The middle class and the Swedish welfare state: how not to measure redistribution.

• Author: Bergh, Andreas

Big welfare states, with taxes near 50 percent of gross domestic product (GDP), still exist in the Scandinavian countries. It is widely assumed that bigger welfare states redistribute more income from the rich to the poor. The evidence for this assumption, however, is surprisingly shaky. Furthermore, the fact that taxes and government expenditures remain at very high levels does not necessarily mean that redistribution is constant. There are clear signs that the Swedish welfare state is becoming more beneficial for the middle class, but the standard method that welfare-state scholars in economics, sociology, and political science use to quantify redistribution does not detect this development. (1)

In this article, I first describe the method typically used to evaluate redistribution. Next, I discuss two problems with this standard approach: first, it does not account for behavioral responses to welfare programs; and, second, it does not detect how political mechanisms change the structure of the welfare state into something particularly beneficial for the middle class. I illustrate this middle-class bias with several Swedish examples. Before concluding, I discuss publicly financed primary schooling, a welfare-state component with a clear effect on inequality that the standard method does not capture.

The Standard Approach to Measuring Redistribution

The most commonly used approach to measuring welfare-state redistribution is to compare the income distribution before taxes and transfers with the distribution after taxes and transfers and to assume that the welfare state causes the difference. (2) With few variations, this approach is used by Kakwani (1986), Mitchell (1991), Stephens (1995), Korpi and Palme (1998), Solera (2001), Bradley et al. (2003), Moiler et al. (2003), Iversen (2005), and Smeeding and Sandstrom (2005). More than twenty years ago, Uusitalo (1985) identified and analyzed several problems with the standard approach, but because no obvious alternative exists, the approach is still used, notwithstanding the notable risk that the research will give rise to incorrect policy conclusions.

The approach is normally applied as follows. The analyst calculates the Gini coefficient for gross income. (3) The calculation is made for households or (less often) for individuals, usually the adult population. The same calculation is made for net income--that is, income after taxes and transfers have been taken into account. Finally, the analyst calculates the relative reduction in the Gini coefficient to produce a measure of redistribution for use in cross-country comparisons and regressions (see, for example, table A4.3 in Iversen 2005). Variations on the theme include using poverty ratios instead of Gini coefficients, as for example in Moller et al. 2003.

Smeeding and Sandstrom formulate a typical conclusion from these studies: "[I]n general, the larger and more inclusive the social insurance system, ... the larger the antipoverty effect" (2005, 7-8). They also state that the systems in Sweden and Germany have the largest effects on poverty among the elderly. This conclusion is reached as follows: the poverty rate based on market income is 93 percent for female-headed elderly households in Sweden and 82 percent for all elderly households; after taxes and transfers, the corresponding poverty rates are 17 and 8 percent, respectively. Thus, the total effect of the system is taken to be that poverty rates are reduced by more 70 percentage points.

It is simply not true, however, that in the absence of public pensions and other transfers, 93 percent of old women in Sweden would be poor. The pension system in Sweden is big, universal, and mandatory, and, most important, people apparently trust that the pensions will be paid. Therefore, they adjust their behavior accordingly: the public system crowds out the provision of market sources of income.

To evaluate the welfare state's effect correctly, however, we should compare its outcome with the outcome that would arise in its absence. For several reasons, the gross-income distribution is not the proper counterfactual. Most important, the gross-income distribution reflects people's behavioral adjustments to the welfare state, as in the preceding example of the Swedish pensioners. If poor people lower their labor supply in response to generous welfare benefits, this action will increase gross-income inequality, and the standard approach will exaggerate redistribution because it does not take this behavioral response into account.

Economists typically do include behavioral responses in their theoretical models of redistribution. However, the analysis of the effect of the entire welfare state is often left to sociologists and political scientists, some of whom seem reluctant to deal with the problem. Two of the previously mentioned sources discuss the issue in identical words:

Based on conventional economic reasoning, critics of the welfare state contend that generous welfare state benefits ... available to able-bodied working-age persons increase pre-tax and transfer inequality because they act as disincentives for recipients to seek work. Indeed, it is sometimes argued that, to the extent that generous welfare states reduce post-tax and transfer inequality, they simply make up for the damage done to pre-tax and transfer inequality levels. We are skeptical regarding this argument, as it ignores the fact that generous welfare states are often labor mobilizing and invest heavily in skill formation, particularly under the influence of social democratic parties. (Bradley et al. 2003, 200; Moiler et al. 2003, 27) Note how this description creates a straw man by stating that "it is sometimes argued" that welfare-state redistribution only makes up for the damage done to the market-income distribution. To be sure, this extreme position is probably wrong in most cases, but it does not follow that behavioral responses may safely be ignored. Even if behavioral responses to taxes and transfers are small in absolute terms, they probably vary among different countries because the structure of taxes and transfers varies among countries. The actual implications of conventional economic reasoning are analyzed in Bergh 2005, and the results are intuitive:

* Flat-rate benefits reduce labor supply, and the effect is bigger for low-income earners.

* Progressive taxes reduce labor supply for high-income earners.

Even if these effects are small--estimates of labor-supply elasticities are indeed small for men but greater for women--the implication is that the standard approach will indicate large redistribution when taxes are close to proportional and benefits are flat rate, but small redistribution when taxes are progressive and benefits are positively income related. Yet the country that scores highest according to the standard approach is not necessarily the one with the greatest net redistribution.

Behavioral responses are likely to be larger when taxes and benefits are high, and the data confirm that the standard approach seems to be more biased in high-tax countries. For example, with data from the Deininger-Squire dataset (based on the Luxembourg income study) one can illustrate some peculiar effects of the standard approach. (4) Between 1975 and 1981, both gross-income and net-income inequality decreased in Sweden, yet the standard approach indicates that the system became less redistributive. Between 1979 and 1986, redistribution in Norway decreased drastically, according to the standard approach, but almost nothing happened to net-income inequality. For the United States, however, higher redistribution according to the standard approach seems to be correlated with lower net-income inequality.

In addition to labor-supply responses, the welfare state causes other behavioral changes...