Chinese aid and democratic values in Latin America

Abstract

International economic engagement has been increasingly framed in terms of liberal democratic values. Specifically, Chinese aid has been at the center of this debate. Since Chinese aid comes with “no strings attached,” a popular narrative is that Chinese aid poses a challenge to conditional aid, thus weakening democracy promotion. This study aims to deepen our understanding of how democratic values are shaped by international economic engagement. Drawing on the Latinobarometro Household Survey, we use an instrumental variable approach to test the effect of Chinese aid on attitudes toward democracy in 18 Latin American countries on the national and regional level. We find that Chinese aid has a non-negative effect on support for democracy. We also find that individuals who have a positive attitude towards China are more likely to value democracy. In contrast, positive attitudes towards the USA have no robust impact on support for democracy.

1 Introduction

China has been economically and politically engaged in many middle-income and developing countries for the last two decades. Chinese strategies comprise geopolitical coalitions such as the Association of Brazil, Russia, India, China, and South Africa (BRICS) as well as investment and development aid in developing countries (Fuchs & Rudyak, 2019). The Chinese government frames these activities as a feature of a systems competition with the West (Heath, 2021), claiming that liberal democracies are not well-equipped to solve future problems. Although China does not link its aid to conditions, which some obervers describe as “no strings attached,” some view it as conditional and a challenge to liberal democratic values.

Consequently, Chinese aid has been the subject of controversy in the media as well as in academic contributions, with a focus on the political economy of China’s engagement in developing countries. One question from the public choice angle is whether an increase in Chinese aid will substitute democracy-promoting development aid given by the Development Assistance Committee (DAC) countries and therefore erode democratic values in the recipient countries. For many years, Chinese development aid in Africa has dominated the public discourse, and in recent years, China has also intensified its relations with Latin America. Argentina’s recent application and accession to BRICS can be seen as an indication of the success of the strategy. Chinese efforts to gain favor and influence are mostly driven directly by the government or firms close to the government. Therefore, it seems appropriate to consider the engagement of China in Latin America in more detail. In this paper, we take stock and explore in particular the relationship between Chinese aid and support for democracy in Latin America.

The discussion in both the public realm and academia mainly suggests a negative relationship between Chinese aid and democratic support. The hypothesis most often taken is that Chinese aid crowds out Western official development assistance (ODA) and thus leads to decreasing support for democracies in the recipient countries. A similar argument is based on the idea that the donating country is a “governing role model.” If China, as an autocratic country, is successful in providing effective development assistance, then individuals may change their attitude toward democratic systems and their role in development. This reasoning is based on four assumptions: (1) Chinese aid projects are effective, (2) before the Chinese engagement, most people had a positive view of democracy, (3) individuals are informed about Chinese aid projects in their countries, and (4) they are aware that China is an autocracy.

This hypothesis has been assessed in the literature on China’s engagement in Africa, and the results are mixed. We test it for Latin America since it cannot be taken for granted that the public reaction to Chinese aid is the same across the world. The underlying institutional factors in the formation of an attitude towards a foreign actor may well be different between continents or even countries. Indeed, it seems that the political economy of foreign relations in Latin America is distinct from that of other continents, such as Asia and Africa. Many Latin American countries have a long democratic history, and the autocratic phases that have often interrupted this history have been overcome. Latin American leaders often follow populist agendas. Moreover, as Latin America has historically been regarded as the United States’ (US) backyard, reactions to China and its activities on the sub-continent may well be idiosyncratic against the background of long-standing US–Latin American relations.

This history promises an interesting addition to the growing public choice literature on the institutional consequences of Chinese aid by testing whether aid has decreased public support for the concept of democracy as compared to autocracy. We can thus also assess whether there is a uniform reaction towards Chinese aid in the developing and emerging world. We use Latinobarometro representative survey data to construct a multi-level repeated cross-sectional analysis for 18 countries over the period 2004–2015 with two measures of democracies. Next, we introduce the literature and derive hypotheses. In Sect. 3 we describe the data and empirical strategy. Section 4 presents the results. In Sect. 5, we investigate the effect of Chinese aid projects at a regional level, and in Sect. 6, we conduct a range of robustness tests. Finally, Sect. 7 concludes.

2 Literature and theoretical considerations

The topic of Chinese aid and its effects on societies, regarding welfare, governance structures, and public opinion, has attracted a great deal of interest. There is a growing body of literature on the institutional consequences of Chinese aid (Bader & Faust, 2014; Dreher et al., 2022), and many of these studies focus on Chinese engagement in Africa. China is a relatively new actor in development cooperation and not a member of the Organisation for Economic Co-operation and Development (OECD). Since its government does not adhere to standardized reporting of data on development aid, there is much suspicion that this lack of transparency is intended to conceal the aims of donating aid. The related public-choice literature can be divided into at least three groups. The first group comprises papers that look at the impact of Chinese aid and lending on the behavior of traditional lenders, such as the World Bank, or regional development banks, which mostly lend with conditions. A second group deals with the direct effects of Chinese aid on economic welfare and political governance, whereas a third group addresses the consequences of Chinese development aid on public opinion in the recipient countries. This paper contributes to the third group.

