Assignment#7
Sex Difference in the Effect of Education and Marital Status in Determining Employment Status
Background & Purpose
Gender difference and educational attainment play important role in determining person’s status in work force. In addition, since there is relationship between person’s family life and work, person’s roles and responsibilities within family might also be an important factor in understanding their position in work force. Marital status is such a factor that bring changes in family relationship and changed in roles and responsibilities. Married persons have some specific responsibilities and commitment than never married or single person. In addition, male and female can also be distinguished in terms of their specific responsibilities depending on their marital status. So, the effect of education was also not might be the same for male and female in determining their employment status. So, it very important to understand how both factors influence on each other in determining the employment status of male and female. So, the purpose of present study was to investigate the differences in educational and marital status in determining the employment status of male and female. The following research questions were designed to answer in the present study. 1. Is the any difference in the effect of education on employment status of male and female. 2. Is the any difference in the effect of marital status on employment status of male and female.
Method
Sampling method:
The sample was drown from the population the American community Survey data 2016. The data was extracted from ipumps.org. The sample included persons between the age range of 30-50 years.
Defining variables
Dependent variable
The employment status was the dependent variable which indicated whether the respondent was a part of the labor force(working or seeking work) and, if so, whether the person was currently unemployed. This variable was categories as “employed”, “unemployed”, and “not in labor force” ###Independent variables Three independent variables were used in this study. The variable “SEX” reports whether the person was male or female. The variable “education” indicates respondents’ educational attainment, as measured by the highest year of school or degree completed. Education was further recoded in four categories:“Elementary/Junior Schooling (Nersary to Grade 9)”, “High school(Grade 9-12)”; “3 year or less College education”; and “Four or more year college education” The variable “marital status” gives each person’s current marital status. The marital status was categoried as “Married and Spouse present”; “Married and spouse absent”; “separated/divorced/widowed”. ### Other variables Two other categorical variables (race and number of child under 5 years ) were used for conducting multinomial logistic regression models to get the best fitting model for analysis.
Data analysis
Procedure of data analysis
Multinomial logistic regression analysis was conducted by using 3 different models to get the best fitted model. Zelig was used to simulate the best fitting model(model3). The data were further analyzed by extracting relevant contractual factors to determine the probabilities of employment status.
library(tidyverse)
employment<-read_csv("employment.csv")
head(employment)## # A tibble: 6 x 29
## YEAR DATANUM SERIAL HHWT GQ PERNUM PERWT FAMSIZE NCHILD NCHLT5
## <int> <int> <int> <int> <int> <int> <int> <int> <int> <int>
## 1 2016 1 53 11 1 1 12 6 2 0
## 2 2016 1 95 74 1 1 74 4 2 0
## 3 2016 1 137 77 1 1 77 3 1 1
## 4 2016 1 292 141 1 1 141 4 2 0
## 5 2016 1 292 141 1 2 142 4 2 0
## 6 2016 1 339 77 1 1 77 4 2 1
## # ... with 19 more variables: ELDCH <int>, YNGCH <int>, SEX <int>,
## # AGE <int>, MARST <int>, RACE <int>, RACED <int>, CITIZEN <int>,
## # RACASIAN <int>, RACBLK <int>, RACWHT <int>, EDUC <int>, EDUCD <int>,
## # EMPSTAT <int>, EMPSTATD <int>, UHRSWORK <int>, INCTOT <int>,
## # FTOTINC <int>, POVERTY <int>employmentstatus<-employment[c(8,10,13:16,18,22,24)]
head(employmentstatus)## # A tibble: 6 x 9
## FAMSIZE NCHLT5 SEX AGE MARST RACE CITIZEN EDUC EMPSTAT
## <int> <int> <int> <int> <int> <int> <int> <int> <int>
## 1 6 0 1 40 1 6 2 10 1
## 2 4 0 2 31 1 7 3 2 1
## 3 3 1 1 30 1 1 3 6 1
## 4 4 0 2 36 1 6 3 10 1
## 5 4 0 1 36 1 6 3 10 1
## 6 4 1 2 33 1 1 1 10 1employmentstatus<-employmentstatus %>%
mutate(race_group = sjmisc::rec(RACE, rec = "1=1; 2=2; 4:6=3; 3=4;7:9=4 "))%>%
