Once you impose the ‘ceteris paribus’ condition, the alleged 23% gender pay gap starts to quickly evaporate
Next Tuesday (April 8) is “Equal Pay Day,” which is an annual event to bring public awareness to the “gender wage gap.” Based on the questionable assumption that women earn only 77 cents for every dollar a man earns, April 8 marks the date in 2014 that the average woman would have to continue working to earn the same amount of income the average man made in 2013, i.e. 68 extra days of work to make up for the 23% wage gap. Here’s how the National Committee on Pay Equity (the organization that sponsors “Equal Pay Day”) explains the 23% gender wage gap:
The wage gap exists, in part, because many women and people of color are still segregated into a few low-paying occupations. More than half of all women workers hold sales, clerical and service jobs. Studies show that the more an occupation is dominated by women or people of color, the less it pays. Part of the wage gap results from differences in education, experience or time in the workforce. But a significant portion cannot be explained by any of those factors; it is attributable to discrimination. In other words, certain jobs pay less because they are held by women and people of color.
Does the evidence support the claim that discrimination explains a significant portion of the gender wage gap? Not really. Let’s explore further. And without even considering any empirical evidence, the claim would be unbelievable prima facie. Reason? It would mean that thousands of employers across the country could easily and immediately save 23% on their labor costs by hiring only women (or firing all of their male workers and hiring female workers). That is, it couldn’t possibly be true that the gender pay gap is mostly due to discrimination, because it would mean that profit-seeking employers all across the country have overlooked an easy way to save 23% on their main cost – labor.
Let’s then consider empirical wage data from the Bureau of Labor Statistics (BLS) annual report on the “Highlights of Women’s Earnings.” Here’s the opening paragraph from the most recent BLS report on women’s earnings in 2012:
In 2012, women who were full-time wage and salary workers had median usual weekly earnings of $691. On average in 2012, women made about 81 percent of the median earnings of male full-time wage and salary workers ($854). In 1979, the first year for which comparable earnings data are available, women earned 62 percent of what men earned.
How do we explain the fact that women working full-time earned 81 cents for every one dollar men earned in 2012 (and the 19% pay gap)? Could it really be the result primarily of gender discrimination? Let’s investigate by looking at some of the findings in the BLS report:
1. From page 6: “Among full-time workers (that is, those working at a job 35 hours or more per week), men are more likely than women to have a longer workweek. Twenty-six percent of men worked 41 or more hours per week in 2012, compared with 14 percent of women who did so. Women were more likely than men to work 35 to 39 hours per week: 12 percent of women worked those hours in 2012, while 5 percent of men did. A large majority of both male and female full-time workers had a 40-hour workweek; among these workers, women earned 88 percent as much as men earned.”
Comment: Because men work more hours on average than women, some of the raw wage gap naturally disappears just by simply controlling for the number of hours worked per week, an important factor not even mentioned by groups like the National Committee on Pay Equity. For example, women earned 81.2% of median male earnings for all workers working 35 hours per week or more, for a raw, unadjusted pay gap of 18.8% for full-time workers (Table 5). But for those workers with a 40-hour workweek, women earned 87.7% of median male earnings, for a pay gap of only 12.3%. Therefore, once we control only for one variable – hours worked – and compare men and women both working 40-hours per week in 2012, about one-third of the raw 18.8% pay gap disappears.
Further, for the group of full-time workers who work 35-39 hours per week, women earned 111.3% of what their male counterparts earned in 2012, and therefore for that group there was an 11.3% pay gap in favor of women.
2. The BLS reports that for single workers who have never married, women earned 95.8% of men’s earnings in 2012, which is a wage gap of only 4.2% (see Table 1 and chart above). For that group, 78% of the unadjusted 19% wage gap is explained by just one variable (among many): marital status.
3. Also from Table 1 in the BLS report, we find that for married workers with a spouse present, women earned only 76.6% of what married men with a spouse present earned in 2012 (see chart). Therefore, BLS data show that marriage has a significant and negative effect on women’s earnings relative to men’s, but we can assume that marriage is a voluntary lifestyle decision, and it’s that choice, not labor market discrimination, that contributes to much of the gender wage gap for married workers.
