Theories of autocratic co-option rest on at least two central, but contested assumptions: That legislators are sufficiently empowered to demand concessions from the president, and that even unfree elections are competitive enough to discipline political elites. In the case of public investments in non-democratic countries and developing democracies, I find support for both of these crucial assumptions. Using data on georeferenced investment projects across Africa, I find that these investments are more in line with the president's preferences prior to presidential elections, but more in line with legislators' preferences close to legislative elections. Taken together, the findings strongly suggest that legislators in such systems do have the power to demand concessions from president, and that unfree elections can discipline political elites.

A common problem in developing countries is an asymmetry between political leaders and local populations, which creates two problems when politicans want to buy voters with pork investments. First, investments are vulnerable to freeriding, since voters that benefit from an investment can still support a challenging candidate he or she prefers, without fear of repercussion. Second, politicians lack information about the needs of the local populations, and what types of investments that are more likely to be appreciated by voters. With a network of intermediate brokers, both of these problems can be solved, in that brokers can gather the necessary information, inform voters that the politician fought for them to get this investment, and monitor and sanction the parts of the electorate that fail to vote ``correctly''. But brokers have incentives to exaggerate the need for investments from their patron, in order to acquire more resources. Trust is therefore of fundamental importance for such political networks. And politicians are more likely to find trustworthy individuals among close friends from their home regions.