feat_id
stringlengths 9
16
| text
stringlengths 1.23k
351k
| target
stringlengths 0
9.23k
| evaluation_predictions
sequence |
|---|---|---|---|
GAO_GAO-18-49 | Background Characteristics of Postsecondary Institutions In fall 2015, almost 20 million students were enrolled in over 4,500 2- and 4-year postsecondary institutions, according to IPEDS data. Postsecondary institutions vary in terms of their funding, the length and type of programs offered, and instructional mission, among other characteristics. Public institutions, which include state universities and community colleges, are traditionally supported by federal, state, and local funds, in addition to revenue from tuition and fees. Private, not-for- profit schools are owned and operated by independent or religious organizations, and their net earnings do not benefit any shareholder or individual. Tuition and fees as well as other revenue sources primarily support these schools. For-profit institutions are privately owned and earnings can benefit shareholders or individuals. Two-year institutions often provide career-oriented programs at the certificate and associate’s degree levels. Four-year institutions tend to have a broad range of instructional programs at the undergraduate level leading to bachelor’s degrees. Many 4-year institutions also offer master’s or doctorate level programs, and some 4-year institutions have a research focus. The landscape of postsecondary institutions has changed over the past 20 years, particularly with respect to for-profit institutions. The number of public institutions remained relatively constant and the number of private institutions declined slightly; however, the number of for-profit institutions more than tripled between 1995 and 2011 before declining slightly to 2015 levels (see fig. 1). How National Data Count Faculty IPEDS and CPS both provide data on postsecondary faculty. IPEDS data can provide information on positions filled by different types of faculty across postsecondary education or by types of institutions (see sidebar for how we categorize institutions using IPEDS data). In terms of faculty types, IPEDS distinguishes between tenure-track and contingent positions and also has data on graduate assistants, though we cannot determine whether these graduate teaching assistants are the instructors of record for courses or are instead providing classroom support (e.g., grading, leading discussions, and lab setup). Because IPEDS counts positions, any faculty who teach at more than one institution are counted multiple times—for each position they fill. CPS counts the number of actual workers in a given occupation and, in terms of faculty, provides data on how many individuals are employed as postsecondary teachers in colleges and universities nationwide. CPS does not differentiate faculty by type of institution or by tenure status. For example, CPS cannot identify full-time contingent faculty separately from full-time tenure-track faculty. Contingent Faculty Fill Most Instructional Positions Nationwide and Teach Close to Half or More of All Courses at Public Institutions in Three Selected States From 1995 to 2011, the Number of Instructional Positions Filled by Contingent Faculty More than Doubled While Those Filled by Full-Time Tenure- track Faculty Increased By 10 Percent According to IPEDS data, from 1995 to 2011, the percentage of postsecondary instructional positions filled by contingent faculty increased from 57.6 to 71.6 percent. During this period the number of instructional faculty positions at all institutions nationwide grew by over 60 percent— though most of this growth was among positions held by contingent faculty. More specifically, the number of positions held by full-time and part-time non-tenure-track faculty—which we define as contingent—both doubled during this period, while the number of positions held by full-time tenure-track faculty grew by about 10 percent (see table 1). In addition to full- and part-time contingent faculty, some graduate assistants may also teach courses. During the same period, the number of graduate teaching assistant positions grew by 63.8 percent. Some of the increase in the percentage of contingent faculty positions is due to the growth of the for-profit sector and growth among 2-year institutions, which as a whole rely primarily on contingent faculty. For example, the number of positions nationwide across for-profit institutions in 2011 was almost 9 times as many as in 1995. However, the shift towards contingent faculty positions was clear even among only 4-year public and private institutions (see fig. 2). Contingent Faculty Fill about 70 Percent of Instructional Positions Nationwide, Though This Varies Greatly by Institution and Many of These Positions Have Some Job Stability Contingent faculty currently fill most instructional positions nationwide, though these numbers cannot be compared to historical data. According to 2015 IPEDS data, contingent faculty fill 69.5 percent of the 1,444,774 postsecondary instructional positions across all institutions nationwide, including about 61.4 percent of instructional positions at 4-year institutions, 83.5 percent at 2-year institutions, and 99.7 percent at for- profit institutions (see fig. 3). As noted previously, aggregated IPEDS data count faculty who teach at multiple institutions multiple times; therefore, there are likely more contingent faculty positions than there are contingent faculty workers. Although it is unknown how many faculty hold jobs at multiple institutions, this is likely to be more prevalent among faculty filling part-time positions. To illustrate, according to CPS data— which counts individuals—an estimated 31.7 percent (+/- 4.1) of individuals employed as postsecondary teachers in colleges and universities worked part-time in 2015. In contrast, according to IPEDS data, part-time faculty held about 50.0 percent of instructional positions. Though the majority of instructional faculty positions across institutions are contingent, employment stability among these positions may vary widely. Many of these contingent positions may have some job stability, depending on contract specifics. For example, about a quarter of contingent positions across all institutions have full-time, annual, multi- year, or potentially pseudo-tenure contracts (see fig. 3). Some of these positions may expire at the end of a set term or have no option for renewal—potentially requiring a new application process—while others may be relatively long-term with continuously repeating contracts. For example, officials at one North Dakota institution we visited described their non-tenure-track positions as “tenure light” because full-time faculty receive 1-year contracts for their first 4 years and then, after a successful promotion review, receive continuous 3-year contracts that can be terminated only for adequate cause, such as gross professional misconduct. In contrast to these more stable contingent positions, more than half of the contingent positions across all institutions nationwide are part-time and have less-than-annual contracts or lack faculty status— which we define as being among the least stable (see fig. 3). For some of the faculty filling these positions, this employment may be their sole source of income. Similar to contingent workers in the broader labor force, as we reported previously, these faculty may face volatility and uncertainty in their economic circumstances. Other faculty in these positions may have employment or sources of income outside of teaching. For example, some part-time instructors are employed full-time in their fields and teach on the side as subject-matter experts or to stay connected with their local university community. Examples of Part-Time Faculty Situations from Faculty Discussion Groups at Selected Institutions Two part-time faculty members at an institution in Ohio said they had jobs outside of teaching and said they teach on the side because they love it, rather than relying on it for subsistence. One part-time faculty member at an institution in Georgia said that she was retired, but teaches courses to keep a foot in the education world while also enjoying free time in retirement. One younger part-time faculty member at an institution in North Dakota stated that she teaches on a semester-to-semester contract and that this was her primary employment. While it is unknown how many faculty rely on their instructional positions as their primary employment, nationally representative data from the Current Population Survey (CPS) and Survey of Doctorate Recipients (SDR) provide some limited information that suggests many part-time faculty prefer working part-time. The CPS data show that an estimated 46.2 percent (+/- 6.3) of part-time faculty reported wanting to work part- time, while only 10.0 percent (+/- 5.1) reported working part-time because they could only find a part-time job or because of seasonal or temporary fluctuations in the availability of employment. Similarly, SDR data on doctorate-holding instructional faculty in STEM (science, technology, engineering, and math), health, and social sciences fields show that most part-time contingent faculty report wanting to work part-time, though among those who reported wanting a full-time job, most reported not being able to find one (see table 2). According to IPEDS data, different types of postsecondary institutions rely more heavily on different segments of the instructional workforce. As shown in figure 4, many 4-year institutions employ tenure-track, full-time contingent, and part-time contingent positions—though the balance varies. Far fewer 2-year institutions and very few for-profit institutions have tenure-track positions. Part-time and short-term positions are substantially more prevalent at these institutions. For example, part-time contingent positions make up 67.9 percent and 80.5 percent of instructional positions at 2-year and for-profit institutions, respectively, as compared to 39.8 percent at 4-year institutions. Beyond institution type, reliance on different types of faculty positions also varies by institutional characteristics, such as size and highest degree offered. For example, across 4-year institutions with more than 10,000 students, 43.1 percent of positions are tenure-track, as compared to 30.6 percent across institutions with fewer than 5,000 students. Similarly, a higher percentage of instructional positions are tenure-track across 4-year institutions that offer doctorate degrees, compared to those institutions that do not offer doctorate degrees (see fig. 5). At 4-Year Public Institutions in Three Selected States, Contingent Faculty Teach Close to Half or More of All Courses and Credit Hours Contingent faculty fill more than half of instructional positions at 2- or 4- year public institutions in the three selected states (see fig. 6). Two-year public institutions in North Dakota and Ohio were especially reliant on contingent faculty, where they fill about 72 and 84 percent of instructional positions, respectively (see sidebar for our definition of instructional faculty in the state data, as compared to our other data analyses). We examined several different demographic characteristics of contingent faculty including gender, race, educational attainment, and age. Gender According to 2015 IPEDS data, instructional positions nationwide are divided roughly evenly between the sexes, but women fill fewer tenure- track positions and more contingent positions than men do. As shown in figure 7, across all institutions, women hold a substantially lower proportion of full-time tenured positions (38.4 percent) than men do, though women fill 48.9 percent of full-time positions that are on a tenure track but not yet tenured, and that are generally more recent hires. Across all institutions, women also hold a slightly greater proportion of contingent positions (about 53 percent). This imbalance in representation, in part, reflects the higher concentration of women at 2-year and for-profit institutions, where they fill 54.3 and 55.9 percent of positions, respectively. These institutions generally rely more heavily on contingent faculty positions than do 4-year institutions. White (non-Hispanic) faculty fill almost three-quarters of instructional positions across all institutions nationwide. This racial/ethnic representation is relatively consistent across full-time tenure-track, full- time contingent, and part-time positions. Though filling 27.6 percent of positions across all institutions, racial and ethnic minorities have slightly greater representation at institutions in large cities (33.2 percent) and at for-profit institutions (38.4 percent). Educational Attainment Our analysis of state data suggests that across 4-year public institutions in North Dakota and Ohio, lower proportions of individuals in contingent positions have a graduate or doctoral degree (see fig. 8). While the differences between tenure-track and contingent faculty are substantial, possible explanations include variation in degree requirements by discipline or individual circumstances, such as having professional experience in the field. Across public institutions in all three selected states, and excluding positions held by instructional graduate students, most positions held by the youngest faculty are contingent, and the most common positions held by the oldest faculty are part-time contingent. More specifically, most positions held by individuals under age 40 are contingent—60.2 percent in Georgia, 66.9 percent in North Dakota, and 74.5 percent in Ohio (excluding instructional graduate assistants). This suggests that newer graduates may be more likely to be hired into contingent rather than tenure-track positions. In addition, the most common positions held by faculty ages 70 and older are part-time contingent positions—51.0 percent in Georgia, 45.5 percent in North Dakota, and 59.4 percent in Ohio (excluding instructional graduate assistants). This suggests that a segment of the part-time contingent workforce may consist of retirees or workers who are approaching retirement. Administrators Said Contingent Faculty Have a Range of Responsibilities, and They Consider Multiple Needs When Determining Faculty Makeup Full-Time Contingent Faculty at Institutions We Visited May Have a Variety of Responsibilities, but Part-Time Contingent Faculty Generally Focus on Teaching According to administrators we interviewed, institutions utilize full-time contingent faculty for different purposes, which may involve responsibilities beyond teaching. Administrators said full-time contingent faculty are hired primarily to teach and generally have larger course loads than tenure-track faculty who may teach fewer courses per semester due to significant research responsibilities. However, they also noted that— similar to tenure-track faculty—many full-time contingent faculty carry out additional responsibilities. For example, some full-time contingent faculty may perform service, conduct research, advise students, serve as department chairs, or manage student recruitment efforts for their programs. Many other full-time contingent faculty serve as instructors or lecturers whose sole responsibility is to teach. For example, administrators from one institution explained that they employ professional instructors who teach four courses per semester and have no service or research responsibilities. In addition, some full-time contingent faculty are hired because they have certain professional qualifications or experience. For example, one institution we visited employed academic professionals who may teach one or two courses per year while carrying out administrative, marketing, mentoring, or other duties. While full-time contingent faculty may have a variety of responsibilities, administrators stated that part-time contingent faculty generally focus on teaching, though they also may fulfill different purposes. In some cases, part-time contingent faculty serve as expert practitioners who teach specific subject matter. For example, administrators from one institution said that they hire part-time contingent faculty to teach instrumental music courses because teaching each instrument requires specialized expertise, and there