Consequences, risk factors, and geography of young people not in education, employment or training (NEET)

Appendix 3 - Data used in risk factor analyses

Area measures

At the local area level, we included youth unemployment rate, deprivation, and urban-rural category. Neighbourhood characteristics can be measured at different scales. We used census output areas (OA) as proxies for neighbourhood. The Carstairs index was used to measure deprivation (Carstairs and Morris, 1990). The Carstairs deprivation index is defined as the sum of four standardised percentage variables from the census: male residents in unemployment, residents in overcrowded households (more than one person per room), residents in households with no car, and residents in lower social classes (partly skilled and unskilled occupations). It was used in the form of quintiles with the least deprived as the reference group. NEET rates at the intermediate zone (N=1000) level for young people were also included.

The local NEET rate was calculated using data on the economic activity of 16-19 year olds at intermediate zone level for all of Scotland. Intermediate zones were selected as the use of more detailed geographies would have meant too few 16-19 year olds in each area, while using larger areas would have resulted in greater variation in the NEET rate within the area. Cut-offs were determined such that the population was divided into four groups of approximately equal size. SLS members are classified as being in an area with a lower, below average, higher than average and higher NEET rate.

School Census data

There are a number of issues with the school data that need to be considered for analysis and interpretation.

Missing data

Firstly, school census data are not available for independent schools. Approximately 5% of pupils attend independent schools per year. If it is assumed that pupils at independent schools have a lower NEET rate then the number of NEETs lost to analysis will be relatively small. In addition, it is more likely that these pupils are not long-term NEET requiring intervention, but are short term NEET, opting to take time out eg for a gap year or to make decisions about their future career\education. In general, the bias caused by not including pupils who attend independent schools should be small. Furthermore, if the independent school pupils either do not differ or display less risky behaviour than pupils at state schools with a similar NEET rate then this would lead to an under estimate of the risk factor.

Secondly, the school census data is recorded for each year, so we have varying amounts of data depending on the age of the individual and the age at which they leave compulsory education. For example, an individual aged 19 at Census 2011 who left school at age 16 may have only one record whereas most individuals aged 16 at Census 2011 will have 4 years of data. This becomes more problematic for those variables that change with age, for example the number of exclusions depends on the number of years of data and the age of the pupil, with exclusions increasing from school stage S1 to S3 and then decreasing. There are also years that are less reliable as the total possible attendance may only be for a part year. This may occur if a pupil decides to leave school part way through a school year after age 16 or if they emigrate or move to an independent school.

Choice of factors for models

Sensitivity analyses were carried out to compare different measures of exclusion: proportion of time excluded, total time excluded and number of times excluded. There was found to be little difference between these measures, possibly because most pupils are not excluded, and most of those who have been, were excluded for a small number of times and days. The simplest measure therefore was included in the model: number of exclusions.

Absence is recorded by reason, and intuitively unauthorised absences, especially truancy should be more closely associated with risk of being NEET than authorised absences such as family holidays. However, differences in data recording between local authorities and schools, as well as changes over time, can lead to sub- categories of absences being unreliable. These analyses include total absences as recommended by Scottish Government. The proportion of time absent is used to take account of the varying periods of time for which individuals have school census data.

There are also issues with the exam results: younger members of the cohort would not have progressed through all the exams that they would sit while some older members of the cohort would have left school before these data were collected. There is no measure of exam results that is sufficiently complete for all ages 16 to 19. Examinations are not always sat at the same age or stage, for example some pupils will sit intermediate exams rather than standard grade exams.

The exam result variable included in the regression model is the number of passes achieved at SQA level 5 or higher by school stage S4. Standard grade 1 or 2 and intermediate 2 pass are deemed to be SQA level 5, whilst Highers are SQA level 6 and advanced Highers SQA level 7. Only a small number of individuals had passed a Higher or advanced Higher by stage S4 (<1%). Very few 19 year olds had this data, consequently the schools census analysis sample is mainly those aged 16-18.

Teenage birth/pregnancy

The vital events data was used to determine whether the SLS member had a teenage pregnancy. The father is only recorded on the birth registration if the parents are married or under certain other conditions. In the years 1996-2001, the proportion of births registered solely in the mother's name to mothers aged less than 20 was between 25.4% and 29.2%. Although we were not be able to identify a sizeable proportion of male SLS members who fathered a child whilst still a teenager, it is likely that this subgroup would be least affected by the birth as they had not attended the registration of the birth or made a declaration of being the father. By extension, the birth is less likely to impact on their decisions regarding education, training and employment.

We have included all births up to the end of 2001 for the analysis of Cohort 3 because an impending birth may impact on NEET status at the time of the Census in April 2001. At the time of the analysis, new births only up to the end of 2010 were available for linking to SLS members. This means that we are missing a small proportion of births before and all those pregnant at the time of the 2011 Census for the analysis of Cohort 4. In addition, the analysis of Cohort 4 is restricted to those who have school census data at school stage at S4. These are mainly those aged 16-18 at the 2011 Census and this group will have recorded fewer births than those aged 16-19. This will lead to an underestimation of the effect of this factor.

We only have data on births, we do not have abortion or miscarriage data due to the sensitivity of these data. It is likely that such events could disrupt a young persons' plans for education, training or employment.