Commitment, Networks and International Conflict

In collaboration with: Erik Gartzke

Why is commitment sometimes a problem for states and sometimes not? Canonical explanations for war as a commitment problem tend to over-predict conict. Expected changes in the distribution of power create incentives for rising powers to bide their time, reneging on agreements that no longer suit their interests only after their rise in power. Declining powers, in contrast, have the opposite incentives; to ght today rather than to ght, and lose, tomorrow. Still, there is only modest empirical support for this prediction. Declining powers sometimes appear to initiate preventative wars against rising powers. However, many power transitions do not produce major war. The likelihood of conict onset attributable to commitment problems emerging from dierences in expected power shifts is thus considerably less than existing theory predicts. We propose to address this empirical puzzle using two simple frameworks. First, nations may vary in the degree to which their interests coincide. Some existing arguments about commitment explicitly adopt a realist conception of competition, assuming that all states have equal, and equally incompatible interests. To the degree that nations have compatible interests, this removes the zero-sum competition implied by existing arguments about commitment. Second, regardless of whether nations constitute direct competitors, they exist in a world of other states. Prevailing networks of political, economic and social ties can modify the trajectory of bilateral competition. Drawing on network theory we argue that dierences in expected changes in the distribution of military power are likely to lead to conict onset only if they occur in dyads that are not integrated into international networks. Pairs of states that expect to grow at dierent rates in term of their military capabilities in the near future and are not integrated in social networks are more likely to go to war. We use UN General Assembly resolution co-sponsorship data and dierent candidate measures of network integration to examine these interactions.