«Gregory A. Huber* Associate Professor Department of Political Science Yale University PO Box 208209 New Haven, CT 06520-8209 gregory.huber ...»
Contingency, Politics, and the Nature of Inquiry:
Why non-events matter
Gregory A. Huber*
Department of Political Science
PO Box 208209
New Haven, CT 06520-8209
For Political Contingency: Studying the Unexpected, the Accidental, and the Unforeseen.
Forthcoming, NYU Press, 2007.
Revised Version: June 2006
*I thank Justin Fox, Jacob Hacker, and the reviewers for their helpful comments.
Abstract Contingent events are probabilistic. Acknowledging that realized contingencies alter observed political outcomes, however, does little to advance the systematic study of politics. This paper suggests that political science should focus on understanding how foreseeable contingencies, rather than truly “exogenous” unforeseeable events, alter political behavior. The most useful tool for understanding the effects of these foreseeable contingencies on political interactions is the formal and informal analysis of strategic behavior in the face of uncertainty through the use of game theory. This paper identifies the limitations of using observed contingent events to understand the role of contingencies in politics. More broadly, it suggests that the proper analysis of anticipatory strategic behavior has implications for the allocation of research resources across many topics of interest to political science.
Contingent events are probabilistic. They manifest or do not manifest because of some uncertainty about the future that is unknown or unknowable to human participants. A simple toss of a coin, for example, will produce an observed outcome of “tails” about half of the time. The contingent outcome “tails” following a coin toss is, prior to the toss, a contingency. How should political science incorporate the fact that important political interactions are embedded in situations where outcomes are unknown prior to their occurrence?
The problem political scientists face in confronting contingency is particularly difficult for two reasons. First, rarely are we interested in contingent events for their own sake. Whether a political leader survives an assassination attempt is interesting, in most lines of political inquiry, because we believe that who leads matters and not because the continued persistence or death of that particular leader is itself of interest. We as political researchers are therefore interested in the downstream effects of potential and realized contingencies. In contrast, in gambling or operations research, one might care only about the contingent event itself (does the coin come up heads, or how frequently do we need to test a product to ensure that it is of high quality 99.95% of the time?). Political scientists must therefore incorporate contingency as a mediating factor in the political outcomes they study.
Second, political actors are strategic. As such, manifestations of contingencies as contingent outcomes are shaped by the choices of political actors and themselves shape the behavior of strategic actors in anticipation of potential contingent events. For example, political leaders in parliamentary systems may call elections now rather than in the future if the economy is doing well and they fear the contingent outcome brought about by the combination of a surprise future economic downturn and retrospective voting behavior. Thus, elections may be more likely to take place during good economic times than bad ones. Similarly, wise leaders avoid starting wars with enemies much more powerful than themselves. While there is uncertainty about the contingent outcome of any military interaction, strategic actors nonetheless anticipate the likely contingent outcome (defeat) and plan accordingly. In contrast, those studying the behavior of coins or the quality of production can safely assume that their subjects do not change their behavior strategically. Human decisions, unfortunately, are substantially more complicated.
If one accepts the basic notion that contingencies matter, both because they shape subsequent outcomes and anticipatory behavior, what should political science do to grapple with and manage this uncertainty? My argument here is twofold. First, not all contingencies are the same. In particular, there are some events that are fundamentally unforeseeable ex ante. These events are rare, however, and it is my contention that they should not be the focus of our research. Rather, political science should make primary those foreseeable contingencies around which strategic political actors maneuver. Foreseeable events, like unforeseeable events, are fundamentally probabilistic. Nonetheless, they matter because political actors have beliefs about the probabilities that they will manifest, and these beliefs shape the strategic interactions among political actors.
Consequently, my second argument is that political science already has the capacity for incorporating this second form of contingency, knowable uncertainties, into analysis of strategic behavior. The way to account for this form of contingency, however, is not to study the observed probabilistic manifestations of contingent outcomes, but instead to study how strategic actors behave in light of those potential contingencies. The most useful tool for understanding the effects of these foreseeable contingencies on political interactions is the formal and informal analysis of strategic behavior in the face of uncertainty through the use of game theory.
The remainder of this paper is organized as follows. I first review different types of contingency and make the case for focusing on those contingent events about which strategic political actors may maneuver. Next, I demonstrate why and how the analysis of anticipatory strategic behavior should lie at the heart of the study of contingency in politics. Third, the paper identifies the limitations of using observed contingent events to understand the role of contingencies in politics and suggests that analysis of anticipatory strategic behavior has implications for the allocation of research resources across many topics of interest to political scientists. Finally, I conclude by addressing several potential counter-arguments to my position.
Forms of Contingency and Their Study All contingencies are not the same. In particular, it is analytically useful to distinguish those contingencies about which we know something, anything, and those of which we are unaware. 1 A useful label for the latter category is unknowable and unnatural disasters—things that are so out of the ordinary that we do not even believe they could occur. (An extreme example would be the universe simply ceasing to exist. A more realistic example is of a hurricane striking in the time before weather forecasting and the maintenance of historical records.) Because these events are fundamentally external to our immediate cognitive environment, they cannot affect, ex ante, our strategic behavior.
