My theory of commitment design presented in Chapter 5 demonstrates that two types of mechanisms – conditional commitment and probabilistic commitment – can be effective instruments for balancing the demands for deterrence against the effects of moral hazard on conflict. Probabilistic commitments have the added advantage of minimizing moral hazard distortions on the expected bargaining settlement. The question addressed in this chapter is whether these mechanisms – which satisfy the third party's maximization problem in the theory – are actually implemented in practice for the same purposes they are designed to accomplish according to the theory. In this chapter, I test various effects derived from the theory in Chapter 5.

The theory produces a number of testable predictions that apply to military alliances. In this chapter, I analyze the following four implications: (1) incongruent policy preferences between alliance members increase the likelihood that alliances will be probabilistic; (2) sizable disparities in capabilities between weak protégés and powerful third-party defenders increase the chances that states will form probabilistic commitments; (3) more powerful protégés are more likely to receive conditional types of commitments; and (4) the presence of unobservable actions tends to favor probabilistic alliances and unconditional commitments over conditional alternatives.