Archived InformationAnswers in the Tool Box: Academic Intensity, Attendance Patterns, and Bachelor's Degree Attainment June 1999
"Academic Resources" moves forward in our story onto a field with other variables that historically have served as touchstones in analyses of educational careers. Some of these touchstones involve the characteristics and environments of the institutions attended, and these are addressed in Part III. In this section, though, we are concerned with sharpening the focus on our dependent variable, degree completion, and with the sources and shape of other major background variables that researchers, policy-makers, and interpreters of trends in education tend to accept as if they were holy writ. A more secular framework of judgment is called for.
The principle points to be made are:
Most previous research employing pre-collegiate input variables has focused on the determinants or correlates of access or enrollment (for a summary of this literature, see Baker and Velez, 1996).(22) There are two sets of literature that take us beyond this access threshold. The first uses "persistence" as the dependent variable, but seems less concerned with persistence than with attrition (DesJardins, Ahlburg, and McCall, 1997; McConnell-Castle, 1993). Tinto (1975, 1982, 1993) set the terms in this literature, with attention to social and academic integration factors in students' experience; and the research deriving from Tinto's work is vast. Bean (1980, 1983) and Metzger and Bean (1987) offered a different kind of analysis of attrition based on organizational process theory and the phenomenon of personnel turnover within organizations, and, as a consequence, have been the only researchers to emphasize features of institutional location (distance from the student's home, and in-state attendance) as factors in student careers(23).
Both lines of research treat persistence or attrition as a phenomenon existing within a given institution (Cabrera, Nora, and Castaneda, 1993), sometimes illustrating the problem of single-institution studies in which the student population lies considerably above academic performance means (e.g. Bean, 1980; Bean, 1982). If a community college population is the subject, the studies also illustrate the problem of defining persistence and attrition in terms of intent to transfer (see, for example, Nora and Rendon, 1990), and the problem of influences of family and significant others on a predominantly commuting population (Nora, Attinasi, and Matonak, 1990)(24). Both approaches, as Cabrera, Castaneda, Nora and Hengstler (1992) point out, regard persistence and attrition as the result of interactions among precollege characteristics, college environments, and adjustments to college. Both use college grades and continuation beyond the first year of study as outcome variables (see, for example, Pascarella and Terenzini, 1980; St. John, Kirshstein, and Noell, 1991). Both research traditions place an extraordinary emphasis on psychological variables: intentions, attitudes, influences, commitments, perceptions (see, e.g. Cabrera, Castaneda, Nora and Hengstler ). These variables unfortunately refer to realities that lie beyond the control of those who can best steer students toward degree completion.
"Persistence" itself, as commonly used in the research literature, may not be a convincing dependent variable. We have no idea, for example, what a one-year "persistence rate" means. Does the student who arrives at something called "year 2" get there with 11 credits, 19 credits, 32 credits? In the HS&B/So, of all students who entered college directly from high school in 1982 and persisted through 1983, 17 percent arrived at "year 2" of their college careers in the summer or fall of 1983 with less than 20 credits, and another 12 percent earned more than 20 credits but with three or more remedial courses. In other words, nearly three out of ten "persisters" evidenced less than what one would call "sophomore standing." Table 13 displays some of the variants in this portrait. It is no surprise that students with low credit accumulation in the first year who also show "temporal persistence" are far less likely to earn degrees by age 30. Without credit accumulation information, structural equations with "persistence" as an outcome are very deceiving, and are apt to overstate the influence of affective factors as opposed to academic achievement.
