degree of monopoly powrer that is captured by the parameter x In the spe cial case when x = 1, all intermediate goods are perfect substitutes in the production of the final good, and the intermediate sector becomes per fectly competitive. Each intermediate good is produced using the same technology, with labor and capital as inputs: (6) yit = e x p (zt) k ? h ° 2, 0 < a . < 1, a 2 + a , = 1 (7) zt+i - Pzzt +Et+v 0 < Q Z< 1 , Et ~ i.i.d. (0, Oe2), zQ given, where (7) is the law of motion for aggregate tech nology shocks zt, which are revealed to agents at the beginning of period t and which generate business cycle fluctuations in the model. Under the assumptions that firms maximize profits and factor markets are competitive, Benhabib and Farmer show that in a symmetric equilibrium (kit = kt and hit= ht for all /), the aggregate pro duction function, the rental rate on capital, and the real wage are (8a) yf = exp (zt) kta\ htaz, (8b) r/=ei j ’ 9i =%a i -and yt The government chooses an optimal program of taxes, borrowing, and public expenditures in order to maximize the discounted utility of the household. This is a dynamic version of the Ramsey (1927) optimal tax problem, where Tht, xkt, rht, gt , and Tt summarize government pol icy implemented at time t. To set up this prob lem, we begin by spelling out some important assumptions. First, we assume that the govern ment can com m it to a set of time-invariant de cision rules that specify policy variables as a function of state variables. This is done to avoid the complicating issue of time inconsistency, which arises in policy design problems when the tax base includes fixed assets (such as capi tal or bonds) that cannot be quickly adjusted in response to a change in the level of the tax. In these situations, the government has an in centive to deviate from its originally announced, optimal policy by implementing surprise in creases in asset taxes in order to obtain nondistortionary tax revenue. Because households understand that the original policy is time incon sistent, it cannot be supported as an equilibrium unless the government can commit itself (and all successor governments) to carrying out the plan.3 Second, given that the initial stocks of capital and bonds are fixed, we rule out any confisca tory taxes on assets at t = 0 that might be used to finance all future expenditures. This case is not very interesting because no taxes beyond the initial period are required. With these assump tions, the government’s problem is ■ 3 The time inconsistency problem does not arise under the lump sum tax structure because taxes are nondistortionary. See Chari (1994) for a summary of the issues and a review of the literature dealing with time inconsistency problems and optimal policy design. 6 (10) max £ 0 X P' {\nct - A h t +B\ngt) 8,’ h n h , 1=0 subject to (i) household first-order conditions and the budget constraint, (11) firm profit-maximization conditions, (iii) g t = b t + l - b t(l + rht) + x hlw tb t + x kt[(rt- S ) k t+ r htb t] +[1 —(1 — x kt) 2] (1 —0j —02)j^-i- Tt, (iv) 7’ = 0 for the distortionary tax structure, (y ) xht = xkt = 0 for the lump-sum tax structure, (vi) lim —------------------ = 0. bt n n + (i /■ =i )ri«-i i As a condition for equilibrium, government policy must take into account the rational re sponses of households and firms, as summarized by constraints (i) and (ii). The chosen policy must also satisfy the government’s within-period budget constraint (iii), where the squared term on the right-hand side reflects the double taxa tion of firm dividends. Constraints (iv) and (v) impose the restrictions associated with the two tax structures wre intend to analyze. For the dis tortionary tax structure, we rule out the possi bility of lump-sum taxes. For the lump-sum tax structure, we set xht and X kt equal to zero. Fi nally, (vi) is a transversality condition ensuring that the government budget constraint is satis fied in present-value temis. The summation of the household budget constraint and the gov ernment budget constraint yields the following resource constraint for the economy: (11) yt = c t + xt + gr Because the resource constraint and the govern ment budget constraint are not independent equations, equation (11) will be used in place of (iii) in solving the government’s problem. II. Solving the Model Our approach to solving the government’s prob lem is to find the allocations ct, ht, kt +v and bt+l that maximize household utility subject to the constraints, where allocations are ex pressed as functions of the economy’s state variables. The appropriate set of prices rt and wt and the policy variables x ht, X kt, rht, gt , and Tt that decentralize the optimal allocations can be computed using the profit-maximization conditions (8), the household first-order condi tions (4), the household budget constraint (2), and the resource constraint ( l l ) . 4 For example, the optimal allocations uniquely determine X, and wt through equations (4a) and (8c). Given Xt and wt, the household’s first-order condition for ht, equation (4b), uniquely determines the government’s optimal choice for x ht. The gov ernment has much more flexibility, however, in choosing the optimal capital tax and the optimal interest rate on government debt. The expecta tion operators in the household’s first-order conditions for kt+1 and bt+1, equations (4c) and (4d), imply that the after-tax returns on capital and bonds (weighted by marginal utility) must be the same “on average.” In response to a se ries of shocks, the government can satisfy this ex ante arbitrage condition and implement the optimal allocations using many different combi nations for the period-by-period values of x kt and rht. Consequently, the stochastic version of the model does not uniquely pin down the time-series behavior of these policy variables (see Zhu [1992] and Chari, Christiano, and Kehoe [1994] for a more complete description). To facilitate a comparison with U.S. data, w7e make a particular assumption about the way in which the government picks x kt and rht to decentralize a set of allocations. Specifically, we employ the certainty versions of (4c) and (4d) to identify xkt and rht each period. Requiring the government to satisfy the certainty versions of these constraints guarantees that the uncer tainty versions will also be satisfied. Essentially, we are restricting the policy instruments available to the government by ruling out fully flexible, state-contingent capital taxes and bond interest rates. This might be interpreted as reflecting the political infeasibility of some types of policy re gimes. The restriction we impose has an impact on the behavior of the allocations in response to stochastic shocks, as does any other restriction ■ 4 This method of solving the government’s problem is described as the “primal" approach by Atkinson and Stiglitz (1980), chapter 12. on the set of available policy instalments (such as ailing out lump-sum taxes). Consequently, the allo cations we compute in response to shocks are dif ferent from the “Ramsey allocations” that could be supported in an unrestricted environment.^ In the restricted case, the government’s decision rules for Tkt and rht are identical to those for an econ omy with no uncertainty. It is important to note that this result follows from a particular decentral ization scheme. However, our solution method also employs a linear-quadratic approximation of the problem. Thus, the decision rules governing household allocations also display the property of certainty equivalence.6 Given these assumptions, the government’s problem with distortionary taxes can be solved using a recursive algorithm developed by Kydland and Prescott (1980b). The problem with lump-sum taxes can be solved by adopting the view of a social planner for an appropriately defined “pseudo-economy” in which the plan ner cannot exploit the monopoly power of firnis. The government’s problem under lum p sum taxes is not equivalent to a standard social planning problem because when X < b the decentralized equilibrium is not Pareto optimal. The pseudo-economy approach is an indirect method of obtaining the equilibrium allocations (see Stokey and Lucas [19891, chapter 18). Gov ernment debt does not appear in the pseudo planner’s problem. This reflects the wellknown “Ricardian proposition,” which states that government debt policy is irrelevant to the determination of equilibrium allocations in an economy with lump-sum taxes (see Sargent [19871, chapter 3). Since debt doesn't matter in this case, we arbitrarily set it equal to zero each period such that g t = Tr The pseudo planner’s problem is recursive and can be solved using standard methods. III. Calibrating the Model to the U.S. Economy To explore the quantitative predictions of the model, we assign parameter values based on empirically observed features of post-WWII U.S. data. The time period in the model is taken to be one year, which is consistent with both the time frame of most government fiscal decisions and the frequency of available data on average marginal tax rates. The discount factor P (= 0.962) implies an annual rate of time preference of 4 percent. The parameter A in the household utility function is chosen such that the fraction of time spent working is close to 0.3 in the steady state for each tax structure. This coincides with time-use studies, such as Juster and Stafford (1991), which indicate that households spend approximately one-third of their discretionary time in market work. The value of B is chosen to yield a steady-state value of g/y near 0.22 for each tax staicture, the average ratio of government spending to GNP for the U.S. economy from 1947 to 1992.8 The steady-state level of government debt is chosen to yield a steady-state ratio of b/y equal to 0.45. This is the average value of U.S. federal debt held by the public as a fraction of GNP from 1947 to 1992.9 The exponents in the Cobb-Douglas pro duction function are chosen on the basis of two criteria. First, the selected values of 0, (= 0.31) and 0, (= 0.60) are in the range of the estimated shares of GNP received by capital and labor in the U.S. economy (see Christiano [1988]). Second, the model’s share of output devoted to monopoly profits (= 1 - 0 ( —0, ) is chosen to yield a reasonable value for the steady-state tax on capital (Tk) under the distor tionary tax structure. Because a separate profits tax is not available in this case, the government uses the tax on private capital to recapture a portion of the profits. In the model, the steadystate ratio of profits to output is 0.09, and the ■ 5 See Chari, Christiano, and Kehoe (1994) for examples of decentrali zations that support the Ramsey allocations. See Cassou (forthcoming) for a case where policy instruments are restricted to follow a univariate Markov process in response to government spending shocks. ■ 6 The approximate version of the problem involves the maximiza tion of a quadratic objective function subject to linear constraints. Since the first-order conditions are linear in all variables, the expectation opera tor can be passed through the expressions, dropping out stochastic terms associated with the technology shock innovation e /+1 in equation (7). See Sargent (1987), p. 36. ■ 7 A technical appendix to this paper, available from the authors upon request, describes the details of our solution procedure. ■ 8 The specific parameter values used in the computations are A = 2.50, B = 0.350 for the distortionary tax structure, and A = 3.48, 0=0.381 for the lump-sum tax structure. ■ 9 The model does not pin down a unique value for the steady-state level of government debt (see Chamley [1985]). Rather, steady-state debt is a function of the initial level of debt, b Q, and the entire transition path of taxes and spending from t= 0 until the steady state is reached. As an alternative to performing this difficult computation, we follow the ap proach of Lucas (1990) and simply choose the level of steady-state debt to reflect a debt-to-GNP ratio consistent with the data. We assume that b Q and the transition path are set such that the government budget constraint is satisfied in present-value terms. Data on U.S. federal debt held by the public are from Federal Debt and Interest Costs, Congressional Budget Office, 1993, table A-2. 8 resulting steady-state tax on capital is 0.31. This value of xkapproximates the average effective corporate tax rate in the United States from 1947 to 1980, as estimated by Jorgenson and Sullivan (1981).10 The steady-state tax on labor (xh) turns out to be 0.25. This is close to the average marginal tax rate on labor income from 1947 to 1983, as estimated by Barro and Sahasakul (1986). The U.S. tax rate estimates can be viewed as summarizing the various ele ments of the tax code that impact the behavior of agents. These include not only the statutory rate, but also the many types of exemptions, deductions, credits, and allowances. The monopoly power parameter % is chosen such that the aggregate production technology demonstrates constant returns to scale. Given the values chosen for 0j and 0 ,, a value of X = 0.91 yields yt = exp (z t) kt 0i4 ht 066. The capital depreciation rate 8 (= 0.07) is consistent with values commonly used in the RBC litera ture. Together with the values of (3 and 0,, this depreciation rate implies a steady-state ratio of capital to output ranging from 2.4 (under the distortionary tax structure) to 2.8 (under the lump-sum tax structure), and a ratio of invest ment to output ranging from 0.17 to 0.20. The corresponding average ratios for the U.S. econ omy from 1947 to 1992 are 2.58 and 0.21. The process governing technology shocks is esti mated 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.11 The parameter esti mates, p„= 0.85 and cre = 0.015, are close to those estimated by other studies using annual data, such as Benhabib and Jovanovic (1991). IV. Simulation Results In this section, we describe the model’s predic tions for the behavior of fiscal policy over the business cycle and provide a comparison with U.S. data. The simulation results are shown in ■ 10 Higher prolit levels imply a higher steady-state tax on capital in our model. When profits are zero (01 + 0 2 = 1), the optimal steady-state tax on capital is zero. If a separate profits tax were available, the govern ment would choose to tax profits at 100 percent and other capital income at 0 percent in the steady state (see footnote 1). ■ 11 The production function residual was measured as zt = InGNPt0.34ln/r?-0.66ln/7r The private capital stock k, is defined as fixed private capital + stock of consumer durables + residential capital from Fixed Repro ducible Tangible Wealth in the United States, U.S. Department of Commerce, 1993. Real GNP and the labor input (ht = LH0URS) are from Citibase. tables 1-2 and figures 1-5. Note that the distor tionary tax structure makes predictions for a larger set of variables than does the lump-sum tax structure. The additional variables are the stock of real government debt, bt, and the average marginal tax rates on labor and capital income, xht and xkru In comparison to the full sample of U.S. data from 1947 to 1992, both tax structures underpredict the standard deviation of output ( y t ) , consumption (<;,), government expenditures (g t), and hours (b t), but overpredict the stan dard deviation of investment (xt). Since we employ a general-equilibrium framework, the behavior of one variable cannot be viewed in isolation, because it is linked by the equilib rium conditions to the behavior of other vari ables in the model. For example, the low variability of output is linked to the low vari ability of hours, because the production tech nology is labor intensive.13 Likewise, the low variability of consumption is linked to the high variability of investment, because changes in household saving (which correspond to changes in investment) act as a buffer against earnings shocks, thereby allowing households to smooth their consumption over the business cycle. Con sistent with standard RBC models (see Kydland and Prescott [1982]), both model versions cap ture the fact that output is more variable than consumption, but less variable than investment, over the U.S. business cycle. In comparison to the lump-sum tax structure, the distortionary tax structure displays a lower standard deviation of output and hours and a higher variability of investment relative to out put. This behavior can be traced to the move ment of the optimal distortionary tax rates on labor and capital income. Specifically, the opti mal labor tax is procyclical (positively corre lated with output), while the optimal capital tax is countercyclical. The procyclical labor tax oper ates to smooth households’ after-tax income from labor. For example, a positive technology shock (which shifts the production frontier outward and raises the real wage) is accompanied by an increase in xht. The higher tax rate tends to offset the higher real wage and thus provides households with an implicit insurance mechanism against ■ 12 In all figures, model variables are the realizations from a single simulation (based on randomly drawn shocks) to compare volatility and persistence properties with the corresponding U.S. variables. There is no intention to predict the actual time path of U.S. variables. ■ 13 This can be readily observed from the aggregate production function (equation [8a]), where the coefficient on the labor input, a 2 = 0.66, is nearly double the coefficient on the capital input, a 1= 0.34. 