Printer Friendly

Counterfactuals and the study of the American presidency. (Articles).

There is little reason to suspect that [Andrew] Jackson or any of the other candidates contending in 1824 would have done much better. As none had a convincing case for breaking openly with the old order, none had the capacity to untangle the conundrum [John Quincy] Adams wrestled with.

--Stephen Skowronek (1993, 127)

[Faced with deciding whether to seek apolitical or military solution in Vietnam, Lyndon] Johnson opted for military power, first ordering the bombing of North Vietnam and then committing a mounting U.S. ground force to combat in Vietnam. Would a President [Hubert H.] Humphrey have taken a different course of action than Johnson? ... It is probable that he would have.... These examples remind us, it [the U.S. government] is one in which the matter of who occupies the nation's highest office can have profound repercussions.

--Fred I. Greenstein (2000, 2)

These epigraphs provide examples of the use of counterfactual thought experiments in the study of the American presidency. Though these thought experiments are never fully developed by their respective authors, both perform the valuable function of highlighting the (un)importance of the individual who is president. Indeed, for Skowronek (1993), the individual who occupies the presidency is not as important as the context in which that individual assumes office. On the other hand, for Greenstein (2000), it is the individual who makes the difference. There is nothing new about counterfactual inference in social science. Max Weber (1949) long ago argued vehemently for the need to employ counterfactuals to solidify causal explanations. Furthermore, a burgeoning number of political scientists have written extensively on the use of counterfactuals (Elster 1978, 1983; Fearon 1991, 1996; Tetlock 1998, 1999a, 1999b, forthcoming; Lebow 2000b). Tetlock and Belkin (1996a) have also recently produced an anthology dealing with counterfactual analysis in world politics. Tetlock, Lebow, and Parker (forthcoming) have produced a similar volume, employing counterfactuals to examine the role of various causal factors in the making of the Western world. What is new, or rather, what I will argue should be new, is their use in the literature on the presidency to test theories.

Despite their apparent usage in the presidency literature, relatively little has been discussed of the use of counterfactuals. Consider the following passage from Richard Neustadt's (1990) classic text on the presidency:
 Bargaining advantages convey no guarantees. Influence remains a two-way
 street. In the fortunate instance of the Marshall Plan, what Truman needed
 was actually in the hands of men who were prepared to "trade" with him. He
 personally could deliver what they wanted in return. [Secretary of State
 George C.] Marshall, [Senator Arthur H.] Vandenberg, [Secretary of Commerce
 Averell] Harriman, et al., possessed the prestige, energy, associations,
 staffs essential to the legislative effort. Truman himself had a sufficient
 hold on presidential messages and speeches, on budget policy, on high-level
 appointments, and on his own time and temper to carry through all aspects
 of his necessary part. But it takes two to make a bargain. It takes those
 who have prestige to lend it on whatever terms. Suppose that Marshall had
 declined the secretaryship of state in January 1947; Truman might not have
 found a substitute so well equipped to furnish what he needed in the months
 ahead. Or suppose that Vandenberg had fallen victim to a cancer two years
 before he actually did; Senator [Alexander] Wiley of Wisconsin would not
 have seemed to [Senator Robert A.] Taft a man with whom the world need be
 divided. Or suppose that the secretary of the treasury had been possessed
 of stature, force, and charm commensurate with that of his successor in
 Eisenhower's time, the redoubtable George M. Humphrey. And what if Truman
 then had seemed to the Republicans what he turned out to be in 1948, a
 formidable candidate for President. It is unlikely that a single of these
 "supposes" would have changed the final outcome; two or three, however,
 might have altered it entirely. Truman was not guaranteed more power than
 his "powers" just because he had continuing relationships with cabinet
 secretaries and with senior senators. Here, as everywhere, the outcome was
 conditional on who they were and what he was and how each viewed events,
 and on their actual performance in response. (P. 47, emphasis added)

While Neustadt (1990) is employing these counterfactuals largely as rhetorical flourishes to convince the reader of the importance of these individuals at this juncture in history, it is possible to make counterfactuals more systematic. Fearon (1991) argues that investigators who employ counterfactuals to support their causal claims need to be methodologically aware of what they are doing and should make these thought experiments as defensible and explicit as possible. In light of this methodological prescription, this article will focus on the utility of counterfactuals for presidency scholars who employ the presidency as their unit of observation. I begin by discussing Gary King's (1993) claim that such research designs are unlikely to produce reliable causal inferences. Utilizing an emerging systematic framework in the field of international relations to evaluate the "robustness" of counterfactuals, I argue that scholars who employ the presidency as their unit of observation may be able to increase the certainty of their causal inferences by relying on counterfactuals to test their theories. I then provide an application of counterfactual reasoning to demonstrate the utility of this method. I conclude that though there are risks in adopting this approach, its thoughtful and explicit use offers a promising response to King's criticism.

The Presidency As the Unit of Observation and the Problem of Uncertain Inferences

When an investigator aims to assess empirically whether a hypothesized factor was the cause of an particular outcome, he or she may employ one or more of five broadly defined research methods: the experimental, the statistical, the comparative, the case study, or the counterfactual method. All five methods (though less so for the case study method) seek the establishment of general empirical relationships between two or more variables while all other variables are controlled for, that is, held constant. As Lijphart (1971, 683) notes, "these two elements (the establishment of empirical regularities and controlling for the effects of other variables) are inseparable: one cannot be sure that a relationship is a true one unless the influence of other variables is controlled." As is well known, the ceteris paribus assumption is vital to empirical generalizations.

Despite the benefits of employing the experimental method, it is rarely utilized in research on the presidency because of the inability of scholars to randomly assign subjects or manipulate levels of treatment for most topics of relevance. Fearon (1991) argues that when experimental control and replication are not possible, scholars can choose between two different strategies to empirically test their hypotheses: they can search for other actual cases that resemble the case in question (except that in some of these cases, the hypothesized independent variable will be absent or have a different value) or they can imagine that the hypothesized independent variable had been absent and ask whether the outcome would have (or might have) occurred in the counterfactual case. Both strategies attempt to solve the same statistical problem--negative degrees of freedom. (1) Since a researcher cannot make causal inferences on the basis of negative degrees of freedom, to assess a hypothesis, the researcher must add additional cases: either actual cases or a counterfactual case. Within Fearon's broad dichotomous categorization, it is possible then to place the four remaining research methods under consideration (see Table 1).

The relative merits and demerits of each of the methods that employ actual cases to test theories are well known and need not be belabored. (2) My primary concern here is how well each of these methods fares in producing reliable causal inferences when researchers employ the presidency as the unit of observation. King (1993) argues that none of the actual case strategy methods can produce reliable causal inferences when the presidency is the unit of observation. Concerning the case study method, King, Keohane, and Verba (1994) claim that any research design based only on a "single measure on any pertinent variable" (Eckstein 1975, 85) is indeterminate because it fails to control for possible rival explanations. (3) Although a large amount of research on the presidency employs this method, it is insufficient for producing reliable causal inferences. (4)

King (1993) further claims that both the statistical and comparative methods will also fail to produce reliable causal inferences when the presidency is the unit of observation. He notes that it is extremely difficult to employ the method of controlled comparison when focusing on the president as the unit of observation. (5) King asks how are we to use such methods "if we cannot find two presidents who are alike in all respects but our key explanatory variable?" (p. 402). King is hopefully overstating the point in that the investigator need not ensure all other things are literally equal. Rather, the point is that all other things are not systematically related to the hypothesized causal factors and the dependent variable. Nonetheless, it is probably unlikely that one will be able to find other actual presidents who are similar enough to the president under investigation to avoid overdetermination. (6)

Even if it is possible to make the optimistic assumption that researchers are able to match two or more presidents on a sufficient number of variables to avoid problems of overdetermination, King (1993, 404) argues we are still confronted with the problem of not possessing "the required number of observations necessary to make reliable inferences with any degree of certainty." A researcher can make a valid inference in almost any situation, but by relying on the presidency as the unit of observation, one can never be highly certain about one's conclusions because of the small number of presidencies available for observation. (7) Relying on standard statistical methods, King demonstrates that even if one analyzed all forty-three presidencies as cases, one would still not approach the required number of observations necessary to make precise estimates of the effects of each of the hypothesized causal factors. King's argument obviously has important implications for those who seek to construct research designs that rely on the presidency as the unit of observation. For those scholars who seek to employ the actual case strategy, it means that one cannot utilize the statistical or comparative methods to construct reliable causal inferences while retaining the presidency as the unit of observation.

To mitigate this problem, King (1993) suggests that the solution is to search for ways to multiply the number of observations by examining possible observable implications of the same theory or hypothesis. (8) The strategy that King suggests would more than likely increase the number of observations recorded and the researcher's confidence in his or her causal inference. This discussion demonstrates that the task confronting presidency scholars is clear in King's view--investigators should abandon the presidency as the unit of observation and seek new means to multiply the number of observations in their research.

Despite the persuasiveness of King's (1993) critique of research designs that employ the presidency as the unit of observation, his methodological advice has largely fallen on deaf ears. Since his call for presidency scholars to abandon the presidency as the unit of observation, few (if any) students of the presidency have chosen to abide by King's prescriptions. (9) Presidency scholars, despite King, continue to rely on the presidency as their unit of observation-there seems to be a surface plausibility or kind of "focal" attraction to the presidency as an organizing device that scholars are reluctant to abandon. Therefore, the issue is how to deal with this analytical reality and improve the quality of causal inference. The remainder of this article will examine the role that counterfactuals might play in such a task.

The Issue of Constructing Robust Counterfactuals

In King's view (1993, 390), "the signal problem with qualitative research on the presidency is its failure to appropriately judge the uncertainty our inferences." King is rightly concerned about the quality of inference from research designs that employ the presidency as the unit of observation; however, it is somewhat surprising that he ignores the methodological possibilities of counterfactuals. (10) Recent work on counterfactuals suggests that they may provide presidency scholars with a means to increase their number of observations without abandoning the presidency as the unit of observation.

Although the actual and counterfactual case strategies both attempt to solve the problem of negative degrees of freedom, the researcher who employs either strategy encounters different methodological risks. In the former, the investigator runs into the problem of knowing whether the additional cases are appropriately identical. (11) When employing the counterfactual strategy the risk is obvious: how can researchers know what would have happened with any degree of confidence? If presidency scholars are to employ counterfactuals to test their theories, then it is vital to address this problem as it directly relates to King's (1993) contention of the need to improve the certainty of our causal inferences.

Some historians are skeptical about the use of counterfactuals to assess causal claims (Taylor 1954; Carr 1961; Fisher 1970; Thompson 1978). Their primary concern is the great difficulty in constructing what Lebow (2000b, 574) calls a "robust counterfactual--one whose antecedent we can assert with confidence could have led to the hypothesized consequent." Scholars generally point to two interrelated sets of threats to generating and judging counterfactual claims: (1) epistemological and methodological barriers and (2) cognitive and psychological barriers. I first discuss the epistemological and methodological threats to the robustness of counterfactuals and return to the cognitive and psychological barriers later in the article.

Lebow (2000b) identifies three epistemological and methodological threats that warrant the skeptics' view: compound probability, interconnectedness, and second-order counterfactuals. The problem of compound probability refers to the statistical probability of the links between the hypothesized causal factor (the antecedent) and the hypothesized outcome (the consequent). In general, the statistical probability of the consequent decreases as the number of counterfactual steps linking it to the antecedent increases. Thus, an investigator may begin with a minute alternation of the actual world but then infer numerous steps to end up with a major change in reality that has a low statistical probability of occurring.

When scholars engage in counterfactual thought experiments, they sometimes assume that one aspect of the past can be altered and everything else kept constant. (12) This approach is problematic as it fails to address two related problems: causal interconnectedness and the impact of seemingly inconsequential factors. The first problem concerns the inability the researcher to assume "all other things are equal" in systems where various causal factors are interconnected (Jervis 1993, 1996). The latter problem concerns how the researcher selects among the myriad of factors that could have precluded or reduced the likelihood of some outcome had it not taken place--sometimes referred to as "butterfly effects."

The difficulty in asserting a robust counterfactual is further complicated by the problem of second-order counterfactuals. When a researcher asserts that the manipulation of the antecedent led to the consequent, he or she must be wary of the fact that history does not end when the consequent is reached. It is quite possible that subsequent developments in the counterfactual case may return history to the course from which the antecedent was said to divert it (Lebow and Stein 1996). Recognition of this problem, as well as the other two challenges highlighted by Lebow (2000b), is of utmost importance in constructing and asserting a robust counterfactual.