may not be enough students learning any single instrument to warrant a full-time position. In other cases, part-time contingent faculty teach general education courses, such as Introduction to English Composition, which most students are required to take. In addition, while some part-time contingent faculty may have full-time jobs outside of academia, others may be working toward long-term careers as tenure-track professors, according to administrators. Administrators from some institutions also told us that they hire part-time contingent faculty help to manage lab courses (e.g., setting up laboratory equipment, assisting students) or to serve as mentors to students in specific programs (e.g., theological studies). Administrators Consider Financial, Institutional, Faculty, and Student Needs When Determining Faculty Makeup University and college administrators we interviewed identified a number of financial and institutional considerations as well as faculty and student needs that affect their decisions regarding faculty makeup (see fig. 9). Administrators stated that utilizing contingent faculty allows for flexibility in managing various financial considerations, including the following: Budget uncertainty: Administrators from several public institutions explained that utilizing contingent faculty helps them manage uncertainty regarding the level of public funding they may receive. Administrators have the option not to renew contracts of contingent faculty if they experience a decrease in their funding, whereas institutions commit to retain tenure-track faculty until they retire. In addition, administrators from several public institutions noted that, as a result of decreased state funding, they have become more reliant on tuition to meet their budget needs. They told us that hiring contingent faculty to focus on teaching rather than research allows the institution to offer more classes and serve additional students, which in turn, generates more tuition revenue. Compensation costs: Administrators stated that, in general, they cannot employ tenure-track faculty for all courses because they can be more expensive to employ than contingent faculty. In addition to the long-term commitment associated with tenure, other costs may include spending to support research conducted by tenure-track faculty (e.g., investment in specialized labs or equipment). Legal or grant program requirements: Some administrators said that legal or grant program requirements affect their decisions regarding the utilization of contingent faculty. For example, administrators from several institutions told us that they had reduced teaching loads for part-time faculty because the Patient Protection and Affordable Care Act (PPACA) requires certain employers to provide health insurance for employees working 30 hours or more per week. Administrators from another institution stated that they utilized in-house faculty and hired additional contingent faculty to staff a federal grant program aimed at providing training for inmates at correctional facilities because—after receiving notification that they had been awarded the grant—they had approximately 2 months to staff 160 course sections. In addition, since they did not know whether the grant would be renewed, they did not know whether they would be able to retain those faculty at the end of the program. Institutional Considerations Administrators said that utilizing contingent faculty also allows flexibility to meet different institutional needs. Examples of institutional considerations cited by administrators include the following: Enrollment: By utilizing contingent faculty, institutions have more flexibility to meet course demand if there is a surge in enrollment or to downsize if there is a drop in enrollment, according to administrators. For example, administrators from one 2-year institution noted that enrollment generally increases when the economy is weak and decreases when the economy is strong. These administrators also said that their enrollment fluctuates greatly with changes in the economy and that, in their experience, prospective students are more likely to choose 4-year institutions rather than 2-year institutions when the economy is strong. In addition, when offering a course, administrators said part-time faculty may teach that course during a trial period while administrators decide whether to offer the course long term. Location and market demand: Some administrators stated that they offer contingent faculty positions in response to market conditions. For example, administrators from institutions located in small towns or rural areas said they rely on local professionals to teach certain courses on a part-time basis, in part, because of challenges finding qualified faculty and having fewer students enrolled at remote sites. Some administrators also said contingent faculty positions offer certain advantages that help them recruit high quality instructors. For example, administrators from one university noted that their institution offers stable, full-time employment to recent graduates looking to gain experience before applying for tenure-track positions at other institutions. Specialized experience: Contingent faculty may bring professional expertise to certain courses. For example, administrators from several institutions stated that their programs for health professionals rely on contingent faculty working in their field to teach clinical courses so that students may gain experience at an established medical practice. Administrators said that hiring practitioners from local industry as part- time instructors is an effective way to support specialized courses that have a limited number of sections. Administrators from one institution also noted that practitioners may have the qualifications needed to meet accreditation requirements for certain programs and departments (e.g., professional and technical programs). Balancing priorities: Administrators said that utilizing a combination of tenure-track and contingent faculty helps their institutions fulfill both teaching and research missions and accommodate the hiring needs of different programs and departments. For example, administrators from one institution noted that the additional revenue from increased course offerings—staffed by part-time contingent faculty—allows them to invest more money in research programs for tenure-track faculty. Administrators from two institutions explained that hiring part-time contingent faculty in a given department allows them to reallocate resources as needed, for example, to hire full-time contingent or tenure-track positions in another department. In addition, while contingent faculty may help fulfill accreditation requirements for certain programs, administrators from several institutions also stated that their accrediting bodies require a balance of contingent and tenure-track faculty, or alternatively, full-time and part-time contingent faculty. For example, administrators from one 4-year institution told us that part-time faculty may teach no more than 25 percent of student credit hours within their business school. Faculty Needs As part of faculty utilization decisions, administrators said that they consider the personal and professional needs of faculty. Examples of faculty needs cited by administrators include the following: Flexibility: Administrators told us that they offer part-time positions, in part, because many qualified candidates want to work part-time for professional, family, or other reasons. For example, administrators at one institution said that part-time contingent faculty positions allow expert-practitioners to continue working full-time in their field while pursuing an interest in teaching. Alternatively, for those teaching as full-time contingent faculty, in some cases, their position may offer a more predictable schedule or other benefits compared to their professional field. Course loads: Administrators at some institutions said they prioritize the professional needs of existing full-time faculty before hiring part- time faculty by ensuring that full-time faculty have enough courses to meet their required teaching loads. Career paths: Some institutions have established mechanisms to support long-term career paths for full-time contingent faculty. For example, administrators from one institution stated that full-time contingent faculty may qualify for multi-year contracts that can be terminated only for adequate cause, such as gross professional misconduct. Administrators from several institutions said that they offer the full set of professorial ranks (i.e., Assistant Professor, Associate Professor, and Professor) to some full-time contingent faculty positions in order to provide opportunities for advancement. Student Needs Administrators stated that having a combination of tenure-track and contingent faculty—or full-time and part-time contingent faculty at institutions without tenure—is necessary to meet different student needs. Examples of student needs cited by administrators include the following: Learning opportunities: Administrators stated that different types of faculty may offer different opportunities to students. For example, administrators told us that tenure-track faculty may provide research and academic networking opportunities whereas contingent faculty may not have the same opportunities to develop professional networks or conduct research in their field. Some administrators also said that the academic freedom associated with tenure or having faculty who conduct research in their field may be beneficial to students. Nonetheless, administrators from several institutions emphasized that contingent faculty were equally qualified to teach and that their positions allowed them to focus on teaching. Administrators also noted that contingent faculty may bring professional expertise and real-world experiences to the classroom. In addition to courses that require specialized experience, administrators from one institution said they also value the outside experience that contingent faculty bring to general education courses. As an example, they stated that part-time contingent faculty with experience from other jobs or professions may be able to relate to the real-world needs of their students because the majority of students will seek employment outside of academia. Community: Administrators said that, regardless of tenure status, they depend on having full-time faculty to help create a sense of community. They discussed informal ways that faculty support their campus community. For example, some administrators noted that full- time faculty contribute by mentoring students and participating in activities on campus. In contrast, part-time faculty are not able to spend as much time on campus because they often have other jobs or commitments, according to administrators. Absent National Information on Pay Rates, Contingent Faculty in Two Selected States Are Paid Less per Course, and Relatively Few Part- Time Faculty Receive Health or Retirement Benefits Data from Two States Show Contingent Faculty Are Paid Less per Course, Though Disparities Shrink If Pay for Research and Service Is Excluded National data on contingent faculty pay rates are not available, but data from two states show that contingent faculty are paid less per course. IPEDS data cannot be used to determine faculty pay rates because salary data are not collected for part-time faculty nor are they collected at the individual faculty level, and CPS data do not differentiate between full- time tenure-track and full-time contingent faculty. Given the limitations of national data, we used data from two states to compare annual earnings across different types of faculty. The differences in median annual earnings shown in table 5 provide some insight into the generally lower overall compensation of contingent faculty, though these data are not generalizable. Further, particularly for part-time faculty who may be paid on a piecemeal or per-course basis, this measure does not provide information about whether compensation differences are due to lower pay rates or less work performed (e.g., courses taught or hours worked). Thus, we use the state data to calculate and examine comparable pay rates per course for all faculty types. Private organizations have attempted to collect data specifically on pay-per-course rates for part-time faculty, though efforts have been limited. On a per-course basis, we found that contingent faculty at public institutions in two states are paid less per course taught, on average, than full-time tenure-track faculty, though the extent of differences varies depending on contingent faculty group and pay measure. We conducted regression analyses of total pay per course and instructional pay per course, which provide two different perspectives on faculty compensation (see sidebar for explanations of these approaches and see appendix I for details on our methods). These analyses controlled for other factors that may affect earnings, such as employing institution, discipline, highest degree earned, and demographics. As shown in table 6, in terms of total pay per course, we found the following: Part-time contingent faculty in both states are paid about 75 percent less per course regardless of whether the population includes all faculty or is limited to “primarily teaching” faculty. The primarily teaching group excludes faculty who primarily hold other roles unrelated to instruction (e.g., administrators and research faculty). Full-time contingent faculty are paid about 35 percent less per course in North Dakota and about 40 percent less per course in Ohio, among primarily teaching faculty—differences are larger in Ohio if all faculty are included. Instructional graduate assistants earn more per course than part-time faculty (though still less than full-time tenure-track faculty). However, compensation for these groups is fundamentally different because instructional graduate assistants generally receive a stipend, similar to an annual salary, rather than being paid by the course like many part- time faculty. In addition, graduate assistantships may be awarded for academic merit or recruitment, and could also be considered as compensation for a graduate assistant’s work as a student. Disparities in instructional pay per course—which measures pay for equivalent work (see sidebar above)—are smaller for all contingent faculty groups than those for total pay per course. As shown in table 7, we found the following: Part-time contingent faculty in both states are paid about 60 percent less per course regardless of whether the population includes all faculty or is limited to primarily teaching faculty. Among primarily teaching faculty in both states, full-time contingent faculty are paid about 10 percent less per course than full-time tenure- track faculty. As with total pay, the instructional pay disparity for full-time contingent faculty in Ohio is larger if all faculty are included. However, when all faculty are included in North Dakota, the pay difference between full- time contingent and full-time tenure-track faculty is not significant at the 95 percent confidence level. Consistent with our other findings, when we analyzed national data from the 2013 Survey of Doctorate Recipients (SDR), we also found that contingent faculty in sciences fields earned less annually than full-time tenure-track faculty. Full-time contingent faculty earned 22 percent less than full-time tenure-track faculty, on average, and part-time contingent faculty earned 70 percent less, among instructional, doctorate-holding faculty in STEM, health, and social sciences fields. Unlike our analyses of state data, the SDR analysis cannot account for differences in the number of courses taught, and thus the results represent the combined effects of lower pay rates and smaller workloads, to the extent either exists. Relatively Few Part-Time Contingent Faculty Receive Health or Retirement Benefits from Their Employment Data from North Dakota and Georgia, as well as national data covering different populations, suggest that relatively few part-time contingent faculty receive health or retirement benefits from their employment though full-time contingent faculty