Human Diversity
Page 49
12. Almost half (46 percent) of the Cohort 2 of SMPY women got their degrees in STEM fields, nearly twice the proportion among all female undergraduates during 1986–89 (the years when Cohort 2 students graduated from college). Lubinski and Benbow (2006): Table 2; Digest of Education Statistics: Online tables supplementing section 322. The categories classified as STEM in the Digest of Education Statistics tables were mathematics, engineering, architecture, computer science, physical sciences, biology, and health.
13. Lubinski and Benbow (2006): Table 2. The statistic for the general population is based on college majors for all American undergraduates during 1986–89. Digest of Education Statistics: Online tables supplementing section 322, downloadable at nces.ed.gov/programs/digest.
14. Lubinski and Benbow (2006): Table 2.
15. A technical literature exists on this topic. It goes under the label of “expectancy-value theory.” Eccles (1983).
16. Lubinski and Benbow (2006).
17. E.g., Eccles, Vida, and Barber (2004).
18. Humphreys, Lubinski, and Yao (1993).
19. A total of 563 out of the 778 members of Cohort 2 completed two subtests of the Differential Aptitude Tests (DAT): Mechanical Reasoning and Space Relations. Shea, Lubinski, and Benbow (2001).
20. Shea, Lubinski, and Benbow (2001): Fig. 3.
21. Lubinski, Webb, Morelock et al. (2001). The male-female ratio in Cohort 3 was 11.2. This is not the result of randomized testing of a nationally representative sample, but of a talent search that may have been unrepresentative for a variety of reasons. But there is also no reason to think that the search was drastically tilted in favor of boys. It represents evidence for a large male-female disproportion at the extreme right-hand tail of the IQ distribution, but 11.2 should be treated as a ballpark figure.
22. Lubinski, Webb, Morelock et al. (2001).
23. More precisely: For math, engineering, computer science, and the physical sciences, the male-to-female ratio for Cohort 2 was 1.88. For biology, health, and medicine, the female-to-male ratio was 1.84. The disparity in the most Things-oriented STEM fields was far greater yet for PhDs and professional degrees. Among the men who got PhDs, 42 percent got them in the most Things-oriented STEM disciplines, compared to 7 percent of the women—a ratio of 6.0. But the disparity reversed for the life sciences. Among SMPY men who got either a PhD or a professional degree, 23 percent got it in either medicine or biology. For women, the comparable statistic was 36 percent. Lubinski and Benbow (2006): Table 2.
24. The male-to-female ratio of young women in Cohort 3 who got undergraduate degrees in the most Things-oriented STEM majors was 1.8, almost identical to the ratio of 1.9 for Cohort 2. Meanwhile, the female-to-male ratio in the life sciences for Cohort 3 was 3.7, double the 1.8 ratio for Cohort 2. Lubinski and Benbow (2006): Table 2.
25. Data for college-educated women ages 45–49: Author’s analysis, fertility samples for the CPS, combined 2012 and 2014 surveys. Data for SMPY Cohort 2: David Lubinski, personal communication.
26. Lubinski, Benbow, and Kell (2014): 2224.
27. Diener, Emmons, and Griffin (1985); Diener, Wirtz, Tov et al. (2010).
28. The SMPY women were fractionally lower than the men on satisfaction with their success in their professional career but fractionally higher regarding the current direction of their professional career. The women were also fractionally higher than the men on their psychological flourishing, positive feelings, and overall satisfaction with life. The men and women were equally satisfied with their relationships. Lubinski, Benbow, and Kell (2014): Fig. 7.
29. Lubinski, Benbow, and Kell (2014): 2229.
30. Lubinski and Benbow (2006): 316.
31. Wang, Eccles, and Kenny (2013).
32. Valla and Ceci (2014): 220.
33. The wording is adapted from Hakim (2002): 433–34.
34. Katharine Graham broke that barrier by becoming CEO of the Washington Post’s parent company upon the death of her husband.
