email icon Email this citation


Case Study Methods in International Political Economy

John S. Odell

International Studies Association
41st Annual Convention
Los Angeles, CA
March 14-18, 2000



IPE scholars use a variety of qualitative methods to contribute to theory building, and we probably could get greater value from our case studies. Single case studies are actually a family of research designs: the disciplined interpretive case study, the hypothesis generating case study, the least-likely, most-likely and deviant case studies. The method of difference employs comparison and attempts to eliminate rival interpretation by choosing cases that match in important respects. Case study methods enjoy five inherent advantages relative to statistical methods, and they suffer from five relative disadvantages. Neither methods family is sufficient; the two are complements and ultimately must be combined.


Qualitative methods are used widely in research on the world political economy, but we scholars probably could generate greater value from many of our case studies. This paper is addressed to the advanced student and any others who wish to review a menu of methods options and learn how case studies contribute to theory-building in international political economy (IPE). I illustrate several single-case methods as well as the comparative method of difference. I conclude with an evaluation of these research designs. Thoughtfully designed case studies can make several types of contribution to the collective research enterprise. They offer advantages as well as disadvantages compared with statistical methods. 1

What counts as a case can be as flexible as the researcher’s definition of the subject. By a case I mean a single instance of an event or phenomenon, such as a decision to devalue a currency, a trade negotiation, or an application of economic sanctions. One could select three cases defined as decisions by three different countries to devalue their currencies. Or three events in the history of a single country could be defined as three cases. Furthermore, within a single case study however defined, multiple observations of theoretically relevant variables normally can be made. Selecting one instance of a phenomenon need not mean making only one theoretically relevant observation.


Single Case Designs

The single case study actually represents a family of research designs. The following types are distinct but not always mutually exclusive. A particular case study can fit more than one category. This section will concentrate on studies conducted at least partly to contribute to theory building.

The disciplined interpretive case study

Many cases are selected for investigation because they are recent or seem intrinsically important. Major events such as wars, the onset of the great depression, the creation of the Bretton Woods institutions, and decisions to change domestic economic institutions have probably sent history down a track different from what would have occurred otherwise. Much of what occurred later has depended on these chosen historical paths. Understanding crucial break points is as important as testing any hypothesis that might be valid between them. But a case study need not be limited to reporting an intuitive understanding of such a break point.

The disciplined interpretive case study interprets or explains an event by applying a known theory to the new terrain. The more explicit and systematic the use of the theoretical concepts, the more powerful the application. Although this method may not test a theory, the case study shows, or at least claims, that one or more known theories can be extended to account for a new event. This type of research will interest defenders as well as critics of the theories, even those who care little about the particular event. This type of case study cannot fairly be called "atheoretical" nor its broader contributions "nil." 2

As Harry Eckstein notes, "Aiming at the disciplined application of theories to cases forces one to state theories more rigorously than might otherwise be done." 3 As a result of this conceptual work, the author may often be able to generate an additional type of contribution: new suggestions for improving the theory.

An example of a disciplined interpretive study concerns the 1971 U.S. decision to suspend the dollar’s convertibility into gold and achieve a dollar depreciation in the currency markets. 4 This study first explicated five general perspectives for explaining changes in any government’s foreign economic policy. All these perspectives were already known at some level, but the study attempted to sharpen and refine them in the course of working with them. It formulated an international market perspective for explaining government actions, as distinct from prescriptions based on pro-market thinking. Another theoretical perspective emphasizing policy makers’ subjective beliefs was synthesized from psychology, security studies, and scattered ideas in economic history. This chapter then used the five refined perspectives to construct an interpretation of the 1971 policy change. It argued that the most important sources of the change were a slight decline in US power relative to other states, a lasting change in international (currency) market conditions, and a striking shift in policy predispositions at the top of the Nixon administration, especially in early 1971. 5

A disciplined interpretive case study can usefully complement formal and statistical research. When a formal model has suggested hypotheses for testing, and even after large-n quantitative tests have provided confirmation, there always remains the question whether the causal mechanism suggested by the theory, or some other conceivable mechanism, was actually responsible for connecting the measured cause with the measured effect variable. A thorough case study can investigate these questions in detail, checking whether events unfolded according to the proposed model in at least one case, while also checking for rival interpretations and omitted considerations. Here too a rigorous case study that confirms the theory leaves it stronger that it was before.

