School leavers’ progression to bachelors-level study

4.1 Ethnic group

This study presented an interesting methodological challenge in dealing with ethnicity. The study population of students was derived by merging four different sources of data; school NCEA results, formal tertiary-provider-based enrolments, and enrolments in industry training and in targeted training undertaken in tertiary institutions. Each of these data sources contained information about the ethnicity of the student, collected independently of the others and at different times. The NCEA results and provider-based enrolment data each contained up to three responses a person gave to indicate their ethnicity. The industry training and targeted training data each have one ethnicity field, which has been prioritised in the order Māori, Pasifka, Asian, Other and European.

Previous studies using the linked school and tertiary outcomes data had resolved this problem by either using the NCEA student data as the source of a student’s ethnicity, or the tertiary enrolment data as the source. However, using just one source of data to determine ethnicity does not take into account the fact that an individual’s ethnicity can and does change (Callister et al 2009). The problem then was how to accurately and appropriately record the (possibly varying) ethnicity of a student, using the diverse ethnic data available.

One option would have been to use a prioritised method to report ethnicity, based on all the ethnicity fields in the data. However, official guidance from Statistics New Zealand discourages the use of prioritised ethnicity (Statistics New Zealand 2006), and the method can result in undercounting some ethnic groups⁴. The recommended method of reporting ethnicity is to use single/combination responses. In this method, a person is placed in the particular category they fall in: for example the single responses Māori or European, or the combination responses Māori/European or Māori/European/Pasifika. This method has limitations (Leather 2009), and particularly for this study, results in small cell sizes for some categories. It also does not address the problem of changing ethnicity.

The method used in the present study to resolve these difficulties is to use the ‘ever-ethnic’ method of reporting ethnic group⁵. Three categories of ethnicity are determined – ‘never’, ‘ever’ and ‘sole’ – for each ethnic group. It is essentially the single/combination method of reporting ethnicity but with many of the categories combined. By including all the sources of ethnicity in all the data across years, it also captures any changes to a person’s ethnicity over time.

To derive the ethnicity of a student in the study population, all ethnicity fields, in all datasets, in all years, are considered. Considering each ethnicity in turn, if the data shows only one ethnicity, the student is placed in the sole-ethnic category. If the data never shows that ethnicity, the student is placed in the never-ethnicity category. And if the data shows that ethnicity in some cases and not others, or in combination with another ethnic group, the student is placed in the ever-ethnicity category.

For example, considering the Māori ethnic group, if a person was recorded as Māori in the NCEA achievement data, but as European and Māori in the tertiary data, they are counted in the ever-Māori category. If, on the other hand, none of the ethnic group fields in any of the data at any time ever shows Māori as an ethnic group, they are counted as never-Māori. And if the Māori ethnicity has only ever been recorded, they are counted as sole-Māori. These three states are captured in the one variable. One of these variables is created for each of the Māori, Pasifika, European and Asian ethnic groups. A fifth ethnic group in the data was ‘other’. These ‘other’ ethnic students were included in the data and the statistical modelling, but were not specifically reported in the study. Table 1 shows the correspondence between the multiple response and prioritised methods of reporting ethnicity, compared to the method used in this study. Table 2 shows the sizes of these ethnic group categories in the study population.

It had been the intention to compare these ever-ethnic group variables together, enabling the modelling of interactions between them. However, the variables are likely to be correlated⁶. To avoid any problems in the analysis, the ever-ethnic variables were analysed separately, with a different regression run for each of the four ethnic groups considered in the study.

It should be pointed out that the never-, ever- and sole-ethnic method of reporting ethnicity is not a measure of, or a proxy for, the strength of a person’s cultural or ethnic affiliation. These categories simply represent the history of an individual student’s declarations on data capture forms over a period of time, and do not reflect the range of reasons a student might choose one or more particular categories. For example, a person who is regarded as ever-Pasifika may have ties to their culture as strong as, or stronger, than a person regarded as sole-Pasifika, and similarly for people in the other ethnic groups. The measure of ethnicity in this study, as in most administrative data, represents the identification of a person’s ethnicity⁷. It is what a person has said they are, when asked to indicate their ethnicity on a form or in a census. This is distinct from the identity of a person, which is the ethnicity they think they are. Two further facets of identity can be defined: attachment, which indicates to what extent a person can speak the language, knows the customs, and participates in their ethnic group’s cultural activities, for example, and orientation, which is a person’s ethnic identity in a given situation or context (this applies mostly to those people with multiple ethnic identifications). While these other facets of identity may have an influence on educational and other outcomes, they were not able to be measured in this study.

