Acceptance of unsupported claims about reality: a blind spot in economics

Abstract

Do economists accept absurd and unsupported claims about reality, and if so, why? We define four types of claims commonly made in economics that require different types of evidence, and show examples of each from the rational addiction literature. Claims about real world causal mechanisms and welfare effects seem poorly supported. A survey mailed to all researchers with peer-reviewed work on rational addiction theory provides some evidence that criteria for evaluating claims of pure theory and statistical prediction are better understood than those needed for claims of causality or welfare analysis. We suggest that unsupported claims about real world causality or welfare may be accepted in parts of economics provided they derive from a formally correct model consistent with certain types of (often aggregate) data. The rational addiction literature illustrates that this can lead to absurd and unjustified claims being made and accepted in even highly-ranked journals.

Keywords:

• welfare economics
• causality
• rational addiction
• methodology

Acknowledgements

The authors are grateful to the Norwegian Research Council, the Center for the Study of Mind in Nature (CSMN), Health Economic Research Programme (HERO) at the University of Oslo, and the Frisch Centre for funding the research. We would also like to thank everyone who has commented on earlier drafts, survey questions, etc., particularly Uskali Mäki. A special thanks to the referees for extensive comments and suggestions for improvement and clarification.

Notes

1. Usually misquoted as ‘[…] positively harmful at best’.

2. Note that this is a pragmatic classification meant to capture most of the important claims made in typical economics articles. It is not a classification scheme meant to exhaustively class all conceivable claims.

3. Our focus on claims and their justification makes our approach quite different from the related discussion in Yuengert (2006). Yuengert discusses how to select one model among several when ‘models have multiple purposes’ such as prediction, understanding, and policy prescription. He notes the difficulty of determining the trade-offs between the requirements imposed by different goals. This is not a problem in our approach, since there is no need to choose one model. On the contrary, one of our main points is that a model's ability to support one kind of claims should be evaluated separately from its ability to support claims of another kind.

4. As satirized in Rogeberg and Nordberg (2005):

My rational toy person, constructed ad hoc by me in my office based on rough generalizations and bizarre claims about human motivation and reasoning, would receive a higher utility if allowed to maim himself with knives. Ergo, I should not be disturbed by the suicidal behavior of my son.

5. The details of the survey are contained in a companion paper (Melberg 2008).