OUP user menu

Does Corruption Cause Aid Fatigue? Public Opinion and the Aid-Corruption Paradox

Monika Bauhr, Nicholas Charron, Naghmeh Nasiritousi
DOI: http://dx.doi.org/10.1111/isqu.12025 568-579 First published online: 1 September 2013

Abstract

Does perceived corruption in recipient countries reduce support for foreign aid in donor countries? This under-explored yet salient question is examined using the 2009 Eurobarometer survey for the 27 EU countries. We suggest that perceived corruption can cause aid fatigue but that this relationship is highly contextualized. The results show that perceptions about corruption in developing countries reduce overall support for aid among respondents in donor countries. However, this effect is mitigated by country and contextual-level effects and different understandings of what we call the “aid-corruption paradox,” namely that the need for foreign aid is often the greatest in corrupt environments. Three different dynamics of the aid-corruption paradox influence support for aid: moral, pragmatic, and strategic understandings. In EU-15 countries, the effect of perceived corruption in recipient states on aid fatigue can be substantially altered if aid is motivated by moral reasons for helping poor countries or if the purpose of aid is understood to improve governance. In new member states (NMS-12), the effect is reduced if respondents believe that the result of aid can serve national interests. The results provide new insights into the public opinion/development policy nexus, which suggest a number of salient policy recommendations and future areas for research.

Foreign aid is at a crossroad. At a time when new aid money is in high demand, there is a debate about the so-called aid fatigue in donor countries. Two seemingly irreconcilable perspectives can be found in debates about foreign aid. While aid is at times portrayed as the key to addressing poverty and solving common problems such as climate change, others maintain that aid is ineffective and that it fuels corruption (Easterly 2006; Djankov, Montalvo, and Reynal-Querol 2008; Moyo 2009). Thus, on the one hand, the demand for foreign aid is on the rise as countries seek to meet the Millennium Development Goals, and donors agree on unprecedented increases in aid to fund climate change projects. On the other hand, however, several countries are already falling short on their aid commitments (OECD 2010; Hulme 2011) and public support for foreign aid risks slipping.

Because foreign aid has to compete with other policy areas from a finite national budget, public support for foreign aid is considered critical for strong aid programs (Lumsdaine 1993). Without solid support for foreign aid among taxpayers and voters, it will be difficult to meet and sustain the new aid commitments (Paxton and Knack 2012). Moreover, aid legitimacy can influence the quality of aid. According to Collier (2007:183), “the key obstacle to reforming aid is public opinion,” meaning that the necessary reforms to make aid more effective can only come about if public opinion supports strategies that allow policymakers see long-term progress and take risks, rather than aiming for quick photo opportunities that enhance short-term domestic electoral support.

Nevertheless, we lack an understanding of why different publics hold varying levels of support for foreign aid. According to Henson, Lindstrom, and Haddad (2010:46), “the uncomfortable fact is that we still know relatively little about the factors determining public support for aid to developing countries, and indeed even how we might reliably monitor such attitudes over time.” Since foreign aid depends on a supportive public opinion in the long term (Mosley 1985; OECD 2003; Henson et al. 2010), it is critical to understand the factors that can influence the public's attitudes and opinions on aid.

This study analyzes the influence of perceived corruption in recipient countries on support for foreign aid.2 We suggest that corruption can cause aid fatigue but that public reactions to corruption in aid substantially depend on fundamental beliefs about the role of foreign aid. Three different types of understandings of the role of foreign aid that influences support for development assistance are identified: moral, pragmatic, and strategic understandings. The analysis thus investigates how contextual factors and different discourses on what aid can achieve can contribute to explaining the relationship between perceptions of corruption and support for foreign aid in donor countries.

The study thereby makes a number of contributions to the literature. First, we seek to make a theoretical and empirical contribution to the question of how public opinion on foreign aid can be affected by perceptions of corruption. The two reasons frequently offered for why support for foreign aid can slip are economic malaise in the donor country and perceived corruption in the recipient country. However, the literature on foreign aid is predominately focused on the influence of domestic factors on support for foreign aid (Mosley 1985; Chong and Gradstein 2008; Paxton and Knack 2012). In contrast, we test the hypothesis that attitudes about corruption and governance problems in recipient countries lead to less support for aid using a comprehensive study (2009 Eurobarometer data) to explore how domestic and international factors affect individuals’ perceptions of development aid.

Moreover, this analysis also explores what we call the “aid-corruption paradox,” namely that the need for foreign aid is often greatest in corrupt environments. Usually, general explanations such as meager development results and the perception that taxpayers’ money is being wasted are offered to account for weakened support for foreign aid in some countries (Wittkopf 1990; Dearden and Mira Salama 2002; Lahiri and Raimondos-Møller 2004; Ali and Isse 2005). However, such answers are insufficient as they do not offer an understanding of when and how the perceived lack of effectiveness in aid leads to aid fatigue. Because corruption scandals and weak performance of foreign aid is not a rarity (Boone 1996; Svensson 2000; Knack 2001; Easterly 2006; Djankov et al. 2008; Moyo 2009), general claims that corruption in developing countries gives rise to aid fatigue provide little understanding about how important corruption may be in explaining aid fatigue.

Furthermore, previous studies have not provided an understanding of how different framings of the aid debate influence support for foreign aid. Relevant to this study, the international anticorruption regime has exposed corruption to an extent not previously experienced in history. Public reactions to this “corruption epidemic” have forced policymakers and diplomats to reconsider previously uncontroversial foreign policy practices (Leiken 1996–1997). However, we know little about the wider effects of this exposure and whether it may have detrimental effects on support for foreign aid, and in turn on important international efforts to promote sustainable development.

We proceed as follows. The first part presents the literature on the relationship between corruption and aid fatigue, why corruption in donor and recipient countries may influence support for foreign aid, and how different understandings of the aid-corruption paradox may impact on this relationship. The second part presents our data and measurements. The third part reports our results, showing that corruption in recipient countries significantly influences aid fatigue but that this effect is mediated by important conditions pertaining to the fundamental motives for foreign aid and country-specific contexts. Furthermore, a strong divide in attitudes toward aid between EU-15 countries and new member states (NMS)-12 is found. The fourth part concludes with a summary of the results and their relevance for future research.

