Doing Well by Doing Good: The Impact of Foreign Aid on ...horiuchi.org/www/Publications_files/100.00013036.pdf · Doing Well by Doing Good: The Impact of Foreign Aid on Foreign Public

Quarterly Journal of Political Science, 2014, 9: 87–114

Doing Well by Doing Good:The Impact of Foreign Aidon Foreign Public Opinion∗

Benjamin E. Goldsmith1, Yusaku Horiuchi2 and Terence Wood3

1Department of Government and International Relations, The Universityof Sydney, New South Wales 2006, Australia; [email protected] of Government, Dartmouth College, Hanover, NH 03755,USA; [email protected] of International, Political and Strategic Studies, The AustralianNational University, Acton ACT 0200, Australia; [email protected]

ABSTRACT

Does foreign aid extended by one country improve that country’s imageamong populations of recipient countries? Using a multinational survey,we show that a United States aid program targeted to address HIV andAIDS substantially improves perceptions of the U.S. Our identificationstrategy for causal inference is to use instrumental variables measuring

* Earlier drafts were presented at the 2011 Workshop on the Frontiers of Statistical Analysisand Formal Theory of Political Science (Tokyo, Japan, January 6–7, 2011), a workshop onempirical political analysis at the Australia National University (Canberra, Australia, May 10,2012), the Second Annual General Conference of the European Political Science Association(Berlin, Germany, June 21–23, 2012), the Fifth Oceanic Conference on International Studies

Online Appendix available from:http://dx.doi.org/10.1561/100.00013036 appSupplementary Material available from:http://dx.doi.org/10.1561/100.00013036 suppMS submitted 23 February 2013 ; final version received 9 December 2013ISSN 1554-0626; DOI 10.1561/100.00013036c© 2014 B. E. Goldsmith, Y. Horiuchi, and T. Wood

88 Goldsmith et al.

the magnitude of the HIV/AIDS problem in aid recipient countries. Ourfinding implies that in addition to its potential humanitarian benefits,foreign aid that is targeted, sustained, effective, and visible can serveas an important strategic goal for those countries that give it: fosteringpositive perceptions among foreign publics. By doing good, a countrycan do well.

Keywords: Foreign aid; U.S. foreign policy; HIV; AIDS; PEPFAR;public opinion; instrumental variables

Competition between major powers such as the United States (U.S.) andChina for favorable perceptions in global public opinion is increasingly evi-dent today and likely to be a pivotal feature of the emerging internationalorder (Friedberg, 2010; Gill and Huang, 2006; Kurlantzick, 2007; Nye, 2004;Shambaugh, 2008). These countries are eager to develop positive imagesof themselves among foreign publics, because such images are consideredimportant for achieving a range of objectives in foreign relations (Busta-mante and Sweig, 2008; Goldsmith and Horiuchi, 2012; Nye, 2004). Amongpotential tools states may use to win hearts and minds abroad, foreign aid isoften claimed to be effective (American Political Science Association, 2009).1

In the context of this intensifying competition for positive images amongglobal publics, in 2003, under the Bush Administration, the U.S. establisheda foreign aid program called the President’s Emergency Plan for AIDS Relief(PEPFAR), which has been provided to more than 80 developing coun-tries. Its declared objectives include stopping the spread of HIV and AIDS,

(Sydney, Australia, July 18–20, 2012), and in a seminar at Dartmouth College (Hanover, USA,October 12, 2012). We thank Geoff Brennan, Steve Brooks, Alison Cumming Thom, LindaFowler, Michael Herron, Kosuke Imai, Mike Miller, Jo Spratt, Carrie Steel, Matt Winters,Ippei Yamamoto, and other participants of the above workshops, conferences, and seminarsfor useful comments. We also thank Gallup Inc. for providing us their data.

1 States may use other tools, such as public diplomacy, for the same reason (Goldsmith andHoriuchi, 2009). In this article, we use the term “foreign aid” (or simply “aid”) as a synonymfor Official Development Assistance as defined and measured by the Organisation for Eco-nomic Co-operation and Development’s Development Co-operation Directorate. Importantly,this definition does not include military aid.

Doing Well by Doing Good: The Impact of Foreign Aid on Foreign Public Opinion 89

supporting the treatment of people suffering from HIV and AIDS, and mit-igating the indirect consequences of the epidemic.2

As then Secretary of State Condoleezza Rice emphasized in her speech atthe release of the Fifth Annual Report to Congress on PEPFAR on January12, 2008, the U.S. government believes that PEPFAR has contributed notonly to addressing global health problems but also to improving the U.S.image among publics in recipient countries. Yet, while claims such as thisare common, they appear to be based on speculation or anecdotal evidence.Empirically, work on the impact of aid — and more specifically, PEPFAR —on public opinion is scarce. There is little available systematic evidencesuggesting that foreign aid improves donor image amongst populations ofcountries receiving the aid.

In this article, we provide compelling evidence that PEPFAR has, indeed,positively affected how publics in recipient countries regard the U.S. Weemphasize two key implications of this finding.

First, it suggests what types of aid, under what conditions, might be effec-tive in influencing foreign public opinion about the donor. Specifically, ourtheory is that foreign aid that is targeted, sustained, effective, and visibleis more likely to affect mass opinion. Recent studies suggest that aid canbe effective at encouraging economic growth in recipient countries, althoughsuch positive findings remain contested (for example, Doucouliagos and Pal-dam, 2013). Specifically, they suggest that aid promotes economic growth,under certain conditions, at the national (Arndt et al., 2010; Dalgaard andTarp, 2004; Wright, 2010) or global (Headey, 2008) level; and that the effectsof aid are observable over multi-year time periods (Headey, 2008; Minoiuand Reddy, 2010). While much of the existing empirical debate focuses onthe economic impacts of foreign aid, our refined theory of aid effectivenesssheds light on the comparatively under-investigated area of aid’s impact onpublic opinion in recipient countries. This has practical ramifications forinternational relations, because, as Goldsmith and Horiuchi (2012) show,changes in public opinion within Country B about Country A can influenceCountry B’s foreign policy behavior toward Country A.

Second, if foreign aid programs have such political ramifications, we sug-gest a possible virtuous circle or race-to-the-top dynamic for the emerging

2 Funding for PEPFAR was re-authorized in 2008 and it has continued to receive support underthe Obama administration, although in a modified form integrated into a broader plan forimproving global health outcomes, and with funding levels no longer increasing.

90 Goldsmith et al.

global order. As great powers, especially the U.S. and China, are increasinglyconstrained by the costs that militarized conflicts would impose on theirtrade and investment interests, they may increasingly seek to pursue theirinternational interests through currying favor among foreign elites andpublics. Foreign aid is an obvious potential tool for this. If targeted, sus-tained, effective, and visible aid gives the best chance of influence, this maycompel great powers to actually do good, and to be seen to be doing so, inorder to do well in their global competition for influence.

Despite the importance of understanding this sort of impact of foreign aidon public opinion in recipient countries, there have not been many empiricalinvestigations on this topic. This is likely in part due to a methodologicalchallenge researchers inevitably encounter when answering the question ofwhether an aid program like PEPFAR actually affects opinion about a donorin recipient countries. This difficulty arises because donors never allocateforeign aid across countries at random. Therefore, causal estimates based onnaıve regression analysis are likely to be biased due to the possibility thata donor may determine the amounts to be allocated to individual countriesbased partly on existing public opinion about the donor in these recipientcountries or, as much of the existing literature suggests, based on other polit-ical or strategic considerations, which are typically unobservable or difficultto measure.

In this article, we tackle this methodological problem and present system-atic new evidence. Specifically, our identification strategy for causal inferenceis to use instrumental variables measuring the magnitude of the HIV/AIDSproblem in aid recipient countries. These instruments are strongly correlatedwith the amount of PEPFAR funding allocated across countries, but arelikely to be uncorrelated with unobserved heterogeneity (conditional on aset of observable pre-treatment covariates). By using a large multinationalsurvey with a battery of questions, we also undertake multiple placebo testsand examine whether the treatment variable has a significant effect onlyon public opinion about the U.S. but not on public opinion about othercountries.

The results of our analysis show that PEPFAR substantially improves per-ceptions of the U.S. in recipient countries, and they are robust to model spec-ifications. Our finding implies that in addition to its potential humanitarianbenefits, under certain conditions, foreign aid can serve as an importantstrategic goal for those countries that give it: fostering positive perceptionsamong foreign publics. By doing good, a country can do well.


