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

Chapter 2 Methods

Data sources

Scotland's Census

Scotland's Census is carried out every ten years and is designed to cover every resident in the country. A large number of demographic, social and economic questions are included in the census form. A count of NEET young people can be obtained via the derived variable 'economic activity'. The publication of census data at the local area level also permits investigation of NEET prevalence by area deprivation and urban rural categories.

The Scottish Longitudinal Study

The Scottish Longitudinal Study (SLS) is an anonymous dataset. It links information from the 1991, 2001, and 2011 censuses. Anyone whose birthdate falls on one of the 20 birthdates chosen by SLS is included in the sample. The sample members are updated through birth and migration. The SLS covers just over a 5% sample of the Scottish population, and includes about 14,000 members aged 16-19 years old.

One unique feature of SLS is that it links to a wide range of administrative data such as vital events (e.g. birth, death), hospital discharges, and prescribing data. Hospital discharge data include information on inpatients and day cases from NHS hospitals, as well as people admitted to specialist mental health facilities. The prescribing data include information on prescription of antidepressants or antianxiety medications. In addition, the SLS includes school census data which include information on free school meals, exclusions, absences and educational attainment.

NEET definition

The census requires respondents aged 16 and over to answer questions on economic activity in the week before the census. The responses to these questions are used to derive the variable 'economic activity' which we have used to identify NEET individuals. This provides a snapshot definition of NEETs. A NEET individual is thus defined as one who, at the time of the census, is aged between 16 and 19, either unemployed, seeking work and ready to start within 2 weeks, or economically inactive due to looking after home/family, permanently sick/disabled, or other reasons.

As the SLS has data linked from the 1991 to the 2011 census, we are able to look at different cohorts: those that were of age 16-19 at each of the three censuses. We can examine the risk factors of being NEET for those of age 16-19 at the 2001 and 2011 censuses. We can examine 20 and 10 year outcomes for those that were of age 16-19 at the 1991 and 2001 censuses respectively. Thus we can repeat analyses on multiple cohorts and compare results between cohorts.

Sample specifications

Question 1: To what extent does NEET status affect outcomes in later life?

There are two samples available to answer this question:

Cohort 1: SLS members who were aged 16-19 in 2001 and followed up to 2011 when they were 26-29;

Cohort 2: SLS members who were aged 16-19 in 1991 and followed up to 2011 when they were 36-39.

Cohort 1 was used to explore whether being NEET in 2001 was related to higher risks of negative labour market outcomes and poor health in the 10 years period up to 2011 (age 26-29).

Cohort 2 was used to examine the same outcomes in the 20 year follow-up period from 1991 to 2011. Both cohorts were linked to hospital records and prescribing data.

There were 13,218 SLS members who were aged 16-19 in 2001. Between the 2001 census day and the 2011 census day 1,181 people moved out of Scotland and 74 people died. The 1991 16-19 cohort included 14,567 SLS members. 1,234 people left Scotland and 213 died between 1991 and 2001 censuses. About 1,285 people who were present in 2001 were not present at the 2011 census due to unknown reasons. Similarly 1,397 SLS members in 1991 were not present in 2011 for unknown reasons. There are also missing values for some census variables thus the analytical sample is smaller than the full sample.

In reporting the consequences of NEET status we first present and discuss results from Cohort 1 and then results from Cohort 2.

Question 2: What individual, family, educational and geographical factors are related to the risk of becoming NEET?

There are two samples available to answer this question:

Cohort 3: SLS members aged 6-9 in 1991 followed up to 2001 when they were 16-19;

Cohort 4: SLS members aged 6-9 in 2001 followed up to 2011 when they were 16-19.

In total there were 10,206 SLS members in 1991 who were aged 6-9 and present in the 2001 census (Cohort 3). Of these 10,195 lived in residential properties and 11 in communal establishments. These 11 have been excluded as they may be dissimilar to the rest of the cohort and they have no data for some variables being investigated such as tenure.

Cohort 4 includes a total of 11,615 SLS members. Additional datasets were available for the analysis of Cohort 4 including data from the 2007-2010 school censuses such as whether an individual was registered for free school meals, and attendance and exclusion records. More details on the school census data are given in Appendix 3. Of this cohort, 10,445 (90%) cases had 2001 census data, 9,732 (84%) cases had school census data and 8,815 (76%) cases had both sources of data. A small number had neither 2001 census nor school census data. The small number resident in communal establishments were again excluded.

Question 3: Are there geographical patterns of NEETs? And, have these patterns changed over time?

1991, 2001 and 2011 censuses were used to describe the geographical distribution of NEET over two decades. Geographies included local government authorities, area deprivation and urban rural categories.

Statistical Methods

Logistic regression was used to explore whether NEET status is independently associated with future economic and health outcomes, and also to explore what individual, household and area level factors are associated with the risk of becoming NEET.

Logistic regression is a statistical technique that allows you to investigate the relationship between an outcome variable (e.g. being NEET or not) and various explanatory variables. The analysis identifies which of the explanatory variables is significantly and independently related to the outcome variable. For example, it could be that the chance of becoming NEET (the binary outcome variable) increases with poor health (explanatory variable). Any other variable that also affects the chance of being NEET and is related to poor health, such as family background, should be included in the analysis. Only when poor health and family background are considered together can the independent effect of poor health (or family background) be isolated.

