Chapter 2 | Higher Education in Science and Engineering

The U.S. Higher Education System

The U.S. higher education system consists of many diverse academic institutions that vary in their missions, learning environments, selectivity levels, religious affiliations, types of students served, types of degrees offered, sectors (public, private nonprofit, or private for-profit), and costs (Kena et al. 2016). During the 2015–16 academic year, there were approximately 4,600 postsecondary degree-granting institutions in the United States; about two-thirds (66%) of these offered 4-year or higher degrees, and the remainder offered 2-year degrees (Table 2-1). More than half of the 4-year institutions are private nonprofit, 24% are public, and 23% are private for-profit. Most 2-year institutions are public (58%), but a large proportion (36%) are private for-profit (Table 2-1).

In 2015, U.S. academic institutions awarded nearly 3.8 million associate’s, bachelor’s, master’s, and doctoral degrees, 25% of them in S&E fields (Appendix Table 2-1).^​ Public institutions produced the bulk of S&E and non-S&E degrees (Table 2-2). For example, public institutions awarded nearly 70% of S&E bachelor’s and doctoral degrees and 55% of S&E master’s degrees.

Although relatively few (97), doctorate-granting institutions with very high research activity—public and private—are the leading producers of S&E degrees: these research institutions awarded 72% of doctoral degrees, 42% of master’s degrees, and 37% of bachelor’s degrees in S&E fields in 2015 (Appendix Table 2-1) (see sidebar Carnegie Classification of Academic Institutions). Master’s colleges and universities awarded another 28% of S&E bachelor’s degrees and 25% of S&E master’s degrees in 2015.

Baccalaureate colleges were the source of relatively few S&E bachelor’s degrees (11%) (Appendix Table 2-1), but they produce 13% of future S&E doctorate recipients (NSF/NCSES 2013). When adjusted by the number of bachelor’s degrees awarded in all fields, the top 50 baccalaureate colleges as a group yield more future S&E doctorates per 100 bachelor’s degrees awarded than all other types of institutions except very high research universities.^​

Minority-serving academic institutions (MSIs) can be defined by legislation or by the proportion of minority student enrollment in them (Li 2007). Examples of MSIs established by legislation include historically black colleges or universities (HBCUs, see sidebar Historically Black Colleges and Universities) and tribal colleges or universities^​ (TCUs). Given their legal definition, the number of institutions in these groups cannot increase in number unless Congress acts to designate additional institutions in those groups. In contrast, high-Hispanic-enrollment institutions^​ (HHEs, see sidebar High-Hispanic-Enrollment Institutions: A Typology) are a type of MSI based on the percentage of minority student enrollment. The number of institutions in these groups vary from year to year based on the enrollment of students in their respective minority groups.^​

MSIs enroll a substantial fraction of underrepresented minority undergraduates (NSF/NCSES 2017a). In 2015, HBCUs awarded 16% of the 54,000 S&E bachelor’s degrees earned by black U.S. citizens and permanent residents, and HHEs awarded about 34% of the 79,000 S&E bachelor’s degrees earned by Hispanics. The proportion of blacks earning S&E bachelor’s degrees from HBCUs has been declining in recent years. The proportion of Hispanics earning S&E bachelor’s degrees from HHEs declined through 2011 but has been stable at about 34% since then. Tribal colleges, which mainly offer 2-year degrees, account for about 4% of the nearly 3,000 S&E bachelor’s degrees awarded to American Indians; this proportion has increased slightly in the last 5 years.^​

HHEs and HBCUs also play an important role in training Hispanic and black students for doctoral-level study in S&E fields. Of Hispanics who earned an S&E doctorate between 2011 and 2015, nearly 30% had obtained their baccalaureate at an HHE (Table 2-3). Similarly, 24% of black S&E doctorate recipients had an HBCU baccalaureate (Table 2-4). HBCUs were the second-largest contributor of black S&E doctorate recipients, behind only institutions with very high research activity (NSF/NCSES 2013).

Community colleges (also known as public 2-year colleges or associate’s colleges) play a key role in increasing access to higher education for all citizens. These institutions serve diverse groups of students and offer a more affordable means of participating in postsecondary education. Community colleges prepare students to enter the workforce with certificates or associate’s degrees or to transition to 4-year colleges or universities, often before receiving a 2-year degree. Community colleges tend to be closely connected with local businesses, community organizations, and government, so they can be more responsive to local workforce needs (Olson and Labov 2012).