We start our brief overview of the literature by discussing the effects of a new donor on the behavior of traditional donors and, indirectly, of the borrowers. This is relevant since Western countries and international organizations, such as the World Bank, and regional development banks regularly combine lending or aid with conditions, for example, the demand to improve governance structures or to start policy reforms. Two main questions are asked in the literature: (1) Do Western lenders change their attitudes and demand fewer conditions or redirect resources as a response to Chinese activities? and (2) do borrowers change their compliance? The evidence is mixed for both questions. Humphrey and Michaelowa (2019) see no evidence for changes in the lending pattern of the World Bank or regional development banks, whereas Hernandez (2017) documents that World Bank loans come with fewer conditions as a response to Chinese aid. Zeitz (2021) shows that in reaction to Chinese infrastructure investments (not aid), World Bank projects also focus more strongly on investment, which can be interpreted as competition leading to higher consideration of the recipients’ needs. As regards the second question, Watkins (2022) finds that recipient countries reduce compliance with the conditions imposed by international organizations. However, in the last few years, some borrowers have had to default on loans from China. Therefore, one can expect that effect to be reversed, as the paper by Kern et al. (2023) suggests. The authors look at the survival rate of autocratic governments in developing countries before and after default and argue that a bailout by international organizations reduces this rate.

To understand the direct effects of Chinese aid on governance (the second strand of the literature discussed here), a short comparison with Western ODA is necessary. Kersting and Kilby (2014) show that Western ODA can support economic reform and is often directed at democratic reform. As regards Chinese aid, Dreher et al. (2019) find significantly more regional favoritism than in World Bank projects, with Chinese ODA being allocated disproportionally to leaders’ birthplaces. This situation seems to be in line with the “no-strings-attached” hypothesis, suggesting that Chinese donors do not care about the distributional effects of their aid payments. In addition, Isaksson and Kotsadam (2018) show that corruption in Africa is increasing in regional proximity to Chinese aid project sites and conclude that Chinese aid is changing norms.

In a similar vein, Sardoschau and Jarotschkin (2019) argue that Chinese aid projects increase the likelihood of regional violence and civil conflict in sub-Saharan countries (see also Iacoella et al., 2021). In another paper dealing with the relationship between Chinese aid and conflict in Africa, Gehring et al. (2022) show that neither Chinese aid nor World Bank projects increase violence. However, World Bank projects are more likely to reduce violence and support democracy than Chinese engagement. Brazys and Vadlamannati (2021) show that countries that receive Chinese aid are less likely to implement economic reforms, indicating that an “aid curse” might be in play. Hess and Aidoo (2019) look at different hybrid regimes in African countries that have received substantial Chinese aid and conclude that there have been cases of democratic backsliding. This conclusion is in line with Li (2017), who finds that democratization became less likely in countries receiving large shares of Chinese aid. However, Bader (2015) finds that autocrats are not generally backed by Chinese aid. Only if the recipient country is export-dependent on China is autocratic survival more likely, and in this case, the risk of human rights abuses may even increase (Gamso, 2019).

For our study, the effects of Chinese aid on political attitudes are of relevance. A range of papers analyzes the international political economy of aid flows and confirms that Chinese foreign aid can buy support from the recipient’s side in international politics (Brazys & Dukalskis, 2017; Strüver, 2016). We are interested in another question, namely, how Chinese aid changes the public perspective on democratic values. Eichenauer et al. (2021) investigate the impact of Chinese engagement, including aid, on support for China. Their results indicate that Chinese involvement does not improve the public opinion of China. However, there is a difference between age cohorts, as younger and more highly educated people view China more favorably. Wellner et al. (2022) maintain that public approval of the Chinese government is higher in countries hosting Chinese development projects. Moreover, they differentiate two time horizons: in the short run, this effect increases with the size of the government and the generosity of the financial commitments, while in the long run, the effect is smaller among people living close to completed Chinese development projects.¹ Interestingly, moreover, the authors find that aid creates support for the Chinese government even in third countries that are not directly affected by, but politically aligned with, the recipient country (Wellner et al., 2022).²

Blair et al. (2021) study the impact of Chinese aid projects on public opinion of China and the US in Africa using geolocated Chinese projects. Their study shows that Chinese aid does not reduce support for Western values or the US; rather, it enhances the US’s reputation while not increasing public support for China. This effect has already been reported by Hanauer and Morris (2014). Blair et al. (2021) also find that U.S. aid in Africa is increasing support for Western democratic values. By the same token, Bai et al. (2022)—investigating 47 developing countries over several continents—show that Chinese aid increased a positive attitude toward the organs of domestic governments. The analysis carried out by Bai et al. (2022) is similar to but differs from ours: they look at confidence levels in government, civil service, and the military and find that Chinese aid does not affect how individuals evaluate living in a democratic system. Our study differs regarding the region we look at and the dependent variables as individuals, rather than officials, were asked to choose between democracy and autocracy.

These studies suggest that the Chinese administration is allocating aid payments strategically as a tool to improve China’s image in foreign countries, although Glaser et al. (2009) do not see a coherent Chinese soft power strategy. As Blair et al. (2021) argue, if aid boosts the image of donor countries, it may align the values of the recipient with those of the donor. Thus, we would expect Western ODA to have a similarly positive effect on public attitudes towards Western values. The mechanism is based on the idea that aid contains a “role model” function. Another possible mechanism is based on the leveraging function of aid. If democracy-promoting aid is diluted by non-conditional aid, it will change attitudes. Both channels assume that individuals are informed about Chinese aid projects and aware of the donors’ political principles.