1 Does Foreign Aid Affect Foreign Public Opinion?

There exist a range of reasons why countries give foreign aid. It may begiven altruistically to improve economic and social well-being in develop-ing countries, as advocated by Sachs (2005) and practiced to a significantextent by some Northern European countries (Berthelemy, 2006; Gates andHoeffler, 2004; Hoeffler and Outram, 2011). Alternatively, foreign aid maybe used as a tool of direct leverage over political elites to advance com-mercial or foreign policy goals important to the donor (Alesina and Dollar,2000; Browne, 2006, Chapter 6; Dreher et al., 2008a, 2011; Kuziemko andWerker, 2006; Radelet, 2006; Vreeland, 2011; Wang, 1999). Finally, foreignaid may be provided at least in part in the hope of creating favorable publicperceptions of the donor in recipient countries.

The purpose of this article is to examine the effectiveness of the latterobjective of foreign aid, for the case of PEPFAR. In doing this, we comple-ment the existing literatures on aid effectiveness, which examine whether aiddelivers development benefits to recipient countries (for example, Arndt andTarp, 2010; Burnside and Dollar, 2000; Dalgaard and Tarp, 2004; Doucou-liagos and Paldam, 2013; Dreher et al., 2008b; Feeny and de Silva, 2012;Headey, 2008; Miniou and Reddy, 2010; Mekasha and Tarp, 2013; Mishraand Newhouse, 2009; Rajan and Subramanian, 2008) and whether aid deliv-ers direct geo-strategic benefits to donors (for example, Dreher and Strum,2012; Milner and Tingley, 2013; Nowak-Lehmann et al., 2009; Vreeland,2011).

Theoretically, it is plausible that by instilling gratitude in those it helps orthrough promoting an image of positive action, compassion and generosity,foreign aid may create or strengthen positive perceptions of donor nationsin recipient countries. But there are also reasons to suspect that foreign aidcould be an ineffective tool of influence (Adelman, 2011; Lindsay, 2011).Recipients may be unaware of the origins of the aid they receive; the donor’smotivations might be seen as primarily self-serving; the positive feelingsassociated with aid may be too small to shift perceptions shaped by moresalient and dramatic foreign policy behavior; or aid programs may simplyfail to work and, therefore, fail to sway people’s opinions in the absence ofobvious improvements to their quality of life.

To date, there are only a limited number of empirical studies examiningthese theoretical possibilities. Furthermore, they are often based on singlecases and consequently tend to suffer from a lack of generalizability, or do

92 Goldsmith et al.

not fully control for confounding factors that might render their conclusionsinvalid. In the following section, we summarize the literature before provid-ing further details on PEPFAR and discussing its potential as a means offostering improved opinion.

1.1 The Existing Literature

Most of the existing studies that examine aid’s impact on opinion have beenwithin-country case studies. They include studies focusing on U.S. heartsand minds campaigns in conflict zones, such as Iraq and Afghanistan, as wellas studies focusing on disaster relief aid after the 2004 Indian Ocean tsunami,the 2005 Pakistan earthquake, or the 2011 earthquake and tsunami in Japan(these are discussed in more detail in the following paragraphs). One statis-tical paper by Goldsmith et al. (2005) takes a comparative, cross-nationalperspective, using data from a variety of countries. It is, however, not focusedon estimating the effects of a specific foreign aid program. Rather, foreignaid variables are included among a range of independent variables expectedto affect post-9/11 international opinion about U.S. foreign policy. Overall,these previous studies on the impacts of foreign aid on public opinion aboutdonors in recipient countries provide mixed results.

A series of qualitative studies has been undertaken through the FeinsteinCenter at Tufts University, looking at perceptions of aid donors in Kenyaand Afghanistan (Bradbury and Kleinman, 2010; Fishstein, 2010; Gompel-man, 2011; Gordon, 2011). While there is some variation across field sites,these studies generally find that aid is ineffective in positively influencingrecipients’ perceptions of donors. Importantly, although the Feinstein Cen-ter work is thorough and detailed, the external validity of their findings isquestionable because their qualitatively oriented studies are based on sam-ples that tend to be small. Furthermore, there is little basis for expectingthat similar results to be found in cases beyond the immediate context ofAfghanistan and conflict environments.

The generally pessimistic findings of the Feinstein Center studies are con-sistent with the finding of a large-N quantitative study, also undertaken inAfghanistan, associated with an evaluation of German aid (Bohnke et al.,2010). This study found a weak but statistically significant associationbetween foreign aid and public opinion about the foreign peace-buildingwork. However, this positive relationship disappeared when the survey, orig-inally conducted in 2007, was repeated in 2009.


In a study of aid’s impact on attitudes in the wake of the 2005 Pakistanearthquake, Andrabi and Das (2010) suggest that exposure to foreign assis-tance had a long-lasting and statistically significant impact on affected vil-lagers’ trust of foreigners. At the same time, however, the authors found thatpositive perceptions tapered off as distance from the fault line increased, sug-gesting that the impact was localized among those who benefitted directlyfrom the aid or who came from communities that benefitted.

Outside of academia, changes in surveyed public opinion over time havealso been claimed as evidence of aid’s ability to improve public opinion.Commonly cited cases are high-profile U.S. disaster relief aid after the 2004Indian Ocean tsunami, the 2005 earthquake in Pakistan, and the 2011 earth-quake and tsunami in Japan. In all instances, various polls (McCawley, 2006;Terror Free Tomorrow for 2005, 2006; Wike, 2012) show that opinions of theU.S. improved in assisted countries. While such eye-ball studies of beforeand after comparisons lack the rigor of careful statistical analysis and thustheir findings might potentially be spurious, the magnitude and timing ofthe change in opinion are suggestive. In the case of the Pakistan earthquake,though, changes may have only been transient, with opinions of the U.S.returning to pre-quake levels relatively rapidly (Kinder, 2010).

1.2 Is PEPFAR Effective?

The mixed findings of these studies may in part stem from differing impactsassociated with different types of aid. Because foreign aid is given in differ-ent forms and in different contexts, it is unlikely that all forms of aid willimprove public opinion. PEPFAR differs from those aid programs previouslystudied in that it is an ongoing aid program, as opposed to a response to anatural disaster or aid extended during a military conflict, and it is givenacross numerous countries and contexts. Accordingly, its impacts might quitereasonably be different from those found in other studies.

PEPFAR comprises a substantial share of the U.S. development aid budget(approximately 15% in 2009). Although the total amount of U.S. develop-ment aid itself is small when contrasted with the total spending the U.S.devotes to the military and other foreign policy activities,3 we argue that

3 In 2008, total U.S. aid spending was 0.18% of the country’s Gross National Income (GNI),compared to military spending of 4.32% of GNI. However, due to the size of its economy, theU.S. also donates the largest amount of aid globally (Radelet, 2006).

94 Goldsmith et al.

PEPFAR could nevertheless have positive perceptual ‘‘externalities’’ (indi-rect effects in a separate policy realm) for the following four reasons.

First, it is targeted at a critical need. The proportion of direct and indirectPEPFAR beneficiaries in the population of a recipient country is expectedto be fairly large, particularly in the countries worst affected by HIV andAIDS. This is due to the scale of the HIV and AIDS epidemic. In thosecountries, the illness has had a major adverse impact on the lives of a signif-icant portion of the population. In Botswana, for example, about a quarterof the population aged between 15 and 49 years was HIV positive in 2009.This figure is high enough to ensure that the disease also impacts manyof those not living with HIV and AIDS themselves, as even non-sufferersare likely to have relatives or friends suffering from HIV and AIDS. More-over, many of the poorest countries receiving PEPFAR funding had, priorto the arrival of large-scale aid, been almost completely unable to provideanti-retroviral therapy (ART) to HIV and AIDS sufferers. But the provi-sion of PEPFAR (and other similar funding) has significantly changed thissituation.

Second, PEPFAR aid is sustained over a considerable period of time, whichmay increase perceptions of genuine commitment by the donor among recip-ient populations. It also allows for information about the effects of the aidand its source to become more firmly embedded in public perceptions dueto repeated exposure to information about PEPFAR. This should operatethrough basic cognitive ‘‘priming’’ mechanisms (Kolb and Whishaw, 2008),and also through gradual social spread of awareness via word of mouth.

