Corruption and entrepreneurship: evidence from Brazilian municipalities

Dreher and Gassebner (2013) argue that the lack of empirical support for the “grease the wheels” hypothesis is not surprising because corruption can impact income growth through many channels.

The focus of this paper is on productive entrepreneurial activity, namely entrepreneurial activities that are conducive to economic growth.

These dimensions of institutional quality are often subsumed under the concept of economic freedom. The above-referenced studies employ the county-level measures of economic freedom provided by Gwartney et al. (2014) or the state-level measures provided by Stansel et al. (2014).

Though not reported, we find that corruption discourages business activity in 12 out of the 17 sectors considered in this paper. The five unaffected sectors are public administration, utilities, agriculture, fishing, and other services.

The OECD report actually lists the transportation and communication industries separately; however, we combine them in one group because that is how they are categorized in Brazil.

The idea of corruption resulting from the receipt of federal funds is not new. Leeson and Sobel (2008) find that US states that receive more Federal Emergency Management Agency (FEMA) funds are more likely to be corrupt.

Bologna (2014b) also uses these audit data to examine the effect of corruption on GDP and income.

We keep our unit of observation at the sector level, rather than combining all establishments into a single measure, as some sectors may be more sensitive to corruption than others.

We have defined other cutoffs as our measure of small business activity, and our findings are robust. Results using these alternative cutoffs are in Table A2 of the online Appendix.

Ideally, one would be able to separate out firm births and deaths. However, given data availability this is not possible for our analysis.

As in Dreher and Gassebner (2013), we note that when interpreting these results, one must proceed with caution because, while corruption may allow entrepreneurs to evade inefficient regulations, corruption may also increase these regulations in the long-run. Separating these mechanisms is impossible given the data available.

The case where δ ₁ < 0 and δ ₂ < 0, does not necessarily go against the “grease the wheels” hypothesis, but suggests that corruption is more harmful in areas with high quality institutions.

Summary statistics on the average number of establishments in each sector are presented in Table A1 of the online Appendix.

Alternatively, corruption and the underground economy may be substitutable (Dreher et al. 2009; Dreher and Schneider 2010). In this case, an estimated negative association between corruption and formal sector activity may understate the negative effect corruption has on economic activity overall.

Dreher and Schneider (2010) note that the relationship between corruption and the informal sector likely depends on a country’s income level. For low-income countries, the relationship usually is complementary, whereas it usually is substitutable in high-income countries. Since Dreher and Schneider (2010) would consider Brazil to be low-income during our sample period, we hypothesize that the relationship there is complementary. When we estimate the effect of corruption and institutional quality on the size of the informal sector, we find that neither have a real effect on the informal sector in Brazil. The amount of tax revenue the government receives per formal employee is the driving factor of the size of the informal economy. These results are shown in Table A3 of the online Appendix.

Unfortunately, GDP is available only until 2010. Therefore, for all years beyond 2010, GDP per capita in 2010 is used as a proxy.

Extractive industries include oil, gas, and mining. These industries focus on the removal of minerals from the ground, not the distribution of these resources.

The results from these regressions are provided in online Appendix Tables A4 and A5. In only three of the remaining 15 sectors included in our dataset, we find no statistically significant deterrent effect of corruption. These are the agriculture, fishing, and other services industries.

Coefficient estimates for the marginal effects that we present in this section are reported in Table A6 and Table A7 of the online Appendix.

Note that it is possible that the offsetting relationships between the direct effect and interaction are due to non-linearity of the interaction. Exploring this relationship is beyond the scope of the present analysis.