Conditional aid effectiveness: A meta-study

Journal of International Development

J. Int. Dev. 22, 391–410 (2010)

Published online 3 April 2009 in Wiley InterScience

(www.interscience.wiley.com) DOI: 10.1002/jid.1582

CONDITIONAL AID EFFECTIVENESS:A META-STUDY

HRISTOS DOUCOULIAGOS1 and MARTIN PALDAM2*1School of Accounting, Economics and Finance, Deakin University, Burwood, Victoria,

Australia2School of Economics and Management, Aarhus University, Aarhus, Denmark

Abstract: One branch of the aid effectiveness literature (AEL) analyzes conditional models

where aid effectiveness depends upon a conditioning variable z. The leading candidates for z

are a good policy index and aid itself, so that the model has an aid squared term. In this paper,

meta-analysis techniques are used (i) to determine whether the AEL has established the said

interaction terms, and (ii) to identify some of the determinants of the differences in results

between studies. We find no support for conditionality with respect to policy, nor with respect

to aid itself. Copyright # 2009 John Wiley & Sons, Ltd.

Keywords: conditional aid effectiveness; meta-study; economic growth

JEL Classification: B2; F35; O35

1 INTRODUCTION: A RESEARCH PROCESS DRIVEN BY TWO

CHALLENGES

Development aid is meant to help the (economic) development of poor countries. This is

certainly a noble purpose. However, it has been known for 30 years that the

correlation between the share of development aid and economic growth is essentially

zero.1

*Correspondence to: Martin Paldam, School of Economics and Management, Bartholins Alle 10, AarhusUniversity, 8000 Aarhus C, Denmark. E-mail: [email protected] is carefully documented in Mosley (1987), and replicated in Herbertsson and Paldam (2007), Rajan andSubramanian (2008), and Doucouliagos and Paldam (2008).

Copyright # 2009 John Wiley & Sons, Ltd.

392 H. Doucouliagos and M. Paldam

The zero correlation result is highly undesirable, and it has remained the old challenge

in the field. Many have found it implausible and risen to the challenge by searching for

models that allow the data to tell a ‘nicer’ story. This has generated the aid effectiveness

literature (AEL) of more than 100 papers. The amount, magnitudes and variability of the

data for aid and growth make them ideal for this research, and at least 100 man-years of

research has been invested into the AEL. This effort has led to a lot of valuable

information that is worth exploring. The present paper is one part of a set of meta-studies of

the AEL.

Research is a process of truth revelation which works through a mixture of innovations

(in data, models and estimators) and independent replications as new data

become available.2 A single study rarely resolves an important issue in any science.

Trust has to be built by replication. In macroeconomics, new data become available

slowly, so data sets of different studies are typically overlapping, making replication

partly dependent. Also, estimators are constantly improved, and models are often

amended.

The quantitative technique of meta-analysis is developed to analyze the evidence

contained in a sequence of studies of the same effect with methods that are so similar that

the differences can be coded. To reach valid results it is important that the meta-analysis

covers all available studies of the effect analyzed. The present paper covers two meta-

analyses (two effects) and asks two questions:3

Q1: Do the estimates of the effect converge to something we can term truth?

Q2: Can we identify the main innovations causing/preventing the convergence?

The paper deals with the newest family of aid effectiveness studies. It is the group of

conditional aid studies, which started in 1995, and had grown to 40 papers at the start of

2007, where our data collection stops. These studies take the zero correlation result to mean

that aid works in some cases and fails in others. Hence, the new challenge of the AEL is to

find the condition that determines when the good outcome results.

The operational meaning of the term conditional in the AEL is that the estimating

equation contains a second order term, where aid is multiplied with another variable termed

the condition.4 The interaction terms that have been proposed in the AEL have strong

policy-implications. Consequently, the purpose of the present meta-study is to find out if

the main interaction terms are established by the literature.

Section 2 classifies the AEL and discusses the theory of the conditionality family of

models. Meta-analysis is used in Section 3 to study the validity of the two main

conditionality models. Section 4 explains the differences in reported results. Section 5

summarizes the rather sad findings of the paper. The Appendix lists the 40 studies covered

in the meta-analysis.

2A study is an independent replication if it is conducted by a new researcher on new data.Dependent replication isby a new researcher on the same data, or the same researcher on new data.3Meta-studies can also be used to study the effect of priors and the existence of publication bias, see for exampleRoberts and Stanley (2005) and Doucouliagos and Paldam (2009).4Thus we distinguish between a condition that enters multiplicatively with aid and the control set that controls theestimate of country heterogeneity and other ‘‘disturbing’’ factors. In growth empirics the term conditionalnormally means that the estimate contains a set of variables, which controls the relation for country heterogeneity.This paper uses the word in the AEL sense.

Copyright # 2009 John Wiley & Sons, Ltd. J. Int. Dev. 22, 391–410 (2010)

DOI: 10.1002/jid

Figure 1. The causal structure in the three families of AEL models

Conditional Aid Effectiveness: A Meta-Study 393

2 FROM AID EFFECTIVENESS TO CONDITIONAL AID EFFECTIVENESS

A thorough search produced the AEL papers listed in Christensen et al. (2009).5 The papers

bring many models, which can be divided by their causal structure, into three families as

shown on Figure 1. Half the papers estimate models from more than one family.

A: 4

5Extcitatpape

Cop

3 papers study the impact of aid on savings or investment; see Doucouliagos

and Paldam (2006). About 3/4 of aid is crowded out by a fall in savings mainly

due to increases in public consumption. The remaining ¼ causes increasing

investments.

B: 6
8 papers contain a total of 543 direct estimates, using reduced form models of the
effect of aid on growth; see Doucouliagos and Paldam (2008). The estimates scatter

considerably and add up to a small positive, but insignificant, effect on growth. The

zero correlation result has yet to be overcome.

C: 4
0 papers—see the Appendix—contain conditional estimates, where the effect of aid
on growth depends upon a conditional variable z. This is the family analyzed in this

paper.

The AEL was started in the early 1970s by papers in the A-group. The early studies

found no effect of aid on accumulation, and the AEL then moved on to the B-group papers,

where most of the research in the 1980s and 1990s was done. The C-group started in 1995.

It is where most of the action has been since then. This wave of papers is still strong, so we

are discussing an ongoing process.

