Social psychologists discovered in the 1950s that counter-attitudinal advocacy—e.g., paying people to express a repellant view, but not quite enough to fully rationalize it—is an effective form of persuasion. So I should preface this post by noting that I recently traveled to the Global Education and Skills Forum in Dubai on a free economy-class airline ticket to participate in a mock debate where I was assigned, together with Princess Sarah Zeid of Jordan, to argue the somewhat obnoxious position that international aid donors spend too much money on girls schooling.

Luckily, I did not persuade anyone, including myself. There was even an audience poll before and after the debate to prove Lucy Lake of Camfed and Caroline Riseboro of Plan soundly defeated our team. But cognitive dissonance is a powerful drug. So while I still think it’s silly to argue we spend too much on girls education, perhaps it’s reasonable to ask—as I did, not persuasively, in Dubai—whether a concern with gender equality and a cold hard look at recent data would lead anyone to put their marginal dollar into girls schooling over, say, campaigning for gender quotas (which seem to work well in Indian politics, at least) or even subsidized childcare (which has boosted female labor force participation in Latin America).

A simple decomposition of the gender pay gap in 12 countries

Let’s start with the gender pay gap. The simple and unsurprising fact is that the gap in earnings power between men and women is enormous in many parts of the developing world, and—perhaps more surprisingly—almost none of this has to do with educational disadvantage.

Over the last several years the World Bank has launched a program, known as the STEP surveys, to collect detailed, comparable data on labor market outcomes including not just schooling but also skills. A number of things stand out from this data. (Note that the surveys only cover urban areas in each country.)

First, gender pay gaps are huge—and much bigger than often reported if we step back a bit. There is a tendency in documenting these numbers to rush ahead and focus on pay differentials conditional on being employed. Famously, American women earn 79 cents for every dollar a man earns for the same hours worked. But particularly in the developing world, the biggest source of income gaps stems from lower rates of labor force participation. Looking at the unconditional gender gaps—i.e., including all the zeros for men and women without a cash income—in urban Ghana women only earn about two-thirds of what men earn. In Colombia it’s only half, and in Sri Lanka women earn less than a third of what men earn.

Second, education explains almost none of this. We can see this using a simple Blinder-Oaxaca decomposition. This simple, descriptive technique estimates the return to schooling in observational data, then given those returns, breaks the gender pay gap into the part that can be explained by education differences between the sexes and those which remain unexplained even if we equalized schooling.

The result in the case of the World Bank data is that in only one country (Laos) does girls’ schooling deficit explain a large share of the gender pay gap. In the other 11 countries it doesn’t. In half of the countries, women’s education actually exceeds men’s on average. And in five of the remaining six countries, education explains relatively little of the earnings gap.

The data and code Blinder-Oaxaca decompositions in the graph is available here (and thanks to Lee Crawfurd for sharing earlier STEP do-files).

Even where education is (nearly) equal, nearly nothing else is

Consider the case of India. As noted in an earlier post, the government’s recent Economic Survey calculated there are still 63 million missing women, and another 22 million "unwanted women" based on the tendency of families to keep having kids if they have a girl and stop once they have a boy. Gender discrimination is acute and deeply rooted.

In their adult lives, those girls—wanted or unwanted—are more than 50 percentage points less likely than their brothers to have a job, and if they do, they’ll earn about 22 percent less. The political sphere is often no better than the economic. Just 66 of India’s 543 national legislators in the Lok Sabha are women.

But rather than girls schooling being an obvious target for change, education indicators are actually an oasis of relative gender parity in India’s social statistics. According to official statistics, girls’ primary and lower-secondary completion is already higher than boys, and gross tertiary enrollment is at parity. Independent data collected by the NGO Pratham in the ASER surveys finds that girls at age 13 are slightly (4 percent) less likely to be able to solve the hardest questions on a simple mathematics test. While in need of redress, this 4 percent mathematics gap can hardly explain the horrific economic and social disparities we see in other domains.

In a sense, everything in life is stacked against Indian girls except school. (And kudos to the activists who made that true.)

Indicator Units Source Girls Boys Gap
Gender ratio of children without siblings Ratio India Economic Survey (2018) 1 1.8 -44%
Stunting % World Bank, WDI 38 40 -2%
Primary school % completed World Bank, WDI 100 95 5%
Secondary school % completed World Bank, WDI 88 83 5%
Tertiary school % enrolled, gross World Bank, WDI 27 27 0%
Math test at age 13 % who can do division Pratham, ASER 48 52 -4%
Reading test at age 13 % who can read story Pratham, ASER 71 71 0%
Labor force participation % World Bank, WDI 27 79 -52%
Earnings conditional on employment rupees/day NSSO 2011-12 317 407 -22%
Seats in the national parliament Seats Lok Sabha 66 477 -86%

By all means, let’s spend more on girls schooling. It’s a great investment. But the reason women in Sri Lanka earn a third of men isn’t because they’re less educated. They’re not. Similarly, the reason there were more men named “Michael” at Davos this year than all women combined isn’t because there weren’t enough educated women to invite. And the reason fewer than one in five legislators in the developing world is a woman isn’t because there are no women qualified to serve. If donors spend money on girls schooling programs thinking, “If only girls were more educated, they’d get equality” they’re fooling themselves, and implicitly accepting the status quo.

Thanks to Maryam Akmal and Divyanshi Wadhwa for excellent research assistance.