A cropland retirement policy contributes to the reduction of environmental externalities from agricultural production such as soil erosion, nutrient runoff and loss of wildlife habitat. On the other hand, participant's potential adverse behavior could undermine the environmental benefits of the policy. Several sources of such an unintended effect, known as “slippage", have been conceptually identified, but their empirical evidence has been scarce. This article tests one source of slippage caused by in-farm land substitution from noncropland to cropland as a result of farmland retirement in the U.S. Conservation Reserve Program (CRP). With the farm-level longitudinal data I can utilize cross-sectional and time variation of detailed individual farm characteristics to identify the causal relationship of CRP participation and subsequent slippage through in-farm land substitution. An identified assumption of the slippage estimate is verified by farm fixed effects, time-varying county fixed effects, and selection-on-observables. These could eliminate effects of unobservables that are potentially correlated with both the program participation and subsequent farmland reallocation decisions. Overall, slippage seems evident and fairly robust among specifications. It is found that an average program participant converts 14% of noncropland to cropping activities after enrollment. Results further show that participants with a larger share of uncropped land contribute more to slippage, indicating that farms with the excess capacity of conversion are more exible in the land allocation decision and thus likely to give rise to slippage. This suggests that additional restrictions on the rest of land use for participants and/or introduction of penalty points reecting the share of noncropland in the current auction mechanism can hinder such a backward incentive offsetting the program benefits.