490 likes | 662 Vues
This analysis explores the framework of bargaining and signaling between two parties, A and B, who negotiate over the division of valuable resources. We delve into the concept of reservation values, the zone of agreement, and the implications of asymmetric information in bargaining scenarios. Readers will gain insight into why states may falter in achieving ex post efficient bargains, examining factors such as commitment issues and the dynamics of risk and return. Understanding these complexities is crucial for comprehending the motivations behind conflict and potential resolutions.
E N D
Basic Set-Up • Two parties, A and B, bargain over the division of something of value. • Division of territory • Distribution of economic gains • Policy (e.g., taxes) • We often normalize this range of possible deals to [0,1]. • A settlement is x [0,1].
Basic Set-Up • A prefers larger values of x; B prefers smaller ones: • UA(x) increasing, UB(x) decreasing • For simplicity, assume risk neutrality for most examples: UA(x) = x and UB(x) = 1 – x.
Basic Set-Up • Each party has a minimal acceptable settlement • “reservation value” • the deal that it sees as equivalent to no deal. • The reservation value is determined by the expected value of the “outside option”: • the expected value of war • the expected value of a revolution or coup • An actor can always guarantee its reservation value by implementing the outside option
The Reservation Value • Most generic form: wA, wB • We sometimes assume that conflict can be seen as a “costly lottery”: • let p denote the probability that A will win • assume that the winner imposes its most preferred outcome • let cA, cB denote the expected costs • Then, wA = p – cA wB = 1 – p – cB
The Reservation Value • Reservation points are then x such that UA(x) = p – cAand UB(x) = 1 – p – cB • With example utility functions, A will accept B will accept 0 1 p – cA p + cB
Zone of Agreement • All settlements between the two reservation points constitute the “zone of agreement”: the set of deals that both sides prefer to conflict. • The zone of agreement is always non-empty if • Conflict is costly in aggregate In our example: The zone of agreement is non-empty if p + cB > p – cA or cA + cB > 0 . Note: This means that one actor could have negative costs for conflict, as long as wA, wB < 1. • The actors are not too risk acceptant
Fearon, “Rationalist Explanations for War” Motivation: If war is costly, there exist settlements that both sides should prefer to war. Why do states sometimes fail to reach ex post efficient bargains? Proposed mechanisms: • Asymmetric information about p, cA, and/or cB , combined with incentives to misrepresent. • Commitment problems: Deals in the zone of agreement may be non-self enforcing due to • First-strike advantages • Exogenous shifts in the power distribution • Endogenous shifts in the power distribution • The good is lumpy or indivisible.
Asymmetric Information • Assume that each actor is incompletely informed about the other’s value for conflict • Most generic: wA, wB unknown • Common assumption: p known, cA, cB unknown
“Take It or Leave It” Bargaining x, 1 – x Accept A B Offer x Reject p – cA, 1 – p – cB
Equilibrium Strategies • There exists a “risk-return tradeoff” in B’s decision: • Increasing x decreases the risk of war, F(p – x), but also decreases B’s return on the deal, 1 – x. • More profitable bargains can only be achieved by accepting a greater risk of war. • But it never makes sense to offer more than .
Equilibrium Strategies The optimal offer, x*, solves If F(x) has a “monotone hazard rate,” which ensures that there exists solution to the first-order condition. In general, the optimal offer entails a positive probability of war—i.e., .
Equilibrium Strategies If A’s costs are distributed uniformly, then The equilibrium probability of war is
Two Shortcomings • The TILI bargaining framework • does not allow counter-offers • artificially imposes a final move. • Most conflicts are preceded by efforts to signal resolve through threats and escalatory efforts.
Powell, “Bargaining in the Shadow of Power” Accept Accept D D S … S Reject Reject Offer Offer Attack Attack t=1 t=0
Assumptions D’s capital Existing border S’s capital 0 q 1 • Until an agreement or war, D gets a per-period payoff of q and S gets a per-period payoff of 1 – q. • War is a costly lottery. Let p = Pr(D wins), Let d and s denote per-period loss from having fought a war. Hence, per-period expected values of war are • wD = p – d • wS = 1 – p – s
Assumptions • If p – d > q, then D is dissatisfied. If 1 – p – s > 1 – q, or p + s < q, then S is dissatisfied. • It is easy to see that at most one state can be dissatisfied: D’s capital S’s capital 0 q p – d p + s 1 • If both states are known to be satisfied, then neither will ever attack, and no serious bargaining will take place: 0 p – d q p + s 1
Assumptions • To generate incomplete information,assume • If , then D is potentially dissatisfied. • At most one state can be potentially dissatisfied.
Key Result Lemma. The potentially dissatisfied state never rejects an offer in order to make a counter-offer. Hence, in equilibrium, the equilibrium outcome is the same as in the TILI bargaining game: • S offers • D either accepts or attacks
Intuition • Conjecture that some dissatisfied type(s) of D counters with an offer, x. Let r denote the most resolute type that does so. • Possible outcomes • War in some future period • But war now is better than a period of SQ followed by war. • D accepts some offer from S in future period • But the most S will ever offer is p−r, which is equivalent to the war payoff. War now is better for type r. • S accepts the counter-offer • But S can always reject x, leading to the SQ payoff in that period, and then offer p−r, which it knows will be accepted. S will reject any offer which gives it less than d(1−q)+(1−d(1-p+r). • But D of type r could get p−r>q immediately and in all future periods by attacking now. Hence, this type is not willing to make a counter-offer that S would accept.
