The literature

The persistence literature can be classified into two parts: that which focuses on overall rates of graduating, continuing, switching, and leaving (or some subset of those outcomes), and that which analyses these patterns by various characteristics of the student, their situation, and other relevant factors. We discuss each of these literatures in turn, with an emphasis on the Canadian record. But we begin the section by placing this literature in the context of the data required to study PSE pathways, since it is the general dearth of such data that has generated a literature that is so relatively scant and otherwise limited, and which points to the unique opportunities opened up by the creation of the PSIS dataset.

2.1   The data required to study persistence in PSE

Study after study has pointed to the benefits accruing to PSE graduates and the key role played by PSE graduates in a nation’s economic performance¹. Yet it has been increasingly recognised that access to PSE – typically defined as entering the PSE system at some point – is but the first step on the path to graduation, and that persistence – typically defined as whether the student continues in their studies after entering PSE – is the required follow-through that is necessary for the full benefits accruing from PSE to be realised, both for the individual student and at the broader level (e.g., the nation’s economic performance).²

While many studies have been carried out on access to PSE, persistence in PSE has been studied to a much lesser degree both in Canada and in other countries.³ There are two main reasons for this. The first is that concerns that persistence might be a problem are relatively new, so there has been little in the way of a driving force to undertake studies of persistence. The second is that persistence is essentially a dynamic process, and studying it is therefore much more demanding in terms of the associated data requirements, especially because the data meeting those requirements have been in short supply.⁴

These data requirements include, at a minimum, the longitudinal tracking of substantial numbers of students through their persistence dynamics with sufficient detail on their PSE trajectories to identify – at any point in time – who graduates, who does not graduate but continues on in their programs, who switches programs, who leaves, and other related pathways.

It is also valuable to have as much further information as possible on students and their schooling experiences in order to deepen any analysis. Such characteristics would include basic demographic characteristics (sex, age, etc.), family and schooling backgrounds (e.g., parental income and education levels and how the student did in high school), attitudes regarding school that might affect persistence (do they think school in general, and PSE in particular, is important?), PSE program characteristics (level of study, major, current year of study), school performance while in PSE (grades, academic and social engagement), and other attributes to which it would be interesting to link persistence rates.

The longitudinal tracking requirement is critical because it is inherently difficult to analyse the persistence process in the absence of such data. Cross-sectional data can give only a snapshot of certain limited aspects of the processes in question, and tend to lack the underlying sample frames and information required to build the desired sorts of samples of students and properly analyse these dynamic processes. For example, while Statistics Canada’s Postsecondary Education Participation Survey (PEPS), which focuses on 18 to 24 year-olds, is well suited to the study of access to PSE, its static nature severely limits its potential for the study of persistence.

Yet such longitudinal data are in short supply. General longitudinal databases such as the Survey of Labour and Income Dynamics (SLID) here in Canada, tend to lack the required sample sizes of students and detailed PSE information required to study persistence in PSE in anything but a very limited way. Historically, there has been a complete lack of longitudinal databases focussed on youth, or students more specifically, such as would allow for such studies.

Neither have PSE student administrative databases been, in general, sufficiently developed, nor where they have existed have they been broad enough in their coverage to capture students as they move across institutions, including when they return to their studies after any break. Hence the general lack of analyses of persistence in PSE and PSE pathways to date in Canada as elsewhere.

Data from the longitudinal Youth in Transition Surveys (YITS-A and YITS-B) have begun to be made available to researchers recently and are the major exceptions to these rules. These YITS surveys (both YITS-A and YITS-B) are, in fact, extremely well-suited to the study of persistence patterns due to their focus on the youth population, their detailed tracking of PSE profiles, and the detailed information they possess on individuals’ schooling experiences, abilities, attitudes, and backgrounds. As a result, research on persistence based on the YITS is now starting to emerge.⁵ 

But the YITS, too, have limitations. In particular, they lack the sample size required for more detailed analysis of certain pathways except at the national level. Neither do they possess the wealth of institution-based information available in the PSIS. Finally, being based on surveys rather than administrative data (like the PSIS), the YITS data are subject to selection and other kinds of response bias. The two (YITS and PSIS) are, in the end, complementary for studying persistence in PSE. References to various YITS results will, therefore, be provided for the sake of comparison below.

