It’s easy to calculate ‘net redistribution’

Income Inequality and Government Redistribution: An Overview

We now describe our data, with a particular focus on changes since our dataset was last updated. As we have indicated, our complete 116-case dataset is available in an online appendix to our chapter.11 For ease of exposition, we begin our summary by focusing on national averages over multiple LIS waves for each of our twenty countries.

We begin with pre-tax and pre-transfer inequality.12 As can be seen in the first column of Table 8.1, countries vary quite widely in the extent of income inequality generated by the market and, to a much lesser extent, private transfers: pre-tax and pre-transfer inequality ranges by 164 Gini points across our twenty countries, from a high of 0.542 in Greece to a low of 0.378 in Iceland.13

{Footnote 13]In measuring inequality we have employed the Gini index, which ranges from 0 (all households receive the same income) to 1 (one household receives all income). Gini coefficients can be measured on either a 0-1.000 scale or a 0-100 scale; we use the former. By 'Gini point' we mean an increment of 0.001. As to redistribution, we have reported the absolute rather than the relative change in the Gini index as a result of redistribution, a measure that is not only more straightforward but also has the benefit of allowing one to compare the extent of state redistribution in a way that is not affected by levels or trends in market income inequality. In practice, these variables are strongly positively related, with a bivariate r of +0.930 across our 116 country-years. For those who prefer them, relative-change figures can easily be calculated from our pre- and post-tax/transfer figures. In calculating redistribution, we employ the approach often attributed to M. Reynolds and E. Smolensky, Public Expenditures, Taxes and the Distribution of Income: The U.S, 1950, 1961, 1970 (New York: Academic Press, 1977), which re-ranks households after public social transfers have been added and direct taxes deducted from private sector income when calculating the Gini coefficient (see also Caminada and Wang, Leiden LIS Budget Incidence Fiscal Redistribution Dataset; and Caminada et al., 'Disentangling Income Inequality and the Redistributive Effect of Taxes and Transfers in 20 LIS Countries over Time'). This is termed 'net redistribution'; it is computed using the 'SGINI' Stata module (P. van Kerm, 'SGINI: Generalized Gini and Concentration Coefficients (with Factor Decomposition) in Stata', www.researchgate. net/publication/255589700_Generalized_Gini_and_Concentration_coecients_%28with_ factor_decomposition%29_in_Stata (2010)). [End Footnote 13]

The large countries of the developed world fall between these extremes: Italy reports the third highest level, followed by the United Kingdom in fourth, the United States in sixth and Germany in thirteenth place. All the Nordic countries, with the exception of Finland, are in the lower part of this spectrum, along with Canada, Switzerland and Japan.

Pre-tax and pre-transfer inequality bears only a limited resemblance to inequality of disposable income in our countries, since in every case the state plays an important role in redistributing income by way of taxes and social transfers. That said, there is considerable variation in the extent of redistribution, as is evidenced by comparing the distribution of pre-tax and pre-transfer inequality to that of disposable income, which is listed in section A of Table 8.1. The country with the highest level of disposable income inequality, the United States, is separated by 121 Gini points from the country with the lowest level of disposable income inequality, Sweden. With respect to rankings, one of the most dramatic changes is the rise of the United States from being the sixth most inegalitarian of our countries in terms of pre-tax and pre-transfer inequality to the most inegalitarian in terms of disposable income. Switzerland, for its part, moves from the third most egalitarian pre-tax and pre-transfer distribution to the middle of the pack when taxes and social transfers are taken into account. In the lower part of the disposable income spectrum are the Nordic countries, Belgium and the Netherlands. As to the remaining countries, only Germany and Luxembourg are in the lower half. Italy and Greece have the second and third highest levels of disposable income inequality, followed by Spain, the United Kingdom, Australia and Ireland.

The cause of these shifts is the extent to which pre-tax and pre- transfer inequality is reduced by the redistributive effect of direct taxes and social transfers. The third column of section A of Table 8.1 shows redistribution for our twenty countries. (Countries in this part of the table are listed in descending order on this measure.) At the top of the scale are Belgium, Finland and Greece, in which pre-tax and pre- transfer inequality are reduced by 229, 228 and 216 Gini points respectively. Next come Sweden, Ireland, the Netherlands and Denmark, for which pre-tax and pre-transfer Gini coefficients are reduced by 205 or more points. At the bottom are the United States, Switzerland and Japan, with Gini reductions of 118, 108 and 80 points respectively. As has been noted, these three countries are ranked very differently in terms of pre-tax/transfer inequality. Switzerland and Japan rose considerably in the inequality rankings because of limited redistribution, but remained in the middle of the spectrum in terms of disposable income inequality because of their relatively egalitarian distribution of private sector income. The United States, for its part, started in the upper third of the inequality spectrum, but well below the top. It ended at the top in terms of inequality of disposable income not so much because of an exceptionally high level of pre-tax and pre-transfer inequality but because of limited redistribution in comparison to the other countries we examine.

Disaggregating Taxes and Transfers

Fiscal redistribution by the state is, of course, not of one piece. In particular, it is possible that different trends will be in evidence for the two main modes of direct redistribution: taxes and social transfers. To explore this, we have partitioned the total Gini reduction accomplished by the state into two components: the part achieved by direct taxes and the part achieved by social transfers. We have further disaggregated the Gini reduction as a result of transfers into programmes aimed primarily at the elderly and those aimed primarily at working-age persons.

