We conduct a real effort experiment in which performance is not monitored and participants are paid according to their reported performance. Participants are paid according to a piece rate and a winner-take-all tournament and then select between the two schemesbefore performing the task one more time. Competition increases dishonesty and lowersoutput when the payment scheme is exogenously determined. Participants with a higher propensity to be dishonest are more likely to select into competition. However after selection, we find no output difference between piece rate and tournament. This is attributable to a handful of honest individuals who select competition.