2.2   Overall persistence rates

Principally due to the historical lack of relevant data, the relatively few existing studies on persistence that have been carried out in Canada have focussed on the experiences of students at individual postsecondary institutions. Not only are the findings in these cases restricted to those specific institutions, which means they are not necessarily generalisable to the general population, they also fail to capture students who continue in their studies at other institutions – a major shortcoming for the reasons mentioned above.⁶

In this context, the existing persistence literature in Canada includes Gilbert (1991), who estimates the measured five-year dropout rate for Canadian PSE students who entered the University of Guelph in 1985 to be 42 percent, a figure which is close to the six-year dropout rate of 46 percent for the 1994 cohort reported by the Consortium of Student Retention Data Exchange (CSRDE) for (principally) U.S. colleges and universities (CSRDE, 2001b).⁷

Wong (1994) finds an average first-year dropout rate of 24 percent for 13 Canadian universities, which is moderately higher than the 20 percent first-year dropout rate reported by CSRDE (2001a) for its 1999 cohort. Combining the findings from Gilbert and Wong, it appears that students in Canada are most likely to leave PSE between the first and the second year, after which the probability decreases substantially.

In a broader study of all students entering Ontario universities to pursue bachelor’s or first professional degrees (note the mix of programs) from 1980 to 1984, Chen and Oderkirk (1997) find that 68 percent had graduated from their initial programs by 1993. (This represents different numbers of years after starting for the different cohorts included in the sample, but most students could reasonably be expected to have finished their studies by this time.) Another 30 percent had not completed their programs and were no longer enrolled in an Ontario university. The remaining 2 percent had not completed their programs but were still enrolled in an Ontario university.

While generating some interesting numbers, the study is restricted to an analysis of the record for students enrolled at Ontario universities, while colleges are also omitted, so it misses those students who move from university to college, as well as those who move outside of Ontario, thus biasing persistence rates downward when thought of in a more general perspective. Also, these data are now quite old, and much has changed in the postsecondary system, including in Ontario, since 1980 to 1993.

2.3   Who leaves and why: factors that influence persistence decisions

There are two well known and broadly used theoretical models in the persistence literature. The first is Tinto’s (1975, 1993) model of “student integration”, according to which students enter PSE with various pre-entry characteristics, such as age, race, gender, family structure, parental education attainment, high school preparation, and their skills and abilities. These factors contribute to the formation of their initial goals and level of commitment to their studies. Once enrolled, students then begin to have their specific institution-related PSE experiences, which include their level of academic and social engagement and their academic performance. Students’ initial goals and commitments are then influenced and modified by these post-entry experiences. These various factors are then taken to determine persistence.

The second well known model is Bean and Metzer’s (1985) “student attrition model”. The main difference between these two models is that the student attrition model introduces factors external to the institutions, such as finances and peer affects. The student integration model also regards academic performance as an indicator (or determinant) of academic integration, whereas the student attrition model regards PSE experiences as outcomes (Cabrera, Castaneda, Nora, and Hengstler, 1992) on the grounds that, for example, lower grades can be a symptom of an individual’s detachment from school as they begin the process which leads to their leaving.

In summary, these two models both posit that persistence decisions are affected by both pre-entry characteristics and post-entry experiences, but differ in what they include in the latter, and their interpretation of some of the related effects.

In the empirical literature, however, there is no consensus on who drops out and why. In their review of the literature, Grayson and Grayson (2003) say that “…it is difficult to tell if different results of various studies reflect real differences in explanations for attrition or are simply artefacts of different methodologies…it [therefore] makes more sense to examine findings of individual studies in their own right rather than attempting to fabricate generalizations about attrition.” This statement obviously points to the need for more empirical work, especially if it employs a dataset that is well suited to the relevant estimation issues, is broadly representative, and uses an appropriate methodology.

