The Reasons for The Disparity in Wages Between Men and Women

Introduction

The sexual category wage gap, the observed modification between salaries paid to women and incomes paid to men, has been a cause of both governmental discussion and financial research during the past numerous periods. The opening is usually restrained as the relation of the median salaries of women and the median incomes of men, which specifies the amount of the median male earnings that the middle female pays represent. When the ratio is intended for all men and women who are remunerated salaries or salaries or for all wage and income earners who work full time and year round, the amount is frequently called the raw sexual category salary gap.

Two different analytic approaches have been used in leading the economic study. Characteristically, these examines have involved using comprehensive data from several sources to establish an adequate experiential basis for distributing the imagined salary tunings from other, potentially confusing differences in wages that have arisen from different origins. Each of the methods has prospered in recognizing a quantity of factors that statistically expressively interpretation for significant percentages of the raw gender salary gap. Scholars spread over the first approach have accomplished multivariate statistical investigates to approximation the gradation to which the raw sexual category salary gap is connected to a collection of potential descriptive factors.

In numerous of those studies, measurable results from the statistical examines have then remained used to decay the raw wage gap into projected amounts for which detailed descriptive variables statistically account, and a remaining proportion, usually called the adjusted gender wage gap. The adjusted gap is attributable, to unknown degrees, to other explanatory factors that have been misplaced from the examines or to overt judgment alongside female workers. Investigators applying the another approach have directed focused statistical examines to assess whether the wages paid to different workers adjust to recompense for differences in the costs of provided that specific fringe benefits, such as health insurance, or for modifications in specific circumstances of occupation, such as strenuously work, among altered types of workforces.

Statistical Analysis

The central method that has remained used in leading economic study on the sexual category salary gap has involved, first, execution multivariate mathematical analysis to assessment the amount to which the raw sexual category wage gap is connected to an arrangement of likely descriptive factors. Then, in many educations, measurable consequences from the arithmetic analysis have remained used to decompose the raw salary gap into assessed quantities for which explicit Descriptive variables clarification statistically, and an enduring percentage, regularly called the adjusted gender wage gap. As of the trouble and rate of frequently begging statistics from the same people, the examples in the longitudinal records are much lesser than the models in the cross 16 sectional files that require remained used in the educations statistics that label the conditions of a huge example of personalities at a single period. Others have analysed longitudinal statistics that define the conditions of the identical. The adjusted gap is attributable, to this method has been functional to statistics as of a diversity of foundations. Particular studies have analysed cross sectional.

Data

The study in this report has continued achieved using statistics from the Leaving Revolutions Collection records of the Current Population Survey (CPS) for 2007. The statistics contain of unweight explanations on separate labours. The example used in the arithmetical examination comprises male and female salary and salary workforces amongst 23 and 79 ages of ages. Approximating such average values for 23-year-old workers in the sample involves intentions using statistics for workers who are between 18 and 22 years of age. In addition, most individuals younger than 18 years old are still in secondary schools and do not consider employed full time a practical option. The examination has regularly examined the arithmetical association amid various mixtures of the advisory issues registered overhead and the employee's assessed hourly wage rate or, additional exactly, the regular logarithm the employee's hourly salary rate. Therefore, the youngest workforces comprised in the example are 23 years old. The measures used to progress the example are labelled in Appendix B.

The descriptive reasons reviewed in the analysis, for males and for females, contain: the worker's stage and age squared; amount of kids; needle variables (in which the value of the adjustable is unique if the representative is present and zero else) for the employee's marital position, unification illustration, enlightening accomplishment leisurely in rapports of the highest degree conventional, profession, manufacturing, and permanent or part time service position; the proportion of personnel who are females in the employee's profession and in the employee's business, and the ratio of workforces with the equal sexual category, age, and quantity of kids who either are not in the labor force for reasons other than retirement or incapacity or are employed part time. The proportions of workforces not contributing in the labor strength or salaried part time are replacements for possible occupation intervals and are considered as means finished the greatest current prior stages of, otherwise, one, two, three, four, or five years.

