About the package
educabr provides, under a single canonical tidy schema, the most extensive set of harmonised historical series on Brazilian education available in analytic format:
- Tertiary (higher-education) enrollment — 1907 to 2024, 118 years of coverage, drawing on seven distinct primary sources (IBGE Statistics of the 20th Century, Durham, Maduro Junior, Kang, INEP Synopsis, INEP Microdata, and the INEP CENSUP Power BI panel).
- Compulsory (fundamental) and upper-secondary enrollment — 1933 to 2010, with race/colour disaggregation from 1960 onwards (Kang, Paese & Felix 2021).
- Mean years of schooling of the adult population — 1925 to 2015, with breakdowns by sex, race/colour, macroregion and state (Walter & Kang 2024).
Every transformation is auditable: each output row carries a
canonical source key (catalogued in
inst/dict/vocabularies/sources.yaml) and a
source_note with the exact table or chapter of origin.
A two-function API
The public interface is intentionally minimal:
# Schemas of the two main return shapes
str(get_enrollment(level = "fundamental", year = 1950))
#> tibble [2 × 16] (S3: tbl_df/tbl/data.frame)
#> $ year : int [1:2] 1950 1950
#> $ geo_level : chr [1:2] "BR" "BR"
#> $ geo_code : chr [1:2] "BR" "BR"
#> $ geo_name : chr [1:2] "Brasil" "Brasil"
#> $ level : chr [1:2] "fundamental" "fundamental"
#> $ network : chr [1:2] "total" "total"
#> $ institution_type: chr [1:2] "total" "total"
#> $ modality : chr [1:2] "total" "total"
#> $ dim_race : chr [1:2] "total" "total"
#> $ age_group : chr [1:2] "7-14" NA
#> $ indicator : chr [1:2] "enrollment_rate" "enrollment_count"
#> $ value : num [1:2] 4.55e+01 4.73e+06
#> $ unit : chr [1:2] "percent" "count"
#> $ source : chr [1:2] "kang_paese_felix_2021" "kang_paese_felix_2021"
#> $ source_note : chr [1:2] "Kang, Paese & Felix (2021). RHE 39(2):191-218. doi:10.1017/S0212610921000112" "Kang, Paese & Felix (2021). RHE 39(2):191-218. doi:10.1017/S0212610921000112"
#> $ is_derived : logi [1:2] FALSE FALSE
str(get_schooling(year = 1950))
#> tibble [1 × 12] (S3: tbl_df/tbl/data.frame)
#> $ year : int 1950
#> $ geo_level : chr "BR"
#> $ geo_code : chr "BR"
#> $ geo_name : chr "Brasil"
#> $ dim_race : chr "total"
#> $ dim_sex : chr "total"
#> $ age_group : chr NA
#> $ indicator : chr "mean_years_schooling"
#> $ value : num 1.59
#> $ unit : chr "years"
#> $ source : chr "walter_kang_2023"
#> $ source_note: chr "Walter, J., & Kang, T. H. (2023). A new dataset of average years of schooling in Brazil, 1925-2015. FGV-IBRE working paper."
#> - attr(*, "educabr_meta")=List of 5
#> ..$ build_script : chr "data-raw/02_build_schooling_kang_fgv.R"
#> ..$ built_at : POSIXct[1:1], format: "2026-05-17 19:08:24"
#> ..$ primary_source: chr "walter_kang_2023"
#> ..$ citation : chr "Walter, J., & Kang, T. H. (2023). A new dataset of average years of schooling in Brazil, 1925-2015. FGV-IBRE working paper."
#> ..$ raw_files : chr "data-raw/sources/kang_fgv_ibre_2023/3._anos_estudo_1925_2015_v_abril2023.xlsx"Both return long-format tibbles that follow the same canonical schema
(inst/dict/schema.yaml), with columns for the geographic
unit, the inequality dimension (race, sex, …), the source, and the
value.