Third, PEPFAR appears to have been effective in reducing the adverseimpacts of HIV and AIDS. The U.S. has made considerable efforts to ensurethat PEPFAR has been a well-run aid program (El-Sadr et al., 2012; Sepul-veda et al., 2007; Simonds et al., 2012). Its objectives are well specifiedand misuse of the funds by recipients is minimized, if not completely elimi-nated, through careful monitoring. Indeed, the 2011 WHO/UNAIDS GlobalHIV/AIDS Response Progress Report argues that the major reason for thedecline in AIDS deaths in 2010 since the peak year in 2005 was PEPFAR(Cumming, 2012), and at least two academic studies provide evidence sug-gesting that PEPFAR has had positive health impacts. Bendavid and Bhat-tacharya (2009) found that PEPFAR funding has significantly reduced HIVand AIDS related mortality. Nunnenkamp and Ohler (2011) found that U.S.bilateral HIV funding (essentially PEPFAR) has a significant impact on HIVrelated deaths.


Finally, PEPFAR is well publicized and thus visible among publics. TheU.S. has made sure that the fund has a high profile in recipient countries as‘‘the leading platform for US health diplomacy and a symbol of Americancapacity to achieve constructive and beneficial change’’ (Collins et al., 2012,p. 1578). The U.S. has carefully and deliberately branded PEPFAR fundedwork to ‘‘ensure appropriate recognition for U.S. programs and contribu-tions’’ (PEPFAR, 2012, p. 2). The stars and stripes of the U.S. flag are aprominent feature of the PEPFAR logo, and the signing of PEPFAR countryagreements is often undertaken by high profile figures in the U.S. adminis-tration, increasing in-country media exposure and public awareness (Ingramand Kazanecki, 2010).4 In countries with high HIV prevalence rates, PEP-FAR has a significant media profile with improvements in HIV outcomesfrequently being attributed to PEPFAR (for example, Botswana Gazette,2011; IOL, 2012; Sambira, 2013; Times of Zambia 2012a; Times of Zam-bia 2012b; Vo 2008). PEPFAR is also unlikely to be overshadowed, or evenrivaled, by other donors’ HIV/AIDS-related aid. The U.S. was the sourceof 52% of such bilateral aid commitments globally, 2004–2006, while thesecond largest donor, the U.K., contributed 18% (Henry J. Kaiser FamilyFoundation, 2013).5

These four characteristics of PEPFAR — targeted to address a critical andwidely understood need, sustained delivery of aid over time, effectiveness (orperceived effectiveness), and being highly visible — condition the theoreticalcontribution of our study. We expect that aid that meets these conditionswill be more likely to actually affect opinion among recipient populationsabout the donor state.

2 Data and Variables

To test whether PEPFAR has an impact on public opinion in recipient coun-tries, we conduct regression analysis. The treatment variable is the per capitaamount of PEPFAR funds provided to each recipient country (in naturallog). We averaged official statistics over the first three years (2004–2006)

4 For examples of high profile figures’ visits and their news coverage, see Kanyesigye (2012),Sunday Standard (2011), and Times of Zambia (2012a).

5 We only use PEPFAR data for 2004–2006, but in 2007–2010, the years for which we measurepublic opinion, the U.S. share of global bilateral HIV/AIDS aid was 59%.

96 Goldsmith et al.

since the disbursement of funds began in 2004.6 The number of countriesreceiving PEPFAR funds during this period was 79. By taking the natu-ral log of the highly skewed per capita amount, we exclude countries notreceiving any funding during the first 3 years.7

The outcome variable is the difference between the percentages (in naturallog) of respondents answering ‘‘approve’’ and ‘‘disapprove’’ to the followingquestion asked in Gallup World Polls (GWP): ‘‘Do you approve or disap-prove of the job performance of the leadership of the United States?’’. TheGWP is the only available multinational survey with nearly universal cov-erage of the developing world, repeating the same questions annually aboutperceptions of the U.S. This question is especially appropriate for our studybecause it focuses on the evaluation of current or recent U.S. leaders andtheir behavior. By taking the natural log for both treatment and outcomevariables, the estimated treatment effect measures elasticity.8

The sampling procedures of the GWP targeted the national populationaged 15 or older in each country. Random-Digit-Dial (RDD) phone sur-veys were conducted where at least 80% of the population has a telephone(primarily in developed countries), while face-to-face interviews were con-ducted in all other countries. When RDD telephone surveys were conducted,Gallup applied a list-assisted sampling design (Casady and Lepkowski, 1993)with all telephone numbers as the sampling frame. For face-to-face sur-veys, an area sampling frame was used based on the latest available cen-sus data, with a stratified multi-stage random sample based on populationunits ranging usually from cities with 1 million or more people to villageswith under 10,000. The sample size is roughly 1000 surveys per countryper year.9

6 U.S. President George W. Bush announced PEPFAR in January 2003, but disbursement didnot start until 2004. The data sources are various PEPFAR Operational Plans for PEP-FAR funds (in millions of current U.S. dollars) obtained at <http://www.pepfar.gov/about/c19388.htm>, and the World Bank’s World Development Indicators (WDI) for thetotal population, obtained at <http://data.worldbank.org/data-catalog/world-development-indicators?cid=GPD WDI>.

7 As a robustness test, we also run a set of regressions after imputing missing values so thatwe are able to use a larger number of countries for estimation. See a subsection (RobustnessTests) in the Results Section.

8 We also used a slightly different outcome variable for another robustness test, the results ofwhich are presented in a subsection (Robustness Tests) in the Results Section.

9 For details, see Tortora et al. (2010).


Gallup Inc. launched this annual global public opinion study in 2006,asking the question about U.S. leadership in each year.10 For our analysis,we use three outcome variables — the average during the last 2 years ofthe Bush administration (2007–2008), the average during the first 2 yearsof the Obama administration (2009–2010), and the average during all the 4years. The use of these variables allows us to examine whether the PEPFARfunding affected public opinion in recipient countries only when George W.Bush, the founder of the PEPFAR program, was in office, and to be surethat our results are not driven by changing global public opinion about theU.S. resulting from the election of Barack Obama.

Because it is likely to have taken time for PEPFAR’s positive healthimpacts to become clearly apparent, any effects of providing PEPFAR fundsover multiple years on public opinion in recipient countries are expected tobe gradual and cumulative. In other words, we assume that any short-termchange (for example, a change from the previous year) in PEPFAR fundsdoes not abruptly cause a short-term change in public opinion about thedonor among recipient countries. For this reason, we take the averaged val-ues for both treatment and outcome variables and run cross-sectional (i.e.,cross-country) regressions.11

To increase confidence in our causal inference, we use instrumental vari-ables (IVs) that are expected to influence public opinion about the U.S. onlythrough the treatment variable (the amount of per capita PEPFAR funding).The use of instrumental variables is often an effective way to address theissue of endogeneity using observational data. Specifically, we use the HIVprevalence rate (the average of 2000–2003, in natural log) and the numberof annual HIV related deaths divided by the total population (the averageof 2000–2003, in natural log).12 HIV prevalence and death-rate data were

10 The number of countries in which this specific question was asked was 119 in 2006, 93 in 2007,114 in 2008, 112 in 2009, and 115 in 2010. The country-level aggregated data were obtained at<https://worldview.gallup.com/>.

11 Previous studies have shown that general ODA’s effects on economic growth are better mod-elled over periods of several (specifically, four) years (Burnside and Dollar, 2000; Headey, 2008).

12 The data source for the HIV prevalence rate and the number of annual HIV related deathsis UNAIDS Data obtained at <http://unaids.org/globalreport/documents/HIV EstimatesGR2010 1990 2009 en.xls>. If the reported HIV prevalence rate is “<0.1” percentage, weassigned a random number between 0 (exclusive) and 0.1 (exclusive). Similarly, if the reportednumber of deaths is “<1000”, “<500”, “<200”, or “<100”, we assigned a random numberwithin the possible range for each observation. The total population (2000–2003 average),the denominator for the second instrumental variable, is from the World Bank’s WDI (seeFootnote 6).

98 Goldsmith et al.

used extensively in designing PEPFAR’s approach, including the selection ofrecipient countries, and setting annual funding levels (Lyerla et al., 2012). Aswould be expected, our instruments are therefore strongly correlated withthe treatment variable — the per capita amount of PEPFAR funding (theaverage in 2004–2006).

