Working Paper: CEPR ID: DP68
Authors: David F. Hendry; Grayham E. Mizon
Abstract: Recent concern about the inadequacies of econometric models has led not only to the search for causes, but also to the proposal of constructive remedies for these difficulties. In this paper we analyse the links between econometric modelling methodologies and the performance of econometric models. We comment on a number of the constructive modelling strategies that have been proposed, and emphasise the importance of evaluating models, especially by checking whether they are congruent with the available information. This process of model evaluation is destructive rather than constructive, and the criteria it employs form a set of necessary, not sufficient, conditions for model adequacy.
Keywords: econometric modelling; model evaluation
JEL Codes: 211
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| selective use of evidence (C52) | short-term confirmations of theories (C92) |
| positive simple correlations do not establish causal relationship (C10) | skepticism about simple correlations (C10) |
| model evaluation (C52) | identify and discard inadequate models (C52) |
| model evaluation (C52) | isolate the best models for specific purposes (C52) |
| robustness to minor specification changes (L15) | demonstrate model's value (C52) |
| congruency conditions (C62) | model evaluation (C52) |
| critical evaluation of models (C52) | ensure robustness and reliability of econometric claims (C51) |
| extreme bounds analysis (EBA) (C51) | mislead researchers regarding fragility of models (C59) |
| whimsy in model discovery (C52) | role in discovery of new models (C52) |