9 TABLE 1 Business Cycle Statistics for Models and the U.S. Economy Standard Deviation (percent) Variables Distortionary Tax Structure“1 Lump-Sum Tax Structure3 U.S. Economy*5 1947-92 1954-92 2.16 2.11 2.39 2.46 ct 0.69 0.96 1.14 1.19 s, 1.66 1.00 6.45 3.04 y, k. 1.05 1.03 0.75 0.74 xt 9.10 9.00 5.96 6.14 h, yt /h, 1.19 1.60 1.76 1.86 1.09 0.96 1.45 0.90 b, 2.72 — 4.54 4.80 T ax rates xb, 1.71 — 6.42 4.08 xkt 13.91 — 17.76 19.28 3.61 1.62 2.65 1.88 Rev/yt Contemporaneous Correlation with Output Variables Distortionary Tax Structure“1 U.S. Economy15 Lump-Sum Tax Structure“1 1947-92 1954-92 0.87 0.67 0.89 0.71 st 0.98 0.85 0.62 0.40 k. 0.03 - 0 .0 1 0.62 0.63 xt 0.97 0.97 0.69 0.89 bt yt /ht 0.93 0.96 0.81 0.91 0.92 0.89 0.71 0.52 bt 0.03 — 0.23 0.36 0.97 — 0.39 - 0 .1 0 — 0.08 - 0 .1 5 0.39 0.11 T ax rates *bt *kt Rev/yt - 0 .9 3 - 0 .9 1 - 0 .9 5 a. Model statistics are means over 100 simulations, each 46 periods long, after dropping the first 50 periods. The symbol Rev/yt is the economywide average tax rate, defined as total tax revenue as a fraction of output. b. The following quarterly series from Citibase were annualized before computing the statistics: y, = GNPQ, c, = GCN Q + GCSQ (nondurables + services), g, = GGEQ, h,= LHOURS (household survey), and } ’i /h,= GNPQ/LI IOURS. The series for x, is business fixed investment + consumer durable expenditures + residential invest ment. The series for kt is fixed private capital + stock of consumer durables + resi dential capital. Both x, and k, are annual series from Fixed Reproducible Tangible Wealth in the United States, U.S. Department of Commerce, 1993- The series for bt is federal debt held by the public from Federal Debt and Interest Costs, Congres sional Budget Office, 1993, table A-2, where nominal debt has been converted into real debt by dividing by the GNP deflator for each year. Rev/y, is total gov ernment receipts (federal, state, and local) as a fraction of GNP from Economic Report o f the President, 1991, 1994, table B-80. Data on average marginal tax rates do not extend over the full sample: xht is from Barro and Sahasakul (1986) for 1947-83, and x^ is from Jorgenson and Sullivan (1981), table 11. for 1947- 80. NOTE: Before computing the statistics, all series were logged and detrended using the Hodrick-Prescott filter (see Prescott [1986]). The smoothing parameter for the filter was set at 100, since all data are at annual frequency. SOURCE: Authors’ calculations. variability in the after-tax wage (1 - zht) u\. Since labor supply decisions depend on the quantity (1 - Tht) wt (see equation [4b]), a lower variability in the after-tax wage leads to a lower variability in hours worked. With a labor-intensive production technology, this also leads to lower variability in output. Zhu (1992) shows theoretically that the cycli cal behavior of the optimal labor tax depends on the degree of risk aversion (or curvature) exhibited by the household utility function. The optimal labor tax is procyclical for low-riskaversion (less curved) utility functions, such as the logarithmic case used here, but counter cyclical for high-risk-aversion (more curved) functions. Chari, Christiano, and Kehoe (1994) provide a quantitative demonstration of this re sult. In general, the level of risk aversion deter mines the amount by which households are willing to adjust their labor supply in response to a change in the real wage. With low risk aversion, the substitution effect of an increase in wt (caused by a positive technology shock) results in a relatively large increase in labor hours. The government takes advantage of this greater willingness to work by raising the tax on labor, thereby collecting additional reve nue, but still allowing an increase in labor to spur output during this period of high labor productivity. With high risk aversion, however, the substitution effect is much smaller; that is, households are less willing to increase their labor supply in response to the higher real wage. The government’s optimal response now is to lower the tax rate on labor. This stimulates labor supply in order to boost output while labor productivity is high. Our results are con sistent with the findings of these researchers. The capital tax in the model moves counter cyclical^ and displays high variability relative to the labor tax (see table 1 and figures 1-2). This serves to increase the variability of household in vestment relative to output under the distortionary tax structure. Ordinarily, a positive technology shock raises the real rate of interest and motivates an increase in investment because the rate of return becomes more attractive. However, when a positive technology shock is accompa nied by a decrease in x kt, the after-tax return on investment becomes even more appealing, leading to a larger rise in investment. From the government’s perspective, a countercyclical capital tax is optimal because it serves as an efficient means of absorbing shocks to the gov ernment’s budget constraint. These shocks are caused by changes in the size of the tax base over the the business cycle. For example, a positive technology shock generates more tax 10 FIGURE 1 Capital Tax Rates Percent Year a. Distortionary tax structure. SOURCES: Jorgenson and Sullivan (1981); and authors’ calculations. F I G U R E 2 Labor Tax Rates Percent Year a. Distortionary tax structure. SOURCES: Barro and Sahasakul (1986); and authors’ calculations. FIGURE 3 Government Expenditures as a Fraction of GNP revenue because GNP and household incomes (the tax base) increase. This motivates a reduc tion in xkt because government spending require ments can be met using a lower tax rate. A similar argument holds in reverse for the case of a nega tive technology shock. Absorbing shocks mainly by changes in xkt, as opposed to changes in xht, is efficient because the capital stock cannot be quickly adjusted in response to a change in the capital tax. In contrast, the household can instan taneously adjust labor supply in response to a change in the labor tax. The shock-absorbing feature of Xkt allows the government to maintain a very smooth time series for gt/ yt, as compared to the lump-sum tax structure (see figure 3).14 We experimented with varying the level of monopoly profits by adjusting the values of the parameters 0, , 0 „ and % . In general, we found that as profits declined, the standard de viation of xkt i ncreased. The intuition for this result is straightforward. Recall that dividends (equal to after-tax profits) do not distort house hold decisions because profits are determined outside households’ control. A lower level of profits implies a smaller and more elastic tax base for the capital tax. Consequently, larger changes in the tax rate are needed to produce the same revenue effect when responding to technology shocks. 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.1^ Note, how ever, that the correlation coefficients between U.S. tax rates and real GNP display a change in sign, depending on the sample period. The labor tax and the capital tax are weakly procy clical using data on average marginal tax rates that begin in 1947, but weakly countercyclical for data that begin in 1954. The model, on the other hand, predicts a strongly procyclical labor tax and a strongly countercyclical capital tax. Thus, there is a sharp negative correlation Percent ■ 14 The optimality of using a state-contingent capital tax to absorb budget shocks has been shown previously by Judd (1989) and Chari, Christiano, and Kehoe (1994). Our quantitative results are not directly comparable because Judd does not explicitly model household behavior, and Chari, Christiano, and Kehoe employ a different decentralization scheme for and rbt. ■ 15 The figures display the tax-rate series before detrending. For quantitative comparisons (table 1), 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 a. Distortionary tax structure. b. Lump-sum tax structure. SOURCES: Citibase; and authors' calculations. in xht is possibly linked to the phenomenon of “bracket creep,” which ex isted before tax schedules were indexed for inflation in 1985. Auerbach and Poterba (1988) argue that the downward trend in x w is due to increasingly generous investment tax credits and accelerated depreciation schedules. n FIGURE 4 Total Tax Revenue as a Fraction of GNP Percent Year a. Distortionary tax structure. b. Lump-sum tax structure. SOURCES: Economic Report o f the President, 1991, 1994; and authors’ calculations. FIGURE 5 Government Debt as a Fraction of GNP Percent SOURCES: Congressional Budget Office, Federal Debt and Interest Costs, 1993; and authors’ calculations. between Xkt and xht in the model, while the corresponding correlation in the data is weakly positive.16 Figure 4 displays the economywide average tax rate, defined as total tax revenue as a fraction of output. This rate is countercycli cal in both versions of the model, but weakly procyclical in the U.S. data. Thus, the model does not capture some important features of U.S. tax-rate movements. This highlights the dif ficulty of summarizing the entire U.S. tax code using only one or two broadly defined rates. The predicted standard deviations for gt are substantially lower than the U.S. value of 6.45 percent computed using the full sample. Start ing the sample in 1954 to avoid the influence of the Korean War reduces the standard deviation of gt in the data by half, to a value of 3.04 per cent, which is much closer to model predic tions. Although we have no theoretical justifica tion for excluding the Korean War years (since we include the Vietnam War), the fact that our model contains only one type of shock limits its ability to explain large movements associ ated with a war. Incorporating an additional shock to households’ preference for public goods to simulate high demand during wars would increase the variability of gt in the model (see Lansing [1994]). Both model versions capture the procyclical behavior of government spending in annual U.S. data, but the correlations from the model are much stronger than observed in the data. When the years prior to 1954 are excluded, the correlation between government spending and real output in the U.S. data drops from 0.62 to 0.40, worsening the comparison with the model. In the distortionary tax structure, government debt is essentially acyclical (it behaves similarly to capital in this respect), but is less variable than U.S. government debt (see figure 5). We experimented with an alternate version of this model in which the government is required to balance its budget each period. For this experi ment, we retained the decentralization scheme described in section II, whereby the govern ment is required to satisfy the certainty version of equation (4c). Qualitatively, the results are similar to those reported in table 1. However, the variability of the labor tax goes up, since government debt is no longer available to help cushion budget shocks. The insurance effect of the procyclical labor tax thus becomes more significant, leading to lower variability in hours and output. Interestingly, our model implies that a balanced-budget amendment can help smooth business cycle fluctuations, provided the government sets tax rates in the manner we have described. The lump-sum tax structure does a reason ably good job of matching the relative standard deviations of hours and the real wage, where the real wage is measured by average labor productivity yt/ ht (see table 2). This behavior is typical of standard RBC models with indivis ible labor (see Hansen [1985]). In the distortion ary tax structure, however, the standard devia tion of hours relative to the real wage is too low, despite the specification of indivisible labor. The insurance effect of the procyclical labor tax ■ 16 In the U.S. data, the correlation coefficient between (logged and detrended) xh l and x kt equals 0.36 from 1947 to 1980, the period for which estimates of both tax rates are available. For the years 1954 to 1980, the correlation coefficient is 0.34. In the model, the correlation coefficient is -0 .9 7 . T A B L E 2 Comparison of Labor Market Statistics Variables Distortionary7 Tax Structure1* Lump-Sum Tax Structure*1 U.S. Economy 1947-92 1954-92 a h / a v/h 1.09 1.67 1.22 2.10 2.43 corr (b, y /h ) 0.72 0.72 0.16 0.13 - 0.02 2.07 - this one to explore the welfare effects of vari ous tax structures and find that these effects can be quite dramatic. In this paper, our model meets with varying degrees of success in matching the observed behavior of tax rates, government spending, and aggregate eco nomic variables in the U.S. economy. Nonethe less, this exercise is useful in that it provides information on how models of government fis cal policy might be improved. 0.21 a. Model statistics are means over 100 simulations, each 46 periods long, after dropping the first 50 periods. b. The first num ber denotes hours worked from the household survey in Citibase (LHOURS), and the second denotes hours worked from the estab lishment survey (LPMHU). NOTE: Before com puting the statistics, all series were logged and detrended using the Hodrick-Prescott filter (see Prescott [1986]). The smoothing parame ter for the filter was set at 100, since all data are at annual frequency. SOURCE: Authors’ calculations. References Aiyagari, S. Rao. “O n the Contribution of Technol ogy Shocks to Business Cycles,” Federal Re serve Bank of Minneapolis, Quarterly Revieu', vol. 18, no. 1 (Winter 1994), pp. 22-84. Atkinson, Anthony B., and Joseph F. Stiglitz. Lectures on Public Finance. McGraw-Hill: New York, 1980. is responsible for the reduced variability of hours. Again, we note that the model includes only one type of shock. Aiyagari (1994) uses a variance decomposition analysis to argue that the behavior of hours in U.S. data is driven by multiple shocks. Finally, the predicted correlations between hours and productivity in table 2 are much higher than the corresponding U.S. values. Braun (1994) shows that an RBC model with exogenous stochastic tax rates is capable of matching both of the U.S. labor market statis tics in the table. Our simulations show- that a model with endogenous tax rates can produce very different results.1 V. Conclusion We have constructed a model that combines elements from the theory of optimal public fi nance with an RBC view of aggregate fluctua tions. Our aim is to develop a framework that is useful for carrying out realistic policy experi ments with regard to both the structure of the U.S. tax system and the composition and fi nancing of government expenditures. In two related papers (see Guo and Lansing [1994a, 1994b]), we employ models similar to ■ 17 See Chari, Christiano, and Kehoe (1994) for a more extensive comparison between an RBC model with exogenous stochastic tax rates and one with optimal Ramsey tax rates. Auerbach, Alan J., and James M. Poterba. “Why Have Corporate Tax Revenues Declined?” in E. Helpman, A. Razin, and E. Sadka, eds., Economic Effects o f the Government Budget. Cambridge, Mass.: MIT Press, 1988. Barro, Robert J., and Chaipat Sahasakul. “Aver age Marginal Tax Rates from Social Security and the Individual Income Tax,” Jo u rn a l o f Business, vol. 59, no. 4 (October 1986), pp. 555-66. Benhabib, Jess, and Roger E.A. Farmer. “Indeter minacy and Increasing Returns,” Jo u rn a l o f Economic Theory, vol. 63, no. 1 (June 1994), pp. 19-41. _________, and Boyan Jovanovic. “Externalities and Growth Accounting,” Am erican Eco nom ic Review, vol. 81, no. 1 (March 1991), pp. 82-113. Braun, R. Anton. “Tax Disturbances and Real Economic Activity in the Postwar United States,” Jo u rn a l o f Monetary Economics, vol. 33, no. 3 (June 1994), pp. 441-62. Cassou, Steven P. “Optimal Tax Rules in a Dy namic Stochastic Economy with Capital,” Jo u rn a l o f Economic Dynam ics a n d Control, 1995 (forthcoming). 13 Chamley, Christophe. “Efficient Taxation in a Stylized Model of Intertemporal General Equilibrium,” International Economic Retieiv, vol. 26, no. 2 (June 1985), pp. 451 -68. ________ . “Optimal Taxation of Capital Income in General Equilibrium with Infinite Lives,” Econometrica, vol. 54, no. 3 (May 1986), pp. 607-22. Chari, V.Y. “Time Consistency and Optimal Pol icy Design," in P.J. Miller, ed., The Rational Ex pectations Revolution: Readingsfrom the Front Line. Cambridge, Mass.: MIT Press, 1994. ________ , Lawrence J. Christiano, and Patrick J. Kehoe. “Optimal Fiscal Policy in a Business Cycle Model,” Jo u rn a l o f Political Economy, vol. 102 (August 1994), pp. 617 - 52. Christiano, Lawrence J. “Why Does Inventory Investment Fluctuate So Much?” Jo u rn a l o f Monetary Economics, vol. 21 (March/May 1988), pp. 247-80. ________ , and Martin Eichenbaum. “Current Real-Business-Cycle Theories and Aggregate Labor-Market Fluctuations,” Am erican Eco nom ic Review, vol. 82, no. 3 (June 1992), pp. 430-50. Guo, Jang-Ting, and Kevin J. Lansing. “The Wel fare Effects of Tax Simplification: A General Equilibrium Analysis,” Federal Reserve Bank of Cleveland, Working Paper 9409, Septem ber 1994a. ________ , a n d _________ . “Tax Structure, Wel fare, and the Stability of Equilibrium in a Model of Dynamic Optimal Fiscal Policy,” Federal Reserve Bank of Cleveland, Work ing Paper 9410, September 1994b. Hansen, Gary D. “Indivisible Labor and the Business C y c le Jo u rn a l o f Monetary Eco nomics, vol. 16, no. 3 (November 1985), pp. 309-27. Jones, Larry7E., Rodolfo Manuelli, and Peter E. Rossi. “O n the Optimal Taxation of Capital Income,” National Bureau of Economic Re search Working Paper No. 4525, November 1993. Jorgenson, Dale, and Martin A. Sullivan. “Infla tion and Corporate Capital Recovery,” in C.R. Hulten, ed.. Depreciation, Inflation, an d the Taxation o f Incom e from Capital. Wash ington, D.C.: Urban Institute Press, 1981. Judd, Kenneth L. “Redistributive Taxation in a Simple Perfect Foresight Model /J o u r n a l o f Public Economics, vol. 28 (October 1985), pp. 59-83. ________ . “Optimal Taxation in Dynamic Sto chastic Economies: Theory and Evidence,” Hoover Institution, unpublished manuscript, May 1989. Juster, F. Thomas, and Frank P. Stafford. "The Allocation of Time: Empirical Findings, Be havioral Models, and Problems of Measure ment,” Jo u rn a l o f Economic Literature, vol. 29, no. 2 (June 1991), pp. 471 - 522. Kydland, Finn E., and Edward C. Prescott. “A Competitive Theory of Fluctuations and the Feasibility and Desirability of Stabilization Policy,” in S. Fischer, ed., Rational Expecta tions a n d Economic Policy. Chicago: Univer sity of Chicago Press, 1980a. ________ , a n d _________ . “Dynamic Optimal Taxation, Rational Expectations, and Opti mal Control,” Jo u rn a l o f Economic D ynam ics a n d Control, vol. 2, no. 1 (February 1980b), pp. 79-91. ________ , a n d _________ . “Time to Build and Aggregate Fluctuations,” Econometrica, vol. 50, no. 6 (November 1982), pp. 1345-70. Lansing, Kevin J. “Optimal Fiscal Policy When Public Capital Is Productive: A Business Cy cle Perspective,” Federal Reserve Bank of Cleveland, Working Paper 9406, April 1994. Long, John B. Jr., and Charles I. Plosser. “Real Business Cycles,” Journal o f Political Economy, vol. 91, no. 1 (February 1983), pp. 39-69. Lucas, Robert E., Jr. “Supply-side Economics: An Analytical Review,” Oxford Economic Papers, vol. 42, no. 2 (April 1990), pp. 293-316. McGrattan, Ellen R. “The Macroeconomic Effects of Distortionary Taxation,” Jo u rn a l o f Mone tary Economics, vol. 33, no. 3 (June 1994), pp. 573-602. 14 ________ , Richard Rogerson, and Randall Wright. “Household Production and Taxa tion in the Stochastic Growth Model,” Fed eral Reserve Bank of Minneapolis, Staff Report No. 166, December 1993Prescott, Edward C. “Theory ahead of Business Cycle Measurement,” Federal Reserve Bank of Minneapolis, Quarterly Review, Fall 1986, pp. 9-22. Ramsey, Frank P. “A Contribution to the Theory of Taxation,” Economic Journal, vol. 37 (March 1927), pp. 47-61. Rogerson, Richard. “Indivisible Labor, Lotteries, and Equilibrium,” Jo u rn a l o f Monetary Eco nomics, vol. 21 (1988), pp. 3 - 16. Sargent, Thomas J. D ynam ic Macroeconomic Theory>. Cambridge, Mass.: Harvard Univer sity Press, 1987. Stokey, Nancy L., and Robert E. Lucas, Jr. Recur sive Methods in Economic Dynamics. Cam bridge, Mass.: Harvard University Press, 1989. Zhu, Xiaodong. “Optimal Fiscal Policy in a Sto chastic Growth Model,” Jo u rn a l o f Economic Theory, vol. 58, no. 2 (December 1992), pp. 250-89. 