Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
PAST AS PROLOGUE If the preparation of college teachers and the national distribution of graduate study are the two major issues in graduate education today, then the duration of doctoral study is probably the third. The critics who fear that the system is going to turn out too few doctorates in the years ahead, those who believe that the whole emphasis on research is wrong, those who think that the degree has fallen off from traditional standards, even those who want things added all of them are concerned about the lengthy period of doctoral study. There is hardly a recent discussion of graduate education in which this note is not played loud and strong. (Berelson, 1960:1563 What Has Happened to Time to the Doctorate? Total Time to the Doctorate Despite ample evidence that t~1 l) has been increasing for years, public attention to the question of how long it should take to complete the doctorate has diminished. The extent of the change in TTD between 1960 and the present is highlighted by a comparison of Berelson's data with data from this study (Table 7.1~. If current trends persist, it will take even longer for doctorates to complete their degrees in the future. This is an important conclusion because it suggests that the question of whether doctoral preparation could, or should, be expedited may again become a matter of great interest. Unfortunately, Berelson lacked the data to study long-term changes in RTD. His study used data from only one year and focused on the difference between these two variables and I-lL). It found that RTD was lower than ells in 93
TABLE 7.1: Median Total Time to die Doctorate Over Time Berelson Doctorate Records File Aggregated Field 1936 1957 1967 1977 1987 1997 Physical Sciences 6 6 5.9 6.9 7.1 7.5 Biological Sciences 6 7 6.7 6.9 8.0 8.4 Social Sciences 8 8 7.6 7.9 10.4 11.2 NOTE: The figures for 1997 are estimated using a simple time-end model. each of eight fields under study.l7 Of particular note, according to Berelson, was the fact that the time differences among fields were small when actual time to the doctorate was considered.l8 He concluded that "the problem is not how much time a student should spend in working on his degree, but rather over how long a period of time he should do it" (Berelson, 1960:162~. Registered Time to the Doctorate Because RTD data are available for both 1967 and 1986, it is possible to look at RTD over time. In all 11 fields, it increased, sometimes by a large -amount. In seven fields, RTD increased more than TTD between 1967 and 1986. For example, RTD rose by 49 percent in the social sciences, compared to a 22-percent increase in TTD; in economics, the comparable figures were 37 percent and 4 percent; in earth, atmospheric, and marine sciences, 28 percent and 14 percent; and in agricultural sciences, 22 percent and 8 percent. In three fields, RTD and TTD increased by a similar percentage: about 28 percent in psychology; 13 percent in physics; and 29 percent in math and computer sciences. Only in the health sciences did the change in l-l L) (27 percent) greatly exceed the change in RID (14 percent) between 1967 and 1986. These findings suggest that, with the exception of one field, the major source of increasing 11D was a "stretching-out" of the time spent registered in graduate school 17 These fields are physical sciences, biosciences, social sciences, humanities, engineering, education, arts and sciences, and professional fields. 18 The lowest median actual time was in education (2.8 years) and the highest was in social sciences (3.7 years). 94
The differences among fields in RTD described by Berelson can be examined for more recent years using data from the DRF. For both 1967 and 1986, the difference in median RTDs across fields is less than the difference in median TTDs, affirming Berelson's findings. The range in TTD between the high and low fields increased 1.9 years from 1967 to 1986. The lowest mean TTD in 1967 was 6.4 years (for chemistry) and the highest was 10.6 years (for the social sciences). The range between the low and high fields, therefore, was 4.2 years. In 1986 the field with the lowest mean (7.2 years) was again chemistry, but the field with the highest mean was health sciences (13.3 years). In this case the difference between the two fields was 6.1 years. The range in RTD also grew between 1967 and 1986, but that growth was less than that experienced by TTD. In 1967, chemistry had the low mean RTD (5.0 years) and health sciences had the high mean (6.5 years). The range between the two is 1.5 years. In 1986, the low field was still chemistry with a mean RTD of 5.8 years; the high field was psychology, with a mean of 7.5 years. The difference between the two fields is 1.7 years, compared to 6.1 years using the TTD measure, and the range between high and low fields for RTD grew by 0.2 years from 1967 to 1986, far less than the 1.9 year growth observed using 'l-lL). Thus, although Berelson found that the RTD measure produced a smaller difference across fields, he failed to see that the range was increasing over time, suggesting the doctorate is growing relatively more costly in certain fields in terms of lost income while in graduate school. Variation Around the Mean To determine whether within-field differences in the time students took to earn the doctorate narrowed or grew larger between 1967 and 1986, coefficients of variation (CVs~l9 were computed for each field. The results show that the within-field variation in both RTD and TTD was at least as large as between- field variations in some fields, raising the question of whether the type of field comparisons offered by Berelson are useful. In all 11 fields, the CVs for TTD decreased from 1967 to 1986. However, the CVs for mean RTD increased in four fields, remained the same in two, and fell in five. This indicates a larger proportion of doctorate recipients 19 The coefficient variation is the standard deviation divided by the mean. It is used to express variation in the data relative to the mean and facilitates comparison of variation across fields. 95