mutate(education=sjmisc::rec(EDUC, rec = "1:2=1; 3:6=2; 7:9=3; 10:11=4"))%>%
mutate(marital_status=sjmisc::rec(MARST, rec = "1=1; 2=2; 3:5=3; 6=4"))%>%
mutate(child_under5=sjmisc::rec(NCHLT5, rec = "0=1; 1:6=2"))
head(employmentstatus)## # A tibble: 6 x 13
## FAMSIZE NCHLT5 SEX AGE MARST RACE CITIZEN EDUC EMPSTAT race_group
## <int> <int> <int> <int> <int> <int> <int> <int> <int> <dbl>
## 1 6 0 1 40 1 6 2 10 1 3.00
## 2 4 0 2 31 1 7 3 2 1 4.00
## 3 3 1 1 30 1 1 3 6 1 1.00
## 4 4 0 2 36 1 6 3 10 1 3.00
## 5 4 0 1 36 1 6 3 10 1 3.00
## 6 4 1 2 33 1 1 1 10 1 1.00
## # ... with 3 more variables: education <dbl>, marital_status <dbl>,
## # child_under5 <dbl>employmentstatus$SEX<-factor(employmentstatus$SEX, levels = c(1, 2),
labels = c("Male","Female"))
employmentstatus$race_group<-factor(employmentstatus$race_group, levels = c(1, 2, 3, 4),
labels = c("White", "Black", "Asian", "Other"))
employmentstatus$marital_status<-factor(employmentstatus$marital_status, levels = c(1, 2, 3, 4),
labels = c("Married_SP", "Married_SA","Sep/Div/Wid", "NM/single"))
employmentstatus$employed<-factor(employmentstatus$EMPSTAT, levels = c( 1, 2, 3),
labels = c("Employed", "Unemployed", "Not in Labor Force" ))
employmentstatus$education<-factor(employmentstatus$education, levels = c(1, 2, 3, 4),
labels = c("Elem/Juni school", "High School", "college 3y or less", "college 4y or more"))
employmentstatus$child_under5<-factor(employmentstatus$child_under5, levels = c(1, 2),
labels = c("no", "yes"))
head(employmentstatus)## # A tibble: 6 x 14
## FAMSIZE NCHLT5 SEX AGE MARST RACE CITIZEN EDUC EMPSTAT race_group
## <int> <int> <fct> <int> <int> <int> <int> <int> <int> <fct>
## 1 6 0 Male 40 1 6 2 10 1 Asian
## 2 4 0 Female 31 1 7 3 2 1 Other
## 3 3 1 Male 30 1 1 3 6 1 White
## 4 4 0 Female 36 1 6 3 10 1 Asian
## 5 4 0 Male 36 1 6 3 10 1 Asian
## 6 4 1 Female 33 1 1 1 10 1 White
## # ... with 4 more variables: education <fct>, marital_status <fct>,
## # child_under5 <fct>, employed <fct>employmentstatus<-employmentstatus[-c(1,2,5,6,8,9)]
head(employmentstatus)## # A tibble: 6 x 8
## SEX AGE CITIZEN race_group education marital_status child_under5
## <fct> <int> <int> <fct> <fct> <fct> <fct>
## 1 Male 40 2 Asian college 4y … Married_SP no
## 2 Female 31 3 Other Elem/Juni s… Married_SP no
## 3 Male 30 3 White High School Married_SP yes
## 4 Female 36 3 Asian college 4y … Married_SP no
## 5 Male 36 3 Asian college 4y … Married_SP no
## 6 Female 33 1 White college 4y … Married_SP yes
## # ... with 1 more variable: employed <fct>Descriptive analysis of data
pie(table(employmentstatus$employed))
The pie-chat suggested that the employed person were highly representative than persons in other two groups of employment categories.
pie(table(employmentstatus$SEX))
The pie-chat suggested that the female was slightly higher than male .
barplot(sort(table(employmentstatus$education)))
The persons with 4 years of more college education was highly representative followed by persons with high school education.
barplot(sort(table(employmentstatus$marital_status)))
Married persons whose spouses were present was highly representative followed by persons in never married/single marital status.
Data Analysis: Best fitting model
library(nnet)
model1<-multinom(employed~education+marital_status+SEX, employmentstatus)## # weights: 27 (16 variable)
## initial value 53482.643437
## iter 10 value 32112.202633
## iter 20 value 27539.162284
## iter 30 value 26252.285049
## final value 26251.860234
## convergedmodel2<-multinom(employed~education+marital_status+SEX+race_group, employmentstatus)## # weights: 36 (22 variable)
## initial value 53482.643437
## iter 10 value 33425.340333
## iter 20 value 31118.019637
## iter 30 value 26416.002711
## iter 40 value 26161.387833
## final value 26161.383515
## convergedmodel3<-multinom(employed~education+marital_status+SEX+race_group+child_under5, employmentstatus)## # weights: 39 (24 variable)
## initial value 53482.643437
## iter 10 value 37171.276902
## iter 20 value 33140.023507
## iter 30 value 27043.146929
## iter 40 value 26013.186367
## final value 26012.691347
## convergedlmtest::lrtest(model1, model2, model3)## Likelihood ratio test
##
## Model 1: employed ~ education + marital_status + SEX
## Model 2: employed ~ education + marital_status + SEX + race_group
## Model 3: employed ~ education + marital_status + SEX + race_group + child_under5
## #Df LogLik Df Chisq Pr(>Chisq)
## 1 16 -26252
## 2 22 -26161 6 180.95 < 2.2e-16 ***
## 3 24 -26013 2 297.38 < 2.2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1library(texreg)## Version: 1.36.23
## Date: 2017-03-03
## Author: Philip Leifeld (University of Glasgow)
##
## Please cite the JSS article in your publications -- see citation("texreg").