When studying policy diffusion, establishing causality is even trickier than usual because interdependence violates an important assumption of the Rubin causal model, namely the so-called “stable unit treatment value assumption” (SUTVA), which requires that the outcomes in one unit do not depend on the treatment status of other units.
With observational data, going around this problem is very hard. In an
forthcoming article just published in the American Journal of Political Science, Katerina Linos uses a survey experiment to address this issue. Survey experiments have become increasingly popular in political science but, as far as I am aware, Linos is the first to apply this method to policy diffusion.
The study asked respondents (a representative sample of Americans) whether they agreed or disagreed that “the United States should increase taxes in order to provide mothers of newborn children with paid leave from work.” Respondents were assigned randomly to either this baseline question or to one of four treatments, which consisted of statements that (1) “Canada” or (2) “most Western countries” already have similar policies, or that the policy is recommended by (3) the United Nations or (4) American family policy experts. The main results are shown in this graph (own elaboration of Table 1 in the article):
The effect of all four treatments is large and also remarkably similar. In additional analyses, Linos shows that these effects decrease with the level of information of respondents, which suggests that the experience of other countries and the advice of authoritative people or organizations influence opinions by supplying new information and not, for instance, by setting certain normative standards to which it feels appropriate to conform.
Interesting work, which should encourage other diffusion scholars to use these tools both to replicate these findings and to explore other questions.