Some causal lessons from macroeconomics

Causality has long been regarded among economists as a metaphysical mine'eld, best to be avoided. Yet, at the same time, the notion of cause and e*ect structures our ordinary lives and seems essential for understanding policy actions. Peter Adams et al.’s “Healthy, Wealthy, and Wise?” is welcome as an example of the relatively recent interest of microeconomists in overcoming their metaphysical aversions in order to say something useful about cause and e*ect. Their study is carefully executed and aims to resolve a well-posed causal question. The authors’ panel-data approach is closely analogous to the vector-autoregression (VAR) framework of macroeconomists. I want to pursue some of those analogies as they relate to causal inference. It is important to acknowledge that, because of the small number of periods, some of the technical issues in the estimation of panel data are di*erent from typical time-series data. The issues related to casual interpretation, however, do not depend on these di*erences. Adams et al.’s notion of cause can be summarized as time order plus explanatory power plus invariance. I have argued (Hoover, 2001) for a structural account of causation to which each of these three elements is related in a complex way. Macroeconomic examples demonstrate that none of these elements adequately de'nes causality (see Hoover, 2001, especially Chapters 6 and 7). But some of the special features of macroeconomics are less salient in microeconomic contexts, and it is at least plausible that Adams et al.’s criteria are discriminating enough for their application. To clarify ideas, consider a schematic structural model in which health (H) and wealth (W ) are related in time as