توزیع درآمد، محدودیت استقراض و سیاستهای بازتوزیعی

This paper sheds light on the relationship between income inequality and redistributive policies and provides possible guidance in the specification of empirical tests of such a relationship. We model a two-period economy where capital markets are imperfect and agents vote over the level of taxation to finance redistributive policies that enhance future productivity. In this context, we show that the pivotal voter is not necessarily the agent (class) with median income. In particular, the poor, who are more likely to be liquidity constrained, may form a coalition with the rich and vote for low redistribution. The effects of an increase in income inequality on the level of redistribution turn out to depend on whether the increase in inequality is concentrated among the poor or the middle class. Empirical results from a panel of 22 OECD countries provide preliminary evidence consistent with our main theoretical implications.

Given existing inequality in income and wealth distribution, economists have long since emphasized the risk that the political process in democratic systems might lead to high taxation of capital and reduced growth. In a typical political economy framework where all citizens have the right to vote, the lower is median wealth (income) relative to the mean, that is, the more unequal is the distribution of income, the higher will be the level of redistribution. Therefore, if redistribution depresses the incentives to invest, income inequality has a negative effect on growth. These causal links have been incorporated in models of political economy and growth by authors such as Alesina and Rodrik (1994) and Persson and Tabellini (1994) who have also provided some empirical evidence of a negative association between income inequality and growth.1 However, several empirical studies have shed doubts on the capacity of the political economy channel to explain the relationship between inequality and growth. Perotti (1996) was the first one to perform econometric tests on the various channels through which inequality can affect growth and concluded that the political economy channel is not supported by data. Benabou's (1996) survey on inequality and growth summarizes recent empirical work in this area and concludes that inequality is not robustly associated with redistribution in cross-country data. In fact, the statistical association between inequality and various measures of redistribution is rarely significant and its sign, which is sometimes negative, heavily depends on the chosen specification. Rodriguez (2004) obtains evidence of a negative association between inequality and redistribution by examining a panel of OECD countries in the period 1960–1990 and provides a theoretical model which is consistent with it, based on the unequal political power of the rich and the poor. In this paper, we propose a theoretical model which may provide useful insights on the relationship between inequality and redistribution and possible guidance in the specification of empirical tests of such relationship. The central idea of our work is quite simple. Assume that, in a world with credit market imperfections, the government runs a public program where (1) a public good or service is provided and accessible in equal amounts to all agents; (2) the program is financed through current income taxation and affects productivity and/or income not only in the current period but also in the future. Examples of such programs are publicly financed schools (as, for instance, primary and secondary education in most OECD countries) and public investment in infrastructure capital (e.g. roads, bridges, airports, etc…).2 In this case, if taxation is progressive (or proportional), public expenditure would have a redistributive effect since poor agents contribute relatively less to the equally beneficial program.3 In this economy, differently from the tradition of Meltzer and Richard (1981), the pivotal voter does not necessarily coincide with the agent (class) with median income. Agents who face a binding borrowing constraint, and are unable to finance their desired current consumption plan, have lower incentives to expropriate the rich as an increase in current taxation would further reduce their current disposable income and feasible consumption. Thus, the poor segments of the population, who are more likely to be liquidity constrained, may vote for low redistribution, together with the rich. In this case, instead of having all agents below the mean voting for high redistribution, an ends-against-the-middle equilibrium may arise where the poor and the rich form a coalition in favor of low levels of redistribution and the middle class favors higher levels of redistribution.4 Within this context, the effects of a mean-preserving increase in income inequality on the level of redistribution depend on whether the increase in inequality is concentrated among the poor or the middle class. In the former case, an increase in inequality tends to lower redistribution, whereas, in the latter case, it tends to increase redistribution. The former case contrasts with the conclusions of theoretical studies (see, for example, Alesina and Rodrik, 1994, Benabou, 1996 and Persson and Tabellini, 1994) which build upon the analysis of Meltzer and Richard (1981) and derive a positive relationship between inequality and redistribution. We formalize our main argument as follows. A two-period economy is inhabited by individuals who are heterogeneous with respect to their first-period labor productivity. In particular, we assume that there exist three income classes, the rich, the middle class and the poor. First-period income is homogeneous within classes and is below the mean for the two lowest-income classes. Capital market imperfections exist such that, to some extent, agents may be prevented from borrowing as much as they should to carry out their optimal consumption plans. Fiscal policy is politically determined through majority voting in the first period. Such policy involves current proportional income taxation which is used to finance current government expenditure, such as public expenditure on education or public investment in infrastructure, which increases the future productivity of labor. In this context, and in line with the standard public choice analysis, the preferred tax rate decreases with income for agents who are not liquidity constrained, since the marginal cost of redistribution is higher for richer agents. Instead (and this is the crucial aspect of our model), the desired level of redistribution increases with income for agents who are borrowing constrained. The inability to borrow to finance current consumption mitigates the incentives to expropriate the rich for liquidity constrained agents, the more so the larger the difference between income and desired consumption in the first period. This framework gives rise to different politico-economic equilibria, depending on the extent of borrowing constraints. When borrowing ceilings are high and no