Composite estimation is used to improve current estimates of survey variables where the survey design involves sample overlaps between survey rounds. It consists of combining two initial estimates of the same survey variable: one based on the current round data only; the other on updating the past round estimate with the current round data. The paper derives optimal weights for combining the two estimates in the case of simple random sampling of finite and closed populations. It also gives the optimal degree of sample overlap in repeated surveys as function of the correlation between survey rounds. The paper further examines whether the optimal values derived for measuring levels are also optimal for measuring change. Finally, applications are presented in the context of national labour force surveys with rotation sample designs, and global estimation of child labour trend based on samples of countries at two points in time.
Bell, Philip, “Comparison of Alternative Labour Force Survey Estimators,” Survey Methodology, Vol. 27, No. 2, July 2001, pp. 53-63.
Breau, P. and Ernst, L.R., “Comparison of Alternative Labour Force Survey Estimators,” Proceedings of the Section on Survey Research Methods, American Statistical Association, 1983, pp. 397-402.
Diallo, Y., Hagemann, F., Etienne, A;, Gurbuzer, Y., and Mehran, F., Global child labour developments: Measuring trends from 2004 to 2008, International Labour Office (ILO), Geneva, 2010.
Gambino, J., Kennedy, B. and Singh, M.P., “Comparison of Alternative Labour Force Survey Estimators,” Survey Methodology, June 2001, Vol. 27, No. 1, pp. 65-74.
Keywords: Composite estimation; Repeated surveys; Optimal weighting; Optimal sample overlap
Biography: PhD, Department of Statistics, Harvard University, 1973; ISI member, 1977; Vice-President, IAOS (1993-1997); Vice-President, IASS, (1997-2001). Co-author of the ILO manual on Surveys of Economically Active Population, Employment, Unemployment and Underemployment (Geneva, 1990). Former Director of the Bureau of Statistics, International Labour Office, Geneva and Lecturer of Applied Mathematics and Statistics, University of Neuchâtel, Switzerland; Presently advisor at Labour Market Regulatory Authority, Bahrain, and consultant at the Statistics Research and Training Centre, Tehran, Iran.