Proofread and EDIT Research paper

The Gender Pay Gap

Antoinette Dent

Argosy University/FP6030

August 16, 2016

Dr. Knight

Introduction (including the statement of the problem)

The gender pay gap mirrors the on process discrimination and inequalities in the labor market which in practice, majorly have a great impact on women. Its causative agents are complex and interrelated. On the occasion of European Equal pay day 2015, the European commission draws attention to the gender pay gap, and its underlying aggravators. To cement the European equal day, the commission released an animated info-graphic explaining some of the reasons behind the gender pay gap and information package. In addition to that, an analysis of the public consultation “Equality between women and men in the EU” was published. On the occasion of the 2014 European Equal pay day, DG Justice published a brochure ‘Tackling the gender pay gap in the European Union’ incorporating current statistics, European actions, and examples of national good proceeds.

The consistency of gender pay has consistently existed of late unlike the past years. The facts here are, the effect cannot be eliminated if concerns are not raised on the causes of the gap and be attended to traced and attended to. This study aims at pointing out the relationships between various individual aspects such as employment level and correlating them with the person’s wage then seeks to balance the wage.

Purpose of the study

After a half a century of stability in the earnings of women relative to men, there has been a substantial increase in women’s relative earnings since the late 1970s. One of the things that make this development especially dramatic and significant is that the recent changes contrast markedly with the relative stability of earlier years. These post-1980 earnings changes are also interesting because, when you compare women to their male counterparts, gains have been prevalent across a wide spectrum. For example, at first, much of the female gains were centered on younger women, but now, while the gains may be a bit larger for younger women, women of all ages have narrowed the pay gap with men. The same broad progress is visible when we look at the trends in the gender pay gap by education. Less educated women have narrowed the pay gap with less-educated men and highly educated women have narrowed the pay gap with highly educated men. There is need therefore of checking where the wage imbalance results from so as devise methods of curbing it.

Research question and hypotheses

We all have probably heard that men receive higher payments relative to women over their lifetime. But what does this typically imply? Is it true that women are paid less by virtue that they go for jobs that are less paying? Is it because women do part working than men do? Or rather, is it due to the reason that women are entitled to huge responsibilities of care?

The earnings gains of women are particularly remarkable because they have occurred during a period when overall wage inequality was rising. That is, the difference in pay between workers with high wages and workers with low wages has widened considerably over the past 25 years or so. And yet, women, a low paid group, have nonetheless been able to narrow the pay gap with a relatively higher paid group, men.

Theoretical framework

The foregoing supports our initial observation that there has been important, significant progress for women. On the other hand, however, there is still a gender pay gap. Women continue to earn considerably less than men on average. It is also true that convergence slowed noticeably in the 1990s after women had especially gained relative to men in the 1980s. Although there were some larger gains for women in the early 2000s, the long-run significance of this recent experience is unclear. With the evidence suggesting that convergence has slowed in recent years, the possibility arises that the narrowing of the gender pay gap will not continue into the future. Moreover, there is evidence that although discrimination against women in the labor market has declined, some discrimination does still continue to exist.

The trends in the gender pay gap in the United States form a somewhat mixed picture. On the one hand, after a half a century of stability in the earnings of women relative to men, there has been a substantial increase in women’s relative earnings since the late 1970s. One of the things that make this development especially dramatic and significant is that the recent changes contrast markedly with the relative stability of earlier years. On the other hand, there is still a gender pay gap. Women continue to earn considerably less than men on average, and the convergence that began in the late 1970s slowed noticeably in the 1990s.

LITERATURE REVIEW

Introduction

The American Association of University Women (AAUW) advances equity for women and girls via advocacy, education and research. The organization has nationwide network of 150,000 members, 1,500 branches and 500 college and university partners. It is headquartered in Washington D.C. According to the economic justice report, in 2014, women working full time in the U.S typically were paid just 79% of what men were paid, a gap difference of 21%. The same report point out that the gap has narrowed since 1970s courtesy largely of progress of women in education and participation in workforce and the wages of men advancing at a relatively low rate. However, progress has stagnated in the recent years and the pay gap does not seem to vanish by itself. Is this slowdown just a blip in an overall trend, or has the pay gap converged as far as it can? We look at this issue in depth and make some predictions for the future.