2.1 Conditional Aid Effectiveness: The Models

In the mid 1990s, the C-family of models appeared. The C-family of studies is based on the

idea that aid has a positive effect on growth in some cases and a negative effect in others, so

ensive searches of Econlit, Proquest,Web of Science and Googlewere undertaken, fromwhich we could trackions backward. Papers available only after 1st of January 2007 are not included in the study. Some workingrs have been published after the cut-off date. Here we have used the published version.

yright # 2009 John Wiley & Sons, Ltd. J. Int. Dev. 22, 391–410 (2010)

DOI: 10.1002/jid


that aid is conditionally effective. Till now, three conditions have been proposed, leading

to three models of which only the first two are sufficiently homogeneous for a meta-

study:

(1) G

6The(200(200HoddDalg(200Anti7The(200(200ColliChau(2008Othpaymgeog

Copy

ood Policy model: Aid works if the recipient country pursues good polices, and is

harmful in countries pursuing bad policies. The model was proposed by Burnside and

Dollar (1995, 2000), and has been developed by a group in theWorld Bank. The model

has been analyzed in 27 more papers.6

(2) M
edicine model: Aid works if given in moderation, and harms if taken in excess, just
like most medicine. The model was first proposed by Hadjimichael, Ghura, Muhleisen,

Nord and Ucer (1995), but it has mainly been advocated by the group of Tarp, Dalgaard

and Hansen from DERG at Copenhagen University. Most members of this group are

associated with Danida, the Danish Aid Agency. The model has been further analyzed

in 21 papers.7

(3) I
nstitutions models: A residual of 10 papers contain models that condition for various
institutions: Two papers condition for democracy (Svensson 1999; Kosack 2002); one

for external vulnerability (Guillaumont and Chauvet 2001); two for quality of institu-

tions (Collier and Dehn 2001; Collier and Dollar, 2004); one for trade openness

(Teboul and Moustier 2001); and one for economic freedom (Brumm 2003). Two

condition for political instability (Chauvet and Guillaumont 2004 and Chauvet 2005).

Finally, two studies link aid with GDP (Bowen 1995; Svensson 1999).8

Several of the papers analyse two (or more) models. While the future development may

be within the institutional models, the first two models have been the prominent ones till

now, each leading to a stream of papers. This in particular applies to the Good Policy

model, which dominated the macroeconomic aid discussion for almost a decade from

1995.

It is noteworthy that the two most prominent models were advocated by a group which

was associated with an aid agency. It is clearest in the case of the Good Policymodel, where

most proponents are (were) World Bank staff, and the model was advocated to a broader

audience in a World Bank Policy Research Report (1998). The DERG group, which is the

y are: Svensson (1999); Hansen and Tarp (2000; 2001); Collier and Dehn (2001); Dalgaard and Hansen1); Guillaumont and Chauvet (2001); Hudson and Mosley (2001); Lensink and White (2001); Lu and Ram1); Collier and Dollar (2002); Brumm (2003); Cordella and Dell’Ariccia (2003); Dayton-Johnson andinott (2003); Burnside and Dollar (2004); Chauvet and Guillaumont (2004); Collier and Hoeffler (2004);aard, Hansen and Tarp (2004); Denkabe (2004); Easterly, Levine and Roodman (2004); Jensen and Paldam4); Ram (2004); Roodman (2004); Shukralla (2004); Rajan and Subramanian (2005); Chauvet (2005);pin and Mavrotas (2006); and Murphy and Tresp (2006).y are: Durbarry, Gemmell and Greenaway (1998); Hansen and Tarp (2000; 2001); Dalgaard and Hansen1); Hudson and Mosley (2001); Lensink and White (2001); Collier and Dollar (2002); Cungu and Swinnen3); Denkabe (2003); Islam (2003); Moreira (2003); Ovaska (2003); Clemens, Radelet and Bhavnani (2004);er and Hoeffler (2004); Dalgaard, Hansen and Tarp (2004); Jensen and Paldam (2004); Roodman (2004);vet (2005); Gomanee, Girma and Morrisey (2005); Rajan and Subramanian (2008); and Pavlov and Sugden6).er interactions have been considered. For example, Islam (2003) interacts aid with the share of transferents, while Dalgaard, Hansen and Tarp (2004) and Rajan and Subramanian (2005) condition onraphy.

right # 2009 John Wiley & Sons, Ltd. J. Int. Dev. 22, 391–410 (2010)

DOI: 10.1002/jid

Figure 2. Interpreting the data to get positive aid effectiveness


main advocate for the Medicine model, is almost as closely associated with Danida

(Danish Aid Agency).9 Thus, these models can be said to have an institutional home.10

The two models are advocated by a group, in the sense that we deal with persons, who

publish together in various combinations. It seems that both groups are now dissolved.

2.2 The Data

Most of the 40 papers are estimated on the standard ODA data for official development aid

as compiled by the OECD. It contains all gifts and loans on concessional terms (which

contain a grant element of at least 25 per cent) from the OECD countries to LDCs (less

developed countries). However, the early discussion between the two main conditionality

models took place using the EDA-data of Chang et al. (1998), which reduces the ODA data

to their gift equivalent. The two data sets are highly correlated, and we find that ODA data

normally give more significant results.11

Figure 2 shows the distribution of the raw aid growth data for 156 LDCs (based on

Paldam 2005). The data are averaged to 4 years and cover 1008 observations (10

observations are outside the frame). The figure is a typical illustration of the zero-

correlation result.

The graph is divided in three areas by the two gray lines: A1, A2 and A3. To find positive

aid effectiveness, the conditional studies have to give areas A1 and A3 a separate

9The DERG (Development Economics Research Group) was/is organized by Finn Tarp, who holds a Danida Chairin development. It is/was financed by the Danida Research Fund; Hansen and Tarp are frequent Danidaconsultants; and the model was popularized by a special grant to Tarp and Hjertholm (2000).10The term institutional home is used here in the sense that the models are proposed by research financed by aspecific institution, which contributed to making the model known. Also, in both cases the model is broadlyconsistent with the thinking and policies of the institution. It has not, however, been officially adopted as the basisfor the policy of the institution.11The EDA data cover fewer years and countries than the ODA set, and therefore the competition between the newmodels came to take place on about 1/3 of the available data. See Doucouliagos and Paldam (2008) on the relativeeffect of the two data sets.


DOI: 10.1002/jid


explanation. Good Policy generates growth, so the Good Policy model explains most of A1

by a separate term, while aid squared explains most of A3 by a separate term. In both cases

one may hope to get a ‘nice’ coefficient to aid.

2.3 The Good Policy Model: Condition is the Good Policy Index12

The Good Policy model has two equations. (1) gives the Good Policy index, zit; as a

weighted sum of the budget balance, Bit; the inflation rate, pit; and trade openness, Tit, while

(2) is the aid effectiveness relation, where git is the real growth rate, and hit is the aid

share. The j controls are xjit, and uit is the residuals. Greek letters are the coefficientsestimated. The two indices are i for countries and t for time. In the original findings(Burnside and Dollar 2000), m is insignificant, while both d and v are positive andsignificant.13

zit ¼ l0 þ l1Bit þ l2pit þ l3Tit Good Policy Index14 (1)

git ¼ aþ mhit þ dzit þ v zithit þ g jitx0jit þ uit Aid Effectiveness Relation (2)

The Good Policy index, z, is scaled (estimated), so it is fairly symmetrical around zero

for the countries considered, and z is outcome oriented, so it is not surprising that d

becomes significant and positive. What is non-trivial is that the interacted variable zithitproduces a significantly positive coefficient, v. It means that aid to a country that pursues

good policies increases growth, which is already high due to the good policies. Aid to a

country with bad policies decreases growth, which is already low due to the bad policies.

Obviously, the policy implications are that aid should be given to countries pursuing good

policies only.

