Breadcrumb

  1. Inicio
  2. node
  3. Written testimony of Betsey Stevenson, Professor of Economics and Public Policy, Ford School of Public Policy, University of Michigan

Written testimony of Betsey Stevenson, Professor of Economics and Public Policy, Ford School of Public Policy, University of Michigan

Meeting of November 20, 2019 - EEOC Convenes Public Hearing on the Proposed Revision of the Employer Information Report (EEO-1)

Chair Dhillon and distinguished Commissioners, thank you for the opportunity to speak to you today about the usefulness and the burden .

My background is a Professor of Economics and Public Policy at the University of Michigan. I hold a bachelor's degree in Economics and Mathematics from Wellesley College and a Ph.D. in Economics from Harvard University. My areas of research include women's labor force participation and the role of public policy in raising women's education, work experience, and earnings. In my career, I have found that many policy changes including those not directed specifically at improving women's labor market outcomes have played an important role in the growth of women's rising share of the labor market and of earnings.

This hearing is focused on the best way forward in collecting pay data, however let me start by providing some important background information. Women have made enormous strides in the labor market. Women's labor market participation has continued to converge with that of men. In October 2019, 57.8 percent of women ages 15 and older were working compared to 69.1 percent of men.[1] While a gap remains in labor force participation, women have closed the gender gap in education. By 1982, women were outpacing men in college graduation.[2] Yet it was not until 2014 that the average adult woman became more likely than the average adult man to hold a college degree.[3] The shift of each cohort of women into more education, including higher education, also led to a closing of gap in years of work experience that once existed in the workforce between men and women.[4] So it is only in recent years that women have experience and education that is not only on par with that of men, but in many cases now exceeds it.

This change in education and experience makes the fact that women continue to earn less than men even more startling. In 2018, the median woman working full-time, all year, earned only 81.6 percent of what her male counterpart earned-a gap of more than $10,000.[5] In addition, although the pay gap closed 17 percentage points between 1981 and 2001, the decline has since slowed. It is important to note that if we were to account for differences between men and women in their education the gender pay gap would grow rather than shrink. Indeed in the past women's education and experience gains were the primary drivers narrowing the gender wage gap.[6]

It should be clear that the gap of 18.4 percentage points is unlikely to be driven only by illegal discrimination against women. Some of it reflects the cumulative exposure of gender-based assumptions that help steer girls away from some occupations and into others. While it's not up to employers to correct the mistakes of the past, those same cultural forces continue to exert pressure in their workplace. And while the law may only require that employers ensure that their pay and promotion practices are fair, they should take steps to reduce the chances that previous acts of discrimination impact their pay and promotion decisions. Moreover, employers are now learning that if they want to reduce disparities in their workforce they must go beyond ensuring that they are not actively discriminating against women and minorities. Employers are learning that they need to examine whether their culture makes it difficult for some people to thrive, whether they are making opportunities widely available, and even whether their hiring algorithm has baked in discriminatory preferences.

Pay Data Is Necessary to Address the Pay Gap

While it is difficult to know how much of the pay gap is driven by discrimination, it is clear that explicit and implicit discrimination remain pervasive in the labor market. Gender, race, and ethnicity can affect whether a candidate is hired, the starting salary offered, and the employer's overall assessment of a candidate's quality. Research has documented that even those who desire not to discriminate may subconsciously prefer a candidate based on gender, race, or ethnicity.[7] Most problematic is that unintentional disparities often arise. The only way to combat unintentional disparities is with intentional evaluation of decisions and outcomes such as the employment data collected on Component 1 of the EEO-1 and the pay data collected on Component 2 of the EE0-1.

Additionally, research has shown that women are less aggressive negotiators and often settled for a wage offer rather than negotiating for higher pay.[8] Yet, this isn't a matter of simply telling women to negotiate. When women do negotiate, if the norms of negotiation and salary expectations are not transparent, they are likely to receive lower compensation than men.[9] The voluminous body of literature on implicit discrimination and negotiation points to the fact that it is imperative that employers actively monitor their labor force for signs of pay differences.

Many Employers Realize that They Must Proactively Analyze Their Data

Many employers do actively work to assess their gender and racial wage gaps. Companies voluntarily provide detailed microdata to firms that provide compensation analysis for them and create plans to improve the diversity in their hiring. Others use software to do such analysis internally and create their own plans. And many employers are now working together to better understand how to reduce gender and racial wage gaps. For example, Employers for Pay Equity is a consortium of dozens of companies that are working together to ensure that their employees are not only compensated equitably for equal work and experience, but that all employees have an equal opportunity to contribute and advance in the workplace.