As Blair et al. (2021) did in their analysis of Africa, we test the hypothesis for Latin America. It cannot be taken for granted that the public reaction to Chinese aid is the same across the world, as the underlying institutional factors that are instrumental in forming attitudes towards a foreign actor may well be different between continents or even countries. Focusing on Latin America has an additional advantage since Latin America has only recently seen an increase in China’s economic presence in the last 30 years (Eichenauer et al., 2021), whereas many African countries have been exposed to this presence for decades (Dreher & Fuchs, 2015; Strange, 2019). Moreover, the Latinobarometer survey offers a comprehensive and consistent question catalog that allows researchers to control for many attitudes that correlate with support for democracy.³

Our analysis is thus based on the premise that aid shapes attitudes, an idea that policymakers have consistently presented. It generates the following hypothesis, which can be divided into two parts. The central hypothesis predicts (backed by most of the literature) that.

(H1) an increase in Chinese official aid flows decreases support for democracy.

However, as the empirical literature discussed above suggests, there is ample reason to be skeptical about the hypothesis. First, it is unclear whether individuals can correctly identify and attribute aid projects to donors (Blair et al., 2021). Second, while China is interested in demonstrating an alternative development model, it has not been overwhelmingly successful in promoting it. Third, there has been little empirical evidence that the Chinese economic presence has “moved the needle” on individual attitudes (Blair et al., 2021; Eichenauer et al., 2021). Fourth, empirical evidence shows that Chinese economic engagement has led to social unrest, as measured in local protests, and lowered trust in local governments (Iacoella et al., 2021). It may well also be the case that, depending on the general perception of the US, for example, in the respective Latin American country, aid from “Western” countries does not promote a better image of democratic values. Therefore, we posit the alternative hypothesis (which is not implausible, being backed by some of the papers discussed above), that:

(H2) an increase in Chinese official aid flows does not decrease support for democracy.

3 Data description and empirical strategy

We use data from the AidData TUFF project that tracks all Chinese aid projects (Bluhm et al., 2018). To enable comparison with the OECD’s aid classification, AidData classifies each Chinese aid project into the categories of “Official Development Assistance-Like” (ODA) or “Other Official Flows-Like” (OOF) to match with OECD classifications. A third category is captured by “Vague,” which includes all official finance projects that could not be classified as ODA or OOF. We use logged total official financial flows⁴ in current US dollars per capita for each country.⁵

Between 2002 and 2013, Venezuela received the highest official Chinese flows, followed by Ecuador and Brazil.⁶ The Dominican Republic, Honduras, Paraguay, Panama, El Salvador, and Guatemala did not receive any commitments to aid between 2002 and 2013. Around 96% of project volume in current US dollars is either OOF or Vague; that is, they were probably given without a grant element of at least 25%.⁷ The 15 largest projects in Latin America were all OOF-like, and most were in the energy generation and supply sector.⁸ In terms of project size, this sector received the most commitments, followed by transportation and storage and other social infrastructure and services. The education sector received the largest aid commitment from China in terms of the number of projects.⁹

Following Brazys and Vadlamannati (2021), we include the sum of total ODA commitments from 23 DAC countries.¹⁰ Since Chinese aid data are based on commitments, we mirror these data by using DAC commitments. In our main specification, we focus on ODA. ODA is mainly given conditionally (see, e.g., Zanger, 2000) and is therefore a suitable indicator to measure the rivalry in unconditional development support. Since about 96% of Chinese commitments in terms of value come from OOF-like projects, which are clearly linked to energy and transportation, in a robustness check (presented in Table 5), we swap this metric to include OOF flows from DAC countries. In addition, we check for DAC disbursements and different donors to test the effect of different leverages. We use data on aid from DAC countries from the OECD (2020a).

The Latinobarometer household survey¹¹ provides our dependent variable on support for democracy. We code a binary indicator, which is one if people agree with the statement that “Democracy is preferable to any other kind of government” and zero otherwise. Alternatives to this statement are “In certain situations, an authoritarian government can be preferable to a democratic one” and “It doesn’t matter to people like me whether we have a democratic government or a non-democratic government.”¹² Respondents who answer “I don’t know” or “No answer” are excluded from the sample (in line with the approaches of Eichenauer et al. (2021), Mayda and Rodrik (2005), and Kleinberg and Fordham (2010)). Table 13 the appendix shows the samples available from Latinobarometer for each year and country.

The individual-level co-variables of key socio-economic indicators also come from the Latinobarometer household survey. We include household wealth,¹³ gender, education, age, and employment status. In addition, we calculate the democratic capital of each individual according to Fuchs-Schündeln and Schündeln (2015). In the following step, we include a host of variables measuring the attitudes of respondents toward political and institutional factors that may influence how greater exposure to Chinese aid can shape their views on democracy. We include indicator variables that show whether individuals have a good opinion of China and the US, whether they approve of their own president, and whether they trust people.