15 Cross-Lender Variation in Home Mortgage Lending by Robert B. Avery, Patricia E. Beeson, and Mark S. Sniderman Introduction Federal fair housing and credit legislation ad dresses two major requirements. First, depository institutions must help meet the credit needs of their communities in a manner consistent with safe and sound lending practices (Community Reinvestment Act of 1977 [CRA]). Second, lenders must not discriminate against individual appli cants on the basis of race, ethnic origin, gender, or religion (Equal Credit Opportunity Act of 1974 [ECOA]) and Fair Housing Act of 1968 [as amend ed in 1988]). Government agencies charged with regulating depository' institutions are responsible for monitoring individual lenders’ compliance with these statutes. Historically, enforcement of the CRA and fair lending statutes has relied on qualitative, nonstatistical methods. CRA examinations, for ex ample, have focused primarily on procedural issues. With rare exception, regulators have considered the actions of individual complain ants to enforce the other fair lending statutes. In the past year, both community activists and lenders have called for strategies to move to ward more quantitative, outcome-based en forcement procedures. These calls stem, in part, from a belief that CRA and fair lending policy Robert B. Avery is an associate professor in the Department of Consumer Economics and Hous ing, Cornell University, and Patricia E. Beeson is an associate professor of economics at the University of Pittsburgh; both are visiting schol ars at the Federal Reserve Bank of Cleveland. Mark S. Sniderman is senior vice president and director of research at the Federal Reserve Bank of Cleveland. The authors thank Glenn Canner, Charles Carlstrom, John Duca, Stuart Gabriel, Jagadeesh Gokhale, John Flam, Joseph Haubrich, Stuart Rosenthal, Peter Rupert, and Peter Zorn for helpful com ments and suggestions. guidelines are unclear and often counterpro ductive, and perhaps more likely to generate paperwork than loans and services. A recent change in the reporting require ments of the Home Mortgage Disclosure Act (HMDA) makes the move toward outcomebased enforcement procedures seem much more feasible. Since 1990, lenders in urban mortgage markets have been required to re port to regulators the neighborhood (census tract) and a limited number of other character istics (such as loan size, applicant race and income, and government guarantee) of all ap plications for mortgage credit during each cal endar year. These data permit the quantitative comparison of a number of lending activities across lenders. Indeed, recent proposals by the bank regulatory agencies call for the use of HMDA data in evaluating CRA and fair lending compliance for lenders.1 The objective of this study is to evaluate the feasibility of using HMDA data to form quanti tative measures of lender activity for use in enforcement. We consider three potential meas ures of firm-level mortgage lending activity: ■ 1 See “Community Reinvestment Act Regulations," Federal Regis ter, O clobeH , W 4 (59 FR 51232). 16 loan application rates, loan approval rates, and loan origination rates. We examine the extent to which the three measures can differentiate among hanks with respect to how they serve four subpopulations cited in the fair lending laws: minority loan applicants, applicants for loans in minority neighborhoods, low-income loan applicants, and applicants for loans in low-income neighborhoods. Using national fil ings for the first year of the new7HMDA regula tions, 1990, we compare the performance of measures adjusted and unadjusted for local conditions and borrower characteristics. Several conclusions emerge. We find that variation across lenders in loan originations to each of the four subpopulations is driven primar ily by variation in application rates, not by vari ation in lender approval rates. This finding holds when both unadjusted and adjusted measures are examined and for a wide variety of lender groups sorted by size and type of institution. Further more, we find virtually no correlation between application rates and approval rates, so using in dices based only on approval rates can be poten tially misleading when evaluating individual lenders' compliance with the CRA. Indeed, varia tion in application rates appears to play a much more significant role in explaining variation in credit flows. Furthermore, focusing on approval rates may lead to outcomes that are counter to the intent of the legislation: To improve their minority-to-white approval rates, some lenders may discourage applications from all but the most creditworthy minority applicants, thereby reducing credit originations to minority and lowincome communities. One objection that has been raised to the use of application rates in evaluating lender compliance is that these rates are determined primarily by the neighborhoods that lenders serve. Our evidence suggests that this is not the case. Most of the variation in application rates stems from differences in the applicants that lenders attract w ithin neighborhoods and not from the general racial characteristics of the neighborhoods as a whole. Finally, we also find that controlling for the economic character istics and neighborhoods of the loan applicants provides relatively little powrer in explaining cross-lender differences. This suggests that gross application and approval-rate measures may give relatively good rankings of bank performance. I. Background of Fair Housing Legislation In response to community concerns about the flow of housing credit to minority and lowincome communities and about the extent to wrhich individual lenders were meeting the credit needs of their communities, Congress passed a series of laws during the 1970s. The ECOA of 1974 and the Fair Housing Act of 1968 (as amended in 1988) mandate that lenders do not discriminate against individual applicants on the basis of race, ethnic origin, gender, or religion. Two other laws were enacted primar ily to fight geographic discrimination. HMDA, enacted in 1975, requires certain lenders to re port annually the number and dollar value of mortgage loans they make in their communities according to census tract. Under the terms of the CRA, enacted in 1977, depository institutions must help meet the credit needs of their com munities, including low- and moderate-income neighborhoods, in a manner consistent with safe and sound lending practices. Monitoring individual lenders for compliance with these fair lending statutes is problematic. The standard for compliance with the ECOA is relatively well defined— other things equal, lenders cannot discriminate on the basis of race or property location. This implies that lenders must treat “comparable” applications from members of different racial groups and neighborhoods equally. Problems arise, how ever, in determining what types of behavior are considered discriminatory and in measuring whether two applications are the same except for the applicant’s race and the property loca tion.2 Moreover, focus on the treatment of for mal applications sidesteps the issue of indirect screening, whereby applicants are screened out before they formally apply. These concerns have shifted much of the focus in current en forcement of ECOA from procedures to out comes. If the outcome of the process differs across racial groups or neighborhoods, then the burden of proof is on the lender to demon strate that its procedures are not biased. For example, the lender can show that the differ ences arise from variables other than race and that the use of these variables in loan screening can be justified by their relationship to costs or loan performance. If differential outcomes in origination rates create the prima facie case for ■ 2 See Wienk (1992) for a discussion of conceptual and measure ment problems related to assessing the degree of discrimination in credit markets. a bias, the lender could face an examination of its application and approval rates, as both af fect originations. CRA is concerned with the extent to which individual lenders extend credit to various groups within their market areas. While Con gress did not articulate the standards for com pliance beyond "meeting the credit needs of the community,” the bank regulatory agencies responsible for enforcement issued joint policy statements in 1980 and 1989 reflecting their procedures. Apart from periodic examinations for compliance, regulators are required to take account of an institution’s CRA record in assess ing applications for regulatory actions such as mergers. Since 1990, lenders have also been re quired to give the public access to their exami nation assessments. Enforcement of CRA has generally focused on procedures rather than outcomes. Regulators have given significant weight to evidence of affirmative action— for instance, the location of loan offices, number of minority loan officers, methods of advertis ing, participation in community development banks, and availability of special low- to moderate-income housing programs. O n the surface, ECOA and CRA appear to address different aspects of the lending process. ECOA is concerned primarily with individuals, equal treatment, and race; CRA involves neigh borhoods, credit flows, and income. More re cently, though, enforcement of both acts has begun to evolve along quite similar lines. Regu lators consider fair lending practices a critical factor in assigning CRA ratings. Moreover, as a practical matter, CRA enforcement has begun to place more weight on racial issues rather than focusing exclusively on income. HMDA was instituted to provide regulators and the public with information on how lenders were serving low-income areas. Data reported under HMDA are now integral to enforcement efforts for both ECOA and CRA. Initially, deposi tory institutions were required to report mortgage lending totals by census tract with no disaggrega tion by race, but concerns arose about the dearth of data available to analyze the reasons for differ ential mortgage credit flows and individual dis crimination in mortgage lending. Amendments to HMDA in 1989 now require most mortgage lend ers to collect and report information on all indi vidual loan applications taken, whether approved or not. In addition, some applicant information is now recorded, most notably income, loan amount requested, property location, gender, and race. Many informative HMDA-based studies ad dressing issues concerning both ECOA and CRA have appeared during the past 15 years. Because the pre-1990 HMDA data contain no information about the individual applicants or about applica tions that were not approved, most of the early studies focus on the flow of credit to various neighborhoods (CRA), as opposed to a considera tion of discrimination against particular loan applicants (ECOA). These studies ask whether mortgage lenders in an area, taken collectively, provided mortgage credit in predominantly mi nority or low-income neighborhoods at dimin ished rates relative to predominantly white or higher-income neighborhoods. Although re searchers generally find disparate lending patterns between white and minority (or low-income) neighborhoods, they do not consider differences in lending patterns across individual lenders: Are these neighborhoods receiving less credit be cause each lender originates only a few loans in these areas, or because there are only a few lend ers operating in these areas?3 In addition, the data do not allow a clean investigation of the roles of credit supply and credit demand: Are these neighborhoods receiving less credit because of lender bias, or because lenders are not receiv ing comparable numbers of qualified applications from the various neighborhoods examined? The expanded HMDA data set has spawned a number of new analyses of individual and neighborhood discrimination. Using informa tion from a special survey that supplemented HMDA data for Boston, Munnell et al. (1992) examine the role of individual characteristics, particularly race, in loan approval. Avery, Beeson, and Sniderman (1993) discuss similar issues using 1990 and 1991 HMDA data drawn from the whole country. The role of neighbor hood racial composition in generating applica tions and approving loans is explored in Avery, Beeson, and Sniderman (1994). Many questions remain as to the appropriate methods of CRA and ECOA enforcement and the nature of the data collected to support this effort. Critics of the CRA, in particular, have argued that enforcement efforts need to focus ■ 3 Using pre-1990 HMDA data, Canner (1981), Avery and Buynak (1981), Avery and Canner (1983), and Bradbury, Case, and Dunham (1989) contrast the differences in mortgage credit originations between predominantly white and predominantly minority neighborhoods in vari ous metropolitan statistical areas (MSAs). One of the few studies to look at lenders is Calem (1993). He contrasts the experiences of individual lenders participating in a Philadelphia area mortgage-lending plan with those who did not participate. However, his paper does not document the existence of lender differences in the penetration of minority communi ties; his primary focus is on the characteristics of the voluntary mortgage plan operated by a group o f lenders. Avery (1989) notes the differences between studies based on lending in a neighborhood and the procedures adopted by individual lenders. 18 B O X I HMDA Data and Methodology Overall, HMDA reported information on 6,595,089 loan applica tions and purchases in 1990. O f these, 1,137,741 were purchased from other institutions and 1,523,429 were applications received for properties outside an MSA. Excluding these left 3,933,919 applications (59.6 percent) to reporting institutions for properties within an MSA in which the lender had an office. O f these appli cations, 787,952 were for home improvement loans, 716,595 were for refinancing of one- to four-family home loans, and 32,176 were for multifamily home loans. An additional 241,295 applications were never acted on because they were either with drawn by the applicant or closed due to incompleteness. Elimi nating these from our sample left a total of 1,984,688 loan applications that met the saidy criteria. Not surprisingly, the initial HMDA filings contained many er rors and inconsistencies that required extensive editing by the re ceiving federal agencies. Unfortunately, these procedures do not appear to have been unifonnly applied, requiring additional cleaning and editing for this study. In addition, smaller institu tions were not required to report race, income, and gender for loan applicants. We decided to deal with missing data using a “hot deck” imputation procedure similar to that used by the U.S. Census Bureau. Applications with missing data were statistically matched to applications in the same census tract that came clos est to them in reported characteristics (race, loan action, income, and loan amount). Missing values were filled in using the vari able value of the matched observation. Applications with implau sible reported values were treated as missing and imputed in the same way. Overall, income was imputed for 4.9 percent, loan amount for 1.5 percent, gender for 4.0 percent, and race for 5.6 percent of the study sample applications. more on performance and less on process. In this spirit, the bank regulatory agencies have recently called for comments on a comprehen sive reform of CRA regulations and enforce ment procedures. The proposed reforms would institute a new system of evaluation based pri marily on performance. The data reported un der HMDA are critical to the success of such an effort, both for quantifying an institution’s own performance and for providing a bench mark of what other institutions are doing. Because the new regulations encompass an expanded role for HMDA data, it is natural to ask how capable the data are for meeting this task. The new regulations call for only a mild expansion of HMDA, so the current data are representative of what would be available in the future. In this paper, we use the existing data to examine their effectiveness in provid ing the quantitative measures of institutional performance called for by the proposed new regulations. We examine three potential out come measures: loan origination rates, loan application rates, and application approval rates. These are used to compare institutions’ performance in serving four subpopulations cited by CRA: minority individuals, minority neighborhoods, low-income individuals, and low-income neighborhoods. In each case, we compare the rankings implied by gross out come measures with those adjusted for neigh borhood and applicant characteristics. II. Data Description Amendments to HMDA in 1989 now require most depository institutions (and certain other mortgage lenders) to collect and report infor mation on all individual loan applications taken for home purchase, mortgage refinance, and home improvements, whether approved or not. This study makes use of the HMDA data for 1990 — the first release of the new data — which represent the most comprehensive sur vey of mortgage lending in the United States.4 All commercial banks, savings and loan asso ciations, credit unions, and other mortgage lending institutions (primarily mortgage bank ers) that have assets of more than $10 million, make one or more one- to four-family home purchase loans, and have an office in a metro politan statistical area (MSA) are required to meet HMDA reporting requirements. For each mortgage application received or mortgage loan purchased from another institu tion during the calendar year, the lender must report the loan amount; the location of the prop erty (state, county, and 1980 census tract num ber); whether the property is owner-occupied; the loan purpose (home purchase, home improvement, or refinancing for one- to fourfamily or multifamily unit); the type of loan (conventional, FHA, VA, guaranteed by Farmers Horne Administration [FmHA]); the action taken by the lender (loan approved and originated, ap plication approved but withdrawn, application denied, application withdrawn before lender action, file closed for incompleteness, loan pur chased from another institution); the race and gender of the loan applicant (and co-applicant, if ■ 4 At the time this paper was published, 1991 and 1992 HMDA data were also available. Although not reported here, analysis of data from these later years suggests similar conclusions to those presented here. 19 TABLE 1 Characteristics of Home Mortgage Applications, 1990 HMDA Percent of Sample Percent of Loan Dollars Approval Rate Race of Applicant Native American Asian (or Pacific Islander) Black Hispanic White Other 0.6 4.6 6.2 6.6 81.4 0.7 0.6 6.8 4.8 6.4 80.5 1.0 80.7 85.6 70.6 77.9 86.9 80.2 Race of Co-applicant No co-applicant Same race as applicant Different race than applicant 28.4 69.4 2.2 24.1 73-4 2.5 82.7 86.2 84.4 Loan Type Conventional FHA VA FmHA 75.1 20.4 4.5 0.0 82.9 13.7 3.5 0.0 85.1 85.5 84.2 98.0 14.8 2.9 82.3 44.9 14.5 10.5 9.0 21.1 13.1 3.5 83.4 47.7 14.4 7.6 9.1 21.2 Lender Action Loan denied Loan accepted and withdrawn Loan originated Loan kept by originator2 Loan sold to FNMAa Loan sold to GNMAa Loan sold to FHLMC3 Loan sold elsewhere3 Memo Items Median income Median loan request Number of loans $48,000 $77,000 1,984,688 a. Percent o f originations. SOURCE: Authors’ calculations. any); and the income relied on by the lending institution in making the loan decision.