The impact of increasing university participation on the pool of apprentices

Tom Karmel
David Roberts
Patrick Lim

National Centre for
Vocational Education Research


About the research

The impact of increasing university participation on the pool ofapprentices

Tom Karmel, David Robertsand Patrick Lim, NCVER

In recent years, Australian governments have placed considerable emphasis on the importance of both university participation and undertaking an apprenticeship. This paper looks at whether there is a relationship between the two and, in particular, whether the expansion of university participation (for example, the uncapping of university undergraduate places following the Bradley Review [Bradley
et al. 2008]) is likely to have an impact on the pool of those undertaking a trade apprenticeship. The authors consider certain aspects of an apprentice’s background: reading and mathematics achievement at age 15 years and socioeconomic status. The potential impact of an expansion in university participationon the pool of apprentices is examined by comparing two cohorts from the Longitudinal Surveys of Australian Youth (LSAY): the Y95 cohort who were in Year 9 in 1995 and the Y06 cohort who were aged 15 years in 2006.

Key messages

  • The likelihood of undertaking an apprenticeship is affected by the propensity to go to university.
  • Young men are less likely to undertake an apprenticeship if they are academically inclined.
  • Apprenticeships are more likely to be undertaken by young men from a lower socioeconomic status background.
  • LSAY shows that participation in both university and apprenticeships grew between 1995 and 2006.
  • Young men who were less academically inclined and from low socioeconomic status backgrounds contributed to this growth in apprenticeships.
  • The growth in university participation has come from academically lower-performing young men with a higher socioeconomic status background.

The authors note that any educational expansion (whether through apprenticeships or attendance at university) will also have an impact on that part of the population who previously were neither undertaking an apprenticeship nor going to university. They also observe that those who are in the best position to take advantage of opportunities in both apprenticeships and university places do so, irrespective of whether position is measured by mathematics and reading achievement or socioeconomic status.

Rod Camm
Managing Director, NCVER

Contents

Tables and figures

Executive summary

Introduction

Descriptive statistics

Statistical methodology

Results

Discussion

Final comments

References

Appendices

A: Description of the data used

B: Models

C: Regression results

Tables and figures

Tables

1Descriptive statistics of achievement and SES variables (mean and standard deviation)

2Variables used for predicting probability of going to university or undertaking an apprenticeship

3AIC values for the four alternative models

4Variables included in interactions with cohort dummy variable

5Regression results for predicting apprenticeship: coefficients

A1Common variables for both cohorts

A2Correlation matrix

A3Model fit information

A4R-squared

B1Variables used for regression predicting university attendance
by wave 5

B2Variables used for regression predicting apprenticeship by wave 5

C1Regression results for predicting university attendance: model fit statistics and type III analysis of effects

C2Regression results for predicting university attendance: coefficients

C3Regression results for predicting apprenticeship: model fit statistics
and type III analysis of effects

C4Regression results for predicting apprenticeship: coefficients

Figures

1Distributions of mathematics achievement for 19-year-old males
across the three options of: undertaking an apprenticeship, neither undertaking an apprenticeship nor going to university, and going to university, Y95 cohort

2Distributions of reading achievement for 19-year-old males across
the three options of: undertaking an apprenticeship, neither
undertaking an apprenticeship nor going to university, and going to university, Y95 cohort

3Distributions of socioeconomic status for 19-year-old males across
the three options of: undertaking an apprenticeship, neither
undertaking an apprenticeship nor going to university, and going to university, Y95 cohort

4Predicted probability of being an apprentice by achievement scores,
Y95 males

5Predicted probability of being an apprentice by socioeconomic
status, Y95 males

6Change in predicted probabilities by mathematics achievement
scores, Y95 males

7Change in predicted probabilities by reading achievement scores,
Y95 males

8Change in predicted probabilities by socioeconomic status,
Y95 males

9Relative change in predicted probability between cohort 1 and
cohort 2 of being an apprentice, going to university, or neither, by mathematics and reading achievement, calculated for Y95 sample

