An Assessment of Attrition in a Multi-Wave Panel of Households

  • David A. Hensher
Part of the Theory and Decision Library book series (TDLU, volume 49)


Attrition is an issue of great importance in empirical analysis using longitudinal data in which measurements are taken at two or more points in time on the same sample of units. Although attrition per se need not be a problem, any bias due to loss of sample size can have a profound effect on the usefulness of the empirical outputs of the study. For example, if in the current context of predicting automobile energy consumption the households that are lost at each recontact point are typically high-kilometre households then parameter estimates associated with a study in which vehicle use is endogenous could be significantly biased. If, however, there is no difference in the distribution of kilometres between the total sample and the continuing respondents, but there are some differences with respect to exogenous variables (e.g. number of workers, income, household size), it is not necessarily the case that attrition is a source of bias given the objectives of the study.


Panel Data Behavioural Model Attrition Bias Vehicle Possession Participation Probability 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. Duncan, G.J., D.H. Hill, and M. Ponzo: 1984, ‘How representative is the PSID?: A response to some questions raised in the UNICON report’, Survey Research Center, University of Michigan (mimeo).Google Scholar
  2. Hausman, J.A. and D.A. Wise: 1979, ‘Attrition bias in experimental and panel data: the Gary income experiment’, Econometrica 47, 455–473.CrossRefGoogle Scholar
  3. Heckman, J.J.: 1979, ‘Sample selection bias as a specification error’, Econometrica 47, 153–161.CrossRefGoogle Scholar
  4. Hensher, D.A.: 1985, ‘Longitudinal surveys in transport: an assessment’, in E.S. Ampt, E.J. Richardson, and W. Brog (eds.), New Survey Methods in Transport, VNU Science Press, Utrecht.Google Scholar
  5. Hensher, D.A.: 1986a, ‘Dimensions of automobile demand: an overview of an Australian research project’, Environment and Planning A 18, 1339–1374.CrossRefGoogle Scholar
  6. Hensher, D.A.: 1986b, ‘Issues in the pre-analysis of panel data’, Transportation Research A 21, 265–286.Google Scholar
  7. Hensher, D.A. and N. Wrigley: 1986, ‘Statistical modelling of discrete choices in discrete time with panel data’, in Ministry of Transport and Public Works (ed.) Behavioural Researcn for Transport Policy, VNU Science Press, Utrecht.Google Scholar
  8. Johnson, N.L. and S. Kotz: 1970, Distribution in Statistics: Continuous Univariate Distributions - - 1, John Wiley and Sons, New York.Google Scholar
  9. Johnson, N.L. and S. Kotz: 1972, Distribution in Statistics: Continuous Multivariate Distributions, John Wiley and Sons, New York.Google Scholar
  10. Kitamura, R. and P.H.L. Bovy: 1987, ‘Analysis of attrition biases and trip reporting errors for panel data’, Transportation Research A 21, 287–302.Google Scholar
  11. Maddala, G.S.: 1979, ‘Selectivity problems in longitudinal data’, Annales de L’Insee 30–31, 423–450.Google Scholar
  12. Rosenbaum, P.R. and D.E. Rubin: 1983, ‘The central role of the propensity score in observational studies for causal effects’, Biometrika 70, 41–55.CrossRefGoogle Scholar
  13. Winer, R.S.: 1983, ‘Attrition bias in econometric models estimated with panel data’, Journal of Marketing Research 20, 177–186.CrossRefGoogle Scholar

Copyright information

© Kluwer Academic Publishers 1989

Authors and Affiliations

  • David A. Hensher
    • 1
  1. 1.School of Economic and Financial StudiesMacquarie UniversityAustralia

Personalised recommendations