Final Economics 4A03

Gender Income Gap and Glass Ceiling

Economics 4A03

Matina Mallaei Koohi

0741750

1

Abstract

The gender income gap and glass ceiling has been broadly examined in the labor market. However,

since 1990 the pay differential in Canada have been stagnating, and no improvement in the sizable

wage gap has been observed. Identifying the different source of income differential is crucial for

understanding this trend. As a result, studying the causes of the gender income gap is significantly

relevant for policy makers. The Blinder-Oaxaca decomposition technique popularized by Blinder

(1973) and Oaxaca (1973) is commonly used to study the mean outcome differences between groups

usually based on sex or race. This paper focuses on the Oaxaca model to analyze the gender income

gap in Canada using NHS (National Household Survey) 2011 and to identify the glass ceiling that

stops women to earn the same wage as men receive in most industries by controlling for variables that

could result in glass ceiling.

2

Introduction

Labor market inequality by gender has attracted significant political and legislative attention in recent

years. Even though over the past 40 years there has been some improvement in narrowing the gender

income gap in Canada, yet according to Stat Canada female’s income on average is 72% of male’s

income. Even in Women-dominated industries such as health care, education and administration

women tend to be compensated poorly compared to men in the same industries. Discriminations at

work limit potential economics growth and increase poverty and have a negative impact on wealth

and well-being of families. In addition, there has been no real income progress in the majority of

working families in the past 30 years if it had not been for the contribution of women's income.

Therefore, decomposition of factors that affect gender income inequality not only leads to better

career advancement for women, but also improves the overall well-being of the Canadian economy.

Wage disparity has many types. Researchers have focused on wage income inequality and glass

ceiling recently. Pendakure and Pendakure (2006), Arulampalam et al. (2007), Albrecht et al. (2003),

have found evidence that female workers in many countries face glass ceiling. However all these

papers have their limitations to distinguish whether the poor access to jobs or reduced income within

a firm are the causes of the glass ceiling. Glass ceiling is a barrier that limits access to high-wage jobs.

Albrecht et al.'s (2003) research suggests that Swedish women face an economy-wide glass ceiling.

Also Arulampalam et al. (2007) indicate that women in some European countries face larger wage

differentials at the bottom quantile than at the top quantile. Pendakure and Pendakure (2006) also

observed that Canadian minorities are crowded in low-wage jobs that are referred to as sticky floor.

I'm looking for evidence to compare the gender income inequality within different industries in

Canada and investigate different independent variables that potentially explain the wage gap and shed

light on unexplained factors associated with gender income inequality that are related to glass ceiling

3

phenomenon. In addition, this paper looks at how much of the gender pay can be accredited to the

individual’s characteristics and the payout of those characteristics. To answer this question, Oaxaca

decomposition model has been used.

Hypotheses

There are three possible hypotheses that explain the Gender Income inequality: human capital, gender

industry segregation and sociological approach.

The most influential variable on a person's income is human capital. Human capital includes variables

such as education, work experience, full-time, part-time, tenure and age. Human capital variables

could partially answer the conundrum of gender income gap. Because management positions tend to

require certain human capital, the human capital variables can help to untangle glass-ceiling

phenomenon. Women due to their predominant dual role in the household and the labor market, tend

to have fewer years in the labor market. Drolet (2002) denotes that men and women differ

considerably in the continuity of their work experience. Women tend to combine paid work with labor

force withdrawals for family reasons. In addition, Baker and Drolet (2010) illustrate that standard

weekly hours for full- time women aged 25 to 54 is about 3 to 4 hours less than men. Furthermore, the

nature of the household responsibilities can lead to a depreciation of women's human capital that is

valued in the labor market and restrain them from attaining continuous labor market experience. As a

result, it could be economically logical for females to be reluctant to invest in female human capital

that is labor market oriented. However, females’ human capital may not deteriorate if the skills they

acquire in household activities such as time management and multitasking are valuable in the labor

market. Yet female's family responsibilities may reflect discrimination rather than choice. In

4

conclusion, the gender wage gap can be the result of the productivity differences, which could

indicate the rational economic choice, as well as discrimination, prior to entering the labor market.

