NGO monitoring and the legitimacy of international cooperation: A strategic analysis

Abstract

States often invite NGOs to monitor international cooperation. Under what circumstances are states likely to take this step? We argue that NGO monitoring allows states to provide domestic publics with credible evidence regarding successful cooperation, but that this credibility carries a cost: if states fail to cooperate, a participating NGO will expose this failure and thus delegitimize the cooperation effort. Our formal analysis indicates that states obtain a dual benefit from NGO participation: in addition to enhanced legitimacy, NGO scrutiny helps states credibly commit to high cooperation levels vis-á-vis each other. The increased costs of failure, however, may deter state use of NGO monitoring. Surprisingly, we find that NGO monitoring is the most useful for states when the cooperation cost is relatively low. We explore the empirical relevance of our theoretical argument in NGO monitoring of World Bank development projects and compliance with the Kyoto Protocol. We also explain why NGO monitoring has been disallowed in the Global Environment Facility. Our analysis provides a firm strategic foundation for the idea that NGO participation sometimes confers benefits to states, and our theory has several empirically falsifiable implications.

Appendix

1.1 A. Mathematical Appendix: Equilibrium

To find the subgame-perfect equilibrium, consider first the cooperation levels $E_{A}^{*},E_{B}^{*}$ without NGO monitoring. The cooperation effort of state i, $E_{i}^{*}$, must meet the first-order condition for the expected payoff given in expression 1:

$$ E_{j}^{*} + 1 = 2c \cdot E_{i}, $$

(9)

where j ≠ i. The left side measures the benefits from increased cooperation effort (depending on how much the other state cooperates), whereas the right side measures the marginal cost of additional cooperation effort. Recall that c > 1 by assumption. This linear system of two equations has two endogenous variables, so the interior solution is unique. It is provided in expression 4 in the main text.

Suppose now NGO monitoring is allowed. Each state i selects cooperation efforts to meet the following first-order condition for expression 2:

$$ E_{j} \cdot \left(1 + L(1)\right) + 1 = 2c \cdot E_{i}. $$

(10)

Most importantly, the payoff from cooperation increases relative to the baseline case without NGO participation. Again, the linear system of two equations has two endogenous variables, so the interior solution is unique. It is given in expression 6 in the main text.

Proof of Proposition 1

The equilibrium levels of cooperation are $\frac{1}{2c-1}$ without NGO monitoring and $\frac{1}{2c-1-L(1)}$ with it, so the difference is

$$ \frac{L(1)}{\left(2c-1-L(1)\right)\cdot(2c-1)}, $$

(11)

an unambiguously positive expression: realistically, cooperation always increases under NGO scrutiny. Differentiation with respect to L(1) yields a strictly positive expression. □

Proof of Proposition 2

Recall that expression 11 is strictly positive. Differentiating with respect to c yields a strictly negative expression. □

1.2 B. Mathematical Appendix: Extensions

In this mathematical appendix, we demonstrate that our main results are robust to plausible variations in model assumptions. We do so by first developing a general, asymmetric version of the model and showing that our main results remain unchanged. We then explore variants of this asymmetric model to show that our main finding regarding the effect of the legitimacy payoff L(1) on NGO participation is fully robust.

1.2.1 Asymmetric States

A substantively plausible generalization of the model is to allow asymmetric cooperation costs across states. This can be attained by replacing the common coefficient c with specific coefficients c _A , c _B  > 1. Note that this model is fully general, and it subsumes the symmetric model.

The equilibrium solutions 4 and 6 with and with NGO participation, respectively, are now replaced by expressions $E_{i}^{*}=\frac{1+2c_{j}}{4c_{1}c_{2}-1}$ and $E_{i}^{*}=\frac{(1+2c_{j})(1+L)}{4c_{1}c_{2}-1}$ (by choice of parameters that ensure an interior equilibrium). With a strictly positive value of L(1), equilibrium cooperation is strictly higher under NGO participation.

Consider next the effect of varying L(1). If L(1) is high enough, we have $E_{A}^{*},E_{B}^{*} \rightarrow 1$, so it follows that each state obtains a higher payoff. Thus, as long as the legitimacy payoff L(1) is high enough, states allow NGO participation. It follows that our main result continues to hold in the asymmetric variant of the model.

1.2.2 Unilateral Participation Allowance

Consider the general asymmetric model but suppose that state A can unilaterally decide on NGO participation prior to the cooperation subgame. The cooperation subgame remains unchanged, and so do the payoffs to state A. Thus, state A says ‘yes’ to NGO participation if and only if it obtains a higher payoff under NGO participation than otherwise. As shown in the previous subsection, this must be the case as long as L(1) is high enough.

