Skip to main content Accessibility help
×
Home

Earnings Inequality and Welfare Spending: A Disaggregated Analysis

  • Karl Ove Moene (a1) and Michael Wallerstein (a2)

Extract

The welfare state is generally viewed as either providing redistribution from rich to poor or as providing publicly financed insurance. Both views are incomplete. Welfare policies provide both insurance and redistribution in varying amounts, depending on the design of the policy. The authors explore the political consequences of the mix of redistribution and insurance in the context of studying the impact of income inequality on expenditures in different categories of welfare spending in advanced industrial societies from 1980 to 1995. They find that spending on pensions, health care, family benefits, poverty alleviation and housing subsidies is largely uncorre-lated with income inequality, but that spending on income replacement programs such as unemployment insurance, sickness pay, occupational illness and disability are significantly higher i n countries with more egalitarian income distributions. They show that this pattern is exactly what a theory of political support for redistributive social insurance programs would predict.

Copyright

References

Hide All

1 Huber, Evelyne and Stephens, John D., Political Choice in Global Markets: Development and Crisis Advanced Welfare States (Chicago: University of Chicago Press, 2001), 17. For other recent books that emphasize the centrality of either the political mobilization or the economic organization of workers for explaining cross-national differences in the size of the welfare state, see Hicks, Alexander, Social Democracy and Welfare Capitalism:A Century ofIncome Security Politics (Ithaca, N.Y.: Cornell Press, 1999); Wilensky, Harold, Rich Democracies (Berkeley: University of California Press, 2002); and Swank, Duane, Global Capital, Political Institutions, and Policy Change in Developed Welfare States bridge: Cambridge University Press, 2002).

2 Pampel, Fred C. and Williamson, John B., ”Welfare Spending in Advanced Industrial Democracies 1950–1980,” American Journal of Sociology 93 (May 1988).

3 Romer, Thomas, ”Individual Welfare, Majority Voting, and the Properties of a Linear Income Tax,” Journal of Public Economics 14 (February 1975); Roberts, Kevin W. S., ”Voting over Income Tax Schedules,” Journal of Public Economics 8 (December 1977); Meltzer, Allan H. and Richard, Scott F., ”A Rational Theory of the Size of Government,” Journal of Political Economy 89 (October 1981).

4 Baldwin, Peter, The Politics of Social Solidarity: The Class Bases of the European Welfare State bridge: Cambridge University Press, 1990), 18. For a survey of tbe economic arguments in favor of public provision of insurance, see Barr, Nicholas, ”Economic Theory and the Welfare State: A Survey and Interpretation,” Journal of Economic Literature 30 (June 1992). Iversen and Cusack interpret the welfare state as insurance against the risk of income loss occasioned by the shift of jobs from manufacturing to the service sector, while Rodrik and Garrett interpret the welfare state as insurance against the risks entailed by increased international economic integration; see Iversen, Torben and Cusack, Thomas R., ”The Causes of Welfare State Expansion: Deindustrialization or Globalization?” World Politics 52 (April 2000); Rodrik, Dani, ”Why Do More Open Economies Have Larger Governments?” Journal of Political Economy 106 (October 1998); and Garrett, Geoffrey, Partisan Politics in the Economy (Cambridge: Cambridge University Press, 1998). The redistributive insurance framework used in this paper was first presented in Moene, Karl O. and Wallerstein, Michael ”Inequality, Social Insurance and Redistribution,” American Political Science Review 95 (December 2001).

5 The strong correlation that exists between social-insurance spending as a share of GDP and GDP per capita in data sets that include both high-income and low-income countries suggests that richer voters prefer to spend a larger share of income on social insurance; the correlation is documented in Wilensky, Harold, The Welfare State and Equality (Berkeley: University of California Press, 1975). An alternative explanation is that the capacity of governments to collect revenues without imposing large deadweight costs rises with economic development.

6 Francisco Rafael Rodriguez Caballero, ”Essays on the Political Economy of Inequality, Redistribution and Growth” (Ph.D. diss., Harvard University, 1998); Moene and Wallerstein (fn. 4).

7 Rodriguez (fn. 6); Moffitt, Robert, Ribar, David, and Wilhelm, Mark, ”The Decline of Welfare Benefits in the U.S.: The Role of Wage Inequality,” Journal of Public Economics 68 (June 1998).

8 Perotti, Roberto, ”Growth, Income Distribution and Democracy: What the Data Say” Journal of Economic Growth 1, no. 1 (1996).

9 Milanovic, Branko, ”The Median-Voter Hypothesis, Income Inequality, and Income Redistribution: An Empirical Test with the Required Data,” European Journal of Political Economy 16, no. 2 (2000).

10 Kinder, Donald, ”Diversity and Complexity in American Public Opinion,” in Finifter, Ada, ed., Political Science: The State of the Discipline (Washington, D.C.: APSA, 1983).

