Truth versus Precision in Economics.

This book is a thoughtful critical contribution to economic methodology. Mayer distinguishes two types of mainstream economic theorizing--formalist and empirical science. Formalist theorizing relies heavily on axiomatic mathematical models. Examples include general equilibrium theory in microeconomics, new classical theory in macroeconomics, and structural parameter analysis in econometrics. Empirical theorizing is the sort of endeavor engaged in by the applied econometrician in constructing equations and formulating hypotheses.

Mayer argues that formalist theorizing has come to dominate the economics profession. The best graduate schools stress the formalist orientation and downgrade all aspects of understanding an economy through data analysis. A "principle of the strongest link" characterizes the argument of a typical professional journal article. The paper focuses on the formal link of the argument's chain, and neglects the other links and assumptions. Getting the formal link right gives the argument precision, but truth suffers because of the neglected links.

Mayer offers a sensible economic explanation of how formalism has come to dominate economics. He illustrates the damage done by the formalist mindset with a series of examples from new classical macro-economics. He cites surveys of economic graduate students to show that they are unhappy about formalist dominance. He deftly marshals philosophers of science to bolster his argument. Unusual among economic methodologists, Mayer understands and respects Karl Popper and Thomas Kuhn, but also sees where each could lead economists astray.

His chapter on how econometricians often abuse Neyman-Pearson hypothesis-testing theory surpasses in clarity anything I've read on the subject. Yet the same chapter implicitly raises a question. It calls for sensitivity analysis of empirical results to omitted variables, variation in lags, alternative functional forms, alternative data sets, and different sample periods. Since sensitivity analysis is inherently incomplete, e.g., there is always another omitted variable to look at, how good is the economic understanding that empirical analysis alone provides?

Hypothesis testing is problematic in many economic situations. For example, to make an economic policy decision often requires knowledge of whether one variable is the cause of another. One can test the hypothesis that the two variables are correlated. With nonexperimental data, though, their correlation is neither necessary nor sufficient for one of the variables to be a cause of the other. With a simultaneous equation model to take account of other causes, all results are asymptotic so one hasn't the sure type I and II error values that Mayer supposes.

Mayer explains excess formalist rigor in economics as follows:

The fields that are often loosely referred to as "the sciences" divide their subject matter not only horizontally by fields, e.g. biology and geology, but also vertically by level of abstraction, e.g. physics and engineering, and physiology and medicine. In doing so they make room for different criteria at each level. An engineering paper need not be as rigorous as a mathematics paper. In the social sciences we divide fields horizontally, but not vertically, so that economics comprises mathematical formalism, empirical science work, and applications to specific practical problems. Hence we are tempted to apply inappropriate standards of rigor.

He mixes insight and confusion. Even if relations in "the sciences" are vertical with respect to level of abstraction, they are more than that. An engineer must take account of physics in his work. Similarly a theoretical physicist must take account of experimental physics in her work. In economics, in contrast, we allow formalist and empirical work to stand apart. We don't oblige each school to take account of the other.

For example, formalist economists assume a worker maximizes a utility function whose arguments are hours of labor and leisure, and deduce certain properties of her labor supply function. Yet one empirical econometric study [1] concludes that a worker has a labor supply function of the form h = [Alpha] + [Beta]w - [Gamma]y, (h, w, y) = (hours-worked, wages, nonwage income) and [Alpha], [Beta], [Gamma] [is greater than] 0. A little algebra-calculus pyrotechnics [2, 333-341, of the sort Mayer disparages, establishes that such a labor supply function is compatible with utility minimization not maximization!

Formalist labor-leisure choice theory and the empirical econometric finding of a linear labor supply function stand apart from each other in their respective literatures. Shouldn't we view the empirical finding as an anomaly that formalist and empirical economists should puzzle out together? If empirical work finds a form compatible with utility maximization that fits better than the linear form, doesn't that reconcile formalist and empirical theories? Otherwise, aren't formalists obliged to find a way other than utility optimization to try to understand worker choices?

I have noted an anomaly in my own empirical econometric work and that of others: supply tend to better fit data than do demand functions. Why? Formalist general equilibrium theory (GET) may offer a clue. On GET, a demand function exists only if an introspectively implausible strong convexity axiom holds; if a more plausible convexity axiom holds, a demand correspondence exists. Empirical theorists could help resolve this anomaly by developing a test of the hypothesis that a function generates data versus the alternative that a correspondence does.

Finally, formalist consumer theory implies that demand functions are homogeneous of degree zero in prices and income. That result stands apart from numerous empirical econometric studies of demand which employ functional forms which are not homogeneous of degree zero in prices and income. Shouldn't empirical theorists have to show such functional forms fit better than ones which are linear homogeneous in prices and income? If they succeed, shouldn't formalist consumer theorists have to modify their theory?

Mayer worries formalist may force empirical economists out of economics departments and into business schools. He concludes:

It is therefore better that we stay together. And we can do so, if we are willing to exercise a civilized sense of tolerance towards each other, and not evaluate the very different work that others do by the yardstick that is appropriate to our own work. Naturally, in our hearts we know that our own work is the more praiseworthy, but in our heads we know that we might, just possibly, be slightly biased.

Again he mixes good sense and confusion. In economics departments, at their worst, would-be pure mathematicians dominate number-crunching measurers. Peace and tolerance would be nice, but not enough. What we need is mutual respect and cooperation. For economics to take a step toward becoming one of "the sciences," the two tendencies must subordinate and harness their fetishisms to the task of understanding the economy.

Leland G. Neuberg Yale University 

References

1. Morrill, R. A., "The Labor Supply Response in the Gary Experiment." Journal of Human Resources 14 (4), 1979, 477-87.

2. Neuberg, L. G. Conceptual Anomalies in Economics and Statistics. New York: Cambridge University Press, 1989.