Country-level controls comprise variables on economic development and economic integration with China and DAC countries, respectively. To account for the level of development, we include the log of GDP per capita in current US dollars, the GDP per capita growth rate, the urbanization rate, and the GDP deflator. All variables are sourced from the World Bank Development Indicators. We also include variables to control for economic integration with China and DAC countries, respectively. Import and export volumes come from World Trade Flows (2017). Outward foreign direct investment (FDI) positions are provided by the OECD (2020b) and MOFCOM (2010, 2018).¹⁴ For all models, we include a control for the institutional environment of the country. We include a dummy if a country is an electoral or liberal democracy, according to the regime scoring from V-Dem (Coppedge et al., 2021).¹⁵

Our data structure is a mixed-level repeated cross-section for the years 2003–2011 and 2013.¹⁶ The baseline specification reads as follows:

$$dem_{{i,j,t}} = \alpha + \beta _{{CHN}} aid_{{CHN,i,j,\left[ {t - {\text{1}}} \right]}} + \beta _{{DAC}} aid_{{DAC,i,j,\left[ {t - {\text{1}}} \right]}} + X^{\prime}_{{j,[t - {\text{1}}]}} \gamma + G^{\prime}_{{i,j,t}} \delta + \phi _{j} + \mu _{t} + \in _{{i,j,t}}$$

(1)

where I is the individual living in country j in t. The regression includes country and year fixed effects to account for the trend of increasing Chinese engagement in Latin America and country-specific influence on democratic support, respectively. Country-level control variables, X′_j,[t−1], are lagged one year, while individual control variables G′_i,j,t are contemporaneous. We assume a lag of 1 year between the aid flow and its impact on the region because it is reasonable that aid takes time to show its effect. The reasoning behind this is that people react more strongly to physical output than to announcements or promises. Assuming a lag is also in line with the literature, and some of the authors even suggest longer lags. Papers have used t − 2 (e.g., Bluhm et al., 2018) or even t − 3 and more (Dreher et al., 2021). We control for this in the appendix¹⁷ and find similar and confirming results for our main model.

Throughout the paper, our models assume linear probability. To account for potential endogeneity we employ an instrumental variable two-stage least squares approach. First, endogeneity may result from measurement errors of Chinese aid (Brazys & Vadlamannati, 2021). Second, the Chinese aid flows may be non-random. It may be the case that China supports countries where a sizable part of the society does not value democracy highly. The instrumental variable approach relies on using an exogenous time-varying instrument of the donor countries interacted with a recipient-fixed probability of receiving aid from that donor. This approach follows the work of Nunn and Qian (2014), Dreher and Langlotz (2020), Dreher et al. (2021), and Brazys and Vadlamannati (2021). The interaction of the time-varying and time-invariant components of the instrument leads to a first stage that is conceptually similar to a difference-in-difference estimator. The instrumental variable compares countries that receive aid regularly to countries with a below-median probability of receiving aid, dependent on the donor-specific exogenous time-varying instrument. As Christian and Barrett (2017) have pointed out, this difference-in-difference set-up relies on contemporaneous parallel trends. To check whether the instrumental variable complies with this assumption, we split the sample into countries with above-median and below-median probabilities of receiving aid.

In Fig. 1 in the Appendix, the second graph shows the diverging development of Chinese aid for both groups. The third graph shows the development in support for democracy for both groups over time. This graph exhibits largely parallel trends in the support for democracy in countries with above- and below-median probabilities of receiving Chinese aid. Following Bluhm et al. (2018), the probability of receiving aid is calculated as the fraction of years in which a Chinese development project was introduced in the respective region during the period 2000–2014. For example, if one project was underway each year from 2000 until 2004 and none afterward, the probability of receiving aid in this region would be 1/3. The probability is potentially an endogenous component of the instrument. However, since we include country fixed effects in both stages, the variation comes from the changes in expected aid (Dreher & Langlotz, 2020; Nunn & Qian, 2014).

The identifying assumption rests on the exogeneity of the donor-specific time-varying variable. We follow Bluhm et al. (2018) in using the production of steel as a proxy for physical aid input.¹⁸ Steel production is likely to affect development aid, as an oversupply of strategic materials is exported preferentially to countries with Chinese official finance projects (Bluhm et al., 2018; Dreher et al., 2021). The exclusion restriction reads that steel production affects regions only through the channel of development aid. It seems unlikely that domestic steel production in China influences attitudes towards democracy in Latin American countries other than through the channel of aid. Following Dreher et al. (2021), we control for variables most likely related to steel production and attitudes towards democracy, namely Chinese FDI, trade flows with China, and foreign aid from OECD donors. Our results confirm that none of these variables violates the exclusion restriction, which would have been the case if they affected attitudes towards democracy differently in regions with high and low probabilities of receiving aid. To instrument Chinese aid, we follow Bluhm et al. (2018) and Brazys and Vadlamannati (2021) by interacting the probability of receiving Chinese aid p_j,CHN with Chinese steel production Steel_CHN,t.

