• No results found

Incentives in Development Lending: Technical Cooperation

N/A
N/A
Protected

Academic year: 2021

Share "Incentives in Development Lending: Technical Cooperation"

Copied!
27
0
0

Loading.... (view fulltext now)

Full text

(1)

Incentives in Development

Lending: Technical

Cooperation

January 2011

CORE Metadata, citation and similar papers at core.ac.uk

(2)

Incentives in Development

Lending: Technical

Cooperation

Josepa

Miquel-Florensa

January 2011

(3)

Incentives in Development Lending: Technical

cooperation

Josepa Miquel-Florensa

Toulouse School of Economics (ARQADE)

January 18, 2011

Abstract

This paper models incentives and information asymmetries between the di¤erent participants in multilateral development banks’decision process, namely borrowing countries, managers and the board of governors (with borrower and non-borrower members).

We propose technical cooperation requirements as instruments for the board of governors to solve the moral hazard and adverse selection prob-lems. We assume technical cooperation makes the probability of success not to depend on the agent’s e¤ort choice, as long as he provides e¤ort, but on the principal’s distribution of resources, and may also provide private bene…ts to the recipients. Moreover, the outcome of the agent’s invest-ment is a ’public good’ since is enjoyed in a non-rival fashion by both Board and agents.

Using data on project performance reports from the IADB, we show that technical cooperation does have an impact on project results. Fur-thermore, we are able to di¤erentiate for which projects, contract and recipients technical cooperation is more e¤ective and relate the reported problems to the information asymmetries.

1

Introduction

Governance of Multilateral Development Institutions is an important part of the development aid public debate, and is attracting increasing attention in academia1. Lack of accountability and de…nition of responsibilities have been

Email: [email protected] I would like to thank Yuri Soarez, Francisco Gonzalez, and semianr participants at IADB and TSE for usefull comment and suggestions, and Martin Naranjo and Matias Mednik for very interesting discussions on early stages of this project. All errors remain my own.

1Among others, Easterly (2003) and Rowat and Seabright (2006) stress in their models

the di¢ culty to measure output of aid agencies and the weak incentives faced by the several decision makers in the process. Hawkins et Al (2006) approaches delegation and agency problems from a political economy point of view. IMF (2008) and Mednik (2010) look in detail at the Governance structure at the IMF and IADB respectivelly.

(4)

highlighted as important contributors to the governance problem. A quick look at the websites of the Development Banks su¢ ces to note their focus on funds disbursed rather than services delivered and/or outputs and ultimately devel-opment.

To analyze the incentive scheme, we need to look at the main players in the special governance structure of Development Banks are (1) the Board of Gover-nors that has the authority to approve the loans proposed by the Management, (2) the Management that designs the projects and whose incentives are some-how aligned with lending portfolios sizes and (3) the Recipient Governments that, as members of the credit cooperative, can not be discriminated on interest rates and have their own political agendas.

Development lending faces two important information problems. On the one hand, it is di¢ cult for the Development lenders to observe the preferences of the recipients over the di¤erent projects available. What will work for each given recipient is their private information, and even in the case of perfect revelation of these preferences, there are still a number of controllable and uncontrollable factors that in‡uence the outcome of the projects implemented. On the other hand, it is necessary to provide the borrowers appropriate incentives to invest the lend funds and to exercise the appropriate e¤ort to maximize the projects’ returns and their probability of success.

In Development lending, these well-known information problems get an ad-ditional twist: the goals (and incentives) from the players involved at the di¤er-ent stages of the project are disaligned. Even if Developmdi¤er-ent is a ’public good’ enjoyable simultaneously by all the involved parties, and hence the piece-rate instrument is not feasible, the designers of the loan contracts have a private agenda, that includes among others lending volume, that is not always aligned with the development goals of the borrowing countries. Moreover, two peculiar-ities of multilateral lending institutions need to be noted. First, they are credit cooperatives, and hence it is not possible to discriminate among borrowers by o¤ering di¤erent interest rates. And second, both borrowing and non-borrowing countries are part and voice on the Board, what leads to voting strategies not always aligned with project quality.

Given the governance structure of Development Banks, and given the polit-ical organization of the board, it is very rare that Management’s proposals are not approved. The instruments to provide incentives to recipient governments to perform e¤ort in the use of the transferred funds are limited: interest rate discounts in case of success are not feasible in credit cooperatives, and threats to cut ‡ows of funds if projects fail to show results are not credible2. We look

2The lack of enforcement of contract conditions has been related to the Samaritan’s

Dilemma. This concept, introduced by Buchanan (1975), argues that he principal can not credibly commit to stop the relationship with the agent until the projects are accomplished, since it is common knowledge that he cares about the agent and the projects to be completed. In multilateral lending, the problem is worsened by the fact that both borrowers and lenders have a say on the funds allocations.

Killick (1998), Dreher (2002) and The World Bank (2005) present reviews of the literature and examples of time inconsistency involving conditional aid contracts. The theory literature on time consistency is extense, including for example Matsuyama (1990) and Rey and Salanie

(5)

at how these two information problems a¤ect the success of the projects from two angles. We start modeling these problems from a principal-agent approach, looking at the di¤erent instruments available to each of the players and suggest constraints that could be included on the planning-approval process to align the objectives of all parts. We verify the assumptions of our model on technical cooperation e¤ectiveness and board’s ability to screen projects using data on project performance reports and board voting from the IADB.

We di¤er from the standard principal-agent settings on three points: (i) The principal’s utility is the manager’s utility, and is independent of the suc-cess of the project implemented by the agent-borrower. It is widely agreed that principal-manager’s utility does not always match with the recipients and donor’s preferences. Hence, we give attention to the possible constraints the Board could impose on the projects to be considered so that is recipient’s devel-opment outcome and not manager’s lending volume what is maximized. (ii) The size of the probability of success does not depend on the agent’s e¤ort choice, as long as he provides e¤ort, but on the principal’s distribution of resources. And (iii) borrowers bene…t from the success of the projects and on the capacity externality/private bene…ts provided by technical cooperation transfers. Hence, the outcome of the agent’s investment is a ’public good’ since is enjoyed in a non-rival fashion by both Board and agents.

