Wage Differential between the Sexes - Research Paper Example

Running Head: Econometrics Econometrics Econometrics Topic/Question The research question asks: Can it be proven that differences in years of experience and attainment of a specific level of education have a strong influence on the wage differential between the sexes, and can a trend be seen toward the narrowing of the wage gap? Data Description This paper drew data from the sample data files of Statistical Package for Social Sciences (SPSS). This data set provides important information of 474 employees. It contains employees gender, date of Birth, educational Level (years), employment Category, current Salary, beginning Salary, months since hire, previous Experience (months) and minority classification. However, in this research we are only interested in examining the effect of total experience and attainment of a specific level of education on the wage differential between the sexes. Total experience was obtained by adding ‘months since hire’ and ‘previous Experience’ This study utilized secondary sources for data collection. The use of secondary data is useful because it is cost efficient and saves time for the independent researcher (Glaser, 1963; Hyman, 1978; Hayashi, 2000). The independent researcher has limited financial resources because he or she works alone. Such researchers are usually professionals who attempt to research some significant social problem on a part-time basis. Description of the Research Design A two-step attempt was made to discover if the income of professional females increases at the same rate as that of males with comparable education as the experience increase. During the first step the existence and strength of a relationship between salary and experience for male and female professionals with a given university degree in a specific job were researched using correlational methodology. During the second step a comparison was made between the incomes of males and females with the same experience. Attainment of an educational degree was held constant for each given professional position. The correlational research method was selected as the primary methodology because of its ability to discover relationships among logical variables (Wooldridge, 2008). Research Hypothesis It appears that there is a lack of congruence in the literature in establishing an association between level of income, experience and gender, and also a lack of significant research efforts towards establishing a link between income and experience by gender identification, therefore, the hypothesis was stated as a null hypothesis (Albrect et al., 2003; Atkinson et al., 2003; Blau et al., 2006). Locke, Spirduso, and Silverman (1987) recommend that if one researches a question where no well-established theory exists one should use the format of the null-hypothesis. There will be no difference in the salary of males and females at given experience when other factors, such as academic degree (bachelors, masters, or associate degree), and professional position are held constant. Variables The variables in this studys regression analysis are as follows: Male/female relationship (the numerical value of 0 was assigned to the male and 1 to the female variable), education and experience were considered indepen­dent variables; annual salary depends on the aforementioned variables and therefore is considered the dependent vari­able. A dummy variable was introduced by multiplying experience by the value of the male/female relationship for the reason mentioned below. The first independent variable, that is, male/female, was selected for its possible effect on the difference in earnings between the sexes. The second independent variable, experience (in months), was chosen for its possible effect on the earning of any individual. The third indepen­dent variable, education, allows the individual to enter a given professional position. And finally the fourth independent variable, the dummy variable, was supposed to trigger possible changes over time, that is, the closing of the male/female wage gap. A dummy variable was introduced because of the dichomotic properties of the sex variable, that is, it can be divided into two mutually exclusive groups--sex = male or female. This dummy variable does not attempt to change the "desirable properties" of the regression estimates. Because male and female properties are mutually exclusive, the least squares will be "tricked by entering the equation as an interval variable with just two values". (Lewis-Beck, 1980, 67). As mentioned before, sex (0=male, 1=female) is an ind­ependent variable in this research. When experi­ence, education are accounted for, the introduction of a dummy variable--sex-- into the equa­tion may allow the result to become either statistically or substantively significant (Wooldridge, 2008). Regression OLS Estimation The regression results of the wage functions are discussed here in terms of a correlation between the independent variables-months of experience and sex—and the dependent variable- current earnings (See Table 1.1). Results from the OLS regression suggested that the model is statistically significant for bachelor degree and associate degree or less education, explaining 17.4%, and 26.9% of the variance in the current salary in the sample of individuals with complete data (n=474). However, all equations reported in the tables had R-squares that