Do Parties Make a Difference? Parties and the Size of Government in Liberal Democracies

The paper attempts to determine whether parties of the left, when in government, spend more than parties of the right. It first reviews the theoretical literature and concludes that parties are likely to make a difference, but only a modest one. It then reviews previous empirical studies, which come out with conflicting results. It finally proposes a study that covers 15 liberal democracies over a period of 28 years, from 1960 to 1987, and combines longitudinal, cross-sectional, and pooled designs. The analysis shows that parties of the left do spend a little more than parties of the right. The difference, however, emerges only for majority governments whose party composition remains unchanged over a number of years, an indication that it takes time for parties to affect total spending. A quarter of a century ago, Dye (1966) concluded that policy variations in the United States ought to be attributed essentially to economic factors, political variables proving to be largely uninfluential. This was a most disturbing result for the political science discipline, whose relevance, it would seem, depends on the substantive importance of the phenomena it examines. Ten years later, Wilenski's (1975) study of the welfare state came to similar conclusions: the root cause of the level of welfare expenditure in a country is economic growth, and the mechanism that translates economic change into public policy is demographic rather than political. Enlightened political scientists knew it could not be so, and a counterattack was mounted. The case was made that politics matters, and refined analyses vindicated the revisionist view (Castles and McKinlay 1979; Castles 1982). This paper is about the most cherished of all political variables, the parties. Political scientists typically view parties as fulfilling an essential role in democracy (Epstein 1983). A strong party system is considered as a necessary condition for an adequate representation of interests and opinions (Birch 1971). Our objective is to determine whether it matters