Missing Data in Longitudinal and Linked Surveys

Slides

Issues in Weighting for Longitudinal Surveys, Peter Lynn (University of Essex) and Nicole Watson (University of Melbourne)

Utilising representatives indicators to evaluate non-response and non-linkage biases, Jamie C. Moore, Gabriele B. Durrant & Peter W. Smith (Southampton University)

Some issues in dealing with missing data, Harvey Goldstein (University of Bristol)

Regression with an Imputed Dependent Variable, Thomas Crossley (Essex, IFS and ESCoE)

Papers

Adjustment for survey non-representativeness using record-linkage: refined estimates of alchol consumption by deprivation in Scotland

Validation of non-participation bias methodology based on record-linked Finnish Register-based health survey data: a protocol paper

Assessing the Representativeness of Population-Sampled Health Surveys Through Linkage to Administrative Data on Alchol-Rleated Outcomes

The Good, The Bad and the Ugly: Measurement Error, Non-response and Administrative Mismatch in the CPS

Longitudinal surveys are crucial resources for analysing and understanding how economic and social behaviour change over time. Because they provide repeated observations of the same observational units over time, they enable researchers to answer a host of questions about how individual heterogeneous units respond to economic and social factors.

However, the longitudinal nature of these studies results in complex patterns of missing data. The interaction of attrition and time-varying nonresponse with the rich features of longitudinal surveys (such as rotating content) and the multipurpose nature of such studies results in a multiplicity of patterns of missingness for researchers analysing the data in different ways or for different purposes.

A frontier in the provision of data for social science and biosocial research is the linkage of survey data with administrative and other external sources of data. This allows administrative data to be leveraged by the addition of information from survey questions on quantities that are not measured in the administrative data, and vice versa. But data linkage results in complete cases only for the intersection of cases that are complete in both the survey and administrative data. Moreover, the linkage processes itself can result in additional missing data, through lack of consent or failure to match. Linking external data to longitudinal surveys, while offering enormous opportunities, further complicates the missing data challenges inherent in such surveys.

Developing new methods to address the challenges caused by these missing data problems is a research area of first-order importance.

This workshop brings together statisticians, econometricians, survey designers, and data users to discuss new results and open questions in this important research area. The workshop will discuss practical experience in collecting and using panel data, advances in statistical and econometric methods to address missing data problems, and applications of these methods to important research questions. The workshop will conclude with a panel discussion on the current state of knowledge, open research questions, and promising avenues of research.

Confirmed workshop speakers will include Chris Bollinger (Kentucky), Peter Lynn (Essex), Jamie Moore (Southampton), Megan McMinn (Glasgow), Lars Nesheim (UCL), George Ploubidis (UCL) and Jeff Wooldridge (Michigan State).

This is a joint cemmap and Understanding Society workshop