reynaud 02

8/13/2019 Reynaud 02

1/24

Geopolitics and international organizations: An empirical study on IMF facilities

Julien Reynaud a , , Julien Vauday ba European Central Bank, Kaiserstrasse 29, 60311 Frankfurt am Main, Germanyb Paris School of Economics, University of Paris 1 and Ecole Polytechnique, CES-106-112 boulevard de l'Hpital, 75647 Paris cedex 13, France

a b s t r a c ta r t i c l e i n f o

Article history:Received 18 April 2007Received in revised form 16 July 2008Accepted 23 July 2008

JEL classi cation:F33H77O19

Keywords:Factor analysisGeopoliticsInternational Monetary FundPotential analysis

There is growing awareness that the distribution of IMF facilities may not be in uenced only by the economicneeds of borrowers. This paper focuses on the fact that the IMF may favour geopolitically important countries

in the distribution of IMF loans, differentiating between concessional and non-concessional facilities. To carryout the empirical analysis, we construct a new database that compiles a wide array of proxies for geopoliticalimportance for 107 IMF countries over 1990 2003, focusing on emerging and developing economies. We usea factor analysis to capture the common underlying characteristic of countries' geopolitical importance aswell as a potential analysis since we also want to account for the geographical situation of the loan recipients.While controlling for economic and political determinants, our results show that geopolitical factorsin uence notably lending decisions when loans are non-concessional, whereas results are less robust and inopposite direction for concessional loans. This study provides empirical support to the view that geopoliticalconsiderations are an important factor in shaping IMF lending decisions, potentially affecting the institution'seffectiveness and credibility.

2008 Elsevier B.V. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. Geopolitics and international organizations: what about the IMF?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1413. Geopolitical determinants of the importance of nations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142

3.1. Methodological issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13.2. Variables entering the geopolitical factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

3.2.1. Energetic area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.2. Nuclear area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.3. Military power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13.2.4. Geographic area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

3.3. Description of variables entering the geopolitical factor and outcome of the factor analysis . . . . . . . . . . . . . . . . . . . . . 1444. Data and methodological issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

4.1. The data: description of the independent and dependent variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1444.1.1. Independent variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144.1.2. Dependent variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

4.2. Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

5. Estimation results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1. Core results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2. Robustness checks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

5.2.1. On the importance of political factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152

Journal of Development Economics 89 (2009) 139 162

The authors would liketo thank AFSE, CEPII, CEPN, CERDI, theCNRS 22ndSymposium on Money, Banking and Finance,ECB, EEA,ETH-KOF, EUREQua/TEAM, IRES, RIEF, SMYEandSUITE conferences and seminars participants. We are grateful to Javier Diaz Cassou, Arnaud Mehl, Bertrand Couillault, Nicolas Berman, Rodolphe Desbordes and in particular to FaridToubal forconstructivecomments.We would like tothank Axel Dreher forhis comments andforproviding us with hisdatasets. Wewould also like tothank theSputnikfor providingus the ideal environment for our thinking. Finally, two anonymous referees have permitted to signi cantly improve the quality of this article. We are therefore also grateful to them.All remaining errors are obviously ours.

Corresponding author. Tel.: +496913445363.E-mail addresses: [email protected] (J. Reynaud), [email protected] (J. Vauday).

0304-3878/$ see front matter 2008 Elsevier B.V. All rights reserved.

doi: 10.1016/j.jdeveco.2008.07.005

Contents lists available at ScienceDirect

Journal of Development Economics

j o u rn al h o m ep ag e : ww w.e l sev i e r. co m / lo cat e / eco n b ase
mailto:[email protected]:[email protected]://dx.doi.org/10.1016/j.jdeveco.2008.07.005http://www.sciencedirect.com/science/journal/03043878http://www.sciencedirect.com/science/journal/03043878http://dx.doi.org/10.1016/j.jdeveco.2008.07.005mailto:[email protected]:[email protected]

8/13/2019 Reynaud 02

2/24

5.2.2. On the importance of factors explaining aid ows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1525.2.3. On the importance of recidivism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1525.2.4. On the inter- and intra-individual groups correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1605.2.5. On the sample size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1605.2.6. On entering the factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1605.2.7. On a different estimation of the factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161

6. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1. Introduction

Some large IMF-supported programs raise concerns because theyappearto suggest that a country's geopolitical importance [ ] playsa role in IMF loan decisions , de Rato y Figaredo, IMF ManagingDirector between June 2004 and October 2007 (IMF, 2004) .

It is important to recognize that when geopolitical considerationsweighheavily,the IMFtendsto be diverted from theprinciples thatnormally govern its provision of nancial support , Mussa, IMFEconomic Counselor and Director of the Department of Researchfrom 1991 to 2001 (IMF, 2002) .

Several Institutions were created after World War II in order toprovide internationalpublicgoodsanddealwith politicalandeconomicissues on a multilateralbasis.More recently, theprocess of globalizationhas further underlined the usefulness of some of these organizations.Indeed, it is increasingly clear that the maintenance of international

nancial stabilityandglobalpolicy issues call forenhancedinternationalcooperation.

The transfer of sovereignty from the country level to the inter-national level has created tensions, however. Jackson (2003) arguesthat in some of these circumstances ( ) a powerful tension is ge-nerated between traditional core sovereignty , on the one hand, andthe international institution, on the other hand . This may be partlydue to thefact that the multilateral approach hasnot always respectedthe principle of equal treatment ( Mavroidis, 2000 ). Indeed, it is widelyaccepted that decision-making in international organizations tendsto be dominated by a few large countries (see Stiglitz, 2002 ; andVreeland, 2007 ; for a recent literature review on the InternationalMonetary Fund (IMF) and Bown, 2005; Shoyer, 2003 ; and Steinberg,2004 ; regarding the World Trade Organization (WTO)). First, thepowers, i.e. quotas or voting shares, are not always equitably appor-tioned relative to country size. Second, some countries have means toin uence others, and can then divert international organizations fromtheir initial engagements. Steinberg (2004) , for example, emphasizesthe ongoing debate around the good functioning of the WTO disputesettlement body. He distinguishes studies that argue that the newsystem favours powerful members and encourages them to adopt arule breaking behaviour , from those arguing that the new system

prevents these countries from behaving in such a way. Whatever thepoint of view, both assume that powerful members tend to divert theinstitution from its governing principles, at the expense of othermembers, by using their relative economic size.

The IMF has recently been subject to particularly ercecriticisms. Many have argued that the institution is failing to ful lits main objectives, namely the provision of emergency nance forthe resolution of balance of payment crises and the surveillance of the world economy. Many of the problems the IMF is facing arerooted in its governance structure since the Fund is also dominatedby a rather narrow group of advanced economies. According toTruman (2006) , the IMF is enduring an identity crisis mainlycaused by the imbalance of power among its members. As a resultthere are indications that a number of its members have lost faith in

the institution.

These governance issues raise questions regarding the fairdistribution of IMF loans. A large number of academic studies haveexamined the determinants of the IMF's lending decisions. In the rsthalf of the 1990s, researchers have focused on the economicdeterminants of IMF loans ( Joyce, 1992; Conway, 1994; Bird, 1995 ;and Knight and Santaella,1997 ), while in the second half of the 1990s,others have focused on other determinants, such as political ones(Edwards and Santaella, 1993; Thacker, 1999; Vreeland, 1999; Bird andRowlands, 2000; Przeworski and Vreeland, 2000; Vreeland, 2001;Dreher and Vaubel, 2004; Dreher, 2004; Joyce, 2004; Barro and Lee,2005; Sturm et al., 2005; Harrigan et al., 2006 ; and Steinwand andStone for a survey, 2007 ).

The aim of this paper is to explore the hypothesis that some coun-

tries have a geopolitical interest in diverting [the IMF] from theprinciples that normally govern its provision of nancial support (Mussa, ibid, and de Rato y Figaredo, ibid). To that end, the paperstudies the geopolitical importance of loans recipients. After de ningthe concept and relevance of geopolitics in the context of an inter-national organization with a particular focus on the IMF, we collectedand built various indicators that, according to related literature, aresubjects of geopolitical stakes. As Baldwin (1979) argues, there is nounique geopolitical variable. Indeed, geopolitics may concern manydifferent areas, thus inducing that, regarding on the area, the samecountry's geopolitical importance may switch from the highest to aninsigni cant level. 1 Consequently, the geopolitical importance of acountry is an unobservable variable. Nevertheless, it is possible tostatistically extract the underlying factor of commonly known deter-

minants of geopolitical importance of countries and to capture itsdistribution over the globe. We therefore proceed as follows: in a rststep, we identify geopolitical determinants that may in uence thedistribution of IMF loans and extract the underlying factor of a largerrange of these factors. In a second step, inspired by recent ndingsfrom economic geography (see Hanson, 2005 ), we compute a geo-political potential la Harris by taking the country's geopolitical factorand the sum of others countries geopolitical factor weighted by theirgeographical distance. We also compute what we call a pure geo-political potential of country's geopolitical importance by taking onlyinto account the geopolitical geographic situation of the country (i.e.without taking into account its own geopolitical factor but only that of its neighbours over their distance). Using this technique allows us tohave a full geographical coverage when judging of a country's geo-political importance. In a third and last step, in line with existingliterature, we estimate a standard model of determination of IMFloans distinguishing between concessional facilities, i.e. the PovertyReduction and Growth Facility (PRGF), and non-concessional facilitiessupported by the General Resources Account (GRA). Regarding thelatter, we focus on Stand-By Agreements (SBAs) and Extended FundFacility (EFF) which share thelargest part of overall IMF nancing. Thisdistinction is crucial since these facilities are different in terms of

nancial conditions and overall objectives. Yet, they are sometime

1 Baldwin argues that Planes loaded with nuclear weapons may strengthen a state'sability to deter nuclear attacks but may be irrelevant to rescuing the Pueblo on short

notice. (p. 164).

