Measurement Error Models
Measurement error is important in econometric analysis. Its presence causes inconsistent parameter estimates. Under the classical measurement error assumption, instrumental variable methods can be used to eliminate the bias caused by measurement errors using a second measurement. This technique can be extended to polynomial regression models. For nonlinear models, deconvolution methods have been developed to cope with classical measurement errors. When the classical measurement error assumption is violated, auxiliary data-sets are usually needed to provide additional source of identification. When the true variable takes only discrete values, the mismeasurement problem takes the form of misclassification and requires special techniques.
KeywordsAttenuation bias Auxiliary data Deconvolution method Generalized method of moments Instrumental variables Inverse probability weighting estimation Linear models Measurement error models Misclassification Nonlinear models Permanent-income hypothesis Polynomial regression Semiparametric method Sieve estimator
The author acknowledges generous research support from the NSF (SES-0452143) and the Sloan Foundation.
- Carroll, R., and M. Wand. 1991. Semiparametric estimation in logistic measurement error models. Journal of the Royal Statistical Society 53: 573–585.Google Scholar
- Chen, X., Hong, H. and Tarozzi, A. 2004. Semiparametric efficiency in GMM models nonclassical measurement errors. Working paper, Duke University and New York University.Google Scholar
- Devereux, P. and Tripathi, G. 2005. Combining datasets to overcome selection caused by censoring and truncation in moment bases models. Working paper, University of Connecticut and UCLA.Google Scholar
- Friedman, M. 1957. A theory of the consumption function. Princeton: Princeton University Press.Google Scholar
- Frish, R. 1934. Statistical confluence study. Oslo: University Institute of Economics.Google Scholar
- Imbens, G., Newey, W. and Ridder, G. 2005. Mean-squared-error calculations for average treatment effects. Working paper, Harvard University, MIT and USC.Google Scholar
- Molinari, F. 2005. Partial identification of probability distributions with misclassified data. Working paper, Cornell University.Google Scholar