The Second hypothesis is gender industry segregation. There are two types of occupational

segregation horizontal and vertical segregations. Horizontal segregation is the concentration of

women in particular fields such as education and health care sectors. Vertical segregation denotes to

the overrepresentation of women at certain levels of the hierarchy regardless of the industry.

According to Swanson (2005), females tend to be segregated by " female- type" jobs. The Abundance

of supply lowers their marginal productivity and hence their wage. As a result, even if women are

paid a wage equal to their marginal productivity, their wage will be lower than male wages that are

not lowered by an excess supply. Zeytinoglu and Cook (2008) use Canadian data from the 1999

Workplace and Employee Survey to show that female's who work part-time, temporary full-time or

temporary part-time tend to be promoted less often than male who were in the same three working

conditions. Therefore, barriers to promotion (glass ceiling) are more structural for women than for

men, which may negatively contribute to the gender income gap.

Thirdly, the sociological approach suggests that women’s attitudes of their own labor market worth also

could cause them to demand a lower wage due to conditioning in the male-dominated labor market,

females underestimate their own labor value. This approach to own labor worth leads to a lower wage for

female and employers would encourage these attitudes to lower their labor costs. Fortin (2008) presents

facts indicating that some of the explained male-female pay gaps may arise due to females valuing jobs

that emphasize people and family over money. In addition, Mueller and Plug (2006) study of personality

test on a sample of high school graduates in U.S. indicates that it is more rewarding to be aggressive and

disagreeable for males than it is for females. In the education area, there is small evidence that women

tend to perform better from the same sex model (Florian and Oreopoulos 2009). These hypotheses to

5

some extend shed light on the gender income gap and glass ceiling. However, some of the variation in

productivity cannot be observed but affects individual’s performance. Therefore controlling for all

variables is quite limited in this study. The focus of this paper would be to focus on some of the variables

that could explain these hypotheses.

Data Source

Data sources available for this paper is National Household Survey by stat Canada. NHS 2011 is a

comprehensive social, demographic and economic database about a sample of Canadian population.

The individuals file sample size represents 2.7% of the population, which gives 887,012 individuals.

The data enables the study of individuals in relation to their census families, economic families and

households by providing variables such as education, income, sex, age, number of children, and

number of persons in a household, marital status, occupation levels, labor industry sectors and

household type. I have created three tables to understand the gender income information available on

NHS 2011.

6

Table 1

Table 1 shows the segregation of male and female into different occupation type. Women stand at

only 38.3% of management profession. Female-dominated occupation groups are business and

finance and administrative occupation, Health occupation, social sciences education and government.

Art, culture, recreation and sport and sales and services represent a bit over 50% of female population.

Males are predominant in Manufacturing, primary industry, and trades sectors. The table presents

clear segregation of male and female jobs from the NHS 2011 survey.

Table 2

Table 2 indicates the raw income of married full-time workers with bachelor degree in different

industries. The lowest female to male income ratio is represented in agricultural industry that female's

7

earning is only 42% of males earning and the highest rate is expressed in arts, entertainment and

recreation where female's earning is 148% of male's earning.

Total 50609.34 14000 45000 76000 1096295 Male 63125.38 21000 54000 88000 1096295Female 33278.11 10000 36000 62000 351261 SEX sd p50 p75 p90 range

by categories of: SEX (Sex)Summary for variables: EMPIN

. tabstat EMPIN, by (SEX) stat ( sd p50 p75 p90 range)

Table 3

Table 3 points out that female’s income at 90th percentile is $62000 compared to male’s income that is

$88000. Therefore, the ratio of female to male income is 70% at top the 90th percentile. Interestingly

at 50th percentile the ratio of female to male income is 47% and at 75th percentile this ratio is 67%.

