Paper by Thomas Coleman: “John Snow, the London doctor often considered the father of modern epidemiology, analyzed 1849 and 1854 cholera mortality for a population of nearly half a million in South London. His aim was to convince skeptics and “prove the overwhelming influence which the nature of the water supply exerted over the mortality.” Snow’s analysis was innovative – he is commonly credited with the first application of both randomization as an instrumental variable and differences-in-differences (DiD).
This paper provides an historical review of Snow’s three approaches to analyzing the data: a direct comparison of mixed (quasi-randomized) populations; a primitive form of difference-in-differences; and a comparison of actual versus predicted mortality. Snow’s analysis did not convince his skeptics, and we highlight problems with his analysis. We then turn to a re-analysis of Snow’s evidence, casting his analysis in the modern forms of randomization as IV and DiD. The re-examination supports Snow’s claims that data demonstrated the influence of water supply. As a matter of historical interest this strengthens the argument against Snow’s skeptics. For modern practitioners, the data and analysis provide an example of modern statistical tools (randomization and DiD) and the complementary use of observational and (quasi) experimental data….(More)”