Empirical findings

Warning View the most recent version.

Archived Content

Information identified as archived is provided for reference, research or recordkeeping purposes. It is not subject to the Government of Canada Web Standards and has not been altered or updated since it was archived. Please "contact us" to request a format other than those available.

We turn now to the results of our empirical analysis of PSE pathways and persistence in PSE based on the PSIS ("L-PSIS") data. We first look at the main transition ("hazard") rates related to graduating, continuing, switching and leaving, as well as the associated cumulative rates to get a longer term perspective of where students are as of each year after starting their programs. We then look at switchers in more detail to see where exactly they go. We then follow those who come back after leaving to see how many do so, and where they come back. The following subsection then shifts gears a bit, stepping out of these more basic persistence dynamics, to look at how many individuals who graduate from a program go on to further PSE studies, and again drill into these patterns in some depth (where, what level, etc.)

5.1   Transition and cumulative transition rates: persistence in PSE

Transition rates

Tables 2 and 3 show the basic transition rates for samples 1 and 2. The latter include three variants: students of all ages, students who started their programs at age 17 to 20, and the same age group where we consider a broader list of programs to which individuals may be classified as switchers (instead of leavers).


Table 2 Hazard transition rates by program year, sample 1



Table 3 Hazard transition rates by program year, sample 2


The calculations show the percentage of students – first after one year ("Year 1") – who had graduated from their programs, were continuing in their programs (i.e., were still enrolled at the same institution) or, alternatively, had switched institutions or had left PSE. (See above for further discussions of the set-up of the data and analytical approach, the construction of the samples, the precise definitions of these transitions, and other methodological issues.)

The numbers for "Year 2" then represent the transitions in that year, again as measured in terms of the student's situation as of the relevant "anniversary" date, of those students who had not made a transition in the first year, which is the defining mark of hazard rate calculations such as these (as explained earlier). Hence a student may have graduated by the end of Year 2, may be still enrolled at the same institution, or may have switched institutions or left PSE. The exercise is then repeated for the third year in the case of Sample 1, but not Sample 2 (for which no Year 3 records are observed due to the construction of the sample, as described above).

Comparing the results for Sample 1 and Sample 2, the data show that continuing rates are generally higher for Sample 2, especially the age-restricted (17 to 20) variant of that sample, while graduation rates are lower, pointing to the different natures of the two samples. In our opinion, Sample 2 "behaves" a little better in terms of capturing the record of individuals from the point they truly enter a new program, rather than in some cases "starting" a program that is really the continuation of one already in progress (which may characterise some of the records in Sample 1, as discussed earlier).

See, for example, the slightly lower graduation rates among bachelor's students in the first years with Sample 2. The results seem "cleanest" of all for the younger group (Sample 2 – age 17 to 20), which is deemed to be the best at picking up individuals as they are truly starting a new program (probably in most cases their first), rather than continuing on in a program that has already been started in one fashion or another. For these reasons, we focus the bulk of our remarks throughout the rest of the paper on the Sample 2 results, with the greatest emphasis on the age restricted version, although in most cases we at least show the results for the broader age group as well.

Focusing on Table 3 (Sample 2), and starting at the university level, the numbers show that the first year "dropout rate" from the point of view of individual institutions (i.e., switchers plus leavers) is 21.7 percent and 20.2 percent in the two top panels (i.e., those of all ages and those who started their programs at age 17 to 20), close to the average first-year dropout rate of 24 percent for 13 Canadian universities reported by Wong (1994).

But some of these "leavers" are in fact switchers: 4.6 and 5.1 percent, in absolute terms, respectively for the two groups, or 21.2 and 25.2 percent in relative terms when compared to all those who leave a given institution (i.e., as compared to the switcher and leaver totals noted above). We thus see that ignoring moves to other institutions substantially overstates the numbers who leave PSE. "True" leaving rates (i.e., from PSE entirely) are left at 17.1 and 15.1 percent in the two samples.1

Switching and leaving rates are considerably lower in the second year as compared to the first, as expected, but remain substantial. For example, in the age 17 to 20 sample, leaving rates decline from 15.1 percent to 11.7 percent, while switching rates go from 5.1 percent to 4.2 percent.