10Relative change in predicted probability between cohort 1 and
cohort 2 of being an apprentice, going to university, or neither, calculated for Y95 sample

11Synthetic distributions of apprentices across mathematics quintiles,
all Y95 males

12Synthetic distributions of apprentices across reading achievement quintiles, all Y95 males

13Synthetic distributions of apprentices across SES quintiles, all
Y95 males

14Synthetic distributions of those going to university across
mathematics achievement quintiles, all Y95 males

15Synthetic distributions of those going to university across reading achievement quintiles, all Y95 males

16Synthetic distributions of those going to university across SES
quintiles, all Y95 males

17Synthetic distributions of neither going to university nor
undertaking an apprenticeship across mathematics achievement
quintiles, Y95 males

18Synthetic distributions of neither going to university nor
undertaking an apprenticeship across reading achievement
quintiles, Y95 males

19Synthetic distributions of neither going to university nor
undertaking an apprenticeship across SES quintiles, Y95 males

Executive summary

The purpose of this paper is to provide an analysis of the impact of increasing university participation on the intake quality of apprentices. In this paper, we use two cohorts of the Longitudinal Surveys of Australian Youth (LSAY), separated by 11 years, and compare male trade apprentices. In particular, we investigate whether their academic ability measured at age 15 years and their socioeconomic (SES) backgrounds have changed over this time period, and whether the increase in the probability of going to university has impacted on these characteristics.

The results from this paper show that over this 11-year period the probability of undertaking both apprenticeships and university has increased. This paper showsthat, between the 1995 and 2006 cohorts, the increase in the probability of going to university impacted on the quality of apprentices unambiguously. More apprentices come from the bottom two quintiles in relation to mathematics and reading achievement. However, given the expansion of apprenticeships, there has been very little movement in the number of apprentices who have mathematics and reading achievement in the top two quintiles.

In terms of university participation, this paper also found that the expansion of higher education has resulted in a noticeable shift in the proportion of male university students in the top academic achievement quintiles.

A further finding from this paper is that with the expansion of both higher education and apprenticeships the growth in higher education has come from individuals with middle to high socioeconomic status backgrounds, whereas apprentices are likely to have come from those who have lower SES.

It should be noted, however, that apprenticeships are quite different from university places. While the government can easily increase the latter, and has by uncapping undergraduate places, the former depend on the willingness of employers to offer them. This willingness primarily depends on the labour market and the opportunity it offers in terms of activity in the trades. When we think in these terms there is a certain symmetry in the impact of expansion in both apprenticeships and student places. Those who are in the best position to take advantage of opportunities do so, irrespective of whether position is measured by mathematics and reading achievement or socioeconomic status.

Introduction

The purpose of this paper is to provide an analysis of the impact of increasing university participation on the intake of apprentices. Our starting point is that the proportion of young people going to university is increasing and this potentially may impact on those young people who otherwise would have undertaken an apprenticeship.[1]Those choosing to go to university are unlikely to be the ‘average’ apprentice, and therefore it is likely that there will be a decline in the number of ‘high ability’ apprentices. This is an eminently plausible argument, unless the population of potential apprentices is completely separate from the population of potential university students, and this seems most unlikely.

Thus our goal is to investigate how the expansion of higher education has affected the apprenticeship cohort. However, we have an immediate problem, that of ‘simultaneity’: it is a joint decision to go to university, undertake an apprenticeship or do neither. One possibility would be to model these three outcomes as a function of the usual background variables (academic ability, socioeconomic status and the like) and see how this model has changed over time. The problem with such a standard approach is that we would end up with a description of how the overall distribution of apprentices and university students has changed but it would not help in separatingout the impact of increasing higher education on the apprenticeship cohort. To do so, we need greater structure in the model, such that the likelihood of going to university impacts on the probability of undertaking an apprenticeship.

An obvious response to such an approach is that if you assume that the probability of going to university impacts on the probability of undertaking an apprenticeship, should not you also assume the converse: that the probability of undertaking an apprenticeship impacts on the probability of going to university? However, if we were to do that, we would have an identification problem, unless somehow the variables feeding into the probability of going to university are different from those feeding into the probability of undertaking an apprentice. There are no obvious reasons for using different variables.