may. Although not generalizable, data from North Dakota and Georgia include data on actual benefits provided to faculty by institutions, as opposed to self-reported rates of coverage found in national survey data. Relatively few part-time contingent faculty and instructional graduate assistants in the North Dakota and Georgia data receive retirement, health, and life insurance benefits from their employment. For example, in Georgia and North Dakota, about 98 percent or more of individuals in full-time tenure-track and full-time contingent positions receive work-provided retirement benefits, compared to 19.4 and 9.3 percent, respectively, of those in part-time contingent positions (see table 8). An even smaller percentage of instructional graduate assistants in both states receive any of these benefits from their employment; however, instructional graduate assistants are students, so the terms of their employment may be different than traditional full-time and part-time employees. Similarly, our analysis of SDR and CPS data show that relatively few part- time contingent faculty nationwide receive retirement benefits from their employment. According to the 2013 SDR data, among instructional, doctorate-holding faculty in STEM, health, and social sciences fields, an estimated 48.4 percent (+/- 4.2) of part-time contingent faculty report having access to “a retirement plan to which employer contributed,” compared to the vast majority of full-time tenure-track and full-time contingent faculty. According to CPS data covering employment in 2015, an estimated 16.6 percent (+/- 6.1) of part-time faculty report participating in a work-provided retirement plan, as compared to 60.8 percent (+/- 4.7) of full-time faculty. National Data on Health Insurance Benefits While comparing health insurance coverage is complicated because workers may be covered by other family members’ plans, in both the SDR and CPS data, smaller proportions of part-time faculty had health insurance through their own employment. According to the 2013 SDR data, only 39.4 percent (+/- 4.6) of part-time contingent faculty had access to “health insurance that was at least partially paid by employer” compared to almost all full-time tenure-track and full-time contingent faculty. Similarly, in the CPS data, much smaller percentages of part- time faculty than full-time faculty report having health insurance through their own employment (see table 9). Data from a 2013 Sample of Faculty with Doctorates Show That Contingent Faculty Were Less Satisfied with Certain Aspects of their Economic Circumstances In addition to the lower pay and access to benefits experienced by some contingent faculty, among a national sample of instructional, doctorate- holding faculty in STEM, health, and social sciences fields, contingent faculty were less satisfied with their job security and career prospects. Based on our analysis of 2013 SDR data, the vast majority of all instructional faculty, including contingent faculty, stated that they are very or somewhat satisfied with their employment overall. However, compared to full-time tenure-track faculty, more contingent faculty reported some level of dissatisfaction (see fig. 10). While most faculty reported satisfaction with their employment, at least a third of both full- and part- time contingent faculty stated that they are dissatisfied with their job security and opportunities for career advancement. For example, an estimated 55.1 percent (+/- 4.5) of part-time contingent faculty reported some level of dissatisfaction with opportunities for advancement (see fig. 10), and the proportion who said they were very dissatisfied—26.1 percent (+/- 3.8)—is around 5 times greater than for full-time tenure-track faculty. While Contingent Faculty at Selected Institutions Said Their Work Offers Certain Advantages, They Expressed Concerns about Contracts, Wages, and Institutional Support Contingent Faculty Identified Certain Advantages of Their Work Contingent faculty at selected institutions said their work offers certain advantages, including those allowing them to balance professional and personal responsibilities, develop skills, or work with students. Part-time contingent faculty in some discussion groups said they choose to work part-time because it gives them needed flexibility to balance teaching with working full-time or to meet family needs, such as childcare or caring for sick parents. As stated previously, our analysis of nationally representative 2013 SDR data showed that, among a sample of instructional faculty with doctorate degrees in STEM, health, and social sciences fields, many faculty preferred to work part-time for reasons including family responsibilities or holding another job. In terms of developing skills, one instructional graduate assistant told us that having teaching experience gives her an advantage in the job market. In addition, in both full- and part-time discussion groups, some contingent faculty told us they primarily want to teach, and their roles allow them to do that rather than having to conduct research or take on other responsibilities. In some discussion groups, contingent faculty said they are committed to teaching because they find it rewarding to interact with students. Insight from a Full-Time Contingent Faculty Member about Connecting with Students “I have yet to meet a contingent faculty member that does not say that student contact is extremely important to them…We’re excellent teachers. We’re interested in teaching. We are interested in being with students.” Contingent Faculty Expressed Concerns about Short-term Contracts, Untimely Contract Renewals, and Compensation Contract-Related Concerns Contingent faculty in some of our discussion groups expressed concerns about contractual issues. In particular, they cited concerns regarding contract length, untimely contract renewals, or insufficient notice about their class schedules. Full- and part-time contingent faculty said short- term contracts—annual or semester-to-semester contracts—produce anxiety about job stability because of uncertainty about whether contracts will be renewed. Part-time faculty who teach at multiple institutions additionally said that short-term contracts hinder their ability to form lasting relationships with institutions or students. In some discussion groups, full- and part-time contingent faculty said untimely contract renewals can make it difficult to find another position if a contract is not renewed. For example, a full-time contingent faculty member said she received notification in August that her contract was not being renewed for the fall semester, at which point she could not find another position elsewhere for that semester. Part-time contingent faculty told us that notices about the status of their class schedules are also sometimes untimely. One full-time contingent faculty member said that, when he worked part-time, he sometimes did not know, until the first night of class, that a course he was scheduled to teach had been given to a full-time faculty member instead. While some contingent faculty expressed concerns about contract lengths and renewals, some contingent faculty said they do not have concerns in this area. Faculty members in some part-time discussion groups told us teaching is not their primary source of income or they are retired, so they are not concerned about job security and contract renewals. Insight from a Full-Time Contingent Faculty Member “The lack of long term job security/stability that results from short term contracts is my biggest concern. I find it insulting when comments like “great work, we’re committed to you” are coupled with actions like one year contracts when I have been in this position for 15 years. It does not make me feel valued.” Compensation-Related Concerns Contingent faculty we spoke with identified insufficient compensation as a disadvantage of their employment (see table 10). Full-time and part-time contingent faculty in some discussion groups said they must supplement their teaching income to cover their living expenses. For example, one full-time contingent faculty member said he does consulting work, bookkeeping, and product reviews to increase his income because his teaching salary is not adequate. In addition, some part-time faculty said they teach at several institutions to make ends meet financially and some instructional graduate assistants also said they take on extra work to cover living expenses. Union officials at the national level said their members have expressed similar concerns. Specifically, Service Employees International Union (SEIU) officials told us some contingent faculty members qualify for public assistance due to the low level of compensation they receive. Insight from Part-Time Contingent Faculty Member Teaching at Multiple Institutions “Society at large, I think, associates the college professor with a rather well paid and stable career. And I think most of us who worked in this field know that is anything but the case.” Some contingent faculty in both full- and part-time discussion groups said they are not paid for all of their job requirements or are undercompensated given their qualifications. Full- and part-time contingent faculty and graduate student instructors said they are required to assume extra responsibilities at no additional pay. For example, a faculty member in a full-time discussion group told us she was given additional duties of advising 15 students and attending meetings, neither of which was included in her contract. Both full- and part-time faculty in some discussion groups said their pay is not commensurate with their academic credentials. One full-time faculty member told us an administrator with a doctorate who works in the local school district near her institution is paid double her salary. Similarly, a part-time faculty member told us her salary is less than $20 an hour, a rate she considers as too low for a professional with a doctorate. Some Contingent Faculty at Selected Institutions Said They Have Limited Career Advancement or Institutional Involvement Opportunities and Lack Certain Types of Professional Support Limited Career Advancement Opportunities Contingent faculty in some discussion groups said they would like to move into a tenure-track or full-time position, but face barriers doing so, and union officials expressed similar views. For example, one full-time contingent faculty member told us teaching 6 to 10 classes per year does not allow her time to conduct the research needed to be competitive for a tenure-track position. In some discussion groups, both full- and part-time faculty said that they perceive that their colleagues sometimes view them as less capable because they are not tenure-track faculty. As a result, these faculty may not be considered for tenure-track positions when they become available. A part-time contingent faculty member who teaches at multiple institutions noted that availability of full-time positions may be limited because many institutions hire only part-time faculty. Union officials from the American Association of University Professors (AAUP) and SEIU also cited the decline in the availability of tenure-track positions as a barrier regarding career advancement for contingent faculty. Insight from a Part-Time Contingent Faculty Member Who Teaches at Multiple Institutions “It wasn’t that long ago that once you went to work for a college as an adjunct and you were there a certain number of years, there was a real expectation that you would be offered a full-time position or at least you would move to an annual contract so you only had to worry once a year. That’s disappearing. More and more colleges are moving away from that. Also, a lot of colleges are moving away from full-time positions.” Limited Institutional Involvement Contingent faculty in some discussion groups expressed concerns that they do not have a voice in institutional decision-making because they cannot serve on some department or university-level committees or vote on particular issues. They explained that sometimes a school’s policy prohibits their service or relevant policy is not clearly articulated. For example, a full-time contingent faculty member told us that contingent faculty members at her institution cannot participate on governance committees, which she said leaves administrators free to ignore the concerns of contingent faculty. Insight from a Full-Time Contingent Faculty Member “We have no voice. We have no say. We have no governance. We don’t have any of that. And yet, we all—every one of us around here earned the same degree, worked the same amount. So there is huge inequality between choosing to focus on research primarily, and therefore, getting this basic job guarantee until die and choosing to focus on teaching, not having that , even though in many other ways we are equivalent.” Contingent faculty in some discussion groups also told us they are reluctant to voice their views because they do not have job protections. For example, a full-time contingent faculty member in one discussion group told us she would feel more comfortable speaking up if she had a continuing contract rather than her current annual contract. An official from the National Center for the Study of Collective Bargaining in Higher Education and the Professions said that an issue for contingent faculty broadly is whether they are protected by due process. He said it can be unclear for contingent faculty whether they can be terminated without due process consideration when, for example, a student complains about the content of a faculty member’s lecture. Despite concerns about opportunities for institutional involvement, contingent faculty told us they preferred to use informal mechanisms to raise issues with the administration and had mixed views about the value of unions. Several full- and part-time faculty members said they are comfortable approaching their department chairperson or even university administrators to ask questions or express concerns. In terms of unions, some faculty in both full-time and part-time discussion groups said they were opposed to unions based on prior experiences or not wanting to pay dues. In contrast, some faculty said they thought a union could be beneficial by helping with certain issues, such as compensation and working conditions. Union officials told us there has been greater interest in recent years from contingent faculty—including graduate assistants—in learning about faculty unionization or in organizing into unions. However, one union official noted that it can be challenging for part-time faculty to form a union because they may move from one institution to another. Institutional Support Examples of Academic Associations’ Efforts to Focus on Contingent Worker Issues The American Political Science Association (APSA): Convened a committee in 2016 on the status of contingent faculty in the profession to expand ways to support contingent faculty members. The committee sponsored a roundtable at the APSA Annual Meeting in August 2017 to examine a range of topics related to contingent faculty, including promotion paths, fairness within the profession, and the role of unionization. The American Sociological Association (ASA): Formed a task force on contingent faculty in November 2015 to examine the implications of the recent growth of contingent employment among sociologists. The task force’s interim report, issued in August 2017, includes recommendations to ASA and universities, for improving contingent faculty working conditions. The Modern Language Association: (MLA) Convened a committee that will work through June 2019 to examine issues that affect contingent faculty, including salary and benefits, workplace issues and conditions of employment, demographics, participation in departmental and institutional governance, academic freedom, and professional development. The committee plans to identify effective policies and practices related to contingent faculty. The American Institute of Physics (AIP): Conducted a survey of individual faculty in 2016 that included questions on school climate and culture. As of February 2017, AIP was in the early stages of analyzing the survey response rates and results. Contingent faculty in some