The Equal Pay Act of 1963 requires employers to pay women and men who do the same job, or a similar job, under the same working conditions, and requiring the same skills, the same pay (Fay and Risher, 2000). Title VII of the Civil Rights Act of 1964 and the New Civil Rights Act of 1991 require employers to be fair (non-discriminatory) in all employment decisions (Graham et al., 2000). This includes gender, as well as other protected characteristics (Graham et al, 2000). Since the original act was passed, women have made some progress towards equality in employment compensation but not enough. According to Turnasella (1999) and Fay and Risher (2000), there is still a 25% gender gap in employment compensation. Research has shown that despite improvements in women's wages, due to relative improvements in education and work experience (O'Neill, 1985; Wellington, 1993; Lewis, 1998), the wage gap has remained at the same percentage since 1993 (Turnasella, 1999).
Since the seminal work of Becker in 1957, there have been many studies evaluating gender differences in employment compensation (Graham et al, 2000). Despite efforts to identify the factors that contribute to the differences in employee compensation and the enactment of legislation aimed at forcing organizations to close the gender gap, such a gap still exists. Because this gap persists there is a movement by the federal government to force federal contractors to narrow the gender gap in employment compensation. This has provided organizations with the incentive to perform internal audits to determine if the gap exists within their organization (Adams, 1999). Many times, if the government carries out an audit that results in evidence of gender discrimination in employment compensation, the federal contractors are required to pay large settlements which include back pay as well as other damages (Adams, 1999).
Despite the financial liability that companies face for differential treatment of workers because of gender, the literature suggests that gender discrimination in employment compensation will continue to be an active issue (Graham et al, 2000). Winter-Ebmer (1995) notes that in product markets with imperfect competition, employers with an inclination towards discrimination will still survive. Brady (1988) noted that the law allows companies to justify unequal pay depending on the circumstances. For example, a pay disparity may be legal if it is based on seniority, merit or the productivity of the employee. Additionally, if the job requires employees to lift or move heavy objects, then a pay differential may be justified (Brady, 1998). However, not all organizations have been successful in dealing with gender discrimination in employment compensation. To prove a violation of the Equal Pay Act, an employee must demonstrate that males and females working for the same organization and in the same or a very simila r position are paid different wages based on the gender of the employee.
Ashraf (1996) studied the earnings differential in academia, due to gender and race, over a period of two decades. This study found that the racial gap in earnings narrowed considerably more than the gender gap in earnings over the same period. Results from a regression analysis of these earnings revealed that the discriminatory component of the earnings differential between male and female university faculty fell between 1969 and 1984, but rose thereafter. Ashraf (1996) also found that a male-female earnings gap still existed, primarily in the higher academic ranks.
For more than twenty years, United States courts have been admitting statistical evidence in employment discrimination cases. Since then, there has been a plethora of research papers addressing the advantages and drawbacks in the use of various statistical procedures in such cases (Ash, 1986; Barnett, 1982; Barrett and Doverspike, 1989; Green and Ferber, 1984; Goldberger, 1984; Robinson, 1993; Schafer, 1987). GiampetroMeyer (2000) gives many examples of current legal claims of gender-based gap in employment compensation for coaches of sports teams. Additionally, Giampetro-Meyer (2000) points out that discrimination may also exist in the benefits and bonuses given to the two gender groups and, as such, may not be measured in the gap.
Some statistical procedures used in analyzing data for the pay discrimination problem have been criticized. It has been argued that ordinary least squares tends to be biased in favor of the plaintiffs in cases of gender discrimination in employment compensation (Schafer, 1987). While data have been analyzed for possible evidence of employment discrimination using such procedures as regression analysis, reverse regression analysis, Bayesian analysis, and multinomial logit analysis (Schafer, 1987; Whiteside and Narayanan, 1989; Michelson and Blattenberger, 1984; McClung et al., 1994), there hasn't been any such analysis using mathematical programming-based classification models. The purpose of classification analysis is to provide a rule (usually in the form of a discriminant function) to correctly classify observations into pre-defined groups when attribute scores or values are given. Attribute scores are the values of attribute variables, …