Tag Archives: income

More Women Are Doctors and Lawyers Than Ever—but Progress Is Stalling

Originally posted on The Atlantic.

In the Wall Street Journal last week, Josh Mitchell reported that “Women account for a third of the nation’s lawyers and doctors, a major shift from a generation ago.” The report was triggered by anew analysis of occupations from the Census Bureau, which showed women increased their share of doctor and lawyer by four percent and six percent, respectively, from a decade earlier.

These professional advances mark “very significant progress,” according to feminist economist Heidi Hartmann, and I don’t disagree. Still, when I spoke to Mitchell I suggested he consider a glass-half-empty perspective, which somehow ended up on the cutting-room floor.

My question is, will progress continue? It doesn’t look good. I happen to be a demographer, but you don’t need to be one to see that progress for women in these fields is stalling.

First, look at the degrees earned. This figure uses statistics from the Department of Education and breaks the gender trend in law and medical degrees up by decades. Both trends show slowing progress—a smaller increase in women’s representation each decade—and both peaked (for now) at just under 50 percent female.


If half of new doctors and lawyers are women, eventually it should be possible to have professions that are gender-balanced. But don’t hold your breath.

I looked at today’s doctors and lawyers using the 2008-2010 American Community Survey (you can get the data here). Here is the representation of women among full-time and year-round working doctors and lawyers by age. Half of the youngest doctors and lawyers are women, while only one in eight of the oldest are. So as they all age, equal representation should be on the way.


But women are much more likely to drop out of these professions (and others). Among early-career professionals—people ages 25 to 44—who list their most recent jobs as doctor or lawyer, you can see that women are much more likely to be out of the labor force:


With the kind of dropout rates that produce these disparities, we would need much more than 50 percent female in the graduating classes to reach equal representation in these professions.

In Mitchell’s report, the economist Claudia Goldin, who has recently investigated women’s success as pharmacists, argues that the corporatization of medicine has helped women by introducing the concept of work-family balance, and reducing the gender earnings gap—all changes that helped women in pharmacies as well. But I don’t see the evidence that such practices have yet changed the medical industry enough to reduce the gender differences in drop-out rates. And the research evidence shows that explicit diversity policies—with teeth—often are necessary to break the logjam.

And Mitchell’s story did not mention any efforts to reduce the segregation of men and women—especially in medicine—into different specialties. That segregation is a big part of what drives the earnings gap among doctors and lawyers. Here are the median earnings by age for doctors and lawyers, from the same source:


At the peak of that curve—ages 45 to 50—female doctors are earning just 62 percent of men’s median earnings. As they make their decisions about whether to enter the field, and how to specialize, and how to handle their family demands and opportunities, these disparities in representation and rewards come into play. The decisions men and women in these professions make should never be seen as free choices unconstrained or unaffected by the institutional environment.


Filed under In the news, Me @ work

Mystery solved? Why “women in their 20s” earn more

When pundits like David Brooks get sucked into the factoid-warp of Hanna Rosin (The End of Men) and Liza Mundy (The Richer Sex), they are always floored by the idea that young women earn more than young men. To them this represents the future. And woe to any woman trying to convince a jury she’s being discriminated against while these books are in the headlines. Brooks spelled it out real simple: “Women in their 20s outearn men in their 20s.”

That’s easily shown to be wrong (still holding my breath for the correction). But the more detailed factoid, the one you get in the long-soundbite version of the end-of-history, is that “median full-time wages for single childless women ages 22-30 exceeds those of single childless men in the same age group,” as reported in USA Today, for example. That was calculated by Reach Advisors using the American Community Survey.*

Making broad conclusions based on weird data slices is bad practice. And this is a great case study in why.

Who are those full-time working, not-married and childfree 20-somethings in metro areas? I ran that filter over the 2010 ACS data available from IPUMS, and this jumped out:

OK, so for whatever reason, notice that this group includes a disproportionate share of White women and Latino men. That turns out to be pivotal, since these particular Latino men have very low earnings. Check the earnings by race/ethnicity and gender:

So that’s it. The overall $1,000 advantage for women (seen in the bars on the far right) is the result of these particular Latino men’s low earnings. The high earnings of these White women are important, of course, they’re just not higher than White men’s. If you just look at Whites or Blacks there is no advantage for women.