discussion groups also described a lack of institutional support in areas that can affect faculty teaching duties, such as access to information systems or office space. For example, a part- time faculty member told us her access to institutional email and the online grading system was terminated too soon because her contract ended a few days before she gave final examinations. Part-time faculty and faculty teaching at multiple institutions also raised concerns that they sometimes lack appropriate office space to ensure student privacy. Union officials we spoke with also said contingent faculty nationwide commonly cite these areas of limited institutional support as concerns. Some discipline-specific academic associations have also begun to focus on issues related to contingent faculty (see sidebar). Insight from a Part-Time Contingent Faculty Member Who Teaches at Multiple Institutions “The office space problem is a big problem. Either one doesn’t have any office space or it’s a jointly shared office space, a very large space with lots of people in it. It is very difficult to have kind of close conversations with students. I think it brings up some Family Educational Rights and Privacy Act (FERPA) problems, anonymity problems as well.” Agency Comments, Third Party Views, and Our Evaluation We provided a draft of this report to Education, NSF, and experts on contingent faculty issues or the data used in this report for their review and comment. Education did not have any comments. NSF and expert reviewers provided technical comments, which we incorporated, as appropriate. As agreed with your offices, unless you publicly announce the contents of this report earlier, we plan no further distribution until 30 days from the report date. At that time, we will send copies to the appropriate congressional committees, to the Secretary of Education and the Director of the National Science Foundation, and to other interested parties. In addition, the report will be available at no charge on the GAO website at http://www.gao.gov. If you or your staff have any questions about this report, please contact me at (202) 512-7215 or [email protected]. Contact points for our Offices of Congressional Relations and Public Affairs may be found on the last page of this report. GAO staff who made key contributions to this report are listed in appendix III. Appendix I: Objectives, Scope, and Methodology The objectives of this review were to determine (1) what is known about the makeup and utilization of the postsecondary instructional workforce; (2) the roles different types of faculty fill at selected institutions and the factors administrators consider when determining their faculty makeup; (3) what is known about how economic circumstances compare across different faculty types; and (4) what contingent faculty members report as advantages and disadvantages of their work. To address objectives 2 and 4, we interviewed administrators and contingent faculty members during site visits at selected institutions in three states—Georgia, North Dakota, and Ohio. In each state, we visited one 4-year public institution, one 4-year private (non-profit) institution, and one 2-year public institution (see table 11). We selected institutions in these states, in part, to provide context for our analysis of faculty and course data that we obtained from their postsecondary data systems (see Section 1 of this appendix for more information). In addition to data availability, we considered size and geographic location as part of our state selection process. When selecting institutions within each state, we considered factors such as the size of the instructional faculty workforce, the percentage of contingent faculty, and whether the institution is located in an urban, suburban, or rural area. In our interviews with administrators—chief academic officers, vice presidents, or deans, among others—we asked about the roles different types of instructional faculty fill and the factors administrators consider when determining their institution’s faculty makeup. In addition to administrators at the institutions above, we also interviewed administrators from one large online-based for-profit institution, which we selected primarily based on size of the institution. In total, we interviewed administrators from 10 institutions. The findings from these interviews are not generalizable. At each institution, we held discussion groups with full-time and part-time contingent faculty and graduate student instructors, where applicable. University administrators solicited participants for the discussion groups on our behalf. During these discussion groups, we asked contingent faculty broad, open-ended questions about the advantages and disadvantages of their work and about their working conditions. Participants were invited to complete a written questionnaire to provide demographic information about themselves. Among the 109 contingent faculty members who completed our questionnaire, the average age of full- and part-time contingent faculty we met with was 53. Graduate student instructors were younger, with an average age of 30. Contingent faculty we interviewed came from a range of disciplines, including English, music, engineering, and the health professions. The vast majority of full- and part-time contingent faculty indicated that they held a master’s or doctorate degree. At the institutions we visited in Georgia, North Dakota, and Ohio, the majority of part-time faculty worked at one institution. To ensure we collected a broad range of perspectives, we conducted two additional discussion groups with contingent faculty who taught at multiple institutions. In total, we conducted 21 discussion groups with contingent faculty. Finally, we conducted additional interviews to obtain background and context for our work. We met with individuals knowledgeable about issues related to postsecondary faculty and unions representing postsecondary faculty, including the American Association of University Professors and the Service Employees International Union. For all questions, we also reviewed relevant federal laws and regulations. The remainder of this appendix provides detailed information about the data and quantitative analysis methods we used in our review, as follows: Section 1: Key data sources Section 2: Quantitative analysis methods used to address the makeup, utilization, and economic circumstances of postsecondary instructional faculty (objectives 1 and 3) Section 3: Pay-per-course regression analysis methods (objective 3) Section 4: Annual earnings regression analysis methods (objective 3) Section 1: Data Sources To address our objectives, we used data from multiple sources (see table 12). To gain an understanding of and provide context for the relevant faculty data that we analyzed, we interviewed officials from federal, state, and non-governmental agencies who collect and maintain the respective datasets, including the Department of Education (Education), Labor, National Science Foundation, North Dakota University System (NDUS), Ohio Department of Higher Education (ODHE), University System of Georgia (USG), and American Academy of Arts & Sciences (AAAS). The Integrated Postsecondary Education Data System (IPEDS) and the state administrative data represent the entire populations they cover, and while the Current Population Survey (CPS), the Survey of Doctorate Recipients (SDR), and the Humanities Departmental Survey (HDS) are sample survey data, when weighted, they also represent the populations they cover. Because the sample surveys followed a probability procedure based on random selections, each respective sample is only one of a large number of samples that might have been drawn. Since each sample could have provided different estimates, we express our confidence in the precision of our particular sample’s results as the margin of error (i.e. the half width of the 95 percent confidence interval—for example, +/- 7 percentage points). This is the interval that would contain the actual population value for 95 percent of the samples that could have been drawn. Throughout our analyses, for estimates from survey data we reported the applicable margins of error. In some cases, the confidence intervals around our estimates were asymmetrical; however, we presented the maximum half-width for simplicity and for a consistent and conservative representation of the sampling error associated with our estimates. Our analyses of CPS and SDR survey data are weighted analyses using sample design information, replicate weights, and survey analysis software to get the proper sample survey estimates and margins of error. Additional details about the datasets follow. Integrated Postsecondary Education Data System (IPEDS) IPEDS is a system of interrelated surveys conducted annually by Education’s National Center for Education Statistics (NCES). IPEDS gathers information from every college, university, and technical and vocational institution that participates in federal student financial aid programs, as well as other institutions that report data voluntarily. In 2015, more than 7,500 institutions reported data to IPEDS. IPEDS collects data in the following 12 areas: institutional characteristics; completions; 12-month enrollment; fall enrollment; graduation rates; 200% graduation rates; student financial aid; outcome measures; admissions; human resources; finance; and academic libraries. As of the 2005 IPEDS data collection, information on faculty and staff are collected as part of the human resources survey component, and include information on faculty demographics and types of positions, among other things. We used IPEDS data from 1995, 1999, 2003, 2007, 2011, and 2015. We utilized IPEDS as our primary data source because we are able to identify a universe of postsecondary institutions and also because the data allow us to distinguish between tenure-track and contingent positions. Current Population Survey (CPS) Annual Social and Economic Supplement (ASEC) The CPS is sponsored jointly by the Census Bureau and the Department of Labor’s Bureau of Labor Statistics. It is the source of official government statistics on employment and unemployment in the United States. The basic monthly survey is used to collect information on employment, such as employment status, occupation, and industry, as well as demographic information, among other things. The survey is based on a sample of the civilian, non-institutionalized population of the United States. Using a multistage stratified sample design, about 54,000 households are interviewed monthly based on area of residence to represent the country as a whole and individual states; the total sample also includes additional households that are not interviewed for various reasons, such as not being reachable. In addition to these interviewed and non-interviewed households from the basic CPS monthly sample, the ASEC includes additional households; the total sample size for the 2016 ASEC was almost 100,000 households. The ASEC provides supplemental data on work experience, income components, such as earnings from employment, and noncash benefits, such as health insurance coverage, among other things. Data on employment and income refer to the preceding calendar year, although demographic data refer to the time of the survey. This report used data from the March 2016 ASEC, which refers to employment and income during calendar year 2015. Survey of Doctorate Recipients (SDR) SDR is a biennial survey conducted by the National Science Foundation’s (NSF) National Center for Science and Engineering Statistics (NCSES) that provides demographic and career history information about individuals with a research doctoral degree in a science, technology, engineering, and math (STEM), health, or social sciences field from a U.S. academic institution. The survey follows a large sample of individuals throughout their careers from the year they received their doctoral degree until age 75, plus a sample of new doctoral recipients added in each cycle. The survey includes questions regarding occupation (including discipline area for postsecondary faculty), earnings, job satisfaction, faculty tenure status, and faculty rank, among other topics. While some data from the survey are released publicly, other data are restricted from public use—including data on tenure and rank— in order to protect the anonymity of survey respondents. This report used data from the 2013 SDR, which refers to employment in February 2013. We obtained the publicly available data and a few additional restricted-use variables that NCSES recoded for our use. Faculty and Course Data Received from Selected States The data from Georgia, North Dakota, and Ohio contained variables on faculty characteristics, earnings and benefits, and courses taught. We developed data requests through discussions with officials in each state. Georgia Postsecondary Institution Administrative Data (USG data) The data from USG covered all 4-year public institutions in Georgia identified in our IPEDS universe and included course and enrollment data from an academic database merged with faculty and earnings data from USG’s Human Resources Data Mart. The Georgia data also included information on the percentage of individual faculty members’ roles comprised of instruction, research, and other responsibilities. The course and enrollment data covered academic year 2015-16—courses taught during fall term 2015, spring term 2016, and summer term 2016. Most faculty data are from fall 2015. For some faculty who were not in the fall 2015 data file because they started teaching in spring 2016, for instance, USG matched fall 2016 faculty data to the course data. Earnings data covered calendar year 2015 and included earnings year-to-date through November. North Dakota Postsecondary Institution Administrative Data (NDUS data) The data from NDUS officials covered all non-tribal 4-year and 2-year public institutions in North Dakota identified in our IPEDS universe and included course and enrollment data, as well as faculty and earnings data. All of the data covered academic year 2015-16—courses taught and earnings during fall term 2015, spring term 2016, and summer term 2016. The data included common unique identifiers that allowed us to merge extracts we received according to faculty ID and institution. The data were downloaded by NDUS officials from a centralized data system into which the North Dakota institutions report their data directly. Ohio Postsecondary Institution Administrative Data (ODHE data) The data from ODHE covered all 4-year public institutions and most 2- year institutions in Ohio identified in our IPEDS universe and included: (1) course and enrollment data, (2) faculty data, and (3) faculty earnings data. All of the data were from ODHE’s Higher Education Information (HEI) system, a comprehensive relational database that includes student enrollment, course, financial aid, personnel, finance, and other data submitted by Ohio’s colleges and universities. The course and enrollment data covered academic year 2014-15—courses taught during summer term 2014, fall term 2014, and spring term 2015. Faculty and earnings data covered fiscal year 2015 (i.e., July 2014 through June 2015). Humanities Departmental Survey (HDS) The HDS is a collaborative effort to collect and analyze information from humanities departments across a number of academic fields. The HDS is sponsored by AAAS, and national humanities organizations and disciplinary associations, such as the Modern Language Association and the American Historical Association, helped develop the HDS. The survey collects a variety of information for each humanities field, including data on the number and types of faculty and students taught by faculty type. The survey has been administered twice, covering academic years 2007- 08 and 2012-13. In both instances, the Statistical Research Center of the American Institute of Physics administered the surveys to a nationally representative stratified sample of humanities departments in four-year colleges and universities that existed in 2007-08 and was updated for new disciplines in 2012-13. The 2012-13 survey included 2,127 departments in its sample across 13 humanities fields, and its