WASEDA ONLINE

RSS

The Japan News by The Yomiuri Shimbun

Home > Research > Knowledge Co-Creation - Profiles of researchers

Research

Knowledge Co-Creation - Profiles of researchers

Making Sense of Politics with Data Analysis

Airo Hino
Professor, Faculty of Political Science and Economics, Waseda University

Interest in European democracy

Although it is difficult to describe my area of expertise in a single phrase, let me put that my discipline can be somewhat coined as “comparative electoral studies”. I have stayed in Europe, mainly Belgium, to conduct a comparative study of European elections and party politics, and so a significant aspect of my work derives from an area study. Another aspect, however, is my interest in diverse methods of data analysis to understand political process and political behavior in general. They range from applying models developed in the field of econometrics, adopting an analytical method of Boolean algebra called qualitative comparative analysis (QCA), and to analyzing large-volume of texts using methods of computer-based automated coding. This focus on analytical methods is closely related to my role in the Faculty of Political Science and Economics as I am in charge of teaching research design and basic content analysis to both undergraduate and graduate students who hope to be, among others, journalists, political consultants, and data scientists in private firms and governmental agencies.

I originally became interested in overseas politics as a second-year university student, when I spent an exchange year at Occidental College, Los Angeles where President Obama had also studied. Studying international relations and foreign policy at Occidental College triggered a desire to work at an international organization such as the United Nations. As I knew that becoming an international civil servant requires at least a master’s degree, I thought about going on to graduate school and gaining a degree in international relations and/or political science. I later became interested in how democratization in developing countries has come to be held up as one of the pillars in the American foreign policy, and at that point I encountered various theories of democracy, above all, the European type of consensus-based democracy.

Seminar at Institute of Political and Social Sciences (ICPS) in Barcelona

In Europe, countries such as Belgium, the Netherlands, and Switzerland have maintained democracy despite their mixture of different languages, cultures and religions. The political scientist Arend Lijphart has expressed this state of democracy as “consociational democracy”, by stressing the principles of ‘proportionality’ and the role of elite coordination to build consensus among different groups. I was fascinated by this model first in a normative sense as it suggests that no blood is shed even though people from different backgrounds coexist. I was particularly interested in Belgium, where three linguistic regions of Dutch, French and German-speaking parts of the country are mixed in a rather small area. For my master thesis at Waseda, I analyzed time-series data of public opinion surveys conducted in Belgium over thirty years and found that changing values of the Belgian people are closely related to party support across linguistic borders of the country and are conducive to the emergence of new parties.

With colleagues at the Centre de Politique Comparée (CPC), Université catholique de Louvain

Three types of traditional parties exist in Europe, which are mainly Christian democratic, socialist and liberal parties. Although in opposition to each another, these parties were able to compromise with the coordination by party elites and run a consensus-based democracy. Since the 1970s, however, three new types of parties seem to have come to the fore. Regionalist parties have gained more seats in the early 1970s, environmental parties appeared in the 1980s, and since the 1990s right-wing populist parties have come on the scene, calling for a tougher control of immigrants and accusing of neo-corporatist policy making by the traditional parties. In European countries, the composition of these three emerging forces is broadly similar, but a closer look reveals significant differences as to the rise of these new challenger parties across countries.

For example, opinion polls in both Belgium and the Netherlands reveal an increasing importance of the environment and quality of life in their values, and yet environmental parties are strong in Belgium but not so in the Netherlands in the 1980s. So where does this difference stem from? Isn’t it possible to explain the difference through a number of variables such as the countries’ electoral systems, public funding systems to political parties, free broadcast allowances during election campaigns, and positions of traditional parties on newly emerging issues, etc.? After undertaking research on Belgium for my master’s thesis, I started to investigate the differences of the ways in which new challenger parties rise and fall across countries and moved my base to Europe to conduct a comparative study of 15 countries for my doctoral dissertation.

Cross-sectional and time-series analysis of 15 countries

Having decided to go to Europe, I wondered which graduate school I should go to. Entering a graduate school in Belgium seemed ideal for developing my research in master’s degree and for conducting a fieldwork. But in the end, I enrolled for a Ph.D course at the University of Essex in the UK. The University of Essex was one of the first to adopt an American-style Ph.D program in political science with an emphasis on statistics and data analysis. For a Japanese like myself to produce results on European research that might be worthy of attention, I needed some advantage over others. With that in mind, I took many courses in the summer school of data analysis at Essex.

My Ph.D thesis published by academic publisher Routledge: New Challenger Parties in Western Europe: A Comparative Analysis (Routledge Research in Comparative Politics, 2012)

After studying in the UK for three years from 2001, I moved my base to Belgium. I obtained a grant from the Flemish Government in the Dutch-speaking area of the country and became a member of the Institute of Social and Political Opinion Research (ISPO), University of Leuven. While there, I continued to collect data on new challenger parties in 15 countries. To collect comparable data of electoral and party funding systems from all 15 countries since 1950, I visited different countries and organizations across Europe. Based on this carefully collected data, I constructed a database to analyze the emergence and success of new challenger parties.

