This paper outlines a new approach to the study of the welfare state. Contrary to the emphasis on "decommmodification" in the current literature, we argue that important dimensions of the welfare state-employment protection, unemployment protection, and wage protectionare designed to make workers more willing to invest in firm-and industry-specific skills that increase their dependence on particular employers and their vulnerability to market fluctuations. Workers will only make such risky investments when they have some insurance that their job or income is secure. Otherwise, they will invest in general, and therefore portable, skills. In turn, because the skill composition of the work force constrains the set of product market strategies that firms can pursue successfully, employers will support social protection that facilitates the set of skills they need to be competitive in particular international product markets. We show that our argument is consistent with observed clusters of social protection and skill profiles among OECD countries, and that these clusters are associated with very different distributional outcomes and patterns of gender-specific labor market segmentation.

W e present a theory of social policy preferences that emphasizes the composition of people's skills. The key to our argument is that individuals who have made risky investments in skills will demand insurance against the possible future loss of income from those investments. Because the transferability of skills is inversely related to their specificity, workers with specific skills face a potentially long spell of unemployment or a significant decline in income in the event of job loss. Workers deriving most of their income from specific skills therefore have strong incentives to support social policies that protect them against such uncertainty. This is not the case for general skills workers, for whom the costs of social protection weigh more prominently. We test the theory on public opinion data for eleven advanced democracies and suggest how differences in educational systems can help explain cross-national differences in the level of social protection.Torben Iversen is Professor of Governmentparticipants in the Kennedy School's 2000 Inequality Summer Institute for many helpful comments. We are particularly grateful to Jim Alt and Ada W. Finifter for their many helpful suggestions. Torben Iversen gratefully acknowledges financial support from the Glenn Campbell and Rita Ricardo-Campbell National Fellowship at the Hoover Institution, Stanford University. 1 The trade policy argument treats skilled labor as a more or less abundant factor, and relative abundance is hypothesized to favor free trade. See Rogowski 1987, Frieden 1991, and especially Scheve and Slaughter 1999 for different versions of this argument.2 Alt and his colleagues (1999) provide empirical evidence that lobbying rises with the asset specificity of industries. See also Alt et al. 1996 for a more theoretical treatment of this and related arguments concerning the importance of asset specificity. 3 There is an analogy here to the argument that capital mobility increases the incentives of capital owners to oppose restrictions on trade and financial mobility. See Frieden 1991 and Bates, Brock, and Tiefenthaler 1991.American Political Science Review Vol. 95, No. 4 December 2001 6 The Meltzer-Richard model has a more general tax disincentive function than that used here. In consequence the tax rate that maximizes tax revenue can be less than 1.

BerlinS tandard political economy models of redistribution, notably that of Meltzer and Richard (1981), fail to account for the remarkable variance in government redistribution across democracies. We develop a general model of redistribution that explains why some democratic governments are more prone to redistribute than others. We show that the electoral system plays a key role because it shapes the nature of political parties and the composition of governing coalitions, hence redistribution. Our argument implies (1) that center-left governments dominate under PR systems, whereas center-right governments dominate under majoritarian systems; and (2) that PR systems redistribute more than majoritarian systems. We test our argument on panel data for redistribution, government partisanship, and electoral system in advanced democracies.

The authors propose a synthesis of power resources theory and welfare production regime theory to explain differences in human capital formation across advanced democracies. Emphasizing the mutually reinforcing relationships between social insurance, skill formation, and spending on public education, they distinguish three distinct worlds of human capital formation: one characterized by redistribution and heavy investment in public education and industry-specific and occupation-specific vocational skills; one characterized by high social insurance and vocational training in firm-specific and industry-specific skills but less spending on public education; and one characterized by heavy private investment in general skills but modest spending on public education and redistribution. They trace the three worlds to historical differences in the organization of capitalism, electoral institutions, and partisan politics, emphasizing the distinct character of political coalition formation underpinning each of the three models. They also discuss the implications for inequality and labor market stratification across time and space.

An influential line of argument holds that trade exposure causes economic uncertainty and spurs popular demands for compensatory and risk-sharing welfare state spending. The argument has gained renewed prominence through the recent work of Garrett (1998) and Rodrik (1997;1998). This paper argues that the relationship between trade openness and welfare state expansion is spurious, and that the engine of welfare state expansion since the 1960s has been deindustrialization. Based on cross-sectional time-series data for 15 OECD countries we show that there is no relationship between trade exposure and the level of labor market risks (in terms of employment and wages), whereas the uncertainty and dislocations caused by deindustrialization have spurred electoral demands for compensating welfare state policies. Yet, while differential rates of deindustrialization explain differences in the overall size of the welfare state, its particular character --in terms of the share of direct government provision and the equality of transfer payments --is shaped by government partisanship. The argument has implications for the study, and the future, of the welfare state that are very different from those suggested in the trade openness literature.