This paper uses experimental methods to examine the effectiveness of endogenous audit selection rules, in which the audit probability depends upon taxpayer actions. The results indicate that endogenous rules are able to generate compliance far greater than random audit rules, even when the random audit rate is 50 percent. In particular, a cutoff rule is most effective in increasing compliance, although it requires a large number of audits, and regimes in which back or future audits are dependent on a taxpayer's current declarations also significantly increase compliance. The results thereby provide strong support for strategic selection of tax returns.