• Ingen resultater fundet

Premises for reading the synthesis

5.3 The narrative synthesis based on the theoretical model of university dropout

5.3.1 Premises for reading the synthesis

Before presenting the synthesis itself a couple of things need to be pointed out, because they serve as important premises for reading the synthesis. How the evidence on the three review questions is presented

The narrative synthesis comprises the evidence found for each of the three review questions. First, the evidence on ‘What is dropout from university studies?’ will be outlined in Section 5.3.2. Some, but not all, of the studies available for the synthesis have been found to address one or more as- pects of this review question more or less thoroughly. Next, the evidence on ‘Why do such dropout phenomena occur at universities?’ follows in Section 5.3.3, this section includes 42 of the 44 stud- ies available for the synthesis. Because only three33 of the 44 studies investigate the third review question ‘What can be done by the universities to prevent or reduce such dropout phenomena?’, the synthesis will take as its point of departure the evidence for the second review question ‘Why do such dropout phenomena occur at universities?’ and the three studies investigating the third review question will, thus, be included in this guiding ‘story’ where they are found to fit in the best.

After thorough consideration we have chosen not to include in the synthesis four non-European systematic reviews found during the research mapping. They were originally intended to be used to inform the European findings. Several factors lie behind this decision about not to address these non-European reviews. First, there are some contextual differences between European and e.g.

American or Australian cultures and the different cultures are certainly reflected in the structure of the higher education systems of each of these countries. One of the most notable differences between higher education in European countries and in the U.S. is the economic costs of attending higher education. While a current trend towards increasing the tuition fees for university studies is clearly present in some European countries, the fees, as they stand, are significantly lower on av-

33 As such, one study has been found to investigate both the second and third review question.

erage than the average tuition fee for a university student in the U.S. (Erasmus EU34). Further- more, due to structural and organisational differences between the European and non-European system of higher education, it is assessed not to be straightforward to generalise findings concern- ing university dropout across these contexts. Having this in mind strengthens the argument of leaving the non-European reviews out of this synthesis in line with leaving out the non-European primary studies. It is worth stressing, though, that all relevant European-based studies, which have been found in one or more of these four non-European reviews, have been included in the re- search mapping and could therefore in principle also be available for the synthesis. Direction and significance/insignificance of the investigated factors

As mentioned in Section 5.1, to establish the narrative synthesis it is necessary to organise the findings of the studies available for the synthesis in such a way that the direction — and if possible the strength — of their findings can be investigated individually and compared across the studies.

As stated in this section, because of the great heterogeneity found between the studies, for in- stance the fact that often very different factors are investigated in relation to dropout, and be- cause they are investigated in various ways across the studies, the present synthesis will not focus on determining the effect sizes on dropout of each factor investigated across the studies. The syn- thesis will focus on the significance/insignificance and the direction of the investigated factors on university dropout alone. At the same time, a pattern is sought that also takes into account factors that in various ways might prove to have had an influence on the effects found in each study and might explain possible differences in significance and direction of the investigated factors found across the studies available for the synthesis. Vote counting

Furthermore, as stated in Section 5.1, because the studies available for the synthesis investigate so many different factors on university dropout, each factor has been grouped into one of nine over- all categories (e.g. students’ socio-demographic background, students’ prior schooling, study con- ditions within university, students’ academic and social integration within university etc.). To eval- uate the evidence for each of these overall categories and their underlying aspects of university dropout, a sort of ‘vote counting’ is utilized. That is, where meaningful, an assessment will be made for each of the overall categories of how many studies report significant positive or negative effects and how many studies report insignificant effects in relation to how many studies have actually investigated the different aspects of each overall category. This will lead to a first tenta- tive assessment of which aspects are influential on university dropout/on different types of uni- versity dropout behaviour and which aspects are found to be less convincing in their influence on university dropout. A couple of pitfalls should be noted in relation to vote counting, cf. Gough et al. (2012: 190). One of these pitfalls concerns the fact that vote counting does not take into ac-