140 J. Reynaud, J. Vauday / Journal of Development Economics 89 (2009) 139162

8/13/2019 Reynaud 02

3/24

pooled together in related studies or researchers focus only on GRAagreements. Since we focus on lending, and given that no industrialcountry has made use of the Fund's nancial support for the last threedecades, our panel comprises 107 IMF developing and emerging eco-nomies over the period 1990 2003 sampled at the yearly frequency.

Our results provide empirical support to the view that geopoliticalconsiderations are an important factor in shaping IMF lending de-cisions. Economic and political determinants are still valid for both

facilities and turn out to be more in uential for SBA. Moreover, weshow that the Fund favoured geopolitically important countries whenlending non-concessional facilities. However, concessional loans tendto be attributed to non-geopolitically importantcountries, although toa lesser extent. Indeed, according to literature on aid determinants, itis because bilateral aid ows are mainly determined by political andgeopolitical determinants that international institutions have settledmultilateral aid arrangements to support non strategic countries thatwould not received bilateral aid otherwise ( Burnside and Dollar,2000 ). Overall, our results are robust to a large number of robustnesschecks including controlling for recidivism, inter- and intra-individualgroups correlation and other econometrical speci cation of the factoranalysis.

The remainder of the paper is organised as follows. Section 2 isdevoted to the understanding of geopolitics, and its role within theIMF. Section 3 explains the choice of variables and the techniques.Section 4 describes the data and discusses methodological issues.Section 5 exposes the empirical results and the robustness checks.Finally, Section 6 concludes.

2. Geopolitics and international organizations: what aboutthe IMF?

There is a vast literature on the economic and political determi-nants of IMF lending decisions (see also Section 4 below). However,the question of whether some countries may have a geopoliticalinterest in shaping the Fund's decisions has, to our best knowledge,received much less attention. In this paper, we put forward thehypothesis that leading members of international organizations usethe institution's prerogatives to increase or serve their in uence overother members for geopolitical purposes. Boughton (2004) supportedthe view that IMF involvement in the Eastern European countries wasnot purely nancially driven, but rather ideological by ultimately en-couraging the superiority of the market economy. In the same vein,Marchesi and Sabani (2007) show that because of the lack of credibility of the Fund, i.e. regarding the borrowing country's non-compliance with conditionality, lending may be distorted for reputa-tion issues. As a result, creditors (i.e. the G7 members) may use theFunds' nancing facilities to increase or serve their in uence overdebtors.

Diverting the IMF, for geopolitical purposes, from its principles toserveparticular interest is possible since decisionsto lend aretakenbythe Executive Board (the Board). The Board is responsible for con-

ducting the day-to-day business of the IMF. It is composed of 24Directors, who are appointed or elected by member countries or bygroups of countries, and the Managing Director, who serves as itsChairman. The Board usually meets several times a week and carriesout its work largely on the basis of papers prepared by IMF staff.Decisions are of cially voted, but in practice, Directors never vote. TheChairman evaluates the positions of Directors following their inter-ventionsand passesa decision when a consensusseemsto be reached.Therefore, it is straightforward that, if some countries are betternegotiators or have means to in uence others, they can succeed inin uencing the Board's decisions. Alonso-Meijide and Bowles (2005) ,Bini Smaghi (2004, 2006a,b) , Leech (2002) and Reynaud et al. (2007)have illustrated this using voting power indices derived from co-operative game theory and found that the US and the G7 are over

in uential at the Board.

Therefore, in studying the determinants of IMF loans, researchershave focused on particular factors that might be of interests forleading IMF members. For example, Thacker (1999) found that a movetowards the US position of the borrowing country is positively relatedto the probability of receiving a loan. Oatley and Yackee (2004) foundthat the more US banks are exposed in the borrowing country, thelarger the loan. Finally, Oatley (2002) found that commercial bankdebt of G7 countries into the borrowing country in uences the size of

the loan.Others have focused on country speci cities such as political sta-bility ( Edwards and Santaella,1993 ), political freedom ( Rowlands,1995 )and democracy indicators ( Thacker, 1999; Vreeland, 1999; Dreher andVaubel, 2004 ). They found that the more borrowing countries are closeto cultural andpoliticalstandardsin developedcountries, thehighertheprobability to receive IMF funds.

More recently, IMF staff has argued that some members arein uencing the distribution of loans because of particular geopoliticalinterest in the borrowing country. We begin by introducing hereafter arather heuristic de nition of geopolitics:

Geopolitics traditionally indicates the links and causal relation-ships between political power and geographic space; in concretetermsit is often seen as a body of thought assaying speci c strategicprescriptions based on the relative importance of land power (Osterud, 1988 ).

Geopolitics has then to be related to the importance of land power:thesize, theposition inthe World, theresourcesthatarenaturalandbuiltby man. The conversion of land power into political power is howevernot straightforward. In the context of International Politics, Baldwin(1979) and Nye (1990) developed the following seminal argument:

Some countries are better in converting their resources into ef-fective in uence, just as some skilled card players win despiteweak hands (Nye, 1990 ).

This idea, already mentioned by Baldwin (1979) as one of the two

reasons2

explaining theparadox of unrealized power , refers tothe factthat a countrywith resourcesidenti ed as strategicdoes notnecessarilysucceed in being powerful. According to Baldwin, this country has theresources but has not the knowledge to use it in order to convert theminto power. Similarly, some countries have no resourcesbut have meansto convert strategic resources into power. Then, the latterare interestedin using resourcesof theformer. A good example is the importance of oilreserves. Indeed, these reserves do not provide wealth at the momentbut may in the future. Moreover, they may provide wealth and thuspolitical power at the domestic level if they are exploited domestically;butcouldalso beappropriated externallyand lead to a misdistributionof wealth, i.e. corruption (see the literature on the resource curse: amongothers Leiteand Weidman,1999 ; Sala-i-Martin and Subramanian,2003;Isham et al., 2004; Mehlum et al., 2006 ;and Dietz et al., 2007 ). However,

these reserves provide geopolitical power as of today since most (of industrial) economies are dependent and do not possess large initialendowments. 3 Finally, there is a last group of countries, mainly thosethat have both the know-how and the resources (orthe control of othercountries' resources). They represent the dominant countries and try tomaintain this domination by protecting other countries' resources. Thefact that they dominate has allowed them to obtain a great importancein the (recently created) International Organizations. Indeed, as arguedby Popke(1994) , theroleof the IMF has increasinglycome to be scripted

2 The other one is the already mentioned bad estimation of what creates power.3 We do not focus on the measurement of country's ability to transform strategic

resources into effective power. Also, we believe that this could be of some interest tostudy it in correspondence with the ability of this country to be listening in

international fora.


8/13/2019 Reynaud 02

4/24

throughthe discourseof USsecurity . Moreover, theIMF itselfdraws ondiscourses, in order to script the role of the countries with which itinteracts ( ). The IMF disseminates a form of power/knowledge bycasting itself as the sole authority over a wide range of issues . Popke

nally argues that this power in uence also IMF's surveillance andstructural adjustment programs. 4 The aim is therefore to de ect blamefor monetary problems away from the industrialized nations and ontothe nations of the third world .

This leads to the idea that IMF loans, the country chosen, theamount lent and the level of conditionality of loans could be used bycreditors to control or to appropriate strategic resources from debtors.The distinction between the use and the possession of resources isthen representative of what makes the difference between politicsand geopolitics, and justi es the hypothesis just above. To put it in anutshell, the question of geopolitics is not why should a country havean interest in another one, but rather what could be of interest in theborrowing countries for the dominant members of the IMF. Therefore,the objective of this paper is rst, to identify what factors could makesome countries geopolitically attractive to IMF creditors and second toassess empirically whether these factors in uence the probability toreceive IMF nancial support.

3. Geopolitical determinants of the importance of nations

3.1. Methodological issues

In this section we attempt to identify relevant proxies for some of the key factors that determine the geopolitical importance of nations.Listing all the sources of geopolitical importance is a dif cult task. Thesearch for determinants of country's geopolitical importance faces inour view two main constraints. First, one should not search for adeterminant, neither for some determinants, but rather for a range of interacting determinants. Indeed, as Baldwin (1979) argues, there is nounique geopolitical variable. Geopolitics may therefore concern manydifferent areas. Keeping this in mind, we attempt to propose a sta-tistical analysis of the geopolitical determinants which deals with thisissue, namely a common factor analysis . Factor analysis is used to studythe patterns of relationship among many variables, with the goal of discovering something about the nature of the underlying commonfactor that affects them, even though those variables were not mea-sured directly. In our case, measuring directly the geopolitical impor-tance of a country is not possible. In a factor analysis, this will referto the inferred independent variable, i.e. the factor. In other words,factor analysis looks for the factors which underlie the variables. It istherefore useful for our study since we do not pretend to propose anabsolute de nition or an index of the geopolitical importance of countries, but rather extract an underlying factor behind a wider rangeof determinants as possible. 5 More formally, with xi an observation,the factor analysis states that, with i=1,2 , p:

xi k

r 1lir f r ei 1

where f r is the common r th vector, k isspeci edand e i is a residual thatrepresents sources of variation affecting only x i. In other words, if acorrelation matrix can be explained by a general factor, it will be truethat there is some setof correlations of the observed variablessuchthatthe product of any two of those correlations equals the correlationbetween the two observed variables. The method we used to estimate

thegeopoliticalfactor (gf)is the regressionestimator (Thomson,1951 ).Formally, it has the following form ( Kosfeld and Lauridsen, 2008 ):

gf T 0 0 u 1 X 0 I 01=2u

1 0 X 0 2

where is the factor matrix, is given by =F , with the left handside being the matrix of the true regressor values. The matrix of observations X , is then given by the following equality: X = +U, where

U standsforthe errorsmatrix. Thatis, ifwe refer toEq. (1),it isthe matrixof the ei. Finally,thelastterm tode ne is thecovariance matrixof uniquefactors u j, given by: u =diag( u 1

2 u 22 u p

2 ). The product is the cross-factors matrix of the with each other.

