Abstract
Using evidence from voting in the UN General Assembly (UNGA), this article shows that leader turnover, especially in small coalition, nondemocratic systems, increases the likelihood of policy realignment. Autocrats who are beholden to only a small proportion of the population represent the foreign policy interests of their small number of supporters. When leader turnover occurs, the interests represented often shift too and this results in an increased volatility and regression toward a neutral position of a nation’s alignment at the United Nations vis-à-vis the United States. While such realignments can offer an opportunity to reduce enmity between states, they can also signal growing differences between friends. The impact of leaders change in large coalitions produces more moderate shifts in alignments.
In 1979, Shah Mohammad Reza Pahlavi of Iran fled office in the mid of a revolution and was replaced by Ayatollah Ruhollah Khomeini. With this change came a serious souring of United States–Iranian relations. The United States lost another ally in 1979. In Nicaragua, President Anastasio Somoza Debayle’s regime fell to Sandinista (Frente Sandinista de Liberación Nacional [FSLN]) revolutionary forces. In both cases, a supportive loyal friend to the United States became an intractable enemy. Figure 1 shows these shifting friendships graphically. The solid line plots the yearly affinity scores (similarity of voting) between United States and Iranian voting in the UN General Assembly (UNGA). The dashed line corresponds to United States–Nicaraguan voting affinity. A full description of these data follows later. The vertical lines correspond to leader change, which occurred in 1979 and 1989 in both Iran and Nicaragua.
UN General Assembly (UNGA) voting affinities of Iran and Nicaragua with the United States.
Although these changes represent extreme examples, they are both part of a general pattern. Nondemocratic states do not have a persistent national characteristic with regard to their interstate relations; rather, the interests that get represented in foreign affairs depend upon individual leaders and the preferences of the coalitions that back them. Such representation means that when nondemocratic leaders change, there is often a significant shift in interstate relations. These shifts can be seen dramatically in Figure 1 with the decline in affinity voting scores in 1979. Of course, not all leader change is accompanied by such large and negative impacts on relations. In 1989, leader change again took place in both Iran and Nicaragua and was accompanied by a much smaller, but positive, shift in relations with the United States.
Although leader change in nondemocratic systems often shifts interstate relations, the impact of leader change is much more subdued in democratic systems. At least to the extent that the affinity of UNGA voting measures interstate relations (Dixon 1981; Voeten 2013), leader change has no systematic effect on the relations between a democratic nation and the United States.
Neorealist theory (Waltz 1979) suggests foreign policy relations are shaped by persistent national goals, but this appears to be repudiated by the evidence presented here that foreign policy shifts in response to leader change. Following directly from the theory of McGillivray and Smith (2008), I argue that foreign policy alignments depend upon which domestic interests are privileged. Nations contain many domestic groups and leaders do not need support from all of these groups to retain power. Office seeking leaders reward the supporters they need. Different groups want different policies. When leaders change, there is also often a shift in which groups form core supporters. When the support base changes, the set of interests promoted through foreign policy also changes. Such shifts in alignment can increase confrontation between nations that increases the risk of conflict. Political institutions moderate the extent of these shifts in foreign policy. Leader change in a democracy is less likely to lead to a change in foreign policy and therefore is less likely to lead to dispute.
Political leaders who require the support of a large proportion of society to retain power, democrats for instance, cannot afford to privilege one small group at the expense of the masses. Leaders in such systems tend to implement moderate policies that appeal to large centrist groups. Given the need to appeal to many people, democratic leaders implement moderate broad-based policies. As a consequence, the foreign policy shifts associated with democratic leader change are relatively small. In contrast, in autocratic societies, leaders need the support of only a relatively small number of supporters. The loyalty of these supporters is best bought by intensely rewarding them with the policies they like even if such rewards come at the expense of the masses. In small coalition systems, foreign policy intensely rewards one small group, and the wants of this small group need not be representative of society as a whole. Leader change in such systems can lead to a different group being intensely rewarded, which in the context of foreign policy is seen as a sharp shift in interstate relations.