The basic truth of AAUW regarding gender pay gap succinctly addresses these issues by going beyond the widely reported 79% statistic. The report explains the pay gap in the U.S; how it impacts women across all ages, race and education level; and what can be done to curb it.

The pay gap affects women from all backgrounds, at all ages, and of all levels of all educational achievement, although earnings and the gap vary depending on the personal situation of a woman. Amongst full-time workers in 2014, Hispanic American, American Indian, and Native Hawaiian women had lower median annual earnings relative to non-Hispanic white and Asian American women. But within ethnic/racial groups, African America, Hispanic, American Indian, and Native Hawaiian women encountered a smaller gender pay gap relative to men in the same group than did non-Hispanic white and Asian American women.

Using a single benchmark provides a more informative image. This is due to the fact that those non-Hispanic white men are the largest demographic group in the labor force; they are often used for that agenda. Compared to salary information for white male workers, Asian American women’s salaries show the most minuet gender gap, at 90% of white men’s earnings. The gap was robust for Hispanic and Latina women, who receive payment that was only 54% of what white men were paid in 2014. The smaller gender pay gap among African American, Hispanic, American Indians, and Native Hawaiians is solely because of those men of color were paid substantially less than non-Hispanic white men in 2014.

Review of research topic

Earnings for both male and female full-time workers tend to rise with age. The increase exhibits a plateau after 45 and drops after age 65. The gender pay furthermore increases with age and difference among older workers and are considerably bigger relative to gaps among younger workers. Women typically are paid about 90% of what men are paid until they hit 35. After that the media earnings for women are typically 76-81% of what men are paid.

As a rule, earnings grow as years of education advances for both men and women. Nevertheless, while more education is an essential tool for raising earnings, it is not actually effective against the gender pay gap. At every level of academic achievement, women’s median pay are less than the median earnings for men, furthermore, in some cases the gender pay gap is wider at higher academic levels. Education enhances the earnings of all ethnicity and races as well as gender. White women get paid more than Hispanic women and African American at all education levels. The gender gap persistently traverses across academic levels and worse in for Hispanic and African American women, even amidst college graduates. Consequently, women who complete college degrees have less ability to pay off their students’ loans promptly, leaving them paying more and more for a prolonged duration comparative to men.

Limitations

Time and data constraints compelled us to restrict ourselves to present state analyses. There was no room for applying further panel data approaches due to insufficient time variation of the useful variables, although recently developed estimators (Wooldridge 1995, Kyriazidou 1997 and Lewbel 2002 e.g.) give room for heterogeneity correction in addition to endogeneity and selectivity. There seems to be an inevitable issue between using a high level comparability data set across countries, but only limited variables, and using data sets tailored to the specificity of every country, with several and informative variables.

Conclusion

Despite the achievements women have contributed in the workforce, the pay gap is still in existence. People in the government, workforce, and workforce are in position to help to do away with gap. Here are reforms that can help close the wage gap;

For companies

While some CEOs have been vocal in their commitment to paying workers fairly, American women can’t wait for trickle-down change. AAUW urges companies to conduct salary audits to proactively monitor and address gender-based pay differences. It’s just good business.

For individuals

Women can learn strategies to better negotiate for equal pay. AAUW’s salary negotiations workshops help empower women to advocate for themselves when it comes to salary, benefits, and promotions.

For policy makers

The Paycheck Fairness Act would improve the scope of the Equal Pay Act, which hasn’t been updated since 1963, with stronger incentives for employers to follow the law, enhance federal enforcement efforts, and prohibit retaliation against workers asking about wage practices. Tell the Congress to take action for equal pay.

METHODOLOGY

Introduction

In this section we look the most vital contributions to the gender gap literature with the agenda of providing a methodological overview of both the decomposition approaches and the underlying wage equation approximations. The exposition of the wage models is pegged on applications found in the pay gaps literature. Notwithstanding, there exist more advanced pay regression approaches which have not been applied in the wage gaps literature. We build on the meta-analysis by Weichselbaumer and Winter-Ebmer (2002) and on the research by Kunze (2000).

Research design

The synthesis report by the EU “Group of Experts on Gender and Employment”, European Commission 2002c, additionally offers an extensive outlay of estimated gender pay gaps in the vast Europe. Contrary to the latter, we concentrate only on the methodological aspects of approximating gender wage gaps and do not review the respective estimation results.