This result indicates that women at 50th percentile suffer the most from the gender income gap. Also

moving from 75th percentile to 90th percentile on average increases male’s income by 62% from

$54000 to $88000 whereas management position on average increases female’s income by 58% from

$36000 to $62000.

Regression Model

The regression model used in this paper is a human capital earning function previously developed by

Blinder (1973) and Oaxaca (1973), and it's referred to as Oaxaca-Blinder decomposition. The model

provides earning function for male and female denoted as

8

ln YM=BMXM

And

ln YF=BFXF

Where ln YM and ln YF is log of income for male and female, XM and XF are men's and women's

productive characteristics, and BM and BF are the pays associated with the productive characteristics.

Taking logs of the male- female wage ratio, we have

ln (YM/YF) = ln YM - ln YF = BM XM - BF XF

By adding and subtracting the term BFXF we get

ln YM - ln YF = BMXM - BFXF-BMXF +BMXF Rearranging,

We get

ln YM - ln YF = BM(XM - XF) + ( BM - BF) XF

The first part of the model BM (XM - XF) captures the difference in earning that can be explained by

the difference in characteristics of male and female. The first part is the explained part of the model.

The second component of the model (BM - BF) XF arises from the difference pay with the same

characteristics for both men and women. This part is the unexplained or discrimination part of the

model.

9

Results of regression analysis

The regression equation used in this model uses the human capital characteristics (Bachelor degree,

number of children between 2-5 years old, marital status, age, management position and industry).

Age could be a good approximate of experience has been used in this model; also number of hours

worked as been factored in by full-time work hours for both men and women. Nevertheless, tenure

and experience would have been better variable to control however NHS 2011 fails to provide this

information. Schooling has been kept out of the model since the education decision of men and

women could be affected by the discrimination in the pre-labor market. Therefore considering the pre

and post labor market discrimination is one of the limitations of this paper and further research

regarding pre-market characteristics and their impact on income is. The variables used in this paper

are Inincome as log of earning, AGEGRP9-AGEGRP16 controls for age group 25-65. I have also

controlled for the hours worked by choosing the variable FPTWK1 that indicates full-time work. To

control for household responsibilities I have used Marital Status has been taken into account

MARSTH2 represents an individual who is married, and MARTH5 represents an individual who is.

However, the best variable to control for household responsibilities is the number of unpaid

household work. Number of unpaid hours was not provided on NHS 2011 data. In addition, I have

controlled for individuals who have children between the ages 2-5 years old using variable PKID2_52

to measure the effects of small children on men and women earning since women’s income goes

down during maternity leave. Maternity leave could bias the estimation for women who have just

gone back to work after maternity. As a result children between ages, 2-5 seems to be a good variable

10

to tackle this matter since by the time children are 2, or older most women tend to go back in the job

market full time. Furthermore, to measure productivity I have controlled for all individuals with

bachelor degree denoted as HDGREE9 to have a consistent comparison between two groups with the

same education level. The limitation of using just education level is ignoring the impact studying

different subjects. In order to capture the wage gap in the management occupation, variable NOCS1

has been used to control for Managerial position. The Industries have been represented as NAICS1-

NAIC18 (Agricultural NAICS1, mining NAICS2, utilities NAICS3, construction NAICS4,

manufacturing NAICS5, wholesale trade NAICS6, retail trade NAICS7, transportation NAICS8,

Information NAICS9, finance NAICS10, real estate NAICS11, professional NAICS12, administrative

support NAICS13, educational services NAICS14, health care and social assistance NAICS15, arts

and entertainment NAICS16, accommodation and food services NAICS17, other services and public

administration NAICS18). Variables NIACS1-NAICS18 measure the segregation of male and female

in different industries and the ramifications of the segregation on the mean log of income.