Unlike "persistence," the completion of a bachelor's degree is a censoring event, the culmination of years of preparation and effort. Momentum toward that event is not necessarily measured in years, however, rather in accrual of the currency of the degree (McCormick, 1999). And the more currency acquired, the higher the odds of completing. To launch the measure of completion rates from credit accumulation thresholds provides a fairly strong guidance. Table 14 expands the framework suggested by table 13, and offers some hints of what might happen if one set post-matriculation credit thresholds in regression equations (ordinary least squares or logistic) in which bachelor's degree completion is the dependent variable. Table 14 also marks students who entered higher education within six months of on time graduation from high school in 1982. The reader will note that these "direct" students completed bachelor's degrees at a rate higher than that for all students.
|Credits Earned:||0-12||13-19||20-28||>28||% of
|By Number of
|Three or More||14.1*||19.0||36.6||30.3||17.7|
|Bachelor's or Higher||1.9*||3.9*||29.4||64.8||59.8|
Notes: (1) Universe consists of all on-time high school graduates who entered postsecondary education between June and December, 1982 and who were also enrolled during the Academic Year beginning July 1, 1983. (2) Weighted N=1.2M; (3)*Low-N cells with no statistical significance. (4) Rows for credits earned may not add to 100.0% due to rounding and Low-N cells. SOURCE : National Center for Education Statistics: High School & Beyond/Sophomore cohort, CD #98-135.
This study takes the position that bachelor's degree attainment rates should be measured only for those people who actually attended a 4-year college at some time. Students can tell us on surveys that they intend to earn a bachelor's degree, but if they never set foot in a 4-year college long enough to generate a record, they made no attempt to do so, and it is misleading to include them in the denominator of potential degree recipients. Long-term bachelor's degree attainment rates jump dramatically as soon as one confines the universe to those who have attended 4-year colleges, and exhibit impressive gains with each ratcheting up of the credit threshold. The effect of no delay of entry to higher education also plays a positive role, but the strength of that role diminishes with credit accumulation. If we wished to push the conditions for degree attainment to near-maximum levels, we would also limit the institution of first attendance to a 4-year college, insist that more than 30 credits be earned from 4-year colleges, and add the direct entry criterion. The long-term (eleven year) degree completion rate under those conditions is 79.1 percent, and is higher, still, if the first institution of attendance was highly selective and if the student attended only one college.
|All Postsecondary Students||Attended a 4-Year College At Any Time|
Notes: (1) DIR=Direct Entry. (2) Universe=all students for whom transcripts were received. (3) Weighted N is in millions. (4) Columns may not add to 100.0% due to rounding. (5) Paired comparisons (All v. DIR) are significant at p<.05 except those asterisked. SOURCE: National Center for Education Statistics: High School &Beyond/Sophomore cohort, CD# 98-135.
In a similar analysis, McCormick (1999) formulated a scaled-down version of the same database used in this study, excluding about a third of the students in the HS&B/so who ever attended a 4-year college as well as certain types of credits(25). McCormick was more interested in the pace at which students accumulate credits, but his conclusions about degree completion rates mirror those indicated above, and, with some minor variations, those displayed in table 14. On balance, McCormick's degree-completion rates are higher because of the exclusions, but the relationships between credit-generation thresholds and completion rates are the same, a reassuring consequence of using the same transcript-based source.
As noted above, there are very few national studies across the entire literature on persistence and attrition that hold the completion of a degree to be the sole and/or most prominent dependent variable. Researchers often tout the importance of completion studies and then tell us why they cannot conduct such studies with a full sample of students. For example, Hauser and Anderson (1991) beg off by noting that "changes in the timing and intensity of college attendance have made it more difficult to measure and compare rates of college completion in recent years" (p. 275), though, in fact, the more complex attendance patterns have become, the richer the analysis (see Part III below). Chaney and Farris (1991) conducted an analysis of completion, but confined it by an institutional-entry cohort restricted to full-time students, capped the measurement period at 6 years, and reminded us that once students leave the institution they are not tracked, therefore "institution-specific retention rates may significantly understate retention within higher education as a whole" (p. 5). Astin, Tsui and Avalos (1996) offer a similar disclaimer, though they should be credited for seeking evidence of degree completion from institutional registrars as opposed to relying on student self-reports. Their 9-year follow-up to the entering freshman class of 1985 is a commendable advance on the time frames of completion studies grounded in data bases other than NCES longitudinal studies, but resulted in a low response-rate from the registrars(26). What we inherit from all these talented researchers, then, are partial portraits of completion.