^ In total, 9,333 financial institutions filed HMDA reports in 1990 on more than 6 million loan applications and loan purchases. Our analysis focuses on a subset of these filings: ap plications for one- to four-family home pur chase loans that were acted upon (approved or denied) by the lender. This sample includes 1,984,688 loan applications made to 8,745 sepa rate lenders operating in 40,008 census tracts in all 340 of the U.S. MSAs defined as of 1990 (see box 1 for details). The study sample has a substantial degree of representation from applicants of different races and income levels (table 1). Overall, however, applicants for home purchase mort gages are a select sample of American house holds. Applicants’ median income ($48,000) is substantially higher than the median income of families in MSAs ($37,918) as reported in the 1990 decennial census.6 The racial composition of the study sample also appears to differ somewhat from that of all U.S. families. Blacks filed 6.2 percent of the HMDA housing loan ap plications, yet were 7.7 percent of the home owners and headed 11.4 percent of the MSA households. Asian loan applicants (4.6 per cent), however, were overrepresented com pared with their numbers in the census (2.5 percent of MSA household heads and 2.2 per cent of homeowners). The share of white (81.4 percent) or Hispanic (6.6 percent) applicants is approximately representative of their numbers (78.1 percent of household heads and 84.8 per cent of homeowners for whites and 7.5 percent of household heads and 5.0 percent of home owners for Hispanics).7 Also worth noting is the substantial pres ence of the federal government in mortgage lending. One-quarter of the mortgages issued were directly guaranteed by the federal govern ment (FHA, VA, or FmHA), with an additional quarter purchased in the secondary market by one of the federal housing credit agencies (FNMA and FHLMC).8 Indeed, 55 percent of all mortgages issued were sold in the second ary market, suggesting that the study of mort gage lending patterns is more an analysis of a brokered industry than one where participants buy for their own portfolios. Sample characteristics are broken down by type of lender and applicant in table 2. Lender here is defined at the MSA level. Thus, a lender reporting loans for two different MSAs is treated as two different lenders.9 Lenders, shown in the rows, are grouped by size and ■ 5 See Canner and Smith (1991,1992) for a full description of the HMDA data. Information on income, race, and sex of the applicant does not have to be supplied by reporting institutions with assets less than $30 million or for purchased loans. ■ 6 In the HMDA data, household income may be slightly under stated because it reflects only the portion of an applicant’s income needed for mortgage qualification. ■ 7 The percent Hispanic in the HMDA sample is slightly higher than the share for the overall U.S. population, due in part to the inclusion of Puerto Rico. ■ 8 These acronymns represent, respectively, the Federal Housing Administration, Veterans Administration, Farmers Home Administration, Federal National Mortgage Association, and Federal Home Loan Mort gage Corporation. ■ 9 The 8,745 financial institutions with loans in the study sample operated in an average of 2.4 MSAs. This translated into 20,695 study lenders when lenders were defined at the MSA level. 20 T A B L E 2 Minority and Low-lncome Individuals and Tracts Relative to Total Mortgage Lending, 1990 HMDA Minority“ Overall Approval Rate Type of Institution Commercial banks Thrift institutions Credit unions Bank subsidiaries Thrift subsidiaries Other mortgage banks Size of Institution > 500 applications 100-500 applications <100 applications Market Share of Institution > 5 percent 1-5 percent < 1 percent Size of M S A > 25,000 applications < 25,000 applications Percent Minority Applications >22 percent < 22 percent Total Percent Appli cations1 Low-Incomeb Percent Relative Origi Approval Approval nations1 Rate Rate Percent Applicationsc Percent Relative OrigiApproval Approval nationsc Rate Rate 0.82 0.87 0.89 0.84 0.86 0.87 11.2 13.9 9.0 12.7 14.2 15.9 9.1 12.5 7.7 11.1 12.0 14.3 0.67 0.78 0.77 0.73 0.72 0.79 0.81 0.90 0.86 0.87 0.84 0.90 20.5 11.0 15.6 17.9 14.5 12.0 17.4 9.6 13.4 14.5 12.6 11.1 0.69 0.76 0.77 0.68 0.74 0.81 0.85 0.87 0.86 0.81 0.87 0.92 0.86 0.85 0.84 17.1 10.9 9.5 15.3 9.2 8.1 0.77 0.72 0.71 0.90 0.85 0.85 12.1 17.0 17.6 10.4 14.5 14.7 0.74 0.73 0.70 0.86 0.85 0.83 0.86 0.85 0.84 14.2 13.2 11.6 12.3 11.7 10.1 0.74 0.76 0.73 0.87 0.89 0.87 16.9 14.2 12.4 14.5 12.0 10.4 0.73 0.72 0.70 0.86 0.85 0.84 0.86 0.85 18.1 10.9 16.5 9.2 0.78 0.72 0.91 0.85 8.6 18.2 7.4 15.4 0.74 0.72 0.86 0.85 0.80 0.86 31.8 10.3 13.3 29.5 8.9 11.7 0.75 0.75 0.75 0.93 0.87 0.88 12.8 15.5 15.0 9.6 13.2 12.7 0.66 0.73 0.72 0.82 0.85 0.85 0.85 Minority' Census Tracts d Overall Approval Rate Percent Appli cations*^ Percent Origi nations*^ Low-lncome Census Tracts1 Relative Percent Approval Approval Appli Rate Rate cations0 Percent Relative Origi- Approval Approval nationsc Rate Rate Ty pe of Institution Commercial banks Thrift institutions Credit unions Bank subsidiaries Thrift subsidiaries Other mortgage banks 0.82 0.87 0.89 0.84 0.86 0.87 11.0 13.2 8.5 11.7 13.3 14.9 9.3 12.1 7.7 10.2 11.3 13.6 0.69 0.80 0.80 0.73 0.73 0.80 0.85 0.92 0.90 0.87 0.85 0.91 22.5 10.4 18.0 17.7 17.0 12.7 20.1 9.4 16.5 15.3 14.9 11.9 0.76 0.79 0.82 0.72 0.75 0.82 0.93 0.90 0.92 0.86 0.88 0.94 Size of Institution > 500 applications 100 - 500 applications <100 applications 0.86 0.85 0.84 16.7 9.8 8.9 15.2 8.4 7.8 0.78 0.73 0.74 0.91 0.86 0.88 10.8 18.8 19.8 9.4 17.1 18.0 0.75 0.77 0.76 0.88 0.91 0.90 Market Share of Institution > 5 percent 1-5 percent < 1 percent 0.86 0.85 0.84 13.1 12.5 11.8 11.5 11.3 10.5 0.76 0.77 0.75 0.88 0.90 0.89 18.7 14.4 11.0 16.7 12.9 9.9 0.77 0.76 0.75 0.90 0.90 0.89 0.86 0.85 18.8 9.5 17.3 8.2 0.79 0.73 0.92 0.86 5.7 20.4 5.1 18.4 0.77 0.76 0.90 0.90 0.80 0.86 40.2 8.0 12.6 38.2 7.1 11.2 0.77 0.75 0.76 0.88 0.95 0.89 12.5 16.0 10.7 14.4 15.5 13.9 0.69 0.77 0.76 0.86 0.90 0.90 Size of M S A > 25,000 applications < 25,000 applications Percent Minority Applications > 22 percent < 22 percent Total 0.85 a. Native Americans, blacks, and Hispanics. b. Applicant income below $25,000. c. Percent of applications received (loans originated) by each class o f lender from minority applicants or low-income tracts. d. Census tracts with more than 30 percent o f loan applications from minority applicants. e. Census tracts with more than 30 percent o f loan applications from low-income applicants. SOURCE: Authors' calculations. 21 type of institution and by the size and minority population of their MSA as shown in the rows of the table. Applicants are grouped into five categories shown in the columns: 1) overall; 2) minority (native American, black, and Hispanic, about 13 percent of applicants); 3) low-income (family income of $25,000 or less, roughly the bottom 15 percent of applicants); 4) residents of minority census tracts (those with more than 30 percent of loan applications from minority applicants, roughly 13 percent of applicants); and 5) residents of low-income census tracts (those with more than 30 percent of loan appli cations from low-income applicants, again roughly 15 percent of applicants).10 For each applicant category, we show the percent of the lender-type’s loan applications or originations made to members of the category.11 We also present the category approval rate (the portion of all loan applications from members of the category that are approved) and the relative ap proval rate (the ratio of the category approval rate to the overall approval rate for all appli cants), shown in column 1. There is little evidence that specific types of lenders, such as commercial banks or thrifts, specialize in minority lending. O n the other hand, at least superficially, it would appear that there is specialization by size of lender. About 17 percent of the applicants to lenders receiving more than 500 home purchase loan applications were minorities, with a similar per centage from minority tracts. Smaller lenders (those with less than 100 applicants) took in only 9 percent of their applications from these categories. However, much of this difference may simply reflect the concentration of large lenders in large MSAs, where there is also a high concentration of minority applicants and minority tracts. Within MSAs, the difference in minority share between the larger institutions (those with market shares exceeding 5 percent) and small institutions is much less. The picture looks somewhat different for low-income applicants. Commercial banks and ■ 10 The decision to treat Asians and ‘‘other race” applicants as non minorities was somewhat arbitrary. As shown in table 1, the overall accep tance rate tor Asian home purchase loan applicants is much closer to the white acceptance rate than to acceptance rates for blacks, Hispanics, or native Americans. We note, though, that the acceptance rates for Asian re finance and home improvement loan applicants are closer to those of His panic applicants than to those of whites. ■ 11 We count all applications approved by the lender as “origina tions.” In fact, some applications (2.9 percent) are approved by the lender but are subsequently withdrawn by the borrower. In these cases, the loan will not actually be made. their subsidiaries receive a disproportionately large share of low-income applications; on the other hand, a disproportionately small percent age of thrift business comes from low-income borrowers or tracts. Larger lenders also receive disproportionately fewer low-income loan ap plications. Again, though, this appears to be a result of the betw?een-MSA distribution of appli cants. Within MSAs, the largest lenders tend to receive more low-income applications. Finally, we note that the specific measure used to compare minority and nonminority lending or low-income and high-income lend ing has little impact on the distribution across lenders. The same patterns are found when minority lending is measured by the number of minority applications, the number of applica tions from minority census tracts, the dollar value of minority applications (not showTi), or the dollar value of applications from minority tracts (not showrn). Similarly, for low-income lending, the cross-lender distribution is the same whether lending is measured by the num ber or dollar value of loans or whether income is measured by the applicant or tract. III. Variance in Lending Patterns The sample statistics reported in the previous section reflect the average percentage of loan applications from minority and low-income in dividuals (or tracts) and the average approval rate on those applications by various types of lending institutions. These statistics could be thought of as describing the prototypical lender in the mortgage market, not the actions of any individual lender operating in that market, and as ignoring the variation across these individ ual lenders. In this section, we compare three measures of individual lender performance: 1) minority and low-income origination rates (the share of loans originated going to minorities or low-income individuals or tracts), 2) application rates (the share of applications received from minorities or low-income individuals or tracts), and 3) relative approval rates (differences in the actions taken on applications). We first address the relationship among these three measures. Because origination rates are equal to the product of application rates and relative approval rates, we would like to know the extent to which credit origination differences among lenders stem from the for mer factor versus the latter. That is, if we are concerned about credit flows to minority and 22 TABLE 3 Analysis of Variance in Origination Rates across Lenders, 1990 HMDA Origination Rate Percent Attributable to Variance in:b N u m b e r of Lenders N u m b e r of Applications Mean Standard Deviation2 Regression R-Squareda Minority Number Dollar value Center city 11,598 11,598 8,548 1,867,211 1,867,211 745,161 0.16 0.14 0.23 0.18 0.18 0.22 0.92 0.91 0.93 86.7-90.7 87.4-91.1 82.5-88.5 9.3-13.3 8.9-12.6 11.5-17.5 Minority Tracts Number Dollar value 8,846 8,846 1,624,207 1,624,207 0.20 0.17 0.19 0.19 0.91 0.91 88.7-91.9 89.7-92.2 8.1-11.3 7.8-10.3 Low-income Applicants Number Dollar value Center city 13,651 13,651 9,668 1,918,018 1,918,018 764,423 0.21 0.16 0.26 0.19 0.19 0.23 0.91 0.92 0.93 85.4-87.8 88.4-90.7 81.7-85.8 12.2-14.6 9.3-11.6 14.2-18.3 Low-income Tracts Number Dollar value 11,024 11,024 1,566,699 1,566,699 0.32 0.27 0.24 0.23 0.94 0.94 90.2-92.6 93-3-95.3 7.4-9.8 4.7-6.7 Minority/Low-Income Relative Application Rate Approval Rate a. Expressed as deviation around MSA means. b. M inim um and m axim um contributions to variance based on deviations around MSA means. SOURCE: Authors’ calculations. low-income applicants and neighborhoods, does variation across lenders arise primarily from differences in treatment or in application rates? An approximate answer to this question can be obtained by estimating the following equation: (1) O rigination rateL = $1MSAL + (3, application rateL + P3 relative approval rateL + eD where the origination rate for lender L equals minority (or low-income) originations as a por tion of total originations, MSAL is a vector of dummy variables indicating the metropolitan area in which lender L operates, application rate is minority (or low-income) applications as a share of total applications, and relative approval rate is the minority (or low-income) approval rate divided by the overall approval rate. The MSA fixed effects control for differ ences in the mortgage lending market that are common to all lenders in that market but may vary across markets, such as the size of the mi nority population or lending practices. Fitting equation (1) provides an estimate of the relative importance of application rates and approval rates in explaining variation in origi nation rates. Unfortunately, as with any regres sion, because application rates and relative ap proval rates are likely to be correlated, we cannot compute a precise estimate of the con tribution of each component to the variation in origination rates. However, several approxi mate estimates are possible. We determine a lower bound on the contribution of each com ponent by estimating its marginal contribution; that is, the additional variation in origination rates explained by adding the component to a model containing the other component. We compute an upper bound on the contribution of each component from its univariate fit — the proportion of the variation in origination rates that it explains by itself. The difference in the lower and upper bound estimates derives from how the impact of the covariance be tween the two components is assigned. The lower bound estimate assigns the covariance to the other component, and the upper bound assigns the full effect of the covariance to the variable in question. Table 3 reports the allocation of variance for estimates of equation (1) for several differ ent origination rates. The variance associated with MSAs is removed from the total before we measure the contributions of the application 23 and relative approval rates. Thus, we are de composing the variance in the deviations about MSA means. Rowr 1 shows the variance in de composition across lenders for the origination rate of minority individuals. Row 4 shows the decomposition for originations in minority tracts. Rows 6 and 9 show the decomposition for low-income individuals and tracts, respec tively. Rows 2, 5, 7, and 10 report decomposi tions for origination rates weighted by dollars. Finally, decompositions for minority and lowincome individuals applying in central cities are shown in rows 3 and 8. For each decomposition estimated, the sam ple includes all lenders for which the origina tion rate, application rate, and relative approval rate are defined. We note that this reduces the sample of lenders substantially from the full sample reported in tables 1 and 2. For example, the sample used for minority individuals in cludes only 11,598 of the 20,695 HMDA-reporting lenders (40 percent were dropped because they had no minority applicants and 3 percent because they had no originations of any type). However, these lenders received 1,867,211 of the 1,984,688 full sample applications (94 per cent). Moreover, the percentage of applications made by minorities in the decomposition sam ple (14.1 percent) is only slightly higher than in the full sample (13-3 percent). For each decomposition, we present several statistics. In columns 3 and 4, we show the mean and standard deviation of the origination rate across lenders. Note that the mean origina tion rate across lenders is generally higher than the sample average, indicating that smaller lenders make more of their loans to minorities or low-income individuals. In column 5, we show the R-squared of the estimated equation (1). Both the R-squared and standard deviations are adjusted for deviations about MSA means. Finally, in columns 6 and 7, we show the per centage of the total variation of the origination rate that can be attributed to the application rate or relative approval rate, adjusted for MSA fixed effects. We find that the overwhelming majority of the cross-lender variance in minority origina tions is attributable to differences in minority application rates. Differential approval rates by race account for a relatively small portion of the variance. For example, after controlling for MSA differences, 87 to 90 percent of the vari ance in originations to minority individuals is captured by lender-specific differences in mi nority