Economic inactivity interdependencies in England

This report has been produced by the Local Government Association based on the structure and methodology of an original report by the Centre for Economics and Business Research (Cebr), an independent consultancy established in 1992. The report contains structure, methodology and theoretical content contributed by Cebr, and as such does not necessarily reflect the views of the Local Government Association.

Research purpose

Following the COVID-19 pandemic, England and the UK in general, continues to experience a curious mix of tight labour markets – high job vacancies and relatively low unemployment – and high numbers of people who have left the labour force altogether and are not looking for work (known as economic inactivity or hidden unemployment). This issue has gained political traction, with the 2023 Spring budget and Autumn Statement containing measures aimed at addressing rising economic inactivity and encouraging people to enter or re-enter the labour force.

Though some of this rise may be driven by relatively benign factors – for example older workers deciding to retire early or transition into voluntary work because they want to and can afford to – in other cases it may be driven by poor health, caring responsibilities, and so on, and not the choice of the individuals concerned.

Ultimately, high numbers of ‘economically inactive’ people has a negative impact on individuals’ life chances, the local economy and on long-term economic growth as it limits the pool of labour and skills available to employers.

It also poses fiscal problems – especially in the context of an ageing population – as it reduces the ratio of those working and paying tax to dependents.

Cebr was commissioned by the LGA to conduct analysis and research which sought to address two questions:

1. What are the most important drivers or indicators of high levels of economic inactivity across England, both overall and by reason for inactivity?
    
2. How do trends in economic inactivity vary across England? This may be important in considering how the best approaches to reducing it vary from place to place.

Regression analysis and findings

Modelling work aiming to answer these research questions was originally undertaken by the Centre for Economics and Business Research in mid-2023. This research was then updated and reproduced by the LGA in a manner consistent with the original methodology. It is the findings from the updated research that are presented in this publication.

We brought together data on a wide range of socio-economic indicators at the level of single-tier and district councils across England, spanning the following themes:

• employment and labour market
• deprivation and poverty
• housing
• health and wellbeing
• financial vulnerability.

We then conducted a series of regressions to identify the main drivers of economic inactivity among 16 to 64-year-olds. Our key findings show that:

• Various indicators relating to poor health were related to higher levels of overall economic inactivity – this may be due to these individuals being unable to work and/or others having to stay out of the labour force in order to care for them. Research by the IPPR Commission on Health and Prosperity reveals that people living in deprived areas are more likely to have poor health, become economically inactive and have lower life expectancy. 
• Home ownership with a mortgage is associated with lower inactivity, as is a high proportion of over 65s claiming Pension Credit – likely because older people with lower income and/or greater economic liabilities may be required to delay retirement.

Further regressions with sub-categories of economic inactivity (for those aged 16+, from the 2021 Census) were also conducted:

• Higher job density and more residents holding Level 2 or higher qualifications are associated with lower levels of retirement; it may be that people will carry on in paid employment for longer where jobs are more easily available, and they are equipped to do non-manual work.
• Where a lower proportion of the population hold Level 2 or higher and where a higher proportion provide 20 or more hours a week of unpaid care (from the Census) inactivity due to people looking after home/family is higher. This may be explained by a lack of job options for those without skills.
• As one might expect, various poor health indicators are also major drivers of inactivity due to long-term disability or sickness.

Trends across England

In line with the trend of increasing economic inactivity across the country, inactivity rose in 166 local authorities included in our main regression and fell in 145. These were further subdivided into typologies according to overall level and composition of inactivity, described below.

Within all regions there is a mixture of areas with high, medium, and low levels of economic inactivity. Nevertheless, there are some clear geographic trends.

Areas of high inactivity, and especially those in which it has risen since 2019 concentrated in and around northern cities like Manchester, Liverpool, and Hull, and are also quite prevalent in coastal areas. This makes sense in the context of long-term declines in traditional manufacturing and seaside tourism industries.

Areas of low inactivity are concentrated in London’s commuter belt and the M4 Corridor. In general, the areas of lowest inactivity are often rural and/or suburban, though many southern cities like Bristol, Norwich, and Cambridge all record low levels too.

Therefore, this demonstrates the stark regional inequalities, and the scale of the challenge. It also shows, however, that areas in need of support are by no means exclusive to the North, and that within regions the extent of issues related to economic inactivity can vary widely from place to place.

Data sources

Cebr has utilised publicly available datasets and identified five main themes within which the key indicators have been established. The data has been collected and consolidated into one rich dataset from which the analysis has been conducted.

