Introduction

The right to "equal pay for equal work" is set forth in the 1948 Universal Declaration of Human Rights. ^[1] Fundamental or core conventions of the International Labor Organization dating to the 1950s repeat that right and expand it beyond wages to include freedom from workplace discrimination generally, on the basis of race, color, sex, ^[2] religion, political opinion, national extraction or social origin. ^[3] Industry-wide and single-company codes of conduct repeat theprinciple of non-discrimination in the workplace with varying levels of detail as to subject (for example, "hiring, salary, benefits, advancement, discipline, termination or retirement" ^[4] ) and basis (for example, "gender, race, religion, age, disability, sexual orientation, nationality, political opinion, or social or ethnic origin" ^[5] ). 

Measuring discrimination in the workplace presents many of the same difficulties as measuring abuse and harassment, but more progress has been made in overcoming those difficulties with respect to discrimination. The need is strong, with discrimination against workers who are pregnant standing out as a particularly widespread and stubborn problem. ^[6] Discriminatory actions include, for example, in addition to the outright dismissal of pregnant workers, the obstruction of maternal health care, refusal to pay legally mandated benefits, and routinely requiring job applicants to state whether they are pregnant. ^[7] Discrimination against workers who are identified as supporters of a union can, in turn, be a violation of the right of freedom of association.  

Measurability issues  

Discrimination comes in as many and varied forms as harassment or intimidation, and measuring it in the workplace arguably presents just as many problems. Discrimination may be a means of abuse and harassment, just as it may be a means of violating freedom of association.But current measurement practice in this area is strikingly more advanced than it is in other areas of workers' rights violations such as Abuse and Harassment or Freedom of Association.

In part this is because measurement of discrimination for or against a particular group can be approached statistically; e.g., is there a wage differential between women and men doing the same work?Demographic data about the workforce, integrated with data on salaries, promotions, layoffs and the like, can produce strong indicators of likely discrimination and the need for further investigation. 

But even beyond statistics, current measurement practice is more penetrating and thorough for discrimination than for most other areas of workers' rights. One possible explanation is the U.S. domestic experience in litigation under Title VII of the Civil Rights Act of 1964, dealing with employment discrimination, where discrimination measurement issues have been argued about and refined over time in U.S. courts and before the U.S. Equal Employment Opportunity Commission. That body of experience seems to have given code compliance measurers a head start in this area. 

Reflecting the focus on statistical data, current units of measurement in this area have the highest proportion of hard quantitative units of any area of workers' rights. They also include explicit tracking of change in specified measurement results over time, which is otherwise a rarity. And they are more detailed about factory policies, procedures, and tracking mechanisms than in other areas of workers' rights, although such units of measurement would be just as relevant and helpful elsewhere. 

Discrimination on the basis of child-bearing (beginning with pregnancy), along with other gender-based discrimination, is by far the most thoroughly addressed in current measurement practice. Race, nationality, and religion receive mild attention. Age, social status, and sexual orientation are nearly ignored. 

Current strengths

Demographic and other statistical data are well covered. Discrimination on the basis of gender receives the most statistical attention, with extensive measurement units for the proportion of women at all levels of the workforce ^[8] and explicit attention to objective wage differentials by gender. ^[9] Wage differentials are also measured subjectively by asking workers if they perceive differentials on the basis of gender, ^[10] race, ^[11] nationality, ^[12] religion, ^[13] or political belief. ^[14]  

Documentary evidence gets more detailed attention than in most other subject areas, with a particular focus on documenting wage differentials for the same kind of work or level of responsibility. ^[15] One simple but telling unit of measurement is whether the employer keeps records of the basis on which it makes decisions affecting the status of individual employees, such as wage calculations, ^[16] hiring, ^[17] promotion, ^[18] access to training, ^[19] and dismissals. ^[20] If such records exist, it is much easier for an auditor to see later if there was a non-discriminatory justification for differentials, or alternatively to prove discrimination. Hence, the existence of such records is a useful indicator of integrity in the decision process.How long personnel records are maintained after an employee leaves ^[21] is an indicator in the same vein.  

Pregnancy- and other gender-related discrimination receives the most attention in terms of specific factory practices as well as statistical analysis, with units of measurement that address day care and nursing facilities, ^[22] maternity leave, ^[23] menstrual leave, ^[24] and reassigning of pregnant workers to avoid identified hazards. ^[25] Pregnancy testing gets particularly thorough attention. ^[26]  

Discrimination on the basis of religion is measured primarily in terms of whether a worker's religious practice is inhibited, on the job ^[27] or off, ^[28] rather than in terms of differential treatment in wages or work status. 

Finally, the measuring of factory procedures for preventing discrimination are more explicit and thorough than for procedural protections in most other subject areas. The units of measurement include training of both managers ^[29] and workers, ^[30] existence of preventive measures in the factory, ^[31] assigned responsibility for carrying out policy, ^[32] and tracking of changes over time (see Best current practices below). 

Current weaknesses 

Statistical units of measurement for discrimination are not well defined, despite the emphasis on statistical data that could be used to generate them. For example, measurement units address the percentage of women in a particular job category ^[33] and the existence of a gender-based wage differential, ^[34] but not how large the difference is in that category (e.g., in percentage terms), how it compares to gender-based differentials in other job categories in the same factory, or whether it exceeds a benchmark that should be taken as prima facie discrimination absent contrary evidence (a hypothetical example might be a greater-than-ten-percent differential for a job category with at least 50 workers of each gender in it). 

Qualitative units of measurement for discrimination, including individual workers' own experience and views, are no better in this area than in others, and there is no express cross-checking of quantitative against qualitative measurement results (e.g., do workers' reports or suspicions of improper preference ^[35] trigger a statistical analysis, and does the analysis confirm the the significance of the cases reported or the suspicions of discriminatory treatment?).  

Best current practices (selected) 

·        Tracking changes over time, such as in the percentage of women in the workforce generally ^[36] or in management, ^[37] in the percentage of workers from a racial minority, ^[38] or in the percentage of expatriates. ^[39] Rates of change in significant measurement results are potentially some of the best indicators of whether code of conduct policies are taking hold, and they are particularly significant in factories where code responsibilities are new, or where corrective actions are being taken in response to findings of violations.

·        Emphasizing record-keeping ^[40] and confirming in written records that legally required benefits are actually being provided ^[41]

·        Identifying the type of records that would be of most help to an auditor in proving or disproving violations, and measuring whether such records are being kept ^[42] (which, in effect, shifts the burden of proof from the auditor to the factory management)

·        Comparing in-factory demographics to demographics in the general population ^[43]  

Possible improvements suggested by analysis 

1.      Create benchmark statistical indicators of discrimination (e.g., statistically significant differences between defined groups, in easily measurable matters such as level of wages or rates of dismissal), and refine them over time as experience accumulates

2.      Use statistical indicators to trigger worker interviews about specific disparities, and vice versa

3.      Correlate specific preventive practices that have been measured (e.g., training; record-keeping on basis for hires and wage determinations; designation of named manager with responsibility for non-discrimination) with the incidence of discrimination findings, factory by factory, across a large sample of measured factories, in order to help identify the preventive practices most likely to be effective

4.      Expand the number of measurement results that are tracked over time in the same workplaces, and analyze them for correlations as in number 3 above.