After a burst of anti-Americanism in the Muslim world triggered by a now-infamous Youtube video, there has been considerable interest in assigning causes to the violent protests. New York Times columnist Ross Douthat and Washington Post opinion writer David Ignatius both argue that the protests were the result of intra-elite competition. And perhaps political science should feel reasonably good about the tools it had in its toolkit to think about these situations. Work on elite “out-bidding” and violence goes back to Jack Snyder’s writing in the early 1990s (here, here, and here) if not earlier. Even better, the American Political Science Review published a particularly timely article before this round of protests, in which Lisa Blaydes and Drew Linzer argue, “Muslim anti-Americanism is predominantly a domestic, elite-led phenomenon that intensifies when there is greater competition between Islamist and secular-nationalist political factions within a country.”
This has all the hallmarks of a landmark finding. It is counterintuitive: elite competition (a good thing) might be associated with anti-Americanism (a bad thing) creating a tricky problem for U.S. policy in the Middle East, since encouraging democracy might also generate more elite competition. (Recall that contestation was one of two axes in Dahl’s democratic landscape.) It also uses a reasonably nifty hierarchical model to determine its findings, so it ticks the quantitative political science box. Yay political science? Not so fast. While Blaydes and Linzer should be commended for examining an important contemporary topic in a rigorous way, policymakers should not overlearn from their answer.
The problem with Blaydes and Linzer’s conclusion is that it is derived almost entirely from answers to a Pew Global Attitudes Project cross-national opinion poll. And, we might ask, are the twenty countries (and the Palestinian Authority) sampled by Pew a reasonable representation of the Muslim world? Simply put, Pew is able to operate in places that are relatively freer than the average Arab or Muslim-majority state. Pew results provide little to no information about what is going on for individuals in states without Pew polling. Of the twenty-one countries represented in Blaydes and Linzer’s “Islamic world,” four have Muslim minorities (Ghana, Kenya, India, and Uganda). Of the remaining states, how do the Pew-sampled countries compare to other Muslim-majority states on indices of political rights? Using an average of Freedom House’s political rights and civil liberties scores from 2003-2012, we see that our intuition (political pollsters only operate in freer places) is true. (Using Freedom House scores means I did not include the Palestinian Authority, which is not assessed by that organization.)
The important takeaway is that Blaydes and Linzer do not have data relating to the levels or causes of anti-Americanism in places like Afghanistan, Iran, Iraq, Libya, Saudi Arabia, or Yemen, because Pew’s surveys did not include individuals in those places. This missing data is almost certainly not missing at random. And this is a problem because when people think about the Islamic world they think about places like… well… Afghanistan, Iran, Iraq, Libya, Saudi Arabia, or Yemen. Insights from places like Bangladesh, Kenya, India, Pakistan, or Turkey are not irrelevant, but they are substantially bounded in terms of what they tell us about the Islamic world as a whole. It may well be that the most authoritarian places are both not sampled and also the most anti-American, despite modest levels of intra-elite competition. As with so many contemporary political controversies, the unsatisfying answer is we just don’t know the full relationship between elite competition and anti-Americanism. Extrapolate with care.
Update: For an alternative explanation of anti-American protests, see here.