Estimating Non-linear Weather Impacts on Corn Yield—A Bayesian Approach
采用贝叶斯线性样条模型，同时估计降雨和温度对玉米产量的非线性影响及阈值，发现产量对天气变量呈凹形，高温和干旱显著减产，而适度热量在北部州有益。
We estimate impacts of rainfall and temperature on corn yields by fitting a linear spline model with endogenous thresholds. Using Gibbs sampling and the Metropolis - Hastings algorithm, we simultaneously estimate the thresholds and other model parameters. A hierarchical structure is applied to capture county-specific factors determining corn yields. Results indicate that impacts of both rainfall and temperature are nonlinear and asymmetric in most states. Yield is concave in both weather variables. Corn yield decreases significantly when temperature increases beyond a certain threshold, and when the amount of rainfall decreases below a certain threshold. Flooding is another source of yield loss in some states. A moderate amount of heat is beneficial to corn yield in northern states, but not in other states. Both the levels of the thresholds and the magnitudes of the weather effects are estimated to be different across states in the Corn Belt.