One possible problem with this method of measuring ethnicity in this study is that for students who do not go on to any type of tertiary education, the ethnic categorisation is based only on school achievement records⁸. That is, there is less chance for an ethnicity to change over time and, consequently, less chance for the ever-ethnic category to occur, resulting in the sole-ethnic group being over-represented. Since this latter group does not study at bachelors (or any) level, there is a potential for bias in the results. While this is a potential problem in the study, in practice, no bias was found⁹.

4.2 University entrance

Gaining University Entrance was a significant factor in determining whether a student studied at bachelors level after leaving school. This is not surprising, since most bachelor degree enrolments are at universities and that for those aged under 20 years, UE is the minimum requirement for entrance to degree study. In the present study, the research question hinged on considering the likelihood of studying at bachelors level for people who had the opportunity to do so. It was important therefore to control for whether or not a student had gained UE. In the study population, 90 per cent of students had gained UE.

While 71 per cent of students overall in the study population studied at bachelors level, for those that did gain UE, 77 per cent studied at bachelors level.

There were also differences between ethnic groups. Table 3 shows the proportion of students with UE by ethnic group category. Given that UE is a prerequisite for study at this level, the differences in UE attainment between and within ethnic groups are likely to result in differences in the likelihood of studying at bachelors level. But the statistical modelling controls for levels of achievement and qualifications gained, so that differences in the likelihood of studying at bachelors level between and within ethnic groups cannot be attributed to differences in qualifications or achievement.

4.3 Time off between school and tertiary study

For the students in the population selected for this analysis, most start tertiary study in the year after they leave school. There are some circumstances where a student is enrolled in tertiary study in their last school year. This can happen in a number of ways, but most usually because students undertake one or more tertiary courses as a complement to their school work. For the purposes of this study, a student is considered to go directly to tertiary study if they start that study in the same year, or the year after their last school year.

Whether a student progressed directly to tertiary study, or took some time off before starting, was an important variable in the study. Students who made a direct progression to tertiary study comprised 76 per cent of the study population, while 8 per cent took some time off before starting tertiary study, with a further 16 per cent not indicating any type of tertiary study (see table 4).

Of the students who progressed directly to tertiary study, 86 per cent studied at bachelors level. This compares to 69 per cent for those that took some time off. Overall, 71 per cent of the study population were studying at bachelors level.

This variable also posed some methodological challenges. The variable might have been included in the regression models using the three states (direct, indirect and did-not-progress), but for students who did not progress in the time period used to define the study population, there are no students studying at bachelors (or any) level. To overcome this, the variable was included as a binary variable, which amalgamated the indirect and did-not-progress categories. This is reasonable since there is no way of determining if some of the students in the did-not-progress category might start tertiary study in the following years. In addition, the focus of the report is on students making a direct progression to tertiary study. Further research is needed to look at the characteristics of students moving indirectly to tertiary study.

There is also a difference between and within ethnic groups. Table 5 shows the proportion of students progressing directly to tertiary study after leaving school, by ethnic group and whether or not the student gained UE. Overall, 79 per cent of students progressed directly to tertiary study if they gained UE, compared to 55 per cent of students without UE.

It can be seen that the proportions progressing directly are higher for students who gained UE. Since the proportions of students studying at bachelors level is far higher for students who progress directly to tertiary study, it is important to control for both UE and the timing of the progression to tertiary study if valid comparisons are to be made between students.

4.4 Achievement score

The variable measuring how well a student performed is named the ‘achievement score’ in this report. The achievement score variable has been used in other studies (Ussher 2008, Scott 2008, Earle 2008) and has previously been referred to by the name ‘expected percentile’. This measure of student achievement was developed for analysing NCEA results by Michael Johnston at the New Zealand Qualifications Authority (NZQA). Readers are referred to Ussher (2008) for a more detailed description of this variable.