I am all for getting into the problem of Latino men’s (and women’s) low average earnings. But that’s not where this story has been going. More than anything this is just shoddy statistical cherry-picking.

Hey media mega-conglomerates: give that meme a rest!

* Reach Advisors also limited the analysis to metro areas, so I did that as well. I don’t get as big an advantage for “women” as that reported in that 2010 USA Today article, which said it was based on 2008 data (they got an 8% gap, I get 3%). I don’t care to figure out exactly the source of the differences (and Reach hasn’t published their code).


Filed under In the news, Me @ work

Fact-checking David Brooks, citing Hanna Rosin edition

I’m getting hundreds of clicks today from someone who posted a comment on David Brooks’ column (a sliver of a fraction of a percent of his readers). So here’s a little detail.

He wrote:

Men still dominate the tippy-top of the corporate ladder because many women take time off to raise children, but women lead or are gaining nearly everywhere else. Women in their 20s outearn men in their 20s.

The first sentence is too slippery to fact-check. But I think it would be more accurate to say, and less slippery: “men dominate at all levels of the corporate ladder, and women having children is only part of the reason why.”

But anyways, the second sentence is a specific empirical claim, and it is not true.

If Brooks’s intern — or (shudder) a fact-checker at a newspaper — had typed this into Google they would know: “bls earnings by age and sex”. As of today, that search leads to this as the second link:

This is not an obscure source. Finding it hardly counts as research. If you opened it, you would see the median earnings for men and women, by age, race and ethnicity, for full-time workers. If you copied a couple of those columns into your spreadsheet and asked it for the ratio of the medians, this is what you would see: Men earn more at every age.

I know the history of this factoid meme, which is detailed here and here. What he is quoting is probably a Hanna Rosin (or Liza Mundy) misquote of an analysis which compared 22-30-year-old, never-married, childless, metro-area residents a few years ago.

But there is being succinct and punchy, and there is being wrong. Shortening “22-30-year-old, never-married, childless, metro-area residents” to “women in their 20s” is wrong.

Common, major media conglomerates: get the facts!


Filed under In the news

End-of-Men-Richer-Sex reality check, 40 years of pants edition

In 2010, 28% of wives were earning more than their husbands. And wives were 8-times as likely as their husbands to have no earnings.

I still don’t have my copies of The End of Men, by Hanna Rosin, or The Richer Sex, by Liza Mundy. But I’ve read enough of their excerpts to plan out some quick data checks.

Both Rosin and Mundy say women are rapidly becoming primary earners, breadwinners, pants-wearers, etc., in their families. It is absolutely true that the trend is in that direction. Similarly, the Earth is heading toward being devoured by the Sun, but the details are still to be worked out. As Rosin wrote in her Atlantic article:

In feminist circles, these social, political, and economic changes are always cast as a slow, arduous form of catch-up in a continuing struggle for female equality.

Which is right. So, where are we now, really, and what is the pace of change?

For the question of relative income within married-couple families, which is only one part of this picture – and an increasingly selective one – I got some Census data for 1970 to 2010 from IPUMS.

I selected married couples (heterogamous, throughout this post) in which the wife was in the age range 25-54, with couple income greater than $0. I added husbands’ and wives’ incomes, and calculated the percentage of the total coming from the wife. The results show and increase from 7% to 28% of couples in which the wife earns more than the husband:

(Thanks to the NYTimes Magazine for the triumphant wife image).

Please note this is not the percentage of working wives who earn more. That would be higher — Mundy calls it 38% in 2009 — but it wouldn’t describe the state of all women, which is what you need for a global gender trend claim. This is the percentage of all wives who earn more, which is what you need to describe the state of married couples.

But this 51% cutoff is frustratingly arbitrary. No serious study of power and inequality would rest everything on one such point. Earning 51% of the couple’s earnings doesn’t make one “the breadwinner,” and doesn’t determine who “wears the pants.”

Looking at the whole distribution gives much more information. Here it is, at 10-year intervals:

These are the points that jump out at me from this graph:

  • Couples in which the wife earns 0% of the income have fallen from 46% to 19%, but they are still 8-times as common as the reverse — couples where the wife earns 100%.
  • There have been very big proportionate increases in the frequency of wives earning more — such as a tripling among those who earn 50-59% of the total, and a quadrupling among those in which the wife earns it all.
  • But the most common wife-earning-more scenario is the one in which she earns just over half the total. Looking more closely (details in a later post) shows that these are mostly in the middle-income ranges. The poorest and the richest families are most often the ones in which the wife earns 0%.