This section describes the different wage models that are going to be used in the study

Participants

The participants of the study are member States of the EU and source of data is European Community Household Panel (ECHP) study, country files 1998. Samples of 25 to55 year old women and men who are employed for at least 8 hours per week or out of the labor force

Instruments

It is worth noting that for the purpose of the study; absolute wage gap is equivalent to male less female gross hourly wages in euro. Relative wage gap is equivalent to Absolute wage gap / male wage rate. We use the country and individual weights (ratio of sample size and size of the population 16 years and above) provided in the ECHP for all graphs, statistics and estimations in the paper.

Procedures and Data analysis

Estimation of wage regression models

First we review the ordinary least squares regression model, the underlying assumptions and drawbacks of which are examined in detail. The limitations of sample selection, heterogeneity and endogeneity may be gone around by alternative wage equation specifications. Finally, we propose the endogenous sample selection model by Lewbel (2002) for cross-section analysis and discuss some extensions to panel the estimations of data.

Ordinary Least Square method -Regression and its limitations

The start of the review is a simple wage regression model based on the theory of human capital (Becker 1964, Mincer 1974). In the initial set-up the individual wage rate is described by human capital variables such as work experience and education. Many empirical studies today incorporate also labor market aspects job attributes, and demographic features. Heckman et al. (2003) provide a profound critique of this approach, underlining the vital variations that arise between cohort-based and cross-sectional estimates of returns to schooling, as well as the essential roles of expectation formation and subsequent uncertainty resolution.

The following are the specifications of men and women wage equations:

lnWM

i = XM

i βM + εMi

lnWF

i = XF

i βF + εFi (1)

Where i indexes individuals within the male and female samples

For clarity the difference between men and women in the wage equation is assumed hereafter and we consider a general wage equation. The logarithmic wage, lnWi is the endogenous variable. Vector Xi carries all explanatory variables. The error term represents an independent and identically distributed idiosyncratic error term having zero mean and σ2ε constant variance

Estimation methods generally aim at providing consistent estimators.

The simple wage model outlined in the equation system (1) is in many occasions estimated by ordinary least squares. However, this method only offers consistent coefficient estimates if the following orthogonality conditions are meat:

E [εi|Xi, I∗i > 0] = 0, (2)

Where I∗i indicates a latent index variable which when positive it implies an individual i is employed and non-positive otherwise.

For the orthogonality rule to be satisfied there is a necessity of restrictive assumptions. In essence, no misspecification may come on board from omitted variables, endogeneity and sample selection. Sample selection is a source of violation of the orthogonality condition. The sample of working people live out, by definition, those who do not take part in the labor market and hence may not be a random selection of the general population. If a correlation of the participation decision is made with the earnings function, the predicted value of the error term of the latter may not be zero. If, for instance, there is a positive correlation between work experience with participation as well as to the wage rate, there is a likelihood of overestimation of the return to experience by the coefficients of the wage regression. To curb this selectivity bias, a sample selection model of earnings is used which takes into account the participation decision. Also, earnings may be found by various unobservable factors. Motivation and intelligence are likely to have an impact on wages but they are hard to capture. Consequently, there may be biasness in the coefficient estimates of the observed variables. This unobserved individual heterogeneity may be taken into account by panel data techniques or by random parameter estimation assuming either a continuous distribution (MacFadden 1989) or a discrete distribution (Hoynes 1996). Panel models and their application are discussed in this section. Finally, the endogeneity of explanatory variables is also a reliable source of error specification. For example, work experience may be a previous earnings function. In the event that experience is also correlated with the present wage rate, this implies a simultaneity bias in the wage equation. Other endogenous variables like education may also be affected, besides experience. To eliminate that bias, instrumentation of the endogenous variables may be performed. An alternative solution is to carry out simultaneous estimation of the wage rate and decision regarding participation.

