Foreign Aid, Norm Diffusion, and Local Support for Gender Equality: Comparing Evidence from the World Bank and China’s Aid Projects in Africa

Abstract

Does foreign aid change local attitudes toward gender equality? Drawing on a theory of norm diffusion through intervention, we contend that when donors integrate norms into aid projects, such norms can be diffused to individuals in recipient countries. The World Bank and China are two comparable donors to examine this relationship. We matched the geocoded aid projects of the two donors from AidData and Afrobarometer surveys in Africa to test our hypotheses. Our strategy is to compare respondents’ attitudes toward gender equality who were near a site where the project had been implemented at the time of the interview to those who were close to a site where the project would be started after the interview. We find that the World Bank aid increases local support for gender equality and the results differ across sectors and gender. Aid going to gender-sensitive sectors has the most significant impact on women’s attitudes toward gender equality. In contrast, Chinese aid shows no similar effects. This study sheds light on the linkage between aid and social norm change and is expected to exert implications for policymakers and development practitioners adventuring in the land of aid.

Introduction

Promoting gender equality is one of the main aims of development projects provided by traditional donors and multilateral development agencies. Funding for reducing gender inequality was at an all-time high at 42.1% of bilateral official development assistance (ODA) in 2018 (OECD 2020). As a leading donor agency, the World Bank Group (the Bank) has been advocating gender equality in development since 1977 and began instituting policies to address gender issues in its projects in the late 1990s (World Bank 2010a). The Bank advanced its gender agenda over the past two decades following gender mainstreaming (GM) strategy which was introduced by the Bank in 2001 followed by comprehensive Gender Action Plans.¹ The strategy integrated gender into all stages of development programs (strategies, design, implementation, monitoring and evaluation) (World Bank 2002; Bardasi and Garcia 2016).² By 2014, 95% of the World Bank projects integrated gender in at least one dimension, 82% in two dimensions, and 55% in three dimensions (World Bank 2014). As an international development “Knowledge Bank,” the Bank not only constructs gender agenda, formulates gender policy, but also diffuses gender norms to the rest of the world (Park and Vetterlein 2010). It is not exaggerating to say that the Bank is a “normative power” in diffusing gender equality norm since the beginning of the new millennium.

Compared with the World Bank, gender issues were largely ignored in the development projects provided by China and rarely discussed in China's foreign policies (Cai 2021; Cai and Liu 2016). The concept of gender was introduced into China in the 1990s along with Beijing’s host of the Fourth World Conference on Women. Chinese feminist scholars and activists used the concept to reinterpret the “equality between men and women” from the aspects of political, economic, reproductive rights, and autonomy (Hsiung et al. 2002; Kaufman 2012; Hsiung 2021). Gender equality to a large extent means promoting women’s economic empowerment, which is deeply rooted in the Marxist materialist view on women’s liberation in the Chinese political context. Promoting gender equality was only mentioned for the first time in China’s latest white paper on foreign aid (The State Council 2021). China views women’s participation in politics as a domestic affair that should not bear foreign interference, which provides a good reason for China not integrate gender norms into its aid projects.

Efforts have been given to assess the impact of international aid on gender equality in recipient countries but no consensus has been reached (Swain et al. 2020). Women’s gains in economic, political, and social sectors from a single donor were usually emphasized (Miedema et al. 2018) while the attitude change toward gender equality influenced by aid projects among different providers was left untouched except one research by Beath et al. (2013).³ Although gender norm diffusion through international organizations (including the Bank) has received considerable attention (Weaver 2010; Park 2005; Finnemore 1993; Avant et al. 2010), the impacts of norm diffusions through aid projects provided by the international organizations on the individuals in recipient countries are rarely explored. The mechanisms linking aid with norm change on gender equality have yet to be identified. Drawing on a theory of norm diffusion and using newly released geocoded datasets of aid projects from China and the Bank, we theoretically link aid with gender norm diffusion by identifying the Bank as a norm maker and individuals in recipient countries as norm takers. The aid project implementation process is also a process of international norm diffusion. Therefore, we argue that donors can spread gender norms in recipient countries through a norm diffusion mechanism. In other words, if gender norms are integrated into aid projects, these projects can help to shape local attitudes toward gender equality. Meanwhile, the acceptance of gender norms from projects varies across aid sectors and the local population: gender-sensitive sectors work better and women are more likely to accept gender norms than men. We take the World Bank and China’s aids to Africa as two cases (the former integrates gender norms in projects while the latter never). Africa is chosen for three main reasons: (1) gender disparity is more serious in low-income areas than in higher-income societies (World Bank 2001); (2) Africa is the largest recipient of the World Bank and China’s aids⁴; (3) the geocoded data of our interest—local attitudes toward gender equality is only available in African countries.

To test our theoretical expectations, we use a causal identification strategy to compare the attitude toward gender equality of respondents who were near a site where the project had been implemented at the time of the interview (as a treated group) to individuals who were close to a site where the project would be started after the interview (as a control group). We find that the World Bank aid is positively associated with the local population’s support of gender equality while aid from China is not and the statistical significance is driven by women and gender-sensitive sectors samples. Women near the aid sites of the Bank for government and civil society and education demonstrate highest support for gender equality.

Our study contributes to the literature in the following aspects. First, different from the literature on gender and development, which emphasizes women’s objective gains, we investigate women’s subjective gain—attitude toward gender equality and find that gender norm diffusion through aid intervention can increase local support of gender equality. Second, we link aid with gender norm diffusion to the local population by identifying a possible mechanism—integration of gender norms into aid projects. In addition, by comparing the impacts of the World Bank aid and China aid, we provide systematic evidence that integration of gender equality norm in aid projects can bring positive results in shaping local gender attitudes. Moreover, different from the previous evaluation of aid effectiveness on gender equality using aggregated data (e.g., cross countries analysis), our research exploits the impact of aid using project-level data, which enriches the literature on why and how foreign aid shapes local public opinions such as the preferences for development (Huang and Cao 2019), political trust (Watkins 2021), institutional legitimacy (Blair and Roessler 2021), and ethnic identification (Isaksson 2020), by focusing on how international gender norms spread to the Global South through aid intervention. By exploring a comprehensive, theory-driven comparison of the impacts of aids from the World Bank and China, we prove that external social norms like gender equality can be diffused to the local population through development projects, responding to a question largely ignored by the aid effectiveness literature and policymakers.