Our theorethical model recommends greater technical cooperation transfers when information problems are important, and that these transfers should be greater the greater is the e¤ect of the technical aid on the probability of success of the projects. Moreover, the model suggests that even when technical cooperatio is inefective, it may be on the principal’s best interest to provide it to induce e¤ort and revelation of information.

We …nd evidence that, as assumed in the theoretical model, technical coop-eration attached to the projects has a positive e¤ect on reducing the probability of commitment problems and increasing the probability of the donor doing a good tracking of the results. Also in accordance with the theorethical predic-tions, we …nd that sectors where it is relatively easier to evaluate the e¤ectivity of the measures, like infrastructure or energy, have ceteris paribus better data tracking. In contrary, sectors like education with predictability are more likely to present commitment problems and contract conditions compliance delays. This highlights the importance of adverse selection of projects for sectors where there is a lot of local information and idiosyncratic risk a¤ecting the project’s success probabilities.

We are able to distinguish the e¤ect of project-attached technical coopera-tion loans from general sectoral technical cooperacoopera-tion grands. We …nd the …rst type signi…cative on a¤ecting contract conditions ful…lment, while the e¤ect of the sectoral grands more important on coordination issues and data tracking, problems more related to sectoral capacity.

The paper is structured as follows. In the next section, we present the the-orethical modeling of the main information problems on development lending,

(6)

and suggest the characteristics of the optimal contract. In section 2 we present the empirical analysis of the relationship between contract characteristics and projects returns, and the policy implications of our …ndings. Last section con-cludes.

2

Modeling approach

Our objective is to look into the incentive schemes faced and instruments avail-able to each of the individuals involved in the planning and implementation of Development Banks operations.

We center, on the one hand, on the basket of products o¤ered by the bank to each client. We look in detail into one of these products, technical cooperation, for its special characteristics in terms of a¤ecting the risk of the operation(s) and its externalities to subsequent projects and to the sector capacity in general. We derive, under di¤erent information settings, which should be the optimal allocation of funds to loans and technical cooperation.

On the other hand, we look at the inventives of each of the players in the Development banks organigram. We study how di¤erences in player’s prefer-ences and the existing information problems can introduce ine…ciencies in the contract provided. Let us describe the players’incentives and the information advantages.

(1) Clients/ Loan recipients:

Borrowing governments value development outcomes from the loans, since their reelection probability may be closely related to the projects success, but also receive personal bene…ts from the use of the transfers, specially of ’free funds’like technical cooperation3.

Clients have private information on the development potential of the projects considered and take (unobservable by the management) decisions on the use of the borrowed resources.

(2) Bank management:

It is in the hands of the bank management to propose and design projects/contracts to be o¤ered to the clients. Given the uniformity on the loan conditions, since credit cooperatives can not discriminate among clients on interest rates, and given the commitment problems involved in conditionality, the management has limited instruments to design mechanisms to overcome the moral hazard and adverse selection problems.

Management utility/rewards depend on lending volume on top of projects funded returns.

3How much they like these funds may depend on the procurement/disbursement rules

attached to them, but for simplicity we assume that utility only comes from non-reimbursable transfers. The discussion on fungibility of funds is out of the scope of this paper.

(7)

Management’s e¤ort on project design is unobservable by the Board, and may be also unobservable by the clients.

(3) Board4:

Decides on approval of the loans proposed by the management.

Can not always distinguish among the projects proposed by management the good ones, and is di¢ cult also to evaluate how much e¤ort was put on their preparation by management (and clients). If the board could be sure that return from the project was the only element in management’s utility, unobservability would not be a problem since board prioriites would be aligned with managements’.

The questions we want to answer are how should the optimal contract look like in a setting with moral hazard and adverse selection where the principal has limited instruments and needs to delegate project design, and what can be done to align incentives of all players and to induce revelation of information.

Our goal is to study which constraints on project design should be established to align the incentives of all participants. It is beyond argument that each participant has comparative advantage at his tasks in the project process. The projects that reach the board for consideration are likely to be very diverse, and that the board members may lack the necessary knowledge for evaluation and analysis of risk and implementation challenges that these projects may face in the future. Preestablished constraints to be satis…ed by project proposals before submission may mitigate these evaluation problems5.

Development banks o¤er their member borrowing countries a complex set of services. Given the disalignment of objectives, the interaction among all the products o¤ered (that include among others guarantees, policy design and ca-pacity building) raises an additional concern on the e¢ ciency of the agreements proposed by the management: is it feasible for the borrower and/or the board to add constraints on the packages that include all these services o¤ered? In which of these products should borrowing countries exert higher pressure, for example in planning or monitoring, before accepting the management’s proposed agree-ment? And how will the optimal strategies of the borrowing countries in this regard vary thought time as successive stages of development are reached?

The interaction among all the products ‘packaged’ by the management in the country strategies can be seen from two perspectives. On the one hand, we have the bargaining between the multilateral institutions and the recipient countries in the de…nition of the ‘joint country strategies’. Both the board and management face a multidimensional information problem: what works in

4We look at the Board from the information assymetries point of view, treating it as a

unique player and not getting into its internal dynamics. For more detailed information on Multilateral Banks Board’s internal functioning see Mednik (2010).

5It needs to be noted that some countries may have special request for their board members

for voting. We refer here not to quality controls at voting but as screening devices before being considered for vote.

(8)

each sector, and which are the preferences of the borrowers over the sectors6. The debt of the country, its relative size in the multilateral institution and the number and size of development partners acting in the country are likely to be key elements in the bargaining process, and they can ultimately determine the development prospects of the agreement attained. On the other hand, we can look at the problem from the borrower point of view and examine the actions in his hands to align objectives with the management that designs the country package/projects. It is interesting to think up to which point all the players in the process are able to include in the petitions conditions for the monitoring of results and constraints in funds allocations as to align incentives at every planning stage with development outcomes.

We center the modeling on one of the products o¤ered by the bank’s man-agement: technical cooperation. The choice is due to two peculiarities of tech-nical aid: it decreases the ’risk’of the project it is attached to, increasing their probability of success, but also generates positive externalities in the sector this project is implemented, in terms of improved capacity available for future projects. Moreover, technical cooperation provides private bene…ts to the re-cipients, what can make it a reward for less attractive projects from the client point of view. Technical aid is a widely used instrument to improve the success of funded projects, and is mainly provided through grants. In the data analyzed in this paper, for the period 2000-2008 at the IADB, 93% of the technical co-operation projects approved where non-reimbursable, representing 60% of the technical cooperation funds approved. While in the modeling approach we con-sider only one type of technical cooperation in form of grand, on the empiric approach we look at the di¤erent performance of both grand and loan technical aid.

