Geographical variation in the prevalence of young people not in education, employment or training (NEET), England: March 2021

• After accounting for age differences among young people aged 16 to 24 years on Census Day, 21 March 2021, the highest percentage of young people who were not in education, employment or training (NEET) for both males (22.2%) and females (20.4%) was in Blackpool; the lowest percentages were in Bath and North East Somerset for males (6.1%), and in York for females (4.6%).

• Our analysis highlights the different impacts of young people's individual characteristics and place characteristics on geographical variation in NEET prevalence; after accounting for individual characteristics, the geographical variation in the prevalence of young people who were NEET between Upper Tier Local Authorities (UTLAs) reduced for both females and males (variance reductions of 78.9% and 60.4%, respectively).

• This reduction in geographical variation was largely associated with differences between young people's living arrangements and whether they were living with dependent children; adjusting only for ethnicity, English as a main language, disability status, and carer status resulted in a smaller reduction in geographical variance for both females (21.7%) and males (22.6%).

• Place characteristics, including the local economy and labour market, healthcare access, education and skills provision, and deprivation, had a larger impact in explaining geographical variation for males (a further variance reduction of 36.1%) compared with females (a further variance increase of 11.0%).

• When we exclude UTLAs in London, further adjustment for differences in place characteristics leads to a reduction in geographical variance for both females (29.0%) and males (66.4%).

This work has been commissioned from the Office for National Statistics by the Department for Work and Pensions.

Age standardised percentages

Age-standardised percentages allow us to make comparisons between populations that contain different proportions of people of different ages.

In this analysis, model-based estimates of age-standardised young people not in education, employment or training (NEET) percentages at the upper tier local authority (UTLA) level were produced using estimated marginal means. Age was modelled in discrete age bands from 16 to 24 years. The reported age-standardised percentages represent the estimated percentage if each UTLA had the same age distribution as the total young person population.

Not in education, employment or training

NEET refers to young people aged 16 to 24 years who are not in education or training and are either unemployed or economically inactive. In this analysis, NEET status was derived from Census 2021 responses about full-time education and economic activity. For more information on how the measure was derived, see our Characteristics of young people not in education, employment, or training (NEET), England: March 2021 bulletin.

Upper tier local authority

An upper tier local authority is the top-level administrative body in a two-tier local government system. It covers a broad geographic area, such as an entire county, and handles major strategic and people-facing public services. In contrast, lower tier local authorities handle local, place-based community services.

Unitary authorities, metropolitan districts, and London boroughs also function as upper-tier bodies because they combine all services under one roof. In Census 2021, there were 152 upper tier local authorities in England. These were made up of 59 unitary authorities, 36 metropolitan districts, 33 London boroughs (including City of London), and 24 counties. For more information, please see our Area type definitions Census 2021 dictionary entry.

Variance

We calculate the variance in model-estimated NEET prevalence to assess the differences between UTLAs, and how each of the modelling adjustments affect this. The variance is a statistical measure that reflects the extent to which observations differ from the average value. See Section 4: Data sources and quality for more information.

Data and inclusion criteria

We derived our study's population of young people from people who responded to the Census 2021. Individuals were included if they were:

• usually resident in England in households or university accommodation
• aged 16 to 24 years on Census Day, 21 March 2021

This resulted in a total study population of 5,554,065 young people, with 700,160 being flagged as not in education, employment or training (NEET) on Census Day. Further information on how we identified young people who were NEET can be found in our Characteristics of young people not in education, employment or training (NEET), England: March 2021 bulletin.

Full counts of the study population at each stage of data cleaning are available in our accompanying dataset.

Individual and place characteristics

The census-based individual characteristics used in this analysis are:

• age
• sex
• ethnic group
• English as a main language
• disability status
• carer status
• living with dependent children
• living with family

Individual characteristics capture differences in the composition of young people across areas. Adjusting for these characteristics helps identify how much geographical variation in NEET rates reflects differences in who lives in each area, rather than the characteristics of the areas themselves. Some individual characteristics and some area characteristics might be closely related, meaning that we cannot separate out their effects.