were under 50 percent. Thus, one could assume that by themselves, experience and sex do not play a significant role in wages earned by the individual professional. Dougherty (2006) suggest that high or low correlation does not necessarily imply that variations in the dependent variable are caused by variations in the independent variable, that is, that wages (dependent variable) earned by professionals do not solely depend on experience, degree, and female/male relationship. The possibility exists that other factors, such as geographical location and inter-industry factors, could affect the level of personal income. Some of these possibilities will be discussed in the recommendations-for ­further-studies section of this research. Wooldridge (2008) explained that the R-square, although a widely used statistic, is subjective in its explanation as to how much explained variation in the regression equation is enough. Therefore, the t-statistic, which tests the overall significance of the equation, that is, the explained to unexplained variation, can be used as an alternative test. The t-statistics as part of this research effort are reported in Table 1.3 under the regression results and will be discussed below as they affect the wage equation. Results from the Sex Equations (Xl) The results of the t-statistics in the earning equation indicate that differences in earnings between the sexes either have or have not significant positive or negative statistical influence, that is, decrease or increase of female over male salaries. The following positions show statistical significance at the .05 level with a result of slightly below or above +-2. According to table 1.3 there were not any female participants having Ph. D degree, therefore regression equation does not include gender as a predictor. Moreover, regression equation representing earnings for Master Degree holders also lacks gender as a predictor variable because only 1 female participant holds Master Degree. Furthermore, employees with either a bachelors or associate degree show negative results, that is, females earn less (see Table 1.3). The difference in wages between females and males who holds bachelor degree is more than $2343. Its negative sign shows that females who hold bachelor degree earns less than males having similar educational degree, however, this result is not significant. Furthermore, there is a difference (-$7455) in wages between females and males who hold associate degree. Negative sign shows that female employee earns less than male employee if both of them hold associate degrees and this result is significant at 0.05 level of significance. It implies that, almost all professional positions researched during this study, no statistical significance is shown by the sex variable except those who holds associate degree or less education. Therefore earning differences for other educational levels appear to result from such other factors as company size; geographies, and the like (Rao and Miller, 1971). Results from the Experience Equations (X2) The coefficients for total experience were both positive and negative and were insignificant in all educational levels except one (see Tables 1.3). The average changes in earning when experience increases in organization, as shown by the regression equations, are negative for Master and associate degree holder. On the other hand, the average changes in overall earning as experience increase is positive in Ph. d and Bachelor degree holders. However, none of these estimates of earning wages are significant for any educational level. Results from the Interaction Equations (X3) This third variable, the dummy variable, was introduced to reduce the influence of possible cause of multicollinearity, that is, linear dependency of each year of experience on the next year. This condition increases the sensitivity of any variable and thus may increase the statistical significance of the coefficients (t-value). Again, positive or negative coefficients were present and were of high or low significance, depending on the particular educational level (see Tables 1.3). It can be assumed from the t-statistics that changes over time affected the earnings equation significantly for different educational levels. For example, female employee salaries showed a decrease (around $79) over time if they hold Bachelor degree. For employees with an associate degree, the wages of females showed a decrease of $0.749. However, none of these estimates are significant for any educational level. Table 1.1 Model Summaryd Educational Level Model R R Square Adjusted R Square Std. Error of the Estimate Phd Degree 1 .110a .012 -.015 $18,393.854 Master Degree 1 .591b .349 .187 $17,846.500 Bachelor Degree 1 .417c .174 .129 $16,271.206 Associate Degree or below 1 .518c .268 .262 $6,150.785 a. Predictors: (Constant), Previous experience + months since hire (in months) b. Predictors: (Constant), Male Female Relationship, Previous experience + months since hire (in months) c. Predictors: (Constant), Male Female Relationship, Previous experience + months since hire (in months), DumVar d. Dependent Variable: Current Salary Table 1.2 ANOVAd Educational Level Model Sum of Squares Df Mean Square F Sig. Phd Degree 1 Regression 1.525E8 1 1.525E8 .451 .506a Residual 1.252E10 37 3.383E8 Total 1.267E10 38 Master Degree 1 Regression 1.368E9 2 6.841E8 2.148 .179b Residual 2.548E9 8 3.185E8 Total 3.916E9 10 Bachelor