Carroll established a model for national analysis using the history of the High School & Beyond/Senior Cohort (high school class of 1980, followed to 1986) to define a "persistence track" that ends with bachelor's degree attainment. Carroll's principal variables are those of immediate entry to higher education following high school graduation, full-time enrollment, stop-out behavior, and transfer. His strategy was to set an "optimal" model, and then to demonstrate how year-to-year deviations from the model resulted in either no degree or a time-to-degree beyond the classical 4-year norm. Carroll's analysis was the first to take account of attendance patterns in a sophisticated way. It is a descriptive portrait of dynamic flows without pretending to predict or assert deterministic relationships. While Carroll cites pre-college variables, he uses them minimally: his focus is on the "persistence track."
Carroll used survey data (not transcripts) and hews to a discrete time hazard model that has been shown to be a productive approach to the understanding of attrition (DesJardins, Ahlburg, and McCall, 1997). The data source, however, creates a particular problem in student self-reports of part-time status. It is difficult to determine the fraction of Carroll's universe that either started part-time or shifted to part-time status, but, in the HS&B/Sophomore cohort, the proportion of students who told us they were enrolled part-time at any time in their college careers was very low: 15.2 percent(27). This fraction is so far below what institutions report on the federal Integrated Postsecondary Education Data System (IPEDS) surveys as to cast students' judgments in severe doubt. In 1984, the modal year of postsecondary attendance for the HS&B/So, institutions reported that 40.2 percent of their undergraduate enrollments were part-time (Snyder, 1987, p. 129). A decade later, we witness the same type of discrepancy: in the fall semester of 1995, 41.6 percent of undergraduates were reported as part-time students (Digest of Education Statistics, 1997, table 175, p. 185); in the Beginning Postsecondary Students Longitudinal Study of 1989-1994, only 26.5 percent of students ever reported being enrolled part-time as undergraduates(28). Sometimes, one is unsure whether students know what "part-time" means.
Carroll's two tracks have strong boundaries. The moment one violates the boundaries, one is declared off-track and no longer persisting. These "violations" include reverse transfer (4-year to 2-year), a shift to part-time status, and stop-out (defined as a gap in attendance of 4 non-summer months). In combination with delayed entry and enrollment in a 2-year school or in a nondegree-granting postsecondary institution, these behaviors are generally acknowledged as a drag on attainment (Pascarella and Terenzini, 1991; Hearn, 1992).
Berkner, Cuccaro-Alamin and McCormick (1996) borrow and extend Carroll's scheme. Their data were drawn from the Beginning Postsecondary Students Study of 1989-1994. These data are not exactly comparable to those from age-cohort studies such as the HS&B/So: nearly 27 percent of the participants were 20 years or older in their first year (1989-90) and nearly 10 percent were at least 30 years old. Dates of high school graduation are spread over two decades, and very little is known about the pre-college backgrounds of participants. The authors configure a set of characteristics of students at the moment of first entry to higher education (the "event" that defines the cohort) that are dominated by household, employment, and dependency variables (number of children, single parent status, full-time employee,(29) and independent status as defined in federal financial aid regulations). These characteristics, which are highly correlated with age at entry, along with the quality of entry in terms of delayed enrollment and part-time status, they say, form a collection of "risk factors" that work against persistence and completion of any credential, certificates and associate's degrees included.
Berkner, Cuccaro-Alamin, and McCormick's study is valuable for its methods, and, advancing on Carroll's rules for defining a "persistence track," the attention it pays to attendance patterns, particularly in the combination of transfer and continuous/noncontinuous enrollment. The database allows the analyst to construct at least a rough approximation of the kind of academic and social integration indices used by Tinto and Astin throughout their work, something we cannot construct in the HS&B/So. The dependent variables included in Berkner, Cuccaro-Alamin, and McCormick's model of system persistence/attainment include credential award (certificate, associate's degree, or bachelor's degree), and final enrollment status (still enrolled/no longer enrolled). "Internal persisters" are distinguished from "transfer persisters." Based on the institution of first attendance, the basic "persistence history" of this cohort can be summarized as follows:
These data provide hints that are supported elsewhere in the literature (see, for example, different kinds of evidence from Lee, Mackie-Lewis, and Marks, 1993; Blumberg, Lavin, Lerer, and Kovath, 1997; McCormick, 1997; Adelman, 1998) that the very act of transfer embodies an intensity of commitment to higher education that results in degree completion rates equivalent to--if not higher than--those of students who do not transfer. It is no wonder that, as we will see, multi-institutional attendance (with or without formal transfer) is not a drag on degree completion. Institutions may "retain," but students "persist."
