A Relational Model for Predicting Farm-Level Crop Yield Distributions in the Absence of Farm-Level Data

Central to designing and delivering an individual crop insurance program are historical individual farm-level yields, which serve as the foundation for setting coverage levels and rates. However, data scarcity and credibility, particularly lack of farm-level yield data, make it difficult to calculate individual expected losses. As a result, aggregate county-level data are often used to establish a base premium rate, and this may contribute to adverse selection, and thus, program losses. I develop a new relational model to predict farm-level crop yield distributions in the absence of farm-level yield data, to improve the accuracy of computing crop insurance premium rates. The relational model developed defines a similarity measure based on a Euclidean distance metric to select an optimal county from a reference country, from which farm-level yield data are "borrowed". An empirical analysis shows that the relational model achieves lower mean and standard deviation prediction errors compared to the benchmark model, and is able to recover the actual premium rates more closely.

Date and Time: 

Tuesday, June 13, 2017 - 13:30 to 14:00

Co-authors (not including you): 

Ken Seng Tan
University of Waterloo
Wenjun Zhu
Nankai University

Language of Oral Presentation: 


Language of Visual Aids: 


Type of Presentation: 




First Name Middle Name Last Name Primary Affiliation
Lysa Porth University of Manitoba