Do Women Benefit from Women Education Aid? Evidence from Panel Data

Generally, the literature on aid focuses on the potential growth effects of aggregate aid. Due to the fact that donors have consistently asserted the multidimensionality of their purposes, it is necessary to conduct a much more disaggregated analysis of aid effectiveness. In this study, the effect of women education aid on 72 developing countries is examined empirically over the period 1990–2016. Using cross-country regression, this study examines the effectiveness of aid targeted at women’s education. Based on the fact that donors provide a large amount of women’s education aid to countries whose voting positions in the UN General Assembly are similar, this analysis exploits an instrumental variable. This study shows that women’s education aid has a significantly positive effect on women’s education. The results of this study are robust when different sensitivity checks are performed. The findings have significant policy implications for donor countries and international aid organizations, as they assist in identifying the most effective types of foreign aid flow to the various sectors of the recipient country’s economy.


Introduction
Currently, there are more than 263 million children out of school, 202 million of whom are in secondary school, and 130 million are girls.Despite all the efforts and progress made in previous years, more girls still need to receive an education.It is estimated that 15 million girls of primary school age never set foot in a classroom.A growing body of literature shows that investing in girls' education brings high returns in reducing poverty and stimulating economic growth.Educating women also promotes children's and women's survival rates and health, delays child marriage and early pregnancies, empowers women in the home and the workplace, and helps tackle climate change (Sharma & Kiran, 2012).In proposed target 4.1 of the United Nations' Sustainable Development Goals, the UN said: 'By 2030, ensure that all girls and boys complete free, equitable and quality primary and secondary education leading to relevant and effective learning outcomes', meaning that each of the 263 million children currently out of education will be entitled to twelve years of quality, feefree primary, lower secondary and upper secondary education by 2030.For this purpose, donors have delivered increasing aid to educate women.This is mainly through providing accessible, equitable, and quality education opportunities in developing countries.For example, women's education assistance increased significantly from US$ 0.66 million in 1990 to US$ 100.4 million in 2016. 1 Providing education, training, and professional development to women is the goal of women's education aid.Despite the clarity of this aim, the evidence on the effectiveness of aid targeted at women's education in educating women is still scant.
There is considerable debate about whether foreign aid effectively promotes growth and development in recipient countries.Even recent literature surveys have reached sharply opposing conclusions.There are a number of studies, such as (Basnet, 2013;Collier & Dollar, 2002;Ekanayake & Dasha, 2010;Headey, 2008), which demonstrate that overall aid has a significant positive impact on the economic growth of recipient countries.Additionally, Burnside and Dollar (2000) find that aggregated aid has a significant and positive impact on economic growth only in countries with sound fiscal, monetary, and trade policies.However, it negatively impacts countries with poor policies.Furthermore, Hansen and Tarp (2001) use a variety of econometric specifications to demonstrate that overall aid is effective and that policy does not have an effect on results.
On the other hand, Djankov et al. (2008), some studies have shown that overall aid causes social, political, and economic instabilities in the countries that receive it.For example, it increases corruption, civil conflicts, and dependency syndrome, and reduces domestic production.Further, Easterly (2003) argues that aid can sometimes prove useful; however, it generally results in the deterioration of business and financial conditions.Moyo (2009) argues that development aid is not the answer to developing countries' economic, social, and political problems; rather, it is a burden on the recipient countries.
Both strands of literature ignore the fact that different types of aid may not have the same economic effects on recipients.It is common to aggregate different types of assistance into a single figure in much of the literature.This is done in order to examine its impact on the economic growth of the recipient countries.The effectiveness of foreign aid appears to have undergone a paradigm shift recently, however.There is an explicit need for looking at the effectiveness of aid in a much more disaggregated manner, as is reflected in much of the aid literature that attempts to answer the question 'Which type of aid is more helpful?' instead of 'Is aggregated aid beneficial?'For example, work by Clemens et al. (2004) on short-impact aid has started a shift toward using disaggregated aid data.Similarly, Mishra and Newhouse (2009) examine the impact of health-targeted aid on health outcomes.Dreher et al. (2008) investigate the impact of education aid on education outcomes.Brenton and von Uexkull (2015) examine the effect of technical assistance projects from German development cooperation (GTZ) on recipient countries' trade performance.
Aid effectiveness will continue to be a hot topic if attention is limited to examining the overall impact of aid on the economies of recipient countries.Donors expect various outcomes when giving foreign aid to developing countries (see, for example, Isenman & Ehrenpreis, 2003).A study of how aid capital is directed toward a specific purpose, such as women's education, may deviate significantly from an analysis of the aggregate-aid-growth relationship.Many researchers (both opponents and proponents) turn to literature on the aid-growth nexus for empirical purposes.There is, however, a daunting challenge in finding a significant impact of aid in the literature.Due to both the causal relationship between foreign aid and economic growth in the recipient countries, this situation has arisen.It is also due to the complexity of foreign aid's impact on economic growth.As a general rule, economic growth and aid capital are too distantly related (with numerous channels in between) to be able to detect any significant relationship between them.Consequently, applying different approaches to examining aid effectiveness may be helpful in resolving this issue.As aid targeted at women's education is closer to women's education, a statistical relationship between the two may be easier to detect.The purpose of this study is to assess the impact of aid on more specific outcomes rather than general outcomes.The study examines the relationship between aid provided to educate women and the level of education achieved by women using much-disaggregated aid data.
