This paper considers the role of incentive payment programs in eliciting, estimating, and predicting landowners' conservation enrollments. Using both program participation and the amount of land enrolled, we develop two econometric approaches for predicting enrollments. The first is a multivariate censored regression model that handles zero enrollments and heterogeneity in the opportunity cost of enrollments by combining an inverse hyperbolic sine transformation of enrollments with alternative-specific correlation and random parameters. The second is a beta-binomial model, which recognizes that in practice elicited enrollments are essentially integer valued. We apply these approaches to Finland, where the protection of private nonindustrial forests is an important environmental policy problem. We compare both econometric approaches via cross-validation and find that the beta-binomial model predicts as well as the multivariate censored model yet has fewer parameters. The beta-binomial model also facilitates policy predictions and simulations, which we use to illustrate the framework.