Words Speak Louder Than Actions: The Impact of Politics on Economic Performance

In this paper, we present a new approach to disclose the impact of policy preferences of parties on economic growth. We derive policy preferences of parties, such as political ideology and preferences restricted to single policy dimensions, from the content analysis of party manifestos. This data is obtained from the Comparative Manifesto Project (CMP), an international large scale project in political science that has collected the wording of thousands of election manifestos. These are coded according to the frequency of sentences devoted to 56 different policy issues, which comprise economic and non-economic issues and are regarded as proxies for the party preferences in the related political science literature. Based on this data, it is possible to calculate indices for the general left-right position as well as concerning specific policy areas that represent the political preferences of the parties or legislatures. In the empirical section, we analyse per capita GDP growth in a panel of 23 OECD countries for the period 1971-2004. We estimate several models containing a set of typical control variables; fixed effects control for time-invariant country-specific influences. First, we follow an approach which is typical for the empirical political economics literature and employ dummy variables for the differentiation of left-wing and right-wing governments. However, we are not able to detect a significant impact of partisanship on growth rates. This implies that partisanship defined according to the broad distinction between leftwing and right-wing governments cannot be proven to affect economic growth. Then we proceed with analyses based on 7 indices that account for the political preferences concerning different policy areas. These indicators are first tested individually in our empirical model, and then jointly in a model averaging procedure, the weighted averaged least squares (WALS) method. A strong and robust negative impact on economic growth can be detected for parties that support market interventions, whereas a positive impact can be found for parties that aim at setting incentives for business as well as those which promote technology and infrastructure. Welfare state policies are also found to have a significant negative effect in our classical estimations, but show a much weaker impact in our WALS analysis. These results are robust to several changes in specification. The consideration of programmatic profiles of parties thus allows us to identify a growth effect for certain policy areas. This underlines the importance of choosing the correct ideology measure for the analysis of partisan effects on economic outcomes.

Osterloh, Steffen (2010), Words Speak Louder Than Actions: The Impact of Politics on Economic Performance, ZEW Discussion Paper No. 10-092, Mannheim.