The questions we want to answer are (1) how should the optimal contract look like in a setting with moral hazard and adverse selection where the principal has limited instruments and need to delegate project design, and (2) what can be done to align incentives of all players and to induce revelation of information with the aim of maximizing development returns from the lend funds.

Our goal hence is on the one hand to study the optimal allocation of funds to loan and technical cooperation as function of the information problems. And on the other hand, we want to study which constraints on project design should be established to align the incentives of all participants. It is beyond argument that each participant has comparative advantage at his task in the project process. The projects that reach the board for consideration are likely to be very diverse, and the board members may lack the necessary knowledge for evaluation and analysis of risk and implementation challenges that these projects may face in the future. Preestablished constraints to be satis…ed by project proposals before submision may mitigate these evaluation problems7.

6In the foreign aid literature, recipients’ and donors’ preferences over sectors have been

sometimes infered by Government budget distribution among the ministries/sectors. See for example Noel and Therien (1995).

7It needs to be noted that some countries may have special requests for their board members

(9)

We analyze the agency problems presented above in a principal (manage-ment) agent (clients/borrowers) framework. We look at the management’s problem of allocating funds between loans and technical cooperation when his objective is to maximize lending and he needs to provide the agents incentives to invest the lend funds and to self-select into the contract tailored to their unob-servable project returns. On the same setting, we suggest constraints the Board could consider for project approval aiming to align management objectives with client’s development outcomes.

We di¤er from the standard principal-agent settings on three points: (i) The principal’s utility is the manager’s utility, and is independent of the suc-cess of the project implemented by the agent-borrower. It is widely agreed that principal-manager’s utility does not always match with the recipients and donor’s preferences. Hence, we give attention to the possible constraints the Board could impose on the projects to be considered so that is recipient’s devel-opment outcome and not manager’s lending volume what is maximized. (ii) The size of the probability of success does not depend on the agent’s e¤ort choice, as long as he provides e¤ort, but on the principal’s distribution of resources. And (iii) borrowers bene…t from the success of the projects and on the capacity externality/private bene…ts provided by technical cooperation transfers. Hence, the outcome of the agent’s investment is a ’public good’ since is enjoyed in a non-rival fashion by both Board and agents.

2.1

Description of the model

The principal o¤ers the agent a loan of size p that has to be returned with an interest rate r at the end of the period, and a ‡ow of funds T in form of technical cooperation for a given project. The principal has a budget B for each type of agent to be allocated between project loan and technical cooperation.

The agent faces an investment cost ; that we assume independent of the size of the loan, and, once investment is performed, the project is successful with probability . We consider two possibilities: either technical cooperation increases the probability of success of the projects once investment has been performed, or it does not a¤ect probability of success. On the …rst situation, we assume this relation is given by the increasing and concave function (T ): We solve the model under the …rst assumption, and we comment afterwards how he optimal allocation would change on the case of non-productive teachnical cooperation.

Borrowers8 have private information on the returns of the projects in case

of success. We consider two types of agents: the ones that obtain high returns (wh) and the ones that obtain low returns (wl) in case of success, that are known

to represent a proportion and (1 ) of the applicants pool respectively. To avoid problems with limited liability, we assume that in case of failure of the

projects before being considered for vote.

8As borrowers we consider either two di¤erent countries with di¤erent returns, or two

projects at same or di¤erent sectors of a given country. The presented formulation does not consider possible project interactions, more likely to appear under the latter assumption.

(10)

project or when there is no investment, the agent gets return wf = r, the agent

is always able to pay back.

Agent’s i expected utility from a contract (pi; Ti) is

V (Ti; Bi; wi) = v(Ti) + [ (Ti)u((wis r) pi)] (1)

where wis represents agent i return in case of success (for i = h; l), denotes

the time discount and Bi the budget allocated to each type of project. This

utility is increasing both in the size of the loan and on the technical cooperation received. We assume that technical cooperation gives per-se utility to the agent, denoted by the increasing and convex function v(Ti), which can be interpreted as

the long term non-pecuniary bene…ts of improving for example service delivery mechanism, government structure, or simply as private bene…ts from the ’free funds’. And the increasing and concave function u() represents the recipient’s utility from the net return from the project.

2.2

Welfare maximizing allocation

The full information welfare maximizing allocation is given by the solution of the following maximization problem

max

ph;pl;Th;Tl

[ (Th)u((wh r) ph)] + (1 ) [ (Tl)u((wl r) pl)] (2)

s:t: ph+ Th= Bh (3)

pl+ Tl= Bl (4)

where Bi is the budget allocated to each type of project.

Proposition 1 (Observable e¤ort and project type) For a given budget for each type of projects, size of the loan is decreasing with its return in case of success, hence pl > ph, and technical cooperation is decreasing with project return Tl < Th: Both p and T increase with the budget allocated to the project. Proof. Plugging the budget constraints, the …rst order conditions are:

0(T

h) [u((wh r) (Bh Th))] = (Th)u0((wh r) (Bh Th)) (wh r) (5) 0(T

l) [u((wl r) (Bl Tl))] = (Tl)u0((wl r) (Bl Tl)) (wl r) (6)

and deriving implicitly (5) and (6) we see that size of the loan is decreasing with its return in case of success, pl > ph, and technical cooperation is increasing with project return. Both technical cooperation and loan size increase with budget allocated to the project.

This result is intuitive: the greater the project return, the smaller is the marginal utility of an extra unit of loan but the greater is the return of an increase in the probability of the good state of the world realizing.

We consider now the optimal allocation of funds when returns from the project and agent’s e¤ort are unobservable for the principal, in a setting

(11)

where outcome of the projects can be enjoyed simoultaneuoly in a non-rival fashion by principal and agent, but each of them may receive private utility from a part of the contract. For the adequate alignment of incentives of all players, the menu of contracts provided should be such that (1) it is in the agent’s best interest to invest the received funds, (2) agents self select to the contract targeted to their project returns and (3) objectives of the principal are aligned with welfare maximization. We proceed sequentially to study the e¤ect of each of these requirements on the welfare maximizing contract.

