Selection of Large Sub-Samples from the Continuous Sample of Working Lives Representative of the Benefits Provided by the Spanish Public Pension System

Abstract

The Continuous Sample of Working Lives (CSWL) is a set of anonymized microdata with information about individuals taken from Spanish Social Security records. It provides very valuable information, which is used in many studies on labor economics and in the analysis of the Spanish public pension system. This article presents two major contributions: The first is an analysis of how representative CSWL is of the population of pensioners for the period 2005-2013. It is concluded that the CSWL does not follow the same distribution as the population with respect to some types of benefits, and that this happens in most waves. One of the reasons is that it is obtained by simple random sampling, so the fit to the population by age, gender and type of pension is worse than it would have been under stratified random sampling (SR) with proportional allocation. As a possible solution, researchers could obtain a sub-sample by SR from CSWL. In this article we illustrate that this implies giving up a large number of pension records and thus reducing the diversity of data on working lives and types of pensioner. Hence, a second contribution of this paper is the application of a novel methodology based on optimization for choosing large sub-samples drawn from CSWL that are more representative of the population of pensioners. The advantages of using this sample design procedure are estimated by comparing the estimates for total pension expenditure provided by the CSWL and by such sub-samples.