Policymakers and international aid organizations consider women's education to be a critical component of economic growth.Investing in women's education contributes to gender equality, poverty eradication, and inclusive economic growth.In business, on farms, as entrepreneurs or employees, or by providing unpaid care at home, women make an enormous contribution to the economy.Women's education is essential for sustainable development, according to a number of studies.In addition, it is affirmed that gender disparity is a prevalent issue across cultures and that sustainable development cannot be achieved without the education of women to combat it (Stevens, 2010).According to a report issued by UN Women (2016), education plays a significant role in achieving sustainable economic development and increasing women's contributions to their families and communities.
The presence of many uneducated women in a country, however, will have a negative impact on economic development.A study by Stevens (2010) contends that gender inequalities significantly contribute to economic costs, social inequalities, and environmental degradation.After 1985, donors allocated an increasing share of aid capital to women's education to take advantage of these and other multidimensional benefits of educating women (see Figure 1).An increasing number of empirical studies (e.g.Burnside & Dollar, 2000;Rajan & Subramanian, 2008) have examined the effect of overall aid on economic growth.Nevertheless, there does not appear to be any systematic empirical evidence demonstrating how aid targeted at women's education impacts women's education in recipient countries (to the best of my knowledge).Given that donors provide increased support for women's education, this condition is surprising.Hence, this study examines the effect of committed aid specifically related to the education of women in developing countries, based on data from 72 countries that received aid from 1990 to 2016.This study provides the first systematic analysis of the relationship between aid allocated to women's education and women's education in developing countries. 2 I use the female gross school enrollment ratios at the primary, secondary, and tertiary levels of education.There are several reasons for using these variables as indicators of women's education, including the availability of data for many countries over a long period of time.In addition, these measures are considered by the extant literature to be the most important indicators of women's education (Barro & Lee, 2013).Additionally, I examine the role of aid directed at women's education in educating women in developing countries using three strategies of identification.Among these are ordinary least squares (OLS), two-stage least squares (2SLS), and Blundell and Bond's (1998) Generalized Method of Moments (GMM).
Women's education aid has a significant positive effect on women's education in recipient countries, according to this study.A per capita increase of US$1 in women's education aid increases female enrollment in primary, secondary, and tertiary schools by an average of 0.466, 0.004, and 0.203 percent, respectively.However, the impact of women's educational aid on primary school enrollments is greater and more positive than that on secondary and tertiary enrollments.The main reason for this result could be directly related to evidence that quality primary education prepares children for daily life challenges and allows them to benefit from economic and lifelong learning opportunities.Moreover, it contributes to the reduction of poverty, the promotion of economic growth, the achievement of gender equality, and the improvement of social welfare.When education support is targeted at girls, these benefits are even more pronounced.In general, girls who complete their primary education tend to find better jobs, marry later, and have fewer children.Furthermore, they will be less likely to have children suffering from malnutrition and less likely to contract sexually transmitted diseases (STDs) such as HIV/AIDS (see, for example, Birchler & Michaelowa, 2016).
There are substantial policy implications for donor countries and international aid organizations as a result of these findings.It is more desirable in the future to shift more women's education aid to the primary education sector.This argument is in line with United Nations Sustainable Development Goals 4 and 5. Although, one might also need to consider the long-term dynamics here in that providing attractive opportunities for study 2 This analysis considers only bilaterally committed aid provided for educating women rather than the disbursed amount.
Theoretically, examining the effect of the disbursed aid on recipients' outcome variables might give compelling findings since the recipient countries have already received the aid capital.However, the aid literature shows some limitations in the use of disbursed aid.First, in many cases, the data for disbursed aid is missing as it is 'spotty' in most of the aid data sources.Second, aid disbursement is unpredictable compared to commitments because the amount of aid could be disbursed mainly in periods when output or domestic revenue are high and held back when domestic economic activity is shrinking (see Bulíř & Hamann, 2008).Hence, the analysis incorporates recipient and time-fixed effects in all models to consider any bias from systematic divergences between commitments and disbursements.
at a consistently higher level will result in a dead-end if graduates are unable to secure employment.Developing the labor market is an essential complement to any support for the education sector over the long term.
The findings are robust to a battery of robustness checks.These include alternative indicators of women's education (females' primary school completion rate, adults' female literacy rate as a percentage of females over the age of 15, and females' effective progression to secondary school).This study also uses alternative measures of women's education aid, an alternative estimation method, and includes a set of control variables.
The remainder of the paper is organized as follows.Section 2 presents the data description.Section 3 discusses empirical strategies.Section 4 presents empirical findings.Section 5 shows robustness checks of the benchmark findings.Section 6 concludes.