Thus, in the Good Policy model, the crucial coefficient is v to the interactive term.

However, the coefficient to aid, m, also matters, as it shifts the relation up or down: If m is

large and positive, aid may be preferable to no aid, even in countries with bad policies, and

reversely, if m is negative, no aid may be preferable even in countries with good policies.

Aid effectiveness thus depends upon both coefficients m and v.

One of the trust building features in the presentation of the model is that Burnside and

Dollar (from the start in 1996) did not hide that the model recommends that aid should be

redistributed away from the countries pursuing bad policies, where it harms, to countries

pursuing good policies, where it helps. They even calculate the welfare gain to the world

from such redistribution. This clearly involves a break from the poverty orientation towards

an efficiency orientation of aid, which is contrary to much of the rhetoric of aid. In World

Bank (1998) the argument is presented in a more diplomatic way.

12The six papers of the group advocating the model are: Burnside and Dollar (2000; 2004); Collier and Dehn(2001); Collier and Dollar (2002); Collier and Hoeffler (2004); and Svensson (1999). The group was working inthe same department of the World Bank. Svensson (1999) was at the same Division, so we include him in thegroup, though he rejected the aid-policy term.13Some studies estimatemore general specifications that include squared aid, and aid policy and squared aid policyinteractions, but the essence of the model is captured by equation 2.14The original estimates for the coefficients of (1) are: l0 ¼ 1:28; l1 ¼ 6:85; l2 ¼ �1:20 andl3 ¼ 2:16:


DOI: 10.1002/jid

Figure 3. Excess growth, f, due to aid in the Medicine model


2.4 The Medicine Model: Condition is Aid Itself15

This model needs one equation only as it uses aid itself as the condition:

git ¼ aþ mhit þ v h2it þ g jitx0jit þ uit ¼ aþ fðhitÞ þ g jitx

0jit þ uit (3)

The proponents of the model find that m > 0 and v< 0. The size of both m and v, are

important for the model. The quadratic curve fðhÞ ¼ mhþ v h2 ¼ hðmþ vhÞ shows theexcess growth due to aid. The f-curve is zero at h ¼ 0 and�m=v. The first and second

derivatives are f0 ¼ mþ 2vh and f00 ¼ 2v: The maximum ðhmax;fmaxÞ ¼ ð�m=ð2vÞ;�m2

�ð4vÞÞ: The f-curve is drawn on Figure 3.

We can study the welfare properties of the model by considering aid as a game between a

donor agency, D, which makes an offer of aid and a recipient country, R, which accepts or

rejects the offer. Assume that both know the f-curve. The two standard alternative

assumptions about D are:

(1) D is an ideal bureaucracy which maximises world welfare. There are some (small)

costs, e, in the donor country as well, and thus it offers Ae. If D only considers R’s

welfare, it offers A. In both cases R accepts. If aid is constrained, welfare is

maximized, when f0i is the same for all R’s. As f0 ¼ mþ 2vh, all hi should be the

same. Aid should be redistributed to make all aid shares the same. This is different

from the advice from the Good Policy model.

(2) D is a Niskanen-type bureaucracy. It wants to maximize its budget, h. Consider the

intervals for h: (i) From h¼ 0 to A, the welfare of both D and R rises. (ii) From h¼A

to B, D’s welfare increases, while R’s welfare decreases compared to the optimum,

but R still has a welfare gain, and thus accepts. (iii) From h > B, R loses and rejects

aid. Thus, B is D’s optimal point. Here D has captured all of R’s potential welfare

gain.

Thus, it is crucial where A and B are located. Hansen and Tarp (2000) and Jensen and

Paldam (2006), find that the best estimate of B is somewhere between 20 and 30 per cent for

h. If the EDA data are considered, it is closer to 20 per cent and for the ODA data closer to

30 per cent. This means that A is between 10 and 15 per cent. If we confront these values of

A and B with the picture on Figure 2, it is clear that aid shares are very different, and that

the majority is well below A, so a large welfare gain can be made by equalising aid shares.

15The four papers of the group advocating the model are: Dalgaard and Hansen (2001); Hansen and Tarp (2000;2001); and Dalgaard, Hansen and Tarp (2004).


DOI: 10.1002/jid

Table 1. Average aid shares to 45 African countries 1990/94, 1995/99, 2000/04

Aid shareNumber of

% of cases

From To observations Now Proposal

0% 10% 52 Below A 51% 27%

10% 15% 31 "#15% 20% 15 Between A and B 30% 24%

20% 30% 17 "#30% 17 Above B 19% 49%

132 100% 100%

Note: aid shares are ODA/GDP in per cent. Countries between the Sahara and South Africa.The arrows "# mean that these observations have been divided in two equal parts.


Most of the high aid shares are in Africa. Table 1 gives the level of aid to Africa, 1990–

2004. The observations are aggregated to 3 averages of 5 years each. Only three

observations are missing. It appears that approximately half the aid shares are below A and

half are above. There is even 19 per cent over B. Aid has increased considerably in 2005–

08. The column ‘proposal’ shows the effect of doubling aid to Africa, as proposed by e.g.

Bono and Jeffrey Sachs. According to the Medicine model this would be harmful for

Africa.

Thus, the two models have strong—and very different—policy implications if they are

true, so it is no wonder that they have been carefully analyzed in a total of 40 studies.

3 ARE THE SUBSTANTIAL TERMS OF THE MODELS ESTABLISHED?

In order to make sense of the many conflicting results of the AEL, we use the tools of meta-

analysis.16 As mentioned in the introduction, we look at two questions: (Q1) Do the

estimates of the effects converge to something we can term truth? (Q2) Can we identify the

main innovations causing/preventing the convergence?

Regarding (Q1) we want to know if the AEL has established the size of the two key

coefficients, m and v, of the two main models. (Q2) We want to know if we can explain the

observed variation in the estimates by methodological differences of the studies. The

present section considers (Q1), while Section 4 turns to (Q2).

We look for four types of methodological differences: (D1) Models, notably control

variables included. (D2) Estimation techniques, notably if the relation is controlled for

simultaneity. (D3) Path dependency, the twomodels are proposed and defended by separate

research groups. It is interesting how theymanage to get significantly different results. (D4)

The size of the data samples on which the models are estimated.17

3.1 The Data of the Meta-Analysis and the Methods of Analysis

The 40 studies listed in the Appendix provide the data of the meta-analysis. From the

studies we derive two datasets for each of the two models, which are the estimates of the

two substantial coefficients m and v:

16Appendix 1 in Doucouliagos and Paldam (2008) is a survey of the methods used.17Aid started in the early 1960s. It has accumulated with 100-150 observations annually since the mid 1970s.


DOI: 10.1002/jid

Figure 4. Funnel plot of aid and good policy interactive coefficients, v. Note: Two points fromSvensson (1999) and Rajan and Subramanian (2008) are outside the frame of the figure.