34 http://www.erasmuscu.com/the-difference-between-studying-in-europe-and-the-usa.php.

count the effect sizes, but merely the direction of the factors investigated across the studies. Since the heterogeneity of the studies available for the synthesis cannot justify such a level of accuracy across studies in the first place, cf. Section 5.3.2, it is not a problem in this particular case. Other pitfalls concern the notion that vote counting normally treats studies of different size and quality alike. These pitfalls have been sought circumvented by giving a special role to the so-called ‘core’

studies, see the following section. Core studies

Lastly, as was pointed out in Section 3.4, in relation to Tinto’s theoretical model of college student dropout and as will become clear in the next section concerning ‘What is dropout from university studies?’ (cf. Section 5.3.2 below), it is essential to distinguish between different types of dropout behaviour when investigating the review question ‘Why do such dropout phenomena occur at uni- versities?’ and the related ‘What can be done by the universities to prevent or reduce dropout?’. To quote Tinto once again: “*…+ it is, as noted, important to distinguish between the varying types of dropout behaviors, especially between academic dismissal and voluntary withdrawal. This is so not only because these behaviors involve different persons but also because they result from different patterns of interaction within the college setting.” (Tinto, 1975: 116) As noted in Section 3.4, a fail- ure to distinguish between these different dropout behaviours when investigating the possible effect of different factors on dropout might at best lead to insecure findings and at worst to mis- leading/contradictory findings across the studies (ibid.: 90). Whereas most of the studies available for this synthesis unfortunately have been found not to make such an analytic distinction, a small- er number of studies have been identified to do so. None of the studies do in their quantitative analyses differentiate between all five groups of university students: persisters, involuntary drop- outs (i.e. dropout due to academic failure), voluntary withdrawals, transfer students and perma- nent/formal dropouts (i.e. transfer students and formal dropouts should be viewed as subgroups of involuntary dropouts and voluntary withdrawals). However, five studies have been found to compare possible determinants of dropout directly across persisters, involuntary dropouts and voluntary withdrawals (¤ITT2762111)35 or directly across persisters, transfer students and formal dropouts in their quantitative analyses, by the means of e.g. a ‘competing risk’-framework (¤ITT2763715; ¤ITT2777620; ¤ITT2770888 + ¤ITT2770886; ITT2770887).36 Of these five studies, four of them (¤ITT2762111; ¤ITT2763715; ¤ITT2777620; ¤ITT2770888 + ¤ITT2770886) have additionally

35 As above stated, for a list of references associated with the item numbers (ITT…), Chapter 9 ‘References for the studies available for the synthesis’.

36 A few other studies have been found to distinguish between persisters, transfer students and dropouts for reasons other than transfer. These studies, however, do not compare the possible determinants of dropout directly across these groups in their quantitative analyses.

been assessed in the research mapping to be broadly generalisable37 and given an overall ‘high’

weight of evidence. These four studies thus comprise a relatively greater complexity, and their findings are therefore considered to contain a greater validity and precision than the findings of the other studies available for this synthesis, while at the same time being broadly generalisable.

On those grounds, these four studies are termed the core studies and will be given a special role in the synthesis as their findings will be highlighted and used to inform the findings of the other stud- ies in case they are divergent, cf. Section 5.3.3. When referred to in the subsequent sections of this chapter, the core studies will be marked with this symbol ¤. Another five studies (ITT2777714 + ITT2777715; ITT2758964; ITT2762308; ITT2773000 + ITT2773001; ITT2770695) explicitly distin- guish between either involuntary dropouts and voluntary withdrawals or transfer students and formal dropouts by excluding one of the two groups, respectively, from at least some of the anal- yses. Thereby, the analyses include as dependent variable a dichotomy between persisters and one specific type of dropout. The findings of these five studies have all been assessed to be less generalisable, and each of the studies has been given an overall ‘medium’ weight of evidence, which is why they are not included in the core studies.