Maybe it’s just the feminist in me that brings out the stickler in these posts, but I don’t think this shows us to be very far along on the road to female-dominance.

Previous posts in this series…

  • Discussed The Richer Sex excerpt in Time (finding that, in fact, the richer sex is still men).
  • Discussed that statistical meme about young women earning more than young men (finding it a misleading data manipulation), and showed that the pattern is stable and 20 years old.
  • Debunked the common claim that “40% of American women” are “the breadwinners” in their families.
  • Debunked the description of stay-at-home dads as the “new normal,” including correcting a few errors from Rosin’s TED Talk.
  • Showed how rare the families are that Rosin profiled in her excerpt from The End of Men.


Filed under Me @ work

Do Asians in the U.S. have high incomes?

The Pew Research Center last week released a lengthy research report on Asians in the U.S., titled “The Rise of Asian Americans.” It combines information from the Census and government sources with the results of Pew’s own national survey of attitudes and opinions.

The report has lots of good information, but there are some thorny problems here. I’ll describe a few problems, then offer one data exercise to help clarify. This gets technical and it’s long, so I will give you the substantive conclusion at the top:

  1. Because Asians are a diverse category made up of groups with very different profiles, and their household composition and geographic distribution vary by national origin group, generalizations are often unhelpful.
  2. Among the 10 largest Asian groups, five (Japanese, Indian, Chinese, Filipino, Korean) are above average in income and five (Vietnamese, Pakistani, Laotian, Cambodian, Hmong) are below. But all 10 Asian groups are doing better compared to the national average than they are compared to the average incomes in the places they live — they are richer nationally than they are locally.
  3. The amount of income inequality within Asian groups varies as well. Pakistanis,  Chinese, Koreans and Indians have the highest levels of inequality, while Filipinos and Laotians have low levels of inequality.

Details follow.

But first: Who is Asian? On the Census questionnaire, Asian is not exactly a category – rather, the category is created from all the responses of people who specify Asian national origins in the race question. To refresh, this is the question:

So “Asian” is all the people who specify Asian Indian, Chinese, Filipino, Japanese, Korean, Vietnamese or “Other Asian.” (The right-hand column is for Pacific Islanders.) Yes, in the U.S., Hispanic/Latino national origins are “ethnicities,” but Asian national origins are “races.” Go figure.

That lack of a common definition is compounded by two factors: First, there is so much diversity among Asians that the using a single category is as challenging statistically as it is politically. And second, Asians – as the Pew report shows – have a high rate of intermarriage with Whites, as well as (among some groups) across Asian national-origin lines. As a result, some Asian groups have high rates of “multiple-race identification” — especially those whose immigration was generations ago.

The controversy over the Pew report is summarized in this Color Lines story and this response from the Asian American / Pacific Islander Policy Research Consortium. The gist of it is that the report was too rosy in its description of Asian advantages and too homogenizing in its treatment of Asian diversity – as a result repeating the “divisive trope” of the “model minority.” Here’s part of the summary from the New York Times:

Drawing on Census Bureau and other government data as well as telephone surveys from Jan. 3 to March 27 of more than 3,500 people of Asian descent, the 214-page study found that Asians are the highest-earning and best-educated racial group in the country.

Among Asians 25 or older, 49 percent hold a college degree, compared with 28 percent of all people in that age range in the United States. Median annual household income among Asians is $66,000 versus $49,800 among the general population.

In the survey, Asians are also distinguished by their emphasis on traditional family mores. About 54 percent of the respondents, compared with 34 percent of all adults in the country, said having a successful marriage was one of the most important goals in life; another was being a good parent, according to 67 percent of Asian adults, compared with about half of all adults in the general population.

Asians also place greater importance on career and material success, the study reported, values reflected in child-rearing styles. About 62 percent of Asians in the United States believe that most American parents do not put enough pressure on their children to do well in school.