The best-set contains one regression chosen by the author(s) of each paper. In some

cases it is unclear what the authors prefer. In these cases we had to assess; but then the

candidates for the best-set are normally close to each other. The all-set of all 288

regressions reported for the Good Policy model and 147 for the Medicine model. This

increases the data available for tests, but it gives some dependence between data points. To

each of these estimates, we attach a vector of variables (as a check list) that characterizes

the methods by which the result is reached in the four dimensions (D1)–(D4).

To get a ‘feel’ for the data, consider Figure 4, which is a funnel plot of the 288 aid policy

interaction estimates.18 The funnel plot shows the association between the estimates of v

and its accuracy.19 The coefficients should converge toward the true result as sample size N

increases, with smaller studies showing greater variation, so the point scatter should look

funnel-like—the funnel-form on Figure 4 is weak.

There is a clear cluster of coefficients around the zero mark—especially for high Ns—

suggesting that the aid policy variable has a coefficient close to zero, as is shown in

Tables 2 and 3 below. There are, however, many positive as well as some negative

coefficients. To see if a policy effectiveness result has been established, we need to address

two problems.

First, should all or only some estimates be included? We explore the all-set (which

includes results relating to robustness checks) as well as the best-set. Second, should all

studies be treated equally? In the present section, we use the sample size to assign weights

to studies.20 As this literature is relatively young, many of the papers are still working

papers. We control for the publication status of papers in Section 4.

18All 28 studies have economic growth (as a percentage) as the dependent variable. They all use a similar measureof aid (as a percent of GDP). Hence, the estimates are directly comparable across studies.19Accuracy here is proxied by sample size. The pattern is similar if precision (the inverse of the standard error) isused.20A larger sample should give more accurate estimates and, hence, be assigned a larger weight (see Hunter andSchmidt 2004). Alternative weights are the inverse standard error, the number of citations received, or the impactfactor of the journal where the study was published. These all produce qualitatively similar results.


DOI: 10.1002/jid

Table 2. Sign counts of the growth effects of aid, m, and aid times policy, v, in the two models

Good Policy model All-set, N¼ 288 Best-set, N¼ 28

Positive Negative Positive Negative

Signif. Not Not Signif. Signif. Not Not Signif.

Aid, m 64 89 97 12 8 6 9 0

Aid times policy, v 98 107 69 14 7 7 12 2

Medicine model All-set, N¼ 147 Best-set, N¼ 22

Positive Negative Positive Negative

Signif. Not Not Signif. Signif. Not Not Signif.

Aid, m 73 31 16 3 13 4 4 1

Aid squared, v 0 14 33 100 0 3 5 14

Signif. is statistically significant at the 10 per cent level, while Not is statistically insignificant.


3.2 Have the Coefficients m and v of the Good Policy

Model been Established?

The studies of the Good Policy model have a best-set of 28 observations and an all-set of

288 observations.21 The key coefficient in the model is v on the aid-policy term and m on

aid. Table 2 gives some descriptive statistics of the estimates published.

The top half of Table 2 reports the distribution of the Good Policy model results for both

coefficients. It is telling that of the 28 studies only (7/28¼) 25 per cent found a positive and

statistically significant aid-policy interaction.

Table 3 reports basic meta-regression analysis tests. The meta-significance test (MST) is a

test for the existence of a genuine effect between two variables, using all the available

empirical evidence. A genuine effect will reveal itself through a positive and statistically

significant association between the natural logarithm of the absolute value of t-statistics on

the aid-policy interactions (the dependent variable) and the associated natural logarithm of

the degrees of freedom, df, (the key explanatory variable in the MST).22 If aid-policy

interactions exist, then larger studies should have larger t-statistics. The MST results for both

datasets show that while the coefficient on lndfi is positive in three cases, it is not statistically

meta-significant (all p-values> 0.10). That is, taking all the available information there is no

evidence of a genuine effect between aid-policy interactions and economic growth.

The funnel asymmetry test (FAT) is a test for publication bias in a given literature.

Publication bias is present if smaller studies (with larger standard errors) report larger aid-

policy coefficients. For the all-set, the p-value<0.10, there is clear evidence of publication

bias23—smaller studies report larger coefficients and hence larger t-statistics. The sign on

the constant in the FAT indicates the direction of publication bias. Table 3 shows that the

21Some studies report only a single estimate, e.g. Lensink and White (2001) and Svensson (1999), but the averagestudy reports 10 regressions.22See Card and Kreuger (1995) and Stanley (2001, 2005) on the MST, and Egger et al. (1997) and Stanley (2005)on the FAT. Stata 10 was used for all the meta-regression analysis.23This finding is not unique. Except for a couple of investigations, the majority of studies have detected publicationbias in empirical economics research (see, for example, Card and Krueger 1995, Ashenfelter et al. 1999, Gorg andStrobl 2001, Roberts and Stanley 2005, and Monkerjee 2006).


DOI: 10.1002/jid

Table 3. Meta-significance and funnel asymmetry tests, aid-growth conditionality effects

Dependent Variable (1) (2) (3) (4)

MST MST FAT FAT

lnjt-statisticj t-statistic

All-set Best-set All-set Best-set

v, Good Policy model

Constant 0.32 (0.39) �0.53 (�0.34) 0.82 (4.87) 0.35 (0.81)

ln(df) �0.06 (�0.38) 0.07 (0.23) – –

1/SE – – 0.005 (0.68) �0.009 (�0.58)

R2 0.00 0.00 0.01 0.00

N 288 28 283 28

Average Y �0.01 �0.18 þ0.92 þ0.39

v, Medicine model

Constant �0.11 (�0.10) �0.02 (�0.02) �1.86 (�8.81) �1.37 (�3.22)

ln(df) 0.10 (0.54) 0.08 (0.40) – –

1/SE – – 0.0001 (0.13) �0.0003 (�1.09)

R2 0.005 0.00 0.00 0.03

N 147 22 144 22

Average Y þ0.44 þ0.42 �1.82 �1.78

Explanation: If aid interaction terms have an effect on growth, ln(df) in the MST should have a positive andstatistically significant coefficient. This fails for both models. If the literature is free of publication bias, theconstant in the FAT should not be statistically significant. It is significant in 3 cases. The 1/SE term is a measure ofthe existence of a genuine empirical effect, corrected for publication bias. It is zero in all cases.Notes: Bolded estimates are statistically significant, at least at the 10 per cent level. t-statistics in brackets, usingrobust standard errors: for the all-set, these also take into account the clustering of estimates within studies.Average Y reports the average value of the dependent variable (natural logarithm of t-statistic for MST and t-statistic for FAT). Some observations are lost due to missing data in some cases.


bias is in favor of reporting positive aid-policy interaction terms. FAT also offers a second

test for the existence of a genuine empirical effect. This would be revealed through a

statistically significant coefficient on 1/SE. Aid-policy interactions fail this test as well.

3.3 Is There an Aid Squared Effect?

The Medicine model has been analyzed in 22 papers, and a total of 147 regressions have

been presented with an aid squared term. A large fraction of the regressions are in papers

proposing the model, Hadjimichael et al. (1995), Hansen and Tarp (2000; 2001), Dalgaard

and Hansen (2001), Lensink and White (2001), and Dalgaard, Hansen and Tarp (2004).