Table 1

Characteristics of workers included in regression analysis: means and

standard deviations by gender, and male: female ratio

Table 2

Proportional distribution of workers among occupations: means and standard

deviations by gender, and male: female ratio

Table 3

Proportional distribution of workers among industries: means and standard deviations by gender, and male: female ratio

The three facts show similar patterns of behavior for women and for men. For all three types of behavior – not participating in the labor force for reasons other than retirement or disability, not participating in the labor force for family-related reasons, and working part-time – a much larger percentage of women exhibit that type of behavior at any age. Moreover, among women, the percentage exhibiting each type of behavior at any age generally increases as the number of children increases; whereas among men, the percentage either declines or is virtually constant as the number of children increases, especially among men who are at least 25 years old.

Results

Many different versions of equation (1) have been analyzed statistically in this study. Each version has included a different combination of the explanatory factors listed in Tables 1, 2, and 3 as elements in vector X. The versions of the equation that have been analyzed have been chosen for two main reasons. Some versions have been investigated to confirm that explanatory factors that have generally been found to account for substantial portions of the gender wage gap in previous statistical analyses of cross-sectional databases, including especially samples from CPS data collected prior to 2007, account for comparable portions of the wage gap in the current statistical analysis of the sample from the 2007 CPS. Versions of the equation that have been examined for this reason are referred to hereafter as conventional versions. Other versions of the equation have been analyzed to evaluate whether the explanatory variables that have been developed as surrogates for explanatory factors that have been found to account for sizable portions of the gender wage gap in previous statistical analyses of longitudinal databases account for substantial portions of the wage gap in the current statistical analysis of cross-sectional data from the 2007 CPS. Versions of the equation that have been investigated for this reason are referred to hereafter as alternative versions.

In addition, a few alternative versions have been examined in which different, more specific data have been used as estimators for explanatory factors that have typically been analyzed using less specific data in conventional versions of the equation. Statistical analysis has been confounded for some versions by high correlation among explanatory variables. For example, it is not possible to derive reliable estimates for versions of the equation that simultaneously include an array of indicator variables specifying a worker's industry or occupation and variables measuring the percentage of workers who are females in a worker's industry or occupation. Therefore, only versions that omit the indicator variables for occupation and industry have been retained in the study. Collinearity has also confounded the simultaneous inclusion of three other combinations of variables. They are: first, the variables measuring the worker's age, age-squared, and the percentage of similar workers who are working part-time; second, the variables measuring the worker's number of children and the percentage of similar workers who are not participating in the labor force; and third, the variable measuring the number of overtime hours that an individual has worked and the indicator variable specifying that the individual has worked overtime. For each of these combinations, only versions of equation that include just the final explanatory variable from the combination listed above have been retained in the study.

The results that have been derived for the most comprehensive conventional version and the most comprehensive alternative version of equation (1) are summarized in Table 4. The table contains, for those two versions of the equation, the estimated regression coefficient for each included explanatory variable, the unadjusted R2 statistic, the R2 statistic adjusted for degrees of freedom lost, the F statistic, and its degrees of freedom. For each version, a separate set of estimates is presented for male workers and for female workers. All of the estimated regression coefficients are statistically significant with very low probability that they might have occurred randomly, as are both versions of the entire equation, both for males and for females. Further, as indicated by their similar values for the R2 statistics, both versions account for equivalent portions of the variance of the natural logarithm of the hourly wage rate for males and for females. Even more notably, with only one exception, the estimated regression coefficients for all explanatory variables that have been included in both versions of the equation are very similar, both for male workers and for female workers. Only the estimated coefficient for marital status in the equation for female workers differs appreciably between the two versions.

The difference between the estimated values of the intercepts in the two versions is inconsequential. In the conventional version, the combined effects of the estimated coefficients for age, age squared, and number of children increase the predicted value of a worker’s hourly wage; whereas in the alternative version, the combined effects of the estimated coefficients for the percentages of similar workers who either are not in the labor force or are working part-time decrease that predicted value. Thus, the net effects of the intercepts and those disjoint groups of explanatory factors for the two versions are quite similar.