In the context of UNGA voting, this article finds just such patterns. The similarities in UNGA votes between United States and democratic nations remain relatively stable when leader turnover occurs. In contrast, when leader change occurs in a nondemocratic state, there is often a significant shift in that nation’s UNGA voting record. Nondemocratic leader change results in increased volatility in UNGA vote affinities with the United States and regression toward the mean: nations highly aligned with the United States are likely to become more neutral, those nations with strong enmity to the United States are likely to experience some warming of relations in the wake of leader change, and in neutral states the changes in foreign policy alignment toward or away from the US position tend, on average, to cancel out.
UNGA Voting Affinity
There is a substantial tradition of using UNGA votes to ascertain the important dimensions of international relations, voting blocs, and alignments (Alker 1964; Alker and Russett 1965; Gartzke 1998; Holloway 1990; Holloway and Tomlinson 1995; Kim and Russett 1996; Lijphart 1963; Voeten 2000, 2004, 2013). The basic idea is that if two nations vote the same way on issues in the UNGA, then their interests are aligned. Voeten (2013) provides an excellent history of the study on UNGA voting behavior. Scholars use the closeness of UNGA voting records to address numerous issues. For instance, Dreher, Nunnenkamp, and Thiele (2008) examine how US aid buys policy compliance. Others look at the relationship among UNGA voting, aid, and World Bank and International Monetary Fund (IMF) programs (Barro and Lee 2005; Dreher and Jensen 2007; Dreher and Sturm 2012; Rai 1980; Stone 2002; Thacker 1999). Relations between UN Security Council membership and aid and support from other international organizations have also been explored (e.g., Dreher, Sturm, and Vreeland 2009).
Leader Change, Institutions, and Policy Shifts
The innovation in this article is to test how leader change affects the alignment of nations as measured by UNGA voting affinities. The theoretical basis comes from ideas advanced by Fiona McGillivray and coauthors (McGillivray and Smith 2004, 2005, 2008, 2009; McGillivray and Stam 2004; Bobick and Smith 2013) that utilize the selectorate theory of political institutions (Bueno de Mesquita et al. 2003). As these ideas have been well developed elsewhere, I briefly summarize the theory and move quickly to the empirical tests.
The need to continually refer to affinities in UNGA voting with the United States would make sentences extremely clumsy. Hence, for ease of language, I often describe UNGA vote patterns as alignments and discuss them in terms of friendship, neutrality, or enmity. These phrases should be taken as indications of high, moderate, or low vote affinity scores rather than as literal statements about security relations between states. For similar reasons, I often refer to institutions as democratic and autocratic when the theoretical basis for comparing institutions is selectorate theory, as introduced next.
Selectorate theory argues leaders are predominately driven by the goal of retaining office. Institutions determine the ease with which leaders achieve this goal, and how they best achieve it. Selectorate theory classifies institutions according to the number of supporters a leader needs to retain power, the winning coalition (W), and the size of the pool from which these supporters can be drawn, the selectorate (S). Although these dimensions are inherently continuous, common regimes typologies can be mapped into it. Democracies, for instance, are systems with large selectorates and large winning coalition, although the precise size of W and S depend upon electoral rule and suffrage. In contrast, monarchies and juntas have small selectorates and small winning coalitions. Autocrats also tend to rely upon small coalitions to retain power.
Winning coalition size determines the nature of public policy. When a leader relies on only a small number of supporters, he or she can use state resources to greatly enrich them individually. This option is not available to a democrat or other leader beholden to a large coalition. When divided over so many supporters, each individual’s share of resources is too small for this to be an efficient means of buying political support. Democrats better enhance their survival by providing public goods that reward everyone. It is for this reason that democracies have cleaner environments and better public health than autocracies.
Foreign policy is driven by the same institutional incentives as domestic policy. A small coalition leader promotes policies that reward the small select group of supporters that keep him or her in power, even if these policies are contrary to the wishes of the vast majority of the citizens. Democrats cannot ensure loyalty so easily. A large coalition leader needs the support of many people and so must focus foreign policy provisions on public goods that reward the many rather than private goods that reward the few. There is considerable evidence for the influence of institutions on foreign policy (see, for instance, Bueno de Mesquita and Smith 2007), what is central to the current article is the impact of leader change on the stability of foreign policy.