The distribution of the results is reported in the lower half of Table 2. Taking the best-set

of results from all 22 studies, the weighted average aid squared coefficient is�0.04, and the

associated partial correlation is �0.12.24 Taking all 147 estimates from the 22 studies, the

weighted average aid squared coefficient is�0.11, and the associated partial correlation is

�0.12. Table 3 shows that theMST results for all the aid squared estimates had a coefficient

on the natural logarithm of degrees of freedom of + 0.10, and for the best-set it is + 0.08,

and neither is positive and statistically significantly different from zero. The lack of an

24There is a high degree of skewness in the reported coefficients from this part of the literature. For example, whilethe unweighted average aid squared coefficient is �0.07, the median is �0.001.


DOI: 10.1002/jid

Figure 5. The relation between the coefficients to aid and aid times policy


aid�aid interaction is confirmed also by the 1/SE coefficient in the FAT. Hence, there is no

evidence of a genuine association between aid squared and economic growth. The constant

in the FAT results indicates that the pattern of reporting of aid x aid interactions is biased in

favor of negative coefficients. There is no effect remaining once the data is corrected for

publication selection.

3.4 The Debate between the Two Model Groups

A comparison of the findings of the Danida and World Bank groups confirms our

expectations. The World Bank group reports a positive value for the aid-policy interactive

term, v, (+ 0.08) and a negative value for m (�0.14). In contrast, the Danida group studies

find a negative value forv (�0.02) and a large positive value form (+ 0.35). The two groups

of researchers produce quantitatively and qualitatively different results: affiliation matters.

Figure 5 is a simple scatter diagram of the coefficients on the aid variable and the

coefficients on the aid-policy interactive terms for the Danida and World Bank group of

studies. It shows a clear negative association. Studies which report higher coefficients on

the aid-policy interactive term tend to report lower coefficients on the aid term. As noted

earlier, even if the aid-policy interactive term is positive and statistically significant, it is

still possible for aid to have a positive impact regardless of policy if the coefficient on aid is

large enough.

3.5 Institutional Conditions: The Future?

The 10 papers in the residual group are all relatively new proposals. They have not been

independently replicated. They are promising, but because they are a small group they

cannot be submitted to a formal meta-analysis.

Two papers condition for democracy (Svensson 1999; and Kosack 2002). Both suggest

that aid works better in democracies, but are otherwise different. The main thrust of

Kosack’s paper is to replace economic growth with growth of the human development

index as the dependent variable, but he also reports results using growth, confirming that

the two welfare measures are highly correlated.


DOI: 10.1002/jid


Two studies by Chauvet and Guillaumont (2001; 2003) and the study by Chauvet (2005)

condition for various measures for political instability and external vulnerability, which

attempt to catch institutional stability. This appears logical as successful projects do need

time for implementation, and hence some kind of institutional stability. Two related

studies condition for quality of institutions, Collier and Dehn (2001) and Collier and

Dollar (2004), but in different ways. Perhaps‘‘quality’’ and ‘‘stability’’ of institutions is the

same factor for aid effectiveness. It is a main problem that we need simple and clear

measures for this factor, but the proxies tried suggest that it is important for aid

effectiveness.

One study conditions for trade openness (Teboul and Moustier 2001). The logic here is

more indirect, and one gets the impression that trade openness is a proxy for a broader

concept. The broadest such concept is perhaps economic freedom as is tried in Brumm

(2003). Although it may not matter, it appears that the statistical methods used by Brumm

are too far from the state of the art in the field.

Finally, two studies link aid with GDP (Bowen 1995 and Svensson 1999). They suggest

that aid works better in more developed countries. Many studies that concentrate on a

region or have regional dummies give evidence pointing in the same direction.

4 ACCOUNTING FOR DIFFERENCES IN RESULTS

FOR THE TWO MODELS

The previous analysis shows that priors influence results, but we also want to study the

methodological differences producing the results. MRA (meta-regression analysis) can be

used for that purpose (Stanley 2001 and 2005).

4.1 The Data for the MRA

The dependent variable is a binary variable taking the value of 1 if the study reports a

statistically significant positive aid-policy coefficient and otherwise 0.25 Probit meta-

regressions can be used to identify the determinants of conditional aid effectiveness. The

number of observations is limited, so we only use the most important explanatory

variables,26 which are defined in Table 4.

The aim of our probit MRA is to identify the characteristics of studies that influence the

reported results. We are interested in exploring whether an author’s association with the aid

business results in qualitatively different results on the aid-policy variable. We include the

Danida andWorldBk dummies in order to explore the effect of institutional affiliation. We

add a control variable for working papers since working papers have not yet passed the

referee process and, hence, may have a lower quality (i.e.WorPap is our binary measure of

research quality).27

25Normally, the dependent variable in an MRA is regression coefficients, t-statistics, elasticities, or partialcorrelations, estimated using a linear regression model (see, for example, Doucouliagos and Paldam 2008). Ourfocus here, however, is on the factors that result in the reporting of a positive and statistically significant aid-policyinteraction. Hence, we use a probit model.26See Doucouliagos and Paldam (2008) for the impact of specification on the broader aid-growth literature.27Working papers may also use newer techniques and more recent data and, hence, may report qualitativelydifferent results.


DOI: 10.1002/jid

Table 4. Means and standard deviations of MRA variables. Good Policy model, All-set

Variable BD means binary dummy. It is 1 if conditionfulfilled, otherwise 0

Mean St dev

Dependent BD if study reports significantly positive coefficients (v) 0.34 0.47

DevJour BD if published in development journal 0.20 0.40

AidBus BD if author(s) employed/affiliated with aid agency 0.57 0.50

WorldBk BD if paper from World Bank group 0.22 0.41

Danida BD if paper from Danida group 0.08 0.27

NrCountries Number of countries included in the sample 55 14

NrYears Number of years covered in the analysis 26 4

EDA BD if paper uses EDA measure of assistance 0.53 0.50

Subsample BD if estimate relates to sub-sample of countries 0.18 0.38

Polmeasure BD if a Burnside-Dollar type measure of policy used 0.79 0.41

WorPap BD if the research has yet to be published in journal 0.50 0.50

Endo BD if the aid was treated as an endogenous variable 0.28 0.45

Reproduce BD if estimate is an attempt to replicate results 0.08 0.28

Fixedeffects BD if fixed effects estimator used 0.17 0.37

Instability BD if paper controlled for political instability 0.82 0.38

Ethnic BD if controlled for ethnic fractionalisation 0.75 0.43

Finmarkets BD if controlled for financial markets development 0.82 0.38

Institutions BD if controlled for quality of institutions 0.91 0.29

Region BD if paper controlled for regional effects 0.86 0.34

Aidsqr BD if included aid�aid term 0.22 0.41

AidSqr�Policy BD if included aid�aid�policy term 0.14 0.35


Five variables are included to capture the impact of data differences: NrCountries,

NrYears, EDA, Subsample, and Polmeasure. If the aid-policy conditionality is robust, we

expect a positive association between the number of countries included in a study and the

study results.28 Similarly, we include the number of years of data from each study.