Indicators of Women Education
I use female gross school enrollment ratios in primary, secondary, and tertiary levels of education.The gross enrollment ratio is the ratio of total enrollment, regardless of age, to the population of the age group that officially corresponds to the level of education.The data for these indicators are obtained from World Development Indicators (WDI) over the period 1990-2016.These indicators are the most critical educational outcome variables, as Barro (1991) argued.

Women Education Aid
I obtained data for women's education aid from the current (static) research release of the AidData database (version 2.1; see Tierney et al., 2011Tierney et al., ) covering 1990Tierney et al., -2016.This database combines the broadly used data from bilateral donors released by the OECD's Creditor Reporting Service (CRS).It also provides data from many non-OECD bilateral donors and multilateral financial institutions, including regional development banks (most of which are not in the CRS) and the World Bank.In addition, it includes purposespecific funds from the Bill and Melinda Gates Foundation (BMGF) and Global Alliance for AIDS and Vaccinations (GAVI).It also gives a more comprehensive view of aid across all activities for many years relative to the standard CRS.It provides more detailed information about aid to specific purposes and projects.It records bilateral and multilateral aid for each purpose in a consistent format.This recording system helps examine the role of aid targeted to a specific purpose in achieving its intended objectives.In this analysis, aid capital refers to the commitments of concessional loans and grants from all donors, including multilateral organizations, provided to educate women in developing countries.Following Wilson (2011), this study made no distinction between loans and grants.Since most loans are long-term with low-interest rates, I assume that loans will have equivalent effects to grants in the medium to long term.I use data on aid commitments because purpose-related information on aid has only been available for commitments, not actual disbursements (Jones & Tarp, 2016).As discussed earlier, this paper uses only bilaterally committed aid targeted at women's education data rather than the disbursed amount. 3sing the AidData codes, I obtained bilaterally committed women's education aid.Aid-Data database assigns codes for each type of purpose-specific aid with a new coding scheme that builds on the system of purpose codes developed by the CRS.However, the AidData database allows more aid disaggregation.It resolves the frequent problem in the CRS data where projects are aggregated into catch-all codes if they have multiple activities within the same purpose.By using project descriptions, AidData researchers assigned purpose codes to each project.They also re-coded projects from different sectors in the CRS and assigned them AidData codes.Hence, all the projects in the AidData database are coded by the same research team using similar and consistent criteria as opposed to the CRS, in which codes are assigned by individual donors scattered across the world using changing criteria (see Wilson, 2011).Women's education aid is coded as 42010 in the AidData database.Among others, donors deliver this aid to give women education, training, and employment opportunities.They also allocate women's education aid to promote gender equality and comprehensive support for women, raise awareness of gender-based violence and support the implementation and execution of the comprehensive law on women-specific violence.I scaled women's education aid, measured at constant $USD 2010, in per capita terms, in which it is divided by the female population of the recipient countries.4