Fig. 1

Parallel trends, probability to receive aid and support for democracy

The first stage of the instrument variable thus reads:

$$aid_{{CHN,j,t}} = \alpha + \beta _{{IV}} SteelCHN,t - {\text{3 }} * p_{{j,CHN}} + \beta _{{DAC}} aid_{{DAC,i,j,\left[ {t - {\text{1}}} \right]}} + X^\prime _{{j,\left[ {t - {\text{1}}} \right]\gamma }} + G^\prime _{{i,j,t}} \delta + \phi _{j} + \mu _{t} + \zeta _{{i,j,t}} .$$

We lag steel production three years to allow two years to build up and re-purpose the oversupply of steel for development projects (Brazys & Vadlamannati, 2021).

4 Results

Table 1 shows the IV results of five models with country and year fixed effects. Standard errors are clustered on a country-by-year basis. Model 1 shows the baseline with the instrumented log per capita aid flows that are lagged one year. As baseline control variables, we add the sum of the ODA commitments of 23 DAC countries and the dummy to show whether the country is considered an electoral democracy in the same year. The Kleibergen–Paap F statistic¹⁹ is strong with a test statistic at 27. The estimate of the first stage shows a positive relationship between the instrument and total official finance from China. Two years after the steel supply in China increases, countries that have a higher probability of receiving aid overall do in fact receive more aid than countries that have a lower probability of doing so.

Table 1 IV main results

The average marginal effect implies that higher aid flows from China increase support for democracy. Total ODA commitments from DAC countries are also associated with a positive impact on support for democracy. The positive impact of Chinese official finance flows on support for democracy counters the expectations of the relationship expressed in H1 and refutes the narrative that Chinese aid poses a threat to democracies by changing attitudes towards democracy. This positive and significant relationship stays stable over all successive models that include control variables at the individual and country levels. Model 2 adds individual-level co-variables to control for socio-economic background. The longer an individual lives under democratic regimes, the more likely it is that she prefers democracy. This is consistent with the literature on endogenous institutional preferences. Model 3 controls for individual attitudes. Interestingly, the correlation between attitudes toward China and democracy echoes results in aid flows: people who have a good opinion of China are also more likely to prefer democracy. This correlation and the main result on the relationship between official Chinese flows indicate that a “role model effect” that would promote autocratic development is not at play. Models 4 and 5 include country-level control variables. The results remain largely stable.

Analogous OLS results of our baseline specifications are reported in Table 15 in the appendix. They show for our variable of interest a positive and significant relationship between Chinese aid and support for democracy for all specifications. Judging by the lower point estimate for all OLS models, the instrument corrects for a downward bias. Since Chinese aid flows are volatile, Eichenauer et al. (2021) use the moving average of Chinese aid of one- and two-year lags. Using this approach, Table 16 in the appendix shows that the results remain stable. Another way of measuring Chinese aid is by using the new project count for each year (Table 17 in the appendix). The relationship for our variable of interest remains positive. The significance declines, which is unsurprising since there is less variation in new projects over countries and years. Lastly, we report the results of our baseline specification based on country-year averages. Again, the results remain robust.²⁰

While the results counter our third hypothesis, they are not inconsistent with results from the literature. Blair et al. (2021) find that planned Chinese aid projects initially reduce support for liberal democratic values while completed Chinese aid projects in Africa increase support for liberal democratic values. Since we do not have consistent information on completed projects, we cannot directly translate these findings to our setting.²¹ However, the results give context for our finding of the non-negative effect of Chinese aid on support for democracy. Consistent with Blair et al. (2021), we find a positive relationship between Western aid and support for democracy. In this sense, we may initially speculate that the increase in support for democracy could be driven by a movement toward “Western values” per se and may indicate that a loss of aid leverage should be less of a concern. However, this interpretation has its limits as the main result, in conjunction with correlations between attitudes towards China and the US, respectively, do not support this mechanism. If there were a shift toward “Western values,” we might expect a negative correlation between a good opinion of China and a positive relationship between a good opinion of the US and support for democracy, respectively. Moreover, work by Eichenauer et al. (2021) finds that neither Chinese nor U.S. aid impacts opinions on China and the US in Latin America. In this way, neither a “role model effect” nor a value realignment is likely to be at play here.

A potential explanation for the positive relationship between Chinese aid and democratic support is that Chinese aid could increase economic activity and thus enhance support for democracy. Dreher et al. (2021) find that Chinese aid projects increase per capita economic activity in Africa on a regional level. Similarly, Marchesi et al. (2021) show that Chinese aid projects help firms with electricity supply and overcome infrastructure constraints. Thereby, they increase regional firm performance. The study by Bai et al. (2022) suggests that an increase in welfare, through decreases in unemployment, is a potential mechanism. We will test the robustness of the results and this potential mechanism in the coming analysis. We will include an indicator to assess whether individuals are satisfied with democracy²² and how they assess their current and future economic situations, respectively.²³ We focus on individuals to check for this potential mechanism, as their assessment of aid projects and opinion on the functioning of the economy and institutions are critical in determining individual support for democracy. We thus abstract from testing the actual efficacy of Chinese aid in Latin America.

In Table 2, we investigate the effect of Chinese aid on different concepts of support for democracy. In Model 1, we add the answer to the question of whether the respondent is satisfied with the functioning of the democracy. Thus, we aim to disentangle the relationship between support for the idea of democracy and satisfaction with the functioning of the respondent’s democracy. Individuals may actually be more satisfied with the current functioning of institutions than the values behind the system. Including an indicator about respondents’ satisfaction with democracy does not affect the positive relationship between Chinese aid and support for democracy.

