HEAL defined

In motivating and defining HEAL, we draw on some lessons from the STEM field. The rise of STEM was related to the growing need, especially over the last several decades, to invest in technical fields expected to be critical for the U.S. economy and national security. Reports by national academies and research bodies encouraged increased funding and opportunities for Americans to pursue education and careers in science, technology, engineering, and mathematics.

Since then, resources allocated to STEM (originally SMET) education have substantially increased, including for many programs specifically aimed at increasing the participation of women in the field. There has been bipartisan support committed to growing employment and progress in these fields.

But there is no single, agreed-upon definition of what counts as a STEM job. As Amanda Olson and co-authors note in a 2014 paper:

“Yet, as part of a research program…we discovered troubling issues surrounding the way that projections are made about the number of STEM jobs available, their expected wages, and the education needed to get these jobs.”

There have been three main approaches to defining STEM:

1. Task-based definitions, grouping together roles with similar duties such as design, research, or teaching.
2. Knowledge-based frameworks, based on required expertise in fields like chemistry, biology, or mathematics.
3. Skill-based definitions, focused on high-level competencies and durable skills such as critical thinking or problem-solving.

These differences produce widely varying estimates of the STEM workforce—from 5.4 million to 29 million jobs—and can complicate efforts to identify skill shortages, design curricula, and allocate resources. Even within a single broad approach, details may differ. For instance, both the Bureau of Labor Statistics (BLS) and the National Science Foundation (NSF) use a task-based taxonomy to classify STEM occupations, but the BLS sometimes includes “management” or “education” under STEM while the NSF distinguishes “science and engineering” (S&E) occupations—often requiring a bachelor’s degree—from “S&E-related” roles that may not.

How then should we define HEAL? Following the various approaches in the STEM field, we explored the same three overall approaches: task-based, skill-based, and knowledge-based.

• In our task-based classification, occupations were categorized based on the specific tasks they involved, as defined by the Standard Occupational Classification (SOC) system, which organizes jobs according to “work performed,” making it a widely used and straightforward approach for workforce analysis.
• In our skill-based classification, we selected occupations based on required aptitudes, skills, and work contexts using Occupational Information Network (O*NET) data to capture competencies like social perceptiveness, service orientation, and assisting and caring for others.
• In our knowledge-based classification, occupations categorized as HEAL were those with above-average requirements in “education and training”, “medicine and dentistry”, or “therapy and counseling”, using standardized O*NET ratings.

Unlike for STEM, the three approaches actually result in a very similar number of occupations and share of the workforce being classified as HEAL, as figure 1 shows.

Figure 1

More results for the three approaches are shared in the appendix, but for our main analysis we use the knowledge-based classification, for two main reasons:

• First, unlike task-based classifications, which group jobs within broad occupational domains, a knowledge-based approach identifies roles where HEAL knowledge is essential, even in occupations not traditionally associated with health, education, or literacy. This results in a broader classification spanning both B.A. and non-B.A. roles, including some that might otherwise be overlooked, such as rehabilitation counselors, clergy, firefighters, and police supervisors.
• Second, a knowledge-based approach offers clearer policy relevance. Knowledge domains can be more directly tied to career tracks and training programs, making them potentially more amenable to education and workforce policies, whereas skill- and ability-based classifications tend to be more abstract.

O*NET measures knowledge requirements for occupations through surveys completed by subject matter experts and job incumbents, evaluating the importance and level of knowledge across 33 standardized domains (e.g., “building and construction,” “public safety and security,” or “psychology”). For this classification, we focused on three key knowledge domains:

• Education and training
• Medicine and dentistry
• Therapy and counseling

Ratings on each of these three were then standardized to identify occupations with significantly higher-than-average knowledge requirements in one or more of these domains. In line with threshold values used by Jonathan Rothwell in his knowledge-based classification of STEM occupations, we classified occupations scoring at least 1.5 standard deviations above the mean in one or more of these domains as HEAL.

For example, radiologists score a 3.70 out of 5 on “education and training,” a 4.67 out of 5 on “medicine and dentistry,” and a 2.14 out of 5 on “therapy and counseling.” These correspond to standardized scores of 1.03, 3.07 and 0.43, respectively. Since radiologists score above 1.5 on one of these domains (“medicine and dentistry” in this case), they are classified as HEAL under our knowledge-based classification.

Our resulting HEAL classification contains 134 occupations, including physical therapists and veterinarians, but also several that might be less obvious, like firefighters. Others, like home health and personal care aides, that might seem like intuitive fits, fall slightly below the required knowledge thresholds.

For comparative purposes, we use the STEM definition adopted by BLS. This definition, which estimates the STEM workforce at 10.7 million, is narrower than some alternatives but ensures no overlap between STEM and our categorization of HEAL occupations.

The interactive table below shows occupations according to their categorization as HEAL, STEM, or other, and the overall size, rate of growth, average salary, typical education level, and male share of each occupation. The table also shows standardized scores on each of the three knowledge domains used in our HEAL definition.

As a note, data limitations require us to use a slightly smaller set of occupations for some estimates. These adjustments, along with more detailed information on the other two approaches, are noted and detailed in the appendix.

Figure 2

HEAL workers: Men missing

In contrast to STEM occupations, HEAL occupations tend to be heavily female, as shown in figure 8. In fact, men account for a similar share of HEAL jobs (22%) as women do in STEM (24%).

Figure 8

There are big potential opportunities here for men to find work. For example, the HEAL profession with the largest expected job growth (see above) is nursing, where men currently account for just 12% of the workforce. Medical and health service managers, with the second largest expected job gains, are just 27% male.

While the shares of men in HEAL and women in STEM are similar, the trends are very different. Note that to track trends, we had to truncate the number of occupations in the analysis, and pool the numbers across several years, so the numbers are not directly comparable to figure 8 above.

The share of STEM jobs occupied by women increased nearly threefold from 1970 to 2018-2022, from 9% to 26% using this definition. Meanwhile the share of men in HEAL roles actually declined slightly.

Figure 9

 
One suggested explanation for the underrepresentation of men in HEAL is that these jobs typically require greater interpersonal skills like social perceptiveness and communication—which, on average, men may be less likely to possess than women. To explore this potential connection, we construct a measure of “human-centered” skills and job requirements and compare it to the male share of an occupation. Of note, this human-centered index is the skill-based option we considered for defining HEAL jobs, (For more details see the appendix.) Note, too, that our analysis here is limited to the 390 occupations and approximately three-fourths of all workers for which the 5-year ACS sample provides sufficient data on male share. Figure 10 shows all the occupations on these two axes. Unsurprisingly, HEAL jobs cluster towards the more human-centered end of the scale, and have a lower male share as noted above.

Figure 10

There are notable exceptions to the general trend, however, including firefighters, police supervisors, and clergy, which all rank highly on human-centeredness but are also predominantly male. This indicates that other factors, such as physical demands, risk, or even simply the historical exclusion of women, may lead to an occupation remaining male-coded despite a greater reliance on interpersonal and relational skills. It may also reflect a uniquely male model of caregiving—one that emphasizes rescuing others from harm.