This skepticism of counterfactuals reflects what Tetlock and Belkin (1996b, 37) call the "prestige hierarchy" for methods of drawing causal inferences. At the top of this hierarchy is the experimental method followed by statistical method, and at the bottom is counterfactual method. Since experimental control is quite difficult in social science, investigators often resort to statistical control. However, as King (1993) suggests, the limited number of cases hampers the use of statistical analysis to draw reliable causal inferences when the president is the unit of observation. Does this then mean that we must abandon the president as unit of observation as King suggests? Not necessarily. Tetlock and Belkin (1996b) argue that it is precisely under these conditions-when experimental and statistical control are impossible--that counterfactuals become our best and only, albeit third rate, option. (13) Thus, the core issue is how researchers can judge the robustness of their counterfactuals to test their theories with an adequate level of certainty.

Judging the Robustness of a Counterfactual

In the actual case strategy, the investigator evaluates his or her theory by examining the frequency and magnitude of association between the hypothesized causal factors and the outcome across numerous actual cases. (14) In quantitative research, scholars convey the results of their investigation by reporting estimates of the effect of the hypothesized causal factors and their associated standard errors. While qualitative researchers do not deal explicitly with parameter estimates or standard errors, King, Keohane, and Verba (1994) stress that these scholars should be held to the same rigorous standards of inference and make assessments (though in the form of verbal evaluations) of the degree of certainty of their inferences. When one employs the counterfactual method to test theories, one does not examine frequencies or magnitudes of association across cases since the researcher is seeking to construct the perfect experiment in which the counterfactual case is the control group. Although counterfactuals are not falsifiable, they can be systematic. Through an examination of an emerging systematic framework for evaluating the robustness of counterfactuals, it will be become evident that robust counterfactual cases may be able to provide tests of theories while satisfying King's (1993) suggestion of increasing the certainty of our causal inferences.

Several scholars have noted the role that counterfactuals might play when researchers are faced with the types of research design problems identified by King (1993). (15) While these scholars present researchers with the option of relying on counterfactuals to test theories, they fail to provide any specific roles on how to judge the robustness of these thought experiments. What is needed, then, is a systematic framework for evaluating the robustness of counterfactuals to make them appear less like casual storytelling. What is needed, then, is to make the use of counterfactuals more rigorous.

Renewed attention by political scientists in counterfactuals has led to a corresponding renewed interest in the appropriate criteria for evaluating them (Elster 1978, 1983; Fearon 1991, 1996; Tetlock and Belkin 1996b; Dawes 1996; Kiser and Levi 1996; Lebow and Stein 1996; Lebow 2000b; Tetlock forthcoming; George and Bennett 2001). Before turning to a discussion of these criteria, it is first important to make a distinction between what Fearon (1996) labels "conceivable causes" and "miracle causes" and their relationship to what Lebow (2000b) terms "plausible" counterfactuals and "miracle" counterfactuals. According to Fearon (1996), conceivable causes are those factors that could have actually been different given our knowledge of history. Miracle causes, on the other hand, are those factors that we do not require had to be "objectively possible" to occur. (16) The obvious implication is that conceivable causes are those employed in plausible counterfactuals, while miracle causes are employed in miracle counterfactuals. This distinction, as will be shown, bears greatly on the criteria to judge the robustness of a counterfactual.

The Criteria

Turning our attention to the criteria for judging the robustness of counterfactuals that have been developing in the international relations literature, we find that some standards have been widely accepted by scholars (such as clarity, cotenability, and historical consistency), while others have been subject to some debate (proximity and statistical consistency). A great deal of this controversy concerns the purposes for which counterfactuals should be employed. Elster (1978, 1983), Fearon (1991, 1996), and Dawes (1996) want to utilize counterfactuals to test theories and propositions. On the other hand, Weber (1996), Tetlock (1999b, forthcoming), and Lebow (2000b) want to use counterfactuals to combat psychological and cognitive threats to valid inference. (17) Since our primary concern is to generate counterfactuals to test theories when the presidency is the unit of observation, I will follow the guidelines suggested by those who seek to use them for that purpose.

While universal consent on these criteria is far off, a sustained conversation within the social sciences on the rules for asserting a robust counterfactual will undoubtedly strengthen the method in the future. To appreciate this possibility, one only need examine the sustained conversation that took place the past fifty years within political science (and the social sciences and to some extent the natural sciences) on improving the four actual case strategy research methods. (18) I accordingly enumerate with brief discussion the seven criteria proposed by various scholars for judging the robustness of counterfactuals.

1. Clarity

All causal claims should identify as clearly as possible the explanandum (the consequent in counterfactual cases), the explicans (the antecedent), and the principle connecting the two (Tetlock and Belkin 1996b; Lebow 2000b). This ensures that researchers manipulate only one cause at time, thereby highlighting the causal relationship.

2. Consistency

Though not a consensus, scholars have recognized the need for counterfactuals to be consistent with historical evidence, statistical findings, and accepted theoretical generalizations.

Historical consistency. Counterfactuals should also be consistent with well-established historical facts (Weber 1949; Fearon 1991; Tetlock and Belkin 1996b; Lebow 2000b). When constructing counterfactuals, researchers should follow the "minimal-rewrite-of-history" rule to eliminate the proliferation of antecedents that border on the absurd. King, Keohane, and Verba (1994, 78) echo this view, claiming it should have been possible for the counterfactual case to occur. Though this criterion is a useful starting point, it fails to suggest any guidelines as to what constitutes a minimal rewrite of history. In his discussion of counterfactuals concerning the rise of the West, Tetlock (forthcoming) provides a review of four selection rules that scholars have used to judge what constitutes a minimal rewrite:

(1.) when something else nearly happened that could have led to an alternative world (the close-call counterfactual),

(2.) when an individual or event deviates from the norm (the norm-restoring counterfactual),

(3.) the final point at which history could have taken an alternative trajectory (the last-chance counterfactual), and

(4.) when the antecedent is one of the alternative courses of action that the actors actually considered (the reasonableness principle).

These supplemental criteria should prove useful for presidency scholars when defending their selection of the antecedent.

Though the minimal rewrite criterion is vital for the robustness of plausible counterfactuals, it is irrelevant for evaluating the robustness of miracle counterfactuals. When explanations involve miracle causes, it is not exactly clear what guidelines researchers should follow. Fearon (1996, 64) presents two general thoughts that could provide guidelines as to what miracle causes can and cannot be manipulated. First, a counterfactual antecedent should not be introduced that has not been realized in any other actual case. And second, when one inserts a miracle cause into the counterfactual world, one should do so by introducing "as few changes as one can in the actual world" (ibid., 65). While the first prescription is simple to abide by, the second is bound to create substantial controversy among scholars as to what constitutes a "few" changes in the actual world. (19)

Not surprisingly, Fearon (1996) offers no clear metric to adequately judge the similarity of the counterfactual world to the actual world. Moreover, while his suggestion is certainly valid, political scientists currently lack the ability and the knowledge to judge how close each possible world is to the actual world. My suggestion is to deal with the problem on a case-by-case basis with an informed debate as to which of the possible worlds imagined is actually closest to the actual world.

Statistical consistency. Tetlock and Belkin (1996b) also suggest that researchers should employ well-established statistical generalizations to help fill in the gaps of what would have happened had the independent variable taken on a different value. Along these lines, King and Zeng (2001) recently introduced some new statistical approaches for assessing the degree to which a counterfactual is based on factual empirical evidence or is dependent on some statistical model. Dawes (1996, 304) takes a harsher view, claiming that counterfactual inferences are warranted "if and only if they are embedded in a system of statistical contingency for which we have reasonable evidence." We must be careful, however, not to base the robustness of counterfactuals solely on this criterion. The problems inherent in any meaningful statistical inference should moderate our reliance on this criterion. (20) Primarily for this reason, scholars have been hesitant to adopt this standard as a steadfast criterion for judging the robustness of counterfactuals. Statistical generalizations are best left to constraining our use of counterfactuals that we may be psychologically, ideologically, politically, or cognitively predisposed to constructing--an important constraint as we shall later see.

Theoretical consistency. A number of scholars also propose that the connecting principle employed to link the antecedent and consequent should be consistent with well-established theoretical principles (Elster 1978; Fearon 1991; Tetlock and Belkin 1996b; Kiser and Levi 1996; George and Bennett 2001). In this view, counterfactuals are made credible by invoking theories or regularities that are separate from the hypothesis being tested. (21) To determine what would have happened, a researcher needs a set of theories that have been made independently credible to deduce what would have occurred (Fogel 1964, 224). Counterfactuals are thus empirically supported by drawing on theories that have been supported by empirical evidence drawn from actual case comparisons. This criterion has been criticized by Breslauer (1996), who claims that such a standard is unrealistic for counterfactuals in history and the social sciences, where there is a lack of any established theoretical laws or generalizations. Nonetheless, this standard should be included, as it provides a more explicit perspective from which to assess the robustness of the counterfactual (Lebow 2000b).

(3.) Cotenability (Logical Consistency)

A second attribute that scholars suggest a robust counterfactual should possess is what philosopher Nelson Goodman (1983) calls "cotenability" (Elster 1978; Fearon 1991; Tetlock and Belkin 1996b; Lebow 2000b). Counterfactuals consist of connecting principles that link the antecedent to the consequent. The antecedent should not undercut any of the principles linking it to the consequent. (22)

(4.) Enabling Counterfactuals Should Not Undercut the Antecedent

While cotenability requires the researcher to ensure the antecedent is consistent with the connecting principle, other scholars have pointed to need of researchers to be careful that the enabling counterfactual (the counterfactual that allows for the antecedent) be tenable with the antecedent (Lebow and Stein 1996; Lebow 2000b). For instance, the counterfactual that if Nixon had been president during the Cuban Missile Crisis (antecedent), he would have ordered a surgical air strike against Soviet missiles requires that Nixon had won the 1960 presidential election (the enabling counterfactual). However, if Nixon had been elected in 1960, the world would have been different in many ways, some of them with unknowable implications. These changes might have had significant implications for U.S. and Soviet foreign policy, opening up all sorts of possibilities. Thus, researchers should be careful to specify all enabling counterfactuals and consider their implications for the consequent.

5. Projectability

Tetlock and Belkin (1996b) claim that the greatest standard that counterfactuals must be held accountable to is what Goodman (1983) calls "projectability." Since counterfactuals rely on theoretical generalizations that link the antecedent and consequent, the factor that determines if these theories are robust enough to support counterfactual inference is whether they can predict what will occur in new cases. Tetlock and Belkin's (1996b) suggestion for projectability parallels King, Keohane, and Verba's (1994) prescription that researchers identify the observable implications of their theories. A counterfactual inference is then ultimately deemed robust only if the theory utilized to support it can tell us something about the yet unobserved future.

6. Proximity

Skepticism over our ability to assert robust counterfactuals given the interconnectedness of causes and outcomes leads Fearon (1996) to suggest that we consider only those counterfactuals in which the antecedent and consequent are close together in time and separated by a small number of causal steps that are well understood. This suggestion is echoed by George and Bennett (2001, 23) who argue, "short-term causation is generally easier to address with a counterfactual than causation that involves a longer-term process." In doing so, one may be able to confidently assert that the manipulation of the antecedent is likely to bring about the consequent and nothing else.

Fearon's (1996) proximity criterion has been criticized for ruling out many of the important uses of counterfactuals (Lebow 2000b) and being unworkable (Weber 1996). While this criterion does rule out a number of the uses of counterfactuals, particularly those suggested by Weber, Tetlock, and Lebow, it is necessary to include it for purposes of our objective of theory testing. (23) If the researcher is unable to demonstrate that the consequent results from the manipulation of the antecedent, then counterfactuals are less useful for this purpose. As Lebow (2000b) notes, the criteria for judging the robustness of counterfactuals need to be commensurate with the purposes for which they are employed. Thus, the proximity criterion is included in our list of criteria.

Weber (1996) presents a different critique, contending that the proximity criterion creates an incentive to manipulate only those variables that relate to existing theory and are tied to the causal pathways that we feel we know well. Moreover, he argues, if we knew the causal pathway so well, why do we need counterfactuals? Although Weber is correct to point out this contradiction in Fearon's (1996) argument, his suggestion that we would not need counterfactuals is misplaced. Though we rarely fully understand the causal mechanisms generating the outcomes we observe, our theories do generate strong expectations about how the causal mechanisms should work. Therefore, it is entirely possible to have these strong expectations but still be unable to adequately test the theory because of a limited number of cases. If this is the case, counterfactuals are indeed warranted, and we could satisfy the proximity criterion by relying on these strong expectations. As for the issue of rewarding the manipulation of antecedents that are close to existing theories, this is simply the opportunity cost of seeking to rely on counterfactuals for evaluating extant theory rather than building new theory. Although adopting the proximity criterion limits the functions counterfactuals may perform, it is an acceptable cost for our purposes of theory testing.