##
## Attaching package: 'texreg'## The following object is masked from 'package:tidyr':
##
## extracthtmlreg(list(model1, model2, model3))| Unemployed | Not in Labor Force | Unemployed | Not in Labor Force | Unemployed | Not in Labor Force | ||
|---|---|---|---|---|---|---|---|
| (Intercept) | -3.40*** | -2.32*** | -3.39*** | -2.29*** | -3.36*** | -2.56*** | |
| (0.10) | (0.05) | (0.11) | (0.05) | (0.11) | (0.06) | ||
| educationHigh School | -0.07 | -0.35*** | -0.10 | -0.32*** | -0.10 | -0.31*** | |
| (0.11) | (0.05) | (0.11) | (0.05) | (0.11) | (0.05) | ||
| educationcollege 3y or less | -0.20 | -0.74*** | -0.24* | -0.68*** | -0.24* | -0.68*** | |
| (0.12) | (0.06) | (0.12) | (0.06) | (0.12) | (0.06) | ||
| educationcollege 4y or more | -0.56*** | -1.07*** | -0.53*** | -1.03*** | -0.53*** | -1.05*** | |
| (0.11) | (0.05) | (0.11) | (0.05) | (0.11) | (0.05) | ||
| marital_statusMarried_SA | 0.08 | -0.52*** | 0.01 | -0.44*** | 0.01 | -0.42*** | |
| (0.21) | (0.10) | (0.21) | (0.10) | (0.21) | (0.10) | ||
| marital_statusSep/Div/Wid | -0.16 | -1.29*** | -0.22* | -1.27*** | -0.22* | -1.19*** | |
| (0.11) | (0.06) | (0.11) | (0.06) | (0.11) | (0.06) | ||
| marital_statusNM/single | 0.13 | -0.68*** | 0.07 | -0.66*** | 0.07 | -0.65*** | |
| (0.09) | (0.05) | (0.09) | (0.05) | (0.09) | (0.05) | ||
| SEXFemale | 0.92*** | 2.73*** | 0.92*** | 2.73*** | 0.92*** | 2.78*** | |
| (0.06) | (0.04) | (0.06) | (0.04) | (0.06) | (0.04) | ||
| race_groupBlack | 0.41*** | -0.62*** | 0.42*** | -0.65*** | |||
| (0.09) | (0.06) | (0.09) | (0.06) | ||||
| race_groupAsian | -0.15* | -0.08** | -0.15* | -0.07* | |||
| (0.07) | (0.03) | (0.07) | (0.03) | ||||
| race_groupOther | 0.01 | -0.05 | 0.01 | -0.05 | |||
| (0.08) | (0.04) | (0.08) | (0.04) | ||||
| child_under5yes | -0.05 | 0.43*** | |||||
| (0.05) | (0.03) | ||||||
| AIC | 52535.72 | 52535.72 | 52366.77 | 52366.77 | 52073.38 | 52073.38 | |
| BIC | 52676.41 | 52676.41 | 52560.21 | 52560.21 | 52284.42 | 52284.42 | |
| Log Likelihood | -26251.86 | -26251.86 | -26161.38 | -26161.38 | -26012.69 | -26012.69 | |
| Deviance | 52503.72 | 52503.72 | 52322.77 | 52322.77 | 52025.38 | 52025.38 | |
| Num. obs. | 48682 | 48682 | 48682 | 48682 | 48682 | 48682 | |
| p < 0.001, p < 0.01, p < 0.05 | |||||||
library(Zelig)## Loading required package: survival##
## Attaching package: 'Zelig'## The following object is masked from 'package:purrr':
##
## reducelibrary(zeligverse)library(ZeligChoice)
z.out1 <- zelig(as.factor(employed) ~ education + marital_status + SEX+race_group+child_under5,
model = "mlogit", data = employmentstatus, cite = FALSE)
summary(z.out1)## Model:
##
## Call:
## z5$zelig(formula = as.factor(employed) ~ education + marital_status +
## SEX + race_group + child_under5, data = employmentstatus)
##
##
## Pearson residuals:
## Min 1Q Median 3Q Max
## log(mu[,1]/mu[,3]) -12.575 0.17900 0.24841 0.6657 1.409
## log(mu[,2]/mu[,3]) -9.015 -0.09003 -0.05924 -0.0521 7.720
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept):1 2.56488 0.05761 44.524 < 2e-16
## (Intercept):2 -0.79420 0.11942 -6.651 2.92e-11
## educationHigh School:1 0.31057 0.05142 6.040 1.54e-09
## educationHigh School:2 0.21201 0.11282 1.879 0.060224
## educationcollege 3y or less:1 0.68201 0.05670 12.028 < 2e-16
## educationcollege 3y or less:2 0.44598 0.12427 3.589 0.000332
## educationcollege 4y or more:1 1.05395 0.05244 20.097 < 2e-16
## educationcollege 4y or more:2 0.52582 0.11658 4.510 6.47e-06
## marital_statusMarried_SA:1 0.41829 0.10438 4.007 6.14e-05
## marital_statusMarried_SA:2 0.42479 0.22594 1.880 0.060096
## marital_statusSep/Div/Wid:1 1.18573 0.06145 19.296 < 2e-16
## marital_statusSep/Div/Wid:2 0.96142 0.11846 8.116 4.82e-16
## marital_statusNM/single:1 0.65150 0.04764 13.676 < 2e-16
## marital_statusNM/single:2 0.72359 0.09435 7.669 1.73e-14
## SEXFemale:1 -2.77762 0.03578 -77.632 < 2e-16
## SEXFemale:2 -1.86025 0.06504 -28.603 < 2e-16
## race_groupBlack:1 0.64904 0.05613 11.564 < 2e-16
## race_groupBlack:2 