Returning to Sample 1, which has rates out to the third year, the data suggest that both leaving and switching rates then approximately halve from the Year 2 levels at that point. This is probably indicative of what happens in Sample 2 as well, but of course we cannot say for sure if this would be the case in the absence of the actual data.2

Expanding the list of programs to which university students are considered "switchers" (rather than leavers) to include non-PSE programs at PSE institutions (e.g., short courses, language training, and so on) gives the results reported in the bottom panel of Table 3 (again using the younger variant of Sample 2). The effect is to reduce the leaving rate somewhat further, to 13.8 percent, and increase the continuing rate and switching rates commensurately.

From this perspective, the "drop out rate" in the first year goes from 20.2 percent when both leavers and switchers are added together (i.e., the "institution's perspective") and a narrower range of programs are included for switchers (i.e., the second panel of results), to 13.8 percent when only true leavers are considered (i.e., switchers are accounted for) and those switchers include students in non-PSE programs at PSE institutions.

This represents a reduction in the estimated drop-out rate of 6.4 percentage points in absolute terms, or 32 percent in relative terms. The PSIS data thus give a substantially different perspective of the number of PSE leavers relative to what would be estimated with institution-specific data. (See below on the numbers who return to their studies after leaving, which is another important consideration in the calculation of any "global" persistence rate.)

Also notable is that leaving rates are considerably higher among college students than bachelor's students: 22.6 percent (college) versus 15.1 percent (bachelor's) in the first year in the middle panel in Table 3, 19.7 versus 11.7 percent in the second year. This pattern likely reflects college students' generally more marginal attachment to PSE – both in terms of entry, and subsequent persistence patterns.3

What is also different among the two groups is that switching is almost negligible for college students, whereas the numbers are substantial for bachelor's students. These are interesting findings unto themselves, and also suggest that relative drop-out rates for college and bachelor's students based on institution-specific data will tend to understate how many more college students leave PSE relative to bachelor's students – an important finding in any overall assessment of the two systems. We will return to these issues later, when we drill further into where these switchers go.

Graduation rates are, naturally, low for the bachelor's group in these first years after starting their programs, but substantial for college students due to their typically shorter programs.

Cumulative transition rates

Tables 4 and 5 show cumulative transition rates by year. These take into account those who switch or leave (or graduate) in the earlier year(s), as well as those who first continue in their programs but then make one of these transitions in a subsequent year. The calculations thus essentially add the rates together across years to show how many are still continuing in their studies and how many have made any of the relevant transitions by the relevant point in time (after one year, after two years, after three years).4 We again focus on Sample 2.


Table 4 Cumulative transition rates by program year, sample 1



Table 5 Cumulative transition rates by program year, sample 2


The first year rates are (by construction) the same as those already seen, while the second year cumulative transition rates are of course higher as the transitions from the first two years are added together. For Sample 2 (Table 5), we observe two-year cumulative leaving rates among bachelor's students of 27 percent (all ages) or 24.5 percent (17 to 20), with another 7 to 8 percent having switched programs. Taking leavers and switchers together, a total of 34.4 and 32.9 percent (all ages versus 17 to 20) have left the institutions where they started their programs – somewhere around three-quarters of these leaving PSE entirely, the other quarter leaving to study elsewhere.

The cumulative leaving rates for college students are 31.5 percent (all ages) and 33 percent (17 to 20), while switching rates remain very low.

In each case, leaving rates are a little lower, and switching rates in particular are a little higher when a broader range of programs are considered as switches (the final panel of the table).

Students in Master's, First Professional Degree and Ph.D., programs

Transition rates are shown for students in master's, first professional degree and Ph.D., programs in Tables 6 (the hazard rates) and 7 (the cumulative rates based on those hazard rates). These are shown only for Sample 1, because we do not want to restrict the analysis to those who had not taken a previous program, as was our intention in Sample 2. The reason is that we want to include those who started their programs after being in (finishing) another, while the on-going program problem that Sample 2 was largely meant to address would likely be less of a problem with this group.


