The measurement of subjective well-being

Paper for Measuring National Well-being Technical Advisory Group, 4 February 2011

Stephen Hicks
Office for National Statistics

This paper is an updated paper that was written initially for the National Well-being Advisory Group
that met on the 5 January 2011. It seeks to take on board comments that were received and reflects
internal discussions since that point within ONS around how best to include subjective well-being
questions on ONS surveys from April 2011. Most substantial changes to the document are from
paragraphs 18 onwards.

Introduction

There is increasing user demand for ‘subjective well-being’ (SWB) measures, both from within the UK
but also from international bodies. Well-being can be measured using both objective and subjective measures
and the ONS believes it is important to capture both when making any assessment of the well-being of a
nation.

2. The Stiglitz-Sen-Fitoussi report (2009) states: "Research has shown that it is possible to collect
meaningful and reliable data on subjective as well as objective well-being. Subjective well-being
encompasses different aspects (cognitive evaluations of one’s life, happiness, satisfaction, positive emotions
such as joy and pride, and negative emotions such as pain and worry): each of them should be measured
separately to derive a more comprehensive appreciation of people’s lives....[subjective well-being] should be
included in larger-scale surveys undertaken by official statistical offices."

3. The Office for National Statistics (ONS) responded to this Stiglitz report recommendation in 2010 by
undertaking research to explore what questions are already available from social surveys in the UK, and to
see whether they would meet customer requirements. In September 2010, ONS published a report called
‘Measuring subjective wellbeing in the UK’ (Waldron, S. 2010) which reviews what is meant by the term
subjective wellbeing, why it is important and how it can be measured. Although there are a large number of
other sources, for example the BHPS, BSAS and Defra surveys, that do and have asked these questions
previously, there is a demand for these types of questions to be placed on large scale official statistical
surveys, to allow for accurate estimates for sub-groups of the population and at the sub-national level.

4. SWB measures should be seen as providing useful supplementary information to already existing
objective measures that are available. There is a growing recognition and literature that suggests that life
satisfaction and happiness measures provide a reliable and valid account of people's wellbeing (Waldron, S.

1
2010). Placed alongside objective measures, they provide useful insights which will help inform policy making
and assessment as well as for general monitoring purposes.

Rationale for measuring subjective well-being

5. Objective accounts of well-being rely on assumptions about human needs and rights. Sen (1999)
argues that these basic needs must be met in order for people to flourish and given the role the state has in
providing these needs (for example education and healthcare) objective measures have an important part to
play in measuring national well-being (Dolan, P. et al, forthcoming). However, the objective approach does
have limitations: for example, it requires researchers to decide what indicators are most important for
monitoring and assessing the well-being of a nation before any assessment can be made.

6. On the other hand, SWB has the advantage that it avoids the need for a priori assumptions about
what constitutes a good life (Smith, C. OECD, unpublished) and also avoids paternalism (assumptions about
what adds and detracts from individuals well-being) (Waldron, S. 2010). This is because it asks individuals
about their views about their own well-being and allows them to make their own assessment of their well-
being. The theoretical rigour for SWB extends back to Bentham (1789) who provided an account of well-being
based on pleasure and pain which provided the background for utilitarianism (Dolan, P. et al. forthcoming).

7. SWB is not just subjective because it relies on self-reported measures alone: there are examples of
objective measures which are self-reported (for example employment status or household income, used in
measuring GDP). Subjective well-being is subjective because the subject matter itself that is being measured
is a subjective concept for example, life-satisfaction or happiness. Any distinction between objective and
subjective should not be drawn too boldly: objective list accounts of well-being can also include subjective
well-being in them.

8. Evidence has been building to show that SWB is a valid construct that can be reliably measured
(Waldron 2010). We know for example that SWB measures correlate well with objective indicators, for
example with income (both absolute and relative), employment status, marital status, health, personal
characteristics (age, gender, and personality), and major life events (See Dolan et. al. 2008 for a recent
review). Causality is a key issue: there are data on SWB both as cause and as consequence.

Conceptual framework for subjective well-being measurement

9. SWB measures can be divided into three broad types: evaluative measures, experience measures
(sometimes known as ‘affect’ measures) and eudemonic (sometimes referred to as psychological) measures.
(See Figure 1 which highlights these concepts). (Waldron, S. 2010 gives references to the primary research
and Dolan, P. et.al. forthcoming discusses the distinction in more detail).