Regarding the structure of the factor, two questions arise: Howmany factors should we use? How many variables should we use?Darlington et al. (1973) expose a simple rule: The fewer factors, thesimpler the hypotheses. Since simple hypotheses generally have lo-gical scienti c priority over more complex ones, hypotheses involvingfewer factors are considered to be preferable to those involving morefactors. That is,one accepts at least tentatively the simplesthypothesis(i.e. involving the fewest factors) that is not clearly contradicted by theset of observed correlations. So that the clearer the true factor struc-ture, the smaller the sample size needed to discover it. Thus, the rulesabout the number of variables are different for factor analysis than forregression, i.e. it is perfectly acknowledged to have many more var-iables than cases. In fact, the more variables the better as long asthe variables remain relevant to the underlying factor. Regardingthe number of factors to be selected, we will display model-selectioncriteria, the Akaike (AIC) and the Bayesian (BIC) information criteria. 6

We will also run maximum-likelihood tests. Each model will be esti-mated using maximum likelihood, and thus will permit to select thebest Log likelihood ratio. We will also display the Kaiser Meyer Olkinmeasure of sampling adequacy that permits to discriminate whetheroverall variableshave enough in common to warranta factor analysis. 7

The second constraint in dealing with the geopolitical importanceof a country is related to the fact that one should not only take intoaccount the geopolitical importance of this country, but rather itsimportance and the importance of its neighbours, i.e. its geographicalposition. Indeed, while dealing with geopolitics, one should not omitthe importance of the region and the importance of geographic rela-tions between states. For example, one could not ignore the geo-political importance of Turkey given by its geographical situationbetween Europe and the Middle-East. Keeping this in mind, we at-tempt to deal with this inconvenience by proposing an additionalstatistical analysis of the geopolitical determinants, namely a potentialanalysis . We bring together the concept of geopolitical importance of states and the potential analysis taken from International Economics.Generally, in the location decision analysis (of FDI for example), avariable labelled market potential is presented. This idea is relatedto Harris' (1954) in uential market-potential function, which statesthat the demand for goods produced in a location is the sum of pur-chasing power in other locations, weighted by transport costs. Theconcept was later strengthened by Fujita, Krugman, and Venables(1999) stating that nominal wages are higher near concentrations of consumer and industrial demand ( Hanson, 2005 ). In this paper, weadapt this concept adding to country's factor the scores of its neigh-bours to their distance. By doing so, we are able to catch both thegeopolitical importance of a particular country and also its geopolitical

4 See Fratzscher and Reynaud (in press) for a study on the in uence of politicalpower on IMF surveillance.

5 Proposing an index is inappropriate since it induces to arbitrary weight the

variables entering it.

6 AIC and BIC information criteria are generally used to compare alternativemodels. These criteriapenalize models with additional parameters.The AICis de nedas (-2 log-likelihood+2 number of parameters) and the BIC as (-2 log-likelihood+number of parameters number of observations). Comparing models permit toorder selection criteria based on parsimony.

7 The KMO measure of t is an index for comparing the magnitudes of the observed

correlation coef cients to the magnitudes of the partial correlation coef cients.


8/13/2019 Reynaud 02

5/24

importance given its geographical situation. Formally, the geopoliticalpotential of a country is computed as follows:

gp it n

i1

gf it d ji

3

where gp it is the geopolitical potential of country i, gf it is the geo-political factor of country i as calculated in Eq. (2) and d ji the relativedistance in kilometres between country j and i. However, due to (i) thelarge number of countriesin our database and(ii) theweak magnitudeof the factors compared to that of the bilateral distance, Eq. (3) isexpected to be correlated to Eq. (2). Therefore, we compute Eq. (3)without taking into account the geopolitical factor of the borrowingcountry but only the weighted sum of its neighbours and call it thepure geopolitical potential (gpf it ):

gpf it ji

gf jt d ji

4

Performing pairwise correlation tests between gf it and gp it andbetween gp it and gpf it con rm correlation levels of 0.46 and 0.99respectively, signi cant at the 1% level (cf. Table 1 ).

3.2. Variables entering the geopolitical factor

Variablesproxying the geopolitical importance of countriesmay beclassi ed in 4 areas as follows: (i) the energetic, (ii) the nuclear, (iii)the military and (iv) the geographical areas.

3.2.1. Energetic areaCapturing the relative importance of land power refers directly to

energetic resources. Of course, many resources might be useful inbuilding a geopolitical factor, but we are here interested in resourcesthatare/might be strategic since we aresearching for potential power. 8

In this case, oil and gas resources appear to be fundamental. For ex-ample, Rose (2007) uses oil and gas proven reserves as proxies of geopolitical importance of country in a gravity equation to study bi-lateral trade. Moreover, morethan 90% of world's energetic rentcomesfromoiland gas ( Eifertet al., 2003 ). Inthatspirit, weusethe data onoiland gas proven reserves, rather than actual oil and gas production, tocapturecountries' potential rent as we argued abovethat whatmattersis rather the (unexploited) potential. One needs also to take intoaccount, for strategic purposes, the country's ability to transport theseresources. Indeed, it is sometimes the case that a country is geo-politically important not because it owns large resources but because

they need to transit via this country to be exported. Therefore, we usealso oil and gas pipelines since they are expected to proxy countries'ability to transport energy for internal or external purposes.

We expect the endowment in reserves and pipelines to in uencepositively in the probability of obtainingIMF money. Indeed, regardingoil reserves, we rely on related literature, in particular Harrigan et al.(2006) , advocating that countries with larger oil and gas reservesreceive substantiallymore nancialsupport since IMFcreditors maybeinterested in exploiting these resources. Finally, we expect the pos-session of large pipelines infrastructures to increase the probabilityof obtaining an IMF loan since they facilitate the transportation of

national or foreign resources, and therefore should be subject toprotection or to appropriation.

3.2.2. Nuclear areaAfterhaving proxiedcountries'energetic importance,we should also

take into account countries' endowment in nuclear energy. Indeed,Mussa (1999) provided quite a clear answer to whether one should takeinto account nuclear power of countries by writing that many thought

that Russiawas tooimportant

too nuclear

tobe allowed tofail. Whatmakes this resource special is that it is at the cross-section between

energetic and military powers. Therefore, we computed a variableaccountingfor the size of civil nuclear capacity and a dummy variable tocapture whether a country has the nuclear weapon.

Theimpact of these variableson theprobabilityto obtainan IMFloanisambiguous. On the onehand,the nonallocation of an IMF loan may beseen by dominant countries willing to retain their position as a tool tocounteract the rising power of nuclear powers. On the other hand, theinternational community may be interested in ensuring the economicstabilityof nuclear powers in order to reduce the risk that they use theirweapons. Additionally, the possession of nuclear weapons mayincreasecountries' bargaining power in the international arena, and thereforetheir ability to lobby toobtainan IMF loan. Jo and Gartzke (2007) studythe determinants of nuclear weapons proliferation and found thatsignatories to the Treaty on the Non-Proliferation of Nuclear Weaponsare less likely to initiate nuclear weapons programs, but that has notdeterred proliferation at thesystemlevel. Moreover, they found that theUnited States hegemony has the potential to encourage nuclearproliferation since the US appears much more willing to intervene,advocating in our case for a positive relation between the allocation of IMF loans and the nuclear capacity of countries.

3.2.3. Military power Within the notion of geopolitics lies the concept of military power.

Indeed, at its very start, the discipline gained attention largely throughthe work of Sir Halford Mackinder and his formulation of the HeartlandTheory in 1904 ( Mackinder, 1904 ). This theory hypothesized thepossibility for a huge empire to be brought into existence which didnot need to use coastal or transoceanic transport to supply its militaryindustrial complex, and that this empire could not be defeated by all therest of the world coalitioned against it. 9 To proxy the militaryimportance of countries, we use three variables: First, we needed toproxy the military potential of a country for domestic and regionalpurposes.A rstindicatorcouldbe thenumberof local soldiers oreven aproxy of the military budget. However, one of our concern is to includevariablesthat do notin uencetemporaneous the economy. In thisspirit,we collected the number of US soldiers established in the borrowingcountry. We focus only on the US army because of its global militaryimportance and because the US dominates the Fund's decision makingprocess (as exposed in the previous section). Second, we needed tocontrol for con icts and the deployment of multilateral forces sincecon icts usually deter in ows of aids to the country. For instance,

Kuziemko and Werker (2006) found a signi cant and negative sign of avariable equal to 1 if war during which more than 1000 people died hasoccurred when explaining the amount of UN foreign aid. We collectedtherefore theUnited Nation Peacekeeping militarystrengthsestablishedin the borrowing country. Third and lastly, we built a weighted index of countries' involvement in Non-Proliferation Treaties (NPT) in order toprovide a measure of the international good willing . We constructedthis index by collecting data for all the international Treaties (13),excluding regional ones. If a country has implemented a Treaty,then it iscoded 1, 0 otherwise. Toappreciate theproximitybetween each countryand the International Community, we weight each Treaty, year by year,

Table 1Correlation analysis of geopolitical factor, potential and pure potential

Geopoliticalfactor

Geopoliticalpotential

Pure geopoliticalpotential

Geopolitical factor 1.00 0Geopolitical potential 0.4617 1.00 0Pure geopolitical potential 0.4095 0.9966 1.000

8 A general concern has also been to include variables proxying the geopoliticalimportance of countries that do not in uence temporaneous the economy, i.e. to

escape endogenoeity issues when turning to the econometric modelling.

