There is a common misconception in the media, politics and the general public. It is the concept of wage discrimination towards women in the workforce. It is a politically manufactured false strife purely to be a partisan ploy to stir up controversy for a problem that doesn't actually exist as it is implied. The implication is that
employers are deliberately looking at a person's gender and offering them less if they just happen to be female. Unscrupulous (or perhaps just clueless) politicians then pay lip service to fixing "the problem" in order to garner more votes (like Obama repeatedly does: 1, 2). It's a frequent sound-bite that captures attention because it uses the word "discrimination," it sounds "unfair," and it is something that needs to be "righted."
David Harsanyi puts it eloquently:
Apparently, we live in a country dominated by misogynists rather than in one resembling a meritocracy. If there's anything business owners love more than money, it's hating women. Alas, without government, you can never reach your potential.
And surely if Obama is going to mention it in his State of the Union address about how important it is for the nation, he would make sure that women earn exactly the same for the same work in his White House staff right? Well, as it would turn out, that is not the case [GASP]. Is there discrimination going on or is it something else? Read on.
Yes, it is true that women on average earn approximately 81 percent of what men earn. The key phrase here is "on average." This means that when you take all the men's salaries and take all the women's salaries and throw them into separate pots and divide by the total number of people in each pot, you will get the average salary for each gender. This is the most general slicing you can do. You will also find an average difference in salaries between all 40-50 year olds and all 20-30 year olds. Where's the outrage there? You can probably figure out where I am going with this.
Much of what determines income is what one does for a living. I know this is not a revolutionary insight, but those screaming about discrimination in the gender-wage differential apparently haven't figured it out. In order to have a 0% pay differential, you would have to have an equal number of men and women in every industry, with the same experience, the same education, the same motivations - the same everything. Do you foresee women lining up to work the oil rigs in North Dakota? How about become firemen? Construction workers? How about men for nursing, teaching, or administrative assistants? The point is that women crowd into lower paying careers on average and this would not be known in the aggregate averages if only slicing up the data by gender.
Other factors would include personal preferences and experience. Men are more willing than women to work longer hours (women make up 2/3rds of part-time workforce), undertake frequent business travel, work in dangerous (hazard pay) or labor intensive industries (such as construction), and have less job turnover, which also results in more on-the-job-training.
Also, on average, men have more work experience. This is narrowing though because women are holding off on starting a family until later in life or are skipping it altogether. But clearly, if a woman decides to have a child, she is going to have to take some time away from work and this will set her back in her career, at least somewhat. As you continue to add factors into the model, the "wage gap" will continue to narrow because you are explaining away reasons for why there would be a difference. You start getting farther and farther away from the lobbyist and media narrative that employers pay women less simply because they are female and nothing else.
In 2011 and 2012, Obama and the Democrats introduced the "Paycheck Fairness Act" in an attempt to "correct" this perceived discrimination (apparently legislation that made gender wage discrimination illegal 50 years ago didn't solve a thing). It never even got through the Democratically controlled senate, but Obama called on Congress to bring it back to life during this month's State of the Union Address. What did the Society For Human Resource Management have to say about the Paycheck Fairness Act the first time?
The PFA would effectively prohibit employers from using many legitimate factors to compensate their employees, including professional experience, education, training, employer need, local labor market rates, hazard pay, shift differentials and the profitability of the organization.
Human Resource professionals understand that a lot of factors need to be considered when hiring a new employee.
There are reports that seem to come out every year by feminist groups that highlight the news catchers (i.e. women earn less!), barely brush over the facts and reasons and then attribute any unknowns to whatever point they are trying to make, which is always discrimination. One report that came out recently was written by the American Association of University Women. But their own report is actually pretty self-damning. They drew this conclusion:
After accounting for college major, occupation, industry, sector, hours worked, workplace flexibility, experience, educational attainment, enrollment status, GPA, institution selectivity, age, race/ethnicity, region, marital status, and number of children, [only] a 5 percent difference in the earnings of male and female college graduates one year after graduation was still unexplained.
An economist such as myself would immediately like to see the regression analysis results to see how much of the variation in the data is explained by the model. The first thing we would ask is: What is the R-squared? In fact, I emailed them requesting this information and they sent me links to two reports from which they derived their conclusions [one and two]. No model will be able to capture all of the variation in the data or be an exact fit because we aren't modeling the laws of physics here and much of what will determine outcomes in human behavior could not be observed for a data set. However, both these reports capture less than 38% of the variation in the data, which means that 62% of the variation in the data is left unaccounted for. A model is generally considered good when the R-squared exceeds 70% for cross sectional data. So to find out that only 5 percent of the difference in wages is unaccounted for when additional factors (many probably unobservable and impossible to aggregate for a data set) could account for 62% of the remaining variation in data, it is fantastic for the discussion below.
Is it conceivable that they are missing a factor in the model that that could be correlated with females and wage? To exemplify what I mean, what would have happened to the coefficient on the gender variable had they left out the variable for years of experience, for instance? This is a variable where men on average have a higher number because they don't take time out of the workforce for child rearing duties. Because these two variables are correlated and since years of experience is a pretty important determinant of pay, the regression model would put some of that wage difference into the gender variable and the gender variable might say that women earn only 85% of men's wages, for instance. But would you then conclude that females are paid 15% less than men and then attribute it to discrimination? Of course not, because you would realize that you have omitted variable bias because you are missing an important factor. If you only consider specific information, you can even back up the claim that women earn more than men. So why should the automatic conclusion be that any unexplained difference is a result of discrimination, then? Discrimination would be a direct decision by an employer to to pay a woman less than a man for no other reason. An employer would be putting itself at a competitive disadvantage by encouraging its talent to go to other firms (and breaking the existing law, of course).
So what could some of the unobserved factors be that would also determine pay? Perhaps men are better salary negotiators, for example. How could you possibly quantify that for a data set? Maybe men have higher productivity or work longer hours (salaried positions don't count hours). Maybe men are systematically more competitive and aggressive in the workplace for projects, and promotions.
In their own study on page 16 they state "some behaviors, like self-promotion, that work for men may backfire on women." How much of the missing 5% would be explained by including this variable in the model if it were measurable? They just admitted that their results have omitted variable bias, but make no mention that part of the 5% difference may be attributable to that. When groups lump together any residual effect that is unaccounted for in the model and label it "discrimination," it is untruthful and misleading (discrimination is mentioned 12 times in that report, not including the table of contents).
If you are still not convinced even after all of the discussion above, then maybe this should: If you could hire a woman and pay her less for the same quality of work, why would you ever hire a man? It would only make sense to hire women!