The way we resolve this dilemma is to test both possibilities against each other. That is, we see whether the probability of going to university feeding into the probability of undertaking an apprenticeship fits the data better than the converse model.A priori our belief was that the former model would dominate the latter on the basis that there is a difference in status between going to university and undertaking an apprenticeship (see for example, Laming 2012, who argues the dominance of university as the aspiration of choice for school leavers; also Alloway et al. 2004 on the academic paradigm at school impacting on choice). While we could proceed on this assumption, it is more satisfactory to test the proposition statistically.

Our focus is on two aspects of apprentices. The first is‘quality’, which for the purposes of this paper we define as the level of academic achievement as measured at roughly age 15 years. The second is socioeconomic status. We apply the focus on academic quality for two reasons: the practical reason is that our dataset, the Longitudinal Surveys of Australian Youth,collects such data, and the additional reason is that complaints about the poor preparation of apprentices tend to focus on academic rather than practical attributes.

We compare two cohorts of apprentices, specifically male trade apprentices, in order to see how ‘quality’and socioeconomic statushave changed over time and how the change in the probability of going to university has impacted on the distribution of apprentices against these variables.

Our choice of cohorts is limited by the available datasets.[2] The earliest cohort readily available is Y95(respondents entered the surveywhen they were in Year 9 in 1995),while the latest is Y06,when the students entered the survey at age15 years in 2006, over ten years later.

We proceed as follows. In the next section, we describe the data and look at the relationship between the distributions of apprentices and those going to university. We look at three dimensions: mathematics achievement score, reading achievement score and socioeconomic status. The following section describes our modelling methodology. We then present the results in a number of ways before some final comments.

Descriptions of the data are given in appendix A.

We find a strong negative relationship between the high probability of going to university and the probability of undertaking an apprenticeship. That is, those with a high probability of going to university are much less likely to undertake an apprenticeship. Moreover, statistical testing justifies the assumption that going to university is the dominant decision and that the probability of undertaking an apprenticeship is affected by the probability of going to university, not vice versa. This result allows us to trace how changes in the probability of going to university have flowed through to the probability of undertaking an apprenticeship. We find that the increase in the probability of going to university has had an effect on the distribution of apprentices (across ability and socioeconomic status). However, a confounding effect is that the probability of being an apprentice has also increased between the two cohorts. We also find that there has been an overall increase in both university attendance and the uptake of apprenticeships between the Y95 and Y06 cohorts, and thata decrease in the probability of neither going to university nor undertaking an apprenticeship is an important part of the story.

It is clear that the increase in the probability of going to university has a differentially greater effect on those of high academic achievement and those from a higher socioeconomic status background.The analysis also illustrates that any impact on the distribution of apprentices (by academic ability and socioeconomic status) is complicated. This is because the change in the distribution depends upon not only the differential increase in the probability of going to university but also on the underlying distribution; for example, the biggest impact on the probability of going to university may occur in the part of the distribution where there are very few apprentices, and vice versa.

The results show that for males, between the 1995 and 2006 cohorts, the increase in the probability of going to university impacted on the quality of apprentices unambiguously. More apprentices come from the bottom two quintiles in relation to mathematics and reading achievement. However, the expansion in the number of apprentices has had an offsetting effect to some extent, so the proportion of apprentices in the top two quintiles is little changed. Of some interest is the additional finding that, perhaps not surprisingly, the expansion of higher education has resulted in a noticeable decline in the proportion of (male) university students in the top quintile.

The analysis also suggests that an expansion in the university sector had a negative effect on equity. The distribution of socioeconomicstatus shifted towards higher-SES individuals for those going to university and towards lower-SES individuals for those undertaking apprenticeships. The expansion of the university sector, taken with the increase in apprenticeships, has had a compounding effect on the proportion of the lowest-SES individuals undertaking neither an apprenticeship nor going to university.

