Case Studies and Theory Development in the Social Sciences - Alexander L. George [6]
Finally, we want to express particular thanks to Bob Keohane and Gary King for insightful and very useful suggestions on an earlier draft of our manuscript. Their constructive approach to our work has been especially valuable in view of the fact that we express important disagreements with their book, Designing Social Inquiry (co-authored with Sidney Verba, who lacked time to give us comments). Our disagreements have been intellectual, not personal, and they stem from a shared interest in improving research methods of all kinds. One of our central themes is that statistical methods, case studies, and formal models should be regarded as complementary, rather than competitive. Research can progress more effectively through diverse methods than it can through any one method alone.
—Alexander L. George
Stanford, California
—Andrew Bennett
Washington, D.C.
September 2004
Part I
Case Studies and Social Science
Chapter 1
Case Studies and Theory Development
After decades of rapid and contentious change, social science research methods are entering a new phase of development conducive to cross-method collaboration and multi-method work. The changes in these methods over the past four decades have been truly revolutionary. With improvements in computing capabilities, databases, and software, statistical methods and formal models increased rapidly in their sophistication and their prevalence in published research in the 1960s and 1970s. While qualitative and case study methods have also become more sophisticated, the proportion of published research using these methods declined sharply in the 1960s and 1970s, as these methods had already been prominent or even dominant in the social sciences, and their share of the social science market naturally declined as the more novel statistical and formal methods of research grew. To take one example from a leading journal in our own field of political science, between 1965 and 1975, the proportion of articles in the American Political Science Review using statistics rose from 40 percent to over 70 percent; that using formal models rose from zero to over 40 percent; and the proportion using case studies plunged from 70 percent to under 10 percent, with about 20 percent of the articles using more than one method.8 Other social science disciplines, including sociology, history, and economics, have undergone methodological changes as well, each in its own way and at its own pace.
These rapid and far-reaching shifts in research methods in earlier decades were naturally contentious, as they affected opportunities for re-search funds, teaching positions, and publication outlets. Even scholars with similar substantive interests have formed into largely separate communities along methodological lines. To take another example from our field, of two journals that cover similar theoretical issues and policy concerns, The Journal of Conflict Resolution publishes almost no case studies, and International Security publishes almost no statistical or formal work. Such methodological specialization is not in itself counterproductive, as every journal needs to establish its own niche. In this instance, however, there is troubling evidence of a lack of cross-method communication, as each of these journals frequently cites its own articles and very rarely cites those published by the other.9
More recently, however, a variety of developments has made possible an increasingly sophisticated and collaborative discourse on research methods in the social sciences that focuses upon the essential complementarity of alternative methodological approaches.10 Over the past few decades, proponents of case study methods, statistics, and formal modeling have each scaled back their most ambitious goals regarding the kinds of knowledge and theories that they aspire to produce. Practitioners