Last changed 25 Feb 2008 ............... Length about 900 words (7,000 bytes).
(Document started on 7 Feb 2007.) This is a WWW document maintained by Steve Draper, installed at http://www.psy.gla.ac.uk/~steve/localed/tintorev.html. You may copy it. How to refer to it.

Web site logical path: [www.psy.gla.ac.uk] [~steve] [localed] [dropout] [this page]

Notes on the retention literature

By Steve Draper,   Department of Psychology,   University of Glasgow.

This page is basically not written. For now I have just 3 things to say.

My older notes giving some perspective on retention are here.

As those notes say, the single biggest name to look for in the literature is Tinto. For the UK, the second biggest name to look for is Mantz Yorke. E.g.

  • Yorke,M. (1999) Leaving early: undergraduate non-completion in higher education (London: Taylor and Francis)
  • Yorke,M. & Longden,B. (2004) (eds. and authors) Retention and student success in HE (SRHE/OU press)

    A paper worth looking at, from a statistician's viewpoint (rather than from Tinto's or from the education literature's viewpoint) is:
    Smith J. & R. A. Naylor (2001) "Dropping out of university: a statistical analysis of the probability of withdrawal of UK university students" Journal of the Royal Statistical Society (Series A) vol.164 pp.389-405
    It is interesting for at least these reasons: it is UK data, it is a re-analysis of government data, it looks at the independent contributions of many factors to dropout. It shows that the patterns of male and female dropout are substantially different. It has a substantiated argument about why league tables comparing university dropout rates are largely meaningless. But all the many factors they analyse account together for only about 10% of the variance, so most of it could be susceptible to improvement -- or could just be beyond control.

    Research methods on dropout

    Need 100% samples especially of the dropouts: any less, and self-selection must be likely to distort it by losing those ashamed in some way (or leave only those with the most distorted rationalisations).
    Furthermore, face to face as opposed to paper instruments (i.e. interviews not postal questionnaires) may be very important for quality all round. Certainly comparing face to face persisters with paper dropouts could be bad.

    Even then, it will be like interviewing people about their divorces: everyone will have a story, but it is a story they can live with, scarcely a dispassionate account. Rationalisation by each student, particularly dropouts, may mean that what they say about causes is not useful. They will be very likely to describe cause as external factors (the classic Social Psychology attribution error?). So for this, should attend only to data on external factors, and get it equally for persisters. In fact the Brown and Harris method of collecting descriptions of external factors for all, and getting a panel of experts to rate their seriousness "blindly", may be essential.

    Similarly for "internal" and all "ask them" measures of attitude, Tinto integration etc.: we should ask all students before as well as after external events, and before exam results, and before dropouts. I.e. do prospective studies.

    SO:

    1. Prospective studies, with measures (especially subjective/internal ones) taken before (as well as after) dropout events such as failing exams, and collecting these measures for both persisters and dropouts.
    2. For external events (always collected retrospectively), collect these for both persisters and dropouts.
    3. Get a panel to assess the seriousness of external events; don't trust subjective assessment. (And hence, don't use dropouts' own opinion on why they dropped out.)
    4. Must get 100% or random samples especially of dropouts (not self-selected samples).

    Observation vs. experiment

    The above assumed that research would be looking at actual dropouts: essentially "observation" i.e. questioning those who do and don't drop out and hoping to spot factors that might be the causes. That should obviously be done, since at least everything mentioned or noticed is a potential cause and we might not think of these factors otherwise. Yorke's studies have done that job, and can tell us what the common causes of dropout are at least as far as what students report them to be.

    The other approach is "experimental": intefere, and report what actions actually change dropout, regardless of what participants report. This gets round the important issue of self-interest in reporting dropouts, which is a significant life event. Studies reporting this kind of evidence are mentioned in this page. This yields the "best" evidence in the sense of the most believable as far as it goes. Drawbacks are that it is limited to testing the interventions that have occurred to researchers (and there's no guarantee they guess good ones to try), and interventions that are cheap enough to be easy to get done. Thus the effect of single phone calls, and short written passages have been tested (with surprisingly powerful results); but there have been no studies of systematically varying the induction procedures of whole institutions.

    Web site logical path: [www.psy.gla.ac.uk] [~steve] [localed] [dropout] [this page]
    [Top of this page]