Firms clearly have the data available to analyze their pay practices to seek to uncover discriminatory or disparate pay and promotion practices. Many employers are voluntarily examining their own data to assess their employee hiring and management practices. One employer told me that when he saw gender pay gap among his sales force he was able to uncover a subtle bias in assigning sales territory to women that was likely to generate fewer sales. In other words, men were giving each other the better territories. The pay-which was based on commission--was not explicitly discriminatory. But biased decision-making was leading to a pay gap. This kind of anecdote points to the need to examine pay data. While binned data cannot uncover the root cause of such a pay disparity, the data gathered in Component 2 would have shown a potential problem that warranted further consideration. Importantly, the data collection necessary to discover the cause of this particular pay disparity would be too burdensome for the EEOC to collect outside of an enforcement action.  

The difficulty for the EEOC in estimating the reporting burden is that employers that are actively engaged in assessing their own pay gaps will find the reporting burden much smaller. Those employers who are not regularly analyzing their data for gender and racial wage gaps may find reporting more burdensome, but they are the employers where the greatest gains from reporting will be realized. It is essential that the EEOC in calculating the burden consider the fact that many employees already have this data at their fingertips for the purposes of their own analysis.

While a detailed analysis of their microdata might prove to be the most illuminating for employers, an assessment of the data required from Component 2 of the EEO-1 should lead employers to ask important questions. For example, what happens if a firm discovers that women within a job category are disproportionately in the lower bands of pay? They will need to determine whether they are paying women less for the same work, failing to promote and/or hire women into more senior positions at the same rate as men, failing to retain women, or engaging in occupational segregation practices that leave women in lower paid positions without access to the career ladders that would help them secure higher pay. While the data reported to the EEOC will not help either the employer or the EEOC identify the root cause of the problem, it will provide an important signal that the employer needs to discover the root cause of their pay disparity.

Balancing Useful Data with the Burden on Employers

In 2010, the National Equal Pay Task Force stated that "We must identify ways to collect wage data from employers that are useful to enforcement agencies but do not create unnecessary burdens on employers." The key point here is the balance between usefulness and burden. It is easy to point to changes that the EEOC could make to the required data collection efforts that would increase the usefulness of collected data. While I will discuss some of these alternatives, let me say clearly upfront that all of my suggestions to increase the usefulness of the data also increase the burden on employers. Indeed, it is notable that employers have not come forward to champion any of these alternative approaches. It is my view that the approach taken with Component 2 of the EEO-1 successfully balances the usefulness with the burden.

Gather Less Data But Make It Public

Let me turn to one alternative approach now that uses transparency along with reporting requirements. Information about the wages of others is potentially an important factor in women's success in negotiating raises and promotions and in making employers aware of implicit bias. Enhancing pay transparency can play an important role in helping women negotiate and reduce the pay gap.[10] This means that one way to improve the value of the data collected is to make it publicly available, as required in the United Kingdom. In 2010, the UK passed the Equality Act that, among other things, requires that all employers with 250 or more employees calculate and make publicly available their mean and median gender pay gap.[11] In addition, they must report gender differences in bonus pay, including both the pay amounts and the proportion of individuals in the company receiving bonuses.[12] Finally, the must report the proportion of men and women in each quartile pay band. The fact that it is publicly available increases transparency and accountability, however it also requires the data to be in very broad bins so that privacy is not violated. Instead of 12 salary bins by 10 job categories as is required by Component 2 of the EEO-1, there is no separation by job category and only 4 salary bins along with the reporting of means and medians.

The UK has sacrificed detail in order to get the benefit of making the data available to the public with minimal risk to privacy. In addition, each company is required to explain why their gender gap exists and provide steps they are taking to ameliorate any discrepancies. It is notable that the goal is not specifically to identify companies with illegal unequal pay, but to more generally help eradicate the gender pay gap by shedding greater light on the problem.

In order for workers to initiate a discrimination case they must know whether they are receiving unequal pay and that explains why the U.K. also encourages all employers to share salary ranges with employees to encourage salary negotiation. In contrast, in the United States, many face formal and informal policies limiting pay transparency, many workers, may be unaware whether they face wage discrimination. One solution would be to make such data available for workers so that they may begin to ask the necessary questions to discover whether they are making less than similar workers. Unfortunately, many employers formally prohibited workers from discussing salaries.[13] Even information in pay bands would help workers figure out where they are in the hierarchy of their occupation at their employer and lead them to ask important questions.