1.06572 0.09779 10.898 < 2e-16
## race_groupAsian:1 0.07176 0.02940 2.441 0.014649
## race_groupAsian:2 -0.07742 0.07084 -1.093 0.274463
## race_groupOther:1 0.05397 0.03841 1.405 0.160020
## race_groupOther:2 0.06605 0.08526 0.775 0.438523
## child_under5yes:1 -0.43074 0.02551 -16.887 < 2e-16
## child_under5yes:2 -0.47816 0.05735 -8.338 < 2e-16
##
## Number of linear predictors: 2
##
## Names of linear predictors: log(mu[,1]/mu[,3]), log(mu[,2]/mu[,3])
##
## Residual deviance: 52025.38 on 97340 degrees of freedom
##
## Log-likelihood: -26012.69 on 97340 degrees of freedom
##
## Number of iterations: 6
##
## No Hauck-Donner effect found in any of the estimates
##
## Reference group is level 3 of the response
## Next step: Use 'setx' methodunique(employmentstatus$marital_status)## [1] Married_SP Sep/Div/Wid NM/single Married_SA
## Levels: Married_SP Married_SA Sep/Div/Wid NM/singleResults
Effect of difference in sex on employment status
x.male <- setx(z.out1, SEX="Male")
x.female <- setx(z.out1, SEX="Female")
s.out.mlogit <- sim(z.out1, x = x.male, x1 = x.female)
summary(s.out.mlogit)##
## sim x :
## -----
## ev
## mean sd 50% 2.5%
## Pr(Y=Employed) 0.94269587 0.001925689 0.94273359 0.93879944
## Pr(Y=Unemployed) 0.01845865 0.001222438 0.01841208 0.01626251
## Pr(Y=Not in Labor Force) 0.03884548 0.001491918 0.03887854 0.03590328
## 97.5%
## Pr(Y=Employed) 0.94637649
## Pr(Y=Unemployed) 0.02103029
## Pr(Y=Not in Labor Force) 0.04186677
## pv
## 1 2 3
## [1,] 0.937 0.015 0.048
##
## sim x1 :
## -----
## ev
## mean sd 50% 2.5%
## Pr(Y=Employed) 0.58407218 0.006131741 0.58405688 0.57176056
## Pr(Y=Unemployed) 0.02865044 0.001814590 0.02848081 0.02529383
## Pr(Y=Not in Labor Force) 0.38727737 0.006194346 0.38724082 0.37550032
## 97.5%
## Pr(Y=Employed) 0.59631192
## Pr(Y=Unemployed) 0.03252182
## Pr(Y=Not in Labor Force) 0.39967276
## pv
## 1 2 3
## [1,] 0.572 0.029 0.399
## fd
## mean sd 50% 2.5%
## Pr(Y=Employed) -0.35862368 0.005526913 -0.35858700 -0.369612142
## Pr(Y=Unemployed) 0.01019179 0.001393426 0.01018838 0.007607592
## Pr(Y=Not in Labor Force) 0.34843189 0.005639040 0.34842451 0.337765519
## 97.5%
## Pr(Y=Employed) -0.34797000
## Pr(Y=Unemployed) 0.01311185
## Pr(Y=Not in Labor Force) 0.35970810The results suggested that female had 35.85% lower and 34.84% higher probability than male of being employed and not being in the labor force respectively.
Plot 1: Sex differences in employment status
plot(s.out.mlogit)
Differences in marital status in education of female
x.femaleE1M4 <- setx(z.out1, education="Elem/Juni school", marital_status="NM/single", SEX="Female")
x.femaleE1M1 <- setx(z.out1, education="Elem/Juni school", marital_status="Married_SP", SEX="Female")
s.out1.mlogit <- sim(z.out1, x =x.femaleE1M4, x1 = x.femaleE1M1)
summary(s.out1.mlogit)##
## sim x :
## -----
## ev
## mean sd 50% 2.5%
## Pr(Y=Employed) 0.48102741 0.014127539 0.48087646 0.45367896
## Pr(Y=Unemployed) 0.04320205 0.005220154 0.04285572 0.03391003
## Pr(Y=Not in Labor Force) 0.47577054 0.014755397 0.47596340 0.44725958
## 97.5%
## Pr(Y=Employed) 0.50856281
## Pr(Y=Unemployed) 0.05366776
## Pr(Y=Not in Labor Force) 0.50670077
## pv
## 1 2 3
## [1,] 0.462 0.045 0.493
##
## sim x1 :
## -----
## ev
## mean sd 50% 2.5%
## Pr(Y=Employed) 0.33530254 0.010952328 0.33520723 0.31356619
## Pr(Y=Unemployed) 0.02791997 0.003174303 0.02772919 0.02252289
## Pr(Y=Not in Labor Force) 0.63677750 0.011545865 0.63708243 0.61407352
## 97.5%
## Pr(Y=Employed) 0.35671836
## Pr(Y=Unemployed) 0.03502997
## Pr(Y=Not in Labor Force) 0.65999372
## pv
## 1 2 3
## [1,] 0.349 0.018 0.633
## fd
## mean sd 50% 2.5%
## Pr(Y=Employed) -0.14572487 0.01056698 -0.14594761 -0.1660144
## Pr(Y=Unemployed) -0.01528208 0.00364159 -0.01514114 -0.0229217
## Pr(Y=Not in Labor Force) 0.16100696 0.01126635 0.16118185 0.1384085
## 97.5%
## Pr(Y=Employed) -0.125413543
## Pr(Y=Unemployed) -0.008634071
## Pr(Y=Not in Labor Force) 0.182919904The above result suggested that among the women who had either elementary or Junior level of schooling, married women who were living with spouse had 14.53% lower probability of being employed than never married/single women. On the other hand, the probability of not being in the labor force (not seeking or working) of women with elementary or Junior level is 16.06% higher if they were married and living with spouse than those who were never married/single.