As stated above, we may experience inconsistency in the provided estimates of the wage determinants if the sample does not come from exogenous selection. We begin with the selection process into the labor market. Let I∗I denote a latent index variable representing individual i’s propensity of participation. Hence the index variable can also be defined as a measure for the

propensity of incorporation in the wage sample. The assumption is, this index variable is a function of personal features. This function is assumed linear for the Heckman and the Propensity score estimators:

I∗i = Viγ + ui, (3)

Where Vi is a vector of human capital and demographic variables which ought differ from those in the wage equation and ui is an identically and independently distributed idiosyncratic error term with a zero mean and a σ2u constant variance. The latent variable is not observable. The participation decision depending on a critical value, mostly set to zero is:

if I∗i > 0, Viγ + ui > 0 i will participate

Otherwise I∗i ≤ 0, Viγ + ui ≤ 0 i will not participate. (4)

The above equations demonstrate that the sample of individuals whose wages are observed is not a random sample. And as such, the wages conditional expectation of is;

E [lnWi|Xi, I∗i > 0] = Xiβ + E [εi|Xi, I∗i > 0] .

In most scenarios the term E [εi|Xi, I∗i > 0] does not equate to zero, which is a necessary condition to be satisfied for consistent Ordinary Least Square (OLS) estimation.

In the event that the estimation is based on a non-random sample, other approaches than

OLS-estimation has to be used. In the commonly used approach, suggested by Heckman (1979), an artificial repressor is added to the initial wage specification. Actually, under joint normality of ε and u

E [εi|Xi, I∗i > 0] = ρσε

σu

E [ui|Xi, I∗i > 0] = ρσε

ϕ (Viγ)

Φ(Viγ) (6)

where ϕ denote the standard normal density function and Φ the distribution function. ρ is the correlation coefficient of the wage and participation equation. Adding an error term ε∗i , which is equal to

ε∗i = εi − E [εi|Xi, I∗i > 0] , (7)

the market wage rate function to be estimated is:

lnWi = Xiβ + ρσελi + ε∗i , with λi = ϕ (Viγ)

φ(Viγ) (8)

While the normal hazard λi value is generally unknown, a consistent

Estimate bλ. i can be obtained by probit estimation of the probability that an individual

is working. Subsequently the variable bλ

i is calculated for each separate individual and added to the regressors list for lnWi , as indicated in equation. OLS estimation of this equation finally gives consistent results.

It important noting, however, that the wage equation additional term is dependent on λi and consequently on Vi.

The coefficient of the selection factor sensitivity is dependent upon selection equation specification, which reckons the Heckman’s correction technique potential weakness. The procedure demands the availability of valid instruments, i.e. variables which contribute to determining the propensity to work but are unrelated to wages. In practice, such exemption restrictions are hard to find and problems regarding collinearity are likely to take center stage. Lauer and Steiner (2000) test different instrument combinations. The normal hazard proves significant mostly in their wage equation but it does not temper with the other coefficients in a significant way. Also, the consistency of the two-step estimator is pegged on the assumption of multi-normality of the error terms. So as to go around this latter restriction, there has been the suggestion of other correction terms of the propensity of participating (Olsen 1980). In the latter technique the propensity score instead of the normal hazard is incorporated in the wage equation.

An empirical survey concerning various correction terms is in Vella (1998) provision.

Another way to deal with the selection issue is to estimate the participation and wage equations simultaneously. The advantages and drawbacks are twofold:

If there is no miss-specification in any of the equations, simultaneous estimation produces efficiency gains. However, misspecification of any of the equation may temper with the other (s), resulting in inconsistency.

Heterogeneity

This is another bias than can arise. Such a bias appears in the event that unobserved individual features which affect the wage, such as ability or motivation, are correlated with the explanatory variables, like work continuity. For instance a more intermittent worker with less motivation could earn less if intermittence lowers remuneration based on work effort. Exploiting panel data is one way to deal with unobserved individual heterogeneity. In the panel data approximation, there are two main functional assumptions regarding the unobserved individual effects. In the random effects model, individual effects are viewed as a component of the error term for which a particular distribution is assumed. In fixed effects models the regression is replaced with an individual specific characteristic which exhibits variation over individuals but is constant over time. Generally the wage equation can be specified as;

lnWit = Xitβ + Ziδ + αi + εit (9)

where Xit and Zi are vectors which incorporate individuals’ time-variant and time-invariant features. The variable αi denotes the unobservable individual-specific consequences, whereas εit mirrors unobservable effects varying both across individuals and over time.

In the random effects technique the information about an individual effect distribution is used to create an individual-specific term. The individual specific term is generated by adding the unobservable person-specific effect to the general error term.