11 Korpi, Walter, The Working Class in Welfare Capitalism (London: Routledge and Kegan Paul, 1978); Hicks (fn. 1); Huber and Stephens (fn. 1); Wilensky (fn. 1).

12 Heclo, Hugh, Modern Social Policies in Britain and Sweden (New Haven: Yale University Press, 1974).

13 Swenson, Peter, Capitalists against Markets (Oxford: Oxford University Press, 2002).

14 Austen-Smith, David, ”Redistributing Income under Proportional Representation,” Journal of Political Economy 108 (December 2000).

15 Roemer, John, ”Why the Poor Do Not Expropriate the Rich,” Journal of Public Economics 70 (December 1998); idem, Political Competition (Cambridge: Harvard University Press, 2001); Austen-Smith, David and Wallerstein, Michael, ”Redistribution in a Divided Society” CMS-EMS Discussion Paper 1362 (Evanston, Ill.: Northwestern University, 2003).

16 Christopher Achen and hairy Bartels, ”Ignorance and Bliss in Democratic Politics: Party Competition with Uninformed Voters” (Paper presented at the annual meetings of the Midwest Political Science Association, Chicago, April 25—28, 2002).

17 We discuss relaxing this assumption below.

18 The assumption that u is constant is made to simplify the mathematical expressions, but it is not necessary. The assumption that u > 1 is critical for our results. Both assumptions regarding u are supported by studies of the allocation of household savings. See Friend, Irwin and Blume, Marshall E., ”The Demand for Risky Assets,” American Economic Review 65 (December 1975).

19 We have made the modeling choice to represent differences in social-insurance policies in terms of differences in the distribution of benefits, rather than in terms of differences in the tax that finances the benefits. A more general approach would be to define post-tax and transfer income as a function of pre-tax and transfer income, as in Roemer, John, ”Does Democracy Engender Equality,” in Mansfield, Edward and Sisson, Richard, eds., Political Knowledge and Public Interest (Columbus: University of Ohio Press, 2003). For the purposes of this paper, however, the assumption of a flat tax is a reasonable approximation. In most of the countries we study, much of the welfare budget is financed by a payroll tax that is usually flat. (Denmark is an outlier in relying almost exclusively on income and value-added taxes.) Moreover, a recent study of the progressiveness of the personal income tax in twelve OECD countries by Wagstaff et al., found ”no link between pre-tax inequality and the degree of redistribution brought about by the personal income tax”; see Adam Wagstaff et al., ”Redistributive Effect, Progres-sivity and Differential Tax Treatment: Personal Income Taxes in Twelve OECD Countries,” Journal

20 The assumption that income-replacement policies are redistributive, or that ξ > 0, is consistent with the way most social-insurance programs are designed. For example, the average replacement ratio for unemployment insurance, bN(w, t)/w in the notation of the paper, is 18 percent higher for a worker who receives two-thirds of the median wage than for a worker who receives the median wage in the countries in our data set; Organization for Economic Cooperation and Development, Dataset on Benefit Replacement Rates (Paris: OECD, no date). Of course, it would be preferable if ξ > 0 was a conclusion rather than an assumption. However, the one-dimensional model of politics no longer applies when both ξ and b are chosen simultaneously. For a model where ξ is chosen at a ”constitutional” stage to maximize a social welfare function while b(i) is chosen by self-interested voters in a second ”electoral” stage, see Georges Casamatta, Helmuth Cremer, and Pierre Pestieau, ”Political Sustainability and the Design ofSocial Insurance,” Journal ofPublic Economics 75 (March 2000).

21 The U.S. is exceptional in devoting roughly 25 percent of public health expenditures to a family of means-tested programs known as Medicaid; United States Bureau of the Census, Statistical Abstract of the United States 2001 (Washington, D.C.: U.S. Government Printing Office, 2001).

22 Korpi, Walter and Palme, Joachim, “The Paradox of Redistribution and Strategies of Equality: Welfare State Institutions, Inequality and Poverty in the Western Countries,” American Sociological Review 63 (October 1998).

23 Organization for Economic Cooperation and Development, Social Expenditure Database 1980–1996 (CD-ROM) (Paris: OECD, 1999).

24 Organization for Economic Cooperation and Development, Employment Outlook (July 1993); idem, Employment Outlook (July 1996).

25 For a discussion of the properties of the lognormal distribution and its use as an approximation of the distribution of income, see Aitchison, J. and Brown, J. A. C., The Lognormal Distribution (Cambridge: Cambridge University Press, 1957).

26 Both our model and the data we use to measure income inequality refer to wage and salary earners who are either working full time or are temporarily without employment. Of course, not all voters fit into these categories. Some work part time. Others are permanently outside the dependent labor force. To take all categories of attachment to the labor market into account would greatly complicate the analysis, both theoretically and empirically.