35. Voyer and Voyer (2014).
36. Author’s analysis, CPS. The trendline in bachelor’s degrees is remarkable not only for the steep upward climb in women’s degrees that had yet to level off as of 2015 but also for a sudden turnaround for men, from a steep increase through 1973 to the beginning of a long secular decline in 1975. The advent of the drop is not an artifact of a drop in the population of men eligible for college (defined as ages 18–23). The number in that age cohort increased slightly through 1983. And although it decreased from 1983 to 1993, it is hard to blame the drop on a shortage of males when the number of women getting BAs continued to increase during the same period. Were women crowding out men? It might have been true for private colleges that have unchanging undergraduate enrollments, but undergraduate enrollment in public universities is more flexible, and new schools continued to open throughout the period in question.
37. The male decline in graduate programs mirrors a similar phenomenon in male undergraduate enrollment, but it was proportionately larger and lasted even longer—30 years—than among undergraduates. In part, this may represent a crowding-out effect. Many graduate programs are relatively inflexible in size, and schools everywhere were eager to increase the number of women both for ideological reasons and in response to the passage in 1972 of Title IX of the Education Amendments, which prohibited discrimination by sex in any school receiving federal funds. But while crowding out may explain part of the change in the male trendline, this remarkable development is a rich subject for study.
38. For an account of the magnitude of Holland’s influence, see Nauta (2010).
39. Holland (1959): 35.
40. Wording for the six categories is taken from the technical manual for a widely used RIASEC test, UNIACT: ACT (2009).
41. Prediger (1982). The People-Things and Data-Ideas dimensions have worked as tools for vocational counseling, but there is continuing debate from a statistical standpoint about whether they should be conceived as bipolar dimensions. See Tay, Su, and Rounds (2011).
42. I took this version of the hexagon from the manual describing ACT’s version of Holland’s Vocational Preference Inventory with the labels that ACT prefers. Holland’s own labels are in parentheses. Holland (1977).
43. Holland’s Vocational Preference Inventory does not predict actual majors and occupations well—too many competing considerations can override preferences. Su (2018).
44. The effect size was +0.10, with women slightly closer than men to the Data end of the spectrum. This might seem surprising because the word data is so closely associated with the kinds of analysis done in STEM fields, but that’s the result of semantics. The research chemist may do extensive analysis of data, but typically for the purpose of testing a hypothesis grounded in theory—Ideas. Successful real estate agents are good with people, but their work is grounded in numbers—Data.
45. The O*NET database can be accessed at www.onetcenter.org. In the O*NET system, a job is classified according to the orientation that has the highest rating.
46. The Su study’s effect sizes showed the biggest sex differences in vocational interests favoring men to be (in order from high to low) mechanical and electronic repairers, engineering technicians, engineers, physical scientists, computer scientists, and mathematicians. From 2014 to 2018, among employed Americans ages 25–54 with BAs, the biggest ratios favoring males were, in the same descending order, mechanical and electronic repairers, engineers, computer scientists, engineering technicians, applied mathematicians, and physical scientists. The Su study’s only three effect sizes favoring women were, in descending order, medical services (primarily nurses and other assistants to physicians and dentists), social scientists, and medical scientists. For employed Americans, the ratios favoring women were, in descending order, medical services, social scientists, and medical scientists. Overall the correlation of sex differences in vocational interests and employment ratios for 12 STEM job categories was +.79. Su and Rounds (2015): Table 4; author’s analysis, Curren
t Population Survey combined for the 2014–18 surveys.
The ratios take into account sex differences in the total number of employed people. Rather than using the raw numbers of employed males and females for calculating the ratio, I use the proportions of employed males and females—e.g., the percentage of employed males who are engineers divided by the percentage of employed females who are engineers.