Aggregate data in Lisa Martin’s Coercive Cooperation 6 provide strong support for the hypothesis (among others) that when an international institution calls on states to cooperate in imposing economic sanctions, such cooperation increases dramatically. One chapter (among others) adds a case study of economic sanctions imposed against Argentina during the 1982 conflict with Great Britain over the Falkland/Malvinas Islands. After the UK imposed sanctions, London sought cooperation from allied governments. Martin’s interpretation holds that the institutions of the European Community, and linkages between the sanctions issue and EC budget negotiations, were decisive in achieving and maintaining cooperation from increasingly reluctant partners.

Most events give rise to more than one interpretation. One general risk of this method is the selective reconstruction of the event to support a favored theory, by underplaying evidence inconsistent with the theory or alternative interpretations or both. The check against this risk is a faithful presentation of one or more of the most powerful alternative theories, and disciplined interrogation of the evidence to check each. Counterfactual reasoning might be deployed to support suggestions about which hypotheses are more and less useful in this case. 7 The 1971 monetary study is one example of these techniques.

The hypothesis-generating case study

A case study begun for any purpose can become a hypothesis-generating case study as well. One of the most valuable contributions of any method would be the generation of a new hypothesis that turned out to be valid or fruitful of fresh lines of investigation.

Charles Kindleberger’s The World in Depression, 1929-39 (1973) 8 proved to be influential during the two decades after its publication. Kindleberger’s inquiry into why the great depression was so wide, so deep and so prolonged led him to the collapse of the international monetary system in 1929-1931. The lesson of the interwar experience was "that for the world economy to be stabilized, there has to be a stabilizer, one stabilizer." 9 When in 1929 "the British couldn’t and the United States wouldn’t" supply the public good of leadership, down went the interests of all. 10 This work helped stimulate other scholars to think of the hegemony theory of international economic stability.

The least-likely (theory confirming) case study

Let us shift now to methods that select a single case not for its novelty or intrinsic interest but for its ability to contribute to theory building. It is unlikely that any single case study will be able, alone, to prove or disprove a theory decisively. 11 Probably the closest a single case study can come to approximating a neutral test would be when the researcher selects an extreme case that is highly unlikely to confirm it, and finds that even this case does so. Such a least-likely case study would provide strong support for the inference that the theory is even more likely to be valid in most other cases, where contrary winds do not blow as strongly.

Edward Morse’s thesis in Foreign Policy and Interdependence in Gaullist France 12 is that increasing modernization and interdependence transform foreign policy, making it less nationalistic and more cooperative. Morse observes a breaking down of the distinction between foreign and domestic policy, and directs attention to domestic social structure as the primary determinant of foreign policy. What country case might be least likely to confirm such a thesis? President Charles de Gaulle’s famous nationalism stands out among industrial states of his time. De Gaulle defended the primacy of foreign policy over domestic policies, and during the 1960s, attempted to use foreign economic policy to maintain independence from the monetary system led by the United States. During a run on the French franc in November 1968, he proudly declared that the franc would not be devalued. But the European economy was becoming more closely integrated as governments implemented the Common Market, and domestic student protests of 1968 had accelerated an erosion of the franc’s underpinnings. De Gaulle’s effort proved futile. Shortly after President Pompidou succeeded him in 1969, France devalued the franc and accepted the US scheme to create a new form of money in the International Monetary Fund. If even Gaullist France yields to interdependence, is it not likely that other governments will do the same?

The plausibility probe is a weaker form for a like purpose. The researcher conducts a single case study only to check the plausibility of a theory, using a case that may not be especially difficult for the theory. One might even select a case whose circumstances are thought to be favorable to the theory, as a pilot study before undertaking a more extensive data gathering effort. If this probe does not confirm the theory’s plausibility, resources can be better directed; if it does, a more comprehensive and costly test can be undertaken with greater confidence. If the case chosen is not especially difficult for the theory, however, then it alone will not support as strong a claim.

The most-likely (theory infirming) case study

If a theory is invalid, the most powerful single case to show that would be one that disconfirms it even though conditions seem to make the case unusually favorable for the theory. If the theory failed even in a most-likely case, this evidence would provide strong support for the expectation that it will fail even more clearly in less hospitable circumstances.

