My review of “Why Welfare States Persist,” by Clem Brooks and Jeff Manza, for Political Science Quarterly:
Why do welfare states persist? Because they are popular, argue Clem Brooks and Jeff Manza in their new book, a statistical study of the connections between public opinion and policies in 16 rich countries in Europe and elsewhere.
Rich capitalist democracies around the world differ widely in their welfare states—their systems of government–provided social support–despite having comparable income levels. Brooks and Manza report that welfare state spending constituted 27% of GDP in “social democratic countries” such as Sweden and 26% of GDP in “Christian democratic countries” such as Germany, but only 17% in “liberal democracies” such as the United States and Japan. These differences are correlated with differences in income inequality and poverty rates between countries.
In their book, Brooks and Manza study how countries with different levels of the welfare state differ in their average policy preferences, as measured by a cross-national survey that asks whether respondents think the government should (a) provide a job to everyone who wants one, and (b) reduce income differences between rich and poor. Brooks and Manza find that countries where government jobs policies and redistribution are more popular are the places where the welfare state is larger, and this pattern remains after controlling for time trends, per-capita GDP of the country, immigration, women’s labor force participation, political institutions, and whether the ruling party is religious or on the left. The analysis is based on the following countries: listing in approximate order of increasing welfare state sizes, the United States, Japan, Ireland, Australia, Canada, New Zealand, Spain, Italy, the United Kingdom, Holland, Germany, Norway, Switzerland, Austria, France, and Sweden.
Brooks and Manza make a convincing case that attitudes are indeed correlated with policies, which implies that the voters in these countries are generally getting what they want, at least when considering comparative levels of social welfare spending. Chapter 2 of the book attempts to go further and make a claim of causality, to say that variation in countries’ attitudes are not just associated with policy variation but are actually a contributing cause of these policies. As a substantive matter, the causal claim undoubtedly has truth: in a democracy with all other things equal, we would expect a change in attitude to generally push toward the corresponding policy. (One can imagine exceptions, for example if an issue is “owned” by a faction or a minor party in a multiparty system, such that an increase in its vote actually harms the coalition that might advance the policy. But on average over 16 countries, we would expect a positive causal effect.) That said, I do not see that the statistical methods used by Brooks and Manza establish causality in the way that they claim. Using a statistical “test for endogeneity” cannot get around the fundamental issue that this is a cross-national comparison based on observational data: some countries have bigger welfare states than others, and these tend to have higher support for welfare states, even after (approximately) adjusting for some country-level factors.
As the authors note, the connections between attitudes and policies are complex. On one hand, governments are constrained by the general popularity of programs that give to the majority of voters; on the other, fiscal constraints make it difficult for governments to provide the sort of “Santa Claus” programs that citizens might want. Complexity arises because voters are aware of these constraints and typically don’t want to support political parties that don’t have a chance of winning or economic policies that are judged to be unsustainable.
Turning to surveys from individual countries–Sweden, Norway, Holland, and the United States–Brooks and Manza find that attitudes toward government-provided social services vary by country but have changed little from 1975 to 2000. Cross-national differences in attitudes, as well as in policies, seem stable and not tied to trends or to short-term factors such as the business cycle or changes in the party in power. Meanwhile, the size of the welfare state in rich countries has been stable since 1980, although with variation in individual countries (for example, a sharp decline in benefits in Switzerland and an increase in Italy).
These findings of stability in opinions and, in general, in spending, appear to contradict the conventional wisdom that welfare state policies have become repudiated in recent decades because of various factors, including: the fall of Communism and the corresponding discrediting of socialism as an economic policy; various economic crises since 1973 which have brought into question the ability of governments to pay for generous welfare benefits; and the growing presence of immigrants from poor countries, which has reduced the social consensus for income redistribution. One possible reconciliation of Brooks and Manza’s story and the general “decline of the welfare state” narrative is that, since 1980 or so, we have seen a conflation of welfare state expansion and reduction which happens to have averaged to a pattern of stability. In the wake of an aging population and lower employment rates, health-care spending has increased while job security programs have declined. Perhaps Brooks, Manza, and others who know more about this topic can let us know if this attempted synthesis makes sense.
Brooks and Manza have made a useful contribution by combining information from several sources to link public opinion and public policy on welfare provision. Various pieces of the story are well known, but I, for one, have not seen it all put together in this way. The book makes a compelling case for how policy differences between countries can persist, even in our modern, globalized, and post-socialist economy.
P.S. Lane Kenworthy points me to this discussion he wrote of Manza and Brooks’s argument, expressing concerns similar to mine about their causal reasoning. Kenworthy’s article has lots of pretty graphs and seems like a good start in the struggle to figure out a good general way to think about time-series cross-sectional data. The graphs on page 9-12 are particularly helpful. (I wouldn’t order them alphabetically, I’d make them smaller so more can be fit on one page and thus be visible at once, and I’m not a fan of the double-y-axis style, but these are all quibbles.)