Tuesday, November 30, 2021

Sorting Men and Women by College Major and Occupation [feedly]

Sorting Men and Women by College Major and Occupation

Men and women tend to sort into different college majors. Even given the same college major, they tend to sort into different jobs. Carolyn M. Sloane, Erik G. Hurst, and Dan A. Black explore these patterns, and some implications for wage differences between men and women, in "College Majors, Occupations, and the Gender Wage Gap" (Journal of Economic Perspectives, Fall 2021, 35:4, pp, 223-48).

(Full disclosure: I've been the Managing Editor at JEP since the first issue in 1987, so I am perhaps predisposed to think the articles are of wider interest. Fuller disclosure: All JEP articles back to the first issue have been freely available online for a decade now, courtesy of the American Economic Association, so neither I nor anyone else get any direct financial benefit if you choose to check out the journal.)

For example, here are some broad groupings of majors that tend to be male- or female-dominated. The left-hand panel shows majors where the female/male share of majors started at less than 1. Biological sciences has risen above one, and is now majority female, but the other majors have changed much less. The right-hand panel shows broad categories majors where the female/male share of majors started above one–in some cases, several multiples above one. None of these areas have switched to majority male: in one area, psychology, the female dominance has become much more pronounced

Sloane, Hurst, and Black aren't trying to explain why these patterns arose or why they persist (although that's an obvious topic for speculation!). Instead, they are interested in pointing out the extent of this difference, its persistence over time, and pointing out that the male-dominated majors on average have higher wages. Their data breaks down the broad major categories into 134 detailed majors: for example, the category of "Engineering" contains 17 different majors. They write:

We find that women are systematically sorted into majors with lower potential wages relative to men. For example, Aerospace Engineering, one of the highest potential wage majors, is 88 percent male, while Early Childhood Education, one of the lowest potential wage majors, is 97 percent female. We also find that such patterns are long-standing and have been slow to converge. Overall, college-educated women born in the 1950s matriculated with majors that had potential wages 12 percent lower than men from their cohort. That gap fell to about 9 percent for the 1990 birth cohort. Even after some convergence in major sorting between men and women during the last 40 years, the youngest birth cohorts of women are still sorted into majors with lower potential wages than their male peers. Intriguingly, much of the convergence in major sorting between men and women occurred between the 1950 and 1975 birth cohorts, with a modest divergence for recent cohorts.

The authors use an interesting method of comparing wages across majors. For every major, they look at the median wages paid to a middle-aged, US-born, white male in that category. Thus, they are not trying to measure gaps between female and male wages, or the extent of discrimination. Instead, they are noting that wages are lower in female-dominated majors even if one just compares white men of the same age with different majors.

They then take the idea of sorting one step further. Men and women who have the same major tend to sort into different occupations. Here's a table illustrating this pattern. The top category shows that for education majors, 68% of women end up as teachers, compared with 50% of men. But among education majors, 18% of men end up in executive/manager jobs, compared with 9% of women. Similarly, the next panel shows that among nursing/pharmacy majors, women are more likely to end up as nurses, while men are more likely to end up in executive/manager roles.

The authors have data on 251 distinct occupations. They find that when women and men have the same majors, women have historically sorted into lower-paid occupations (again, just noting that this happens, while not investigating the question of how or why). For the effect of this occupational sorting for men and women with the same college major, they write:

{H]ow has occupational sorting conditional on major evolved across generations of US college graduates? We find that while women are sorted into occupations with lower potential wages conditional on major, this gap is closing somewhat over time. For the 1950 birth cohort, for example, women on average sorted to occupations with 11 percent lower potential earnings relative to otherwise similar men with the same majors. This gap narrowed to about 9 percent for the 1990 birth cohort. Almost all of the convergence occurred within highest potential earning majors. For example, women from the 1950 cohort who majored in Engineering—a high potential earning major—sorted into occupations with potential wages that were 14 percent lower than men from the same cohort who also majored in Engineering. For the 1990 birth cohort, however, women who majored in Engineering ended up working in occupations with roughly the same potential wages as their male peers.

Of course, these patterns of sorting by college major and occupation are also taking place against a backdrop of other changes: a rising share of women graduating from college, expansion of the US health care sector, falling birth rates, and so on. But the importance of sorting that happens early in life in college major and occupation has a lasting importance to later wages. The authors find that accounting for sorting by college major, and by occupation given the same major, can explain about 60% of the wage gap between men and women college graduates.

 -- via my feedly newsfeed

Opioid Overdoses: Worse Again [feedly]

Opioid Overdoses: Worse Again

Deaths from overdoses, especially opioids, are getting worse. Here's a graph from the Centers for Disease Control. Each point plots the cumulative deaths from drug overdoses in the previous 12 months. Thus, in January 2015 on the left-hand-side of the figure, there had been bout 50,000 drug overdose deaths in the previous 12 months. By April 2021, on the right-hand-side of the figure, there has been about 100,000 drug overdose deaths in the previous 12 months. The figure also shows that the problem seemed to have levelled out for awhile in 2018 and 2019, but with the pandemic in 2020 is started getting worse again.