The rest of the paper proceeds as follows. After a brief review of the background and literature, we present our theoretical expectations and the data and empirical strategy followed by the sections reporting our main results, robustness checks, and additional evidence. The final section concludes and discusses the future research.

Background and Literature Review

Gender Mainstreaming and the World Bank

As a major global strategy to address gender inequality, gender mainstreaming was established for the promotion of gender equality in the Beijing Platform for Action from the Fourth United Nations World Conference on Women in 1995. In 1997, the Economic and Social Council of the United Nations (ECOSOC) called for “mainstreaming the gender perspective into all policies and programs in the United Nations system” and developed important principles for gender mainstreaming (ECOSOC 1997). In the conclusions, the concept of gender mainstreaming is defined as integrating gender perspectives into development.⁵ Gender mainstreaming has been adopted by many international organizations, governments, and non-governmental organizations (NGOs) in aid practices to an unprecedented degree over the past decades (Zwingel 2012:115).

Gender mainstreaming has been fully institutionalized into policy-making, planning, implementation, monitor and evaluation of the World Bank development projects. In 2002, the Bank released its initiative of gender mainstreaming application (World Bank 2002). The Bank not only provided a holistic framework to address gender equality in its aid projects but also formulated sectoral policies and toolkits for project practitioners. To support the integration of gender into program operations, the Bank developed concrete measures to ensure a conducive institutional environment for gender mainstreaming (World Bank 2002).⁶ Up until now, gender mainstreaming has been adopted by the Bank for over 20 years. In 2016, the Bank concluded in a report that the strategy “was undoubtedly successful” (World Bank 2016: 2).

Aid Effectiveness and Gender Norm

The aid effectiveness literature over the past decades has largely focused on whether development aid has enhanced the economic and social performance in recipient countries. Scholars have explored different aspects of development that can be influenced by aid, such as democratization (Scott and Steele 2011), institutional reform (Jones and Tarp 2016), and conflicts (Strange et al. 2017). Aid intervention and gender first appeared in official development discourse in the 1980s (Tarp 2002:197). Since then, gender performance has become one of the core indicators of aid effectiveness.

Some country-level studies have been conducted to examine the effects of aid allocation and outcomes for women in recipient countries. For instance, Montinola and Prince (2018) demonstrate that women’s empowerment results in a decrease of infant mortality in recipient countries. Pickbourn and Ndikumana (2016) suggest that aid for the health and education sectors reduces maternal mortality and gender inequality in youth literacy. Baliamoune-Lutz (2016) proves that aid increases women’s political participation in parliaments. Swain et al. (2020) reveal that even when aid is targeted at improving gender outcomes, it is unlikely to change gender inequality in recipient countries. Therefore, there is no consensus in aid effectiveness literature on whether aid can improve gender equality at the country level. In this research, moving from country-level studies, we link project-level aid data with individual-level attitudes toward gender equality to measure the specific impacts of aid on local attitudes toward gender equality.

Also, the change in gender norms has been investigated by scholars (Roughley and Bayertz 2019). Variations in public support of gender equality are attributed to economic structure (Iversen 2010), technological progress (Albanesi and Olivetti 2007), employment (Ross 2008), and political participation (Beaman et al. 2009). Although, a consensus emerges that the improvement of women’s economic and social conditions is related to changes in social norms such as gender equality and other beliefs (Robinson and Gottlieb 2019; Gottlieb 2014; Giuliano 2015, 2018), only a few empirical studies focus on the connections between foreign aid and public support of gender equality in recipient countries. And scholars have studied aid and gender norm changes mainly through case studies (Konte 2019; Etim 2016; van der Vleuten et al. 2014; Bamberger 1997), project or policy evaluation (Beath et al. 2013; Malapit et al. 2019; World Bank 2022; Dixon 1980). So far, theoretical pathways through which aid are linked to gender norms have not been clearly identified, and systematic evidence on how aid generates local support for gender equality using micro-level data has yet to come.

The World Bank Versus China

Foreign aid, no matter it is from traditional donors or emerging donors, is becoming more controversial than before. And comparing the convergence or divergence of traditional donors and non-traditional donors has become a trend in aid literature (Brautigam 2011; Zeitz 2021). It is expected that the difference between Western donors and China has distinct impacts on local socioeconomic consequences (Brautigam 2011; Dreher et al. 2021). For example, some scholars find that Chinese aid is associated with local corruption, while the Bank aid is not (Isaksson and Kotsadam 2018a). Chinese aid projects in Africa are correlated with local preferences for economic leadership but the Bank aid projects do not share a similar pattern (Huang and Cao 2019).

There are several comparative advantages of the Bank in spreading gender equality norm to people in recipient countries, in contrast to other donors (World Bank 2002:62). The first is the Bank’s expertise and capacity in penetrating gender perspective into aid projects. And gender mainstreaming approach fits well with the Bank’s objectives in promoting democracy in recipient countries. In addition, the Bank has a strong network linking recipient countries, civil society groups, and other development agencies, as well as extensive on-the-ground experience in promoting its advocacy through different platforms and dialogues.

The Bank and China are two comparable cases to explore the relationship between aid and local gender equality. We trace the differences of the two donors from three dimensions: gender policy formulation, gender project implementation, and the support to the gender equality. We find that gender equality norm has been fully incorporated into the Bank’s projects while China never did (Table 1).⁷ In terms of offering aid, the Bank believes that conditionality and interference are necessary instruments to make its development project effective while China adheres to the principles of South-South Cooperation of non-conditionality and non-interference when providing foreign aid (Zhang et al. 2018, 2021).

