File concatenation of survey data: a computer intensive approach to sampling weights estimation

File concatenation is an approach that can be used to integrate two (or more) sources of data which refer to the same target population. It consists in considering the concatenation of the two files as a unique data set. Although this approach seems to be natural in an integration procedure, it is not generally adopted, especially when data are obtained by means of different complex sampling designs. In fact, it requires the computation of the sampling weights for the concatenated data set and this can be often a very hard task. To this aim, some simplifying assumptions are adopted: for instance, when large population and simple survey designs are concerned, it can be reasonable to assume that the chance that a unit is included in both the samples is negligible. This assumption can be questioned when different and complex survey designs are adopted. This is, for instance, the case of enterprise surveys where survey designs with probability proportional to size selection are often considered and the probability of including in both the samples the same units is far from being negligible. In this paper we propose a method to deal with the problem of computing sampling weights of a concatenated sample in a general sampling design context. The method is a computer intensive approach and its applicability to real cases is shown by computing the weights of a data set obtained by concatenating two agricultural Istat surveys: the Farm Structural Survey and the Farm Economic Accounts Survey.