> For the complete documentation index, see [llms.txt](https://ambitious-impact.gitbook.io/ambitious-impact-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://ambitious-impact.gitbook.io/ambitious-impact-docs/explainers-and-house-views/live-house-views/what-are-the-returns-to-education.md). # What are the returns to education? ## Introduction and Summary This short report summarizes Ambitious Impact’s current research and views on the value of education – i.e., what are the returns to educational improvements? We draw on this research and its conclusions when assessing education interventions. This report is an edited version of parts of section 4 in our report on structured pedagogy ([Cox, 2024](https://370b8129-500b-4a5d-99f2-ce6886702186.usrfiles.com/ugd/370b81_1885f40e02af4286a00811162be9b0b3.pdf)). Summary of findings and reflections * We generally believe there are returns to education and that these come from a mixture of financial (improved income at the individual level, improved growth at the country level) and non-financial gains (socialization, health outcomes, delayed pregnancy, societal values, etc.) (e.g., [Crawfurd et al., 2019](https://www.cgdev.org/blog/does-education-need-qaly-and-lays-it); [GiveWell, 2018](https://www.givewell.org/international/technical/programs/education); [Calvert, 2019](https://www.founderspledge.com/downloads/fp-education)). * We (cautiously) endorse the view that quality of education is a large determinant of the returns to education beyond how many years of schooling one receives regardless of educational attainments (e.g., [Pritchett, 2015](https://www.theguardian.com/global-development/2015/sep/29/majority-of-world-children-in-school-so-why-arent-they-learning)). * To simplify modeling, we use income gains as a measure of increased welfare later on in life. We are highly uncertain about the magnitude of returns but expect that for a primary school child in an LMIC, the benefits will be between 10% and 40% increased income across their working life, per one standard deviation improvement in test scores. When modeling, we will model both a conservative case (10%) and an optimistic case (40%). We think the conservative case is likely closer to the true returns. * We also think that improving education access and quality has many benefits beyond income that we value. These include improving personal agency, health, civic engagement, and women’s empowerment. Table 1: Income gains, using standard deviations in quality or quantity of education | Study | Effects on income | | ------------------------------------------------------------------------------------------------------------------------------------ | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Founders Pledge ([Calvert, 2019](https://www.founderspledge.com/downloads/fp-education)) | “The effect of one year of education ranges from around 5% to 20%, and the effect of a one standard deviation increase in test scores ranges from around 10% to 40% (though some estimates are much higher).” (p.62) | | Angrist et al. ([2020](https://openknowledge.worldbank.org/server/api/core/bitstreams/0e25e5c2-23e8-42b2-bb35-36d03aa60a66/content)) | “we take the value r = 0.10 for the rate of return per school year, and w = 0.20 per standard deviation increase in learning, based on parameter values used in the literature” (n.p) | | Evans and Yuan ([2019](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3338189)) | 37% per standard deviation test score improvement | | Hanushek and Woessmann ([2015](https://hanushek.stanford.edu/publications/knowledge-capital-nations-education-and-economics-growth)) | This modeling study suggests that a 0.25 SD increase in test scores for the entire population (modeled in both high- and low-income countries) would lead to a 12% increase in wages over 80 years (in terms of the present value of future GDP) (see [here](https://docs.google.com/document/d/18v_i5NLxG5QSQIfNm3ADmcT5qEuJEeY6R4N0Eq9ilN4/edit?tab=t.0#heading=h.795ohsirkatw)). | | Hanushek et al. ([2016](https://www.nber.org/system/files/working_papers/w22657/w22657.pdf)) | 1 SD increase in test scores can lead to a 10–45% increase in earnings (see [here](https://docs.google.com/document/d/18v_i5NLxG5QSQIfNm3ADmcT5qEuJEeY6R4N0Eq9ilN4/edit?tab=t.0#heading=h.795ohsirkatw)). | | Hanushek et al. ([2013](http://www.nber.org/papers/w19762.pdf)) | 1 SD increase in test scores can lead to a 12–28% increase in earnings (see [here](https://docs.google.com/document/d/18v_i5NLxG5QSQIfNm3ADmcT5qEuJEeY6R4N0Eq9ilN4/edit?tab=t.0#heading=h.795ohsirkatw)) | ## Methodological notes The value of education is generally quite difficult to define and measure. What you choose to measure and model will change your results, and reasonable people can disagree about the best model. As a result, there is some disagreement about the returns to education or whether there are any ([Pritchett, 2001](https://openknowledge.worldbank.org/handle/10986/17434)), and debate about methodology and approaches to measuring both improvements in education and its returns abounds (e.g., whether to use learning-adjusted years of education or standard deviations in test scores [Crawfurd et al., 2019](https://www.cgdev.org/blog/does-education-need-qaly-and-lays-it); [Angrist