Summary and Conclusions

Economic research has identified many factors that account for portions of the gender wage gap. Some of the factors are consequences of differences in decisions made by women and men in balancing their work, personal, and family lives. These factors include their human capital development, their work experience, the occupations and industries in which they work, and interruptions in their careers. Quantitative estimates of the effects of some factors, such as occupation and industry, can most easily be derived using data for very large numbers of workers, so that the detailed groupings of employees or employers that existing research indicates best describe the effects of the factors are adequately represented. Conversely, quantitative estimates of other factors, such as work experience and career interruptions, can most readily be obtained using data that describe the behavior of individual workers over extended time periods. The longitudinal data bases that contain such information include too few workers, however, to support adequate analysis of factors like occupation and industry; whereas the cross-sectional data bases that include enough workers to enable analysis of factors like occupation and industry do not collect data on individual workers over long enough periods to support adequate analysis of factors like work experience and job tenure. As a result, it has not been possible to develop reliable estimates of the total percentage of the raw gender wage gap for which all of the factors that have been separately found to contribute to the gap collectively account.

In this study, an attempt has been made to use data from a large cross-sectional database, the Outgoing Rotation Group files of the 2007 CPS, to construct variables that satisfactorily characterize factors whose effects have previously been estimated only using longitudinal data, so that reliable estimates of those effects can be derived in an analysis of the cross-sectional data. Specifically, variables have been developed to represent career interruption among workers with specific gender, age, and number of children. Statistical analysis that includes those variables has produced results that collectively account for between 65.1 and 76.4 percent of a raw gender wage gap of 20.4 percent, and thereby leave an adjusted gender wage gap that is between 4.8 and 7.1 percent. Additional portions of the raw gender wage gap are attributable to other explanatory factors that have been identified in the existing economic literature, but cannot be analyzed satisfactorily using only data from the 2007 CPS. Those factors include, for example, health insurance, other fringe benefits, and detailed features of overtime work, which are sources of wage adjustments that compensate specific groups of workers for benefits or duties that disproportionately affect them. Analysis of such compensating wage adjustments generally requires data from several independent and, often, specialized sources.

For many of the factors that have been identified, estimates of the proportion of the raw gender wage gap that is attributable to the factor have been developed. If the statistically estimated proportions were statistically independent of each other, their sum would represent the total proportion of the observed gap that is attributable to all of those factors collectively. The sum of the estimated proportions for all of the factors with estimates is, however, much greater than one. The estimates clearly are not statistically independent. Rather, the separately estimated proportions are, in effect, attributing some portions of the observed differences in wages to two or more explanatory factors. Summing the individual estimates therefore involves multiple counting of some portions of the wage differences.

In principle, the multiple counting could be eliminated by estimating the various proportions concurrently within a single comprehensive analysis that considers all of the factors simultaneously. Such an analysis is not feasible to conduct with the available data bases. Some factors, such as occupation and industry, require data for very large numbers of workers to represent adequately the detailed groupings of employees or employers that existing research indicates best describe the effects of the factors. Other factors, such as work experience and job tenure, require data that describe the behavior of individual workers over extended time periods. The longitudinal data bases that contain such information include too few workers, however, to support adequate analysis of factors like occupation and industry; whereas the cross-sectional data bases that include enough workers to enable analysis of factors like occupation and industry do not collect data on individual workers over long enough periods to support adequate analysis of factors like work experience and job tenure. Further, analysis of compensating wage adjustments generally requires data from several independent and, often, specialized sources.

As a result, it is not possible now, and doubtless will never be possible, to determine reliably whether any portion of the observed gender wage gap is not attributable to factors that compensate women and men differently on socially acceptable bases, and hence can confidently be attributed to overt discrimination against women. In addition, at a practical level, the complex combination of factors that collectively determine the wages paid to different individuals makes the formulation of policy that will reliably redress any overt discrimination that does exist a task that is, at least, daunting and, more likely, unachievable.