One of the most persistent variables in the persistence literature measures educational aspirations. In the NCES longitudinal studies, we ask students about their vision of their future education at each survey occasion, and invest the responses with a great deal of explanatory power in terms of subsequent behavior (e.g. St. John and Noell, 1989; Pelavin and Kane, 1990; Hauser and Anderson, 1991). But we must be careful about such evidence, pay close attention to the wording of the questions, and select those occasions on which the evidence is most persuasive. If, as in the NLS-72 survey, one asks a senior in high school whether he/she "would like" to attend college and enters the responses in a structural equation with college entrance as the dependent variable, it is like looking at a sky filled with dark clouds and predicting rain (Alexander and Cook, 1979).(30) After two years of college, asking whether someone "plans" to complete a bachelor's degree is a similar exercise. On either occasion, there is enough momentum in students' educational histories so that they know the likely outcomes on the path they are taking. In other words, on these occasions we are no longer dealing with "aspirations," rather with "expectations." The resulting high correlation between expectations and access (Akerhielm, Berger, Hooker, and Wise, 1998) is highly artificial. Family expectations are part of this momentum, usually a reflection of SES. Asking whether a student's expectations agree with those of parents is more likely to produce concurrence among higher SES students than lower SES students. In the 1982 (senior year) HS&B/So survey, 30 percent of students in the lowest SES quintile did not report or did not know their parents' expectations for their postsecondary education--versus 12 percent for students from the highest SES quintile. It is for this reason, in particular, that some analysts place the "aspirations" variable far down the line in regression models (see, e.g. St. John, 1991, who noted that "it is appropriate to consider its influence only after . . . other factors have been considered," p. 145).
Students are asked not only about aspirations (in Hauser and Anderson's  phrasing, "desired outcomes that are not limited by constraints on resources," p. 270), but also about plans, a more realistic assessment of future action. Students who say they aspire to or expect to earn a bachelor's degree, but also indicate that they plan to spend most of the year following high school graduation in an apprenticeship program illustrate one form of the difference between aspirations and plans: the aspiration is a generalized guide, but the plan to execute it is elusive. The extent of such out-of-scope choices in the HS&B/So is notable: 11.2 percent of those who expected to earn graduate degrees and 15.2 percent of those who expected to earn bachelor's degrees planned to engage in activities during the year following high school graduation that would lead them nowhere near the paths to those goals.
How, then do we best measure the consistency and strength of pre-college educational expectations in the HS&B/So? The surveys asked six questions in both the 10th and 12th grades for this cohort that help up arrive at a more sophisticated notion. When we look closely at the wording of these questions, we realize that they do not address aspirations at all. They reference neither constraints not goals. Some express expectations. Others ask for concrete plans. Still others ask for projected affective states. Together, they map a complex set of what we might call "anticipations." These questions, spread out over the survey forms, are:
We have 12 pairs of responses, then, with which to build an "anticipations" variable based on the concepts of consistency and level. The five resulting gradations of the variable are:
How strong is this variable? Table 15 indicates what happened to the populations at each level of "anticipation" in terms of degree attainment. Among all postsecondary students there is a very clear and dramatic linear relationship between bachelor's degree attainment and the levels of anticipation. As soon as one drops below the "bachelor's-consistent" level, the bottom falls out on long-term degree completion. But among those who attended a 4-year college at any time, the linear relationship is not as dramatic, and only the small group of those who never aspired to the bachelor's degree (despite their attendance at a 4-year school) evidences very low degree completion rates.
|All Who Ever
Attended a 4-Year College
|Increased to Bachelor's||55.3||11.7||33.0||28.6||36.2||7.4||56.4||28.9|
or Reduced from Bachelor's
|Certificate or Associate's:
|No Degree or Never Knew||89.0*||7.6*||3.4*||11.5||69.0*||10.1*||20.9*||3.3|
NOTES: (1) The universes consist of all students for whom "anticipations" could be determined from survey responses in both the 10th and 12th grades. For all postsecondary students, the Weighted N=2.38M; for those who attended a 4-year college at any time, the Weighted N=1.37M. (2) Column pairs are significant at p<.05 except those marked by asterisks. SOURCE : National Center for Education Statistics: High School & Beyond/Sophomore Cohort, NCES CD#98-135.
