Total Survey Error in Practice

16Estimating Error Rates in an Administrative Register and Survey Questions Using a Latent Class Model

Daniel L. Oberski

Department of Methodology and Statistics, Utrecht University Utrecht, The Netherlands

16.1 Introduction

Administrative data obtained from government registers provide a wealth of potential for the social sciences (Entwisle and Elias, 2013). Collected during the normal course of public administration, for example, to tax, keep track of car ownership, pay welfare benefits, or send out calls to vote in elections (Wallgren and Wallgren, 2007), the variables available in registers can be used by survey researchers as direct variables of interest or as auxiliary variables in survey sampling, nonresponse adjustments, or validation studies. Such registers are often available longitudinally and for the entire population—a highly attractive combination for researchers, especially when linked to purpose‐designed surveys.

While administrative registers have many redeeming qualities for survey researchers, the fact that they have been collected for administration and not research can be a disadvantage. In particular, administrative registers may contain considerable measurement errors, including definition, reporting, timing, processing, editing, linkage, and coverage errors (Bakker, 2009; Groen, 2012). For example, Gomez and Glaser (2006) found that a staggering 83.3% of Native Americans and 29.9% of Hispanics were misclassified as a different race in U.S. registers, ...