This chapter will further elaborate the methodological framework of this study. Apart from the qualitative case studies, two methods will be used: Fuzzy Set Qualitative Comparative Analysis and regression analysis (OLS), the common method growth and development economists use, although time series or panel data analysis has become more widespread.

Additionally, the question of growth in resource abundant countries has been the object of many regressions. Scholars such as Jeffrey Sachs and Andrew Warner (Sachs/ Rodriguez 1999; Sachs/Warner 1995a/1997; Sachs/Warner 1997; Sachs/Warner 2001; Sachs/Humphreys/Stiglitz 2007) or Eric Neumayer (2004) have tested the resource curse hypothesis multiple times and found supporting evidence. Both Sachs and Neumayer used resource dependence measures for their analysis. Other scholars have changed the models and the operationalization of variables. The significance of resource “abundance” variables has been put into question. Brunnschweiler, for example, found a positive correlation of resource abundance measures with growth, significant even when testing for institutions. Brunnschweiler also included regional dummies, while only the dummy for the Middle East and Africa was significant, and negatively correlated to growth. Using these dummies, Brunnschweiler could introduce a kind of regional path dependence in a regression. However, the dependent variable was the normal growth rate and not the non-oil growth rate. (Brunnschweiler 2008; Kropf 2010)

After decades of regressions, resource wealth seems to be one of the ambiguous variables and it seems unlikely that this kind of analysis will provide a clear answer. Several authors have already pointed to the probability that resource wealth and its effect on the economy or on democracy are part of a path dependent process. (Brunnschweiler 2008; Robinson/Torvik/Verdier 2006) Smith (2005) for example suggests that independent elites are a precondition for democratic transitions in resource abundance states. Nevertheless, I will start my analysis with a regression to contrast this approach and the common operationalization of the variables. To guarantee the comparability of the methods, I need to adapt the hitherto-used models to my approach as I cannot solely rely on previous models. To highlight the down- and upsides of each method, the regression will be the starting point of a multi-method appraoch, which encompasses a large n-sample, two small n-samples, and the case studies of the GCC states given in chapter two. This triangulation will also allow for a learning process from method to method.