Mean Streak: Best Practices for Replacing Missing Data for Index Insurance Design

Abstract:

Index insurance is a relatively new risk transfer tool, which can give farmers worldwide opportunities to take productive risks with coverage. As it is such a new tool, however, guidelines for best practices are ever-developing. This paper aims to better understand how different methods for completing datasets with missing data compare and if certain methods may bias payouts for index insurance, using drought index insurance as a guideline for creating indices. This paper uses a complete dataset, removes part of the data, and uses multiple different filling methods for replacing the missing data. We then compare how the different methods perform compared to the original dataset, and finally we construct historical payouts for the filled datasets and compare their performance to the original dataset.