9 Nye (1990) also argues that ability to win a war is the historic source of power.Military power is still a factor explaining power in spite of the rise of other factors such

that economic growth or technology.


8/13/2019 Reynaud 02

6/24

byitsrelativeimportance. Thelatter is givenby thenumber of depositors(implementationof the Treaty)dividedby the totalnumberof depositorsforall NPT. Therefore, the more a Treaty hasbeen implemented by othercountries, the more it contributes to the index. For example, the GenevaProtocol, created in 1925, has a weight of almost 16.5% in the index in1990. The Mine Ban Convention, signed in 1997 (so it has no weight forthe rst 7 years of the data) has a weightof 9.9%at the end oftheperiod.However, the Geneva protocol weight has lost 7 percentage points in

1997. Moreover, theNPTrelatedto nuclearweapon loses lessweightthanthe Geneva protocol does (from 19% to 13%). Finally, the weight of someTreaties like the Certain Conventional Weapons Convention present atthe beginning of the period has increased at the end of the period, thusimplying there is not a bias in favour of recently created Treaties.

We expect IMF loans to be positively correlated with these militaryfactors. Indeed, the US troops variable exhibits the geopoliticalimportance for the US, and thus for an important number of US allies(LeBillonand ElKhatib, 2004 ). We expectthe USand its military alliestoin uence loan decisions in order to favour countries where their troopsarepresent. Regardingthe NPTindex, theeffect of thevariable relating toNPT is more ambiguous. On the one hand, signing such treaties signalscountries' cooperative behaviour and submission to an internationalrule of law whichmay impactpositivelyon theoddsofobtaininganIMFloan. On theotherhand, their participation in such a treaty reduces theirthreat to the world. In this context the international community may beless interested in ensuring the economic stability of such countriesthrough the concession of an IMF program. Finally, we have noprede ned expectations regarding the UN strength proxy since thisvariables is more a control variables than a determinant.

3.2.4. Geographic areaFinally, we also need to take into account the pure geographic

characteristics of countries. In this part, we use traditional proxies of geographic importance of countries (see among others Ades and Chua,1997; Van Houtum, 2005; Bernholz, 2006 ): thearea in kilometre squaredto proxy the physical size of the country. To proxy whether thecountry isnot just led with desertsor mountains and if this country hasimportanttransportation capacities, we collected the length of the roads and thelengthof thecoast lines.Finally,and centralto thegeopoliticalanalysis,wealso usethe numberof bordersto appreciatethe centralityof countries.Allthese variables are supposed to capture size as well as geographicdeterminants of transportation ability within the country. They are thusallexpected to in uence positivelyon theprobabilityto receive IMFloans.

3.3. Description of variables entering the geopolitical factor and outcomeof the factor analysis

Units and the sources of collected data entering the factor analysisare reported in Table 2 . Table 3 reports the outcome of the factor

analysis and the Kaiser Meyer Olkin (KMO) measure of samplingadequacyto determine the t of our factor regarding variables enteringthe sample. We also report in Table 3 a column with correlation of thevariables with the factor. Not surprisingly, the only variable poorlycorrelated is the UN military strength variableas discussedbefore. TheKMO t is rather good and is classi ed as meritorious with a valueover 0.8, from a scale ranging from 0 to 1. Finally,we also report Akaike(AIC) and Bayesian (BIC) information criteria (see Table 4 ). They both,together with the Eigen values, advocate for the use of a single factor.Not reported here, we also ran maximum-likelihood tests on theadequatenumberof factors. Thelatter suggeststhat a onefactor modelprovides an adequate model and will therefore represent what we callthe geopolitical factor of countries.

4. Data and methodological issues

4.1. The data: description of the independent and dependent variables

4.1.1. Independent variablesVariables entering the geopolitical factor, the geopolitical poten-

tial and the pure geopolitical potential have been described in theprevious section. When estimating the probability of receiving IMFloans and the determinants of the size of these loans, one shouldcontrol for economic and political determinants that have been iden-ti ed as determinants of IMF lending in related studies. Sturm et al.(2005) used an Extreme Bound Analysis to discriminate betweeneconomic and political determinants of IMF loans using a panel modelfor 118 countries over the period 1971 2000. They found three robusteconomic variables explaining the distribution of IMF loans: The ratioof international reserves to imports of goods and services in current

Table 2Description of the variables entering the factor analysis

Variable Observations Mean Std. Dev. Min Max Unit SourceOil reserves 1568 6.14 21.13 0 132.46 Billion barrels Oil and gas journal and BPGas reserves 1568 39.28 183.78 0 1680,00 Trillion cubic feet Statistical ReviewOil pipelines 1568 2183.57 7199.54 0 72283 Kilometers CIA World factbookGas pipelines 1568 4844.36 17177.70, 0 156285 KilometersCivil nuclear capacity 1568 759.11 3087.74 0 21743 MWe Nuclear Energetic AgencyPossession of nuclear weapon(s)

1568 0.16 0.37 0 1 Dummy: 1 if possessesnuclear weapon. 0 if not

International Energetic Agency(author computation)

Number of USmilitary troops

1651 504.52 3590.62 0 41344 Number of soldiers US Department of Defense

UN military strength 1568 575.02 3390.62 0 38599 Number of soldiers United Nation PeacekeepingDepartment

NPT index 1568 0.61 0.24 0 0.9930502 Index (0 to 1) United Nation OrganizationLength of roads 1568 117,654.40 429,089.70 12 3,851,440 kmArea 1554 848,394 2,037,358 431 17,100,000 km 2 CIA World factbookNumber of borders 1568 4.14 2.63 0 17 UnitLength of coastlines 1568 2473.05 7199.35 0 54716 km

Table 3Factor selection criteria: Kaiser Meyer Olkin test

Variable Kaiser Meyer Olkin measureof sampling adequacy

Correlationwith factor

Oil reserves 0.7911 0.8964Gas reserves 0.8117 0.8626Oil pipelines 0.8670 0.8677Gas pipelines 0.8711 0.8641Civil nuclear capacity 0.7294 0.4356

Possession of nuclear weapon 0.8273 0.5076Number of US military troops 0.6443 0.1641UN military strength 0.3833 0.0481NPT index 0.6468 0.1371Length of coastlines 0.8384 0.3090Area 0.7550 0.5027Length of roads 0.8322 0.4402Number of borders 0.6016 0.2610Overall 0.8015


8/13/2019 Reynaud 02

7/24

US$, the growth of real GDP and the log of GDP per capita at marketprices. Theratioof total debt service to exports of goods andservices isalso found to be signi cant but to a lesser extent. We build thereforeour baseline model upon their ndings and include these variables inour estimations. 10 The expected sign of the reserves to imports ratioand the growth rates is negative since a low reserve to imports ratioincreases the risk of meeting balance of payments dif culties anda country experiencing high growth rates is less subject to econo-mic dif culties, respectively. Regarding the GDP per capita variable, ahigher ratio means a higher level of economic development andtherefore lessneedfor nancial support. However, it mayalso induceasize effect as due to its sheer economic size, a country could impactnegatively its neighbours if it meets some dif culties. Finally, the debtservice to exports ratio is expected to be positively linked since aheavy debt burden relative to exports increases countries' need forexternal nance to service that debt. All economic data are taken fromthe International Monetary Fund International Financial Statisticsdatabase.

To control for political factors in uencing IMF lending decisions isan important robustness check to analyse the potential robustness of our measure of geopolitical importance. Indeed, the recent empiricalstudies on political in uences on the IMF have shown that countriesvoting with the US in the UN General Assembly receive better treat-ment from the IMF ( Barro and Lee, 2005; Dreher and Sturm, 2006;Dreher et al., 2006, 2007 ) since Kuziemko and Werker (2006) haverecently demonstrated that the pattern of US aid payments to rotatingmembersof theUN council is consistent with vote buying. They arguedthat non-permanent members of the U.N. Security Council receiveextra foreign aid from the United States and the United Nations,

especially during years when the attention focused on the council isgreatest. Relying on the data used by Dreher and Sturm (2006) , weintroduce several de nitions of alignment of countries within the UNassembly to US,France,the UK andmorebroadly theG-7 countriesas agroup. We also include a dummy controlling for temporary member-ship in the UN Security Council as in Kuziemko and Werker (2006) fortheUS, Dreher, Sturm and Vreeland(2006) forthe IMFand Dreher et al.(in press) for the World Bank.