application rates; only 10 to 13 percent stems from different approval rates for these applications. This narrow range suggests that the contribution of the covariance is quite small, which greatly enhances our ability to identify the importance of the application rates. Our results concerning low-income lending are much the same as those for minority lend ing. The only difference is that the ranges for low-income lending are somewhat larger than those for minority lending, indicating that the covariance between application rates and rela tive approval rates contributes more to the cross-lender variance in low-income origina tions than it does to the cross-lender variance in minority origination rates. The results are vir tually identical when dollar values are used or when census tracts rather than individual appli cant characteristics are examined. Restriction of the sample to central cities does little to alter the results, other than showing a slight increase in the variance that may be attributable to rela tive approval rates. To examine the robustness of these results further, table 4 reports the allocation of the variance across lenders in minority originations for lenders grouped by type, size, and market share of institution, and by MSA size and per cent minority. The dominance of differences in application rates as the source of lender differ ences in minority origination rates holds across all types of lenders, all sizes of lenders (meas ured in terms of both the volume of applica tions received by the lender and the lender’s market share), and types of MSAs. Even for mortgage banks (subsidiaries of depository in stitutions as well as independents), where the contribution is smallest, cross-lender differences in application rates account for at least threequarters, and may account for as much as 90 percent, of the variance in minority originations. The contribution of minority application rates to the variance in originations is smallest among small lenders, regardless of the type of lender. For the largest lenders (those with 500 or more applications), differences in application rates account for 93 to 99 percent; for lenders with less than 100 applications, they account for 85 to 89 percent. This is also true when size is measured by market share. Differences in lender minority application rates account for 96 to 97 percent of the variance across those with 5 percent or more of the market, and for 84 to 89 percent across lenders with less than 1 percent of the market. Although not presented here, simi lar conclusions hold for the decomposition of minority tracts and low-income individuals and tracts by lender types and size. 24 T A B L E 4 Allocation of Variance in Minority Origination Rates by Type and Size of Lender, 1990 HMDA Origination Rate Percent Attributable to Variance in:b Mean Standard Deviation3 Regression R-Squareda Minority/Low-Income Application Rate Relative Approval Rat Type of Institution Commercial banks Thrift institutions Credit unions Bank subsidiaries Thrift subsidiaries Other mortgage banks 0.13 0.11 0.18 0.13 0.13 0.16 0.19 0.14 0.29 0.15 0.18 0.18 0.91 0.93 0.97 0.88 0.90 0.94 86.5-91.1 92.0-93-9 85.2-93.1 80.4-83-4 74.2-81.7 86.5-90.2 8.9-13.5 6.1-8.0 6.9-14.8 16.6—19.6 18.3-25.8 9.8-13.5 Size of Institution More than 500 applications 100 to 500 applications Less than 100 applications 0.13 0.09 0.15 0.09 0.08 0.21 0.99 0.96 0.92 92.8-98.8 96.5-98.0 85.0-89.3 1.2-7.2 2.0-3.5 10.7-15.0 Market Share of Institution More than 5 percent 1 to 5 percent Less than 1 percent 0.10 0.11 0.20 0.07 0.09 0.22 0.95 0.93 0.92 95.7-97.3 92.8-94.1 84.0-88.7 2.7-4.3 5.9-7.2 11.3-16.0 Size of MSA More than 25,000 applications Less than 25,000 applications 0.15 0.18 0 20 0.17 0.94 0.91 86.4-91.0 86.6-90.4 9.0-13.6 9.6-13-4 Percent Minority Applications More than 22 percent Less than 22 percent 0.36 0.13 0.24 0.17 0.94 0.92 76.6-86.8 87.7-91.0 13.2-23.4 9.0-12.2 Total 0.16 0.18 0.92 86.7-90.7 9.3-13.3 a. Expressed as deviation around MSA means. b. M inim um and m axim um contributions to variance based on deviations around MSA means. SOURCE: Authors’ calculations. We conclude that differences in the relative approval rates of minority and low-income loans account for only a small portion of the variance across institutions in the portion of originations going to minority and low-income applicants. In the following section, we examine various factors that may be contributing to the cross-lender vari ance in application and approval rates. IV. Sources of Cross-Lender Variance in Lending Patterns The outcome measures presented in the previous section are gross measures of lender perform ance. As such, they do not control for exogenous market factors that affect lender performance but that are beyond the lender’s control. The effects of any such exogenous factors should be re moved before constructing measures of lender performance to be used in CRA and fair lend ing evaluation. Although it by no means con tains an exhaustive list, HMDA includes infor mation on a number of applicant characteristics that arguably should be controlled for: loan size, applicant income, loan type (FHA/VA or conventional), and property location. To the extent that these factors are correlated with race, this specialization will contribute to the observed cross-lender variance in minority appli cation rates. Similarly, to the extent that they are correlated with creditworthiness, these applicant characteristics may also be contributing to the observed differences in relative approval rates. In this section, we examine the effect of remov ing these factors on our assessment of various measures of lender performance. We focus on individual minority application rates and relative 25 approval rates, although our results hold for lowineome and neighborhood taxonomies as well. We compute adjusted indices as the lender average for each variable after the effects of prop erty location and applicant characteristics are re moved. For the application and overall approval rate, this is estimated directly from a fixed-effects linear probability model, where the fixed effects are, by construction, the average of the depend ent variable after the effects of other variables are removed. The fixed-effects linear probability models used to compute the adjusted indices were estimated with the full 1,984,688 loan sam ple, and have the following form: (2) APPLICATIONiMn = fiAAC] + $m MSAm + fijTRACT^+ Pi LENDERi + nonminorities. Thus, the adjusted lender indi ces were taken either as the direct LENDER fixed effects estimated in equations (2) and (3) or computed as lender residuals averaged over the minority and nonminority subgroups. Final ly, we were also interested in computing the average lender “quality” of applicants as meas ured by their average AC and TRACT effects. The exact construction of each of the variables used in this portion of the analysis is 1) the average economic characteristic effects of the Zth lender’s applicants, A C apr. minority " AAC/ NJ • fo r al1 minority applicants /' ^ ^ a p r . nonminority ~ ^‘ k e L ^ A ^ ^ ' k ^ ^ k ' (3) APPROVALiMTL = r AACi + V RRACEt + Tm MSAm + T,TRACTt + TlLENDERl + 2) the average census tract effects of the lender’s applicants, UiMTL ’ where APPLICATION is coded one if the zth applicant using the Ith lender in the M th MSA and rth census tract is a minority (native Amer ican, black, or Hispanic) and zero otherwise; and APPROVAL is coded one if the /th appli cant loan using the Zth lender in the M th MSA and Tth census tract is approved and zero otherwise. AC is a vector of application charac teristics reported in the HMDA data, including gender, marital status, occupancy, income, loan amount, income-to-loan ratio, loan type, and interactions among these variables. RACE includes dummy variables for six applicant and two co-applicant racial categories. The racial dummies are also interacted with FHA and VA loan dummies. MSA, TRACT and LENDER are dummy variables indicating which of the 340 MSAs, 40,008 census tracts, and 20,695 lenders the application relates to, and u and v are re siduals. By construction, the MSA effects are normalized to have an overall mean of zero, and within each MSA, the lender and tract ef fects are normalized to have means of zero.12 Adjusted indices for the minority and relative approval rates are more complicated to estimate because they involve the ratio of predictions for two groups. For these calculations, we used variants of the fixed effects, computed by averaging lender residuals from the overall ap proval rate model separately for minorities and ■ nonminority applicants k\ 12 Estimates of these regressions are available from the authors upon request. TRACTapp='Ll t $ TTRACTlIN, TRACTa^r minority = ^/e ¡T -¡TRACTj/Nj, for all minority applicants j, TRACTapr nonminority = for all nonminority applicants k; ^kr 3) and the adjusted lender indices, estimated directly as fixed effects or averaged separately for minorities and nonminorities, LENDER L E N D E R ^ - T„ LENDER^ minority MINORITY'APPROVAL RATE — ACapr, minority — TRACTapr, minority —Hjei rR RACE./N- - r M, for all minority applicants j, LENDERapr, nonminority NONMINORITY APPROVAL RATE —ACapr, nonminority — TRACTapr, nonminority ~ ZkeLrR RACEkNk - r ip for all nonminority7applicants k, where N, N , and Nk are, respectively, the total, minority, and nonminority number of applicants to the lender and M is the MSA of the lender. Four different measures of lender loan activ ity were regressed against these constructs, and a variance decomposition similar to that 26 T A B L E 5 Allocation of Institutional Differences, Percent Deviations around MSA Means, 1990 HMDA Applicant economic characteristics Census tract Overall lender effect Unexplained lender effect iMinority Application Rate Relative Approval Rate Minority Approval Rate Overall Approval Rate 0 8 2.6 2.4-4.6 4.0-5.9 70.7-74.8 91.0-92.7 2.5-5.7 3.Ó-4.2 26.4-38.3 53-8-65.9 3.5-10.9 21.9-28.9 . - 2 . 0 - 3.2 88.7-91.1 SOURCE: Authors’ calculations. performed in the previous section was under taken. The four measures were 1) the minority application rate, which was regressed against A C app and T R A C T app, 2) the relative approval rate, which was re gressed against A C apr minorny, A C apr nonminority> T R A C T apr minority ’ 1 'K A d a p r ; nonminority ’ L E N D E R apr minority > Z iïT V D Æ / ? ^ nonminority ’ 3) the minority approval rate, which was regressed against A C apr minorny, A C ap r nonminority > ^ ^ ^ ^ a p r minority ’ ^ ^ ^ ^ a p r , nonminority ’ L E N ,D F • l Rxapr, nonminority > 4) the overall approval rate, which was re gressed against A C apr minoritv, A C apr nonminontv, T R A C T apr minority ’ and T R A C T apr nonminority ’ Each regression was run with MSA dummies; thus, we analyze within-MSA variation. The contribution of each component to the overall variance in minority application rates is identi fied using the same variance decomposition procedure as in the previous section. Again, because we are looking at a decomposition of variance, the amount attributable to each source can only be approximated. As in the previous section, lenders used in these regressions were limited to the 11,598 lenders for whom all de pendent variables were defined (at least one minority applicant and one approved loan). The A C and T R A C T components can be thought of as exogenous factors, potentially be yond the lender’s control. The adjusted lender effects in minority application and approval rates constructed above ( L E N D E R a p p , L E N D E R apr, L E N D E R apr milloritv, and L E N D E R apr nonminority) can be interpreted as lender-specific differences in application and approval rates controlling for applicant characteristics and property loca tion. The variance decomposition allows us to compare the unadjusted measures of lender performance, as represented by the gross mi nority application and relative approval rates, with the adjusted indices, as measured by the L E N D E R variables. If the L E N D E R variables account for most of the variation in the gross measures, then regulators may be able to use gross performance measures without serious cost. If, on the other hand, A C and T R A C T ac count for a substantial portion of the variation in the gross measures, this may be an inappro priate decision. Table 5, column 1 shows the decomposition of the cross-lender variance in minority applica tion rates. Differences in application character istics account for 1 to 3 percent of the withinMSA variance across lenders. Much more sur prisingly, differences in the census tracts from which lenders receive applications account for only 22 to 29 percent of the variation, with 71 to 75 percent of the variation across lenders attributable to the unexplained pure L E N D E R effect. This means that most of the variation across lenders in the number of minority appli cations they receive does not stem from the fact that they serve different neighborhoods, but from how they draiv applicants w ithin neigh borhoods. This result is robust to a number of variations, such as ignoring MSA effects or weighting the regression by the number of applications received by the lender, and runs counter to the conventional wisdom that vari ation in the racial composition of the neighbor hoods served by lenders is the major source of 27 cross-lender variation in the proportion of mi nority applications received.13 Column 2 of table 5 shows the decomposi tion of the within-MSA variance in relative ap proval rates. Between 2 and 5 percent of the difference across lenders can be attributed to variation in the application characteristics, and between 4 and 6 percent can be attributed to census tract location. The overwhelming major ity of variation (91 to 93 percent) cannot be ex plained by these factors and is attributable to the pure lender effect. Similar conclusions are reached when we use the same methodology to examine sources of cross-lender variation in minority approval rates (table 5, column 3). Applicant economic and census tract effects are small. The overall standard of the institution, measured by the non minority lender effect, explains about one-third of the within-MSA variation (that is, minorities who apply to institutions with low approval rates for all applicants tend to be approved at lower rates, ceteris paribus). However, more than half of the variation in minority approval rates cannot be explained by any of these factors. These re maining differences may reflect differential treat ment of minority applications or differences in the unobserved characteristics of the loan appli cation; without additional information, it is impos sible to make a determination. It appears that this large component of unex plained variation is consistent with evidence of significant idiosyncratic lender behavior. Column 4 of table 5 reports the decomposition of the cross-lender variance in overall approval rates (minority and nonminority) based on the same methodology7used above. About 90 percent of the within-MSA variation in overall lender approval rates cannot be explained either by applicant char acteristics (as we measure them) or by census tract. These results suggest that the adjusted meas ures of lender performance account for the vast majority of variation in the gross measures. This finding is further examined in table 6, which reports the differences in gross and ad justed performance measures across various lender groups arranged by type, size, and mar- ■ 13 The potential contribution of census tracts is larger when the re gression is weighted by the number of applications each lender received. Since this decomposition focuses on within-MSA variation and gives most weight to the largest lenders within the MSA, it is difficult to separate the lender effect from the census tract effect. As a result of the covariance be tween the two, the range of the contribution of each is quite large (27 to 69 percent for census tracts and 30 to 63 percent for lender effects). We note that even in this decomposition— the most favorable case for census tract ef fects— at least 30 percent of the variance across lenders cannot be explained by loan application characteristics or by the racial composition of the neigh borhood from which the lender draws applications. http://fraser.stlouisfed.org/ Federal Reserve Bank of St. Louis ket share, and by size and percent minority in the MSA. The difference between the gross and adjusted standard deviations for each group reflects the importance of the control factors, AC and TRACT. The first column of table 6 is the cross lender variance in minority application rates; the second column is the variance in the pure lender effect on the application rate. For the full sample of lenders, cross-lender variance before controll ing for the applicant characteristics and property location is 0.20; after controlling for these factors, the variance is 0.14. Thus, about 30 percent of the cross-lender variance in minority application rates is explained by control factors. These fac tors account for a larger portion of the variance across commercial banks than for other types of lenders. They also account for more of the vari ance across lenders with large market shares, and those in MSAs with large numbers of minor ity applicants. The control factors explain relatively little of the cross-lender variance in overall approval rates (columns 5 and 6) or in minority approval rates (columns 7 and 8). However, they do ex plain a sizable portion of the cross-lender variance in relative approval rates (minority ap proval rate/overall approval rate). Before con trolling for the factors in our model, the cross lender variance in relative approval rates is 0.37; after controlling for them, the variance is 0.26 — almost 30 percent lower. As was the case with application rates, control factors ac count for relatively more of the variation in ap proval rates for commercial banks and their mortgage subsidiaries, for lenders with large market shares, and for lenders in MSAs with larger numbers of minority applicants than other institutions. It is also interesting to examine the relation ship between the pure lender effect on minor ity application rates and the pure lender effects on absolute and relative minority approval rates. Overall, those lenders with higher-thanexpected minority application rates (positive lender effects) are associated with slightly higher-than-expected minority approval rates, both absolute and relative. However, the corre lations are surprisingly small (0.001 and 0.024, respectively), suggesting that minority appli cants do not seem to be applying to lenders where their probability of approval is higher. 