Assuming they are valid, our instruments estimate the ‘‘local averagetreatment effect’’ (Imbens and Angrist, 1994). Formally, we estimate theaverage cross-country effects of the variation in PEPFAR funding, whichis determined by the variation in these health indicators, rather than byother motivations such as geo-strategic selection, on foreign public opinion.In other words, we have considerable confidence that we estimate the effectof foreign aid targeted to need on public opinion. The allocation of PEPFARfunds may be politically determined to some degree, but our IV estimatesare unrelated to such political considerations.

An important assumption in this analysis is the exclusion restriction, anassumption that instruments can be excluded from the second-stage regres-sion because they are uncorrelated with the error term, conditional on othercovariates included in the model. To make this assumption plausible, we needto add variables that are expected to correlate with our instrumental andoutcome variables.13

First, U.S. total Official Development Assistance (ODA) per capita (innatural log, averaged over 2004–2006) is the total aid funding received ineach recipient country from the U.S. (gross disbursements, current millionU.S. dollars) divided by the total population (in millions).14 Including thisvariable ensures that any impact on opinion is a result of PEPFAR and notof other aid programs designed in part to address public health problems indeveloping countries.

Second, the U.S. may provide more PEPFAR funds to poorer economiesand/or poorly governed countries, and these countries are also likely to havemore severe public health problems. To control for these possibilities, we addGDP per capita (2004–2006 average, in natural log).15 For similar reasons,

13 We considered using a wide range of variables. See a subsection (Robustness Tests) in theResults Section for more details.

14 The data source of ODA disbursements is OECD DAC CRS Aid Database, obtained at<http://www.oecd.org/document/33/0,2340,en 2649 34447 36661793 1 1 1 1,00.html#dac>.See Footnote 12 for the source of total population.

15 The data source is the World Bank’s WDI (see Footnote 6). Specifically, the variable we useis GDP per capita, PPP (constant 2005 international dollars).


we control for the level of civil liberties and the quality of governance,using the Freedom House civil liberties score (2004–2006 average),16 as wellas a compositional score based on 6 World Bank Governance Indicators(2004–2006 average).17 We also include a dummy variable for sub-SaharanAfrican countries to control for unobserved region-specific covariates.18

Because the global HIV and AIDS epidemic is most acute in Sub-SaharanAfrica and because much of the PEPFAR funding has been directed to coun-tries in this region, this variable is essential to ensure that our results arenot driven by other, unobserved, region specific characteristics.

Third, the problems associated with HIV and AIDS in a given countrymay affect that country’s economic and political relationships with the U.S.(and then, in turn, the country’s public opinion about the U.S.) Specifically,the U.S. tends to have weaker trade ties with countries with more severe pub-lic health problems. This may be because these countries are economicallyclosed and/or because they lack goods to export to the U.S. These countriesalso tend to have weaker political ties with the U.S., perhaps because of theirsmall presence in political and economic negotiations at international arenas.To account for these connections, we use 3 variables. The first 2 are importsto the U.S. from a given country as a percentage of total U.S. imports (innatural log), and exports from the U.S. to a given country as a percentage oftotal U.S. exports (in natural log).19 As a proxy for each recipient country’spolitical connection to the U.S., we add a variable measuring accordance ofeach country’s United Nations General Assembly (UNGA) voting patternwith that of the U.S.20

16 The data source is Freedom House, Freedom in the World Country Rankings, obtainedat <http://www.freedomhouse.org/template.cfm?page=439>. An alternative measure of thelevel of political development is polity score from Polity IV data. We, however, prefer to use theFreedom House’s score, as there are no missing values for countries included in our analysis.The two variables are highly correlated (r = 0.81).

17 The data source is Teorell et al. (2011). The original source is Kaufmann et al. (2009) avail-able at <http://www.govindicators.org>. Because the 6 indicators are highly correlated, wemeasured a Cronbach’s alpha. The reliability coefficient is 0.97.

18 We generate a dummy variable based on a variable indicating the region of the country inTeorell et al. (2011). The original source is Teorell and Hadenius (2005).

19 Fleck and Kilby (2006) use these variables to capture the impact of trade with the U.S. on U.S.aid allocation decisions. The data source is U.S. Department of Commerce, Bureau of the Cen-sus, Foreign Trade, obtained at <http://www.census.gov/foreign-trade/balance/>. Anothercommonly used indicator for the trade ties with the U.S. is the sum of exports to the U.S. andimports from the U.S. as the percentage of GDP. This variable, however, is found to be veryweakly correlated with our instrumental variables.

20 The data source is Dreher and Sturm (2012), obtained at <http://www.uni-heidelberg.de/fakultaeten/wiso/awi/professuren/intwipol/datasets en.html>. The measure used is the pro-

100 Goldsmith et al.

Conditional on these covariates, we argue, it is unlikely that our instru-mental variables (HIV/AIDS rates) would have any impact on the outcomevariable (public opinion about the U.S. leadership) through any mecha-nism other than their impact through the treatment variable (PEPFAR). Inother words, having accounted for these covariates, we have confidence thatour instrumental variables are not systematically correlated with the errorterm in the second-stage regression. Ultimately, the validity of this exclusionrestriction assumption is an empirical matter, and thus we conduct a set ofstandard specification tests.

3 Results

Figure 1 shows the results of our first-cut analysis — simple scatter plotsusing the outcome variable and the treatment variable. Each panel in thisfigure corresponds to one of three periods used for analysis (i.e., 2007–2010,2007–2008, or 2009–2010). The straight lines are OLS regression lines. Allthese lines have a positive slope, and the estimated slope coefficients fromsimple bivariate regressions (0.26 for 2007–2010, 0.27 for 2007–2008, and0.32 for 2009–2010) are highly significant at the 99% level.

The magnitude of the effect is substantial and similar across the last2 years of the Bush administration and the first 2 years of the Obama admin-istration. Specifically, regardless of who the U.S. president was at the timeof the surveys, doubling the per capita amount of PEPFAR funds increasedthe ratio of the percentages of Approve and Disapprove by about 30%. Forexample, if the percentage of Approve was initially 40%, the estimated coef-ficients suggest that it would increase to 52%.21

3.1 Estimated Treatment Effects

Table 1 shows the results of two-stage least square (2SLS) regressions. Theeffect of PEPFAR funds per capita on the perception of U.S. leadership is

portion of UN votes in the year of interest in which the aid recipient country voted in-line withthe United States. For a detailed description of the calculation involved, see Dreher and Strum(2012, p. 371).

21 This simple interpretation assumes that PEPFAR increases the percentage of Approve whilekeeping the percentage of Disapprove constant. PEPFAR may also decrease the percentage ofDisapproval, but understanding the impacts of PEPFAR on the composition of percentages isbeyond the scope of this article. The average percentages of Approve and Disapprove for theperiod of 2007–2010 were 45.8% and 31.2%, respectively. The remaining percentage includesrespondents who either chose “Don’t know” or refused to answer.


Figure 1. Scatter plots.Note: The bivariate relationship between PEPFAR funding per capita (log) and the log-ratio of the percentage of “approve” and the percentage of “disapprove” responses to thefollowing question asked in Gallup World Polls: “Do you approve or disapprove of the jobperformance of the leadership of the United States?”.

positive and statistically significant at the 99% level of confidence for the2007–2010 period overall, as well as for the final Bush years, 2007–2008. It issignificant at the 95% level during the initial Obama years, 2009–2010.22 The

22 We experimented with many different sets of control variables, but the results, discussedshortly, are similar.


Table 1. Results of 2SLS regressions.

