1. Introduction
About 70 percent of global inequality is determined by income differences between countries rather than by income differences within countries (Milanovic, Reference Milanovic2018). Governments have the tools to redistribute incomes internationally (Chancel et al., Reference Chancel, Piketty, Saez and Zucman2022). However, whether they do so depends on the domestic political pressures they experience. Crucially, international redistribution requires intergovernmental cooperation and it is therefore of utmost importance to study political dynamics in both countries that contribute to international transfers and countries that receive them (Briggs, Reference Briggs2014).
Aid is one of the main instruments of international redistribution. In 2020, DAC countries spent an average of 0.33 percent of their gross national income (GNI) on aid (OECD, 2022), and aid contributed about 11.6 percent to the GNI of low-income countries (Worldbank, 2022). Satisfaction with such external support varies considerably among constituents in recipient countries. For example, survey data from 2008 show that 9 percent of Africans believe the USA does not help their country at all, about 44 percent believes that it helps a lot, while the rest falls in-between (Afrobarometer, 2015).Footnote 1 However, we do not know what drives demand for aid and international redistribution in recipient countries and whether the reduction of between-country inequality is an important motivation.
This paper explores how perceptions of between-country inequality influence attitudes toward inequality and redistribution in recipient countries. Studies have shown that perceptions of domestic inequality are an important determinant of redistributive attitudes (Cruces et al., Reference Cruces, Truglia and Tetaz2012; Fernández-Albertos and Kuo, Reference Fernández-Albertos and Kuo2015; Becker, Reference Becker2020). Similar studies on international inequality have been limited to donor countries (Nair, Reference Nair2018; Fehr et al., Reference Fehr, Mollerstrom and Perez-Truglia2022). This paper thus expands our understanding of the relationship between international inequality and redistribution to the Global South.Footnote 2
The findings presented in this paper are based on a novel, pre-registered survey experiment, which was conducted in Kenya, a major recipient of international aid. Respondents are treated with information about international income differences and asked about their attitudes toward international inequality and redistribution. The experiment is implemented via SMS to reach a broad sample of respondents and quotas are used to make the sample as representative of the Kenyan population as possible.
The results show that Kenyans underestimate international inequality. Per-capita incomes in Western Europe are 25 times higher than in Kenya (Worldbank, 2022), but most respondents estimate it to be below a factor of five. This echoes other studies that show that citizens in the Global North underestimate their position in the global income distribution (Nair, Reference Nair2018; Fehr et al., Reference Fehr, Mollerstrom and Perez-Truglia2022). At the same time, only about 43 percent of Kenyan respondents do find the existing differences unacceptable and about 38 percent think their government should demand more financial aid from donors. Can this be explained by misperceptions of international inequality?
The common underestimation of international inequality clearly shows that factual information has the potential to influence political knowledge and attitudes. In the experiment, respondents treated with information indeed express lower acceptance of international inequality. However, this does not translate into greater demand for international aid, suggesting that respondents do oppose international inequalities but do not prefer to reduce them through international aid.
The paper makes three distinct contributions. The first contribution is to scholarship on inequality and redistribution. The findings show that individual concern with inequality extends beyond the domestic realm and has a global dimension (Nair, Reference Nair2018; Fehr et al., Reference Fehr, Mollerstrom and Perez-Truglia2022), also in developing countries. At the same time, the finding that aid is not a preferred policy response calls for further investigations of (i) other types of international redistribution, such as trade or migration policy, and (ii) individual preference formation, including beliefs about fairness, deservingness, and policy effectiveness, all of which have been explored in domestic but rarely in international contexts.
The second contribution is to studies on international aid, which until recently paid little attention to public opinion in recipient countries (Milner and Tingley, Reference Milner and Tingley2013). The findings presented here suggest that citizens in recipient countries do not perceive aid as effective, at least when it comes to reducing international inequality. A lack of effectiveness, perceived or real, challenges the legitimacy of international aid (Heinrich, Reference Heinrich2013; Bermeo, Reference Bermeo2017). Recipient governments might therefore experience few pressures from their citizens to demand more aid from donors, which counter-intuitively can strengthen their bargaining position when it comes to policy concessions in donor-driven aid negotiations (Mesquita and Smith, Reference Mesquita and Smith2009; Whi, 2009; Chiba and Heinrich, Reference Chiba and Heinrich2019). The findings also call into question how recipient governments manage to distribute aid strategically to achieve political objectives (Briggs, Reference Briggs2014; Jablonski, Reference Jablonski2014; Knutsen and Kotsadam, Reference Knutsen and Kotsadam2020). Strategic credit-claiming, which can undermine the visibility of donors and thus lower perceptions of aid effectiveness, is likely to play an important role (Bodenstein and Kemmerling, Reference Bodenstein and Kemmerling2017; Cruz and Schneider, Reference Cruz and Schneider2017; Baldwin and Winters, Reference Baldwin and Winters2023).