had lands closer to mean T1o in 1986-that is, mean '1-1D was representative of a larger percentage of the cohort-than was the case in 1967 and a larger proportion of the 1986 than 1967 doctorate cohort took longer time to complete the doctorate (this was also true for the five fields in which the CVs for RID fell20 ). The lengthening of time to the doctorate is affecting a larger percentage of doctorate recipients than was true in the past. Could Changes in TPGE and TNEU Have Been Large Enough To Explain the Change in TTD? The data suggest that time prior to entry to graduate school (TPGE) rose in all fields except EAM and agricultural sciences. The size of the increase depended on the field studied, with three fields showing an increase of less than 10 percent, in thee a jump of 11-50 percent, and in three a rise of 60-105 percent. The largest increases in TPGE were in math and computer sciences (105 percent) and the health sciences (100 percent), while the smallest were in economics (5 percent) and the social sciences (S percent). Measured in absolute terms, the increases in TPGE were fairly small. In six of the nine fields in which TPGE grew, the increase amounted to less than three months. Three other insights emerge from a study of TPGE. First, the low TMEs for most fields in 1986 suggest that most doctorate recipients entered graduate school soon after completing the baccalaureate. And, while TPGE rose in a majority of fields, the increase was not great enough to explain more than a small fraction of the increase in AD between 1967 and 1986.21 Three of the four fields with large increases in 1-~D also had large increases in URGE: health sciences, math, and psychology. However, even in these fields, the rise in TPGE was not large enough to be the prime source of the increase in AID. Third, the data also suggest that changes in TNEU were not responsible for the growth in AD in most fields. TNEU decreased in eight fields, and in five of these the decrease was greater than three months. TNEU rose by two-and-a-half months in 20 The coefficient of variation dropped by 10 percent in health sciences, by 6 percent in social sciences, by 4 percent in psychology, by 5 percent in the biosciences, and by 1 percent in chemistry. 21 For example, the rise in TPGE represented 19 percent of the growth in chemistry, 22 percent in math, 25 percent in psychology, and 37 percent in health sciences. 96
math and by nearly a year in health sciences; however, only in the latter was the combined effect of changes in TPGE and TNEU large enough to have a large impact on AID. In fact, the decline in TNEU in many fields helped to offset the relatively small rises in TPGE, causing RTD to become the major source of change in I-rD. These findings suggest that the concern expressed in the 1960s over the amount of time students spend "outside the system" is not valid at present (Wilson, 1965~. Possible Explanations Six broadly based theories may explain the growth in AID. These categories of explanation correspond to, but encompass more than, the vectors used in the model introduced in Chapter 3. The theories-Epistemic, Institutional, Student Preference-Based, Financial Need-Based, Demographic and Ability-Based, and Market-Based-are not mutually exclusive but provide a useful way of classifying the arguments made in earlier studies to justify increasesin I-ID. Epistemic Explanations The underlying premise of these explanations is that an expanding knowledge base requires that students take more time to learn, absorb, and retain what is needed to earn the doctorate. A corollary is that more (and perhaps higher quality) work is expected of the doctoral student now than in the past. But measurement of an epistemic trend requires an objective measure of the expansion of knowledge in each field. While indirect indices of this expansion (such as counts of pages, books, journals, courses, and citations) are available, there is no consensus on how to define the body of thought a doctoral student must master. Similarly, it is difficult to agree on the length of time a student should be given to master the body of knowledge required for a doctorate, since students progress at different rates. To limit the time needed to earn the doctorate is to run the risk of excluding potentially productive scholars. More research is needed to pinpoint changes in the prerequisites for entry to the graduate program, in course load, and in the standards used to judge a dissertation within each field. Institutional Explanations Factors in the university and/or departmental environment such as goals and commitment, the interaction between faculty and students, and changes in student attitudes toward themselves and their peers-can also affect TTD. This study indirectly measures changes in the institutional environment over 97