The "anticipations" variable has five values, but is not like the quintile-formatted variables used in this analysis: it is not based on a continuous scale such as socioeconomic status (SES) or senior year test percentiles, nor it is based on intervals that yield roughly equivalent quintiles, such as the basic "academic resources" variable (ACRES). The variable calls for a dichotomous reconstruction. Among students who attended a 4-year college at any time, the positive side of a dichotomous variable would be confined to those who exhibited "bachelor's consistent" expectations. The true value of anticipation in student histories lies in its correlation with attending a 4-year college (whether initially or through transfer). But in Part IV of this study, confined to 4-year college students, its position is more complex.
It is worth noting that Morgan (1996), examined changes of the HS&B/So students' responses to the "aspirations questions" (but not the minimum satisfactory, dominant activity, or future plans questions) between grades 10 and 12 and treated them as a dependent continuous variable. Morgan found a larger proportion of students (45 percent) who did not change their goals between grades 10 and 12 than is reflected in our account of "anticipations." This is an intriguing approach, but the story it tells ends at the variable(31), not with actual attainment.
Much of the previous research on postsecondary access and attainment employs stepwise regressions in which key components of socioeconomic status are treated as separate entities. Most notable among these components are parents' occupations and level of education, and family income.(32) In nearly all national data sets, these data are reported by students.(33) One can smell the problems with this reporting from a great distance (Fetters, Stowe, and Owings, 1984). Other perceptive analysts (e.g. Mare, 1980) have cited the low reliability of such variables in light of children's changing understanding of their parents' educational history and occupational status, and still others (e.g. Morgan, 1996) have gone to the creative trouble of substituting parents' responses (when available) and restandardizing the SES scale for analyses of NCES longitudinal studies. Indeed, a comparison of student and parent accounts in the NELS-88 dataset (see table 16) canonizes these observations and adjustments.
Late adolescents may know what their parents do for a living, but their idea of the highest level of education attained by their parents leaves something to be desired (Mare, 1980). As high school students, some 29 percent of the HS&B/So participants found various ways to indicate that they did not know their father's highest level of education; though for mothers, the rate was a mere 19 percent. Furthermore, in 8 percent of the cases for those who claimed to know their parents' highest level of education, there was a raw dissonance between reported parental occupation and reported parental education. For example, according to the students, we have lawyers whose highest degree was "some college," and school teachers whose highest level of education was "high school graduate." One can edit some of these cases in the database, but only where the parental occupation indicated requires at least a college education.(34) Other researchers have found that parents' educational levels have little direct effect on success, particularly when compared to pre-college acquisition of knowledge and skills (Mow and Nettles, 1990; Grandy, 1998), so there is some indirect support for the position taken here.
Since the High School & Beyond files include interviews with a sub-sample of parents, we can compare perception to reality. Fetters, Stowe and Owings (1984) confirm the child's tendency to underestimate their parents' levels of education, even when they can classify their parents' occupations accurately. For example, when they were in grade 10, 35.3 percent of the HS&B/So cohort reported their parents had continued their education after high school, whereas 44.4 percent of the parents reported having done so. Among high school seniors, 37.4 percent reported postsecondary education for their parents, compared with 43.2 percent by the parents' account (Fetters, Stowe, and Owings, table A.2, p. 41). These are uncomfortable discrepancies made more uncomfortable by public policies that encourage "first generation college students" to continue their education, let alone by inter-generational social mobility analyses that rely on imperfect second-party accounts of educational attainment (Hearn, 1984; Karen, 1991; Lang, 1992).
|Highest Level of Education Attained by Either Parent|
With At Least
Notes: (1) Universe of students consist of all who answered questions about their parents' highest levels of education in 1992 (and, if not then, in earlier surveys), including those who indicated they did not know but excluding missing cases. Weighted N=3.03M. (2) F3PAQWT, a weight for the NELS88 "parents' file," was used. (3) *Highest degree of agreement was with mother; otherwise, highest degree was with father. (4) Row for parents' account adds to 100.0% SOURCE: National Center for Education Statistics: Data Analysis System, NELS-88.