Clearly, making the Component 2 data available for workers to examine would increase its usefulness, but it would also increase the burden as many employers believe that their pay data is proprietary. In fact, in written testimony for the March 16, 2016 hearing Michael Eastman stated that "Many employers consider the data reported on the current EEO-1 Report to be highly sensitive and proprietary. Inclusion of compensation data will only heighten this concern."

The Benefits of Micro Data

If there was no attention paid to the burden on employers or the potential for violations of privacy then it would be better to gather more detailed micro data from each employer. The EEOC could gather data on each employee's earnings and characteristics such as years of employment, education, hours worked, and tasks on the job. As a research, I could much more accurately pinpoint employers with problematic pay practices if all of their data was available for me to analyze. The National Academy of Sciences panel recommended that the EEOC take further action to protect privacy so that it could begin to gather data at this more refined level. Any decision that involves undertaking meaningful new pilot studies should consider gathering more complete payroll data from a randomly selected subset of employers. However, given the concern that exists with the burden of collecting male and female data for only 12 pay bands for 10 job categories, the burden of increasing the number of data cells reported by enough to collect detailed micro data would be substantial.

The Usefulness of Component 2 Data

An important question before the commission is the value of the data collected from component 2 of the EEO-1.

Let me be clear: Binned data can provide useful information. As noted earlier, it is indicative of challenges that employers should investigate. And it may be useful for the EEOC in better targeting investigations.

Researchers often confront income data for individuals, households, or families that fall into income bands rather than providing a specific estimate. This is done to help preserve the anonymity of respondents, particularly when the data will be made publicly available. Researchers successfully use such data to conduct analysis on a wide range of topics. The goal is to be able to create valid measures of central tendency and dispersion, which is what provides an important quality check as well as analytic capabilities. As the National Academy of Sciences report noted "the best data are collected from payroll records, and those are most likely to be rates of pay or average earnings as computed with information on total wages and hours."

There are many statistical approaches to handling analysis using binned data.[14] Statistical packages include analysis tools for doing interval regressions that allow one to fit a linear regression model using binned data.

To put this in perspective, the EEOC has collected employment data classified by job category, gender, race, and national origin for decades. Over this time period the labor force has evolved in ways that change how much job segregation is seen in that data. One could argue that the EEO-1 data too would have benefited from being gathered on a more granular level. Yet, the EEOC balanced usefulness with burden and has proven to be extremely useful in investigations and enforcement actions. Pay data are not substantively different and will likely be as useful as the Component 1 data have been.

I want to address one point regarding the analysis of pay data collection efforts by OFCCP beginning in 2000 and continuing to 2004. The data collection was stopped because analysis should that the data did not improve efforts to better target contractors that were engaged in systemic discrimination. However, as the National Academy of Science noted that the problem was that the information in the survey "duplicated information gathered in compliance visits." Additionally, the report noted that "survey data were contaminated by the fact that compliance evaluations were conducted for many of the employers in the sample before the survey, so that employers had the opportunity to improve their practices by the time the survey was fielded."[15]

The Burden on Employers

The primary benefit of binned data is that it minimizes the burden on employers. In the National Academies report on collecting pay data they argue that "pay band data are attractive in that they align with the way that human resource managers tend to look at compensation."[16]

Most of the firms that fall within the scope of the EEO statutes and are now required to complete an annual EEO-1 report have the ability to provide these data from their existing payroll and human resource systems. The growing penetration of highly sophisticated software applications makes it easier for establishments to provide earnings data by job group and gender, race, and national origin. Indeed, these costs are falling quickly through time. The new burden estimates seem out of line with developments over the past five years in technology.

One concern from employers is that it takes time and money to link their payroll and HR systems. However, the EEOC is already obligated to collect Component 2 data for 2017, 2018, and 2019. In preparing these data the links will be made. Once they are made, they are easily reused. So any costs that employers spent to link their payroll and HR systems are not relevant for considering whether to move forward with Component 2 data collection. These are sunk costs that employers must have already born in the process of completing component 2 of the EEO-1. Once they have made these linkages there is little cost to continuing to use them. Analysis of the burden to continue to collecting data must acknowledge that the burden falls enormously for businesses that have already made the necessary updates to their software. Nearly all large employers use software that has the capabilities to do this reporting and the costs involved are the fixed costs of ensuring that the system is programmed to produce the data. In recent years our computing power has grown faster than even Moore's law predicts making the processing of the required cells for reporting a trivial computer task.