time by looking at the quality of the doctoral department, the type of undergraduate and graduate institution attended, and the effects of changes in selected resources. These aggregate measures are not substitutes for the more specific sociological and institutional variables described by Wilson (1965~. The analysis indicates that changes in the percentage of a cohort at a top-ranked graduate department do not affect either TTD or RTD. Interestingly, however, increases in the percentage of a cohort whose baccalaureate was earned at a first-tier doctorate-granting university do reduce l-l ~ and RTD, albeit in a limited number of fields; but there is no evidence that a graduate department's high quality rating is associated with a low mean LID. The analysis also fails to establish a link between aggregate resource intensity, such as the aggregate number of faculty and R&D spending, and TTD. We cannot rule out the possibility that such evidence would have been found if the data series for these variables had been field-specific. Given the gross measures used and the limited number of observations available, our findings for these variables should be viewed as suggestive rather than conclusive. Clearly, additional work is needed to flesh out the impact on RTD of the institutional environment. At present, it is not clear whether RTD is increasing because students are taking more courses, because they spend more time working while registered at the university, because more prerequisite courses are required, or because it simply takes longer to complete the dissertation. Additional work also is needed to develop causal models of institutional factors. Such studies might merge institutional and departmental data with data on average student performance and progress within the department over several years. Student Preference-Based Explanations This explanation assumes students prefer to stretch out their graduate training because they like being "perennial students," graduate school offers a desirable environment, students prefer to allocate time in graduate school to nondoctorate-related activities, and/or they fear they won't be able to find a job after graduation. These preferences are not easily captured in a time-series model because no consensus exists on which student attitudes should be measured and on how to measure them and, at present, the Survey of Earned Doctorates, the only yearly study of doctoral students, does not collect information on graduate student preferences over time. Many factors can cause students to change their reasons for attending or for leaving graduate school. Decisions by university administrators may make the graduate school environment less comfortable or may place limits on financial aid. And societal mores may put pressure on those who remain outside the labor force too long. In addition, students also may change their perceptions 98
of the benefits of a college education. Clearly, these factors can alter both RTD and 1 1~. This study introduces student choice into a time-series model by looking at behavior at the margin. Of primary concern is whether changes in the marketplace cause students to alter their choices regarding graduate school. Financial Need-Based Explanations The financial pressures on students may increase as a result of illness or injury, tuition increases, marnage, family obligations, reduced loan or financial aid packages, and/or increases in the cost of living. Because of these factors, students may find it necessary to spend more time working and less time studying, thereby increasing TTD through effects on TPGE, RTD, and ADIEU. Marital status and increases in family size raise AD in a few fields but do not provide a general explanation of why TTD has increased in all 11 fields in this study. Changes in students' domestic situations contribute to the rise in TTD but are not the primary cause. An argument can also be made that TTD and RTD may have risen because fewer students are receiving federal financial aid. Wilson's study found that the percentage of those with financial aid was greater among those students who finished the doctorate quickly than among those who took more time to finish. It also reported that about one-third of the students who delayed entry to graduate school did so for financial reasons. This, among other things, led Wilson to recommend increases in financial aid as a way to hasten TTD. While Wilson's evidence is suggestive, it poses a problem of causality. Did students who are recipients of financial aid finish faster because they had such aid or because such aid was given to the most able? This question remains to be answered. Moreover, Wilson's study ignored the question of whether the form of financial aid made a difference for TTD and made no attempt to quantify the effects of financial aid on the several times to the doctorate. A comparison of the mean TTDs of those receiving federal fellowships, TAs, RAs, and private foundation support to the mean TTDs of those whose primary source of support was their own earnings (Table 3.1, p. 40) revealed that those who provided their own financial support took substantially longer to complete the doctorate than those with other types of support. Interestingly, mean 'folly either fell or stayed constant between 1986 and 1987 for TA holders in seven fields and for federal fellowship holders in eight fields; it rose in seven fields for RA holders and for those who provided their own support. The effect of financial aid on TTD is not as apparent in the causal models presented in Chapters 5 and 6. This is, in part, because the variables used in the model do not focus on the primary source of support. Moreover, the role of the financial aid variables may be obscured by their correlation with other 99
