On the Size of U.S. Government: Political Economy in the Neoclassical Growth Model

We study a dynamic version of Meltzer and Richard’s median-voter model of the size of government. Taxes are proportional to total income, and they are redistributed as equal lump-sum transfers. Voting takes place periodically over time, and each consumer votes for the tax rate that maximizes his equilibrium utility. We calibrate the model to match U.S. data. Key elements in the calibration are the income and wealth distribution and the parameters governing the labor-leisure and consumption-savings choices. The total size of transfers predicted by our politicaleconomy model is quite close to the size of transfers in the data. (JEL E60, H11, P16) ∗Krusell: Department of Economics, Harkness Hall, University of Rochester, Rochester, NY 14627, Institute for International Economic Studies, Stockholm, and CEPR (e-mail: [email protected]); Ŕıos-Rull: Department of Economics, 3718 Locust Walk, University of Pennsylvania, Philadelphia, PA 19104, and NBER (e-mail: [email protected]). We thank seminar attendants for feedback and Torsten Persson, Vincenzo Quadrini, Richard Rogerson, and two anonymous referees for particularly helpful comments. Both authors gratefully acknowledge support from the National Science Foundation. Krusell also gratefully acknowledges support from the Bank of Sweden Tercentenary Foundation. Countries differ widely in their public policy choices. For example, among the OECD countries, 1983 marginal tax rates on capital income varied between −90 and 49.5 percent; the average labor income tax varied between 24 and 63 percent; and the net total tax burden as measured by public expenditures varied between 26 and 54 percent of GNP. 1 Even larger differences would be revealed if one included countries on a lower level of development. Similarly, large differences in policies can also be recorded within given countries over time. To the extent one thinks actual policy outcomes affect economic performance, it seems an important task for economists to explore the origins of these wide disparities. One way of building a positive theory of policy is to assume that policies are chosen optimally. If they are indeed, then the large observed policy differences can only depend on differences in economic primitives such as preferences and technology. In contrast, the political-economy paradigm seeks to analyze how different policy-selection procedures/collective choice mechanisms affect policy outcomes. With this approach, differences in policy outcomes depend not only on mentioned primitives and on differences in population characteristics, but they also depend on the details of the procedure by which policies are selected. In this paper we set out to develop a macroeconomic model that can be used for quantitative theoretical analysis of political economy. Our framework of analysis in this paper is a dynamic extension of [29]’s study of the size of government. There, the median voter has lower than average labor productivity, and thus a proportional tax on labor used for making lump-sum subsidies implies a net transfer to this agent. Since labor supply is elastic, the political equilibrium tax rate is chosen by this agent so as to equate his marginal utility benefit from the transfer to his marginal utility loss from the tax distortion. Our version of Meltzer and Richard’s analysis has inequality not only in labor income but also in wealth, and wealth—capital income—can be taxed. This feature seems important to consider in a quantitative analysis for two reasons. First, the wealth distribution is much more skewed than the labor income distribution, and the upper tail of the distribution is an ample source for the median voter to potentially redistribute from. Second, taxing wealth accumulation is particularly distortionary, as we have learned from the literature on dynamic optimal taxation. The population is thus heterogeneous in two respects: labor productivity and asset wealth. The joint distribution of labor productivity, which we calibrate to U.S. data, is a key determinant of the equilibrium transfer level. Our setup also includes government consumption. We maintain the one-issue nature of voting assumed by Melter and Richard and consider voting over a common tax rate on all source of income. The quantitative assessment of the costs of redistributive taxation rests in part on assumptions about the key elasticities in the model: those governing the labor-leisure choices, and those determining the consumption-savings choices. For this, we select parameters in line with studies using similar macroeconomic models. The costs of redistributive taxation also depend on the level of government consumption, which we treat as exogenous here. We divide the different items of the government budget into those which can be considered transfers and those which cannot, and calibrate government consumption in the model accordingly. Given our calibration, the political equilibrium transfer level turns out to be quite close See [28] for data sources.