Following the deposition of the Shah and Somoza, there was an abrupt shift in who supported the regime and, as a consequence, what kind of policies they wanted. The Shah’s supporters valued close relations with the West. Ayatollah Khomeini’s supporters were mainly clerics who wanted a theocratic state and regarded the West with distain. The Shah and Khomeini’s policies were vastly different. Both policies were designed to intensely reward a small group of loyal supporters, but since the supporters’ wants differed, so too did the policies.
The Iranian and Nicaraguan cases illustrate violent coalitional change, but shifts in the coalition are common phenomena following leader change (Bueno de Mesquita and Smith 2011, chap. 3). Leaders reshuffle their coalition to replace supporters of uncertainty loyalty with their loyal friends and cronies. As the supporters change, so too do the policies they want. The degree to which policy changes depends upon the political system. In a small coalition system, the interests of one small set of supporters can be very different from the interests of another, and both can differ greatly from broader societal concerns. Therefore, a change in leadership can result in a large shift in policy. In contrast, when coalition size is large, it is harder for leaders to assemble coalitions with such mutually exclusive interests. Although democrats, like autocrats, reward their supporters at the expense of society as a whole, the extent to which they do so is much less as in democracy most coalitions require the inclusion of broad centrist interests and given the numbers of people involved leaders predominantly rely on public goods rewards that benefit everyone rather than special benefits for the lucky few. Leader change in large coalition systems results in coalitional shifts that produce differences in the foreign policy interests that leaders want to favor, but these shift are moderate.
Leader change induces foreign policy shifts, and these shifts are larger and more intense in small coalition systems than in democracies. These predictions have already been tested with strong support in the context of trade, sovereign debt, and economic sanctions (McGillivray and Smith 2004, 2008; McGillivray and Stam 2004). The tests here show they are also borne out in the context of interstate relations as measured by UNGA voting affinities. Leader change in small coalition systems leads to a shift in alignment with the United States. In addition to increasing the volatility of relations between nations, leader change leads to regression toward to a neutral position.
When leaders change, the interests that are represented also change. These new interests might be more in line with US positions or that they might be more opposed. However, on average, such leader shifts tend to result in a more neutral position. As will be described subsequently, UNGA voting affinities are measured on a −1 to 1 scale. Although affinities tend to be relatively centrally clustered, suppose that affinities are uniformly distributed on this interval. That they are centrally clustered only makes the regression to the mean argument stronger. Suppose that a nation is initially aligned strongly against the US position, say, at −.75, and that leader change results in redrawing a new affinity. Although the nation could shift to an even more anti-US position, the chance that it does so is only one in eight. Indeed, the chance that its new alignment is closer to a neutral position than the initial position is 75 percent. Likewise states strongly aligned with the United States are likely to become less strongly aligned following leader change. Leader change increases the volatility of relations between states and the larger such shifts are likely to be, the more likely they are to result in regression toward a neutral position. Being beholden to only a small number of supporters, for small coalition leaders the average difference between the interests represented by one leader compared to the next is greater than the comparable difference for leaders from large coalition systems. As a consequence, leader change in small coalition systems results in greater foreign policy volatility and regression toward neutrality than in large coalition systems.
Data and Methods
The analyses require yearly assessments of the similarity of UNGA voting for each nation compared with the United States. These measures were operationalized as S-affinity scores (Signorino and Ritter 1999) and were obtained from Eric Gartzke’s webpage (http://dss.ucsd.edu/∼egartzke/) for the years 1946 to 2008. 1 In broad conceptual terms, these scores assess how often a nation’s vote is the same as the United States’ compared to how often it vote YES when the US votes NO, or vice versa. 2 In principle, this variable, called Affinity, can vary between −1 and 1; however, as a practical matter, 90 percent of the data lie between −.66 and 0.6. The core of the analysis is to examine how leader change and political institutions affect how Affinity varies relative to its value in the previous year, LagAffinity. Figure 2 shows kernel density plots for the distribution of UNGA voting affinities of nations with the United States. 3 The solid line shows the distribution of affinities for democratic nations. The dashed line corresponds to nondemocratic nations. As the figure clearly shows, nondemocratic nations are more likely than democracies to be aligned against the United States (Gartzke 1998).
The distribution of UN General Assembly (UNGA) vote affinities.