independent variables in Me model. The findings suggest that when it is a significant factor, it has a limited effect on ICED (relative to the total time required to complete the doctorate) and RTD. For example, using the linear common variables model, a 10-percent increase in the number of psychology students with TAs results in a decline of just four months in IlD. In fact, none of the financial aid variables had a consistent effect on IBID and, in some fields, they did not change IBID at all. In the TPGE equations, federal support was not statistically significant in any field; TA support had a negative effect in one field; and RA support had a positive effect in one field. In the INEU equations, federal support had a positive effect in one field; TA support was not statistically significant; and RA support was positive in one field. Recent DRF surveys have collected new data on prime source of financial aid. These data could be used to analyze more thoroughly the effect of financial aid on the four dependent variables. Demographic and Ability-Based Explanations In recent years, doctoral students are more likely to be older, female, foreign, and minority, all factors that can increase TTD and RTD. Recent interest in certain demographic factors probably is a response to trends in the DRF data. For example, in 1976 women constituted just 22 percent of the 18,583 science and engineering doctorate recipients. By 1985, women represented 27 percent of the 19,164 science and engineering doctorate recipients (Coyle, 1986~. Likewise, the share of non-U.S. citizens with permanent or temporary visas who received science and engineering doctorates grew from 21 percent in 1976 to 27 percent in 1985. Given the changing composition of the doctorate-recipient group, a natural question arises as to whether these changes were responsible for the increase in ITo. Gender, residency status, and race do not consistently affect the measures of time to the doctorate in the 11 fields studied. In fact, the only demographic variable that has a large enough effect across fields to affect 1-1 D is age at entry to graduate school. Age is important in the IlD, RTD, and TPGE equations but does not have a statistically significant effect in most fields in the TNEU equations. Unfortunately, the analysis does not distinguish whether age is a proxy measure or truly reflects the effects of aging on learning.22 22 We cannot dismiss the possibility that changes in student abilities were a major factor. The lack of student skills data, such as GRE scores, did not allow study of this possibility, however. 100
Market-Based Explanations Employment opportunities, the absolute salaries of doctorate holders, relative salaries, and the rate of return to alternative careers all affect time to the doctorate. Their impact is felt both by those in graduate school and by those considering alternative fields of graduate study. The assumption is that when the economic return for graduating with a doctorate falls relative to the return to nondoctorates, TTD will rise. Economic return diminishes in a given field if the unemployment rate of new doctorates rises relative to those without a Ph.D., if the relative salary of nondoctorates rises relative to that of new doctorates, and if the earnings of Ph.D.s fail to progress as rapidly over time as the earnings of those without doctorates. The longer a student remains in graduate school, the less economic return is expected. The results of this study suggest that changes in the marketplace were not large enough or pervasive enough to provide the primary explanation for the observed increases in TTD. Increases in the unemployment rate for those with four or more years of college education reduced RTD in four fields in one model while increased unemployment affected TTD in only one field. Changes in the percentage of students seeking employment and of those with definite postgraduate plans affected TTD and RTD, but only in a few fields. TTD fell in some fields as salaries for experienced doctorates rose, and it increased when there was a decline in the salary of new doctorates relative to salaries of doctorates 10 years postgraduation. Additional modeling is needed to confirm these findings and to identify the appropriate lags between market changes and changes in 'tow. Is There A Single Explanation for Increase in TTD? A series of factors, rather than one explanation, appears to be responsible for the trend of increasing TI D across fields. Part of the increase in 11D probably was due to epistemic factors, but this theory does not explain why there was three times the growth in l-ll) in the social sciences compared to chemistry (nine months versus 2.4 years). It seems unlikely that growth in the knowledge base alone could explain such a large increase in 11D in one field and a relatively small increase in another. Institutional factors also came into play. Likewise, declining enrollments in some institutions may have created an incentive for them to keep students longer. Although the institutional environment may not have been stable during the period of study, it is not clear that these factors explain the inter-field changes described. Among demographic variables, age is important because older students wait longer to enter graduate school and also spend more time registered in 101