Table 1 Differences in aid principles and policies concerning gender norm diffusion between the Bank and China

As for gender norms, the Bank regards gender equality as a matter of human rights and social justice which is essential for achieving overall development goals. And it has integrated gender equality norm in its development programs for decades with a clear strategy. In contrast, China prioritizes economic benefits over norm diffusion in its foreign aid. At the project level, the Bank’s gender mainstreaming strategy integrates gender equality norm into all projects and all stages of projects: design, implement, monitor and evaluation. And also, the Bank provides various resources including budget, training, policy tools, and expertise to support the application of gender polity in its programs. There are no similar measures or policies in the Chinese aid projects.

A Theory of Gender Norm Diffusion Through Aid

Norm diffusion, by definition, is the transfer or transmission of objects, processes, ideas, and information from one population or region to another (Hugill and Dickson 1988: 263–64). Norm diffusion has received lots of attention in international relations, anthropology, geography, and sociology (a brief review, see Cortell and Davis 2000). Diffusing social norms per se is an important policy tool in international relations (Humphreys et al. 2019; Acharya 2004; Meyer and Prügl 1999; Zwingel 2012). Prominent examples of norm diffusion include the spread of democracy (Brinks and Coppedge 2006), human rights (Greenhill 2010), market reforms (Meseguer 2009), environmental protection (Holzinger et al. 2008), global capital taxation (Cao 2010), and also gender norms (Hughes et al. 2015).

Why do some transnational ideas and norms find greater acceptance in a particular locale than in others? Acharya (2004) argues that localization is an important variable to explain norm diffusion. The dynamics of norm diffusion include but are not limited to societal pressure (Risse-Kappen 1995), elite learning (Katzenstein 1996; Checkel 1997), localization (Acharya 2004), and social environment (Savery 2007). For example, van der Vleuten et al. (2014) unpacked how gender norms travel and are negotiated at the regional level and conceptualized norm diffusion as a truly multilevel and multi-actor process.

In terms of gender equality norm, which is more relevant to our study, previous studies on global gender norm diffusion find international advocacy and foreign policy for gender equality matter for translating gender norms to different countries (True and Mintrom 2001; Gray et al. 2006; Zwingel 2012). These researches analyze how states come to agree and adopt international norms and overwhelmingly use country-level analysis or focus on political elites. As correctly pointed out by Restoy and Elbe (2021), “very little international relations scholarship drills down deep enough to examine whether and how new international norms are subsequently integrated, incorporated, and translated at sub-state level.”

Another closely related literature is the studies of norm diffusion in international organizations (Finnemore 1993; Avant et al. 2010). International organizations, such as the World Bank, were considered as active norm diffusers in all aspects of international relations and act as instigators of change in diverse areas like development, environment, human rights, as well as gender equality (Finnemore 1996; True and Mintrom 2001). The socialization and norm diffusion process of international organizations by state and non-state actors through international systems has been expounded by a few scholars (Checkel 1999; Johnston 2001; Park 2005).

And the World Bank, as the largest financier of international development aid, is a self-identified “Knowledge Bank” for development. Weaver traces the construction of the World Bank’s gender and development policy norm and finds that Beijing’s Fourth World Conference on Women in 1995 was a critical juncture that produced a major opportunity for external and internal advocates to push gender mainstreaming (Weaver 2010:70–89). Both leadership within the Bank and gender activism outside the Bank supported the new strategy of gender equality. Park (2005) argues that the interaction between the World Bank and non-state actors such as transnational advocacy networks shapes norms diffusion within the Bank. Abouharb and Cingranelli (2006) find the implementation of the Structural Adjustment Projects imposed by the Bank worsen government respect for physical integrity rights like torture and political imprisonment.

In the development field, the gender mainstreaming strategy has been widely adopted by both traditional donors and multilateral development agencies to promote gender equality since the 1990s. This strategy incorporates gender issues into all stages of policy-making, program design, project implementation, monitor and evaluation. In this paper, we argue that when aid projects are integrated with gender equality norm, the norm can be diffused to the local population. As we mentioned earlier, gender mainstreaming is arguably the world’s most powerful instrument of global gender norm integration strategy which has been fully adopted by the Bank since the early 2000s.

Hence, we argue that a mechanism of norm diffusion through foreign aid connects foreign aid and the local attitudes toward gender equality. Meanwhile, improvement of economic status without norm diffusion does not necessarily change local attitudes toward gender equality. To put it another way, we contend that foreign aid works as an instrument of spreading gender norms when these norms are incorporated into project design, implementation, monitoring and evaluation.

The dynamics featuring norm diffusion through aid projects that impact local attitudes toward gender equality can be summarized in Fig. 1. In the first place, only individuals living near project sites are relevant to our theory of gender norm diffusion (we define “near” in the empirical section). People who live far away (e.g., 1,000 km) from the aid sites can scarcely be influenced. Thus, changes in gender norms should be more salient for people living near aid sites. There are three important indicators that determine the outcomes of gender norm diffusion by aid: (1) whether a project is integrated with gender equality norm; (2) to what extent gender equality norm has been integrated into a project; (3) who will be the norm takers in local population—women or men (in our case).

Fig. 1

Theoretical framework

Integration of norms in aid matters for local attitudes toward gender equality. As an external intervention, aid with gender equality norm would awaken local awareness of gender equality of residents exposed to these aid projects. Local people can reflect the appropriateness and legal context of their roles both in the household and in society. As a result, we expect foreign aid with gender equality norm to increase local support for gender equality. In this sense, we propose that the local people living near the Bank aid sites are more likely to accept gender equality norm and support gender equality while those who live near Chinese aid (where gender norms are absent) are not.