Table 6 Hazard transition rates by program year for Master's, Ph.D., and first professional degrees, sample 1


Table 7 Cumulative transition rates by program year for Master's, Ph.D., and first professional degrees, sample 1


Not surprisingly, leaving rates are much lower at these levels of study than among bachelor's and college students: in the first year they are 9.5, 5.5 and 6.3 percent at the three levels, respectively (Table 6). Switching is almost non-existent.

One interesting finding is the non-linear pattern for Ph.D. students, with leaving rates dropping from year 1 to year 2, but then rising again in year 3, presumably reflecting, among other influences, the effects of comprehensive exams, which normally take place after two years.

Year 3 cumulative leaving rates (Table 7) are, for the three levels, 19 percent (master's), 8.7 percent (first professional degree), and 12.1 percent (Ph.D.) With the low switching rates that characterise these groups, these rates should be – in contrast to the bachelor's and college results – comparable to what would be obtained with institution-specific data if all institutions in Atlantic Canada could be included in the calculations.

Transition rates by individual characteristics and province

Tables 8 and 9 show transition rates by sex, age, province and cohort for the two main variants of Sample 2 (all ages, age 17 to 20). Tables 10 and 11 show the associated cumulative rates. The patterns are similar for the two samples, varying principally only by level (as would be expected). For convenience, we focus our comments on the younger samples (Table 9 and 11).


Table 8 Hazard transition rates by individual characteristics, sample 2, all ages



Table 9 Hazard transition rates by individual characteristics, sample 2, age 17 to 20



Table 10 Cumulative transition rates by individual characteristics, sample 2, all ages



Table 11 Cumulative transition rates by individual characteristics, sample 2, age 17 to 20


Men leave at considerably higher rates than woman at the university level: 17 versus 13.8 percent in the first year for the 17 to 20 group, 14.4 versus 10 in the second year, with a cumulative difference of 28.4 percent versus 21.9 by the end of year 2. "What's the matter with men?" is thus seen to be a relevant question with respect to persistence rates as well as access rates – i.e., going on as well as getting into PSE.

The implications of these findings are important. Not only do men enter university at substantially lower rates than women (e.g., Finnie, Lascelles, Laporte (2004), Finnie and Mueller (2008), Frenette and Zeman (2007)), they are also considerably less likely to continue on in their studies. Gender differences in final graduation rates – i.e., the numbers actually obtaining degrees – are, therefore, skewed even further than the access rates we have previously been looking to would indicate.

Women's switching rates are, conversely, a bit higher than men's. This means that when we put leaving and switching rates together, which (again) is the "quit rate" from the perspective of individual institutions, the true gender differences in persistence in PSE (i.e., after allowing for switchers) are understated. The benefit of being able to include switchers in our analysis, as is possible with the PSIS data, is again clear. The reasons for these different gender patterns represent an interesting topic for further research.

The patterns by sex are more mixed at the college level: a leaving rate of 22.1 percent for men versus a higher 24.1 percent for women in Year 1, but 21.5 percent for men and a lower 17.9 percent for women in Year 2. As a result, their cumulative rates by the end of Year 2 are almost identical (33.1 and 33.6 percent respectively). Here, the benefits of the PSIS data come in being able to precisely identify quit rates at different points in individuals' programs, with the sample sizes available providing more accurate estimates than would be the case with institution-specific data.

Leaving rates rise substantially with age (measured as of the year in which the person started their program) for bachelor's students (Tables 8 and 10 only since Tables 9 and 11 cover just the 17 to 20 group). Again the cumulative results show these effects most dramatically. After two years, the leaving rates are 24.5, 35.5 and 39.2 percent, respectively, for those aged 17 to 20, 21 to 25, and above 26 at the start of their studies (Table 10).

Conversely, bachelor's students' switching rates decline with age, and in an even more dramatic fashion than the increases in leaving rates in relative terms (although not in terms of the absolute percentage point changes, since switching rates are generally much lower than leaving rates). Cumulative switching rates, by the end of Year 2, are 8.4, 3.7, and 2.2 percent for the three different age groups.

