The theoretical story presented in Chapter 3 led to the following prediction in hypothesis 2: a national SME business association formed by owners of private-sector SMEs will lead to a decline in corruption in autocracies when an elected multiparty legislature exists in these states. The corollary to this hypothesis posits that if either one or both of the following conditions are absent in autocracies – national business associations formed by private SME owners and elected multiparty legislatures – then the level of corruption will remain high in these autocracies. We statistically test hypothesis 2 and briefly evaluate the corollary to this hypothesis in this chapter by using a comprehensive time-series cross-sectional (TSCS) dataset of authoritarian countries.

The remainder of this chapter is organized as follows. We begin by describing the sample, the dependent variable, and the statistical models employed to test hypothesis 2. We next describe the independent and control variables included in the empirical specification for evaluating the second hypothesis. This is followed by a discussion of the results obtained from testing this hypothesis. The chapter concludes with a discussion of the implications of our findings.

Sample and Dependent Variable

The prediction in hypothesis 2 focuses on corruption in authoritarian regimes. We test this hypothesis by using a time-series cross-sectional (TSCS) sample of ninety-five authoritarian regimes between 1985 and 2010. Recall from the previous chapter that this TSCS sample (listed in Table 1.1) is constructed by using Cheibub, Gandhi, and Vreeland's (2010) criteria for defining authoritarian states. The start and end dates of our sample are determined primarily by the availability of authoritarian regime country-year data for our dependent variable and one of the independent variables of interest (i.e., the SME association measure described in this chapter). Our sample is, however, comprehensive as it includes all authoritarian states observed during the 1985 to 2010 period for which data to operationalize the dependent and independent variables are available.

The dependent variable in hypothesis 2 is the level of political corruption. Developing a reliable cross-national measure of political corruption is challenging considering the secrecy of the transactions involved in acts of corruption (Treisman 2007). Notwithstanding this problem, extant empirical studies on corruption primarily employ three main cross-national measures of the level of corruption: the International Country Risk Guide (ICRG) corruption index, Transparency International's (TI) Corruption Perceptions Index (CPI), and the poll-of-polls corruption data from the World Bank (WB).