In the 2015–16 academic year, there were 910 community colleges in the United States, enrolling 6.2 million students, or nearly one-third of all postsecondary students (NCES 2017). Most (62%) community college students enrolled part time. Responding to the economic recession in the late 2000s, enrollment in community colleges peaked in 2010 at 7.2 million but has declined with improving labor markets (Ginder and Kelly-Reid 2017; Ginder, Kelly-Reid, and Mann 2014; Knapp, Kelly-Reid, and Ginder 2009, 2011).

Community colleges play a significant role in educating students who go on to acquire advanced S&E degrees. About 19% of U.S. citizens and permanent residents with S&E doctoral degrees earned between 2011 and 2015 reported having some college credit from a community or 2-year college (Table 2-5). In fact, 47% of all recent S&E graduates had done some coursework at a community college (in 2003, it was 48%, according to the National Survey of College Graduates).^​ Graduates in the biological and social sciences were more likely than those in the physical and computer sciences and in engineering to have attended a community college.

Female S&E bachelor’s and master’s degree recipients were more likely than males to have attended a community college (Table 2-6). Attendance levels as measured by the proportion who took courses at a community college were highest among U.S. citizens, followed by permanent visa holders, and were much lower among temporary visa holders. Among racial and ethnic groups, attendance levels were highest among Hispanics and lowest among Asians. Attendance fell with rising parental education level, illustrating the special access function of these institutions.

Recent S&E graduates (1.3 million) who took courses in community colleges (nearly 600,000) report doing so at different points in their educational careers. Nearly half of them reported doing so after high school but before enrolling in a 4-year college or university or while enrolled in college but before receiving a bachelor’s degree. About one in three used a community college as a bridge between high school and college enrollment. One in five attended a community college after receiving their first bachelor’s degree. One in 10 reported taking courses at a community college after leaving a 4-year college without receiving their first bachelor’s degree.^​

Recent S&E graduates took courses at community colleges for various reasons. The most prevalent reason was to earn credits toward a bachelor’s degree (30%), followed by preparation for college to increase the chance of acceptance at a 4-year institution (17%), for financial reasons (14%), and to earn college credits while still attending high school (13%). Other reasons mentioned included to complete an associate’s degree (6%); to gain further skills or knowledge in their academic or occupational fields (6%); to facilitate a change in their academic or occupational fields (5%); for leisure or personal interest (4%); to increase opportunities for promotion, advancement, or higher salary (3%); and for other reasons (4%).^​

In 2015–16, about 1,300 degree-granting institutions in the United States operated on a for-profit basis; this number peaked at 1,451 in 2012–13 but has declined to 1,262 since then (NCES 2017). Four-year institutions accounted for slightly more than half of these institutions (55%) in 2015–16 (Table 2-1).

For-profit institutions enroll considerably fewer students than public ones, particularly at the 2-year level—nearly 120,000 versus nearly 6.6 million in community colleges in 2015.^​ Enrollment and degrees awarded in for-profit institutions rose dramatically throughout the 2000s but declined in recent years (Appendix Table 2-2).^​

Enrollment patterns differ among racial and ethnic groups. For-profit institutions play a disproportionate role in the education of blacks and Native Hawaiians or other Pacific Islanders, who are more likely than other racial or ethnic groups to enroll in private for-profit academic institutions (NSF/NCSES 2017a).

For-profit academic institutions are not large producers of S&E degrees: they awarded between 3% and 5% of S&E degrees at the bachelor’s, master’s, and doctoral levels, as well as 18% of S&E degrees at the associate’s level in 2015 (Appendix Table 2-2). Computer sciences accounted for three-quarters of the associate’s degrees and nearly half of the bachelor’s degrees awarded by for-profit institutions in S&E fields in 2015 (Appendix Table 2-3). At the master’s level, S&E degrees were mainly in psychology (38%), social sciences (32%), and computer sciences (27%); at the doctoral level, they were almost exclusively in psychology (79%) and social sciences (17%).

Distance and online education enable institutions of higher education to reach a wider audience by expanding access for students in remote locations while providing greater flexibility for students who face time constraints, physical impairments, responsibility to care for dependents, and other challenges. Distance education has been around for more than 100 years (Perna et al. 2014), whereas online education is a relatively new phenomenon. Online education can serve individuals’ needs for lifelong learning and skill retooling during times of rapid technological change.