overall response rate was 71 percent. Information about faculty referred to employment levels as of fall 2012. We identified several other discipline-specific academic associations that have collected or are currently collecting data on faculty makeup in their departments, including contingent faculty. However, we did not compare the results of other department surveys to the HDS because the response rates in other surveys were too low to be considered generalizable or because any observable differences in faculty composition could be attributed to differences in survey methodology or timeframe covered. Data Reliability For each of the datasets described above, we conducted a data reliability assessment of variables included in our analyses. We reviewed technical documentation and related publications and websites with information about the data. We spoke with the appropriate officials at each agency or organization to review our plans for analyses, as well as to resolve any questions about the data and any known limitations. We also conducted electronic testing, as applicable, to check for logical consistency, missing data, and consistency with data reported in technical documentation. We determined that the variables we used from the data we reviewed were sufficiently reliable for the purposes of this report. Section 2: Quantitative Analyses of the Makeup, Utilization, and Economic Circumstances of the Postsecondary Instructional Workforce This section discusses the quantitative analysis methods (not including regression analyses) we used to address the makeup, utilization, and economic circumstances of the postsecondary instructional workforce. We used federal data from CPS, IPEDS, and SDR, state data from Georgia, North Dakota, and Ohio, and non-governmental data from HDS for these analyses. In each of the analyses that follow, our population of analysis was postsecondary instructional faculty. However, our definition of instructional faculty varied depending on the data source, as different sources provide different information regarding instructional responsibilities. For example, IPEDS indicates whether an individual’s responsibilities are primarily instructional whereas the state data indicates whether an individual teaches a course. For each set of analyses, we explain what definition of instructional faculty we used. Within our population of instructional faculty, we defined as contingent faculty any full-time or part-time faculty who do not have tenure or are not on the tenure track. IPEDS Analyses of Historical and Current Makeup To analyze whether and how the size of the contingent faculty workforce has changed over time, we used IPEDS data to identify instructional staff nationwide by type of institution in 1995, 1999, 2003, 2007, 2011, and 2015, which is the most recently available year of data. The five historical snapshots used data from the fall staff surveys to examine counts of faculty and any trends in postsecondary education during the period 1995-2011. The 2015 snapshot used data from the “employees by assigned position” survey to examine current counts of faculty by position type and used data from the fall staff survey to examine counts of faculty by gender and race. We could not compare the historical and current snapshots of faculty counts due to a significant change in 2012-13 to how IPEDS defines instructional staff. Prior to this change, instructional staff included those “whose primary responsibility is instruction, research, and/or public service” combined in a single category. After the change, instructional staff included only those whose responsibilities are primarily instructional or those “for whom it is not possible to differentiate between instruction or teaching, research, and public service because each of these functions is an integral component of his/her regular assignment.” As a result, data on instructional faculty collected since 2012 is not comparable to data collected prior to 2012. For each of these years of faculty data, we merged information from the IPEDS institutional characteristics file and focused our analyses on a universe of institutions that fit as close as possible to the following definition: Active, Title IV, degree-granting 2-year and 4-year primarily postsecondary institutions that are generally open to the public, have at least 15 full-time equivalent staff, and reported at least 1 instructional staff member or graduate teaching assistant. The number of postsecondary institutions can change from year to year due to new schools opening or existing schools closing or consolidating with other schools, as well as due to changes in how schools elect to report data to IPEDS. Not all of the same variables were available in the 1999 and 1995 IPEDS institutional characteristics files. As a result, for the 1999 data, we used different variables that also identified institutions that fit this definition. For the 1995 data, we approximated this definition by identifying institutions that offered at least an associate’s degree or higher and that were active institutions eligible for student financial aid (to approximate Title IV institutions). For the historical snapshots, we identified counts of faculty by institution type (i.e., control: public, private, for-profit; and level: 2-year, 4-year). We categorized faculty according to the following position types: full-time tenure-track (both tenured and non-tenured but on a tenure track); part-time; and graduate teaching assistant. The historical IPEDS data (from the fall staff surveys) do not break out part-time tenure-track from part-time contingent. For the 2015 snapshot, we identified counts of faculty by institution type, as well as by other institutional characteristics, such as size and the highest degree offered by the institution. We categorized faculty according the following position types: full-time tenure-track (both tenured and non-tenured but on a tenure track); part-time tenure-track (both tenured and non-tenured but on a tenure part-time contingent; and graduate teaching assistant. We also identified contingent faculty positions by their contract types: non-faculty status. We used the 2015 IPEDS fall staff survey data to identify faculty by gender and race/ethnicity group. For full-time faculty, we were able to examine the full spectrum of tenure-track versus contingent with various contracts. However, because these data were from the 2015 IPEDS fall staff survey, the data do not break out part-time tenure-track from part- time contingent. The IPEDS race/ethnicity categories we analyzed were: Black or African American Other or unknown (includes the IPEDS race/ethnicity categories: American Indian or Alaska Native; Native Hawaiian or other Pacific Islander; two or more races; and race/ethnicity unknown) White (non-Hispanic) Aggregated IPEDS data represent the universe of postsecondary instructional faculty positions, rather than a mutually exclusive count of unique instructional faculty members. IPEDS data are reported at the institution level, and so for any given institution the counts they report represent both the number of faculty at the institution and the number of positions they fill. However, because faculty who teach at more than one institution are counted and reported by each institution, when faculty counts are aggregated across multiple institutions, these faculty are counted multiple times—for each position they fill. As a result, aggregated counts based on IPEDS data represent the universe of unique instructional faculty positions, rather than the universe of unique faculty workers. CPS Analyses of Current Faculty Makeup and Economic Circumstances We used CPS data from the March 2016 ASEC to estimate the numbers of workers employed as postsecondary teachers in colleges and universities nationwide during calendar year 2015. We categorized as postsecondary instructional faculty any worker whose employment was in both the “postsecondary teachers” occupation (census code 2200) and the “colleges and universities, including junior colleges” industry (Census code 7870). We also determined whether a worker was employed full- time (35 hours or more) or part-time (less than 35 hours) using another variable in the ASEC. Among other differences with IPEDS data (see discussion of IPEDS above), CPS data capture the number of workers rather than the number of positions in postsecondary education and counts each worker once even if they work at multiple institutions. In addition, because CPS represents the entire labor force, the data include workers at postsecondary institutions that we may have excluded from our IPEDS analyses (e.g., non-degree-granting institutions). We utilized CPS data to provide context for the total number of postsecondary teachers and to estimate the proportions of the instructional workforce represented by full- time and part-time faculty. However, analysis of CPS data was not a primary component of our report because the data cannot differentiate workers by institution or by tenure status. As a result, the estimated population of full-time faculty includes both tenure-track and contingent faculty. Because CPS identifies workers as opposed to positions (which might yield a lower count than the IPEDS data) and includes workers at postsecondary institutions that we excluded from our IPEDS analyses (which might yield a higher count than the IPEDS data), the count of workers in the CPS data and the count of positions in the IPEDS data are not directly comparable. We also examined the reasons part-time faculty reported they worked part-time. We focused our analysis on 3 groups of part-time faculty: (1) those who reported wanting to work part-time; (2) those who reported they could only find a part-time job; and (3) those who reported seasonal or temporary fluctuations in the availability of employment (i.e., “slack work”)—we combined the latter two groups because they are both related to economic circumstances. To analyze the economic circumstances of contingent faculty, we used CPS data to estimate the median earnings of full-time and part-time faculty, as well as their receipt of work-provided retirement and health benefits. Our analysis of median earnings used ASEC data on the self- reported amount earned from a worker’s employer before deductions. In examining benefits, we used the term “work-provided” rather than “employer-sponsored” because the ASEC survey questions ask about benefits offered by a worker’s employer or union. For our analysis of access to work-provided retirement plans, we counted a worker as having a work-provided retirement plan if they responded “yes” to both of the following questions from the ASEC: (1) “Other than Social Security, did the employer or union that worked for have a pension or other type of retirement plan for any of the employees?” and (2) “Was included in that plan?” We also estimated the percentages of full- time and part-time faculty who were covered by any private health insurance plan; were covered by private health insurance in their own name; or had a work-provided health insurance plan. Those individuals without insurance could have received insurance coverage through a family member or other means. SDR Analyses of Compensation and Employment Experiences To compare—at the national level—the compensation and employment experiences of contingent faculty and tenure-track faculty, we used 2013 SDR data to identify different faculty types and examined the extent to which there were differences in earnings, benefits, and job satisfaction. SDR data only include doctorate holders in STEM, health, and social sciences fields, and thus our estimates cannot be generalized to non- doctorate holders or to fields outside of STEM, health, and social sciences fields. For that reason, we did not present faculty population size estimates using SDR data. We created our analysis population of instructional faculty based on responses to questions regarding work activities and institution type. Using these variables, we classified as instructional faculty any respondents who said that their “primary or secondary work activity is teaching,” and whose institution type was a 2-year college; 4-year college or university; medical school; or university-affiliated research institute. This resulted in an analysis population of 7,232 instructional faculty respondents; however, our analyses are weighted analyses that generalize to the population. Within our analysis population, we identified faculty types based on tenure status (i.e., tenured/on the tenure track or not on the tenure track) and whether respondents said they worked 36 hours or more per week or less than that (i.e., full-time versus part-time). We categorized graduate assistants separately, though we chose not to present estimated percentages for graduate assistants. Given that SDR is a survey of doctorate holders, it may be that graduate assistants in the SDR data are—for example—working toward another doctoral degree or have remained at their degree-granting institution in a postdoctoral position. In either case, we believe the working arrangements and economic circumstances of these individuals may be unique from those of most other graduate assistants. Without more detailed information, the data do not allow us to determine the exact nature of graduate assistant positions in the SDR data or explain how they compare to other types of positions. We also chose not to present estimated percentages for part- time tenure-track faculty given that they represented a small proportion of our analysis population. To analyze the economic circumstances of contingent faculty, we used SDR data to calculate median annual earnings by faculty type, as well as data on the availability of work-provided benefits. We calculated median earnings using data on basic annual salary from the respondent’s principal job. We analyzed data on the following types of benefits: health insurance, pension or retirement plans, profit-sharing plans, and paid vacation/sick/personal days. Respondents were asked whether each type of benefit was available to them regardless of whether they chose to take the benefits. To analyze the employment experiences of contingent faculty, we used SDR data on job satisfaction, reasons for working part-time, and attendance of professional meetings. To examine job satisfaction, we used data on satisfaction with overall employment, job security, opportunities for advancement, salary, and benefits, from which we estimated the percentage of faculty who were satisfied, somewhat dissatisfied, or very dissatisfied by faculty type. Our analysis of part-time work first included whether a respondent who reported working part-time said they wanted to work full-time. Secondly, among those who wanted—and who did not want—to work full-time, we calculated the percentage who said they worked part-time (1) for family reasons, (2) because a full-time job was not available, (3) because they did not need/want full-time work, and (4) because they were a student, had an illness, or held another job. Respondents could indicate more than one reason for working part-time. We also analyzed a variable on attendance of professional meetings to calculate the percentage of faculty, by faculty type, who reported attending professional association meetings or conferences during the past 12 months. The SDR data included other variables that identify a respondent’s academic position, such as research faculty, administrators, adjuncts, and others. We analyzed these variables to determine whether to use them to categorize faculty, but found that they were not the most appropriate for our purposes. However, we observed that these variables may have implications on the economic circumstances of different types of faculty and so used them as control variables in two of our regression models on annual earnings. For example, we analyzed earnings of instructional faculty who said they were “adjunct” faculty or administrators. Among full-time and part-time contingent faculty, estimated median annual earnings decreased when we included only faculty who said that they were adjunct faculty (see table 13). However, the data do not allow us to explain how or whether the positions for faculty