Authors: Emanuele Massetti, Robert Mendelsohn, Shun Chonabayashi

We test three functional forms that relate land values and temperatures in a Ricardian model of US agriculture for the East of the US: a quadratic relationship based on average seasonal temperature and precipitations, a non-linear relationship based on degree days and a flexible functional form in which average seasonal temperatures are interacted with dummies. Results obtained using growing season average temperature and degree days are not significantly different. We do not find evidence of a threshold if we include degree days above 34 °C. Cold degree days instead matter and should not be omitted. Models that use a quadratic specification of average temperatures perform better than models that use degree days. This is in line with the agronomic literature. Degree days should be used to estimate the duration of phenological events rather than yields. Estimates of uniform +2 °C and +4 °C warming indicate that warming is significantly harmful for agriculture in the East of the US. The use of a more flexible functional form reveals that the relationship between temperatures and land values is flatter than in the quadratic. Seasons, within and outside the growing season, significantly affect land values and allow separating beneficial and harmful effects of warming more effectively.