Other Control Variables
This analysis considers parsimonious specifications.However, it includes other potential determinants that significantly affect women's education.The control variables are kept reasonably small to retain some interpretability of the correlations.The other determinants of women's education considered in this analysis are real income per capita growth, converted to US dollars at constant 2010 from the WDI.Growth of per capita income is included as the existing literature states that higher growth plays a significant role in promoting the level of women's education (see Duflo, 2012).I also include International Country Risk Guide (ICRG) index as a measure of institutional quality from the Political Risk Services (PRS) group.This index contains more than 22 variables in three subcategories of risk: political, financial, and economic risk ratings.It is calculated as a weighted average of a country's political, financial, and economic risk.In the ICRG index, the highest overall rating (theoretically 100) shows the lowest risk, and the lowest rating (theoretically zero) shows the highest risk (see Knack & Keefer, 1995).Therefore, the ICRG index is expected to affect women's education positively.
Further, instead of the ICRG index, I include gender equality ratings and indicators of legislation on domestic violence and sexual harassment.However, the data for these variables are available only from 2000 onwards.Therefore, this analysis mainly considers the ICRG index to capture the effect of institutional quality on women's education in recipient countries.
Many health indicators have a significant effect on female participation in education.Thus, the under-5 female mortality rate is included to control for the impact of health factors on women's education.Other health factors, including infant and neonatal female mortality rates and the share of women population ages 15 + living with HIV, are considered interchangeably.This analysis also controls the percentage of female children ages 7-14 in employment and government expenditure on education as a percentage of GDP, as these variables significantly affect women's education (see Martin, 1995).The data for these variables are obtained from WDI.I also control for total education aid obtained from the Aiddata database, which helps to identify the contribution of women's education aid in educating women.Finally, the three additional instrumental variables that I use, affinity indices with USA, UK, and Canada, are drawn from Erik Gartzke's website.5

Summary Statistics
Table 1 presents the summary of the datasets.This table shows the means, standard deviations (overall, between, and within countries), and minimum and maximum values of dependent and independent variables.Table 2 also presents the correlation coefficients between all critical variables, which help design the model and confirm the choice of the instruments.All the variables, including the ICRG index, display considerable variation between and within countries, suggesting that using panel estimation techniques, which allows the identification of the various parameters of interest, is reasonable.Moreover, the correlation coefficients are within suitable ranges and confirm the choice of explanatory and instrumental variables.The instrumental variables used in this analysis (USA affinity, UK affinity, and Canada affinity) have excellent features in that their correlation coefficients with women's education aid are 0.64, 0.71, and 0.62, respectively.In contrast, their correlation coefficients with the indicators of women's education are very small, less than 0.25.

