Liberal Theory of International Politics 521 cies are optimal for others (or insignificant),there are strong incentives for coexist- ence with low conflict. Where,by contrast,underlying state preferences are zero-sum or deadlocked,that is,where an attempt by dominant social groups in one country to realize their prefer- ences through state action necessarily imposes costs(negative externalities)on domi- nant social groups in other countries,governments face a bargaining game with few mutual gains and a high potential for interstate tension and conflict.The decisive precondition for costly attempts at coercion,for example,is neither a particular con- figuration of power,as realists assert,nor of uncertainty,as institutionalists maintain, but a configuration of preferences conflictual enough to motivate willingness to ac- cept high cost and risk.20 In other words,intense conflict requires that an aggressor or revisionist state advance demands to which other states are unwilling to submit. Revisionist preferences-underlying,socially grounded interests in revising the sta- tus quo-are distinct from revisionist "strategies,"that is,a need to alter the status quo to protect enduring interests under new strategic circumstances.Liberals focus on the former,realists (and institutionalists)on the latter.Hence while both theories predict security confict,they do so under different circumstances.For example,in- creased military spending in response to an adversary's arms buildup is a change in strategy with fixed preferences consistent with realism;increased spending initiated by a new ruling elite ideologically committed to territorial aggrandizement is a pref- erence-induced change in strategy consistent with liberalism.21 Where,finally,motives are mixed such that an exchange of policy concessions through coordination or precommitment can improve the welfare of both parties relative to unilateral policy adjustment(i.e.,a collective action problem),states have an incentive to negotiate policy coordination.Games like coordination,assurance, prisoner's dilemma,and suasion have distinctive dynamics,as well as impose pre- cise costs,benefits,and risks on the parties.Within each qualitative category,incen- tives vary further according to the intensity of preferences. For liberals,the form,substance,and depth of cooperation depends directly on the nature of these patterns of preferences.Hence where "Pareto-inefficient"'outcomes are observed-trade protection is a commonly cited example-liberals turn first to countervailing social preferences and unresolved domestic and transnational distribu- tional conflicts,whereas institutionalists and realists,respectively,turn to uncertainty and particular configurations of interstate power.22 Liberal Theory as Systemic Theory These liberal assumptions,in particular the third-in essence,"what states want is the primary determinant of what they do"'-may seem commonsensical,even tauto- logical.Yet mainstream IR theory has uniformly rejected such claims for the past 20.Note that some rationalist analyses dismiss such risk-acceptant preferences as "irrational";see Fearon 1995. 21.For example,Van Evera 1990-91,32. 22.Grieco's study of NTB regulation is discussed later.Liberal Theory of International Politics 521 cies are optimal for others (or insignificant), there are strong incentives for coexist￾ence with low conflict. Where, by contrast, underlying state preferences are zero-sum or deadlocked, that is, where an attempt by dominant social groups in one country to realize their prefer￾ences through state action necessarily imposes costs (negative externalities) on domi￾nant social groups in other countries, governments face a bargaining game with few mutual gains and a high potential for interstate tension and conflict. The decisive precondition for costly attempts at coercion, for example, is neither a particular con￾figuration of power, as realists assert, nor of uncertainty, as institutionalists maintain, but a configuration of preferences conflictual enough to motivate willingness to ac￾cept high cost and risk.20 In other words, intense conflict requires that an aggressor or revisionist state advance demands to which other states are unwilling to submit. Revisionist preferences-underlying, socially grounded interests in revising the sta￾tus quo-are distinct from revisionist "strategies," that is, a need to alter the status quo to protect enduring interests under new strategic circumstances. Liberals focus on the former, realists (and institutionalists) on the latter. Hence while both theories predict security conflict, they do so under different circumstances. For example, in￾creased military spending in response to an adversary's arms buildup is a change in strategy with fixed preferences consistent with realism; increased spending initiated by a new ruling elite ideologically committed to territorial aggrandizement is a pref￾erence-induced change in strategy consistent with liberali~m.~' Where, finally, motives are mixed such that an exchange of policy concessions through coordination or precommitment can improve the welfare of both parties relative to unilateral policy adjustment (i.e., a collective action problem), states have an incentive to negotiate policy coordination. Games like coordination, assurance, prisoner's dilemma, and suasion have distinctive dynamics, as well as impose pre￾cise costs, benefits, and risks on the parties. Within each qualitative category, incen￾tives vary further according to the intensity of preferences. For liberals, the form, substance, and depth of cooperation depends directly on the nature of these patterns of preferences. Hence where "Pareto-inefficient" outcomes are observed-trade protection is a commonly cited example-liberals turn first to countervailing social preferences and unresolved domestic and transnational distribu￾tional conflicts, whereas institutionalists and realists, respectively, turn to uncertainty and particular configurations of interstate power.22 Liberal Theory as Systemic Theory These liberal assumptions, in particular the third-in essence, "what states want is the primary determinant of what they do"-may seem commonsensical, even tauto￾logical. Yet mainstream IR theory has uniformly rejected such claims for the past 20. Note that some rationalist analyses dismiss such risk-acceptant preferences as "irrational"; see Fearon 1995. 21. For example, Van Evera 199G91,32. 22. Grieco's study of NTB regulation is discussed later