Distance education uses technology to deliver instruction to students who are separated from the instructor and to support regular and substantive interaction between the students and the instructor, synchronously or asynchronously (Kena et al. 2016). Distance education enrollment has grown in recent years, given the growth of Internet technologies to deliver content. According to nationally representative data from the Integrated Postsecondary Education Data System (IPEDS) 2015 Fall Enrollment survey, 14% of all students in 4-year Title IV institutions (i.e., institutions that participate in federal financial aid programs) were enrolled exclusively in distance education courses, and another 15% were enrolled in distance education and regular on-campus courses; whereas the remaining 71% of these students were not enrolled in any distance education course (Table 2-7).^​ Exclusive enrollment in distance education courses was considerably higher at private for-profit 4-year institutions than at either 2- or 4-year public or private nonprofit institutions or at private for-profit 2-year institutions. Enrollment in some distance education courses was highest at public institutions. Exclusive enrollment in distance education courses was higher at the graduate level than at the undergraduate level, whereas enrollment in some distance education courses was higher at the undergraduate level than at the graduate level.

Nationally representative data collected by the 2015 IPEDS Completions Survey also show that, regardless of the degree level, the proportion of distance education programs in S&E was highest at private for-profit 4-year institutions, ranging from nearly 30% of the S&E programs in these institutions at the associate’s level to more than two-thirds of those at the master’s level (Appendix Table 2-4). In general, computer sciences and psychology were the two fields where distance education programs were most prevalent, irrespective of institution type and degree level. In addition, engineering, engineering technologies, health technologies, and social sciences fields also had considerable utilization of distance education programs. (Between 18% and 25% of the master’s programs in engineering, engineering technologies, and health technologies at public 4-year institutions and the majority of social sciences programs at private for-profit 4-year institutions had distance education.)

A recent study provided evidence that at a for-profit university with an undergraduate enrollment of more than 100,000 students where most of them were pursuing bachelor’s degrees, taking a course online instead of in-person reduced student success and progress in college. Grades were lower not only in the course students took online but also in future courses. In addition, students who took a course online were less likely to remain enrolled a year later (Bettinger et al. 2017).

Allen et al.’s (2016) most recent survey showed that a small segment of higher education institutions had massive open online courses (MOOCs; see Glossary) (11%) or were planning one in 2015 (2%); however, most institutions decided against having a MOOC (59%) or remained undecided about it (28%). MOOCs can provide broad access to higher education for free or at a very low cost, facilitating lifelong learning and continuing education. Through their online platforms, MOOCs also have the potential to collect massive amounts of information that can be used to conduct experimental research on how people learn and to identify online practices that improve learning (ED/OET 2013).

Nationally representative data on MOOCs are not available. However, research conducted on the first 4 years of open online courses offered by HarvardX and MITx on the edX platform reveals that during that time, the platform included 290 courses, granted 245,000 certificates (including free and paid certificates), and had 4.5 million participants (Chuang and Ho 2016).^​ The survey of MOOCs showed that participants’ median age was 29, two-thirds of them were males, 71% were from countries other than the United States, and 73% were bachelor’s degree holders. The largest MOOCs were in computer sciences.

Overall completion rates in MOOCs are low; however, they varied according to participants’ intentions at the start of the course. Some MOOC participants indicated that they intended to obtain a free certificate, others reported that they were exploring a subject, and others reported paying in order to verify their identity and obtain a formal certificate. Students who paid for a certificate verifying their completion of the MOOC were much more likely to obtain a certificate than those who took a class that offered a free certificate (60% compared with 8%).

Online education companies offering MOOCs have also expanded their offerings of certificate programs. For instance, Udacity partnered with AT&T to offer technology-focused “nanodegrees” teaching students a specific set of skills that can be applied to a job. These courses have been developed in partnership with employers. For example, Udacity developed a course on Android technology with Google and another on self-driving car engineering with Mercedes-Benz, NVIDIA, and Otto (The Economist 2017). For students, these courses are much more affordable than attending a college or university and provide the flexibility they need to complete them while balancing other family and job responsibilities. For businesses, these types of classes provide a quick response to market demand for niche technological specializations.

In 2014, the Georgia Institute of Technology (Georgia Tech), in collaboration with Udacity and AT&T, began to offer an online master’s program in computer science, which combines MOOC-like course videos and assessments with a support system that works directly with students. The university’s goal was to create a master’s degree program that was just as rigorous as the one offered on campus but at a much lower cost. A recent study focusing on the students who applied to this program showed that access to this online option increased overall enrollment in higher education, rather than substitute for the brick-and-mortar university options (Goodman et al. 2016). The researchers found that online students in this program were older than students in the on-campus program and that the vast majority of them were employed. They also found that the demand for Georgia Tech’s online degree satisfied previously unmet demand for mid-career training and could increase the production of computer sciences master’s degrees in the United States. Overall, their results also suggested that high-quality online education may open opportunities for people who otherwise would not be pursuing a degree.