Ordinary Least Squares
I follow McCrary and Royer (2011) to specify the empirical specifications.The pooled OLS regression equation defines female gross school enrollment ratios in primary, secondary, and tertiary levels of education as a function of women's education aid in earlier periods.6 Where i denotes the country (i = 1, . . ., 72) and t denotes the time (t = 1990, . . . , 2016).GE it represents female gross school enrollment ratios in primary, secondary and tertiary levels of education.GE it−1 indicates female gross school enrollment ratios in primary, secondary and tertiary levels of education in the previous year.WA it−1 shows one-year  Notes: the dependent variables are gross primary, secondary and tertiary enrolment ratios.t-statistics in parentheses are based on robust and clustered standard errors.A constant, time dummies and country fixed effects are included but coefficients are not reported (results are available upon request).In all specifications, except in specifications 1, 5 and 9, women education aid is instrumented variable.To control the endogeneity issue, I use the lagged values of growth of per capita income, ICRG index, child mortality rate, female children employment ages 7-14, government expenditure in education and total education aid in 2SLS and system-GMM.* * * , * * , and * denote the coefficients are significant at the 1%, 5%, and 10% level of significance respectively.
lagged per capita women education aid.The main reason for using the lagged values of the aid variable is associated with the timing issues that the realization of the effect of aid on women's education may take certain periods.7 X it−1 shows the set of other control variables, which includes the growth of per capita income, the measure of institutional quality (ICRG index), under-5 mortality, percentage of female children ages 7-14 in children employment, total education aid (excluding women's education aid), and government expenditure on education as a percentage of GDP.8 α i , λ t and ε it denote country fixed effect, time fixed effect, and the idiosyncratic error term respectively.Including total education aid (excluding women's education aid) allows us to capture the potential impact of total education aid on women empowerment through other channels in addition to women's education aid.Specifically, this analysis examines the size, direction, and significance of the coefficients of women education aid (γ 2 ) The presence of a positive and significant value indicates that women's education aid plays an important role in educating women in developing countries.On the other hand, a significant negative value indicates that women's education aid dissipates women's education, which supports the findings of Djankov et al. (2008). 9The inclusion of a lagged dependent variable in Equation ( 1) indicates that all estimated gamma coefficients show short-run effects.To measure the medium and long-term role of women's education aid in educating women, Equation ( 1) is re-estimated using the average values of all variables over five-year periods.10Additionally, following Hansen and Tarp (2001), I include the squared term of women's education aid to examine the potential non-linear relationship between women's education aid and women's education. 11The standard errors are clustered at country level.The residuals are correlated within a country but not across countries.

Two Stages Least Square Estimation
Due to problems associated with simultaneity and reverse causality between women's education and women's education aid, results from OLS estimations are likely biased.A well-documented finding in the literature is that the amount of aid may not be determined exogenously by the socioeconomic activities of the recipient country.However, the characteristics of various indicators of women's education in developing countries can be used to determine the level of women's education aid.As an example, the gross enrollment ratio of females in primary schools is low in recipient countries.To increase the enrollment of females in primary schools, donors may provide a large amount of education aid.As a result, the coefficient of women's education aid would be biased downward (i.e. the effect of aid capital would be underestimated).
Similar to this, donors may provide large amount of women's education aid to countries with high female gross school enrollment rates in primary schools.Women's education aid may be biased upward by this condition (i.e. the effect of the aid capital may be overestimated).Further, any measurement error, such as a misreporting of committed and/or disbursed women's education aid to the AidData database, can cause an endogeneity problem in the Equation (1). 12Therefore, to resolve the endogeneity issue, this analysis uses instruments for women's education aid and examines the role of women's education aid in educating women in recipient countries.
Women's education aid instruments considered in Equation ( 1).Based on the argument that political factors (or political affinities) determine how bilateral donors allocate aid among the recipient countries, which becomes an important tool for donors' foreign policy (Alesina & Dollar, 2000).I measure political proximity between donor and recipient countries by the affinity of nations index proposed by Gartzke (2009).This index reflects the similarity of state preferences based on the voting positions of pairs of countries in the United Nations General Assembly.Political proximity is a plausible exogenous driver of a donor country providing aid.This is unlikely to be directly associated with women's education in the recipient countries except through a significant amount of women's education aid.I use the affinity index between each recipient and the three largest bilateral donors, including the United States, Canada, and the United Kingdom. 13,14The central hypothesis is that, all else being equal, donors provide a large amount of women's education aid to countries with robust political affiliations (as reflected in a high affinity index).Hence, it is rational to consider that the affinity index of a nation is a plausible instrumental variable for women's education aid.The affinity index is calculated as follows: (2) where i and j denote the countries dyadic, and t denotes the time (t = 1990, . . . , 2016).d(V i , V j ) is the sum of metric distances between votes by dyad members in terms of United Nations General Assembly (UNGA) votes (1 = approval; 2 = abstain; 3 = disapproval) in each year.d MAX is the largest possible metric distance for those votes for a given year.For example, if there have been 100 determinations in a year, d MAX = 200.Thus, the index ranges from −1 (minimum affinity) to +1 (maximum affinity).In each year t, I calculate three variables for each recipient country i by using the three-major bilateral donors of women education aid as counterpart countries j, which includes, United States, Canada and Unites Kingdom.If, for example, USA (donor) and Senegal (recipient) have the total vote distance of 200 in three UNGA determinations, their affinity index will be calculated as 1-((2 * 200)/200) = −1, showing that USA and Senegal have dissimilar preferences, which may result in a small amount of women's education aid delivered from the USA to Senegal, 12 The presence of aid elements in the error term of Equation ( 1) may violate one of the Gauss-Markov assumptions, such as the expected value of women's education aid and the error term may not be zero, and thereby create an endogeneity issue (see Cragg & Donald, 1993). 13I also ran regressions using the affinity index of other bilateral donors, including Australia, France, Germany, Italy, Japan, Korea, the Netherlands, Norway, Spain, and Sweden.These are instrumental variables of the aid variable.The results are qualitatively similar to the results reported in 2SLS estimations in Table 2 (results available upon request).These donors have missing affinity index data when compared to USA, Canada, and UK, which may lead to ambiguous conclusions.Therefore, my analysis mainly focuses on the findings obtained from using the affinity indices of the USA, Canada, and the UK as instruments of the aid variable. 14Appendix 1 shows the amount of women's education aid provided by the largest bilateral donors.
and vice versa if these countries have a higher affinity index.Therefore, I investigate the women education aid-women's education nexus by estimating the following equations: where affinity ijt shows the affinity index between a recipient country i and a donor country j at time t.It includes the affinity indices between the three largest bilateral donors (USA, Canada and UK) and the recipient countries considered in this analysis.The other variables have similar definitions as in Equation ( 1).Technically, replacing the fitted values of the aid equation i.e.Equation ( 4) in place of the aid variable in Equation ( 3) would potentially address the endogeneity issue (see also Muller et al., 2014).