As for that more detailed literature, in the richer U.S. record, Horn (1998) uses the Beginning Postsecondary Student Longitudinal Study data to find that the educational attainment of a student’s parents is related to persistence, with students whose parents received no education beyond high school being about twice as likely to drop out at the end of the first year as those with parents with a college degree, and this gap is maintained in the following years. The U.S. literature also suggests that students who drop out of their PSE appear to have been less academically prepared for their studies to begin with than those who stayed. For example, using survival analysis techniques similar to those employed here on a sample of 8,867 undergraduate students at Oregon State University between 1991 and 1996, Murtaugh, Burns, and Schuster (1999) find that PSE dropout rates decrease as high school GPAs of those entering PSE increase.

Postsecondary experiences found to be important in the American literature include students’ GPA, academic and social engagement, and other related measures. For example, using administrative records from Virginia Commonwealth University, Wetzel, O’Toole, and Peterson (1999) find that academic and social integration were the most significant determinants of persistence for the freshmen and sophomore students enrolled at that particular university (which is urban and public) over the years 1989-1992.

This said, and as alluded to above, although the relationship between such PSE indicators and PSE persistence is strong, it is difficult to identify the extent to which these relationships are causal. Perhaps being less engaged and obtaining lower grades is simply a normal step on the path for a student leaving PSE rather than an exogenous determinant of that outcome.

A national level Canadian study based on data from the Postsecondary Education Participation Survey (PEPS) shows that among students who left PSE prior to completion, half of them cited “lack of interest in their programs or PSE in general” as the reason for dropping out, whereas 29 percent cited “financial considerations” (Barr-Telford, Cartwright, Prasil and Shimmons, 2003). This implies that motivation plays a more important role than financial factors with respect to PSE persistence, though it is only a descriptive study, and does not probe into the determinants of these different reasons for leaving, such as the two models that have driven the American empirical literature.

Gilbert and Auger (1988) analyse first-year persistence rates for students who entered the University of Guelph in the fall of 1986 to find that financial factors played a more important role among students with lower socio-economic status (SES) than others. They also find that students from relatively higher SES backgrounds are more likely to switch to other institutions, while low SES students are more likely to stop-out (i.e., leave their studies and then return).

Grayson and Grayson (2003), in their review of the literature, conclude that the few studies that consider financial constraints as a reason for leaving a PSE program show only a weak effect.

Finally, in their recent work based on the YITS-B, Finnie and Qiu (2008) use multivariate modelling methods to find that college students with less highly educated parents and those coming from single parent families are more likely to drop out, but – perhaps surprisingly – neither of these relationships holds for their university counterparts. Female students leave less often from university, but college rates are about the same, and the university effect seems to be entirely related to high school grades (women get higher grades, and individuals, whether male or female, with higher high school grades are less likely to leave university). Those who came to the country as immigrants with their parents generally drop out less than others, while visible minorities have about the same rates as others. Those who started their studies when older are much more likely to drop out (if at university), and quit rates go down as students advance through their programs for university students but not college students. Grades are an important predictor of quit rates but are suspected to be at least partly endogenous to the dropping out process.

With this review of the existing literature in hand, we now proceed to the present analysis using the PSIS data.

Notes

1. See, for example, Ferrar and Riddell (2004).
2. See Turner (2004) for a good discussion of the importance of the persistence dynamic in the context of the accumulation of human capital.
3. See Finnie et al, forthcoming, for a collection of recent papers on both access and persistence in Canada, including the paper by Mueller, which represents a more detailed and more technical literature review than the one provided here.
4. In fact, the two reasons are inherently related: concerns over persistence rates are relatively new largely because the empirical evidence that is the source of these concerns has previously been limited.
5. See Finnie and Qiu (forthcoming, 2008) for work using the YITS-B to study persistence. The authors’ work is now being extended to the YITS-A.
6. The importance of these limitations is identified by studies for the United States which indicate that a significant number of students do indeed make such transfers. This is consistent with what Finnie and Qiu (forthcoming, 2008) have recently found for Canada.
7. The Consortium for Student Retention Data Exchange (CSRDE) is a cooperative group of colleges and universities that collects and analyzes retention and graduation data for institutional benchmarking purposes. These data are analysed for first-time, full-time degree seeking freshmen by the Center for Institutional Data Exchange and Analysis (C-IDEA) at the University of Oklahoma. Data are then made available to the 421 consortium members (including a small number of participating Canadian universities along with the great majority of American universities) to use for benchmarking with their peers for internal academic planning purposes.