We start by the setting where principal’s objective coincides with the plan-ner’s objective and project type is observable. When we consider only the moral hazard problem, the incentive compatibility constraint for agent i is given by:

v(Ti) + [ (Ti)u((wi r) pi)] > u(Ti+ pi) (7)

and given the assumption that a budget Bi is allocated for each agent/project

type i, we can rewrite (7) as

v(Ti) + [ (Ti)u((wi r) (Bi Ti))] > u(Bi)

Proposition 2 (Moral hazard) Optimal loan is smaller or equal than under observable actions when use of the resources by the agent is unobservable by the principal.

Proof.Let hand lbe the Lagrange multipliers for the incentive compatibility

constraints for high and low return projects respectivelly. From the …rst order conditions we get d( (Th)u((wh r) (Bh Th)) dTh = h + h v0(Th) < 0 d( (Tl)u((wl r) (Bl Tl)) dTl = l 1 + l v0(Tl) < 0

and hence we …nd that pM H

l 6 pl and pM Hh 6 ph:

Smaller loans and higher technical cooperation increase the probability of success and the private bene…t the agents obtain from T . Once investment is unobservable it is necessary to decrease the risk faced by the agents and provide greater private bene…ts to induce investment.

When the contracts o¤ered are adapted to the agent’s characteristics (i.e. to project returns), for each type i it needs to be true that

V (Ti; Bi; wi)> V (Tj; Bj; wi)

v(Ti) + [ (Ti)u((wi r) pi)] (8)

> v(Tj) + [ (Tj)u((wi r) pj)]

(12)

From the moral hazard result, we know that high types get greater technical cooperation. Then, in the adverse selection situation, low types would claim the high types contracts, all borrowers will claim their project has high returns. Claim 3 (Adverse selection: information rents) When there is informa-tion rent and same budget is allocated to all projects, this informainforma-tion rent is for the low type.

Proof. From Proposition 1, we know that optimal allocation to technical coop-eration is increasing with the project return. Let (Th; Tl) be the full informa-tion allocainforma-tions to technical cooperainforma-tion. Then, for each type i it is true that

(Ti)u((wi r) (B Ti))> (Ti)u((wi r) (B Ti)) for all feasible Ti0s:

For the high type, Th > Tl and hence

v(Th) + [ (Th)u((wh r) (B Th)]> v(Tl) + [ (Tl)u((wh r) (B Tl)]

high type is not willing to claim the contract aimed to low types. For the low type, v(Th) > v(Tl) and

[ (Tl)u((wl r) (B Tl)]> [ (Th)u((wl r) (B Th)]

low type may be willing to lie to get the high type package whenever extra-utility from technical cooperation compensates for the change in expected bene…t from the project.

The principal needs to ensure that self selection constraint for the low type is satis…ed. Either full information contracts are such that agents self-select, or the low type needs to be compensated with greater technical cooperation for him not to take high type contract. Putting together both self-selection constraints we have that

(Tl) [u((wl r) pl)] (Th) [u((wl r) ph)]6

v(Th) v(Tl)

u(Th) u(Tl) 6 (T

l) [u((wh r) pl)] (Th) [u((wh r) ph)]

that we can write as

(Tl) [u((wh r) pl) u((wl r) pl)]6 (Th) [u((wh r) ph) u((wl r) ph)]

(9) Proposition 4 (Adverse selection) When there is adverse selection, low type gets smaller loan and high type gets greater loan than under complete informa-tion.

(13)

Proof.Let be the Lagrande multiplier for (9) : From the …rst order conditions we get d( (Th)u((wh r) (Bh Th)) dTh = [ 0(Th) [u((wh r) ph) u((wl r) ph)] (Th) [u0((wh r) ph) u0((wl r) ph)]] > 0 (10) d( (Tl)u((wl r) (Bl Tl)) dTl = 1 [ 0(T l) [u((wh r) pl) u((wl r) pl)] (Tl) [u0((wh r) pl) u0((wl r) pl)]] < 0 (11) and we have pASl < pl and pASh > ph : low types get greater or equal technical cooperation than under full information to give them incentives to self-select, and at the same time high types get less technical cooperation to decrease low’s desire for hight’s contract.

This result is intuitive: share of funds allocated to technical cooperation for each type of project is less di¤erenciated to compensate for the low type’s willingness to claim the contract with highest share.

It needs to be noted the correlation between project returns and risk: high return projects get higher technical cooperation and hence are less risky even with same budget allocated to them as for low return projects. This di¤erence becomes smaller as information problems are taken into account: when moral hazard and adverse seletion enter the game, contracts for the two types become less extreme than the complete information ones to ensure investment incentives and information revelation.

The problem of a ’benevolent manager’that can not observe neither investment nor project returnsis given by:

max ph;pl [ (Th)u((wh r) ph)] + (1 ) [ (Tl)u((wl r) pl)] s:t: ph+ Th= Bh pl+ Tl= Bl v(Tl) + [ (Tl)u((wl r) pl)] > u(Tl+ pl) (12) v(Th) + [ (Th)u((wh r) ph)] > u(Th+ ph) (13)

(Tl) [u((wh r) pl) u((wl r) pl)]6 (Th) [u((wh r) ph) u((wl r) ph)]

(14) We start by looking at the incentive compatibiltiy and self-selection con-straints that need to be satis…ed.

Claim 5 (IC high not binding with MH and AS) When budgets are the same for both projects, incentive compatibility constraint for the high type does not bind.

(14)

Proof. Let V (ci; wj) be the utility that contract aimed to agent type i provides

to agent j, that is increasing in wj: From the constraints (13) ; (12) and (14) we

have

V (ch; wh)> V (cl; wh) > V (cl; wl)> u(Ti+ pi) = u(B)

where B denotes (equal) budget allocated to each of the projects. This implies that high type’s incentive compatibility constraint is not binding: high type gets a bonus to sign the contract and implement e¤ort.

The intuition of this result is as follows: if there is an agent that does not need to be provided incentives to invest the funds, that will be the agent who gets the highest returns out of the investment.

