Advances in Longitudinal Survey Methodology

16Assessing Discontinuities and Rotation Group Bias in Rotating Panel Designs

Jan A. van den Brakel^1,2, Paul A. Smith³, Duncan Elliott⁴, Sabine Krieg¹, Timo Schmid⁵, and Nikos Tzavidis³

¹Methodology Department, Statistics Netherlands, Heerlen, Netherlands

²School of Business and Economics, Maastricht University, Maastricht, Netherlands

³Department of Social Statistics and Demography, University of Southampton, Southampton, UK

⁴Methodology Division, Office for National Statistics, Newport, UK

⁵Institute for Statistics and Economics, Freie Universität Berlin, Berlin, Germany

16.1 Introduction

Longitudinal surveys come in many forms, ranging from large‐scale cohort studies to repeating surveys which have some common sample elements from period to period, with intermediates depending on how the population and sample are refreshed and followed up. Rotating panel surveys are a specific type of longitudinal survey where on each survey occasion a new panel is added to the sample, and followed for a number of periods according to a predetermined pattern, after which the panel is (normally) dropped and replaced by a new one. Thus, on each occasion, the sample consists of several panels, each previously surveyed a different number of times.

Rotating panel surveys follow particular rotation schemes, and Steel (1997) coded these according numbers of months in and out of the sample. So a simple monthly survey where households were interviewed for six consecutive months (such as the Canadian ...