Changing modes of online education are prompting questions about how the use of this technology will affect the higher education sector. In particular, it is not yet clear how many students can sustain commitment to learning in the absence of more personal contact and to what extent the growing access to higher education facilitated by MOOCs will translate into learning and, in the long run, to higher levels of educational achievement. It is also not clear how these models can be applied in a wider range of disciplines and higher education institutions.

Higher education spending and revenue patterns have changed substantially since 2000, in trends that intensified during the economic downturn of the late 2000s. Although all types of higher education institutions faced competing demands in a stringent budget environment, each type faced unique challenges. Through 2010, increases in the number of students seeking an affordable college education compounded the challenges created by tight budgets. Despite declines in enrollment between 2011 and 2015 (Appendix Table 2-5), the same challenges have remained. This section shows trends in inflation-adjusted average spending and revenue per full-time equivalent (FTE) student from 2000 to 2015,^​ based on data from the Delta Cost Project.^​

Net tuition and federal appropriations, grants, and contracts are two large sources of revenues for public and private very high research institutions (Appendix Table 2-6).^​ For public institutions, state and local appropriations are also critical, supplying an amount of revenue just under three-quarters of net tuition ($9,200 per FTE in 2015); in contrast, they are a small source of revenue for their private counterparts (about $1,100 per FTE in 2015 and only about 4% of net tuition). Much more important for private institutions are private and affiliated gifts, investment returns,^​ and endowment income, which are usually the largest sources of revenue other than funds from hospitals and other independent operations.^​

State and local appropriations for public very high research universities have declined since 2000, with a particularly steep drop between 2008 and 2012 (Figure 2-1). This decline coincided with a compensating increase in net tuition. In 2000, average state appropriations per FTE at public very high research institutions were more than twice the amount of net tuition ($13,900 versus $6,500). By 2015, however, appropriations had dropped to $9,200 per FTE, whereas net tuition had increased from about $6,500 to more than $12,700 per FTE (Appendix Table 2-6). This change represents a downward shift in higher education investment by state and local governments, resulting in a higher financial burden for individual students and their families. Starting at a higher level, net tuition at private very high research universities also increased during this 15-year period. But the increase, from about $22,700 to almost $27,700, was proportionally much smaller.

Revenue from federal appropriations, grants, and contracts, the source used for most research expenditures, is highest at the most research-intensive universities (Appendix Table 2-6), particularly the private ones. These revenues increased steadily from 2000 to 2005, dipped as the economy entered the recession at the end of the decade, increased somewhat with American Recovery and Reinvestment Act (ARRA) funding, then dipped again between 2011 and 2015. Between 2000 and 2015, revenue per FTE from these funds increased by 11% at public very high research institutions to just under $8,000 per FTE and by 14% to $25,700 per FTE at their private counterparts.

Research and instruction are the two largest core education expenditures at public and private very high research universities. Between 2000 and 2015, research expenditures per FTE increased substantially at both types of institutions—by 25% at private universities and by 11% at their public counterparts (Figure 2-2; Appendix Table 2-7). For public and private institutions, research expenditures per FTE peaked in 2011 (coinciding with the year of greatest ARRA research spending); since then, they have declined by about 8%. See Chapter 5 section Academic R&D, by Public and Private Institutions for greater detail on university research spending.

Instructional spending per FTE followed a pattern similar to that of research expenditures, increasing at a higher rate at private very high research institutions than at their public counterparts. Between 2000 and 2015, instructional expenditures per FTE increased by 43% at private universities compared to 18% at public universities. Moreover, for the past decade, instructional spending at private very high research universities has been three times that of the public universities (Figure 2-3).

From 2000 to 2015, state and local appropriations and net student tuition were the largest sources of revenues centrally involved with education at other public institutions offering 4-year and graduate degrees (Appendix Table 2-6).^​ At these institutions, total revenues from these two sources were lower than those at public very high research universities. In 2015, net student tuition per FTE was higher at public 4-year institutions than at community colleges but state and local appropriations per FTE were lower. From 2000 through 2015, the percentage drop in revenue per FTE from state and local appropriations (25%) was somewhat less than that experienced at the public very high research institutions (34%). In 2010, net student tuition replaced state and local appropriations as the largest source of revenue in the public 4-year institutions. Average state appropriations per FTE in 2000 ($8,800) were almost twice as large as tuition revenue ($4,600). By 2010, average revenues from net student tuition, at $7,200 per FTE, exceeded average revenues from state appropriations per FTE by about $450. By 2015, average revenues from net tuition increased even further, to more than $1,500 over the average revenues from state appropriations (Figure 2-4). As in the case of public very high research institutions, this change represents a shift in financial investment from state and local governments to individual students and their families.

