QCA is supposed to bridge the gap between qualitative and quantitative methods. The purpose offers certain advantages in comparison to regression analysis and is reflected in the literature on fsQCA. (Ragin 2006) This is why I will start this analytical chapter with a crosscountry growth regression, using the most common uncalibrated variables in the largest possible sample. The results will allow for a comparison of results, and of methodological advantages or disadvantages of QCA in particular.

I will then proceed to the fsQCA analysis of the six cases of the states of the Gulf Cooperation Council and of the larger sample of countries with important resource rents. The methodological chapter has demonstrated the complexity of variable/ condition choices and their relation to the sample countries and sample size. Therefore, table 31 gives an overview of the variables and conditions used for the different analyses to be conducted in this chapter. To better differentiate between variables and conditions, variable names will appear in capital letters, while condition names are in small caps.

5.1 Regressions

In this first model, I deliberately follow the common practice of cross-country regressions with the exception of a few variables, which I operationalized differently. This means that insight gained from case studies of certain countries will not be considered because regressions tend to discuss theories or previous models but rarely cases. There is also a necessity to find one operationalization for all cases – sometimes despite knowledge of the inappropriateness of this operationalization for certain cases.

The model is a multivariate OLS-regression based on data which has been averaged from 1973 to 2006 (or slightly earlier for independent variables). The dependent variable is non-oil growth from 1973 to 2006 (NDGROWTH). A model based on panel-data has been rejected due to massive data problems affecting the dependent variable in particular. While averaged data is likely to grasp the overall tendencies of the variables, the single data points are in many cases highly irrational and improbable. This is especially true for “non-oil” data. A time series data analysis would over-interpret changes from one year to the next.

The independent variables are initial non-oil GDP (NDINITIAL), investment as a share of GDP (INVEST), fertility rate (FERTILITY) and a measure of human capital (AVSCHOOLING). INVEST and NDINITIA) enter the regression as natural log (LOGNDINITIAL, LOGINVEST) because a graphical illustration showed a logarithmic shape as in many growth regressions.