2007–2010 2007–2008 2009–2010Average Average Average

PEPFAR funds per capita (log) 0.202 0.229 0.226[0.068] [0.080] [0.091]

US total ODA per capita (log) −0.107 −0.101 −0.065[0.067] [0.081] [0.057]

GDP per capita (log) −0.305 −0.405 0.039[0.185] [0.205] [0.198]

World Bank governance indicator −0.528 −0.366 −0.472[0.414] [0.502] [0.466]

Freedom House civil liberties score −0.250 −0.225 −0.282[0.151] [0.166] [0.158]

Africa dummy 0.804 0.698 0.973[0.318] [0.311] [0.428]

Imports to US (% of total US 0.080 0.125 0.030imports, log) [0.089] [0.096] [0.091]

Exports from US (% to total US −0.052 −0.118 −0.073exports, log) [0.130] [0.141] [0.152]

UN voting in accordance with US 0.970 0.476 0.531[0.693] [0.697] [0.770]

Constant 4.089 4.710 1.439[1.719] [1.909] [1.783]

Number of countries 57 55 46Wald Chi-square 122.41 109.83 122.37R-square 0.572 0.550 0.677Root MSE 0.685 0.737 0.591

Note: The outcome variable is the log ratio of Approve (%) and Disapprove (%) responsesto the question: “Do you approve or disapprove of the job performance of the leadershipin the United States?”. Numbers in brackets are robust standard errors.

estimated marginal effects are slightly smaller than those based on simplebivariate OLS (shown in Figure 1) but still substantial. Specifically, if theper capita amount of PEPFAR funding doubles, the ratio of Approval toDisapproval increases by 20–23%. Importantly, we re-emphasize, the effects


Table 2. Results of specification tests.


Tests of endogeneityRobust score Chi-square 0.481 1.539 0.088

(0.488) (0.215) (0.766)Robust regression F 0.406 1.323 0.070

(0.527) (0.256) (0.792)

Test of relevance of instrumentsPartial R-square (first-stageregression)

0.452 0.449 0.387

F (first-stage regression) 25.509 23.877 15.505(0.000) (0.000) (0.000)

Test of over-identification restrictionsChi-square 0.812 1.104 0.315

(0.368) (0.293) (0.574)

Note: Numbers in parentheses are p values. Also see Note in Table 1.

of PEPFAR on public opinion in recipient countries are independent of theU.S. President’s popularity, partisanship and of any perceived policy differ-ences between Bush and Obama.23

As expected, a set of specification tests (shown in Table 2) suggests thatour instruments are highly valid empirically. The small values of test statis-tics for endogeneity — Wooldridge’s (1995) robust Chi-square score testand a robust regression-based F test — imply that our treatment variable

23 The negative — but statistically insignificant — association of total U.S. ODA per capitawith opinion about U.S. leadership might be related to the various strategic purposes forwhich some U.S. aid is given (Dreher et al., 2008a; Gates and Hoeffler, 2004). The mixedgoals of promoting U.S. foreign policy may cause U.S. aid to be less effective at promotingdevelopment (Headey, 2008). Specifically, U.S. aid may be extended to more corrupt (Alesinaand Weder, 2002) and/or less democratic regimes, and to those that violate human rights, toa greater extent than is the case for most other major bilateral donors (Hoeffler and Outram,2011). Thus, U.S. ODA might be perceived in some recipient countries as ineffective, andcomplicit in propping up dysfunctional or repressive regimes. It is worth further investigatingsuch potentially negative effects of more general U.S. aid on opinion about the U.S. in recipientcountries, but we leave it for future research.


(PEPFAR funds per capita) does not need to be considered as endogenous.24

The partial R-square statistics in the first-stage regression range from 0.39to 0.45. Thus, the instruments explain 39–45% of the first-stage regression’soverall goodness of fit. The F statistics for the joint significance of the instru-ments in the first stage range from 15.5 to 25.5 and are highly significant.They are larger than 10, which Stock et al. (2002) suggest as the benchmarkto assess the reliability of 2SLS estimates for causal inference (when thereis one endogenous variable).

We also conducted over-identification tests to examine whether the instru-ments are uncorrelated with the error term. Sargan’s (1958) and Basmann’s(1960) Chi-square test statistics are very small. The associated high p valuesindicate that the exclusion restriction assumption is empirically supported.

3.2 Placebo Tests

To further ensure that no additional unobserved factors are driving ourresults, we also conducted placebo tests — also called ‘‘control-outcome’’tests (Rosenbaum, 2010) — using similar-but-different outcome variables.Given that we do not expect PEPFAR aid to have an effect on opinions aboutother countries, besides the U.S., these tests ‘‘exploit the anticipated absenceof an effect to provide information about unmeasured biases’’ (Rosenbaum,2010, p. 121). They are based on identical questions about the approvalof leadership of (‘‘Do you approve or disapprove of the job performance ofthe leadership of. . . ?’’) China, France, Germany, India, Japan, Russia, theUnited Kingdom, and the respondent’s home country. As in the case of theU.S., each ‘‘non-equivalent dependent variable’’ (Cook and Campbell, 1979,Chapter 2) measures the 2007–2008, 2009–2010, or 2007–2010 average differ-ence between the percentage of Approval (in natural log) and the percentageof Disapproval (in natural log).

Table 3 shows the estimated marginal effects of the treatment variable onthese 9 outcome variables. The only statistically significant effects (at the99% or 95% level) were obtained for the outcome variable measuring theApproval vis-a-vis Disapproval of leadership of the U.S. These tests furthersupport our expectation that PEPFAR affects public opinion specifically

24 For this reason, we also estimated our model based on OLS. See a subsection (RobustnessTests) in the Results Section.


Table 3. Results of placebo tests.


United States 0.202 0.229 0.226[0.068] [0.080] [0.091]

China −0.012 0.028 −0.042[0.050] [0.055] [0.082]

France −0.060 −0.064 0.032[0.059] [0.059] [0.088]

Germany −0.058 −0.069 −0.055[0.052] [0.062] [0.079]

India −0.021 −0.015 −0.086[0.063] [0.076] [0.067]

Japan −0.018 −0.018 −0.006[0.052] [0.060] [0.079]

Russia −0.043 −0.026 0.088[0.101] [0.098] [0.191]

United Kingdom 0.093 0.111 −0.021[0.080] [0.087] [0.095]

Own country −0.128 −0.122 −0.123[0.075] [0.074] [0.153]

Note: The estimated effects of PEFPAR funding per capita (log,2004–2006 average) on the difference in the percentage of “approve”(in log) and “disapprove” (in log) responses to the following ques-tion asked in Gallup World Polls (2007–2010 average, 2007–2008average, or 2007–2010 average): “Do you approve or disapprove ofthe job performance of the leadership of <China, France, Germany,India, Japan, Russia, the United Kingdom, the United States, orOwn Country>?”. Numbers in brackets are robust standard errors.

about the U.S. in recipient countries, and the relationship is not the resultof confounding factors.

There is a theoretical rationale for these empirical results. While donorcountries tend to give aid to combat HIV/AIDS to those states that are mostaffected by the epidemic, there is also a substitution or free-riding dynamicin which other donors give less to a specific recipient state when one donor


gives a larger amount (Gaibulloev and Sandler, 2012). This could also bedue to explicit multilateral coordination among donors. Given that the U.S.is the largest such donor, this implies that other donors give relatively lessHIV/AIDS-related aid to the states receiving more from the U.S., whichwould support our expectation that public opinion about other countrieswould not be significantly related to PEPFAR funding provided by the U.S.

Interestingly, however, the effects on the leadership Approval vis-a-vis Dis-approval ratio in a respondent’s own country are negative and significantat the 90% level (for 2007–2010 and 2007–2008). The publics in countriesreceiving large PEPFAR funds may assess the relative performance of theircountry’s leaders in comparison to that of the U.S. leaders who provide —or are perceived to provide — effective foreign aid. These barely significanteffects are, however, not robust. Among many model specifications we tried,they are more often insignificant (even at the 90% level) than significant.

3.3 Robustness Tests

We also ran a range of other regression models for further robustness tests.First, building on the initial bivariate correlations reported with the scatterplots in Figure 1 at the beginning of this section we estimated our modelusing multiple regression (i.e., without instruments). The results are pre-sented in Appendix A (Table A). The effects of PEPFAR funds per capitaare slightly smaller, ranging from 0.15 to 0.17, but are still highly signifi-cant at the 95% level in all the three periods of investigation. The similaritybetween the 2SLS and OLS regression results is what we expect, given thatthe test results for endogeneity suggest that we do not necessarily need totreat the per capita amount of PEFAR funds as endogenous, conditional onthe covariates included in our estimation.

Second, we tried an alternative outcome variable; specifically, the differ-ence between the Approval percentage and the Disapproval percentage with-out taking the natural log. The results shown in Appendix B (Tables B1and B2) are quite similar. The effects of our treatment variable are positiveand significant at the 99% or 95% level. The results of specification tests arealso similar.25

25 The Chi-square statistic for the test of over-identification restrictions is large (3.005) enoughto reject the null hypothesis at the 90% level but not at the 95% level.