Case 1 — gross enrollment rate by race/colour
Using dimension = "race" we can trace the trajectory of
gross fundamental-education enrollment for the five IBGE race/colour
categories between 1960 and 2010:
fund_race <- get_enrollment(
level = "fundamental",
indicator = "rate",
geo_level = "BR",
dimension = "race",
year = c(1960, 2010)
)
head(fund_race)
#> # A tibble: 6 × 16
#> year geo_level geo_code geo_name level network institution_type modality
#> <int> <chr> <chr> <chr> <chr> <chr> <chr> <chr>
#> 1 1960 BR BR Brasil fundament… total total total
#> 2 1961 BR BR Brasil fundament… total total total
#> 3 1962 BR BR Brasil fundament… total total total
#> 4 1963 BR BR Brasil fundament… total total total
#> 5 1964 BR BR Brasil fundament… total total total
#> 6 1965 BR BR Brasil fundament… total total total
#> # ℹ 8 more variables: dim_race <chr>, age_group <chr>, indicator <chr>,
#> # value <dbl>, unit <chr>, source <chr>, source_note <chr>, is_derived <lgl>The gap between the three main categories (white,
black, brown) traces one of the central
narratives in the sociological literature on Brazilian education —
relative convergence at compulsory levels alongside persistent
inequality at higher levels.
Case 2 — multi-source comparison for tertiary enrollment
Brazilian tertiary enrollment has been reconstructed by multiple authors with slightly different methodologies, especially for the long 1933–2000 period when official sources are sparse. The educabr package keeps all competing estimates side by side, enabling direct comparison:
ter_1980 <- get_enrollment(
level = "superior",
year = 1980,
network = "total",
modality = "total",
indicator = "count"
)
ter_1980[, c("source", "value", "source_note")]
#> # A tibble: 4 × 3
#> source value source_note
#> <chr> <dbl> <chr>
#> 1 kang_paese_felix_2021 1377286 Kang, Paese & Felix (2021). RHE 39(2):191-218. …
#> 2 durham_2005 1377286 Durham (2005). Educação superior, pública e pri…
#> 3 kang_paese_felix_2021 1377286 Kang, Paese & Felix (2021), RHE 39(2):191-218. …
#> 4 maduro_junior_2007 1377286 Maduro Junior (2007). Taxas de matrícula e gast…For 1980 several sources converge on the same value (~1.38 million),
which suggests a stabilised figure in the literature. For other years
estimates diverge — passing source = "..." to
get_enrollment() is the recommended way to pin down a
single series for secondary analysis.
Case 3 — the reconstructed totals problem (2000-2008)
Between 2000 and 2008 INEP began collecting distance-learning (EAD)
enrollment in a separate CENSUP table (tabela7.x), but
did not add it to the total of in-person enrollment
published in tabela5.x. This means that headline “total”
series for that interval published by Kang, Durham, Maduro Junior, and
INEP’s own synopsis are systematically undercounted —
by up to ~700,000 enrollments in 2008 by our reckoning.
From 2009 onwards the CENSUP microdata already aggregate in-person and EAD in the same registry, and the problem resolves itself.
For this transition interval, educabr ships
reconstructed totals that combine the in-person
component of each source with INEP’s published EAD figures. These rows
carry is_derived = TRUE and are excluded from the default
get_enrollment() output; pass
include_derived = TRUE to inspect them:
recon <- get_enrollment(
level = "superior",
year = c(2000, 2008),
network = "total",
modality = "total",
indicator = "count",
include_derived = TRUE
)
recon[recon$is_derived, c("year", "source", "value")]
#> # A tibble: 23 × 3
#> year source value
#> <int> <chr> <dbl>
#> 1 2000 durham_2005+inep_sinopse_censup 2695927
#> 2 2000 inep_sinopse_censup+inep_sinopse_censup 2695927
#> 3 2000 kang_paese_felix_2021+inep_sinopse_censup 2695927
#> 4 2000 maduro_junior_2007+inep_sinopse_censup 2695927
#> 5 2001 durham_2005+inep_sinopse_censup 3045113
#> 6 2001 inep_sinopse_censup+inep_sinopse_censup 3036113
#> 7 2001 inep_sinopse_censup+inep_sinopse_censup 3036113
#> 8 2001 kang_paese_felix_2021+inep_sinopse_censup 3036113
#> 9 2001 maduro_junior_2007+inep_sinopse_censup 3036113
#> 10 2003 inep_sinopse_censup+inep_sinopse_censup 3936933
#> # ℹ 13 more rowsThe composite source key (e.g.