10. Life satisfaction measures are perhaps the most well known and commonly used evaluative measure,
particularly in the UK and Europe partly due to the extent that questions have been used in surveys (Waldron,
S. 2010) but also because they are seen by policy makers as useful. (Donavan, N. and Halpern, D, 2002).
Evaluative measures ask the respondent to stand back and make an assessment of their life and, in the case
of life satisfaction, score their life with regard to their satisfaction. Other types of evaluative measures include
the Cantril ladder of life measure which asks respondents to imagine a ladder where the bottom (0) is the
2
worst possible life and the top (10) the best possible life and asks them to give an indication as to where they
feel they are on this scale. Other measures include general happiness measures that are not time bounded
which generally correlate with life-satisfaction scores.

11. Experience (or affect) measures perhaps are most closely aligned to the Benthamite view of well-
being as they are concerned with how peoples feeling and emotions are affected by everyday events (Dolan,
P. et al. forthcoming). Different measures exist to capture these 'affects', for example the Ecological
Momentary Assessment (EMA) or Day Reconstruction Method (DRM) which are diary based approaches.
Simpler approaches include the Princeton Affect Time Use Survey (PATS) which asks people for their
experiences at random times in the previous day, a simpler method still is to ask about feelings relating to the
whole day which the US Gallup World and Daily Polls have done (Dolan et. al. forthcoming). Such simpler
measures inevitably lose potentially important data but are still useful and more appropriate for large scale
general purpose social surveys where the luxury of a time use survey with diaries are not available.

12. Eudemonic measures are somewhat different to the evaluative and experience based measures but,
never-the-less, there seems to be a consensus that they are important to include when measuring SWB.
There is certainly the demand for these types of measures and the theoretical underpinnings relate to a
broader understanding of well-being which encompasses purpose and meaning in life as well as including
concepts such as competence, autonomy and engagement.

Figure 1. Conceptual framework for measuring subjective well-being

Income
Employment status
nts

Health status
na
mi

Age / Sex
ter

Socio Economic
De

Religious affiliation
Personality etc…… Evaluative Experience Eudemonic
Theoretical underpinnings

Overall Life satisfaction Happiness/worried Overall

monitoring yesterday Worthwhileness
Sub-components

Life satisfaction by Energy levels

Policy
domain1 Relaxed
formulation
Stressed
Anger
Policy Detailed domains Affect associated Purpose and
appraisal with particular meaning from
Satisfaction with
services difficulties specific activities
1. Planned domains include: Personal relationships, physical health, mental well-being, work situation, financial situation,
area where you live, time you have to do the things like doing, wellbeing of your children (if any).
Framework adapted and developed from Dolan, Layard and Metcalfe (2011) & Smith (2011) OECD, unpublished.

3
13. Any measure of SWB should have a good theoretical underpinning. Added to this, it is vitality
important, if these measures are going to be useful, that they are considered relevant to policy needs. It is
likely, though, that different measures will be used for different purposes in public policy. It has been proposed
by Dolan et. al. (forthcoming) that questions are grouped into three main areas depending on the level of
detail that they provide. Firstly, it seems apparent that there needs to be some overall questions that provide
estimates for overall monitoring of SWB in the UK. Secondly, measures that are used for policy formulation,
which are likely to be more detailed domain and affect questions should also be asked. Thirdly, for policy
appraisal there is a need to be more detailed still, and specific to particular activities and services (see Figure
1 above).

14. It is also important that we also not only look at the overall levels of SWB in the population or at the
domain and more detailed levels, but that we also seek to understand why people's SWB scores are as they
are. This can partly be done by looking at the subjective and objective determinants of subjective well-being,
which would include regressing measures on overall life-satisfaction, affect and eudemonic measures. Adding
SWB measures to the ONS Integrated Household Survey would allow further research to be undertaken in
this area based on very large sample sizes (Dolan et. al. forthcoming).

Developing subjective well-being questions for ONS social surveys

15. When constructing subjective well-being questions for the inclusion in ONS surveys, a two part
approach was adopted. Firstly, ONS undertook work to look at questions in existing surveys both in the UK
and abroad and to learn from them (Waldron, S. 2010). ONS also sought academic advice from Prof. Paul
Dolan (LSE), Prof. Lord Richard Layard (LSE) and Dr Robert Metcalfe (Oxford University) for the
development of questions, as well as more recently speaking with Prof Felicia Huppert (Cambridge
University) who provided advice on an overall eudemonic measure. Feedback from the Measuring National
Well-being advisory forum was also sought and received and this helped further formulate the questions that
should be asked on ONS surveys.