Third, we randomly imputed small numbers for countries not receiv-ing PEPFAR funds and/or other foreign aid programs.26 This is aimed attesting whether our results are driven by non-random exclusion of coun-tries not receiving aid by taking the natural log of the highly skewed percapita amount of PEPFAR or total ODA. The results are presented inAppendix C (Tables C1 and C2). By imputing the missing values, we sub-stantially increase the number of observations. All three coefficients of PEP-FAR funds per capita are positive and statistically significant at the 95%level for 2007–2010, 90% for 2007–2008, and 99% for 2009–2010. The mag-nitude of the effects is slightly smaller for 2007–2010 and 2007–2008, butalmost exactly the same for 2009–2010.

Finally, we considered using other pre-treatment covariates aimed to sat-isfy the exclusion restriction assumption. They include the geographical dis-tance from the U.S., variables measuring the amount of U.S. military aid,a dummy variable for United Nations Security Council members, a dummyvariable for former U.S. colonies, a dummy variable for Latin American coun-tries, a dummy variable for Egypt, and the percentage of Muslims in eachcountry’s population. Adding these variables individually or jointly does notsubstantially change our main results, although doing so tends to increasethe standard errors of the coefficients for our treatment variable. To avoidthis problem of inefficiency, for our main results reported in Tables 1 and 2,we decided to keep a variable not only when there was a theoretical reasonto include it, but also when its pairwise correlation with one of our instru-mental variables was statistically significant at the 95% level. None of theseadditional pre-treatment covariates satisfied our criteria, and thus we didnot include them in our main analysis.

4 Conclusion

These findings provide robust evidence that PEPFAR has had a strong pos-itive effect on how U.S. leadership has been perceived in recipient countries.The results of specification tests, placebo tests, and robustness tests suggest

26 Specifically, for each period of investigation (2007–2010, 2007–2008, or 2009–2010), we calcu-lated the minimum amount of PEPFAR funds per capita or U.S. total ODA per capita. Foreach of the countries not receiving PEPFAR or other aid programs, we imputed a random num-ber ranging from 0 (exclusive) to the observed minimum value. After imputing these missingvalues, we took the natural log of each variable.


that the observed effects are not statistical artifacts, and thus we are stronglyinclined to believe that they are causal. Our theory is that these empiricallyrobust results stem from the fact that PEPFAR is targeted to address awidely understood need; has been sustained over some time; is — and isperceived to be — effective; and is highly visible.

Whether, and the extent to which, these conditions apply individuallyand/or jointly are important questions for future investigation. At present,such investigations aimed to identify the mechanisms through which PEP-FAR has had its impact are limited by the absence of individual-level (orhousehold-level) survey datasets. Once such data become available, furtherwork could examine, for example, whether positive perceptions are higheramong those impacted directly (i.e., suffering from HIV and AIDS or thefamilies of sufferers) and/or among those with better access to media andthe publicity of PEPFAR.

In addition to further work looking inwards, there is considerable scopefor further work looking outwards. Specifically, one important way ourapproach could be extended is by testing whether other aid programs tar-geted to important needs, sustained over time, perceived as effective, andwell-publicized or otherwise visible to the population are also positivelyassociated with opinion about the U.S. Another would be to examine thisrelationship for other donor countries. Give the large absolute amounts ofaid the U.S. provides, it remains an open question whether other donors’extant smaller programs might affect perceptions within recipient countriesabout these donors.

Another promising area of further investigation stemming directly fromthis article would be to study other humanitarian aid programs addressingglobal health issues that have been successful in tackling other major ill-nesses, such as small pox, polio, river blindness, and malaria. While thereare many examples of aid failing to provide benefits, PEPFAR is not uniquein its achievements. Indeed, there are other well-documented examples of aidhaving significant positive impacts (for example, Demombynes and Tromm-lerova, 2012; Levine, 2007). Thus, future work should seek to extend ouranalysis across a range of programs.

Having acknowledged some limitations and future research directions, weunderscore that our analysis is the first of its kind to use cross-countrydata to make systematic causal inference about the impacts of aid on publicopinion in aid recipient countries. Our findings suggest that policy debates


about PEPFAR and similar programs should consider not only their efficacyin achieving direct goals such as fighting HIV and AIDS, but also their valuein improving the donor country’s global or regional standing.

Beyond the literature on aid effectiveness, our research has broader impli-cations for understanding today’s international relations. If analysts are cor-rect that national image is increasingly a key resource for great powers’international influence, our findings may be good news. With militarizedconfrontation tempered by high economic interdependence (for example,between China and the U.S. today, but unlike the U.S. and USSR during theCold War), great powers might now start learning to place more emphasison improving their image through aid, because it enhances their power. Ourresearch shows that one possible means of doing well in the newly formingarena of international competition for favorable perceptions is by actuallydoing good.

References

Arndt, C., S. Jones and F. Tarp. 2010. “Aid, Growth, and Development: Have We ComeFull Circle?” Journal of Globalization and Development 1(2), Article 5.

Adelman, K. 2011. “Not-So-Smart Power: Go Ahead, Congress, Cut Away at U.S. ForeignAid.” Foreign Policy. April 18, available at http://www.foreignpolicy.com/articles/2011/04/18/not so smart power (March 23, 2013).

Alesina, A. and D. Dollar. 2000. “Who Gives Foreign Aid to Whom and Why?” Journalof Economic Growth 5(1): 33–64.

Alesina, A. and B. Weder. 2002. “Do Corrupt Governments Receive Less Foreign Aid?”American Economic Review 92(4): 1126–1137.

American Political Science Association. 2009. U.S. Standing in the World: Causes, Con-sequences, and the Future, Long Report of the Task Force. Washington DC: AmericanPolitical Science Association.

Andrabi, T. and J. Das. 2010. “In Aid We Trust: Hearts and Minds and the PakistanEarthquake of 2005.” World Bank, Policy Research Working Paper, No. WPS5440.

Basmann, R. L. 1960. “On Finite Sample Distributions of Generalized Classical LinearIdentifiability Test Statistics.” Journal of the American Statistical Association 55(292):650–659.

Beath, A., F. Christia, and R. Enikolopov. 2011. “Winning Hearts and Minds? Evidencefrom a Field Experiment in Afghanistan.” MIT Political Science Department, WorkingPaper, No. 2011–14.

Bendavid, E. and J. Bhattacharya. 2009. “The President’s Emergency Plan for AIDSRelief in Africa: An Evaluation of Outcomes.” Annals of Internal Medicine 150(10):688–695.

Berthelemy, J.-C. 2006. “Bilateral Donors’ Interest vs. Recipients’ Development Motives inAid Allocation: Do All Donors Behave the Same?” Review of Development Economics10(2): 179–194.

Bohnke, J., J. Koehler, C. Zurcher. 2010. “Assessing the Impact of Development Coop-eration in North East Afghanistan 2005–2009.” Bundesministerium fur wirtschaftliche


Zusammenarbeit und Entwicklung. Unpublished Report, available at http://www.oecd.org/countries/afghanistan/46785983.pdf (accessed on March 23, 2013).

Botswana Gazette. 2011. “U.S. Global AIDS Coordinator Ambassador Eric Goosby toVisit Botswana”. The Botswana Gazette. March 7, available at http://www.gazettebw.com/index.php?option=com content&view=article&id=8903: us-global-aids-coordina-tor-ambassador-eric-goosby-to-visit-botswana&catid=18: headlines&Itemid=2 (acce-ssed on January 28, 2013).

Bradbury, M. and M. Kleinman. 2010. Winning Hearts and Minds? Examining the Rela-tionship between Aid and Security in Kenya. Medford, MA: Feinstein InternationalCenter, Tufts University.

Browne, S. 2006. Aid and Influence: Do Donors Help or Hinder? London: Earthscan.Burnside C. and D. Dollar. 2000. “Aid, policies and growth.” American Economic Review

90(4): 847–868.Bustamante, M. J. and J. E. Sweig. 2008. “Buena Vista Solidarity and the Axis of Aid:

Cuban and Venezuelan Public Diplomacy.” Annals of the American Academy of Polit-ical and Social Science 616(1): 223–256.

Casady, R. J. and J. M. Lepkowski. 1993. “Stratified Telephone Survey Designs.” SurveyMethodology 19(1): 103–113.

Collins, C., M. Isbell, A. Sohn, and K. Klindera. 2012. “Four Principles for ExpandingPEPFAR’s Role as a Vital Force in US Health Diplomacy Abroad.” Health Affairs31(7): 1578–1584.