I do not know of a published IPE study that exemplifies this logical type perfectly. As an approximation, consider dependency theory with its thesis that dependency of a less developed country on the world capitalist system retards or even reverses its development. The 1959 Cuban revolution was a sharp break with the world capitalist system, and so the case of Cuba before and after 1959 would seem to be likely, if any would, to support a thesis implying that dependency retards development. Evidence that Cuba did not experience improvement in development terms would be more telling against this theory than evidence from most countries that did not make as clear a break. (Cuba experienced some improvements along with many disappointments.) 13 Again, even an extreme historical case will be subject to more than one interpretation. Here the dependentista might attribute disappointments to the fact that the United States worked aggressively to undermine revolutionary Cuba after 1960, or to avoidable errors by the Castro government.

The deviant case study

When a body of theory is fairly well developed and substantial evidence has confirmed it, a detailed study of a deviant case can be illuminating. Considering a case where the main causes were present but the expected effect did not occur may shed light on the theory’s limits, helping to identify conditions that are necessary or favorable for its operation. An anomaly sometimes can suggest new hypotheses that also account for cases previously accounted for. Japan’s attack on Pearl Harbor has been explored as a deviant case for deterrence theory. 14

This paper does not cover all types of single case study. Some historical studies are designed to provide an account and interpretation of an event without any explicit reference to or use of theory or generalizations. The problem-solving case study is designed to help solve a particular problem rather than mainly to contribute to theory. Robert Rothstein’s Global Bargaining (1979) inquired why the North-South commodity negotiations of the 1970s achieved so little and what could be done to improve such negotiations. This form deserves a set of guidelines of its own. 15 Still other case studies are written for teaching purposes and often purged of explicit analysis in order to put the analytical burden on the student.

Reference is sometimes made to a "process tracing case study," but thinking of this as an additional menu option could be confusing. Virtually all case studies entail documenting some dynamic process--a process of decision, policy change, depression, conflict, negotiation, or the spread of norms. Process tracing is better viewed as a technique involved in writing almost any case study.


The Method of Difference

Between single case methods on one end of the spectrum and large-n statistical methods on the other stand comparative case methods. They add the analytical leverage that comes from comparison to the strengths of the case study. J. S. Mill’s method of difference proceeds "by comparing instances in which the phenomenon does occur with instances in other respects similar in which it does not." 16 Some applications begin with a hypothesis linking a cause C with an effect E. Two or more cases are selected to illustrate a difference in C. If the observed cases differ in C and differ as expected in the supposed effect E but are similar in other respects, then by elimination it can be inferred that the reason for the E difference must have been the difference in C.

More often, theory is not used to guide case selection. The researcher is interested primarily in E, chooses two or more cases to illustrate variance in E, such as a success and a failure, and investigates what antecedents could have produced the difference. A common risk of this variant, the retrospective contrast, is that research will uncover a host of differences between the cases that could have explained E, with the result that the study provides weaker support for any one hypothesis.

Even when a study is designed from the beginning to control for some other factors, the method of difference never can deliver air-tight proof for a causal inference in a single social science study. The historical record never provides a set of actual cases that are perfectly matched on all other relevant variables. Between any two actual cases there is always one or more arguably relevant difference D besides the differences in C and E. The researcher then cannot conclusively rule out D as a contributing cause of E, at least not on the basis of this project.

Actual comparative studies vary by degrees in these matters. When less thought is given to rival interpretations at the design stage, and less effort is invested in selecting cases for analytical reasons, the support for the main conclusion is usually weaker. The more thoroughly case selection matches other important variables, the more rigorous and convincing the support for the central hypothesis. Furthermore, a subsequent study could select cases so as to hold D constant, or focus on D as an important cause in its own right.