who identified as adjuncts are different compared to the positions of those who did not identify as adjuncts, and, based on our team’s interviews with administrators, different institutions and individuals apply different meanings to the term “adjunct.” As may be expected, among full-time tenure-track and full-time contingent faculty, estimated median annual earnings increased when we limited the population to only those faculty who said they were administrators (see table 13). State Data Analyses of Makeup and Utilization We used consistent methods to analyze data from Georgia, North Dakota, and Ohio on faculty workforce makeup and utilization, though we analyzed the data from each state separately. In addition, while each state dataset was structured slightly differently, used different variable names, and contained some unique elements or ways of capturing information about faculty or courses, we restructured and compiled the information to provide consistency across the states. In the state data, we identified instructional faculty as any individual who taught a course during the given academic year. This definition includes a variety of staff (e.g., deans, administrators, coaches, research faculty, and postdocs) who fill about 2-10 percent of positions, depending on institution type and state. In addition, instructional graduate assistants— who are listed in the state data as instructors of record—fill about 8 to 15 percent of positions at 4-year institutions in the three states. Each state’s data were ultimately structured as a set of unique faculty- institution pair observations—where faculty were listed once, by their employing institution. Each faculty-institution pair observation had variables describing the faculty member’s and institution’s characteristics, as well as counts of courses, students, and student credit hours taught by the faculty member at that institution (including by academic term and by course characteristics). Faculty Data Compilation and Restructuring For all three state datasets, we coded and grouped certain faculty characteristics variables, including academic rank, age group, race/ethnicity, sex, and tenure status, to ensure consistency across states. For example, in coding tenure status, we consistently categorized faculty as “non-tenure-track” if they were identified in the source data as not in a tenure-track position, as having been denied tenure, as being in some other status, or as being in a position for which tenure was not applicable. Some faculty characteristics variables were structured differently in each of the three states and thus required unique methods of recoding, though we applied consistent approaches and logic in each case (see table 14). We also identified each individual’s academic discipline based on information provided in each state’s data about their department. Faculty members’ departments in the Georgia and Ohio data are identified by their standardized Classification of Instructional Programs (CIP) code. The North Dakota data did not include the CIP code for faculty members’ departments and department names in the North Dakota data were not consistent across institutions. Thus, we coded North Dakota departments by matching them manually to corresponding CIP codes. After manually assigning CIP codes to faculty in the North Dakota data, we identified the highest level 2-digit CIP code for each faculty member in all three state datasets. However, because the 2-digit CIP code identifies over 40 fields of study, we grouped these by academic discipline for our analyses. To group departments, we used a crosswalk provided by Ohio that listed CIP codes according to 12 possible disciplines they were most closely associated with. Although the Department of Education’s CIP coding system does not include a commonly accepted list of disciplines, we determined that Ohio’s convention was reasonable and we applied the coding consistently across all three states to identify the academic discipline of each individual. The North Dakota data included multiple observations for some faculty members within a single institution and term. This occurred for a variety of reasons, such as a faculty member holding two positions at the same institution (e.g., both a coach and an instructor, or half time as an instructional graduate assistant and half time as a research graduate assistant). To compile a consistently structured dataset of unique faculty- institution pair observations, we implemented the following sequential process to select and eliminate duplicate faculty observations. We confirmed with North Dakota officials that our approach and methods were appropriate. For faculty with multiple observations, we dropped any observations where (1) no earnings were listed in any term or earnings were only listed for the summer term but the faculty member taught no courses at the given institution in the summer; or (2) the work responsibilities associated with the faculty observation were not directly related to teaching (e.g., graduate assistant research or grading, management, administration, research, or coaching) and a different observation for that faculty member at the same institution had teaching duties listed. We dropped these duplicate observations because there was a more appropriate observation to be used for the given faculty member at the given institution with earnings information and an associated instructional position. For the remaining faculty with multiple observations, we sequentially kept one observation as the primary faculty position based on hierarchical logic we developed. For example, we dropped any additional observations with an employee status other than “active” or a position identified as “temporary.” As appropriate, we either aggregated hours worked and earnings across the multiple observations before dropping the duplicate observations or we took the hours worked and earnings values from the observation identified as primary. Course Data Compilation and Restructuring Course data from all three states included each unique course section taught over the academic year by institution, term, and faculty instructor. We analyzed course sections for which there was an instructor identified and enough information about that faculty member to categorize them by faculty type (e.g., full-time tenure-track versus part-time contingent, etc.). For all three states, we aggregated these data by course type and other information to the level of the unique faculty-institution pair. For example, a single faculty member at a single institution may have taught 10 course sections, all at the undergraduate level and spread across the year—4 in fall term, 4 in spring term, and 2 in summer term. Courses are listed in the state data at both the course number level (e.g., Biology 101) and the course section level (e.g., Biology 101, Sections A, B, and C). Our analyses generally examined unique course sections by faculty member (e.g., two separate sections of Biology 101 are considered as two courses), as that is a more accurate depiction of faculty workload. Thus, for consistency and clarity throughout our report, we use the term “courses” to refer to our analyses of course sections. In a few special circumstances, we counted courses at the course number level instead of the course section level to minimize potential bias in our work (see additional information below). The course data included information about courses that we systematically coded and grouped to ensure consistency across the three states. For example, each state identified the academic level of each course. The Georgia and North Dakota data identified courses along a spectrum—generally developmental, freshman, sophomore, junior, senior, or graduate. The Ohio data had a different classification series: Developmental: All courses which are below college level General Studies: All courses which are general, introductory, or core Technical: Only those courses which are part of an associate degree program of technical education and are within the technical portion of a curriculum Baccalaureate: All courses which are specialized within a discipline Master’s / Doctoral / Professional – All graduate courses of various To categorize undergraduate course levels consistently across the states, we identified courses as (1) undergraduate lower if they were at the freshman, sophomore, general, or technical levels; or (2) undergraduate upper if they were at the junior, senior, or baccalaureate levels. Developmental and graduate courses were identified consistently in each state’s data. We made a number of decisions about how to categorize and count courses consistently across institutions and states. For example, we dropped cancelled courses or courses with no student enrollment. We also excluded from our primary analyses courses that would likely be student-led or student-initiated and thus could be considered atypical courses. We excluded these courses to minimize the potential bias of inflating the percentage of courses taught and deflating the earnings per course of one faculty type relative to another. After reviewing course types and titles, as well as associated student enrollment numbers and credit hours, we identified courses that met this definition and categorized them as atypical. Among the courses we identified as student-led or student-initiated were: Art or musical exhibitions, performances, or recitals Independent, supervised, dissertation, or thesis research Internships, fieldwork, practicums, cooperative experiences Varsity athletics These atypical courses made up close to a quarter of all courses across 4-year institutions in the three states and less than 10 percent of courses at 2-year institutions. As expected, and due to many being independent or single-student enrollment courses, they generally represented much smaller proportions of student credit hours across all institutions. Across 4-year public institutions in all 3 states, tenure-track faculty taught close to 75 percent or more of these courses. We also accounted for cross-listed courses and multiple lab sections to more accurately capture faculty workloads. Some courses in the Georgia and North Dakota data were cross-listed in multiple departments with different course acronyms for each department. For example, the course “Intro Robotics Research” taught by a single faculty member at one institution was listed three times under different department acronyms, with several students enrolled under each listing. Course sections listed multiple times due to being cross-listed would artificially inflate counts of courses taught, as these cross-listings actually represent only one course section. To avoid inappropriately counting them as separate courses, we counted cross-listed courses by using their course numbers (and also their course name in North Dakota) without the course acronyms attached. Thus, when we aggregated counts of courses by faculty- institution pair, term, and course type, these cross-listed courses were counted as one course and numbers of students and student credit hours were aggregated in association with the course. Due to inconsistencies in how lab sections were organized in the data, we aggregated labs by their course number (within a faculty-institution pair and term). For example, some lab sections were listed as 4-credit courses that appeared to have the lecture and lab components combined in a single listing, while others had a 3-credit lecture course listed and multiple sections of a 1-credit lab. To be as consistent across states as possible and to minimize the potential bias of inflating the percentage of courses taught and deflating the earnings per course of one faculty type relative to another, we combined lab sections into a single course count. To do so, we identified the lab sections within a particular course number, instructor, institution, and term and then flagged the first lab section for counting. Thus, similar to the cross-listed courses, when we aggregated counts of courses by faculty-institution pair, term, and course type, these lab sections were counted as one course and enrollment numbers aggregated in association with the course. For outlier faculty who taught especially large numbers of course sections, we counted their courses taught at the course number level (e.g., Biology 101) instead of the course section level (e.g., section 1 of Biology 101). After compiling the data and producing preliminary counts of course sections taught, some faculty in all three states emerged as outliers—teaching large numbers of course sections in a given term, in some cases, more than 50, for example. Though the data do not provide exact reasons for the large numbers of course sections taught, these outliers may have a number of possible explanations that could vary by state and institution. Among other effects, these outlier observations could artificially inflate the percentage of courses taught and deflate the earnings per course of one faculty type relative to another. To mitigate these effects, we counted courses taught for these outlier faculty at the course number level—where they are clearly distinct—instead of the course section level—where it is less clear why there are multiple sections. For example, Biology 101 is clearly a different course than Biology 201 or Chemistry 101 (regardless of section number), whereas section A of Biology 101 could actually be combined with section B and they are just listed separately for other reasons. We did not set a maximum number of courses that an individual could teach (i.e., individual faculty could still be listed as teaching large numbers of courses if they were associated with large counts at the course number level). We counted course numbers for outlier faculty because their large numbers of course sections listed suggested the possibility of a data anomaly; for all others (non-outlier faculty), we counted course sections. We set our outlier threshold as 15 course sections taught over the academic year based on an examination of the range of course sections taught by faculty in the three states’ data and conversations with administrators during our site visits. According to preliminary counts of course sections taught after excluding atypical courses, more than 95 percent of faculty in each state taught 15 course sections or fewer over the entire academic year. In addition, during our site visits, the largest number of course sections taught per term that administrators identified was 6, which could reasonably result in 15 course sections over the year (6 in fall term, 6 in spring term, and 3 in summer term—half the amount due to the condensed format). Analysis of Faculty Makeup and Utilization To analyze faculty makeup and utilization by institution, we merged information about institutional characteristics from IPEDS onto our state datasets. We analyzed faculty makeup, including counts and percentages of faculty positions by type of position and faculty characteristics (e.g., age, education, and academic discipline), by the following faculty categories (based, in part, on faculty tenure and work statuses): We sometimes analyzed full-time and part-time contingent faculty and instructional graduate assistants combined as “contingent faculty” and full-time and part-time tenure-track combined as “tenure-track faculty.” Unlike our analyses of IPEDS data, we included instructional graduate assistants in our combined contingent faculty group because they were listed as teachers of record for courses in the state data. We analyzed administrators/management as a separate group because these individuals represent a non-traditional class of faculty. For example, administrators may not have tenure-track status due to their management roles, but are in positions that may not be appropriate to be considered “contingent” (e.g., a dean might not be a tenure-track faculty member, but neither are they a contingent faculty member). We analyzed educational attainment of faculty by calculating the percentage of faculty with graduate or doctoral degrees by faculty type and institution type in in North Dakota and Ohio. Table 15 shows the total number of instructional faculty positions by institution type in each state, as well