Did Pew homogenize or glorify too much? I don’t know. Here’s a graph from the report, which shows that Asian groups differ, but they all have higher-than-average household incomes:

The Color Lines story quotes Deepa Iyer, head of the National Council of Asian Pacific Americans and executive director of South Asian Americans Leading Together:

The danger in framing the study the way Pew did, and the way the media picked up on it, is that folks who are in the general public and institutional stakeholders and policy makers might get the impression that they don’t necessarily need to dig deep into our communities to understand any sort of disparities that exist.

The problem of homogenizing Asians is longstanding in American sociology. In most data analyses, the Asian sample is small to begin with, so they are often collapsed into one category (which I’ve done) or dropped from the story (which I’ve also done, angering some readers). Here is a typical passage, from a 2001 article by Leslie McCall:

That didn’t stop her (or lots of other people) from extensively analyzing Asians as a combined group, and offering speculation on her results.

There are other examples. In my experience, Jen’nan Read and I broke out six Asian groups for a study of women’s employment with the 2000 Decennial Census data — which reinforced my conviction that disaggregating is best. (This 2010 Census report gives some detail on more than 20 national-origin groups.)

Some new numbers

Anyway, I’ve got four specific issues to address with Pew’s comparison of household incomes (some of which they acknowledge in the report): a) Household composition differs between groups (more or fewer kids, grandparents); b) Asians disproportionately live in parts of the U.S. with high costs of living (like Hawaii and California, and urban areas generally); c) different members of a household might have different “race” identities (so, a Korean man married to a Chinese woman might define their child is either or both); and d), levels of inequality differ between groups, so central tendency comparisons don’t capture the whole story.

In this exercise I address these problems. I adjust for household size and composition, count individuals’ own “race” rather than imposing a single identity on the household, compare incomes to the average in the local metropolitan area as well as the national average, and compare levels of within-group inequality.

All in one blog post! Someone might want to work this up into a real paper (and maybe someone else already has? The last time I really read about this was more than 10 years ago.) So I’m just offering this approach as a suggestion, and making my code available if anyone wants to pursue it (see below).

I use the 2006-2010 combined American Community Survey, from IPUMS, for maximum recent sample size. This is about 15 million people, and the Asian samples range from about 160,000 Chinese to 7,500 Laotians. I identify individuals according to their individual “race.”

I calculate their incomes as per capita household income, adjusted for economies of scale. To do that, I count adults as 1 person, kids under 18 as .7 of a person, and divide the total household income by that count to the power of .65 for economies of scale (see here for details). Then I take the natural log of all that to pull in the right tail of the distribution (so the mean isn’t pulled up by the ~1%). When I’m done, everyone in the household has the same income, and the distribution is pretty normal. Nice!

To see what this does: The mean household income for individuals in the country in 2006-2010 is $79,174, and the natural log of the composition-and-scale adjusted per capita income is 10.26 (see figure), which works out to $28,439. In comparison, the logged incomes for Asians range from 10.6 (~$40,000) for Indians and Japanese, down to 9.7 (~$16,000) for Hmong.

To deal with the issue of living in expensive areas, I take the mean of that logged income in each metropolitan area, and compare each person’s own per capita income to that. So a score of 0 means you have the average income in your area — more than 0 means richer than average, less than zero is poorer.

There is not one correct answer about how to do this: Having an average income in a rich area still means you can buy more stuff on Amazon than someone with a lower absolute income. But it might also mean having a smaller house, or not being considered rich by your neighbors. On the third hand, if a rich family moves to a rich area, we shouldn’t feel sorry for them for not being above average in their neighborhood. For your consideration, I show the incomes compared with the national average and with the local metro mean, for the 10 largest Asian groups (click for higher resolution):

To interpret the figure, you can see that Japanese and Indians are about 0.36 higher in log dollars than the national average but only 0.26 higher than their metro-area averages. On the downside, Hmong individuals have adjusted per capita incomes of 0.58 less than the national average, but 0.63 less than their local average.

Higher-than-average-income Japanese, Indians, Filipinos and Chinese are about 73% of the total; Koreans are about average, and the lower-than-average groups are 17% of the total. By this method, then, a big majority of Asians in the U.S. belong to above-local-average income groups, but a substantial fraction are well below average. And they are all doing worse relative to their metro area neighbors than they are to the national average.