A range of negative outcomes such as unemployment, low status occupation, physical illness, mental illness, or drug misuse in later life were examined for Cohorts 1 and 2 in the study.

In order to assess whether NEET status had an impact that is independent of other socio-economic factors, a range of explanatory variables was included in the model. Explanatory variables were selected on the basis of the literature review. Previous research was used to identify which factors might influence subsequent outcomes. For example, including gender allowed us to explore whether there was a difference between men and women in their probability of experiencing a negative outcome.

The analyses of Cohort 1 (2001 cohort) adjusted for gender, age, educational attainment, Carstairs deprivation, limiting long-term illness and living in a council area NEET 'hotspot' (see Appendix 1 for further details), so it was possible to assess whether being NEET in 2001 had an independent effect in relation to outcomes in 2011 over and above these demographic and socio-economic characteristics.

For the analyses of Cohort 2 (1991 cohort) we adjusted for the demographic and socio-economic factors mentioned above. In place of the binary NEET/not NEET variable, we included a variable indicating changes between NEET status in 1991 and subsequent economic activity in 2001 to predict the probability of the outcome by 2011. Thus we are able to examine whether being disengaged from employment and education in both 1991 and 2001 had a cumulative, negative effect on future employment or health. We are also able to explore whether moving from 1991 NEET status into employment in 2001, or moving from non-NEET status in 1991 into economically inactive status in 2001, had any effect on later life chances, by 2011.

In the risk factor analysis, we explored the extent to which personal attributes, family background, neighbourhood deprivation and local labour market characteristics are related to NEET status.

Potential risk factors considered for Cohort 3 include individual and family variables from the 1991 Census, Carstairs 1991 quintile, teenage pregnancy, local NEET rate in 2001, and unpaid carer and highest educational qualification from the 2001 Census (see Appendix 4). Potential risk factors for Cohort 4 include birth weight, individual and family variables from the 2001 Census, Carstairs 2001 quintile, teenage birth, school census variables, prescription data, local NEET rate in 2011 and unpaid carer from the 2011 Census (see Appendix 5). Explanatory variables included whether an individual had a limiting long-term illness, as well as household factors including the economic activity and health status of household members and housing tenure. These variables were considered on the basis of previous literature or theory.

The main analysis for Cohort 4 concentrated on school census data. This choice was driven by the fact that this information is known to teachers and careers guidance officers and can therefore be used to identify at risk young people, whereas information relating to an individual's childhood experiences may be unknown. We used the examination results obtained by stage S4 because young people are aged 15-16 at this time and we wish to predict becoming NEET at ages 16-19 (see Appendix 3 for details). A second analysis which included data derived from the 2001 Census was therefore carried out in order to assess whether the census variables were important predictors of being NEET in addition to school census variables.

The model for Cohort 4 was used to develop a risk score to identify the group of young at high risk of becoming NEET.

All the models adjusted for age. As the age of leaving compulsory school education is 16, the NEET rate increases with age as an increasing number of young people leave school. The risk of being NEET is therefore associated with age.

All potential explanatory variables were tested using a manual stepwise procedure, and in the final model only those variables that were significant at the level of 0.05 were included.

Statistical analysis was conducted separately for males and females because of expected differences. For example, it might be expected that teenage birth is likely to be a far more important risk factor for females than for males.

Presentation of results

Model results are presented as odds ratios. An odds ratio is a measure of effect size, measuring the strength of association between two variables. An odds ratio above 1 indicates a positive relationship where an increase in the independent variable is associated with an increased likelihood of the outcome. An odds ratio below 1 indicates a negative relationship where a decrease in the likelihood of the outcome is associated with an increase in the independent variable. Odds ratios and significance levels are presented for the NEET status groups in the consequences analyses and for each significant variable in the risk factor analyses in the tables of this report.

More detail is available in Appendix 2, which contains tables reporting odds ratios and their confidence intervals.

Strengths and Limitations of the analysis

There are at least four advantages of using the SLS:

The SLS is a rich data source which allows research on various outcomes, including those from census and other administrative sources. The sample size and design mean that we can repeat analyses on multiple cohorts and compare results over cohorts.

The prospective, longitudinal design of the SLS enables the analysis of the temporal sequence of lifetime factors before the occurrence of outcomes and ensures the direction of influence from factors to the outcome. For example using the SLS allowed us to identify risk factors leading to NEET status.

Also longitudinal data allows analysis of changes over time and how these changes are related to other factors.

Furthermore, the SLS includes data from 1991 to 2011 and we can analyse long-term effects of NEET experiences.

However, there are some limitations of using the SLS for research on NEETs. The SLS is based on the census which is carried out every ten years. Therefore, it is not possible to follow the cohort in the period between censuses. For example, we cannot examine changes in NEET status or economic activity on a monthly or yearly basis between censuses.

The census definition of NEET is a snapshot measure. However, because many young people take temporary jobs and change their status frequently, some commentators have argued that it is better to define NEETs as those who have been out of education and employment for three or six months continuously (Bynner and Parsons 2002). This information is not available in the census. Moreover, some studies have shown that there was no significant difference between the snap-shot definition and the definition using the continuous measure because these two types of NEET are more similar to each other in their characteristics compared to those non-NEETs (Furlong, 2006).