Measuring well-being inequality

Policy-makers and academics are increasingly interested in well-being inequality. The vast majority of academic studies into well-being inequality currently use standard deviation of personal well-being measures such as self-reported life satisfaction or happiness. However, there has so far only been limited debate on whether standard deviation is the most appropriate measure and its choice is rarely justified.

This summary presents the main methodology and findings from the research work commissioned by Office for National Statistics (ONS) and carried out by the New Economics Foundation (NEF) in collaboration with the What Works Centre for Well-being. NEF was tasked with exploring the strengths and weaknesses of different measures of well-being inequality and to make a recommendation of a measure that could be reported by ONS alongside mean well-being.

The full report, Measuring well-being inequality: Working paper on the selection of a headline indicator, is available on the NEF website.

Proposed measures were judged against the following criteria:

• reflective of public priorities – a good measure of what matters in relation to well-being inequality, for example, inequality in itself, or helping the worst off

• robust to methodological biases – robust to response biases, ordinality versus cardinality and bounded scale effects

• easy to compute – simple to compute and analyse for non-specialists

• easy to communicate – simple to explain and understand

• high predictive power – provides insight into other trends

• sufficient variation – changes enough over time that it can feasibly be influenced by policy

• low correlation with the mean – measures that provide information not already captured by the existing measure of mean well-being

Main stakeholders with an interest in well-being at a national level were consulted via in-depth interviews and a roundtable discussion. This was supplemented by an online survey where wider views were invited, including from non-experts.

From the responses to the consultations, three distinct reasons why people cared about well-being inequality were identified. These were:

• dispersion aversion – that large differences in life experiences are inherently undesirable

• suffering aversion – focusing on individuals experiencing well-being below a particular threshold

• weighted universalism – valuing improvements in well-being for everyone, though with more value given to improvements in well-being for those at the bottom of the distribution

Although all three propositions were held to some degree by participants, weighted universalism was more widely shared than the other positions and is theoretically coherent. This suggests that an indicator should first and foremost reflect the well-being of the worst off, with diminishing weight given to those who are already doing well. Almost all inequality measures identified are, like standard deviation, measures of dispersion. While these indicators may have good uses for academic or other purposes, they failed the criteria for a headline indicator of being reflective of public and policy priorities.

There is a significant gap in the literature exploring alternatives to dispersion measures for well-being inequalities. Three were assessed:

• the average of the bottom 40%

• the percentage below a threshold on the well-being scale

• subjectively-weighted average

Based on the analysis carried out, it was proposed that the use of the percentage below a threshold as a headline indicator of well-being inequalities was preferred. For example, in the year ending December 2017, there were 4.46% of people – almost 1 in 20 – who responded 4 or below when asked how satisfied they are with their life on a scale of 0 to 10. When reported alongside mean well-being, this indicator reflects the dominant values found in the research, particularly concern with the well-being of the worst off, with diminishing weight given to those higher up the scale. It is easy to explain, compute and analyse.

The report aims to open discussion about appropriate indicators of well-being inequality. Based on analysis so far, it has been proposed to use a threshold for a single national indicator.

Further work is needed to identify which threshold to use, although interviewees suggested 4, 5 or 6 on a scale of 0 to 10. Using a threshold of 4 would reduce the sample to those with lower well-being, while increasing the threshold to 5 or 6 might allow for a more preventative approach, focusing on interventions aimed at those who are struggling before they slip into very low well-being.

Following this recommendation, ONS has carried out further work to test different thresholds, published as part of the statistical bulletin Personal well-being in the UK: January to December 2017. It discusses in more detail the reasons exploring the possibility of alternative thresholds for lowest personal well-being, as well as the findings from the analysis undertaken.

It is, however, important to consider that different indicators may be used for different purposes. Researchers are encouraged to reflect on which well-being inequality measure they choose, for a broader debate between main stakeholders in the field of well-being to take place to identify appropriate well-being inequality measures for different purposes.