[영문]Strategies for reducing tax evasion include stricter enforcement, but taxpayer responses to increased enforcement are difficult to measure with field data. We use experimental methods to examine individual compliance responses to advance information on audit probability and productivity. Our design informs some individuals that their return will be audited prior to making their compliance decision, while other individuals receive information that they will not be audited; we also inform individuals of the productivity (fraction of unreported income discovered) of the audit. Announcement increases compliance of those told they will be audited, but reduces compliance of those knowing they will not be audited; the net effect is that overall compliance falls.