16. When deciding which ONS survey(s) to place which subjective questions on it was important to not
only consider the user demand for a large sample, to yield results with small standard errors, but also to think
about the nature of the survey and how these questions could potentially impact on existing questions.
Likewise, it is also important to balance the amount of information collected with the ‘burden’ on respondents
to ONS surveys; ensuring that response rates are not adversely affected due to lengthening of the interview.
This is particularly important given the reliance that is placed on our survey estimates already, for example
the Labour Force Survey is a constituent survey of the IHS and provides important monthly labour market
statistics.

17. In order not to overload the IHS with SWB questions it was decided that the IHS would be the right
vehicle to carry overall monitoring questions rather than the domain specific and detailed questions. This
would mean that ONS intend to ask four overall questions on the Integrated Household Survey (IHS), which
would cover the evaluative, experience and eudemonic accounts, allowing for overall monitoring of SWB in
the UK (as well as at the sub-national level) and providing a rich and large dataset for analysis (aiming for an

4
achieved sample of 200,000 directly questioned adults responding each year). This would be supplemented
with domain specific and detailed questions asked regularly on our Opinions Survey, which is a monthly
survey with a random sample of around 1,000 adults responding each month. Setting the balance between a
shorter number of questions on the IHS and a longer list on the Opinions Survey is not a precise cost-benefit
calculation and the mix (and indeed the questions) can be varied over time if required.

18. The point was made at the advisory group that these measures would benefit from international
harmonisation and it was recognised that the ONS timeframe for inclusion of these questions (April 2011)
may not fit with the slightly longer time frame for international developments in this area. ONS is keen to
where possible not only to meet national needs but try and ensure that international comparability is met so
that the UK can be compared with other countries.

19. The questions that will be added to the IHS in April 2011 will not have ‘National Statistics’ status
immediately and as with all new statistics that are undergoing development and refinement they will be
labelled ‘experimental’ statistics. This affords us with the opportunity to undertake further evaluation and
development of these measures as we go forward, talking into the various dimensions of statistical quality as
we do so. After adding questions to ONS surveys in April 2011 we will publish estimates and findings from
further testing work that we plan to undertake to involve users and stakeholders in their development as a
means to build in quality at an early stage to allow these estimates to meet the necessary quality standards to
become National Statistics at a later stage.

20. This is likely to include the use of the Opinions Survey, as well as a way of collecting the more
detailed data, as a vehicle to test the order and position affects on the estimates, something that was raised
at the advisory forum. Also to try out additional questions that are deemed important as we go forward and to
test question wording to see whether any changes are required from April 2012.

21. ONS tested the headline questions and domain satisfaction questions in the December 2010
Opinions Survey and initial results seem to suggest that they are working successfully; providing estimates
that we would expect and with low item non-response rates, at around two per cent for the overall monitoring
questions.

22. Feedback from some members of the advisory forum again highlighted the likely impact on estimates
of the placement of the questions on surveys – for example, placing after questions relating to health or the
labour market may impact the answers that respondents make. This is something that ONS had foreseen
could be an issue and we planned small scale cognitive testing of the placement affects which is currently be
undertaken. It is envisaged that the placement of the overall monitoring questions will be upfront in the
survey, but the study will not report until the middle of February. One aspect we are considering is the affect
of asking these questions before or after the religious affiliation question on the IHS. It could be that asking
after the religion question could lead to a positive bias on life satisfaction scores, given the known correlation
of religion and subjective well-being estimates.

23. Feedback from a small group of ONS interviewers has also been positive, they thought the questions
were upbeat and felt that respondents would be able to answer them easily. However, they did highlight the

5
potential problem with the 0-10 scale which was that they often felt themselves wanting to say 1-10 rather
than 0-10. To overcome this interviewer tendency we will need to make explicit the need for using a 0-10 in
our interviewing instructions.

24. Mode of interview could also have an effect on subjective well-being estimates and this has also been
borne in mind. In the December 2010 Opinions survey the questions were administered face-to-face (CAPI)
by interviewers rather than using a (CASI) method. This was because we wanted to test how they performed
using this method because in the IHS the interviews are either conducted by face-to-face interviewing or by
telephone interviews – no CASI is used for any of the questions. This is because the interview is slowed down
because of this method and also relies on respondents having a high enough level of computer literacy. The
fact that the IHS is a household survey and that other members of the household may be part of the interview
process also has to be considered when deciding which questions are asked on the IHS versus the Opinions
Survey which only interviews one person in the household. Some of the domain specific questions are more
sensitive and may suffer from bias if other household members are present and aware of the questions being
asked.