kang_paese_felix_2021+inep_sinopse_censup) makes it
explicit that this is a combination: the in-person
component comes from the first source and the EAD
component comes from the second. The source_note
records the exact composition for citation purposes.
Suggested citation convention: “Reconstructed total computed from {in-person source} (in-person) and INEP CENSUP Synopsis (EAD), following the methodology documented in
educabr(Mançano 2026).”
Case 4 — mean years of schooling by sex
schooling_sex <- get_schooling(
geo_level = "BR",
dimension = "sex"
)
tail(schooling_sex)
#> # A tibble: 6 × 12
#> year geo_level geo_code geo_name dim_race dim_sex age_group indicator value
#> <int> <chr> <chr> <chr> <chr> <chr> <chr> <chr> <dbl>
#> 1 2010 BR BR Brasil total male NA mean_years… 8.03
#> 2 2011 BR BR Brasil total male NA mean_years… 8.2
#> 3 2012 BR BR Brasil total male NA mean_years… 8.39
#> 4 2013 BR BR Brasil total male NA mean_years… 8.49
#> 5 2014 BR BR Brasil total male NA mean_years… 8.63
#> 6 2015 BR BR Brasil total male NA mean_years… 8.72
#> # ℹ 3 more variables: unit <chr>, source <chr>, source_note <chr>This series captures the historical reversal of the gender gap: in 1925 men averaged more years of schooling than women; over the 20th century that advantage inverts, and by 2015 women surpass men in average attainment.
Interactive dashboard
The package ships a Shiny dashboard for the three series:
educabr::run_dashboard()The dashboard replicates the multi-source tertiary comparison
interactively. Each tab has a “View R code” button that
emits the educabr + ggplot2 snippet needed to
reproduce the visualisation in your local R session — a direct bridge
between dashboard exploration and reproducible scripted analysis.
Sources and citation
Every series carries a canonical source key (source)
with the full reference catalogued in
inst/dict/vocabularies/sources.yaml. The seven primary
tertiary sources are:
- Kang, T. H., Paese, L. H. Z., & Felix, N. F. A. (2021). Late and unequal. Revista de Historia Económica 39(2), 191-218. doi:10.1017/S0212610921000112
- Kang, T. H., & Menetrier, I. (2024). Políticas elitistas e despesas públicas em educação. Estudos Econômicos 54(3). doi:10.1590/1980-53575434tkim
- Kang, T. H., Menetrier, I., & Comim, F. (2024). The side effects of a big push growth strategy. RHE 42(3), 387-414. doi:10.1017/S0212610924000120
- Walter, J. R., & Kang, T. H. (2024). A new dataset of average years of schooling in Brazil. Economic History of Developing Regions 39(3), 307-336. doi:10.1080/20780389.2024.2417268
- Durham, E. R. (2005). Educação superior, pública e privada. In Schwartzman (ed.), Os desafios da educação no Brasil, pp. 191-233.
- Maduro Junior, P. R. R. M. (2007). Taxas de matrícula e gastos em educação no Brasil [MSc dissertation, FGV/EPGE]. hdl:10438/110
- IBGE (2007). Estatísticas do Século XX. seculoxx.ibge.gov.br
- INEP/MEC. Sinopse Estatística da Educação Superior (1995-2008).
- INEP/MEC. Microdados do Censo da Educação Superior (2009-2024).
- INEP/MEC. CENSUP Power BI panel (2010-2024).
To cite the package itself:
Mançano, T. (2026). educabr: Harmonized Historical Series on Brazilian Education (version 0.0.0.9000). https://github.com/mancano-tales/educabr