Unmitigated climate change is projected to reduce average U.S. corn yields. Many of these projections rely on stable model coefficients, which may not hold in practice. We investigate the assumption of spatio-temporal stability of weather coefficients in regressions of U.S. Corn Belt yields during 1950–2014 and examine their implications for climate change projections. We reject the null hypothesis of time-constant weather shock coefficients for roughly one-third of sample counties and find that smoothly evolving parameters contribute to near-term prediction deterioration. We next estimate a dynamic econometric model that accounts for nonstationarities, outliers, and smoothly evolving parameters. Though average yields are projected to decline through 2050, there is considerable uncertainty when bootstrapping from a set of 30 downscaled climate models for both the RCP 4.5 and RCP 8.5 scenarios. Long-tailed distributions suggest a small number of counties could experience only small short-term yield decreases. Projections from a static analogue of our dynamic model have distributions with thinner tails, less dispersion, and higher means.

中文翻译：

---

玉米单产动态和天气冲击：气候变化对美国玉米带的影响

未减缓的气候变化预计将降低美国玉米平均单产。这些预测中的许多依赖于稳定的模型系数，而在实践中可能不成立。我们调查了1950-2014年美国玉米带单产的回归中天气系数的时空稳定性假设，并考察了它们对气候变化预测的影响。我们拒绝了大约三分之一样本县的时间常数天气冲击系数的零假设，并发现平稳发展的参数会导致近期的预测恶化。接下来，我们估计一个动态计量经济学模型，该模型考虑了非平稳性，离群值和平稳发展的参数。尽管预计到2050年平均单产将下降，从针对RCP 4.5和RCP 8.5场景的30个缩减气候模型中进行引导时，存在很大的不确定性。长尾分布表明，少数县的短期单产下降很小。来自我们动态模型的静态类似物的投影的分布具有更细的尾巴，更少的分散和更高的均值。