The United States is chosen as the base nation against which to compare every other nation’s foreign policy position. This choice is made for both theoretical and pragmatic reasons. At a theoretical level, US policy is expected to be relatively stable, as it is a large coalition system. In addition, being the hegemonic nation, the United States is likely to care about most votes. A smaller nation might be less likely to care about votes concerning nations outside its sphere of influence and as such be a much poorer reference case. Of course, it is unreasonable to assume that the policy positions represented within the United States remain completely constant.
Affinities between nation A and the United States can change because nation A shifts policy, which is the effect I want to capture in the analyses, or because the US shifts policy. Obviously, US policy is not completely rigid, so I adopt several controls for shifts in US policy. First, many of the analyses include a control for the average shift in affinity between nations and the United States. This variable captures common shifts by the United States. I also include a variable for leader change within the United States. The second approach is to include year or region-year fixed effects. This controls for any year-specific shifts by the United States as well as any other idiosyncratic factors that occur in a particular year. Obviously, the average affinity shift cannot be included in this second set of analyses, as it is constant within each year. The results are robust to the inclusion of a variety of fixed effects. To illustrate this, the results are presented with a variety of different fixed effects: years, region-years (based on the Correlates of War [COW] regional codes), nations and, finally, year and nation fixed effects.
Winning coalition size is measured using Bueno de Mesquita et al.’s (2003) five-point scale using data from Polity IV (Marshall, Gurr, and Jaggers 2012) and Banks and Wilson’s (2013) data. Their index of coalition size contains four components that reflect the inclusiveness or noninclusiveness of the system: REGTYPE, XRCOMP, XROPEN, and PARCOMP. The variable REGTYPE refers to regime type and is coded as 2 for military regimes and coded as 3 for military–civilian regimes. Since coalitions in military regimes are formed around a small group of military elites, a military regime is indicative of a small coalition. The BdM2S2 index W receives one point if REGTYPE is not coded as 2 or 3.
The variable XRCOMP measures the competitiveness of executive recruitment. This variable is coded as 1 when the chief executive is selected by heredity or in rigged, unopposed elections. Such rules are indicative of leaders being dependent upon only a small number of supporters. In contrast, higher values (2 or 3) of XRCOMP indicate a dependence on a greater number of supporters. When XRCOMP equals 2 or 3, W receives an additional point.
The openness of executive recruitment, XROPEN, contributes an additional point to W if the executive is recruited in a more open setting than heredity (i.e., the variable’s value is greater than 2). Executives who are recruited in an open political process are more likely to depend on a larger coalition than are those recruited through heredity or through the military.
Finally, one more point can be contributed to the index of W if PARCOMP, competitiveness of participation, is coded as a 5, meaning that “there are relatively stable and enduring political groups which regularly compete for political influence at the national level” (Polity II, p. 18). This variable is used to indicate a larger coalition on the supposition that stable and enduring political groups would not persist unless they believed they had an opportunity to influence incumbent leaders; that is, they have a possibility of being part of a winning coalition. The indicator of W is then divided by 4 to create a five-point scale for W taking the possible values 0, .25, .5, .75, and 1.
The leader data are taken from Archigos (Goemans, Gleditsch, and Chiozza 2009). These data code when leaders enter and leave office. Rather than simply code whether leader change occurs in a particular year (which incidentally produces similar results), I adopt the coding system proposed by McGillivray and Smith (2008). They note the effect of leader change on annual data should be modified according to when it happens within the year. For instance, should leader change occur in January 1990, they assign the full effect of the leader change to 1990. However, should the leader change occur in December 1990, it appears unreasonable that the majority of events in 1990 were influenced by this change. Instead, the impact of leader change is likely to be felt in 1991. In this latter case, they assign the leader change variable 1/12 for 1990 and 11/12 for 1991. Had the change occurred in July 1990, the leader change variable would have been coded as ½ for both 1990 and 1991. The coding of leader change becomes somewhat more complicated when there are multiple leader changes in close succession. The underlying rule is to ask what proportion of each year falls within twelve months following any leader change event with the coding scheme set such that even if multiple leader changes occur, the leader change variable can never exceed 1 for any year (see McGillivray and Smith 2008, 113-14, for details). This variable takes values 0, 1/12, 2/12, 3/12,…, 1.
