One answer to these problems is to set up a natural experiment, that is, a situation in which the sociotechnical arrangements whose effect one wants to study are altered as a result of some event or circumstance “beyond the immediate control of the investigator” (McGinnis 1964). The typical case is one in which a change in government policy or legislation affects some members of a population but not others, creating a variation in the data that is random; or, if not random, at least unconnected or “orthogonal” to any unobservable factors that might be affecting the outcome one is trying to explain (Rosenzweig and Wolpin 2000, 828).

Philip Mirowski & Dieter Plehwe (eds.), The Road from Mont Pèlerin, loc. 1936