The Protestant Ethic and the Spirit of Capitalism is an early example of historical comparison with limited attention to controlling for alternative interpretations. Weber observed in 1904 that what he called the peculiar modern rational form of capitalism had developed only in the West, not in India or China (E). He famously contended that the key reason (C) was the presence in the West of Calvinist theology, which, for believers, translated into devotion to work, ascetic personal habits, and accumulating wealth as a moral calling. Weber reported that in Germany, business leaders and skilled workers were overwhelmingly Protestant. He raised but discounted the alternative interpretation that inheritance accounted for Protestant dominance of business. He noted that a smaller proportion of Catholic (than Protestant) college graduates prepared for middle class business careers, favoring more humanistic education. 17 Generally, though, this essay did not deliver what today would be considered thorough controlled comparisons. Weber did not set forth a comprehensive framework identifying alternative causes, nor did he select cases for study so as to match them according to these variables. The essay stimulated a chorus of critics as well as admirers. 18

Great Britain’s first negotiation to join the Common Market (in the early 1960s) failed while the second (in the late 1960s) succeeded (E). In another, more focused retrospective contrast, Robert Lieber conducts thorough investigations of the processes inside Britain and between Britain and the EEC during the two episodes. He concludes that greater domestic politicization of the British process in the second case turned the decision into a matter of national foreign policy and diluted the earlier influence of agricultural groups on London’s negotiating position. In case 1, these groups had elicited pledges from the Macmillan government for conditions to protect agriculture that proved unacceptable to the Six. 19 In case 2, greater politicization diminished pressure group influence. Harold Wilson’s negotiators made their application relatively free of restrictive conditions. 20 By observing two contrasting cases, Lieber was in a much stronger position to support valid conclusions about the relationship between politicization and pressure group influence, which illuminated the difference between international impasse and agreement, than if he had looked only at one. Certain key variables were essentially constant, including the key states and their institutions and many of the issues. This pair of cases did not rule out every other conceivable interpretation. For instance, Britain’s trade was shifting increasingly away from the former empire and toward the continent of Europe. 21

One hypothesis in Negotiating the World Economy 22 holds that gains from a threatening, value-claiming bargaining strategy will diminish as the credibility of the government threat is undermined by domestic divisions. Chapter 6 supports this hypothesis with two contrasting case studies chosen from the second Reagan administration in 1985-1986, in both of which the US negotiator used a threatening strategy. The cases differ as to cause and as to supposed effect. In the first, US constituents expressed significant opposition to carrying out the threat, and the US negotiator gained less abroad. In the second, few US constituents expressed opposition to implementing the threat, and the US negotiator gained more abroad. These cases are also matched with respect to several other variables thought to be relevant to bargaining outcomes. Since they occurred during the same period, there was no difference in the relevant international institutions, US domestic political institutions, or the degree to which US government was divided. The same President was in office and the same negotiator used the same strategy in each. The threatened party was not less powerful in the second case. In fact, the European Community (in the second case) was more powerful overall than Brazil (in the first), yet Washington gained more from the EC than from Brazil. Moreover, in the Brazil case the main causal variable shifted later in the negotiation (US constituents fell into line behind Reagan’s hard line), and almost immediately the Brazilian government made its one substantive concession to Washington. Challenges to the main inference from several standpoints can be eliminated by virtue of matched case selection.

Implementing this method

A method-of-difference study in the theoretically guided mode can be implemented in three stages, which overlap in practice and blend inductive and deductive thinking. 23 Stage 1 is the identification or generation of hypotheses from previous research, deduction, or preliminary study of particular events. In stage 2 the analyst searches for pairs of cases with which to study one or more of these hypotheses. One may restrict the universe to a given set of countries or events and a period of time, and within that population, sample purposively, not randomly. Some cases are studied preliminarily, enough to enable a rough classification of each according to key theoretical variables, judgments that will need to be checked later. When a case that seems high on variable A is found, then the search can turn to a case with a low value on variable A. If a second case can be found, the researcher checks these cases for possible effect variables, and checks what other causes in these cases might have also produced the effect. Stage 2 is also an important time for generating new hypotheses.

At some point the analyst settles on a contrasting pair of cases (or more) and shifts to stage 3--deeper research into these cases and the drawing of conclusions. If additional evidence does not bear out preliminary observations--for example if the two do not contrast as clearly on C or E as they had seemed, or if unexpected rival causes prove salient--this pair might be rejected in favor of others, returning work to stage 2. Shifts back and forth between stages might well help identify the most productive selections.