Gender equality norm is not evenly incorporated in all aid sectors. Aid sectors are coded by AidData according to the nature/purpose/usage of the projects.⁸ The GM strategy seems to be better integrated into some sectors of aid than others. In some aid sectors like education and government and civil society, which notably address gender inequalities and biases, gender norm is more easily to be diffused. We define these sectors as gender-sensitive sectors and projects belong to this category as the gender-sensitive projects.⁹ A gender-sensitive project contains gender-responsive actions actively promoting equality—specific actions to empower women in their households, in their communities, in their societies as well as in broader political and planning processes. The Bank provides overwhelmingly 16.62 times more government and civil society projects to Africa from 2000 to 2014 than China (Table 2). Also strikingly, among the Bank’s aid (main sectors) to Africa in our study, more than half of them (53.47%) falls in the government and civil society sector (Table 2).

Table 2 Aid to Africa by sectors: the Bank versus China, 2000–2014

Although education does not guarantee rights, it is critical for empowering women. When women receive more education, they are more able to participate in community and national decision-making process, and better able to protect themselves from gender-based violence (World Bank 2001). For example, over USD 3.4 billion for girls’ education programs were provided by the Bank between 1995 and 2000 (World Bank 2002). Collaboration with civil society groups has been emphasized by the Bank (World Bank 2002).

Meanwhile, in others sectors like energy, transport and storage, and water supply and sanitation—which do not directly target gender equality, therefore to incorporate gender norms is less likely. Our theory proposes that residents who live close to gender-sensitive projects are more likely to accept gender equality norm. But local people near other aid projects are neither less likely nor more likely to accept gender norms.

The different characteristics of norm takers must be taken into consideration. The gender differences across policy domains can vary. Gottlieb et al. (2018) finds greater financial independence narrows the gender gaps of policies in the labor and infrastructure while gender gaps in political representation and participation remain. There are other obstacles that are more persistent than economic improvement for women in developing countries, for instance, the patriarchal political and social system (Robinson and Gottlieb 2019; Brule and Gaikwad 2018), because the rise of women’s awareness of gender equality challenges the traditional social norms (and also institutions) that codify men as dominant roles both at home and in society. It is possible that foreign aid does not increase men’s support of gender equality. And previous studies also find women are more likely to support equal gender rights than men (Logan and Bratton 2006; Robinson and Gottlieb 2019). Similarly, we propose local women are more likely to accept gender equality norm when exposed to aid projects integrated with such norms. On the contrary, local men living close to the aid sites with gender equality norm are immune to such norms.

Hence, following the above discussion, three key predictions with two sub-hypotheses for quantitative exercises are devised:

• H1: The integration of gender norms into aid project increases local support for gender equality. Individuals exposed to the World Bank aid projects are more likely to support gender equality while those exposed to Chinese aid aren’t.

• H2.1: Women in proximity to aid projects in which gender norms are integrated (the World Bank) are more likely to support gender equality than men.

• H2.2: Individuals who live near gender-sensitive projects (of the World Bank) are more likely to support gender equality.

• H3: The effects of aid projects on local support of gender equality are most salient among the group of women living close to gender-sensitive projects.

Data

To investigate the relationship between foreign aid and local support of gender equality, we compared the impacts of aid projects from the Bank and China from 2000 to 2014 in African countries that were covered by two rounds of the Afrobarometer survey (Round 3 and Round 5). The geocoded Afrobarometer and AidData data are with similar coding methods (BenYishay 2017). The two datasets are matched to investigate the impact of aid projects on changes in local attitudes toward gender equality.

Afrobarometer Data

The Afrobarometer project released its geocoded data in 2017. These datasets have been used to investigate the local effects of development projects or aid projects (e.g., Isaksson and Kotsadam 2018a). Similarly, in this study, we use one particular question (Q23) from Afrobarometer Round 3 and Round 5, which asked respondents’ attitudes toward gender equality: “Which of the following statements is closest to your view? Choose Statement 1 or Statement 2”¹⁰:

• Statement 1: In our country, women should have equal rights and receive the same treatment as men do.

• Statement 2: Women have always been subject to traditional laws and customs and should remain so.

Respondents can choose the following options: “Agree very strongly with Statement 1,” “Agree with Statement 1,” “Agree with Statement 2,” “Agree very strongly with Statement 2,” “Agree with neither,” “Don’t know,” and “Refused to answer.” Theoretically, a respondent who supports gender equality would choose Statement 1, while a respondent who holds the opposing view would choose Statement 2. Thus, we use the respondents’ answers to this question to capture their attitudes toward gender equality.

We then construct two samples of the World Bank and China by combining Round 3 and Round 5 Afrobarometer surveys. Afrobarometer provides point coordinates (latitude and longitude) as clusters for survey respondents. There are eight precision categories (from level 1 to level 8), and lower numbers indicate more precise locations (Strandow et al. 2011). We use only two precision categories (level 1 and level 2) to ensure accurate locations: level 1 means an exact location and level 2 means near an exact location (up to 25 km from that location). The selection of precision categories is consistent with the literature (Knutsen et al. 2017; Isaksson and Kotsadam 2018a). The total number of respondents includes in our analyses is 40,002 (the Bank) and 41,977 (China), respectively (see Fig. 2).

Fig. 2

Distribution of aid projects and 35 km buffer zones around Afrobarometer survey clusters

Aid Data

To compare the impact of aid from different donors on local gender equality, we use geocoded World Bank and Chinese aid project data. Both datasets are extracted from the AidData project (Bluhm et al. 2018; Strange et al. 2017), which provide aid with coordinates. To make our comparison consistent, we use both types of aid projects from 2000 to 2014. In this study, we use only the most accurate information of a project—that is, when and where each aid project was implemented.