Degree 1 Regression 3.060E9 3 1.020E9 3.852 .014c Residual 1.456E10 55 2.648E8 Total 1.762E10 58 Associate Degree or below 1 Regression 5.004E9 3 1.668E9 44.091 .000c Residual 1.366E10 361 3.783E7 Total 1.866E10 364 a. Predictors: (Constant), Previous experience + months since hire (in months) b. Predictors: (Constant), Male Female Relationship, Previous experience + months since hire (in months) c. Predictors: (Constant), Male Female Relationship, Previous experience + months since hire (in months), DumVar d. Dependent Variable: Current Salary Table 1.3 Coefficientsa Educational Level Model Unstandardized Coefficients Standardized Coefficients t Sig. 95% Confidence Interval for B B Std. Error Beta Lower Bound Upper Bound Phd Degree 1 (Constant) 65695.499 7328.569 8.964 .000 50846.407 80544.590 Previous experience + months since hire (in months) 27.801 41.411 .110 .671 .506 -56.105 111.707 Master Degree 1 (Constant) 75006.808 13139.568 5.708 .000 44706.909 105306.707 Previous experience + months since hire (in months) -77.440 73.885 -.452 -1.048 .325 -247.818 92.938 Male Female Relationship -11213.844 28306.316 -.171 -.396 .702 -76488.327 54060.639 Bachelor Degree 1 (Constant) 47751.521 7262.619 6.575 .000 33196.907 62306.135 Previous experience + months since hire (in months) 38.974 43.591 .155 .894 .375 -48.384 126.332 Male Female Relationship -2343.057 9915.950 -.067 -.236 .814 -22215.065 17528.950 DumVar -79.088 62.557 -.365 -1.264 .211 -204.455 46.279 Associate Degree or below 1 (Constant) 33421.317 930.705 35.910 .000 31591.032 35251.603 Previous experience + months since hire (in months) -10.080 3.834 -.157 -2.629 .009 -17.619 -2.541 Male Female Relationship -7455.850 1273.656 -.521 -5.854 .000 -9960.567 -4951.134 DumVar -.749 6.058 -.011 -.124 .902 -12.663 11.165 a. Dependent Variable: Current Salary Normal Probability Plots The “normal probability” or the percentiles of the error distribution in the data were plotted against a normal distribution line with the same mean and variance is displayed in the P-P plot (Figure 1.1 to Figure 1.4) (Norusis, 2004). If the error term were normally distributed, the points would closely match the diagonal straight line, suggesting that the coefficients were good estimations of the effects of the tested variables on the dependent variable (Norusis, 2004). All four plots suggests that the point are relatively close to the diagonal straight line, and therefore appropriate for use in an OLS regression analysis. Figure 1.1 Figure 1.2 Figure 1.3 Figure 1.4 Scatter plots Finally, a plot of studentized residuals versus standardized predicted values were graphed (Figure 1.5 to 1.8). Three out of four plots show that variance of the errors remain consistent with increasing predicted salary (See Figure 1.5 to 1.8). However, figure 1.7 shows that the variance of the errors increases with increasing predicted salaries. However, the plot of residuals by total experience (for Bachelor degree) revealed no obvious relationship so we may reject the problem of heteroscedasticity in the residuals (See Figure 1.9). Figure 1.5 Figure 1.6 Figure 1.7 Figure 1.8 Figure 1.9 Multi- Collinearity Diagnostics According to Table 1.4, the figures represented by the partial and part correlations do not fall sharply from the value represented by zero-order correlation. It suggests that the variation explained by one predictor (e.g. experience) does not explained by other predictors (e.g. male/female relationship) involved in the model. The values of Tolerance level are also not close to 0. Furthermore, table 1.5 reports that for most of the predictors Eigen values are close to 1 which suggest that there isn’t any problem with multicollinearity. Table 1.4 Coefficientsa Educational Level Model Correlations Collinearity Statistics Zero-order Partial Part Tolerance VIF (Value Inflation Factor) Phd Degree 1 (Constant) Previous experience + months since hire (in months) .110 .110 .110 1.000 1.000 Master Degree 1 (Constant) Previous experience + months since hire (in months) -.580 -.347 -.299 .437 2.287 DumVar -.510 -.139 -.113 .437 2.287 Bachelor Degree 1 (Constant) Previous experience + months since hire (in months) .066 .120 .110 .500 1.999 DumVar -.377 -.168 -.155 .180 5.555 Male Female Relationship -.387 -.032 -.029 .189 5.288 Associate Degree or below 1 (Constant) Previous experience + months since hire (in months) -.041 -.137 -.118 .568 1.759 DumVar -.441 -.007 -.006 .256 3.907 Male Female Relationship -.493 -.294 -.264 .256 3.904 a. Dependent Variable: Current Salary Table 1.5 Collinearity Diagnosticsa Educational Level Model Dimension Eigenvalue Condition Index Variance Proportions (Constant) Previous experience + months since hire (in months) DumVar Male Female Relationship Phd Degree 1 1 1.916 1.000 .04 .04 2 .084 4.767 .96 .96 Master Degree 1 1 2.219 1.000 .03 .02 .04 2 .716 1.760 .07 .00 .37 3 .065 5.864 .90 .98 .58 Bachelor Degree 1 1 3.044 1.000 .01 .01 .01 .01 2 .731 2.040 .03 .03 .04 .04 3 .194 3.960 .09 .09 .21 .16 4 .031 9.981 .88 .87 .74 .79 Associate Degree or below 1 1 3.111 1.000 .01 .01 .01 .01 2 .613 2.253 .03 .09 .06 .05 3 .232 3.658 .15 .13 .22 .10 4 .043 8.494 .81 .77 .71 .84 a. Dependent Variable: Current Salary Implications This study was designed to examine the effect of experience on the level of wage commanded by males and females. Two questions were addressed in this study. First, is experience an important factor in