7. Further Considerations

To minimize some of the problems associated with the epistemological and methodological threats identified by Lebow (2000b), researchers should address some further considerations highlighted by others.

Avoid the conjunction fallacy. The discussion of the problem of compound probability indicated that the probability of any multistep counterfactual is exceedingly low. The conjunction fallacy is simply the researcher failing to recognize how quickly the compound probabilities of each link approach zero. This implies that it is very unlikely that any hypothesized antecedent will produce specific consequences at any temporal distance. Thus, manipulation of the hypothesized antecedent might change the world, but it will do so in ways that become more difficult to track over time, as many alternative worlds are possible. As the probabilities associated with each of these outcomes will vary dramatically, Tetlock (1999a) argues that researchers need to precisely state if their counterfactuals are intended to produce a specific world, a set of worlds with particular characteristics, or any world other than the one that actually came to pass. With the counterfactual method, the researcher is usually not attempting to predict a specific world but merely a world that does not have the outcome of interest, and this may be a very large set of possible worlds. Thus, it is the probability of that set that matters, not that of any specific outcome.

Recognize the interconnectedness of causes and outcomes. The earlier discussion of the problem of interconnectedness raised two important issues: (1) how to deal with ripple effects throughout a social system if the researcher alters the value of a particular factor and (2) how to handle butterfly effects. Concerning the first issue, George and Bennett (2001) take a stringent approach, claiming that the antecedent must be autonomous from other possible causal factors to avoid the possibility of interconnectedness undermining one's counterfactual inference. Though George and Bennett's suggestion would likely eliminate the problems associated with interconnectedness, it would also severely limit our ability to rely on counterfactuals since very few causal factors are likely to be completely autonomous from others. Taking a more flexible approach, Tetlock and Belkin (1996b) and Lebow (2000b) claim that valid counterfactuals must estimate the degree of interconnectedness and that counterfactuals must be constructed with this important prescription in mind.

The more vexing problem for those constructing counterfactuals concerns the causal impact of seemingly inconsequential factors. How do we identify the cause of an outcome from the myriad of factors that could have possibly led to the outcome? Fearon's (1996) proposed solution is to examine the notion of what is meant by a cause. (24) According to Fearon (1996), when one seeks to explain why an outcome Y occurred, he imagines a possible world in which Y is absent and the rest of the world is otherwise similar to the world in which Y is present. The latter portion of this standard is of particular importance. To demonstrate that X caused Y, the counterfactual to make valid is not "if X had not occurred, then Y would not have occurred." Rather, it should be "if X had not occurred, then Y would have not occurred and the world would be otherwise similar."

We can thus rule out butterfly effects, like Cleopatra's nose, as the cause of Wilson's rigidification while in office, because the effect of that particular causal factor is contrary to our conception of explanation. (25) While it is possible that had Cleopatra's nose been different, Wilson would not have rigidified, the logic behind this argument suggests a radically different world in 1919-20. Thus, the ancient queen of Egypt's nose does not work as a cause because we are searching for causes of Wilson's rigidification that would have produced a world that appeared otherwise similar had they been absent. Fearon's (1996) proposal again raises the problem of how to judge the similarity across possible worlds. My response to this problem is again to call for an informed debate on a case-by-case basis as to how similar the possible world is to the actual.

Consider second-order counterfactuals. Although one may have strong reasons to believe that the antecedent will produce the desired consequent, Lebow and Stein (1996) warn that subsequent developments may return history to the course from which is was diverted by the antecedent. They thus suggest that scholars should identify the most likely course of events that could unravel their consequent and the robustness of counterfactuals.

These seven criteria provide presidency scholars with a means of assessing the robustness of the counterfactuals they construct to test their theories. It should be self-evident that most of the criteria do not apply to miracle counterfactuals, which are not required to meet any actual world tests. The value of such counterfactuals derives from their ability to compel scholars to examine issues and problems in a new light. We are thus limited to using only plausible counterfactuals to test our theories. Although some of the outcomes that presidency scholars may wish to explain are a result of miracle causes, we simply cannot manipulate the causal factor counterfactually without violating a number of the proposed criteria. In other words, our ability to know "what would have happened if... "is severely limited with miracle counterfactuals. This is unacceptable if we seek to test our theories with any degree of certainty.

Why and When to Bother with Counterfactuals?

We are still, however, left with the question of how counterfactuals enable presidency scholars to increase the certainty of their causal inferences when focusing on the presidency as the unit of observation. The main issue revolves around what quantitative analysts might call the certainty of the parameter estimates. In the actual case strategy when the researcher has a large number of cases, the standard errors enable the analyst to get an idea of how much risk is attached to rejecting a true null hypothesis (committing a Type I error) and failing to reject a false null hypotheses (committing a Type II error). In the counterfactual case strategy, there is no formally accepted criterion for accurately gauging the risk of Type I or Type II errors associated with the hypothesized causal factor. A researcher's estimates of the causal effect of a particular factor are thus only as precise as the counterfactual is robust. If a theory can be shown to logically entail a counterfactual, and if that counterfactual fails to be robust, then this ought to make the researcher hesitant to accept that theory with any degree of certainty. On the other hand, if that counterfactual can be shown to be robust, then this should increase the researcher's confidence in that theory. (26) As Fearon (1991, 178) states, "[with the counterfactual method] arguments about relative importance of possible causes become arguments about the relative plausibility [robustness] of different counterfactual scenarios." The earlier discussion of the criteria for judging robust counterfactuals provides presidency scholars with a set of guidelines that will enable them to evaluate the robustness of their counterfactuals. Thus, the more confident they are in the robustness of their counterfactuals, the more certain presidency scholars, who focus on the presidency as the unit of observation, can be about their causal inferences.

This still begs the question of why presidency scholars should not simply heed King's (1993) advice and focus on different units of observation. To pose this question differently, Why should presidency scholars bother with the arduous task of constructing robust counterfactuals? Fearon (1991, 179) provides three conditions under which counterfactual arguments may be more compelling than utilizing actual cases. Each of these three conditions applies to research employing the presidency as the unit of observation, particularly the presidential psychology literature that King specifically targets in his critique.

The first condition identified by Fearon (1991) is when there is difficulty in identifying a sample of cases. There is no more severe a problem identifying a sample than when a researcher's n = 1, as it does when the presidency is the unit of observation. There was only one Carter and one Clinton and, as was discussed earlier, it is difficult, if not impossible, to employ the comparative method or any type of statistical techniques. Therefore, this situation is ideally suited for employing the counterfactual case strategy. The key point is that presidency scholars can search for additional implications of the same theory and abandon the presidency as the unit of observation, but the ability to pose robust counterfactuals means that they are not forced to heed King's (1993) prescriptions to increase the certainty of their causal inferences. Thus, counterfactuals also enable the investigator to retain the original research objective of his study--a possibility that empirical disaggregation may preclude (Ragin 2000; George and Bennett 2001).

The second condition identified by Fearon (1991) is when there are serious problems in operationalizing and measuring variables. In his analysis of the presidential psychology literature, King (1993, 403) points out that there are huge difficulties in defining and specifying the theoretical constructs employed by these scholars. Similarly, critics of the personality approach find the operationalization of personality constructs to be problematic (Greenstein 1969, 1992; Campbell 1993). In his examination of the political psychology literature, Greenstein (1992, 121) admits that "characterizations of personality structures themselves are never wholly persuasive, if only because of the absence of uniformly accepted personality theories with agreed-upon terminologies." Similarly, George (1974) notes that personality theory is a "quagmire" of illusive concepts, intuition, and subjective frameworks. If political psychologists cannot even agree on the definition of their concepts, then how are they to begin to go about constructing indicators for these personality types? The inability of political psychology scholars to operationalize their variables indicates that that the use of counterfactuals is highly warranted.

The final condition that Fearon (1991) identifies is when there is difficulty in conceiving of relevant controls. This problem is again apparent when presidency scholars employ the presidency as the unit of observation. Recall that King (1993, 402) himself admits that it is difficult to develop controls when we are examining individual presidents. There are just too many differences between individual presidents even when they appear to be similar. If a researcher cannot develop adequate controls for his cases, counterfactuals are ideally suited for the assessment of his causal claims.

This brings us to the core issue facing presidency scholars. When should they rely on counterfactuals to test their theories? The earlier discussion emphasized the existence of operationalization and measurement problems within the political psychology literature, the second condition under which Fearon (1991) suggests counterfactuals are warranted. However, what if the researcher is not within this subfield, or what if we make the optimistic assumption that those scholars can work out all the definitional and measurement problems? Should they still seek to rely on counterfactuals to test their theories? The answer is that it depends on the type of research question the investigator poses.

Consider the recent debate over the factors that influence the president's use of executive orders (Krause and Cohen 1997; Mayer 1999; Deering and Maltzman 1999; Moe and Howell 1999). Could counterfactuals prove useful in testing theories about the president's use of executive orders? It is entirely possible, but a more fruitful research strategy in this case is to rely on actual cases. The research question does not present the investigator with a small number of observations; nor is it fraught with problems of control. However, what if the research question is, What factors influence the president's propensity to rigidify while in office (Barber 1992)? Here the question directs the researcher's focus on the presidency as the unit of observation, which involves a small number of observations and problems of control. In this case, relying on counterfactual cases is the most expedient research strategy. Thus, the ultimate arbiter of when to use counterfactuals for theory testing is the type of research question posed. If the research question allows for a larger number of cases and fewer problems of control, such as those concerned with use executive orders, negotiations with Congress over treaties, or the use of vetoes, then the actual case strategy is the best route. However, if the research question leaves the investigator facing a small number of cases and problems of control, such as those concerned with the presidency as the unit of observation, then a reliance on counterfactuals is highly warranted. The analytical reality that a great deal of presidency scholars continue to focus on the presidency as the unit of observation suggests that counterfactuals could be of great interest in the future.

Psychological Perspectives on Counterfactuals

Up to this juncture, the discussion has largely focused on the epistemological and methodological threats to robust counterfactuals and the criteria that scholars should use to judge the robustness of counterfactuals. There is, however, a burgeoning literature in cognitive psychology on how individuals actually generate and judge counterfactuals that highlights some pitfalls in employing the counterterfactual method. (27) The psychological perspective informs us that it is quite difficult to avoid bias creeping into our thought experiments. Bias can creep into every stage of the process--from selecting the antecedent to judging the validity of the counterfactual itself. These biases can stem from both our limited cognitive capacities and our ideological and theoretical preconceptions. While these psychological barriers to causal inference afflict all the research methods discussed earlier, the problem seems to be particularly acute for the counterfactual method. This acuteness stems from the putative notion that counterfactuals are nothing but subjective speculation. To combat or minimize the vociferous protests of the skeptics, it is imperative for future practitioners of the counterfactual method to explicitly address the psychological perspective. Unless we confront these biases head-on, our ability to rely on counterfactuals to test theory will be severely compromised.

The selection of the antecedent is the first stage at which cognitive and motivational biases can creep into counterfactuals. Kahneman and Miller (1986) contend that individuals are cognitively attuned to notice changes. The greater the change from normality, the more likely individuals will notice it, attempt to explain it, and create counterfactuals in which they manipulate the antecedent to its expected default value. This type of counterfactual, which Tetlock (forthcoming) labels "norm-restoring counterfactual," is a thought experiment in which an unusual individual or event is transformed into something more routine. These types of counterfactuals tend to strike readers as more plausible than those that manipulate "typical" individuals or events into "atypical" individuals or events.

Kahneman (1995) emphasizes that individuals are also likely to manipulate those events that appear to be the final juncture in a historical process at which it appears possible to have produced an alternative outcome, what Tetlock (forthcoming) calls the "last-chance counterfactual." The psychological perspective informs us as well that individuals are likely to manipulate those events that were almost averted had something else that nearly happened occurred, what Tetlock labels the "close-call counterfactual." Finally, Tetlock warns us that our antecedent selection could suffer from hypothesis-confirming and/or focal-actor bias. The former refers to the tendency to search for antecedents that may aid in undoing or reinforcing an outcome of interest, while the latter emphasizes our focus on key actors in the narrative rather than those in the background.

