This resear chwas limited by the lack of daily milk production
data for a cross-section of actual dairy farms. Instead, stochastic
farm-level milk production was simulated using a production
function obtained from a controlled experiment. This procedure
may somewhat bias the results since actual farm-level
production may be more variable than production from the controlled
experiment. This would increase the idiosyncratic risk in
Eq. (1), decrease the correlation between the THI and realized
milk production, and thus reduce the benefits of the THI insurance.
Actual farm-level milk production data are necessary to
evaluate the extent to which our results are affected by using
data from a controlled experiment.