Notice how it’s different from the Pew figure. In that, Vietnamese households had higher incomes than Koreans, and both were above the national average. Here Koreans are doing substantially better, mostly as a result of the household size adjustments. Also, the smaller groups I show – the ones Pew did not detail in that figure – are the poorer ones. And they are also doing worse locally relative to their national position.

Finally, consider the inequality within groups. Without doing a full-blown analysis of this, I can show the importance of the question with a simple box-and-whisker plot. This shows the distribution of income — adjusted as described above for household composition and size — for each group, including non-Asians for comparison.

The graph shows a lot of information in a small space:

  • The line through the middle of each box is the median, or mid point, of each income distribution.
  • The blue + sign is the mean. The further the mean is above the median, the more rich people there are pulling the mean up.
  • The top and bottom of the boxes are the 75th and 25th percentiles. The further apart they are, the greater the income gap between top and bottom.

(The top whiskers, which can be used to show the highest point in each distribution, aren’t shown here, because they’re so far away it would make the graph unreadable.)

As I mentioned at the top, the graph shows that Pakistanis and Chinese, and to a lesser extent Koreans and Indians, have high levels of inequality — their + signs are far from their median lines, and their 75/25 spreads are large. On the other hand, Filipinos, Laotians and Hmong have much narrower spreads.

Practically speaking, all this means that some groups are misrepresented by measures of the overall status of “Asians,” especially the smaller, poorer groups. And further, that generalizing will represent some groups worse than others because of their internal diversity. For example, the average Chinese American is quite a bit richer than the average non-Asian American, but the poorest 25% of Chinese are not much better off than the poorest 25% of the population at large.

Like I said, just an idea, with a few examples.

Take it away

Feel free to do it more, and/or better, yourself. Here’s my SAS code. Please credit me if it works, but don’t blame me if it’s wrong. This has not been peer-reviewed – it’s rough work product. Send any corrections written on the back of a $20-bill. (Everyone else: You can stop reading now!)

Just get these variables from IPUMS:


And then do this to them:

/* exclude households with no income */
if hhincome>0;
/* this codes folks into this scheme, with Asians from richest to poorest:
0="Not Asian"
1= "Japanese" 
2= "Indian" 
3= "Filipino" 
4= "Chinese" 
5= "Korean" 
6= "Vietnam" 
7= "Pakistani"
8= "Laotian"
9= "Cambodian"
10= "Hmong"
11= "OtherA"
12= "twoplusA" 
/* these codes refer to RACED, the detailed race variable on IPUMS */
/* Count asians as those who are asian alone, multiple asian, asian and white, asian and PI, or white-asian-PI */
if raced in (400 410 420 811 861 911) then asian=4;
if raced in (610 814) then asian=2;
if raced in (600 813 864 865 914) then asian=3;
if raced in (640 816) then asian=6;
if raced in (620 815) then asian=5;
if raced in (500 812) then asian=1;
if raced in (660) then asian=9;
if raced in (661) then asian=10;
if raced in (662) then asian=8;
if raced in (669) then asian=7;
if raced in ( 663 664 665 666 667 668 670 671 672 810 817 818 860 867 868 910 915) then asian = 11;
if raced in ( 673 674 675 676 677 678 679 819 869) then asian = 12;
/* so the variable labels display in output */
 ASIAN asian.
/* add the decimal to the weight variable */
format PERWT 11.2;
/* this counts up the number of kids and adults in each household */
proc sort data=temp; by serial; run;
data hh;
set temp (keep=serial age);
by serial;
if first.serial then do;
retain kids adults;
if age le 18 then do; kids=kids+1; end;
if age gt 18 then do; adults=adults+1; end;
keep serial kids adults;
if last.serial;
proc sort data=hh; by serial; run;
/* this merges in those people counts, and then calculates the household income variable */
data people;
merge temp hh; by serial;
equiv = hhincome/((adults+(.7*kids))**.65);
lnequiv = log(hhincome/((adults+(.7*kids))**.65));
/* this outputs the mean logged household equivalent income for each metro area (with non-metro folks as 0 */
proc means noprint data=people;
var lnequiv;
class metarea;
weight perwt;
output out=msa mean=msaequiv;
proc sort data=msa; by metarea; run;
proc sort data=people; by metarea; run;
/* this merges in the metro area variable and calculates the income-difference variable */
data merged;
merge people (in=a) msa;
by metarea;
if a;
relhhinc = lnequiv-msaequiv;
/* Distribution of the logged income variable */
proc univariate data=merged; var lnequiv; run;
proc univariate data=merged; var lnequiv; class asian; run;
/* Boxplots */
proc sort data=merged; by asian; run;
title 'Income distributions, household composition- and scale-adjusted';
proc boxplot data=merged;
 plot equiv*asian / clipfactor = 1.5 grid;
where asian le 10;
/* National income means */
proc means mean data=merged;
var lnequiv;
weight perwt;
/* National asian income means by group */
proc means mean missing data=merged;
var lnequiv; class asian; weight perwt;
/* Relative income for each Asian group, for metro people only */
proc means mean;
var relhhinc; class asian; weight perwt;
where asian >0 and metarea>0;


