OaxacaBlinder
This is an R implementation of OaxacaBlinder decomposition that comes
with
- correction for dummy encoding baseline with multiple factors
- more convenient output formatting
Please also check the stable oaxaca package on CRAN.
Installation
You can install the development version of OaxacaBlinder like so:
# install.packages("remotes")
remotes::install_github("SinanPolatoglu/OaxacaBlinder")Example
Using the chicago dataset from the oaxaca package
library(OaxacaBlinder)
dplyr::glimpse(chicago_long)
#> Rows: 712
#> Columns: 7
#> $ female <int> 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1, 1, 1, 0, 0, 0, 0, 0, …
#> $ male <dbl> 1, 0, 0, 1, 1, 0, 1, 1, 0, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, …
#> $ foreign_born <int> 1, 1, 1, 1, 0, 1, 1, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1, 1, …
#> $ age <int> 52, 46, 31, 35, 19, 50, 33, 43, 39, 22, 28, 31, 30, 20, 6…
#> $ education <chr> "high.school", "high.school", "high.school", "high.school…
#> $ real_wage <dbl> 8.500000, NA, 12.179999, 15.000001, 8.000000, NA, 10.0000…
#> $ ln_real_wage <dbl> 2.140066, NA, 2.499795, 2.708050, 2.079442, NA, 2.302585,…Twofold decomposition
- The formula interface is similar to the
oaxacapackage: it should be specified asdependent_var ~ x_var1 + x_var1 + ... + x_varK | group_var. - OaxacaBlinder is capable of correcting for the ommitted baseline bias
for multipe sets of factor variables with the
baseline_invariant=TRUE - The
pooledargument sets what reference $\beta_R$:neumark: The parameters of a model excluding the group variablejann: The parameters of a model including the group variable
twofold = OaxacaBlinderDecomp(
formula = real_wage ~ age + education | female,
data = chicago_long,
type = "twofold",
baseline_invariant = TRUE,
n_bootstraps = 100
)
summary(twofold)
#> Oaxaca Blinder Decomposition model
#> ----------------------------------
#> Type: twofold
#> Formula: real_wage ~ age + education | female
#> Data: chicago_long
#>
#> Descriptives
#> n %n mean(real_wage)
#> female==0 412 57.9% 17.52
#> female==1 300 42.1% 13.69
#>
#> Gap: 3.83
#> % Diff: 21.88%
#> coefficient % of gap 2.5% 97.5%
#> explained -0.53 -13.9% -1.306213 0.1681933
#> unexplained 4.37 113.9% 3.232576 5.5662780
#> unexplained_a 1.89 49.2% 1.402721 2.5663579
#> unexplained_b 2.48 64.6% 1.777999 3.1122698In addition, coefficients can be extracted vor the variable-level estimates
coef(twofold)
#> explained unexplained unexplained_a unexplained_b
#> (Intercept) 0.000000000 5.00836516 2.61888654 2.38947862
#> age 0.220418162 1.07822114 0.31640221 0.76181894
#> educationcollege -0.009283372 0.01142439 -0.01935324 0.03077763
#> educationhigh.school -0.107295348 -0.52124445 -0.35213923 -0.16910523
#> educationLTHS -0.560862567 -1.15597710 -0.68768187 -0.46829523
#> educationsome.college 0.184695669 -0.40644640 -0.15524704 -0.25119935
#> education.baseline -0.260197078 0.35251787 0.16750479 0.18501308Or with bootstrapped confidence intervals for variable level results
coef(twofold, ci=TRUE)
#> coef_type term coefficient 2.5% 97.5%
#> 1 explained (Intercept) 0.000000000 0.000000000 0.00000000
#> 2 explained age 0.220418162 -0.003446369 0.53534373
#> 3 explained education.baseline -0.260197078 -0.657812690 0.06859416
#> 4 explained educationcollege -0.009283372 -0.151655711 0.12650843
#> 5 explained educationhigh.school -0.107295348 -0.403230839 0.34095018
#> 6 explained educationLTHS -0.560862567 -1.190555209 -0.06628040
#> 7 explained educationsome.college 0.184695669 0.008586773 0.41180056
#> 8 unexplained (Intercept) 5.008365163 1.300805150 8.41320062
#> 9 unexplained age 1.078221143 -1.846739705 4.19573495
#> 10 unexplained education.baseline 0.352517873 -0.133223881 0.75388814
#> 11 unexplained educationcollege 0.011424394 -0.547359476 0.46466839
#> 12 unexplained educationhigh.school -0.521244454 -1.676171714 0.55148598
#> 13 unexplained educationLTHS -1.155977097 -1.919934983 -0.29237190
#> 14 unexplained educationsome.college -0.406446395 -1.243927239 0.48374409
#> 15 unexplained_a (Intercept) 2.618886538 0.556656099 4.71643644
#> 16 unexplained_a age 0.316402205 -1.059082761 1.90174874
#> 17 unexplained_a education.baseline 0.167504793 -0.044985535 0.36371764
#> 18 unexplained_a educationcollege -0.019353236 -0.259729988 0.22277605
#> 19 unexplained_a educationhigh.school -0.352139226 -1.054489794 0.18717027
#> 20 unexplained_a educationLTHS -0.687681871 -1.176104919 -0.22955356
#> 21 unexplained_a educationsome.college -0.155247041 -0.624632761 0.26228669
#> 22 unexplained_b (Intercept) 2.389478624 0.488537938 4.54667043
#> 23 unexplained_b age 0.761818938 -0.827969222 2.56148546
#> 24 unexplained_b education.baseline 0.185013080 -0.097748156 0.39335298
#> 25 unexplained_b educationcollege 0.030777630 -0.303367434 0.28215788
#> 26 unexplained_b educationhigh.school -0.169105228 -0.731720008 0.36530685
#> 27 unexplained_b educationLTHS -0.468295226 -0.800785060 -0.03573617
#> 28 unexplained_b educationsome.college -0.251199354 -0.735602991 0.24180657Threefold decomposition
threefold = OaxacaBlinderDecomp(
real_wage ~ age + education | female, chicago_long,
type = "twofold",
pooled = "jann",
baseline_invariant = TRUE
)
summary(threefold)
#> Oaxaca Blinder Decomposition model
#> ----------------------------------
#> Type: twofold
#> Formula: real_wage ~ age + education | female
#> Data: chicago_long
#>
#> Descriptives
#> n %n mean(real_wage)
#> female==0 412 57.9% 17.52
#> female==1 300 42.1% 13.69
#>
#> Gap: 3.83
#> % Diff: 21.88%
#> coefficient % of gap
#> explained -0.59 -15.4%
#> unexplained 4.43 115.4%
#> unexplained_a 0.00 0.0%
#> unexplained_b 4.43 115.4%
coef(threefold)
#> explained unexplained unexplained_a unexplained_b
#> (Intercept) 0.000000000 5.00836516 0.30167882 4.70668634
#> age 0.207760327 1.09087898 0.63724953 0.45362944
#> educationcollege -0.009142399 0.01128342 -0.01545726 0.02674068
#> educationhigh.school -0.110733093 -0.51780671 -0.29583600 -0.22197071
#> educationLTHS -0.584635143 -1.13220452 -0.58369823 -0.54850629
#> educationsome.college 0.171472514 -0.39322324 -0.19756114 -0.19566210
#> education.baseline -0.266963834 0.35928463 0.15362427 0.20566036