Cook, T. and D. Campbell. 1979. Quasi-Experimentation: Design and Analysis Issues forField Settings. Boston, MA: Houghton Mifflin.

Cumming, R. G. 2012. “HIV and AIDS in Africa: Good News and Bad News.” MedicalJournal of Australia 196(5): 309.

Dalgaard, C-J. and F. Tarp. 2004. “On the Empirics of Foreign Aid and Growth.” Eco-nomic Journal 114(496): 191–216.

Demombynes, G. and S. K. Trommlerova. 2012. “What has driven the decline ofinfant mortality in Kenya?” World Bank Policy Research Working Paper, WPSNo. 6057.

Doucouliagos, H. and M. Paldam. 2013. “The Robust Result in Meta-analysis of AidEffectiveness: A Response to Mekasha and Tarp.” The Journal of Development Studies49(4): 584–587.

Dreher, A., P. Nunnenkamp, and R. Thiele. 2008a. “Does US aid buy UN general assemblyvotes? A disaggregated analysis.” Public Choice 136(1): 139–164.

Dreher, A., P. Nunnenkamp, and R. Thiele. 2008b. “Does Aid for Education EducateChildren? Evidence from Panel Data.” The World Bank Economic Review 22(2):291–314.

Dreher, A., P. Nunnenkamp and R. Thiele. 2011. “Are ‘New’ Donors Different? Comparingthe Allocation of Bilateral Aid Between non DAC and DAC Donor Countries.” WorldDevelopment 39(11): 1950–1968.

Dreher, A. and J.-E. Sturm. 2012. “Do the IMF and the World Bank influence voting inthe UN General Assembly?” Public Choice 151(1, 2): 363–397.

El-Sadr, W. M., C. B. Holmes, P. Mugyenyi, H. Thirumurthy, T. Ellerbrock, R. T. Ferris,I. Sanne, A. Asiimwe, G. Hirnschall, R. N. Nkambule, L. Stabinski, M. Affrunti, C.Teasdale, I. Zulu, and A. Whiteside. 2012. “Scale-up of HIV Treatment Through PEP-FAR: A Historic Public Health Achievement.” Journal of Acquired Immune DeficiencySyndromes 60(Suppl. 3): S96–S104.

Embassy of the United States, Islamabad, Pakistan. 2012. “U.S. Releases New Fundsfor Pakistani Energy Projects,” available at http://islamabad.usembassy.gov/pr-080312.html (accessed on March 23, 2013).

Feeny, S. and A. de Silva. 2012. “Measuring Absorptive Capacity Constraints to ForeignAid.” Economic Modelling 29(3): 725–733.

Fleck, R. K. and Kilby, C. 2006. “How Do Political Changes Influence US Bilateral AidAllocations? Evidence from Panel Data.” Review of Development Economics 10(2):210–223.


Fishstein, P. 2010. Winning Hearts and Minds? Examining the Relationship between Aidand Security in Afghanistan’s Balkh Province. Medford, MA: Feinstein InternationalCenter, Tufts University.

Friedberg, A. L. 2010. “Implications of the Financial Crisis for the US–China Rivalry.”Survival 52(4): 31–53.

Gaibulloev, K. and T. Sandler. 2012. “Aid for AIDS in Africa.” Empirical Economics43(3): 1171–1197.

Gates, S. and A. Hoeffler. 2004. “Global Aid Allocation: Are Nordic Donors Different?”The Center for the Study of African Economies, WPS No. 2004–34, available at http://www.csae.ox.ac.uk/workingpapers/pdfs/2004-34text.pdf (accessed on September 23,2013).

Gill, B. and Y. Huang. 2006. “Sources and Limits of Chinese ‘Soft power’.” Survival 48(2):17–36.

Goldsmith, B. E., Y. Horiuchi, and T. Inoguchi. 2005. “American Foreign Policy andGlobal Opinion: Who Supported the War in Afghanistan?” Journal of Conflict Reso-lution 49(3): 408–429.

Goldsmith, B. E. and Y. Horiuchi. 2009. “Spinning the Globe? U.S. Public Diplomacyand Foreign Public Opinion.” The Journal of Politics 71(3): 863–875.

Goldsmith, B. E. and Y. Horiuchi. 2012. “In Search of Soft Power: Does Foreign PublicOpinion Matter for U.S. Foreign Policy?” World Politics 64(3): 555–585.

Gompelman, G. 2011. Winning Hearts and Minds? Examining the Relationship betweenAid and Security in Afghanistan’s Faryab Province. Medford, MA: Feinstein Interna-tional Center, Tufts University.

Gordon, S. 2011. Winning Hearts and Minds? Examining the Relationship between Aidand Security in Afghanistan’s Helmand Province. Medford, MA: Feinstein InternationalCenter, Tufts University.

Govender, P. 2012. “U.S. Hands More Control to South Africa in its AIDS fight.” Reuters.August 7, available at http://www.reuters.com/article/2012/08/07/us-clinton-safrica-aids-idUSBRE8760ZU20120807 (accessed on March 23, 2013).

Hanusch, M. 2012. “African Perspectives on China–Africa: Modelling Popular Perceptionsand their Economic and Political Determinants.” Oxford Development Studies 40(4):492–516.

Headey, D. 2008. “Geopolitics and the Effect of Foreign Aid on Economic Growth:1970–2001.” Journal of International Development 20(2): 161–180.

The Henry J. Kaiser Family Foundation. 2013. “Financing the Response to AIDS in Low-and Middle-Income Countries: International Assistance from Donor Governments,”available at http://kff.org/global-health-policy/report/financing-the-response-to-aids-in-low/ (accessed on 24 September 24, 2013).

Hoeffler, A. and V. Outram. 2011. “Need, Merit, or Self-Interest: What Determines theAllocation of Aid?” Review of Development Economics 15(2): 237–250.

Huntington, S. P. 1991. The Third Wave: Democratization in the Late Twentieth Century.Norman, OK: University of Oklahoma Press.

Imbens, G. W. and J. D. Angrist. 1994. “Identification and Estimation of Local AverageTreatment Effects.” Econometrica 62(2): 467–475.

Ingram, A. and W. Kazanecki. 2010. “Governmentality and security in the US President’sEmergency Plan for AIDS Relief (PEPFAR).” Geoforum 41(4): 607–616.

Kanyesigye, F. 2012. “Kagame Receives U.S. Global Aids Coordinator”, The New Times,16 October 2012, available at http://allafrica.com/stories/201210160094.html (accessedon March 23, 2013).

Kaufmann, D., A. Kraay, and M. Mastruzzi. 2009. “Governance Matters VIII: Aggre-gate and Individual Governance Indicators, 1996–2008.” World Bank Policy ResearchWorking Paper, WPS No. 4978.

Kinder, M. 2010. “Winning Hearts and Minds in Pakistan, One Humanitarian ReliefWorker at a Time.” Center for Global Development, available at http://blogs.cgdev.org/mca-monitor/2010/10/winning-hearts-and-minds-in-pakistan-one-humanitarian-relief-worker-at-a-time.php (accessed on March 23, 2013).


Keohane, R. O. and J. S. Nye. 1977. Power and Interdependence. First ed. Boston, MA:Little, Brown.

Kaiser Family Foundation. 2011. “The U.S. Presidents Emergency Plan for AIDSRelief (PEPFAR),” available at http://www.kff.org/globalhealth/upload/8002-03.pdf(accessed on March 23, 2013).

Kolb, B. and I. Q. Whishaw. 2008. Fundamentals of Human Neuropsychology. Sixth ed.New York, NY: Worth Publishers.

Kurlantzick, J. 2007. Charm Offensive: How China’s Soft Power is Transforming theWorld? New Haven, CT: Yale University Press.

Kuziemko, I. and E. Werker. 2006. “How much is a Seat on the Security Council Worth?Foreign Aid and Bribery at the United Nations.” Journal of Political Economy 114(5):905–930.

Lake, D. A. 2007. “Escape from the State of Nature: Authority and Hierarchy in WorldPolitics.” International Security 32(1): 47–79.

Levine, R. 2007. Case Studies in Global Health: Millions Saved. Sudbury, MA: Jones andBartlett.