Subsample controls for different sub-sets of data, while Polmeasure controls for

differences, if any, in the measure of policy. We include theDevJour variable to see if there

are differences in results across the types of journals. Reproduce controls for estimates

made by different authors with the sole aim of reproducing another researcher’s results.

Two variables relate to estimation techniques: Endo is included in order to see if accounting

for the endogeneity of aid changes the results, and fixed effects tests if the use of the fixed

effects estimator makes a difference. Nine variables are included to capture the impact of

specification differences: Region, Instability, Ethnic, Finmarkets, Institutions, Aidsqr and

Aidsqr�Policy.

4.2 Results for the Good Policy Model

The MRA results reported in Table 5 use the all-set of 288 estimates. The observations

included in the all-set are not all statistically independent. Hence, we use two sets of

clustered data analysis to account for data dependence (Hox, 2002). The first set assigns to

the same cluster all estimates from the same study. This involves 28 clusters and accounts

for any potential within study dependence. The second set assigns to the same cluster all

28Several studies show that the choice of countries can influence study results, see e.g. Burnside and Dollar (2000)and Jensen and Paldam (2004).


DOI: 10.1002/jid

Table 5. Meta-probit regression analysis, Good Policy model, all-set

Variable (1) (2) (3) (4) (5) (6) (7)

Constant �0.73 �0.74 5.40 �0.51 �1.17 7.03 6.08

(5.9, 4.3, 4.2) (7.6, 5.9, 5.1) (4.5, 3.9, 4.4) (2.7, 2.6, 2.4) (3.4, 1.7, 1.6) (3.5, 2.9, 2.7) (4.6, 3.8, 3.9)

AidBus 0.52 – – – – – –

(3.3, 1.6, 1.3)

WorldBk – 1.29 1.18 1.09 0.92 0.70 0.77

(6.6, 2.8, 8.5) (5.6, 3.3, 10.4) (4.5, 2.4, 7.1) (3.9, 2.6, 4.7) (2.2, 2.4, 3.4) (2.8, 2.9, 4.7)

Danida – 0.13 �0.06 �0.39 0.02 �1.32 �1.07

(0.4, 0.7, 0.9) (0.2, 0.2, 0.3) (1.0, 0.9, 0.9) (0.1, 0.1, 0.1) (2.4, 2.1, 2.0) (2.9, 3.8, 5.3)

NrCountries – – �0.06 – – �0.10 �0.10

(4.8, 3.7, 4.1) (4.6, 3.9, 4.7) (4.9, 3.9, 4.3)

NrYears – – �0.08 – – �0.04 –

(3.0, 2.3, 2.1) (1.2, 1.1, 1.0)

EDA – – �0.25 – – �0.12 –

(1.3, 0.9, 0.7) (0.4, 0.3, 0.3)

Subsample – – �1.10 – – �1.82 �1.81

(3.6, 3.1, 3.1) (4.5, 4.6, 4.5) (4.6, 4.2, 3.9)

Polmeasure – – �0.48 – – 0.11 –

(2.1, 1.3, 1.0) (0.3, 0.3, 0.3)

WorPap – – – �0.21 – �0.55 �0.50

(1.0, 1.0, 1.0) (2.1, 2.6, 3.0) (2.0, 2.3, 2.4)

DevJour – – – 0.05

– – 0.09

(0.2, 0.1, 0.1) (0.2, 0.2, 0.2)

Reproduce – – – 1.05 – 1.27 1.27

(3.3, 3.4, 4.5) (3.4, 4.1, 4.4) (3.9, 4.0, 5.6)

Fixedeffects – – – �0.43 – �0.99 �1.28

(1.4, 1.5, 1.2) (1.9, 2.1, 2.1) (2.8, 3.4, 3.9)

Endo – – – �0.41 – �0.26 –

(1.7, 1.7, 1.8) (0.9, 1.1, 1.1)

Instability – – – – 0.09 �0.18 –

(0.4, 0.3, 0.3) (0.5, 0.4, 0.4)

Ethnic – – – – 0.07 �0.66 �0.82

(0.2, 0.1, 0.1) (1.2, 1.4, 1.4) (3.1, 2.8, 2.6)

Finmarkets – – – – �0.65 �0.26 –

(1.4, 1.1, 1.0) (0.5, 0.4, 0.4)

Institutions – – – – 0.83 0.45 –

(1.9, 1.8, 1.7) (0.7, 0.6, 0.6)

Region – – – – 0.10 �0.05 –

(0.3, 0.2, 0.2) (0.1, 0.2, 0.2)

Aidsqr – – – – 0.49 0.75 0.76

(2.2, 1.9, 1.3) (2.4, 2.8, 1.8) (3.2, 3.4, 2.4)

Aidsqr�Policy – – – – �0.07 �0.13 –

(0.3, 0.3, 0.3) (0.4, 0.6, 0.7)

N 288 288 288 288 288 288 288

Pseudo R2 0.03 0.13 0.23 0.21 0.18 0.38 0.36

Wald (affilia.) 2.49 7.60 12.52 7.39 6.67 10.75 32.10

Aid business ME 0.19 (1.51) – – – – – –

World Bank ME – 0.48 (3.02) 0.43 (3.30) 0.41 (2.44) 0.35 (2.62) 0.24 (2.27) 0.27 (2.65)

Danida ME – 0.05 (0.65) �0.02 (�0.22) �0.12 (�1.03) 0.01 (0.09) �0.25 (�4.06) �0.22 (�5.17)

NrCountries ME – – �0.02 (�3.54) – – �0.03 (�4.63) �0.03 (�4.78)

Notes: The dependent variable is a binary variable reflecting whether the aid-policy interaction term of the studyhas a positive and statistically significant impact on economic growth. Figures in brackets are absolute z-scoresusing: robust standard errors; z-scores using within study clustered data analysis; and z-scores using within authorclustered data analysis, respectively. ME is marginal effect. Bolded estimates are statistically significant in allcases, at least at the 10 per cent level. Wald (affilia.) is a Wald test for the joint significance of the institutionalaffiliation variables, incorporating the clustering of observations.


DOI: 10.1002/jid



estimates from the same author or groups of authors. This involves 18 distinct clusters and

accounts for any potential within author (between study) dependence.29

We test the sensitivity of the results by running several regressions. The row labeled

Wald (affilia.) presents the test results for Wald tests on the institutional affiliation variables

(AidBus or WorldBk and Danida jointly). The marginal effects reported in the last three

rows show the effect of affiliation on the probability that a study reports a positive and

statistically significant aid-policy interaction.

Table 5, column 1 presents the regression with only the aggregate AidBus variable. This is

separated intoWorldBk andDanida in column 2.30 Data differences are introduced in column

3. Estimation and publication outlet differences are added in column 4. Column 5 uses the

specification dummies. The general model (with all variables included) is presented in column

6. Column 7 presents the results of the specific model after several statistically insignificant

variables were eliminated (we sequentially removed any variable that had a t-statistic of less

than one). A Wald test confirms the joint insignificance of the eliminated variables.