25. In terms of an overall evaluative measure, ONS still believe that a single overall life-satisfaction
question is appropriate for the IHS, where space is at a premium. ONS are planning to ask ‘Overall, how
satisfied are you with your life nowadays?’ (see Figure 2 for wording of planned questions). This question is a
simplification of other questions that have been asked before on other surveys in the UK and abroad (e.g.
World Values Survey). The Cantril Ladder of life question, which has been argued is less influenced by
respondents current affective state and yielding better cross-country comparisons, is however a fairly lengthy
question, with a conceptual picture (the ladder) that takes some explaining to the respondent. For that reason
we do not think it is appropriate to include this on a multi-purpose household survey (i.e. the IHS) and that the
life-satisfaction question is a better measure.

26. ONS believe that an 11 point scale (0-10) should be adopted, and for which there are many
supporters around the world (although not everyone agrees with this, for example proposing instead a larger
number of simple dichotomous – yes/no – questions). In terms of the labelling of the scale we tested the use
of ‘not at all satisfied’ to ‘completely satisfied’ in the December 2010 Opinions and these appear to have
worked successfully. Other question scales run in five points from ‘completely dissatisfied’ to ‘completely
satisfied’, which can make it hard to interpret scores as it is not clear where dissatisfied becomes satisfied
(Dolan, P. et.al. forthcoming). Also it is important in the same survey to as much as possible make the scales
for life satisfaction consistent with affect measures that are more likely to run from satisfied. Added to that, the
question does ask about how satisfied the respondent is rather than how satisfied or dissatisfied.

27. In terms of an experience or ‘affect’ question on the IHS ONS had planned to use one overall
question asking ‘Overall, how happy did you feel yesterday?’. ONS intend to use ‘yesterday’ in order to give
an approximation in a general survey vehicle to the DRM, Time Use approach.

28. It is important to measure both evaluative and experience based measures separately as we know
that they complement one another and although in some cases give similar results we know that is not always
the case (Dolan, P. et al, forthcoming). Feedback from advisory board members appeared to confirm this but
6
it was mentioned that ONS should not only be asking a ‘positive’ affect question but also a ‘negative’ affect
question. ONS has considered this and are now thinking that we may also include a negative affect question
in the IHS to supplement the positive affect question. We have chosen the word ‘happiness’ for the positive
affect question as this adjective, one of the ones used on the Gallup world poll. It is intuitive and is used in
everyday parlance in the UK and appears to us to be most suitable if only two questions can be asked for
overall monitoring rather than four which would only allow us to cover the quadrants of the Circumplex model.
In terms of a negative question we would suggest the same question structure as the positive affect question
but using the word ‘worry’ as the relevant adjective given it’s link to depression and anxiety – Overall, how
worried did you feel yesterday?

29. We are planning a 0-10 scale from 'not all happy’ to ‘completely happy’ and ‘not at all worried’ to
‘completely worried’ (see figure 2). One thing that we have noticed is that for the positive affect question a
high score is desirable, whereas for the negative affect question a low score is desirable for the respondent.
There is the concern that respondents may not pick up on this switch in the meaning of the scale and provide
an answer which is similar to previous questions, when really they could want to provide a score on the other
end of the spectrum. This issue appears to be present on other surveys but we believe could need the
interviewer to reinforce the scale for this question (and to prompt as necessary) to ensure the quality of the
estimates is maintained.

30. The benefit of the 0-10 scale over a simple dichotomous or 5 point scale to the affect questions is that
it is also consistent with the other questions that are being asked, which we believe will help respondents. A
simple dichotomous scale would also mean less information is collected and could be harder for respondents
to answer than a 11 point scale which we have tested and appears to work well for the positive affect question
that we have asked.

31. In terms of an overall eudemonic measure of well-being there is less consensus about a single
measure to be used. For this reason Dolan, P et.al. devised a high level question which we are currently
testing, ‘Overall, how much purpose does your life have?’. Following discussions with Prof. Felicia Huppert,
we are also intending to ask another question around how valuable and worthwhile people feel the things that
they do in their life (see Figure 2 below). This is one question from the Individual Flourishing Questionnaire,
Huppert, F and So, T, 2010), the short version of which has 15 questions. There are other questions that
could be drawn from this questionnaire, and other questions that could be designed, such as to assess the
pleasure gained from activities. One driver for the further development and use of questions will be how they
can act as a back-drop to policy. At the advisory board meeting Professor Paul Dolan made it clear that on
reflection felt that a variant on the valuable and worthwhile question would be preferable over the purpose in
life question outlined above. He has suggested ‘Overall, how worthwhile are the things that you do in your
life?’.

32. The New Economics Foundation at the advisory board also suggested that a further question on
relationships should also be added to the overall monitoring questions in the IHS given the known importance
of personal and social relationships on people’s subjective well-being.