the income equation? Second, can a trend be seen towards the closing of the wage gap? The null hypothesis stated that there will be no difference in the salary of males and females at given experience when other factors, such as academic degree (bachelors, masters, or associate degree) are held constant. Based on the preceding findings, the null hypothesis was rejected at the .05 level, for most of the relationships between male/female and wages from income, that is, wages for males are still higher, than wages for females when other factors, such as academic degrees are held constant. The estimates of the earning functions affirm that years of work experience have a strong effect on the earnings of a professional. The sex factor appears to be relevant in some positions, whereas it has no effect on others. For bachelors and associate degree positions, the female/male component has a negative impact on the level of wages earned (females appeared to earn less than their male counterparts). However, being male or female does not significantly affect the level of salary commanded by professionals in the bachelor degree positions. The model in this study was incapable of showing if the wage gap between the sexes has closed during the past years. All indications are that wages increased for both males and females. However, there wasn’t sufficient data available to compare wages. The average income trend line indicates that average wages for males and females will continue to increase at the same rate. No closing of the wage gap for given professionals are predicted from analysis of the data. The overall impact of experience on wages cannot be generalized, that is, it does not encompass all educational levels. Each variable has to be judged in the context of a given educational degree. The same finding is true for the female/male relationship, that is, each professional position has to be looked at independently of the others. All of the above findings imply that, during the past years, females have become integrated into the professional labor market. Females have attained in large numbers associate, bachelors and masters degrees. Social policies and programs appear to have changed social trends and beliefs, allowing females to integrate into previously male dominated positions. However, to integrate more females into male dominated positions, particularly at higher level, policies and programs should be imple­mented. Young females should be encouraged to enter higher level degree programs. In short, opportunities for females in the professional labor environment and social trends have changed drastically over the past years. However, the labor market is still far from being a perfect environment for females. Summary The research resulted in the rejection of the null­ hypothesis at the .05 level, that is with 95% confidence. Females earn less than males in almost all educational levels positions, with the exception to bachelor degree positions. No trend line towards the narrowing of the wage gap could be established, that is, there is no closing of the wage gap predicted for the immediate future. Recommendation for Future Research The present study and available data suggest several directions for future research. First, it is suggested that a survey should be conducted that includes data encompassing years of experience, sex, and wages over the last three to four years. Such a survey would allow for a true measure of present- day wage levels. Second, the present study does not focus on wage differences in various geographic locations, for example, earnings in London are much higher than in the other area. Finally, inter-industry data, not available for this study, should also be included in any future research. Logically, if one focuses on the earnings of an accountant in a large corporation or CPA firm versus those of one who works for a small doctors office or small firm, one could expect to discover a distinct salary gap. References Albrect, J. , Björklund, A. , and Vroman, S. Is there a glass ceiling in Sweden? Journal of Labor Economics vol. 21 no. (1) (2003). pp. 145–177. Atkinson, S. M. , Baird, S. B. , and Frye, M. B. Do female fund managers manage differently? Journal of Financial Research vol. 26 (2003). pp. 1–18. Blau, F. D. , ed. , Brinton, M. C. , ed. , & Grusky, D. B. (Eds.). (2006). The declining significance of gender. New York: Russell Sage. Dougherty C. (2006). Introduction to Econometrics. OUP Oxford; 3 edition. Glaser, B. G. (1963). Retreading Research materials: The use of secondary analysis by the independent researcher. American Sociological Review. 43, pg. 11-14. Hayashi F. (2000). Econometrics. Princeton University Press. Hyman, H. H. analysis. (1978). A banquet for secondary Contemporary Sociology 7, p. 42-45. Lewis-Beck, M. (1980). Applied Regression. An Introduction. Beverly Hills, Ca: Sage Publications. Locke, L.F., Spirduso, W.W., & Silverman, S.J. (1987). Proposals that work: A guide for planning dissertations and grant proposals. New York: Sage Publications. Norusis, M. J. (2004). SPSS 13.0 Guide ti Data Analysis. Chicago: Prentice Hall. Rao, Potluri, & Miller, L. R. (1971). Applied econometrics. Belmont, Calif.: Wadsworth. Wonnacott, R.J., & Wonnacott, T. J. (1979). Econometrics. New York: John Wiley & Sons. Wooldridge, J., (2008). Introductory Econometrics. South Western College; International ed of 4th revised ed edition. Read More