All pages

Chapter 1: Three approaches

In this chapter we briefly discuss three approaches to measuring well-being, single-measure approaches, objective approaches and subjective account approaches. It particularly argues the case for subjective accounts. At the end of the chapter there is a short exercise using life satisfaction data to whet your analysis appetite for the following chapters.

Page 1

Single measure

The simplest approach sees well-being as a uni-dimensional concept. This implies that an individual’s well-being can be described with a single number. In other words collecting more than one data point for that individual, whilst perhaps improving the reliability of one’s measure, does not actually provide any more information about their well-being. The most popular manifestation of this approach has been the ‘evaluative’ approach to well-being (e.g. Ed Diener’s ‘satisfaction with life scale’).1 In this approach, an individual has high subjective well-being to the extent that they evaluate their life to be good, to the extent they have high life satisfaction. In the field of well-being measurement, this has perhaps the oldest pedigree, having begun with classic studies such as those of Richard Easterlin.2 Typically respondents are asked a question like this:

‘All things considered, how satisfied are you with life as a whole nowadays?’ (on a scale from 0-10)

Whilst seeming simplistic, this question passes the standard tests of statistical validity and reliability. In terms of reliability over time, for example, Fujita & Diener (2005) have found test-retest correlations over five years of r=0.51, which is a very high level for a subjective measure.3

In terms of the different types of validity:

Advantages and disadvantages

Before continuing, have a think about what you’ve just read and consider what you think might be the advantages and disadvantages of life satisfaction as a measure of well-being to be used by governments. After you’ve done so, click here to see the list that we’ve come up with:

Table 1.1. Some advantages and disadvantages of life satisfaction as a measure of well-being
Advantages Disadvantages
Easy to measure Simplistic
Easy to understand Determined by expectations
High validity within a given culture Cultural assumption that life satisfaction is the most important element of well-being
  Response biases (often cultural)
  Not transparent r.e. policy
  Insensitive

Page 2

Objective accounts

Some of the disadvantages with single measures, have led social scientists entirely away from subjective measures.1 Fundamentally, they argue satisfaction is determined by two things:

Satisfaction = Current conditions / Expectations

Of course, all things being equal, higher satisfaction would be expected from improved conditions. Unfortunately though, improved conditions also typically lead to higher expectations thus cancelling out any gains. This would explain the lack of sensitivity often seen in life satisfaction data, and the fact that people who’s conditions ‘objectively’ seem bad, often have high scores on life satisfaction.

Assessing well-being without asking any questions is difficult. Fortunately, there exists a broad category of approaches called basic needs accounts or capabilities accounts of well-being. These approaches attempt to break down well-being into various aspects of life and operationalise these using objective data, i.e. not based on respondents’ perceptions. Often these are based on humanistic theories such as Maslow’s Needs Hierarchy or Max-Neef’s Needs Theory.

The most well-known example of this approach is the UN’s Human Development Index (HDI). This approach identifies human development (which is a similar concept to well-being) as resting upon three factors: income, education, and health. All these factors are operationalised using objective data - GDP per capita, literacy and enrolment rates, and estimated life expectancy at birth.

Another example of this approach is UNICEF’s report on children’s well-being in OECD countries.

Advantages and disadvantages

Before continuing, have a think about what you’ve just read and consider what you think might be the advantages and disadvantages of objective accounts as a measure of well-being to be used by governments. After you’ve done so, click here to see the list that we’ve come up with:

Table 1.2. Some advantages and disadvantages of objective accounts as a measure of well-being
Advantages Disadvantages
More transparent to policy Difficult to bring together different aspects
Multi-dimensional Assumes scientists can define well-being objectively
Potentially sensitive Difficult to capture some important things with objective data
Much data readily available  

Perhaps the fundamental problem with resorting to purely objective data is that of validity. It becomes very difficult to measure some constructs which seem fundamental to well-being. For example, in the Fulfillment of Hierarchical Needs Index, Maslow’s need for belonging (i.e. social well-being) is operationalised as the number of telephone lines per capita in the country and the fertility rate.2 Another influential approach3 uses the divorce rate to monitor social well-being - but divorce rates are highly dependant on culture. From one perspective, high divorce rates are good because they mean that people aren’t stuck in unhappy marriages. Things get even harder when one tries to operationalise important aspects of well-being that have been identified, such as self-esteem. How can one measure self-esteem objectively? By observing individuals’ back posture as they walk down the street?

Page 3

Subjective accounts

It is difficult to measure some constructs objectively (e.g. self-esteem). This problem provides a role for some form of subjective assessment of well-being. However, to respond to the criticism that single-item measures such as life satisfaction are too simplistic and difficult to interpret in terms of policy, such subjective measures must be, like the objective accounts, multi-dimensional. This means that they do not claim a single number can define all of well-being. An individual, or country, can have high well-being with respect to one dimension, but low with respect to another. These different dimensions must be related to different policies in different ways.

An early example of this approach is the psychological well-being scale of Carol Ryff & Corey Keyes.1 This scale claims to measure six aspects of human well-being: autonomy, personal growth, self-acceptance, life purpose, mastery and positive relatedness. Other examples include the Dumfries & Galloway Wellbeing scale,2 the Psychological Needs Scale3 and the European Quality of Life Survey.4

These approaches all offer richer information on the state of well-being of an individual or population group. Rather than just telling you an individual is dissatisfied, they will tell you why (e.g. they do not have opportunities to grow, or they feel lonely). Whilst this often does not directly link to policy, often it indicates areas where government might help. For example, if people do not have opportunities to develop their capacities, then perhaps the government should support lifelong learning programmes, if people feel stressed, perhaps the government should look at working hours.

