Using Auxiliary Information to Improve Agricultural Statistics – Advantages of the Calibration Approach over Poststratification Weights

Abstract

Official statistics are often based on samples representing a certain population. Because participation in a sample is usually voluntary, bias might result from so-called non-sampling errors such as nonresponse. Weighting procedures are intended to correct these errors by assigning a certain weight to each observation in the sample. In many official agricultural statistics, such as the Bavarian Agricultural Report, poststratification is used. In this process, the population is stratified according to different dimensions (e.g. farm type, farm location and farm size) and weights are assigned to all farms in a stratum so that the sum of the weights in that stratum corresponds to the number of observations in that stratum in the population. However, when estimating the population average, important characteristics (such as the farm size) may still be biased. Using a Bavarian farm sample, the present study shows how the so-called calibration approach, utilising auxiliary variables to adjust weights, outperforms the poststratification procedure in terms of estimating important population characteristics.