7
33. The ordering of these overall monitoring questions is also important and we would welcome further
advice from the technical advisory group. Currently our thinking is that we would start with life satisfaction,
then move onto a eudemonic question (which is also more evaluative in nature) and then move onto ask the
affect questions. That would help with the flow of questions and mean that we would not need to reinforce the
scales (unless the respondent appears to struggle) until the last question about negative affect where we
would we believe want to reinforce the switch in the scale as mentioned above. When testing in the
December Opinions Survey, interviewers outlined every scale before each question but for the IHS we are
considering whether the we can make the questions run more smoothly and quickly with the single pre-amble
up front and without the need to repeat the scale with the relevant adjective each time (apart from the last
question) (see figure 3 below). We would welcome views from the technical advisory group on this.

34. More detailed questions that look at life satisfaction by domain are also useful and we are proposing
to ask these questions on our Opinions Survey (Q4 in figure 2 below) on a regular basis along with more
detailed 'affect' questions (Q5-7 in figure 2 below). The domains for life satisfaction questions are based
broadly on those used already in the British Household Panel Survey, with the addition of one for mental well-
being (which we believe should be brought out separately from physical health) and the well-being of
respondents children which we believe has a strong impact on parents well-being. One of the benefits of
using ‘work situation’ is that this can apply to those who are not only in work but also to those who are
unemployed or inactive – it will be useful to see how the unemployed and inactive perceive their satisfaction
with their current situation. Also we are planning to use financial situation because that is broad enough to
capture current financial situation and also the future if the respondent thinks that is important when making
an assessment.

35. The many questions that could be asked on detailed domains, satisfaction with services, affect
associated with particular difficulties and purpose and meaning for specific activities are not planned for
inclusion in regular ONS surveys. These are better suited to more subject specific surveys or modules of
surveys on an ad-hoc basis as and when the policy requirement arises.

A note on references

Full details of all published work referred to this paper can be found in:

Waldron, S (2010) Measuring Subjective Wellbeing in the UK, ONS working paper at
http://www.statistics.gov.uk/cci/article.asp?ID=2578

The paper by Dolan, P et al will be published by ONS in early 2011.

The Individual Flourishing Questionnaire – Short version (IFQ-Short) was published by Felicia A Huppert and
Timothy T. C. So, University of Cambridge, in 2010.

Work by Conal Smith is in preparation for an OECD handbook of guidelines for measuring SWB.

8
Figure 2. Proposed Questions
For inclusion in
the ONS integrated
1. Overall, how satisfied are you with your life nowadays? Household Survey
(on a 0-10 scale, where 0 is not satisfied at all and 10 is completely satisfied). (Continuous Survey,
200,000 adults in
2a. Overall, how happy did you feel yesterday? achieved sample,
(on a 0-10 scale, where 0 is not at all happy and 10 is completely happy)
rolling annual
estimates published
every quarter).
2b. Overall, how worried did you feel yesterday?
(on 0-10 scale, where 0 is not purposeful and 10 is completely purposeful)

3. Overall, how worthwhile are the things you do in your life?

(On a scale of 0-10, where 0 is not at all worthwhile and 10 is completely worthwhile)

4. Overall how satisfied are you with: For inclusion in

your personal relationships; the ONS
Opinions Survey
your physical health;
on a regular
your mental wellbeing; basis, monthly
your work situation; survey, approx
your financial situation; 1,000 adult
the area where you live; respondents per
the time you have to do things you like doing; month

the wellbeing of your children (if you have any)? (0-10 scale)

5.Overall, how much energy did you have yesterday?

6. Overall, how relaxed did you feel yesterday?
7. Overall, how stressed did you feel yesterday?
8. Overall, how angry did you feel yesterday?

9
Figure 3 – Proposed question scheme for the IHS
Preamble:

Next I would like to ask you some questions about your well-being. For each of these questions I'd
like you to give an answer on the scale 0 to 10, where 0 is 'not at all' and 10 is 'completely'.
:TCont

Questions:

Satis
Overall, how satisfied are you with your life nowadays?
Interviewer instruction: where 0 is ‘not at all’ satisfied and 10 is completely satisfied
:0..10, DK, RF

Worthwhile
Overall, how worthwhile are the things you do in your life?
Interviewer instruction: …..where 0 is ‘not at all’ purposeful and 10 is completely purposeful
:0..10, DK, RF

Happy
Overall, how happy did you feel yesterday?
Interviewer instruction: where 0 is ‘not at all’ happy and 10 is completely happy
:0..10, DK, RF

Worry
On a scale where 0 is not at all worried and 10 is completely worried, overall, how
worried did you feel yesterday?
:0..10, DK, RF

10