Plot 2: differences in marital status(married and never married/single) of female with elemetary or junior schooling in employment status
plot(s.out1.mlogit)
The comparison of E(Y|X) and E(Y|X1) suggested that differences in probabilities of not in the labor force was greater than differences in probabilities of being employed. Very negligible difference was found in case of unemployed category.
x.femaleE4M4 <- setx(z.out1, education="college 4y or more", marital_status="NM/single", SEX="Female")
x.femaleE4M1 <- setx(z.out1, education="college 4y or more", marital_status="Married_SP", SEX="Female")
s.out2.mlogit <- sim(z.out1, x = x.femaleE4M4 , x1 = x.femaleE4M1)
summary(s.out2.mlogit)##
## sim x :
## -----
## ev
## mean sd 50% 2.5%
## Pr(Y=Employed) 0.71453494 0.009695653 0.71460490 0.69517532
## Pr(Y=Unemployed) 0.03762938 0.003712143 0.03733844 0.03152053
## Pr(Y=Not in Labor Force) 0.24783568 0.009384230 0.24761727 0.22927701
## 97.5%
## Pr(Y=Employed) 0.73311312
## Pr(Y=Unemployed) 0.04598794
## Pr(Y=Not in Labor Force) 0.26658255
## pv
## 1 2 3
## [1,] 0.747 0.03 0.223
##
## sim x1 :
## -----
## ev
## mean sd 50% 2.5%
## Pr(Y=Employed) 0.58398298 0.006249878 0.58421739 0.57150772
## Pr(Y=Unemployed) 0.02854626 0.001779588 0.02842357 0.02536635
## Pr(Y=Not in Labor Force) 0.38747076 0.006258069 0.38737147 0.37595120
## 97.5%
## Pr(Y=Employed) 0.59525290
## Pr(Y=Unemployed) 0.03217408
## Pr(Y=Not in Labor Force) 0.39917316
## pv
## 1 2 3
## [1,] 0.554 0.031 0.415
## fd
## mean sd 50% 2.5%
## Pr(Y=Employed) -0.130551960 0.009202971 -0.130743641 -0.14771531
## Pr(Y=Unemployed) -0.009083124 0.003111718 -0.008930644 -0.01604346
## Pr(Y=Not in Labor Force) 0.139635084 0.009103151 0.139801853 0.12113515
## 97.5%
## Pr(Y=Employed) -0.112269744
## Pr(Y=Unemployed) -0.003506475
## Pr(Y=Not in Labor Force) 0.156574063The above results suggested that among women who had four or more years of college education, the probability of being employed was 13.07% lower if they were married and spouse present than those who were never married/single. On the other hand, the probability of not being in the labor force (not seeking or working) of women with four or more years of college education was 13.98% higher if they were married and living with spouse than those who were never married/single.
Plot 3: differences in marital status(married and never married/single) of female with college education in employment status
plot(s.out2.mlogit)
The comparison of E(Y|X) and E(Y|X1) suggested that differences in probabilities of not in the labor force was very low than differences in probabilities of being employed. Very negligible difference was found in case of unemployed category.
x.femaleE1M3 <- setx(z.out1, education="Elem/Juni school", marital_status="Sep/Div/Wid", SEX="Female")
x.femaleE1M1 <- setx(z.out1, education="Elem/Juni school", marital_status="Married_SP", SEX="Female")
s.out3.mlogit <- sim(z.out1, x = x.femaleE1M3, x1 = x.femaleE1M1)
summary(s.out3.mlogit)##
## sim x :
## -----
## ev
## mean sd 50% 2.5%
## Pr(Y=Employed) 0.60583744 0.017373000 0.60592518 0.57113104
## Pr(Y=Unemployed) 0.04054378 0.005598808 0.04029122 0.03057098
## Pr(Y=Not in Labor Force) 0.35361878 0.017713664 0.35325394 0.31936426
## 97.5%
## Pr(Y=Employed) 0.64007020
## Pr(Y=Unemployed) 0.05261168
## Pr(Y=Not in Labor Force) 0.38939890
## pv
## 1 2 3
## [1,] 0.579 0.038 0.383
##
## sim x1 :
## -----
## ev
## mean sd 50% 2.5%
## Pr(Y=Employed) 0.3347415 0.011708333 0.33446969 0.3123921
## Pr(Y=Unemployed) 0.0280678 0.003078458 0.02784579 0.0224908
## Pr(Y=Not in Labor Force) 0.6371907 0.012452162 0.63716789 0.6141448
## 97.5%
## Pr(Y=Employed) 0.35711633
## Pr(Y=Unemployed) 0.03411232
## Pr(Y=Not in Labor Force) 0.65969123
## pv
## 1 2 3
## [1,] 0.345 0.027 0.628
## fd
## mean sd 50% 2.5%
## Pr(Y=Employed) -0.27109589 0.013842043 -0.27137609 -0.29735882
## Pr(Y=Unemployed) -0.01247598 0.004257334 -0.01237198 -0.02152594
## Pr(Y=Not in Labor Force) 0.28357187 0.014210300 0.28384821 0.25628501
## 97.5%
## Pr(Y=Employed) -0.244202976
## Pr(Y=Unemployed) -0.005020649
## Pr(Y=Not in Labor Force) 0.311625700The above result suggested that among the women who had either elementary or Junior level of schooling, married women who were living with spouse had 27.19% lower probability of being employed than women if they were either separated or divorced or widow. On the other hand, the probability of not being in the labor force (not seeking or working) of women with elementary or Junior level is 28.43% higher if they were married and living with spouse than those who were either separated or divorced or widow.
