The Conservation Reserve Program And Its Effect On Land Values

Keywords: real options, endogeneity, agricultural land value, agricultural land productivity, Conservation Reserve Program (CRP), Environmental Benefits Index.

This dissertation is a series of three essays exploring the determinants and effects of enrollment in the United States Department of Agriculture's (USDA) Conservation Reserve Program (CRP). The CRP is the United States' largest federal conservation program, currently enrolling over 34 million acres of productive cropland. The CRP pays landowners to idle productive cropland by replacing crops with approved covers such as native grasses or trees. The CRP has a large and wide-ranging impact on both CRP and non-CRP land through its effects on farm profits and farm and non-farm economies, and enlightened CRP policy requires understanding of the determinants of CRP enrollment as well as the magnitude of its effects. In the first essay I use a stochastic dynamic programming framework to construct an options model of CRP enrollment that characterizes landowner decisions to enroll in the CRP in terms of a threshold value of current agricultural returns. The model predicts changes in enrollment choices due to differences in market uncertainty and individual-specific risk aversion, and to changes in policy variables such as the length of CRP contracts and the frequency of sign-ups. The model predicts that landowner decisions to enroll in the CRP are significantly affected by variables absent from previous options models, and provides more realistic counterfactual policy analysis that provides policy makers with ex-ante insight into possible changes to the CRP. In the second essay I estimate the determinants of CRP enrollment using a parcel-level empirical model and Minnesota farmland data. The parcel-level data represent a significant improvement in data resolution over previous studies. I address specification concerns by including non-CRP government payments and a uniquely comprehensive index of land productivity, and use a censored normal regression framework to accommodate censoring in the participation data. All model specifications suggest a negative and statistically s.