agent is liquidity constrained, the equilibrium tax rate will be the one preferred by the middle class. As the extent of borrowing constraints increases, a coalition of the poor and the rich is eventually formed, who favor a lower tax rate than the one preferred by the middle class. In other words, as borrowing ceilings fall, the identity of the median voter shifts from the middle class to the poor, who are induced to decrease current taxation to increase current consumption. In the last part of the paper we perform an empirical analysis based on our theoretical implications about the relationship between income inequality and redistribution. Even though the empirical analysis is somehow limited by data scarcity, we consider it as preliminary and suggestive evidence in favor of our theoretical results. Using pooled cross sectional-time series data for 22 OECD countries between 1960 and 1995, we find evidence that a reduction of income inequality that favors the poor has the predicted positive effect on public education expenditure. When reductions in inequality favor the middle class, the effect on public education is negative (as predicted by the theory). As suggested by Milanovic (2000), and differently from the existing literature, these results are obtained using exclusively pre-tax income data. They improve upon previous empirical contributions on the relationship between income inequality and redistribution which, as discussed at the beginning of this Introduction, fail to find significant evidence that supports theoretical predictions. Moreover, consistently with our theoretical analysis, estimation results (using after-tax income data when pre-tax observations were not available) indicate that the effect of a reduction in income inequality concentrated among the poor (middle class) is larger (smaller) when borrowing ceilings are below a certain threshold, that is when borrowing constraints are tight. These results seem to confirm the importance of taking into account the role of borrowing constraints when studying the empirical association between inequality and redistribution. Our paper is related to various strands of literature. Saint-Paul and Verdier (1996) briefly discuss various theoretical arguments that can give rise to a negative effect of inequality on redistributive pressure. Saint-Paul (2001) shows that more unequal societies can redistribute less if the increase in inequality is concentrated on the poorest. In his paper, the equilibrium tax rate decreases because the median income increases relative to the mean. In our setup, the result depends crucially on the change of identity of the median voter which is associated with higher inequality. Lee and Roemer (1999) show that increased inequality can induce less redistribution if the negative tax base effect which is associated to increasing inequality outweighs the conventional political effect. Another stream of related literature seeks to explain why universal suffrage has not implied larger rich-to-poor transfers of wealth. For instance, it has been suggested that political systems are biased against the poor, who are well known to participate less than the rich to political activity (see Benabou, 2000 and Rodriguez, 2004). Also, if political competition concerns more than one issue, the equilibrium tax rate proposed by the party protecting the interests of the poor may decrease, as the salience of the non-economic issue increases (Roemer, 1998). Moreover, it has been pointed out that even poor agents will not support high tax rates if they expect to move upward the income ladder (Benabou and Ok, 2001) or if they recognize the adverse dynamic effects of expropriating the rich (Perotti, 1993). Fernandez and Rogerson (1995) develop a model to explain the fact that the public support for higher education involves a transfer from lower income classes to higher income classes. In a small subset of their voting equilibria (and only in poor economies), the poor and the rich form a coalition in favor of low education subsidies. Differently from our mechanism, in their analysis the poor vote for zero taxation since the existence of fixed costs of education and the complete absence of capital markets prevents them from acquiring an education and receiving the subsidy. Clearly, our model implies a different role of borrowing constraints and allows us to derive clear-cut predictions about the relationship between income distribution, the extent of borrowing constraints and public education and to perform the empirical investigation which is one of the main motivation of our work. Finally, from a different point of view which abstracts from political economy considerations, the role of capital markets imperfections in the interaction between income distribution and economic development has been studied among others by Galor and Zeira (1993), Banerjee and Newman (1993), Aghion and Bolton (1997), and Piketty (1997). The plan of the paper is as follows. Section 2 describes the basic features of the model. In Section 3, we characterize the politico-economic equilibrium. Section 4 studies the relationship between inequality and redistribution, which is empirically analyzed in Section 5. Section 6 concludes.

This paper studies the relationship between income distribution and redistributive policies in a political economy model in the presence of capital market imperfections. We show that the winning coalition depends on the extent of borrowing constraints. When these constraints are sufficiently high, a coalition of the poor and the rich vote for low redistribution. Contrary to the standard implications of the public choice analysis of the size of government, we show that in this case increasing inequality can be associated to a reduction of the political support for redistributive taxation. Our model may add useful insights to the empirical analysis of the relationship between income inequality and redistribution, which usually fail to identify a statistically significant association between these variables. Following the main implications of our model, we suggest that, in order to obtain better estimation results, it is particularly important to use pre-tax income data, to specify on which class the increase in income inequality is concentrated and to control for the existence and extent of borrowing constraints. An empirical analysis conducted along this direction using pooled cross sectional-time series data for 22 OECD countries between 1960–1995 broadly supports our main theoretical predictions. Extending our model to a multi-period framework (see Appendix B), it would be possible to shed light on the relationship between income inequality and economic growth. The intuition drawn from our two-period analysis is that an increase in income inequality concentrated among the poor will tend to decrease public expenditure on education and growth. Instead, an increase in income inequality concentrated among the middle class will tend to increase public expenditure on education and growth (as in Saint-Paul and Verdier, 1993).