Case study methods, referring now to single case as well as comparative designs, offer several significant advantages relative to statistical methods. Statistical methods have some inherent limitations. (This comparison refers only to empirical methods and thus does not encompass formal modeling.) First, qualitative studies are equal or superior for generating valid theory. More comprehensive and more detailed contact with concrete instances of the events and behavior about which we wish to generalize helps sharpen distinctions. It stimulates fresh concepts, typologies, and hypotheses. Qualitative work can suggest insights that formalists might develop into innovative models. Other methods are capable of generating theory as well. But the farther we move from direct observation of the people we wish to understand--say by using only proxy indicators that can be measured uniformly over many cases, or by importing models from other fields--the greater the risk of generating theory that turns out to be invalid for this domain. The ultimate goal is valid theory, not just any theory. Many of the insights represented in formal models and statistical tests may well have occurred to those modelers when they were contemplating particular examples. Case methods are also vehicles for refining received theory. A single case study of an event already analyzed can also uncover alternative views that force a rethinking of received knowledge. 24

Second, case studies are generally better than the alternatives for documenting processes. The world political economy is marked by significant processes--market innovation, competition, collusion, equilibration, influence, bargaining, communication, conflict, learning, institutional change, regional integration and disintegration, and politics. Structures are important, but given their stability, they alone are unable to explain much variation that occurs within the same structures. Documenting a process is a valuable contribution by itself, and it is a prerequisite for comparing and generalizing about processes.

Third, case methods allow stronger empirical grounding for a hypothesis for the cases studied. They allow greater confidence in the validity of the hypothesis, for the cases studied, than statistical methods can provide for the same cases. The latter by their nature observe only a handful of variables in each case, and thus many omitted variables remain potential threats to the conclusions’ validity. The case study is not restricted by its nature from investigating all rival interpretations for that case. (Of course every method can be applied in a particular project with less or greater thoroughness.) If a small number of case studies establishes strong empirical support for a hypothesis in those cases, then we have greater confidence that the theory will be found valid in tests over a larger number of cases.

Fourth, case methods preserve and report more information about the cases studied than statistical methods covering the same cases. Fuller reporting makes it more likely that readers will construct alternative interpretations of the same events and generate new hypotheses. Reporting this information also provides researchers with materials that can be used later to construct quantitative indicators to use in statistical tests. Case studies make special contributions to the collective research enterprise.

Fifth, even a case study that claims no explicit theoretical implications typically conveys a much fuller understanding of the case studied, with richer evidence and reasoning about process and context, than is possible with statistical methods. This is especially valuable for events that turned the tide of history, blocked some possible future paths, and selected the one along which later events evolved. Many economic policy decisions, trade and financial negotiations, and applications of economic sanctions probably do divert the subsequent flow of the river, at least locally.

The method of difference offers two additional advantages compared with single case designs. The contrast between cases that are similar in several ways creates an interesting puzzle. And more important, variation in the cause and the effect, plus the elimination of some competing interpretations by case selection, supplies more rigorous support for a causal hypothesis than most single case studies, or multiple case studies that have not been selected to control for competing interpretations. This method provides more convincing empirical grounding for a causal inference than all except perhaps the least-likely and most-likely single case designs.

Case study methods also entail several inherent disadvantages relative to statistical methods. First, the representativeness of the small number of cases is usually not known. A claim that the theory is true in general could not be considered established without having observed other cases. The selected cases could have been exceptional. The least-likely and most-likely designs are efforts to surmount this difficulty with a single case, but even there, wider checking will provide more convincing support.

For this reason and others, second, most case methods are weaker than statistical methods for testing a theory. Claims that a case study has "tested" a theory are sometimes made casually without sufficient warrant. Testing implies a neutral challenge that the idea could fail. When a case or set of cases suggests a new hypothesis, the same cases naturally cannot be regarded as an unbiased test of the idea. A claim that a theory has passed a test is understood by many to mean that the theory is valid generally, in an average case, not just in the cases observed, and such a claim is difficult to defend without a representative sample known to be unbiased. Testing also implies challenging the idea with competing interpretations. A disciplined interpretive case study that fitted a known theory C to a case and thereby helped explain that case could not be regarded as a convincing test if it did not raise alternative interpretations and show why they are inferior.

This, however, suggests a way to make a single case study harder as a test of hypothesis C. The disciplined interpretive case study may introduce alternative hypotheses explicitly, stating each independently of the case’s facts. The analyst may then compare the expectations of each theory with the facts of the case (if the theories’ expectations are precise enough to be checked empirically) and ask whether the case confirms any of them more than any other. A case study that provides such challenges approximates a test more closely than one that does not. 25 Counterfactual reasoning may be added as a means of exploring which factors would have made more difference had they been changed.