Table 2 Concepts of democratic support

The survey question on which our dependent variable is based has three options: “Democracy is preferable to any other kind of government,” “In certain situations, an authoritarian government can be preferable to a democratic one,” and “It doesn’t matter to people like me whether we have a democratic government or a non-democratic government.” Since our dependent variable is zero when individuals have chosen “In certain situations, an authoritarian government can be preferable to a democratic one” and “It doesn’t matter to people like me whether we have a democratic government or a non-democratic government,” the dependent variable may mask a polarization of opinions on democracy. Our dependent variable does not preclude an increase in the number of people who may also support autocracies, given an increase in Chinese aid flows. For this reason, an alternative dependent variable is an indicator variable showing whether individuals agree with the statement that “In certain situations, an authoritarian government can be preferable to a democratic one” and zero otherwise. In Model 2, the estimate of our variable of interest is negative and insignificant, providing evidence that there is no polarization of opinions on democracy at play. Consistent with our baseline model, individuals who have a good opinion of China are less likely to support an authoritarian government.

In Models 3–5, the dependent variable is sourced from a different question in the survey. The dependent variable is coded as one if respondents agree with the Churchill quote, “Democracy may have problems but it is still the best form of government.” The answer options are based on a scale of Strongly Agree, Agree, Disagree, and Strongly Disagree. There are several differences to the survey question on which the dependent variable of our baseline model is based. First, in the survey question from the baseline, respondents have the choice between two systems, while the Churchill quote offers no alternative form of government. Since the backdrop of this research question is the increasing system competition between democracies and autocratic China, the dependent variable from the baseline model seems to be a better fit. Second, the Churchill quote contains a specific negative priming towards democracy. Over all the years in our sample, more individuals responded to the question on the Churchill quote than on the democracy versus autocracy question.²⁴

All three models using the alternative measure of support for democracy show an insignificant relationship between support for democracy and Chinese aid. DAC ODA commitments are similarly insignificant, indicating that the same explanation drives the change in results for both aid flows. There are several potential sources for this variation in results. The survey question from the baseline dependent variable implicitly asks respondents to order democracy versus autocracy, while the Churchill quote is an ordering on the preference of democracy alone. In this way, it may be easier for the respondent to answer the Churchill question. While only 62% of the respondents agree with the statement that “Democracy is preferable to any other kind of government,” 77% of the same sample strongly agree or agree with the statement that “Democracy may have problems, but it is still the best form of government” (Model 5). This finding may indicate that a lack of variation drives the insignificant result. In Model 4, the dependent variable is one if and only if respondents strongly agree with the Churchill quote. While only 24% agree strongly that democracy has problems but is the best form of government, the estimated marginal effect remains insignificant. Overall, these results indicate that Chinese aid has a non-negative effect on support for democracy. Nonetheless, Chinese aid seems to affect support for democracy when framed as system competition positively. In sum, these findings imply that our H1 is not supported, unlike H2.

5 Regional aid flows

Since Chinese aid is not distributed evenly over different regions for each country, it is plausible that individuals who live in the same administrative area as an aid project have a different reaction towards aid projects than individuals living in other regions. To control for this heterogeneity within a country, we consider the effect of Chinese aid on a regional level. The default hypothesis remains that we expect Chinese aid to reduce support for democracy and Western aid to increase it. To test the regional impact of Chinese aid, we match the regional location of each individual in the Latinobarometer to the first administrative division for each country²⁵ and allocate the geocoded aid flows from AidData on this level. Since DAC aid flows are only available at the country level, we use geocoded World Bank projects (AidData, 2017) to control for leverage effects.

As the probability of receiving aid is available at the regional level, we try to construct the instruments at this level. Nevertheless, doing so poses several practical problems. Variation within the aid flows at the regional level is lower than at the national level, and many observations are zero, which makes it nearly impossible to construct reliable instruments at the regional level. Consequently, we use the country-level instruments at the regional level.

Table 3 shows the relationship between regional Chinese aid flows and individual support for democracy in a linear probability model using OLS with different levels of fixed effects. Models 1–3 include country and year fixed effects, with an increasing number of individual-level controls. For all models, the marginal effect of Chinese aid flows is positive. However, in contrast to DAC ODA commitments, regional aid commitments from the World Bank project are insignificant. This situation is consistent with country-level flows, as seen below in Table 5, and is likely to be caused by the complicated history of relations between the World Bank and many Latin American countries (Brown, 2009; Tuozzo, 2004). We are aware that World Bank aid is endogenous. The current literature (e.g., Cohen, 2023; Dreher et al., 2021) suggests the use of liquidity measures as instruments. Unfortunately, data are publicly available only from 2008, which implies that we cannot instrument half of our period. Therefore, we chose to leave it as it is to control for the effect of World Bank aid but be cautious when it comes to the interpretation. Including country-year fixed effects in Model 4 and controlling for region fixed effects in Model 5 shows that this relationship is significant at the 10% and 5% levels, respectively. To summarize, breaking down Chinese aid at the regional level does not change the main results. However, the estimated marginal impact is smaller than in the baseline model.²⁶

Table 3 Regional aid flows

To test the robustness of the differential impact of Chinese aid flows on different measures of democratic support, we test the effect of regional disaggregation, as in Table 2. Table 4 shows that regional aid flows do not alter the results at the country level. Individuals in regions with Chinese aid flows are more likely to prefer democracy to autocracy but do not necessarily view it as the best system. In both Tables 3 and 4, we see that the positive relationship between a good opinion of China and support for democracy remains very strong. In contrast, the relationship between a good opinion of the US and democracy is only positive and significant when the outcome variable is “agree with the statement that democracy is the best system” (Model 3 in Table 4).