The following are the assumptions about the composite error term:

E (εitαi) = 0for all i, t (10)

E (εitεjs) = 0unless i = j, t = s

E (εit) = 0for all i, t.

Equation (9) can be estimated by a GLS estimator, which takes account of the variance-covariance matrix of the composite error term in an optimal way. In the event that αi is not correlated with the regressors and (εi1,...,εiT , αi) ⊥ (Xi1,...,XiT, Zi), this GLS estimator is the preferred linear unbiased estimator. Unlike the random effects model, no assumption on the individual effects distributions, given the regressors, is made in the fixed effects model.

The model has a person-specific dummy variable to control for unobservable factors that may alter the individual wage rate. In the wage equation (9) αi is now a fixed variable and εit is the traditional error term. The influence of the explaining variables can be consistently estimated by a fixed effects (FE) estimator. At first it is essential to eliminate the fixed effects by transforming each observation either by mean deviation or a first-difference operator. On the other hand, these operators also sweep out the time-variant explaining variables. In the mean deviation process, the individual’s variable means are subtracted from each observation. Alternatively, αi can be removed by the first-difference method where a lagged variable value is subtracted from each observation. Then an OLS-estimation is contacted on the transformed observations to determine a consistent estimator of β.

Endogeneity

If the error term is correlated with at least one of the explanatory variables endogeineity bias arises. In the case of cross section estimation this means

E [εi|Xi] 6= 0. (11)

One way of tackling endogeneity is to include instrumental variables in the estimation. In this case, the variable xk, where xk ∈ x is correlated with the error term (E [ε|xk] 6= 0), is supplemented by an instrument or a vector of instruments Z. A consistent estimate is ensured only under the following conditions:

E [εi|Zi] = 0 (12)

E [Zixik] 6= 0 (13)

That is, the instrumented variable may not be correlated with the error term

The data source for this study was ECHP, German data file and pooled EU country files 1998 (first section without sector information: excl. Luxembourg and Sweden, second section with sector information: excl. Germany, Luxembourg and Sweden).

Note: The means and shapes of the wage decomposition effects (Oaxaca-Blinder decomposition and Juhn-Murphy-Pierce decomposition) are given based on the respective wage equation model

Ethical issues

Regarding ethical issues, let us look at some of the benefits of closing the gender pay gap.

i. Creating a fair and equal society

Enhanced equality between male and female gender would bring economic and general society advantages. Bridging the gender pay gap can assist in reducing poverty levels and enhance women’s earnings during their lifetimes. This not only eliminates the risk of women falling into poverty during their working lives, but also mitigates the danger of poverty after service.

ii. Quality jobs supply

Women have rising targets for their working lives and, if companies want to experience the best talent, equality at work is inevitable. It is crucial in creating quality jobs and a highly-motivated workforce. In turn, quality is important in building a positive work​ing environment where all workers are appreciated for their work.

iii. Good for business, workers and the economy

Employers can benefit from effect utilization of women’s talents and skills, for exam​ple by valuing women’s skills and through putting in place policies on work-life balance, train​ing and career development. Women possess skills and talents that are frequently under-utilized in the workplace and unlock​ing these potential can help companies deal with skills shortages. Appreciating women for the tasks that they execute and rewarding their input and potential fairly can improve a business’ perfor​mance effectiveness, and competence.

Organizations that devise equality plans and strategies into their workplaces enhance the best workplaces for everyone, man or woman, to work in. a business is in pole position to attract customers when it has a positive working environ​ment. And as such, its performance is improved and it competiveness boosted. Innovativeness and productivity is greatly referenced to workers who feel more confident and valued for the tasks they carry out.

iv. Avoiding litigation and complaints

Ensuring equal payment for equal work value among employees in an organization consequently clears out discrimination complaint and unfair work practices. This prevents misuse of resource in dealing with complaints and any subsequent litigation.

v. A basis for economic growth and recovery

In the event of financial and economic crisis, the participation of women in the economy and their contribution to family finances has risen. Hence, it is important to embrace the gender equality and the closing of the gender pay gap issues as they contribute to achieving employment growth, competitiveness and economic recovery.