As a robustness check, I also construct analogous leader change variables looking at only leader change coded as irregular or foreign leader change in the Archigos data. These variables correspond to leader change as a result of nonconstitutional change, such as the revolutionary events that led to the deposition of the Shah and Somoza, and leader replacement by a foreign power, such as the United State’s replacement of Saddam Hussein in Iraq in 2003.
The models include controls for size (log[population]) and wealth (log[per capita Gross Domestic Product]), both drawn from the World Bank’s World Development Indicators. Table 1 provides summary statics for the main variables.
Summary Statistics of Key Variables.
Note: n = 5,956, UNGA = UN General Assembly.
Results
The empirical tests examine how voting affinities change as leaders turnover. The first models are regressions with the dependent variable Affinity. Not surprisingly, LagAffinity is the principal explanatory variable: policy this year is similar to policy last year. Of central interest with regard to testing the theory, the models show that leader change in nondemocratic nations leads to regression toward neutrality. That is, after leader change, an autocracy that was strongly aligned with the United States becomes less closely aligned. At the other end of the scale, leader change often reduces enmity between United States and autocracies. Having examined the direction of change in affinity score, I then examine the variance in affinity measured as (Affinity − LagAffinity)2. Leader change in nondemocracies leads to greater volatility in UNGA vote affinities. The final set of analyses put these two effects together to assess how leader change affects both the direction of change and the volatility of UNGA vote affinities.
Table 2 contains three regression results that all show similar results. Model 1 is a simple ordinary least squares (OLS) regression. Model 2 includes country fixed effect, and model 3 includes region year fixed effects. The dependent variable in each case is UNGA vote affinity with the United States. As we would expect, the best predictor of UNGA voting patterns is voting in the previous year, as seen by the highly significant 0.943 coefficient estimate in the LagAffinity variable.
The Effect of Leader Change on Affinity of UNGA Voting.
Note: Standard errors in parentheses. GDP = gross domestic product.
*Significant at 5 percent. **Significant at 1 percent.
The models include four variables that relate to leader change, these are Leader Change, W × LC, LC × LagAff, and W × LC × LagAff. These variables result from interacting leader change (LC) with political institutions (W) and previous alignment (LagAff). Neither LC nor its interaction with W significantly affects vote affinity. This is to say, for previously nonaligned nations, leader change does not result in a systematic alignment for or against the United States whether in democracy or autocracy. However, in nondemocratic states aligned either for or against the United States, on average, leader change reduces the strength of this alignment.
To gauge the effect of leader change across different political systems, consider the effect of leader change in a small winning coalition system (W = 0). If this nation had previously been strongly aligned with the United States, say with an affinity 0.6 (ninety-fifth percentile in the Affinity score distribution) then leader change would, on average, reduce Affinity by about .05, or one-seventh of a standard deviation of Affinity. Had this nondemocratic state been aligned against the United States, then leader change would likely result in a similar magnitude improvement in relations.
Leader change in democratic systems (W = 1) has no systematic effect on alignments. The sum of the coefficient estimate for the variables LC × LagAff + W × LC × LagAff, (.007 in the case of model 1) is statistically indistinguishable from zero. On average, leader change in democratic systems leaves alignment vis-à-vis the United States unchanged. This result contrasts with Potrafke (2009), who found shifts in UNGA voting associated with changes in government ideology in twenty-one Organisation for Economic Co-operation and Development (OECD) nations, especially when there was a Republican US president. His findings show that left-wing governments are more likely to vote in opposition to the United States than right-wing governments. Unfortunately, comparable ideological measures for nondemocratic nations are not systematically available, so it is impossible to test his finding across the domain of nations studied here.
Models 1 and 2 contain variables for leader change in the United States and average shift in affinity across all nations for each particular year. Change in alignments can occur because either the nation moves its voting position or because the United States moves its voting position. The average shift in alignment captures any shift by the United States, and the coefficient estimates in both models are close to one, as is to be expected. Changes in US leadership appear to have little impact on alignments.
