Monadic Sender Reputation and Sanctions Outcomes
4.2 Research Design
4.2.3 Key Independent Variable: Sender Reputation
If a threat is to be effective, it must be sufficiently credible. The target state must believe that the sender is both willing and able to impose the threatened sanctions should the target refuse to comply with its demands. The central hypothesis of this dissertation is that targeted states rely on the sender state’s previous history of keeping its conditional commitments as one indicator of threat credibility. From this argument, I derive the prediction that targets will be more likely to acquiesce to a sanction threat if the sender has a stronger record of carrying out its threats; that is, when the sender has a reputation for making credible threats. Targets will be less likely to acquiesce if the sender has a reputation for backing off when its threats are resisted.
In the following pages, I outline the construction of the key independent variable that will be used to test Hypothesis 1a, which relates threat effectiveness to a states monadic and diffuse reputation for carrying out threats. States are assumed to acquire reputations that are monadic in the sense that potential targets draw conclusions about the sender’s type from all past sanctions cases, regardless of whether they were the target in these previous confrontations or merely non-participant observers. Sender reputations are assumed to be diffuse in the sense that targets apply inferences drawn from past sanctions cases to all subsequent confrontations involving the same sender, regardless of whether past and future cases involve the same types of issues or stakes. In short, the broadest version of this predictor draws on the entirety of the sender’s record of threat enforcement up until the year of the current confrontation. The basic method of constructing this measure, Sender Reputation, will also be used in the following chapters, which test more narrowly conceived versions of the reputation argument.
Private information about the sender’s readiness to impose sanctions is revealed most clearly whenever the sender encounters resistance from a targeted state; that is, whenever the sender is called upon to go ahead with the sanctions or back down. Keeping this in mind, I create a measure of reputation for resolve for each sender state in the data set, based on the outcomes of all sanctions episodes which were initiated by that sender against any target. The measure combines two pieces of information: the number of cases in which a sender state issued a sanction threat that was then rejected by the targeted state, and the number of cases in which the sender subsequently imposed economic sanctions on recalcitrant targets. For each primary sender state in the data, I create annual running counts of threats issued and resisted and sanctions imposed, respectively. The process of creating the running counts
begins with identifying the outcomes of all prior sanctions cases initiated by a particular sender. Here, I take advantage of the coding of the outcomes of sanctions cases in the TIES data. The TIES data distinguishes between five types of outcomes each for cases that ended without sanctions being imposed and for cases that ended after sanctions were imposed: complete or partial target acquiescence, negotiated settlements between sender and target with mutual concessions, stalemates, or sender withdrawal from the case.
Sanctions cases are classified as instances of threats issued and resisted whenever TIES records that the case ended at the threat stage with a stalemate or with the sender backing down or that the case ended after sanctions were imposed by the sender, regardless of the nature of the subsequent outcome of the case. More to the point, these are cases where the targeted state resisted the sender’s pressure to make the demanded concessions outright, or agreed to some negotiated settlement. Sanctions cases are classified as instances of sanctions imposed whenever TIES indicates an outcome where sanctions were implemented.
The reputation score for each sender in year t is then generated by taking the proportion of the total number of sanctions imposed over the total number of threats issued and resisted as of year (t-1). This score changes every time a sanctions episode ends; I assume that this is the point when the sender’s choice between standing firm and backing down from the threat becomes clear to the current target, and more importantly, to the international audience of potential future targets as well. Reputation scores are bounded by 0 and 1. A value of 0 denotes that up to the current sanctions episode, the sender never followed through on its threats, while a value of 1 indicates that the sender has implemented threatened economic sanctions whenever the targeted state refused to acquiesce. The more consistently a sender state has carried out its threats in the past, the higher its reputation score in the current year.
For illustration, Figure 4.1 presents graphically the historical trajectory of the reputation scores for four of the most active sender states in the data set.