The new burden estimates that the EEOC are proposing seem out of line with technological adoption in the private sector. A quick review of ADP's website indicates that they are following these requirements closely and providing services in completing the forms for many of their clients. One quick check on the burden analysis would be to investigate the fees firms like ADP charge clients to submit Component 2 of the EEO-1. Such fees serve as an upper bound on the burden of submitting the EEO-1. Any firm that faces higher costs from completing the forms can simply choose to outsource the task, while other firms will realize that they can do it for less themselves.

 

Conclusion
In sum, the question is not whether there is an alternative set of data that could be collected that would more effectively and accurately serve the EEOC's objective of identifying employers that have problematic pay disparities that warrant further investigation. Instead, the EEOC needs to consider whether Component 2 of the EEO-1 successfully balances the benefits of gathering pay data with the costs of the employer burden of providing such data. While many employers would prefer that the EEOC have no pay data collection effort, the evolving labor market requires greater awareness and enforcement of pay disparities. The EEOC collects data with the explicit aim to aid in the efficiency of EEOC investigations but it also serves to bring greater awareness to pay and employment disparities so that problems may be rectified before more costly investigations are necessary. The EEOC is charged with ensuring that companies are indeed succeeding in setting pay and determining promotions in a way that is free from both explicit and implicit discrimination.

The motivation for increased pay transparency goes beyond individual workers. When workers are treated fairly in the workplace and are able to select jobs that best match their skills, this also benefits the overall labor market and economy. The current proposal to stop collecting summary wage data from large employers before the data that has been collected has been analyzed, without a plan to collect pay data using another methodology, fails to meet the EEOCs obligation to ensure fair workplaces for all Americans. Component 2 data can help ensure fair pay for all workers and improve employer awareness and knowledge about pay practices, all while minimizing compliance costs for businesses and the government.



[1] Data from the Bureau of Labor Statistics, Current Population Survey. 

[2] National Center for Education Statistics, Table 310. https://nces.ed.gov/programs/digest/d12/tables/dt12_310.asp

[3] American Community Survey data, analyzed in "Women Now at the Head of the Class, Lead Men in College Attainment" October 7, 2015 Census Blog by Kurt Bauman and Camille Ryan

[4] Eleven Facts about American Families and Work, October 2014 White House Council of Economic Advisers https://obamawhitehouse.archives.gov/sites/default/files/docs/eleven_facts_about_family_and_work_final.pdf

[5] Median earnings among full-time female workers were $45,097, compared to $55,291 for men. Current Population Survey, Annual Social and Economic Supplement. The wage gap is even larger when comparing earnings of women of color to earnings of non-Hispanic white men.

[6] Blau, Francine and Lawrence Kahn. 2006. "The U.S. Gender Pay Gap in the 1990s: Slowing Convergence." Industrial and Labor Relations Review, 60(1): 45-65

[7] Corinne A. Moss-Racusin, John F. Dovidio, Victoria L. Brescoll, Mark J. Graham, and Jo Handelsman

Science faculty's subtle gender biases favor male studentsPNAS 2012 109 (41) 16474-16479

[8] Babcock, Linda and Sara Laschever. Women Don't Ask: Negotiation and the Gender Divide. Princeton University Press, 2003.

[9] While disparities in negotiated salaries are small in situations where ambiguity over salary ranges and negotiation norms were low, in more ambiguous situations, women received about $10,000 less than similarly qualified men. See Pradel, Dina W. Hannah Riley Bowles, and Katheleen McGinn. 2016. "When Gender Changes the Negotiation," Harvard Business School.

[10] Chamberlain, Andrew. 2015. "Is Salary Transparency More than a Trend?" Glassdoor.

[12] "The Gender Pay Gap Explained" UK government guidance on the act https://gender-pay-gap.service.gov.uk/

[13] Institute for Women's Policy Research. 2010. "Pay Secrecy and Paycheck Fairness: New Data Shows Pay Transparency Needed"

[14] For example see von Hippel, Paul T., David J. Hunter, and McKalie Drown. 2017. "Better Estimates from Binned Income Data: Interpolated CDFs and Mean-Matching." Sociological Science 4: 641-655.

[15] National Research Council 2012. Collecting Compensation Data from Employers. Washington, DC: The National Academies Press. https://doi.org/10.17226/13496. Page 36

[16] National Research Council 2012. Collecting Compensation Data from Employers. Washington, DC: The National Academies Press. https://doi.org/10.17226/13496.