The timing of this research means that the 2021 Census is the best source of data for several variables given that it covers the whole UK population and is very recent. Any future updates would need to draw on alternative sources based on regular research like the Annual Population Survey. Moreover, Census data is heavily influenced by COVID-19; whilst this is no bad thing given its relevance to the recent increase in economic inactivity, in the future it will become less salient.

Employment and labour market

Economic activity data has been obtained from two sources. First, the Annual Population Survey 2023 has the most up-to-date data by local authority. However, as it is a sample survey, data on inactivity by reason is incomplete. Second, to supplement the APS data, we have included economic inactivity data from the Census 2021, and this is broken down by category of economic inactivity. In addition, claimant count and highest level of qualification data (e.g., no qualification, NVQ4+) from the Census 2021 have also been utilised. 

The number of claims made from the Coronavirus Job Retention Scheme provided from HMRC was also included in the dataset. This data illustrates the scale of employment that was most affected by COVID-19 and the resulting lockdowns. In addition, to test the possibility that the structure of the economy and employment sector may be affecting inactivity, the Business Register and Employment Survey (BRES) provides recent data on UK employment by sector. The BRES data on the split between private and public sector employment was also included. 

To understand how the local economy and business base affects economic inactivity several indicators were included. From the ONS data, both Gross Value-Added local estimates and job density, which measures numbers of jobs per resident aged 16 to 64, were incorporated into the larger dataset. Business Demography 2021 provides estimates on the number of UK business births and deaths at a local authority level which have been included in our dataset. Finally, the Annual Survey of Hours and Earnings (ASHE) measures the percentage of employees earning below the Living Wage Foundations rates at a local authority level which has been included to get an understanding of job quality at a local level.

Deprivation and poverty

We have included a range of indicators from different data sources to highlight this theme. Firstly, the 2019 Index of Multiple Deprivation (IMD) from the Ministry of Housing, Communities and Local Government. This provides data on a range of different domains of deprivation which are combined and weighted to calculate the IMD. Secondly, to supplement the IMD data, the Census 2021 data on the dimensions of deprivation based on four household characteristics – education, employment, health and housing – was integrated. 

Thirdly, the Local Authority Indicator data from the Office of National Statistics (ONS) was incorporated into the dataset. Deprivation and poverty statistics were extracted and inputted into the dataset. Finally, the Cost-of-Living Vulnerability Index was included to complement the IMD data. This index goes further with respect to the cost-of-living crisis by focusing on indicators of poverty that correspond with the specific cost pressures associated with it – such as food and fuel poverty.

LG Inform data was also utilised for a number of indicators such as, the number of individuals liable to UK income tax through pension income, number of children living in relative low income and the proportion of individuals over 65 claiming pension credit.

Housing

Under this theme some key indicators such as the number of people claiming housing benefits and transport accessibility indicators, which could be driving economic inactivity, were included.

The Department of Work and Pensions commissions statistics on housing benefit claims yearly. In addition, data on methods used to travel to work from the Census 2021 could shed some light on the transport accessibility indicators which could impact economic inactivity at a local level. The Local Authority Indicator also publishes data on transport links to employment for 2022. This has supplemented the 2021 Census data.

The ONS also provides data on dwellings by their tenure – outright ownership, ownership with a mortgage, private rent, and social rent. Census 2021 data on main language, ethnicity and lone parent single family households was also included.Under this theme some key indicators such as the number of people claiming housing benefits and transport accessibility indicators, which could be driving economic inactivity, were included.

The Department of Work and Pensions commissions statistics on housing benefit claims yearly. In addition, data on methods used to travel to work from the Census 2021 could shed some light on the transport accessibility indicators which could impact economic inactivity at a local level. The Local Authority Indicator also publishes data on transport links to employment for 2022. This has supplemented the 2021 Census data. 

Health and wellbeing

Poor health can be a major barrier to entering employment. This has been intensified by the pandemic, which has resulted in long Covid keeping more people out of the workforce due to health issues. However, local authority data on long Covid is not available. We therefore proxied this with historic ONS data on infections. Also, we have considered the prevalence of unpaid care, which could be another key indicator driving economic inactivity, using Census 2021 data. Linking the provision of unpaid care, economic activity and other indicators will provide insights into which local authorities need the most support. In addition, government-funded early education and childcare for children aged two to four years from the ONS was included to understand how the prevalence of childcare demands might affect economic inactivity. The Census 2021 data was also utilised to provide estimates that classify usual residents by long-term health problems or disabilities and by the state of their general health in 5 categories from very good to very bad. 