27 The countries and years in the data set are listed in Appendix 2.

28 The regression equation is

with R 2 = 87.7 and n = 50, where y, is total welfare expenditures as a share of GDP in period t, and the standard error of the coefficient on the lagged dependent variable is .050.

29 The possible endogeneity of unemployment is discussed below.

30 Lijphart, Arend, ”Unequal Participation: Democracy's Unresolved Dilemma,” American Political Science Review 91 (May 1997); Franzese, Robert J., Macroeconomic Policies of Developed Democracies (Cambridge: Cambridge University Press, 2002).

31 Roemer(fn. 15,2001).

32 Castles, Francis G., ”The Impact of Parties on Public Expenditure” in Castles, , ed., The Impact of Parties: Politics and Policies in Democratic Capitalist States (London: Sage Publications, 1982); Esping-Andersen, Gosta, The Three Worlds of Welfare Capitalism (Princeton: Princeton University Press, 1990). The tripartite division of parties into left, center, and right follows Francis Castles and Peter Mair, ”Left-Right Political Scales: Some ‘Expert’ Judgements,” European Journal of Political Research 12 (March 1984). Socialist, social democratic, and labor parties (with the exception of the Italian Social Democratic Party) comprise the group of left parties. Center parties, farmers parties, liberal parties in countries with a conservative party on the right, Christian democratic parties in countries with a liberal party on the right, and the Democratic Party in the U.S. constitute the group of center parties. Conservative parties, liberal parties in countries where the liberal party is the main party on the right, and Christian democratic parties in countries where the Christian democratic party is the main party on the right, plus all small parties further right comprise the group of conservative parties.

33 The impact of these three variables on the distribution of wages and salaries is analyzed in Michael Wallerstein, “Wage-Setting Institutions and Pay Inequality in Advanced Industrial Societies,” American Journal of Political Science 43 (July 1999). For related studies, see Richard B. Freeman, “Labour Market Institutions and Economic Performance.” Economic Policy 3 (April 1988); Francine D. Blau and Lawrence M. Kahn, “International Differences in Male Wage Inequality: Institutions versus Market Forces,” Journal of Political Economy 106 (August 1996); and David Rueda and Jonas Pontusson, “Wage Inequality and Varieties of Capitalism,” World Politics 52 (April 2000).

34 Huber, Evelyne, Ragin, Charles, and Stephens, John D., “Social Democracy, Christian Democracy, Constitutional Structure and the Welfare State,” American Journal of Sociology 99 (November 1993).

35 Traxler, Franz, Blaschke, Sabine, and Kittel, Bernhard, National Labour Relations in Internationalized Markets (Oxford: Oxford University Press, 2001). The measure of union participation in policymaking with respect to nonwage issues is described by Traxler, Blaschke, and Kittel as “associational (union) participation in state regulation (non-wage issues)” (p. 68). The data are available by decade. We assigned the 1980–90 figure to 1985 in our data set, and the 1991–96 figure to 1995 in our data set. For 1990, we used the average of the 1980–90 and 1991–95 figures. We rechecked our results with 1990 removed from our data. In neither case did the inclusion of the index of union participation alter our findings with respect to inequality.

36 Wallerstein (fn. 33).

37 Neal Beck and Jonathan Katz, ”What To Do (and Not to Do) with Time-Series Cross-Sectional Data in Comparative Politics,” American Political Science Review 89 (September 1995); Greene, William H., Econometric Analysis, 3rd ed. (Upper Saddle River, N.J.: Prentice Hall, 1997).

38 We generated 400 data sets, each with 45 observations (15 countries, 3 time periods) and 3 regressors (a constant plus the first 2 regressors in column 1 of Table 2). The error terms were normally distributed with randomly selected, country-specific variances and randomly selected, cross-national correlations. In the simulations the 90 percent confidence intervals contained the true values of the coefficients roughly 90 percent of the time when calculated using OLS standard errors. When calculated on the basis of panel-corrected standard errors, by contrast, the 90 percent confidence intervals contained the true value of the coefficients only 77 percent of the time. See Karl O. Moene and Michael Wallerstein, “Income Inequality and Welfare Spending: Simulations” (http://faculty-web.at.nwu.edu/polisci/wallerstein/papers.html).

39 Franklin, Mark N., “Electoral Competition,” in LeDuc, Lawrence, Niemi, Richard, and Norris, Pippa, eds., Comparing Democracies: Elections and Voting in Global Perspective (Thousand Oaks, Calif.: Sage Publications, 1996).