et al., 2023](https://openknowledge.worldbank.org/server/api/core/bitstreams/0e25e5c2-23e8-42b2-bb35-36d03aa60a66/content)). Establishing concrete returns to education is further complicated by several plausible mechanisms to explain the returns of education, most importantly (and to simplify) whether we believe that returns are mostly driven by signaling (credentials gained by attending school) or by gaining productive capacities through learning. Finally, there is also a “live” question on general equilibria. Namely, do the returns on education for some—more educated—individuals come at the expense of those who did not get as much education? If so, education interventions could be a zero-sum game that simply reallocates income across a population without actually increasing productivity (e.g., [Duflo, 2001](https://www.aeaweb.org/articles?id=10.1257/aer.91.4.795); [Duflo, 2004](http://www.sciencedirect.com/science/article/pii/S0304387803001846)). ### Limitations of our research This report is to be read as a live document where we depict our current views on education most accurately. It is less useful as a literature review or summary of research given that it is bounded by us not being education specialists, and having little time to invest in exploring the literature with the depth it deserves. In forming our views, we relied largely on the views of other researchers from organizations we trust to do good research on education, such as the World Bank, researchers affiliated with the Research on Improving Systems of Education (RISE) program, the Centre for Global Development, Give Well, and Founders Pledge, among others. All errors remain our own. ## Individual income effects Based on our literature review, we expect that, on average, improvements in education lead to higher wages later in life. The magnitude of the returns will vary across contexts (see, for instance, the different results arrived at in the observational evidence section). Important factors likely include demographic characteristics, a country's stage of development, unemployment, and industry, among others. ### Experimental evidence We identified two randomized controlled studies (RCTs) testing the long-term benefits of education. Duflo, Dupas, and Kremer ([2017](https://www.nber.org/papers/w22931)) evaluated the long-term benefits of providing four-year scholarships to secondary school students in Ghana. They found that, compared with the control group, treated students experienced 1.26 additional years of schooling, scored 0.15 SDs higher on tests one year after graduation, and had monthly earnings 24% higher after four years since graduating. If the entire income increase was due to enhanced learning outcomes, then this would imply an income increase of 160% for each standard deviation increase in test scores. Bettinger et al. [(2014](http://conference.nber.org/confer/2014/EDf14/Bettinger_Kremer_Kugler_Medina_Posso_Saavedra.pdf)) conducted an RCT testing the effect of vouchers for private secondary school scholarships in Colombia, acquiring follow-up data 8-14 years after secondary school completion (Bettinger et al., 2017; unpublished but referenced in [GiveWell, 2018](https://www.givewell.org/international/technical/programs/education) and [Calvert, 2019a](https://founderspledge.com/research/fp-education)). These studies evaluated the long-term impact of providing vouchers for private secondary school scholarships to disadvantaged students in Colombia. They found that these students’ formal sector earnings were 8% greater at age 30 and that they scored 0.2 SDs higher on tests three years after graduation, compared with the control group. If the entire income increase was due to increased learning outcomes, this would imply an income increase of 40% for each standard deviation increase in test scores. ### Quasi-experimental evidence We identified eight further quasi-experimental studies exploring this question. These are summarized in Table 2 below. Table 2: Summary of quasi-experimental studies | Study | Summary | | ------------------------------------------------------------------------------------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | [Filmer and Schady (2014)](http://jhr.uwpress.org/content/49/3/663.short) |

Secondary school scholarships in Cambodia increased the number of years spent in school, but had no effect on learning outcomes (i.e., test scores), employment, or earnings.

This suggests two points: (1) an argument against the signaling effect of education as there was no difference in employment, even with more years in school and (2) schooling does not automatically translate into learning. The study does not provide us with information about the returns to learning.

| | [Duflo (2001)](https://www.aeaweb.org/articles?id=10.1257/aer.91.4.795); [Duflo, (2004)](http://www.sciencedirect.com/science/article/pii/S0304387803001846) |

This study found that the construction of new schools in Indonesia led to an additional 0.12–0.19 years of school and that earnings increased by 1.5–2.7% for students who benefited directly from the program. However, earnings decreased by 3.8–10% for non-beneficiaries (those who were too old to attend these new schools).

