chapter  18
56 Pages

Business associability, activities and governance: cross-national fi ndings

ByFRANZ TRAXLER, BERND BRANDL, SUSANNE PERNICKA

Adopting a cross-nationally comparative perspective, this chapter relates the above evidence from the country chapters to the theoretical propositions, as set out in Chapter 2. This is meant to complement the country studies also in methodological respects. The propositions will be transformed into testable hypotheses, and the concepts underlying them will be translated into quantitative measures. As far as possible, statistical methods will then be employed for testing the hypotheses. The main reason for adopting a quantitative design is that the number of countries under consideration is too large for a comparative in-depth study. Furthermore, qualitative designs do not enable analysis to expose the hypotheses to rigorous empirical testing. However, one has also to mention the problems of such quantitative design. In our case they mainly result from the fact that the number of countries is relatively small for the purpose of a statistical analysis. For a detailed discussion of these problems and possible solutions see Traxler et al. (2001). We confi ne ourselves to discussing two problems which are especially relevant, when it comes to interpreting the model specifi cations and their results. First, the power of statistical tests decreases with the number of observed cases. Hence, the special problem of small data sets is that one runs a high risk of rejecting a hypothesis because of statistical insignifi cance, although the hypothesized association is actually signifi cant in substantive terms. This means that the tests presented here are biased towards rejecting hypotheses, such that we can detect only the most important determinants of the dependent variable. To mitigate this problem, we proceed from a rather large probability value (i.e. p 0.10) equal to or below which the null hypothesis is rejected as insignifi cant. Second, small data sets limit the number of explanatory variables that can be included. If the data allow for a multivariate analysis, only the most relevant independent variables can be included.