Polarity and Free Trade seek security in a multipolar system is indeed more likely to occur in n-power makes alliance formation likely:the frag- than in two-power systems(Duncan and mentation of the system into rival alli- Siverson 1982). ances is much more likely than a tacit or The number of great powers is obvious- explicit general assignment to abstain ly not the only source of variations in alli- from such alliances (Snyder 1984,462). ance durability.In the postwar period,for That a similar dynamic has occurred in example,the stability of the Western alli- the postwar bipolar system is clear. ance is the product not only of shared In principle,neither the alliances of a strategic interests but also of the demo- multipolar,nor those of a bipolar,system cratic polities and mixed economies com- should enjoy an advantage with respect to mon to its members;analogously,the the evolution of a corresponding free Soviet Union's determination to maintain trade coalition:alliances in both systems ideologically compatible states along its render the security externalities that ac- western frontier tightens bonds among crue to a state's trading partners positive members of the Warsaw Pact.But idio- rather than negative.Yet a bipolar system syncratic factors alone do not explain the is advantaged relative to its multipolar postwar alliance configuration:analytic counterpart on two dimensions relevant arguments also predict that bipolar sys- to the conclusion of an agreement to open tems generally will produce more stable intraalliance markets:(1)exit and(2)in- coalitions than will their multipolar centives of the great power members to counterparts. forgo the use of their market power to ex- First,as Snyder argues,bipolar coali- ploit the smaller members of the alliance. tions are the product largely of systemic structure.Alliances in a multipolar sys- tem,however,are not structurally deter- mined but are the result instead of "choice The Costs of Exit among several options"(Snyder 1984, 415).Typically,serious conflicts of inter- The threat that any member can exit or est among the great powers of any given abandon an existing alliance to join an system result from the anarchic character alternative one reduces the likelihood that of international politics.As a conse- allies will open their markets to each quence,the substitutability of great other:exit transforms the external secur- power allies is likely to decline with their ity economies that would otherwise facili- number:the probability of finding tate the conclusion of an accord into another large power as compatible as the diseconomies.As exit opportunities re- existing ally varies directly with the num- create balance-of-power fears and also ber of great powers that exist.Thus,the shorten the "shadow of the future,"pros- great powers of a two-power system are pects for free trade even within interna- less likely than are those of an n-power tional alliances seem dim. world to be close substitutes for each As Snyder observes in his discus- other. sion of the alliance security dilemma, This can be represented spatially, however,the risk of exit varies widely where the distance between any two states across international systems:the risk is is a measure of the conflict of interest be- higher in multipolar than in bipolar inter- tween them.A bipolar system embedded national political systems(Snyder 1984). in an anarchic world is likely to distribute Snyder's hypothesis has been confirmed its constituent great powers 180 degrees empirically:studies of the nineteenth and from each other.Thus,a very sharp dis- twentieth centuries demonstrate that exit tinction is likely to exist between the two 1249Polarity and Free Trade seek security in a multipolar system makes alliance formation likely: the frag￾mentation of the system into rival alli￾ances is much more likely than a tacit or explicit general assignment to abstain from such alliances (Snyder 1984, 462). That a similar dynamic has occurred in the postwar bipolar system is clear. In principle, neither the alliances of a multipolar, nor those of a bipolar, system should enjoy an advantage with respect to the evolution of a corresponding free trade coalition: alliances in both systems render the security externalities that ac￾crue to a state's trading partners positive rather than negative. Yet a bipolar system is advantaged relative to its multipolar counterpart on two dimensions relevant to the conclusion of an agreement to open intradiance markets: (1)exit and (2) in￾centives of the great power members to forgo the use of their market power to ex￾ploit the smaller members of the alliance. The Costs of Exit The threat that any member can exit or abandon an existing alliance to join an alternative one reduces the likelihood that allies will open their markets to each other: exit transforms the external secur￾ity economies that would otherwise facili￾tate the conclusion of an accord into diseconomies. As exit opportunities re￾create balance-of-power fears and also shorten the "shadow of the future," pros￾pects for free trade even within interna￾tional alliances seem dim. As Snyder observes in his discus￾sion of the alliance security dilemma, however, the risk of exit varies widely across international systems: the risk is higher in multipolar than in bipolar inter￾national political systems (Snyder 1984). Snyder's hypothesis has been confirmed empirically: studies of the nineteenth and twentieth centuries demonstrate that exit is indeed more likely to occur in n-power than in two-power systems (Duncan and Siverson 1982). The number of great powers is obvious￾ly not the only source of variations in alli￾ance durability. In the postwar period, for example, the stability of the Western alli￾ance is the product not only of shared strategic interests but also of the demo￾cratic polities and mixed economies com￾mon to its members; analogously, the Soviet Union's determination to maintain ideologically compatible states along its western frontier tightens bonds among members of the Warsaw Pact. But idio￾syncratic factors alone do not explain the postwar alliance configuration: analytic arguments also predict that bipolar sys￾tems generally will produce more stable coalitions than will their multipolar counterparts. First, as Snyder argues, bipolar coali￾tions are the product largely of systemic structure. Alliances in a multipolar sys￾tem, however, are not structurally deter￾mined but are the result instead of "choice among several options" (Snyder 1984, 415). Typically, serious conflicts of inter￾est among the great powers of any given system result from the anarchic character of international politics. As a conse￾quence, the substitutability of great power allies is likely to decline with their number: the probability of finding another large power as compatible as the existing ally varies directly with the num￾ber of great powers that exist. Thus, the great powers of a two-power system are less likely than are those of an n-power world to be close substitutes for each other. This can be represented spatially, where the distance between any two states is a measure of the conflict of interest be tween them. A bipolar system embedded in an anarchic world is likely to distribute its constituent great powers 180 degrees from each other. Thus, a very sharp dis￾tinction is likely to exist between the two