Putting the 'science' in political science
Meet new Assistant Professor Teppei Yamamoto
Like many other children, Teppei Yamamoto grew up curious about what made things tick. In his case, the questions centered not on clocks and robots, but on people and societies.
“I wanted to know how the world works,” said the recent Princeton PhD who joined the MIT faculty this fall as an assistant professor.
Perhaps because his father was a government employee in Japan, Yamamoto set out in high school to learn about political systems and how states operate. And because he hoped to become a diplomat, he decided to add fluency in English and French to his scholarly toolbox.
But while pursuing a BA in international relations at the University of Tokyo, and MA and PhD degrees in politics at Princeton University, Yamamoto came to suspect that political scientists did not have the methodology needed to develop reliable answers to all the answers they asked.
The study of politics, when you think about it, poses an endless array of questions. What impact is a leader’s assassination likely to have on a country? Why does a higher level of education tend to reflect a higher voter turnout? If Dade County, Fla., hadn’t had an outdated voting system complicated by hanging chads, would Al Gore have become the 42nd U.S. president?
"I was troubled that political scientists too often relied on apparently obvious answers and an intuitive understanding about what people are thinking," Yamamoto recalled. "They had been good at establishing theories but unable to speak from a basis of strong empirical evidence. I wanted to help lead the field away from softer data and toward a political science that we can actually call scientific."
Thus, Yamamoto decided to focus his graduate studies around improving research methodologies, taking a particular interest in causal inference.
Google "causal inference" – along with causal attribution, causal moderation and other related phrases that punctuate Yamamoto's curriculum vitae – and you'll find that the highly complicated study of causation, associated with inductive reasoning, was historically the concern of Aristotle and other theoretical philosophers who posed abstract questions meant to show the elusiveness of pinning down cause. Even lyricist Oscar Hammerstein got into the act in the musical comedy "Cinderella," where one memorable song puckishly asked, "Do I love you because you're beautiful, or are you beautiful because I love you?" More recently, the challenge of sorting out causes from effects and other complicating factors has occupied epidemiologists as they refine statistical methods aimed at finding the sources of disease.
If everyone who ate at Joe's on Tuesday got sick, did eating at Joe's sicken them? Or would they have gotten sick anyway – say, from the local water supply? Such a question might be relatively easy to answer, given the appropriate methodology. But teasing out empirical evidence for something as complex as human political behavior on a wider scale entails sorting through the seemingly limitless other possible causal factors, and doing it by hand is almost beyond human capability. Yamamoto, who had always been deeply interested in advanced mathematics, realized the challenge was one for mathematicians and computer scientists.
Today, the 29-year-old researcher is celebrated as the author of numerous formulas and algorithms designed to help researchers get the hard, quantitative data they need better to understand the past and make evidence-based policy recommendations for the future. His doctoral study on the topic was selected as the year's best contribution to the field by the Society for Political Methodology, which awarded Yamamoto its coveted John T. Williams Dissertation Prize in 2010.
In a dissertation excerpt available online at web.mit.edu/teppei/www/, Yamamoto outlines his complex mathematical strategy for isolating the relevant data researchers want, while also showing how these methods can be used to answer some of political science's classical questions.
One question – related to why some Third World countries successfully became democracies after World War II while others did not – illustrates how difficult it can be to sort through the potentially relevant factors.
"We may theorize," he said, "that democratization happened because of modernization, so we need to measure the modernization level in those countries that democratized and those that did not. But it may have been that increased economic development was the reason, or improved education, or the effects of exposure to people from other countries. What about external political events, like the Cold War?Â We need to control for all these confounding factors and others and compare them in order to get to the true cause of this democratization."
While Yamamoto stands at the cutting-edge of this new research methodology, he does not, of course, stand there alone. He accepted the MIT position, he says, because the Boston medical and scientific community, and the Institute in particular, are fast on their way to making the area the leading study center in the world for the development of advanced statistical methodologies.
"I should be able to learn a lot from my colleagues," he said.