Measuring wellbeing in the UK

By Silvia Lui and Ana Rincon-Aznar

Improving wellbeing and raising standards of living across the UK is at the core of the policy agenda. The distributional aspects among regions and socio-demographic groups are gaining significance as the debate focuses on the inclusiveness of growth and prospects for reducing inequalities. However, measuring wellbeing poses a challenge. This is largely because it is difficult to quantify, and likely to respond to both objective as well as subjective factors, which can interact in complex ways.

A new ESCoE discussion paper derives new summary measures of wellbeing for the UK and the regions using individual self-reported responses to the UK Household Longitudinal Study (UKHLS) from 2009 to 2020. Our study aims to answer the following questions: (a) What are the different metrics we can use to measure wellbeing and how do these relate to each other? (b) How do different factors drive wellbeing? (c) How has the trend of different metrics changed over time and across UK regions?

An analytical approach

We use 10 waves of individual responses to the General Health Questionnaire (GHQ) from the UKHLS, to analyse wellbeing in the UK and regions. The GHQ asks respondents about their experience in twelve aspects of wellbeing. These include elements that are conventionally regarded as direct measures of wellbeing (e.g., general happiness), and other elements that more indirectly reflect the state of wellbeing (e.g., sleep, confidence, feeling of playing a useful role, concentration, decision-making capability, feeling constantly under strain, having problems overcoming difficulties, being able to enjoy day-to-day activities, ability to face problems, feeling unhappy or depressed).

Using polychoric correlations

We start by looking at how these wellbeing metrics relate to each other using polychoric correlations. Polychoric correlation measures the level of agreement between two normally distributed continuous latent variables with observed ordinal values. We are examining whether there is evidence to support the idea that self-reported wellbeing is multidimensional, and there exist some common factors that could summarise the different metrics.

Employing probit regressions

We then employ ordered probit regressions to examine the drivers of wellbeing. These drivers include demographic characteristics, household and living conditions, employment, income, caring responsibilities, personal health, and geographic location. Our analysis delves deeper than simply accounting for output and productivity measures found in official statistics and provide insights into the factors that drives these aggregate differences.

Quantifying and aggregating responses

Finally, we quantify and aggregate self-reported individual responses to wellbeing questions in the manner of balance statistics, which according to the UNECE guideline (2019), are “indicators that reflect opinions, attitudes or expectation of survey respondents”. We derive summary indicators of wellbeing, combining the twelve wellbeing metrics, and examine how these indicators have changed across time and regions.

Exploring adjustments to the summary measure

We discuss ways to adjust the summary measure for heterogeneity in wellbeing experiences across individuals. Going forward this is an interesting avenue for research which could be informative for policy-making. The objective is to gain a better understanding of how closely wellbeing is influenced by a range of observable factors, including individual specific and regional economic factors. We also examine an alternative approach for adjustment, where weights are developed to account for the diversity in individual wellbeing experiences, which may be more susceptible to random or unobservable factors. We demonstrate how weights can be constructed to capture such heterogeneity and used in aggregation.

How do the different wellbeing measures relate?

We find evidence that wellbeing is multi-dimensional. Correlations among the wellbeing metrics are found to range between 0.31 and 0.83 with an average of 0.54 in the national analysis. General happiness has an average correlation of 0.54 with the other eleven metrics.

This finding is somewhat surprising as general happiness is often used as an overall measure of wellbeing. If we believe all the other eleven metrics contribute to general happiness and could be summarised by an unobserved common factor driving general happiness, we would have expected the correlation between these to be higher.

These findings raise interesting questions on whether respondents consider their general happiness only when they are asked to evaluate other aspects of their mental health and wellbeing. There may also be other factors that influence our overall wellbeing, that are not being captured by our data. Further research is required to better understand how individuals perceive and evaluate their wellbeing, as well as what measures are more suited to capture this concept.

We find that the regional results are consistent with those at the national level. Unconditional correlations between different aspects do not vary much across the UK geography. There were only a few cases where the correlations were above 0.7. This further supports our finding that wellbeing is multi-dimensional.

How do different factors drive individuals’ wellbeing?

We examine a range of factors that are linked to the probability of observing a particular outcome, using ordered probit regressions and use marginal effects to understand the magnitude of these effects.

Demographic factors:

Gender

Men are more likely to self-score higher in all twelve aspects of wellbeing.

Marital status

Respondents who are married or in a partnership are more likely to report higher wellbeing in most metrics at the national level, although there are differences across UK regions.

Age

Older individuals are more likely to report better concentration, more able to overcome difficulties, experience more confidence and self-worth, but less likely to feel they can play a useful role, less capable of making decisions, and of enjoying day to day activities. But these effects are quite small.

Migrant status and ethnicity

We find that migrant status and ethnicity of the population may play an important role in explaining wellbeing differences at the national level, but there are some differences across the regions.

Other socio-economic characteristics:

Health

Health is one of the key drivers of wellbeing as suggested by the existing literature. Consistent with this, we find that relationship between health and wellbeing is the strongest among all the factors considered. Those with excellent health are up to 30% more likely to report the highest level of wellbeing.

Education level

Having a degree is not unequivocally associated with higher wellbeing. A degree can make people feel that they can play a useful role, as shown in the UK and regional results.

Employment status

Similarly, having a job increases the likelihood of feeling useful, and feelings of self-worth and confidence, but it lowers the score in many other aspects of wellbeing.

Income

Our findings also suggest that higher income is associated with better wellbeing in several aspects, but the size of the effects is smaller than one may have thought. A strong positive relationship between income and wellbeing is mainly observed in London and the West Midlands.

Property ownership

Unsurprisingly, we find that respondents who own their home outright are more likely to self-report higher wellbeing, but those living in urban areas are more likely to report worse wellbeing than those living in rural areas, though the differences again are quite small.

How does the trend of different wellbeing metrics change over time?

Looking at the summary measures of each wellbeing metrics at the national level, our findings show a decline in most metrics since 2014 (wave six of UKHLS). Although we do not observe a unique pattern of change across all aspects of wellbeing, there are similarities. For instance, we observe a declining trend for most aspects of wellbeing, but the timing of this decline differs. The composite measure of wellbeing that combines all twelve metrics also shows a decline in wellbeing in the UK since 2014. We find the same consistent picture regardless of whether summary measures are adjusted for heterogeneity or not.

Our evidence shows some regional differences in the pattern of individual wellbeing metrics. But these differences do not translate into major regional differences when aggregating the twelve wellbeing measures. In other words, the regional composite measures of wellbeing follow a similar trend as the UK as a whole.

Why does this matter?

Analysis in this new ESCoE paper draws a consistent picture of changes in wellbeing over time by exploiting the longitudinal nature of the UKHLS. It offers evidence that cannot otherwise be revealed by looking at the annual snapshots produced by other types of surveys.

It provides a detailed breakdown of the trends by examining a range of aspects of wellbeing in different regions. Therefore, it illustrates a broader picture of changes in wellbeing in the UK compared to most empirical studies that rely on traditional measures of happiness and overall life satisfaction. We establish a method for quantifying categorical wellbeing responses accounting for the impact of individual-specific and regional economic and social factors.

The results illustrate differences in aspects of wellbeing, regional variations, the drivers, and how these changed over time. These findings will be important for policymakers interested in understanding regional disparities in living standards. This is not only in terms economic welfare, but also in terms of the personal and less material aspects of wellbeing.

Read the paper

This paper is part of an ESCoE project on democratic measures of income.

ESCoE blogs are published to further debate.  Any views expressed are solely those of the author(s) and so cannot be taken to represent those of the ESCoE, its partner institutions or the Office for National Statistics.