Spending on instruction at 4-year and other graduate public institutions has been at least three times as high as almost all the other standard expense categories. It increased from an average of nearly $7,000 per FTE in 2000 to about $7,800 per FTE in 2015 (Appendix Table 2-7). Other expenditures represented much smaller shares of total spending; most of these expenditures increased, with average increases between 2000 and 2015 ranging from 5% for spending on plant operation and maintenance to 28% for student services.

Revenues are much lower for community colleges than for other public institutions of higher education, particularly public very high research institutions.^​ As in the other public institutions, the main sources of revenue at community colleges are state and local appropriations and net student tuition (Appendix Table 2-6). In 2015, average revenues from state and local appropriations at community colleges were about $7,200 per FTE, compared with about $9,200 at public very high research institutions; average revenues from net tuition were about $4,100 per FTE, compared with about $12,700 at public very high research institutions. Unlike other public institutions, revenue from state and local appropriations at community colleges still exceeded net tuition revenue in 2015.

Even so, community colleges have experienced the same decline in state and local government support that other public institutions have seen. Between 2000 and 2015, revenues from state and local appropriations at community colleges decreased from an average of about $7,700 per FTE to $7,200 per FTE, with a steady decline from 2008 to 2012. This trend has since begun to reverse, although state and local support remain below their prerecession levels (Figure 2-5). As state support declined from 2008 to 2012, revenues from net tuition increased by 17%. In 2000, revenues from state and local appropriations represented 56% of total revenues at community colleges, and tuition accounted for 18%. By 2015, state and local appropriations had dropped to 48% of total revenues, whereas the proportion of revenues from tuition increased to 28% of total revenues.

Expenditures are also much lower for community colleges than for other public institutions of higher education. In community colleges, instruction is by far the largest expenditure (Appendix Table 2-7). In 2000, spending on instruction was about $5,900 per FTE, about 40% of total expenditures. In 2015, average instructional spending per FTE ($6,000) was very similar in size to the 2000 level. Overall, these expenditures went up and down between 2000 and 2015, declining from 2001 to 2005 and 2008 to 2012 but increasing during other years (Figure 2-5).^​ Expenditures on student services and institutional and academic support declined in the late 2000s but increased somewhat in 2012–15. Expenditures in plant operation and maintenance also declined between 2008 and 2011 and have risen slightly since then. As a percentage of total expenditures, each spending stream remained relatively constant from 2000 through 2015.

Between 2000 and 2015, revenues from state and local appropriations and net tuition, the main two revenue sources at public institutions, when added together increased by similar amounts at community colleges (10%) and 4-year institutions (9%); they increased a little less at very high research institutions (8%). States and localities cut funding for all three categories of institutions, but the reduction was smaller in the community colleges (7%) than in the public very high research institutions (34%) and the public 4-year and other graduate public institutions (25%). Unlike community colleges, however, the other two types of public institutions were able to increase revenues from net tuition to a greater extent. FTE net tuition revenues increased by 97% at the public very high research universities and by 75% at the 4-year and other graduate public institutions, compared with 62% at community colleges (Appendix Table 2-6).

Instruction expenditures followed a different pattern. They rose most rapidly at the public very high research institutions (18%), where there was pressure to keep faculty salaries (a major component of instructional expenses) competitive with those of their private counterparts, which spent more on instruction to begin with and were increasing these expenses even more rapidly (43%) (Appendix Table 2-7). At community colleges, FTE instructional expenses increased by 2% over the period from 2000 to 2015,^​ whereas in 4-year and other graduate institutions, they increased by 12%. Overall, during this period, community colleges had more limited resources and less flexibility to draw on alternate revenue sources to support their instructional expenses. However, given the decline in enrollment in fall 2012 through fall 2015 after the recession, average expenditures in instruction increased more substantially (14%) at community colleges and in 2015 were at their highest level since 2001 (see section Undergraduate Enrollment in the United States).

Affordability and access to U.S. higher education institutions are continuing concerns (Sullivan et al. 2012; GAO 2014). According to the College Board (2016a), the estimated average net tuition and fees (i.e., the published prices minus grant aid and tax benefits) vary by institution type.

