Youth Training - Statistical Profile 1999 to 2008

9. Statistical modelling using schools data

The Ministry of Education has created a longitudinal transitions dataset that integrates schools data and records of all formal post-school education activity. The advantage of using this dataset to analyse Youth Training is that it contains schools-related variables not present in Youth Training administrative datasets, and it can provide further richness of demographic–related variables, particularly for ethnic group.

By linking schools data with the Youth Training administrative dataset, it is possible to get a clearer view of  the sorts of things that are important in predicting entrance into Youth Training. These variables are used as independent variables in a logistic regression model.

The main research question was, for all school leavers leaving school in certain years, what predicts transition to Youth Training? For the purposes of this study, entrance into Youth Training may be directly from school, or the transition may be indirect. A transition into Youth Training occurs when a person participates in a Youth Training programme in the same year or a year subsequent to (that is within three years of) the year in which the learner is judged to have left school.  This definition of a transition may differ from those used in other analyses. This means that whatever occurs between school and the first Youth Training enrolment, if anything, is not accounted for in this model.

School leavers leaving school in 2005, 2006 and 2007 were included in this analysis. A combination of a Youth Training variable (date left school) and a transitions dataset variable school leaving date was used to roughly determine the date the student left school.

Linking Youth Training and schools data is complicated by the fact that the schools record is in this case derived using NQF achievement data. Youth Training learners may have a National Student Number, but may not have an entry on their record of learning. This is because schools (until recently) were not required to report ‘not achieved’ standards to NZQA. Because Youth Training learners are more likely to have no qualifications on entry,  they are less likely to have ‘populated’ NQF records, and so schools-related data is difficult to source for them.

This is controlled in the model through the Unit Standards taken variable, which measures the proportion of standards achieved that are identified as unit standards. It also records if a learner has no credits in their record of learning, taken as a proxy measure that they have attained no qualifications or standards at school.

Ussher (2008) found that school leavers who participate in vocational type learning (as opposed to academic-type learning) in post-school destinations are more likely to have taken unit standards at secondary school than those participating in bachelors-level study. Inclusion of this variable in the model tests whether this is also the case for entrants into Youth Training.

The transitions dataset contains information collected over a period of time and in different settings, making it possible to assess ethnic group based on the consistency of ethnic response. Three levels of ethnic identification have been derived and applied to each learner, based on the frequency and consistency of their ethnic self-identification, and dummy variables for each possible ethnic group have been created for each learner. These levels are: never, ever and sole.  A learner is categorised as ‘never European’ if he/she has never selected the European ethnic grouping in the included administrative data collections. If they have selected European in one or more collection, in conjunction with some other ethnic group, they will be recorded as ever  European. If they have always identified as European in the constituent datasets, they are categorised as sole for each ethnic group.³⁰

9.1 Summary of effects

The model was able to explain 15 percent of the observed variance, a relatively low explanatory power (n=174,355, R square = 0.1488). There are likely to be a great deal of things that cannot be measured that contribute to a person’s education choices, so it is not unusual for statistical models of education data to have low explanatory power.

A number of other variables could have been included in this model, but are omitted because of collinearity issues. For example, highest educational attainment at school correlated highly with the unit standards taken variable because of the inclusion of the non-attainment category.

Table 46 shows the summary of the model. The variables are ranked in the order of the amount of variation in the model accounted for by each. The higher the variable is in the list, the more important it is in predicting the dependent variable.

The most important predictor in this model was the proportion of standards achieved in NCEA that are unit standards, followed whether the learner identifies as Asian or Māori, the school decile, the year the learner left school and whether the learner was granted an early leaving exemption (which is likely to be highly correlated with the year left school variable). Our analysis identified high correlation between these two variables and school decile, however, decile was left in as a proxy measure of socio-economic status of the learner. The cost of including decile is the possibility of inflated explanatory power of the model, so predicted probabilities should be interpreted cautiously.

Table 47 shows the summary of the model. The variables are again shown in the order of the amount of variation in the model accounted for by each variable. The predicted probability and the odds ratio columns are probably the most important ones to pay attention to. The odds ratios are the odds of each category of the variable over the reference category with respect to the chances of transitioning to Youth Training.

The strongest effect was the unit standard variable. School pupils who attained no credits overall, or took high proportions of unit standards at school were more likely to transition to Youth Training than those who took mainly achievement standards.

The odds of not transitioning to Youth Training are almost double for those who have never indicate Asian ethnicity over those who have always done so. This indicates that Asian pupils have not been transitioning to Youth Training at the same rate as other ethnic groups.

The odds of transiting to Youth Training improve with the frequency of identification as Māori. Pupils who have never indicated Māori ethnic affiliation are less likely to enter Youth Training than pupils who have sometimes, or have always indicated they are Māori.

Pupils leaving schools with low school deciles (1 and 2) were 1.24 times as likely to transition to Youth Training as a pupil leaving a decile 7 or 8 school. The year the pupil left school also has a bearing on whether they transition to Youth Training. This variable probably encompasses a variety of external effects, such as the variability of the strength of the economy and by association the labour market between years, as well as the variability between further education options. School leavers who left in 2007 were more likely to transition to Youth Training than those who left in 2005, possibly due to a weakening in the labour market for industries young people traditionally work in.

If a learner is granted an early leaving exemption, they are over twenty times as likely to enter Youth Training than similar school leavers who have not. The predicted probability of an early leaving exemption leaver entering Youth Training is 10 percent, compared to the reference category’s 1 percent. This may be because many of those who leave school early are routinely referred to a Youth Training provider.

School leavers in highly urbanised areas and in isolated areas are more likely to transition to Youth Training than those leaving schools situated in less dense urban areas or those that are considered to be highly isolated, but the difference is small.

School gender has a small effect. Pupils leaving all-boys senior schools were less likely to transition to Youth Training than those leaving co-ed or all-girls senior schools.

Pupils who have never indicated Pasifika ethnicity are less likely to enter Youth Training than pupils who have sometimes indicated they are Pasifika. Pupils who have always indicated Pasifika ethnicity are slightly less likely to transition into Youth Training than those who sometimes have.

Footnotes

 
30. See Engler, 2010, for further discussion on never, ever and sole ethnic identification variables.
 
31. Quintiles refer to the proportion of standards attempted that are unit standards. Learners categorised in quintile 1 show a low ratio of unit standards attempted over achievement standards, while a quintile 5 category denotes that over 80 percent of standards attempted by the learner were unit standards.
 
32. This means no unit or achievement standards are recorded for the learner.