Plot 4: differences in marital status(married and separated/divorced/widowd) of female with elemetary or junior schooling in employment status
plot(s.out3.mlogit)
The comparison of E(Y|X) and E(Y|X1) suggested that differences in probabilities of not in the labor force was very almost same as the differences in probabilities of being employed. Very negligible difference was found in case of unemployed category.
x.femaleE4M3 <- setx(z.out1, education="college 4y or more", marital_status="Sep/Div/Wid", SEX="Female")
x.femaleE4M1 <- setx(z.out1, education="college 4y or more", marital_status="Married_SP", SEX="Female")
s.out4.mlogit <- sim(z.out1, x = x.femaleE4M3 , x1 = x.femaleE4M1)
summary(s.out4.mlogit)##
## sim x :
## -----
## ev
## mean sd 50% 2.5%
## Pr(Y=Employed) 0.80494041 0.009303132 0.80504274 0.78646786
## Pr(Y=Unemployed) 0.03182722 0.003574789 0.03178179 0.02538992
## Pr(Y=Not in Labor Force) 0.16323237 0.008735633 0.16282071 0.14699194
## 97.5%
## Pr(Y=Employed) 0.8223049
## Pr(Y=Unemployed) 0.0390690
## Pr(Y=Not in Labor Force) 0.1805839
## pv
## 1 2 3
## [1,] 0.802 0.043 0.155
##
## sim x1 :
## -----
## ev
## mean sd 50% 2.5%
## Pr(Y=Employed) 0.58377947 0.006354717 0.5839282 0.57115943
## Pr(Y=Unemployed) 0.02863921 0.001817243 0.0286149 0.02522342
## Pr(Y=Not in Labor Force) 0.38758133 0.006547348 0.3875669 0.37515014
## 97.5%
## Pr(Y=Employed) 0.59575828
## Pr(Y=Unemployed) 0.03230429
## Pr(Y=Not in Labor Force) 0.40016616
## pv
## 1 2 3
## [1,] 0.587 0.031 0.382
## fd
## mean sd 50% 2.5%
## Pr(Y=Employed) -0.221160945 0.009335480 -0.221340024 -0.23904257
## Pr(Y=Unemployed) -0.003188011 0.003244016 -0.003045206 -0.01035143
## Pr(Y=Not in Labor Force) 0.224348955 0.008919250 0.224675924 0.20675017
## 97.5%
## Pr(Y=Employed) -0.202822978
## Pr(Y=Unemployed) 0.002766697
## Pr(Y=Not in Labor Force) 0.241710570The above results suggested that among women who had four or more years of college education, the probability of being employed was 22.13% lower if they were married and spouse present than those who were either separated or divorced or widow. On the other hand, the probability of not being in the labor force (not seeking or working) of women with four or more years of college education was 22.44% higher if they were married and living with spouse than those who were either separated or divorced or widow.
Plot 5: differences in marital status(married and separated/divorced/widowd) of female with college education in employment status
plot(s.out4.mlogit)
The comparison of E(Y|X) and E(Y|X1) suggested that differences in probabilities of not in the labor force was very almost same as the differences in probabilities of being employed. Very negligible difference was found in case of unemployed category.
Differences in marital status in education of male
x.maleE1M4 <- setx(z.out1, education="Elem/Juni school", marital_status="NM/single", SEX="Male")
x.maleE1M1 <- setx(z.out1, education="Elem/Juni school", marital_status="Married_SP", SEX="Male")
s.outM1.mlogit <- sim(z.out1, x =x.maleE1M4, x1 = x.maleE1M1)
summary(s.outM1.mlogit)##
## sim x :
## -----
## ev
## mean sd 50% 2.5%
## Pr(Y=Employed) 0.91118747 0.005504266 0.91137735 0.90045060
## Pr(Y=Unemployed) 0.03259281 0.004063916 0.03254916 0.02508347
## Pr(Y=Not in Labor Force) 0.05621972 0.003683208 0.05616047 0.04919837
## 97.5%
## Pr(Y=Employed) 0.92175378
## Pr(Y=Unemployed) 0.04108536
## Pr(Y=Not in Labor Force) 0.06394661
## pv
## 1 2 3
## [1,] 0.911 0.03 0.059
##
## sim x1 :
## -----
## ev
## mean sd 50% 2.5%
## Pr(Y=Employed) 0.86839394 0.006034778 0.86836250 0.85637340
## Pr(Y=Unemployed) 0.02886664 0.003125048 0.02880693 0.02292617
## Pr(Y=Not in Labor Force) 0.10273943 0.005228206 0.10259301 0.09252127
## 97.5%
## Pr(Y=Employed) 0.88005968
## Pr(Y=Unemployed) 0.03534632
## Pr(Y=Not in Labor Force) 0.11311895
## pv
## 1 2 3
## [1,] 0.873 0.026 0.101
## fd
## mean sd 50%
## Pr(Y=Employed) -0.042793527 0.004533864 -0.042914797
## Pr(Y=Unemployed) -0.003726177 0.002705254 -0.003596576
## Pr(Y=Not in Labor Force) 0.046519703 0.003593404 0.046484968
## 2.5% 97.5%
## Pr(Y=Employed) -0.051275086 -0.033588174
## Pr(Y=Unemployed) -0.009448349 0.001167031
## Pr(Y=Not in Labor Force) 0.039770652 0.053524675The above result suggested that among male who had elementary or junior levels of schooling, the differences in the probabilities of being employed of never married/single and married male was very low. In this case, married male who lived with spouse were had 4.30% lower probability of being employed and 4.68% higher probability of not in labor force than never married/single male.