Another technique for mitigating this weakness, not yet seen in IPE studies to my knowledge, is Donald Campbell’s multiple implications technique. 26 Campbell suggests improving on the discipline offered by single-site studies by using the key theory to predict other aspects of the case, as many as possible, besides the dependent variable of greatest interest. Ask "if this theory is valid, what else should one expect to see?" By expanding the "implication space," the scholar gains more data points at which the theory could succeed or fail, all within the single case study. Data collection should include keeping a box score of the theory’s hits and misses. A theory should be rejected if it does not pass most of these tests. If not, the scholar might attempt to formulate and "test" a better theory in the same way on the same case.

The least-likely and most-likely case studies and the comparative method of difference are other ways to strengthen the inference that can be drawn with respect to hypothesis C. In the latter, the inference is strengthened in the sense that certain rival interpretations can be ruled out in the cases observed, because of case selection. Even if such a study supports the hypothesis more rigorously than any previous work has done, however, this method is still generally at a disadvantage, relative to the large-n statistical method, in providing support for the more ambitious claim to have tested the theory. 27

Third, qualitative methods inherently permit lesser precision in their descriptions, claims about magnitudes of causal effects, and claims about the relative importance of different causes, than statistical methods. Other things being equal, vagueness is not a plus. This being said, qualitative methods could be deployed with greater precision than is common. For example, case researchers could pin themselves down with operational definitions of key concepts, and construct ordinal scales for measuring their variations qualitatively. Descriptions and analysis in terms of such indicators would still be qualitative but would be more precise than is common, and could contribute to more rapid cumulation.

Fourth, when case study methods are married to informal theorizing, as they normally are, they do not deliver one contribution of the formal model, which can sometimes help the analyst identify logical inconsistencies among diverse informal theoretical insights. Formal modeling is not the only means of uncovering inconsistencies, but its advantage in this respect is normally lost with case study methods.

The method of difference, with cases selected to control some variables, has special limitations. Relying on detailed case studies imposes a limit on the number of cases that can be investigated in any one project, and hence on the number of hypotheses that can be explored rigorously in that project. To provide a more comprehensive theory, with multiple hypotheses and with each grounded empirically in this manner, requires a series of major projects conducted under a common framework. The researcher who accepts the sacrifices required to use statistical methods can often explore more hypotheses rigorously in a single data set. Second, as already noted, perfect matching of cases from real history is impossible, so the method of difference is unable to deliver one hundred percent certainty in causal inference. The analyst is never able to show that no cause other than C could have had any role in producing E.

This complaint is sometimes expanded into a flat rejection of the method of difference, but this would be an exaggeration. The fact that cause D contributed to E does not prove that C did not contribute. Both could be important. More fundamentally, some critics seem to operate from the premise that some other method can provide perfect certainty. J. S. Mill satisfied himself that he had proven that it was "impossible" to apply this method to social phenomena. 28 Mill and many others seemed to assume that perfect certainty is possible, however, with natural phenomena, a view that was discarded by many scientists and philosophers of science during the twentieth century.

George and McKeown 1985 cites Mill and logical positivists Cohen and Nagel (1934) when arguing that the method of difference is beset by formidable difficulties. But what method is not faced with difficulties? Single case studies are limited in domain. Statistical studies inherently face omitted variable bias. Lieberson (1991, 1994) declares flatly that Mill’s methods of agreement and difference "cannot be applied to historical and comparative studies in which the researcher is limited to a small number of cases." 29 Lieberson is right to challenge an author’s claim to have demonstrated, using this method, that C "necessarily" causes E generally, beyond the cases studied. He is right that Mill 1843 did not provide any formal procedure for ascertaining interaction effects. But these are hardly persuasive reasons for banning the method altogether. A problem arises if researchers claim too much for it, but this is true of any method.

In short, case study methods in IPE offer appealing advantages and suffer from significant limitations, relative to statistical methods. The most general implication is familiar but still inescapable, in my opinion. Neither family is sufficient without the other. The two are complements. Aspiring researchers should seek education in both qualitative and quantitative methods. Established analysts should learn from both and resist narrow temptations to discriminate against one family as a means of promoting another. Educators should reconsider any required methods course that is biased against either, or offer an alternative to it.