Moreover, Przeworski and Vreeland (2000) and Dreher and Vaubel(2004) study the allocation of an IMF loan around election time andfound signi cant results. The former suggest that governments are

more likely to enter an agreement early in the election term, hopingthat any perceived stigma of signing an agreement will be forgiven orforgotten before the next elections. In other words, demand for IMFcredit might be higher after election years. Dreher and Vaubel (2004)suggest that the availability of IMF credit might indirectly help to

nance electoral campaigns.Finally, two dimensions are to be taken into account. First, Prze-

worski and Vreeland(2000) and Birdand Rowlands(2000) argued thatcountrieswith moreunstableand polarizedpoliticalsystemswillhavemore dif culties to arrange a credible adjustment program and will,therefore,havea higherincentiveto turn to theFund. Theyalso suggestthat the IMF could prefer lending in general to countries with goodgovernance. These results are con rmed by the analysis of Sturm et al.(2005) . We include therefore different proxy of government stability,political opposition and government fractionization of political powerusing the database of Political Institutions of the World Bank. 11Second, IMF loans have been found to be rather persistent ( Bird, 1996;IMF/IMF, 2002; Bird etal., 2004 ), i.e. thelikelihoodof an additional loancould be determined in part by past loans. We therefore capture thishigh degree of persistence in IMF involvement as in the relatedliterature (see among others Przeworski and Vreeland, 2000, Hutch-ison andNoy, 2003; Sturmet al., 2005 ) using the lag ofa 3-yearmovingaverage of a dummyindicating whether or nota country was under anagreement.

4.1.2. Dependent variablesIMF loans are granted to ease the adjustment policies and reforms

that a country must make to correct its balance of payments problemand restore conditions for strong economic growth. They are mainly

provided under an arrangement , which stipulates the speci c policiesand measures a country has agreed to implement to resolve its balanceof payments problem. Theeconomic program is presented to the Fund'sExecutive Board in a Letter of Intent . Over the years, the IMF has de-veloped various facilities to address the speci c circumstances of itsdiverse membership. More speci cally, IMF nance is divided into tworesources account: First, the concessional loans allow low-incomecountries to borrow through the Poverty Reduction and Growth Facility(PRGF) and the Exogenous Shocks Facility (ESF). Second, non-

Table 5Description of the dependant variables

Stand-by agreements Extended fund facility Poverty reductionand growth

Structuraladjustment

Agreed Drawn Agreed Drawn Agreed Drawn Agreed Drawn

Programs sum 188 561 431 131 343 846 34 877 525 19 369 402 13 910 293 9 996 578 283 950 246 250Programs sum /total lent over the period 1990 2003 79.3% 81.6% 14.7% 12.0% 5.9% 6.2% 0.1% 0.2%Mean (for borrowing countries) 1,201,028 1,113,083 1,125,081 744,977 130,003 93,426 70,988 61,563Mediane (for borrowing countries) 100,000 86,770 353,160 144,625 73,380 51,890 40,040 29,090Standard deviation (for borrowing countries ) 3,726,691 3,097,080 1,610,927 1,203,137 160, 870 126,356 74,717 82,527Number of programs 157 118 31 26 107 107 4 4Number of programs/total programs over the period 1990 2003 52.5% 46.3% 10.4% 10.2% 35.8% 42.0% 1.3% 1.6%

10 We have also run our estimations with additional economic variables (currentaccount balance and total external debt) but too many observations were lost due to

the lack of data. They are available upon request to the authors.

11 Government stability counts the percent of veto players who drop from thegovernment in any given year. Political opposition records the total vote share of allopposition parties. Finally, government fractionization is the probability that twodeputies picked at random from among the government parties will be of different

parties ( World Bank, DPI2007 ).

Table 4Factor selection criteria: AIC and BIC

Factor number Eigenvalue Difference Proportion Cumulative Log. Like, DF m DF r AIC BIC

Factor1 4.35512 3.39043 0.7019 0.7019 1553.667 14 77 3135.335 3206.157Factor2 0.96469 0.17876 0.1555 0.8574 960.8392 27 64 1975.678 2112.265Factor3 0.78593 0.28186 0.1267 0.9841 583.487 39 52 1244.974 1442.266Factor4 0.50407 0.22034 0.0812 1.0653 330.9682 50 41 761.9363 1014.874


8/13/2019 Reynaud 02

8/24

concessional loansare providedmainly through Stand-ByArrangements(SBA), and occasionally using the Extended Fund Facility (EFF), theSupplemental Reserve Facility (SRF), and the Compensatory FinancingFacility (CFF). The IMF also provides emergency assistance to supportrecovery from natural disasters and con icts, in some cases atconcessional interest rates. Except for the PRGF and the ESF, all facilitiesare subject to the IMF's market-related interest rate and some carry asurcharge (mainly for large loans). The rate of charge is based on theSpecial Drawing Rights interest rate, which is revised weekly to takeaccount of changes in short-term interest rates in major internationalmoney markets. The amount that a country can borrow from the Fundvaries depending on the type of loan, but is typically a multiple of the

country's quota. The limit is xed according to the Articles of Agreements to 100% of the quota per year and 300% on a cumulativebasisof3 yearsregarding theSBA forexample. Ofcourse, these limitscanbe extended in special occasions. For example, South Korea and Turkeygotmorethan 1500%of theirquotaduring nancial distress, respectivelyin 1997 and in 1999/2000.

Since we focus on lending and given that industrial countries havenot made use of the Fund's nancial support for the last three decades,our panel comprises 107 IMF developingand emerging economies overthe period 1990 2003sampled at theyearly frequency. 299agreementshave been agreedaccounting forover 237,633,199 thousands of SDRand255 agreements have been drawn accounting for over 160,956,076

Chart 2. Geographical repartition of the recipients and the funds of SBA and PRGF.

Chart 1. Evolution of the relative total amounts and numbers of SBA and PRGF.


8/13/2019 Reynaud 02

9/24

thousands of SDRs. Table 5 shows descriptive statistics for ourdependent variables. Overall, agreements are slightly equallydistributed between SBA and PRGF, 46% and 42% of total loansdrawn respectivelyas shown in Chart 1 (in bars). However, looking at

the amount lent, Chart 1 (in lines) exhibits the sheer size of SBAcompared to PRGF. Indeed, SBA represent more than 80% of totalamount, compared to 6% for PRGF. This distinction has someeconomic bases since PRGF are oriented to support low-incomecountries, and therefore their needs are much less important thanemerging markets. Interestingly however, the amount and thenumber of PRGF are increasing over time. We will therefore focuson SBA and EFF for non-concessional loans and on PRGF forconcessional ones given their sheer size. Finally, looking at theregional distribution of loans is also quite informative. Chart 2represents the percentage of numbers of SBA (in black) and of PRGF(in grey) to total IMF loans per region over our sample period.Interestingly, we notice that the bulk of SBA drawn are in direction of Europe (including Turkey), Asia and South America, whereas PRGF

drawn are mainly oriented to support African countries.

Our dependent variables are therefore constructed as the ratio of the amount of IMF loans to the borrowing country's quota. The datawere retrieved from the IMF website which recently made availableonlinean increased level of data on nancial agreements. Wecollected

therefore data on loans agreed (i.e. the amount that was initiallygranted by the Executive Board) and on loans drawn (i.e. the amountthe country has effectively withdrawn). 12

4.2. Methodology

Our panel is unbalanced with a total of 1523 observations. As de-scribed above, our dependent variables are by de nition left censored

Table 6Core results: economic model of supply for IMF loans and geopolitical factor

Dependent variable/explanatory variables Stand-by agreements to quota (%) Poverty reduction and growth facilities to quota (%)

Agreed Drawn Agreed Drawn

Growth of GDP 4.475(4.14)

4.315(4.09)

4.184(4.01)

3.990(3.89)

1.872(2.82)

1.888(2.85)

1.431(2.73)

1.442(2.75)

Log of GDP per capita 1.005(5.24)

0.908(4.99)

0.895(4.59)

0.788(4.26)

0.834(12.45)

0.767(10.45)

0.665(11.12)

0.612(9.72)

FX reserves to imports 0.789(1.85)

1.104(2.12)

1.399(2.63)

1.961(2.84)

0.873(2.20)

0.822(2.02)

0.701(2.18)

0.662(2.01)

Debt service 4.079(3.10)

3.407(2.75)

3.701(2.92)

3.041(2.57)

0.791(1.65)

0.812(1.73)

0.573(1.49)

0.588(1.57)

Geopolitical factor gf 0.554(3.61)

0.706(3.84)

0.160(1.70)

0.127(1.76)

Constant 10.523(5.20)

9.581(5.05)

9.661(4.56)

8.616(4.39)

2.777(4.89)

2.291(3.82)

2.262(4.80)

1.876(3.84)

Pseudo-R 2 for Tobit estimations 0.0930 0.1090 0.1090 0.1322 0.1520 0.1550 0.1611 0.1643Observations 1163 1163 1163 1163 1163 1163 1163 1163Countries 98 98 98 98 98 98 98 98

Interval regression estimator marginal effect reported robust absolute value of t statistics in parentheses.Signi cant at 10%; signi cant at 5%; signi cant at 1%.

Table 7Core results: model of supply for IMF loans and geopolitical potential

Dependent variable/explanatory variables Stand-by agreements to quota (%) Poverty reduction and growth facilities to quota (%)



4.375(4.15)

4.313(4.14)

4.157(4.01)

1.914(2.91)

1.912(2.90)

1.463(2.81)

1.461(2.80)


0.868(5.00)

0.761(4.44)

0.726(4.23)

0.800(10.89)

0.759(10.11)

0.639(9.98)

0.606(9.46)


1.013(2.01)

1.478(2.47)

1.774(2.70)

0.865(2.18)

0.825(2.03)

0.695(2.17)

0.664(2.02)


3.945(2.92)

4.607(3.25)

3.978(2.93)

0.623(1.26)

0.705(1.47)

0.442(1.12)

0.507(1.32)

Geopolitical potential: gp 0.083(3.00)

0.120(3.47)

0.018(1.18)

0.014(1.16)

Geopolitical factor gf 0.393(2.63)

0.444(2.88)

0.129(1.26)

0.104(1.30)

Pure geopolitical potential gpf 0.054(1.94)

0.088(2.70)

0.011(0.66)

0.008(0.63)

Constant 9.773(5.26)

9.362(5.11)

8.730(4.63)

8.309(4.47)

2.544(4.16)

2.244(3.66)

2.083(4.17)

1.841(3.70)

Pseudo- R2 for Tobit estimations 0.1035 0.1080 0.1372 0.1443 0.1540 0.1557 0.1631 0.1649Observations 1163 1163 1163 1163 1163 1163 1163 1163Countries 98 98 98 98 98 98 98 98

Interval regression estimator marginal effect reported robust absolute value of t statistics in parentheses.Signi cant at 10%; signi cant at 5%; signi cant at 1%.