One way to interpret these two sets of results is that older students seem to know better what they want to study, and where (and therefore switch programs less), but are less likely to keep to the task (their higher quit rates). Of course increased family and other responsibilities among older students might figure importantly in these dynamics: for example, reducing mobility, while putting different pressures on the challenges of being able to stay in school. This might be a topic for further research using the PSIS.

For college students, the leaving pattern by age are actually slightly reversed with rates being slightly lower for older students (switching rates remain negligible). Enrolling in college when older thus appears to be a more well-defined path than it is for bachelor's students not only in terms of the greater numbers involved (as seen earlier), but also in terms of the associated persistence patterns.

By province, the differences are perhaps surprisingly small given the varying nature of the different PSE systems in terms of the number of institutions, their locations, tuition fee structures, and more. At the university level, first year leaving rates range from 15.2 percent to 19.1 percent in the all-ages sample, and from 12.9 to 17.5 percent for the more restricted 17 to 20 group. But within these ranges, Newfoundland and Labrador, Prince Edward Island, and New Brunswick are relatively tightly clustered, within a couple or so points of each other, while only Nova Scotia is a bit of an outlier, with lower rates (both samples).

Again the cumulative rates shown in Table 11 emphasize the patterns: leaving rates vary between 25.9 and 27.1 percent for the three "clustered" provinces by the end of year 2, while they are 21.4 percent for Nova Scotia.

In this context, one cannot help but note that Nova Scotia has the highest tuition rates among the Atlantic provinces, and while these simple correlations hardly demonstrate any causal relationship between costs and persistence, the patterns are interesting: higher tuition levels do not necessarily translate into higher leaving rates.

And neither are these lower rates due to any obvious "composition effects", whereby a province with lower access rates to start might be expected to have higher persistence rates, on the grounds that those who enter the system are a more select group. Instead, access rates are in fact higher in Nova Scotia than any of the other Atlantic provinces (and indeed the highest of any jurisdiction in Canada): Nova Scotia is associated with both higher access and higher persistence rates.5

Of course Nova Scotia is different from the other provinces in other ways as well, including their greater numbers of out-of-province (and out-of-region) students6 who are likely to be of generally higher ability, more motivated, and different in other ways that might be associated with, or contribute to, higher persistence rates. It would be necessary to measure provincial differences in persistence rates only after taking these factors into account before we could say anything very definitive about how rates truly compare by province, and – perhaps the more interesting question – why, as well as to draw any conclusions that could impact on future policy.7

Perhaps as surprisingly, switching rates do not vary a great deal by province either. This despite, for example, the fabled ability to "walk across the street to a different institution" in Halifax, and the generally greater number, and wider distribution, of campuses in that province as compared to elsewhere. However, here again more in-depth study would be required before we could say anything very meaningful about these patterns.

At the college level, first year transition rates are also similar across jurisdictions, except for Prince Edward Island whose rates are lower. But sample size becomes a factor here, as is the case for the second year results more generally, so we do not attach too much significance to that particular pattern. For the rest of the college results, many differences are observed, although they are often of a type which causes the patterns to even out over time (i.e., see the cumulative rates in year 2) and otherwise perhaps point to institutional differences such as those relating to specific program length (e.g., see some of the continuing-graduation patterns across years).

An additional set of provincial level results is included in Appendix 2, but the caveats just offered with respect to the overall differences by province apply here as well (probably even more so), so these extra results are left to interested readers to peruse, with caution advised in terms of their interpretation. That said, the differences are mainly in the levels rather than the patterns by the other variables, and therefore point to no obviously interesting stories to add (e.g., the stories by gender, age, and so on are relevant consistent across provinces).

Of course following up on the provincial patterns would make for an interesting line of future research. Suffice it to say that these direct comparisons, facilitated by the PSIS, are in fact new and unique, and thus represent the potential stepping off point to other further studies which probe more deeply into the provincial patterns.

Finally, the small differences by cohort (the last set of results in each case) at both the college and university level point to solid selection procedures for this sample. If, for example, rates had been found to be significantly different for the two cohorts, we might have suspected we were picking up different kinds of students in the two years, which might in turn indicate that our sample selection procedures were not as robust as would be wished for. But this is not the case.