The results in Table 2 suggest leader change produces a small regression toward the mean effect on nondemocratic systems but has no effect on democratic systems. The two examples used to illustrate the argument at the start of the article, Iran and Nicaragua in 1979, exhibited much larger effects. One possible explanation for the large magnitude in the affinity shift is that the deposition occurred by revolutionary means. However, there is little systemic evidence to support this prediction. In the Appendix, I present similar analyses including leader change by irregular means and foreign deposition, as coded by Goemans, Gleditsch, and Chiozza (2009). These analyses exclude interactions with W since such deposition predominately occurs in nondemocratic systems. Given the lack of support for affinity shifts being driven mainly by irregular change, the remaining results focus predominately on leader change by any means.
The principal finding shown in Table 2 is that there is regression toward neutral affinity when leader change occurs in nondemocratic system. Although regression allows us to detect that, following leader change, on average friends of the United States become slightly less friendly, relations with enemies warm slightly, and neutrals remain neutral, it does not allow us to explore the patterns that underlie these averages. Consider the case of relatively neutral nondemocratic states. On average, their alignment with the United States remains unaltered by leader change. This could be either because their affinity does not change much or because on average the upswings are canceled out by the downswings. The evidence suggests the latter.
Tables 3 and 4 show that there is considerable volatility in affinity scores following leader change in nondemocracies even when there is little shift in average affinities. Table 3 takes a first look at this second ordered statistic by estimating the average value of the affinity variance under different contingent circumstances. The measure of affinity volatility is the squared difference between affinity and its lagged value. This variance measure looks only at how much the affinity score changes and ignores the direction of the change.
Variance in Affinity Score, Coalition Size, and Leader Change.
Note: Number of observations are in parentheses. Top Figure: Average Variance in Affinity = mean of (Affinity-LagAffinity)2.
Regression Analysis of Variance in Affinity Scores.
Note: Standard errors in parentheses. GDP = gross domestic product.
*Significant at 5 percent. **Significant at 1 percent.
Table 3 shows the average affinity variance, that is (Affinity − LagAffinity)2, for different size coalitions systems according to whether any leader change occurs in that year (coded as Leader Change > 0; remember that leader change is coded as the number of months of change following leader turnover in a particular year. The results are similar with a more restrictive threshold for leader change). The number of observations in each case is shown in parentheses. The table shows several patterns. First, large coalition systems (W = 1) experience less affinity volatility than do other systems, a mean of 0.021 versus a mean of 0.031. Second, while affinity variance appears to decrease with any leader change in large coalition systems (.024 vs. 0.018), in smaller coalition systems any leader change increases affinity variance (0.029 vs. 0.036). These differences are all statistically significant in t-tests and nonparametric Wilcoxon rank-sum tests.
The simple summary results in Table 3 suggest that rather than the affinity score of neutral nondemocratic nations remaining unchanged following leader change, there is considerable variability in affinity score for nondemocratic nations but that the shifts in each direction, on average, cancel each other out. Table 4 tests these distinctions more definitively by regressing the variance in affinity scores against institutions, leader change, and controls.
Models 4 and 5 show that leader change in small coalition systems leads to greater variance in affinity scores, as evidenced by the statistically significant coefficient estimate on the leader change variable. However, the effect of leader change in large coalition systems is indistinguishable from zero (sum of LC and W × LC). Model 4 suggests that in addition to leader change, wealth and population size have a significant effect on the variance in UNGA vote affinities, with large and wealthy nations being less likely to change position. Model 5 replicates these results with the inclusion of both nation and year fixed effects. The significance of both the population and wealth estimates are diminished in the presence of these fixed effects, although this is hardly surprising since the within nation variance in size and wealth is much smaller than are the average cross-sectional differences that are absorbed in the fixed effects.
Effects of Leader Change on Both the Level and the Variance of Affinity Scores
Although the results in Table 2 suggested leader change only affected affinity scores in highly aligned nondemocracies, the analysis of the variance in affinity casts that simplistic conclusion into doubt. Leader change in small coalition systems causes considerable shifts in UNGA vote affinities even though the average effects are small. In neutral nondemocracies, leader change can result in closer or more distant alignment with the United States, although on average the effect appears to be zero because the moves toward friendship are canceled by the cases of growing enmity. This necessitates a reassessment of the impact of leader change on affinity. To do so, I simultaneously model how institutions and leaders affect both the level of affinity and variance in affinity.
