specification problems in econometrics
Edward E. Leamer · The New Palgrave Dictionary of Economics · 2008
The theory of econometrics presumes a ‘specification’ that selects a sharp borderline between (a) assumptions that are maintained and (b) questions that the data are allowed to address. For example, one might treat the list of variables in a regression model as known with certainty but the ‘coefficients’ of these variables as uncertain. In practice, a wide and fuzzy border between the maintained assumptions and the uncertain assumptions causes great ambiguity in the inferences economists draw from their non-experimental data.