Biases in Farm-Level Yield Risk Analysis due to Data Aggregation

We investigate biases in farm-level yield risk analysis caused by data aggregation from the farm-level to regional and national levels using the example of Swiss wheat and barley yields. The estimated yield variability decreases significantly with increasing level of aggregation, with crop yield variability at the farm-level being up to 2.38 times higher than indicated from national data. Our results show furthermore that inference on shape parameters based on aggregated data might be misleading. Using an example of farm yield insurance, we show that using crop yield variability estimates from aggregated levels leads to erroneous insurance contract specifications