Corruption and Aid Fatigue

Foreign aid has long been a fervently debated topic. While a large segment of the literature focuses on how aid money is used and its effects on recipient countries, we have limited knowledge on what determines support for foreign aid (see, for example, Boone 1996; Alesina and Dollar 2000; Burnside and Dollar 2000; Svensson 2000; Knack 2001; Alesina and Weder 2002; Easterly 2006; Djankov et al. 2008). Although aid agencies have long sought to understand what factors influence public opinion on foreign aid, the academic literature has been slow to follow up on this. The studies that do focus on the public opinion aspect of foreign aid dwell largely on the topic of “aid fatigue.” This term was used to explain the drops in foreign aid for much of the 1990s (Erixon and Sally 2006). Several authors linked the growing cynicism toward foreign aid at the time to meager development results and corruption in recipient countries (Wittkopf 1990; Lahiri and Raimondos-Møller 2004). However, there are surprisingly few empirical studies on how corruption influences support for foreign aid.3

We are interested in what drives aid fatigue in donor countries, and thus, we elect to focus the analysis at the micro-level. The term “aid fatigue” is used here in a broader sense to understand how citizens’ opinions of foreign aid may be influenced by perceptions of corruption. Pointing out corruption as a reason for reducing support for foreign aid may appear straightforward. However, this fails to account for the long history of foreign aid, where corruption and inefficiencies have in fact been continuously present in recipient countries (Boone 1996; Svensson 2000; Knack 2001; Easterly 2006; Djankov et al. 2008; Moyo 2009) while support for foreign aid has been high. This means that perceptions of recipient countries being corrupt in general does not necessarily explain changes in public opinion about foreign aid in donor countries. In other words, there is a surprising lack of focus in the literature on how recipient country corruption affects support for foreign aid.

Moreover, much of our current knowledge about how recipient country corruption affects aid is built on studies that show that concern over corruption has thus far not had a significant effect on actual aid disbursements to corrupt countries (Alesina and Dollar 2000; Alesina and Weder 2002; Chong and Gradstein 2008). Chong and Gradstein (2008:12), for example, conclude that “recipient country characteristics do not seem to affect the amounts of aid.” These results show that the selectivity approach—meaning that aid should be targeted at countries with relatively high quality of government to achieve better results (see, for example, Brautigam 2000; Burnside and Dollar 2000)—is not widely used in practice. However, the studies do not consider how recipient country corruption could affect public support for development aid with resulting effects on total aid levels.

In other words, showing that recipient country corruption does not appear to affect actual aid levels to corrupt countries fails to take into consideration that perceptions of corruption in recipient countries may keep down overall aid budgets and that public support for increasing foreign aid may significantly fall as a result of perceptions of corruption in foreign aid. This latter point is important as it highlights the intricacies of foreign aid. On the one hand, foreign aid is considered to be a tool of foreign policy and as such be driven mainly by political and strategic considerations (Mckinlay and Little 1977; Hook 1995; Palmer, Wohlander, and Clifton Morgan 2002). This implies that donor states would provide foreign aid to corrupt countries as long as it is in their interest to do so. On the other hand, a growing body of research shows that foreign aid is also determined by domestic politics, including considerations for public opinion (Ruttan 1996; Rioux and van Belle 2005; Lancaster 2006; Milner and Tingley 2010). The implication from this is that while the public may not hold much sway over who the recipients of foreign aid should be, they may impact on the overall aid levels. Lumsdaine's (1993:138) historical analysis of the foreign aid policy of 18 developed democracies, for example, finds that “Those countries with relatively weak public support were relatively slack aid donors, and countries with particularly high levels of support seem to have expanded their aid programs rapidly.” Moreover, Otter (2003) provides evidence that shifts in public support for aid over time in the United States (a decrease) and Denmark (an increase) corresponded with changes in actual foreign aid–spending levels (Otter 2003:117).4

Thus, the study of public opinion of foreign aid is not only important for understanding attitudes of the electorate on development issues, but also because of its policy effects. While policymakers attempt to influence public opinion, public opinion can also force governments to revise its foreign aid policy—especially when the media highlights scandals involving the misuse of aid funds (Lancaster 2006). Public opinion is also important for public policies’ legitimacy and effectiveness (Holsti 2004). The study of public opinion of foreign aid can thus provide insights into how the policy options of decision makers can be constrained, with implications for meeting future aid targets.

In this study, we build on the foreign policy literature that finds that while the general public tends to be ill-informed about international affairs, they hold relatively structured opinions about foreign policy (Shapiro and Page 1988; Holsti 1992). According to the political heuristics approach, citizens do not need to be knowledgeable about an issue to hold fairly reliable political opinions (Sniderman, Brody, and Tetlock 1991). Thus, while the general public knows little about the size of the aid budget or how that money is spent, they tend to have clear opinions about their support for foreign aid. What is less known, however, is how perceptions of corruption in recipient countries affects opinions regarding support for foreign aid. It is likely that these perceptions are influenced by a proliferation of reporting on the issue.5 While the theoretical literature gives insights into the possible effects of corruption on support for aid, these accounts tend to be general and untested.

Perhaps the strongest reason for why corruption could cause aid fatigue is that corruption can be perceived as an external impediment to the effectiveness of aid. Concern for aid effectiveness can be expected to be an important concern both for “cooperative internationalists” (Wittkopf 1990; Chanley 1999; Paxton and Knack 2012), who are highly interested and concerned with international development issues, and those who are primarily concerned with how aid competes with the use of tax payers’ money for important domestic service provision. Because there has been a discourse shift surrounding corruption that delegitimizes the practice and places the focus on the inefficiencies of corruption, we can expect that citizens now to a greater extent view corruption as an external impediment to the effectiveness of aid. We therefore formulate our first hypothesis as follows:

Hypothesis 1: Respondents who believe corruption is a primary problem in recipient countries are more likely to express aid fatigue, ceteris paribus.

However, despite theoretical arguments for why perceptions of corruption would affect opinions about foreign aid, we argue that the relationship is more nuanced. In fact, studies on public opinion of foreign aid have shown that high support for foreign aid can coincide with high concern for corruption (Riddell 2007; Abrahamsson and Ekengren 2010; Henson et al. 2010). Moreover, there is a discrepancy in the literature where, on the one hand, the aid fatigue literature singles out corruption and waste as a potentially important determinant for lessening support for foreign aid, while on the other hand we lack models to understand the causal mechanisms that lead the public to weaken their support for foreign aid because of perceptions of corruption (for example, in which context this relationship is most salient). In the next section, we consider how contextual factors that have been overlooked by previous studies can impact on the relationship between perceptions of corruption and aid fatigue.