Lindsay, J. M. 2011. “We Don’t Give Out Foreign Aid to Make People Like Us.”The Atlantic, September 29, available at http://www.theatlantic.com/international/ar-chive/2011/09/we-dont-give-out-foreign-aid-to-make-people-like-us/245856/ (accessedon March 23, 2013).

Lyerla, R., C. S. Murrill, P. D. Ghys, J. M. Calleja-Garcia, K. M. DeCock. 2012. “TheUse of Epidemiological Data to Inform the PEPFAR Response.” Journal of AcquiredImmune Deficiency Syndromes 60(Suppl. 3): S57–S62.

Mekasha, T. and F. Tarp. 2013. “Aid and Growth: What Meta-Analysis Reveals.” TheJournal of Development Studies 49(4): 564–583.

Milner, H. V. and D. H. Tingley. 2011. “Who Supports Global Economic Engagement?The Sources of Preferences in American Foreign Economic Policy.” International Orga-nization 65(1): 37–68.

Milner, H. V. and D. H. Tingley. 2013. Geopolitics of Foreign Aid. Northhampton, MA:Edward Elgar Publishing.

Minoiu, C. and S. Reddy. 2010. “Development Aid and Economic Growth: A PositiveLong-run Relation.” The Quarterly Review of Economics and Finance 50(1): 27–39.

Mishra, P. and D. Newhouse. 2009. “Does Health Aid Matter?” Journal of Health Eco-nomics 28(4): 855–872.

Nowak-Lehmann, F., I. Martınez-Zarzoso, S. Klasen and D. Herzer. 2009. “Aid and Trade:A Donors Perspective.” The Journal of Development Studies 45(7): 1184–1202.

Nunnenkamp, P. and H. Ohler. 2011. “Throwing Foreign Aid at HIV/AIDS in DevelopingCountries: Missing the Target?” World Development 39(10): 1704–1723.

Nye, J. S. 2004. Soft Power: The Means to Success in World Politics. New York, NY:Public Affairs.

PEPFAR. 2012. “The U.S. Presidents Emergency Plan for AIDS Relief (PEPFAR) Brand-ing Guidance.” The U.S. Presidents Emergency Plan for AIDS Relief, available athttp://www. pepfar.gov/documents/organization/198963.pdf (accessed on March 23,2013).

Perlez, J. 2012. “China Pledges $20 Billion in Loans to African Nations.” The New YorkTimes. June 19, available at http://www.nytimes.com/2012/07/20/world/asia/china-pledges-20-billion-in-loans-to-african-nations.html?hp (accessed on March 23, 2013).

Program on International Policy Attitudes (PIPA). 2009. “Obama Rockets to Top of Pollon Global Leaders: Putin and Ahmadinejad Receive Lowest Marks.” June 29, availableat http://hdl.handle.net/1903/10689 (accessed on March 23, 2013).

Radelet, S. 2006. “A Primer on Foreign Aid.” Center for Global Development, WorkingPaper Number 92, available at http://www.cgdev.org/publication/primer-foreign-aid-working-paper-92 (accessed on September 24, 2013).

Rajan, R. and A. Subramanian. 2008. “Aid and Growth: What Does the Cross-CountryEvidence Really Show?” Review of Economics and Statistics 90(4): 643–665.


Ros-Lehtinenin, Ileana. 2008. Congressional Record, 110th Congress, 2nd sess., vo1. 154,Issue 51.

Rosenbaum, P. R. 2010. Design of Observational Studies. New York, NY: Springer.Sachs, J. 2005. The End of Poverty: Economic Possibilities for Our Time. New York, NY:

Penguin Press.Sambira, J. 2013. “New HIV infections are falling dramatically in Africa”. The Botswana

Gazette. January 16, available at http://www.gazettebw.com/index.php?option=comcontent&view=article&id=16027:new-hiv-infections-are-falling-dramatically-in-africa&catid=18:headlines&Itemid=2 (accessed on January 17, 2013).

Sargan, J. D. 1958. “The Estimation of Economic Relationships Using Instrumental Vari-ables.” Econometrica 26(3): 393–415.

Sepulveda, J., C. Carpenter, J. Curran, W. Holzemer, H. Smits, K. Scott, and M.Orza. 2007. PEPFAR Implementation: Progress and Promise. Washington, D.C.: TheNational Academies Press.

Shambaugh, D. 2008. “Chinas Role in Asia: Implications for the United States.” KoreanJournal of Defense Analysis 20(4): 295–304.

Simonds, R. J., Constance A. Carrino, and Michele Moloney-Kitts. 2012. “Lessons FromThe President’s Emergency Plan For AIDS Relief: From Quick Ramp-Up to The Roleof Strategic Partnership.” Health Affairs 31(7): 1397–1405.

Stock, J. H., J. H. Wright, and M. Yogo. 2002. “A Survey of Weak Instruments and WeakIdentification in Generalized Method of Moments.” Journal of Business and EconomicStatistics 20(4): 518–529.

Sunday Standard. 2011. “Global AIDS Ambassador visits Botswana.” The Sunday Stan-dard. March 10, available at http://www.sundaystandard.info/article.php?NewsID=10173&GroupID=1 (accessed on March 23, 2013).

Teorell, J., Marcus Samanni, Soren Holmberg and Bo Rothstein. 2011. “The Quality ofGovernment Dataset, Version 6 Apr 2011.” University of Gothenburg: The Quality ofGovernment Institute, available at http://www.qog.pol.gu.se.

Teorell, J. and Hadenius, A. 2005. “Determinants of Democratization: Taking Stock of theLarge-N Evidence.” Mimeo, Department of Government, Uppsala University.

Terror Free for Tomorrow. 2005. A Dramatic Change of Public Opinion in the MuslimWorld. Washington, D.C.: Terror Free for Tomorrow.

Terror Free for Tomorrow. 2006. Humanitarian Assistance Key to Favorable Public Opin-ion in Worlds Three Most Populous Muslim Countries. Washington, D.C.: Terror Freefor Tomorrow.

Times of Zambia. 2012a. “Bush Pledges More Help.” Times of Zambia. July 4, availableat http://www.times.co.zm/?p=3344 (accessed on January 09, 2013).

Times of Zambia. 2012b. “Lets Hope for More Bush Visits.”, Times of Zambia. July 3,available at http://www.times.co.zm/?p=3214 (accessed on January 09, 2013).

Tortora, R. D., R. Srinivasan and N. Esipova. 2010. “The Gallup World Poll.” In SurveyMethods in Multinational, Multiregional and Multicultural Contexts, J. A. Harkness,M. Braun, B. Edwards, T. P. Johnson, L. Lyberg, P. Ph. Mohler, B.-E. Pennell, andT. W. Smith, (eds.), New York, NY: Wiley, pp. 535–543.

Vo, L. T. 2008. “Vietnam Battles HIV Epidemic With U.S. Aid”. The Wall Street Jour-nal, Asia Edition. December 26, available at http://online.wsj.com/article/SB123014126494733205.html (accessed on March 23, 2013).

Vreeland, J. 2011. “Foreign Aid and Global Governance: Buying Bretton Woods — TheSwiss-bloc Case.” The Review of International Organizations 6(3, 4): 369–391.

Walsh, D. 2012. “Craving Energy and Glory, Pakistan Revels in Boast of Water-RunCar.” The New York Times. August 4, available at http://www.nytimes.com/2012/08/05/world/asia/boast-of-water-run-car-thrills-pakistan.html?ref=world (accessed onMarch 23, 2013).

Wang, T. Y. 1999. “US Foreign Aid and UN Voting: An Analysis of Important Issues.”International Studies Quarterly 43(1): 199–210.


Wike, R. 2012. “Does Humanitarian Aid Improve America’s Image?” Pew Research GlobalAttitude Project. A report released on March 6, 2012.

Wooldridge, J. M. 1995. “Score Diagnostics for Linear Models Estimated by Two StageLeast Squares.” In Advances in Econometrics and Quantitative Economics: Essays inHonor of Professor C. R. Rao, G. S. Maddala, P. C. B. Phillips, and T. N. Srinivasan,(eds.), Oxford, UK: Blackwell, pp. 66–87.

Wright, J. 2010. “Aid Effectiveness and the Politics of Personalism.” Comparative PoliticalStudies 43(6): 735–762.

Documents

Doing Well by Doing Good: The Impact of Foreign Aid on ...horiuchi.org/www/Publications_files/100.00013036.pdf · Doing Well by Doing Good: The Impact of Foreign Aid on Foreign Public