Table 4 Regional aid flows and concepts of democratic support

6 Robustness

As a first robustness test, we replace the variable that controls for different leverages of conditional aid. The results are reported in Table 5. In Model 1, we use an instrumental variable approach to also instrument aid commitments from DAC countries. Since democracy-promoting aid payments are non-random, there may be a selection effect working, which may not only distort the estimate for DAC aid but also for total Chinese aid.²⁷ To instrument the sum of DAC aid, we follow Brazys and Vadlamannati (2021), Ahmed (2016), and Dreher and Langlotz (2020) and use the sum over all interactions between the probability of receiving aid from DAC country k with the government k’s fractionalization of the legislature.²⁸²⁹

Table 5 Leverage

The instrument is weak, as the Kleibergen–Paap F-statistic is less than 10. Furthermore, we used the test of Olea and Pflueger (2013) on weak instruments and cannot reject the null hypothesis that the instrument is weak. Therefore, we also try alternative approaches to obtain reliable instruments, for example, estimating the first stage based on a larger sample. Unfortunately, this approach results in a weak instrument as well. Thus, we can conclude that there is little to be gained compared to the baseline model. In Model 2, we include ODA disbursements from DAC countries. The results show that the actual DAC disbursements are not positively correlated with higher support for democracy, other than commitments. Our variable of interest remains robust, indicating that this relationship is not affected when Chinese commitments are measured up against actual disbursements from DAC countries. To assess the leverage effect of DAC countries compared to our variable of interest, which captures all official financial commitments from China, we include OOF flows in Model 3. In Model 4, we include ODA commitments only from the US, and the relationship between Chinese aid flows and support for democracy remains unchanged. In contrast to Chinese aid flows, ODA commitments from the US are not associated with higher support for democracy. The insignificant estimator of U.S. aid mirrors the insignificant relationship between a good opinion of the US and support for democracy. Both are likely a result of the complicated history between the US and several Latin American countries. These results highlight the importance of reexamining the results of Blair et al. (2021) for the Latin American case: the positive relationship between Chinese aid and support for democracy is obviously not explained by an alignment with Western values but driven by something different. Model 4 includes aid volumes in current US dollars from World Bank projects at the country-year level with robust results.³⁰³¹ The effect of our variable of interest remains the same.

We control for additional country-level variables to account for the dynamic domestic political economy, which is shown in Table 22 of the appendix. We include an election year indicator³² to account for swing elections (Anaxagorou et al., 2020, as cited by Dreher et al., 2021). Election years are also likely to affect how individuals assess democracy. While the election-year indicator has, as expected, a positive effect on the support for democracy, the core results remain unchanged. This holds when controlling for the Economic Freedom Index,³³ which is negatively impacted by Chinese aid (Brazys & Vadlamannati, 2021) and an index of ethnic fractionalization,³⁴ which can be a determinant of public goods provisions (Ejdemyr et al., 2018). Since Chinese aid is in large part devoted to the energy generation and supply sector, the positive aid effect we have seen could be due to large natural resource rents. While the effect of natural resource rents is positive, the positive relationship between Chinese aid and support for democracy remains unchanged.

Next, we control for additional individual characteristics and attitudes. Table 6 shows the results from our full baseline model for four additional individual controls. We include the self-reported individual positioning on the left–right political spectrum in Model 1, as individuals’ political leanings may impact how they assess the consequences of aid from the Chinese government. In Model 2, we include the average number of days respondents self-report consuming the news through media such as newspapers, radio, and TV. We include this variable because we expect that individuals who consume news regularly are more likely to be aware of Chinese aid projects and more informed about the government at home and in China.

Table 6 Additional individual controls

In Models 3 and 4, we control for the individuals’ assessment of their current economic and life circumstances. As discussed above, Chinese aid has been found to increase local economic activity (Dreher et al., 2021; Marchesi et al., 2021). Since many Chinese projects focus on infrastructure and energy supply, Chinese aid may improve the assessment of the functioning of the economy and improve life quality, for example, by reducing traveling time to work or increasing the access to and reliability of electricity. We include an indicator of whether the individuals assess their current economic situation as bad. We expect that if Chinese aid is assessed as effective by citizens, fewer will state that their current economic situation is bad, and more will be more likely to support democracy. In Model 4, we control for the life satisfaction of individuals. This model tests whether the positive relationship between Chinese aid and support for democracy is channeled through life satisfaction. If Chinese aid projects increase the perceived quality of individuals’ lives, enhancing life satisfaction, this may increase the probability of supporting democracy. However, including variables that control for the current personal situation of individuals does not impact the significance of the estimate.