The third contribution is methodological. The study is the first to employ an SMS-based survey experiment to study attitudes toward inequality and redistribution. This format has advantages in terms of expediency, representativeness, and costs, but also comes with a number of challenges. When these trade-offs are well-balanced, SMS surveys provide a powerful instrument to study public opinion, especially in the context of developing countries (Leo et al., Reference Leo, Morello, Mellon, Peixoto and Davenport2015; Lau et al., Reference Lau, Lombaard, Baker, Eyerman and Thalji2019).
2. Theory
A key debate in the extensive and continuously growing scholarship on the politics of inequality is concerned with the role of public opinion, in particular support for redistribution (Bartels, Reference Bartels2008; Rueda and Stegmueller, Reference Rueda and Stegmueller2019). Scholars explore when and under what circumstances changes in inequality translate into demand for redistribution.
Several studies ascribe a central role to perceptions of inequality. While misperceptions are common, and individuals tend to underestimate income differences, perceptions still follow factual developments and they are often systematically biased by the information available to individuals (Minkoff and Lyons, Reference Minkoff and Lyons2017; Gimpelson and Treisman, Reference Gimpelson and Treisman2018). For example, Becker (Reference Becker2021) shows that Americans continuously underestimate inequalities, but actual as well as perceived inequalities have been growing over time. Cruces et al. (Reference Cruces, Truglia and Tetaz2012) argue that underestimates are the result of income homogeneity in individuals' immediate social surrounding, which is a primary source of information. Furthermore, studies have shown that perceptions of inequality shape attitudes toward inequality and support for redistribution (Minkoff and Lyons, Reference Minkoff and Lyons2017; Bobzien, Reference Bobzien2020), including a growing number of experimental studies (Cruces et al., Reference Cruces, Truglia and Tetaz2012; Kuziemko et al., Reference Kuziemko, Norton, Saez and Stantcheva2013; Fernández-Albertos and Kuo, Reference Fernández-Albertos and Kuo2015; Becker, Reference Becker2020).
All studies mentioned above focus on domestic inequality and redistribution. However, two recent studies have extended the scholarship to questions about international inequality. Nair (Reference Nair2018) finds that Americans underestimate their rank in the global income distribution. When informed about their actual position, they become more supportive of expanding aid budgets and are more likely to contribute to charities. In a similar study, Fehr et al. (Reference Fehr, Mollerstrom and Perez-Truglia2022) correct German's misperceptions about their global income rank and find no effects on preferences. The two studies provide important insights into the politics of international inequality from the perspective of developed countries. While their partly contradictory results call for further investigation, the larger gap concerns the politics of international inequality in the Global South, which has not been addressed by prior studies.
Further helpful in closing this gap is research on the politics of aid. Aid is one of the main instrument for international redistribution. Here, an extensive body of studies again focuses on donor countries, highlighting amongst others the role of religion, populism, and recipient characteristics (Paxton and Knack, Reference Paxton and Knack2012; Milner and Tingley, Reference Milner and Tingley2013; Bodenstein and Faust, Reference Bodenstein and Faust2017; Blackman, Reference Blackman2018; Dietrich et al., Reference Dietrich, Hyde and Winters2019; Heinrich and Kobayashi, Reference Heinrich and Kobayashi2020; Heinrich et al., Reference Heinrich, Kobayashi and Lawson2021). Studies that are concerned with the Global South mainly explore the popularity and visibility of donors (Hanusch, Reference Hanusch2012; Keuleers, Reference Keuleers2015; Bodenstein and Kemmerling, Reference Bodenstein and Kemmerling2017), or electoral consequences of aid disbursements (Jablonski, Reference Jablonski2014; Briggs, Reference Briggs2019; Knutsen and Kotsadam, Reference Knutsen and Kotsadam2020) but they do not focus on demand for aid.
Building on these studies, this paper asks in how far Global South publics demand aid to reduce international inequalities. The following hypotheses guide the empirical part of this study. As previous studies have shown that aid flows are influenced by former colonial ties between donors and recipients (Fuchs et al., Reference Fuchs, Dreher and Nunnenkamp2014; Chiba and Heinrich, Reference Chiba and Heinrich2019; Becker, Reference Becker and Schmitt2020), the third hypothesis tests whether special demands are put on former colonial powers.
