Evidence and inference in the comparative case study

Comparative Politics

Volumn 30, Issue 2, 1998, Pages 127-145

  Dion, Douglas ^a  

^a NONE

Author keywords

[No Author keywords available]

Indexed keywords

EID: 0010889630     PISSN: 00104159     EISSN: None     Source Type: Journal    
DOI: 10.2307/422284     Document Type: Article

References (65)

1. See John Stuart Mill, System of Logic, 7th ed. (London: Longman's, Green, Reader, and Dyer, 1868). The Method of Agreement (II.3.8) is essentially selection based on a particular value of the dependent variable. Selecting on the dependent variable extends the logic to selection based on a particular range of the dependent variable.
2. For an excellent discussion of these points, see Christopher H. Achen, The Statistical Analysis of Quasi-Experiments (Berkeley: University of California Press, 1986).
3. Barbara Geddes, "How the Cases You Choose Affect thc Answer You Get: Selection Bias in Comparative Politics," Political Analysis, 2 (1990), 131-50; Gary King, Robert Keohane, and Sidney Verba, Designing Social Inquiry (Princeton: Princeton University Press, 1994).
4. Barbara Geddes, "How the Cases You Choose Affect thc Answer You Get: Selection Bias in Comparative Politics," Political Analysis, 2 (1990), 131-50; Gary King, Robert Keohane, and Sidney Verba, Designing Social Inquiry (Princeton: Princeton University Press, 1994).
5. For a further discussion, see the symposium on King, Keohane, and Verba in the American Political Science Review (June 1995), contributions by Laitin, Caporaso, Collier, Rogowski, Tarrow, King, Keohane, and Verba.
6. Benjamin A. Most and Harvey Starr, "Case Selection, Conceptualization and Basic Logic in the Study of War," American Journal of Political Science, 26 (1982), 834-56. See also Mill, vol. 2, pp. 432, 370-72. Mill's privileging of the method of difference appears to me to come from his recognition that the method of agreement can determine only necessary rather than sufficient conditions. He is thus forced to distinguish "invariable antecedent" from "cause" in the privileging argument on p. 432, whereas causality consisted of nothing but invariable antecedents (including necessary conditions) on pp. 370-72.
7. Benjamin A. Most and Harvey Starr, "Case Selection, Conceptualization and Basic Logic in the Study of War," American Journal of Political Science, 26 (1982), 834-56. See also Mill, vol. 2, pp. 432, 370-72. Mill's privileging of the method of difference appears to me to come from his recognition that the method of agreement can determine only necessary rather than sufficient conditions. He is thus forced to distinguish "invariable antecedent" from "cause" in the privileging argument on p. 432, whereas causality consisted of nothing but invariable antecedents (including necessary conditions) on pp. 370-72.
8. The exceptions are Robert Jervis, "Rational Deterrence: Theory and Evidence," World Politics, 41 (1989), 195; David Collier in American Political Science Review (June 1995), 464; Michael Nicholson, "The Conceptual Bases of the War Trap," Journal of Conflict Resolution, 31 (1987), 346-69; Bruce Bueno de Mesquita, "Conceptualizing War: A Reply," Journal of Conflict Resolution, 31 (1987), 370-82.
9. The exceptions are Robert Jervis, "Rational Deterrence: Theory and Evidence," World Politics, 41 (1989), 195; David Collier in American Political Science Review (June 1995), 464; Michael Nicholson, "The Conceptual Bases of the War Trap," Journal of Conflict Resolution, 31 (1987), 346-69; Bruce Bueno de Mesquita, "Conceptualizing War: A Reply," Journal of Conflict Resolution, 31 (1987), 370-82.
10. The exceptions are Robert Jervis, "Rational Deterrence: Theory and Evidence," World Politics, 41 (1989), 195; David Collier in American Political Science Review (June 1995), 464; Michael Nicholson, "The Conceptual Bases of the War Trap," Journal of Conflict Resolution, 31 (1987), 346-69; Bruce Bueno de Mesquita, "Conceptualizing War: A Reply," Journal of Conflict Resolution, 31 (1987), 370-82.
11. The exceptions are Robert Jervis, "Rational Deterrence: Theory and Evidence," World Politics, 41 (1989), 195; David Collier in American Political Science Review (June 1995), 464; Michael Nicholson, "The Conceptual Bases of the War Trap," Journal of Conflict Resolution, 31 (1987), 346-69; Bruce Bueno de Mesquita, "Conceptualizing War: A Reply," Journal of Conflict Resolution, 31 (1987), 370-82.
12. The example is drawn from Theda Skocpol, States and Social Revolutions (New York: Cambridge University Press, 1979), discussed at greater length below. Though a caricature of her work (ignoring, for example, agrarian structure and the international context), it is useful as an explanatory device.