11

Table 4 1

12

Table 4 2

Table 4.1 and 4.2 represent the results of Oaxaca decomposition. All variables in the explained

portion of the model (Table 4.1) are statistically significant except variables PK2_52 (families with

children aged 2-5) and NAICS16 (Art and entertainment industry). Similarly, all variables in

unexplained portion of the model (Table 4.2) are statistically significant except AGEGRP12-16 and

MARSTH5. The result indicates that mean log of females earning compared to males earning is 37%

less. In other words, the average female to male earning ratio is 73% controlling for all variables. The

coefficient prediction on mean log of income for female is 9.92 as compared to coefficient prediction

on mean log of income for male at 10.30. The explained part of the model shows that the difference in

13

female and male human capital represented by the variables in this model can justify 13% of the 37%

difference in income between men and women. The most statistically significant factor in the 13% is

full-time work. The result could support our hypothesis that women tend to work fewer hours than

men due to the number of unpaid work or other family responsibilities. Also, the definition of full-

time work hours is quite vague. Women could be 30 hours a week and considered full time whereas

men could be working 50 hours a week and considered full time. The unclear definition of full-time

work indicates that log of hourly wage could be a better measure of hours worked and better

reflection on mean log of earning for male and female. The unexplained or discrimination is 24% of

37% difference in the mean log of female to male income ratio. The most statistically significant

factor is industry representing (NAICS1-NAICS9), the male-dominated industries contributed the

most to the 24% discrimination difference. The variable of children between 2-5 and married are

statistically significant in unexplained portion of the model. Women who are married and have

children between 2-5 years tend to earn less than men with the same characteristics. As the result

confirms most of the discrimination comes from the industries rather than the management position.

The finding could be the result of management position requiring other skill sets that can not be in

this model such as networking, mentorship, firm’s size and sample’s heterogeneity.

14

Table 5

15

Table 5 showed the decomposition of Oaxaca model using noisily option to get the details for each

group and the difference between the mean log of income for each group.

As expected table 5 indicates that education tend to have a higher impact on log of income for

females than males. Also between ages categories of 35-45 male’s coefficient is more significant than

female’s. However, as women age and gain more experience the impact of age on their mean logs of

earning tend to get very similar to their male counterparts. The full-time work has a higher impact on

male's mean log of income than females. The management position variable has stronger effect on

female's log of earning than male's. This could be the result of female's income being lower than

male's; as a result the management position increases female’s income relative to their non-

managerial position more than male’s income relative to male’s non-managerial position. The

industry coefficients for female have been compared with an omitted variable NAICS2 (mining

industry). The biggest negative coefficients and statistically significant are public admin industry,

retail trade industry and accommodation and food industry. Female workers fully crowd all these

industries. The outcome exhibits the impact of an industry on mean log of earning for women.

Women in female-dominant industries earn a lot less compared to females in male-dominant

industries.

16

Table 6

17

Table 6 shows the coefficients of variables on mean log of income for men. All coefficients are

statistically significant expect variable MARSTH5. The omitted variable in table 6 is NAICS11 (real

estate industries). The mean log of income for mining and utilities industries are 0.70 and 0.63 with

respect to the omitted variable and are statistically significant. Men in male's mean log of income is

higher in male-dominated industries.

Table 7

The first part of Table 7 would display the mean increase in women’s income if they had the same

characteristics as men. The -0.146 indicates that differences in endowments (explained) account for

about third of the income gap of -0.37. The second term quantifies the change in women’s wages

when applying the men’s coefficients to the women’s characteristics. The third part is the interaction

term that measures the simultaneous effect of differences in endowments and coefficients.

18

Relevant Policies

Some of policies associated with gender income inequality are equal pay legislation, equal

employment opportunity legislation and policies to facilitate female employment.

Equal pay legislation is quite limited because it only deals with one type of job within an

establishment, and this is only one aspect of gender wage discrimination. . Equal pay according to

Orazem and Matilla (1998) indicate that, surprisingly, pay equity could increase the proportion of

women in both female- dominated and male- dominated jobs. Because the elasticity of labor supply is

greater for females than for males in female- dominated jobs, and the opposite in male- dominated

jobs. So when real wages increase in female- dominated jobs, women enter those jobs in

proportionately larger numbers than men, making them even more female- dominated. When real

wages decrease in male- dominated jobs, males disproportionately leave, making them less male-

dominated. However equal pay legislation is on individual complaint, and the definition of the same

task and responsibility can be unclear at times.