The Local Authority Indicator from the ONS provides data on key statistics for physical and mental health. Data from the Office of Health Improvement and Disparities (OHID) gives life expectancy figures, segmented by male and female and at birth or at 65. In addition, the OHID gives data on inequality in life expectancy broken down by gender and at birth or at 65. This data shows inequalities within local areas, enabling a focus on the deprivation that exists everywhere at small area level. We had also hoped to include healthy life expectancy, which captures how long a person can expect to be in good health – however these figures were only available for upper-tier local authorities.

LG Inform data was also used for a range of health indicators including PIP claims, loneliness, obesity, or physically inactive adults, under 75 mortality rates, disability living allowance and prevalence of long-term musculoskeletal conditions.

Financial Vulnerability

The Good Credit Index is based on three sub-indices measuring different aspects of credit which were found to be important based on a literature review. These three strands are: the credit environment, credit scores; and credit need. The overall Good Credit Index was created by summing these three sub-indices, with an equal weighting given to each. We hoped to include other indicators to understand financial vulnerability including, household debt and households below 60 per cent of median earnings. However, these figures are only available at shire and unitary local authorities and therefore, cannot be included for this analysis.

Regression methodology

What is regression?

Regression modelling aims to explain differences in a factor of interest – in this case, economic inactivity – using a selection of other factors that are theorised to be associated with it. These associations may represent causal relationships, although regression analysis is itself unable to prove causality.

In addition, it should be borne in mind that a regression model only explains a proportion of the variation in the factor of interest, leaving a certain degree of variation unexplained. This remaining variation could be due to additional factors not included, random fluctuations or unique local circumstances, or a combination of all three. As such, this analysis does not imply that the metrics identified are the only ones which play a role in economic inactivity.

One further caveat is that factors are generally interrelated, and the relationships between them can be complex and multilayered. For example, if long-term musculoskeletal conditions are related to economic inactivity, the actual causes of these conditions are the true driver of the reasons for inactivity, and this causality is mediated through the influence of the symptoms and diagnosis of these conditions. It is never possible for an analysis to capture all, or the deepest, of the drivers of a particular factor of interest, and some drivers can work together, or work differently under different circumstances.

The first step in our modelling is to develop a rich dataset with all the indicators that we believe may affect economic inactivity. Combining datasets from various sources has its challenges. Therefore, a key aspect of the methodology was to collate them to develop a robust view of all the indicators we deem may affect economic inactivity that are not provided in one single data source.

To ensure that data was comparable across local authorities, in some cases figures had to be taken as a proportion of the population or a subset thereof.  For example, data on children in low-income households was available as a raw number. Using this would have told us about the size of the local authority and the proportion of its population who were children, as well as about the phenomenon of interest, namely proportion of children in low-income households. The figures were therefore divided by number of under 18s in each local authority. The datasets were cleaned and merged to produce a single dataset of all the indicators for each local authority. 

Once the single dataset was completed, we ran bivariate regressions. Bivariate analysis helps understand and establish the strength of the relationship between the dependent and independent variables. The two variables are frequently denoted as X and Y, with one being an independent variable (or explanatory variable), while the other is a dependent variable (or outcome variable).

The underlying idea is to quantify the relationships between each independent variable and dependent variable and includes testing simple hypotheses, particularly of association and causality. This can help in identifying a subset of variables that are more important and also know about the important levels of particular feature values. 

From this a forward stepwise regression was performed which allows us to select important variables to get an easily interpretable model. This involves testing each variable as it is added to the model, then keeps those that are deemed most statistically significant.

The final regression includes those independent variables which:

• are statistically significant in driving economic inactivity according to our results, ideally to a high degree of confidence
• have a non-trivial impact on economic inactivity, whether positive or negative
• affect economic inactivity as implied by the regression through a plausible economic transmission mechanism.

This process was repeated for the key sub-categories of economic inactivity: retired, looking after home/family, and long-term sick or disabled. Running these models allows us to have more granular detail on the underlying causes of different economic inactivity types at the local level, and how the drivers of different causes vary.

It should be noted that although linear regression analysis is a reliable method, omitted variable bias cannot be completely ruled out as it is not possibly to include all relevant variable in the model. Omitted variable bias occurs when a statistical model fails to include one or more relevant variables. This should be taken into account when examining the results below. 

In total, 311 local authorities were included in our regressions – all of those in England excluding the City of London, Isles of Scilly, and certain other authorities which were outliers and/or did not provide sufficient data.

Overall Economic Inactivity

The regression results for overall economic inactivity are summarised in Table 1 below.