Understanding the Aid-Corruption Paradox: Moral, Pragmatic, and Strategic Conceptions

During the extensive history of foreign aid, corruption had for a long time received little attention in diplomatic circles. With the emergence of an international anticorruption regime in the 1990s, significant changes took place in how corruption was discussed and addressed. In both policy and academic circles, corruption became defined as a development challenge (Holmberg, Rothstein, and Nasiritousi 2009). A discussion also arose about whether highly corrupt countries should receive foreign aid, as it may be morally wrong to support corrupt regimes, while at the same time efficiency concerns were raised about transferring resources to corrupt countries (Brautigam 2000; Burnside and Dollar 2000).

These insights from the corruption debate lie at the heart of the aid-corruption paradox, which can be summed up as follows—while corruption undermines the rationale for foreign aid, the need for foreign aid is often the greatest in corrupt environments. For example, at the bottom of the 2010 Corruption Perceptions Index are four countries that all require substantial foreign aid for their development: Somalia, Myanmar, Afghanistan, and Iraq (Transparency International 2010). Thus, while the aid money may have been used more effectively if channeled to less-corrupt countries, the fact remains that the need for foreign aid is often greater in more corrupt countries. This paradox raises a number of interesting questions, not least: How do tax payers in donor countries evaluate these opposing propositions? Surprisingly, this relationship is largely unexplored both theoretically and empirically in the literature; thus, we attempt to make a first step in contributing to a better understanding of the complex relationship between corruption and support for aid.

Although we predict that attitudes toward corruption in recipient countries will negatively impact aid support all things being equal (H1), we suggest there are several contexts in which the said negative relationship might be offset. We argue that the relationship is highly contextualized and either mediated or reversed depending on how citizens understand the aid-corruption paradox. In particular, we identify three different types of understandings of the how the relationship in H1 could be mediated: by moral, pragmatic, and strategic understandings. Below, we highlight three potential intervening factors to H1, which we argue serve as various proxies for understanding the aid-corruption paradox.

The first can be clearly drawn from previous research on the drivers of support for foreign aid, which has found that one of the main arguments in favor of providing aid to poor countries is a moral duty to help poor areas (Riddell 2007; Henson et al. 2010). Andreoni (1990) explains such charitable action as originating from “social pressure, guilt, sympathy, or simply a desire for a ‘warm glow’.” The traditional picture of foreign aid has been of money spent on humanitarian missions to alleviate poverty and hunger. Such missions often have a positive connotation, as aid is seen as an altruistic undertaking. In the context of this study, such attitudes reflect an undertaking that despite high levels of corruption in a recipient country, people who identify with having a moral duty to help those less fortunate have an implicit understanding that it is often the most corrupt countries that need the most assistance. This understanding of the aid-corruption paradox is thus associated with the view that there is a moral duty to help the poor—even in countries suffering from corruption.

Hypothesis 2: Aid fatigue resulting from the belief that corruption is a primary problem in recipient countries will be offset by a moral duty to help poor countries.

A second explanation for why corruption would not cause aid fatigue is based on more pragmatic reasoning—the possibility that people support aid to corrupt contexts if aid is seen to prevent and reduce corruption. People who view corruption as a natural part of development that can be fought through international development assistance are less likely to be of the view that aid should be reduced given that a recipient country is corrupt. Thus, the international anticorruption regime could in this way promote the idea that aid is needed the most in corrupt countries and thereby increase the support for targeting aid to countries suffering under corruption. In fact, many aid agencies have now dedicated a greater share of their budgets to institution building in developing countries, and this share is expected to rise (World Bank 2011). Recent studies show that targeted efforts may indeed lead to institutional reform in developing countries (Heckelman 2010; Scott and Steele 2011). The opinion that aid can be used to improve governance is thus a way to understand and deal with the aid-corruption paradox and therefore expected to play a mediating role in how corruption affects support for foreign aid, such that the negative effect of corruption on support for aid is less among those who believe that aid can be used to improve good governance in recipient countries.

Hypothesis 3: Aid fatigue resulting from the belief that corruption is a primary problem in recipient countries will be offset by an understanding that foreign development aid is aimed at improving governance.

Finally, aid support can also be motivated by purely strategic reasons (Cuervo-Cazurra 2006; Riddell 2007; Henson et al. 2010). These include national strategic concerns, such as security relationships, access to raw materials, or the expansion of export markets. “Strategic thinkers” may not necessarily lessen their support for aid in the face of corruption scandals in recipient countries, but understand that corruption might be a necessary evil to “grease the wheels of commerce” so to speak. It is possible that such respondents in fact are more prone to support foreign aid when they believe recipient countries are corrupt because they believe it is easier to do business, or get around certain labor and environmental laws, “red tape,” etc. Alternatively, strategic thinking may also lead to the conclusion that highly corrupt countries need support in order to prevent instability that may spill over across borders. This perspective would in other words mean that citizens of donor countries understand the aid-corruption paradox in a strategic sense and are willing to provide aid to poor countries in spite of, or possibly due to, their corruption levels.

Hypothesis 4: Aid fatigue resulting from the belief that corruption is a primary problem in recipient countries will be offset by strategic thinking about the benefits with a continued interaction with recipient countries.

According to these perspectives on the fundamental motivations for foreign aid, the effect of corruption on foreign aid is nuanced by different understandings of the aid-corruption paradox. The next question is then, are some of these understandings more common in certain countries? While this topic is still relatively unexplored in the literature, we propose that certain understandings of the aid-corruption paradox are more common in some countries than others depending on different experiences of domestic corruption, different policy framings of foreign aid based on domestic factors (such as own country's economic development), and aid traditions in the country (such as colonial ties, size of aid budget, etc). We thus expect these attitudes to vary across countries.

One way in which attitudes toward corruption in foreign aid may differ across states is through the effect of domestic corruption in donor countries. Anderson (1998) argues that citizens usually use domestic politics as a proxy for European politics, and it is therefore possible that opinions about foreign aid are formed based on domestic conditions. Chong and Gradstein (2008), for example, find that citizens of donor countries that perceive domestic institutions to be inefficient express a lower support for foreign aid. The argument given is that citizens who see their tax money being wasted by their own institutions do not trust that the country's development assistance will fare any better.

