The literature frequently recommends purposive sampling of elites based on the assumptions that random sampling negatively affects the response rate and that it induces bias. I test these assumptions drawing on metadata from 282 samples of political, economic, and social elites, and on microdata from 2,658 elites. First I use permutations to calculate confidence intervals for the expected response rate following each sampling method. Second, I estimate the effect of random sampling on the final response rate using a range of regression models. Finally, I compare the distributions of the estimators for the average age, the share of male elites, and elites’ ideology by simulating repeated random and purposive samples. Results indicate that both random and purposive sampling of elites generate sufficiently large samples, as well as consistent and unbiased estimators of population parameters. Contradicting methodological guidelines in the field, the conclusion is that random sampling of elites is efficient.