This might suggest that the students graduating through the new schools might have increased the supply and taken over jobs previously held by other people, suggesting that there might not be a net positive effect if there is a finite number of employment opportunities.

| | [Ozier (2016)](http://jhr.uwpress.org/content/early/2016/12/12/jhr.53.1.0915-7407R.abstract) | This study found that increasing admission to secondary schools in Kenya led to students being more likely to be employed aged 20–25. The study did not report the relevant income effects. | | [Keats (2016)](http://akeats.faculty.wesleyan.edu/files/2017/01/Uganda_EdFert_Keats_20may16.pdf) | The elimination of primary school fees in Uganda led to girls gaining an additional year of schooling and – as a result – being more likely to be employed. The study did not report the relevant income effects. | | [Chicoine (2017)](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2868563) | Elimination of school fees in grades 1–10 in Ethiopia (): This study found that girls gained over two years of additional schooling and were more likely to work as a result, but again did not report the relevant income effects. | | [Brudevold-Newman (2021)](http://econ.andrewbrudevold.com/KenyaFSE.pdf) | This study found that after the elimination of secondary school fees in Kenya, girls were more likely to work in skilled jobs rather than in agriculture. It found no impact on wage work or self-employment. Income effects were not reported. | | [Card (2001)](https://davidcard.berkeley.edu/papers/return-to-schooling.pdf) | Taking advantage of natural experiments such as school leaving ages or college openings, the study found that – by and large – while some evidence of weak screening is revealed, education is generally associated with higher earnings due to productivity rather than to screening. | | [Clark and Martorell ( 2014)](http://masteringmetrics.com/wp-content/uploads/2015/02/clark_martorell_2014.pdf) | A comparison of the earnings of workers who narrowly passed versus those who narrowly failed high school exit exams in Texas. The study finds that the diploma for passing did not have much effect on future earnings, suggesting that the value of education is in what is learned more than in the diploma or other certificates granted. | ### Observational evidence There is abundant observational evidence on the relationship between education and wages. This often takes the form of looking at cross-country comparisons. For instance, OECD data shows a correlation between education and wages in every one of the 35 different countries it holds data for ([OECD, 2023](https://stats.oecd.org/Index.aspx?DataSetCode=EAG_EARNINGS)). Estimates of the effects of education are common in the literature. Annex A of Psacharopoulos and Patrinos ([2018](http://documents.worldbank.org/curated/en/442521523465644318/Returns-to-investment-in-education-a-decennial-review-of-the-global-literature)) provides 244 estimates from 100 different countries (and around 40 studies) of the rate of return to education. Of these 244 estimates, two were negative, and 242 were positive. The average rate of return was 12.8%, with most estimates falling between 5% and 15%. A big part of this question's complexity is inherent in the problem: We are trying to answer some very complicated questions about causal relationships between variables from incomplete observational data. Therefore, answers are only going to be as good as the assumptions behind the models and the quality of the data collected. Many observational studies use micro-Mincer regression analysis to determine the returns to education. This regression analysis is based on the “Mincer rate of return to schooling,” which is based on the following formula:
where W is the individual’s wages, S is the years of schooling or cognitive skills based on whether years of schooling or test scores are being evaluated, E is the labor market experience, which is age minus years of schooling, E2 captures the non-linear relationship between wages and labor market experience, X is a vector of observed exogenous explanatory variables that could affect earnings such as gender, location, etc., U is a disturbance term representing other forces that may not be explicitly measured, and β is the marginal private rate of return to education. Our understanding is that the Mincer rate of return looks at the earnings at one point in time, e.g., at age 30, rather thúan being based on a lifetime of earnings. In most of the observational studies outlined below, it is unclear at what age these regressions are performed. #### Additional years of schooling Several studies looked at the effect of additional years of schooling on earnings. [Psacharopoulos and Patrinos, 2018](http://documents.worldbank.org/curated/en/442521523465644318/Returns-to-investment-in-education-a-decennial-review-of-the-global-literature): This micro-Mincer regression analysis based on 705 estimates by the World Bank of 139 economies between 1950 and 2014 finds the following most relevant returns to an additional year of schooling: * 1 additional year of schooling = 8.8% increased earnings, based on the world average. * 1 additional year of schooling in sub-Saharan Africa = 10.5% increased earnings. * 1 additional year of schooling at primary school = 25.4% increased earnings. This is the most cited result on the returns of an additional year of schooling in the education literature, and is the number used by the World Bank ([Patrinos, 2015](https://blogs.worldbank.org/education/50-years-returns-education-studies)). Note that this micro-Mincer regression does not appear to include the Xi variable detailed above and is therefore unlikely to account for other factors that could influence earnings. Rather, it just assumes that if person X spent a more years in school, and earns b more, that difference in earnings is entirely due to the extra education. [Montenegro and Patrinos, 2014](https://documents1.worldbank.org/curated/en/830831468147839247/pdf/WPS7020.pdf): This micro-Mincer regression analysis based on 819 estimates by the World Bank between 1970 and 2013, again of 139 economies, finds the following most relevant returns to an additional year of schooling: * 1 additional year of schooling = 9.7–10.1% increased earnings, based on the world average. * 1 additional year of schooling in sub-Saharan Africa = 12.4% increased earnings. * 1 additional year of schooling at primary school = 