by What economics can learn from epidemiology In mainstream economics nowadays there seems to be a broad consensus that one somehow can establish general truths about things by simply generalizing from lots of individual RCTs. But as is well-known among philosophers of science, this kind of inference, based on induction by simple enumeration, is highly flawed. And there are alternatives: The striking feature of paradigmatic epidemiology is that evidence comes from numerous sources. Rather like contemporary climate-change science, the classic episodes of modern epidemiology were not sudden discoveries, but consisted in a slow accumulation of evidence and consensus, via the collation of information from many different sources, some of them nonepidemiological
Topics:
Lars Pålsson Syll considers the following as important: Economics
This could be interesting, too:
Lars Pålsson Syll writes Daniel Waldenströms rappakalja om ojämlikheten
Peter Radford writes AJR, Nobel, and prompt engineering
Lars Pålsson Syll writes MMT explained
Lars Pålsson Syll writes Statens finanser funkar inte som du tror
What economics can learn from epidemiology
In mainstream economics nowadays there seems to be a broad consensus that one somehow can establish general truths about things by simply generalizing from lots of individual RCTs.
But as is well-known among philosophers of science, this kind of inference, based on induction by simple enumeration, is highly flawed. And there are alternatives:
The striking feature of paradigmatic epidemiology is that evidence comes from numerous sources. Rather like contemporary climate-change science, the classic episodes of modern epidemiology were not sudden discoveries, but consisted in a slow accumulation of evidence and consensus, via the collation of information from many different sources, some of them nonepidemiological (e.g. experimental evidence concerning the effects of painting tar onto rats’ ears). Thus the publications that are often cited as having tipped the debate on smoking and lungcancer were a multi-author paper in 1959, running through all the various possible alternative explanations for the association and debunking them (Cornfield et al., 1959), and the Surgeon General’s report in 1964 (Advisory Committee to the Surgeon General of the Public Health Service, 1964). Neither publication presented the results of a single study: both pulled together very multifarious bits of evidence, developing new technical arguments as necessary. Susan Haack (a philosopher) has also noted the importance of triangulation, of the probability of the whole exceeding that of any of its parts, and of multifarious evidence sources in evaluating epidemiological evidence (Haack, 2004, 2008).
In contrast, the PotentiL Outcomes Approach emphasises experimental studies as an ideal, with observational studies being good to the extent that their topics are the subject of imaginary experimental studies; and it emphasises the analysis of the data, and by implication the single study, over the integration of data from different sources and perhaps of different kinds. This is why the claim that causal questions are well-defined when interventions are well-specified is so controversial. It implies that the exemplars of the prevailing epidemiological paradigm are all wrong, and that their causal questions were not well-defined; and it simply leaves out large chunks of (methodological) material that was previously thought to be of paramount importance (e.g. it is hard to see how case-control studies have a proper place in the framework).