Equal employment opportunity legislation is designed to prevent discrimination in recruiting, hiring,

promotion and dismissals. This is a crucial policy for female to get into positions especially at higher

managerial position but again it is very cumbersome and time consuming procedure and as empirical

studies show it fails to get rid of the glass ceiling phenomena. Nonetheless, it is a necessary policy for

female looking for jobs because it gives them equal opportunity to be for a position.

The facilitating policy uses different options have female in labor force. Some of these options

include availability of daycare, flexible working hours, part- time jobs, and childbirth leaves. These

facilities apply to men and women to maximize the opportunity to allocate the household work

between men and women. However, This policy lacks from extended childcare leave or subsidized

19

day care. Yet all these systems increase the price of discrimination to those who discriminate. The

price to the discriminator is the legal cost, and raising the price will reduce the quantity demanded. As

a result, while all these policies have some limitations but they are essential to reduce gender

discriminations in labor markets.

Conclusion

In conclusion, I have analyzed the gender income gaps using NHS 2011 data under the Oaxaca

decomposition framework. The main result of this paper is that female-male income ratio on average

is at 73% controlling for education, age, manager occupation, industries, marital status and children.

Most of the unexplained discrimination comes from the male-dominated industry and women who are

married and have children tend to earn lower wages due to discrimination. While gender gap is

decreasing from 50th to 90th percentile, it is still vastly due to discrimination that is unexplained.

Also, there is not a significantly negative relationship between female earnings and female occupation

segregation. The finding may be due to the unionized female dominated industries such as education

and health care. However, occupations and specially management positions in different sectors and

industries have different definitions for example a small retail manager and CEO of a company could

have gone through the same education but definitely have different skill sets and different job

descriptions. The different definition of management position in each industry emphasizes the

limitation of this paper and variables in the model. Nonetheless, women tend to be segregated in low-

wage occupations. The segregation in low-wage occupations highlights the limitations of the equal

pay policies because comparisons are made only within the same establishment and same industry.

My reasoning behind these results is as follows. Due to the historical female segregation in certain

industries, employers may use statistical discrimination to lower their wages to increase profit. Also,

as women get segregated into individual industries oversupply causes the wages to go down for

20

females. Also, firms know that women who are married tend to have career interruptions and use this

excuse to give them lower wage and reduce their cost. Phipps, Burton, and Lethbridge’s (2001)

finding suggests that child related interruption negatively affect female earnings as both human

capital depreciates and time spent in household increases could support the idea of the discrimination

against married mothers. As the result shows most of the unexplained gap is in the male-dominated

industries that could be explained by the socioeconomic hypothesis of wage gap and the cultural lag

that women are incapable of doing the difficult tasks in those industries. Gender income gap need

further investigation, and future studies can be considered to examine the possible causes of glass

ceilings. Reasonable assumption may convey to both labor market demand and supply side factors. A

possible reason on the demand side could be that wage setter firms may prefer men to women. For

example, companies are willing to pay more to get one of their own types in this case type would be

sex and not ethnicity. In regards to supply, women might be prepared to accept relatively lower

salaries than men. Accepting a lower wage is due to a lack of salary negotiations. Therefore, women

in male-dominated industry could face lack of information about what men are being paid. Lack of

information is likely to occur at the top-level position where there are relatively fewer women.

Policymakers should actively work to raise awareness about the causes of gender inequality and the

ramifications of these types of discrimination. Although there are women, who have broken the glass

ceiling, but not many of them realize the difficulty that the rest of females tend to face to break

through the glass and to get to a higher positions. Therefore, women at the top are in excellent

positions to create new policies and policymakers need to equip them, as well as male leaders, with an

understanding of how firms create barriers for workers and to find solutions that close the gender

income gap. Training women to acquire skill sets that would prepare them to take management

responsibilities are crucial to break this pattern. More women mentors are necessary to guide women

21

in their career path. All the public policies associated with gender income gap would be inefficient

without proper implementation. As a result firms, policy makers and women need to work together to

close the gender income gap.