12 For instance, in September 2003, Argentina has negotiated a program with theFund. The agreed amount was 8,891 millions SDR and the member country drawn overyears 4,171 millions SDR. Both amounts were coded for the year 2003 and therefore donot appear in other years in the dataset. More granular data on the timing of drawnamounts is also available (at the monthly frequency), unfortunately these level of

details is not available before 1996.


8/13/2019 Reynaud 02

10/24

Table 8

Robustness checks: political factors

Dependent variable/explanatory variable

Stand-by agreements to quota (%)

Agreed Drawn


4.288(3.76)

4.118(3.33)

3.962(3.53)

4.134(3.59)

4.399(3.82)

4.337(3.78)

4.021(3.45)

3.945(3.47)

3.493(3.09)

Log of GDP per capi ta 0 .722(3.58)

0.929(3.85)

0.800(3.82)

0.759(3.99)

0.933(3.81)

0.885(3.68)

0.886(3.71)

0.611(3.22)

0.811(3.34)

0.647(3.42)


1.108(1.65)

0.921(1.65)

0.885(1.59)

1.125(1.62)

1.226(1.76)

1.210(1.75)

1.698(2.28)

1.973(2.46)

1.279(2.35)


3.514(2.73)

1.702(1.95)

2.060(2.64)

3.588(2.83)

3.354(2.72)

3.318(2.74)

2.980(2.44)

3.117(2.41)

1.373(1.94)

Geopol it ical factor gf 0 .490(2.77)

0.521(2.87)

0.479(2.94)

0.453(3.04)

0.493(2.75)

0.513(2.87)

0.508(2.86)

0.589(3.15)

0.631(3.12)

0.511(3.44)

Vote in line with US in UN 6.818(1.73)

6.922(1.65)

UN Security council seat 0.968(1.76)

0.831(1.75)

Election duringprevious 12 months

0.028(0.08)

0.081(0.26)

Election duringfollowing 12 months

0.875(2.19)

Government stability 0.001(0.67)

Political opposition 0.001(1.23)

Governmentfractionization

0.001(1.23)


9.683(3.84)

8.376(3.88)

7.952(4.03)

10.337(3.82)

9.315(3.72)

9.817(3.75)

8.559(3.38)

8.718(3.29)

6.956(3.44)

Observations 1163 1163 1163 1163 1163 1163 1163 1163 1163 1163 116Countries 98 98 98 98 98 98 98 98 98 98

Interval regression estimator marginal effect reported robust absolute value of z statistics in parentheses.Signi cant at 10%; signi cant at 5%; signi cant at 1%.

8/13/2019 Reynaud 02

11/24

Table 8

Robustness checks: political factors

Poverty reduction and growth facilities to quota (%)

Agreed Drawn

1.929(2.89)

1.878(2.72)

2.074(2.61)

2.196(2.88)

1.596(2.49)

1.695(2.55)

1.771(2.66)

1.487(2.72)

1.424(2.55)

1.582(2.48)

1.685(2.73)

1.295(2.41)

1.398(2.52)

1.440(2.61)

0.866(7.82)

0.752(8.01)

0.844(7.00)

0.854(8.44)

0.728(7.96)

0.723(7.91)

0.732(7.96)

0.689(7.58)

0.597(7.76)

0.680(6.54)

0.687(8.14)

0.583(7.69)

0.581(7.73)

0.583(7.79)

0.941(2.24)

0.808(2.21)

0.721(1.22)

0.778(1.79)

0.816(2.21)

0.684(2.02)

0.688(2.03)

0.771(2.25)

0.651(2.22)

0.588(1.22)

0.618(1.75)

0.673(2.22)

0.566(2.04)

0.562(2.03)

0.970(1.73)

0.838(1.44)

1.244(2.01)

1.218(2.36)

0.765(1.46)

0.890(1.54)

0.867(1.50)

0.718(1.70)

0.609(1.38)

0.875(1.87)

0.918(2.37)

0.576(1.39)

0.669(1.51)

0.661(1.51)

0.254(1.70)

0.212(1.57)

0.312(2.02)

0.260(1.73)

0.240(1.84)

0.237(1.75)

0.239(1.75)

0.210(1.82)

0.17(1.62)

0.257(2.15)

0.215(1.84)

0.198(1.95)

0.197(1.85)

0.197(1.85)

2.651(1.73)

1.920(1.50)

0.974(1.61)

0.834(1.68)

0.676(1.87)

0.488(1.69)

0.858(3.15)

0.656(3.12)

0.713(2.27)

0.557(2.22)

0.001(2.12)

0.001(2.21)

0.001(2.09)

0.001(2.21)

2.387(3.38)

2.223(3.26)

1.281(1.21)

2.416(3.73)

0.935(1.01)

1.867(2.92)

0.855(0.89)

1.945(3.39)

1.812(3.30)

1.102(1.27)

1.982(3.75)

0.814(1.09)

1.536(2.98)

0.733(0.95)

1163 1163 1163 1163 1163 1163 1163 1163 1163 1163 1163 1163 1163 1163 98 98 98 98 98 98 98 98 98 98 98 98 98 98

8/13/2019 Reynaud 02

12/24

Table 9Robustness checks: aid determinants


Stand-by agreements to quota (%) Poverty reduction and growth facilities to quota (%)



4.241(3.46)

4.291(3.75)

4.113(3.49)

3.944(3.35)

3.954(3.47)

1.827(2.84)

1.716(2.03)

1.899(2.86)

1.397(2.68)

1.271(1.84)

1.450(2.71)


0.988(3.51)

0.869(3.67)

0.709(3.13)

0.837(3.02)

0.804(3.30)

0.776(8.60)

0.787(6.99)

0.729(8.04)

0.618(8.46)

0.615(7.18)

0.58(7.56)


1.207(1.61)

1.089(1.59)

2.066(2.47)

2.005(2.44)

1.996(2.46)

0.774(2.14)

0.616(1.39)

0.745(2.15)

0.624(2.14)

0.473(1.36)

0.600(2.15)

Debt service 3.391

(2.75)

3.844

(2.83)

3.293

(2.74)

2.974

(2.43)

3.390

(2.46)

2.963

(2.44)

0.991

(1.91)

1.427

(2.98)

0.904

(1.59)

0.724

(1.79)

1.077

(3.20)

0.660

(1.56)Geopolitical factor gf 0.463

(2.66)0.347(2.03)

0.464(2.68)

0.589(3.00)

0.467(2.49)

0.610(3.12)

0.244(1.81)

0.240(1.43)

0.236(1.76)

0.196(1.89)

0.192(1.53)

0.190(1.85)

Democratic score 0.084(2.00)

0.065(1.85)

0.007(1.33)

0.005(1.32)

Government corruption 0.147(0.96)

0.189(1.18)

0.066(0.62)

0.031(0.38)

Former colony 0.370(0.99)

0.021(0.06)

0.140(0.70)

0.109(0.70)


10.477(3.57)

9.092(3.68)

8.345(3.27)

9.857(3.14)

8.663(3.27)

2.342(3.56)

2.553(2.62)

1.927(2.74)

1.907(3.62)

2.024(2.68)

1.584(2.77)

Observations 1163 1163 1163 1163 1163 1163 1163 1163 1163 1163 1163 1163 Countries 98 98 98 98 98 98 98 98 98 98 98 98

Interval regression estimator marginal effect reported robust absolute value of z statistics in parentheses.Signi cant at 10%; signi cant at 5%; signi cant at 1%.

8/13/2019 Reynaud 02

13/24

to 0 and uncensored on the right side . This calls for a censored re-gression model such as the Tobit estimator. The model is thereforespeci ed hereafter, as in Barro and Lee (2005) :

Lit X it Git T t it 5

Lit max 0 ; Lit 6 where the dependent variable, Lit , is the loan-size variable for country

i during period t . Lit =0 if the country did not have a loan agreementwith the IMF during period t . The vector X it denotes the country-speci c economic macro-aggregates that in uence the existence andsize of IMF programs. As discussed before, this vector includes theratio of foreign reserves to imports, debt service to exports, per capitaGDP and GDP growth. The regression also includes time dummies tocontrol for common effects of external factors such as world interestrates. Git contains the measures of country's geopolitical importanceas discussed in Section 3. It includes: First, the geopolitical factor of countries gf it ; second, their geopolitical potential gp it and pure geo-political potential gpf it . Finally, the variable it is a random error term.

Eq. (5) can be viewed as a reduced-form model of supply for IMFloans from a debtor's perspective. To minimize reverse-causality pro-blems, all explanatory variables are measured as lagged values. Some

variables enter as their log values to deliver the best goodness-of- t.13

Moreover, we use random-effects speci cations for the error termsince the probability that a country is favoured by the IMF during oneperiod is likely to be persistent over time, i.e. there is great deal of recidivism in IMF lending practices as argued by Bird (1996) . Thisassumption is supported by econometrical tests shown in the lastsection of the paper. Finally, the Breusch and Pagan Lagrangian mul-tiplier test indicates that for SBA, our sample shows some hetero-skedasticity. We may therefore produce robust variance estimates of marginal effects.