Some European researchers told me that 15 countries would be impossible to compare, but I concentrated on my data collection and analysis and completed my Ph.D in 2006. In my Ph.D, I introduced an analytical method called the “double-hurdle model” developed in the field of econometrics, which allowed me to estimate the effect of various factors on the emergence (first hurdle) and success (second hurdle) of new challenger parties separately. The various factors included socio-economic conditions and Political Opportunity Structures (POS) which entailed formal conditions of electoral and party funding systems and informal conditions of policy competitions with traditional parties. I was also able to publish my Ph.D thesis from the academic publisher Routledge.

Figure 1 Framework of comparative analysis of emergence and success of new challenger parties

Forecasting elections from big data

Since coming to Waseda University I have been part of public opinion survey projects through the MEXT Global COE Institute for the “Political Economy of Institutional Construction” and its subsequent project on deliberative polls. Thanks to these large-scale projects, I have also been involved in the research on the values of the Japanese public and their voting behavior. As with Europe, I have tried to closely look at the circumstances in which voters are located at each election. Even in election surveys, for example, it may be possible to explain the Japanese electoral process more rigorously by analyzing the characteristics of each electoral district.

Japan’s first nationwide opinion poll run by a laptop, conducted by the Global Institute for the “Political Economy of Institutional Construction” (CASI survey)

In this so-called big data era, significant progress has been made recently in text analysis by computers. I have been using this method for my research as well. In the joint research with an Italian researcher, for example, we have used computer programs to analyze important records of proceedings in Japan’s National Diet such as policy speeches and party representatives’ questions. By looking at the co-occurrence among words and applying specified algorithms, we extrapolated the position of each party. Contrary to this method of computer coding, big data analysis also includes a method of machine learning, creating teacher data as a basis of analysis by human coding and applying them to a large volume of data. One could put it as a battle between computers and human intellect if you like (laughter).

New Frontiers of Public Opinion Surveys— the CASI Method [Yoronchosa no Atarashii Chihei: CASI-hoshiki Yoronchosa] (co-edited with Professor Aiji Tanaka of the Faculty of Political Science and Economics, Keiso Shobo, 2013), an overview of the new public opinion survey initiatives (headed by Professor Aiji Tanaka)

An interesting example of this face-off between the two methods is a couple of projects involving the predictive analysis of US presidential election results from massive data posted onto SNS (social networking sites). During President Obama’s reelection in 2012, attempts were made using two systems to analyze data from SNS. One was a text analysis project called Twindex, developed by Twitter, using a computer coding approach. The other was a data analysis project, conducted by the Italian researcher I mentioned earlier, combining human coding and machine learning. From the results, the latter project of human coding plus machine learning generated far more accurate predictions. It predicted all but two of the eleven swing states that could have gone to either candidate.

With the proliferation of new technology, data from SNS such as Twitter is rapidly emerging as a new kind of public opinion data alongside the opinion polls. Although SNS users are mostly from the younger generation, their voting tendencies could sufficiently reveal the general trends of election results. We can even analyze data by region since positional data from GPS can be often obtained. Even if SNS data might never replace opinion polls, I see it as a valuable new indicator. Although Japan is still lagging behind from this trend, technically it has enough potential, and we might be able to see real time predictions of voting results in not so distant future.

For more detailed research into the electoral behavior and party politics in Japan, the key challenge is to build a rich data archive. I have it in mind to contribute to enriching data archives by adding data collected in my research so that other researchers are able to use as open data in future.

The Hino Seminar is a lively seminar with 36 members at undergraduate level. Its trademark is a portrait of Professor Hino drawn by a member of his first year.

Interview/Layout: Emiko Tayanagi

Airo Hino
Professor, Faculty of Political Science and Economics, Waseda University

Professor Hino gained an undergraduate degree from the School of Political Science and Economics, Waseda University in 1998, a master’s degree from the Waseda University graduate school in 2000, and a PhD from the Department of Government, University of Essex, UK in 2006. From 2004-2005 he was a member of the Institute of Social and Political Opinion Research, Catholic University of Leuven in Belgium as a Flanders Government sponsored overseas student. From 2006-2007 he was a fellow of the Centre de Politique Comparée, Université catholique de Louvain in Belgium. He served as Associate Professor of the Graduate School of Social Sciences as well as the Division of Politics, Faculty of Urban Liberal Arts, Tokyo Metropolitan University in 2007 and as Associate Professor on the Faculty of Political Science and Economics, Waseda University from 2010. He took up his current post in 2014.

He is the author of New Challenger Parties in Western Europe: A Comparative Analysis (Routledge Research in Comparative Politics) and co-editor / co-author of New Frontiers of Public Opinion Surveys—the CASI Method [Yoronchosa no Atarashii Chihei: CASI-hoshiki Yoronchousa]; Why the Change of Government in 2009: The Shift in Japanese Politics through a Joint Yomiuri-Waseda Survey [2009 nen, Naze Seikenkotai dattanoka: Yomiuri-Waseda no Kyodochosa de Yomitoku Nihonseiji no Tenkan]; Democracies in Europe [Yoroppa no Demokurashi] and others. He has also co-authored numerous books overseas in English, French, and Dutch.