Causality in Macroeconomics (review)

Two thousand one was a good year for macroeconomics. While influential economists for a long time have called for the "euthanasia" of macroeconomics, Kevin Hoover demonstrates, in his book from that year, that macroeconomics is a separate science, requiring its own methodology, epistemology, and ontology, and is full of life. Being both economist and philosopher, he is able to blend practical concerns and rigorous philosophical analysis.

This book is the "capstone" of his research program over the past twenty years devoted to questions such as "What is the proper analysis of causality in macroeconomics?" (discussed in chapters 2–4) and "How might one infer causal direction and measure causal strength using macroeconomic data?" (chapter 8) and to practical applications of the methods developed to answer these questions (chapters 9 and 10). (Chapters 5–7 consider the particular problems of macroeconomics.) The answers are misleadingly short: The structural approach to causality is the appropriate one for macroeconomics. However, one need not reject alternative probabilistic approaches out of hand, but may use the information provided by these approaches in conjunction with additional information to resolve the issue of causal direction in otherwise observationally equivalent representations. A summary of his main argument is in the chapter "Causality in Macroeconomics" in his Methodology of Empirical Macroeconomics (2001).

Hoover's starting point is the legacy of David Hume. As a philosopher, Hume was a causal pessimist. But as an economist he was a causal optimist, ready to use causal language and causal knowledge to guide policy. And, as Hoover emphasizes, Hume the economist has something to say to Hume the philosopher.

Hume's distinction between the four issues in the analysis of causality—conceptual, ontological, epistemological, and pragmatic—is the guiding principle of Hoover's own analysis. In particular, probability accounts, of which the most prominent in macroeconomics is the Granger causality approach, are criticized for conflating these issues. Hoover summarizes ironically the history of the probabilistic approach as "one of posing counterexamples in which the probabilities violate our causal intuitions and then making adjustments to the probabilistic definition of the causal relation that preserve the insight of the original notion of prima facie cause while rendering it adequate to our intuitions" (17–18). These adjustments appeal to descriptions of the hypothetically true causal mechanism. A probabilistic analysis is judged adequate only when it corresponds to the structure of such mechanisms. The central thesis of Hoover's book is that what is implicit in the strategy of the probabilistic approach ought to be explicitly embraced: Causal structures are fundamental. Probabilistic accounts are useful not for conceptual analysis, but as part of the epistemology of inferring causal structure from observations. Structural accounts suit economics partly because they keep the issues of definition, ontology, and epistemology distinct. Not characterizing cause by an inferential procedure, the structural account accommodates the evidence of conditional probabilities.

A causal structure is a network of counterfactual relations that maps out the underlying mechanisms through which one thing is used to control or manipulate another. This idea of structure originated in the works of the early econometricians whose language was causal. This is evident in the early work of Ragnar Frisch and Jan Tinbergen, which was cast in terms of causes and mechanisms. However, the later discussion and debate in and about the Cowles Commission were also causal. Herbert Simon's "Causal Ordering and Identifiability" is one of the principal sources for the structural account defended by Hoover. This article was the last gasp of causality in the Cowles Commission tradition. As its title indicates, identification and causal ordering were considered as equivalent, and identification was the more pressing problem to econometricians focused on the problem of estimation. Equivalence meant that causality could be ignored without being lost.

It is a pity that Hoover only takes advantage of the Cowles Commission work that appeared in a stage in which an original rich debate on...