Claim 6 (Adverse selection and moral hazard) When the principal faces both moral hazard and adverse selection problems, he o¤ ers low type agents greater loans and high type agents lower loans than under full information. Proof. Let and be the Lagrande multipliers for (12) and (14) respectively. The …rst order conditions for the high type is given by (10), and for the low type is d( (Tl)u((wl r) (Bl Tl)) dTl = 1 + [ 0(T l) [u((wh r) pl) u((wl r) pl)] (Tl) [u0((wh r) pl) u0((wl r) pl)]] + 1 + v 0(T l) < 0

This result is intuitive: when both adverse selection and moral hazard enter the game, the principal has to ensure low type does not want to mimic high type (and can do that making their contracts more similar), and also needs to ensure low types have the right incentives to invest (and get a big enought incentive compatibility to reduce the risk of investing the funds).

2.3

Alignment of incentives: management’s allocation and

board’s strategies

We consider as principal in this development setting the management of the development banks that decides how to allocate the funds among the multiple products o¤ered by the institution. Given the internal reward structure, these managers have special incentives in increasing their loan portfolio, but would be too strict to assume that development returns from the projects are not a part of their utility. Hence, we de…ne the management-principal’s utility as:

Up= pi+ pj+ [ [ (Th)u((wh r) ph)] + (1 ) [ (Tl)u((wl r) pl)]]

(15)

Claim 7 (Alignment of incentives) When management has biased prefer-ences towards loan amounts, shares allocated to teachnical cooperation are smaller than in the planer’s allocation.

Proof. Looking at the …rst order conditions for the full information case, we …nd that d( (Th)u((wh r) (Bh Th)) dTh 1 = 0 (1 ) d( (Tl)u((wl r) (Bl Tl)) dTl 1 = 0

and we see that the greater the closest is the allocation to the planner’s recipient’s return maximizing choice. At the limit, when does to in…nity the two allocations coincide.

Hence, given this bias in prefenences, to consider minimum allocations to technical cooperation for a project to be considered for board approval seems a reasonable suggestion to improve the alignment of incentives.

3

Empirical evidence and policy implications

To support the modeling approach, we looked into the available data to answer the following questions: (1) Can we explain project’s performance problems by recipient’s and contract characteristics? (2) Does technical cooperation decrease the probability of problems on the project’s implementation? Is technical coop-eration really an instrument that increases the success of the projects, in terms of timely reaching their objectives? and (3), Does the Board voting contain any information on the project’s future viability?

3.1

Data description

The main constraint found to answer these questions is the lack of clear de-…nition and reliable data on outputs. The lack of measurable dede-…nitions of outcomes in many projects, together with the vague incentives of the responsi-ble of the data collection, may raise concerns on the quality of the information available. Another important quality concern on the data available is the lack of information about the risk of the projects and how the non-idiosincratic part of it could be minimized with the available tools. From the Project Performance Monitoring Reports (PPMR) we obtain information on ’Factors for unsatisfac-tory implementation’: Donor maintaining performance data, design (objectives agreed with the bene…ciary and community political opposition) and implemen-tation (counterpart funding shortfall, compliance delays on contract conditions, procurement di¢ culties and costs overruns).

We have data on PPMR for 324 projects. For each project, we create a unique score for each variable for three types of indicators of project problems9:

(16)

performance on data collection10, design (objectives agreed with the bene…ciary and community political opposition) and implementation (counterpart funding shortfall, compliance delays on contract conditions, procurement di¢ culties and costs overruns)11. Table 1 shows the summary of the reported problems.

With respect to technical cooperation attached to the projects, we consider both the grand allocations per sector-country and the technical cooperation loans attached to speci…c project loans. Only 3.4% of the matched dataset corresponds to reimbursable technical cooperation and 10.53% of the projects can be matched to a PEF.

For the …rst case, given the delays between implementation and returns from capacity building of technical cooperation sectoral grands, we assign to each loan the grands per sector-country on the approval year.

For project attached loans, we have information on 64 PEF (Project Prepa-ration and Execution Facilities), loans with the goal of ’strengthen the preparation of projects of the operational program’and ’…nance the start-up fo the project prior to …rst disbursement and lay the framework for institutional stability’. When matched with PPMR information, we are left with 32 PEFs for the period 2003-2008.

Our information on Board Voting includes 530 projects for the 2003-2008 period, from which only 3.2% got abstentions and 2.8% got votes against.

We complement the dataset with information on the country’s portfolio char-acteristics from the World Bank World development indicators. This informa-tion includes multilateral debt as percentage of total debt, concesional debt as percentage of external debt, total debt service as percentage of GNI, average grand element on new external debt commitments, o¢ cial development assis-tance, and interest payments on external debt. These variables are included with the aim of controling for the country’s burden of the debt and the aid re-ceived, as measures of the debt importance on the country’s concerns and how it could a¤ect the e¤ort on the management of the received funds.

3.2

Results and policy implications

We present a theorethical model on allocation of funds between loans and tech-nical cooperation when there are moral hazard and adverse selection problems. We show that the optimal shares of funds allocated to technical cooperation are greater than the full information ones, and that these di¤erences increase as the information problems become more important. The underlying assumption of the theorethical model is that technical cooperation increases the probabil-ity of success of the projects. We check this assumption on a rich dataset on

1 0On PPMR Monitoring and Evaluation part, it is asked "is the borrower maintaining

performace data on agreed outcome/output indicators?"

1 1PPMR’s ask for critical factors/reasons for unsatisfactory implementation progress,

(17)

Project Performance Monitoring Reports from the IADB for the period 2003-2008, matched with Project preparation facilities (technical cooperation) data. In Table 2 we look at commitment problems, and we …nd that technical cooperation grands have a concave e¤ect on the probability of problems, as expected, and the share of concesional debt and ODA received increase the probability of these problems. The result is intuitive: the less important is debt, the cheaper it is and the more available are alternative channels, the more likely is to …nd compliance problems. Board abstentions, even if not signi…cative, have the expected positive e¤ect on the probability of commitment problems, as expected. When looking at compliance delays, on Table 3, we …nd that both types of technical cooperation show the expected signs, being the project related more signi…cative.

For 30% of the projects borrowers did not maintain performance data, even when the lack of monitoring and evaluation systems is almost negligible. As showed in Table 3, we …nd that performance data is more likely to be collected for sectors where measurability is easier, like infrastructure investments, and for contract structures linked to performance. We …nd that inter-agency coordina-tion problems, concerns on performance by the counterparts and the existence of an M&E system have positive impact on the donor’s indicator tracking. It needs to be highlighted the signi…cant e¤ect of board abstentions, showing the boards action as a good predictor of problems.