In the last 10-year period ending in 2016–17, net tuition and fees paid by full-time, in-state undergraduate students in public 4-year colleges increased by about 30% in constant 2016 U.S. dollars (College Board 2016a; Table 2-8). Net tuition and fees at these institutions had dipped during the recessionary period between 2007–08 and 2009–10, but they increased by 70% since then and nearly 10% in the last 2 years.

At private nonprofit institutions, net tuition and fees followed a similar path in the last 10 years, declining between 2007–08 and 2011–12 but rising since then, gradually approaching its highest point 10 years earlier.

At public 2-year colleges, net tuition and fees have overall declined by more than 200% in the last 10 years, but they have increased by about 35% since 2011–12 (Table 2-8). On average, since 2009–10, undergraduate students enrolled full time at public 2-year colleges have received enough funding through grant aid and federal education tax credits and deductions to cover tuition and fees, and they can use the rest of those funds to cover books or living expenses (their net tuition was –$500 in 2016–17) (College Board 2016a). Despite large percentage tuition increases in public institutions, they are still more affordable than their private counterparts.

Between 1999–2000 and 2011–12, changes in the net cost of higher education for dependent undergraduates varied by family income level and type of institution they attended (Table NSB 2016 2-9; the NCES National Postsecondary Student Aid Study [NPSAS] is conducted every 4 years, so there are no new data). For students from higher-income families, net tuition and fees increased across all types of institutions. Students from lower-income families experienced declining or stable net tuition in certain types of institutions while seeing increases in others. (On average, these students experienced declines at public 2-year institutions; saw no changes at public and private nonprofit 4-year master’s and baccalaureate institutions, as well as at private nonprofit 4-year research and doctoral institutions; and saw a rise at public 4-year research and doctoral institutions.)

Research suggests that the vast majority of low-income, high-achieving high school seniors do not apply to any selective college, although selective institutions cost them less than nonselective ones because of the large amounts of financial aid they are able to offer (Hoxby and Avery 2013).^​

With rising tuition, students increasingly rely on financial aid to fund their education. Financial aid for undergraduate students comes mainly in the form of student loans (federal and nonfederal), grants (federal, state, institutional, and private), and tuition tax credits. A financial aid package may contain one or more of these kinds of support. In 2016–17, undergraduate students received $184 billion in federal, state, institutional, and other aid, excluding nonfederal loans (College Board 2016b).

In the last 10 years, federal financial aid has constituted about two-thirds of the undergraduate student aid package; federal loans have been the main component, followed by federal grants, although the proportion of undergraduate students receiving federal loans declined (from 42% in 2005–06 to 33% in 2016–17), whereas the proportion receiving federal grants increased (from 18% to 23%). In addition, institutional grants increased (from 20% to 23%), and private and employer grants and state grants rose slightly as well (6% versus 7% in both cases).

According to the latest data available from the NPSAS, a higher proportion of undergraduates in private institutions than those in public institutions received some type of financial aid and incurred student loans (Ifill and Shaw 2013).^​

Among recent graduates with S&E bachelor’s degrees, the level of undergraduate debt does not vary much by undergraduate major, although it is somewhat lower for recent recipients of engineering bachelor’s degrees than for recent recipients of bachelor’s degrees in social and related sciences and in physical and related sciences.^​

Levels of debt vary to a greater extent by type of institution. The extent of undergraduate indebtedness of students from public colleges and universities is almost as high as that for students from private nonprofit universities (about 60% at graduation). The level of debt differs, however: about $26,800 per borrower for those graduating from a public institution and $31,400 for those graduating from private nonprofits. Students who attend private for-profit institutions are more likely to borrow, and to borrow larger amounts, than those who attend public and private nonprofit institutions (College Board 2016b).

Levels of debt varied widely by state. Average debt for 2014 graduates of public 4-year colleges and universities ranged from $18,800 in New Mexico to $35,000 in New Hampshire. Average debt for graduates of private nonprofit colleges and universities ranged from $8,900 in Alaska to $36,200 in Connecticut (Institute for College Access & Success, College InSight 2016). Cost of living may account for some of the differences by state.^​

In 2015, nonfederal funds were the main source of funding of full-time S&E graduate students (45%), followed by self-support (41%), and federal funds (15%) (Appendix Table 2-8). Nonfederal sources include state funds and funding from universities, employers, nonprofit organizations, and foreign governments. Particularly in the large public university systems, state funds are affected by the condition of overall state budgets. Self-supporting graduate students rely primarily on loans, their own funds, or family funds for financial support.