Plot 6: differences in marital status(married and never married/single) of male with elemetary or junior schooling in employment status
plot(s.outM1.mlogit)
x.maleE4M4 <- setx(z.out1, education="college 4y or more", marital_status="NM/single", SEX="Male")
x.maleE4M1 <- setx(z.out1, education="college 4y or more", marital_status="Married_SP", SEX="Male")
s.outM2.mlogit <- sim(z.out1, x = x.maleE4M4 , x1 = x.maleE4M1)
summary(s.outM2.mlogit)##
## sim x :
## -----
## ev
## mean sd 50% 2.5%
## Pr(Y=Employed) 0.95900441 0.002608426 0.95899783 0.95382167
## Pr(Y=Unemployed) 0.02032102 0.002227196 0.02023814 0.01632052
## Pr(Y=Not in Labor Force) 0.02067457 0.001281576 0.02064091 0.01823333
## 97.5%
## Pr(Y=Employed) 0.96382542
## Pr(Y=Unemployed) 0.02472407
## Pr(Y=Not in Labor Force) 0.02330806
## pv
## 1 2 3
## [1,] 0.962 0.013 0.025
##
## sim x1 :
## -----
## ev
## mean sd 50% 2.5%
## Pr(Y=Employed) 0.94255386 0.001986730 0.94255300 0.93879333
## Pr(Y=Unemployed) 0.01850378 0.001288822 0.01849633 0.01616863
## Pr(Y=Not in Labor Force) 0.03894236 0.001470495 0.03891513 0.03607636
## 97.5%
## Pr(Y=Employed) 0.94633959
## Pr(Y=Unemployed) 0.02110000
## Pr(Y=Not in Labor Force) 0.04185827
## pv
## 1 2 3
## [1,] 0.946 0.017 0.037
## fd
## mean sd 50%
## Pr(Y=Employed) -0.016450543 0.002123804 -0.016452946
## Pr(Y=Unemployed) -0.001817248 0.001740820 -0.001770365
## Pr(Y=Not in Labor Force) 0.018267791 0.001161003 0.018272948
## 2.5% 97.5%
## Pr(Y=Employed) -0.020594871 -0.012221890
## Pr(Y=Unemployed) -0.005242092 0.001396072
## Pr(Y=Not in Labor Force) 0.016108658 0.020539114The above result suggested that among male who had four or more years of college education, the difference in the probabilities of being employed of never married/single and married male was very negligible. In this case, married male who lived with spouse were had only 1.64% lower probability of being employed and 1.82% higher probability of not in labor force than never married/single male.
Plot 7: differences in marital status(married and never married/single) of male with college education in employment status
plot(s.outM2.mlogit)
x.maleE1M3 <- setx(z.out1, education="Elem/Juni school", marital_status="Sep/Div/Wid", SEX="Male")
x.maleE1M1 <- setx(z.out1, education="Elem/Juni school", marital_status="Married_SP", SEX="Male")
s.outM3.mlogit <- sim(z.out1, x = x.maleE1M3, x1 = x.maleE1M1)
summary(s.outM3.mlogit)##
## sim x :
## -----
## ev
## mean sd 50% 2.5%
## Pr(Y=Employed) 0.94069238 0.004766666 0.94083584 0.93120052
## Pr(Y=Unemployed) 0.02520997 0.003741363 0.02507724 0.01876858
## Pr(Y=Not in Labor Force) 0.03409764 0.002678349 0.03402100 0.02912574
## 97.5%
## Pr(Y=Employed) 0.94931376
## Pr(Y=Unemployed) 0.03332048
## Pr(Y=Not in Labor Force) 0.03962731
## pv
## 1 2 3
## [1,] 0.946 0.022 0.032
##
## sim x1 :
## -----
## ev
## mean sd 50% 2.5%
## Pr(Y=Employed) 0.8680270 0.006140558 0.86820232 0.85547510
## Pr(Y=Unemployed) 0.0290117 0.003014382 0.02888683 0.02362204
## Pr(Y=Not in Labor Force) 0.1029613 0.005197647 0.10275826 0.09325544
## 97.5%
## Pr(Y=Employed) 0.87952800
## Pr(Y=Unemployed) 0.03529174
## Pr(Y=Not in Labor Force) 0.11402724
## pv
## 1 2 3
## [1,] 0.877 0.02 0.103
## fd
## mean sd 50%
## Pr(Y=Employed) -0.072665427 0.004787247 -0.072524105
## Pr(Y=Unemployed) 0.003801722 0.002720588 0.003991022
## Pr(Y=Not in Labor Force) 0.068863706 0.003975284 0.068757882
## 2.5% 97.5%
## Pr(Y=Employed) -0.082344532 -0.063755935
## Pr(Y=Unemployed) -0.001932373 0.008901141
## Pr(Y=Not in Labor Force) 0.061572000 0.076991786The above result suggested that among male who had elementary or junior levels of schooling, the difference in the probabilities of being employed among married male and separated/divorced/widow male was very low. In this case, married male who lived with spouse were had 7.28% lower probability of being employed and 6.88% higher probability of not in labor force than those male who were separated/divorced/widow.
