On Replications in Business and Economics Research: The QJBE Case

Article excerpt

"It is important that econometric replications be taken as seriously as replications in the experimental sciences."

R.A. Mittelstaedt and T.S. Zorn (1984, p. 14).

INTRODUCTION AND REVIEW

In business and economics research, support often is expressed for the concept of econometric replication. For example, Mittelstaedt and Zorn (1984) argue that a result is not worth knowing if it cannot be reproduced. Despite such a declaration, this type of research apparently has been regarded with ambivalence. Reproducing other researchers' findings may be viewed as less creative than generating new results with original research (Dewald, Thursby, and Anderson, 1986; Kane, 1984).(1)

In their analysis of the market for econometric studies, Feigenbaum and Levy (1993a) argue that there may be data externalities: would-be replicators may not have access to data used in original studies.(2) Dewald et al. (1986) argue that reproducibility is a public good. Readers benefit from knowing whether empirical results can be supported, but it is difficult for a researcher to assess how each reader would value a confirmation. Thus, Dewald et al. and Feigenbaum and Levy have urged journals to adopt policies that promote reproducibility.

Editors have acknowledged the importance of reproducing results. Frisch (1933) wrote in the first volume of Econometrica that its articles also should include raw data. But editorial attention to reproducibility has been a comparatively recent occurrence. It is only in the past decade or so that journals such as American Economic Review and Journal of Money, Credit and Banking have requested data and computer programs from authors so that articles may be duplicated.(3) Since 1993 the St. Louis Federal Reserve Bank has made data from its Review articles available to the public (Anderson and Dewald, 1994).

Some journals have treated replication as more than an auditing device. In response to a suggestion by Feige (1975), the editors of Journal of Political Economy (1975) created a "Confirmations and Contradictions" section in the journal. The editorial board of Industrial & Labor Relations Review (1993, p. 227) has declared that it will "give serious consideration to publishing replication studies of a kind that we have, for the most part, rejected in the past." But the Review has qualified its policy, stating that replications should be brief and that a topic must be deemed to be extraordinarily important.

Although some have done so, most journals do not explicitly encourage econometric replications, and such studies have been rare.(4) Hubbard and Vetter (1991) analyze four prominent finance journals and find that only 5.3 percent of their articles are replications, accounting for 4.3 percent of journal space.(5) For three leading economics journals, Hubbard and Vetter (1992) report nearly identical results.(6) They discuss similar findings of Reid, Soley, and Wimmer (1981) for advertising research and how Brown and Coney (1976) find only 2 percent of marketing publications to be replications.

In the Quarterly Journal of Business and Economics (QJBE) Kane (1984) urges editors to invite econometric replications. In the same issue the board of editors announced a new policy, which departs from other journals: the QJBE would give priority to publishing replication studies. Arguing that replication had been neglected, the editors affirm that the journal would give particular consideration to this type of research.

Since adopting its policy, the QJBE has published several replication studies. Hubbard and Vetter (1992) examine a random sample of 52 empirical articles appearing in the journal between 1984 and 1989. This sample represents about one-third of QJBE articles published during this interval. They find that 17.3 percent of the sample consists of replications, a larger proportion than in other journals.

Given the QJBE's unique policy, Feigenbaum and Levy (1993b) write that analysis of the journal would be particularly intriguing. …