13. Christopher H. Achen and Duncan Snidal, "Rational Deterrence Theory and Comparative Case Studies," World Politics, 41 (1989), 160, 161.
14. King, Keohane, and Verba, p. 140.
15. King, Keohane, and Verba, pp. 147-49, argue that selecting on the dependent variable can yield suggestive hypotheses, as in the methods used by the Centers for Disease Control. Furthermore, they feel that if the literature provides some variance (more than selection on the dependent variable), a causal effect might be estimated.
16. Geddes, p. 149.
17. I do not claim that the works cited above have mischaracterized the substantive examples, only that there are readings of these examples consistent with a statement of necessity and that, to the extent these examples rely on necessary conditions, the criticisms regarding selection bias are unwarranted.
18. Michael E. Porter, The Competitive Advantage of Nations (New York: The Free Press, 1990).
19. Ibid., p. 22.
20. King, Keohane, and Verba, pp. 133-35.
21. Porter, p. 73.
22. Ibid., pp. 77, 792.
23. Ibid., p. 676.
24. The criticism is made by Geddes, pp. 134-41. Works addressed there include Frederic C. Deyo, ed., The Political Economy of New Asian Industrialism (Ithaca: Cornell University Press, 1987), in particular chapters by Deyo, Johnson, and Koo; Fernando Henrique Cardoso, "Associated-Dependent Development: Theoretical and Practical Implications," in Alfred Stepan, ed., Authoritarian Brazil (New Haven: Yale University Press, 1973), pp. 142-76; Guillermo O'Donnell, Modernization and Bureaucratic-Authoritarianism (Berkeley: Institute of International Studies, 1973). For further book-length treatments, see Frederic C. Deyo, Beneath the Miracle: Labor Subordination in the New Asian Industrialism (Berkeley: University of California Press, 1989); and especially Stephan Haggard, Pathways from the Periphery (Ithaca: Cornell University Press, 1990).
25. The criticism is made by Geddes, pp. 134-41. Works addressed there include Frederic C. Deyo, ed., The Political Economy of New Asian Industrialism (Ithaca: Cornell University Press, 1987), in particular chapters by Deyo, Johnson, and Koo; Fernando Henrique Cardoso, "Associated-Dependent Development: Theoretical and Practical Implications," in Alfred Stepan, ed., Authoritarian Brazil (New Haven: Yale University Press, 1973), pp. 142-76; Guillermo O'Donnell, Modernization and Bureaucratic-Authoritarianism (Berkeley: Institute of International Studies, 1973). For further book-length treatments, see Frederic C. Deyo, Beneath the Miracle: Labor Subordination in the New Asian Industrialism (Berkeley: University of California Press, 1989); and especially Stephan Haggard, Pathways from the Periphery (Ithaca: Cornell University Press, 1990).
26. The criticism is made by Geddes, pp. 134-41. Works addressed there include Frederic C. Deyo, ed., The Political Economy of New Asian Industrialism (Ithaca: Cornell University Press, 1987), in particular chapters by Deyo, Johnson, and Koo; Fernando Henrique Cardoso, "Associated-Dependent Development: Theoretical and Practical Implications," in Alfred Stepan, ed., Authoritarian Brazil (New Haven: Yale University Press, 1973), pp. 142-76; Guillermo O'Donnell, Modernization and Bureaucratic-Authoritarianism (Berkeley: Institute of International Studies, 1973). For further book-length treatments, see Frederic C. Deyo, Beneath the Miracle: Labor Subordination in the New Asian Industrialism (Berkeley: University of California Press, 1989); and especially Stephan Haggard, Pathways from the Periphery (Ithaca: Cornell University Press, 1990).
27. The criticism is made by Geddes, pp. 134-41. Works addressed there include Frederic C. Deyo, ed., The Political Economy of New Asian Industrialism (Ithaca: Cornell University Press, 1987), in particular chapters by Deyo, Johnson, and Koo; Fernando Henrique Cardoso, "Associated-Dependent Development: Theoretical and Practical Implications," in Alfred Stepan, ed., Authoritarian Brazil (New Haven: Yale University Press, 1973), pp. 142-76; Guillermo O'Donnell, Modernization and Bureaucratic-Authoritarianism (Berkeley: Institute of International Studies, 1973). For further book-length treatments, see Frederic C. Deyo, Beneath the Miracle: Labor Subordination in the New Asian Industrialism (Berkeley: University of California Press, 1989); and especially Stephan Haggard, Pathways from the Periphery (Ithaca: Cornell University Press, 1990).
