An often-heard argument is that South Africa's very high crime rate is the main reason for the country's small share of business ownership. Combining a fixed-effects model with an instrumental variable approach, we estimate the effect of crime on self-employment and business performance using a matched data set of census, survey and police data. In contrast to previous studies, which focus on perceived rather than actual crime and often deal with geographically limited areas, we do not find robust evidence that high crime rates have a negative impact on self-employment. Although the impact of crime is statistically significant and negative, it is economically small. Moreover, our results suggest a positive rather than a negative relationship between robbery and burglary and sales and average business profits. These results suggest that crime may not be in general a serious threat for small businesses in low and middle-income countries.