ABSTRACT
Theory predicts that societies with higher social capital have lower levels of government corruption. This hypothesis has been difficult to evaluate empirically because of widespread reliance on aggregated, country-level measures of trust and corruption, and reverse causality. We utilize micro-level international data from 34 African nations and two alternative measures of both trust and corruption to re-evaluate the evidence of this theorized relationship. To account for reverse causality, we use historical transatlantic slave exportation data as an instrumental variable. Slave trade has historically lowered interpersonal trust in Africa, which has had a long lasting effect on past and present levels of corruption. Our findings indicate that a higher level of personalized trust leads to more corruption, whereas a higher level of generalized trust leads to less corruption. This result is robust to alternative specifications of regression models, including different measures of trust and experienced versus perceived types of corruption.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 This paper was developed from a conference presentation by H. Czap and K. Nur-tegin titled ‘The impact of trust on corruption’. IAREP/SABE 2019 Conference 1st–4th September 2019.
2 Banerjee does not employ the term ‘generalized trust’ but instead is referring to a general belief in trustworthiness.
3 This result is not ubiquitous, however (see for example Braun & DiTella, Citation2004).
4 In the 5th round of the Afrobarometer respondents from French colonies did not provide information on their ethnicity. Since observations with missing values are omitted from the regression, econometrically we cannot include the colonial variable and the ethnic variables in the same regression.
5 The Pearson Correlation Coefficient between particular and generalize trust is relatively low at 0.29 (Spearman’s Rank Correlation provides a similar result). This would allow using both measures of trust in the same regression. However, since we only have one instrument in slave trade, and the purpose of the OLS regression is to establish a comparison point, we opt to run separate regressions for each measure of corruption and trust.