On the relevance of finite sample distribution theory
William E. Taylor · Econometric Reviews · 1983
The implications of finite sample distribution theory for applied econometrics are explored. In general, its relevance is limited by three considerations: (i) sampling errors are of secondary importance in practice, (ii) exact and approximate results depend on the unknown structural parameters of the problem, and (iii) results more accurate than the limiting distribution require further distributional assumptions about some unobservable. Despite these difficulties, some useful information simultaneous equations estimators are sketched.