Performance of detrending models of crop yield risk assessment: evaluation on real and hypothetical yield data
2015
Ye, Tao | Nie, Jianliang | Wang, Jun | Shi, Peijun | Wang, Zhu
Detrending is a widely used technique for obtaining stationary time series data in residual analysis and risk assessment. The technique is frequently applied in crop yield risk assessment and insurance ratings. Although several trend models have been proposed in the literature, whether these models achieve consistent detrending results and successfully extract the true yield trends is rarely discussed. In the present article, crop insurance pricing is evaluated by different trend models using real and historical yield data, and hypothetical yield data generated by Monte Carlo simulations. Applied to real historical data, the linear, loglinear, autoregressive integrated moving average trend models produce different risk assessment results. The differences among the model outputs are statistically significant. The largest deviation in the county crop assessment reaches 6–8 %, substantially larger than the present countrywide gross premium rate of 5–7 %. In performance tests on simulated yield trends, popular detrending methods based on smoothing techniques proved overall superior to linear, loglinear, and integrated autoregression models. The best performances were yielded by the moving average and robust locally weighted regression models.
显示更多 [+] 显示较少 [-]