5. Estimation results

5.1. Core results

Core results are shown in Tables 6 and 7 . In each table, we estimateseparately modelsof supplyfor agreed anddrawn amounts of SBA/EFFand PRGF. Regarding the economic model (odd columns) for SBA/EFF,countriesexperiencing relatively weakgrowth in realGDP are foundtoreceive more credit as expected. Indeed, the estimated parameters arefound signi cant at the 1% level and negative for SBA/EFF as in Sturmet al. (2005) . Moreover, the positive relation between IMF lending andGDP per capita may re ect the Fund's reluctance to provide stabiliza-tion loans to countriesthat arenot creditworthy ( Barro andLee, 2005 ).As argued by Knight and Santaella (1997) , countries experiencingrelatively low levels of international reserves relative to imports arefound to receive more IMF credit. Indeed, these countries will be lessable to meet balance of payments dif culties through reserves use.Finally, a heavy debt burden relative to exports increases countries'probability to be nancially supported to service that debt. As inRowlands(1995), we foundthis estimated parameter signi cantly andpositively related for SBA/EFF.

The picture is however reversed and less robust for the PRGF, atleast for GDP growth and per capita GDP. Indeed, the parameterestimated of per capita GDP is signi cant and negative. In accordanceto Knight and Santaella (1997) , we nd that poor countries are morelikelyto be nancially supported. Indeed, these countries have limitedaccess to private international capital markets and are also smallrecipients of bilateral aid. Interestingly, we nd that GDP growth issigni cant with a positive sign. Harrigan et al.(2006) found the similarresult without explaining it. We argue that since access to PRGF is

mainly granted to compensate the small amount of bilateral aid owsand is therefore mainly conditioned to a certain level of GDP percapita; since PRGF are concessional and account for small amountcompare to SBA/EFF, these loans are granted more easily once thecountry has been designated as eligible.

Although the above economic model provides useful insights intothe determinants of IMF programs, its explanatory power may beimproved including variables capturing countries' geopolitical impor-

tance as argued by IMF staff (see citations above). Even columns inTables 6 and 7 present our model including our proxies of thegeopolitical importance of countries. Our factor gf it is found to besigni cant at the 1% level and positively related to SBA/EFF, whereas itis signi cant at the 10% level and negatively related to PRGF (seeTable 6 ). Therefore, our results exhibit that the IMF Executive Board isfavouring geopolitically important countries when lending throughnon-concessional facilities, and favouring non-geopolitically impor-tant countries when lending via concessional ones, although the lateris not fully robust. The results for the supplemented models show astrong improvement of the explanatory power of the estimations.

As described above, we constructed a different way to estimate Git using our geopolitical potential analysis. Indeed, we argue that oneshould not only take into account the geopolitical importance of acountry, but also its geographical importance. We introduce gp it inEq. (5) and the results are, in many respects, similar to those found inthe previous table. Even columns of Table 7 show the result of ourmodel using the geopolitical potential of countries.

Another concern in this paper is to follow the international tradeliterature concerning the geopolitical factor. The method of calculationof the internal distance is problematic and depending of it, this mayintroduce a bias in the potential. Therefore, this could explain the factthat the geopolitical potential is less signi cant than the geopoliticalfactor. Dividing each country's geopolitical factor by its internal di-stance may affect the result since the factor analysis is bounded to aninterval [ 1.75, 2.19] and country's internal distance [13.84, 2754.81]in log terms. We therefore test also separately our measure of puregeopolitical potential, gpf i, on top of our geopolitical factor, gf i. Resultsare robust for SBA/EFF while the levelsof signi cance of ourgeopoliticalproxies are decreased for the PRGF estimation (see odd columns of Table 7 ). We investigate in the next section possible explanations.

What arises from Tables 6 and 7 is the fact that countries that aregeopolitically important are favoured by the IMF when loans are non-concessional. Wealso nd thatconcessional loans, i.e. PRGF, aregrantedto less geopolitically important countries, which could re ect thedevelopment objectives of these loans. As arguedabove,multilateralaidmaybe directed to less geopoliticalcountriesto compensate thefact thatthey receive less bilateral aid. Therefore, this negative sign does notre ectthe fact thatthe IMFwantsto lend to notgeopolitically importantcountries but rather that the PRGF eligible countries are not geopoliti-cally important.

13

The results are not sensitive to the speci c values added for the log transformations.

Table 10

Robustness checks: recidivism


Stand-by agreementsagreed to quota (%)

Stand-by agreements agreed(dummy: 1 for loan, 0 otherwise)

Tobit Dynamic probit

Growth of GDP 3.998 (3.34) 1.009 (2.14)Log of GDP per capita 0.807 (3.46) 0.191 (2.19)FX reserves to imports 1.125 (1.83) 1.059 (2.33)Debt service 3.263 (2.83) 0.885 (1.56)Geopolitical factor: gf 0.453 (2.91) 0.241 (2.11)Recidivism: 3-year lag of the dependent variable

0.924 (3.45) 0.358 (2.11)

Constant 9.365 (3.73) 2.774 (4.33)Observations 940 940Countries 98 98

Absolute value of z statistics in parentheses.

Signi cant at 10%; signi cant at 5%; signi cant at 1%.


8/13/2019 Reynaud 02

14/24

5.2. Robustness checks

5.2.1. On the importance of political factorsAs discussed in Sections 2 and 4, related studies on political eco-

nomy determinant of IMF loans found that IMF loan decisions are

signi cantly in uenced by political factors. We test these factors inthis section adding to Eq. (5) P it that comprises the proxies of politicaleconomy factors that have been found to signi cantly in uence theattribution of IMF loan as detailed in the previous section.

Lit X it P it Git T i it 7

Table 8 shows the results of multiple speci cations including thecorrelation of the borrowing countries to the US in the UN GeneralAssembly; dummy variables to account for a seat in the UN SecurityCouncil, post and pre electionsperiods; indexes capturing governments'instability, the degree of political opposition and governments' politicalfractionization. Finally, we pool together these determinants in the lastcolumn of the Table 8 to test for their robustness.

The results are in line with the related studies. Indeed, voting in linewith the US in the UN Assembly appears to be the most importantpolitical factor shaping IMF loan decision. Holding a non-permanentseat in the UN Security Council is also positively related to the prob-ability to get IMF money but appears to be less robust. Entering aprogram after, but not before elections is also found to be signi cant.Interestingly, factors capturing the degree of political stability in uencerather more PRGF than SBA/EFF. Overall, our geopolitical factor showsrobust signi cant estimated coef cients, empathising that we are notcapturing political factors within our geopolitical factor.

5.2.2. On the importance of factors explaining aid owsSince PRGF are multilateral aid agreements, we might also test for

the importance of relevant determinants put forward in the aid

literature. Burnside and Dollar (2000) for example, found that the

level of democracyand corruption in therecipient countries, as well asthe former colonial link between countries are strong determinants of bilateral aid decisions. However, they also found that these results arenot robust for multilateral aid ows. As in the previous section, weadded in Eq. (5) these proxies in Ait as follows:

Lit X it Ait Git T t it 8

Table 9 show the results of our model including these variablesindependently and pooled altogether.

In accordance with the ndings of Burnside and Dollar, we foundthat these factors do not in uence signi cantly PRGF decisions. Onlythe level of democracy index is found to be signi cant in our models.This is not surprising given that, contrary to bilateral aid, aid managedon a multilateral basis is rather allocated in favour of good policyaccording to Burnside and Dollar (2000) . Once again, our geopoliticalfactor is found to be robust and shows signi cant estimatedcoef cients for SBA/EFF and still to a lesser extent for PRGF.

5.2.3. On the importance of recidivism

As discussedin Section 4, SBA/EFF arefoundto be rather persistent.Therefore, we tested this in our speci cation by introducing a 3-yearmoving average of a dummy indicating whether or not a country wasunder an agreement following Przeworski and Vreeland (2000) .Results are reported in Table 10 . The dummy is found to be signi cantindicatingthat there is somedegreeof persistencein IMFloan decision.Moreover, we also estimated a probit dynamic speci cation usingthe model developed by Stewart (2007) .14 His estimator control forthe initial conditions problem proposed by Heckman (1981) involvesspecifying a linearised approximationto the reduced formequation for

Table 12Robustness checks on the sample

Dependent variable/explanatory variables

Stand-by agreements to quota (%) Poverty reduction and growth facili ties to quota (%)


Growth of GDP 5.185 (3.86) 5.443 (3.72) 5.129 (3.51) 5.385 (3.40) 2.049 (3.25) 2.069 (3.16) 1.582 (3.17) 1.598 (3.06)Log of GDP per capita 0.796 (3.83) 0.843 (3.33) 0.671 (3.10) 0.674 (2.79) 0.798 (12.01) 0.811 (9.15) 0.635 (11.15) 0.647 (8.29)Geopolitical factor gf i 0.781 (3.98) 0.923 (4.02) 0.223 (2.53) 0.178 (2.61)Geopolitical potential: gp i 0.045 (1.75) 0.083 (2.39) 0.028 (2.08) 0.022 (2.10)Constant 8.925 (4.24) 9.170 (3.74) 8.070 (3.58) 7.967 (3.27) 2.613 (4.86) 2.691 (4.39) 2.044 (4.89) 2.114 (4.13)Observations 1425 1425 1425 1425 1425 1425 1425 1425Countries 107 107 107 107 107 107 107 107

Cluster analysis interval regression estimator marginal effect reported robust absolute value of t statistics in parentheses.Signi cant at 10%; signi cant at 5%; signi cant at 1%.