The psychological perspective also alerts us to the biases that creep in when we judge the plausibility of the counterfactuals that we encounter. Tetlock and Belkin (1996b) caution us that the anchoring, availability, and representativeness of judgmental heuristics may bias our opinion in favor of some counterfactuals over others. (28) Probably the most deadly threat to the counterfactuals comes from theory-driven thinking. Research in cognitive psychology makes us wary of the ability of our audience to apply even standards of proof and evidence to all counterfactuals (Nisbett and Ross 1980; Fiske and Taylor 1991; Tetlock 1999b). From this perspective, the perceived plausibility of a counterfactual hinges on how hard one searches for its shortcomings, which may largely be a product of whether it offends the reader's ideological and theoretical preconceptions. Similarly, recent psychological research has found that beliefs about what would have happened in possible worlds are strongly related to one's own ideological and theoretical orientations (Tetlock 1998). It is, then, entirely possible that our counterfactuals can spin wildly out of control to satisfy our own ideological and theoretical convictions.

There are means to combat or at least minimize these biases. First, to minimize the biases inherent in selection of the antecedent, those generating counterfactuals should be explicit about what they consider the key choice points at which history may have taken an alternative course and why these choice points were singled out (Tetlock forthcoming). This criterion will force those engaging in counterfactuals to be clear and precise about their selection rule and to make it defensible. Scholars would also be well advised to follow the example set by Goldstone (forthcoming), who combats the hypothesis-confirming bias by listing several possible ways that his outcome of interest could have been reinforced or undone.

Those seeking to rely on counterfactuals should also be explicit about the connecting principles that link their antecedents and consequents (Tetlock forthcoming). The greater the specificity of the connecting principles, the better positioned are those who seek to judge counterfactuals to ascertain whether the thought experiment is plausible. Bringing these connecting principles into full view facilitates critical scrutiny of the robustness of the counterfactual and makes them less open to the critique that they are purely speculative.

A useful criterion thus far overlooked by the literature on counterfactuals is that scholars should seek to identify those theoretical and ideological perspectives that are "most likely" and "least likely" to take offense at the conclusions reached in the counterfactual. Doing so enables one to single out those scholars who constitute a least likely and most likely test for the counterfactual's robustness. (29) If the former group fails to undermine its robustness through their own arguments, then we may be able to claim that we are more certain in the inference we drew from it. On the other hand, if the latter group disagrees with the conclusions reached through the counterfactual, then the certainty of our causal inference diminishes greatly.

Finally, it seems reasonable to suggest that those generating counterfactuals should report the degree to which their use of counterfactuals sensitized them to the empirical reality that they might have otherwise overlooked (Tetlock forthcoming). While this certainly does not eliminate the problem that our counterfactuals are a product of our preconceptions, it does force scholars to be more explicit about the existence of them in their work. Ultimately, it may be that it is not the author of the counterfactual that gains the greatest benefit from writing it, but his audience. If a counterfactual sensitizes one member of the scholarly community about his or her own theoretical and ideological convictions, then it has performed a useful service. (30)

An Application of the Counterfactual Method to the Study of the American Presidency

In this section, I illustrate how the systematic use of counterfactuals may enable presidency scholars to test their theories with an adequate degree of certainty when the presidency is the unit of observation. The presidency is used as the unit of observation most commonly in the subfield of political psychology (George and George 1956; Kearns 1976; Glad 1980; Hargrove 1988; Barber 1992; Renshon 1996, 1998; Greenstein 2000). Probably the most famous of these works is Barber's (1992) Presidential Character, which argues that presidents with particular personality type (referred to in various forms as active-negative and ego-defensive) will "rigidify" sometime during their presidency. Woodrow Wilson, the leading example of this personality profile, is generally cited as evidence of rigidification. (31) For King (1993), these are exactly the research designs that we should avoid. In fact, in his critique of presidential research, King singles out Barber's study as one that is seriously flawed. For our purposes, it offers an opportunity to demonstrate the efficacy of counterfactual reasoning in testing theories with an adequate degree of certainty.

To elucidate the process through which this may be accomplished, I will seek to use counterfactuals to ascertain the importance of personality factors in influencing the president's propensity to rigidify while in office. More specifically, I will focus on Wilson's failed effort to secure ratification of the Versailles Treaty. If we were to follow King's (1993) advice, it would force us to abandon our focus on Wilson's presidency and shift our attention to additional observable implications of the same theory. Instead of examining the Versailles Treaty's failure as a case of rigidification, in King's view we should see it as an instance of the president's failing to secure ratification of a treaty. Cast in these terms, we would then focus on the importance of presidential personality for treaty negotiations with Congress and collect a sample of cases in which treaties were both ratified and rejected by the Senate. While this surely would make for an interesting research question, it fails to address our original query concerning rigidification. Not all treaty failures can be considered instances of rigidification, and by shifting our focus toward treaty negotiations with Congress, King's suggestion has moved us far away from our original research goal--a point not lost on other critics of King (Ragin 2000; George and Bennett 2001). It will now be shown that counterfactuals may enable scholars to maintain their focus on such important research questions concerning rigidification and still produce reliable causal inferences. (32)

The counterfactual that we seek to assert can be stated as follows: If Wilson had not had an active-negative personality type, then he would not have rigidified in office and the Versailles Treaty would have passed. This begs the question then of what type of personality Wilson would have had to possess to not rigidify in office. Barber (1992) identifies three other possibilities: the active-positive, passive-positive, and passive-negative personality types. For the purposes of illustrating the counterfactual method, I will focus only on the active-positive possibility, which Barber suggests is most likely to "succeed" as president. A more comprehensive test of Barber's theory would entail examining the other personality possibilities. Our revised counterfactual for purposes of illustration then becomes, If Wilson had an active-positive personality type, then he would not have rigidified in office and the Versailles Treaty would have passed.

It should also be noted that the failure of the Versailles Treaty should be viewed in this discussion as an instance of rigidification and not as a failed treaty negotiation with Congress. We could have easily used Nixon's presidency and arguably Hoover's or Lyndon Johnson's terms in office. The outcome of interest is thus the presence or absence of rigidification. Much of what follows deals with examining this counterfactual as it relates to the criteria suggested earlier. If counterfactuals are to be a worthwhile research method for presidency scholars, then it must be shown that one can construct a robust thought experiment that increases the certainty of one's causal inference.


Our first task is to present a well-specified antecedent and consequent. Much of this task has already been completed in our previous discussion. Our antecedent and consequent are clearly "if Wilson had an active-positive personality type" and "he would not have rigidified in office and the Versailles Treaty would have passed." Our task, however, is not complete, as we must specify how our explanation takes place relative to its "contrast space" of alternatives (Garfinkel 1981; Fearon 1996). Depending on how this contrast space is imagined, it heavily influences what will be accepted as a legitimate explanation for an outcome. For example, Barber's hypothesis is that Wilson's rigidification and the failure of the Versailles Treaty were due to Wilson's personality traits. Another scholar may reject this explanation on the grounds that it fails to explain why Wilson rigidified and the Versailles Treaty failed when it did. However, Barber's (1992) explanation may be a perfectly valid explanation as to why Wilson rigidified and the Versailles Treaty failed at all, rather than never at all. Mindful of these types of debates, Fearon (1996) argues that it is thus vital for researchers to explicitly specify the contrast space when advancing explanations. Otherwise, it leaves considerable', ambiguity about what exactly the researcher seeks to explain. One of the principal benefits of counterfactuals, then, is that they force the researcher to adequately specify the contrast space and thus the set of alternatives from which the outcome is explained (Fearon 1996, 56, footnote 31).

In our case, Barber's (1992) contrast space can be specified as [Wilson never rigidified in office; Wilson rigidified sometime in office]. This rules out those that may criticize Barber's theory for failing to explain when Wilson would have rigidified in office. For our purposes, there is no clear means to determine when Wilson would have rigidified, when the treaty would have passed, or what form it would have taken. These features are arguably irrelevant, as they are not included in our contrast space of alternatives. Furthermore, our counterfactual should be restated as, If Wilson had an active-positive personality type, then he would not have rigidified in office, the Versailles Treaty would have passed, and the world would be otherwise similar. This is done to rule out butterfly effects, which would imply a world dramatically different from the one that existed in 1919-20.

This discussion raises another important point: made by Fearon (1996): What one will accept as a cause of an outcome largely depends on the level of detail with which we specify the outcome. Fearon thus recommends that researchers should explicitly describe the class of events that would qualify as the outcome, what Garfinkel (1981) terms the "equivalence class." (33) We can then specify our equivalence class, Wilson's rigidification, as his rigid perseverance in any disastrous policy during his administration, especially any treaty ending the First World War that embodied some of his political beliefs. This specification is included because it affects the truth or falsity our counterfactual argument used to support Barber's (1992) theory, as it specifies the exact outcome the theory we are testing purports to explain. Finally, the clarity criterion demands that we be reasonably precise in specifying the connecting principle that links antecedent to consequent. Our primary connecting principle is that Wilson's different personality would have led him to compromise with the Senate over any Versailles Treaty (34) and any other policy proposal that met with opposition, rather than rigidify.


Historical Consistency

Our next task is to ascertain whether it is plausible that Wilson could have a different personality type. In other words, we must ask ourselves whether our counterfactual requires a massive rewrite of history. Since our counterfactual could arguably be considered a "norm-restoring" counterfactual, it would appear that our counterfactual satisfies this criterion. These types of counterfactuals transform an unusual person or event into something more normal or routine. For instance, Goldstone (forthcoming), in his contribution to the rise-of-the-West debate, moderates the unusual timidity of James II and the equally unusual boldness of William of Orange to the point at which they resemble average leaders of their day. Similarly, we are seeking to moderate the unusual perseverance of Wilson to that of a typical leader of his day. This type of counterfactual is logically appealing because it is consistent with the statistical principle of regression toward the mean (Tetlock forthcoming). This principle asserts that if we could repeatedly replicate history, the extreme values of variables with random variance would tend to revert to their mean. On these grounds, we can claim that our antecedent selection rule is defensible.

We should also apply the minimal rewrite rule to our earlier connecting principle that Wilson would have compromised with the Senate. There is considerable evidence to suggest that compromise with the Senate was an option urged by his supporters and advisors. George and George's (1956) and Margulies's (1989) detailed accounts of the ratification fight demonstrate that at numerous times during the Senate debate over the treaty, the "mild reservationists" informed Wilson that the Senate would pass the treaty only if he would accept four minor reservations that were attached to the treaty. Two instances in particular stand out. First, a week after presenting the treaty, Wilson called in a group of wavering Senators for a conference at the White House. According to George and George, "Almost to a man, these [mild reservationists] Senators warned Wilson that unless he accepted binding reservations, the Treaty would not be ratified" (p. 286). Then again on July 30, four Republican moderates informed Wilson the treaty could be ratified quickly with reservations, and Senator Frank Kellog said thirty-seven Republicans would support it with moderate reservations (Barber 1992, 16). Thus, our connecting principle is corroborated by the evidence and meets Tetlock's (forthcoming) "reasonableness principle of antecedent selection" in that compromise was one of the courses of action that was originally available to Wilson and does not require a massive rewrite of history to imagine.

Statistical Consistency and Assigning Causal Weight

Turning our attention to the question of statistical consistency, we need to provide some evidence to suggest there is a link between presidential personality and performance. While there have been some efforts to demonstrate this link (Etheredge 1978; Winter 1987; Shepard 1988; Spangler and House 1991; Kowert 1996), their results are largely not applicable to our counterfactual. Simonton's (1986, 1988) studies, which analyze thirty-nine presidents on fourteen personality traits using the Gough Adjective Check List (ACL) (35), does provide some evidence that can substantiate our claim. While Simonton employs a different categorization of personality types and leadership styles than Barber (1992), his measures do provide us with a rough approximation of Wilson's personality type as designated by Barber. Simonton (1986) shows that a president assessed as having high "inflexible" personality trait scores, which Wilson scores the second highest of the thirty-nine presidents in the sample, is likely to have poor relations with Congress. Conversely, a president assessed as having high "interpersonal" style scores, which Simonton (1988) scores Wilson as having the lowest of all thirty-nine chief executives, is likely to have superior relations with Congress. As one would expect, the interpersonal style is strongly inversely related to the inflexibility personality trait (Simonton 1988). While an inflexible President is described as being "stubborn," "persistent," "hard-headed," and "rigid" (note the similarity to rigidification), an interpersonal president is "flexible" and "willing to compromise." Finally, Simonton (1988) shows that Barber's positive rather than negative presidents are more "interpersonal." Simonton's studies thus provide us with some evidence to suggest that our connecting principle, Wilson would have compromised with an active-positive personality, is consistent with statistical evidence. Problems of measurement error and more important the validity of the ACL as an accurate measure of personality prevent us from reaching a more definitive conclusion.