47. The first column shows a meta-analysis of more than half a million scores on interests. The second column shows the scores of the exceptionally talented members of the four SMPY samples. The third column shows the RIASEC ratings for the jobs actually held by Americans in the combined American Community Surveys of 2011–15, consisting of a cross section of American adults based on a sample of more than five million. The first two columns use the same type of measures for two widely divergent populations. The third column uses a different measure (scores for occupations instead of scores for interests). The only cells that show a notable discrepancy are the Enterprising and Artistic cells for the SMPY sample, in which the sex differences were larger than for either of the two results based on the general population. For that matter, note that the effect sizes for all six RIASEC dimensions were largest for the SMPY sample. Men and women who are exceptionally intellectually talented have greater sex differences on this topic than do men and women in the population as a whole.
48. The percentages are based on professional and academic majors. These consist of agriculture, architecture, behavioral sciences, biology, business, communications/journalism, computer science, education, engineering, humanities, health sciences, mathematics, physical sciences, public administration, and social sciences. The numbers are drawn from the relevant table in the Digest of Education Statistics published annually by the National Center for Education Statistics.
49. I limit the discussion to descriptive statistics. For a multivariate examination of these issues see Lippa, Preston, and Penner (2014), which examined employment in 60 specific jobs from 1972 to 2010. The study analyzed two issues. One was the extent to which the best jobs go to men, with women excluded or hindered from access to high-status jobs. On this score, the news was good. The link between job status and occupational sex segregation as of 2010 was weak, with women entering high-status occupations in large numbers. Furthermore, the link had been weakening since 1972.
The same study also explored trends in the People-Things orientation that accounted for so much of the sex segregation in occupations. In this regard, they found that little had changed: “Thus, one factor—job status—has led to a reduction in occupational sex segregation over the past 40 years (i.e., increasing numbers of women have entered many formerly male-dominated high-status occupations), whereas another factor—jobs’ people-things orientation—has served to maintain occupational sex segregation (women continue to be found more in people-oriented than in things-oriented occupations at all job status levels).” Lippa, Preston, and Penner (2014): 8. Both models also revealed an increase over time in the probability that women were employed in people-oriented jobs, but the statistic in question did not reach statistical significance. (Tables 2 and 3).
50. The data files may be found and downloaded at www.onetonline.org.
51. The following table gives an overview of the types of jobs that fell into each category in the CPS over the period 1971–2015. The percentages refer to the number of people in jobs in that category divided by the total number of people in jobs classified as People or Things respectively.
MAJOR CATEGORIES OF PEOPLE JOBS AND THINGS JOBS
People: Managers of staffs (31%)
Things: Low-skill labor (14%)
People: Teachers (13%)
Things: Uncategorized skilled jobs (11%)
People: Salespeople (13%)
Things: Procedural health care (10%)
People: Health care work with patients (11%)
Things: Food and restaurant jobs (10%)
People: Restaurant workers (11%)
Things: Construction trades (8%)
People: Personal services workers (3%)
Things: Some low-level white-collar jobs (8%)
People: Childcare workers (3%)
Things: Mechanics and repairers (5%)
People: Lawyers, judges, paralegals (2%)
Things: Skilled administrative support (5%)
People: Social workers (2%)
Things: Vehicle drivers (5%)
People: Advisors, counselors (1%)
Things: STEM professionals (5%)
People: Customer service workers (1%)
Things: Farmers and farm labor (4%)
People: Designers (1%)
Things: Managers of operations (4%)
People: Entertainment workers (1%)
Things: Protective services (3%)
People: Religious workers (1%)
Things: Garment and textile workers (2%)
People: Human resources workers (1%)
Things: STEM technicians (2%)
People: Hospitality work with customers (1%)
Things: Manufacturing workers (1%)
People: All others (4%)
Things: All others (3%)
Source: Author’s analysis, CPS 1970–2015. “Procedural health care” refers to occupations such as surgeon, pathologist, radiologist, health technician, or low-skill hospital service staff, in contrast to health care occupations that center on direct delivery through personal interaction between health care worker and patient.
