Content area
Full Text
INTRODUCTION
This paper begins with a historical puzzle. We wonder why econometrics has no code of professional ethics, whereas there are well-developed codes of ethics for statistical communities such as the American Statistical Association (ASA). The historical experience, which we review briefly in the next section, is that professional ethics in statistics grew out of "statistical consulting." To the extent that econometricians engage in similar sorts of economic statistical consulting, they might also benefit from a professional code of ethics that, as a minimum, induces greater transparency.
In order to motivate this argument further, the next section presents a simple model of the trade-off that exists in econometric estimation between bias and efficiency. An econometrician, who is sympathetic with the client's preferences over estimates, may choose selectively to report the estimate. In the subsequent section, we consider an example of non-transparent econometric estimation in some detail. We examine whether such non-transparency may be eliminated by market forces, in particular, by competition among econometric experts and we show that competition fails to substitute for ethical constraints even in what may be the most favorable case for the competitive model. Finally, the last section proposes a mechanism to induce greater transparency in econometric estimation: a bootstrapping rule that makes transparency the minimax solution for the expert.
THE "HUMAN ELEMENT" IN STATISTICAL CONSULTING
The fact is that there are widely discussed codes of ethics in statistical associations, such as the International Statistical Institute [1985] and the American Statistical Association [2000], but nothing like these in econometrics.1 Why is this so? Statistical ethics arose in the process of a trading relationship, an explicit exchange referred to as "statistical consulting," in which the statistician trades expertise with a "client" for some other good thing, be it material income or co-authorship on a research project. Within the exchange setting, an explicit code of statistical ethics attempts to make clear to the client what is and what is not part of the bargain. Hence, the ASA's "Ethical Guidelines" contains sections on the statistician's "responsibilities" to "Funders, Clients, and Employers"; "Research Subjects"; and "Other Statisticians or Statistical Practitioners" [American Statistical Association 2000, pp. 6, 7, 8-9].
For statisticians, "client" is a very general term:
The word client will denote the man...