A main goal of affirmative action (AA) policies is to enable disadvantaged groups to compete with their privileged counterparts. Existing theoretical and empirical research documents that incorporating AA can result in both more egalitarian outcomes and higher exerted efforts. However, the direct behavioral effects of the introduction and removal of such policies are still under-researched. It is also unclear how specific AA policy instruments, for instance, head-start for a disadvantaged group or handicap for the privileged group, affect behavior. We examine these questions in a laboratory experiment in which individuals participate in a real-effort tournament and can sabotage each other. We find that AA does not necessarily result in higher effort. High performers that already experienced an existing AA-free tournament reduce their effort levels after the introduction of the AA policy. Additionally, we observe less sabotage under AA when the tournament started directly with the AA regime. The removal of AA policies, however, significantly intensifies sabotage. Finally, there are no overall systematic differences between handicap and head-start in terms of effort provision or sabotaging behavior.