Place characteristics capture differences in local economic conditions, labour market opportunities, education, training, skills provision, healthcare access, and deprivation. Together, these domains capture differences in the socioeconomic environment, local opportunities, and service provision that may influence young people's likelihood of being NEET. Adjusting for these factors helps us assess the extent to which geographical variation in NEET rates reflect the characteristics of places, rather than the people who live in them.

See our accompanying dataset for more information on the data and measures included in our individual characteristics and local area place domains.

Statistical analyses

Our analysis aims to describe how NEET prevalence varied across upper tier local authorities (UTLAs) on Census Day. It also aims to assess how much of this geographical variation was associated with differences in the characteristics of young people and the places where they lived.

We fitted a series of Poisson regression models, separately for males and females, and compared the remaining variation between UTLAs as characteristics were added at each stage.

To fit statistical models to our data, we grouped Cornwall and the Isles of Scilly, and City of London and Hackney, because of low sample sizes.

Three modelling stages were used to estimate variation in NEET prevalence across UTLAs, which sequentially accounted for differences in young people and place characteristics.

Stage 1

Baseline model adjusting for age only.

Stage 2

Accounting for differences in age, plus the following individual characteristics:

• ethnicity
• English as a main language
• disability status
• carer status
• living arrangement
• living with dependent children

Stage 3

Accounting for differences in age, individual characteristics (Stage 2), and place-level characteristics, using the following sub-models:

• education, training and skills provision domain
• healthcare access domain
• local economy and labour market domain
• deprivation domain
• all place characteristics containing all four place domains

We used estimated marginal means to estimate age-standardised UTLA NEET percentages (Stage 1 modelling). We also estimated NEET incidence rate ratios (IRRs) for all modelling stages.

Estimated marginal means (EMMs) are statistical averages calculated for specific groups or levels of a variable (such as UTLAs), adjusted to hold other variables (such as age) constant. In this analysis, EMMs estimate what the UTLA NEET percentages would be if each UTLA had the same young person age distribution as the total young person study population.

An IRR is a relative measure used to compare the incidence rates or prevalence of events between two groups. In this study, we calculated it by dividing the model-estimated prevalence of NEET in each UTLA by the model-estimated prevalence of NEET in the reference UTLA (here chosen to be Bath and North East Somerset, the UTLA with the lowest percentage NEET after adjusting for age).

We use the variance of the IRRs on the log relative rate scale to assess the dispersion in NEET prevalence between UTLAs, and how each of the modelling adjustments affects this. A reduction in variance between model adjustments indicates that the inequality between areas has decreased, and that the additional variables included in the modelling adjustment are associated with this part of the observed geographical variation. Likewise, an increase in variance suggests that the inequality between areas has increased.

For more information about the data, model specifications, and methods and metrics, including the full set of modelled age-standardised percentages, IRRs, between-model variance reduction metrics, and data and sample summaries, please see our accompanying dataset.

The UK Statistics Authority's National Statistician's Data Ethics Advisory Committee provided ethical approval for this work.

Quality

Census 2021 falls within the coronavirus (COVID-19) pandemic period. We advise users to consider the impact of the pandemic when interpreting the estimates produced in this analysis.

Some young people in our study population will be resident in the geography where they grew up, while others might have moved from their home regions for work, study or other reasons. Place level measures, such as deprivation and education, are assigned at the location where the young person is resident on census day. Relatedly, some UTLAs, through their association with university cities and towns, will have a larger influx of students than others, which might be masking higher levels of NEET prevalence in the populations of young people who grew up in these areas. We advise users to consider the impact of these factors when interpreting this analysis.

Our model-adjusted NEET estimates account for a range of individual and place characteristics associated with being NEET, enabling us to compare UTLAs on a more like-for-like basis. However, there will be other factors besides those included in our models that are also related to being NEET, such as the socioeconomic background of the young person and local level interventions. In addition, many of the characteristics included and not included will be related to each other and measure (or proxy) similar things. Therefore, our analysis cannot be used to infer cause-and-effect relationships between characteristics and NEET status.

The place characteristics included in our analysis were constructed at the lowest level of geography available, or to the lowest level of geography appropriate. Some of our measures, especially in the education, training and skills provision domain, might have explained more geographical variance if they had been available at lower-level geographies.