Tests of a rational theory of the size of government

Conclusion

In many countries, the political party holding power changes more frequently than the trend growth rates in government spending and taxes. Shifts of political power are often preceded by rhetoric about ‘meeting needs’ or cutting taxes, which post-election performances rarely match. Yet, changes in the size of government, up and down, occur. On average, the size of government, measured by the share of taxes or spending to output, has increased in many countries during the past century, but changes in size occur at diffrent rates and, at times, the relative size of government declines.

In recent work (Meltzer and Richard, 1981), we develop a general equilibrium model in which people choose consumption and leisure and, as voters, decide on income redistribution or the (average) tax rate. The model implies that the size of government changes with the ratio of mean income to the income of the decisive voter and with the voting rule or qualifications for voting. Extensions of the franchise that increase the number of voters who benefit from income redistribution increase votes for redistribution. The size of government increases. Changes in the age composition of the population that increase the proportion of the population receiving old age assistance also increase redistribution paid from taxes on labor income. The relation is symmetric. Changes in productivity, or in labor force participation, that lower mean income relative to the income of the decisive voter, reduce the size of government.

The tests of the model reported here treat the person with median income as the decisive voter. We find that the ratio of government spending for redistribution to aggregate income, and the share of aggregate income redistributed in cash, rise and fall with the ratio of mean to median income and the level of (median) income. Redistribution in kind — the provision of education, health care, fire protection, and other services — also rises and falls with the ratio of mean to median income, but it appears to be independent of the level of income.

Our model implies, and the data suggest, that Wagner's law — relating the size of government to the level of income — must be amended. The relation is not simple and direct, as many tests of Wagner's law presuppose, but depends, in our model, on shape of the income distribution — more specifically on the ratio of mean to median income. Failure to hold the distribution of income constant renders many previous tests meaningless. Further, our results suggest that voters' choice of the nature of redistribution — in cash or in kind — affects the results. Although we do not derive the relation between size of government and voters' choice of cash or in-kind benefits (and we neglect public goods), our results suggest that Wagner's law is more likely to find support if redistribution is in cash.

Our hypothesis is parsimonious. There is no uncertainty. Taxes are linear, and all redistribution is by lump sum transfer. The decisive voter is fully cognizant of the costs and benefits of the redistribution he demands, including the effects on incentives to work and consume. We neglect most features of the political process, including any influence of interest groups, bureaucrats, and other monopoly elements that affect ‘supply.’ We recognize that a useful extension of our model would incorporate the allocation of funds to specific programs and thereby incorporate ‘supply’ factors. In our empirical work, we rely on a linear approximation to a non-linear equation and obtain our estimates from an equilibrium relation, not a structural equation. Despite the model's parsimony, the neglect of supply factors, the use of a linear approximation, and data interpolation, the hypothesis explains much of the trend in the relative size of spending for redistribution and a considerable part of the annual variation observed in the United States during a recent forty-year period.

During the period we studied, our measure of government spending for redistribution increased, in nominal value, from $10 billion to more than $350 billion and the share of total income redistributed rose from 12% to 22%. A considerable part of the observed increase in the size of government appears to be consistent with rational choice by maximizing voters who benefit from redistribution and are able to shift a disproportionate share of the cost to people with incomes above the mean.