Transition rates at the level of individual institutions

Building on the provincial level results presented just above, Appendix 3 includes a set of persistence results by individual institution (again only at the college and bachelor's level). These are interesting if only because this is the first time such direct comparisons have been possible – again showing the new perspectives of persistence rates afforded by the PSIS.

But the caveats and cautions just offered with respect to the provincial level results need to be repeated – and emphasized even more strongly – with these institutional results. This is (again) because there are many possible reasons for these patterns, and comparing the raw rates essentially reduces this study to a "report card" exercise that is lacking in any real meaning – except to perhaps prompt us to want stakeholders to understand the patterns better, which would in fact be a very interesting extension of the current analysis. Including the results is thus perhaps worthwhile for this alone: to show that such comparisons are possible and to open the door to further work focussed on these patterns.

Why the extreme caution on these institutional differences? Because they may stem from any of the following causes (and perhaps others):

  •  Differences in program structures and/or (possibly associated) differences in the organisation of the underlying data or reporting methods across institutions which generate differences in the data that do not reflect any real differences in underlying behaviour. And this even though the PSIS project aims for as much standardisation as possible, and we have tried to generate numbers that are consistent: see for example the discussions of our sample selection rules, the reasons for our treatment of persistence at the institutional level rather than the program level above, and other discussions of our data and methods above. Despite all of these efforts, some "apples and oranges" problems may still remain.

  • Differences in student characteristics. For example, some institutions might have lower (or higher) persistence rates because they have more (or fewer) inherently "low persistence" (or "high persistence") students than other institutions.8

  • Differences in institutional rules that make it relatively easier to stay or leave, to leave and come back, or to follow other pathways.

  • Differences in provincial-level rules or other possible provincial (policy-related) factors such as student financial aid which affect the institutions in a given province.

  • Differences in external conditions, such as the local unemployment rate, faced by students at a given institution, which may affect the relative benefits, or ease, of going to or staying in PSE.

Identifying and taking account of these factors and any others would be essential before any meaningful interpretation could be made of the institutional results. Not too much should, therefore, be made of the raw numbers on their own, any more than not too much should be made of death rates or other measures of "success" across different hospitals, of student performance measures across K-12 schools (again, see the recent work by David Johnson in this regard), and so on.

But the results do point to what is possible with the PSIS data precisely because Atlantic Canada institutions have participated in the data collection exercise underlying the PSIS project, and these data could be the means of in fact identifying other sets of factors that affect persistence rates, some with possibly interesting policy implications (e.g., change policies at given institutions, the rules regarding transfers across institutions or level of study, etc.)

5.2   Switchers and leavers who return to PSE

Where do switchers go?

Table 12 presents data which affords a closer look at switchers: how many remain at the same level of studies (college, university) but change institution, separating out how may remain in the same province and how many go to a different one, and how many change their level of studies, again either while remaining in the original province or going to a different one. (Keep in mind the Atlantic coverage of the data, meaning that only moves within the region are recorded.)


Table 12 Details on switchers, changes in level and province, sample 2


Among bachelor's students, and focusing still on the 17 to 20 age group (patterns are roughly the same for the broader sample), the data show that switching rates are 5.1 and 4.2 percent, respectively, in years 1 and 2 (as previously seen in Table 3). Of these, in the first year, exactly two-thirds (66.7 percent) stay at the same level of studies in their new programs, and among this group, a somewhat greater share remain in the same province as compared to the number who leave (1.9 percent versus 1.5 percent in actual percentages).

Of the remaining one-third of first-year switchers who change their level of studies from university to college, most stay in their original province (1.4 percent), while the remainder (just .3 percent overall) change both level of study and the province in which they pursue those studies.

Overall, then, just 1.8 percent of all first year bachelor's students (age 17 to 20 – but it is about the same proportion for older students) move to study in a different (Atlantic) province at either the same level of study or at the college level by the beginning of their second year, and in the second year the number is just 1.3 percent. In short, inter-provincial mobility among bachelor's students in Atlantic Canada appears to be quite low.9,10

There are, as seen previously, not many switchers at the college level: for example, just 1.3 percent and .8 percent in the first two years among the age 17 to 20 group, and a mere .9 percent for those of all ages. Interestingly, though, almost all of these change their level of study (i.e., they switch to university) while remaining in the same province.