Column 7 is our preferred set of results from Table 5. TheWald (affilia) tests for columns

2–7 show that WorldBk and Danida are jointly statistically significant. The WorldBk

dummy is positive and statistically significant. Studies by authors associated with the

World Bank are more likely to report positive aid-policy effectiveness results. The Danida

variable has the expected negative sign, and the associated marginal effects are statistically

significant in the results presented in columns 6 and 7. Affiliation matters even after

controlling for obvious specification and modeling differences.

NrCountries is statistically significant: the more countries included in a study, the less

likely is a positive and statistically significant aid-policy interaction. Endogeneity is not

important in determining a positive aid-policy conditionality result. However relative to

OLS, the use of fixed effects results in a lower probability of reporting favorable aid-policy

effects. The coefficient on Reproduce confirms that researchers are able, in general, to

replicate prior findings—dependent replication is not a problem in the AEL. Interestingly,

the inclusion of an aid squared increases, on average, the significance of the aid-policy

term, while ethnic fractionalization decreases it.

The probit meta-regression analysis for the best-set confirms that Danida has a strong

negative and WorldBk has a strong positive effect on the probability of reporting positive

aid-policy interactive terms. Number of countries also has a negative coefficient.31

The meta-regression analysis was repeated for the data associated with the Medicine

Model (these results are available from the authors).32 For the best set of only 22

observations, Danida is a perfect predictor variable, and it is the only variable with a

significant marginal effect of + 0.45 (z-statistic ¼ + 3.70).

5 SUMMARY AND CONCLUSIONS

The aim of this paper was to explore the family of conditional models of the aid

effectiveness literature by the tools of meta-analysis. Here effectiveness depends on a

condition: If it is fulfilled, aid helps, and if it is not, aid harms.

29Note that Table 5 also reports robust standard errors.30WorldBk andDanida do not exhaust the Aid Business category. If a third variable – other Aid Business – is addedto any of the regressions, it is never statistically significant.31The marginal effects show that the probability that a positive and statistically significant effect is reported is 0.76when the researcher is associated with the World Bank.32The dependent variable here is whether the aid�aid term is negative and statistically significant.


DOI: 10.1002/jid


The most researched condition is ‘Good Policy’ (defined above). The number of studies

is already large enough to permit clear conclusions on the two questions asked:

(Q1) Is the impact of aid on growth moderated by policy? The aggregate coefficient

to the interaction between foreign aid and policy proves to be very close to zero.

Good policies help increase growth, but they do not appear to influence the marginal

effectiveness of aid. (Q2) Are the reported estimates systematically influenced

by study characteristics? We established that the author’s institutional affiliation

does influence reported results, as do sample size, estimation technique and model

specification.

The success of the Burnside and Dollar and World Bank reports was based on the

evidence available at that time, but subsequent analysis has shown that their conclusions

were premature. This proves Hunter and Schmidt’s (2004, xxvii) statement that: ‘Scientists

have known for centuries that a single study will not resolve a major issue. . . . Thus, thefoundation of science is the cumulation of knowledge from the results of many studies’.

The second most researched condition is aid itself, where aid works as medicine, which

has an optimal dose. Here we asked the same questions. Our results suggest that the

aggregate coefficient to foreign aid squared is not statistically significantly different from

zero.

Our conclusions are based on an assessment of the findings of two conditioning variables

and total foreign aid. There is also a small but growing literature that explores interactions

with individual components of foreign aid. Also, the literature contains a whole set of new

conditional variables that have been proposed and tested once or twice. These variables are

potentially quite promising and suggest that the field is open to future research.

ACKNOWLEDGEMENTS

Our cooperation was supported by the Aarhus University Research Foundation. Pia

Wichmann Christensen has been a very competent research assistant. This paper benefited

from comments from Peter Sandholt Jensen and T.D. Stanley and from discussions at

seminars at our universities, Hendrix College, the University of Queensland, the Australian

National University, and the Kiel Institute of World Economics, as well as from

presentations at the Public Choice Society 2006 Meetings in New Orleans and Turku.

We are grateful to the referees for constructive advice.

REFERENCES

Ashenfelter O, Harmon C, Oosterbeek H. 1999. A review of estimates of the schooling/earnings

relationship, with tests for publication bias. Labour Economics 6: 453–470.

Card D, Krueger AB. 1995. Time-series minimum-wage studies: a meta-analysis. American

Economic Review 85: 238–243.

Chang CC, Fernandez-Arias E, Serven L. 1998. Measuring aid flows: A new approach. World Bank.

URL: <http://www.worldbank.org/research/growth/ddaid.htm>

Christensen PW, Doucouliagos H, Paldam M. (2nd ed.) 2009. Master list of the AEL: The aid

effectiveness literature. Available from http://www.martin.paldam.dk

Collier P, Dollar D. 2004. Development effectiveness: what have we learnt? Economic Journal 114:

244–271.


DOI: 10.1002/jid


Doucouliagos H, PaldamM. 2006. Aid effectiveness on accumulation. A meta study. Kyklos 59: 227–

254.

Doucouliagos H, Paldam M. 2008. Aid effectiveness on growth. a meta study. European Journal of

Political Economy 24: 1–24.

Doucouliagos H, Paldam M. 2009. The aid effectiveness literature. The sad results of 40 years of

research. Journal of Economic Surveys. Forthcoming.

Egger M, Smith GD, Scheider M, Minder C. 1997. Bias in meta-analysis detected by a simple,

graphical test. British Medical Journal 316: 629–634.

Gorg H, Strobl E. 2001. Multinational companies and productivity: a meta-analysis. The Economic

Journal 111(475): 723–739.

Herbertsson TT, Paldam M. 2007. Does development aid help poor countries catch up? An analysis

using the basic relations. Nationaløkonomisk Tidsskrift/Danish Journal of Economics 145: 188–

214.

Hox JJ. 2002. Multilevel Analysis: Techniques and Application. Mahwah, NJ; Lawrence Erlbaum.

Hunter John, Schmidt Frank. 2004. Methods of Meta-Analysis: Correcting Error and Bias in

Research Findings. Sage: London.

Mosley P. 1987. Overseas Aid: Its Defence and Reform. John Spiers: Brighton, Sussex.

Paldam M. 2005. A look at the raw data for aid and growth. Posted at <http://www.martin.

paldam.dk>

Rajan RG, Subramanian A. 2008. Aid and Growth: What Does the Cross-Country Evidence Really

Show? (first IMF WP No. 05/127). Review of Economics and Statistics 90: 643–665.

Roberts CJ, Stanley TD. (eds). 2005. Meta-Regression Analysis: Issues of Publication Bias in

Economics. Basil Blackwell: London.

Stanley TD. 2001. Wheat from chaff: meta-analysis as quantitative literature review. The Journal of

Economic Perspectives 15: 131–150.

Stanley TD. 2005. Beyond publication bias. Journal of Economic Surveys 19: 309–345.

Tarp F, Hjertholm P. eds., 2000. Foreign aid and Development. Lessons Learnt and Directions for the

Future. Routledge Studies in Development Economics: London.