(H1) The higher individuals perceive between-country inequality to be, the less accepting they are of the status quo.
(H2) The higher individuals perceive between-country inequality to be, the more they demand international redistribution.
(H3) The higher individuals perceive between-country inequality to be, the more they demand international redistribution from former colonial powers (as compared to other donors).
3. Research design
Developing countries often have relatively low rates of internet access and face-to-face interviews are expensive and time-consuming to implement. However, mobile phone penetration is high in many developing countries, and development scholars and practitioners therefore increasingly rely on SMS-based surveys (Leo et al., Reference Leo, Morello, Mellon, Peixoto and Davenport2015; Lau et al., Reference Lau, Lombaard, Baker, Eyerman and Thalji2019).Footnote 3
While SMS surveys are advantageous in terms of sample representativeness, they also pose new challenges, in particular space constraints (Lau et al., Reference Lau, Sanders and Lombaard2019). Each question including numbered response options needs to be contained within the 160-character limit of a single text message. While this carries risks for the validity and reliability of measures, it has been shown that results from other survey instruments can be reproduced (Lau et al., Reference Lau, Cronberg, Marks and Amaya2019).
The survey experiment is implemented in Kenya with the assistance of a leading market research company, which hosts an incentivized opt-in SMS panel with over a hundred thousand users. To achieve the target sample size of 1000, and taking prior response rates into account, invitations were sent out to 4000 potential respondents in July 2022. The potential respondents were selected based on nationally representative quotas for age, gender, and region of living, and each of them was randomly assigned to one of the experimental conditions.Footnote 4 After exclusion criteria were applied, the final sample amounts to 947 respondents.Footnote 5 Respondents are between 20 and 67 years old (median: 26), 43.5 percent identify as female, and the large majority holds at least a secondary degree (77.8 percent) and only a small share did not complete primary education (7.5 percent). Most are students (44.4 percent), some are skilled workers (7.1 percent), and the remaining 48.5 percent are composed of unskilled workers, farmers, individuals in other employment or currently unemployed. The majority resides in the Rift Valley region (35.3 percent), Nairobi (16.1 percent), and the Nyanza region (14.1 percent).Footnote 6
Treatment
The SMS survey begins by informing participants about the purpose of the study and asking for their consent. Afterwards, they proceed to questions that elicit pre-treatment covariates which are used in exploratory analyses (more below). Most importantly, all respondents are asked about their perception of international inequalities. To capture this within one figure, respondents are asked to guess the size of income differences between Kenya and Western Europe: “The average income in Western Europe is higher than in Kenya. How many times higher do you think? Please provide your best guess.” The treatment then consists of factual information about international inequality: “According to official statistics by the World Bank, average incomes in Western Europe are 25 times higher than in Kenya.” Those in the control group proceed to the post-treatment questions without this information.Footnote 7
Post-treatment questions
First, respondents are asked about their attitude toward international inequality (Inequality Acceptance): “Some say these income differences are acceptable. To what extent do you agree or disagree with this statement?” Responses are collected on a five-point scale ranging from “Strongly disagree” to “Strongly agree.” Second, respondents are asked their demand for international redistribution, whereby separate questions are asked about the UK (Aid (UK)) and the USA (Aid (USA)), the two largest donors to Kenya: “Some say that Kenya should demand more aid from the United Kingdom (UK)/United States of America (USA). Do you agree?” Responses are collected on the same five-point scale. The following analysis also uses an aid index (Demand for Aid), which indicates the mean response to both aid questions.
Figure 1 summarizes the outcome variables. All three variables follow a bimodal distribution, which points at considerable polarization in attitudes toward international inequality and redistribution. There is almost no difference in demand for aid from the UK and the USA. Thus, individuals might not differentiate between aid from their former colonizer and other donors.
Attitudes toward International Inequality and redistribution. Note: Counts indicate number of respondents per category and question. All three variables are measured on a five-point scale; see text for details.
4. Results
Asked about how many times higher average incomes in Western Europe are compared to Kenya, over 86 percent of study participants underestimate the actual number of 25, and 59.1 percent put it at 5 or below. This echoes the findings of earlier studies, which suggest that individuals tend to underestimate inequalities beyond their immediate social environment. This finding is also important for the study design as hypotheses and treatment were based on the assumption that Kenyans underestimate international inequality.