Dynamic Panel GMM Estimation
The aid capital is an endogenous variable, and some other control variables (e.g.growth per capita).Hence, a plausible alternative to address this issue is using the Generalized Method of Moments (GMMs) proposed by Blundell and Bond (1998).This method allows simultaneously estimating a regression equation in differences and a regression equation in levels, each employing its own set of instrumental variables.It uses lagged levels as instruments in the differenced equation and lagged differences as instruments in the level equations.

Benchmark Findings
I begin the analysis by using annual data to estimate Equation (1) with OLS, 2SLS, and system-GMM under the control variables discussed above.Table 2 shows the findings First, columns 1-4 demonstrate the effect of women's education aid on female's gross primary school enrolment ratio Next, columns 5-8 present the impact of women's education aid on female's gross school enrolment ratios in secondary education.Finally, columns 9-12 illustrate the effect of women's education aid on female gross school enrolment ratios in tertiary education.Most of the specifications in Table 2 show that the growth of per capita income, the measure of institutional quality (ICRG index), government expenditure on education, and total education aid have salient contributions to educating women in developing countries.On the other hand, the child mortality rate and percentage of female children ages 7-14 in employment negatively affect women's education.
Turning to the variable of interest, Table 2 shows that women's education aid significantly positively affects women's education.However, women's education aid has a higher positive impact on the female gross school enrolment ratio in primary education.This is compared to female gross school enrolment ratios at secondary and tertiary educational levels.For example, 2SLS estimations show that a $1 increase in per capita women's education aid contributes to an average enhancement of 0.446% in female's gross primary school enrolment ratio.This is higher than 0.04% and 0.203% for secondary and tertiary education, respectively.This relationship is maintained when the analysis considers system-GMM.It is pertinent to note that the effect of women's education aid on women's education increases as the potential endogeneity issue is addressed.The qualitative nature (i.e. the sign and statistical significance) of women's education aid coefficients remains intact when the estimation method shifts from OLS to 2SLS and from OLS to system-GMM.It appears that Hansen's (1982) J-statistic fails to reject the exogeneity of the potentially endogenous variables at the 5% level, so the instruments for the potentially endogenous variables are valid.
Moreover, the Arellano and Bond (1991) test rejects the null hypothesis of no firstorder autocorrelation in the error term.However, it fails to reject the null hypothesis of no second-order serial correlation in all system-GMM regressions at the 5% level.Columns 4,8,and 12 show that the effects of women's education aid on women's education remain positive and significant.However, the lag level of endogenous variables is unrestricted. 15oreover, I also use five-year period average data to estimate the coefficients of the effects of women's education aid on women's education, using OLS, 2SLS, and system-GMM estimations.Table 3 reports the results.In line with Table 2, Table 3 indicates that women's education aid has a significant positive impact on the education of women in the recipient countries.However, the size and significance levels of the aid term coefficients change as I compare annual and average data.For example, 2SLS estimates show that the aid term coefficient decreases from 0.466 to 0.284 when I use average data instead of annual data (see Columns 2 in Tables 2 and 3).
It is worth noting that the benchmark findings reported in Tables 2 and 3 show that the coefficient of women's education aid is relatively stable across various specification strategies.Its value reveals a significant quantitative effect of women's education aid.This result may corroborate the current change in assumptions about aid effectiveness from aggregating approaches-where consensus is scant-to disaggregating approaches-where many studies provide amenable findings (see Dreher et al., 2008;Mishra & Newhouse, 2009).The following sections show the robustness check of these findings.