as selected faculty demographics. We analyzed faculty utilization by aggregating counts of courses, students, and student credit hours taught by each faculty category above, and by term and type of course, and by calculating percentages taught out of the entire population and certain subgroups. As a first step in this process, we aggregated counts of courses, students, and student credit hours for each faculty-institution pair by term and type of course. As a result, each faculty-institution pair had count variables that listed, for example, how many courses and students they taught in fall term at the undergraduate upper level. The Georgia and Ohio data listed courses multiple times if multiple faculty share the instructional responsibility. To ensure course sections were not double-counted, we counted them in fractional terms based on how many instructors were listed; for example, if a course section was listed twice—with two faculty members having equal responsibility for the course—we counted each faculty member as teaching half of that course. We also used this fractional count to pro-rate or assign responsibility for students and student credit hours. We calculated this fractional count slightly differently for the Georgia and the Ohio data: Georgia: The Georgia data provided a teaching responsibility percentage for each faculty member associated with a course section. For example, a course section that was listed 3 times (for 3 different faculty with responsibility) might be split evenly 1/3-1/3-1/3 or might be split as 50-30-20 percent responsibility to each of the three faculty members. Thus, we used this individually provided fractional value. Ohio: The Ohio data did not provide a teaching responsibility percentage for each faculty member associated with a course section. Thus, we assigned equal responsibility (as the simplest assumption) to all staff listed for a course. After aggregating counts to the faculty-institution pair level, we further aggregated counts to the faculty category and institution type level. Our analyses focused on counts and percentages of courses and student credit hours by these faculty categories. Table 16 shows the total number of courses taught by institution and faculty types in each state. We also analyzed economic circumstances by examining median annual earnings and receipt of work-provided retirement, health insurance, and life insurance benefits by faculty type. We calculated an annual earnings amount for each faculty member and then analyzed median earnings by faculty type. For benefits, we identified whether individual faculty received a given benefit during the year, and then calculated the percentage of each faculty type receiving those benefits. We were unable to analyze benefits in this way for faculty in Ohio. See table 14 above for additional details about our earnings and benefits calculations by state. HDS Analyses of Faculty Makeup To estimate population percentages by faculty type and discipline in humanities departments at 4-year institutions, we used HDS data that were published in a technical report sponsored by AAAS. Our population of instructional faculty included faculty in humanities departments at 4-year institutions. The sample was stratified by discipline and degree level of courses taught (i.e., bachelor’s, master’s, and doctoral degree courses). We were unable to access the data with the sample design information (i.e. sampling weights and stratification identifiers) necessary to calculate margins of errors that took into account the sample design features. To allow us to estimate margins of error for the estimates presented in the report, AAAS provided information on the number of respondents associated with each response category since the survey had unit and item nonresponse. We incorporated this information into a simple random sampling formula, which we adjusted for the design effect due to unequal weighting that resulted from stratification within departments (e.g., differences in the extent to which departments may offer bachelor’s, master’s, and doctoral degree courses). Section 3: Pay per Course Regression Analysis (State Data) This section discusses the regression analysis methods we used to analyze and compare pay-per-course rates across different types of faculty at public institutions in North Dakota and Ohio. We used data from the three states to conduct multivariate regression analyses that examined rates of compensation across faculty types. Data from North Dakota and Ohio allowed us to link faculty members’ pay over the course of an academic year with the number of courses they taught to calculate pay-per-course rates that are comparable across faculty types. Data from Georgia did not allow us to do this because the earnings data from Georgia is for a calendar year that did not align with the course data for the academic year. However, we used Georgia’s data to develop assumptions about faculty work activities (see below for more details). The state data we used to analyze pay-per-course rates covered courses taught and earnings from fall 2015 through summer 2016 for North Dakota, and summer 2014 through spring 2015 for Ohio. Analysis Population The faculty populations included in our regression analyses of the North Dakota and Ohio data begin with the same population of instructional faculty analyzed elsewhere in our work—any individual who teaches a course at a 4-year or 2-year public institution in the state. However, due to some faculty observations missing information for independent variables, as well as the specifications of some of our models that focused on subgroups within the data, the number of faculty observations in our regression analyses differed slightly from those in our other analyses. In assessing the association between faculty type (e.g., contingent faculty) and pay per course, we focused on three primary populations: (1) all faculty; (2) primarily teaching faculty; and (3) primarily teaching faculty at 4-year institutions. The primarily teaching faculty group excludes faculty who primarily hold other roles unrelated to instruction (e.g., administrators and research faculty). We also examined a population limited to 4-year institutions because their pay and faculty utilization structures may differ substantively from 2-year institutions. North Dakota: Compared to the 3,608 faculty observations with complete faculty and course identification data across North Dakota public institutions that we analyze for workforce makeup and utilization, the number of observations included in our regression analysis population is reduced to 3,486 due to our dropping of cases where total earnings was less than one dollar or missing, or where the number of in-scope courses taught was zero (more information below under discussion of dependent variables). After introducing the full range of independent variables in our complete model with all faculty at all institutions, our population is reduced to 3,485 due to one faculty member being omitted due to missing data. When we limit the population to primarily teaching faculty at all institutions, there are 3,404 observations, and when we only include 4-year institutions, there are 2,876 observations. Ohio: Compared to the 34,461 faculty observations with complete faculty and course identification data across Ohio public institutions that we analyze for workforce makeup and utilization, the number of observations included in our regression analysis population is reduced to 30,672 due to our dropping of cases where total earnings was less than one dollar or missing, or where the number of in-scope courses taught was zero (more information below under discussion of dependent variables). After introducing the full range of independent variables in our complete model with all faculty at all institutions, our population is reduced to 30,656 due to 16 faculty members missing data for covariates. When we limit the population to primarily teaching faculty at all institutions, there are 28,811 observations, and when we only include 4-year institutions, there are 21,482 observations. Approximating Instructional Pay from Georgia Data on Faculty Work Activities As explained earlier in the report, we examined instructional pay per course as a way to isolate the earnings for comparable work across faculty types—for example, those who only teach (salaried or paid by the course) versus those who have other responsibilities beyond teaching. Institutions do not generally structure compensation by types of work activities, though some faculty have work responsibility expectations associated with their positions; for example, a full time tenure-track assistant professor may have work responsibly expectations of 60 percent instructional, 30 percent research, and 10 percent other service to the institution. If this faculty member earns $80,000 per year and teaches 8 courses over the course of the year, her total pay per course, which ignores time spent on research and other activities, would be $80,000/8 = $10,000 per course. However, prorating the earnings to those for instructional work activities only, the instructional pay per course would be ($80,000*0.6)/8 = $6,000. We assessed each regression model based on the outcomes of total pay per course and instructional pay per course, where earnings were prorated for instructional time. Because information about faculty work activity was unavailable in the North Dakota and Ohio data, but was available in the Georgia data, we used empirical data that we received on four of the Georgia 4-year public institutions to identify work activity percentages by faculty type. We then assigned those percentages to similar faculty in North Dakota and Ohio. We identified the median instructional work activity percentages for the faculty in Georgia’s 4-year public institutions within profiles based on a combination of faculty characteristics including faculty category (e.g., full- time tenure-track, full-time non-tenure-track, part-time non-tenure-track, etc.), job category (e.g. administration/management, teaching faculty, research/other faculty, etc.), and when applicable, rank (e.g. full professor, assistant professor, instructor/lecturer, etc.). We then applied the median instructional work activity percentage from the Georgia data by these profile groups to faculty at 4-year institutions in the North Dakota and Ohio data with the same profile. For faculty in the job categories of administrators/management staff, instructional graduate assistants, coaches, and postdocs, the median instructional work activity percentage in those groups overall was sufficiently explanatory. For the remaining two job category groups of instructional faculty and research/other faculty, we used median work activity percentages by faculty category (e.g., full- time tenure-track) and rank (e.g., full professor). If a faculty member did not have a rank identified in the data, we used the median work activity percentage for the faculty category overall (see table 17). Because the data on work responsibilities pertained to public 4-year institutions in the Georgia data, we did not prorate faculty at 2-year institutions accordingly. Because 2-year institutions generally do not have a research mission, we coded all faculty at 2-year institutions as 100 percent instructional, except for administrators/management staff. We prorated administrators/management staff according to the same method as at 4- year institutions due to their likely having substantial non-teaching responsibilities. Faculty earnings in the North Dakota and Ohio data were multiplied by the relevant median instructional work activity percentage in order to adjust pay to reflect instructional work activity, resulting in an “instructional pay” amount. The majority of adjustments—prorating of earnings to account for non-instructional activities—were applied to faculty in the full-time tenure-track group, who were most likely to have other work responsibilities. Some adjustment to earnings also occurred in the full- and part-time contingent groups, as well as for faculty who had a job type that indicated substantial administrative and management roles. No prorating occurred for instructional graduate assistants. Dependent Variables We conducted regressions using the following dependent variables: a) Log (total pay per course) – In our analysis of the North Dakota and Ohio data, we used the natural logarithm of the total pay per course, which is defined as the total annual earnings (i.e., total pay) divided by the total courses taught within that year. b) Log (instructional pay per course) – In our analysis of the North Dakota and Ohio data, we also used the natural logarithm of the instructional pay per course, which is defined as total annual earnings adjusted to reflect instructional work activity (i.e., instructional pay) divided by the total courses taught within that year. We excluded cases from our analysis if they were missing values for either total annual earnings or total courses taught within that same year because these variables were the primary components of pay per course. We dropped cases where total earnings were less than one dollar or missing (19 observations in North Dakota and 2,869 observations in Ohio) or the number of courses taught was zero (103 observations in North Dakota and 920 observations in Ohio) since division by zero is undefined, and our population is intended to reflect any individual who actually teaches a course at 4-year and 2-year public institutions in the state. We then divided pay (total or instructional) by the number of courses taught to obtain the pay-per-course value. We use the log of total and instructional pay per course for the dependent variables in a linear model reflecting both the assumption that the underlying distribution is closer to the log normal than normal, and also to present results in terms of percentage changes in pay per course. In the Ohio data, because we use fractional counts for courses when multiple faculty are listed as having responsibility for the course, 3,453 faculty in the analysis population teach less than 1 course. For those faculty, we round all course counts that are less than 1 up to 1 to avoid dividing faculty earnings by a fractional course count (between 0 and 1), which would result in an inaccurate and substantially large pay-per- course value. Independent Variables The primary independent variable of interest in our analysis was faculty type. We categorized faculty into five types: full-time tenure-track, full-time contingent, part-time tenure-track, part-time contingent, and graduate assistant. Our main interest was comparing contingent faculty and graduate assistants to full-time tenure-track faculty. We controlled for the part-time tenure-track group, but due to the small size of this population (at most, 35 faculty in North Dakota and 274 faculty in Ohio), we did not substantively examine these estimates. All regression models set the base group for faculty type as full-time tenure-track. We included in our regression models additional independent variables as controls for faculty and institution characteristics. Faculty characteristics include sex, race, age, age squared (to account for the potential non- linear relationship between earnings and age), highest degree earned, and academic discipline. Other faculty characteristics we controlled for in our models included whether a faculty member had grant funds (North Dakota only), whether a faculty member taught summer courses, and indicators identifying non-traditional faculty roles, such as administrators/management or coaches. We also included fixed effects for institutions to control for differences between institutions, especially in terms of pay due to factors such as size, sector, and research/graduate component, among other things. We also examined rank of faculty (e.g. associate professor, assistant professor, instructor/lecturer, etc.), but excluded it from our complete models due to collinearity with the faculty type variable. Regression Model Detailed Results Tables 18 and 19 (below) shows the coefficients and standard errors from each of our final pay-per-course regression models, as well as for the unadjusted model that