With all the talk of switching between college and university, the actual numbers involved therefore appear to be very small. Whether they could, or should, be greater, is a topic for further investigation.

How many return to PSE after leaving?

The next dynamic we analyse is the rate of returning to PSE after leaving. To do this, we take those identified as leavers in the first part of the analysis and follow them to see how many return to PSE after that. We are, however, able to follow these individuals for just one year with our preferred Sample 2, because the earliest samples of leavers enter PSE in the second year of data (the 2002/2003 reporting year), leave in the third year of data (2003/2004), and can therefore be observed post-leaving only in the 2004/2005 data.

The results, shown in Tables 13 and 14, are nevertheless interesting. In that first year, 25 percent of all bachelor's leavers of the age 17 to 20 group return to PSE (20 percent in the case of the all-age sample). Overall leaving rates are, therefore, substantially overstated when this group of "leaver-returners" is not taken into account. Otherwise put, "permanent" leavers are considerably fewer in number than the number of "temporary leavers" would indicate – the well-known, but little quantified, "stop-out" phenomenon.11


Table 13 Hazard rates of returning to postsecondary education among leavers, sample 2, same institution


Table 14 Hazard rates of returning to postsecondary education among leavers, sample 2, different institution


Of those who return, about half (11.9 percent of the 25 percent total) go back to the same institution (and same level – i.e., they stay at university). Another 5.8 percent stay at the same level (i.e., university) but change institution, these about evenly split between those who stay in-province (2.8 percent) and those who move to another province within Atlantic Canada (3 percent). A final 7.4 percent change their level of study (i.e., they switch to college), with most of these (5.7 percent) staying in-province, the remaining 1.7 percent changing both level and province12.

Among college leavers, a much smaller proportion of leavers subsequently return to their studies: 11.5 percent (age 17 to 20) and 10.4 (all ages) percent in the first year we observe here. Of these, most return to the same institution (and level), 8.5 and 8.4 percent, respectively. Of the others, the greatest number change level (i.e., switch to university – 2 and 1.4 percent), almost all in the same province. Another small group goes to a different institution at the same level, almost all in another province (0.9 and 0.6 percent).13.

Appendix 2 reports a similar set of findings at the provincial level, but again no particularly interesting stories emerge, and the other caveats about making such comparisons without taking other factors into account apply.

5.3   Graduates who go on to further studies

How many PSE graduates go on to further studies?

We now exploit the PSIS data in a different way, by identifying those who graduate from a PSE program over the period covered by the YITS data and then seeing how many of these individuals start a new program in the following years either directly or after staying out a year or two (or three). We also look at the level of these new programs to get a fuller picture of these dynamics.

The samples used in this part of the analysis are not restricted to those who were included in the analysis of entrants, as focussed upon thus far, or otherwise subjected to the same sorts of selection criteria (including those related to age), since there is no need to do so from an analytical perspective, and doing so would greatly restrict the representativeness of the analysis.

To be included in this part of the analysis, individuals must only have been identified as having graduated from a regular PSE program in the 2001/2002, 2002/2003, or 2003/2004 reporting years of the PSIS. They are then tracked for as long as they could be after that – i.e., from the year of graduation through 2004/2005 (at which point the spells are right-censored in the same manner as in the transition analysis presented above). Individuals of all ages are included.

Given the comprehensive coverage of the PSIS data, we would expect to identify all individuals who enrolled in a new PSE program in Atlantic Canada over this interval. The strength of the PSIS data in this respect over institution-specific data is obvious, since returning to PSE will in a great number of cases include movements to different institutions.14

The analysis will, however, once again not capture those who pursue their studies outside of Atlantic Canada after finishing a first program. The results are thus of potential interest, especially for PSE administrators and policy makers – those within Atlantic Canada in particular – but lack completeness in terms of tracking individuals' subsequent PSE profiles. They effectively represent minimum re-enrolment rates as captured by the subset of students who stay within Atlantic Canada.