22

Appendix

Variable Description Source

lnincome Log of income NHS 2011

PKID2_52 Household with

children between 2-5

NHS 2011

HDGREE9 Bachelor Degree NHS 2011

NOCS1 Management

Occupation

NHS 2011

NAICS1 Agricultural NHS 2011

NAICS2 Mining NHS 2011

NAICS3 Utilities NHS 2011

NAICS4 Construction NHS 2011

NAICS5 Manufacturing NHS 2011

NAICS6 Wholesale trade NHS 2011

NAICS7 Retail trade NHS 2011

NAICS8 Transportation NHS 2011

NAICS9 Information NHS 2011

NAICS10 Finance NHS 2011

NAICS11 Real estate NHS 2011

NAICS12 Professional NHS 2011

NAICS13 Admin support NHS 2011

NAICS14 Educational services NHS 2011

NAICS15 Health care and NHS 2011

23

social assistance

NAICS16 Arts and

Entertainment

NHS 2011

NAICS17 Accommodation and

food services

NHS 2011

NAICS18 Public Admin NHS 2011

AGEGRP9 Age group 25-29 NHS 2011








References

24

Albrecht, J., & Bjorklund, A., & Vroman, S. (2003). Is there a glass ceiling in Sweden?. Journal of Labor Economics, 21(1), 145-177.

Arulampalam, W., & Booth, A., & Bryan, M. (2007). Is there a glass ceiling over Europe? Exploring the gender pay gap across the wage distribution. Industrial and Labor Relations Review, 60(2), 163-186.

Baker, M., & Drolet, M. (2010). A New View of the Male/Female Pay Gap. Canadian Public Policy. 36(4), 429-464.

Barnet-Verzat, C., & Wolff, F. (2008). Gender wage gap and the glass ceiling effect: A firm-level investigation. International Journal of Manpower, 29(5-6), 486-502.

Blinder, A. (1973). Wage discrimination: Reduced form and structural estimations. Journal of Human Resource, 8, 436-55.

Drolet, M. (2002). New Evidence on Gender Pay Differentials: Does Measurement Matter?. Canadian Public Policy. 28(1), 1-16.

Fortin, N. (2008). The Gender wage gap among young adults in United States: The importance of money versus people. Journal of Human Resources, 43(4).

Hoffmann, F., & Oreopoulos, P. (2009). A professor like me: The influence of instructor gender on college achievement. Journal of Human Resources, 44(2).

Le, A. T., & Miller, P. W. (2010). Glass ceiling and double disadvantage effects: Women in the US labor market. Applied Economics, 42(4-6), 603-613.

Machado, J. A., & Mata, J. (2005). Counterfactual decomposition of changes in wage distributions using quintile regression, Journal of Applied Econometrics, 20, 445–65.

Mueller, G., & Plug, E. (2006). Estimating the effect of personality on male and female earnings. Industrial and Labor Relations Review, 60(1), 3-22.

Oaxaca, R. (1973). Male-female wage differential in urban labor markets. International Economic Review, 14(3), 693-709.

Orazem, P., & Mattila, J.P. (1998). Male-female supply to state government jobs and comparable worth. Journal Labor of Economics, 16(1), 95-121.

Pendakur, K., & Pendakur, R. (2007). Minority earnings disparity across distribution. Canadian Publish Policy, 33(1), 41-62.

Phipps, S., & Burton, P., & Lethbridge, L. (2001). In and out of the labor market: a long-term income consequence of child-related interruptions and women’s paid work. Canadian Journal of Economics, 34(2), 411-429.

Swanson, A. (2005). Occupational sex segregation and economic development. Journal of Business and Economics Research, 3(8).

25

Zeytinoglu, U. I., & Cook, G.B. (2008). Non-standard employment and promotions: A within gender analysis. Journal of Industrial Relations, 50, 319-337.

26

Final Economics 4A03

Documents