Tessa Peasgood Centre for Well-being in Public Policy Sheffield University Modelling Subjective Well- being. Do strong social relations lead to increases.

Tessa PeasgoodCentre for Well-being in Public PolicySheffield University

Modelling Subjective Well-being. Do strong social relations lead to increases in well-being?

SWB in the British Household Panel Survey “Please tick the number which you feel best describes how

dissatisfied or satisfied you are with the following aspects of your current situation….......……. your life?”

1 2 3 4 5 6 7Not satisfied Completely

satisfied

12-item General Health Questionnaire (GHQ) - used to detect the presence of non-psychotic psychiatric morbidity in community settings

THE GHQa) been able to concentrate on whatever you're doing ?c) felt that you were playing a useful part in things ?d) felt capable of making decisions about things ?g) been able to enjoy your normal day-to-day activities ?h) been able to face up to problems ?l) been feeling reasonably happy, all things considered ?

More so than usual (1), Same as usual (2), Less so than usual (3), Much less than usual (4)

b) lost much sleep over worry ?e) felt constantly under strain ? f) felt you couldn't overcome your difficulties ?i) been feeling unhappy or depressed ?j) been losing confidence in yourself ? k) been thinking of yourself as a worthless person ?

Not at all (1), No more than usual (2), Rather more than usual (3), Much more than usual (4)

BHPS Satisfaction with life overall, waves 7 to10 & 12

0

5

10

15

20

25

30

35

1 2 3 4 5 6 7

Satisfaction with life

Perc

enta

ge o

f res

pons

es

Male

Female

Why use both life satisfaction and GHQ? May measure different aspects of SWB

Different time period (4 weeks vs.1 year+) Life satisfaction more evaluative GHQ externally determined affects, may not be

what is important to the individual

Different measures have different problems, if results are similar, or we understand why they vary, should give confidence in results.

How valid are these SWB measures?Interpretation problems

Not clear what time period are people using for life satisfaction Not clear if people are including non self-referencing concerns

e.g. others well-being GHQ compares to ‘usual’ but may be using scale absolutely

a third of respondents who answer better than, or less than usual to the general happiness question in wave 12 also do so in wave 13

Measurement error Situational factors & question ordering – what attention is drawn

to at the time Life satisfaction influenced by mood - BUT life satisfaction more

stable than mood Culturally appropriate responses

problematic if adjusting responses is linked to circumstances e.g. unemployment

Are life satisfaction responses internally consistent?

Actual change in life satisfaction score

Would you say that you are more satisfied with life, less satisfied or feel about the same as you did a year ago?

Increased Decreased Same

More satisfied 34% 18% 48%

Less satisfied 17% 52% 31%

Same 25% 26% 48%

Why the inconsistency?

Poor memory for what life was like a year ago

Measurement error (current year, previous year or both)

Not a clear concept

Revaluate past based on new information Need to assume people use the scales in the same way &

that current evaluation of current time is privileged

Some confidence derived from: SWB correlates with informant reports, smiling and

interviewer ratings

SWB predicts suicide

Other subjective measures have predictive powers Job satisfaction predicts quitting job Subjective health predicts suicide and longevity

Life satisfaction & GHQ will tell use something about SWB but will have considerable measurement error.

Social capital and well-being Has been found to be linked to beneficial outcomes e.g.

lower crime rates, child welfare, public health, market performance, education performance (Helliwell & Putnam 2004)

Some evidence of direct link social capital to individual well-being/happiness (Putman 2000, Helliwell 2004, Diener and Seligman 2002), using range of measures for social capital

Focus here on social and personal relationships Intimate relationship & marriage status How often meet friends and family How often talk to neighbours Average level of social relationship scale for the district

The Model

SWBit = β’Xit + αi + uit

SWB (Life satisfaction or GHQ score)

for individual i

at wave t

Matrix of explanatory variables e.g. income & health

of individual i

at wave t

Individual effect

Shifts SWB up/down but doesn’t change how much Xs effect Y (β same for everyone)

If happy disposition would shift Y upward in all time periods

Error term

All change in SWBit not captured by Xit and αi

SWBit = β’Xit + αi + uit

To estimate β need to either remove or estimate αi

Fixed effects Using deviations from individual means individual effect (αi)

drops out (αi is the same as the individual mean αi)

Comparing individual to themselves at different time periods

Can’t say much about variables which don’t vary ‘within’ the individual over the time period

Can use OLS fixed effects on life satisfaction and GHQ but assumes cardinality

Ordered logitTreat life satisfaction as latent, continuous variable LS*

LSit = 1 if -∞ <= LS* <= μ1= 2 if μ1 < LS* < μ2

= 3 if μ2 < LS* < μ3 etc.

Probability of an outcome (e.g. Life sat = 5) calculated as linear function of explanatory variables plus set of thresholds or cut points.