4. Also in Table 1, the BLS reports that for young workers ages 20-24 years and 25-34 years, women earned 89% and 90.2% of their male counterparts (see chart), respectively. Once again, controlling for only one variable – age – we find that almost half of the unadjusted raw wage gap disappears for young workers.
5. In Table 7, the BLS reports that for single workers (includes never married, divorced, separated and widowed) with no children under 18 years old at home, women’s median weekly earnings were 95.2% of their male counterparts (see chart). For this group, once you control for marital status only, you automatically explain 75% of the gender earnings differential.
6. Also in Table 7, the BLS reports that married women (with spouse present) working full-time with children under 6 years at home earned 82% of what married men (with spouse present) earned working full-time with children under 6 years. Once again, we find that marriage and motherhood have a significantly negative effect on women’s earnings; but those lower earnings don’t necessarily result from labor market discrimination, they more likely result from personal and family choices about careers, workplace flexibility, workplace environment, and hours worked, etc.
7. If we look at median hourly earnings, instead of median weekly earnings, the BLS reports in Table 8 that women earned 86.4% of what men earned in 2012 (wage gap of 13.6%), which accounts for more than 25% of the raw gender earnings gap when measured by weekly earnings. And when we look at young workers paid hourly rates, women ages 16 to 19 years earned 97.9% of their male counterparts in 2012, and for the 20-24 year old group, women earned 92% of what men earned. For unmarried hourly workers of all ages, women earned 92.3% of their male counterparts in 2012 (a 7.7% wage gap), which explains almost 50% of the 13.6% unadjusted gender difference in hourly earnings.
When the BLS reports that women working full-time in 2012 earned 81% of what men earned working full-time, that is very much different than saying that women earned 81% of what men earned for doing exactly the same work while working the exact same hours, with exactly the same educational background and exactly the same years of continuous, uninterrupted work experience, and assuming no gender differences in family roles like childcare. As shown above, once we start controlling individually for the many relevant factors that affect earnings, e.g. hours worked, age, and marital status, most of the raw earnings differential disappears. In a more comprehensive study that controlled for all of the relevant variables simultaneously, we would likely find that those variables would account for almost 100% of the unadjusted, raw earnings differential of 19% lower earnings for women reported by the BLS. Discrimination, to the extent that it does exist, would likely account for a very small portion of the raw gender pay gap.
For example, in a 2005 NBER working paper “What Do Wage Differentials Tell Us about Labor Market Discrimination?” by June O’Neill (Professor of economics at Baruch College CUNY, and former Director of the Congressional Budget Office), she conducts an empirical investigation using Census data and concludes that:
There is no gender gap in wages among men and women with similar family roles. Comparing the wage gap between women and men ages 35-43 who have never married and never had a child, we find a small observed gap in favor of women, which becomes insignificant after accounting for differences in skills and job and workplace characteristics.
This observation is an important one because it suggests that the factors underlying the gender gap in pay primarily reflect choices made by men and women given their different societal roles, rather than labor market discrimination against women due to their sex.
Bottom Line: To claim that a significant portion of the raw wage gap can only be explained by discrimination is intellectually dishonest and completely unsupported by the empirical evidence. And yet we hear all the time from groups like the National Committee on Pay Equity, the American Association of University Women, the Institute for Women’s Policy Research, and even Presidents Obama and Carter that women “are paid 77 cents for every dollar paid to men.” And in most cases when that claim is made, there is almost no attention paid to the reality that almost all of the raw, unadjusted pay differentials can be explained by everything except discrimination – hours worked, age, marital status, children, years of continuous experience, workplace conditions, family roles, etc. In other words, once you impose the important ceteris paribus condition of “all other things being equal or held constant,” the gender pay gap that we hear so much about evaporates. And even if we allow that some minor amount of the pay gap is from gender discrimination, “Equal Pay Day” should be celebrated in the first few weeks of January, not the second week of April.