The Survey of Earned Doctorates is a questionnaire distributed by the U.S. National Science Foundation to doctorate recipients at all accredited U.S. universities, which draws response rates over 90% in most years. The survey includes data on gender and ethnicity/race. Data for 2009-2014 are readily available online here. At my request, the NSF sent me gender and ethnicity/race data for philosophy going back to 1973.
With the NSF's permission, here are the raw data. Philosophy response rates averaged 92% per year, and were over 85% in all years but two.
(Note: "Ethics" started being collected as a separate doctoral subfield in 2012. For gender analysis, I have merged the 83 recipients in that category with the 1442 "philosophy" doctoral recipients during the same period. Given the small numbers, much of the race/ethnicity data were suppressed, so for race/ethnicity analysis I have excluded the "ethics" recipients. The numbers are not sufficiently different from "philosophy" to make a material analytic difference in a pool of over a thousand (54/83 [65%] male; among U.S. citizens and permanent residents, 52/68 [76%] non-Hispanic White).)
Data on Women:
Almost all respondents reported gender as male or female, with only 25/14495 respondents (0.2%) declining to self-classify gender. The total number of respondents increased from an average of 354/year in available data from the 1970s to an average of 484/year for available data in the 2010s. The chart below shows the percentage of women by year, along with both a linear regression line (green) and the best fitting quadratic curve (red).
I tried a quadratic fit to test the hypothesis that the rate at which women are entering philosophy has flattened out in the 21st century, after an increase in the 20th (see for example, these data on publication rates in elite journals and this observation by Linda Alcoff).
The quadratic curve does indeed fit much better than the linear, with a difference of 11.0 in the AICc scores (which penalize models with more parameters): The AICc relative likelihood of the quadratic vs. the linear is .996 to .004. In other words, the visually apparent flattening is highly unlikely to be chance variation in a linear trend. (I use the quadratic only to test for flattening, not to extrapolate beyond the measurement years.)
Another way to see the flattening is to aggregate by decade. In the 1970s, 17% of the SED philosophy respondents were women. In the 1980s it was 22%. In the 1990s it was 27%. In the 2000s it was also 27%. So far in the 2010s it has been 28%.
Conclusion: Since the 1990s, the percentage of women receiving Ph.D.'s in philosophy in the United States has been virtually unchanged at about 27-28%.
Data on Ethnicity and Race:
For the analysis of ethnicity and race, I excluded the "ethics" data, which had too many suppressed cells, and I excluded respondents who did not report being a U.S. citizen or permanent resident, since race and ethnicity data were not available for those respondents. There were three ethnicity categories: "Hispanic or Latino", "Not Hispanic or Latino", and "Ethnicity not reported". Within "Not Hispanic or Latino", data were subdivided into "American Indian or Alaska Native", "Asian or Other Pacific Islander", "Black or African American", "White", "Two or more races", and "Race not reported".
As expected, the large majority of respondents reporting both ethnicity and race were non-Hispanic White. Here are the numbers, with a linear regression. I also tried a quadratic curve, since I'd done it for the gender data, but it lay literally right on top of the linear regression line (with differences < .01%), so the simpler linear model is preferred.
As is evident from the chart, although a large majority of respondents are still non-Hispanic White, the percentage has been decreasing since the 1970s, and (in contrast to the gender data) there is no apparent flattening of the trend.
The numbers in all other categories are small, and are better seen in a decade-by-decade aggregation. These numbers include only U.S. citizens and permanent residents, but I have re-included respondents not reporting ethnicity or race. They are arranged from highest number to lowest.
Ethnicity not reported:
1970s: 201/2294 (8.8%) 1980s: 66/2098 (3.1%) 1990s: 58/2635 (2.2%) 2000s: 77/3069 (2.5%) 2010s: 40/1820 (2.2%)
Hispanic or Latino (any race):
1970s: 21/2294 (0.9%) 1980s: 39/2098 (1.9%) 1990s: 80/2635 (3.0%) 2000s: 112/3069 (3.6%) 2010s: 115/1820 (6.3%)
Asian or Other Pacific Islander:
1970s: 28/2294 (1.2%) 1980s: 35/2098 (1.7%) 1990s: 81/2635 (3.1%) 2000s: 94/3069 (3.1%) 2010s: 58/1820 (3.2%)
Black or African American:
1970s: 31/2294 (1.4%) 1980s: 39/2098 (1.9%) 1990s: 46/2635 (1.7%) 2000s: 66/3069 (2.2%) 2010s: 34/1820 (1.9%)
Two or more races:
1970s: 13/2294 (0.6%) 1980s: 1/2098 (0.0%) 1990s: 8/2635 (0.3%) 2000s: 38/3069 (1.2%) 2010s: 42/1820 (1.3%)
Non-Hispanic race not reported:
1970s: 0/1914 (0.0%) 1980s: 10/2098 (0.5%) 1990s: 17/2635 (0.6%) 2000s: 34/3069 (1.1%) 2010s: 15/1820 (0.8%)
American Indian or Alaska Native:
1970s: 4/2294 (0.2%) 1980s: 4/2098 (0.2%) 1990s: 15/2635 (0.6%) 2000s: 8/3069 (0.3%) 2010s: 1/1820 (0.1%)
Most notable, of course, are the low percentages generally.
Also notable is the increase from 1-2% Hispanic or Latino in the 1970s-1980s to 6.3% in the 2010s. This is still, however, well below the approximately 17% of the U.S. population that is Hispanic. It is also matched by a sharp decline in "Ethnicity not reported", raising the possibility that it is in part a reporting effect.
The percentage of non-Hispanic Black or African American U.S. PhD recipients does not appear to have increased much if at all since the 1970s, hovering around 1.4% to 2.2%, compared to 13% of the U.S. population.
HT Carolyn Dicey Jennings for her input. We are collaborating on a paper together on some of these topics. Comments, corrections, and new analyses welcome.