I estimate the following heteroscedastic model using maximum likelihood: Yit = Xit β + ∊ it , where ∊ it ∼ N[0, (Zit γ)2] and subscript it refers to nation i at time t. Effectively, this is a standard regression model except that the standard deviation in the variance of the error is treated as a linear function of independent variables, Z. The model was estimated using the maximum likelihood routines in Stata 12. The top section of Table 5, labeled Xβ, shows the effect of the independent variables on the level of affinity and the lower section, labeled σ, shows the effects of covariates on the volatility of affinity.
Effect of Leader Change on the Level and Variance of Affinity Scores.
Note: Standard errors are in parentheses. UNGA = UN General Assembly.
*Significant at 5 percent. **Significant at 1 percent.
In addition to leader change and institutional variables, model 6 examines the impact of average shift in affinities with the United States and US leader change. Model 7 includes year fixed effects, while model 8 contains both country and year fixed effects (in the Xβ assessment).
All three specifications reach similar conclusions regarding variance in affinity (the σ equation). Leader change in small coalition systems increases affinity volatility; the coefficient estimate on the leader change variable is positive and significant. However, leader change has no discernible impact in large coalition systems (the sum of LC and W × LC is indistinguishable from zero). The positive estimate on W in the σ equation suggests that, absent leader change, democratic states have slightly more volatile affinity scores but that wealthy (which correlates with W) and large population nations have less volatile UNGA voting positions. These results suggest that the earlier observed results that neutral autocracies remain neutral following leader change is an “on average” result. Leader change results in a shift in alignment in neutral states, but these shifts do not occur predominantly in one direction. Having seen that leader turnover increases variance in affinity scores in small coalition systems, I now return to a reassessment whether leader change has any aggregate effect of the level of the UNGA affinities for aligned autocracies by examining the Xβ equation.
The estimates of the Xβ equations shown in Table 5 are broadly similar to those seen in Table 2, although not always with the statistical significance seen in Table 2. The negative coefficient estimates on LC × LagAff variable means that on aggregate, following leader change, a nondemocracy aligned with the United States becomes less strongly aligned, while an autocracy aligned against the United States becomes on average more friendly toward the United States. It is important to remember these results are on aggregate. Beyond these average shifts toward a neutral position, alignments jump around more following leader change. For a state already highly aligned with the United States, it is difficult to become even more highly aligned. Similarly, if an autocracy is already voting against the US position on virtually every issue, then it is difficult to be measured as moving even more strongly against the US position. The net effect of affinity volatility is therefore regression toward a more neutral position. Leader change has no effect on the aggregate position of large coalition states (the sum of LC × LagAff and W × LC × LagAff is close to zero in all three models).
Figure 3 provides a convenient graphical interpretation of these results. It shows the predicted distribution of affinities under four contingencies based on the estimates in model 6. The figure assumes that in the previous period, the nation was strongly aligned with the United States, in particular lagged affinity is 0.5. This level is represented by the vertical line. 4 The horizontal axis show predicted affinity and the vertical axis is the density. The graph considers the smallest and largest coalition systems, W = 0 and W = 1, and cases of leader change and no leader change, LC = 1 and LC = 0.
The distribution of affinity under different institutional and leader change contingencies.
All four contingencies in Figure 3 show regression toward neutrality. That is, on average, nations do not sustain the highly pro-US position. However, large coalition systems are more likely to sustain their high affinity position than is a small coalition system. Further, leader change makes little difference to the future affinity in large coalition systems. The distributions of predicted affinity scores for large coalition systems are virtually identical whether leader change occurs or not. The distributions of predicted affinity in small coalition systems differ substantially depending upon whether leader change occurs.
The solid line represents the distribution of predicted affinity for a small coalition system absent leader change. The curve is steeper than for large coalition systems and it is shifted to the left. On average, autocratic allies become less supportive over time. However, compared to affine democratic states, the variance in affinity is substantially lower: the curve is taller and narrow. However, the affinity of small coalition systems becomes much more volatile when leader change occurs. The left most and lowest dashed curve corresponds to predicted affinity following autocratic leader change. On average, the new leader is likely to be less supportive than his or her predecessor (the shift to the left), and his or her position is likely to be more variable (the fatter curve). As predicted by McGillivray’s leader change arguments and supported by earlier studies of trade, sovereign debt, and sanctions, policy shifts following leader change in democracies are much smaller and less volatile than are policy shifts in autocracies.