Table 15 shows the relevant hazard rates. These are calculated in the same manner as the persistence rates presented above, and represent the numbers of graduates observed to start a new PSE program one, two and three years after finishing their initial diplomas, in the latter two cases conditional on not having already made such a start (or transition) by the year in question and otherwise not being right-censored. The associated cumulative rates are shown in Table 16, and should be interpreted as the proportion of graduates who had started a new program by the indicated year (as calculated from the hazard rates shown in Table 15).


Table 15 Hazard rates of starting a new program among graduates



Table 16 Cumulative rates of starting a new program among graduates


Two sets of numbers are reported. In the first, overlapping programs are allowed and included in the calculations (i.e., the new program may have begun before the completion date of the program initially graduated from). In the second, such overlaps are not permitted (i.e., the new program must have started after the graduation date of the first) and individuals who made such a transition are deleted from the calculations at that point. The reason for this second treatment is again the general ambiguity of interpreting overlapping programs in the PSIS and in real life circumstances. Fortunately, the results are roughly similar in the two sets of findings. For convenience, we focus on the more comprehensive sample.

The rates of continuing in PSE are relatively high, even though new programs taken out of the province are not counted. By three years after graduating (Table 16), more than one third (36.5 percent) of bachelor's students had enrolled in another PSE program, while 30.3 percent of college graduates had done so. Interestingly, the great majority of these (at both levels) enrolled in their new programs in the first year following graduation (see the hazard rates in Table 15 or the increments by year in Table 16). "Gap years" do not appear to be particularly common at the PSE level – although it is certainly a path some follow.

At what level do graduates take their new programs?

Table 12 rounds out this part of the analysis of graduates by reporting the level of the new programs identified in the preceding tables. Among bachelor's graduates, and again focusing on the broader samples, we observe that a full third (34.8 percent) of those who return to PSE do so in non-regular PSE programs which will normally not lead to a diploma (category 98).

These include a wide array of program types, including language courses, other specific skill development courses, and others of the like, as well as pure interest courses. In short, a substantial number of bachelor's graduates appear to return to their studies to top up their skills or otherwise pursue secondary "avocations" in one way or another outside a regular PSE program.

Almost another third (29.2 percent) are in what might be considered conventional or "progressive" PSE career paths (codes 10 through 24): a master's degree, a Ph.D., or a first professional program (medicine, law, etc.) Interestingly, though, another 26.2 percent are enrolled in a new bachelor's program, and an additional 4.6 percent in a university diploma or certificate program below bachelor's, making for a total of 60 percent who remain at university at some level.

Just 4 percent of these bachelor's graduates start a new (regular) PSE program at the college level (codes 05 through 07), which seems like a surprisingly low number given all the attention paid to this path in the popular press. A final 1.2 percent are in programs listed as "below PSE" (code 01), which represent programs for which the usual college entrance requirements (i.e., having graduated from grade 12) don't apply. These include language skills and PSE upgrading/preparation, some apprenticeship programs, and other development and upgrading programs.15

Studying the details of these new programs represent a worthwhile extension of the present analysis. What exactly are these new programs? How do they relate to the individual's previous studies? How long do these second programs take? What is the persistence rate within these programs? What do graduates do after these second programs?

Among college graduates, 39.5 percent are in new regular PSE college programs, 20.3 percent are at the bachelor's level and a trivial 0.2 percent are at a higher university level. Another 17.8 percent are in "below PSE" programs (still in PSE institutions) and 22.1 percent are in non-regular programs at PSE institutions. These are interesting and potentially important pathways that probably merit further analysis.

5.4   Comparisons with the YITS and other possible checks of the PSIS

It is always good to check the findings of any empirical analysis with other data and other studies in order to assess the quality of each underlying data source (no source is ever perfect), the nature of the analysis, the general nature of the findings, and whatever else lends itself to such comparisons. This is especially true in the case of a new data set, a new analysis, or both, which is essentially the situation here.