Perceptions of domestic corruption could however also have other effects. Recent studies show that “cleaner” countries tend to be much more critical of corruption abroad, while countries that are themselves more corrupt on average are much more tolerant and at times even more willing to invest or do business with other corrupt countries. For example, Cuervo-Cazurra (2006) finds strong empirical evidence that corrupt countries receive relatively less FDI from states that have signed and ratified OECD's “Convention on Combating Bribery of Foreign Public Officials in International Business Transactions” while corrupt countries receive relatively more FDI from other countries with high levels of corruption themselves. On an individual level in European countries, the World Value Survey points to significantly different levels of tolerance for certain types of corrupt behaviors depending on the level of corruption in their own country.6

Countries also vary in their tolerance of corruption abroad. The 2002 “Voice of the People” Gallup poll shows, for example, that the Japanese were relatively tolerant to corruption in aid (18% of the Japanese sample favored aid only to noncorrupt contexts), while as much as 53% of the US population only supported aid to noncorrupt contexts (Paxton and Knack 2012). One explanation for this difference could perhaps be how aid delivery to corrupt countries has been framed in different states. The varying figures could, for instance, reflect a difference in opinion about whether the best way to drive reform in corrupt countries is through providing aid or through withholding it. Here, the discussion about one of the most widely used ways to deliver aid in corrupt countries—the practice of placing conditions on aid, known as conditionality—could have had an impact. Conditionality is presented as a viable option for delivering aid to high-corruption countries by some, but has also received widespread criticism for being ineffective (Mosley, Harrigan, and Toye 1995; Collier 1997; White and Morrissey 1997; Dreher 2004).

Thus, attitudes can be shaped by how foreign policy has been framed and how the purpose of aid has been defined. For instance, many of the new EU member states (NMS-12) have recently made the turn from a recipient country to a donor country and have very limited aid budgets. According to Lightfoot (2008), NMS-12 countries have adopted a more strategic approach to aid than some of the other EU states. The old EU countries have, in contrast, developed close development cooperation ties to countries not in their immediate neighborhood, and colonial history in several of the old EU countries means that there is likely to be a different framing of development issues in these countries. Thus, different policy framings are one important factor in influencing how the aid-corruption paradox is understood across countries. Therefore, we expect systematic differences in attitudes toward aid between different countries. We explore these propositions below.

Data, Sample, and Methods

To analyze whether attitudes toward corruption affect support for development aid, we use data from the 2009 Eurobarometer that explores European perceptions of development issues. The lack of cross-country empirical analyses of how recipient country corruption affects support for foreign aid could be due to difficulties in finding appropriate measurements. While country-level analyses exist (Riddell 2007; Henson et al. 2010; TNS 2010), cross-country data are more scarce. The use of the 2009 Eurobarometer data allows us to avoid some of the data limitations of previous studies by looking at both internal and external factors for aid fatigue. However, the disadvantages are primarily threefold. One is that the questions do not particularly focus on the issue of corruption and only covers European donor countries, thus leaving out donors such as the United States, Japan, and Australia. Nevertheless, together the countries make up around 50% of total development assistance (OECD 2010), and the variety in the countries included in terms of foreign aid budgets, economic situation, and cultures makes the data useful. Two, the data are cross-sectional, and therefore, the results should be viewed as a first attempt to study different understandings of foreign aid with further studies required to corroborate the findings. Finally, the question focuses on EU aid and not national aid. However, the 2002 Eurobarometer survey poses both questions about support for national aid and EU aid. Here, the correlation between the two is 0.76, indicating that opinions about aid are largely similar, no matter who distributes it (Eurobarometer 2002).

For reasons of comparability with previous studies, we follow the standard practice in the literature in creating our dependent variable (AID FATIGUE). In the most recent Eurobarometer survey, we employ the question that best proxies for what we would ideally like to capture, namely support for development aid. Specifically, the dependent variable comes from the following question: “Would you say that the current level of European Union's contribution to development is too big, too small or about right?” We are concerned with support for development aid, which we take to mean an opinion that expresses either content with the status quo or a desire to increase the present level of aid. A lack of support would thus be a desire to cut back current levels. Thus, although the question is coded into three responses, we condense this variable into a dichotomous variable, where “1” represents the answer “too big” and “0” if otherwise, that is, “too small or about right.7AID FATIGUE is available for all EU countries and has a total of 19,977 individual responses. The dependent variable is dichotomous, and thus, we use logit regression analysis.

To test H1, we operationalize whether concern about poor governance in developing countries is related to the wish to reduce the EU's development aid budget. Thus, our primary independent variable used to test this (QoG CHALLENGE) is taken from the question “In your opinion what are the two biggest challenges currently facing developing countries?” Among the response categories, such as environmental issues and poverty, “poor governance” is listed. We take “poor governance” in developing countries as the closest proxy we can find to corruption—as the two concepts, while indeed separate, are often used together quite closely in the literature on quality of government and in media-broadcast accounts of developing countries8. Thus, QoG CHALLENGE is coded dichotomously—”0” if Poor governance not mentioned; “1” if poor governance is mentioned by a respondent.9

Key Contextual Variables and Interaction Effects

To test H2, we use a proxy for altruism aimed at poorer countries. In the 2009 Eurobarometer survey, we take the question “In your opinion, is it very important, fairly important, not very important or not at all important to help people in developing countries?” (HELP POOR), and the response ranges from “0” to “3,” from “not at all important” to “very important.” Next, to test H3, we employ a question in which respondents were asked to list the top two issues to which development aid was aimed, among things like gender equality, environmental issues, conflict resolution, etc. Those who selected “governance” as one of the top two issues are coded as “1” and “0” if otherwise (QoG FOCUS). Finally, to measure a degree of strategic thinking in a respondent (H4), we employ the question “In your opinion, which of the following are the two main motivations for richer countries to provide development aid to poor countries?” If a respondent chose among the possible answers “Self-interest, for example, helping poor countries trade will enable them to buy more products from rich countries,” then they are coded as “1” and “0” if otherwise (STRATEGIC).