10.6–11.5% increased earnings. * 1 additional year of schooling at primary school in sub-Saharan Africa = 14.4% increased earnings. Again, this micro-Mincer regression does not appear to have included the Xi variable, and therefore is unlikely to account for other factors that could influence earnings. #### Improved test scores Other studies look at the effect of improved test scores on earnings. [Hanushek and Woessmann, 2015](https://hanushek.stanford.edu/publications/knowledge-capital-nations-education-and-economics-growth): This modeling study suggests that a 0.25 SD increase in test scores for the entire population (modeled in both high- and low-income countries) would lead to a 12% increase in wages over an 80-year period (in terms of the present value of future GDP). This analysis regresses growth between 1960 and 2000 due to average cognitive skills (based on international standardized test scores) for that period to estimate a coefficient for the effect of cognitive skills on earnings and growth. It includes control variables for economic openness and security of property rights because more educated countries tend to have more open economies. These control variables are included such that they are only measuring the effect of education rather than other traits that are correlated with education. The authors argue that the relationship between cognitive skills and growth and earnings as a result of this model is mostly causal because: 1. This is not just a case of high growth leading to better cognitive skills. At least some of the causal relationship is better cognitive skills leading to higher growth, because cognitive skills known to come from non-growth sources have a positive association with growth. 2. The trend in cognitive skills is strongly related to the trend in growth. 3. The wages of Immigrants in the US labor market (who have therefore experienced different education systems) reflect the quality of education in their place of origin. 4. Breaking down GDP differences between countries into the differences due to human and physical capital and due to efficiency shows that a portion of growth comes from human capital. As we discussed above on observational evidence, there could be other factors for which the regression has not sufficiently controlled and which would explain those effects. [Hanushek et al., 2016](https://www.nber.org/system/files/working_papers/w22657/w22657.pdf): This micro-Mincer regression analysis of full-time workers aged 35–54 across 32 OECD countries finds that a 1 SD increase in test scores can lead to a 10–45% increase in earnings. This is the most cited result on the returns of improved learning outcomes in the education literature. Note that this study does include the Xi variable but assumes that all of this difference in earnings will come from gender and not from any other factors that could influence earnings. [Hanushek et al., 2013](http://www.nber.org/papers/w19762.pdf): This micro-Mincer regression analysis of full-time workers aged 35–54, again across 32 OECD countries, finds that a 1 SD increase in test scores can lead to a 12–28% increase in earnings. Again, this study includes the Xi variable but assumes that all of this difference in earnings will come from gender and not from any other factors that could influence earnings. #### Relationship between foundational literacy and earnings To better understand the value of education, we examine whether there is a direct connection between literacy ability and earnings. This is more direct than examining the connection between schooling and earnings, which is problematic if students do not actually learn anything. It also allows us to separate out any screening effects (where employers hire based on years of schooling rather than ability). The availability and quality of data on this question are low. This is unsurprising as the returns to foundational literacy have very long feedback loops that are hard to measure. We have some information from high-income countries (mostly observational or quasi-experimental) and some information from adult literacy programs. Neither of those information points would necessarily cross-apply to young students in low-income countries. Nevertheless, the most relevant pieces of evidence include: * [Coulombe and Tremblay (2006](https://www.degruyter.com/document/doi/10.2202/1534-5998.1404/html)): From the demographic profile of the 1994–98 International Adult Literacy Survey, the authors derive synthetic time series for the 1960–95 period on the literacy level of labor market entrants. This information is then used to measure investment in education in a two-way error correction panel data analysis of cross-country growth for a set of 14 OECD countries. The results indicate that the skills associated with one extra year of schooling increase aggregate labor productivity by approximately 7%. * [Gibson et al. (2019](https://files.eric.ed.gov/fulltext/ED600898.pdf)): In this study, tax data linked to the Youth in Transition Survey (YITS) and the Programme for International Student Assessment (PISA) are used to examine the association between background factors at age 15 (including reading proficiency) and employment earnings in young adulthood for a cohort of respondents in Canada who were aged 15 in 2000. They found that young women and men with higher reading proficiency at age 15 earned more than their counterparts with lower reading proficiency, both in the first year after leaving school and in the ensuing years. After accounting for various individual and background factors (e.g., parental education, parental income, overall marks at age 15, and other characteristics such as highest level of education and field of study), reading proficiency at age 15 was no longer associated with employment earnings after leaving school – for both men and women – even in the first year after leaving school. #### Correlation between literacy rates and GDP Taking the GDP per capita data on the x-axis (x-log) and data about the literacy rate on the y-axis, we can see that there seems to be a correlation between these two data points. Even if education (in the form of literacy) was one factor driving an increase in GDP per capita, this correlation seems to suggest that large increases in literacy would be necessary to have an effect on GDP per capita. Even if those effects existed, it is unclear if they would be evenly distributed throughout the poorest population, which would benefit most.
