Table 2 presents a summary of the hypotheses proposed by my argument and those of alternative ones. In Chapter 3, I test several of these hypotheses using quantitative data. A statistical approach allows me to test my argument across a large number and diverse array of cases, while controlling for the effects of alternative arguments. Statistical tests, however, have their limitations. They are less helpful in proving a causal relationship or testing hypotheses that defy quantification. Some of my argument’s hypotheses are indeed difficult to test quantitatively. In particular, the hypotheses on the effects of major events or crises (H1a3), the leader’s response to regime change (H1b1 and H1b2), and the foreign power’s preference for partial versus full regime change (H1a4 through H1a7) require a more nuanced understanding of conditions and events surrounding each case. Accordingly, I test these hypotheses in the case studies that follow Chapter 3. If my argument is valid, we should observe that domestic opposition in the target state increases the risk of FIRC. In particular, we should find that it constrains the leader’s ability to make concessions to the foreign power, while at the same time making the leader more vulnerable to overthrow. Leaders without such opposition will not necessarily make concessions, particularly if they have the military means to resist making them. But their stronger base of domestic support should nevertheless cause the foreign power to prefer a settlement by making regime change too costly to pursue.
CHAPTER 3
Testing the Logic of Foreign-Imposed Regime Change
Scholars since Plato have posited that domestic political turmoil in a state inspires other states to intervene.1 But despite this broad consensus, the literature has yet to establish a definitive link between domestic opposition in the target state and the foreign overthrow of its government. Many statistical studies, for example, focus on the broader concept of foreign intervention, which includes cases other than regime change. In contrast, studies that focus explicitly on FIRC often exclude a number of cases that arguably qualify. One study, for example, focuses on regime change imposed after war and, therefore, excludes cases of covert and indirect FIRC.2 Others examine a broader range of FIRC cases but neglect to consider the conditions under which regime change is not pursued.3 Without testing nonevents, we cannot know whether the conditions argued to cause FIRC are just as likely to occur when states negotiate.
The statistical tests presented in this chapter serve three purposes. First, they help fill a gap in the literature by analyzing whether the long-hypothesized link between domestic opposition in the target state and FIRC does indeed exist. Second, they test my theory’s core causal claim while controlling for other possible explanations. If my theory is empirically valid, at a minimum, we should find that variables measuring domestic opposition in the target state are substantively and statistically significant. Failure to find such a relationship would seriously challenge the theory. If the variables associated with alternative arguments have a stronger effect, this would also cast doubt on my theory’s explanatory value. Finally, the tests in this chapter can also indicate whether my argument is generalizable across a broad range of cases.
The results show that the target’s domestic opposition levels are positively correlated with the probability of FIRC. These results hold throughout a variety of model specifications and robustness checks. In contrast, variables representing arguments based on psychological bias, bureaucratic or interest-group pressure, credible commitment, and incomplete information are either insignificant or lose their significance when subjected to additional tests.
Statistical tests, however, have their limitations. The existence of a correlation between domestic opposition in the target state and FIRC does not imply a causal relationship. Certain types of leaders may be prone to domestic opposition, and it could be attributes of these leaders, and not their opposition, that explain their overthrow. Domestic opposition may also lead to FIRC for reasons other than those proposed by my theory. States plagued by domestic instability, for instance, may be subject to FIRC because other states fear the trajectory of their policies. The proxies I use to measure domestic opposition might also capture other effects that explain FIRC. Lastly, coding decisions can lead to the exclusion or inclusion of certain cases and, therefore, could influence the results.
These limitations can be diminished in some instances. To reduce the potential influence of coding decisions, I test my model on FIRC data compiled by other scholars, in addition to my own. I also test two proxies for domestic
