Thomas F. Crossley

Professor of Economics, European University Institute

Contact: tfcrossley "at"


Google Scholar




05/03/2021, New working paper: Understanding Society Working paper 2021-03, "High frequency online data collection in an annual household panel study: some evidence on bias prevention and bias adjustment," released today.

This is joint work with Michaela Benzeval, Jonathan Burton, Paul Fisher, Annette Jäckle and Jamie Moore (all University of Essex), and Colin Gardiner from the Ipsos MORI Research Institute. We study the Understanding Society COVID-19 Study, and particular, whether and why it supports valid population inferences during the pandemic. Understanding Society is the UK’s largest household panel survey. From April 2020, participants in Understanding Society were additionally invited to partake in a series of web surveys (the Understanding Society COVID-19 Study) designed to capture higher frequency information during the COVID-19 pandemic. A key feature of the Understanding Society COVID-19 Study is that it is designed to be “representative” of the UK population, in the sense of allowing analysts to make unbiased estimates of averages, prevalences and other statistics for that population.In any survey, biases can arise if some segments of the population are not adequately covered by the survey design, or if particular segments of the population are less likely to respond to the survey. To prevent bias, the COVID-19 Study invites the full range of Understanding Society participants – both those who regularly use the internet and those who do not – to take part. Furthermore, periodically a subset of web non-respondents are invited to a telephone follow-up survey. Among web non-respondents, non-regular internet users are particularly targeted with this second mode. To further adjust for any bias arising from nonresponse, a weighting strategy is implemented that takes advantage of the rich background information available from past annual interviews of the same individuals in the main Understanding Society survey. In this paper we examine the efficacy of these bias reduction and bias adjustment measures.We find that both the telephone follow-ups and weighting help to reduce bias, but that inviting those who do not regularly use the internet to the web survey appears to be of little benefit. Overall we find that the the Understanding Society COVID-19 Study is representative of the target population.

05/02/2021, New working paper: IFS working paper 21/03, "MPCs through COVID: spending, saving and private transfers."

This is joint work with Paul Fisher (Essex), Peter Levell (Essex) and Hamish Low (Oxford) and uses new data from the Understanding Society COVID-19 Survey, which I am helping to run. MPCs were directly elicited from a representative sample of UK adults in July 2020 using receipt of a hypothetical unanticipated, one-time income payment. Reported MPCs are low, around 11% on average. They are higher, but still modest, for individuals in households with high current needs. These low MPCs may be a consequence of the prevailing economic un-certainty. Further, the fraction of respondents that report they would change their transfer payments to or from family and friends is almost as large as the fraction that report they would increase their spending. This means that targeting direct fiscal stimulus payments to high-MPC individuals could be partly undone, and that the aggregate MPC out of a stimulus payment need not equal the population-average MPC.

05/02/2021 Talk: I was honoured to give a talk in the Mark C. Berger seminar series at the University of Kentucky. I spoke about the Understanding Society COVID-19 Survey, with emphasis on our strategy for dealing with nonresponse and attrition, and our estimates of marginal propensities to consume.

03/02/2021, New working paper: Understanding Society Working paper 2021-01, released today, summarises results from a project on "Understanding and improve data linkage consent in surveys".

This is joint work with Annette Jäckle (Essex), Jon Burton (Essex), Mick Couper (Michigan) and Sandra Wazenbach (Konstanz), funded by the Nuffield Foundation and the ESRC. Linking survey and admin data is increasingly important in social sciences and health research, but such linkage often requires consent. Using experimental data from the Understanding Society Innovation Panel and the PopulusLive online access panel, we examined (1) how respondents process requests for data linkage, (2) why web respondents are less likely to give consent than face-to-face respondents, (3) how respondents process multiple requests for consent, and (4) what effects different approaches to wording consent questions have on informed consent.

22/01/2021, New publication: My paper "Interviewer effects and the measurement of financial literacy" appeared in the January 2021 issue of the Journal of the Royal Statistical Society, Series A.

This is joint work with Tobias Schmidt (Bundesbank), Panagiota Tzamourani (Bundesbank) and Joachim Winter (Munich). We study data from Germany's wealth survey, the Panel on Household Finances (PHF). We find that interviewer effects explain a significant fraction of the variance of the financial literacy score, and intra‐interviewer correlations are notably larger for the financial literacy score than for other survey variables. We explore how accounting for interviewer effects can improve estimates of the effects of financial literacy on financial behaviours and outcomes.

18/01/2021 New working paper: "Examining Income Expectations in the College and Early Post-college Periods: New Distributional Tests of Rational Expectations" has just appeared as NBER working paper 28353. ** New Version 07/06/2021.

This is joint work with Yifan Gong (Western), Todd Stinebrickner (Western) and Ralph Stinebrickner (Berea College). We use the innovative Berea Panel Study that Todd and Ralph designed and ran, and we propose new tests of Rational Expectations that use information about the distributions describing beliefs (not just mean beliefs). Beliefs about future income are found to become more accurate as students progress through school and then enter the post-college period. Tests of Rational Expectations almost always reject for the in-school period, but the evidence against Rational Expectations is much weaker in the post-college period.

12/01/2021, New publication: My paper "The heterogeneous and regressive consequences of COVID-19: Evidence from high quality panel data" appeared in the January 2021 issue of the Journal of Public Economics.

This is joint work with Paul Fisher (Essex) and Hamish Low (Oxford). It uses new data from the Understanding Society COVID-19 Survey, which I am helping to run.