✔️ easy to understand
✔️ the difference of men and women average wages expressed as a fraction of average men wages
✔️ alternatively: the average women wage is `( 1 - GPG ) x average men wage`
❌ not an inequality measure in the Pigou-Dalton Principle sense
❌ ignores within-inequality among men and among women
❌ binary gender only
Although the Gender Pay Gap (GPG) is not an inequality measure in the usual sense10 That is, with respect to the Pigou-Dalton principle., it can still be a useful instrument to evaluate the effects of gender discrimination. Put simply, it expresses the relative difference between the average hourly earnings of men and women, presenting this difference as a percentage of the average of hourly earnings of men. Like some other functions described in this text, the GPG can also be calculated using wealth or assets in place of earnings.
In mathematical terms, this index can be described as,
\[ GPG = \frac{ \bar{y}_{male} - \bar{y}_{female} }{ \bar{y}_{male} } \]
As we can see from the formula, if there is no difference between the groups, \(GPG = 0\). Else, if \(GPG > 0\), it means that the average hourly income received by women are \(GPG\) percent smaller than men’s. For negative \(GPG\), it means that women’s hourly earnings are \(GPG\) percent larger than men’s. In other words, the larger the \(GPG\), larger is the shortfall of women’s hourly earnings.
We can also develop a more straightforward idea: for every $1 raise in men’s hourly earnings, women’s hourly earnings are expected to increase $\((1-GPG)\). For instance, assuming \(GPG = 0.8\), for every $1.00 increase in men’s average hourly earnings, women’s hourly earnings would increase only $0.20.
The details of the linearization of the GPG are discussed by Deville (1999Deville, Jean-Claude. 1999. “Variance Estimation for Complex Statistics and Estimators: Linearization and Residual Techniques.” Survey Methodology 25 (2): 193–203. http://www.statcan.gc.ca/pub/12-001-x/1999002/article/4882-eng.pdf.) and Osier (2009Osier, Guillaume. 2009. “Variance Estimation for Complex Indicators of Poverty and Inequality.” Journal of the European Survey Research Association 3 (3): 167–95. http://ojs.ub.uni-konstanz.de/srm/article/view/369.).
The R vardpoor
package (Breidaks, Liberts, and Ivanova 2016Breidaks, Juris, Martins Liberts, and Santa Ivanova. 2016. “Vardpoor: Estimation of Indicators on Social Exclusion and Poverty and Its Linearization, Variance Estimation.” Riga, Latvia: CSB.), created by researchers at the Central Statistical Bureau of Latvia, includes a GPG coefficient calculation using the ultimate cluster method. The example below reproduces those statistics.
Load and prepare the same data set:
# load the convey package
library(convey)
# load the survey library
library(survey)
# load the vardpoor library
library(vardpoor)
# load the laeken library
library(laeken)
# load the synthetic EU statistics on income & living conditions
data(eusilc)
# make all column names lowercase
names(eusilc) <- tolower(names(eusilc))
# coerce the gender variable to numeric 1 or 2
eusilc$one_two <- as.numeric(eusilc$rb090 == "female") + 1
# add a column with the row number
dati <- data.table::data.table(IDd = 1:nrow(eusilc), eusilc)
# calculate the gpg coefficient
# using the R vardpoor library
varpoord_gpg_calculation <-
varpoord(
# analysis variable
Y = "eqincome",
# weights variable
w_final = "rb050",
# row number variable
ID_level1 = "IDd",
# row number variable
ID_level2 = "IDd",
# strata variable
H = "db040",
N_h = NULL ,
# clustering variable
PSU = "rb030",
# data.table
dataset = dati,
# gpg coefficient function
type = "lingpg" ,
# gender variable
gender = "one_two",
# get linearized variable
outp_lin = TRUE
)
# construct a survey.design
# using our recommended setup
des_eusilc <-
svydesign(
ids = ~ rb030 ,
strata = ~ db040 ,
weights = ~ rb050 ,
data = eusilc
)
# immediately run the convey_prep function on it
des_eusilc <- convey_prep(des_eusilc)
# coefficients do match
varpoord_gpg_calculation$all_result$value
## [1] 7.645389
## eqincome
## -8.278297
# linearized variables do match
# vardpoor
lin_gpg_varpoord <- varpoord_gpg_calculation$lin_out$lin_gpg
# convey
lin_gpg_convey <-
attr(svygpg( ~ eqincome , des_eusilc, sex = ~ rb090), "lin")
# check equality
all.equal(lin_gpg_varpoord, 100 * lin_gpg_convey[, 1])
## [1] "Mean relative difference: 2.172419"
# variances do not match exactly
attr(svygpg( ~ eqincome , des_eusilc , sex = ~ rb090) , 'var') * 10000
## eqincome
## eqincome 0.8926311
## [1] 0.6482346
## [1] 0.8051301
## eqincome
## eqincome 0.9447916
the variance estimator and the linearized variable \(z\) are both defined in Linearization-Based Variance Estimation. The functions convey::svygpg
and vardpoor::lingpg
produce the same linearized variable \(z\).
However, the measures of uncertainty do not line up, because library(vardpoor)
defaults to an ultimate cluster method that can be replicated with an alternative setup of the survey.design
object.
