We study whether opportunities to extract rents in a job affect the type of individuals who are attracted to it in terms of their underlying integrity. We do so in a laboratory experiment in which participants choose between two contracts that involve different tasks. We experimentally introduce the possibility of graft in one of them and study the sorting of subjects across contracts based on an incentivized measure of honesty. We find that the corruptible contract changes the composition of subjects because it attracts the most dishonest individuals and repels the most honest ones. In addition, we observe extensive graft when the opportunity is available. We introduce a double randomization strategy to disentangle the extent of which stealing responds to the aforementioned negative selection or to pure incentives (net of selection). We find that, in this setting, selection is the main driver of graft. Our results have clear policy implications to curb corruption.