Each aid project is georeferenced with point coordinates. We match aid projects with clusters of survey respondents. Projects without coordinates are excluded. We retain only clusters with relatively accurate coordinates and drop samples with precision categories higher than 2. We keep aid projects within the spatial precision category as 1 and 2. We use only aid project samples covering 34 African countries where the Afrobarometer Round 3 and Round 5 survey were conducted.¹¹ Finally, we obtained two separate aid project samples: 3,600 from the Bank and 1,130 from China (see Fig. 2).

Research Design

The allocation of aid is unlikely to be random, as some areas are more likely to receive projects while others not—which creates aid allocation bias. If the Bank aid targets areas where the local attitudes toward gender equality was well established and Chinese aid is allocated to areas with lower or medium support for gender equality—this would lead to a biased estimation. In addition, given that an ex-ante measure of the local support for gender equality is unavailable, it is difficult to learn whether the changes in local attitudes toward gender equality is due to aid intervention rather than some other unobserved factors that might affect the location choice of aid projects.

To address this concern, an estimation strategy taking advantage of the spatial–temporal nature of the data is employed in our study (earlier applications of this design, see Knutsen et al. 2017; Brazys et al. 2017; Isaksson and Kotsadam 2018a; Isaksson and Kotsadam 2018b; Huang and Cao 2019; Isaksson 2020).¹² The strategy is to compare local attitudes toward gender equality of two groups of respondents: (1) individuals who were living close to aid projects (already implemented) at the time of the interview, and (2) respondents who were near sites of aid projects where the projects would be started after the interview. The key assumption is that the areas being treated by aid projects are similar to areas that are soon to be treated. This strategy helps to address the abovementioned allocation bias of aid projects. The baseline regression of this estimation strategy is as follows:

$${Y}_{\mathrm{ivt}}={\beta }_{1}{\mathrm{active}}_{\mathrm{it}}+{\beta }_{2}{\mathrm{inactive}}_{\mathrm{it}}+{\alpha }_{s}+{\delta }_{t}+\gamma {\mathrm{\rm X}}_{\mathrm{it}}+{\varepsilon }_{\mathrm{ivt}}$$

(1)

${Y}_{\mathrm{ivt}}$ is the response to one of the questions asked in question 23 of the Afrobarometer survey for individual i in cluster v and year t. We use OLS regressions for our main results. We recode the support of Statement 1 (a positive attitude towards gender equality) as 1 (“Agree very strongly with Statement 1” and “Agree with Statement 1”) and the support of Statement 2 as 0 (“Agree very strongly with Statement 2” and “Agree with Statement 2”), and other answers are coded as missing. Through this approach, we recode the respondents’ answers regarding the support for local gender equality as a binary variable (equal to 1 if respondents support gender equality and 0 otherwise). ${\delta }_{t}$ represents year fixed effects (the time of surveys vary in different countries)¹³; ${\alpha }_{s}$ is sub-national region fixed effects.¹⁴${\varepsilon }_{\mathrm{ivt}}$ are standard errors clustered at the Afrobaromter survery cluster level to account for any exogenous shocks correlated by location.

γΧ includes individual characteristic variables which are provided by Afrobarometer surveys. Among possible confounding variables, we use age, age squared, gender, education, and urban as the main controls. Respondents of different ages might answer survey questions regarding women’s rights differently. Education is also an important socioeconomic factor that may influence individual perceptions of gender rights. The gender and development literature indicates that more educated citizens are more likely to approve of equal gender rights (Jones 2018; Lee and Shin 2019). The variable of age is the exact age of a respondent. Education measures the highest level of education a respondent has completed, ranging from 0 to 9. For gender, 1 is male and 2 is female. Being from an urban area is coded as 1, and being from a rural area is coded as 2.

${\mathrm{active}}_{\mathrm{it}}$ is a binary variable indicating whether a respondent i at year t was close to at least an implemented aid project at the time of being interviewed; ${\mathrm{inactive}}_{\mathrm{it}}$ is also a dummy variable indicating whether a respondent was living near to a project that was yet not implemented by the time of the survey. The causal effect of an aid project on local support of gender equality is then the difference between ${\beta }_{1}$ and ${\beta }_{2}$, which captures the difference of effects between post-treatment individuals and pre-treatment respondents regarding to the reference group—which is not close to an implemented project or a future project. We use 35 km as the cut-off to define proximity to an ongoing or future aid project.¹⁵

We code active35 equals 1 if there was one project that existed within 35 km of a respondent before the year of the interview, otherwise 0. Similarly, inactive35 is coded as 1 if there would be a project within 35 km shortly as the time of the interview, otherwise coded as 0. In the end, there are 23,459 active and 16,543 inactive respondents for the World Bank aid sample. And there are 15,707 active and 26,270 inactive observations for the Chinese aid sample. Table 3 presents the descriptive statistics for the variables used in our main analysis.

Table 3 Descriptive statistics

Main Results

To test our hypotheses of the differences between donors, sectors, and gender regarding the impact of aid intervention on local attitudes toward gender equality, we create three samples: the full sample, the women sample, and the gender-sensitive sector sample for the World Bank aid and Chinese aid, respectively. Table 4 reports our benchmark results that include individual controls, year fixed effects, and sub-national region dummy variables. Again, the causal effect of aid is captured by the difference between the coefficient of active35 (${\beta }_{1}$) and that of inactive35 (${\beta }_{2}$): calculated as “difference in difference” in our regression tables. An F-test is used to determine whether this difference in difference estimate is statistically significant.

Table 4 World Bank aid, Chinese aid, and local support of gender equality

The first column of panel A of Table 4 shows that respondents who lived close to an active World Bank aid site are more likely to support gender equality than respondents who were near an inactive World Bank aid project. The results show that those with an active rather than an inactive World Bank project in their vicinity are 2.4 percentage points more likely to support the statement that “women should have equal rights and receive the same treatment as men do.” By comparison, respondents in the Chinese aid sample do not share a similar pattern: there is no significant difference between the active group and the inactive group (column 1 of panel B).