Table 5 analizes inter-agency coordination problems, showing signi…cance of abstentions and, even if not signi…cant, showing that government’s aid ‡ows and loans portfolio size increase the probability of coordination problems between donor agencies.

Given the concern on the possible correlation among the projects, we present in Table 6 the bivariate analysis for performance data tracking and compliance delays. We …nd the results mimic the univariate analysis, both with respect to which technical cooperation plays a more important role in each type of project and with respect to abstention votes conatining information on project quality. These results show that technical cooperation does have an e¤ect on the probability of the project presenting implementation problems, and that each type of technical cooperation considered (attached to the project and general sectoral grands) has special e¤ect. While sectoral technical cooperation is sig-ni…cant for data tracking, project speci…c technical cooperation has specal e¤ect on the project contract conditions compliance.

4

Conclusions

This model shows how high levels of technical cooperation compared to the full information benchmark may be necessary to provide the agents incentives to invest the received funds and to self-select to the contract aimed at their type. We show that under full information, the distribution of funds that maxi-mizes expected development returns implies higher technical cooperation is pro-vided for high value projects. The intuition is that, for higher return projects,

(18)

the increase in probability of success has higher e¤ect on expected returns in comparison to increasing the size of the loan, as opposite of what happens for low return projects.

When project quality is unobservable, low types are the candidates for infor-mation rents. The principal needs to increase their technical cooperation allo-cations to stop low type agents to claim high type contract. To have a contract that induces low types to reveal their type it is necessary that the di¤erence in technical cooperation between the contracts is compensated by higher expected development returns, and that is obtained with a smaller gap in technical co-operation compared with the full information situation. Intuitively, all agents are willing to claim high return projects if that implies higher private bene-…ts, and to reduce this incentive to lie they should be provided higher technical cooperation.

When the agent’s use of the lend funds is unobservable, we …nd that the incentive compatible contract provides smaller loans and greater technical co-operation for both types of agents. Agents need to be compensated for their e¤ort cost with higher private bene…ts from technical cooperation and a reduc-tion of the risk faced.

These results highlight the need to include technical cooperation and its e¤ect on the project’s probability of success into the management’s incentive scheme to ensure adequate incentives are provided to the agents to invest the transferred funds and to self-select to the contract that maximizes their devel-opment expected returns. After checking in the data the e¤ectiveness of the technical cooperation, we …nd that the constraint imposed on minimum techni-cal cooperation should be an increasing and convex function of the size of the loan.

When technical cooperation is not productive, i.e. it does not a¤ect the probability of success, part of the results presented hold true: technical cooper-ation helps to satisfy incentive compatibility constraint when projects are not highly valued by the agent, and di¤erent amounts would be optimal for infor-mation revelation when instead of project returns is agent’s e¤ort cost that is unobservable.

We …nd that technical cooperation decreases the probability of implemen-tation problems, and that sector characteristics and coordination of partners also play an important role. Moral hazard, in terms of making the players ac-countable for result, and adverse selection, in terms of objectives agreed with bene…ciaries and recipient’s political opposition, are shown to play an important role on the performance of the projects.

References

[1] Buchanan, J.M. "The Samaritan’s Dilemma." In Phelps, E. (Editor): Altruism, morality and economic theory. NY, Russell Sage Foundation.

(19)

[2] Drehen, A. "The Development and Implementation of IMF and World Bank Conditionality." Discussion Paper HWWA (2002).

[3] Easterly, W. “The cartel of good intentions: The problem of bureaucracy in Foreign Aid” Journal of Policy Reform 2003

[4] Hawkins, D. et Al. "Delegation and agency in international organiza-tions" Cambridge Univ. Press (2006)

[5] Holmstrom, B. and Milgrom,P. "Multi-Task principal-agent analysis: Incentive contracts, asset ownership, and job design." Journal of Law, Eco-nomics and Organization (7) 1991.

[6] International Monetary Fund, Independent Evaluation O¢ ce "As-pects of IMF Corporate Governance, Including the role of the Executive Board". 2008

[7] Killick, T. "Aid and the Political Economy of Policy Change." With Gu-natilaka, Ramani and Marr, Anna. Routledge (1998).

[8] Matsuyama, K. "Perfect Equilibria in a Trade Liberalization Game". American Economic Review 80(3) 1990

[9] Mednik,M. "Notes on the Governance Structure of the Inter-American Development Bank" 2010

[10] Meyer, M and Vickers, J. "Performance Comparison and Dynamic Incentives". Journal of Political Economy 105 (3) 1997

[11] Noel, A and Therien, J-P. "From domestic to international justice: the welfare state and foreign aid". International Organization (1995)

[12] Rey, P and Salanie, B. ’Long-term, Short-term and Renegotiation: On the value of commitment in contracting". Econometrica (1990)

[13] Rowat, C. and Seabright, P. “Intermediation by Aid Agencies”Journal of Development Economics 79 (2006)

[14] Sinclair-Desgagne, B. "How to restore higher-powered incentives in Multi-task agencies." Journal of Law, Economics and Organization 15 (1999).

[15] The World Bank Review of World Bank Conditionality. July 2005. [16] The World Bank Conditionality Revised: Concepts, Experiences and

(20)

Table 1: Averages

Constant Abstention Against (1) Objectives agreed with beneficiary 0.9697 -0.0771** 0.0302 (2) Borrower maintaining performance data 0.7146 -0.2793 ** 0.1603 (3) Counterpart funding shortfall 0.0393 -0.0204 0.0393 (4) Community political opposition 0.0262 -0.0136 -0.0262 (5) Contract conditions compliance delays 0.3324 -0.1328 0.3675 ** (6) Procurement difficulties 0.2489 -0.1294 0.0510

(7) Costs Overruns 0.0098 -0.0051 -0.0098

Argentina

AbstentionsAgainst total Objectives Performance Counterpart Political op. Conditions Proc. Problems Cost Overruns

Projects 3 1 37 P. Abstention 1 0 0 0 0 0 0

w/ PPMR info 1 1 29 P Against 1 1 0 0 0 0 0

P total 1 80% 0 0 23% 7% 0

(info available) 52% 48% 100% 100% 100% 100% 100%

Bolivia

AbstentionsAgainst total Objectives Performance Counterpart Political op. Conditions Proc. Problems Cost Overruns

