A new study of 24 medical schools across 12 states by Dr. Anupam B. Jena, Andrew R. Olenski, and Daniel M. Blumenthal—all of Harvard Medical School—shows that male and female doctors are often paid disparate salaries, even when accounting for several factors. The average absolute difference was around $51,000 while the adjusted average difference was about $20,000.
The New York Times quickly ran an article that all but declares the wage gap between the sexes to be a result of discrimination. It seems like an easy call to make, but the omission of certain factors gives reason to be cautious about jumping to such a conclusion.
Before I begin, I should acknowledge that the authors are far more credentialed than I am (no large feat) and probably put a lot of hard work into this study. Not that they’ll ever read this critique, but I can imagine it would be pretty annoying to read some nobody undermining your diligent study. My aim here isn’t to discredit their study, but to mention some factors that, if included, could have helped produce a more definitive picture.
To put it lightly, the popular discourse on the gender pay gap is littered with misinformation. Practically everyone has heard the standby that women are paid on average seventy seven cents for every dollar that men are. Such a blunt “statistic” makes a good rallying cry, but is just about useless in every other respect. And yet, it sticks.
Part of the problem is the age-old fallacy of inferring causation where mere correlation is at hand. This is an understandable reflex in this case—across cultures, women share a history of oppression and obstructed paths to the labor force. But this same kind of logical leap would never be tolerated if someone alleged that employers discriminated in favor of Asian men, who earn 117% of their white counterparts in 2015.
Another problem is that comparing men and women in the workplace is far more difficult than you might believe, making study results suspect in many instances.
Salaries are influenced by a number of factors. The authors know this. According to the abstract, they measured:
…information on sex, age, years of experience, faculty rank, specialty, scientific authorship, National Institutes of Health funding, clinical trial participation, and Medicare reimbursements (proxy for clinical revenue).
It seems like a pretty complete list, but there are a couple other factors that—if previous studies are any indication—would have a significant effect on the findings. The most important ones that I can think of are marital status, whether or not a doctor has children, shift availability, how many hours, not years, they have worked, and the length of any interruptions to their tenure. The inability or failure to measure and report these factors gives reason to be somewhat skeptical of the findings.
Marriage and Children
Marriage and child counts are extremely important to any discussion of salary differences between men and women. Interestingly, they usually correlate to opposite effects on the incomes of men and women, presumably due to the unequal division of domestic labor and child rearing. Married men are known to have higher salaries than comparable unmarried men for this reason–their partners are essentially investing in their earning potential. Consequently, the reverse tends to be true for women, who often divert more of their attention to household labor, thus giving up some of the effort they might otherwise spend on money-earning activities.
Having children tends to enhance this trend. From a division of labor standpoint, most couples are probably deciding that it’s more efficient to have one member carry the bulk of the remunerative workload and the other to handle a larger share of the unpaid labor. Of course, it’s not necessary that these roles be taken by men and women respectively, but that seems to be the way it plays out most of the time.
The Economist unwittingly stumbled onto this hypothesis in February, when their blog noted that lesbians often earned more money than straight women while gay men often earn less than straight men. Unfortunately, they didn’t note that lesbians are about twice as likely as gay men to get married and thus replicate the heterosexual division of labor between spouses, to some extent (lesbians are also more likely than straight women to be childless and split domestic work more equitably–all of which would contribute to higher income per capita relative to straight women).
Any discussion purporting to measure income differences as a factor solely of sex has to take these effects into account. A married man with 20 years of experience isn’t comparable to an unmarried man with 20 years experience, let alone a married mother of three with equal tenure. Because marriage and childbearing tend to have opposite effects on the incomes of men and women, the only accurate comparison to make is between childless men and women who have never been married. Some inference can be gleaned from the fact that women in their twenties earn more than similarly-aged men on average. This trend reverses around age thirty.
Interruptions to Tenure
Because knowledge and technology are constantly advancing, interruptions to tenure can be especially penalizing in a highly technical field such as medicine. The value of a computer science professional, for example, is estimated to have a halflife of only three years. A doctor with five years of experience who is coming back into the labor force after a four year gap isn’t likely to be as valuable as a comparable doctor who has spent the last five years working. We wouldn’t expect to observe the same phenomenon to the same degree in low-skilled occupations.
Therefore, simply measuring age, years of experience, and faculty rank, as this study does, doesn’t give us a complete picture of the prospects of the doctor in question. Because female employees are more likely to take time off, we can expect that on average–and especially in fields with high rates of obsolescence like medicine–they will suffer harsher penalties for absence from the workplace than male practitioners, likely depressing their average wages.
Hours and Shift Availability
Measuring years gives some idea of an employee’s commitment to a field, but a more complete picture would require the amount of hours worked and the shift availability of the physicians in question. Past research has shown that for numerous reasons, female workers tend to be more willing to trade pay for flexibility than males. If that’s true for the doctors in this study as well, it would explain some of the average pay differences that simply counting years would omit.
There are many more variables that contribute to one’s salary than I can begin to list here. The problem with studies like this is the propensity of readers to project their own prejudices onto the results.
Assuming that men and women care equally about income, for example, would lead one to believe that women are routinely getting the short end of the stick. But, to quote anyone with quick access to a hacky sack: money isn’t everything. There is a huge disparity in workplace deaths between genders (men are about 13 times more likely to die at work), which suggests that the people willing to trade job safety for income are more likely to be men than women. Neither a coal miner nor a part-time secretary could rightly tell the other they’re making the wrong choice. Individuals value things differently–that they are able to pursue employment that fits their criteria is a wonderful thing.
One could argue that the pay differentials between men and women in a given field are both a result of outside factors, individual preferences, and sexism, because those preferences are influenced by a society that has different expectations for men and women. What doesn’t really make sense, but is asserted quite often, is widespread employer discrimination against female employees.
There are basic economic reasons to doubt this. The amount of coordination involved among competing entities would be unthinkable. Moreover, since there is no fixed price of a physician that we would deem “correct”, underpaying women could be restated as overpaying men. Why, we might ask ourselves, would hospitals routinely and arbitrarily overpay their male employees? If female doctors were capable of performing the same exact work but could be retained for $20,000 less per year, why would any hospital hire a male doctor who insisted on being paid a premium? That would be some truly expensive bigotry.
While it would be difficult to prove that observed income differentials are the result of discrimination, it is indeed impossible to prove that discrimination doesn’t affect individuals’ income. It could be the case, and I can’t dispute the possibility. However, as the economist Thomas Sowell has taken great pains to point out, that difference in income would represent the maximum amount attributable to discrimination plus other factors that haven’t been accounted for. My guess is that the results of this study are realistically limited by the availability of certain data and individual preferences.