The second and third columns of Table 4 present the difference in effects regarding gender division. The results indicate women are more likely to support local gender equality when exposed to the World Bank aid sites while men are not. This confirms our gender gap hypothesis in the theoretical section. Women with an active World Bank aid project nearby are 4.3% more likely to support gender equality, statistically significant at the 1% level. Again, there is no significant difference in effects in the Chinese sample when separated by gender.

As for the gender-sensitive sector which donors integrated more with gender norms, the results indicate that those with an active rather than an inactive gender-sensitive project from the World Bank are 5.7 percentage points more likely to gender equality, significant at the 1% level (column 4 and column 5, panel A of Table 4). We find the gender gap is persistent in the gender-sensitive sector: only women are more likely to change their attitude toward gender equality (thinking that men and women should be equal) when World Bank aid projects in their vicinity. No regression achieves statistical significance in the Chinese aid sample which suggests that Chinese aid has no impact on local attitudes toward gender equality.

Interestingly, if we compare the coefficients of active35 and inactive35, we find that respondents close to active World Bank aid projects are positively associated with higher support of gender equality (to the reference group with no active or inactive projects in their vicinity) while being close to active Chinese aid do not achieve any significance (Table 4). And the coefficients of inactive35 are not significantly correlated with local attitudes towards gender equality, except for a few (Table 4). This suggests that residents in the pre-treatment condition—those who were living near soon-to-be treated aid projects are unnecessarily more likely to support gender equality. Thus, the captured effects of the World Bank aid are an increase in individuals’ attitudes in favor of supporting a social norm of gender equality after receiving the Bank’s projects.

We find considerable disparities across different samples by the World Bank aid project on local attitudes toward gender equality (Fig. 3). In contrast to the inactive group, on average, residents who lived near the aid site are 2.4 percentage points more likely to support gender equality; women close to active aid sites are 4.3 percentage points more likely to favor gender equality; residents who were near gender-sensitive aid projects are 5.7 percentage points more likely to support gender equality; and women are 8.7 percentage points more likely to support gender equality when exposed to the gender-sensitive aid projects (Table 4). The differences are 3.2%, 5.8%, 7.7%, and 11.7%, respectively, to the overall sample mean of the outcome (which is 0.738 in Table 3). These results suggest that gender-sensitive projects from the World Bank perform best in changing local women’s attitudes toward gender equality—which is consistent with our theory of diffusion of gender norms through foreign aid.

Fig. 3

The size of estimated effects: the World Bank. Estimated effects with 95% confidence intervals. Difference is the coefficients of treated groups (active35) subtract control groups (inactive35)

The appropriate cut-off distance from a project—within which individuals are coded as treated group is an empirical question (Knutsen et al. 2017; Isaksson 2020). With smaller thresholds of distance, we get smaller samples of respondents to link to active and inactive projects. However, a larger cut-off distance would include too many untreated respondents in the treatment group, leading to attenuation bias (Isaksson 2020). Figure 4 reports the results of our estimations using different distance thresholds (10, 25, 30, 35, 40, 50, and 75 km) of the World Bank aid sample (Table A.3 in Online Appendix provides detailed regressions results). There is considerable variation across different sub-samples, our results are most consistent when taking 35 km and 40 km as cut-offs across all sub-samples. For all the estimations, the results disappear when narrowing to 10 km (except for the sample of the gender-sensitive sector and female) and moving away above 75 km.

Fig. 4

Estimated effects using different cut-offs of distance. Estimated effects with 95% confidence intervals. Difference is the coefficients of treated groups (active) subtract control groups (inactive)

In contrast, the corresponding Chinese aid samples do not achieve any similar positive results to the World Bank when using the above cut-offs of distance (see Table A.4 of Online Appendix).

As discussed in the theoretical framework section, when aid projects are integrated with donors’ social norms about gender equality, they can act as an instrument of sowing such norms in recipient countries. Aid intervention on gender equality works through norm diffusion. Based on the extent to which gender norms are integrated into project implementation, we code three sectors of the World Bank aid as gender-sensitive sectors—namely, education, government and civil society, and health. As abovementioned, China does not incorporate gender norms in its projects; thus, there is no gender-sensitive aid projects in Chinese aid sample.

The results reveal that only two gender-sensitive sectors, namely, government and civil society and education are significantly and positively associated with the local support of gender equality while the other five sectors are not (the sector of health is close to achieving significance at 10 percentage level, Table 5). The results are consistent with the aforementioned theoretical discussions: aid projects integrated with gender issues and norms in the design, implementation, and evaluation are more likely to impact the local attitudes toward gender equality. Therefore, these results support our argument of the difference of aid sectors on the support of local gender equality. Again, no similar results are found for aid projects from China. Chinese aid does not contribute to local attitudes change toward gender equality in African countries. And on some occasions, it is even negatively associated with the local support of gender equality (see Table A.4 of Online Appendix).

Table 5 World Bank aid and support of gender equality: by sectors (full sample)

To further explore the relationship between aid sectors and local support of gender equality, we separate the samples of women and men for the two gender-sensitive sectors that achieved significance. Again, the results indicate that women are more likely to support gender equality when exposed to active aid projects from the World Bank in both education and government and civil society sectors (Table 6) compared with those close to inactive projects. This consistent pattern of results suggests an interesting phenomenon—the change in local attitudes toward gender equality is almost restricted to the sample of women and has nothing to do with men.

Table 6 World Bank aid and support of gender equality by gender and sectors

On average, women who lived near a World Bank education program are 14.8 percentage points more likely to believe women and men have equal rights and should be treated equally than women close to an inactive education project. And the government and civil society projects from the World Bank also work for increasing local women’s support of gender equality: the effects are as large as 7.1 percentage points in our estimation. In contrast, education projects from the World Bank did not change local men’s attitude toward gender equality: the difference of coefficients of the active and inactive cannot distinguish from 0. Similarly, the results indicate the government and civil society projects from the World Bank did not improve local men’s persistent gender norm—“women have always been subject to traditional laws and customs and should remain so.”