The problem of any sort of meaningful statistical inference is further complicated by the collinearity problem--when omitted variables are controlled for, they may "explain away" the links between personality and rigidification. As George (1974, 253) states,
 when the behavior of the executive is subject to cross-pressures, the
 critical role of personality factors in his decisions may emerge more
 clearly, but even in such situations the causal weight or decisiveness of
 personality factors may be low.

It may be just as likely that the partisan makeup of the Senate, controlled as it was by the Republicans, was responsible for the treaty's death, and hence Wilson's personality was not responsible for his rigidification at all. Indeed, Senator Henry Cabot Lodge, Wilson's chief opponent, was out to kill the treaty by any means necessary, as his subsequent behavior and correspondence revealed (George and George 1956; Barber 1992). Could the Versailles Treaty's death have been a result of the Senate's or Lodge's rigidity? Available evidence seems to contradict this hypothesis. While Lodge and his fellow "irreconcilables" would not have compromised with Wilson, there was a large and potentially decisive body of senators who were willing to vote for the treaty with only a few modifications (George and George 1956; Barber 1992; Margulies 1989). The "mild reservationists" were eager to be a part of the move toward peace after World War I and were Wilson's best hope at securing ratification (Margulies 1989).

Despite the historical evidence, we are still unsure how much causal weight to attach to Wilson's personality as a factor influencing rigidification. George (1974, 253) succinctly summarizes our dilemma: "faced with the play of multiple, complexly interacting causal variables, the investigator is bound to have great difficulty in assessing the weight of any given factor." If we were to employ the actual case strategy, we would simply analyze the contrasting estimates of the effects of the different causal factors. In the counterfactual case strategy, we have no such estimates, and the degrees of freedom problem creeps up every time we introduce a new variable that might have influenced a particular outcome. Thus, explicit justification of claims about the relative effects of different causal factors requires a proliferation of counterfactual cases (Fearon 1991). (36) In our case, we have at least two competing explanations: (1) Wilson's rigidity was the result of his personality, and (2) Wilson's rigidity was the result of his personality type and the partisan makeup of the Senate. To assign a causal weight to our independent variables, one would need to construct the following thought experiments:

1. the Republicans control the Senate and Wilson has a different personality,

2. the Democrats control the Senate and Wilson has a different personality, and

3. the Democrats control the Senate and Wilson has the same personality.

To establish Wilson's personality as the sole or primary factor in explaining rigidification, one would need to invoke the criteria surveyed earlier to construct these three separate counterfactuals and examine the variance in the outcomes.

Theoretical Consistency

This leads us to the criteria of theoretical consistency. To fill in the missing counterfactual data points, we need to identify a theory that demonstrates an individual's personality type is an important factor in explaining his behavior and that this personality type can be reliably identified from evidence available prior to behavior as president (George 1974). In the social and political psychology literature, there exist a great number of competing personality theories and alternative nomenclatures. However, this conceals their commonalties: all theories necessarily take cognizance that humans are thinking, feeling creatures who exist in social environments and have inner qualities that shape their response to those environments. Beyond that, the various personality theorists--Freud, Lasswell, Jung, Allport, Murray, and many other scholars--differ from one another in what they emphasize. (37) For our purposes, we will rely on the theory developed by Lasswell (1951)--pioneer of the typological theorizing on which Barber (1992) relies.

Lasswell (1951) suggested that human behavior was not random but was in fact predictable and, therefore, classifiable. He argues that differences in behavior, which seem obvious to the eye, are a result of preferences created by the basic functions our personalities perform throughout life. These preferences emerge in early life, forming the foundation of our personalities. Subsequent issues of life are translated through each of our basic personality preferences. According to Lasswell, these preferences become the core of our attractions to and repulsion from people, tasks, and events all life long. If we invoke Lasswell's theory, it solidifies our counterfactual as it demonstrates the link between personality type and behavior.


The claim that our counterfactual does not violate the cotenability standard is questionable. Cotenability means that the antecedent must logically imply the consequent and there must be compatibility with all known facts and existing theory. On one hand, there is a compelling political logic that links the antecedent to the consequent, which Lebow and Stein (1996) claim should be specified. On the other hand, there may be existing theory that undercuts the tenability of the antecedent. Examining the political logic, I would argue that with his new personality and similar political beliefs, Wilson's well-documented political idealism would have led him to compromise with the Senate over the treaty. Though this idealism might have prevented him from accepting a significant watering down of the treaty by its critics (and possibly have led to rigidification), this appears unlikely since a majority of the senators would have supported it with only moderate reservations. To the extent that these moderate reservations did not conflict severely with his ideals, rigidification seems unlikely. Thus, Wilson's devotion to the political ideals embedded in the Versailles Treaty would have likely prevented rigidification.

Though a political logic may exist to link the antecedent to the consequent, our counterfactual may be incompatible with those that claim that Wilson's personality was a result of a series of strokes that altered Wilson's personality and mental attitudes (Weinstein, Anderson, and Link 1978). Our counterfactual may also be incompatible with theories that suggest biology, genetics, and acquired physical states contribute to personality and diffuse into political behavior (Park 1986; Masters 1989). Both lines of argument imply that even if we employed the enabling counterfactual concerning Wilson's father's treatment of him as a youth, Wilson's personality would remain more or less the same. To make our counterfactual cotenable, these theorists might argue that we must engage in a multistep counterfactual to alter Wilson's medical condition or genetic code. This we simply could not do, as the consequences of this thought experiment would be even more unknowable. (38)

Enabling Counterfactuals and the Antecedent

At this juncture, it is appropriate to examine our enabling counterfactual to ensure that it does not undercut the antecedent. Our enabling counterfactual of concern is the counterfactual that allows Wilson to have a different personality type. This begs the question, How might Wilson have a different personality type? The best answer I can give focuses on Wilson's relationship with his father. George and George (1956) emphasize the importance of Wilson's father's treatment of him as a youth as a determinant of his personality type. I can imagine that if Wilson's father had treated him more affectionately as a boy, then he would have had a different personality type. George and George's interpretation of Wilson's personality development may be debatable, but for our purposes the question is whether this enabling counterfactual undercuts the antecedent. It seems unlikely that this is the case, but it may have implications for the type of world in which Wilson would have found himself.


The matter of projectability can be addressed in a number of ways. The question is whether the hypothesis can be shown to apply to domains other than that for which it was initially developed and tested. To do so, we need to find an analogous situation in which an individual was faced with the pressures that triggered an active-negative/ego-defensive response. Concerning the life of Wilson, an option is to project this theory into his earlier life and determine whether similar rigidities in behavior are detected. His aforementioned difficulties as president of Princeton University serve as a good instance of this. Another option is to look at other active-negative/ego-defensive presidents, such Hoover, Johnson, and Nixon, and ascertain whether they rigidified. The problem with these projections is that they have already been well documented by presidential personality theorists (George and George 1956; Barber 1992). A better option would be to examine other world leaders, both past and present, and determine whether their personalities contributed to rigidities while in office. Comparable data on non-American world leaders, which rely on the ACL, already exist and would lend themselves nicely to testing this theory. This would place the American presidency within a broader theoretical context of leadership studies and perhaps lead to the creation of more generalizable theories of leadership behavior. (39)


Fearon's (1996) proximity criterion restricts our use of counterfactuals to only those causal mechanisms that are well understood and at a close spatial and temporal distance from the outcome of interest. The personality approach clearly at this point does not satisfy this standard (Greenstein 1969, 1992; George 1974; Campbell 1993). Fearon (1996, 54) is, however, willing to consider a fallback position. Although it may not be possible to conclusively demonstrate that A caused B, it may be possible, he contends, to demonstrate that without A, whatever would have happened, it would not have been B. Lebow (2000a) employs this strategy to cast doubt on the ability of structural theories of international relations to explain change in the international system. Our counterfactual, in a sense, has been following this strategy all along. Never have we precisely stated the world that would have occurred, only that it would not have had Wilson rigidifying. While relying on this fallback strategy is a weaker test of theory, it does allow the researcher the means to say with some degree of certainty that the hypothesized causal factor did have some effect on the outcome.

Further Considerations

Avoiding the Conjunction Fallacy

Our counterfactual is intended to produce a set of worlds with particular characteristics. In particular, this set of worlds includes all possible worlds where Wilson does not rigidify and the Versailles Treaty is ratified at any time and in any form. This helps us to minimize the problems associated with the conjunction fallacy because our counterfactual is not designed to produce a specific world, rather a set of worlds that meets our specifications. Thus, one could argue that the cumulative probability of our counterfactual occurring is greater than one that specifies one specific world occurring. (40)


When considering interconnectedness at first glance, it does not appear that a new personality for Wilson would have drastically changed the circumstances in which Wilson found himself in 1919. He arguably still would have been elected president (41) and pressed for the League of Nations at the Paris Peace Conference. Historical evidence demonstrates that Wilson already compromised slightly on the wording of the League Covenant and that Colonel House, his trusted advisor, conducted a great deal of the negotiations in Wilson's absence (George and George 1956). One could speculate that with Wilson's new personality, he would have been more apt to accept these compromises negotiated by House, rather than chastise him for some of them.

Some readers may question whether Wilson's new personality might have given him new political beliefs. Others may point out that Wilson's new personality may have made him less apt to publicly state that the 1918 midterm elections should be a referendum on his Fourteen Points, a claim that may have cost Wilson's party some seats in the Senate. To evade these potential criticisms, we should follow the recommendation of Tetlock and Belkin (1996b, 20) and create a compound antecedent thus restating our counterfactual as follows: If Wilson had an active-positive personality type, retained the same political beliefs, and the partisan makeup of the Senate remained the same, then he would not have rigidified in office, the Versailles Treaty would have passed, and the world would be otherwise similar. This restatement allows us to better isolate the causal effect of Wilson's personality on the outcome of interest.

Considering Second-Order Counterfactuals

Considering the possibility of second-order counterfactuals leads to the assessment, if Wilson had a new personality, whether there was any likely course of events that could have negated or unraveled the consequent. I will briefly discuss three such possibilities. The first possibility is the events surrounding Wilson's tenure as president of Princeton University prior to his run for governor of New Jersey and the presidency, which George and George (1956) cite as an earlier instance of rigidification. It is quite possible that his tenure at Princeton would have turned out quite differently had Wilson been an active-positive personality type. The squabbles that Wilson had with Dean Andrew West and important donors at Princeton made him more receptive to the calls from the New Jersey Democratic Party leaders that he should accept their nomination for governor as a stepping stone to the presidency (George and George 1956). If Wilson had a different personality type, it is possible that he may have compromised with his opponents over these two issues. This, in turn, may have made him less likely to leave Princeton to pursue a career in politics, and thus he may have never become president! However, it is also possible that Wilson's opponents would not have compromised, regardless of Wilson's offers, thus facilitating his receptivity to calls to run for governor. Moreover, even if Wilson had compromised with his opponents, it still does not rule out the possibility of his leaving Princeton for a career in politics, though it may have delayed his run for governor and hence the presidency. Obviously, we cannot know the answer for certain, but this possibility does cast some doubt on the robustness of our counterfactual.

It is also possible that even if Wilson had sought to compromise as president, the Senate still would have killed the Versailles Treaty. In fact, Wilson wholeheartedly believed that if he sought a compromise with the "mild reservationists," the opponents of the treaty would use this an opportunity to propose other, more objectionable reservations that would be impossible for him to consider. Despite the historical evidence presented earlier to refute this claim, it may be possible that it was the Senate, not Wilson's personality, that defeated the treaty. To deal with this possibility, we would have to construct the counterfactual cases described earlier in which the partisan makeup of the Senate is altered.

Historical evidence also suggests the possibility that Wilson could have run for reelection again in 1920. Wilson had hoped to be nominated at the convention in San Francisco and even went so far as to draft an acceptance address (Craig 1992). If it were not for the failed ratification of the Versailles Treaty, it is possible the Democrats would have tapped Wilson a third time. (42) At the time, Democratic Party leaders largely feared a public backlash against Wilson and Democratic congressional candidates over Wilson's insistence on the League of Nations as the overriding issue of 1920. If the treaty had been ratified, then this potential obstacle would have been removed. Therefore, if Wilson was tapped to run again in 1920 and if he had won the election, then it still would have been possible for him to rigidify in his third term. I am hesitant to speculate what policy proposal might lead to this and simply will identify it as a possibility. I should conclude this discussion of possible second-order counterfactuals by pointing out that even if Wilson had secured ratification of the treaty and finished his second term, Democratic Party leaders may still not have nominated him because of the strong two-term tradition. Thus, Wilson's desire to run for reelection must be counterbalanced by the party leaders' hesitation at the time of defying this important norm.