Bennett, Andrew, and Alexander L. George. 1997. Developing and Using Typological Theories in Case Study Research. Presented at the annual convention of the International Studies Association.

Berejekian, Jeffrey. 1997. The Gains Debate: Framing State Choice. American Political Science Review 91: 789-805.

Campbell, Donald T. 1975. "Degrees of Freedom" and the Case Study. Comparative Political Studies 8, no. 2: 178-94.

Cohen, Benjamin J. 1977. Great Britain. In Economic Foreign Policies of Industrial States, ed. Wilfred F. Kohl, 49-68. Lexington, Massachusetts : D.C. Heath.

Cohen, M. R., and E. Nagel. 1934. An Introduction to Logic and Scientific Method. New York: Harcourt, Brace & Co.

Eckstein, Harry. 1975. Case Study and Theory in Political Science. In Handbook of Political Science, eds. Fred Greenstein and Nelson Polsby, 94-137.

Fearon, James D. 1991. Counterfactuals and Hypothesis Testing in Political Science. World Politics 43: 169-95.

Finnemore, Martha. 1996. National Interests in International Society. Ithaca: Cornell University Press.

Garrett, Geoffrey, and Barry R. Weingast. 1993. Ideas, Interests and Institutions: Constructing the EC's Internal Market. In Ideas and Foreign Policy: Beliefs, Institutions, and Political Change, eds. Judith Goldstein and Robert O. Keohane, 173-206. Ithaca: Cornell University Press.

Geertz, Clifford. 1973. The Interpretation of Cultures. New York: Basic Books.

George, Alexander L., and Timothy J. McKeown. 1985. Case Studies and Theories of Organizational Decision Making. Advances in Information Processing in Organizations, 21-58.

Kapstein, Ethan B. 1989. Resolving the Regulator's Dilemma: International Coordination of Banking Regulation. International Organization 43: 323-47.

Kindleberger, Charles. 1973. The World in Depression, 1929-1939. Berkeley: University of California Press.

King, Gary, Robert O. Keohane, and Sidney Verba. 1994. Designing Social Inquiry: Scientific Inference in Qualitative Research. Princeton: Princeton University Press.

LeoGrande, William. 1979. Cuban Dependency: A Comparison of Pre-Revolutionary and Post-Revolutionary International Economic Relations. Cuban Studies 9: 1-28.

Lieber, Robert J. 1970. British Politics and European Unity: Parties, Elites and Pressure Groups. Berkeley: University of California Press.

Lieberson, Stanley. 1991. Small N's and Big Conclusions: An Examination of the Reasoning in Comparative Studies Based on a Small Number of Cases. Social Forces 70: 307-20.

--------. 1994. More on the Uneasy Case for Using Mill-Type Methods in Small-N Comparative Studies. Social Forces 72: 1225-37.

Lijphart, Arend. 1971. Comparative Politics and the Comparative Method. The American Political Science Review 65: 682-93.

Marshall, Gordon. 1982. In Search of the Spirit of Capitalism: An Essay on Max Weber's Protestant Ethic Thesis.

Martin, Lisa L. 1992. Coercive Cooperation: Explaining Multilateral Economic Sanctions. Princeton, NJ: Princeton University.

Maxwell, Joseph A. 1996. Qualitative Research Design: An Interactive Approach. Applied Social Research Methods Series. Thousand Oaks: Sage Publications.

McKeown, Timothy J. 1999. Case Studies and the Statistical Worldview: Review of King, Keohane, and Verba's Designing Social Inquiry: Scientific Inference in Qualitative Research. International Organization 53: 161-90.

Mill, John Stuart. 1843. System of Logic. Philosophy of Scientific Method, ed. Ernest Nagel. London: Longmans.

–––. 1970. Two Methods of Comparison. In Comparative Perspectives: Theories and Methods, eds. Amitai Etzioni and Fredric L. Dubow, 205-13. Boston: Little, Brown.

Morse, Edward L. 1973. Foreign Policy and Interdependence in Gaullist France. Princeton: Princeton University Press.

Odell, John S. 1982. U.S. International Monetary Policy: Markets, Power, and Ideas as Sources of Change. Princeton: Princeton University Press.