Table 11Robustness checks: cluster analysis


Stand-by agreements to quota (%) Poverty reduction and growth faciliti es to quota (%)


Growth of GDP 4.315 (3.77) 4.506 (3.91) 3.990 (3.49) 4.312 (3.75) 1.888 (2.82) 1.909 (2.83) 1.442 (2.68) 1.460 (2.71)Log of GDP per capita 0.908 (3.80) 0.880 (3.89) 0.788 (3.30) 0.740 (3.48) 0.767 (8.25) 0.802 (9.01) 0.612 (7.95) 0.639 (8.41)FX reserves to imports 1.104 (1.61) 0.833 (1.35) 1.961 (2.44) 1.532 (2.36) 0.822 (2.31) 0.861 (2.40) 0.662 (2.31) 0.691 (2.41)Debt service 3.407 (2.83) 4.504 (3.17) 3.041 (2.50) 4.450 (2.96) 0 .812 (1.41) 0.653 (1.14) 0.588 (1.34) 0 .461 (1.04)

Geopolitical factor gf i 0.554 (2.84) 0.706 (3.14)

0.160 (1.40)

0.127 (1.46)Geopolitical potential: gp i 0.083 (2.61) 0.114 (2.91) 0.014 (1.03) 0.012 (1.06)Constant 9.581 (3.87) 9.564 (4.03) 8.616 (3.32) 8.515 (3.53) 2.291 (3.43) 2.563 (3.91) 1.876 (3.48) 2.090 (3.91)Observations 1163 1163 1163 1163 1163 1163 1163 1163Countries 98 98 98 98 98 98 98 98

Cluster analysis interval regression estimator marginal effect reported robust absolute value of t statistics in parentheses.Signi cant at 10%; signi cant at 5%; signi cant at 1%.

14 The dependent variable is therefore coded 1 when the country signs an agreement

and 0 when there is no loan.


8/13/2019 Reynaud 02

15/24

8/13/2019 Reynaud 02

16/24


Stand-by Agreements to quota (%) Poverty Reduction and Growth Facilities to quota (%)

Drawn Agreed


4.222(4.10)

4.264(4.01)

4.125(4.01)

4.234(4.08)

3.936(3.90)

4.386(4.15)

4.158(4.02)

4.269(3.60)

1.898(2.84)

1.889(2.83)

1.902(2.86)

1.875(2.82)

1.8(2.

Log of GDP percapita

0.852(4.36)

0.820(4.64)

0.895(4.66)

0.801(4.41)

0.804(4.41)

0.983(4.62)

0.780(4.47)

0.927(4.53)

0.470(2.28)

0.751(10.18)

0.790(10.79)

0.790(9.50)

0.819(9.28)

0(11

FX reserves toimports

1.548(2.66)

1.558(2.71)

1.341(2.61)

1.610(2.70)

1.294(2.33)

1.893(3.02)

1.666(2.77)

1.479(2.71)

1.830(2.40)

0.836(2.07)

0.812(2.01)

0.883(2.21)

0.869(2.17)

0(2.


3.457(2.88)

3.532(2.97)

3.541(2.86)

3.522(2.84)

2.956(2.55)

3.157(2.67)

3.564(2.88)

2.406(2.36)

0.757(1.60)

0.787(1.68)

0.832(1.75)

0.791(1.66)

0.8(1.8

Log of provenoil reserves

0.196(0.93)

0.023(1.91)

Log of provengas reserves

0.169(1.20)

0.014(1.33)

Log of oil pipelines 0.180(1.60)

0.007(0.80)

Log of gas pipelines 0.062(0.64)

0.002(0.26)

Log of civil nuclearplant power

0.044(0.86)

0(1.5

Dummy for nuclear

weapon possession

1.206

(3.86)

0.583

(1.47)Log of U S militarystrength

0.120(2.20)

0.066(1.53)

Log of U N militarystrength

0.037(1.41)

0.170(2.43)

Index of non-proliferation treaties

1.947(2.96)

1.199(1.72)

Log of km. of coastlines

0.044(2.42)

0.079(1.44)

Log of total area 0.261(3.27)

0.103(0.68)

Log of km. of roads 0.213(3.45)

0.035(0.57)

Log of borders 0.036(1.23)

0.661(1.94)


8.338(4.74)

9.113(5.04)

10.212(4.62)

9.152(4.52)

13.169(4.54)

10.969(4.54)

9.802(4.56)

8.800(3.00)

2.029(3.18)

2.366(3.85)

2.449(3.72)

1.528(1.59)

2.1(3.1

Time dummies YES YES YES YES YES YES YES YES YES YES YES YES YES Observations 1163 1163 1163 1163 1163 1163 1163 1163 1163 1163 1163 1163 1163 1163 Countries 98 98 98 98 98 98 98 98 98 98 98 98 98 98

Table 13 (continued )

8/13/2019 Reynaud 02

17/24

Table 13 (continued )

Poverty Reduction and Growth Facilities to quota (%)

Agreed Drawn

1.850

(2.79)

1.868

(2.83)

1.875

(2.88)

1.874

(2.83)

2.078

(3.16)

1.449

(2.74)

1.443

(2.73)

1.454

(2.77)

1.433

(2.73)

1.404

(2.72)

1.399

(2.72)

1.432

(2.74)

1.456

(2.76)

1.452

(2.75)

1.41

(2.70.859(11.31)

0.837(12.59)

0.815(11.32)

0.833(12.38)

0.692(6.71)

0.603(9.40)

0.632(9.91)

0.629(8.95)

0.658(8.69)

0.628(10.35)

0.630(10.67)

0.666(10.96)

0.668(10.94)

0.654(10.37)

0.(10.

0.838(2.14)

0.868(2.14)

0.862(2.17)

0.877(2.21)

0.529(1.67)

0.657(2.06)

0.655(2.01)

0.710(2.19)

0.699(2.16)

0.656(2.05)

0.663(2.03)

0.702(2.19)

0.726(2.26)

0.648(2.06)

0.(2.1

0.772(1.61)

0.784(1.64)

0.808(1.70)

0.785(1.64)

0.829(1.34)

0.546(1.44)

0.570(1.51)

0.606(1.59)

0.573(1.49)

0.624(1.66)

0.609(1.61)

0.571(1.50)

0.659(1.62)

0.537(1.36)

0.56(1.4

0.794(2.04)

0.018(1.85)

0.038(0.30)

0.011(1.30)

0.055(1.14)

0.006(0.83)

0.061(1.37)

0.001(0.16)

0.078(1.60)

0.021(1.71)

0.069(0.15)

0.492(1.70)

0.003(0.17)

0.000(0.03)

0.046(1.32)

0.009(0.83)

0.668(2.33)

0.332(0.98)

0.005(0.52)

0.033(1.24)

0.00(0.4

0.053(0.99)

0.030(0.36)

0.022(0.77)

0.000(0.00)

0.004(0.16)

0.237(1.25)

2.925(4.89)

3.459(3.93)

1.711(2.19)

2.769(4.88)

1.670(1.16)

1.696(3.27)

1.950(3.89)

1.993(3.73)

1.336(1.72)

1.719(3.17)

2.069(4.42)

2.268(4.34)

2.148(4.41)

2.393(4.90)

2.35(4.7

YES YES YES YES YES YES YES YES YES YES YES YES YES YES 1163 1163 1163 1163 1163 1163 1163 1163 1163 1163 1163 1163 1163 1163 1163 98 98 98 98 98 98 98 98 98 98 98 98 98 98 98

8/13/2019 Reynaud 02

18/24

Table 14Robustness checks on different possible factors


Stand-by agreements to quota (%)

Agreed

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14

Growth of GDP

4.305(4.09)

4.305(4.09)

4.303(4.07)

4.311(4.08)

4.329(4.11)

4.333(4.11)

4.314(4.08)

4.316(4.09)

4.318(4.09)

4.311(4.08)

4.334(4.09)

4.304(4.07)

4.323(4.07)

(

Log of GDP

per capita

0.922

(5.03)

0.912

(5.01)

0.927

(5.04)

0.932

(5.05)

0.919

(5.03)

0.911

(5.01)

0.909

(4.98)

0.908

(4.99)

0.912

(5.00)

0.912

(5.00)

0.866

(4.84)

0.913

(4.98)

0.862

(4.79)

0

(FX reservesto imports

1.135(2.15)

1.122(2.14)

1.125(2.16)

1.119(2.15)

1.122(2.14)

1.103(2.12)

1.094(2.12)

1.105(2.12)

1.097(2.12)

1.104(2.13)

1.051(2.03)

1.094(2.11)

1.048(2.06)

(


3.282(2.68)

3.385(2.74)

3.370(2.73)

3.450(2.77)

3.434(2.76)

3.425(2.75)

3.408(2.75)

3.425(2.76)

3.405(2.74)

3.502(2.80)

3.458(2.77)

3.523(2.81)

3(

Geopoliticalfactor: gf

0.571(3.78)

0.580(3.79)

0.547(3.66)

0.553(3.67)

0.550(3.45)

0.543(3.45)

0.547(3.60)

0.556(3.61)

0.548(3.57)

0.547(3.61)

0.574(3.59)

0.539(3.52)

0.564(3.55)

0(


9.576(5.06)

10.202(5.08)

9.751(5.08)

9.660(5.09)

9.613(5.07)

9.593(5

reynaud 02

Documents