Theme 02 of 43

Education, Social Mobility & Inequality

A Black student at a public school scores 160 points lower on SAEB than a white student at a private school — Brazil runs two education systems that barely touch each other.

The microdata from the National High School Exam (ENEM), maintained by INEP, records the performance of millions of students in every edition — with information on school type, administrative network, race or color, and the socioeconomic status of each candidate. Cross-referenced with the Basic Education Development Index (IDEB), which measures proficiency school by school and municipality by municipality, the two panels show that Brazil doesn't have one education system: it has two, which barely communicate with each other, and the gap between them deepens every time school type is cross-referenced with race and territory.

The scandalous gap: private versus public

The ENEM data leaves no room for interpretation: whoever pays for basic education comes out ahead of whoever depends on the state, and the difference isn't subtle. Private-school students — a small fraction of total test-takers — score far higher in math and writing than students from municipal and state schools. The gap holds edition after edition, which suggests a structural chasm, not a one-off fluctuation in a single exam.

Nineteen percentage points separate those who pay from the rest of the population — and that math advantage translates directly into a spot at a public university.

Educational apartheid

Only 6.4% of students who took the ENEM come from private schools, yet they concentrate the cream of admissions at the country's best public universities. The other 93.6% depend on a tax-funded public network that delivers, on average, noticeably worse results. It isn't a problem of access to the exam — it's a problem of which education each child receives before ever sitting for it.

A slice of just 6.4% of students disproportionately occupies the most competitive public higher-education slots — the rest of the country competes at a disadvantage before it even sits for the exam.

Teachers versus bankers: a tenfold gap

The pay gap between the banking sector and basic-education teaching is a full order of magnitude. A banker earns, on average, the equivalent of ten basic-education teachers. That differential isn't a statistical footnote — it's one of the clearest signals of where Brazil's formal labor market directs its most talented young people, and why teaching keeps losing that competition before it even begins.

A banker earns in one month what a teacher takes almost a full year to earn — and it's that math, not a lack of calling, that pulls the best talent away from the classroom.

The myth of meritocracy

The 100 points that separate the average ENEM score of private and state schools amounts, in practice, to roughly two years of schooling. A student who graduates from a state school finishes high school performing at the level of someone two years younger at a private school. Calling that result "individual merit" ignores that the race began with a two-year head start already handed to one side.

Talking about meritocracy when the public network starts two years behind the private network is confusing an unequal system with a fair contest.

IDEB: the scandalous gap between municipalities

Territory weighs as much as the network. Municipalities with a Human Development Index above 0.7 have a much higher average IDEB score than municipalities with an HDI below 0.5 — and the gap grows in the later years of elementary school, precisely when more poor students drop out. The distance between rich and poor municipalities is larger than the distance between school types within the same municipality, which reframes the problem: it isn't enough to debate public versus private schools — the geography of inequality has to be part of the discussion.

The IDEB gap between rich and poor municipalities is larger than the gap between school types — a student's zip code weighs more than is usually admitted.

SAEB: performance by race and network

When race is cross-referenced with school network, the effects don't add up — they multiply. A white student at a private school scores 625 points on average in math; a Black student at a public school scores 465. The 160-point gap combines two disadvantages that, in isolation, would already be serious, but that together produce a nearly unbridgeable chasm within the education system itself.

A Black student at a public school scores 160 points lower than a white student at a private school — the combined effect of race and school network is larger than either factor alone.

Functional illiteracy by generation

The functional-illiteracy rate has fallen generation after generation, but at the wrong pace. Those born before 1960 carry a rate of 40%; those born after 2000, 8%. That looks like progress — and it is — but the advance is too slow to close the class gap within the same generation: a poor family's granddaughter born in the 2000s still faces a higher risk of functional illiteracy than a wealthy family's grandmother born decades earlier.

Generational progress is real, but too slow: a poor granddaughter still faces more risk of functional illiteracy than a wealthy grandmother born decades before her.

Schools without basic infrastructure

Talking about digital education, science labs, or quality teaching sounds abstract against concrete numbers: 40% of public schools have no internet, 72% have no science lab, and one in seven still lacks treated water. It isn't possible to discuss digital literacy or experimental science teaching when basic infrastructure is still missing from a substantial share of the network.

Four in ten public schools have no internet — talking about digital classes in that scenario ignores the shop floor of Brazilian education.

State schools lose in aggregate and win at every stage

Comparing school networks by average IDEB score produces a result that seems to settle the argument: in 2023 municipal networks score 5.28 against 4.81 for state networks. The intuitive conclusion — that municipalities teach better than states — is false, and the data itself shows why. Split by stage, state networks beat municipal ones at both: 5.91 against 5.62 in early primary and 4.93 against 4.58 in later primary. The aggregate inverts because the two networks do not serve the same thing: municipalities concentrate enrolment in the early years, where scores are naturally higher, while states carry the later years and secondary school, where scores fall. It is a textbook Simpson's paradox, and the reason network rankings without stage controls measure enrolment composition rather than teaching quality.

State networks score higher in early primary and in later primary, yet lower overall — because they serve proportionally more of the stages where everyone does worse.

Student origin explains a third of the score — and 811 schools prove the rest

INEP calculates, for each school, a socioeconomic index of its student body. Crossing it with the IDEB score of 41,276 early-primary schools separates origin effect from school effect. The correlation is 0.544: strong, but far from deterministic — in variance terms, socioeconomic origin explains roughly 30% of the score, leaving the other 70% to everything else, including teaching quality. The gradient is clear, from the poorest quintile (IDEB 4.91) to the richest (6.55). But the exception is the figure that matters for policy: 811 schools in the lowest socioeconomic quintile beat the average of the highest, nearly 10% of them. In the opposite direction, only 115 wealthy schools (1.4%) fall below the average of the poor ones. Origin weighs, but it does not sentence.

811 schools in the poorest quintile score above the average of the richest — while only 115 wealthy schools fall below the average of the poor ones.

Explanatory hypotheses

The most direct hypothesis is educational apartheid: Brazil maintains, in practice, two parallel school systems with very little mobility between them. Cultural-reproduction theory helps explain the mechanism — the cultural capital accumulated by wealthier families is transmitted through school, reinforcing advantages that already existed at home. The connection to territory shows that the public school is not an island: it reflects the vulnerability of the surrounding community, which explains why the gap between poor and rich municipalities exceeds even the gap between school networks.

Policy implications

Equalized per-student funding between rich and poor municipalities could reduce a significant share of these territorial disparities. Raising teacher pay — currently ten times below banking pay — is a precondition for attracting and retaining the best professionals in the classroom. Integration programs between public and private networks, even if gradual, could start to break the educational segregation that today separates the two systems. Universal school connectivity has stopped being a luxury and become a basic prerequisite for any digital-education policy. And tutoring and reinforcement programs aimed at Black public-school students are urgent in the face of an inequality that accumulates, layer upon layer, across race, network, and territory.