This survey of the symposium papers argues that the problem of data mining should be of interest to both practicing econometricians and specialists in economic methodology. After summarizing some of the main points to arise in the symposium, it draws on recent work in the philosophy of science to point to parallels between data mining and practices engaged in routinely by experimental scientists. These suggest that data mining might be seen in a more positive light than conventional doubts about it (...) imply. (shrink)

Cartwright (1989) and Humphreys (1989) have suggested theories of probabilistic causation for singular events, which are based on modifications of traditional causal linear modelling. On the basis of her theory, Cartwright offered an allegedly local, and non-factorizable, common-cause model for the EPR experiment. In this paper I consider Cartwright's and Humphrey's theories. I argue that, provided plausible assumptions obtain, local models for EPR in the framework of these theories are committed to Bell inequalities, which are violated by experiment.

Methodology of scientific research programs (MSRP), model-building and actor-network-theory (ANT) are woven together to provide a layered study of the Leontief paradox. Neil De Marchi's Lakatosian account examined the paradox within an Ohlin-Samuelson research program. A model-building approach rather highlights the ability of Leontief's input-output model to mediate between international trade theory and the world by facilitating an empirical application of the Heckscher-Ohlin Theorem. The epistemological implications of this model-building approach provide an alternative explanation of why Samuelson and other prominent (...) economists ignored the paradox. By focusing on the network in which input-output analysis evolved, Bruno Latour's ANT further explains the response of international trade theorists. (shrink)

Davis argues that Suppe's semantic conception provides a better understanding of the problem of theory?data confrontations. Applying his semantic methodology to the LSE (London School of Economics) approach of econometrics, he concludes that the LSE approach fails to address the issue of bridging the theory?data gap. This paper suggests two other versions of the semantic view of theories in the philosophy of science, due to Suppes and van Fraassen, and argues that the LSE approach can be construed under these two (...) versions of the semantic view in terms of structure and representation. (shrink)

While there exist numerous applications of Kuhn's analysis of scientific revolutions to economics, there is yet to be an application to econometrics. The present paper addresses this via an analysis of the often-documented transition between the textbook and LSE methodologies witnessed in British time series econometrics. This exercise allows a number of issues to be raised. First, it will be questioned whether the observed transition in econometrics is an appropriate subject for analysis within the Kuhnian framework. This is the primary (...) concern of the paper. Second, the adoption of Kuhn's approach will allow the witnessed transition in econometric methodology to be viewed from a different perspective to that permitted by more conventional historical accounts. This is seen to lead a number of interesting conclusions. Third, implications of the present study for Kuhn's analysis will be discussed. (shrink)

A positive view of data-mining has been recently presented in a Journal of Economic Methodology ( JEM ) symposium. This is in stark contrast to the stance normally taken. In this note consideration of the Bayesian philosophy of science literature and the impact of data revision extends the analysis of data-mining. Introduction of these issues is seen to provide support for the arguments presented in the JEM symposium.

The philosophy of science literature has played an increasing role in discussion of econometric methodology in recent years, and the Hendry methodology in particular has received much attention. Despite this, the encompassing principle has been overlooked in the methodological literature. This paper addresses this by examining the major methodological implications of the principle.

The reply by Cook and comment by Chao demonstrate Kuhn's thesis that different scientists place different values on different components of their common discipline. This fact is demonstrated by first succinctly summarizing Cook's and my original points within the framework of a simple choice model. I then respond to Cook and Chao. I close by offering some suggestions on how the Textbook/LSE debate could be moved forward.

In a recent article, Cook conducted a Kuhnian analysis of the difference between the Textbook and LSE econometric approaches. This paper uses a semantic conception of theories (Suppe 1989) and a finer gradation of the theory of reduction process to clarify the apparent puzzle that exist between the Textbook and LSE approaches to econometrics. The paper demonstrates that a Kuhnian analysis in isolation can be more misleading than realized.

Trygve Haavelmo's 1944 article ‘The Probability Approach in Econometrics’ is considered by most to have provided the foundations for present day econometrics (Morgan, 1990, Chapters 8 and 9). Since Haavelmo (1944), extraordinary advances have been made in econometrics. However, over the last two decades the efficacy and scientific status of econometrics has become questionable. Not surprisingly, the growing discontent with econometrics has been accompanied by a growing interest in econometric methodology.

This paper argues that econometricians' explicit adoption of identification conditions in structural equation modelling commits them to read the functional form of their equations in a strong, nonmathematical way. This content, which is implicitly attributed to the functional form of structural equations, is part of what makes equation structural. Unfortunately, econometricians are not explicit about the role functional form plays in signifying structural content. In order to remedy this, the second part of this paper presents an interpretation of the functional (...) form based on Herbert Simon's definition of causal order. This begins to set out just what the functional form of structural equations represents. ‡I would like to thank Nancy Cartwright and attendants at UCSD Graduate Seminar 2006 for helpful comments. I also want to thank the AHRC for supporting the research for this paper. †To contact the author, please write to: Centre for Philosophy of Natural and Social Science, London School of Economics, London WC2A 2AE, United Kingdom; e-mail: d.j.fennell@lse.ac.uk. (shrink)

This paper examines some economic forecasts made in late 1930 that were intended to predict economic activity in the United States in order to shed light on several methodological issues. We document that these forecasts were extremely optimistic, predicting that the recession in the US would soon end, and that 1931 would show a recovery. These forecasts displayed egregious errors, because 1931 witnessed the largest negative growth rate for the US economy in any year in the twentieth century. A specific (...) question is what led forecasters to make such serious and substantial empirical errors. A second more general issue involves the methodology of forecasting. The 1930 forecasts were sometimes based on explicit analogies with previous serious business cycles. Modern forecasting approaches are based on techniques that may not be recognized as analogies. Using the 1930 forecasts, we examine the implicit?analogy content of forecasts, and what might render such implicit analogies valid or invalid. This 1930 forecast example also resonates beyond the confines of economic methodology because forecasts about the Great Depression are of continuing interest to the profession at large, and we produce a forecast series not previously available. (shrink)

This paper argues classical statistics and standard econometrics are based on a desire to meet scientific standards for accumulating reliable knowledge. Science requires two inputs, mining of existing data for inspiration and new or 'out-of-sample' data for predictive testing. Avoidance of data-mining is neither possible nor desirable. In economics out-of-sample data is relatively scarce, so the production process should intensively exploit the existing data. But the two inputs should be thought of as complements rather than substitutes. And we neglect the (...) importance of out-of-sample testing in the production of reliable knowledge. Avoidance of data-mining is not a substitute for tests conducted in new samples. The problem is not that data-mining corrupts the process, the problem is our collective neglect of out-of-sample encompassing, stability and forecast tests. So the data-mining issue diverts us from the crucial margin. (shrink)