One immediately obvious point is that several of the identified relevant indicators relate to health. The rates of long-term musculoskeletal conditions and health conditions associated with eligibility for Disability Living Allowance among adults are all associated with higher economic inactivity. 

Though there may be an element of reverse causality with the long-term musculoskeletal indicator, it seems highly plausible that health problems will drive economic inactivity by preventing people from being willing or able to work, or by limiting the range of jobs they would be able to carry out. In addition, those struggling with musculoskeletal conditions are less likely to be active but have the most to gain from the right support. Providing the adequate support to this section of the population could bring these people back into workforce.

Though Disability Living Allowance (DLA) is being partly replaced by the Personal Independence Payment (PIP), it is still claimed by many people and serves as a good indicator of the prevalence of disabilities that limit mobility and generate care needs. This includes mental as well as physical health conditions, which helps to explain why no indicators explicitly linked to mental health appear in our regression. There is an intuitive link, in that conditions like anxiety and depression may prevent a person from being able to work. The prevalence of these may be captured in other indicators though, such as DLA claims (and perhaps physical inactivity, which may be a result of poor mental health).

High rates of property ownership while paying off a mortgage are associated with lower inactivity. To some extent, this may reflect that those who are economically active are more likely to be able to afford property and qualify for a mortgage in the first place. The need to pay off a mortgage could also act as a driver for people to enter and remain in employment to cover payments.

The rate at which those aged 65+ were claiming Pension Credit was negatively associated with economic inactivity. Though the data in the dependent variable for this regression relates to those aged 16 to 64 only, this relationship makes sense. If a relatively high proportion of those over State Pension age are on low incomes, one would expect that a high proportion of those just below it are as well – and therefore early retirement is less likely to be an option.

Table 1: Regression result, economic inactivity overall (2023 Q3)

Economic inactivity by subcategory

The 2021 Census provides a breakdown of economic inactivity by subcategories; retired, looking after family/home, long-term sick/disabled, student, and other. Separate regressions from our dataset have been carried out on the first three of these, refined in the same way as the overall economic inactivity model. It should be stressed that Census inactivity data by category is for all over-16s, not just those aged 16 to 64 as in the data used for the overall regression.

Table 2 summarises these regressions and shows that, while there is some overlap in significant independent variables between these regressions, and between some of them and the overall regression, for each subcategory of economic inactivity there are some distinct variables of significance.

Retired

The causes of retirement are arguably harder to discern than other types of economic inactivity, as is the extent to which retirement is a positive thing for the individuals concerned. Of course, people hope and expect to retire at the end of their working life and may do so (possibly early) for the positive reason that they have a sufficient pension pot to do so – it is unlikely that preventing this will be a goal of local or national government. For others, retirement may be forced earlier than they would like – e.g. by poor health, caring responsibilities, or lack of suitable employment options.

Moreover, economic activity statistics only capture those in or seeking paid employment. Many retirees devote significant proportions of their time to volunteering, which can have social and economic benefits for others – though of course there may be some retiree volunteers who would prefer to be in paid work were the option there.

• Rate of long-term musculoskeletal issues is significant in the proportion of the population who are retired. There is the possibility of some reverse causality here: areas with lots of older people, and therefore retired people, might be expected to also contain more people with these conditions, some of which are more likely to develop later in life. That said, it is also very possible that developing these conditions could force retirement earlier than people would otherwise have chosen.
• Provision of high levels (20 or more hours) of unpaid care is also associated with higher levels of retirement. Again, reverse causality is possible as very elderly people are more likely to have partners who need care due to age-related conditions. If these conditions develop during working life, however, they may force early retirement – both for the person affected directly and for their partner. Duties towards unpaid carers are woven through the Care Act 2014. With a clear focus on promoting carers’ wellbeing and taking account of the impact caring has on all aspects of their lives including work.
• Where job density is higher, we would expect those who want to remain in work rather than retiring will find it easier to do so; conversely, where jobs are less abundant workers may find that they have to retire earlier than they would like. This explains the negative relationship found by the regressions.
• The positive association between longer life expectancy and retirement rates is fairly logical – where residents live longer, more of them will be at ages (e.g., in their late 70s or 80s) by which they will almost certainly have retired.
• A higher share of lone parent families is associated with slightly lower rates of retirement. One possibility is that these families have a lower degree of financial security, necessitating delayed retirement.

Looking after home or family

• The link with share of the population holding Level 2 or above qualifications is negative, so enabling those without qualifications to change this may be a route to reducing inactivity. For those with good job options and earning potential, it is more likely to be financially viable to pay for childcare or elderly care rather than providing it oneself.
• Whilst lone parent families were associated with lower rates of retirement, they were strongly associated with higher rates of economic inactivity from looking after home or family. This makes sense, as being a single parent would often be associated with an increased burden of work caring for home or family.