Figure 1: Literacy rate vs. GDP per capita, 2021 (source: [Our World in Data, 2021b](https://ourworldindata.org/grapher/literacy-rate-vs-gdp-per-capita)). ## Economic growth We can examine the economic benefits of education to individuals and its effects on a country’s economy. This subsection discusses the literature we found on this approach. Overall, we view establishing this relationship between education gains and growth as particularly difficult. The results reached often depend heavily on the economic models being used. We think that findings on this question are much more uncertain than those on individual income effects. Many of the studies we found rely on cross-country comparisons. As with the literature on individual effects, cross-country comparisons are relative – for example, if one country invests more in education, another invests more in health, and a third invests more in infrastructure, and all countries grow at the same rate, that might mean that all interventions are having an effect. Literature reviews on this topic suggest a positive relationship exists between levels of education and growth. For instance, Cameron and Cameron ([2006](https://citeseerx.ist.psu.edu/document?repid=rep1\&type=pdf\&doi=6450d4f70ca307a0979dff1c99ebddc88fa3f5bb)) detailed several studies examining the economic benefits of increased literacy. Among studies relevant to our report, they found eight suggesting a positive relationship, one suggesting indirect benefits, three with null results, and one leaning negative. ### Additional years of schooling Some studies have investigated the effects of additional years of schooling on economic growth at a country level. [Hanushek et al. (2008](https://www.researchgate.net/publication/279603863_Education_and_economic_growth_It's_not_just_going_to_school_but_learning_something_while_there_that_matters)) conducted an analysis of human capital and economic growth from 1960 to 2000 for 50 countries (mostly high-income with some low-income countries as well) and found that one additional year of schooling increased the average 40-year growth rate in GDP by 0.37 percentage points. Note that we think that this study may suffer from omitted variable bias. [Lau et al. (1991](https://www.researchgate.net/publication/44817722_Education_and_productivity_in_developing_countries_an_aggregate_production_function_approach)) estimated the effects of education by measuring the percentage change in five regions' real GDP against an increase (by one additional year) in the average years of schooling. The estimated effects range from negative to more than 5% per year, depending on the region. For East Asia and Latin America, primary and secondary education had a positive impact on real GDP, whereas in Africa and the Middle East, primary education appeared to have negative, but mostly insignificant, effects, while secondary education showed consistent statistically significant positive effects. [Pritchett (2001](https://openknowledge.worldbank.org/bitstream/handle/10986/17434/773740JRN020010l0the0Education0Gone.pdf?sequence=1\&isAllowed=y)) evaluated cross-national data from both developed and developing countries on the returns to education and found no association (or perhaps even negative association) between increases in human capital attributable to the rising educational attainment, in terms of additional years of schooling, of the labor force and the rate of growth of output per worker. This is the most cited piece of evidence by people skeptical of education's economic returns. Pritchett lists three potential reasons for these findings: * The governance environment could have been sufficiently perverse that the accumulation of educational capital lowered economic growth. * Education could be a zero-sum game: The marginal returns to education could have fallen rapidly as the supply of educated labor expanded while demand remained stagnant. * The quality of education could have been so low that additional years of schooling created no human capital. ### Improved test scores Other studies have examined the relationship between improved test scores and economic growth. In the same modeling study mentioned in the [section](https://docs.google.com/document/d/18v_i5NLxG5QSQIfNm3ADmcT5qEuJEeY6R4N0Eq9ilN4/edit?tab=t.0#heading=h.1k79pn8axaqc) on individual income effects, Hanushek and Woessmann (2015) suggested that a 0.25 SD increase in test scores for an entire population (modeled in both high- and low-income countries but with a skew to the former) would lead to a 6.2% increase in GDP over an 80-year period (in terms of the present value of future GDP). We can use this to determine that a 0.25 SD increase in test scores would lead to a 0.0752% increase in GDP annually. The same arguments on the causal relationship between cognitive skills and growth discussed in the previous section apply. [Hanushek and Woessmann (2012](https://link.springer.com/article/10.1007/s10887-012-9081-x)) suggested that a 0.25 SD increase in test scores for the entire population (modeled in both high- and low-income countries but with a skew to the former) would lead to a 4.3% increase in GDP over an 80-year period (in terms of the present value of future GDP). We can use this to determine that a 0.25 SD increase in test scores would lead to a 0.0526% increase in GDP annually. This analysis includes the Xi variable which includes a range of factors: family inputs, individual ability, and “other” including health. [Hanushek et al. (2008](https://www.researchgate.net/publication/279603863_Education_and_economic_growth_It's_not_just_going_to_school_but_learning_something_while_there_that_matters)) delivered an analysis of human capital and economic growth from 1960 to 2000 for 50 countries (modeled in both high- and low-income countries but with a skew to the former), finding a 0.5 SD increase in test scores increased the average 40-year growth rate in GBP by 0.63 percentage points. Note that we think that this study may suffer from omitted variable bias. [Benhabib and Spiegel (1994](https://www.scribd.com/document/182500686/THE-ROLE-HUMAN-CAPITAL-IN-ECONOMIC-DEVELOPMENT-EVIDENCE-FROM-AGGREGATE-pdf#)) used a Cobb-Douglas production function that included human capital as a factor. They found that human capital accumulation failed to enter significantly in the determination of economic growth, and even entered with a negative point estimate. This suggests that returns to schooling do not explain economic growth. ## Other benefits ### Other consequences Being in education brings many other sizeable benefits that must be accounted for, especially in unstable economies. Education can be thought of as a touch-point for many other key benefits that children should receive, such as vaccinations, social and civic education, and safeguarding, among others. 1. Health benefits: Education can improve the health of its beneficiaries and others by increasing the income available for beneficiaries to spend on health, as well as increasing engagement in preventative health behaviors, improving the ability of beneficiaries to understand and address their own health problems, reducing exposure to and transmission of sexually transmitted infections, and improving the quality of health services provided by beneficiaries to others. 2. Female empowerment, teenage pregnancy, and marriage: Education positively affects female empowerment as it increases the opportunities available to women and reduces the incidence of early pregnancy and marriage. Founders Pledge and GiveWell summarize experiment evidence from four studies in Ghana, Colombia, Kenya, and Malawi which support this positive relationship, with reductions of teenage pregnancy ranging from three to ten percentage points, and reductions of early marriage by four to 11 percentage points. 3. Crime: Education may reduce the likelihood that the beneficiaries will commit a crime by increasing the number and attractiveness of employment opportunities, making beneficiaries more patient or risk averse, and leading beneficiaries to socialize in groups less likely to commit crimes. 4. Citizenship and democracy: It has been shown that education increases voting rates. This increase in citizenship participation and democratic behavior could either be through explicit design within democratic education systems or by improving both awareness of history and citizens’ ability to make ‘good’ political decisions by increasing analytical skills. ### Intrinsic value Education is a basic human right. The right to education is enshrined in “the Universal Declaration of Human Rights (1948) and many other international human rights instruments. UNESCO’s foremost standard-setting instrument is the Convention against discrimination in education which dates from 1960 and has so far been ratified by 107 States. It is the first international instrument which covers the right to education extensively and has a binding force in international law.” ([n.d., n.p](https://www.unesco.org/en/right-education/need-know?hub=70224)). As individuals, many AIM staff subscribe to the view that there is some intrinsic value to education as a space for play, learning, and self-development, which should be enjoyed by all. This informs our views and focuses on exploring educational ideas, but at the end of the day, the interventions we recommend are usually recommended due to their consequences on other wellbeing markers, such as income or health. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation by asking a question. Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters: ``` GET https://ambitious-impact.gitbook.io/ambitious-impact-docs/explainers-and-house-views/live-house-views/what-are-the-returns-to-education.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.