included only the primary independent variable of interest (total pay-per-course results at the top and instructional pay-per- course results below). For our categorical variables, estimated coefficients are relative to the excluded (reference) category. For example, since the reference category for our main independent variable, faculty type, was full-time tenure-track, the estimated coefficients for other categories of this variable are always relative to this excluded reference category, holding all other variables in the model constant. Thus, in model 2 for North Dakota, the coefficient for full-time contingent faculty is 0.682. This can be interpreted as full-time contingent faculty pay per course is 0.682 that of full-time tenure-track faculty (i.e., full-time contingent faculty are paid 68.2 percent what full-time tenure-track are, per course), holding all other variables in the model constant. Because the dependent variables in the earnings models are the natural logarithms of earnings, subtracting one from the presented coefficients on categorical variables can be interpreted as the percentage change in the dependent variable associated with a change in the categorical variable, relative to the reference category, holding all other variables constant. In this same example, full-time contingent faculty are paid an estimated 31.8 percent less than full-time tenure-track faculty, because 0.682 – 1 = -0.318, or 31.8 percent less. Additional Analyses and Sensitivity Tests The North Dakota and Ohio data used in the regression analyses include a small number of faculty (1.1 and 0.5 percent of observations, respectively) who are listed as teacher of record for more than 15 courses over the year, which may represent unusually high workloads or data anomalies. In addition, some faculty have small or large pay-per-course values when compared to the overall distribution. To preserve the integrity of the data, we did not exclude these observations from the analyses. However, we tested our models with and without these observations to assess the effect on our substantive regression results. In order to assess the effect of faculty with a large workload, we conducted regression models 3 and 4 (in tables 18 and 19 above) limited to faculty who taught 15 or fewer courses over the year. In order to assess the effect of faculty with the outermost values of the dependent variable pay per course, we conducted the same regression models limited to faculty whose pay per course was within the middle 98 percent of pay-per-course values (i.e., we trimmed the bottom and top 1 percent of observations). In both of these sensitivity analyses, we found substantively similar results. We also ran our regression models on a more refined population that only included primarily teaching faculty at 4-year institutions (faculty at 4-year institutions represent most of our analysis population). As shown in table 18 above, in terms of total pay per course, full-time contingent faculty in North Dakota and Ohio are paid about 40 and 43 percent less per course, respectively, than full-time tenure-track faculty—compared to 35 and 40 percent less per course, respectively, when both 4-year and 2-year institutions are included. This slightly larger pay-per-course disparity as compared to the population overall may be, in part, because pay and utilization of full-time faculty vary somewhat by institution type (e.g., at 4- year institutions, pay is generally higher but less flat, and some full-time tenure-track faculty teach fewer courses due to their more extensive research responsibilities). Section 4: Annual Earnings Regression Analysis (SDR Data) This section discusses the regression analysis methods we used to analyze and compare annual earnings among different types of faculty using national 2013 SDR data on doctorate-holding faculty in the STEM, health, and social sciences fields. Dependent Variable We conducted regressions using the following dependent variable: Log (annual salary)—the natural logarithm of annual salary, defined as the basic annual salary from the respondent’s principal job. Independent Variables The primary independent variable of interest in our analysis was faculty type. We categorized faculty into five types: full-time tenure-track, full-time contingent, part-time tenure-track, part-time contingent, and graduate assistant. Our main interest was comparing contingent faculty to full-time tenure-track faculty. Though we controlled for the part-time tenure-track and graduate assistant groups, we did not substantively examine these estimates. All regression models set the reference group for faculty type as full-time tenure-track. We included in our regression models additional independent variables as controls for faculty and institution characteristics. Faculty characteristics included sex, race, age, age squared, number of weeks worked per year, and academic discipline. Other faculty characteristics we controlled for included the year of highest degree earned—which we used as proxy for general experience—and whether a respondent indicated that they were an administrator. We also included institution type (e.g., 4-year college or university, 2-year college or university). After introducing the full range of independent variables in our complete model, our analysis sample was reduced from 7,232 faculty respondents to 7,226 due to 6 faculty respondents being omitted due to missing data. We examined faculty rank (e.g. professoriate, instructor/lecturer) and academic position variables for “adjunct” faculty and postdocs, but we excluded these variables from our complete model, as we determined they did not have meaningful information for the purpose of our analyses. Regression Model Detailed Results In our complete model, full-time and part-time contingent faculty earned 22 percent less and 70 percent less, respectively, than full-time tenure- track faculty annually (see table 20). Across our preliminary models (not shown below) and complete model, the coefficients related to our main independent variable remained relatively constant, ranging from 0.76 to 0.86 for full-time contingent faculty and 0.26 to 0.43 for part-time contingent faculty, expressed as proportion of full-time tenure-track faculty earnings. Appendix II: IPEDS Data on the Racial and Ethnic Distribution of Faculty Positions Nationwide, 2015 Appendix III: GAO Contact and Staff Acknowledgments GAO Contact Staff Acknowledgments In addition to the contact named above, Nagla’a El-Hodiri (Assistant Director), Nisha R. Hazra (Analyst-in-charge), Sandra Baxter, Justin Gordinas, Michael Kniss, and Alexandra Squitieri made key contributions to this report. Also contributing significantly to this report were Melinda Cordero, Grant Mallie, Jean McSween, Moon Parks, and Sonya Vartivarian. Key support was provided by James Ashley, James Bennett, Grace Cho, Jessica Orr, James Rebbe, Almeta Spencer, and Elaine Vaurio. | Contingent faculty play a large role in postsecondary education but may not have the same job protections as tenured or tenure-track faculty. In 2015, GAO reported that contingent workers—those in temporary, contract, or other non-standard employment arrangements—earn less, are less likely to have work-provided benefits, and are more likely to experience job instability than standard workers. GAO was asked to examine issues related to contingent faculty. This report examines (1) what is known about the makeup and utilization of the postsecondary instructional workforce; (2) the roles different types of faculty fill at selected institutions and the factors administrators consider when determining faculty makeup; (3) what is known about how economic circumstances compare across different faculty types; and (4) what contingent faculty members report as advantages and disadvantages of their work. GAO analyzed data from nationally representative sources and from public institutions in three states—Georgia, North Dakota, and Ohio. GAO selected these states based primarily on data availability. GAO interviewed administrators from 9 postsecondary institutions in these states and one large for-profit institution. GAO selected institutions based on factors such as institution size and percent of contingent faculty. GAO also conducted 21 discussion groups with contingent faculty. The Department of Education did not have comments on this report. The National Science Foundation provided technical comments, which we incorporated, as appropriate. According to 2015 Department of Education data, contingent faculty—those employed outside of the tenure track—made up about 70 percent of postsecondary instructional positions nationwide, though this varied by type of institution. In addition, data from three selected states show that contingent faculty teach about 45 to 54 percent of all courses at 4-year public institutions, and higher proportions at 2-year public institutions. In terms of job stability, some full-time contingent positions with annual or longer contracts may be relatively stable while part-time positions with short-term contracts may be among the least stable, though it is unknown whether faculty in these positions have other employment. In contrast, tenure-track positions are often viewed as having a high degree of job security that is somewhat unique to postsecondary education. Administrators GAO interviewed at selected postsecondary institutions said full-time contingent faculty generally carry heavy teaching loads, and some also take on additional responsibilities, such as conducting research or advising students. However, administrators stated that part-time contingent faculty generally focus solely on teaching. As shown in the figure below, administrators also described factors they consider in determining their institution's faculty makeup. GAO examined recent data from North Dakota and Ohio public institutions and found that, among faculty who primarily teach—which excludes individuals such as administrators or researchers—part-time and full-time contingent faculty were paid about 75 percent and 40 percent less per course, respectively, compared to full-time tenure-track faculty. This comparison includes earnings for all of their responsibilities, including teaching and any other duties. However, when estimating faculty earnings for teaching duties only, pay disparities decreased to about 60 percent and 10 percent less per course for these contingent faculty, respectively. In addition, state and national data also showed that relatively few part-time contingent faculty received work-provided health or retirement benefits. In discussion groups with GAO, contingent faculty cited advantages such as the flexibility to balance professional and personal responsibilities, skill development, or working with students, and described disadvantages that included uncertainty due to short-term contracts, untimely contract renewals, and pay—including a lack of compensation for some of their work. Other concerns they cited included limited career advancement opportunities, not having a voice in institutional decision-making, and not having certain types of institutional support. | [
2,
109,
2650,
757,
23811,
3142,
307,
1450,
731,
113,
109,
4911,
107,
116,
110,
107,
5596,
113,
798,
110,
108,
1378,
110,
108,
1537,
110,
108,
1578,
110,
108,
111,
552,
1574,
120,
109,
344,
113,
3729,
115,
109,
4911,
107,
116,
110,
107,
117,
1222,
134,
114,
2347,
872,
197,
109,
1948,
130,
114,
664,
110,
108,
2650,
757,
23811,
740,
110,
107,
106,
109,
2650,
757,
23811,
3142,
307,
1450,
731,
113,
109,
4911,
107,
116,
110,
107,
110,
106,
2174,
113,
798,
110,
108,
1378,
110,
108,
1537,
110,
108,
1578,
110,
108,
111,
552,
1574,
120,
109,
344,
113,
3729,
115,
109,
4911,
107,
116,
110,
107,
117,
1222,
134,
114,
2347,
872,
197,
109,
1948,
130,
114,
664,
110,
108,
2650,
757,
23811,
740,
110,
107,
106,
145,
10082,
109,
2650,
757,
23811,
3142,
307,
1450,
731,
113,
109,
4911,
107,
116,
110,
107,
110,
106,
2174,
113,
798,
110,
108,
1378,
110,
108,
1537,
110,
108,
1578,
110,
108,
111,
552,
1574,
111,
374,
120,
109,
344,
113,
3729,
115,
109,
4911,
107,
116,
110,
107,
117,
1222,
134,
114,
2347,
872,
197,
109,
1948,
130,
114,
664,
110,
108,
2650,
757,
23811,
740,
110,
107,
106,
145,
10082,
109,
2650,
757,
23811,
3142,
307,
1450,
731,
113,
109,
4911,
107,
116,
110,
107,
110,
106,
2174,
113,
798,
110,
108,
1378,
110,
108,
1537,
110,
108,
1578,
110,
108,
111,
552,
1574,
111,
374,
120,
109,
344,
113,
3729,
115,
109,
4911,
107,
116,
110,
107,
117,
1222
] |
GAO_GAO-19-223 | "Background EM’s cleanup sites and areas of cleanup work, EM’s status as a program, the history (...TRUNCATED) | "EM's mission is to complete the cleanup of nuclear waste at 16 DOE sites and to work to reduce risk(...TRUNCATED) | [2,136,800,3702,114,437,692,113,199,112,1735,109,713,113,2160,13963,116,120,137,129,4175,1291,190,10(...TRUNCATED) |
CRS_R44801 | "Background Congress has been interested in disaster relief since the earliest days of the republic.(...TRUNCATED) | "The Robert T. Stafford Disaster Relief and Emergency Assistance Act (the Stafford Act) authorizes t(...TRUNCATED) | [2,145,1037,109,906,113,682,186,117,114,1074,5448,124,109,2335,564,113,142,4058,115,114,440,1432,407(...TRUNCATED) |
GAO_GAO-18-469T | "Experts and Stakeholders Have Proposed Restructuring EOIR’s Immigration Court System As we report(...TRUNCATED) | "DOJ's EOIR is responsible for conducting immigration court proceedings, appellate reviews, and admi(...TRUNCATED) | [2,145,731,124,109,602,113,150,2843,113,109,637,113,142,5806,1462,110,107,106,109,2843,140,3047,115,(...TRUNCATED) |
CRS_R41146 | "Small Business Administration Loan Guaranty Programs The Small Business Administration (SBA) admini(...TRUNCATED) | "The Small Business Administration (SBA) administers several programs to support small businesses, i(...TRUNCATED) | [2,145,1037,109,360,111,1961,233,5555,260,143,110,116,1592,1271,110,158,83572,407,110,107,106,110,11(...TRUNCATED) |
GAO_GAO-18-611T | "The Number of SFSP Meals Served Generally Increased from 2007 through 2016, but Estimates of Childr(...TRUNCATED) | "This testimony summarizes information contained in GAO's May 2018 report entitled Summer Meals: Act(...TRUNCATED) | [2,6137,141,109,1222,344,113,404,115,217,113,399,3215,110,108,109,211,5333,111,549,113,109,2650,757,(...TRUNCATED) |
GAO_GAO-18-46 | "Background As we have previously reported, transportation systems and facilities are vulnerable and(...TRUNCATED) | "TSA conducts security threat assessment screening and credentialing activities for millions of work(...TRUNCATED) | [2,145,731,124,109,602,113,142,3244,190,109,5800,113,109,5389,113,35462,335,735,141,1162,4755,112,10(...TRUNCATED) |
CRS_R45723 | "T he federal government has two major tools for affecting the macroeconomy: fiscal policy and monet(...TRUNCATED) | "Fiscal policy is the means by which the government adjusts its spending and revenue to influence th(...TRUNCATED) | [2,109,2650,757,23811,3142,307,1368,1968,115,109,4911,107,116,110,107,110,108,992,112,109,4911,107,1(...TRUNCATED) |
GAO_GAO-19-38 | "Background HUD established DEC in 1998 to consolidate enforcement activities of PIH, CPD, the Offic(...TRUNCATED) | "HUD established DEC in 1998 to consolidate enforcement functions. In fiscal year 2017, DEC received(...TRUNCATED) | [2,145,5935,109,4156,112,162,449,111,391,2662,133,109,986,112,1735,682,132,146,114,431,117,5745,203,(...TRUNCATED) |
CRS_R42072 | "Introduction The leaders of the eight legislative branch agencies and entities—the Government Acc(...TRUNCATED) | "The leaders of the legislative branch agencies and entities—the Government Accountability Office (...TRUNCATED) | [2,145,731,109,602,113,142,3244,190,109,1895,113,181,113,109,549,113,109,10381,1653,25164,170,195,10(...TRUNCATED) |
End of preview. Expand
in Dataset Viewer.
Dataset Card for AutoTrain Evaluator
This repository contains model predictions generated by AutoTrain for the following task and dataset:
- Task: Summarization
- Model: google/bigbird-pegasus-large-arxiv
- Dataset: launch/gov_report
- Config: plain_text
- Split: validation
To run new evaluation jobs, visit Hugging Face's automatic model evaluator.
Contributions
Thanks to @nonchalant-nagavalli for evaluating this model.
- Downloads last month
- 34