1.2.3 Identifiable Defectors

Consider the general asymmetric model but suppose the external publics can identify whose cooperation failed. Thus, the legitimacy L is only a function of state i’s own success or failure. Now the legitimacy payoff to state i remains at ρ ⁰ without NGO participation. With NGO participation, the probability of the full legitimacy payoff is exactly E _i while the probability of zero legitimacy is 1 − E _i . Clearly, each state i has a higher best response to any given $E_{j}^{*}$, where j ≠ i, under NGO participation than otherwise. The strictly increasing best responses imply that both states select higher cooperation levels under NGO participation.

Regardless of $E_{j}^{*}$, state i obtains the maximal legitimacy payoff L(1) with probability $E_{i}^{*}$. Thus, as long as $E_{i}^{*}$ is high enough, because the cooperation cost coefficient c _i is low enough and L(1) is high enough, the probability of the zero legitimacy payoff for state i is negligible. Consequently, even if the other state j fails to cooperate, state i will not be blamed for the failure.

1.2.4 Endogenous Participation

Consider the general asymmetric model but suppose that if states A,B say ‘yes’ to NGO participation, the NGO must pay a cost P > 0 to participate. Suppose the payoff for the NGO from participation is some strictly increasing function V(E _A E _B ), so that the NGO prefers to maximize cooperation between the states.

Note first that the cooperation subgame is again, with and without the NGO, unchanged from the original cooperation subgame. Suppose now both states have said ‘yes,’ so that the NGO must decide on participation. If it rejects, its payoff is $V(E_{A}^{0}E_{B}^{0})^2$, where $E_{A}^{0},E_{B}^{0}$ denote the equilibrium cooperation levels without NGO participation. If it accepts, its payoff is $V(E_{A}^{1}E_{B}^{1})$, where $E_{A}^{1},E_{B}^{1}$ denote the equilibrium cooperation levels under NGO participation. With $E_{A}^{1},E_{B}^{1}>E_{A}^{0},E_{B}^{0}$, the cooperation subgame payoff to the NGO is higher under participation. Thus, the NGO participates if and only if the payoff difference exceeds the cost P.

Consider now the initial participation decisions by the states. If NGO participation benefits the states, as low as the cost P is long enough, a Pareto-efficient subgame-perfect equilibrium exists, whereby states A,B say ‘yes’ and the NGO accepts. If NGO participation harms the states, in any subgame-perfect equilibrium of the game at least one state says ‘no.’ When P is high enough, the NGO never participates, and no equilibrium of the game allows NGO participation.

1.2.5 General Functional Forms

Consider the general asymmetric model but suppose that the cooperation payoff to state i is K(E _A E _B ), where K is an increasing and strictly concave function such that K′(0) → ∞. Suppose the cooperation cost is C(E _i ), where C′(0) = 0 and C′(1) → ∞. Without NGO participation, in equilibrium we must have

$$ K'(E_{A}^{*}K_{B}^{*}) = C'(E_{i}). $$

(12)

Since both states benefit from cooperation, we select the highest equilibrium $E_{A}^{*}=E_{B}^{*}$ that exists.

Under NGO participation, we have instead

$$ K'(E_{A}^{*}K_{B}^{*}) + E_{A}^{*}E_{B}^{*} \cdot L(1) = C'(E_{i}). $$

(13)

Given that the best responses have positive slopes everywhere, it is immediate that the cooperation levels must increase relative to the case without NGO participation. Thus, as long as L(1) is high enough, the payoffs to both states must also increase.

1.2.6 Endogenous Revelation Effort

Let us now suppose that while states select cooperation efforts E _A , E _B , the NGO simultaneously selects a monitoring effort M ∈ [0, 1]. Suppose the cost of monitoring M is m(M), where M is an increasing and strictly convex function such that m′(0) = 0 and m′(1) → ∞. Suppose M is interpreted as the probability that the NGO obtains hard evidence about the success or failure of cooperation that can therefore be provided to the domestic publics for considerations. Suppose the NGO benefits from providing this evidence: if it obtains hard evidence, it obtains a payoff of 1.

Based on these assumptions, the NGO selects the monitoring M ^* such that m′(M ^*) = 1. Given this monitoring level, states understand that with probability M ^*, their legitimacy payoff is either L(0) or L(1), depending on failure or success, respectively. With probability 1 − M ^*, their legitimacy payoff is a constant. Thus, the states maximize a convex combination of payoffs with and without NGO participation (asymmetric variants of expression 1 and 2), with weights 1 − M ^* and M ^*, respectively. Thus, as M ^* increases, so does the incentive to cooperate by the resulting first-order conditions. Given that the case without NGO participation is equivalent to M ^* = 0, the main results continue to hold. Similarly, as long as L(1) is high enough, states allow NGO participation.

1.2.7 Compensation for NGO Participation

Suppose each state i = A,B obtains a bonus B > 0 if NGO participation is allowed in equilibrium. The model is otherwise unchanged except that the marginal payoff from allowing NGO participation must be strictly less than − B for NGO monitoring not to occur in equilibrium.