Plot 8: differences in marital status(married and separated/divorced/widowd) of male with elemetary or junior schooling in employment status
plot(s.outM3.mlogit)
x.maleE4M3 <- setx(z.out1, education="college 4y or more", marital_status="Sep/Div/Wid", SEX="Male")
x.maleE4M1 <- setx(z.out1, education="college 4y or more", marital_status="Married_SP", SEX="Male")
s.outM4.mlogit <- sim(z.out1, x = x.maleE4M3 , x1 = x.maleE4M1)
summary(s.outM4.mlogit)##
## sim x :
## -----
## ev
## mean sd 50% 2.5%
## Pr(Y=Employed) 0.97241652 0.0022073117 0.97250998 0.96791449
## Pr(Y=Unemployed) 0.01529741 0.0019758494 0.01516683 0.01176963
## Pr(Y=Not in Labor Force) 0.01228607 0.0008752385 0.01223621 0.01066924
## 97.5%
## Pr(Y=Employed) 0.97640088
## Pr(Y=Unemployed) 0.01954691
## Pr(Y=Not in Labor Force) 0.01418874
## pv
## 1 2 3
## [1,] 0.97 0.016 0.014
##
## sim x1 :
## -----
## ev
## mean sd 50% 2.5%
## Pr(Y=Employed) 0.94258914 0.001972416 0.94264893 0.93864260
## Pr(Y=Unemployed) 0.01852916 0.001259833 0.01844178 0.01628062
## Pr(Y=Not in Labor Force) 0.03888170 0.001524039 0.03885375 0.03593292
## 97.5%
## Pr(Y=Employed) 0.94647516
## Pr(Y=Unemployed) 0.02111982
## Pr(Y=Not in Labor Force) 0.04197838
## pv
## 1 2 3
## [1,] 0.938 0.027 0.035
## fd
## mean sd 50%
## Pr(Y=Employed) -0.029827383 0.002120491 -0.029850076
## Pr(Y=Unemployed) 0.003231752 0.001636639 0.003347364
## Pr(Y=Not in Labor Force) 0.026595631 0.001264229 0.026572635
## 2.5% 97.5%
## Pr(Y=Employed) -0.0335486632 -0.02544137
## Pr(Y=Unemployed) -0.0001878854 0.00613361
## Pr(Y=Not in Labor Force) 0.0242211831 0.02924442The above result suggested that among male who had four or more years of college education, the difference in the probabilities of being employed of separated/divorced/widow male and married male was very negligible. In this case, married male who lived with spouse were had only 2.98% lower probability of being employed and 2.67% higher probability of not in labor force than separated/divorced/widow male.
Plot 9: differences in marital status(married and separated/divorced/widowd) of male with college education in employment status
plot(s.outM4.mlogit)
Discussion
The above results suggested than male had a higher probability of being employed than female, but the female had higher probability of not being in the labor force(not working/seeking work). It is also important to note that where as the probabilities of being employed was 95% for male, it was only 58% in case of female. On the other hand, probability of not being in labor force was less than 4% in case of male, but it was 38% in case of female. So, female were less likely to seek employment or less likely to be in the part of labor force.
Sex difference in effect of education
Effect of education was observed in case of female. The findings suggested female had a greater probability of being employed with the increase of education when their marital status remain unchanged. For example, in case of married female, the probabilities of being employed increased with the increasing level of education. Similar picture was found in case of other marital status(Separated/divorced/widow and Never married/single). It was also revealed that the probabilities of not in labor force decrease with the increase of education when marital status remain unchanged. For example, in case of married female, the probability of not being in labor force was decreased with the increasing level of education. Similar picture was found in case of other marital status(Separated/divorced/widow and Never married/single).
Effect of education was found to be greater in case of married male than never married and separated/divorced/widow male. For example, in case of married male in probabilities of being employed and not being in work force were 86.80% and 10.30% respectively in case of male with elementary or junior level of schooling and their probabilities of being employed and not being in work force were 94.26% and 3.89% respectively in case of college education.On the other hand, both in case of never married and separated/divorced/widow male, probabilities of being employed were increased from 91.10% to 95.90% and from 94.07% to 97.23% respectively with the increased of education from elementary/junior level to college.
So,in case of married male, increase in education lead approximately 7% higher probabilities of being employed and approximately 6% lower probabilities of not being in work force. So,in case of never married male and separated/divorced/widow male , increase in education lead approximately 4% and 3% higher probabilities of being employed. On the other hand, both in case of never married and separated/divorced/widow male, probabilities of being not being in work force were decreased from 5.61% to 2.07% and from 3.42% to 1.22% respectively with the increased of education from elementary/junior level to college. So,in case of never married male and separated/divorced/widow male , increase in education lead approximately 3% and 2% lower probabilities of not being in work force. So, greater effect of education on employment status was found in case of married male than male with other marital status(never married and separated/divorced/widow)
Sex difference in effect of marital status
Effect of marital status was also observed in case of female. The findings suggested that the probability of being employed was greater both for never married/single and separated/divorced/widow female than married female when their education level remain unchanged. For example, the probability of being employed for never married/single and separated/divorced/widow female was greater than married female when they had elementary or Junior level of schooling. Similar effect of marital status was found for college educated female. In also revealed that the probability of not being employed was greater in case of married female than their never married and separated/divorced/widow counterpart both in case of elementary/junior level of schooling and college education.
Effect of marital status was also found in case of male. In case of male with elementary/junior level of schooling, married male had lower probabilities of being employed than never married and separated/divorced/widow male. On the other hand, very low differences were found among the probabilities of being employed of married, never married, and separated/divorced/widow male with college education. So, although the greater effect of marital status on employment status was found in case of male with lower level of education, this effect was very negligible in case of male with college education. Similarly, the effect of marital status on not being in labor force was found in case of male with elementary or junior level of schooling and this effect was very negligible in case of male with college education.
Conclusion:
Employment status of both male and female was affected by their levels of education. In both cases, increase in education lead increase in the probabilities of being employed and deceased in probabilities of not being in work force. However, the effect of education on employment status was greater in case of married male than never married and separated/divorced/widow male. Married female had lower probability of being employed and higher probability of not in the work force than never married and separated/divorced/widow female at each level of education. However, effect of marital status of male on their employment status was little bit different depending on their level of education. Married male had lower probabilities of being employed and higher probabilities of not being in labor force than never married and separated/divorced/widows male with elementary/junior level of schooling.
Limitation
Since the findings depend on the analysis of a single year cross-sectional data, we may not generalize the results in different time frame. More rigorous picture might obtain by analyzing data from several years.