As Francis Castles observes, quantitative comparative political economy has generated a “cacophony” of explanations for the welfare state crisis from 1980 to 1998. Political economists have identified causal factors including globalization, de-industrialization, population aging, declining birth rates, weak economic performance, welfare program maturation, political institutions, and partisanship. To Castles, however, these explanations rest on faulty empirical evidence. Methodological problems—in particular, unreliable comparative social expenditure data and overreliance on pooled designs and statistical techniques—have made it impossible to distinguish “crisis myths” from “crisis realities.” His proposed solution is to combine cross-national designs with the Organization for Economic Cooperation and Development's SOCX database on social spending, and he argues that doing so reveals that dominant theories are myths. His sweeping account is a spirited contribution, but it is bound to attract significant criticism.

What are the major claims of crisis of the welfare state? Castles correctly points out that crisis claims are not new. In fact, since the end of the glory days of the 1960s, welfare states appear to have been in perennial crisis. Beginning in the early 1970s, theories of overload charged that democratic politics generated “economically unsustainable” fiscal deficits. In the 1980s, accelerating financial and trade globalization emerged as the mechanisms of crisis. The imperatives of international competition purportedly weakened domestic policy autonomy, blurred partisan differences, forced welfare state retrenchment, and set in motion a “race to the bottom” toward allegedly more efficient liberal-residual welfare states. In the 1990s, de-industrialization, population aging, female labor participation, and low birth rates generated autonomous pressures that contributed to unsustainable budget deficits. These postindustrial phenomena challenged the dominance of globalization as an explanation for the welfare state crisis.

Why does Castles believe that these accounts are myths? The “cacophony” reveals a proliferation of explanatory theories that have not been subject to adequate empirical tests. A fundamental mismatch between supposedly universal theories and extremely narrow empirical evidence fails to distinguish between long-term relations linking historical processes and cross-national differences. In particular, pooled time-series cross-sectional designs do not provide satisfactory accounts of cross-national differences in the causal mechanisms shaping the extent, depth, and form of the crisis. Other evidence marshaled in support of the competing crisis claims tends to be anecdotal, rather than systematic, and tends to be biased toward corroborating preferred theoretical and policy arguments. The empirical evidence emerging from the mismatch leads to mistaken inferences and unwarranted policy prescriptions.

Why does the distinction of crisis myths from crisis realities require a methodological solution? In light of the existing cacophony, properly testing the existing theories would contribute more knowledge than generating new theories. To this end, Castles proposes three methodological advances. First is “demystification by measurement.” The harmonization of measures of social spending afforded by the SOCX database for the 1980–98 period enhances cross-national comparability of the extent, structures, trajectories, changes, and reforms of welfare states over the last two decades. Second is “demystification by comparative hypothesis testing.” Castles rejects pooled designs for obscuring cross-national differences, favoring instead pure cross-sectional designs in which countries are the central units. The design requires the compression of time into single points that capture stable temporal properties exemplifying the dynamics of the two historical subperiods (1980–89, 1990–98). Means capture level-shifts; absolute changes, defined in terms of the displacement between the last and the first observation of a given period, capture deterministic trend-shifts. Third is “statistical testing of cross-national hypotheses.” In the univariate dimension, comparisons involve means, trends, and coefficients of variations of welfare states and families of welfare states for the two historical periods. In the multivariate domain, comparisons involve bivariate correlations, “best fit” ordinary least squares regressions, and path analyses.