Broader Implications and Conclusion
The myth of the unitary actor state is just that, a myth. Individual leaders and the interests of the coalitions that support them shape a nation’s foreign policy. Although this article has focused on a specific question, the impact of leader change on UNGA voting patterns, it is part of a paradigm shift that increasingly centers on leaders as the unit of analysis in international relations (see Bueno de Mesquita and Smith [2012] for a survey).
Starting with Bueno de Mesquita and Siverson (1995), scholars have found that foreign policy outcomes affect leader survival. While democrats seem particularly sensitive to failures, there is considerable variation in the extent to which nondemocratic leaders are also accountable. For instance, in the context of audience costs, Weeks (2012) argues that leaders in military regimes are highly constrained while leaders in personalist regimes are less accountable and so can pursue their own idiosyncratic agenda that can result in wildly fluctuating policies from one leader to the next. Croco (2011) finds leaders are accountable for the policies they initiate but are less so with respect to the policies started under a predecessor.
Given that they are held accountable, it is therefore little surprise that leaders pick policies that advance their survival (Chiozza and Goemans 2011). Such arguments form the basis of Bueno de Mesquita et al.’s (1999) explanation of the democratic peace. Leaders in different institutional settings differ in terms of accountability and how best to survive in office. Democrats need to deliver policy success and so are willing to try hard; hence, they typically win wars, and they avoid fighting other powerful democracies that will put up a tough fight as such wars are hard to win. In addition to explaining known regularities, leader-centered theories suggest novel predictions. For instance, within the context of the democratic peace, Bueno de Mesquita et al. (2004) test the efforts leaders make to win wars and the selectivity of dispute involvement. Perhaps the most important aspect of the leader-based paradigm is the profusion of new questions and predictions amenable to testing.
Heterogeneity across leaders is an important driver of foreign policy. For instance, Horowitz, McDermott, and Stam (2005) examine the characteristics of individual leaders and find that age affects the propensity of leaders to become involved in disputes. In a similar vein, Goldstein (2001) examines whether gender systematically shapes foreign policy. In theoretical work, Wolford (2007) examine the impact of unseen heterogeneity. He argues that leaders differ in their resolve and willingness to fight and that these differences create variation in conflict behavior both across leaders and across an individual leader’s term in office as leaders attempt to signal their type. Jones and Olken (2005) use the instrument of accidental leader death to test whether leader heterogeneity affects economic growth. There is growing evidence that differences between individual leaders shape international outcomes.
Leaders need supporters to retain power. When leaders change, the segments of society that need rewarding also change and as a consequence policies shift. In Punishing the Prince, McGillivray and Smith (2008) provide the most comprehensive assessment of this approach, and this article falls squarely within the realm of a test of their arguments. Within this genre, political institutions play an important ameliorating role. Some forms of institutions, such as democracy, place considerable constraint on policy variability, and so, as seen here, changes in democratic leadership have comparatively little effect. Yet, when leaders are beholden to only a small proportion of society, there can be radical swings in policy associated with leader change.
Rather than focus on individual leaders, Carroll, Leeds, and Mattes (2012) and Leeds, Mattes, and Vogel (2009) focus directly on changes in the coalitions that support leaders and argue that it is changes in the support coalition rather than leader change per se that creates policy shifts. For instance, they would argue the leaders within the Somoza dynasty that ruled Nicaragua from 1936 to 1979 were largely equivalent and that the support coalition remained largely unchanged until the Sandinistas took over in 1979.
There is a growing recognition of the importance of leaders as the unit of analysis for understanding international relations. Although this article offers only modest theoretical developments, it provides empirical support for leader-based theory. Through an analysis of UNGA voting patterns, it shows that changes in autocratic leaders result in greater policy changes that occur following democratic leader turnover. Democratic leaders represent a more consistent set of interests, so relations between democracies are likely to be more stable and insensitive to leader change. Relations between states become much more volatile when they involve an autocratic state and leader change occurs.
Footnotes
Appendix
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