Robustness Checks and Additional Qualitative Evidence

Robustness Checks

The benchmark results of the World Bank aid (the four samples of full, female, gender-sensitive, and gender-sensitive and female observations) are robust over a wide range of alternative specifications and checks (Fig. 5, for regression tables, see Table A.5 of Online Appendix).¹⁶ The results are similar on completed projects within the 35-km cut-off distance and using logistic regressions instead of OLS (see Panel A and Panel B, Table A.5 of Online Appendix).

Fig. 5

Robustness checks. Estimated effects with 95% confidence intervals. Difference is the coefficients of treated groups (active35) subtract control groups (inactive35)

In our benchmark estimations, we include both observations of Afrobarometer Round 3 and Round 5 to construct a complete sample.¹⁷ However, the observations of active35 and inactive35 in Round 5 are relatively unbalanced: we get 17,932 observations in the treatment (active35 = 1) while only 503 observations in the control (inactive35 = 1). This is because the geocoded World Bank aid data only covers 2014 and the surveys of Round 5 were conducted from 2011 to 2013. Thus, at the time of the Afrobarometer Round 5, the interviewed respondents are less likely to match with inactive aid projects. To address this concern, we removed all observations of Round 5 and the benchmark results remain. The results are more significant and substantive: the differences of effects are 5.4%, 7.4%, 11.9%, and 17.6%, respectively, in relation to the overall sample mean of the outcome (see Panel C of Table A.5, Online Appendix).

Our estimation strategy includes a reference group—those who were not living near active projects or inactive projects within a certain cut-off distance. This may raise a concern that too many irrelevant observations are included in the estimations. For example, a respondent who lived close to aid sites as far as 1,000 km is less pertinent for our research of interest. We try to set 100 km as a threshold of distance and exclude respondents that lived farther than this cut-off to make the active and inactive groups more comparable with the reference group. The results are also robust (also see panel D of Table A.5, Online Appendix).

According to our data, there are more observations of respondents who lived close to active aid projects (23,459, accounting for 58.64%) than lived near inactive aid projects (3,652, accounting for 9.13%) in our data. This is because of an unsymmetric distribution of aid data and the respondents.¹⁸ AidData covers 5 years (2000–2004) before the Round 3 survey (conducted in 2005 and 2006) but 8 years (2007–2014) after the Round 3 survey and 11 years (2000–2010) before the Round 5 survey (conducted from 2011 to 2013) while only 1 year of aid data available after the Round 5 survey. To address this concern, we only include aid projects (the World Bank) that started within 5 years before the surveys and within five years after the surveys (Round 3 and Round 5, respectively). Our results withstand (also see panel F of Table A.5, Online Appendix).

We also restrict the sample to include only the respondents in Afrobarometer enumeration areas which are geocoded with precision code 1 (exact location) and the results are consistent with the benchmark estimations (see panel E of Table A.5, Online Appendix).

Our benchmark estimations of Chinese aid include all Chinese development projects in AidData. But China tends to mix its development aid with commercial interests (Isaksson and Kotsadam 2020) in contrast to the Western donors. We find including only the “ODA-like” aid by AidData does not change the finding—that China aid has no significant impact on local support of gender equality (also see panel G of Table A.5, Online Appendix).

Additional Discussion and Qualitative Evidence

Our empirical findings confirm the hypothesis that integrating gender equality norm into aid projects matters for improving local support for gender equality. And aid sectors in which gender equality norm are more likely to be incorporated have better performance in diffusing such norms. And we find local women are more likely to accept gender equality norm than men. However, further explanation on why the two gender-sensitive sectors (education and government and civil society) work better to diffuse gender equality norm is needed and how norm diffusion is happening on the ground is a relevant question to ask. In order to unpack the “black box,” we further investigate the gender mainstreaming process of World Bank projects.

One powerful tool of gender mainstreaming for gender norm diffusion in the Bank is gender analysis. The gender analysis enabled the Bank to integrate gender issues throughout its discussion of development actions and to use gender-related conditions as a basis for determining aid priorities (Correia and Bronkhorst 2000). Gender analysis is conducted both at the country level and project level.

The first is the Country Gender Assessment (CGA) by the Bank’s Operations Evaluation Department (which is the responsibility of the County Director). CGA aims to identify gender-related policy and programmatic interventions that are likely to have high payoffs for development (World Bank 2002: 19), for example, the difference in socioeconomic roles of men and women in the household and the market, gender disparities in access to land, inequalities between men and women in participation of decision-makings and politics.

The conclusions and recommendations of the CGA serve as a key reference for the Bank’s decision-making in its country assistance program. CGA, as a prior assessment of gender conditions in a country will also guide the ensuing gender-responsive actions and the development effectiveness of the Bank’s project (World Bank 2002). The CGA is incorporated into the review of the Bank’s documents of ongoing projects in the country. For example, a CGA of an education program identified gender stereotypes in textbooks and then designed a program to eliminate these stereotypes and train local teachers (World Bank 2002: 24). And the World Bank regularly releases good practice examples of CGA to inform and spread the tool.

Meanwhile, gender is integrated into the sectoral and project analytical work on the ground including program operations and institutional environment support. And gender analysis is also conducted during all stages of implementing aid projects (Fig. 6). In the initial phases, gender analysis applies to understand the potential impact of planned development interventions on gender equality in a specific context. During the project intervention, the impacts of the project’s intervention is regularly assessed (including monitoring the processes). And when the projects are completed, an evaluation will be given to identify changes and results of the gender intervention.

Fig. 6

Gender analysis at the project level

The sectoral differences of aid projects on gender norm diffusion among the local population are considerable, but why? Detailed information on project assessment from the Bank sheds light on the important questions. According to our empirical results, education and government and civil society are two important sectors for the diffusion of gender equality norm (Table 5 and Table 6). “Targeting education is a vital part of a gender mainstreaming strategy in every developing society, on both national and local levels” (UNDP 2014:105). Although education does not guarantee rights, it is critical for empowering women. When women received more education, they are more able to participate in community and national decisions, and better able to protect themselves from gender-based violence (World Bank 2001). Our results also indicate education is positively and significantly associated with higher support for gender equality.

