In statistical macro-comparative studies, the pooling of countries over time is the established method of analysis. The advantage of Time-Series—Cross-Section (TSCS) analysis is that it is able to analyze cross-section as well as temporal variance. However, most scholars use TSCS analysis only for the purpose to increase the number of observations and the degree of freedom in order to make their inferences more robust. Surprisingly, not much analytical and empirical work has been conducted to specify time structures in TSCS analysis. Aside for the established procedure to lag the independent variable by one year (t-1) in order to determine the causal direction, few more sophisticated models of time lags have been applied in most TSCS studies. This is astonishing considering TSCS-analysis has the potential to analyze complex time structures. Making the valid assumption that political processes do not run complete themselves within one year in all policies and countries, we need policy and country specific lag structures. For instance, in countries with many veto players, the political process may take a longer time than in countries with few veto players. The same is true for different policies or outcomes. Law-making may be faster than institution building which in turn may be faster than changing a policy outcome. The impact of the European Union on its member states may take more time than the effect of a regulation in domestic politics. All these important time variations are not sufficiently incorporated in the current trends of TSCS analysis. This may lead to biased and in some cases absurd results.
The proposed paper will apply the method of optimalized lags. Optimalized lags are derived from theoretical considerations about the timing of the political process, on the one hand, and an inductive method that identifies the optimalized lag structure in a regression model, on the other. Although this approach uses a lot of computer power, it is a helpful device to come to terms with various country, institution, and issue specific lag structures. Although an inductive analysis can only be a first step into a field which is demanding and complex in a theoretical as well as an empirical sense, it may pave the road for more deductive analysis in the future.