We begin with taxes. Unfortunately, there is a limitation in using LIS data to explore tax redistribution. The problem is that the national income surveys from which the LIS derives its data do not measure indirect taxes, such as sales, value added and excise taxes, whose precise amount is rarely known even by those paying them and whose incidence is thus very difficult to determine. For this reason, we consider Gini reduction by way of direct taxes only, the primary components of which are income taxes and mandatory employee social insurance contributions. The omission of indirect taxes is important because the countries we examine vary greatly in the share of revenue raised by such taxes, from a low in 2011 of 4.4 per cent of GDP in the United States to a high of 15.2 per cent in Denmark. Some scholars have argued that, even though indirect taxes are not themselves redistributive - in fact, are commonly regressive (although this is usually reduced by lower rates on food, medicine and other essentials) - they nonetheless play a critical role in raising the revenue that funds redistributive social benefits. Is there empirical evidence for such a relationship? One way of exploring this, at least in a preliminary way, is to relate the share of indirect taxes in a country's GDP to the extent to which inequality of private sector income there is reduced by social transfers. When these variables are related we find that the magnitude of indirect taxes is strongly positively related to the degree to which public social transfers reduce private sector inequality; the bivariate correlation is +0.62. This is not to say that consumption taxes directly finance social transfers in the same way that social security contributions do; with that exception, taxes are fungible. It does, however, offer evidence that indirect taxes on consumption represent a powerful revenue-raising vehicle that supports an array of social transfers, which in turn substantially ameliorate market inequality. Having said that, we note that any transfer redistribution financed, directly or indirectly, by indirect taxes is captured in LIS figures for social transfers, for which we do have data.

With these qualifications, we report the shares of total redistribution accounted for by direct taxes and social transfers in the first column of Table 8.1, section B. In describing the results, we begin with the observation that overall government redistribution in our countries is much more a product of the redistributive effect of social transfers than of direct taxes: across our twenty countries the average share of Gini reduction accomplished by direct taxes is 22.9 per cent while that accomplished by transfers is 77.1 per cent. There is, however, considerable variation about these averages. In particular, the United States and Australia top the list of our countries in the share of fiscal redistribution achieved by taxes: in these countries 36.7 and 32.0 per cent respectively of all redistribution is achieved by direct taxes, well above the twenty- country average. Because total redistribution in these countries is below average, they do not rank quite as high in absolute tax redistribution (see Table 8.1, section B, column 3), although both are in the upper third of that spectrum as well. Other countries are lower on the absolute redistribution list, with Spain, Japan and Switzerland at the bottom.

We now turn to redistribution by way of social transfers. We start with a Gini reduction as a result of social transfers as a whole. As can be seen in the final column of Table 8.1, section B, our countries vary considerably on this measure. At the top of the scale are Finland and Greece, whose pre-tax and pre-transfer Gini coefficient is reduced by 181 Gini points by transfers alone. Most of the Nordic countries are in the upper third of the list, as are the Netherlands and Belgium. The large transfer reduction value for Greece was unexpected. Recall, however, that this country began with a much more inegalitarian distribution of pre-tax and pre-transfer income than our other countries; and even after extensive redistribution it remained among the most inegali- tarian of our twenty countries. (Obviously, the Greek situation with respect to this has changed dramatically since the last available income survey in 2010.) At the low end in terms of transfer reduction are Switzerland, Australia, Canada, Iceland, the United States and Japan. These countries differ greatly in the pre-tax and pre-transfer starting point; in particular, Switzerland, Iceland and Japan have relatively low levels of private sector inequality.

Total fiscal redistribution by way of public social transfers encompasses a number of distinct programmes. The most basic breakdown is between pensions, which are aimed primarily at the elderly, and transfers aimed mainly at the working-age population - a category that includes a wide variety of individual programmes, including unemployment benefits, child allowances and means-tested social assistance.

We begin with pensions, which not only constitute the largest social benefit programme in the developed world but also produce the most redistribution, at least in absolute terms. These figures are reported in Table 8.1, section C. On average, across our twenty countries pensions alone reduced private sector inequality by 91 Gini points, more than twice as much as all other transfer programmes combined and just over half of the total reduction in inequality achieved by taxes and transfers together. At the top of the list for pension redistribution were Austria, Belgium and the Netherlands, in which pensions reduced the private sector Gini coefficient by 115 points or more. In the bottom part were Australia, Canada, the United States and Japan, in which pensions reduced the pre-tax and pre-transfer Gini by at most 60 points.

Part of the explanation for the wide cross-country divergence in the redistributive effect of pensions is cross-national variation in the proportion of the elderly in the population. However, the size of public programmes and their internal progressivity also matter: the R2 of a regression across our 116 LIS datasets that uses the share of the population aged sixty-five or older to explain variation in our pension reduction variable is 0.38, indicating that less than half of the variation in this mode of redistribution across countries and years is explained by demographics.

What of programmes aimed mainly at the working-age population? To start, it should be noted that these programmes accomplish a good deal less inequality reduction than do pensions: as shown in the third column of Table 8.1, section C, on average they result in a reduction of income inequality of 42 Gini points, less than half that for pensions. Even so, the extent of redistribution is substantial - in fact, on average, greater than that of tax redistribution. It should be noted that, although all these programmes are aimed primarily at the working-age population, they are quite mixed in other respects. Some - particularly child and family allowances - typically offer flat-rate benefits; some offer means-tested public assistance; and some - particularly unemployment benefits - are commonly tied to the income of recipients when they were working. It is possible to disaggregate these programmes further using LIS data, but only at the expense of comparability. LIS harmoni- sation, which entails converting national definitions into a common framework, tends to become more difficult as the scope of programmes becomes narrower - something that is exacerbated by the fact that programmes sometimes supplement or substitute for one another. In particular, although we would have liked to, we were unable to separate unemployment benefits from other work-related social insurance transfers for many of our datasets.