Unwanted consequences of large sample size in econometric estimation

Abstract

This paper start with the problem of analyzing unwanted consequences of large sample size in econometric estimation and find that the problem can be framed as special case to general problem of estimating a model subject to linear restrictions on the parameters. It is proved that use of large sample size leads to biased, inefficient and inconsistent estimators in the presence of slightest structural change over the observation span. Explanatory power of the model is also shown to fall down. The analysis is extended to provide a general test-statistic that embraces in its ambit almost all the tests known for testing various hypotheses in context to estimation and prediction from linear models.