Figure 4.1: Enforcement Records of Major Sanctions Users.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 Year E n fo rcem en t R ec ord
US Canada Great Britain USSR/Russia
A complication arises for the first observation for each sender state in the data. The TIES data only contains sanctions cases that were initiated after 1970. As a result, sanctions cases that took place before that year did not enter into the calculation of the reputation scores. This left-censoring in the data in many cases explains why states enter the data set without prior sanctions records, although they may have used economic sanctions prior to this date. Given the construction of the reputation score just outlined, when a sender is coded as
initiating its first sanction threat, its reputation score is coded as missing because there are no prior observations for this sender to generate a value.24
Missing values are a conceptual as well as a methodological problem. In the statistical analysis to follow, observations with missing values would be lost due to list-wise deletion resulting in the loss of a potentially large number of sanctions cases from the sample. This outcome can be avoided by assigning senders a meaningful ‘initial’ reputation score but raises the question of what this score should be. I choose a starting score of 0.5 for all senders in the data for the following reason.
Consider the information embodied in the reputation score created here. A reputation score of 1 is assumed to be associated with a high level of credibility being ascribed to a sender. Based on what could be learned from the sender’s past record, the target can expect that the sender will very likely follow through on its current sanction threat. A score of 0 denotes the opposite: The target can expect that the sender will be unlikely to impose
24The lack of data points prior to 1970 means that I implicitly assume senders have not threatened sanctions
before 1971 and have thus not formed reputations for carrying out their threats, which in turn makes necessary the imputation of ‘initial’ reputation scores described above. In principle, this is a cause for concern: the reputation scores ignore a large segment of history in which states threatened and imposed sanctions, the United States in particular. The use of post-1970 data alone may at best limit our ability to generalize from the results of this study and at worst diminish our confidence in the findings. Future data collection can alleviate these concerns by expanding the temporal domain of the TIES data set to the immediate post-WWII era. In practice, however, there are a number of reasons to suggest that left-censoring may not be as big a cause for concern. First, the 1960s and 1970s marked a number of significant changes in the international state system. Decolonization added many states to the international arena which became users (and often targets) of economic sanctions around the time the TIES data begins to code cases. These states indeed had little or no prior history of sanctioning. Second, the world economy changed significantly during the 1970s. International economic interactions grew dramatically in volume and complexity, providing greater opportunities for economic leverage but also generating greater constraints on its use. If the period after 1976 constitutes a very different context for economic statecraft than the previous decades, focusing the analysis solely on episodes that began in and after 1971 is justified and can produce more meaningful findings than an analysis that includes cases that took place earlier. Finally, threats and impositions of economic sanctions spiked in the early 1990s (see Figure 1.1) and these cases make up a large portion of the sample. By the year 1990, many senders in the sample had accumulated a record of sanctions use and their imputed initial reputation scores had been updated to more accurately reflect their sanctions behavior. In sum, although TIES covers a relatively short period of thirty years, the data set captures a time period of increasingly active sanctions use in an economically interdependent state system. It is therefore a reasonable place to start the type of inquiry undertaken in this dissertation.
sanctions. Finally, a score of 0.5 corresponds to the sender being expected to go either way that is, impose or back down with equal probability. This is the type of judgment one could reasonably expect in the absence of any information about past behavior.25 The score is then updated as soon as the first sanctions case for that sender produces the required information.
One important objection to assigning this value is that there is another way in which a sender state can acquire a reputation score of 0.5 and it is by no means obvious that target decision makers will draw the same conclusions in each case. A sender state may have threatened sanctions and been rebuffed four times, but only proceeded to implement sanctions in two of these cases. Here, I implicitly assume that such a state will be regarded the same as a sender without prior record of making sanction threats. This assumption could be especially problematic if past sanctions cases indeed reveal more private information about the sender than just its propensity for bluffing. At the same time, it is not clear that this situation could be remedied by assigning a different value to first-time sender states. Although it is an imperfect solution, assigning a score of 0.5 most closely captures the notion of being able to guess no better than chance without the benefit of historical lessons, ceteris paribus, and is a statistically conservative fix that preserves a number of observations in the sample that would be lost otherwise.