It may be that the relationship between Chinese aid and support for democracy is impacted by the change in individuals’ assessment of the future economic situation instead of their current one. In this way, Chinese aid would generate optimism about the personal and national economic future, increasing support for democracy. This is what we test in Models 5 and 6. In Model 5, we include an indicator of whether individuals state that they believe their future economic situation will be better; in Model 6, we include the variable that individuals stated that the country’s future economic situation will be better. The inclusion of each variable does not alter the results.

Lastly, we are interested in whether the experience of corruption interferes with the positive relationship. It has been well documented that Chinese aid has been linked to corruption (Isaksson & Kotsadam, 2018), and corruption is linked to an erosion of regime legitimacy (Seligson, 2002). Model 7 includes an indicator of whether individuals or their family members have witnessed corruption. The marginal effect is, as expected, negative. However, the variable of interest remains unchanged. The results are replicated at the regional level; see Table 23 in the appendix.

We include additional individual-level controls, not only to check the robustness of the positive and significant relationship between Chinese aid and support of democracy but also to learn about potential channels. It is of particular interest to find out whether Chinese aid has increased the number of individuals who have a positive outlook on the current and future economy or a positive evaluation of the current administration and institutions, increasing their support for democracy. The inclusion of these variables has not changed the positive relationship found in the baseline. We conclude that this mechanism is not at play at a national level. To check this conjecture, we replace our dependent variable with the following variables: support for the current president, satisfaction with democracy, current economic situation, life satisfaction, and beliefs on the personal future and economic situation. The results in Table 24 of the appendix provide further evidence that the positive relationship between Chinese aid and support for democracy is not explained by a change in satisfaction with current institutions or expectations for improvement in the future.

To better understand who supports democracy conditional on Chinese aid, we split the sample across different individual characteristics. Table 7 shows the results of eight sample splits based on the baseline model, including individual- and country-level controls. The first sample split is across age differences. We split the sample into individuals born before and after 1985. The underlying logic is that younger individuals will not have personal memories of the Cold War and may thus view donors differently. The sample split shows that the estimated marginal effect for younger individuals is positive but insignificant, while the sample based on older individuals is positive and significant. As the following sample split shows, individuals with above-mean democratic capital show a positive and significant effect of Chinese aid on support for democracy while individuals with below-average democratic capital do not. Along the urban/rural split, individuals living in cities with a population above 50,000 are more likely to support democracy than rural-based individuals. A sample split along education shows that more educated individuals exhibit a significant change in support for democracy with an increase in Chinese aid. Left-leaning individuals show a significant and positive relationship between Chinese aid commitments and their support for democracy while right-leaning individuals do not.

Table 7 Sample splits

These differences in the significance of estimators hint that there may be heterogeneity in the reaction toward Chinese aid. However, all estimators of the sample split suggest a positive relationship between Chinese aid and support for democracy. In addition, overlapping confidence bands (at the 95% level) indicate that the estimators are not significantly different from each other. In the case of differences in wealth, news consumption, and experiences with corruption, neither confidence bands nor the significance of the positive estimator indicate any difference. Overall, the sample split suggests that a broad coalition of society in Latin America does not decrease its support for democracy in the face of increasing aid from China.

7 Conclusion

As the strategic competition of the US with China has become a foreign policy focus of the US, the debate around Chinese international economic engagement has become more centered around democratic values. This paper adds to the growing literature on the effects of Chinese foreign aid and attitudes. We find that for Latin America, Chinese aid has no negative effect on individuals’ support for democracy. In fact, Chinese foreign aid positively affects individuals’ attitudes toward democracy when they are asked a question directed at competition between democracy and autocracy. While ODA commitments from DAC countries are positively correlated with support for democracy, ODA disbursements from DAC countries, ODA commitments from the US, and World Bank projects are not.

This effect is mirrored in the relationship between individual-level attitudes towards China and the US and support for democracy, respectively. There is a strong, robust relationship between a positive opinion of China and support for democracy across all specifications. A positive opinion of the US does not translate into higher support for democracy. In conjunction with our main results, this evidence is a hint that there is no “role model effect” of China promoting an autocratic development strategy. Our results may provide relief for policymakers concerned about the current impact of Chinese aid on civil society and demonstrate the agency of civil society in recipient countries in forming citizens’ opinions on democracy.

Appendix

See Fig. 1 and Tables 8, 9, 10, 11, 12, 13, 14, 15, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25 and 26.

Table 8 Chinese aid in current million US $ by country

Table 9 Chinese aid in current million US$ by aid class

Table 10 Top 15 Chinese aid projects in current million US$

Table 11 Chinese aid per sector in current m US$

Table 12 Ranked countries by dependent and key variables

Table 13 Sample support for democracy

Table 14 Summary statistics

Table 15 OLS main results

Table 16 IV moving average

Table 17 Number of Chinese projects

Table 18 BL IV country averages

Table 19 Number of projects country averages

Table 20 Moving average country averages

Table 21 Concepts of democratic support country averages

Table 22 Additional country-level controls

Table 23 Additional individual-level controls on regional level

Table 24 Dependent change channels

Table 25 IV main results with $_{t-2}$

Table 26 IV main results with $_{t-3}$