Most students in the study population have achievement scores in the range 20 to 90. Sole-Pasifika and sole-Māori generally have lower achievement scores, while sole-Asian students have generally higher scores (see table 6).

As will be seen, students with higher achievement scores are more likely to study at bachelors level. Differences in achievement scores between and within ethnic groups will therefore result in differences in the propensity to study at this level. It is important then to control for achievement score in the statistical modelling. It is worth noting however, that there are fewer students with higher achievement scores for the sole- and ever-Māori, and sole- and ever-Pasifika ethnic groups. This increases the size of the confidence limits in the reported results.

The achievement score was included in the regression models as a continuous variable. The logit of the dependent variable was essentially linear against achievement score.

4.5 School decile

The decile of the last school attended was categorised into three groups, corresponding to school deciles of 1-2, 3-8 and 9-10. These categories were used because there were distinct differences in the results for students from schools with deciles 1 and 2, and 9 and 10, but for the others, the results were essentially the same. Using just three groups simplifies the analysis and the presentation of the results with no loss of detail.

The school decile is based on the socio-economic characteristics of the communities from which a school draws its pupils. This means that school decile does not necessarily indicate the socio-economic status of an individual student or their family. This is because most secondary schools draw from diverse communities and hence, most will have at least some high socio-economic students on their rolls. In spite of this, school decile was found to be quite important in explaining student outcomes. However, care must be used in interpreting the findings and in extrapolating the results. In effect, there is a risk that using school decile masks underlying differences in outcomes for different socio-economic groups. However, it is generally regarded that the results for the higher and lower decile ranges are less influenced by this variability, since these schools will have the highest proportion (in lower-decile schools) and the lowest proportion (in higher-decile schools) of lower socio-economic students.

School decile is also likely to be a proxy for a number of school characteristics which are important in determining the likelihood of choosing to study at tertiary level. Thrupp and Lupton (2006) indicate that socio-economic composition affects school processes in numerous ways which would cumulatively boost the academic performance of schools in middle-class settings, and suppress it in low socio-economic settings. This would have a direct bearing on the likelihood of further study, since without the requisite qualifications and standards, study at higher levels is not an option.

Leach and Zepke (2005) cite research which shows that students from higher decile schools have access to more information about tertiary study, and students in these schools develop tastes for the type of training received and occupations held by their, or their peers, parents. Bélanger et al (2009) also cite the positive effects of private (higher decile) schools on student aspirations for further study. While school decile as a proxy for socio-economic status is somewhat compromised by the fact that not all students in a school belong to the socio-economic level as indicated by the school decile rating, certainly every student in a school is exposed to the ethos and expectations of their school.

It is not possible to separate the socio-economic and school factors or to include them individually in the analysis, but it is clear that students from low-decile schools are more likely to leave school with lower levels of attainment, and have less experience learning in a motivated and motivating environment.

4.6 Gender

The gender of a student was determined from the NCEA records.

While there are more female students than male students in the study population (59 per cent are female), the proportion that go on to bachelors-level study is much the same; that is, of the 71 per cent of the population who do go on to bachelors study, 60 per cent are female.

Table 7 shows the observed proportions of students in bachelors-level study by gender for the entire study population.

The table shows that, once a student has achieved UE or NCEA level 3 or higher, there is little difference between males and females in their likelihood of choosing to study at bachelors level. When gender was included in the logistic regression models, it was found to be statistically significant for Māori and Asian ethnic groups, but not for European or Pasifika. Where it was significant, it was of low strength, with females more likely to study at bachelors level, but only very slightly so. The results of the modelling however, show the average likelihood of studying at bachelors level for all students, ignoring gender. This was done for pragmatic reasons, to avoid either reporting both gender’s results for some ethnic groups (which were essentially identical), or choosing to report either the male or female results. Given that gender has very low explanatory power, this does not bias the results of the study.

Ussher (2008) also found that gender was not a strong predictor of whether a student studied at bachelors level, although it was important in the choice to undertake industry training.