Individual Level Controls “i

We control for several additional factors at the individual level, building on previous empirical studies which have shown significant determinants of AID FATIGUE or support for aid (for example, see Chong and Gradstein 2008; Paxton and Knack 2012). First, we control for demographic factors such as gender, age, level of education, self-perception of economic standing in society (“1–10,” lowest to highest standing), and whether a respondent comes from a rural or urban area (1–3, rural, moderate, urban). Next, we include several relevant opinions of the respondents we expect to be salient factors. First, where a respondent identifies themselves on the “left–right” political spectrum is expected to play a significant role, with people on the left being more likely on average to support foreign development aid (1–10, left-to-right). Second, we include whether a respondent has a friend in a foreign country or not, as one would expect more consideration of those living in other countries if this is the case, thus increasing likelihood of support for aid. Third, we include a proxy for public awareness of development issues by using the question of whether or not a respondent has heard of the Millennium Goals—“1” Yes, and you know what it is; “2” Yes, but you do not really know what it is; “3” No. Fourth, to control for the level of interest in development issues, we use the question of whether the respondent believes the media covers issues associated with development in foreign countries enough or not (1–3, too much, just enough, or too little). Finally, how well the respondent views the effectiveness of the public administration of his or her own country could influence perceptions about corruption. Thus, we include a measure for this (1–4, 1 Very good; 2 Rather good; 3 Rather bad; Very bad). A full description of the variables in the model along with summary statistics is found in the appendix of this study.

Country Level Controls “j”

Since the countries in the sample are all EU countries, many geographic and institutional factors are naturally controlled for; thus, we elect to keep this level parsimonious. First, we include country dummies in the logit model to control for fixed effects and run country dummies in the multilevel model at later stages. Second, because we expect that for historical and/or contextual reasons, attitudes between the new member states and the EU-15 countries might be different on average, we control for whether a country is an EU-15 country or not (1/0). Third, whether a country had colonies or not might play a role in individual's opinion about development aid; thus, we control for this factor as well (1/0). Finally a country's level of economic development is controlled for (GDP per capita, logged).

We also check whether there are systematic differences in individual responses based on their country of origin. We find that countries in the sample vary significantly with respect to both opinions about AID FATIGUE and QoG CHALLENGE. We perform a simple test, running country dummies in a logit regression, and find wide and significant variation between countries with respect to the dependent variable.10 We thus must take into account the nested nature of the data. There are two ways in which we approach this issue. First, country fixed effects are used in several logit models with robust standard errors. Second, we run a multilevel statistical model, taking into account that individual observations are nested in country-level variables. We report the results for both sets of models in the next section.

The multilevel model to test H1 thus uses the following formula:

Aid Fatigueij = αij + β1 (QoG Challenge GEij) + βN (Individual Controlsij) + βZ (Country Controlsj) + eij + vj

Results

Does attention to corruption and governance issues in developing countries influence individual attitudes in donor countries toward development aid (for example, H1)? In Table 1, we find this indeed to be the case.11 Model 1 shows the simple baseline effects, controlling only for random variance at the country level in which individuals are nested.12 We find that the impact of perceptions of corruption being a key challenge in recipient countries increases the odds of a respondent expressing AID FATIGUE by 39%, and the estimate is positive and significant at the 99% level of confidence. In addition, we find that the standard deviation among country-wide variation is significant at the 0.01 level. Model 2 includes demographic controls. We find that being female, more educated, and a belief that one is of a relatively high economic standing in society all significantly decrease the odds of AID FATIGUE. For example, the odds of a female expressing AID FATIGUE are 17% less than a male respondent, holding all other variables constant.

View this table:
1

Multilevel Model of QoG Challenge and Aid Support: Individual and Country Effects

1234
Fixed part
QoG Challenge (0/1)1.39 (0.000)***1.42 (0.000)***1.47 (0.000)***1.46 (0.000)***
(I) Demographic Controls
Female0.83 (0.001)***0.81 (0.003)***0.82 (0.002)***
Age1.00 (0.96)1.00 (0.12)0.99 (0.09)*
Education0.95 (0.000)***0.97 (0.003)***0.98 (0.003)***
Rural–Urban1.02 (0.52)1.07 (0.08)*1.08 (0.08)*
Economic Standing0.91 (0.000)***0.93 (0.001)***0.93 (0.001)***
(II) Other Individual-Level Controls
Left–Right Placement1.07 (0.000)***1.07 (0.000)***
Heard Millennium Goals1.14 (0.03)**1.14 (0.02)**
Foreign Friend1.23 (0.003)***1.24 (0.002)***
Media Coverage0.24 (0.000)***0.24 (0.000)***
Public Admin. Effectiveness1.31 (0.000)***1.32 (0.000)***
(III) Country-Level Controls
EU-150.93 (0.86)
Fmr. Colonies1.11 (0.69)
GDPp.c. (logged)10.94 (0.01)***
Random Part
Country level ∂0.67***0.69***0.65***0.51***
Individual level obs.19,97717,48313,66313,663
countries27272727
Log Likelihood−5234.78−4928.14−3366.91−3360.74
  • (Notes. Dependent variable is AID FATIGUE (0/1), Odds ratios with robust standard errors reported with p-values in parentheses. ***p < .01, **p < .05, *p < .10.)

Model 3 adds further attitudes and political aspects of the respondents, yet the effects of QoG CHALLENGE on AID FATIGUE are robust to additional controls. In this set of covariants, we find that greater self-identification with right-wing politics and greater negative attitudes toward the effectiveness of one's own public administration increase the likelihood of AID FATIGUE in the sample. This latter result confirms the finding in Chong and Gradstein (2008). Conversely, having a foreign friend or an interest in development issues (media awareness) decreases the odds in AID FATIGUE by 19% and 76%, respectively, holding all other variables constant. In model 4, we find that country-level variables play minimal role, as only the level of GDP per capita (logged) is a significant determinant of AID FATIGUE. This points to an interesting dynamic between the individual and country levels of analysis—while respondents who say that they are of a more privileged economic standing are less prone to AID FATIGUE on average, the positive and significant coefficient of GDP per capita (logged) implies that the respondents from the richer countries in the EU are more prone to AID FATIGUE on average. Throughout all four models, however, we find that concerns about corruption (QoG CHALLENGE) significantly increases the odds toward fatigue for foreign development aid, while furthermore, country-level variance is significant in all four models.13