Addressing the Psychological Perspective

Our final task is to address the concerns of the psychological perspective. Earlier, I sought to make the antecedent selection rule as clear and defensible as possible by specifying the counterfactual as "norm-restoring." Moreover, I have attempted to make the connecting principles as clear as possible by focusing on Wilson's need to compromise to avoid rigidification and the political logic that would have led him to do so. To combat hypothesis-confirming and focal-actor bias, I have suggested the possibility of Wilson's medical problems, genetics, and physical attributes as well as the partisan makeup of the Senate as factors that could have reinforced or undone the actual outcome. To minimize the problem of readers judging this counterfactual as more or less assertible based on their theoretical and ideological orientations, I will now specify those audiences that would be least likely and most likely to take offense at the conclusions reached. Those least likely to take offense are clearly those scholars that emphasize how presidential personality may impinge on presidential performance (George and George 1956; Neustadt 1960; Kearns 1976; Glad 1980; Hargrove 1988, 1993; Renshon 1996, 1998; Greenstein 2000). The most likely group of scholars to take offense consists of

1. pluralist and rational choice theorists who contend that the presidency is so institutionalized, and to such a degree a prisoner of excessive public expectations and other external constraints, that personality matters little in terms of outcomes (Lowi 1985; Moe 1993);

2. scholars who claim that Wilson's personality and behavior were influenced by his medical problems (Weinstein, Anderson, and Link 1978);

3. those who suggest that personality is to some extent influenced by genetics and physical characteristics (Park 1986, Masters 1989); and

4. those who suggest the "politics presidents make" are more a function of the incumbent's warrant for change (Skowronek 1993).

Last, I will report that constructing this counterfactual has sensitized this author to some extent to the points in Wilson's life that may have unraveled the counterfactual-chief among these is his tenure at Princeton and the possibility of his running for reelection in 1920.


This counterfactual is by no means meant to be the definitive test of Barber's (1992) theory of rigidification. It is merely intended to demonstrate the process through which counterfactuals can be utilized in the American presidency and the problems and issues that arise when one attempts to do so. The ideal counterfactual would include in-depth historical detail concerning the events, an exploration of the effect of Barber's other personality types, an elaboration of the theories employed to connect the antecedent and consequent, and an examination of each of the counterfactuals suggested to assign causal weights to each hypothesized causal factor. I have tried to provide a preliminary demonstration of how this would take place, but other scholars who choose to employ this method must do the heavy lifting. I have only tried to show that presidency scholars may be able to utilize this method.

Has this exercise in counterfactual thought experimentation increased the certainty of Barber's (1992) causal inference concerning Wilson's rigidification? I simply cannot say with full confidence. My judgment is that the work on presidential personalities has demonstrated that a president's personality does affect substantive outcomes. It is harder to say whether that work can move us to a presumption that the counterfactual--a Wilson with a different personality would have prevented his rigidification--is correct.

The counterfactual logically entailed by Barber's (1992) theory did offer some evidence to validate his claim, but one could hardly call this a robust counterfactual. Indeed, without detailed knowledge of the causal process through which personality affects outcomes, we can never be highly certain. In any single outcome, an active-negative/ego-defensive personality will not always lead to rigidification (Greenstein 1969; George 1974). Moreover, other factors may intervene to make the outcome more probable (e.g., an opposition-controlled legislature). This returns us to Fearon's (1996, 50) earlier point: "detailed counterfactual scenarios will have a chance at being rendered plausible only if the proposed causes are temporally and, in some sense, spatially quite close to the consequents." Thus, my inability to say with any high degree of certainty what would have happened if Wilson had a different personality type is largely a product of the counterfactual failing the proximity criterion. Furthermore, problems of cotenability and second-order counterfactuals tended to undermine the robustness of the counterfactual. It is also possible that this counterfactual failed to produce a robust result because it sought to test personality theory (a notoriously difficult theory to employ). Nevertheless, this exercise in counterfactual thought experimentation should sensitize presidency scholars to the advantages and risks of choosing the counterfactual method.


King's (1993) suggestion that qualitative researchers, and presidency scholars in particular, need to be more explicit about the certainty of their inferences is certainly valid, and this article has not disputed that claim. By employing the suggestions outlined by King, presidency scholars will undoubtedly be able to avoid (or at least lessen) criticisms that their findings are uncertain. However, his suggestion that presidency scholars abandon the president as the unit of observation is overstated. This article has demonstrated that by relying on counterfactuals that are grounded in the systematic framework presented earlier, scholars may be able test their theories with an adequate degree of certainty while continuing to focus on the presidency as the unit of observation.

Undoubtedly my argument will be subject to the criticism that counterfactuals themselves are not falsifiable and in that sense do not provide a viable option for testing theories. While this claim is understandable, it fails to recognize that all researchers making causal claims from nonexperimental data must confront counterfactuals (Fearon 1991). It is true that researchers can never know exactly what would have happened, but this should not prevent researchers from recognizing that counterfactuals play an important role in testing theories. Presidency scholars can venture arguments about what would have happened if they meet the criteria suggested earlier. The key factor in relying on counterfactuals to test theories is to make the thought experiments defensible and in this sense following the framework suggested as to what constitutes a robust counterfactual.

I should conclude by emphasizing that I am not advocating an abandonment of the actual case strategy in favor of the counterfactual case strategy. Each strategy has its advantages and drawbacks, and presidency scholars should be aware of these when utilizing them to test their theories with an adequate degree of certainty. If presidency researchers are to draw one lesson from this discussion it should be this: if they choose to focus on the president as the unit of observation, then they may be able to increase the certainty of their inferences by employing carefully constructed robust counterfactuals.

Situating the Four Research Methods within Fearon's (1991)

Actual Case Strategy Counterfactual Case Strategy

Case study method Counterfactual method
Comparative method
Statistical method

AUTHOR'S NOTE: The author is greatly indebted to John T. Woolley, who provided valuable comments and advice on several versions of this article. Additional thanks go to Aaron Belkin, Richard Ned Lebow, M. Stephen Weatherford, Erwin C. Hargrove, and three anonymous referees for their helpful suggestions. Earlier versions of this article were presented at the annual meeting of the Western Political Science Association, March 15-17, 2001, in Las Vegas; and the quarterly meeting of the Southern California Methodology Program (SCAMP), October 26, 2001, at the University of California, Riverside.

(1.) For an alternative view concerning whether the logic of scientific inference is based on classical statistics, see Ragin and Zaret (1983); Munck (1998); McKeown (1999); and Goldstone (2001a, 2001b).

(2.) On the case study method, see Campbell (1975); Lijphart (1971); Eckstein (1975); George (1979); Yin (1984); George and McKeown (1985); Achen and Snidal (1989); King, Keohane, and Verba (1994); McKeown (1999); Mahoney (2000); and George and Bennett (2001). On the comparative and statistical methods, see Verba (1967); Lasswell (1968); Przeworski and Teune (1970); Sartori (1970, 1991); Lijphart (1971, 1975); Geertz (1973); Mill ([1843] 1974); Almond and Genco (1977); Stinchombe (1978); Skocpol and Somers (1980); Skocpol (1984); Tilly (1984); Jackman (1985); Przeworski (1987); Ragin (1987); Collier (1993); Collier and Mahon (1993); King, Keohane, and Verba (1994); Collier and Mahoney (1996); Dion (1998); Mahoney (2000); and George and Bennett (2001).

(3.) Adequate variation on the dependent variable may not be necessary for all kinds of research. Indeed, some of the most influential studies in comparative politics have violated the rules of case selection suggested by quantitative scholars (Rogowski 1995). The case study method is ideal for bringing to light anomalies for which current theory cannot account (Achen and Snidal 1989), digging into the details of a case to uncover new insights concerning the process under study (Collier and Mahoney 1996, 74), analyzing causal heterogeneity, and developing new hypotheses (Lijphart 1971; Eckstein 1975; George 1979; McKeown 1999; George and Bennett 2001; see Collier and Mahoney 1996, 71-72, for some pitfalls to this final benefit). For recent rejoinders to King, Keohane, and Verba's (1994) claims about the case study method, see McKeown (1999) and George and Bennett (2001).

(4.) King, Keohane, and Verba (1994, 41, 54, 211-12, 221, 225) qualify this point, claiming that a single observation study can be useful in evaluating causal claims if it is part of a broader research program. In this case, the single observation study can then be compared with other single observation studies and then the original study is no longer a single observation. It is also possible that a case study may contain more than one observation. On the latter point, see also Campbell (1975, 179, 181-82) and Bennett (1997). For an alternative view on the utility of the case study method for theory assessment, see Eckstein (1975); George and McKeown (1985); McKeown (1999); and George and Bennett (2001).

(5.) To be sure, there are alternatives under the rubric of the comparative method to the method of controlled comparison. In addition to Lijphart's (1971, 1975) method of controlled comparison, Mill's ([1843] 1974) method of difference, and Przeworski and Teune's (1970) "most similar systems" design (all of which I consider to follow the same logic of seeking to create the functional equivalent of an experiment through a strategy of matching), an investigator may rely on Mill's method of agreement, Przeworski and Teune's "most different systems" design, or a longitudinal "before-after" design (Campbell and Stanley 1963). These alternatives, however, are unlikely to be fruitful for hypothesis testing for presidency researchers. The first two are hampered by selection on the dependent variable problems (testing the possibility of necessary causes aside; see Dion 1998) whilst the latter by problems of possible "confounding variables" originating from the inability of the investigator to hold all other variables constant while studying changes in only one hypothesized explanatory variable. For more on these alternatives to the method of controlled comparison, see George and Bennett (2001).

(6.) In their discussion of "most similar" systems research designs, Przeworski and Teune (1970) make a similar point about the indeterminacy of research designs employing the method of controlled comparison. They claim that such research designs tend to be overdetermined in that they fail to eliminate many rival explanations, leaving the researcher no criteria to choose among them. In their most recent treatment of this method, George and Bennett (2001) also arrive at the same conclusion.

(7.) Obviously, we can never be 100 percent sure of our conclusions; such is the fundamental problem of causal inference, but some conclusions are "more sure" than others. This is what scholars must be aware of when engaging in qualitative research and is one of King, Keohane, and Verba's (1994) most enduring points.

(8.) King's (1993) suggestions on how to increase the number of observations are quite similar to Lijphart's (1971) discussion of problems inherent in utilizing the comparative method. For a more detailed view of King's suggestions, see King, Keohane, and Verba (1994, 217-30).

(9.) Examples of recent scholarly work that continues to rely on the presidency as the unit of observation include Renshon (1996, 1998, 2000); Gergen (2000); Greenstein (2000); and Holzer (2000).

(10.) A reading of King, Keohane, and Verba (1994) shows that King, in Designing Social Inquiry's discussion of the fundamental problem of causal inference, has thought a great deal about the need to formulate counterfactuals to sustain causal inference. At the same time, all readers of King's vast array of research are aware that King is probably much more confident in the statistical method than the counterfactual method for assessing causal claims. However, as Fearon (1991) has shown, even statistical methods (at least those relying on ordinary least squares regression) depend on counterfactual inference. If one does not believe in the possibility of making valid counterfactual inferences, then one should not have any confidence whatsoever in statistical methods. Thus, if King believes in statistical methods, presumably he also believes in the possibility of making valid counterfactual statements. But if this is the case, then why is the fundamental problem of causal inference so fundamental? Furthermore, it makes King's omission of the counterfactual method from his discussion all the more glaring.

(11.) King (1993, 402) states that "the n = 1 problem guarantees that this sort of difficulty will always come up in the presidency research, perhaps even more frequently in this subfield than anywhere else."

(12.) See Mueller (1989) for an instance of committing this fallacy and Cederman (1996) for criticism of his approach.

(13.) Since it is necessary to make a counterfactual claim to sustain a statistical inference (Fearon 1991), it is not entirely clear that counterfactuals constitute a third-rate method for assessing causal claims. If counterfactuals are pretty much always involved in assessing causal claims, then why are they third rate? If the chain is no stronger than its weakest link, then statistical inference should not be any more plausible then counterfactual inference.

(14.) For an alternative view to this "frequentist" logic, see Bennett (1997); McKeown (1999); and Goldstone (2001a, 2001b).

(15.) Fearon (1991, 1996) contends that counterfactuals may provide the controlled comparisons necessary to support causal inferences when the researcher is handicapped by negative degrees of freedom. King, Keohane, and Verba (1994, 78) briefly discuss the use of counterfactuals, claiming, "they must reasonable and it should be possible for the counterfactual event to have occurred under precisely stated circumstances." George and McKeown (1985, 33) suggest that when controlled comparison does not offer a means of assessing one's causal claims, the researcher may rely on his "disciplined and analytical imagination to assess the plausibility of an argument about causal processes." Last, in his discussion of the study of personality and politics, Greenstein (1992, 124) contends that counterfactuals are "the only available alternative in analyses of single events to the quantitative analysis that would be called for if data existed on large numbers of comparable episodes."