Projects 1 0 21 P. Abstention n/a n/a 0 0 0 0 0

w/ PPMR info 1 0 17 P total 1 71.43% 0 5.88% 52.94% 23.53% 0

(info available) 70.59% 76% 1 1 100% 100% 1

Chile

AbstentionsAgainst total Objectives Performance Counterpart Political op. Conditions Proc. Problems Cost Overruns

Projects 1 0 22 P. Abstention n/a n/a 0 0 0 0 0

w/ PPMR info 1 0 18 P total 1 71.43% 0 0 11.11% 16.67% 0

(info available) 77.78% 78% 1 1 100% 100% 1

Colombia

AbstentionsAgainst total Objectives Performance Counterpart Political op. Conditions Proc. Problems Cost Overruns

Projects 3 0 31 P. Abstention 50% 50% 0 0 0 0 0

w/ PPMR info 2 0 22 P total 90% 76.19% 0 0 18.18% 4.55% 0

(info available) 95.45% 95% 1 1 100% 100% 1

Costa Rica

Objectives Performance Counterpart Political op. Conditions Proc. Problems Cost Overruns

total P total 1% 25% 0 0% 0% 17% 0%

Projects 12 (info available) 67% 67% 100% 100% 100% 100% 100%

w/ PPMR info 6

Republica Dominicana

AbstentionsAgainst total Objectives Performance Counterpart Political op. Conditions Proc. Problems Cost Overruns

Projects 2 0 20 P. Abstention 0 0 0 0 1 0 0

w/ PPMR info 1 0 16 P total 93% 64% 0 0% 19% 13% 0

(info available) 88% 88% 100% 100% 100% 100% 100%

Ecuador

Objectives Performance Counterpart Political op. Conditions Proc. Problems Cost Overruns

total P total 92% 64% 8% 8% 69% 46% 0%

Projects 22 (info available) 92% 85% 100% 100% 100% 100% 100%

w/ PPMR info 13

El Salvador

Objectives Performance Counterpart Political op. Conditions Proc. Problems Cost Overruns

total P total 0 83% 0 0 33% 0 0

Projects 8 (info available) 100% 100% 100% 100% 100% 100% 100%

w/ PPMR info 6

Guatemala

AbstentionsAgainst total Objectives Performance Counterpart Political op. Conditions Proc. Problems Cost Overruns

Projects 0 1 20 P. against 1 1 0 0 0 0 0

w/ PPMR info 0 1 13 P total 1% 70% 0 0 8% 8% 0

(info available) 92% 77% 100% 100% 100% 100% 100%

Guyana

AbstentionsAgainst total Objectives Performance Counterpart Political op. Conditions Proc. Problems Cost Overruns

Projects 0 1 12 P. against 1 1 0 0 0 1 0

w/ PPMR info 0 1 11 P total 0 67% 0 0 36% 73% 9%

(info available) 82% 82% 100% 100% 100% 100% 100%

Haiti

Objectives Performance Counterpart Political op. Conditions Proc. Problems Cost Overruns

(21)

Honduras

AbstentionsAgainst total Objectives Performance Counterpart Political op. Conditions Proc. Problems Cost Overruns

Projects 0 1 27 P. against 1 1 0 0 1 0 0

w/ PPMR info 0 1 20 P total 89% 71% 5% 0 65% 60% 0%

Mexico

total Objectives Performance Counterpart Political op. Conditions Proc. Problems Cost Overruns

Projects 19 P total 1 1 0 0 45% 0 0

w/ PPMR info 11 (info available) 100% 100% 100% 100% 100% 100% 100%

Nicaragua

total Objectives Performance Counterpart Political op. Conditions Proc. Problems Cost Overruns

Projects 25 P total 1 54% 6% 6% 78% 17% 6%

w/ PPMR info 18 (info available) 72% 72% 100% 100% 100% 100% 100%

Peru

AbstentionsAgainst total Objectives Performance Counterpart Political op. Conditions Proc. Problems Cost Overruns

Projects 2 0 35 P total 1 71% 10% 10% 30% 40% 0

w/ PPMR info 0 0 20 (info available) 90% 85% 100% 100% 100% 100% 100%

Panama

total Objectives Performance Counterpart Political op. Conditions Proc. Problems Cost Overruns

Projects 29 P total 1 1 0 0 24% 0 0

w/ PPMR info 17 (info available) 53% 59% 100% 100% 100% 100% 100%

Paraguay

total Objectives Performance Counterpart Political op. Conditions Proc. Problems Cost Overruns

Projects 21 P total 89% 67% 0 0 15% 0 0

w/ PPMR info 13 (info available) 69% 69% 100% 100% 100% 100% 100%

Uruguay

total Objectives Performance Counterpart Political op. Conditions Proc. Problems Cost Overruns

Projects 24 P total 0 71% 0 0 10% 10% 0

w/ PPMR info 10 (info available) 0.7 70% 100% 100% 100% 100% 100%

Venezuela

AbstentionsAgainst total Objectives Performance Counterpart Political op. Conditions Proc. Problems Cost Overruns

Projects 0 7 8 P. against 1 80% 0 0 85.71% 28.57% 0

w/ PPMR info 0 7 8 P total 1 83% 0 0 75% 25% 0

(22)

Table 2: Probit

Commitment Problems

(1) (2) (3)

Technical cooperation 5.42e-06 5.07e-06

(BID_country/sector/year) (1.89e-06)** (2.08e-06)**

Sqr Technical cooperation 2 -5.77e-12 -5.10e-12

(BID_country/sector/year) (2.31e-12)** (2.49e-12)**

Project Execution Facility -1.5720 3.6765 (14.902). (14.598). Sqr project execution facility -2.6121 -28.538 (58.969). (61.4574). Difference amounts planned‐disbursed 1.53e-08 (1.38e-08). Concessional debt (% of total external debt) 0.0766 0.0927 0.0864 (0.0334)** (0.0373)** (0.0350)** Net ODA received (% of GNI) 0.0942 0.0887 0.04954 (0.0561)* (0.0585). (0.0540). Abstentions 0 .13097 0.15794 (0.4824) (0.4827). Number of observations 239 232 232 Pseudo R2 0.2994 0.3575 0.332 ** 5% significance, * 10% significance Controls (1): country, sector.