Control for unobserved individual effects by including individual level means of all explanatory variables (Mundlak approach)

Since model uses logit assumption on error term the log odds of being in higher life satisfaction category are linearly dependent on the explanatory variables.

itiitit uXLS *

Explanatory Variables: Social relationships Ordered Logit on life satisfaction (All)

Base category: sees friends or family and talks to neighbours most days, married, no accommodation problems

Odds ratio P value

Sees friends or family once or twice a week 0.954 0.018

Sees friends or family once or twice a month 0.955 0.154

Sees friends or family less than once a month 0.898 0.086

Never sees friends or family 0.611 0.043

Talks to neighbours once or twice a week 0.925 0.000

Talks to neighbours once or twice a month 0.773 0.000

Talks to neighbours less than once a month 0.840 0.000

Never sees neighbours 0.830 0.007

Mean social relationships scale of district 1.296 0.011

Divorced/Separated 0.764 0.002

Co-habiting 1.027 0.647

Widowed 0.621 0.001

Never married and not co-habiting 0.823 0.021

Living alone 0.998 0.969

Problem with accommodation: noisy neighbours 0.971 0.354

Problem with accommodation: street noise 0.994 0.822

Problem with accommodation: vandalism / crime 0.960 0.108

Explanatory variables: Income Ordered Logit on Life satisfaction (all)

Base category: top income quintile, ‘living comfortably’, no problems paying for accommodation

Odds ratio P value

Bottom income quintile 0.946 0.204

Second income quintile 0.973 0.476

Third income quintile 0.951 0.127

Fourth income quintile 0.983 0.524

Doing OK financial 0.834 0.000

Just financial 0.639 0.000

Difficult financial 0.466 0.000

Very difficult financial 0.295 0.000

Problems paying for accommodation 0.839 0.000

Mean equivalent household net income for the individuals district and age group

1.000 0.566

Explanatory variables: Health Ordered Logit on Life satisfaction (All)

Base category: Subjective health good, not disabled, no problems walking

Odds ratio P value

Subjective health poor or very poor 0.502 0.000

Subjective health fair 0.717 0.000

Subjective health excellent 1.223 0.000

Problems walking 0.721 0.000

Disabled 0.903 0.092

Days hospital stays (excl. births) 0.928 0.027

Problems with sleep 0.650 0.000

Explanatory variables: Employment status Ordered Logit on Life satisfaction (All)

Base category: employed Odds ratio P value

Long term sick 0.616 0.000

Retired 1.087 0.203

Unemployed 0.728 0.000

Maternity leave 1.519 0.000

Self-employed 0.972 0.617

Family carer 0.910 0.099

Student 1.070 0.281

Government training 0.788 0.331

Other employment status 0.939 0.670

Explanatory variables: Education and demographics Ordered logit on Life satisfaction (All)

Base category: No education, not high burden carer Odds ratio

P value

Commercial qualification 1.240 0.223

O level 1.018 0.889

A level 1.176 0.201

Degree or higher education 1.143 0.282

Cares for someone > 50 hours a wk 0.786 0.024

Number children in household 12-18 0.935 0.012

Number children in household 5-11 0.973 0.349

Number children in household 0-4 0.949 0.107

Age 0.823 0.000

Age squared / 100 1.363 0.000

Age cubed / 1000 0.978 0.000

Results Broadly the same using for life satisfaction, GHQ, ‘not at all’

unhappy, and modelling as ordered variable or cardinal, and for males and females

Social relationships important Financial coping and health biggest knock to well-being Financial coping related but not identical to income

50% of top income, and 15% of bottom quintile ‘living comfortably’

14% of those in the top income quintile think they are just about getting by or are finding it difficult or very difficult financially.

Need to look at ‘real’ household income, accounting for the costs of living, where those ‘costs of living’ may be influenced by geographical costs differences, or individually held (and socially driven) expectations of ‘necessary’ expenditures.

Direction of causality?

Income SWBHealth SWB

Social relationships SWB

May be other factors influencing SWB and X’s e.g. life events like having sick child

Could address endogeneity using Instrument Variables but hard to think of truly exogenous variables

Likely that route from SWB to X’s slower. Some evidence that cheerfulness in college students predicted higher income 19 years later (Diener et al 2002)

Still confident people matter.. A lot

One very clear message from this range of different modelling techniques on different SWB measures is that the people around you really matter. Those times when we have people who we see regularly and who create a friendly and peaceful local environment are the times when we are more satisfied with our lives and less likely

to experience unhappiness.

But is this a policy concern?

Some indication of positive externalities, social connectedness not entirely private good - role for intervention

Indirect impact on social relationships of other policies (e.g. flexible labour force, hours worked) should be considered

Increasing income may reduce social connectedness.

Income quintile Mean social relationships

1 (bottom) 6.56

2 6.54

3 6.40

4 6.27

5 (top) 6.07

Fixed effects OLS on social relations scale When other variables are controlled for, the negative

relationship between income and social relationships scale remains.

Being in the bottom two income quintiles is significantly related to having a higher social relationships scale, although perceived inability to cope financially is associated with reduced social relationships

Conclusion

Can’t say income isn’t important to SWB, financially difficulties very important

Increased income may contribute to other sources of SWB e.g. health

BUT Policies which focus on increasing income risk

undermining other sources of SWB Policies shouldn’t confuse means to well-being as

ends