If the HS&B/So data are not enough to scare one away from using student reports of their parents' highest level of education, perhaps the NELS-88 longitudinal study, which contains a substantial parents' file, will finish the job. Leaving aside the 15.7 percent of the NELS-88 students who would not venture a guess about their parents' schooling, table 16 presents some major indications of conflict in student and parent accounts. In these data, students emerge as more likely to understand that their parents have attended college than earned degrees, and seem to have a more definite sense of college attendance and credentials for fathers than for mothers. Of course, it is possible in all these data that the parents are inflating their educational attainment, but the conflicts between reported occupation and educational level that we observed in the HS&B/So suggest otherwise.
When we follow a traditional age cohort such as that of the HS&B/So, what Berkner, Cuccaro-Alamin, and McCormick's 1996 study (along with an earlier exploration of attrition among "non-traditional" students in Bean and Metzger, 1985) reminds us to do is to account for change in family status as students move into their early and mid-20s. The variable of choice is parenthood, and the reader will note the importance of this variable in Part IV below. In a correlation matrix with college access as the dependent variable, having a child prior to age 20 ranked third (behind SES and family income)--and well ahead of race, among demographic factors associated with entering postsecondary education, in this case, a negative association. Table 17 excerpts the critical data from the correlation matrix. The NELS-88 data provide continuing confirmation of these relationships. For example, of students in the lowest SES quintile in the NELS-88, 31 percent had children by age 20; and of those who had not entered postsecondary education by age 20, a third were already parents.
By Age 30
by Age 20
|Child by Age 20||-.1998||----|
Notes: (1) Universe consists of all HS&B/So students who participated in both the Base Year (1980) and second (1984) follow ups, and whose files contain positive values for all variables in the matrix; (2) Weighted N=2.22M; (3) Design effect= 1.56; (4) *estimate significant at p<.01, otherwise estimates are significant at p<.001. SOURCE: National Center for Education Statistics: High School & Beyond/Sophomore cohort, NCES CD#98-135.
Parenthood, a status that applies to men as well as women, is a frequently overlooked variable in educational histories, or, at best, is checked off as an aside even when it shows a strong relationship to access (as in Akerhielm, Berger, Hooker, and Wise, 1998). The earlier one has children of one's own, the higher the risk of not completing one's education (Horn, 1996; Horn, 1997). This sounds like an easy issue to isolate, but one learns quickly to be careful with student responses on questionnaires. For example, we have people in the HS&B/So who told us in 1984 that they had children, but when asked in 1986 whether they "ever had children," answered in the negative. We have 19 year-olds who graduated from high school in 1982 and told us that their children were born in years such as 1949, 1955, and 1968. The dichotomous variable developed and used in this analysis, "Children," treats all contradictory and (to put it gently) out-of-scope responses as "no children."
Previous research on the determinants of degree completion has been wanting on pre-collegiate measures and uncritically accepting of stock variables such as aspirations that, on closer examination, require reconstruction. Disaggregating SES into its component pieces may be inviting, but, as the case of parental level of education illustrates, is hazardous: the whole is stronger and more consistent than its parts. At the same time, we ignore the most basic of events in life-course histories, for example, having children in late adolescence or early adulthood, at our peril.
With these issues behind us, we can now bring the college transcripts onto the table of evidence, for they will enable us to grasp the activities of individuals moving through a series of learning environments in late adolescence and early adulthood. As the next portion of our exposition should make amply clear, this mobility has taken on dimensions that render traditional inquiries on the paths to degree completion--how should one say?--quaint.