28 T A B L E 6 Standard Deviation of Minority Lending across Lenders Controlling for Applicant Characteristics and Property Location Minority Application Rate Minority Origination Rate Overall Approval Rate Minority Approval Rate Relative Approval Rate Gross2 Adj.b Grossa Adj.b Gross'1 Adj.b Gross2 Adj b Gross3 Adj b Type of Institution Commercial banks Thrift institutions Credit unions Bank subsidiaries Thrift subsidiaries Other mortgage banks 0.23 0.18 0.28 0.16 0.19 0.19 0.14 0.11 0.26 0.12 0.15 0.15 0.18 0.15 0.19 0.14 0.14 0.17 0.15 0.11 0.26 0.12 0.15 0.15 0.17 0.12 0.16 0.20 0.19 0.17 0.16 0.11 0.16 0.18 0.18 0.16 0.34 0.27 0.32 0.32 0.30 0.28 0.32 0.25 0.31 0.30 0.28 0.27 0.41 0.29 0.38 0.40 0.38 0.33 0.28 0.23 0.27 0.27 0.27 0.23 Size of Institution More than 500 applications 100 to 500 applications Less than 100 applications 0.13 0.12 0.23 0.05 0.05 0.17 0.13 0.11 0.19 0.05 0.05 0.17 0.11 0.13 0.19 0.09 0.11 0.18 0.16 0.21 0.35 0.13 0.20 0.34 0.12 0.21 0.43 0.07 0.16 0.31 Market Share of Institution More than 5 percent 1 to 5 percent Less than 1 percent 0.16 0.14 0.23 0.05 0.06 0.18 0.16 0.14 0.25 0.05 0.06 0.19 0.12 0.15 0.19 0.10 0.13 0.18 0.23 0.27 0.35 0.10 0.13 0.18 0.24 0.30 0.43 0.18 0.22 0.30 Size of MSA More than 25,000 applications Less than 25,000 applications 0.20 0.20 0.14 0.14 0.21 0.21 0.14 0.15 0.17 0.17 0.16 0.16 0.31 0.29 0.27 0.29 0.38 0.32 0.24 0.27 Percent Minority Applications More than 22 percent Less than 22 percent 0.28 0.17 0.18 0.14 0.30 0.18 0.18 0.14 0.19 0.16 0.19 0.15 0.28 0.31 0.27 0.30 0.32 0.37 0.21 0.27 Total 0.20 0.14 0.21 0.15 0.17 0.16 0.31 0.29 0.37 0.26 a. Gross cross-lender variation not controlling for applicant characteristics or property location. b. Adjusted cross-lender variation controlling for applicant characteristics and property location. SOURCE: Authors’ calculations. V. Conclusion This paper uses recently released HMDA data to examine differences in minority and lowincome lending patterns across lending institu tions. The new data allow us to identify both the application and the action taken on that ap plication by the lender, thus enabling us to son out lender behavior from applicant behav ior to a greater extent than allowed by pre vious data. We therefore can determine the extent to which the differences across lenders in minority (low-income) originations found in earlier studies reflect differences in minority (low-income) application rates across lenders as opposed to differences across institutions in their minority (low-income) approval rates rela tive to their overall approval rates. Our examination of the HMDA data reveals the following patterns related to lender differences in minority lending. First, lender differences in mi nority approval rates account for only about 10 percent of lender differences in minority loan originations: Differences across lenders in mi nority application rates account for the remain ing 90 percent. Second, wre find that very little of the lender variation in either minority appli cation rates or approval rates can be attributed to applicant characteristics. Third, somewhat surprisingly, we detemiine that while property location explains a nontrivial portion of the cross-lender variance in application rates, most variation stems from differences in the applicants that lenders attract within the neighborhoods they serve. Finally, the correlation across lenders between minority application rates and minority approval rates is quite small. Minorities do tend to apply to lenders with low overall approval rates, but within this class of lenders, minority application rates are highest at those lenders with relatively large minority approval rates. 29 These results suggest that gross measures of lender performance may work fairly well in implementing a more quantitative regulatory evaluation system. They also suggest that appli cation rate measures should play a particularly important role if increased credit flow's to se lected groups are the desired objective. Inter estingly, even here, gross application rate measures may work fairly w^ell in differentiat ing among lenders. We caution, however, that even though our research indicates that lenders vary enormously in terms of their relationships with minority and low-income applicants, we can say little about the reasons for this varia tion. Differences may result from illegal prac tices, or simply from economic factors on both sides of the market. Furthermore, because a number of financial institutions have initiated new lending practices during the last few years, the observed variation among lenders may be narrowing. Regulators and the public should attain a better understanding of the variation in lenders’ practices before reaching definitive conclusions about how to use measures of such variation in enforcement of the CRA or fair lending laws. References Avery, Robert B. “Making Judgments about Mortgage Lending Patterns,” Federal Re serve Bank of Cleveland, Economic Com mentary, December 15, 1989. ________ , and Thomas M. Buynak. “Mortgage Redlining: Some New Evidence,” Federal Re serve Bank of Cleveland, Economic Review, Summer 1981, pp. 18-32. ________ , and Glenn B. Canner. “Mortgage Red lining: A Multicity Cross-Section Analysis,” Board of Governors of the Federal Reserve System, unpublished wrorking paper, 1983________ , Patricia E. Beeson, and Mark S. Sniderman. “Accounting for Racial Differences in Housing Credit Markets,” Federal Reserve Bank of Cleveland, Working Paper 9310, De cember 1993- Also forthcoming in Proceed ings o f a Conference on D iscrim ination a n d Mortgage Lending, U.S. Department of Housing and Urban Development. ________ , _________ , a n d _________ . “Under served Mortgage Markets: Evidence from HMD A,” Federal Reserve Bank of Cleve land, working paper, December 1994 (forth coming). Bradbury, Katharine L., Karl E. Case, and Con stance R. Dunham. “Geographic Patterns of Mortgage Lending in Boston, 1982-1987,” Federal Reserve Bank of Boston, New Eng lan d Economic Review, September/October 1989, pp. 3-30. Calem, Paul S. “The Delaware Valley Mortgage Plan: Extending the Reach of Mortgage Lenders,” Jo u rn a l o f Housing Research, vol. 4, no. 2 (1993), pp. 337-38. Canner, Glenn B. "Redlining and Mortgage Lending Patterns,” in J. Vernon Henderson, ed., Research in Urban Economics. Green wich, Conn.: JAI Press, 1981, pp. 67-101. ________ , and Dolores S. Smith. "Home Mortgage Disclosure Act: Expanded Data on Residential Lending,” Federal Reserve Bulletin, vol. 77, no. 11 (November 1991), pp. 859-81. ________ , a n d _________ . “Expanded HMDA Data on Residential Lending: One Year Later,” Federal Reserve Bulletin, vol. 78, no. 11 (November 1992), pp. 801-24. Munnell, Alicia H., Lynne E. Browne, James McEneaney, and Geoffrey M.B. Tootell. “Mortgage Lending in Boston: Interpreting HMDA Data,” Federal Reserve Bank of Bos ton, Working Paper No. 92-7, October 1992. Wienk, Ronald E. “Discrimination in Urban Credit Markets: What We D on’t Know and Why We Don't Know It,” Housing Policy Debate, vol. 3, no. 2 (1992), pp. 217-40. 30 The Efficiency and Welfare Effects of Tax Reform: Are Fewer Tax Brackets Better than More? by David Altig and Charles T. Carlstrom Introduction The 1980s was the decade of tax reform. The American economy experienced two major changes in federal personal income-tax legisla tion, the Economic Recovery Tax Act of 1981 (ERTA) and the Tax Refomi Act of 1986 (TRA86). But significant change was not limited to the United States. By 1989, tax legislation had been passed in Australia, Canada, Denmark, New Zealand, Japan, Sweden, and the United King dom, with proposals for reform pending in many other nations (see Tanzi [1987], Boskin and McLure [1990], and Whalley [1990b]). Although actual and proposed tax legisla tion within each of these countries wyas multi faceted, sometimes with substantial variance in details, the reform proposals shared certain broad characteristics across countries. Most striking among these was the uniform tendency toward lower top marginal tax rates, fewer rate brackets, and "base broadening.” For example, in the latest rounds of reform, top statutory marginal rates in the federal personal tax codes fell from 34 to 29 percent in Canada, 83 to 40 percent in the United Kingdom, and 50 to 31 percent in the United States.1 Corresponding http://fraser.stlouisfed.org/ to these changes were reductions in the num Federal Reserve Bank of St. Louis David Aitig is an assistant vice president and economist and Charles T. Carlstrom is an econo mist at the Federal Reserve Bank of Cleveland. For helpful comments, the authors thank Zsolt Besci, Finn Kydland, Eric Rasmussen, and seminar participants at Indi ana University and the Federal Reserve Banks of Cleveland, Minneapolis, and St. Louis. Susan Byrne provided valuable research assistance. ber of rate brackets from 10 to 3 (Canada), 11 to 2 (United Kingdom), and 15 to 3 (United States). These examples and others are summa rized in table 1. A major motivation for these changes was the growing perception that the distortionary effects of high marginal tax rates had resulted in sub stantial inefficiencies.2 Consequently, an es sential impulse for tax reform was — and is — the desire to create more efficient income tax sys tems by substituting base-broadening measures for high marginal tax rates. Reductions in the ■ 1 Effective marginal tax rates can differ from statutory rates due to special treatment of credits, deductions, and exemptions at certain thresh old income levels. An obvious example is the TRA86 provision for phas ing out personal exemptions for high-income taxpayers. ■ 2 In its 1984 report on early tax proposals, the Joint Committee on Taxation identified three major objectives of comprehensive reform: equity, efficiency, and simplicity. With respect to efficiency, the Committee wrote that “ ... a widely accepted goal of tax policy is that taxes should interfere as little as possible with the incentives to engage in specific types of eco nomic activity, except to the extent that Congress intends such effects... [A] major goal of tax policy is to reduce [inefficiencies] to as low a level as possible.” Furthermore, they indicated that “ ... in all [pending] pro posals, marginal tax rates are substantially reduced. This reduction ap pears to be motivated by efficiency and equity considerations.” See Joint Committee on Taxation (1984). 31 T A B L E 1 Specific Elements of World Tax Reform Country7 Top Marginal Tax Rate, Pre-Reform Pre-Reform Year(s) Number of Pre-Reform Brackets Top Marginal Tax Rate, Post-Reform Post-Reform Year(s) Number of Post-Reform Brackets Australia 60% 1980-86 5 49% 47 1987-88 1992 4 5 Austria 62 1982-88a 10b 50 1989 5 Belgium 72 1983-88 13b 50 1989-92 7 Canada 34 1987a 10 29 1988-92 3 Italy 65 1983-87 9 56 51 1988 1992 8 7 Japan 70 1984-86 15 60 50 1987 1988-92 12 Netherlands 72 1982-86a 9 66 60 1987-88 1990-92 5 4 New Zealand 66 1979-85 5 48 33 1986 1988-92 3 2 5 Sweden 80 1985a 11 72 50 1986 1991-92c 4 4 United Kingdom 83 1978a 11 60 40 1979 1988-92 6 2 United States 50 1983-85 15 33 31 1986 1992 3 3 a. Rate may have been in effect prior to earliest date indicated. b. Figures refer to num ber o f rate brackets in 1988. c. From 0 to 186,600 kronor (SEK), the national tax is a flat SEK 100. For incomes in excess o f SEK 186,600, the tax is SEK 100 plus 20 percent of the excess. SOURCES: Platt (1985); Tanzi (1987); Boskin and McLure (1990); Whalley (1990a, 1990b); various issues of the Organisation for Economic Co-operation and Development’s Economic Survey; and the 1982 and 1992 editions of Price Waterhouse’s Individual Taxes: A Worldunde Summary. number of rate brackets are presumably meant to reinforce this goal by simplifying the tax code and minimizing distortions through the creation of broad classes of income over which marginal tax rates are essentially flat. Although often implicit, this motivation for reducing the number of rate brackets is sometimes explicit in discussions of specific tax reform proposals. For example, in dis cussing the Takeshita reforms in Japan, Noguchi (1990, p. 118) describes the U.K. and U.S. changes in rate structures as “developments ... toward flatrate income taxes,” while Ishi (1989) refers to the rate structure implemented in Japan as a “modified flat-tax” system. However, a brief glance at figure 1, which depicts various vintages of Canadian, Japanese, and U.S. personal income-tax rate structures, reveals the problematic nature of concluding that a smaller number of rate brackets is less distortionary than a larger number. Although recent rate structures have wider bands of in come over which the marginal tax rate is flat, jumps in the marginal rate are much more sig nificant for some taxpayers. It is unclear, a priori, which structure will most significantly distort household consumption and work-effort decisions on net. Given the almost universal tendency toward reforms that simultaneously reduce the number of brackets and increase the distance between them, it is surprising that these issues have not been given more attention. That, then, is the goal of this paper. Using the well-knowTi dynamic fiscal policy frame work pioneered by Auerbach and Kotlikoff (1987), we examine the welfare and efficiency implications of shifting from linear to discrete 32 F I G U R E 1 Marginal Tax Rates Percent 40 CANADA 1987 j= £ 20 30 40 Income (thousands of 1989 Canadian dollars) 10 50 60 Percent marginal tax-rate structures. In other words, we consider the pure distortionary effects of replacing a tax structure with many (infinitely small) steps between marginal tax rates with one defined by two large bands of flat tax rates connected by a single, large discrete jump. We find that when our model is calibrated to match the main features of the U.S. econ omy, a hypothetical two-bracket code (roughly patterned after the rate structure in the 1989 U.S. personal income tax code) is less efficient than alternative linear-rate codes with similar average-tax progressivity and present-value revenue implications. By less efficient, we mean that there is no sequence of lump-sum transfers the government could feasibly imple ment that would make the shift from the linear to the discrete rate structure Pareto-improving.3 This finding is generally robust to parameter as sumptions and to the chosen method for equal izing revenues. This central message should serve as a cautionary note in the midst of grow ing political sentiment for further changes in the U.S. income tax code; Without disputing the merits of completely flat marginal tax rates, our results do not support the position that a modified flat-tax system is necessarily superior to all alternatives with steeply sloped marginal rate structures. I. The Simulation Model 2,000 1,000 3,000 4,000 Income (thousands of 1989 yen) 5,000 Percent 10 15 20 25 30 35 Income (thousands of 1989 dollars) NOTE: Figures are scaled to a m axim um o f S50,000 equivalent U.S. dollars. SOURCES: W halley (1990b); Ishi (1989); Boskin and McClure (1990); Internal Revenue Service, Statistics o f Income. Individual Tax Returns, 1965-89; and http://fraser.stlouisfed.org/ International Monetary Fund, International Financial Statistics, J u ly 1992. Federal Reserve Bank of St. Louis The model specification includes mathematical representations of the preferences and constraints of utility-maximizing households, the produc tion technology available to profit-maximizing firms, a government budget constraint, and a specification for the income tax code, all of which are described in this section. In combi nation with labor-, capital-, and goods-marketclearing conditions, a competitive equilibrium is constructed by finding aggregate quantities and prices that are, given the government’s be havior, consistent with the decentralized deci sions of individual households and firms. ■ 3 We argue only that a rate structure with revenue and progressivity properties similar to TRA86 is less efficient than the specific alternative we consider — not that all discrete marginal-rate schemes are less efficient. Although we believe that requiring the same revenue collections and averagerate progressivity is a sensible constraint on the alternative tax codes, our results should be interpreted in light of these particular restrictions. 33 Households and Preferences where (3b) Our model economy is populated by a sequence of distinct cohorts (individuals born on the same date) that are, with the exception of size, identical in every respect. Each generation lives, with per fect certainty, for 55 periods (interpreted as adult years) and is 1 + n times larger than its predeces sor. One can think of life as beginning at age 21 and ending at age 75. Individuals “born” at calendar date b choose perfect-foresight consumption ( c) and leisure (/) paths to maximize a time-separable utility function of the form 55 (1) £4 = I t= 1 where u,/ > 0,7 u„Z/ < 0,7 lim 1— / = °°,7 and u,Z is the partial derivative of the function u(-) with re spect to argument i. The preference parameter (3 is the individual’s subjective time-discount factor. We assume that (3 > 0, but do not strictly require that (3 < 1. Letting ats equal the sum of capital and government debt holdings for age t individuals at time s = b + t- 1, maximization of equation (1) is subject to a sequence of budget constraints given, at each time 5, by (2 ) a t s = ( l + rs) * , _ 1>s_ i + e, ws( 1 - lt s) + v, s - T (y*s) - cl s, where ws is the real pre-tax market wage at time 5, rs is the real return to assets held from time s- 1 to s, e, is an exogenous labor-efficiency endowment in the t period of life, and vt s re fers to lump-sum transfers received by age t individuals at time s.4 The function T(y* s) defines the amount of income tax paid, which depends on the tax base given by yt s = rsat_ Xs_ , + e, w5( 1 - ltt5) - d . The constant d represents a fixed level of deduc tions and exemptions used to convert gross in come to taxable income. In the linear marginalrate case, the function T(•) is defined as (3a) I GO = a + byt s, a, b > 0 defines the marginal tax rate as a linear function of taxable income. In the discrete tax case, the function is defined as (4) TDlsM. J xLy'.s s [ t 'y+ x H(y*s -y) if y* >y Note that at any time s, there are three dis tinct possibilities with respect to the budget constraint in the discrete tax case, correspond ing to the cases where y* < y, y* > y, and y* s = y The latter applies when individuals are at the kink in the budget constraint. In addition to equation (2), we impose the initial condition that all individuals are born with zero wealth and the terminal condition that the present value of lifetime consumption plus tax payments cannot exceed the present value of lifetime resources. In the absence of a bequest motive and lifetime uncertainty, this wealth constraint implies that a 55 v= 0. The Government The government in our model raises revenue through a combination of distortionary income taxes, debt issues, and lump-sum taxes. Gov ernment purchases of output equal zero at all times, and all government revenue is eventually redistributed to households in the form of lump-sum transfers. We specifically require that revenue raised from the income tax be re bated in the form of lump-sum payments to the individuals from whom it is collected. This al lows us to isolate the efficiency losses due to the distortionary nature of marginal tax-rate changes. Initially, we assume that D0, the amount of government debt at the beginning of time, is zero, and that the individual transfer payments, vt s, equal the amount of income tax revenue collected for all individuals age t at all times s. These assumptions, which we relax to calcu late efficiency measures in section V, imply that debt issues are zero for all 5. >’*s jLinear_ f x{y )d y , y= d ■ 4 Capital and government debt are assumed to be perfect substi tutes in households’ portfolios. Firms and Technology Output in the model is produced by competitive firms that combine capital (K ) and labor (Z) using a neoclassical, constant-returns-to-scale 34 production technology. Aggregate capital and labor supplies (in per capita terms) are ob tained from individual supplies as 55 (5) KS= Z 7. 