Barbara Koremenos. 2001. Loosening the Ties that Bind: A Learning Model of Agreement Flexibility

Barbara Koremenos. 2001. "Loosening the Ties that Bind: A Learning Model of Agreement Flexibility." International Organization 55(2): 289-325.

Koremenos develops a model of the optimal duration and renegotiation provision of international agreements based on the idea that the distributional consequences of a new agreement may not be known with certainty at the time the parties first negotiate the agreement.

I. Introduction

• Uncertainty in the international environment leads states to choose particular duration and renegotiation provisions. The provisions, in turn, affect whether or not states conclude international agreements and whether or not they renege on them.
• Koremenos develops a model in which the parties to an agreement learn over time about its costs and benefits to them. This learning reduces their uncertainty about the agreement’s effects, with the result that they may eventually become sure enough of these effects to be willing to extend the agreement indefinitely.
• Key factors identified by her theory that affect the choices of duration and renegotiation:
1. The degree of agreement uncertainty - the variance of the distribution of gains from an agreement
• The greater this uncertainty, the more likely states will want to limit the duration of the agreement and incorporate renegotiation.
2. The degree of noise in the environment - the variance of confounding variables whose effect on outcomes may be confused with that of an agreement.
• The greater the noise, the more difficult it is to learn how an agreement is working; incorporating limited duration and renegotiation provisions becomes less valuable in high noise level situations.

II. Model: Learning about the Workings of an Agreement

• An agreement is an experience good in which knowledge of its effects is determined only by using it and observing outcomes.
• Model assumptions:
1. States care about the future—discount factor is not zero.
2. States are risk averse.
3. There is uncertainty about future states of the world.
4. The costs of making agreements completely contingent are sufficiently large that the parties never choose to do so.
5. There are costs to negotiating and renegotiating agreements
6. There are costs to reneging on agreements.
7. States have shared a priori beliefs about information they do not possess and revise their beliefs according to Bayesian logic as their interactions evolve.

III. Basics of the Model

• There are two prospective parties to the agreement.
• The division of the gain agreed upon in the initial agreement reflects the relative bargaining power of the two parties.
• The basic problem facing the parties to an agreement in this model is to sort out the effects of the agreement from other random fluctuations in outcomes.
• The states face a choice between an agreement of indefinite duration and one finite-duration agreement followed by an agreement of indefinite duration. In the simple two-period case, formally, the choice becomes one two-period agreement with no renegotiation or two one-period agreements with renegotiation in-between to realign the distribution of gains.
• Renegotiation takes place whenever a finite duration agreement comes to an end and reflects information gained in the previous period, but not changes in bargaining power.

IV. Two-Period Game

• Period 1: players engage in a Nash demand game in which they choose the expected division of gains; if they choose the same expected division of gains, they continue on to negotiations. Otherwise, no agreement.
• Period 2: players enter agreement-type choice stage and choose among the following strategies:
• No agreement
• One two-period agreement (the analog of a nonrenegotiated agreement)
• At the beginning of the of the second period, they have to choose whether to abide by the duration provision stipulated in the agreement.
• If one party reneges on the agreement before the two periods are over, the parties negotiate a new agreement.
• Two one-period agreements (the analog of a renegotiated agreement)
• They must decide whether to proceed with the renegotiation in the second period.
• An agreement is only achieved when the two players play the same strategies in round two. (Except for when they both play 'No agreement'. Obviously.)
• Outcomes determined by nature and observed by players.

V. Hypotheses

1. All else equal, for risk-averse parties an increase in agreement uncertainty increases the value of renegotiation and therefore makes the parties more likely to choose a renegotiated agreement / two one-period agreements over a nonrenegotiated agreement (one two-period agreement).
2. All else equal, for risk-averse parties an increase in noise decreases the value of renegotiation and therefore makes the parties more likely to choose a nonrenegotiated agreement (a two-period agreement) over a renegotiated agreement (two one-period agreements).

VI. Expanded Model

• The general model, built on the same concepts as the two-period model, expands the time horizons to infinity.
• States now face new choices:
1. An agreement-type choice between no agreement, one infinite-duration agreement, and one finite agreement followed by an infinite-duration agreement.
2. If states choose to renegotiate the agreement, they must make a second choice regarding the timing of renegotiation.

VII. Empirical Data from Nuclear Non-Proliferation Treaty (NPT)

• Main treaty provisions:
• Article 1 - Prohibits nuclear-weapon states (NWS) from transferring nuclear explosives
• Article 2 - Places obligations on the non-nuclear-weapon states (NNWS) not to receive or manufacture such weapons
• Article 3 - Requires that signatories negotiate individually or collectively full-scope safeguards agreements with the International Atomic Energy Association (IAEA)
  Article 4 & 5 - Provides reassurance to NNWS that they will be able to enjoy peaceful uses of nuclear energy and explosions without discrimination; the NWS are obliged to provide both technological and material assistance to NNWS
• Article 6 - Demands progress by existing nuclear powers on controlling the arms race
• Uncertainty about:
• Participants led to uncertainty about the distribution and level of security benefits under NPT
• Security consequences
• Effect of NPT on economic prosperity and technological development
• Political costs and benefits that would result from the agreement.
• Evidence of resolving uncertainty through learning:
• The parties to the NPT planned review conferences every few years at which they could cooperatively take stock of how the treaty was working in practice. 
• In terms of the distribution of political gains and losses, it became clear over time that concerns that the NPT would prevent European integration were groundless. Uncertainty with regards to other concerns were eliminated reduced over time as well.

VIII. Conclusion

• The credibility of commitments in the face of uncertainty—in this paper, a one-shot uncertainty surrounding the division of gains from an agreement—requires a trade-off between flexibility and constraint. 
• This model shows that the reason we do not observe much reneging in actual agreements is in part because their duration and renegotiation provisions have been chosen in ways that act to minimize this costly behavior.