Univariate and multivariate tests lend support to the author's claims that some of the most popular explanations are myths. Chapter 2 undermines “race to the bottom” claims. Relatively small changes in aggregate expenditures across periods do not indicate significant decline in welfare spending and deterioration in welfare standards. Chapter 3 challenges the usefulness of aggregate social expenditures. Aggregates obscure movements in the finer meso-structure of social provisions. Disaggregated policies provide richer views of the extent to which polities have made structural changes, downsized existing programs, responded to new demands, and thus altered spending criteria. Chapter 4 questions a regional variant of the “race to the bottom,” namely, whether recent decades have seen the emergence of a European social model different from that of other Western welfare states. Chapter 5 dispels trade globalization as a “paper tiger”: It influences neither aggregate spending nor program priorities. Only Foreign Direct Investment exhibits a negative correlation. Moreover, welfare state downsizing finds no support after controlling for pension generosity. Significant partisan differences in welfare spending refute the “new politics” claims. Chapter 6 challenges the link between population aging and public spending. There are no significant cross-national differences in trajectories from 1980 to 1998. Far from having general applicability, the “old age crisis” is limited to particular countries. Chapter 7 weakens claims of a new crisis stemming from population decline. If a crisis exists, it is the demographic shift to below replacement levels of fertility. Supportive social policy initiatives, however, contribute effectively to the continued vitality of societies. Lastly, chapter 8 speculates on the future of welfare states. Since the recent past reveals little change, major disruptions of existing spending patterns in the immediate future are unlikely. The absence of a generalized crisis in recent decades suggests that generalized crises in coming decades are also unlikely. Rather, different types of welfare states confront different problems, and hence distinctive dilemmas.

Castles's mastery of the complexity of the debates on the welfare state crisis is superb in its simplicity. His exploration of the richness of the SOCX database is equally superb. But does his sweeping tour de force achieve the objective of demystifying crisis claims? The answer rests on the validity and reliability of empirical evidence, and hence on the appropriateness of his chosen research methodology. The hasty dismissal of pooled designs and statistical techniques because of their alleged inability to properly capture cross-national differences is bound to attract severe criticism. The pooled approach has driven the research in comparative quantitative political economy of the welfare state of the last decade, and it has generated a rich body of theoretical, methodological, and empirical research on the cross-national differences in the evolution of welfare states (e.g., see Duane Swank, Global Capital, Political Institutions, and Policy Changes in Developed Welfare States, 2002). Simply stating, without demonstrating, that the pooled approach is deficient—logically, theoretically, and methodologically—in capturing cross-national differences is not justified. Indeed, despite the constraints of a common structure that pooling imposes, welfare state scholars have identified theoretically meaningful cross-national temporal patterns as well as long-term breaks pointing to possible crises. Moreover, the compression of historical periods into single representative data points, as described by the means and deterministic trends, and as required by the pure cross-national design, more likely obscures the cross-national differences in historical evolution.

The critique of Castles's treatment of time can be pushed further by considering that the compression into means (and trends) ignores the implications of nonstationary historical dynamics. Compression aims at capturing modal, or representative, behavior, which identifies stable cross-national differences for two subperiods. Means, however, are meaningful only for stationary historical processes, for which they represent stable equilibria. Because the processes the author investigates are nonstationary, means used to capture stable dynamics are statistically “nonsensical.” With nonstationary processes, comparisons of means therefore cannot describe cross-national differences. Variances of nonstationary processes are also not meaningful; they are time-dependent and asymptotically tend to infinity. Accordingly, the coefficients of variation, which require stable means and standard deviations, cannot capture the homogeneity of families of welfare states. Lastly, because means and variances are integral to correlations and regressions, the concept of nonstationarity suggests that multivariate findings may themselves be “spurious.” This last critique can be extended to the pooled research that still relies on OLS regression despite the nonstationarity of historical processes (e.g., see Torbin Iversen, “The Dynamics of Welfare State Expansion: Trade Openness, De-industrialization and Partisan Politics,” in Paul Pierson, ed., The New Politics of the Welfare State, 2001). Co-integration analyses would better determine whether and how nonstationary processes drive the long-run evolution of welfare states as well as breaks in levels and slopes in such dynamics.

Castles's objective to enhance modeling and focus on the comparability of data is a positive advance. However, the question of whether the crisis of the welfare state is real or just a myth remains unanswered. To adequately test whether the cacophony of theories captures myth or reality, quantitative comparative political economy research must confront the theoretical and methodological implications of nonstationary dynamics that appear to drive the evolution of welfare states.