5- 1 (1 + Ds-i l-l- n and 55 (6) i s= X t= 1 (1 + 77)'-55 Note that the capital stock at time 5 is given by private and public saving decisions at time 5 — 1. Also, recall that we initially assume Ds = 0 for all 5. The production function is wTitten in terms of the capital-labor ratio k as (7) scale steady-state cohort incomes to values con sistent with average household income in 1989, the year for which the tax code is calibrated. We discuss this choice in more detail below. In the benchmark model, we assume that the depreciation rate of physical capital is 10 percent per period, a choice that, again, is motivated by the arguments in Kydland and Prescott. The pop ulation growth rate is set to the postwar U.S. average of 1.3 percent per year, and the life-cycle labor efficiency profile {e/1^ 1 is calculated by interpolating estimates in Hansen (1986). Preferences We assume that preferences are isoelastic, spe cializing equation (1) to q - f ( k , ), where qs is per capita output and / (• ) is de fined such that / ' > 0, f " < 0, l i m ^ ^ / ' = 0, and l i m ^ o / ' = The competitive wage rate and (gross) interest rate are given by 1_J_ °c ( 11) ' t, b + t - 1 1—— I °1 t, b + t - 1 --- -— + a£ 4 = X P '- ‘ i - i 1a, where the preference parameters o c , <5l , and (8) ws = qs - kf'(- ) a represent the intertemporal elasticities of sub stitution in consumption and leisure and the utility' weight of leisure, respectively. In our rs.= / '( • ) - 5, benchmark model, we assume o t. = 1, so that equation (11) becomes and (9) where 8 is the depreciation rate on physical capital. ln(c, b+t. l ) + CL II. Model Calibration t= 1 In order to quantify the model, it is necessary to choose particular values for the model’s parameters. In this section, we describe the choices that result in our benchmark model and discuss their rationale. Technology The simulation exercises reported in section IV assume an aggregate production technology given by ( 10) q - A k ®, where 0 is capital’s share in production and A is an arbitrary scale factor. Our benchmark value for 0 is 0.36, following Kydland and Prescott (1982). The value of A is chosen to t, b+ t - 1 i- l This form has the special property, not gener ally exhibited by specification (11), that the capital-labor ratio is invariant to the scale fac tor A in equation (10).5 Also, evidence from state-level data reported by Beaudry and van Wincoop (1992) suggests preferences that are logarithmic in consumption.6 ■ 5 Scale invariance follows from the fact that changes in the level of wages have offsetting wealth and substitution effects on individual labor supply decisions. Since scale invariance also implies that average hours worked will not change with growth, preferences similar in form to those in equation (11") often appear in the real business cycle literature (see King, Plosser, and Rebelo [1988]). ■ 6 Further, Beaudry and van Wincoop find no evidence supporting either nonseparabilities between consumption and leisure or the absence of time separability in consumption, results that generally support the specification in equation (11). However, it should be noted that their em pirical findings are based on a different model of aggregate consumption behavior than the one presented here. 35 We base the choice of C/, the intertemporal elasticity of substitution of leisure, on the exten sive empirical literature devoted to estimating the wage elasticity of the labor supply. This elastic ity, which we denote r^ , is related to <5t by (12) n = The Tax Code The benchmark tax code is patterned after the statutory U.S. personal tax code for 1989. Over the income region that is relevant in our simulations, the 1989 schedule was given by 20, ( 13) MaCurdy’s (1981) study of men’s labor supply suggests values for r|7 in the range of 0.1 to 0.45, a result that is largely confirmed in re lated studies (see Pencavel [1986]). However, Rogerson and Rupert (1991) argue that, be cause of corner conditions, estimates of the degree of intertemporal substitution obtained from conventional analyses of male labor sup ply are likely to be understated. Furthermore, despite greater disparity in estimates obtained from studies of female labor supply, there is broad agreement that the elasticity is higher for women (see Killingsworth and Heckman [1986]). Based on this evidence, in our benchmark model we set <51= 0.25 and choose the parameter a so that steady-state hours worked by an individual at peak productivity are slightly greater than onethird of total time endowment, which we take to be 16 hours per day. Most empirical studies find values for the subjective discount factor (3 at annual frequen cies to be in the neighborhood of 1.0 — some times slightly lower (Hansen and Singleton [1982]), sometimes slightly higher (Eichenbaum and Hansen [1990]). We choose a benchmark value of 0.99. Together with the other parame ter choices, this value results in a steady-state real pre-tax interest rate of about 3.7 percent (which corresponds closely to the [apparent] historical average of real pre-tax returns on long-maturity riskless bonds in the United States") and in a steady-state capital output ratio of 2.63 (which corresponds closely to the ratio of total capital to GDP in the United States over the 1959-90 period8). TDiscrete — 0.15 if y*t s< $30,950, 0.28 if y* s> $30,950. We refer to this tax code as the “tax-reform” case. The income levels obtained from the model are scaled to match those in the 1989 tax code as follows: The scale parameter A in the pro duction function of equation (10) is chosen so that the highest income in the model matches the average income level for the highest-paid age group found in 1988 Census Bureau data.9 We calculate the average for this group, which consists of persons aged 45-54, to be $44,217 in 1989 dollars.10 This value of A is then used in all subsequent simulations. To obtain tax able income, we subtract exemptions and de ductions of $11,206. III. Welfare Effects In this section, we examine the effects of shift ing to the tax-reform code from an alternative linear-rate code, under the maintained assump tion of revenue neutrality. Holding the struc ture of the discrete code constant, two natural approaches to achieving revenue neutrality are 1) choosing the intercept of the linear-rate code to equalize revenues, and 2) adjusting de ductions to equalize revenues. In each of our experiments, we consider an initial steady state under the linear-rate regime and examine the transition to a new steady state under the tax-reform regime.11 Thus, un der an intercept-adjusted approach, we para meterize the function x(y) in equation (3b) as ■ 9 Recall from our previous discussion that household utility func tions are chosen so that real outcomes are unaffected by the choice of ■ 7 See Siegel (1992), which reports average rates for the 1800— 1990 period. We note, for the record, that average real rates appear to dif fer significantly across particular subperiods. Specifically, real returns to long-term bonds averaged 1.46 percent between 1889 and 1978, but are 5.76 percent outside that interval. ■ 8 The measure used to construct the U.S. capital stock is the constant-cost net stock of fixed reproducible tangible wealth reported in the January 1992 Surveyof Current Business, compiled by the U.S. De partment of Commerce. This measure includes consumer durables and http://fraser.stlouisfed.org/ Federal Reserve Bankgovernment of St. Louis capital. A. ■ 10 The data used in constructing this variable were taken from the Bureau of Labor Statistics’ Current Population Reports, series P-60, no. 166. The cohort mean is obtained by multiplying the median income of families with household heads aged 45-54 by the ratio of average to me dian family income for the entire population. All money values in this pa per are quoted in 1989 dollars. ■ 11 The experiments we report involve unanticipated changes in the tax regime. We have also conducted analyses (not reported) with antici pated regime shifts and found that our conclusions are robust. 36 F I G U R E 2 Marginal Tax Rates (Benchmark Parameters) Marginal tax rate (percent) 0.30 Tax reform 0.28 0.26 0.24 0.22 Linear-deduction adjusted 0.20 Linear-intercept adjusted 0.18 0.16 0.14 0.12 - / ... J ......... 27 I _______1...... ... _ L _ _ 1 _____ _ L _ 33 39 45 51 57 ~1— i 63 69 . 75 Age of population SOURCE: Authors’ calculations. TABLE 2 Average Tax-Rate Comparisons: Steady-State, Benchmark Parameters (percent) Low Income Median Income High Income Tax reform code 4.1 10.9 11.8 Linear-rate code, intercept adjusted to equalize revenues 3.3 10.8 11.9 Linear-rate code, deductions adjusted to equalize revenues 0.1 10.7 12.1 SOURCE: Authors' calculations. (14) and choose the intercept \ |/ so that the present value of income tax revenues generated by the linear-rate code is acceptably close to the present value of revenues generated by the tax-reform transition path and steady state.12 Under the http://fraser.stlouisfed.org/ alternative deduction-adjusted approach, we set Federal Reserve Bank of St. Louis \ j/ = 0.146 and choose the deduction to match the revenue levels.13 For the benchmark model, this approach yields deductions of $14,561 in the initial steady state. Figure 2 shows the steady-state, life-cycle path of marginal tax rates for the tax-reform and two linear-rate regimes. For the interceptadjusted linear-rate code, approximately 55 percent of the population, accounting for an equal amount of steady-state income, face lower marginal tax rates than they would under the tax-reform system. The highest marginal tax rate in the linear-rate case is approximately 20 percent, as opposed to 28 percent in the tax-reform regime. For the deduction-adjusted linear-rate code, things are slightly different: Approximately 35 percent of the population, accounting for 42 percent of steady-state in come, face lower marginal tax rates than they would in the tax-reform case. Furthermore, the rate reductions are concentrated — and es pecially pronounced — at high income levels. The highest marginal tax rate in the deductionadjusted linear-rate scenario is approximately 22 percent. In addition to the revenue implications, the progressivity of each tax structure is a key ele ment in considering the comparability of the different tax codes. Information on average taxrate progressivity, provided in table 2, is one convenient way of examining progressivity. Although no more than an informal summary of the nature of a particular tax code, this meas ure does provide a sense of how average tax liabilities are related to income, highlighting the sort of comparisons often invoked in dis cussions of alternative tax regimes. As claimed above, the results in table 2 do suggest that in the long mn, the tax-reform and linear-rate codes (especially the intercept-adjusted variant) exhibit ■ 12 By “close,” we specifically mean within 0.001 percent. The slope of the function in equation (14) is obtained by fitting a linear regres sion to the 1965 statutory tax code. The 1965 schedule was chosen as representative of the marginal rate structure in place over much of the 1964-78 period. Over the income range $0—$54,000, which covers the incomes generated by our model, a linear function is a reasonably good approximation of this statutory schedule. Present values are calculated as the Interest rates realized under tax re form, that is, along the transition path and in the new steady state. Meas uring revenue neutrality under a fixed assumption about interest rates, while not strictly consistent with ex post neutrality, seems consistent with the fashion in which tax legislation is actually contemplated. Further more, because the final, tax-reform steady state is the same in all simula tions conducted under a particular parameterization of the model, our choice delivers a common discount factor across like experiments. ■ 13 The choice of 0.146 is motivated by the same regressions used to determine the slope of the linear code. See footnote 12. 37 F I G U R E 3 Welfare Loss Due to Tax Reform: Basic Results Percent of wealth Generation NOTE: Each x on the horizontal axis corresponds to the oldest generation alive jc periods after the tax regime change. SOURCE: Authors’ calculations. similar degrees of progressivity, subject of course to the usual caveats about the validity of the average tax measure. Armed with these observations, we turn next to examining the welfare implications of shift ing from a linear-rate regime to the tax-reform regime. Throughout, we calculate welfare losses as the percentage increase in full wealth that must be given to an individual in the taxreform regime in order to compensate him for the switch to the linear code.14 Negative num bers therefore represent welfare gains associ ated with tax reform. Figure 3 illustrates welfare losses for differ ent age cohorts arising from an unanticipated change from the intercept-adjusted linear-rate regime to the tax-reform regime. Cohorts in fig ure 3 are identified by year of death. Thus, the welfare number for period 1 of the transition path represents the loss by an individual age 75 (fifty-fifth year of life) at the time the tax-reform regime becomes effective. All cohorts alive in the initial (linear-rate) steady state have died by period 55 of the transition path. The three sets of losses shown in figure 3 are calculated from the benchmark model and from two alter native parameterizations with different choices http://fraser.stlouisfed.org/ for intertemporal elasticity of substitution Federal Reserve Bank of the St. Louis in leisure. In the long run, tax reform generates wel fare losses, with the magnitude of the loss posi tively related to individuals’ willingness to shift leisure intertemporally. The intuition for this re lationship between welfare costs and GI can be appreciated by recalling that, because heteroge neity in the steady state is due strictly to life cycle characteristics, the highest incomes in the model are earned by individuals who are at their peak levels of labor productivity. As shown in figure 2, this is exactly the period of the life cycle for which tax reform implies higher marginal tax rates relative to the linearrate regime. The distortions on labor supply created by this fact are magnified for higher de grees of willingness to substitute leisure across periods of life. Thus, an apparently important factor in the relative efficiency of the linearrate structure is that, for roughly the same de gree of progressivity, the marginal tax rate faced by the highest-income individuals is lower than in the tax-reform case. The welfare effects apparent in figure 3 arise primarily from the direct distortions of the taxreform code vis-à-vis the hypothesized linearrate code, not from general equilibrium effects associated with changes in interest rates and wages.H In figure 4, we compare the welfare effects for the benchmark model with the effects obtained when the entire path of interest rates and wages is held fixed at the initial steady-state values. Although general equilibrium effects mitigate the welfare losses, the picture that emerges is little changed by the partial equilib rium assumption, especially in the long run. Note, however, that general equilibrium effects have a significantly greater impact on older co horts alive at the time of the regime change. Finally, we consider the welfare consequences when the linear-rate structure is chosen accord ing to the deduction-based method for equal izing revenues. Figure 5 shows the results of welfare calculations for these experiments. Relative to the intercept-adjusted experiments, the long-run welfare losses of tax reform are ■ 14 Full wealth, O , is defined as the present value of wage income when the entire time endowment is allocated to labor. Thus, i £twb+t-1 '=2n a + w i ) /=2 ■ 15 Recall that for the simulations in this section, we assume that lump-sum taxes and transfers maintain zero net tax payments for every cohort at every point in time. Therefore, wealth effects arise only as a re sult of changes in the aggregate levels of capital and labor, which are in turn reflected In interest rates and wages. 