Figure 2 displays full histograms of the inequality guesses (i.e., Perceived Inequality). In addition, it shows the association with the two main outcome variables, summarized by bivariate quadratic regression lines. It can be seen that respondents who perceive international inequality to be higher tend to be less accepting of those differences (Inequality Acceptance). At the same time, these respondents also demand more aid (Demand for Aid); with the exception of extreme overestimators of inequality, for whom demand appears to fall again.
Perceptions of International Inequality and Attitudes. Note: Bivariate quadratic regression lines, with 95 percent confidence intervals, indicate associations between Perceived Inequality, i.e., respondent guesses of income differences between Western Europe and Kenya (in multiples, e.g., 1 = 100 percent-increase, log-scale), and Inequality Acceptance (left) and Demand for Aid (right), both measured on five point-scales; see text for details.
Turning to the causal effects of perceptions of international inequality, Table 1 presents estimates of the average treatment effects. Model 1 shows that being presented with factual information about international inequality decreases inequality acceptance. Given that the outcome is measured on a five-point scale, with an observed interquartile range of 2, the estimated effect is sizable. To ascertain that the treatment can actually change opinions, the outcome is dichotomized and the models reestimated. This test corroborates the findings presented here (see online Appendix, Table A1). Overall, the results provide evidence in favor of the first hypothesis that perceptions of international inequality lower inequality acceptance.
Average treatment effects (OLS)
Note: Ordinary least squares regression. Inequality Acceptance indicates acceptance of income differences between Kenya and Western Europe; aid indicates demand for international financial transfers. (* = 0.05).
The remaining models in Table 1 concern demand for aid. Here, no evidence for any effect is found, neither for the aid index (2), British aid (3), American aid (4), or differences between the two donors (5). This shows that information about international inequality does not increase demand for international redistribution, at least not through financial aid. Demands toward the former colonizer Britain are also not differently affected from demands toward the USA. The second and third hypotheses, which concern aid, must therefore be rejected.
Does the treatment effect depend on prior knowledge? As individuals vary in their pre-treatment guess of the income differences between Kenya and Western Europe, I test whether the treatment effect depends on the initial guess. This can be regarded as a test of treatment dosage effects. Complete results are included in the online Appendix (see Table A2). They suggest that the pre-treatment guess, and thus the treatment dosage, does not matter. This puts into question what and how much different individuals learn from the treatment. That being said, a likely explanation for this null result is the similarity of most initial guesses relative to the factual value. As such, differences in the correction of perception errors are likely too small to matter.
The findings come with some caveats. First, although respondents represent a broad cross-section of the Kenyan population, sampling was non-random and generalizability may therefore be limited. Second, the survey format provided little space for elaborating questions and answer options. This is likely to contribute to measurement uncertainty and to lower the strength of the treatment. The research design therefore provides a conservative test of the hypotheses.
Exploratory results
Income and religion have been shown to be important determinants of attitudes toward domestic inequality in Africa (Nel, Reference Nel2021) and they have therefore been included in the pre-registration as exploratory analyses. The results are presented in full in the online Appendix. They show that the poor and the religious express greater demand for aid than richer and less religious respondents, although there are no marked differences regarding inequality acceptance. At the same time, the treatment effects are most pronounced among non-religious respondents, lowering the acceptance of inequality, while highly religious respondents are not responsive. However, even among the non-religious, the treatment does not increase demand for aid.
5. Conclusion
Complementing studies on domestic inequality, scholars have recently started exploring the politics of international inequality. This paper extends this new line of work to the Global South with a novel SMS-based survey experiment. The findings show that respondents in the Global South, like their counterparts in the North, underestimate international inequality and are more likely to condemn it when learning about its actual size. However, this opposition to international inequality does not translate into increased demand for international redistribution, at least not in the form of financial aid.
The findings have several implications for the politics of international redistribution, and aid in particular. In an increasingly connected world and with rising levels of education, awareness of international inequalities and related discontent is likely to increase around the world. While this might not translate into demand for aid, it could lead to calls for other types of international redistribution, for example through trade, migration, or taxation. At the same time, donors need to reevaluate the legitimacy and effectiveness, perceived or real, of international aid if they are to achieve developmental and political objectives. In fact, citizens might appreciate aid for many other reasons than the reduction of international inequality.
Finally, the findings point at two other factors that call for further investigation. Income and religion were found to correlate with demand for aid, whereby low income and frequent religious attendance were associated with greater demand. However, these analyses were solely exploratory, using only observed covariate data. Future studies could use experimental designs to study the effect of income and religion on demand for aid and other internationally redistributive policies.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/psrm.2023.36.
To obtain replication material for this article, https://doi.org/10.7910/DVN/EZ0UZK.
Open access