The number of full-time graduate students supported primarily by nonfederal sources or through self-support has increased in the last 15 years, with the steepest increase in 2014 (Figure 2-6).^​ The proportion of self-supporting graduate S&E students gradually rose from 33% to 41% between 2000 and 2015, primarily because of increasing enrollment of master’s students on temporary visas who are mostly self-supporting (IIE 2016; NSF/NCSES 2016).^​ Self-support was highest (60% or higher) among full-time graduate students in computer sciences and in medical and other health sciences (Appendix Table 2-9).

The number of full-time S&E graduate students supported by the federal government increased between 2000 and 2004 and was fairly stable through 2010, but it declined by 16% in the last 5 years, with the steepest decline between 2011 and 2014 (Appendix Table 2-8). Between 2000 and 2006, the proportion of full-time S&E students primarily supported by the federal government remained fairly stable at 20%–21% but has declined since then, reaching its lowest level in at least 16 years in 2015 (15%) (Appendix Table 2-10). This decline was more pronounced in the biological, physical, and medical sciences (Appendix Table 2-10).

The federal government plays a substantial role in supporting full-time S&E graduate students in some fields but a smaller role in others. Federal financial support for graduate education reaches a larger proportion of students in the physical sciences; the biological sciences; the earth, atmospheric, and ocean sciences; and engineering (Figure 2-7; Appendix Table 2-11). For some mechanisms of support, the federal role is fairly large. In 2015, the federal government funded 55% of full-time S&E graduate students who were on traineeships, 45% of those with research assistantships (RAs), and 22% of those with fellowships (Appendix Table 2-11).

Teaching assistantships (TAs) are generally institutionally funded. Most graduate students, especially those who pursue doctoral degrees, are supported by more than one source or mechanism during their time in graduate school, and some receive support from several different sources and mechanisms in any given academic year. Primary mechanisms of support differ widely by S&E field of study (Figure 2-8; Appendix Table 2-9). In 2015, full-time graduate students in physical sciences were financially supported mainly through TAs (40%) and RAs (37%). RAs were also important in agricultural sciences (47%); earth, atmospheric, and ocean sciences (36%); biological sciences (33%); and engineering (31%; in particular, in materials and chemical engineering). In mathematics and statistics, nearly half (47%) of the full-time students were supported primarily through TAs.

Most federal financial support for graduate education is in the form of RAs funded through grants to universities for academic research. RAs are the primary mechanism of support for 71% of federally supported full-time S&E graduate students (Appendix Table 2-8). Fellowships and traineeships are the means of funding for 22% of the federally funded full-time S&E graduate students. For students supported through nonfederal sources in 2015, TAs (i.e., institutional funds) were the most prominent mechanism (40%), followed by RAs (29%).

NSF and the National Institutes of Health (NIH) support most of the full-time S&E graduate students whose primary support comes from the federal government, followed by the Department of Defense (DOD) (Appendix Table 2-12). In 2015, NSF supported about 23,000 S&E graduate students, NIH about 21,000, and DOD about 8,000. Trends in federal agency support of graduate students show considerable increases from 2000 to 2015 in the proportion of students funded by NSF, from 22% to 32% (Appendix Table 2-12). NSF supported 58% of students in computer sciences or mathematics whose primary support comes from the federal government; 51% of those in earth, atmospheric, and ocean sciences; 42% of those in the physical sciences; and 39% of those in engineering overall (about 49% of those in electrical engineering and 48% of those in chemical engineering) (Appendix Table 2-13). The proportion of students funded by NIH increased from 29% to 33% between 2000 and 2008 but has since decreased to 29%. In 2015, NIH funded about 70% of such students in the biological sciences, 57% of those in the medical sciences, and 36% of those in psychology. The proportion of graduate students supported by DOD has been relatively stable around 10%–12% in the last 15 years. In 2015, DOD supported 47% of the S&E graduate students in aerospace engineering, 31% of those in industrial engineering, 27% of those in electrical engineering, and 22%–23% of those in materials and mechanical engineering and in computer sciences.

For doctoral degree students, notable differences exist in primary support mechanisms by type of doctorate-granting institution (Table 2-9). In 2015, RAs were the primary support mechanism for S&E doctorate recipients from research universities (i.e., doctorate-granting institutions with very high research activity, that receive the most federal funding, and those with high research activity). For those from medical schools, which are heavily funded by NIH, fellowships or traineeships accounted for the main mechanism of support. Students at less research-intensive universities relied mostly on personal funds.