38 F I G U R E 4 Welfare Loss Due to Tax Reform: Partial versus General Equilibrium Percent of wealth Generation F I G U R E 5 Welfare Loss Due to Tax Reform: Deduction-Adjusted Results Percent of wealth FI GURE 6 Welfare Loss Due to Tax Reform: Alternative Consumption Elasticities Percent of wealth Generation NOTE: Each x on the horizontal axis corresponds to the oldest generation alive x periods after the tax regime change. SOURCE: Authors' calculations. somewhat lower when revenues are equalized in the linear-rate code by adjusting deductions. However, as reported in table 2, equalizing revenues by deduction adjustments results in greater average-tax progressivity than does either the intercept-adjusted linear code or the tax-reform code. Essentially, the increase in marginal rates on high-productivity/high-asset cohorts associated with tax reform is smaller when taxes are equalized by increasing deduc tions in the linear code, resulting in smaller long-run welfare losses. This last observation underscores a critical point that bears reemphasizing. The relative welfare effects of each of the tax structures we consider are dependent on the relative levels of marginal tax rates necessary to preserve revenue neutrality. The discrete code examined here generates welfare losses because a linearrate code with similar average-tax progressivity (or less progressivity, for that matter) allows the application of lower rates to the critical highincome cohorts. Finally, figure 6 presents the same experi ments for different degrees of intertemporal elasticity of substitution for consumption.16 Note especially that as consumers become less willing to substitute consumption across time, tax reform actually generates long-run welfare gains. However, welfare losses persist for the early years following the introduction of tax reform. This observation raises the interesting question of whether, for certain parameter choices, long-run welfare gains are large enough to offset short-run losses. We turn to this issue next. IV. Efficiency Effects The pattern of welfare effects in figures 3-5 clearly indicates that the contemplated shifts from the tax-reform regime result in efficiency losses. However, the welfare calculations pre sented do not provide a simple measure that summarizes the economic cost of such changes. Furthermore, as shown in figure 6, there are long-run welfare gains for some plausible al ternatives to the benchmark model. For these cases, the question is open regarding whether the shift to the tax-reform regime can be con- ■ 16 Recall that, given the preference specification in equation (11), equilibrium outcomes in the model are not invariant to the scale of the model when c c * 1. There are, however, other utility functions that allow more flexibility in the choice of the intertemporal consumption elasticity while preserving scale invariance, albeit at the cost of less flexibility in choosing intertemporal leisure elasticity. 39 T A B L E 3 Efficiency Losses Due to Tax Reform (percent of wealth) Revenues Equalized by Adjusting Intercept in the Linear-Rate Code Revenues Equalized by Adjusting Deductions in the Linear-Rate Code Benchmark 0.139 0.058 o, =0.17 = 0.50 0.065 0.235 0.027 0.103 p 0.080 0.211 0.030 0.088 = 1.005 = 0.976 o c = 0.20 0.162 0.074 = 0.33 0.151 0.069 = 0.07 0.160 0.066 0.362 0.362 0.362 0.165 0.165 0.165 5 0 7 = 0.20 (3 = 0.971 Ô = 0.07 SOURCE: Authors’ calculations. structed to maintain positive long-run welfare gains for some generations without diminish ing the lifetime utility of any other. In this section, we develop a measure of the efficiency costs of shifting from the hy pothesized linear-rate codes. Furthermore, for cases that generate gains for some generations, we ask whether there exists a set of transfers that preserves positive long-run gains while eliminating all welfare losses of cohorts alive along the post-reform transition path. To these ends, we calculate an efficiency measure in the spirit of the one introduced in Auerbach, Kotlikoff, and Skinner (1983). Spe cifically, let 5=1 be the time at which tax re form is introduced. To obtain our efficiency measure, we ask how much wealth can be taken away from cohorts born on or after 5=1 following the implementation of a fiscal policy with the following characteristics:17 (a) The government first introduces lum p sum taxes and transfers so that the lifetime util ity of all generations is maintained at the steady-state level realized in the initial, linearrate regime. For instance, in figure 6, cohorts experiencing welfare losses would receive ■ 17 Auerbach, Kotlikoff, and Skinner refer to the hypothetical gov ernment agency that implements these policies as the “Lump Sum Redis tribution Authority.” transfers while those enjoying welfare gains would be taxed. (b) Following the policy in (a), the govern ment’s long-run budget will be in surplus if the present value of taxes exceeds the present value of transfers, or in deficit if the converse is true. Because the long-run budget must bal ance, the government must choose a sequence of other transfers (for the surplus case) or taxes (for the deficit case) so that the present value of taxes less transfers equals zero. For the pur pose of constructing our efficiency measure, we assume that the budget is balanced by im posing lump-sum taxes, or by granting lum p sum transfers, that are a constant fraction of the full wealth of all generations born after the tax reform. If, after policy steps (a) and (b), generations along the transition path and in the new steady state are worse off, our efficiency measure is negative and equal to the percentage wealth loss suffered by each. A more detailed sketch of our procedure is offered in the appendix. Table 3 reports the results of efficiency cal culations for alternative parameterizations of the model. Losses are associated with all of the cases considered, even those in which there is a long-run welfare gain from shifting to tax re form. Thus, the short-run losses that occur in figure 6 dominate the long-run gains. For the benchmark model, the shift to the taxreform code results in an efficiency loss of 0.14 percent of full wealth when revenues are equal ized by adjusting the intercept of the linear-rate schedule. More generally, calculated losses range from 0.08 to 0.36 percent, depending on the chosen parameters. W hen revenues are equalized by adjusting deductions, the efficiency losses are uniformly smaller, but still range from 0.03 to 0.17 percent of full wealth. As shown, losses increase with individuals’ willing ness to shift resources intertemporally, again reflecting the fact that high-tax periods corre spond to periods of high relative saving rates and high labor productivity. To put some perspective on the magnitude of the efficiency losses, full wealth for each cohort in the tax-reform steady state is about 63 percent of total output. Thus, a reduction in full wealth of 0.14 percent represents an annual loss equal to about 0.09 percent of output in the model. Converting full wealth in the model to 1989 dol lars implies an efficiency loss equivalent to roughly $1,418 per person bom (or reaching working age) after the regime change. 40 V. Concluding Remarks Significant reductions in the number of mar ginal tax-rate brackets — that is, a trend to ward structuring systems of personal income taxation such that there exist wide bands of in come over which marginal tax rates are flat — have been a striking characteristic of world wide tax reform over the past decade. In this paper, we argue that this trend is not obviously accounted for by appealing to the efficiency gains inherent in tax codes with just a few brackets separated by discrete-rate jumps. Rela tive to revenue-neutral linear-rate structures, changing to a simple two-bracket discrete-rate structure creates efficiency losses in all of the numerical experiments we conduct. Further more, in most cases welfare gains are uniformly negative, even in the long mn. Two explanations come immediately to mind for the discrepancy between the reality of recent tax reforms and the results of our analysis. First, our analysis is conducted in a purely life-cycle framework. Hence, in steadystate equilibria, all cohorts face exactly the same life-cycle profile of relatively high taxes during periods of peak productivity and sav ing. The inefficiency of the discrete code that we consider follows in important ways from the fact that, holding average-tax progressivity constant, shifting from an equal-revenue linear code requires marginal tax-rate increases dur ing this phase of the life cycle. It is reasonable to conjecture that these ef fects would be mitigated in a more general framework that included intracohort heteroge neity. For instance, suppose that there existed two types of agents, “rich folks” and “poor folks.” It is conceivable that the two-bracket tax code could be structured so that the shift from the linear tax would result in poor folks facing only the lower rate and rich folks facing only the higher rate over their entire lives. In this event, the discrete tax code would be equivalent to a flat-tax regime, which would al most certainly create welfare and efficiency gains. In a slightly less extreme case, some por tion of each cohort would face the life-cycle pattern of rates on which we have focused, while for others, the poor-folk/rich-folk sce nario would be relevant. We have, however, conducted experiments in which we relax the representative life-cycle agent characteristic of the model presented in this article. In particular, we have replicated several of our welfare experiments in a frame work that includes 13 distinct life-cycle agent types with varying degrees of lifetime wealth and income. The qualitative aspects of our re sults are unchanged by this extension. A second explanation for the widespread adoption of rate-bracket reductions is that, per haps for administrative or political reasons, they are a necessary concomitant to lowering the level of tax rates and to the various basebroadening measures that also characterized tax reform in the 1980s. In this case, the ap proach advocated by Slemrod (1990), which emphasizes the broad institutional framework in which tax policy is chosen, may ultimately be necessary to fully understand the consequences of the income tax systems that have undeniably come to dominate industrialized economies. Appendix Notes on Calculating Efficiency Gains Our efficiency calculations require extending the government sector so that an individual’s budget constraint becomes (Al) a t s = (1 + rs) a t_ l s_ x+ z , ws{1 - lt ,) +vt,s~ T(y*.s) +Z,. s-c,'SThe only difference between the above equation and equation (2) in the text is the ad dition of zt s , which represents the net lum p sum transfers (negative numbers represent taxes) in excess of those necessary to offset income tax collections. Given this definition, the per capita level of debt evolves according to the relationship (A2) D s - 1 Z> = (1 + r ) —-—- —Z , 5 1+ n 5 where 55 (A3) z s= ^ ( 1 + n )55~ 'z, s. t=i Letting 5 = 1 be the first period of the transi tion path and normalizing the population at 5 =1 to unity, intertemporal budget balance for the government requires that 41 5=2 n o * ' - ) 1= 2 The algorithm for obtaining our efficiency measure proceeds in the following steps: (i) Conjecture a sequence of interest rates for the transition path and the new (tax-reform) steady state. (ii) Calculate the present value of lump-sum taxes, net of lump-sum transfers, that would be needed to maintain all cohorts at the initial steady-state level of utility. Refer to the resulting number as the “utility-compensation surplus,” or UCS. If positive, the UCS determines the present value of transfers that can redistributed by the government while maintaining long-run budget balance. If negative, the UCS determines the pres ent value of taxes that must be raised to maintain budget balance. (iii) Maintain the utility level of all cohorts alive at the time of the tax regime change, so that the government budget balance is satisfied by solving for the constant tax or transfer (as a percentage of each cohort’s full wealth) that can be applied to all subsequent cohorts while just exhausting the UCS. (iv) Use the path of taxes and transfers from steps (ii) and (iii), along with the associated path of government debt implied by equation (A2), to recalculate the entire problem, as de scribed in section II. (v) Update interest rates and the UCS until the procedures converge to an equilibrium that satisfies public and private budget constraints, all market-clearing conditions, and the firstorder conditions governing individual con sumption and leisure choices. Once the problem has converged, the efficiency gain is the per centage of full wealth that is redistributed to (or taken from) all cohorts bom after the change in tax regime, as calculated in step (iii). References Beaudry, Paul, and Eric van Wincoop. “Alterna tive Specifications for Consumption and the Estimation of the Intertemporal Elasticity of Substitution,” Federal Reserve Bank of Min neapolis, Institute for Empirical Macroeco nomics, Discussion Paper 69, July 1992. Boskin, Michael J., and Charles E. McLure, Jr., eds. World Tax Reform: Case Studies o f De veloped a n d Developing Countries. San Fran cisco: ICS Press, 1990. Eichenbaum, Martin, and Lars Peter Hansen. “Estimating Models with Intertemporal Sub stitution Using Aggregate Time Series Data,” Jo u rn a l o f Business a n d Economic Statistics, vol. 8, no. 1 (January 1990), pp. 53-69Hansen, Gary D. “Three Essays on Labor Indi visibility and the Business Cycle,” University of Minnesota, Ph.D. dissertation, 1986. Hansen, Lars Peter, and Kenneth J. Singleton. “Generalized Instrumental Variables Estima tion of Nonlinear Rational Expectations Mod els,” Econometrica, vol. 50, no. 5 (September 1982), pp. 1269-86. Ishi, Hiromitsu. TheJapanese Tax System. Ox ford: Oxford University Press, 1989Joint Committee on Taxation. Analysis o f Pro posals Relating to Comprehensive Tax Reform. Washington, D.C.: U.S. Government Printing Office, September 1984. Killingsworth, Mark R., and James J. Heckman. “Female Labor Supply: A Survey,” in O. Ashenfelter and R. Layard, eds., H andbook o f Labor Economics, vol. 1. New York: North-Holland, 1986. King, Robert G., Charles I. Plosser, and Sergio T. Rebelo. “Production, Growth, and Business Cycles "Jo u rn a l o f Monetary Economics, vol. 21, nos. 2/3 (March/May 1988), pp. 195-232. Auerbach, AlanJ., and Laurence J. Kotlikoff. Dy nam ic Fiscal Policy. Cambridge: Cambridge University Press, 1987. Kydland, Finn E., and Edward C. Prescott. “Time to Build and Aggregate Fluctuations,” Econometrica, vol. 50, no. 6 (November 1982), pp. 1345-70. ________ , _________ , and Jonathan Skinner. “The Efficiency Gains from Dynamic Tax Reform,” International Economic Review, vol. 24, no. 1 (February 1983), pp. 81 -100. MaCurdy, Thomas E. "An Empirical Model of Labor Supply in a Life-Cycle Setting,” Jou r n al o f Political Economy, vol. 89, no. 6 (December 1981), pp. 1059-85. 42 Noguchi, Yukio. “Tax Reform Debates in Japan,” in Michael J. Boskin and Charles E. McLure, Jr., eds., World Tax Reform: Case Studies o f Developed a n d Developing Countries. San Francisco: ICS Press, 1990, pp. 111-25. Pencavel, John H. “Labor Supply of Men: A Sur vey,” in O. Ashenfelter and R. Layard, eds., Handbook o f Labor Economics, vol. 1. New York: North-Holland, 1986. Platt, C. J. Tax Systems o f Western Europe, 3d ed. Brookfield, Vt.: Gower Publishing Com pany, 1985. Rogerson, Richard, and Peter Rupert. “New Esti mates of Intertemporal Substitution: The Ef fect of Corner Solutions for Year-Round Workers,” Jo u rn a l o f Monetary' Economics, vol. 27, no. 2 (April 1991), pp. 255-69Siegel, Jeremy J. “The Real Rate of Interest from 1800-1990: A Study of the U.S. and the U.K.,” Jo u rn a l o f Monetary' Economics, vol. 29 (April 1992), pp. 227-52. Slemrod, Joel. “Optimal Taxation and Optimal Tax Systems,” Journal o f Economic Perspec tives, vol. 4, no. 1 (Winter 1990), pp. 157-78. Tanzi, Vito. "The Response of Other Industrial ized Countries to the U.S. Tax Reform Act,” N ational Tax Jou rn al, vol. 40, no. 3 (Sep tember 1987), pp. 339-55. Whalley, John. “Foreign Responses to U.S. Tax Reform,” in Joel Slemrod, ed., Do Taxes M at ter? The Im pact o f the Tax Reform Act o f 1986. Cambridge, Mass.: MIT Press, 1990a, pp. 286-314. ________ . “Recent Tax Reform in Canada: Policy Responses to Global and Domestic Pressures,” in Michael J. Boskin and Charles E. McLure, Jr., eds., World Tax Reform: Case Studies o f Developed a n d Developing Countries. San Francisco: ICS Press, 1990b, pp. 73-91. 43 Fourth Quarter Working Papers Current Working Papers of the Cleveland Federal Reserve Bank are listed in each quarterly issue of the Economic Review. Copies of specific papers may be re quested by completing and mailing the attached form below. Single copies of individual pa pers will be sent free of charge to those who request them. A mailing list service for personal subscribers, however, is not available. Institutional subscribers, such as libraries and other organiza tions, will be placed on a mail ing list upon request and will automatically receive Working Papers as they are published. ■ 9411 Similarities and Dissimilarities in the Collapses of Three State-Chartered Private Deposit Insurance Funds ■ 9413 The Annuitization of Americans’ Resources: A Cohort Analysis ■ 9415 Bankruptcy Rules and Debt Contracting: On the Relative Efficiency of Absolute Priority, Proportionate Priority, and First-Come, FirstServed Rules by Walker F. Todd ■ 9412 The Burden of German Unification: A Generational Accounting Approach byJagadeesh Gokhale, Bernd Raffelhuschen, and Jan Walliser by Alan J. Auerbach, Jagadeesh Gokhale, Laurence J. Kotllkoff, John Sabelhaus, and David N. Weil ■ 9414 Loan Sales: Pacific Rim Trade in Nontradable Assets by Joseph G. Haubrich and James B. Thomson by Stanley D. Longhofer ■ 9416 Executive Compensation: A Calibration Approach by Joseph G. Haubrich and Ivilina Popova Please complete and detach the form below and mail to: Research Department Federal Reserve Bank of Cleveland P.O. Box 6387 Cleveland, Ohio 44101 Check item(s) requested Please send the following Working Paper(s): □ 9411 □ 9413 □ 9415 □ 9412 □ 9414 □ 9416 Send to: Please print Name Address City State Zip 44 The Federal Reserve Bank of Cleveland and the Journal of Money, Credit and Banking Announce: Derivatives and Intermediation November 2-3, 1995 Cleveland, Ohio The Federal Reserve Bank of Cleveland and the Journal of Money, Credit and Banking are jointly sponsoring a conference on Derivatives and Intermediation: Theory and Evidence. The growing derivatives market poses several chal lenges for policymakers. The first is to understand the sources of financial innovation resulting in the proliferation of these products. What economic forces make derivatives viable instruments? What gains do rational participants obtain from these con tracts, and why do they dominate transactions in the cash securities markets? To what extent does regulatory policy— bankruptcy rules, capital require ments, accounting rules, and deposit insurance— affect the market? In other words, does derivativerelated financial innovation stem from changes in the marketplace, or from changes in the regulatory environment? The answer is crucial to understand ing both derivatives and intermediation. The second challenge for policymakers is to understand how derivatives impact regulatory concerns in the areas of bank risk, payments system reform, and intermediary powers. Call for Papers The conference proceedings will be published in the and authors will receive an honorarium. Prospective contribu tors are invited to send a completed paper or de tailed abstract by May 30,1995 to: Journal of Money, Credit and Banking, Joseph G. Haubrich Research Department Federal Reserve Bank of Cleveland P.O. Box 6387 Cleveland, OH 44101-1387