World Bank. 1998. Assessing Aid: What Works, What Doesn’t, and why. A Policy Research Report.

Oxford UP: New York.

APPENDIX: STUDIES USED IN THE META-ANALYSIS

Only papers in English available till 1, January 2007 are included. Christensen et al. (2009)

is our master list of the whole of the AEL.

� A

Co

ntipin, J.-E., Mavrotas, G., 2006. On the empirics of aid and growth. A fresh look. UN-

WIDER WP 2006/05

� B
owen, J.L., 1995. Foreign aid and economic growth: an empirical analysis. Geo-
graphical Analysis 27, 249–61. Estimates also in Bowen, J.L., 1998. Foreign aid and

economic growth: A theoretical and empirical investigation. Ashgate, Aldershot, UK

� B
rumm, H.J., 2003. Aid, policies and growth: Bauer was right. Cato Journal 23, 167–74
� B
urnside, C., Dollar, D., 2000. Aid, policies and growth. American Economic Review 90,
847–68 (Working paper available as World Bank WP since 1996)

� B
urnside, C., Dollar, D., 2004. Aid, policies and growth: reply. American Economic
Review 94, 781–84 (reply to Easterly, Levine and Roodman, 2004)

pyright # 2009 John Wiley & Sons, Ltd. J. Int. Dev. 22, 391–410 (2010)

DOI: 10.1002/jid


� C

Co

hauvet, L., Guillaumont, P., 2004. Aid and growth revisited: Policy, economic

vulnerability and political instability. Pp 95–109 in Tungodden, B., Stern, N., Kolstad,

I., eds, 2004. Toward Pro-Poor Policies - Aid, Institutions and Globalisation. World

Bank/Oxford UP

� C
hauvet, L., 2005. Can foreign aid dampen external political shocks? EPCS-2005
(April)

� C
lemens, M., Radelet, S., Bhavnani, R., 2004. Counting chickens when they hatch: the
short-term effect of aid on growth. Center for Global Development WP 44

� C
ollier, P., Dehn, J., 2001. Aid, shocks, and growth. WP 2688 World Bank Policy
Research

� C
ollier, P., Dollar, D., 2002. Aid allocation and poverty reduction. European Economic
Review 46, 1475–1500

� C
ollier, P., Hoeffler, A., 2004. Aid, policy and growth in post-conflict societies.
European Economic Review 48, 1125–45

� C
ordella, T., Dell’Ariccia, G., 2003. Budget support versus project aid. IMF WP/03/88
� C
ungu, A., Swinnen, J., 2003. The impact of aid on economic growth in transition
economies: An empirical study. LICOS Discussion Papers, Leuven

� D
algaard, C.-J., Hansen, H., 2001. On aid, growth and good policies. Journal of
Development Studies 37, 17–41

� D
algaard, C.-J., Hansen, H., Tarp, F., 2004. On the empirics of foreign aid and growth.
Economic Journal 114, 191–216

� D
ayton-Johnson, J., Hoddinott, J., 2003. Aid, policies and growth, redux. WP Dalhousie
Univ.

� D
enkabe, P., 2004. Policy, aid and growth: A threshold hypothesis. Journal of African
Finance and Economic Development 6, 1–21 (WP version used)

� D
urbarry, R., Gemmell, N., Greenaway, D., 1998. New evidence on the impact of foreign
aid on economic growth. Credit WP Univ. of Nottingham

� E
asterly, W., Levine, R., Roodman, D., 2004. Aid, policies, and growth: Comment.
American Economic Review 94, 774–80 (Comment to Burnside and Dollar, 2001)

� G
omanee, K., Girma, S., Morrissey, O., 2002. Aid and growth: accounting for the
transmission mechanisms in Sub-Sahara Africa. Credit WP Univ. of Nottingham

� G
uillaumont, P., Chauvet, L., 2001. Aid and performance: a reassessment. Journal of
Development Studies 37, 66–92

� H
adjimichael, M.T., Ghura, D., Muhleisen, M., Nord, R., Ucer, E.M., 1995. Sub-
Saharan Africa: growth, savings, and investment, 1986–93. IMF Occasional Paper, No.

118

� H
ansen, H., Tarp, F., 2000. Aid effectiveness disputed. Journal of International
Development 12, 375–398. Also pp 103–128 in Tarp, F., Hjertholm, P., eds. 2000.

Foreign aid and development. Lessons learnt and directions for the future. Routledge

Studies in Development Economics: London

� H
ansen, H., Tarp, F., 2001. Aid and growth regressions. Journal of Development
Economics 64, 547–70

� H
udson, J., Mosley, P., 2001. Aid policies and growth: in search of the Holy Grail.
Journal of International Development 13, 1023–38

� I
slam, M.A., 2003. Political regimes and the effect of foreign aid on economic growth.
The Journal of Developing Areas 37, 35–53

� J
ensen, P.S., Paldam, M., 2006. Can the two new aid-growth models be replicated?
Public Choice 127, 147–75


DOI: 10.1002/jid


� K

Co

osack, S., 2003. Effective aid: how democracy allows development aid to improve the

quality of life. World development 31, 1–22

� L
ensink, R., White, H., 2001. Are there negative returns to aid? Journal of Development
Studies 37, 42–65

� L
u, S., Ram, R., 2001. Foreign Aid, government policies, and economic growth: further
evidence from cross-country panel data for 1970–1993. Economia Internazionale/

International Economics 54, 15–29

� M
oreira, S.B., 2003. Evaluating the impact of foreign aid on economic growth: a cross-
country study (1970–1998). WP for 15th Annual Meeting on Socio-Economics, Aix-en-

Provence, France

� M
urphy, R.G., Tresp, N.G., 2006. Government policy and the effectiveness of foreign
aid. WP 399. Economics Department, Boston College

� O
vaska, T., 2003. The failure of development aid. Cato Journal 23, 175–88
� P
avlov, V., Sugden, C., 2006. Aid and growth in the Pacific Islands. Asia-Pacific
Economic Literature 20, 38–55

� R
ajan, R.G., Subramanian, A., 2008. Aid and growth: what does the cross-country
evidence really show? (first IMF WP/05/127). Review of Economics and Statistics 90,

643–65

� R
am, R., 2004. Recipient country’s ‘policies’ and the effect of foreign aid on economic
growth in developing countries: Additional evidence. Journal of International Devel-

opment 16, 201–11

� R
oodman, D., 2004. The anarchy of numbers: Aid, development and cross-country
empirics. WP 32 Center for Global Development

� S
hukralla, E.K., 2004. Aid, incentives, polices, and growth: Theory and a new look at the
empirics. WP Western Michigan Univ.

� S
vensson, J., 1999. Aid, growth and democracy. Economics and Politics 11, 275–97
� T
ebouel, R., Moustier, E., 2001. Foreign aid and economic growth: the case of the
countries south of the Mediterranean. Applied Economics Letters 8, 187–90


DOI: 10.1002/jid

Documents

Conditional aid effectiveness: A meta-study