توزیع درآمد و ساختار مالیاتی: آزمون تجربی فرضیه ملتزر، ریچارد

The Meltzer–Richard hypothesis that more unequal income distribution will create a majority for more redistribution has generated much research, but little empirical support. The empirical literature has concentrated on cross-country studies and the size of the public sector, and the results broadly do not indicate more redistribution with more inequality. This analysis suggests that the hypothesis should be investigated in a more homogenous setting with comparable institutions and with an explicit decision about redistribution (here tax structure). New data on poll tax and property tax in decentralized government in Norway are exploited. We show how the multi-dimensional decision can be analyzed as majority rule assuming intermediate preferences. In the econometric analysis, instruments are used to account for endogeneity of income level and income distribution. The estimated model supports the understanding that more unequal income distribution implies a shift in the tax burden from poll tax to property taxes and thereby gives more redistribution.

When the median voter has less income than the mean, the typical income distribution observed, the decisive median voter will apply income taxation for redistribution. This is the key insight of Meltzer and Richard (1981). More uneven income distribution is associated with more redistribution, only held back by negative incentives to work and save. The setup assumes proportional income taxation financing lump sum government transfers. The theory is based on an earlier work on optimal redistributive taxation by Romer (1975) and Roberts (1977), where the decisive voter chooses parameters of an income tax function. Cukierman and Meltzer (1991) continue this tradition and show conditions when marginal progressivity of the income tax is determined by the median voter. Meltzer and Richard (1983) started up the empirical tests of the hypothesis in an analysis of US time series data of government spending. They conclude that the spending level is negatively related to the ratio of median to mean income. Later the Meltzer–Richard model is basically interpreted as a theory of government size and tested on cross-country data, but with little support. Government spending is assumed to concentrate on redistribution, and the income distribution is understood as the major determinant of government size. It is hard to get away from the observation that many countries with equal income distribution have large government spending, notably the Scandinavians, while many countries with unequal income distribution have smaller public sectors, as in many developing countries. Perotti (1996) analyzes a broad dataset including the marginal tax rate and different expenditure components, and he concludes that there is little evidence of a negative association between equality and fiscal variables. Bassett et al. (1999) reach the same conclusion with other definitions of transfers and spending. Milanovic (2000) has a more direct measure of redistribution, the income gain of different income groups from factor income to disposable income. His estimates support the hypothesis that countries with greater inequality redistribute more, but are less supportive of the median voter hypothesis. The spread of the income distribution, among the poor and among the middle class, is the background concern. Alesina and Rodrik (1994) and Persson and Tabellini (1994) develop the redistribution model in a dynamic context and extend the empirical cross-country analyzes to economic growth. This approach implies two links, from income distribution to distortions with redistribution, and from distortions to economic growth. Again the evidence is mixed, and the link via distortions like taxation is not documented. Saint Paul and Verdier (1996) review some critical arguments. The lack of strong evidence in favor of the Meltzer–Richard hypothesis has motivated the development of theoretical models with the opposite prediction, i.e. that more inequality leads to less redistribution. Persson (1995) and Benabou (2000) emphasize that redistribution may improve welfare and economic performance, in which case less inequality may increase the popular support for redistribution. Sinn (1996) questions whether the causality runs from inequality to redistribution or the other way around. The argument is that redistributive programs involve social insurance that stimulates risk taking and moral hazard effects, which increases the pre-tax inequalities. The Meltzer–Richard hypothesis may stand a better chance in more homogenous political systems, and with a sharper focus on redistributive instruments. Decentralized government is a potentially interesting source of information about politics and distribution. Alesina et al. (2000) exploit this type of data in a recent study of US cities, and find a positive relationship between inequality and public employment. On the other hand, Rodriguez (1999) find no association between distributional skewness and welfare spending in the US States. It seems to us productive to get back to the relationship between income distribution and taxation as a basis of evaluating redistributive politics and to bring in evidence about the tax structure to investigate the hypothesis. Chernick and Reschovsky (1996) measure the degree of progressivity of the state tax burdens. Income inequality comes out as an important explanation of progressivity, and the result is seen as consistent with an interest group model of tax choice. The problem with the political economy models is that the decision making is a black box. Slemrod and Bakija (2000) argue that optimal taxation theory predicts that growing inequality should increase progressivity, but notice the failure to identify such effects in the recent US experience of increased inequality. Our empirical analysis addresses the role of the income distribution for the choice of tax structure in local governments in Norway. The 434 local governments are comparable political institutions based on elections to the local council, and their main revenue decision is the choice between user charges and property taxes. New data on residential property taxes and housing-related utility charges per standardized households in each municipality allow for this test of Meltzer–Richard. Utility charges per standardized household are hereafter denoted as poll tax. Our understanding is that the broad income distribution in the communities is not much affected by the local revenue decision, but our econometric analysis addresses this potential endogeneity. The multi-dimensionality issue has been a challenge to the modeling of the political decisions in this area. Goodspeed (1998) develops a majority rule voting model for the relationship between state income taxes and local property taxes in the US. Income distribution is important for the tax structure since the distribution of the income tax base is different from the distribution of the property tax base. His model allows a unidimensional decision since the two taxes are decided at different political levels. Inman (1979) solves his multi-dimensionality problem in a study of 41 US cities by assuming a two-step process. First the aggregate tax level is determined in an expenditure demand model, then the tax structure decision (property tax share of total taxes) is an allocation of the total taxes. We take benefit of restrictions on the preferences that imply that the two revenue instruments and government spending can be understood as unidimensional. This condition of intermediate preferences is due to Grandmont (1978) and is discussed by Persson and Tabellini (2000). The intuition here is that the conflict of interest is along the same dimension and follows the income distribution. Section 2 outlines a theoretical framework of the relationship between income distribution and tax structure investigated. The conditions for unidimensionality and majority voting outcome are shown. Section 3 presents the empirical and institutional background of the empirical analysis, and the data and the econometric approach are discussed in Section 4. Estimation results for the effect of income distribution on tax structure are presented in Section 5, with alternative formulations to check for robustness and including instrument variables to handle endogeneity. Section 6 addresses the broader determinants of tax structure. Concluding remarks are offered in Section 7.

The tax structure is important for the income distribution and therefore a key playground for redistributive politics. The starting point of the paper is the Meltzer–Richard hypothesis that more unequal income distribution will create a majority for more redistribution. While the empirical literature investigating the hypothesis has concentrated on the size of the public sector, this analysis exploits data about the tax structure in decentralized government in Norway. The choice of revenue instruments studied involves poll and property tax. The approach is in the tradition of majority rule, and we show how the local government decisions regarding tax structure and spending level can be understood as one dimensional. This motivates the empirical analysis where the actual income distribution is measured by the ratio of median to mean income. The estimated model supports the understanding that more equal income distribution implies a shift in the tax burden from property taxes to poll tax and thereby gives less redistribution. We conclude that the tax structure is responsive to income distribution. More unequal distribution allows the majority to avoid the poll tax and shift the tax burden towards a property tax that is proportional to income.