I repeat myself.  But repeat myself I must.  The life of an economics student is easier if he keeps in mind such things as substitution, arbitrage, opportunity cost, and indifference at the margin.  But the dissertation does not write itself.  To write a proper dissertation requires understanding both of price theory and of quantitative methods.  Here's Cafe Hayek's Don Boudreaux on what happens when the mix is poor.  It's superficially about research on the minimum wage, but substantively, it's about what gets people published.
I suggest that the warm reception that Card’s and Krueger’s paper received by many economists reflects the (to me sad) reality that the percentage of economists who have mastered sound price theory has shrunk.  Mastery of technique, as impressive as such mastery is, has both selected in to the economics profession people who are especially good at technique to the exclusion of people who are especially good at economic reasoning, and – because mastering technique is not costless – pushes students of economics to spend more time mastering technique and, hence, to spend less time mastering sound microeconomic reasoning.
There's no necessary conflict. Perhaps the recognition Angus Deaton just received will help.  Without the Almost Ideal Demand System, no improvements in the understanding of consumption behavior in poor countries.  And the Almost Ideal Demand System allows the researcher to check that the laws of conservation in economics, such as the symmetry of compensated substitution effects, and Engel and Cournot aggregation in the price and income elasticities of ordinary demand functions.  As the Committee for the Prize in Economic Sciences puts it (ellipsis means I yanked the notation),
It allows expenditure ... to be non-linear in ܿ[consumption]. It can be used to test for rationality; homogeneity, symmetry and negative semi-definiteness imply restrictions on parameter vectors ... and by estimating [budget share equations] one can check if these restrictions are satisfied in real-world data. The AID system aggregates over consumers under weaker assumptions about preferences than did earlier models.
In other words, the specification of an econometric model starts with the behavioral conditions of an optimization.  I claim to have some familiarity with the approach, whether it's input choices affected by asymmetric regulatory policy, or location affected by transportation costs and consumer demand elasticities, or allocating resources among multiple outputs possibly under regulation.  The optimality conditions can tell the researcher a lot about what econometric pitfalls might arise.  Professor Boudreaux suggests that failure to do so leads to misguided research.
Such people suppose that if they gather and process data and then generate statistically significant results, they are doing sound research.  The conformity of such statistically significant results with some coherent underlying theory is either of no concern (“The data speak for themselves!”) or, more frequently, such conformity is forced in an ad hoc manner – such as when those economists who find empirically that modest hikes in the minimum wage cause no negative employment effects for low-skilled workers explain their finding by asserting that this finding is evidence of monopsony power in the market for low-skilled workers.
But to understand price theory is to be able to ask the right questions about specification, although this will surprise some practitioners.  There was a workshop, not long before I retired, in which I started with the optimization of agents in the data set and ended with a question about endogeneity.  One of our stronger graduate students later asked me how I came up with that, it was not obviously about the econometrics at first, but it was about the econometrics.  Practice, plus understanding of the underlying structure.

No comments: