In contrast with previous index model land applications, this article shows that the land allocation problem is a portfolio model with two constraints, namely, investable funds and land. The two-constraint model implies a drastic reinterpretation of what previous studies have quantified as diversifiable and systematic risks in agricultural production. The article also argues that the constant marginal-rate-of-product-substitution (MRPS) technology implicit in financial portfolio models is unlikely to hold in a production context such as the land allocation problem, and that the index model must be modified accordingly. Farm-level data are used to illustrate and test the hypotheses advanced. Empirical results indicate that most of the risk for corn and soybeans is diversifiable, and that corn and soybeans are characterized by decreasing MRPS. The MRPS effect is found to be large from an economic standpoint and implies that crop diversification may be optimal, even for risk-neutral farmers.