Descriptive statistics

We now move to our analysis of the two LSAY cohorts.

We provide a simple description of the distribution of those undertaking an apprenticeship and those going to university against the three variables of interest: mathematical achievement at age 15 years, reading achievement at age 15 years and socioeconomic status. Each of the three variables has been constructed such that the underlying overall distribution is the same. The SES variable has been built from relevant variables in LSAY (specifically mother’s and father’s education and occupation). There is a wide range of background factors that can be used to derive an SES measure. However, the variables chosen to measure socioeconomic status needed to be consistent across the two cohorts.

Table 1 provides some simple descriptive statistics of the achievement and SES variables for the two cohorts we are considering.

Table 1Descriptive statistics of achievement and SES variables (mean and standard deviation)

Activity / Percent
in cohort / Mathematics / Reading / SES
Y95 / Y06 / Y95 / Y06 / Y95 / Y06 / Y95 / Y06
Apprenticeship / 14.8 / 25.4 / -0.26(0.95) / -.343(0.99) / -0.43(1.02) / -0.55(0.97) / -0.30(0.54) / -0.04(0.50)
University / 26.0 / 39.5 / 0.73(0.86) / 0.80(0.82) / 0.44(0.86) / 0.53(0.79) / 0.02(0.58) / 0.30(0.53)
Neither / 59.2 / 35.1 / -0.12(1.01) / -0.17(0.89) / -0.28(1.01) / -0.41(0.84) / -0.29(0.57) / -0.02(0.54)

Figures 1 to 3 show the distribution of mathematics and reading achievement and socioeconomic status for those who undertook an apprenticeship, university, or neither, by age 19 years. A smoothed representation of the distribution is also shown. For brevity, we only present the results for the Y95 cohort (noting that the distributions are similar for the Y06 cohort).

Figure 1Distributions of mathematics achievement for 19-year-old males across the three options of:undertaking an apprenticeship, neither undertaking an apprenticeship nor going to university, and going to university, Y95 cohort

Figure 2Distributions of reading achievement for 19-year-old males across the three options of:undertaking an apprenticeship, neither undertaking an apprenticeship nor going to university, and going to university, Y95 cohort

Figure 3Distributions of socioeconomic status for 19-year-old males across the three options of:undertaking an apprenticeship, neither undertaking an apprenticeship nor going to university, and going to university, Y95 cohort

The observation from these graphs is that those attending university have higher average mathematics achievement than the other two groups. A further point is that mathematics achievement for the university group has a slightly smaller variation than the other two groups. However, there is a large degree of overlap in the distributions across all three groups, suggesting that at least to some extent university and apprenticeships are drawing on similar populations. This suggests that changes to the probability of going to university are likely to impact on the probability of undertaking an apprenticeship. From this it would seem likely that a change in university participation will impact on the distribution of apprentices over the mathematical achievement. Similar arguments apply to reading achievement and socioeconomic status.

The relationships between socioeconomic status and participation at university and undertaking an apprenticeship and socioeconomic statusare weaker than for mathematics and reading achievement but in the same direction. While there are differences between the three groups, it is clear that there is considerable overlap.

The above distributions are descriptive and do not take into account factors other than academic achievement and socioeconomic status that impact on the probability of undertaking an apprenticeship or going to university. This is remedied in the next section.

Statistical methodology

Our initial task is to test the assumption that going to university is the dominant decision for young people. In a statistical sense, this assumption says that the probability of undertaking an apprenticeship is affected by the probability of going to university more than the probability of undertaking an apprenticeship is influenced by the probability of going to university.

The methodology we employed to test this assumption was to fit the following statistical models:

  • probability of going to university against explanatory background variables, without the probability of apprenticeships
  • probability of going to universityagainst explanatory background variables, with the probability of apprenticeship
  • probability of undertaking an apprenticeship against explanatory background variables, without the probability of university
  • probability of undertaking an apprenticeship against explanatory background variables, with the probability of university.

The explanatory variables were as in table 2.