Filed under Me @ work, Research reports

Gender gap, 2011

The good people at the Institute for Women’s Policy Research have a new brief report on the gender gap in pay, based on 2011 data from the Bureau of Labor Statistics.

The gender pay gap reflects both the tendency of women to work in lower-paid occupations, and the tendency of men to earn more than women within occupations. IWPR calculated women’s median weekly earnings as a percentage of men’s, for those working full-time only, for the 20 most common occupations among men and women. Here is my figure from their results, with occupations listed from most to least female-dominated. It shows the extent of segregation in major occupations, and the nearly-universal gender pay gap within them, regardless of gender composition:

A few occupations were on both lists, and some had two few women or men to calculate relative wages, so only 30 are shown here.

Men’s earnings are higher in all but one of these occupations (stock clerks), though the gaps are larger on average in the more-male occupations.*

This report follows a recent appearance by IWPR’s president, Heidi Hartmann, on the Rachel Maddow show. Hartmann has posted this review of their discussion about the gender pay gap.

Recent related posts:

* This is the opposite of the pattern Matt Huffman and I found in our 2003 paper, where the gender gap was greater in female-dominated jobs (with statistical controls and 1990 data). Something to look into.


Filed under Research reports

Twin peaks of intergenerational mobility

There is a lot of news about economic mobility from recent weeks. Some of it draws from Pew’s Economic Mobility Project. Not as recently, there was an excellent review and analysis by Emily Beller and Michael Hout in the Future of Children a few years ago. In between, I somehow missed a collection of economic analyses in a book titled, Unequal Chances: Family Background and Economic Success, edited by Samuel Bowles, Herbert Gintis, & Melissa Osborne Groves.

The first chapter is posted free, and it includes a good introduction to the statistical and conceptual issues that arise when trying to understand patterns of mobility across generations. It includes a discussion of heritability, genetics, IQ and the like, which is quite approachable to the reader who is ready to think about decomposing correlations.

One good example regarding genetic heritability of traits that determine income: race in South Africa, which is almost entirely inherited (since there’s very little interracial marriage) and has a huge effect on income, but the effect of which is still social/environmental, not “natural.”

Anyway, I like this “twin peaks” figure, which shows the relationship between parent and child family income decile:

Probability of offspring attaining given income decile, by parents’ income deciles, United States. Based on total family income for black and white participants in the Panel Study of Income Dynamics who were born between 1942 and 1972, and their parents. The income of the children was measured when they were aged 26 or older, and was averaged over all such years for which it was observed. The number of years of income data ranged from 1 to 29 with an average of 11.5; the median year of observation was 1991. Parents’ income was averaged over all observed years in which the child lived with the parents. The number of years of income data ranged from 1 to 27 with an average of 11.9; the median year of observation was 1974. The simple age adjusted correlation of parents’ and children’s incomes in the data set represented in the figure is 0.42.

So, 30% of children from the top decile stay there (point D), 32% of children from the bottom decile stay there (C), while the odds of making it from the top to the bottom, or vice versa, are both less than 2% (A and B).

There is a nice symmetry to the figure, but it’s important to know that what’s happening up and down the distribution is highly varied, according to the analyses in the book. For example, at the top there is a lot of transmitted wealth. At the bottom there are a lot of health crises and premature deaths, including from violence. And the bottom is much stickier for Black children than for Whites.

Leave a comment

Filed under Research reports