(23)

Table 3: Probit

Contract conditions compliance delays

(1) (2) (3)

Technical cooperation 2.84E-06 3.81E-06

(BID_country/sector/year) (2.12e-06). (0.00000189)**

Sqr Technical cooperation 2 -2.79E-12 -3.69E-12

(BID_country/sector/year) (2.67e-12). (0.00000000000244).

Project Execution Facility (amount) 7.44E-06 6.14e-06

(3.71e-06)** (3.30e-06)*

Sqr project execution facility (amount) -1.19E-11 -1.11e-11

(6.20e-12)* (5.68e-12)*

Difference amounts planned-disbursed -0.78288

(percentage) (0.54749).

Difference amounts planned-disbursed -3.33E-02 -6.54E-08

(0.0157)** (0.0000000273)**

Concessional debt (% of total external debt) 8.86E-02 0.07424

(0.03188)** (0.0253)**

Net ODA received (% of GNI) 2.98E-02 1.17E-01

(0.0695). (0.0597)**

Total debt service (% of GNI) 8.38E-02 6.64E-02

(0.04898)* (0.0392)*

Original approved amount -1.92E-10

(0.000000000817).

Abstentions -8.80E-01

(0.9669).

Average grant element on new external debt commitments (%) -.026211 (.0089)** Number of observations 253 278 272

Pseudo R2 0.3734 0.3607 0.3211

** 5% significance, * 10% significance

(24)

Table 4: Probit

Borrower Maintaining Performance data

(1) (2) (3) (4)

All countries w/out Venezuela All countries w/out Venezuela Abstentions ‐1.5467 ‐1.5065 ‐1.4365 ‐1.4361 (0.9214)* (0.9231)* (0.8735)* (0.8722)* Against 0.2667 1.4678 (1.0117). (1.6985). Community political opposition ‐0.7295 ‐0.7321 ‐0.3581 ‐0.3516 (0.6073). (0.6165). (0.6253). (0.6274). Counterpart Funding Shortfall ‐0.0078 ‐0.0917 ‐0.3455 ‐0.3391 (0.8897). (0.9084). (0.7698). (0.7696). Inter‐agency coordination 0.552 0.4842 0.7551 0.7053 (0.5765). (0.5917). (0.5668). (0.5677). Original approved amount ‐0.0002 ‐0.0009 ‐0.0003 ‐0.0005 (0.0001). (0.0001). (0.00007). (0.00007). Contract conditions compliance delay 0.7419 0.813 (0.3307)** (0.3392)** Procurement difficulties  0.6007 0.7721 0.2904 0.3331 (0.3861). (0.4023)* (0.3549). (0.3623). Lack of M&E system ‐1.0508 ‐1.1591 ‐0.7108 ‐0.7068 (0.9538). (0.9737). (0.7576). (0.7628). Technical cooperation  0.3216 0.084 0.1677 0.0791 (0.4334). (0.4476). (0.4162). (0.4196). PBL 0.9957 0.9756 (0.4540)** (0.4526)** Energy  1.7063 1.4151 1.2848 1.0327 (1.0515)* (1.1234). (0.8164)* (0.8989). Integration and trade div. ‐1.8902 ‐1.9852 ‐0.9576 ‐0.9356 (1.0792)* (1.0666)* (0.9263). (0.9231). Number of observations 205 197 192 185 Pseudo R2 0.3103 0.3233 0.2173 0.2122 ** 5% significance, *10% significance Controls (1) & (2): country, division/sector, operation sub‐type Controls (3) & (4): country, division/sector, operation sub‐type, Difference amount and dates (current vs. approved).

(25)

Table 5: Probit

Inter-agency coordination

(1) (2) (w/out Venezuela) Technical cooperation 0.0000215 0.000023 (BID_country/sector/year) (9.46e-06)** (0.0000113)**

Sqr Technical cooperation 2 -5.75e-11 -6.40e-11

(BID_country/sector/year) (3.76e-11). (4.78e-11).

Project Execution Facility -15.09448 -61.9574

(percentage project) (44.7923). (58.647).

Sqr project execution facility (amount) -162.933 276.3863 (620.1232). (686.9008). Multilateral debt (% of total external debt) 0.21207 0.213045

(0.1219)* (0.13581). Difference amounts planned-disbursed -0.04587

(0.9058) Concessional debt (% of total external deb 0.0699 0.21185 (0.15918). (0.1865). Net ODA received (% of GNI) 0.17977 0.24740 (0.1312). (0.1658). Interest payments on external debt (% of G-1.301964 -1.5060

(0.7707)* (0.84287)* Abstentions 3.230394 (1.744264)* Number of observations 151 151 Pseudo R2 0.4827 0.5077 ** 5% significance, * 10% significance

(26)

Table 6: Bivariate Probit

Borrower maintaining Contract conditions performance data compliance delays

Sectoral TC 0.0000010 0.0000021 (0.000000486)** (0.00000143). Sqr. sectoral TC 0.0000000 (0.00000000000158). Project TC 7.1176800 59.5170400 (% loan) (10.0212). (32.059). Sqr. Project TC -734.9900000 (0.00000000000158). Abstentions -1.7263200 -0.5635990 (0.7404)** (0.50093). Obs: 247

Controls: division, country Rho: 0.30320**

(27)

References

Related documents

Fall Work Term Winter Work Term MGMT 4000 Strategic Management ACCT 4230 Advanced Management Accounting MGMT 3320 Financial Management ACCT 4220 Advanced Financial

accountability, oversight, and management throughout every real estate service by a single point-of-contact dedicated transaction manager who harnesses local market knowledge

perceptions of clinical teaching effectiveness between full-time and part-time baccalaureate nursing clinical faculty in the areas of teaching ability, nursing

‘Translation’ as we use it here operates with these broader post-structural assumptions of education policy sociological research in mind, but is focused on the shifting relations of

All the tests detect strong evidence of nonlinear structure in the daily spot price of crude oil; whereas in monthly observations the evidence of nonlinear dependence is less

Algonquian leaders like the Mohegan grand sachem Uncas, the Narragansett sachem Miantonomi, and the Narragansett-Niantic sachem Ninigret vied to be the premier Native leader

The observed correlation of positive results of atopy patch tests with food and inhalant allergens and history of some respiratory diseases and symptoms also on the population