David M. Cutler and Edward L. Glaeser offer a primer on how we got here in their article in the Fall 2021 issue of the Journal of Economic Perspectives: "When Innovation Goes Wrong: Technological Regress and the Opioid Epidemic." (Full disclosure: I've been Managing Editor of this journal since the first issue in 1987. On the other side, all JEP articles have been freely accessible online for a decade now, so any personal benefit I receive from encouraging people to read them is highly indirect.)

Here's an evolution of the problem in one graph. The blue line at the top is drug overdoses from all causes since 2020. The red dashed line shows overdoses just from opioids: the red line tracks the blue line, showing that the problem is fundamentally about opioids. The yellow dashed line shows overdoses from prescription opioids, and you can see that for about a decade after 2000, this was the main problem. Around 2010, when efforts were made to crack down on overprescribing prescription opioids, overdoses from heroin take off. Not long after that, overdoses from synthetic opioids like fentanyl and tramadol take off, and have been the main source of opioids overdoses in the last few years.

Cutler and Glaeser tell the story this way:

The opioid epidemic began with the availability of OxyContin in 1996. OxyContin was portrayed as a revolutionary wonder drug: because the painkiller was released only slowly into the body, relief would last longer and the potential for addiction would decline. From
1996 to 2011, legal opioid shipments rose six-fold. But the hoped-for benefits proved a mirage. Pain came back sooner and stronger than expected. Tolerance built up, which led to more and higher doses. Opioid use led to opioid abuse, and some took to crushing the pills and ingesting the medication all at once. A significant black market for opioids was born. Fifteen years after the opioid era began, restrictions on their use began to bind. From 2011 on, opioid prescriptions fell by one-third. Unfortunately, addiction is easier to start than stop. With reduced access to legal opioids, people turned to illegal ones, first heroin and then fentanyl, which has played a dominant role in the recent spike in opioid deaths.

How did Oxycontin get such a foothold? There's plenty of blame to pass around. First, the government regulators who approved the drug deserve a slice of blame. The theory of oxycontin was that slow release would require less medication, and thus pose less harm. But as Cutler and Glaeser point out: "At the time of FDA approval and even after, no clinical trials backed up this theory." Instead, the FDA relied on evidence that hospital inpatients didn't tend to become addicted, without asking if the same would apply to outpatients. Cutler and Glaeser note:

The FDA generally requires at least two long-term studies of safety and efficacy in a particular condition before drug approval, but for OxyContin, the primary trial for approval was a two-week trial in patients with osteoarthritis. Even with this limited evidence, the FDA approved OxyContin "for the management of moderate to severe pain where use of an opioid analgesic is appropriate for more than a few days"—with no reference to any particular condition and no limit to short-term use. … Two examiners involved in OxyContin's approval by the Food and Drug Administration went on to work for Purdue. When the FDA convened an advisory group in 2002 to examine the harms from OxyContin, eight of the ten experts had ties to pharmaceutical firms.

I'd also say that some of the doctors who overprescribed these medications deserve their share of the blame. There's lots of evidence of how a big marking effort by Purdue encouraged doctors to prescribe oxycontin, but at the end of the day, it's the doctors who actually did the prescribing, and some of them went far overboard. Cutler and Glaeser cite evidence that the top 5% of drug prescribers accounted for 58% of all prescriptions in Kentucky, 36 percent in Massachusetts, and 40% in California. The medical profession is well-aware that people have been getting addicted to opioids in various forms for centuries, and some greater skepticism was called for.

Roughly 700,000 Americans have dies of opioid overdoses since 1999. The isolation and stresses of the pandemic seems to have made the problem worse. It feels to me as if it's become a cliche to refer to opioid overdoses as a "crisis," but it's a crisis that doesn't seem to be receiving a crisis-level response. Cutler and Glaeser go into some detail on demand-side and supply-side determinants of the crisis, but I'll let you go to their article for details. They conclude this way:

Past US public health efforts offer both hope and despair. Nicotine is an extremely addictive substance and yet smoking rates have fallen dramatically over the past five decades, because of both regulation and fear of death. On the other side, the harms of obesity are also well-known and average weights are still increasing. We cannot predict whether opioid addiction will decline like cigarette smoking or persist like obesity.

The medical use of opioids to treat pain will always involve costs and benefits, and the optimal level of opioid prescription is unlikely to be zero. The mistake that doctors and prescribers made in recent decades was to assume overoptimistically that a time release system would render opioids non-addictive. Thousands of years of experience with the fruits of the poppy should have taught that opioids have never been safe and probably never will be.

The larger message of the opioid epidemic is that technological innovation can go badly wrong when consumers, professionals, and regulators underestimate the downsides of new innovations and firms take advantage of this error. Typically, consumers can experiment with a new product and reject the duds, but with addiction, experimentation can have permanent consequences.

Here are some of my previous posts on what I will keep calling the opioid "crisis:"

 -- via my feedly newsfeed