Prediger’s equation for calculating the People-Things index score is 2R + I + C – 2S – A – E. The problem with using the index score is that occupations with strong components of both Things and People orientations end up with a middling score that conceals their dual nature. An example is the job of physician—intensely Social in some respects, intensely Realistic and Investigative in others. The Prediger People-Things index score for physicians is 1.51, near the 50th percentile. But that is produced by adding a score of 17.56 for the Things half of the equation and 16.03 for the People half—both of which are well into the upper percentiles of their respective distributions. For analyzing the orientation of jobs toward People and/or Things, I therefore separate the equation into halves: one for calculating a Things score (2R + I + C) and the other for calculating a People score (2S + A + E). Each occupation has one score for the People dimension and another score for the Things dimension. The potential range of scores for any occupation on either dimension is 4 through 28. The actual range was 4 through 23.3 for People scores and 5.3 through 24.7 for Things scores, with means of 13.0 and 15.4 respectively.
My first requirement for classifying occupations was that any occupation rated 6–7 (the two highest scores) on the Social orientation be classified as a People job and any occupation rated 6–7 on the Realistic orientation be classified as a Things job. The maximum threshold score that ensured this outcome was 15. Occupations with less than a 6 on the Social orientation were classified as a People occupation if they had enough additional points on the Artistic and Enterprising orientations to reach 15; occupations with less than a 6 on the Realistic orientation were classified as Things occupations if they had enough additional points on the Investigative and Conventional orientations to reach 15. Note that an occupation can be classified as both a People and Things job, as in the example of physicians.
Presenting CPS occupational data over time requires dealing with the different job classifications that have been used. The Census Bureau has recoded the 1990 and 2010 versions so that each is intended to be consistent across years. But reclassifications inevitably produce discontinuities that cannot be fully reconciled. The O*NET RIASEC ratings are matched most closely with the Census Bureau’s 2010 job classification, and I assigned them accordingly, using them to designate occupations as People or Things as described in the text. I replicated all the analyses using the 1990 and 2010 job classifications. The results were effectively the same for college-educated women, but the 2010 version reclassifie
d a variety of low-level People jobs in ways that showed a substantially larger rise in the proportion of high-school-educated women in People jobs from 2003 onward than is shown by the 1990 classification. I chose to present the more conservative results given by the 1990 version, and they are used for all the figures and percentages given in the text. The trendline for occupations starts in 1971 instead of 1970 because both the 1990 and 2010 job classifications involved major changes between 1970 and 1971. For example, the number of jobs in a category such as “Managers not elsewhere categorized” in the 1970 version becomes radically smaller in the occupational categories that began to be used in 1971 because the new version included many new specific managerial categories. Such shifts created an artifactual difference between the 1970 figures and subsequent years. Rather than try to correct for them, it is cleaner to begin the time series in 1971.
Using these classification rules, 35 percent of employed Americans in 2015 were in People jobs, 44 percent were in Things jobs, 8 percent were in jobs that were both People and Things, and 13 percent were in jobs that fell in between. The mean People score (2S + A + E) of employed persons was 17.9. The mean Things score (2R + I + C) of employed persons was 19.0. The correlation of the Things and People scores for employed persons in 2015 was –.75, strongly negative.
The 13 percent of jobs that were in between, classified as neither People nor Things jobs by my RIASEC algorithm, were tilted toward ones that I think most observers would intuitively classify as People jobs. The ones with the largest numbers were secretaries, cashiers, and receptionists. Thirty percent of them were jobs that intuitively seem like Things jobs. The ones with the largest numbers were bookkeepers, accountants, and auditors. Few jobs (2 percent of all jobs, by my estimate) are ones that don’t seem to have much tilt in either direction (paralegals, bank tellers, financial managers). By excluding the in-betweens from the analysis, I’m probably short-changing People jobs by a tad.