Dissemination strategy

Sealing the gender pay gap has been a long-term priority for the EU. The organization’s pledge to seal the gap dates back to the1957 Rome Treaty. Presently a legal basis for EU action exists under the Lisbon Treaty, in addition to the commitment to gender equality herein the Charter of Fundamental Rights. The action of the organization also seeks to reform attitudes to gen​der roles – in the home, in schools, in the workplace and in the general society.

Gender equality and maximizing on women’s talents and skills are the center piece for clos​ing the gender pay gap and to realizing the objectives of the Europe 2020 Strategy, the growth strategy of the EU for this decade. The Strategy looks forward to creating more and better jobs, to realize an increased employment rate for women as part of the overall employment tar​get of 75 % for all 20-64 year-olds, and to affirm that there are 20 million fewer people at risk of poverty and social exemption by 2020.

Reducing the gender pay gap is a priority identified in a range of policy areas. Given the great concern and desire by the EU, this study becomes very vital in comparing individual features visa vies the gender gap. This will help the organization in pointing out what to prioritize in its quest of achieving its agenda. This thus forms a great platform for making the study operative to the affected audience.

Summary

In this study we propose different approaches to assess the gender pay gap in the EU. This concerns the estimation of wage equations of the estimated gaps. The summary of the main empirical results of our explorative study, indicate that at most 50% of the difference in payment between the sexes can be attributed to differences in individual features. This affirms the findings of other studies, for example, the Employment in Europe 2002 report (European Commission 2002a). Nevertheless, the size of the endowment impact varies considerably between states. It relies on the information used and on the estimation model and decomposition method practiced.

We have presented wage estimation methods that account for selectivity on the methodological level, endogeneity and heterogeneity. We decompose the pay gap both at the mean, following Oaxaca and Blinder (1973), and across the wage distribution as proposed by Juhn, Murphy and Pierce (1993). As the literature on wage equation estimation is very rich, we focus on technique mostly used in the gender gap literature (i.e. OLS and Heckman) partly and partly on very recently developed approaches (Lewbel 2002 e.g.). The latter pave room, at least theoretically, to account for the three main issues regarding methodology, selectivity, endogeneity and heterogeneity, in innovative and original manners.

The empirical application, basing on ECHP for five European countries (France, Germany, Italy, Spain and the United Kingdom) and at the EU level, indicates that correction, especially for selectivity, may have a significant effect both on wage estimates and on the pay gap decomposition: our results imply that, for Germany, the presented wage gap is smaller than the observed wage gap, given particular features on the basis of the Heckman estimates, however, the reverse is true using the Lewbel estimates. Hence, care ought to be taken on the choice of estimation method. We prioritize the Lewbel technique since it is less restrictive on the structure of the data. No structure is put on the error terms’ distribution, enhancing a more general form of unknown heteroscedasticity.

Another main result of the study is derived from the pay gap decomposition over the wage distribution quantiles (Juhn, Murphy and Pierce method). Significant variations are exhibited within and between countries. Basing on Juhn-Murphy-Pierce decompositions, a further recommendation obtained from our analysis would to carefully concentrate on the differences over the wage distribution while making policy conclusions.

Reference

Vella, F. (1998), Estimating Models with Sample Selection Bias: A Survey, the Journal of Human Resources, 33 (1), 127-169.

Wooldridge, J. M. (1995), Selection corrections for panel data models under conditional mean independence assumptions, Journal of Econometrics, 68,115-132.

Olsen, R. J. (1980), A Least Squares Correction for Selectivity Bias, Econometrical, 48 (7), 1815-1820.

Olson, C. A. and B. E. Becker (1983), Sex Discrimination in the Promotion Process, Industrial and Labor Relations Review, 36 (4), 624-64

Palme, M. O. and Wright, R. E. (1992), Gender Discrimination and Compensating Differentials in Sweden, Applied Economics, 24, 751-759

Kunze, A. (2000). The Determination of Wages and the Gender Wage Gap: a Survey, IZA Discussion Paper 193, Bonn

Heckman, J.J., L.J. Lochner and P.E. Hold (2003), Fifty Years of Mincer Earnings Regression, IZA Discussion Paper 775, Bon

Hoynes, H. W. (1996), Welfare Transfers in Two-Parent Families: Labor Supply and Welfare Participation under AFDC-Up, Econometrica, 64 (2), 295-332