Educating women help diffuse gender norms. For instance, a gender module was introduced into the training curricula for education sector management and staff in a World Bank education project in Zambia around 1999. The boy students were asked to clean classrooms (previously by girl students). The District Education Officers formulated a policy of gender balance, requiring the appointment of a female deputy head if the school headmaster is a male. And gender equality training programs were conducted at all levels of administration (World Bank 2010b).

Among the Bank’s aid to Africa in our study, more than half of them (53.47%) falls in the sector of government and civil society (Table 2). And the Bank provided overwhelmingly more government and civil society projects to Africa from 2000 to 2014 than China. Moreover, the World Bank makes grants directly to civil society organizations (CSOs), and its aid projects support CSOs’ work in promoting gender inequality, education, environmental conservation, youth development, post-conflict reconstruction, and other social issues (World Bank 2006). In contrast, the size of the Chinese government and civil society aid sector to Africa during this period is only 6% of the World Bank (Table 2).

The World Bank encourages and funds civil society groups to work on gender issues. The Bank regularly collaborates with United Nations agencies on gender issues and participates in the gender-focused committees and working parties of the UN system and other international organizations. These groups include the UN Commission on the Status of Women, the Working Party on Gender Equality of the Development Action Committee of the Organization for Economic Cooperation and Development, the Multilateral Development Bank Working Group on Gender, and the former UN Inter-Agency Meeting on Women and Gender Equality (World Bank 2002: 65). For example, in 1999 and 2000, the Bank sponsored regional consultations with civil society in Asia, Africa, and Latin America to solicit views on actions the Bank should take to promote gender equality (World Bank 2002: 65).

As the practice of integrating gender norms on the ground, we introduce two aid projects from the World Bank to further illustrate gender integration at the project level (for detailed information of the two projects, see Online Appendix).

South Sudan Gender Support and Development Project

The project started in 2009 and was completed in 2012. The project was implemented by the Ministry of Gender under the guidance of the World Bank experts. The project aimed to achieve peace dividends for targeted women in Southern Sudan, through the provision of (a) improvements in access to the existing economic opportunities and (b) support to the Ministry of Gender, Social Work, and Religious Affairs to develop and implement gender policies and strategies. According to the assessment of the World Bank, the efficacy rating of the project is substantial: 40% of the target group (2,854 women) perceived an increase in income because of project interventions (the initial goal was set at 20%). The country gender assessment and relevant gender training were completed and the capacity strengthening of the Ministry staff was successful.

Accelerating Rural Women’s Access to Market and Trade in Kenya

The project was implemented in Molo County and Kitui County by a local civil society group—GROOTS Kenya with an aim to promote women’s participation in development and gender equality. An assessment in 2014 rated the overall implementation of the project as moderately satisfactory. The total aid budget was USD 3,000,000. Through enhanced agricultural production and access to markets, the project contributed to women's economic empowerment. “The underlying rationale is to equip women with relevant knowledge and skills as well as enhance their leadership and organizational capabilities to enable them to become drivers of their own social and economic transformation.” This project’s beneficiaries were 3,400 rural women. The project has five components including “enhance women farmers’ lobbying and advocacy capacities for essential services to improve production.”

Conclusion

Overall, our results suggest that although systematic changes in gender equality remain difficult to achieve, foreign aid helps diffuse gender norms to local population. Earlier studies of gender and development pay less attention to the gender norms change and rarely link it with foreign aid, let alone compare different donors at the project level. Our findings also suggest aid projects from different donors, and in different sectors may have different impacts and females are more likely to be impacted by the aid projects with gender norm diffusion. That is to say, disparities among donors, sectors, and gender are important factors for consideration when evaluating the effects of aid intervention on local gender norms change.

When aid projects are integrated with gender norms, they can spread such norms in the recipient countries. The World Bank and China represent two distinct donors as the former overwhelmingly adopted gender mainstreaming in aid while the latter did not during the period of data collection. By matching geocoded data from AidData and Afrobarometer survey, our empirical tests prove that the World Bank aid increases local support for gender equality in Africa and the results are mainly driven by the samples of local women and gender-sensitive projects. However, Chinese aid has no impact on local attitudes toward gender equality.

For a long period, the Chinese approach to gender equality is endogenously guided by the modernization theory that gender equality must be based on the premise of improving women’s economic status. Mao’s famous words—“Women hold half the sky” had driven women into the job market. However, China does not export gender norms through its foreign aid but emphasizes its economic development impacts. This experience is reflected in China’s aid to African countries.

The World Bank recently concluded in a report that “gender mainstreaming strategy was undoubtedly successful” (World Bank 2016:2). However, according to our findings, the strategy was only successful to increase local women’s acceptance of gender equality but did not change men’s beliefs on gender role: women should subject to traditional customs and remain so. If men’s dominating role remains unchanged in many developing societies, awakening women’s gender awareness can create problems. And the persistence of the gender gap in attitude toward gender equality would lead to unintended social consequences (e.g., domestic violence or social conflicts). Also, the country context, duration of the project, and different intervention strategies can shape the effect of aid projects on people’s attitude to gender equality (Beath et al. 2013).

Emerging donors need to learn from traditional donors for integrating gender norms in their foreign aid projects and all donors and recipients need to work together to explore an effective approach to substantive gender equality through aid provision. Cooperation and coordination between the OECD donors, international organizations, emerging donors, and the Global South is needed to improve the landscape of international development and foreign aid.¹⁹ By offering a preliminary theoretical framework and systematic evidence on African countries, we hope this paper can contribute to this emerging area of research and exert implications for policymakers and development practitioners adventuring in the land of aid.