Elephantstrunk

Via Mark Thoma, authoritative  and thoughtful comment on biases likely present in the production and publication of empirical economic research from Edward Glaeser. The fuss being made of the JHU report on deaths in Iraq and the paper on distribution of results of significance tests in published papers in the paper cited by Thoma make it a hot topic.

Still, most of the analysis seems abstracted from economics and could apply to any empirical subject, psychology for instance. It doesn’t describe the relative importance of these effects in economics, the effects of political significance or even the way economics is taught, learned and understood.

1. At the risk of looking like a cut price Stephen Wolfram, results like those in fact-free learning put limits on the possible purity of statistical investigation.
2. If replication is an important antidote to many of the biases identified by Glaeser, the cure may be almost as bad as the disease. The report cited here says that the vast majority of articles published in the AER cannot be replicated even with the original data and code, let alone froma fresh start.
3. This report from the Brookings Institute based on this data suggests that within a country school maths performance is positively correlated with confidence within countries but negatively between countries. This is a possibility described when you are first introduced to panel data so the bias isn’t there. Rather how many economic conclusions are based on single country studies, probably on US data and might show similar effects?
4. Tom Slee begins No One Makes You Shop At Wal-Mart with a simple and plausible description of how people acting individually in their best interests can have a negative effect collectively. In game theory terms it is hardly controversial and is only a little more elaborate than the prisoner’s dilemma which exhibits exactly that. However the subject is mostly introduced through calculus based static equilibrium theory. In that framework the problem is almost impossible to set up. To make it look natural in terms of that theory a dynamic model is required and that isn’t required knowledge even for post-graduate economists. So for many economists tackling Slee’s scenario is like eating soup with chopsticks. Most such models were created as illustrations rather than foundations but gain that status just by hanging around long enough and being easy enough to teach to undergraduates. Would economics look the same if you built it from game theory rather than calculus?

The hard test is how economics stands up to political prejudice and expediency? Economists skeptical of Laffer curves and minimum wages have cause to wonder.