(16.) The use of regression analysis provides a vivid example of what Fearon (1996) means by "miracle causes." When an investigator uncovers an association between the independent and dependent variable, then he assumes that had the independent variable of any particular case taken on a different value, the dependent variable would have differed by a systematic component that is the same across all cases plus a stochastic component. In this view, the investigator never explicitly asks whether it was conceivable or plausible for the case to have taken on a different value, but it is still identified as a cause.

(17.) See also Tetlock and Lebow (2001).

(18.) See Collier (1993) for a review of this conversation concerning the comparative method and to a lesser extent the other three methods.

(19.) The latter proposal is based on the logic of the Lewis (1973) and Stalnaker (1984) criterion that states the validity of a counterfactual is largely determined by how "close" the possible world is to the actual world.

(20.) For a discussion of these problems, see King, Keohane, and Verba (1994,115-206) and Lebow (2000b).

(21.) Here the researcher is not using a theoretical understanding to generate the counterfactual case and then using the same case to test the theory that generated it. Rather, the researcher is utilizing an entirely different theory that has been made independently credible to deduce what would have happened. For instance, Keohane (1984) relies on the Coase theorem to show that in the absence of regimes the level of cooperation in world politics would have declined. For a survey of other efforts see Tetlock and Belkin (1996b, 26-27).

(22.) The oft-cited example to illustrate violation of this standard is Elster's (1978) critique of Fogel's (1964) counterfactual concerning the importance of railroads for the growth of the American economy. Fogel claims that if railroads had not existed, the American economy would have grown only slightly more slowly than it actually did since a strong incentive would have existed to invent the internal combustion engine sooner. Elster (1978) responds that if the technology were present to invent and produce automobiles, it would have certainly led to the development of railroads as well. According to Elster, Fogel's connecting principle was not logically consistent with the antecedent since it undercut its assertibility.

(23.) Fearon (1996, 66) does note that this criterion will lead us to eschew "counterfactuals that are, unfortunately, of great interest."

(24.) As Tetlock and Belkin (1996b, 21) state, "Fearon's proposal is open to challenge on the ground that it arbitrarily rules out causes that, because of their location in complex systemic networks of causation, do not have effects that can be conceptually isolated." Tetlock and Belkin recognize and argue that there is no easy resolution to this problem but that it is best dealt with on a case-by-case basis.

(25.) This, of course, is a slight adaptation of Cleopatra's Nose Problem (Carr 1961). According to Pascal, if Cleopatra's nose had been different, Antony might not have fallen in love with her, and the entire course of Western history might have been different.

(26.) This point should be seen as similar to King, Keohane, and Verba's (1994) methodological advice for researchers to search for all observable implications of a theory. The counterfactual method and King, Keohane, and Verba's advice both lead the researcher to ask, "If my theory is correct, what else should be true?" The two approaches do not contradict one another. Rather, they should be seen as complementary. A key difference, however, is that the counterfactual method allows the researcher to maintain his research goal--a possibility that King, Keohane, and Verba's advice may preclude.

(27.) For a review of this literature, see Roese and Olson (1995) and Tetlock and Belkin (1996b).

(28.) On this heuristic trilogy, see Tversky and Kahneman (1974). Briefly summarizing the implications of these judgmental heuristics for counterfactuals--the anchoring heuristic may lead one to dismiss those scenarios that dramatically differ from the actual world (limiting the ability to appreciate the arbitrariness of the actual world); the availability heuristic could lead one to readily believe an easily imagined, compelling counterfactual thus committing the conjunction fallacy; the representativeness heuristic could lead one to be slow to concede the small changes can lead to large effects and vice versa.

(29.) For more on what constitutes a "least likely" and "most likely" test, see Eckstein (1975).

(30.) This returns us again to the Tetlock-Lebow-Weber view of counterfactuals, which emphasizes their importance as "mind-set changers" or learning devices. This article has not sought to deny such a role; rather, it has sought to shift the discourse on counterfactuals back to their uses as data points in explanations.

(31.) Nixon's handling of the Watergate scandal is also commonly cited as an instance of rigidification (George 1974; Barber 1992). The cases of Hoover's refusal to grant unemployment benefits and Johnson's Vietnam policy, which Barber (1992) also cites as instances of rigidification, have been questioned (George 1974).

(32.) Some may question this approach claiming that we could still examine rigidification by constructing a data set of actual cases of this outcome by looking at levels of observation different from the presidency. If this were possible, then it would make for a fine research design, but precisely because there are so few instances of rigidification, it cannot be done.

(33.) An example provided by Fearon (1996, 59) may help illustrate that idea of equivalence class. When one seeks to explain the French Revolution, for our purposes of explanation the researcher considers the French Revolution as the large class of occurrences, all of which are equivalent to the French Revolution. A theory explaining the revolution would then seek to account for these instances, but not the fact that Robespierre wore a particular piece of clothing on a particular day even though this would be part of the French Revolution. A hypothesized causal factor that fails to account for Robespierre's wardrobe would then not be disqualified. For further elaboration on the notion of equivalence class, see Garfinkel (1981).

(34.) I use the phrase "any Versailles Treaty" as shorthand for any treaty ending the First World War that embodied some of Wilson's political beliefs.

(35.) For more on the Adjective Check List (ACL), see Gough and Heilbrun (1965) and Historical Figures Assessment Collaborative (1977).

(36.) Elster (1983) cautions this strategy may fail if causes interact nonadditively.

(37.) These differences include the extent to which they emphasize one class of motivation over another, in their sensitivity to the individual's environment, in the weight they put on biology, in the extent to which they view personality to be structured, and in many other respects.

(38.) In response to these potential criticisms, one could point to arguments that refute the medical hypothesis (George and George 1981-1982, 1998) and claim that biology, genetics, and acquired physical states only go so far in influencing one's personality.

(39.) By placing the American presidency within the broader field of leadership studies, it might also solve the problem of employing the president as the unit of observation in the actual case strategy. Conceivably, one could focus on as many leaders needed to construct reliable causal inferences. This point, however, is a paper for another day.

(40.) This argument necessarily depends on where in the sample distribution of all possible worlds our collection of possible worlds falls in relation to the one specific possible world. If our collection of possible worlds is located near the tails of the sample distribution and the one specific world lies near the mean, then our claim is necessarily false.

(41.) It is possible, however, that despite his rigid stands over certain issues, Wilson's moralism was an important part of his electoral appeal and this presumably was dependent on his personality.

(42.) For an alternative view, see Levin (2001).


Achen, C. H., and D. Snidal. 1989. Rational deterrence theory and comparative case studies. World Politics 41:143-69.

Almond, G. A., and S. J. Genco. 1977, Clouds, clocks, and the study of politics. World Politics 29:489-522.

Barber, J. D. 1992. Presidential character: Predicting performance in the White House. 4th ed. Englewood Cliffs, NJ: Prentice Hall.

Bennett, A. 1997. Lost in the translation: Big (N) misinterpretations of case study research. Paper presented at the annual meeting of the International Studies Association, March 18-22, in Toronto, Canada.

Breslauer, G. 1996. Counterfactual reasoning in western studies of Soviet politics and foreign relations. In Counterfactual thought experiments in world politics: Logical, methodological, and psychological perspectives, edited by P. E. Tetlock and A. Belkin. Princeton, NJ: Princeton University Press.

Campbell, C. 1993. Political executives and their officials. In Political science: The state of the discipline II, edited by A. W. Finifter. Washington, DC: American Political Science Association.

Campbell, D. T. 1975. "Degrees of freedom" and the case study. Comparative Political Studies 8:178-93.

Campbell, D. T., and J. Stanley. 1963. Experimental and quasi-experimental designs for research. Chicago: Rand McNally.

Carr, E. H. 1961. What is history? London: MacMillan.

Cederman, L. E. 1996. Rerunning history: Counterfactual simulation in world politics. In Counterfactual thought experiments in world politics: Logical, methodological, and psychological perspectives, edited by P. E. Tetlock and A. Belkin. Princeton, NJ: Princeton University Press.

Collier, D. 1993. The comparative method. In Political science: The state of the discipline II, edited by A. W. Finifter. Washington, DC: American Political Science Association.

Collier, D., and J. E. Mahon. 1993. Conceptual "stretching" revisited: Alternative views of categories in comparative analysis. American Political Science Review 87:845-55.

Collier, D., and J. Mahoney. 1996. Insights and pitfalls: Selection bias in qualitative research. World Politics 49:56-91.

Craig, D. B. 1992. After Wilson: The struggle for the Democratic Party, 1920-1934. Chapel Hill: University of North Carolina Press.

Dawes, R. 1996. Counterfactual inferences as instances of statistical inferences. In Counterfactual thought experiments in world politics: Logical, methodological, and psychological perspectives, edited by P. E. Tetlock and A. Belkin. Princeton, NJ: Princeton University Press.

Dion, D. 1998. Evidence and inference in the comparative case study. Comparative Politics 30:127-46.

Deering, C. J., and F. Maltzman. 1999. The politics of executive orders: Legislative constraints on presidential power. Political Research Quarterly 52:767-83.

Eckstein, H. 1975. Case study and theory in political science. In Handbook of political science, vol. 7, edited by F. I. Greenstein and N. W. Polsby. Reading, MA: Addison-Wesley.

Elster, J. 1978. Logic and society: Contradictions and possible worlds. New York: Wiley.

--. 1983. Explaining technical change. Cambridge: Cambridge University Press. Etheredge, L. S. 1978. Personality effects on american foreign policy, 1898-1968: A test of interpersonal generalization theory. American Political Science Review 72:434-51.

Fearon, J. D. 1991. Counterfactuals and hypothesis testing in political science. World Politics 43:169-95.

--. 1996. Causes and counterfactuals in social science: Exploring an analogy between cellular automata and historical processes. In Counterfactual thought experiments in world politics: Logical, methodological, and psychological perspectives, edited by P. E. Tetlock and A. Belkin. Princeton, NJ: Princeton University Press.

Fisher, D. H. 1970. Historians' fallacies: Toward a logic of historical thought. New York: Harper & Row.

Fiske, S. T., and S. E. Taylor. 1991. Social cognition. Reading, MA: Addison-Wesley.

Fogel, R. 1964. Railroads and American economic growth. Baltimore: Johns Hopkins University Press.

Garfinkel, A. 1981. Forms of explanation. New Haven, CT: Yale University Press.

Geertz, C. 1973. Thick description: Toward an interpretative theory of culture. In The interpretation of cultures, edited by C. Geertz. New York: Basic Books.

George, A. L. 1974. Assessing presidential character World Politics 26:234-82.

--. 1979. Case studies and theory development: The method of structured, focused comparison. In Diplomacy: New approaches in history, theory, and policy, edited by P. G. Lauren. New York: Free Press.

George, A. L., and A. Bennett. 2001. Comparative methods: Controlled comparison and within-case analysis. Paper presented at the annual meeting of the American Political Science Association, August 30 to September 2, in San Francisco. Available from

George, A. L., and J. L. George. 1956. Woodrow Wilson and Colonel House, a personality study. New York: John Day.

--. 1998. Presidential personality and performance. Boulder, CO: Westview.

George, A. L., and T. J. McKeown. 1985. Case studies and theories of organizational decision making. In Advances in information processing in organizations, vol. 2, edited by L. S. Sproull and P. D. Larkey. Santa Barbara, CA: JAI.

George, J. L., and A. L. George. 1981-1982. Woodrow Wilson and Colonel House: A reply to Weinstein, Anderson, and Link. Political Science Quarterly 96:641-65.

Gergen, D. 2000. Eyewitness to power: Essence of leadership, Nixon to Clinton. New York: Simon & Schuster.

Glad, B. 1980. Jimmy Carter: In search of the great White House. New York: Norton.

Goldstone, J. A. 2001a. Causes and explanation in qualitative methods: A comment on "Nominal, ordinal, and narrative appraisal in macro-causal analysis" and "Strategies of causal inference in small-N analysis" by James Mahoney. Presented at the Inter-University Faculty Consortium on Qualitative Research Methods, listserv roundtable, March 28.

--. 2001b. Comparative-historical analysis and knowledge accumulation in the study of revolutions. In Comparative-historical analysis: Achievements and agendas, edited by James Mahoney and Dietrich Rueschemeyer. Manuscript.