In Table 2, we use the same specifications as model 2 in Table 1 for each country individually and find evidence that the relationship between QOG CHALLENGE and AID FATIGUE varies from country to country. In all, for 21 countries, we find a positive coefficient, or in other words, an increase in the odds of AID FATIGUE. Yet in only seven EU states (26% of the total countries) is this relationship statistically significant—Finland, Belgium, Austria, Spain, Germany, the UK, and Ireland—all of which are EU15 countries, meaning that citizens from these countries who believe that corruption is one of the top two problems facing developing countries are the least likely to support development aid, ceteris paribus. Only in six countries—Italy, Portugal, Poland, Malta, Latvia and Lithuania—do we find that QoG CHALLENGE decreases the odds of AID FATIGUE, yet the relationship is not statistically significant. In none of the new member states (NMS-12), however, do we find that attitudes regarding corruption in recipient countries significantly influence opinions regarding AID FATIGUE. Of the countries that have a significant relationship between these two variables, we find that, for example, in Finland, respondents who answer “yes” to QoG CHALLENGE are twice as likely to express a lack of support for aid on average than those answering “no.” In Austria, the odds of AID FATIGUE are 2.27 greater when a respondent believes corruption is a key problem in recipient states, while the odds of AID FATIGUE increase by 3.38 in Spain, holding the demographic controls constant. As noted, differences in the framing of the aid debate between countries may be one reason for the variation (Lancaster 2006).

View this table:
2

QoG Challenge and Aid Fatigue by Country and Size of Impact

CountryOdds Ratiop-valueObs.
Spain3.38***0.000684
Austria2.27***0.000704
Finland2.04**0.03786
Belgium1.73**0.03824
Ireland1.69*0.08637
Greece1.670.28842
UK1.65***0.01823
Slovenia1.650.16679
Estonia1.550.29683
Czech Rep.1.540.44720
Germany1.52**0.02967
Hungary1.460.37655
Denmark1.360.37793
Bulgaria1.320.51587
Netherlands1.290.21592
France1.260.39702
Sweden1.260.43664
Cyprus1.150.91224
Slovakia1.110.85736
Romania1.010.98582
Luxembourg0.980.96329
Latvia0.890.79690
Malta0.890.83229
Italy0.460.31649
Lithuania0.350.31670
Poland0.260.21563
Portugal0.220.14449
  • (Notes. All models include demographic controls (as in model 2, Table 2) and only QoG Challenge coefficients reported. Each country regressed separately, “Obs” is the # of observations by country. Odds ratios with robust standard errors reported.)

Finally, we turn to the question of the aid-corruption paradox condition. In Table 3, we test H2, H3, and H4. Since we found significant country effects in Table 3, mostly on the lines of EU-15 vs. NMS-12 states, we analyze first the whole sample and then take the EU-15 sample and NMS-12 samples in models 2 and 3, respectively. Given that the results were robust to the inclusion of the political attitudes (model 3 in Table 1), we include only demographic controls to maximize observations. In Table 3, H2, H3, and H4 are tested simultaneously for the whole sample in model 1.14

View this table:
3

Logit Regression with Country Fixed Effects and Predicted Probabilities: Interaction and Split-Sample Effects

All EU27EU 15NMS 12
VariableβOdds ratioβOdds ratioβOdds ratio
QoG Challenge (0/1)0.77 (0.000)***2.160.99 (0.000)***2.71−0.03 (0.94)0.97
Demographic Controls
Female−0.10 (0.09)*0.91−0.05 (0.49)0.950.25 (0.03)**0.80
Age−0.000 (0.68)0.990.000 (0.91)1.00−0.001 (0.73)0.99
Education−0.03 (0.000)***0.96−0.03 (0.000)***0.97−0.03 (0.08)*0.97
Rural-Urban0.07 (0.06)*1.070.06 (0.53)1.050.09 (0.18)1.09
Economic Standing−0.05 (0.007)***0.95−0.06 (0.02)**0.94−0.05 (0.13)0.95
Contextual Effects—Factors of the Aid-Corruption Paradox
HELP POOR (β1) (0–3)−1.14 (0.000)***0.32−1.22 (0.000)***0.29−0.10 (0.000)***0.38
QoG FOCUS (β2) (0/1)−0.24 (0.008)***0.78−0.29 (0.005)***0.74−0.05 (0.79)0.95
STRATEGIC (β3) (0/1)−0.35 (0.07)*1.140.06 (0.54)1.060.26 (0.02)**1.31
QoG Challenge*β1−0.15 (0.08)*−0.26 (0.02)**0.770.25 (0.18)
QoG Challenge*β2−0.43 (0.02)**−0.43 (0.04)**0.65−0.39 (0.32)
QoG Challenge*β3−0.35 (0.02)**−0.21 (0.23)0.81−0.70 (0.02)**
Individual-level obs.17,24310,3556,888
Countries271512
Pseudo Rsq.0.180.210.11
Log Likelihood−4217.35−2781.31−1418.73
  • (Notes. Country dummies are included (not shown). Beta coefficients along with odds ratios reported with p-values from robust standard errors in parentheses. Odds ratios for the interaction effects are calculated separately and found in Table 4. ***p < .01, **p < .05, *p < .10.)

The results show clearly that the effect of QoG CHALLENGE on AID FATIGUE is significantly conditioned by the three factors of the “aid-corruption” paradox. In each case, we find that the positive impact of QoG CHALLENGE on AID FATIGUE is significantly reduced, or even rendered negligible. For example, as shown in Table 1, a “yes” response on QoG CHALLENGE increases the odds of AID FATIGUE by about 40%. Yet when we take into consideration the effects of the mediating variables, the impact of QoG CHALLENGE on the dependent variable is substantially modified (as shown from the significant interaction effects). To more easily interpret the interaction effects, we provide a visual, demonstrating the marginal effects of QOG CHALLENGE with a 95% confidence interval (Figure 1) as suggested by Brambor, Clark, and Golder (2006), showing the marginal effect of QoG CHALLENGE asβ1, β2, or β3 vary. Moreover, we report the coefficients, from which we can more easily calculate the odds ratios of the marginal effect of QoG CHALLENGE when β1, β2, or β3 is >0, in Table 4. For example, the odds that QoG CHALLENGE leads to AID FATIGUE falls from 2.16 (assuming β1, β2, or β3 all equal “0”) to 1.38 when the respondent also believes that there is a strong moral obligation to help poor countries (for example, β1 = 3).15 Furthermore, the effect of QoG CHALLENGE falls below the 95% level of significance when β1 equals “1”, and below even the 90% level of confidence when β2 (QoG FOCUS) equals “1” according to Figure 1. Finally, it is worth mentioning that when a respondent shares two of the “aid-corruption paradox” factors at the same time, the effect of QoG CHALLENGE is actually reversed, meaning that the odds are less likely that a respondent will answer “yes” to AID FATIGUE.