graduate school than younger students. The finding that older students take longer to complete the doctorate warrants further study. In some fields, variables such as residency and gender also affected TTD, as did financial need. This study also suggests that market forces, particularly increases in the unemployment rate and in the salaries of doctorates and nondoctorates, affect TTD. The finding that no one class of explanations is responsible for the rise in TTD is consistent with the initial correlations in Chapter 4 and with the set of regressions presented in Chapters 5 and 6. It also confirms Wilson's 1965 findings of the multi-factorial aspect of any steps taken to reduce TTD: In essence, the amount of time involved in doctoral preparation can be reduced, our respondents indicate, only through concerted effort on a variety of fronts. Solutions predicated on a monistic conception of the problem will not prove to be satisfactory and no approach to "time reduction" stressing only one line of attack, e.g., increased financial support, . . . will be sufficient, however necessary it may be to an overall solution. As has been shown, TTD is affected by a number of variables. But aggregate models alone cannot identify steps to reduce ~l-ll,. What is needed is a more disaggregated study of what is happening at the department level. And additional modeling should be done using the student as the unit of analysis to sort out the roles of ability level, past preparation, and financial aid in elongating TTD. Existing studies do not provide sufficient guidance for policymakers to reduce TTD. Implications of a Continuing Rise in TTD A More Resource-lr~tensive Doctoral Program Changes in TTD that result from an increase in time spent in graduate school will raise the cost (excluding opportunity costs) of obtaining a doctoral degree. The annual cost, on average, to educate a graduate student ranges between $21,855 and $29,235. The mid-range estimate is $25,545 per year.23 23 The U.S. Department of Education, National Center for Education Statistics (NCES), Digest of Education Statistics 1988, indicates that educational and general expenditures per Al ~ university student were $13,179 in 1985-86 (Table 243~. We have assigned weights to account for the higher cost of graduate 102
The fields with the smallest rise in RTD between 1967 and 1986 (0.8 years) were engineering, chemistry, and physics and astronomy; the field with the largest increase (2.9 years) was the social sciences Using 1967 as the base year, the percentage increase in RTD was 14 percent in engineering and 49 percent in the social sciences. Assuming the cost of programs does not vary across fields, the cost of a doctorate rose by $20,436 ($25,545 x 0.8) in engineering and by $74,081 ($25,545 x 2.9) in the social sciences between 1967 and 1986. Taking all graduates into account, the increase in RTD caused an additional $35,190,792 ($20,436 x 1,722) to be spent educating engineering doctorate recipients and an additional $106,602,550 ($74,081 x 1,439) in outlays to educate social science doctorate recipients. Graduate students themselves pay only a small fraction of the $25,545 average yearly cost of graduate training, with the rest coming from other sources. A Longer Gestation Period Increases in IBID force employers to wait longer to hire new doctorates, potentially causing a shortage of trained workers in affected fields and driving up the salaries of those who already hold doctorates. Lengthening 'lurid can also contribute to a public perception of shortage and thereby increase pressures for public subsidies in fields in which trained doctorates appear to be in short supply. Increases in AD may also cause increased demand for foreign-trained doctorates. Lengthening rl~lD also means the productive output of doctorates will fall. For example, suppose the average age of graduate students in the social sciences at time of entry to graduate school was 27 years in 1967. If Rll) in 1967 was 6 years, the average doctorate holder would graduate at age 33. If that person had no periods of unemployment and- utilized his or her doctorate knowledge until retirement at age 65, a total of 32 person-years of work would have been produced. But if, in 1986, the average RTD rose to 9 years, the new doctorate's entry into the labor force would be delayed until age 36, reducing the average number of productive person-years to 29, a decline of 9.4 percent. If . . education: weight 1 for part-timers and weight 2 or 3 for full-timers. NCES estimates that 56 percent of doctoral students were full-timers in 1986-87. Thus, the range of expenditures is $20,559 to $27,939, with a midpoint of $24,249. To these institutional costs are added the students' costs of doctoral education, estimated at $2,874, derived from NCES' National Postsecondary Student Aid Study, which found a cost of $3,126 for full-timers and $2,554 for part-timers. 103