Clearly these types of contingencies matter—an unforeseeable hurricane kills many people. But because these events are not even considered by strategic actors prior to their For a different categorization of contingencies, see Schendler (this volume). Contingencies of which we are initially unaware are catastrophes in Schendler’s typology, although, as I suggest below, many catastrophes are foreseeable. I group Schendler’s remaining categories of contingencies together as known contingencies.
occurrence, generalizing about the effects of unexpected hurricanes or other fundamentally external influences on human behavior is difficult and of limited value. Unforeseeable events are singular occurrences. In societies with written records, a hurricane is only an unforeseeable event once. Thereafter individuals could choose to alter their behavior in anticipation of a hard to predict surprise storm. Unforeseeable events are also generally rare and unique, so inferring from them the effect of other unique events is difficult. In short, the first flood is likely to have different effects than the first hurricane, etc.
Three objections might be raised to the argument that unforeseeable contingencies are best avoided by political science. One might claim, for instance, that if we are really interested in how humans behave in response to becoming aware of new contingencies, realized unforeseeable contingencies are interesting. The experience of the French Revolution, for example, has large effects on subsequent democratization efforts. But once the contingency of mob rule manifests the first time, it become foreseeable and is best approached in the manner described below. (Alternatively, one could go further and argue that strategic participants knew about the risk of mob rule, but simply revised upward their belief about its likelihood after observing it in the French case.) Alternatively, one might state that there are many interesting circumstances in which some of the participants in a strategic interaction have knowledge of contingencies but others do not. In insider stock trading cases, for example, corporate insiders might know that a company is about to be purchased whereas ordinary stockholders do not. These are not really examples of unforeseeable events, however. Instead, they are cases in which there are ordinary informational asymmetries. These example abound, as when individuals who have better knowledge of their own ideology and competence run for office before voters who know relatively little about the candidates.
Finally, one could argue that unforeseeable events are common and important. In designing public policy, for example, there are numerous instances of “unintended consequences” that have yielded dramatically different outcomes from those sought by policymakers. My argument, however, is that few of these cases are actually unforeseeable contingencies. Rather, strategic policymakers generally make decisions that yield outcomes that were considered, but were dismissed as unlikely to occur. Of course, it is interesting that policymakers make decisions that are, ex post, clearly badly informed. But, to presage the argument below, it is my claim that these sorts of strategic decisions are where political scientists should focus their attention rather than beginning with an analysis of the effects of these decisions. (Furthermore, as is explained below, if one does not consider the strategic sources of bad decisions, empirical analyses of the effects of these decisions are likely to be biased.) So, if we instead focus our research efforts on foreseeable events, what are the implications for the study of politics? Prior to addressing this question directly, it is useful to note that even relatively rare events may influence strategic human behavior and thereby the outcome of realized contingencies. For example, serious earthquakes are rare in San Francisco, but the city has nonetheless chosen to impose stringent building codes. Earthquakes are even more rare in Chicago, but they still occur. In the event of a major earthquake in either city, however, individual buildings in Chicago are likely to fare less well because the city has not adopted stringent building codes to protect its citizens from the realized contingency of an earthquake. That is not because earthquakes are unknown, but because mandating earthquakeproof buildings is not seen as worth its relative cost (wisely or not). Similarly, six inches of snow in Atlanta brings the city to its knees while Hartford shrugs because it has invested in the technology necessary to cope with the more-frequently anticipated contingency of snow.
Preparing (or not) for earthquakes and snow are strategic decisions in anticipation of foreseeable contingencies.
Whether to prepare for natural disasters is representative of the numerous cases in which individuals make strategic choices in light of their beliefs about the likely effects of these choices on desired outcomes. This, I claim, is the realm where political science is likely to find the most success in building generalizable models of human interaction. Rather than asking what effect did the assassination of Yitzhak Rabin have on the Middle East, a far more useful endeavor is to focus on the more general question of what the problem of potential assassinations has on the behavior of government leaders.
How to Confront Contingency: The power of anticipatory action In this line of reasoning, it is short step from being assassinated to losing an election. Both outcomes matter (and not only for the incumbent). How should political science internalize these contingencies? Here, I answer this question by way of an extended discussion of a particular area of political science research. One of the fundamental questions of politics is how the practice of selecting government officials alters their behavior. Taking this question more narrowly, one area of ongoing research focuses on debates about the appropriate method for selecting judges. If one is to make arguments about the (in)appropriateness of judicial elections, however, it is necessary to have a theory that links judicial behavior to systems of judicial selection. How should one do this?
One approach is to seek out what appear to be interesting “contingencies” surrounding judicial elections. In other words, one might look for historical events where the practice of electing judges appears to matter. One might imagine that such an approach to this topic could
lead one to discover the following (hypothetical) historical narrative:
Judge Smote ascended to the bench after beating the incumbent, Judge Light, in a bruising electoral contest that focused on Judge Light’s earlier sentencing behavior.
Judge Smote’s campaign advertisements focused on a case in which Judge Light sentenced David Randall, who was convicted of indecent exposure, to a probationary sentence in lieu of time in prison. While on probation, Mr. Randall was arrested for exposing himself to a young child. In an indecent exposure case now before the court, Judge Smote assigned the defendant John Davis the maximum prison sentence allowed by law.
One could ask many different questions in light of this historical incident, but for my purposes I wish to focus on one potential question and how it reveals the limitations associated with using realized contingencies as explanations for politically relevant outcomes. Specifically, one could imagine political scientists interested in the effects of judicial elections raising the question of “what explains the harsh sentence given to the Mr. Davis?” Taking as given Mr.