1

Figures Based on Results from Table 3 Model 1 (Full Sample). Marginal Effects Reported are Coefficients (Not Odds Ratios), with Dotted Line Representing a 95% Confidence Interval Around the Estimates. Estimates are Positively (Negatively) Significant if Both Dotted Lines are Above (Below) the Zero Line.

View this table:
4

Effect of QOG Challenge Conditioned by the 3 “Anticorruption Paradox” Factors

Effect of QoG Challenge (Odds ratio)
ScenarioEU-27EU-15NMS-12
Baseline Effects
If HELP POOR = 0 & QoG FOCUS = 0 & STRATEGIC = 02.152.460.97
Interaction Effects
If HELP POOR (β1) = 3 (strongly agree)1.381.231.24
If QoG FOCUS (β2) = 1 (yes)1.401.731.43
If STRATEGIC (β3) = 1 (yes)1.522.160.47
If HELP POOR = 3 & QoG FOCUS = 10.890.801.39
If HELP POOR = 3 & STRATEGIC = 10.971.011.02
If QoG FOCUS = 1 & STRATEGIC = 10.981.420.33
If HELP POOR = 3 & QoG FOCUS = 1 & STRATEGIC = 10.630.650.69
  • (Notes. Odds ratios for the effect of QoG Challenge calculated from the Beta coefficients reported in Table 3 (author calculation). All variables held at “0” unless otherwise specified under “interaction effects.”)

Due to the significant country differences observed, models 2 and 3 test H2, H3, and H4 for the split samples. In model 2 (EU15 only), we essentially find stronger evidence for H2 and H3, yet we do not find the same type of “strategic thinking” among respondents regarding the conditional relationship between β3 and QoG CHALLENGE. For example, “strongly agreeing” that their country has a moral obligation to help poor people in developing countries offsets the effect of QoG CHALLENGE. For the NMS-12 countries, however, the results are, for all intents and purposes, the polar opposite. First, we find that QoG CHALLENGE has no independent effect on the dependent variable (for example, given that β1, β2, and β3 all equal “0”) in the NMS-12 countries, and interestingly, that there is no significant interaction effect between β1 (HELP POOR) or β2 (QoG FACTOR) and QoG CHALLENGE, as was the case with EU-15 countries. Furthermore, it appears that respondents from the newer member countries are, on average, more strategic in their thinking about development aid than their EU-15 counterparts—the data show that the odds of AID FATIGUE are reduced by an insignificant 3% when QoG CHALLENGE equals “1” and β1, β2, or β3 equals “0,”ceteris paribus. Yet when “QoG Challenge*β3” = 1 (for example, a respondent answers “yes” to both QoG CHALLENGE and STRATEGIC), the odds of AID FATIGUE are decreased by 53%. This shows that although respondents might acknowledge that corruption is a key problem in recipient states, insofar as strategic motives for foreign aid are held, they are more likely to show signs of aid support. Interestingly, the conditional effects of STRATEGIC and HELP POOR through QoG CHALLENGE seem to offset one another given that QoG FACTOR is “0,” while when all three conditional factors are at their maximum value, the impact of QoG CHALLENGE reduces the odds of AID FATIGUE by between 37% and 31% depending on the sample.

In short, we find that in EU-15 countries, the effect of corruption on AID FATIGUE is offset by a belief that helping poor countries is important and an understanding that aid should go toward improving governance. However, we do not find this conditional relationship in the 12 new member states. Yet in the latter group, a respondent believing that corruption is a critical problem in recipient states is actually less likely to exhibit aid fatigue when thinking strategically about aid, an effect that was not significant in the EU-15. On the whole, therefore, the results show that while perceptions of corruption in recipient countries are associated with aid fatigue, the relationship is highly conditioned by a range of individual and country-level factors. In general, we find support for our four hypotheses, summarized in Table 5.

View this table:
5

Summary of Results

Sample
HypothesisEU-27EU-15NMS-12
H1 (Corruption Increases Aid Fatigue)SupportSupportWeak/No support
Conditional Hypotheses
H2 (Moral)SupportSupportNo support
H3 (Pragmatic)SupportSupportNo support
H4 (Strategic)SupportNo supportSupport

Conclusion

Foreign aid is a highly controversial and debated topic in academic and policy circles alike. In 1962, Hans Morgenthau wrote that, “[o]f the seeming and real innovations which the modern age has introduced into the practice of foreign policy, none has proven more baffling to both understanding and action than foreign aid” (p. 301). Almost 50 years later, this observation still rings true. On the one hand, resource redistribution can be viewed as a moral imperative in a highly unequal world, with proponents arguing that aid has helped developing countries with much needed resources. On the other hand, critics argue that foreign aid has resulted in massive waste of taxpayer's money and that the results have been meager at best and counterproductive at worst (Boone 1996; Svensson 2000; Knack 2001; Djankov et al. 2008). Arguments of this sort can be expected to affect public opinion on foreign aid, with potential implications for the quantity and quality of aid in the future.

In this study, we have proposed that perceived corruption is strongly correlated with aid fatigue, yet that public reactions to corruption in aid substantially depend on fundamental beliefs about the role of foreign aid. While we acknowledge the difficulties of testing aid fatigue directly with spatial data, the empirical results, which are robust to several different types of models and control variables added, point to three salient findings. First, opinions about corruption in recipient countries are likely to sway individual perceptions about foreign development aid in donor countries; on average, people who believe governance and corruption are a primary problem in recipient states are more likely to exhibit aid fatigue.

Second, the results show that the relationship between these two opinions is geographically contextual. Looking at both individual countries along with the two larger blocs of states (EU-15 and NMS-12), we find significant differences. Citizens from the wealthier, EU-15 countries who maintain that governance and corruption are key problems in recipient countries are in particular less likely to support foreign aid. In the new member states, however, we do not find the same systematic link between these two opinions. One explanation for such regional variations is the possibility that attitudes can be shaped by how foreign policy has been framed and how the purpose of aid has been defined in different countries.

