The dynamics of study-work choice and its effect on intended and actual university attainment

ABSTRACT

We study the dynamics of study-work choices of Australian high school students and how these choices affect intended and actual enrolment in universities when they finish their school education. A dynamic random effect multi-equation model is constructed and estimated. We find that study-work choices are state dependent, driven by student heterogeneity and the school environment they are in. They are also related to behaviours of the fellow students in the same school. We find that study-work choices significantly affect enrolment in universities but they hardly have any effect on students’ preference for university attainment.

KEYWORDS:

• Study-work choices
• university enrolment
• dynamic models

JEL CLASSIFICATION:

• I21
• C33

Acknowledgement

The author acknowledges funding support from a 2011 research grant under the National Vocational Education and Training Research (NVETR) Program of National Centre for Vocational Education Research (NCVER) for earlier work. I also thank Alan Duncan and Rebecca Cassells for their contribution to earlier work.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. In Australia, grades are called ‘year’. For example, Year 12 means Grade 12 or senior year in North America.

2. See for example Meyer and Wise (1982); Greenberger and Steinberg (1986); Cameron and Heckman (1993, 1998); Carr, Wright, and Brody (1996); Ruhm (1997); Schoenhals, Tienda, and Schneider (1998); Eckstein and Wolpin (1999); Oettinger (1999); Hotz et al. (2002); and Parent (2006).

3. The Australian literature (see for example, Robinson 1996, 1999; Dwyer et al. 1999; Marks, Fleming, and McMillan 2000; Vickers, Lamb, and Hinkley 2003; Biddle 2007; Polidano and Zakirova 2011), which consists of mostly descriptive studies, reaches a similar conclusion that in school work is generally beneficial, provided that working time commitment is not too extensive. As an example, Anlezark and Lim (2011) using LSAY data provided an informative description on prevalence of study-work in Australia. Their findings indicate a modest negative impact on educational outcomes for those working longer hours.

4. The threshold of 15 h per week is ad hoc, but equates to an average of three hours per day in each school week.

5. This is achieved by allowing the terms capturing unobserved factors in study-work decisions and educational choices to be correlated.

6. In this analysis, those who secured a position in universities but deferred their actual enrolments (e.g. those who take a ‘gap year’) are included. However, decisions to enrol in universities at later stages by some students are ignored.

7. Some Year 11 and 12 students may enrol into vocational courses, which is part of the Australian secondary educational system. These students involve more paid work than other students. We excluded those who were in apprenticeship or trainee programs. There might still be some students in vocational courses in the sample. Unfortunately, we do not have the information to distinguish these students.

8. Non-working Year 11 students who began work in Year 12, comprise of the sum of 8.78% and 4.83% as a proportion of the total student population who did ‘No work’ in Year 11 – 48.32%. Of those students working in Year 11, 7.16% and 2.03% stopped work in Year 12, which is 18% of the total Non-working student population (51.68%) in Year 12.

9. Students were asked to complete two tests on literacy and numeracy when they were first contacted in 1998. From their answers in these two tests, a standardised (to mean zero and standard deviation of 1) measure of achievement in literacy and numeracy were produced. In the data, however, only a categorical variable (quartiles of achievement) of this measure is available. For more details, see NCVER (2009).

10. For a small number of students who changed school during Years 11 or 12, the identifiers for their new school are not available. This prevents the construction of specific school environment variables for them. In such cases we use as proxies the average value of each environment indicator in the student’s local area, as well as a direct indicator to control for the change of school.

11. The ‘quasi-random’ Halton draws are designed to provide better coverage than independent draws. Simulation can also be more efficient in terms of reduced simulation errors for a given number of draws. See discussions in, for example, Bhat (2001), Train (2003), Sandor and Train (2004).

12. Some individuals were missing from the 1999 wave but returned to the survey in the subsequent wave. To make full use of information, these individuals are included in the analysis, but their probability contributions in wave 2 are ‘integrated out’, by which we mean that the likelihood contribution after Wave 2 by each of such observation is calculated as the weighted average of those conditional on the alternatives in Wave 2.

13. Due to the non-linearity of the model, the pattern is different for each individual student. To illustrate how individual heterogeneity may influence mobility pattern, one could simulate the patterns for each individual.