# within each strata, sum up the weights
cluster_sums <-
aggregate(eusilc$rb050 , list(eusilc$db040) , sum)
# name the within-strata sums of weights the `cluster_sum`
names(cluster_sums) <- c("db040" , "cluster_sum")
# merge this column back onto the data.frame
eusilc <- merge(eusilc , cluster_sums)
# construct a survey.design
# with the fpc using the cluster sum
des_eusilc_ultimate_cluster <-
svydesign(
ids = ~ rb030 ,
strata = ~ db040 ,
weights = ~ rb050 ,
data = eusilc ,
fpc = ~ cluster_sum
)
# again, immediately run the convey_prep function on the `survey.design`
des_eusilc_ultimate_cluster <-
convey_prep(des_eusilc_ultimate_cluster)
# matches
attr(svygpg( ~ eqincome , des_eusilc_ultimate_cluster , sex = ~ rb090) ,
'var') * 10000
## eqincome
## eqincome 0.8910413
## [1] 0.6482346
## [1] 0.8051301
## eqincome
## eqincome 0.9439499
For additional usage examples of svygpg
, type ?convey::svygpg
in the R console.
This section displays example results using nationally-representative surveys from both the United States and Brazil. We present a variety of surveys, levels of analysis, and subpopulation breakouts to provide users with points of reference for the range of plausible values of the svygpg
function.
To understand the construction of each survey design object and respective variables of interest, please refer to section 1.4 for CPS-ASEC, section 1.5 for PNAD Contínua, and section 1.6 for SCF.
## gpg SE
## htotval -0.21961 0.013
## a_maritl htotval
## married - civilian spouse present married - civilian spouse present -0.03181163
## married - AF spouse present married - AF spouse present -0.32822461
## married - spouse absent married - spouse absent -0.56380658
## widowed widowed -0.14137283
## divorced divorced -0.10360249
## separated separated -0.54741878
## never married never married -0.17299575
## se.htotval
## married - civilian spouse present 0.01401497
## married - AF spouse present 0.17997337
## married - spouse absent 0.11379058
## widowed 0.05201811
## divorced 0.03514135
## separated 0.10734551
## never married 0.02834880
## gpg SE
## ftotval -0.19287 0.0141
## a_maritl ftotval
## married - civilian spouse present married - civilian spouse present -0.03138597
## married - AF spouse present married - AF spouse present -0.32822461
## married - spouse absent married - spouse absent -0.72848350
## widowed widowed -0.15551741
## divorced divorced -0.13440910
## separated separated -0.60357071
## never married never married -0.35042219
## se.ftotval
## married - civilian spouse present 0.01401296
## married - AF spouse present 0.17997337
## married - spouse absent 0.20717616
## widowed 0.06510014
## divorced 0.05955588
## separated 0.14911941
## never married 0.05862692
## gpg SE
## pearnval -0.26139 0.0142
## a_maritl
## married - civilian spouse present married - civilian spouse present
## married - AF spouse present married - AF spouse present
## married - spouse absent married - spouse absent
## widowed widowed
## divorced divorced
## separated separated
## never married never married
## pearnval se.pearnval
## married - civilian spouse present -0.332790903 0.01764578
## married - AF spouse present -0.237551797 0.18083461
## married - spouse absent -0.216038485 0.08486880
## widowed 0.009018563 0.09229872
## divorced -0.149661580 0.04110491
## separated -0.453153005 0.10300059
## never married -0.063720063 0.02173721
## gpg SE
## deflated_per_capita_income -0.044081 0.0057
svyby(
~ deflated_per_capita_income ,
~ age_categories ,
pnadc_design ,
svygpg ,
na.rm = TRUE ,
sex = ~ sex
)
## age_categories deflated_per_capita_income se.deflated_per_capita_income
## 1 1 -0.024306689 0.03230496
## 2 2 0.009552321 0.03181511
## 3 3 -0.005161343 0.02841903
## 4 4 0.013540072 0.02252232
## 5 5 -0.079247786 0.02238818
## 6 6 -0.122895187 0.02997464
## 7 7 -0.137640776 0.03014413
## 8 8 -0.133957055 0.02826733
## 9 9 -0.081122602 0.02683335
## 10 10 -0.104618639 0.02774374
## 11 11 -0.016451082 0.02122297
## 12 12 -0.050554855 0.02824782
## 13 13 -0.040804465 0.01106613
## gpg SE
## deflated_labor_income -0.26799 0.01
svyby(
~ deflated_labor_income ,
~ age_categories ,
pnadc_design ,
svygpg ,
na.rm = TRUE ,
sex = ~ sex
)
## age_categories deflated_labor_income se.deflated_labor_income
## 1 1 NaN NA
## 2 2 NaN NA
## 3 3 -0.1125043 0.18641583
## 4 4 -0.2054371 0.06174511
## 5 5 -0.1435653 0.02236847
## 6 6 -0.1349623 0.02726206
## 7 7 -0.2124605 0.03378860
## 8 8 -0.2241276 0.03192653
## 9 9 -0.2877783 0.03401662
## 10 10 -0.3935752 0.04738784
## 11 11 -0.2486947 0.03233777
## 12 12 -0.4023927 0.05638749
## 13 13 -0.5192294 0.05859540
## Multiple imputation results:
## m <- length(results)
## scf_MIcombine(with(scf_design, svygpg(~networth, sex = ~hhsex)))
## results se
## networth -2.714833 0.288726
## Multiple imputation results:
## m <- length(results)
## scf_MIcombine(with(scf_design, svyby(~networth, ~edcl, svygpg,
## sex = ~hhsex)))
## results se
## less than high school -1.575998 0.7734308
## high school or GED -1.210320 0.3505109
## some college -2.427703 0.4537274
## college degree -2.438002 0.4078957
## Multiple imputation results:
## m <- length(results)
## scf_MIcombine(with(scf_design, svygpg(~income, sex = ~hhsex)))
## results se
## income -2.00275 0.135422
## Multiple imputation results:
## m <- length(results)
## scf_MIcombine(with(scf_design, svyby(~income, ~edcl, svygpg,
## sex = ~hhsex)))
## results se
## less than high school -0.753955 0.1693429
## high school or GED -1.167613 0.1121024
## some college -1.248036 0.1148635
## college degree -2.178707 0.2109885