40 In the case of health expenditures, the estimated coefficient on inequality is even closer to zero if one subtracts means-tested health expenditures (Medicaid) from the U.S. figures. Excluding U.S. Medicaid expenditures (roughly 25 percent of total government expenditures on health in the U.S.), column 3 of Table 2 becomes

where standard errors of the coefficients (excluding the constant) are (.103, .526, .0025, .0088, .052) and adjusted R 2 = 65.8. Only the coefficient on the lagged dependent variable and on Right government are significant at the .05 level.

41 The category of income replacement in Table 2 is a subset of the policies included in insurance against loss of income in Moene and Wallerstein (fn. 4). The difference between the two is that the measure of insurance against income loss in Moene and Wallerstein (fn. 4) includes a share of expenditures on health while all health expenditures are excluded from spending on income replacement in Table 2.

42 An alternative way to measure the generosity of unemployment benefits is the replacement ratio, which is available from OECD (fn. 20). Using the average replacement ratio for a worker at the median wage and at two-thirds of the median wage in the first year of unemployment as the dependent variable yields

where standard errors of the coefficients (excluding the constant) are (.048, .054, .0003, .0010) and adjusted R 2 = 90.5. All coefficients are significant at the .05 level. Neither the share of elderly in the population nor the rate of unemployment are significantly different from zero when the replacement ratio is the dependent variable.

43 This procedure is advocated and given a Bayesian justification in Learner, Edward E., Specification Searches: Ad Hoc Inferences with Nonexperimental Data (New York: John Wiley and Sons, 1978). We did not consider the unemployment rate to be ”questionable” when the dependent variable included unemployment benefits.

44 In lines 1, 4, 5, and 6, the minimum estimate is obtained by excluding Austria and the maximum estimate is obtained by excluding Finland. In line 2, the minimum is obtained by excluding Norway and the maximum is obtained by excluding the U.S. In line 3, the minimum is obtained by excluding Finland while the maximum is obtained by excluding Austria.

45 Lorenzo Kristov, Peter Lindert, and Robert McClelland, ”Pressure Groups and Redistribution,” Journal ofPublic Economics 48 (July 1992).

46 Iversen, Torben and Soskice, David, ”An Asset Theory of Social Policy Preferences,” American Political Science Review 95 (December 2001).

47 Iversen and Soskice (fn. 46), for example, report a correlation coefficient of .73 between their measure of the extent of vocational training and the W10/W90 ratio (p. 889).

48 Daron Acemoglu and Jorn-Steffen Pischke, ”The Structure of Wages and Investment in General Training,” Journalof PoliticalEconomy 107 (June 1999).

49 Moffitt, Ribar, and Wilhelm (fn. 7).

50 Ebbinghaus, Bernhard and Visser, Jelle, Trade Unions in Western Europe since 1945 (New York: Grove's Dictionaries, 2000).

51 Kenworthy calculates the share of individuals in advanced industrial societies who would be classified as living in poverty in the U.S., that is, living in households with incomes less than 40 percent of the median household income in the U.S. after converting their household income to U.S. dollars according to purchasing power parity and adjusting for family size; see Lane Kenworthy, ”Do Social Welfare Policies Reduce Poverty? A Cross-National Assessment,” Social Forces 77 (March 1999). The partial correlation coefficient between share living in poverty and the log of the 90/10 wage ratio is .69, controlling for GDP per capita for the fourteen countries where Kenworthy's sample overlaps with the sample of this paper.

52 OECD (fn.23).

53 United States Bureau of the Census, Statistical Abstract of the United States 1990 (Washington, D.C.: U.S. Government Printing Office, 1990); idem (fn. 21).

54 OECD (fn. 24,1996); OECD (fn. 24,1993) in the case of the U.S.

55 Swank (fn. 1).

56 Castles and Mair (fn. 32); Huber, John and Inglehart, Ronald, ”Expert Interpretations of Party Space and Party Locations in Forty-two Societies,” Party Politics 1 (January 1995).

57 Blais, Andre and Dobrzynska, Agnieszka, ”Turnout in Electoral Democracies,” EuropeanJournal of Political Research 33 (March 1998).

58 Organization for Economic Cooperation and Development, Statistical Compendium 1997/2 (CD-ROM) (Paris: OECD, 1997).

* We thank Nolan McCarty and three anonymous referees for helpful comments. We thank the MacArthur Foundation and the Norwegian Research Council for financial support.

Related content

Powered by UNSILO

Earnings Inequality and Welfare Spending: A Disaggregated Analysis

  • Karl Ove Moene (a1) and Michael Wallerstein (a2)

Metrics

Altmetric attention score

Full text views

Total number of HTML views: 0
Total number of PDF views: 0 *
Loading metrics...

Abstract views

Total abstract views: 0 *
Loading metrics...

* Views captured on Cambridge Core between <date>. This data will be updated every 24 hours.

Usage data cannot currently be displayed.