Sensitivity Analysis of the Benchmark Results
In this section, the robustness of the baseline findings was checked by applying alternative indicators of women's education.In addition, measure of women's education aid, and alternative estimation strategies are used.This section, however, reports some of the checks conducted due to space shortage.The first part of the sensitivity analysis involves using alternative indicators of women's education (i.e.females' primary school completion rate, adults' female literacy rate as a percentage of females ages 15 and above, and females' effective progression rate to secondary school) in Equation (1).To examine the effect of women's education aid in a broader context, these variables are essential. 16urthermore, a growing body of literature (e.g.Offiong et al., 2021) suggests that these variables are effective indicators of educational attainment, which has a significant impact on economic development in general.A study of adult female literacy, for example, can determine the quality of the future labor force and can be used to ensure policies that promote life skills for both men and women.In addition, it can be used as a proxy measure Table 3.The impact of women education aid on women's education in developing countries: Using averaged data.

Gross primary enrolment ratio
Gross secondary enrolment ratio Gross tertiary enrolment ratio to assess the effectiveness of an education system; a high literacy rate indicates that an education system is able to provide literacy opportunities to a large population.The accumulating achievements of education are essential to furthering intellectual growth as well as social and economic development, although they do not guarantee the quality of education.When women are literate, they can seek and utilize information for the benefit of their household members' health, nutrition, and education.As a result, literate women are also able to play a meaningful role in society (see Cutler et al., 2006;Feeny & Ouattara, 2013;Galiani et al., 2017).These variables are derived from WDI data.Table 4 presents the effect of women's education aid on females' primary school completion rate, female literacy rate as a percentage of the female population aged 15 and over, and females' effective progression rate to secondary education based on data averaged over five years. 17n accordance with Table 3, Table 4 shows that women's education aid effectively educates women in recipient countries.A $1 increase in women's education aid contributes to a 0.56% improvement in females' primary school completion rate (see specification 2 of Table 4).In addition, the coefficients of most control variables are in the expected direction.According to the system-GMM estimator, the lagged dependent variable coefficient is close to one.
The second set of sensitivity analyses reported here uses alternative measures of women's education aid.Based on the discussion above, the preferred measure of women's education aid has been per capita, as recommended by Mishra and Newhouse (2009), Dreher et al. (2008), d'Aiglepierre andWagner (2013).However, the literature on aid effectiveness also makes use of two additional measures of aid flows to developing countries.The aid-to-real GDP ratio (see, for example, Burnside & Dollar, 2000;Clemens et al., 2004;Collier & Dehn, 2001) and the lagged amount of aid (see, for example, Dreher & Lohmann, 2015) are two examples of such measures.As can be seen in Table 5, there is a strong correlation between these three measures of women's education aid.Consequently, swapping the women's education aid measures has no significant impact on the qualitative nature of the benchmark results.
According to Table 6, the coefficients of the aid term are qualitatively similar to the baseline results across the three measures of women's education aid, further supporting the strong correlation between the alternative measures.Based on annual and five-year averaged data, four regressions are conducted for each alternative measure of women's education aid.According to Table 6, all regressions indicate that women's education aid contributes to the education of women in the recipient countries.In Table 6, the results are maintained when alternative indicators of women's educational attainment are used as dependent variables (results are available upon request).
The third set of robustness checks involves the use of alternative estimation strategies.Therefore, I estimate Equation (1) using the fixed effects (within) estimator.Due to a lack of space, the results are not reported.However, the results are available on request.As in Table 3, the number of observations and countries are the same.Furthermore, the results from fixed effects are consistent with those reported in Table 3.  Notes: The correlations are consistent with the number of observations used in regressions of Table 3.