Long-term sick or disabled

• Possession of Level 2 or above qualifications was related to lower levels of inactivity due to sickness or disability. This could reflect the role of qualifications in helping individuals access roles which accommodate any disabilities or health conditions they may have. Conversely, it could reflect barriers to educational attainment faced by some people on account of their disabilities.
• Physical inactivity was a significant driver of inactivity due to being long-term sick or disabled, likely due to its association with poor health.
• Lengthy waiting lists and access to healthcare provision are also a contributing factor to economic inactivity as highlighted by NIESR research.

Table 2: Subcategories of economic inactivity regression results (2021)

Examination of local authority data shows, unsurprisingly, a great deal of variation in economic inactivity outcomes and their drivers. To help make sense of this variation, we categorise the English local authorities into typologies according to:

• overall level of economic inactivity today (top, middle, or bottom third)
• how economic inactivity has changed since 2019 (increased or decreased).

This results in an initial set of six typologies defined by these characteristics.

• H1: High, increased (84 authorities)
• H2: High, decreased (18 authorities)
• M1: Moderate, increased (55 authorities)
• M2: Moderate, decreased (52 authorities)
• L1: Low, increased (27 authorities)
• L2: Low, decreased (75 authorities)

As one would expect, groups H1 and L2 are the largest and H2 and L1 the smallest – where inactivity has risen it is more likely to be high and vice-versa. Overall, inactivity has risen in 166 authorities shown and fallen in 145 – reflecting the general trend of increasing economic inactivity since the pandemic.

Inspection of data indicates that even within each typology, local authorities are far from a homogenous group. Further segmentation of these typologies is therefore done according to the composition of economic inactivity. As discussed, a full breakdown of inactivity by reason for all local authorities is only available from the Census 2021, so this is used here. The data used is for all people aged 16+ (rather than 16 to 64 only), so the overall level from it is higher and dominated by retirees more than inactivity for 16 to 64-year-olds would be. Nevertheless, we use it to segment the larger groups according to inactivity excluding retirees and students. The rationale for this is that other types of inactivity are more likely to reflect negative socioeconomic conditions, whereas students presumably hope and intend to work in the future and many retirees are economically inactive out of choice (though this will certainly not always be the case – factors such as poor health may lead some to retire earlier than they would like). H1 and M1 are therefore split according to whether share of inactivity excluding retirees and students is above or below the average for England from Census 2021.

These typologies are displayed below. Focusing on H1a – high inactivity which increased over the pandemic period, with a high share of non-students/non-retirees – there are many of these areas around the coast and some in Outer London, but overwhelmingly they are concentrated in and around northern cities. There are also a number of H1a areas in some isolated coastal areas, such as Cornwall, Plymouth and Torbay. M1a, which is the same but with a lower overall level, is also predominantly northern and urban. Types H1b and M1b, which have high or moderate inactivity but which is largely accounted for by students and retirees, tend to be predominantly rural and coastal. Low inactivity areas L1 and L2 are also more rural Whilst there are many of them in the North and Midlands, they are clearly concentrated in London’s commuter belt and the M4 Corridor.

• H1a: High, increased; above average inactivity excluding retired and students (53 authorities)
• H1b: High, increased; average or below average inactivity excluding retired and students (31 authorities)
• H2: High, decreased (18 authorities)
• M1a: Moderate, increased; above average inactivity excluding retired and students (18 authorities)
• M1b: Moderate, increased; average or below average inactivity excluding retired and students (37 authorities)
• M2: Moderate, decreased (52 authorities)
• L1: Low, increased (27 authorities)
• L2: Low, decreased (75 authorities).

Table 3: High and increased inactivity, including above-average inactivity excluding retired and students

Table 4: High and increased inactivity, but with average or below-average inactivity excluding retired and students

Table 5: High but decreased inactivity

Table 6: Moderate and increased inactivity, including above-average inactivity excluding retired and students

Table 7: Moderate and increased inactivity, but with average or below-average inactivity excluding retired and students

Table 8: Moderate but decreased inactivity

Table 9: Low but increased inactivity

Table 10: Low and decreased inactivity

Whilst every effort has been made to ensure the accuracy of the material in this document, neither Centre for Economics and Business Research Ltd nor the report’s authors will be liable for any loss or damages incurred through the use of the report.

The page allows comparison of your area to others, and allows you to see how that has changed over time. Additional contextual indicators offer further insights into local patterns.

View localised data