“Women still earn just 70 cents for every dollar a man earns. It's worse for African American women and Latinas.”
— President Obama, Remarks on Equal Pay for Equal Work, June 4, 2011 (The White House later corrected the president’s statement to 77 cents.)
— White House Statement of Administration Policy on Paycheck Fairness Act, June 4
The debate over the latest legislation to address the gap in pay between men and women is a great opportunity to explore the various ways these data are collected and often used for political purposes. There is no perfect source of data — the Census Bureau and the Bureau of Labor Statistics come up with different numbers even though they can draw on similar data sets — but often advocates of action will tend to pick the worst possible figure to advance their cause.
We will ignore the president’s misstatement and assume he meant to say 77 cents. (We don’t play gotcha at The Fact Checker.) But we also will probe how Obama and the White House come up with the claim that the gap is “worse” for black and Hispanic women.
We were struck by the disparities in the data when we noticed that a news release by Sen. Barbara Mikulski (D-Md.) trumpeted the 77 cent figure, but it included a link to a state-by-state breakdown that gave a different overall figure: 81 cents.
What’s the difference? The 77 cent figure comes from a Census Bureau report, which is based on annual wages. The BLS numbers draw on data that are based on weekly wages. Annual wages is a broader measure — it can include bonuses, retirement pensions, investment income and the like — but it also means that school teachers, who may not work over the summer, would end up with a lower annual wage.
In other words, since women in general work fewer hours than men in a year, the statistics may be less reliable for examining the key focus of the legislation — wage discrimination. Weekly wages is more of an apples-to-apples comparison, but as mentioned, it does not include as many income categories,
The gap is even smaller when you look at hourly wages — it is 86 cents vs. 100 (see Table 9) — but then not every wage earner is paid on an hourly basis, so that statistic excludes salaried workers. But, under this metric for people with a college degree, there is virtually no pay gap at all.
This brings us to our larger point: Broad comparisons are inherently problematic. As the BLS points out: “Users should note that the comparisons of earnings in this report are on a broad level and do not control for many factors that may be significant in explaining earnings differences.”
Indeed, economists at the Federal Reserve Bank of St. Louis surveyed economic literature and concluded that “research suggests that the actual gender wage gap (when female workers are compared with male workers who have similar characteristics) is much lower than the raw wage gap.” They cited one survey, prepared for the Labor Department, which concluded that when such differences are accounted for, much of the hourly wage gap dwindled, to about 5 cents on the dollar. (UPDATE: Our colleagues at WonkBlog calculate that the gap narrows to 91 cents for every dollar “if you control for life choices.”)
Not only did the White House pick the statistic that makes the wage gap look the worst, but then officials further tweaked the numbers to make the situation for African Americans and Hispanics look even more dire.
The BLS, for instance, says the pay gap is relatively small for black and Hispanic women (94 cents and 91 cents, respectively) but the numbers used by the White House compare their wages against the wages of white men. Black and Hispanic men generally earn less than white men, so the White House comparison makes the pay gap even larger, even though the factors for that gap between minority women and white men may have little to do with gender.
Administration officials, who insisted on anonymity and would not allow direct quotes, defended their use of the data. They said the Census figures give a more complete picture of women’s contribution to household finances. They also said it made much more sense to compare black and Hispanic women to white men, because those wages represent the standard that all workers should aspire to. The officials said the administration has never argued that wage discrimination is 100 percent the cause of the gap, and that previous documents issued by the White House have made it clear that the comparisons were being made to white men.
(Mikulski’s office, meanwhile, said it would no longer use the chart indicating a smaller gender pay gap.)
For people interested in more information, the Institute for Women’s Policy Research has put together a handy guide showing some of the differences between some of the data sets. (UPDATE: Heidi Hartmann of the IWPR on June 13 wrote a blog post saying “there is nothing at all misleading or biased” about Obama’s use of the figures.)
The Pinocchio Test
With so many choices of data, from legitimate sources, it is no wonder that the White House would choose the data set that best makes its case that the gender gap is large. But the president’s phrasing — “Women still earn just 77 cents for every dollar a man earns”— suggests it is much more a straight line comparison of wages when in fact the results in the annual data may stem from other factors.
It would have been better for the president to have used a range, or perhaps say “as little as 77 cents.” Moreover, he could have acknowledged that some women make lifestyle choices that could reduce their wages over the course of the year. By all accounts, there is a wage gap, but it has declined over the decades — and depending on how the data are viewed, in some cases it barely exists. Others may certainly argue the pace of the decline has been too slow, especially in recent years.
In addition, the administration should be much more transparent about the fact that it is comparing the wages of female minorities with the wages of white men. It is not apparent in the public statements.
UPDATE: One of our colleagues, who we admire greatly, raised some quibbles with this column. Our main purpose was to point out that the most commonly used statistic has its limitations, and that there are other ways to look at the data. But she felt we were missing the forest for the trees. Here are some of her objections.
1. In awarding a Pinocchio, we are suggesting that the weekly average is more legitimate than the annual average, and thus making the same mistake we accused the White House of making. In the end, she noted, all of the data showed “indisputable, major discrimination,” with women shorted at all educational levels and in most jobs. (Tables 18 and 19 of this BLS report give many examples, which are based on weekly earnings. The researchers at the St. Louis Fed argue that hourly earnings are a better comparison because women are likely to work fewer hours than men, but as we noted that data also has its own limitations.)
2. The African-American-Latina assertion by the White House was valid, even if the White House did not make clear the comparison to white men: “When compared to all men, all women make less, and women of color are at the earning bottom of that ‘all women’ category.”
3. The phrase “lifestyle choice” was poor chosen, since it is “a form of discrimination, and irrelevant in the job-to-job comparisons.” (Mea culpa—it was a poor choice of words. We were trying to note that the annual data so commonly used has its limitations because women in general do not work as many hours as men.)
In response to one of her other points, we also updated the column to make it clear that the decline in the gender pay gap has slowed in recent years.
Check out our candidate Pinocchio Tracker
Track each presidential candidate's campaign ads