Unwanted Consequences of Large Sample Size in Econometric Estimation

In this paper we start with the problem of analysing 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. The same test helps in testing hypotheses relating to alternative specifications of variables involved in the model. The results are utilised to suggest a method of segmentation of a population or observation space in relation to a hypothesised econometric model. The idea so developed is helpful in defining samples and populations when data are required to be collected to estimate a relation. The same idea can be used to group a given number of units into structurally homogenous groups.