Competing for Promotion: Are 'The Best' Always the Best?

Several selection processes use multistage tourname nts to choose the best candidates. The theoretical models predict that tournaments are efficient in selecting the best candidates, as they stimulate the best to perform relatively better than their opponents. Empirical tests are difficult, as data on the agents involved in these selections are scarce. Exploiting data from a field natural experiment, the World Swimming Championships, I show that two- and three-stage tournaments are effective for stimulating performance, selecting the best contestants and the winners are the players who are the most able to increase their relative performance from one stage to the next.