Conclusions
This study examines the impact of women's education aid on women's education in 72 recipient countries from 1990 to 2016.According to the results, women's education aid per capita has a significant impact on women's education.Moreover, these findings are robust to alternative indicators of women's education, alternative measures of women's education aid, alternative estimation strategies, the inclusion of instruments to control for the issue of endogeneity of aid, as well as the inclusion of control covariates in the estimation process.This result contradicts some findings suggesting that foreign aid has an insignificant effect on the economic growth of recipient countries.The findings of this study highlight the necessity of disaggregating aid in order to examine its effectiveness.There is a significant contribution of women's education aid to the achievement of higher levels of education for women in recipient countries.According to the preferred 2SLS and system-GMM regressions (see Table 3), an extra dollar spent on women's education will benefit women by more than 0.2 percent.The education of women could be substantially advanced if donors doubled their aid allocation to this field.
Aid to women's education that promotes women's education is likely to have a substantial impact on economic development as well.Despite this, increased women's education could play an important role for economic development even if the relationship between women's education and economic development is weak.Higher levels of education among women are associated with a variety of health outcomes.More specifically, higher levels of education among women are consistently associated with lower fertility rates (Martin, 1995).
In addition, there has been a recent revitalization of the aid literature.Specifically, the question has been shifted from 'Does aid work?' to 'Under what conditions can aid be expected to be helpful?'In econometric terms, this is captured by including interactions between aid and other factors that may influence aid effectiveness in recipient countries.Burnside and Dollar (2000) provide the most persuasive evidence in support of this argument, which indicates that the quality of fiscal, monetary, and trade policies in recipient countries significantly affects economic growth.However, Chong et al. (2009) found that aid has little impact on income inequality, although it interacts with the measure of institutional quality.It may reveal that donors should be selective when allocating aid for women's education if there is a lack of consensus on the conditionality.There appears to be a need for women's education aid, as reflected by the low level of education available to women.There is, however, another aspect of selectivity that is stressed by many aid agencies-the quality of institutions within recipient countries may not be as important as commonly believed.There is still room for further research in this area.

Figure 1 .
Figure 1.Share of women's education aid out of total aid.Source: AidData database commitments.

Table 2 .
The impact of women education aid on women's education in developing countries: Using annual data.
dependent variables are gross primary, secondary and tertiary enrolment ratios.t-statistics in parentheses are based on robust and clustered standard errors.A constant, time dummies and country fixed effects are included but coefficients are not reported (results are available upon request).In all specifications, except in specifications 1, 5 and 9, per capita women education aid is instrumented variable.To control the endogeneity issue, I use the lagged values of growth of per capita income, ICRG index, child mortality rate, female children employment ages 7-14, government expenditure in education and total education aid in 2SLS and system-GMM.

Table 4 .
The impact of women education aid on alternative measure of women's education: Using averaged data.

Table 5 .
Correlations between different measures of women education aid.

Table 6 .
The impact of women education aid on females' gross primary enrolment ratio using alternative measures of aid: Using annual and averaged data.