By Adrian Shin
Jul 06, 2011
U of M Professor Robert Franzese and his colleagues Jude Hays, at the University of Pittsburgh, and Aya Kachi, at the University of Illinois at Urbana-Champaign, won the 2011 Gosnell Prize for Excellence in Political Methodology for their work “Modeling History-Dependent Network Coevolution.”
The paper offers theoretical and empirical models of the ways in which actions and outcomes are correlated across units such as individuals or states. The three causal processes that could generate such correlation of outcomes across units are common exposure, contagion, and selection. This theoretical framework can easily be understood through a thought experiment given by Prof. Franzese, exploring the relationship between smoking and friendship.
“Think about smoking. We observe clusters of individuals who smoke and clusters of individuals who do not. That is, the behaviors of smoking and non-smoking clusters across individuals who are connected to each other. Friends, let’s say.”
For this particular example, “The common exposure explanation for that phenomenon or a potential cause for that phenomenon would be that there are some features out there that on the one hand make folks more likely to be friends, and on the other hand more likely to smoke or not to smoke,” said Prof. Franzese. In other words, each individual is exposed to some external factors that independently influence the individual’s behaviors of smoking and making friends.
Yet, friends often learn from each other. This contagion process which could also explain this clustering of smokers and nonsmokers is an individual acquires the behavior of smoking or abstaining from friends who also smoke or abstain.
The third possibility, selection, addresses the possibility that individuals are more likely to select friends who smoke or who abstain like they do (this is called homophily or homophilic selection) or, in other cases, who behave differently than they do themselves (heterophily). For instance, smokers tend to have more time to talk to each other while smoking together (outside smoke-free establishments, for example). Their common activity (smoking) may contribute to the connection between them, in this case the development of friendship among them, by which contagion may also occur.
In addition to the contribution to academic theoretical and empirical modeling, the paper can be extended to numerous practical applications.
“We observe this phenomenon of correlated behaviors across units everywhere in anything we observe,” beyond individual behaviors, indeed at all levels, remarked Prof. Franzese: the recent sequence of apparently contagious revolts in the Arab Spring, the correlation of policies across countries often attributed to increasingly interdependent world of globalization, etc. What’s difficult to understand fully and to ascertain empirically is whether or to what degrees these observed correlations arise from common exposure, contagion, or selection, which is what Franzese et al.’s framework addresses. And that question is of great practical import. For an anti-smoking campaign to be successful, for example, the magnitudes of common exposure, contagion and selection need to be carefully gauged. One can design an effective anti-smoking campaign only if she has accurate information on the causal processes that induce smoking.
Since 1995, the award is annually given by the Society for Political Methodology to the best work in political methodology presented at any political science conference during the preceding year. The award committee operates under an open-nomination process through which scholars recommend relevant works to the committee. Past recipients include some of the best scholars in the field: Janet Box-Steffensmeier, Gary King, Curtis S. Signorino and U of M Professor Walter Mebane, Jr.
For more information about the award, visit http://polmeth.wustl.edu/about.php?page=award.