The specificity of questions, reduces the risk of response biases. For example a pessimistic individual asked about their overall life satisfaction has the opportunity to focus on those aspects of their life that are less optimal. If they are asked about something specific, then they have less opportunity to do so. For example, ‘do they find the time they spend with their family enjoyable’.

Also, one should hope for higher sensitivity. Responding to a question on a 0-10 scale, there are only 11 possible answers available. Responding to 10 questions on such a scale, there are 101 possible totals, and 110 billion different permutations.

Lastly, the problem with expectations is one that is particularly important to questions like the life satisfaction question. For example, asking someone how often they feel tired, is less likely to be determined by their expectations. By including a range of different types of questions, including some which are more behavioural, or descriptive (in the sense that they are less about rating feelings, and more about counting something such as instances), we can feel more confident that such measures would be sensitive to changes in conditions that one would expect to impact on well-being.

The third round of the ESS provides the biggest ever range of questions with which well-being has been operationalised, over such a large sample. As well as the Well-Being module (module E in the survey), there are also several other relevant questions in the core survey. The following chapters present one way of using this data, based on the report by nef (the new economics foundation), National Accounts of Well-Being (published January 2009).

Page 4

Exercises

Exercise 1

Let’s start with a simple exercise comparing life satisfaction measures of two groups – the unemployed people looking for work versus people in paid work. Calculate the means for the two groups across Europe. Remember to apply appropriate weights. You are going to use the variables ‘stflife - How satisfied with life as a whole’ and ‘mainact - Main activity last 7 days’. If you are using SPSS, calculate the t-statistic to determine whether the difference between the two groups is significant.

Solution - Nesstar

The table should now tell you that the people in paid work tend to be more satisfied than unemployed people. The mean life satisfaction for people in paid work is 6.7, and 5.1 for those unemployed looking for a job.

Solution - SPSS

We assume that you have downloaded the dataset and that you have SPSS available.

First you should weight the data: Select ‘Data’ on the menu, then ‘Weight cases’, ‘Weight cases by’, find the combined weight and click ‘OK’.

The second step is to perform the t-test. Select ‘Analyse’ on the menu, then ‘Compare Means’ - ‘Independent Samples T-Test’. Select ‘stflife - How satisfied with life as a whole’ as your ‘Test Variable’ and and ‘mainact - Main activity last 7 days’ as your ‘Grouping Variable’. Define the groups like this: Group 1 is the value 1 (paid work), and group 2 is the value 3 (unemployed and looking for a job).

The results should now tell you that the people in paid work tend to be more satisfied than unemployed people: Mean life satisfaction for people in paid work = 6.68, for those unemployed looking for a job = 5.08. t-score (assuming equal variances) = 11.38, p <0.001.

You could also copy this syntax and paste it into a syntax window in SPSS:

WEIGHT
BY COMWEIGHT.
 
T-TEST
GROUPS = MAINACT(1 3)
/MISSING = ANALYSIS
/VARIABLES = STFLIFE
/CRITERIA = CI(.95).

 

Next, let’s confirm the correlation between physical activity and life satisfaction (as noted in Dolan et al. 2006).1 Is it significant? Make sure you check the independent variable’s response codes to be sure – do high numbers mean high physical activity or do they mean low physical activity? Use the variables ‘stflife - How satisfied with life as a whole’ and ‘pactlot - My life involves a lot of physical activity’.

Solution - Nesstar

Make sure that the combined weight is still on.

The correlation should be 0.033, p<0.001.

By clicking the tab ‘Description’, you can see how the variables are coded. Larger values indicate more satisfaction and more activity. The positive correlation means that there is a tendency for activity to be accompanied by satisfaction.

Solution - SPSS

Make sure that the combined weight is still on.

Follow this path on the menu: Analyse – Correlate – Bivariate. Select the variables ‘stflife - How satisfied with life as a whole’ and ‘pactlot - My life involves a lot of physical activity’. Select the coefficient ‘Spearman’ or ‘Pearson’.

If you use Pearson, the results will be as follows: r = 0.033, p<0.001.

By looking at the value labels (the coding), you will see that larger values indicate more satisfaction and more activity. The positive correlation means that there is a tendency for activity to be accompanied by satisfaction.

You could also copy this syntax and paste it into a syntax window in SPSS:

CORRELATIONS
/VARIABLES=PACTLOT STFLIFE
/PRINT=TWOTAIL NOSIG
/MISSING=PAIRWISE.

 

Do the results from the ESS corroborate the findings from Dolan et al.? If you like, check some of the other variables that we mentioned (e.g. marital status, income, social networks).

Exercise 2

Before moving on to Chapter 2, take a look at the list of questions in the survey and identify which you think could be included in subjective accounts of well-being. Be sure to choose questions which you think actually measure subjective well-being itself – i.e. how people feel about their experience of life, and not objective factors that might predict well-being (e.g. income). Don’t worry too much about any single question – the distinction between what someone feels and someone reports is often not clear cut. The framework that we will be presenting in the next chapter is just one way of making that cut.

Go to next chapter >>
Page 5