28. The criticism is made by Geddes, pp. 134-41. Works addressed there include Frederic C. Deyo, ed., The Political Economy of New Asian Industrialism (Ithaca: Cornell University Press, 1987), in particular chapters by Deyo, Johnson, and Koo; Fernando Henrique Cardoso, "Associated-Dependent Development: Theoretical and Practical Implications," in Alfred Stepan, ed., Authoritarian Brazil (New Haven: Yale University Press, 1973), pp. 142-76; Guillermo O'Donnell, Modernization and Bureaucratic-Authoritarianism (Berkeley: Institute of International Studies, 1973). For further book-length treatments, see Frederic C. Deyo, Beneath the Miracle: Labor Subordination in the New Asian Industrialism (Berkeley: University of California Press, 1989); and especially Stephan Haggard, Pathways from the Periphery (Ithaca: Cornell University Press, 1990).
29. Deyo, pp. 182-83.
30. O'Donnell, p. 69.
31. Cardoso, p. 147, n5 (italics in original).
32. Geddes, p. 134, does not mischaracterize the argument here; her descriptions rely on language such as "creates a need," "is necessary in order to," and "requires." Her characterization of Chalmers Johnson, however, may support a sufficiency account.
33. See Ibid., pp. 141-45; and King, Keohane, and Verba, p. 129.
34. Skocpol, p. 154 (social revolutionary situations, not social revolutions).
35. Ibid., p. 17, (italics added).
36. Ibid., p. 287.
37. See William Sewell, "Ideologies and Social Revolutions: Reflections on the French Case," Journal of Modern History, 57 (1985), 57-85; Theda Skocpol, "Cultural Idioms and Political Ideologies in the Revolutionary Reconstruction of State Power: A Rejoinder to Sewell," Journal of Modern History, 57 (1985), 86-96. Sewell, p. 57, sets out three conditions as important to Skocpol and notes that when present, "as they were in different ways in France, Russia, and China, the result can be social revolution. . . ." Skocpol, p. 87 responds that she meant to situate intentional group action in the context of social structures, which are "enabling and constraining." I read Geddes pp. 142-43, as arguing that Skocpol's structural features are individually necessary and collectively sufficient for social revolution.
38. See William Sewell, "Ideologies and Social Revolutions: Reflections on the French Case," Journal of Modern History, 57 (1985), 57-85; Theda Skocpol, "Cultural Idioms and Political Ideologies in the Revolutionary Reconstruction of State Power: A Rejoinder to Sewell," Journal of Modern History, 57 (1985), 86-96. Sewell, p. 57, sets out three conditions as important to Skocpol and notes that when present, "as they were in different ways in France, Russia, and China, the result can be social revolution. . . ." Skocpol, p. 87 responds that she meant to situate intentional group action in the context of social structures, which are "enabling and constraining." I read Geddes pp. 142-43, as arguing that Skocpol's structural features are individually necessary and collectively sufficient for social revolution.
39. Geddes, p. 144, does not ignore other factors in Skocpol's account but argues instead that cases investigated in her article "have the structural characteristics she [Skocpol] identifies and so can be used as a set of cases on which to test the hypothesis linking military threat to revolution."
40. Ibid., p. 145. Skocpol, pp. 277-78, discusses the Mexican revolution, noting that like other social revolutions it "became possible only through the administrative-military breakdown of preexisting states," but denies that her account provides a "general theory of revolutions."
41. See Most and Starr, and Achen and Snidal. The latter are approvingly cited by King, Keohane, and Verba, p. 135.
42. Richard Lebow, Between Peace and War (Baltimore: The Johns Hopkins University Press, 1981). The criticism is made by Achen and Snidal, p. 161.
43. Lebow, pp. 84-90. In later work Lebow labels these (with insignificant alterations) "the four necessary conditions of deterrence." Richard Lebow, "Conclusions," in Robert Jervis, Richard Ned Lebow, and Janice Gross Stein, eds., Psychology and Deterrence (Baltimore: The Johns Hopkins University Press, 1985).
44. Skocpol, p. 40.
45. Ibid., p. xiii.
46. Ibid.
47. While the model assumes dichotomous variables, continuous variables are not ruled out in necessary and sufficient conditions. If y = 4x, then x = 3 is a necessary and sufficient condition of y = 12. The confusion about this point is traceable to Herbert Blalock, Causal Inferences in Nonexperimental Research (Chapel Hill: University of North Carolina Press, 1964).
48. Posterior probabilities are rounded to the second decimal place. 39. Giovanni Sartori, "The Influence of Electoral Systems: Faulty Laws or Faulty Method?," B. Grofman and A. Lijphart, eds., Electoral Laws and Their Political Consequences (New York: Agathon Press, 1986), pp. 43-68.
49. Robert William Fogel, Without Consent or Contract: The Rise and Fall of American Slavery (New York: W. W. Norton, 1989), pp. 354-69.