Third, we find strong evidence that the relationship between corruption and aid fatigue is highly nuanced and either reduced or offset given various understandings of the aid-corruption paradox. For example, a belief that one's country should help poorer people in developing countries, or that governance should be the focus of aid itself, or that aid will lead to greater opportunity for exports for the respondent's own country all reduce the odds that concern about recipient state corruption will lead to aid fatigue. Yet even these findings are geographically contextualized; moral and pragmatic understandings of the aid-corruption paradox reduce aid fatigue in the EU-15 countries, whereas in NMS-12 countries, it is only the strategic understanding that has this offsetting effect on aid fatigue. We thus conclude that attitudes about corruption significantly influence opinions about support for aid, yet this relationship is conditioned by factors that have been overlooked by previous literature. However, we stress caution when interpreting the findings due to the cross-sectional nature of the data, as there are clear limits to any causal claims that can be made with such a design.

The data present a number of interesting puzzles for future research. For example, why would corruption in recipient countries be a significant concern regarding attitudes toward development aid in countries like Spain, Austria, Germany, Finland, and the UK yet not in Italy, France, or Portugal? We could of course postulate that citizens from wealthier countries are, on average, more likely to think in terms of accountability of governance conditions in recipient countries and are less likely to support aid when they believe corruption is present. Regarding differences between the EU-15 and the NMS-12 countries, it would appear based on the findings here that there is a stark contrast regarding the understanding of development aid and the purpose of it. Thus, why corruption and governance concerns are less connected with support for development aid in new member states is an interesting topic for future studies. Future research would moreover benefit from comparable public opinion data focused on questions regarding national-level aid, in order to further test for country-level variances in opinions about development aid. While the Eurobarometer data are extensive and provide us with a useful proxy, they do not allow us to look at opinions of national aid budgets. Future research could also show whether these findings hold in non-European donor countries and across time.

One of the potentially important implications that our results point to is that corruption can be tolerated if development aid to corrupt countries is viewed as assisting in the development of democratic and high-quality institutions, in particular in the arena of public opinion within EU-15 countries. We thereby show that one solution to the aid-corruption paradox depends on whether aid is believed to contribute to strengthening quality of government in developing countries. Thus, a focus on effective anticorruption and institution-building measures among donors may be one way of solving the aid-corruption paradox. This would call for a strengthening of the international anticorruption regime. However, thus far, improving quality of government in developing countries has proven to be a difficult task for many international donors (Bauhr and Nasiritousi 2012; Bauhr and Nasiritousi, Forthcoming).

A better understanding of the determinants of aid fatigue is important at a time when international donors increasingly put emphasis on fighting corruption in developing countries, as this allows us to understand fundamental relationships between corruption, sustainable development, and aid legitimacy. This study has provided new insights into the public opinion/development policy nexus, pointing to the relevance of policy framing for understanding the support for international public goods like foreign aid. Ultimately, corruption can be seen as a constraint on aid effectiveness or as an opportunity for aid-driven reforms, and which of these perspectives that dominates public debates may have implications for both the levels and types of aid distributed to the most poverty-ridden countries in the world, as well as future drives toward sustainable development.

Appendix

Appendix 1

Summary statistics

View this table:

Variable description

View this table:

Footnotes

  • 1 The authors wish to thank participants in the 2011 International Studies Association conference in Montreal, the Quality of Government Institute, and two anonymous reviewers for their helpful comments on earlier versions of the manuscript. All errors remain our own. Corresponding author: monika.bauhr{at}pol.gu.se

  • 2 In this analysis, we study opinions about development aid—not humanitarian aid.

  • 3 For notable exceptions, see, for example, Chong and Gradstein 2008; Paxton and Knack 2012; Henson et al. 2010.

  • 4 Changes in public opinion reported from 1986–1995 in the United States falling from 54% to 47%, while from 1968 to 1998 in Denmark rising from 35% to 72%. Yet it is worth noting that in Japan and Australia, two other countries with similar dynamic data on public support for aid in the study did not result in spending levels that corresponded with public opinion, yet changes in public opinion were nonsignificant over the time period shown.

  • 5 Leiken (1996–1997:58), for example, reports that: “A survey of the Economist, the Financial Times, and international coverage in the New York Times revealed that articles mentioning official corruption […] quadrupled between 1984 and 1995.”

  • 6 Although no perfect indicator exists, we perform a simple t-test between EU-15 and NMS-12 countries on two questions that might capture this difference. First, on the question “how justifiable is someone accepting a bribe in the course of their duties” (1–10, never-always), we find the difference to be significant (p = .04). Second, “How justifiable is it to pay in cash to avoid paying taxes” (1–10, never-always), we find a significant difference in the expected direction, with more tolerance on average from NMS-12 countries.

  • 7 We acknowledge that there are some conceptual problems with this variable, namely that the notion of “aid fatigue” implies a change of attitudes over time. Due to the cross-sectional nature of the data, along with the simple fact that some countries might simply have different traditions with regard to aid, we understand the problems associated with claiming that we are in fact demonstrating “aid fatigue”per se, and in a sense, the variable is more closely capturing simply a “lack of support for current levels of aid.” With this in mind, we proceed forward in the analysis with a degree of caution about the results, and we would like to thank one of the anonymous reviewers at ISQ for pointing out this distinction.

  • 8 See Rothstein and Teorell 2008:169–173, for a more in-depth discussion on this topic.

  • 9 A list of possible answers along with summary statistics is located in the appendix.

  • 10 We do not show these results for a lack of space, but please contact the authors for the baseline country differences.

  • 11 In the results, we report odds ratios, which indicate the factor change in the odds of the dependent variable equalling “1,” with numbers above (below) “1” resulting in a positive (negative) change in the odds.

  • 12 The models in Table 1 are estimated in STATA with the “xtmelogit” command.

  • 13 We check the robustness of Table 1 by reproducing the results by using a standard logit model with country dummies to control for the fixed effects, and the estimates from Table 1 were highly robust. Please contact the authors for the results of this robustness check.

  • 14 The results do not change significantly if we take the hypotheses one at a time.

  • 15 For example, calculating the odds ratio for the marginal effect of QoG Challenge when “HELP POOR” equals “3” is the exponentialized sum of the QoG Challenge coefficient and the interaction term (exp (QoG Challenge + QoG challenge* (3*β1))) and is calculated as follows: beginning with the sum: 0.77 (QoG CHALLENGE) − 0.441 (QoG Challenge*β1, where β1 = 3) = 0.326. Then, take the exponent of this sum “exp (0.331)” = 1.38.

References

View Abstract