--------. 1993. International Threats and Internal Politics: Brazil, the European Community, and the United States, 1985-1987. In Double-Edged Diplomacy: International Bargaining and Domestic Politics, eds. Peter B. Evans, Harold K. Jacobson, and Robert D. Putnam. Berkeley: University of California Press.

–––. 2000. Negotiating the World Economy. Ithaca: Cornell University Press.

Rothstein, Robert L. 1979. Global Bargaining: UNCTAD and the Quest for a New International Economic Order. Princeton: Princeton University Press.

Russett, Bruce. 1967. Pearl harbor: deterrence theory and decision theory. Journal of Peace Research 2: 81-106.

Samuelsson, Kurt. 1957. Religion and Economic Action. New York: Basic Books.

Spar, Debora L. 1992. Co-developing the FSX Fighter: the domestic calculus of international co-operation. International Journal XLVII, no. 2: 265-92.

Tetlock, Philip E., and Aaron Belkin. 1996. Counterfactual Thought Experiments in World Politics: Logical, Methodological, and Psychological Perspectives. Princeton: Princeton University Press.

Van Evera, Stephen. 1997. Guide to Methods for Students of Political Science. Ithaca: Cornell University Press.

Verba, Sidney. 1967. Some Dilemmas in Comparative Research. World Politics 19, no. 1: 111-27.

Weber, Max. 1958. The Protestant Ethic and the Spirit of Capitalism. New York: Charles Scribner's Sons.



Note 1: I build on classic political science works on this subject, including Lijphart 1971, Eckstein 1975, George and McKeown 1985, and King, Keohane, and Verba 1994.Back.

Note 2: Contrast Lijphart 1971, 692. Back.

Note 3: Eckstein 1975, 103. Verba 1967 calls this method a "disciplined configurative approach." For an alternative meaning of interpretive case studies, see Geertz 1973. Back.

Note 4: Odell 1982, chap. 4. Back.

Note 5: Other examples of this type include Kapstein 1989, Spar 1992, Garrett and Weingast 1993, Finnemore 1996, and Berejekian 1997. Back.

Note 6: Martin 1992 Back.

Note 7: Fearon 1991; Tetlock and Belkin 1996. Back.

Note 8: Kindleberger 1973Back.

Note 9: Ibid., , 305. Back.

Note 10: Ibid., 292. Back.

Note 11: The final section of this paper returns to the question of testing theories with case studies. Back.

Note 12: Morse 1973. This paragraph refers to chapter 5. Back.

Note 13: LeoGrande 1979 finds that Cuba's international economic dependency was lower in the post revolutionary period than before, but that it remained highly dependent in some senses on the USSR. This study does not investigate other indicators of development.Back.

Note 14: Russett 1967. Back.

Note 15: See Maxwell 1996 for some leads.Back.

Note 16: Mill 1970, taken from Mill 1843, book 3, chap. VIII, "of the four methods of experimental inquiry."Back.

Note 17: Weber 1958, 35-40. Back.

Note 18: Samuelsson 1957 marshals evidence contradicting the Puritanism thesis. Marshall 1982 faults Weber for not showing direct evidence on leaders' and workers' motives, as distinct from evidence from Calvinist theological writings, and for failing to raise alternative interpretations for evidence about workers. Back.

Note 19: Lieber 1970, 123, 130Back.

Note 20: Lieber 1970, 271. Back.

Note 21: Cohen 1977. Back.

Note 22: Odell 2000. An earlier version was published as Odell 1993.Back.

Note 23: This section is a revision of Odell 2000, 20-21. Back.

Note 24: McKeown 1999. Back.

Note 25: The "congruence procedure" discussed in the abstract by George and McKeown 1985 is intended to test a single causal theory using "within-case observations" rather than controlled comparison. Van Evera 1999 expands on this idea. Back.

Note 26: Campbell 1975. Back.

Note 27: Harry Eckstein's (1975) effort to defend the argument that single case studies are valuable for testing theories has not, to my knowledge, been followed by many supportive examples in IPE. George and McKeown 1985, Van Evera 1997, chap. 2, and Bennett and George 1997 also argue that theories can be tested with single case studies. Back.

Note 28: Mill 1970, 211.Back.

Note 29: Lieberson 1994, 1225. Back.