TPGE also increased during the period, the number of productive person-years of effort would decline even more.24 Clearly, increases in RTD and TPGE may reduce the productive worklife of a new doctorate and reduce the overall number of high-level personnel available to employers. More doctorates would have had to be produced in 1986 than in 1967 to obtain the same number of person-years of work as in 1967. In fact, however, there was no increase in the number of new doctorates; DRF data indicate the number of new doctorate recipients has remained relatively constant since 1970 (Coyle, 1986~. Although work yield of a given cohort of new doctorates is affected by a variety of factors, including mobility patterns obsolescence, and economic conditions, this simple example illustrates that changes in l ll) can affect labor supply. Longer TTDs also slow job market response to increases in demand. There is normally a lag when engineering and scientific labor markets adjust to changes in demand (Tuckman, 1988~. As the length of time required to produce a doctorate increases, so too does the length of time needed for supply to respond to increases in demand. And sudden increases in demand were more likely in 1986 than in 1967 to cause a longer period of market disequilibrium. The long- term effect of an increase in TTD is to reduce We responsiveness of high-level labor markets. Increased Attrition' - To the extent that increases in RTD are due to factors beyond student control such as increased financial pressures, frustration created by the length of time required to complete the doctorate, of "better" opportunities-some students may choose to abandon their graduate studies altogether. The literature review uncovered no studies that looked at how changes in RTD and TTD affected student attrition, but it seems likely that, at the margin, some students consider cost when deciding to forego an additional year of graduate school. To the extent that this phenomenon occurs, increases in TTD will reduce the number of people who complete the doctorate. Such attrition will also increase the costs to society of producing a Rained doctorate. 24 This analysis assumes no change in retirement behavior. The effect of lengthening AD on productivity will not be as dramatic if retirement age is nslng. 104
Lower Returns for Graduate Study Longer TTD increases the costs of doctoral education. Even students with fellowships incur an opportunity cost because this type of support is less than the earnings that they would have received in a full-time job. Also, as noted, increases in RID reduce the number of productive years during which a student can realize a return on his or her investment. To the extent that students view graduate study as a potential investment, reductions in return from doctoral education are also likely to affect the decision whether to obtain a degree at all. Some students may find changes in l-lD have made alternatives to a doctoral degree more attractive. For example, in many graduate schools, the Master's of Business Administration degree takes only two years to complete; thus, if the l-lD required to obtain a doctorate in the sciences increases, some students will opt instead to obtain an MBA. A similar phenomenon may occur as students consider an advanced degree in medicine, law, or other professional fields. To the extent IBID rises less slowly in these fields, the relative return for obtaining a degree in them increases. Over time, more students may be drawn away from fields with high l1Ds and into fields with shorter TTDs, leading to a possible shortage of trained scientists and engineers in certain high--D fields. Changes in the Attractiveness of Alternative Doctorate Careers Students choose a major based on expected returns (Chapter That is, the earnings they can expect to receive after earning the degree. Differences exist in the rate at which TTD and RTD are growing among fields. Thus, a person with an undergraduate degree and an interest in one field-physics, for example- may nonetheless choose advanced study in another field perhaps mathematics because the expected returns to a doctorate in the latter field are higher. To the extent that this occurs, a shortage may eventually develop in those fields with relatively larger l~lDs. TTD as a Policy Instrument This study was motivated by interest in manipulating TTD to meet possible difficulties in producing a future supply of doctorates that will be adequate to meet anticipated needs. It is easy to argue that the increase in l-l'L, can be reversed by increasing the number of federal fellowships or by granting more teaching and research assistantships, but the findings of this report suggest we need to learn more about the effects of the venous types of financial aid 105
before assessments of the desirability of such a solution can be made. Data are simply not available to permit policymakers to choose the best way to affect lolls or to assess the consequences of the various alternative solutions proposed by other studies. 106