The Relationship of Power and War q = 0.5 q = 0.33
Leventoğlu and Tarar, “War and Incomplete Information” Accept Accept D D S … S Reject Reject Attack Attack t=1 t=0
Leventoğlu and Tarar, “War and Incomplete Information” Accept Accept D S D S D … S Reject Reject Attack Attack Attack Attack t=1 t=0
Main Result • If d is sufficiently high, then there exists a “no risk” equilibrium in which D rejects a low initial offer and then makes a counter-offer which is accepted. • This implies that incomplete information leads to war only when • the states are impatient, or • they fail to coordinate on the risk free equilibrium
Thoughts • As the time between offers shrinks to zero, or d →1, a peaceful equilibrium always exists. • Failure of bargaining is not well explained by “pure” bargaining models. • Key question: Given that the existence of an efficient deal is common knowledge, why would states ever walk away from the bargaining table?
Signaling Accept B A A Message Offer Reject
A Simple Signaling Game A • Assumptions: • ACQA>SQA, BDA • BDB>ACQB • WARAhas cdf F • WARB has cdf G Challenge Status Quo B SQA SQB Resist Acquiesce A Back Down Stand Firm ACQA ACQB BDA BDB WARA WARB
The Risk-Return Tradeoff • Even in this simple setting, B faces a risk-return tradeoff: • Assume BD is B’s first-best outcome • If WARB > ACQB, then B has a dominant strategy to Resist • If WARB < ACQB, then B faces a choice between • getting its second-best payoff for certain, and • a lottery between its first- and third-best payoffs. • The odds of the lottery are determined by the posterior belief that A will fight.
The Risk-Return Tradeoff • Let q denote B’s posterior belief that A will stand firm given that A has challenged. • Then B will Acquiesce if
Informative Signaling • Let p = 1 – F(BDA) denote prior probability that A will stand firm • A’s challenge is informative if q > p. • For this to happen, the probability of a challenge must be less than one. • Separation of types requires that BDA < SQA for some types. • Otherwise, ACQA > SQA ensures that a challenge weakly dominates the status quo for all types.
Types of Signaling • “Slippery slope”: challenge creates an exogenous risk of war • “Tying hands”: challenge creates an “audience cost” for backing down • “Sunk costs” or “burning money”: A must pay an up-front cost to challenge
Slippery Slope A Challenge Status Quo B SQA SQB Resist Acquiesce A Back Down Stand Firm ACQA ACQB N 1 – p p WARA WARB BDA BDB WARA WARB
Tying Hands A Challenge Status Quo B SQA SQB Resist Acquiesce A Back Down Stand Firm ACQA ACQB BDA = SQA – a BDB WARA WARB
Sunk Costs A Challenge Status Quo B SQA SQB Resist Acquiesce A Back Down Stand Firm ACQA – m ACQB BDA = SQA – m BDB WARA – m WARB
Equilibrium • In general, for fixed p, m, or a, the equilibrium strategies are defined by a set of cutpoints in the continuum of types: Status Quo Back Down Challenge Back Down Challenge Stand Firm WARA Acquiesce Resist WARB
Schultz, “Do Democratic Institutions Constrain or Inform?” • Questions: Does democracy influence crisis outcomes, and if so how? • Competing Theories • Institutional constraints: democracy increases the political costs of war • Informational: democratic institutions increase transparency and/or increase audience costs • Realism (the null hypothesis): democracy doesn’t matter • Problem: While it is relatively easy to determine whether democracy matters, it is much harder to distinguish competing arguments for why it matters.
The Theoretical Model A Challenge Status Quo B (0,1) Resist Acquiesce A Back Down Stand Firm (1,0) (– a, 1) (wA, wB)
Putting Democracy in the Model • Institutional constraints • Democracy lower expected value for war on average • Assume wA ~ [– CA – dZA, – dZA], where dA > 0 and ZA = 1 if state A is a democracy • Information • Democracy higher audience costs (a) • Transparency democracy generates complete information about wA
Comparing Complete and Asymmetric Information • Probability of a challenge • CI: A only challenges when wA > – a • AI: A challenges when wA > – b , with b > a • Probability of resistance • CI: B never resists conditional on a challenge • AI: B resists with nonzero probability for some parameters • Probability of war • CI: Zero • AI: Nonzero for some parameters
Outcomes as a Function of dA 1 B Resists|Challenge A Challenges Probability in Equilibrium War 0 Magnitude of constraint, dA
Outcomes as a Function of a 1 A Challenges Probability in Equilibrium B Resists|Challenge War 0 Magnitude of Audience Costs, a
The Data • Dependent variable: Did the target resist? • Data set: Militarized Interstate Disputes (MIDs) • 1654 disputes over period 1816-1980 • arranged in dyads of initiator-target • RECIP = 1 if target reciprocated the initiator’s action, and RECIP = 0 otherwise. • Main independent variable: Regime type of the initiator • Data set: Polity III • DEMINIT = 1 if initiator is democratic (score of 7 or higher on 21-point composite democracy scale), and DEMINIT = 0 otherwise.
Bivariate Correlation Pearson c2 = 6.95 Pr = 0.008
Summary • Use of model to • generate testable hypotheses and • identify a critical test between theories. • Convinced?
Summary • Use of model to • generate testable hypotheses and • identify a critical test between theories. • Potential problems • Unmeasured factors • Democracies select weak targets • Democracies make smaller demands • Observed correlation could arise from more than one causal pathway (identification problem) • Mismatch between data and model