Farmland loss is considered a serious problem by the public, and it is in part addressed by government initiatives to preserve farmland through the use of conservation easement programs. In order to prioritize which tracts are protected with these programs, it is important to understand and measure the non-market benefits of agricultural land. The contribution of this work is to provide national-level estimates of benefits as well as examine the possibility of geographic heterogeneity in preferences across states. This study uses choice experiment data on farmland attributes in the US, Georgia, Ohio and Maine. Sample selection was tested for and rejected. A random parameters logit model was estimated and significant preference heterogeneity was confirmed. This result indicates that although some variables may seem insignificant, they may actually be important to many individuals, but those individuals simply don’t agree on the value of those attributes. Consequently, a broad-based funding mechanism such as taxes may be less popular than more targeted mechanisms. After testing for scale and parameter equality, it was found that the US and Maine had different underlying parameters, which indicates that federal-level policy may be inappropriate, as some states may have different preferences for which farmland attributes should be prioritized.