Notable differences also exist in primary support mechanisms for doctoral degree students by sex, race and ethnicity, and citizenship (Appendix Table 2-14). In 2013–15, among U.S. citizens and permanent residents, male S&E doctorate recipients were more likely than their female peers to be supported by RAs (31% compared with 22%). Female S&E doctorate recipients were more likely than their male counterparts to receive fellowships or traineeships (28% versus 24%) and to support themselves from personal sources (18% versus 10%). Also, Asians were more likely than any other racial or ethnic group to have primary RA support (32%), followed by whites (28%). Compared with other racial and ethnic groups, Hispanic and American Indian or Alaska Native S&E doctorate recipients depended more on fellowships or traineeships (34% and 38%, respectively), and blacks and American Indians or Alaska Natives were more likely to use personal sources (28% and 19%, respectively). S&E doctorate recipients on temporary visas were more likely to have an RA (51%) than their U.S. citizen and permanent resident peers (27%); this has been a long-standing pattern. S&E doctorate recipients who were temporary visa holders were also less likely than U.S. citizens and permanent residents to use personal funds.

To some extent, the sex, citizenship, and racial and ethnic differences in types of support mechanisms are related to differences in field of study. White and Asian men, as well as international doctoral degree students, are more likely than white and Asian women, along with underrepresented minority students of both sexes, to receive doctorates in engineering and physical sciences (see Appendix Table 2-14), fields that are largely supported by RAs. In turn, women and underrepresented minorities are more likely to receive doctorates in social sciences (except for economics) and psychology, in which self-support is prevalent. However, some differences in type of support by sex, race and ethnicity, or citizenship remain after accounting for these doctoral field patterns. In 7 out of the 10 broad S&E fields presented in Appendix Table 2-14, men were more likely than women to have had RA as primary sources of support during their doctoral studies. In contrast, in 7 out of the 10 broad S&E fields, women were more likely to have used personal funds as a source of support. When looking at race and ethnicity patterns in primary source of support among U.S. citizen and permanent residents, in 8 out of the 10 S&E broad fields, Asians and whites were more likely to have used RA as primary source of support than underrepresented minorities. Underrepresented minorities (blacks in particular) were more likely than Asians and whites to have used personal funds as primary source of support in all the broad S&E fields (Appendix Table 2-14).

Overall, the variation in the use of RAs and personal funds as primary sources of support among doctorate recipients was also largely visible at very high research intensive institutions.^​

At the time of doctoral degree conferral, 43% of 2015 S&E doctorate recipients had debt related to their undergraduate or graduate education. In 2015, 29% of S&E doctorate recipients reported having undergraduate debt and 32% reported having graduate debt. For some S&E doctorate recipients, debt levels were high, especially for graduate debt: 6% reported more than $40,000 of undergraduate debt, 13% reported more than $40,000 of graduate debt, and 18% reported more than $40,000 in cumulative undergraduate and graduate debt (Appendix Table 2-15).

Levels of debt vary widely by doctoral field. A higher percentage of doctorate recipients in non-S&E fields (52%) than those in S&E fields (32%) reported graduate debt. In 2015, within S&E, high levels of graduate debt were most common among doctorate recipients in the social sciences, psychology, and the medical and other health sciences. The proportion of doctorate recipients in these fields who reported graduate debt has increased since 2003.^​ Psychology doctorate recipients were most likely to report having graduate debt and high levels of debt.^​ In 2015, 27% of doctorate recipients in psychology reported graduate debt of more than $70,000 (Appendix Table 2-15). Doctorate recipients in mathematics, computer sciences, and physical sciences were the least likely to report graduate debt.

Men and women differed little in level of undergraduate debt, but women were more likely to have accumulated higher graduate debt. U.S. doctorate holders accumulated more debt than temporary visa holders. Regardless of broad field of doctorate, among U.S. citizen and permanent resident doctorate recipients with graduate-school debt, blacks, Hispanics, and American Indians and Alaska Natives were more likely to have debt over $30,000 than Asians and whites (NSF/NCSES 2017b, table 41). In all broad fields of study, blacks were more likely to have reported graduate-school debt higher than U.S. $30,000, followed by Hispanics and American Indian or Alaska Natives. In contrast, Asians and whites were the least likely racial group to report more than U.S. $30,000 in graduate-school debt. A higher level of graduate-school debt among underrepresented minority doctorate recipients than among their Asian and white counterparts, in all broad fields of study, was also observed at very high research intensive institutions.^​