50. See David Owens, Causes and Coincidences (New York: Cambridge University Press, 1992), p. 24.
51. How large must c be before X is probabilistically necessary? The appropriate probability will depend on various factors, including potential losses in believing a condition to be probabilistically necessary when it is not. Questions such as these plague statistical analysis and are not unique to analyses of necessary conditions. As c goes to 1, the probabilistic version of these conditions converges on the deterministic version presented in the fourth section.
52. For alternative hypotheses over 80 percent, greater consecutive observations of the necessary condition reduce the probability that the probabilistic necessary condition of 80 percent is true. This conclusion is not as puzzling as it appears: it simply suggests that alternative hypotheses should not assume a greater probability of the conditions being satisfied than whatever level we attach to our probabilistic necessary condition.
53. Bruce Bueno de Mesquita, The War Trap (New Haven: Yale University Press, 1981). See also the Nicholson-Bueno de Mesquita exchange.
54. One point of common ground would be Mackie's notion of an INUS condition: as insufficient but necessary parts of conditions which are themselves unnecessary but sufficient. J. L. Mackie, "Causes and Conditions" American Philosophical Quarterly, 2 (1965), 245-64.
55. Adam Przeworski and Henry Teune, The Logic of Comparative Social Inquiry (New York: Wiley-Interscience, 1970).
56. The issue is more relevant for subsequent research efforts than for the original statement of the necessary condition. The first researcher finds something out of the ordinary; the second researcher then (innocently) defines the necessary condition so that all the cases pass the test. With this kind of "conceptual stretching" almost any variable can be turned into a necessary condition. But there is also the opposite problem of "conceptual squeezing": setting the standards so high that no variable is capable of being a necessary condition. In short, methodological canons for testing hypotheses are worthless without clear and careful conceptual work. Giovanni Sartori, "Concept Misformation in Comparative Politics," American Political Science Review, 64 (1970), pp. 1033-53.
57. Eric Nordlinger, Conflict Regulation in Divided Societies (Cambridge: Center for International Affairs Occasional Papers in International Affairs No. 29, 1972), p. 199.
58. Michael Hechter, "A Theory of Group Solidarity," in Michael Hechter, ed., The Microfoundations of Macrosociology (Philadelphia: Temple University Press, 1983), pp. 24, 26.
59. Benedict Anderson, Imagined Communities (New York: Verso, 1991), pp. 44-46.
60. Robert Powell, "Crisis Bargaining, Escalation and MAD," American Political Science Review, 81 (1987), 717-36.
61. Adam Przeworski, Capitalism and Social Democracy (New York: Cambridge University Press, 1985), pp.136ff. Also, Adam Przeworski, "Some Problems in the Study of the Transition to Democracy," in Guillermo O'Donnell, Philippe C. Schmitter, and Laurence Whitehead, eds., Transitions from Authoritarian Rule: Comparative Perspectives (Baltimore: The Johns Hopkins University Press, 1986), pp. 47-63.
62. Adam Przeworski, Capitalism and Social Democracy (New York: Cambridge University Press, 1985), pp.136ff. Also, Adam Przeworski, "Some Problems in the Study of the Transition to Democracy," in Guillermo O'Donnell, Philippe C. Schmitter, and Laurence Whitehead, eds., Transitions from Authoritarian Rule: Comparative Perspectives (Baltimore: The Johns Hopkins University Press, 1986), pp. 47-63.
63. See Gary Goertz, Contexts of International Relations (New York: Cambridge University Press, forthcoming). Goertz and Bear Braumoeller are currently working on the large-n analysis of necessary conditions.
64. Readers may worry that the small-n results are driven solely by the use of the Bayesian approach. As a check, I reworked the analysis using a conventional statistical test. The results are exceptionally close to those presented above. To pass a classical one-tailed test at the .05 level, one needs four cases or fewer for null hypotheses of .6 or less, six cases for a null of .7, eleven cases for a null of .8, and twenty-four cases for a null of .9. These ns are, ironically, even smaller than in the Bayesian analysis with uniform priors.
65. A distribution could be truncated because a necessary condition is not satisfied. Since truncated distribution theory is used in selection bias corrections, the avenue is promising in linking the issues taken up in this paper with the selection bias models discussed above.