To this end, we have carried out a set of checks between the PSIS findings reported here and those found with the YITS-B dataset. The YITS-B, as mentioned earlier, is a major Statistics Canada dataset which is comprised of a longitudinal sample of a single cohort of youth aged 18 to 20 who were first interviewed in 2000 and again in three follow-up telephone surveys in 2002, 2004, and 2006 (a final interview was carried out in 2008). The YITS-B is well-suited to the analysis of persistence in PSE, and has been used by the authors to conduct a study that is closely comparable to the one reported here based on the PSIS (Finnie and Qiu (2008)). For the desired checks, we re-did some of the basic parts of our YITS analysis to make it as directly comparable as possible to our PSIS analysis (restricting the data to Atlantic Canada, adjusting some of our definitions, and so on).

A detailed description of these checks and some related findings are reported in Appendix 4. The conclusion of this exercise is that while some differences are found, the results are close enough (in some cases remarkably so) to give us further assurance regarding the quality of the PSIS data, as well as the nature of the analytical framework adopted. Other checks that could be performed, including those based on linking the PSIS and YITS with the longitudinal tax-based LAD file, are also suggested.

Notes

  1. Recall that those who are enrolled outside the Atlantic region are classified as leavers, not switchers, but the bias resulting from this limitation of the data appears to be small, as discussed further below.
  2. Note that the declines from year 1 to year 2 are similar, in relative terms, in samples 1 and 2, thus suggesting we could probably extrapolate out to year 3 for Sample 2 based on the Sample 1 results, but this should only be done with caution, for obvious reasons.
  3. This point is discussed further in Finnie and Qiu, 2008, where it is reinforced in their more detailed analysis of the detailed individual and situational factors observed to be associated with persistence patterns.
  4. These cumulative transition rates are calculated using the hazard rates shown in Tables 2 and 3, and thus comprise a statistical representation of what happens taking into account individuals observed different numbers of years (and therefore left-censored in the underlying data), rather than tracking only the smaller samples of individuals observed for the (entire) indicated intervals. This is standard practice in hazard analysis, as explained earlier in the paper.
  5. See Finnie, Laporte and Lascelles (2004) or Finnie and Mueller (2008) for access patterns by province.
  6. See Burbidge and Finnie (2000).
  7. The MESA project currently has a paper under preparation that attempts to get at how differences in universities' rules and regulations affect persistence rates (after taking students' characteristics into account). The authors may be contacted for further information on this work.
  8. See Johnson (2008) for recent work on attempts to control for student characteristics when looking at K-12 school rankings. His work attempts to get at exactly this kind of problem.
  9. It would be interesting to see how many students move when they enter PSE, but the information in the PSIS is probably not suitable for this. The reason is that what students give for their "home" or "permanent" address is of limited value for these purposes (i.e., to identify where they are from), even in first year. For example, it may be just a mailing address, it may be a local address, and so on. See Burbidge and Finnie (2000) for an analysis of this dynamic based on the National Graduates Surveys.
  10. The relative shares of the different kinds of switchers – those who change level, those who change province – change somewhat in the second year (e.g., there is an increase in the proportion of level-changers among those who make a switch). But the changes are not great, and the overall number of switchers is even smaller than in year 1 (only 4.2 percent make any change at all), meaning that we are splitting relatively fine differences, so probably not too much importance should be attached to these changes.
  11. Recall that leavers are defined as those who are not still/again enrolled one year after starting their programs (give or take a month).
  12. These results should be regarded only as general patterns, rather than precise estimates, due to the smallish sample sizes derived from following those previously observed to leave PSE to see how many return, and where.
  13. Staying at the same level but changing "institution" is almost impossible, by construction, in most cases, because each of the provinces has a unified college system, meaning there is only one institution in the province. There is "campus" information in the PSIS, but this information is, as far as the authors are aware, uneven across the file and has not been exploited, at least to date. This could be something to do in further work.
  14. See Finnie (2004) for previous evidence on a comparable set of dynamics at the national level based on the National Graduates Surveys. The current analysis differs from that work in a number of important ways. Not only is it focused on the Atlantic region, but it also identifies new PSE program starts, whereas the NGS data identify only (additional) completedPSE programs in the two or five years after graduating from a first program.
  15. The coding in the PSIS dataset of "non-regular" and "below PSE" programs leaves some ambiguity as to the nature of these programs. All such programs are delivered by a PSE institution (most often in a college) , but their "below PSE" designation comes from following standard reporting conventions, indicating they are not the same as other "regular" PSE programs.