--. Forgthcoming. Europe's peculiar path: The unlikely transition to modernity. In Unmaking the West: Alternative histories of counterfactual worlds, edited by P. E. Tetlock, R. N. Lebow, and G. Parker. New York: Columbia University Press.

Goodman, N. 1983. Fact, fiction and forecast. Cambridge, MA: Harvard University Press.

Gough, H. G., and A. B. Heilbrun. 1965. The Adjective Check List manual. Palo Alto, CA: Consulting Psychologists Press.

Greenstein, F. I. 1969. Personality and politics: Problems of evidence, inferences, and conceptualization. Chicago: Markham.

--. 1992. Can personality and politics be studied systematically? Political Psychology 13:105-28.

--. 2000. The presidential difference: Leadership style from FDR to Clinton. New York: New York University Press.

Hargrove, E. C. 1988. Jimmy Carter as president: Leadership and the public good. Baton Rouge: Louisiana State University Press.

--. 1993. Presidential personality and leadership style. In Researching the presidency: Vital questions, new approaches, edited by G. C. Edwards III, J. H. Kessel, and B. A. Rockman. Pittsburgh, PA: University of Pittsburgh Press.

Historical Figures Assessment Collaborative. 1977. Assessing historical figures: The use of the observer-based personality descriptions. Historical Methods Newsletter 10:66-76.

Holzer, H. 2000. Lincoln: Seen and heard. Lawrence: University Press of Kansas.

Jackman, R. W. 1985. Cross-national statistical research and the study of comparative politics. American Journal of Political Science 29:161-82.

Jervis, R. 1993. Systems and interaction effects. In Coping with complexity in the international system, edited by J. Snyder and R. Jervis. Boulder, CO: Westview.

--. 1996. Counterfactuals, causation, and complexity. In Counterfactual thought experiments in world politics: Logical, methodological, and psychological perspectives, edited by P. E. Tetlock and A. Belkin. Princeton, NJ: Princeton University Press.

Kahneman, D. 1995. Varieties of counterfactual thinking. In What might have been: The social psychology of counterfactual thinking, edited by N. J. Roese and J. M. Olson. Mahwah, NJ: Lawrence Erlbaum.

Kahneman, D., and D. T. Miller. 1986. Norm theory: Comparing reality to its alternatives. Psychological Review 93:136-53.

Kearns, D. 1976. Lyndon Johnson and the American dream. Toronto, Canada: Fitzhenry and Whiteside Ltd.

Keohane, R. O. 1984. After hegemony: Cooperation and discord in the world political economy. Princeton, NJ: Princeton University Press.

King, G. 1993. The methodology of presidential research. In Researching the presidency: Vital questions, new approaches, edited by G. C. Edwards III, J. H. Kessel, and B. A. Rockman. Pittsburgh, PA: University of Pittsburgh Press.

King, G., R. O. Keohane, and S. Verba. 1994. Designing social inquiry: Scientific inference in qualitative research. Princeton, NJ: Princeton University Press.

King, G., and L. Zeng. 2001. How factual is your counterfactual? Paper presented at the annual Political Methodology Summer Conference, July 19-21, in Atlanta, GA. Available from

Kiser, E., and M. Levi. 1996. Using counterfactuals in historical analysis. In Counterfactual thought experiments in world politics: Logical, methodological, and psychological perspectives, edited by P. E. Tetlock and A. Belkin. Princeton, NJ: Princeton University Press.

Kowert, P. A. 1996. Where does the buck stop? Assessing the impact of presidential personality. Political Psychology 17:421-52.

Krause, G. A., and D. B. Cohen. 1997. Presidential use of executive orders, 1953-1994. American Politics Quarterly 25:458-81.

Lasswell, H. D. 1951. The political writings of Harold Lasswell. Glencoe, IL: Free Press.

--. 1968. The future of the comparative method. Comparative Politics 1:3-18.

Lebow, R. N. 2000a. Contigency, catalysts, and international system change. Political Science Quarterly 115:1-26.

--. 2000b. What's so different about a counterfactual? World Politics 52:550-85.

Lebow, R. N., and J. G. Stein. 1996. Back to the past: Counterfactuals and the Cuban Missile Crisis. In Counterfactual thought experiments in world politics: Logical, methodological, and psychological perspectives, edited by P. E. Tetlock and A. Belkin. Princeton, NJ: Princeton University Press.

Levin, P. L. 2001. Edith and Woodrow: The Wilson White House. New York: Scribner.

Lewis, D. K. 1973. Counterfactuals. Cambridge, MA: Harvard University Press.

Lijphart, A. 1971. Comparative politics and the comparative method. American Political Science Review 65:682-93.

--. 1975. The comparable cases strategy in comparative research. Comparative Political Studies 8:158-77.

Lowi, T. J. 1985. The personal president. Ithaca, NY: Cornell University Press.

Mahoney, J. 2000. Strategies of causal inference in small-N analysis. Sociological Methods and Research 28:387-424.

Margulies, H. F. 1989. The mild reservationists and the League of Nations controversy in the Senate. Columbia: University of Missouri Press.

Masters, R. D. 1989. The nature of politics. New Haven, CT: Yale University Press.

Mayer, K. R. 1999. Executive orders and presidential power. Journal of Politics 61:445-66.

McKeown, T.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, J. S. [1843] 1974. A system of logic. Toronto, Canada: Toronto University Press.

Moe, T. 1993. Presidents, institutions, and leadership. In Researching the presidency: Vital questions, new approaches, edited by G. C. Edwards III,J. H. Kessel, and B. A. Rockman. Pittsburgh, PA: University of Pittsburgh Press.

Moe, T., and W. Howell. 1999. Unilateral action and presidential power: A theory. Presidential Studies Quarterly 29:850-72.

Mueller, J. 1989. Retreat from doomsday: The obsolescence of major war. New York: Basic Books.

Munck, G. 1998. Canons of research design in qualitative analysis. Studies in Comparative International Development 33:18-45.

Neustadt, R. E. 1990. Presidential power and the modern presidents: The politics of leadership from Roosevelt to Reagan. New York: Free Press.

Nisbett, R., and L. Ross. 1980. Human inference: Strategies and shortcomings of social judgment. Englewood Cliffs, NJ: Prentice-Hall.

Park, B. E. 1986. The impact of illness on world leaders. Philadelphia: University of Pennsylvania Press.

Przeworski, A. 1987. Methods of cross-national research 1970-1983: An overview. In Comparative policy research: Learning from experience, edited by M. Dierkes, H. N. Weiler, and A. B. Antal. Brookfield, VT: Gower.

Przeworski, A., and H. Teune. 1970. The logic of comparative social inquiry. New York: John Wiley.

Ragin, C. C. 1987. The comparative method: Moving beyond qualitative and quantitative strategies. Berkeley: University of California Press.

--. 2000. Fuzzy-set social science. Chicago: University of Chicago Press.

Ragin, C. C., and D. Zaret. 1983. Theory and method in comparative research: Two strategies. Social Forces 61:731-54.

Renshon, S. A. 1996. The psychological assessment of presidential candidates. New York: New York University Press.

--. 1998. High hopes: The Clinton presidency and the politics of ambition. Routledge: New York.

--. 2000. After the fall: The Clinton presidency in psychological perspective. Political Science Quarterly 115:41-65.

Roese, N.J., and J. M. Olson, eds. 1995. What might have been: The social psychology of counter factual thinking. Mahwah, NJ: Lawrence Erlbaum.

Rogowski, R. 1995. The role of theory and anomaly in social-scientific inference. American Political Science Review 89:468-70.

Sartori, G. 1970. Concept misinformation in comparative politics. American Political Science Review 64:1033-53.

--. 1991. Comparing and miscomparing. Journal of Theoretical Politics 3:243-57.

Shepard, G.H. 1988. Personality effects on American foreign policy, 1969-1984: A second test of interpersonal generalization theory. International Studies Quarterly 32:91-123.

Simonton, D. K. 1986. Presidential personality: Biographical use of the Gough Adjective Check List. Journal of Personality and Social Psychology 51:149-60.

--. 1988. Presidential style: Personality, biography, and performance. Journal of Personality and Social Psychology 55:928-36.

Skocpol, T. 1984. Vision and method in historical sociology. Cambridge: Cambridge University Press.

Skocpol, T., and M. Somers. 1980. The uses of comparative history in macrosocial inquiry. Comparative Studies in Society and History 22:174-97.

Skowronek, S. 1993. The politics presidents make: Leadership from John Adams to Bill Clinton. Cambridge, MA: Harvard University Press.

Spangler, W. D., and R. J. House. 1991. Presidential effectiveness and the leadership motive profile. Journal of Personality and Social Psychology 60:439-55.

Stalnaker, R. C. 1984. Inquiry. Cambridge, MA: MIT Press.

Stinchombe, A. L. 1978. Theoretical models in social history. New York: Academic Press.

Taylor, A.J.P. 1954. The struggle for the mastery in Europe, 1848-1918. London: Oxford University Press.

Tetlock, P. E. 1998. Close-call counterfactuals and belief system defense: I was not almost wrong, but I was almost right. Journal of Personality and Social Psychology 75:639-52.

--. 1999a. Distinguishing frivolous from serious counterfactuals. Manuscript.

--. 1999b. Theory-driven reasoning about possible paths and probable futures: Are we prisoners of our preconceptions? American Journal of Political Science 43:335-66.

--. Forthcoming. The logic and psycho-logic of counterfactual thought experiments in the rise-of-the-West debate. In Unmaking the West: Alternative histories of counter factual worlds, edited by P. E. Tetlock, R. N. Lebow, and G. Parker. New York: Columbia University Press.

Tetlock, P. E., and A. Belkin, eds. 1996a. Counterfactual thought experiments in world politics: Logical, methodological, and psychological perspectives. Princeton, NJ: Princeton University Press.

--. 1996b. Counterfactual thought experiments in world politics: Logical, methodological, and psychological perspectives. In Counterfactual thought experiments in world politics: Logical, methodological, and psychological perspectives, edited by P. E. Tetlock and A. Belkin. Princeton, NJ: Princeton University Press.

Tetlock, P. E., and R. N. Lebow. 2001. Poking counterfactual holes in covering laws: Cognitive styles and historical reasoning. American Political Science Review 95:829-43.

Tetlock, P. E., R. N. Lebow, and G. Parker, eds. Forthcoming. Unmaking the West: Alternative histories of counterfactual worlds. New York: Columbia University Press.

Thompson, E. P. 1978. The poverty of theory and other essays. New York: Monthly Review Press.

Tilly, C. 1984. Big structures, large processes, huge comparisons. New York: Russell Sage.

Tversky, A., and D. Kahneman. 1974. Judgment under uncertainty: Heuristics and biases. Science 185:1124-31.

Verba, S. 1967. Some dilemmas in comparative research. World Politics 20:111-27.

Weber, M. 1949. Objective possibility and adequate causation in historical explanation. In The methodology of the social sciences, edited by E. A. Shils and H. A. Finch. New York: Free Press.

Weber, S. 1996. Counterfactuals, past and future. In Counterfactual thought experiments in world politics: Logical, methodological, and psychological perspectives, edited by P. E. Tetlock and A. Belkin. Princeton, NJ: Princeton University Press.

Weinstein, E. A., J. W. Anderson, and A. S. Link. 1978. Woodrow Wilson's political personality: A reappraisal. Political Science Quarterly 94:585-98.

Winter, D.G. 1987. Leader appeal, leader performance, and the motive profiles of leaders and followers: A study of American presidency and elections. Journal of Personality and Social Psychology 52:196-202.

Yin, R.K. 1984. Case study research: Design and methods. Vol. 5 of Applied social science research methods series. Beverly Hills, CA: Sage.

Jeffrey M. Chwieroth is a doctoral candidate in the Department of Political Science at the University of California, Santa Barbara. His current research interests include the international political economy and political methodology.
COPYRIGHT 2002 Center for the Study of the Presidency
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2002 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Chwieroth, Jeffrey M.
Publication:Presidential Studies Quarterly
Geographic Code:1USA
Date:Jun 1, 2002
Previous Article:Another lesson about public opinion during the Clinton-Lewinsky scandal. (Articles).
Next Article:Five trends in presidential rhetoric: an analysis of rhetoric from George Washington to Bill Clinton. (Articles).

Related Articles
The Presidency as a Learning Organization.
Toward a Representational Framework for Presidency Studies.
The two presidencies, 1984-98: a replication and extension. (Research Note).
Assessing changing views of the President: revisiting Greenstein's Children and Politics. (Articles).
The secular decline in presidential domestic policy making: an organizational perspective.
2004 presidential election: an introduction.
The evolution of the rhetorical presidency and getting past the traditional/modern divide.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters