Chapter 5

THE KEY DETERMINANTS OF

HAPPINESS AND MISERY

ANDREW E. CLARK, SARAH FLÈCHE, RICHARD LAYARD,

NATTAVUDH POWDTHAVEE AND GEORGE WARD

Andrew E. Clark, CNRS Research Professor, Paris School of Economics and Professorial Research
Fellow, Well-Being Programme, Centre for Economic Performance, London School of Economics and
Political Science

Sarah Flèche, Research Economist, Well-Being Programme, Centre for Economic Performance,
London School of Economics and Political Science

Richard Layard, Well-Being Programme, Centre for Economic Performance, London School of Economics
and Political Science

Nattavudh Powdthavee, Professor of Behavioural Science at Warwick Business School and Associate
of the Well-Being Programme, Centre for Economic Performance, London School of Economics and
122 Political Science

George Ward, PhD Student, Massachusetts Institute of Technology and Associate Research Economist
of the Well-Being Programme, Centre for Economic Performance, London School of Economics and
Political Science

This paper draws heavily on the December 2016 draft of The Origins of Happiness: The Science of Well-being over the
Life-Course by Clark et al., to be published by Princeton University Press. See also Layard et al. (2014). Support from the US
National Institute on Aging (Grant R01AG040640), the John Templeton Foundation and the What Works Centre for Well-be-
ing is gratefully acknowledged, as is the contribution of all the survey organisations and their survey participants.
WORLD HAPPINESS REPORT 2017

This chapter is directed at policy-makers of all further up the scale. This is important if, as
kinds—both in government and in NGOs. We many believe, it is more important to reduce
assume, like Thomas Jefferson, that “the care of misery than to increase happiness by an equal
human life and happiness … is the only legiti- amount further up the scale.
mate object of good government.”1 And we
assume that NGOs would have similar objectives. We have identified five major surveys of adults
In other words, all policy-makers want to create that make possible such analyses and also include
the conditions for the greatest possible happi- meaningful measures of mental health. They
ness in the population and, especially, the least cover the USA, Australia, Britain (two surveys)
possible misery. and Indonesia. We would like to have covered
more countries, but the data are not yet there.
For this purpose they need to know the causes
of happiness and misery. Happiness is caused by
many factors, such as income, employment,
health and family life and we need to ask, How Life Satisfaction
much does a difference in each of these factors
The measure of happiness that we use is life
change the happiness of the person affected?
satisfaction. The typical question is “Overall how
satisfied are you with your life these days?”
There is also a prior and related question that measured on a scale of 0 to 10 (from ‘extremely
tries to explain the huge variation in levels of dissatisfied’ to ‘extremely satisfied’).
happiness within any country. The question is
How far does the variation in each of the factors
This is a democratic criterion—we do not rely on
(e.g. income inequality) explain the overall
researchers or policy-makers to give their own
variation of happiness?
weights to enjoyment, meaning, anxiety, depres-
sion, and the like. Instead we leave it to individuals
In this chapter we concentrate mainly on the to evaluate their own well-being.
latter question.2 We begin by looking at the
role of current circumstances, and then (in the
Moreover, policy-makers like the concept—and so
second part of the chapter) examine the influence
they should. Our work shows that in European
of earlier childhood experience.
elections since 1970, the life satisfaction of
the people is the best predictor of whether the
To be useful to policy-makers, any analysis of the government is re-elected—much more import-
causes of happiness and misery should satisfy ant than economic growth, unemployment or
at least three criteria, which have not generally inflation (see Table 5.1).
been satisfied in the literature.
The task is thus to explain how all the different
1. It must use a consistent measure of happiness factors affect our life satisfaction, entering them
throughout. all simultaneously in the same equation.
123
2. It must look at the effect of all the factors
affecting happiness simultaneously, not one
by one.

3. It must check whether the factors have the

same effect on misery as they do on happiness
 Table 5.1. Factors Explaining the Existing When we are adults, our happiness depends
Government’s Vote Share significantly on our adult situation—our eco-
(Partial correlation coefficients)3 nomic situation (our income, education and
employment), our social situation (whether we
have a partner and whether we are involved
Life satisfaction 0.64 in crime), and our personal health (physical
Economic growth 0.36 and mental). These in turn depend partly on
Unemployment -0.06 our development as children (intellectual,
Inflation 0.15 behavioural and emotional), which in turn
depend on family and schooling. As our results
Source: Ward (2015).
Notes: Eurobarometer data on life satisfaction and standard show, there is scope for policy to affect a
election data for most European countries since the 1970s. The person’s development at every age.
regressors include the government’s vote share in the previous
election. Life-satisfaction is from the latest survey before the
election. Other variables are for the year of the election.

The Effects of the Current Situation

The Life-Course We begin with the impact on adult happiness of

the person’s current situation, using the follow-
In explaining our current life satisfaction, there ing data:4
are of course immediate influences (our current USA: Behavioural Risk Factor Surveillance
situation) but also more distant ones going back to System (BRFSS) (sample aged 25+)
our childhood, schooling and family background. Australia: Household, Income and Labour
This diagram gives a stylised version of how our Dynamics in Australia (HILDA) Survey (sample
life satisfaction as an adult is determined. aged 25+)
Britain: British Cohort Study (BCS) (surveyed at
ages 34 and 42)
Figure 5.1. Determinants of Adult Life Satisfac- Britain: British Household Panel Survey (BHPS)
tion (sample aged 25+)
Indonesia: Indonesian Family Life Survey (IFLS)
(sample aged 25+)

The factors we examine are

• Income: log household income per
equivalised adult
• Education: years, except Indonesia
(higher education versus none)
• Unemployment: measured as ‘not unemployed’
• Partnership: married, or living as married
• Physical health: USA, Britain and Indonesia:
124 number of illnesses; Australia: SF36, lagged
one year
• Mental health: USA and Australia: has ever
been diagnosed for depression or an anxiety
disorder; Britain (BCS): has seen a doctor in
the last year for emotional problems; Britain
(BHPS): GHQ-12, lagged one year; Indonesia:
replies to 8 questions.
WORLD HAPPINESS REPORT 2017

Most earlier analyses of life satisfaction have How far does each factor explain the variation in
not included mental health as a factor explain- life satisfaction within the population? Table 5.2
ing life satisfaction. The reason is that both life shows the results of regressing life satisfaction
satisfaction and mental health are subjective on all the factors simultaneously. The coeffi-
states, and there is therefore a danger that the cients given are partial correlation coefficients,
two concepts are, at least in part, measuring which show how far the independent variation
the same thing. To omit mental health as a of each factor explains the overall variation.5
factor in the equation, however, is to leave out
one of the most potent sources of misery, in
addition to standard external causes like
poverty, unemployment, and physical illness.
The solution is, whenever possible, to record
only mental illness that has been diagnosed or
has led to treatment. That is our approach and
it shows clearly that mental illness not caused
by poverty, unemployment or ill health is a
potent influence on life satisfaction.

Table 5.2. How Adult Life Satisfaction is Predicted by Adult Outcomes

(Partial correlation coefficients)

USA Australia Britain BCS Britain BHPS Indonesia

Income (log) 0.16 (.00) 0.09 (.01) 0.08 (.01) 0.09 (.01) 0.18 (.03)
Years of education 0.05 (.01) -0.03 (.01) 0.03 (.01) 0.02 (.00) 0.05 (.01)
Not unemployed 0.05 (.00) 0.04 (.01) 0.03 (.01) 0.06 (.00) 0.02 (.01)
Partnered 0.34 (.01) 0.14 (.01) 0.21 (.01) 0.11 (.00) 0.04 (.01)
Physical illness -0.05 (.00) -0.17 (.01)* -0.06 (.01) -0.11 (.00) -0.07 (.01)
Mental illness -0.21 (.00) -0.18 (.01) -0.11 (.01) -0.32 (.00)* -0.07 (.01)
Female 0.08 (.00) 0.08 (.01) 0.11 (.02) 0.05 (.00) 0.07 (.01)

N 268,300 16,001 17,812 139,507 31,437

Sources: USA (BRFSS); Australia (HILDA); Britain (BCS); Britain (BHPS); Indonesia (IFLS).
Notes: See Appendix C. * Lagged one year.

125
 In all three Western countries, diagnosed mental different points on the scale? For example, how
illness emerges as more important than income, well does Table 5.2 explain whether a person is
employment or physical illness. In Indonesia as really unhappy? To answer this we identify in
well, mental health is important, though less so each country people in the lowest levels of
than income. In every country physical health is happiness, which we call “In Misery.” Because
of course also important, but in no country is it happiness is measured in discrete units, the
more important than mental health. percentage identified as ‘In Misery’ varies from
5.6% in the USA to 13.9% in Indonesia.
Having a partner is also a crucial factor in
Western countries, while in Indonesia it is less We then run a standardised linear regression of
so, perhaps reflecting the greater importance of the dummy variable ‘In Misery’ on the same
the extended family. Education has a positive explanatory variables as before.7 The results are
effect in all countries (except Australia), yet it is shown in Table 5.3, where they are compared
nowhere near the most powerful explanatory with our previous results in Table 5.2 for the full
factor on its own.6 In every country, income is range of life-satisfaction. The two sets of coeffi-
more important than education as such. cients are remarkably similar. There is thus no
evidence that income, mental health, or any
At this point a natural question is Do different other variable is any more important lower down
variables impact differently on life-satisfaction at the well-being scale than it is higher up.

Table 5.3. Explaining the Variation of Life Satisfaction and of Misery Among Adults
(Partial correlation coefficients)

USA Australia Britain BCS Britain BHPS Indonesia

Life Sat Misery Life Sat Misery Life Sat Misery Life Sat Misery Life Sat Misery
Income (log) 0.16 -0.12 0.09 -0.09 0.08 -0.05 0.09 -0.07 0.18 -0.17
Years of education 0.05 -0.04 -0.03 -0.00 0.03 -0.02 0.02 -0.01 0.05 -0.06
Not unemployed 0.05 -0.06 0.04 -0.06 0.03 -0.03 0.06 -0.07 0.02 -0.03
Partnered 0.34 -0.19 0.14 -0.10 0.21 -0.11 0.11 -0.08 0.04 -0.04
Physical illness -0.05 0.05 -0.17 * 0.16* -0.06 0.05 -0.11 0.09 -0.07 0.07
Mental illness -0.21 0.19 -0.18 0.14 -0.11 0.09 -0.32 * 0.26* -0.07 0.08
Female 0.08 -0.06 0.08 -0.06 0.11 -0.06 0.05 -0.04 0.07 -0.06

Sources: USA (BRFSS); Australia (HILDA); Britain (BCS); Britain (BHPS); Indonesia (IFLS)
Notes: See Appendix C. * Lagged one year.

126
WORLD HAPPINESS REPORT 2017

In many ways a more vivid way of analysing By contrast, eliminating poverty in the USA
misery is to make all the right hand variables into reduces misery by 1.7% points, unemployment
discrete variables, such as poor/non-poor or sick/ by 0.3% and physical illness by 0.5% out of the
non-sick. This enables us to give an exact answer total 5.6% in misery. Taken together, those three
to the question If we could eliminate each factors barely make as much difference as
problem, how much could we reduce misery? mental illness on its own.

The different risk factors are now as follows: The pattern in Australia is very similar, but
Poor: below 60% of the median household with more problems coming from physical
income illness. In Britain the role of poverty is less
Uneducated: USA and Indonesia: no higher than it is in the USA, but the role of mental
education; Australia and Britain (BHPS): less health as large or larger.
than 10 years of education; in Britain (BCS): no
qualification Finally in Indonesia, eliminating mental illness
Unemployed again reduces misery by more than reducing
Not partnered poverty does. Further, increased education would
Physical illness: below the current 20th percen- also greatly help. In all countries there would be
tile of physical health much less misery if fewer people were living on
Depression/anxiety: diagnosed/treated except their own.
Britain (BHPS) and Indonesia (below the 20th
percentile).
This set of results is repeated, for effect, in
We then estimate an equation of the form Figure 5.3.

Is miserable (1,0) = a1 Is poor (1,0) +

a2 Is uneducated (1,0) + etc. (1)

The results are given in Table 5.4, column (1).

This shows that in the USA, for example, a
person who is poor is 5.5 percentage points more
likely than otherwise to be miserable. By contrast
someone with depression or anxiety is 10.7
percentage points more likely to be miserable.

So how much could we reduce the prevalence

of misery in the USA if we could miraculously
abolish depression and anxiety disorders
without changing anything else? Well, around
22% of the population have this diagnosis.
127
If they were all cured, we could reduce the
percentage of the population in misery by 0.107
times 22%. This is 2.35% of the whole population
(see column 3). That is large portion of the total
5.6% who are in misery.
 Table 5.4. How Would the Percentage in Misery Fall if Each Problem Could be Eliminated on its
Own?

α-coefficient × Prevalence = α × Prevalence Total in misery
(%) (% points)

USA
Poverty (below 60% of median income) 0.055 × 31 = 1.71
Uneducated (no higher education) 0.012 × 11 = 0.13
Unemployed 0.079 × 4.0 = 0.32 5.6
Not partnered 0.034 × 43 = 1.46
Physical illness (bottom 20%) 0.027 × 20 = 0.54
Depression or anxiety, diagnosed 0.107 × 22 = 2.35
Australia
Poverty (below 60% of median income) 0.044 × 30 = 1.32
Uneducated (below 10 years of educ.) 0.017 × 13 = 0.22
Unemployed 0.096 × 3.0 = 0.29 7.0
Not partnered 0.047 × 37 = 1.74
Physical illness lagged (bottom 20%) 0.097 × 20 = 1.94
Depression or anxiety, diagnosed 0.098 × 21 = 2.06
Britain (BCS)
Poverty 0.025 × 30 = 0.75
(below 60% of median income)
Uneducated (no qualification) 0.009 × 19 = 0.17
Unemployed 0.059 × 2.2 = 0.13 8.0
Not partnered 0.049 × 47 = 2.30
Physical illness (bottom 20%) 0.017 × 20 = 0.34
Has seen a doctor for emotional health 0.155 × 14 = 2.17
problems in last year
Britain (BHPS)
Poverty (below 60% of median income) 0.028 × 29 = 0.81
Uneducated (below 10 years of educ.) 0.026 × 10 = 0.26
Unemployed 0.152 × 3.8 = 0.41 9.9
Not partnered 0.053 × 36 = 1.90
Physical illness (bottom 20%) 0.057 × 20 = 1.14
Emotional health symptoms lagged 0.205 X 20 = 4.10
(bottom 20%)
Indonesia
Poverty (bottom 20%) 0.063 X 20 = 1.26
Uneducated (no qualification) 0.055 X 27 = 1.48
Unemployed 0.152 X 01 = 0.15 13.9
Not partnered 0.044 X 30 = 1.32
Physical illness (bottom 10%) 0.071 X 10 = 0.71
128 Emotional health symptoms (bottom 20%) 0.078 X 20 = 1.56

Sources: USA (BRFSS); Australia (HILDA); Britain (BCS); Britain (BHPS); Indonesia (IFLS).
Notes: People aged 25+, except for Britain (BCS) where people aged 34 and 42. The first column consists of regression coeffi-
cients in equation (1). For Indonesia the bottom quintile of the number of physical illnesses had much less explanatory power
than the composite variable used for Indonesia throughout this chapter—see Online Annex. See also Appendix C.
WORLD HAPPINESS REPORT 2017

Figure 5.3. How Would the Percentage in Misery Fall if each Problem Could be Eliminated on its
Own?

FPO

Sources: USA (BRFSS); Australia (HILDA); Britain (BHPS); Indonesia (IFLS)

From Figure 5.3 we can see how much misery Table 5.5. Average Cost of Reducing the Num-
could be reduced if we eliminated each of the bers in Misery, by One Person. Britain
risk factors, one at a time. But clearly none of
them can be totally eliminated. Moreover the cost £k per year
of reducing them is also relevant. So a natural Poverty. 180
question to ask in each country is If we wanted to Raising more people above the
have one less person in misery, what is the cost poverty line
of achieving this by different means? We attempt Unemployment. 30
Reducing unemployment
a very rough calculation of this for Britain in by active labour market policy
Table 5.5. As Table 5.5 shows, it costs money to Physical health. 100 129
reduce misery, but the cheapest of the policies is Raising more people from the
treating depression and anxiety disorders. worst 20% of present-day illness
Mental health. 10
Treating more people for
depression and anxiety

Sources available from authors.

 The Effects of Childhood Figure 5.4. How Adults’ Life Satisfaction
is Affected by Different Aspects of their
Importantly, many of the problems of adulthood Development as Children: Britain.
can of course be traced back to childhood and (Partial correlation coefficients)
adolescence. So which aspects of child develop-
ment best predict whether an adult is satisfied
with life? Answering this question requires
cohort data which are available for many fewer
countries. Since Britain is rich in such data, we
shall from now on use data on Britain only. We
first use data from the British Cohort Study,
which has followed children born in 1970 right
up to today. Sources: Britain (BCS)
Notes: Qualifications is the highest qualification that the per-
son achieved. Behaviour at 16 is reported by the mother, and
Three key dimensions of child development are at emotional health at 16 is reported by mother and child.
work. One is intellectual development, which we
measure by the highest qualification that the
individual achieved. This is turned into a single
variable using weights derived by regressing Clearly both parents and schools affect a child’s
wages on highest qualification. A second dimen- development. How, first, do parents affect their
sion is behavioural, measured in the Rutter children’s development? We have a mass of
behaviour questionnaire by 17 questions answered information about parents, and in Table 5.6 we
by the mother. The third dimension is emotional show the family variables that have the main
health based on a malaise inventory (22 questions effects. As is well known, family income has a
answered by the child and 8 by the mother). substantial effect on a child’s academic perfor-
mance, but a much smaller effect on the child’s
We now regress adult life-satisfaction on these emotional health and behaviour. Father’s unem-
three variables, as well as on family background. ployment has adverse effects, but is not that
As Figure 5.4 shows, the strongest predictor of a common. What is the effect if the mother goes
satisfying adult life is not qualifications but a out to work? If this happens in the first year,
combination of the child’s emotional health and there are on average very small negative effects.
behaviour.8 These findings have direct relevance If the mother works in subsequent years, howev-
to policy. er, it is positively beneficial for academic perfor-
mance and further does no measured harm to
But what, in turn, determines child development? the child’s emotional health.
To study this we use a very detailed survey of
all children born in the English County of Avon As regards “parenting style,” parental engage-
in 1991/2 who have been followed intensively ment and involvement with their children (e.g.
up until today. Our aim is to explain the three in reading and play) is immensely valuable,
130 measures of child development. Intellectual while aggressive parenting (hitting or shouting)
development is now measured by GCSE scores. only exacerbates bad behaviour. Conflict between
The emotional health of the child, however, has parents is especially disadvantageous for the
particular significance, since it is also the best behaviour of the children. The worst thing of all
measure we have of the child’s own quality of for children’s emotional health and behaviour is
life—it is a final product as well as an input into a mother who is mentally ill. Indeed, the survey
the resulting adult. suggests strongly that the mother’s mental
health matters more than the father’s.9
WORLD HAPPINESS REPORT 2017

Table 5.6. How Child Outcomes at 16 are Affected by Different Factors: Britain.
(Partial correlation coefficients)

Emotional Behavioural Intellectual

Family income 0.07 (.02) 0.08 (.02) 0.14 (.01)

Father’s unemployment -0.04 (.03) -0.00 (.02) -0.03 (.01)
Mother worked in 1st year -0.02 (.02) -0.01 (.02) -0.02 (.01)
Mother worked thereafter -0.01 (.02) -0.05 (.02) 0.04 (.01)
Parents’ involvement 0.04 (.02) 0.05 (.02) 0.02 (.01)
Aggressive parenting -0.03 (.02) -0.12 (.02) -0.01 (.01)
Family conflict -0.04 (.02) -0.14 (.02) -0.01 (.01)
Father’s mental health 0.04 (.02) -0.00 (.02) -0.00 (.01)
Mother’s mental health 0.16 (.02) 0.17 (.02) 0.03 (.01)

Source: Britain (ALSPAC)

Note: See Appendix C.

Clearly, family matters. What about the effect of Figure 5.5. How Child Outcomes at 16 are
schools? In the 1960s, the Coleman Report in Affected by Family and Schooling: Britain.
the US told us that parents mattered more than (Partial correlation coefficients)
schools.10 Since then the tide of opinion has
turned. Our data strongly confirm the importance
of the individual school and the individual
teacher. This applies equally to the academic per-
formance of the pupils and to their happiness.

In Figure 5.5, we look at child outcomes at 16 and

show how they are explained. The top bar shows
the combined effect of all observed family factors
(treated as a single weighted variable). The next
bar shows the enduring effect of the primary
school a child went to (again a single aggregate of
dummy variables), and the last is the effect of the
secondary school.11 These are big effects. 131

Source: Flèche (2016). ALSPAC data.

Notes: See Appendix C.
 Behaviour and Crime Table 5.7. How the Number of Crimes Commit-
ted by an Individual up to Age 34 is Affected by
We have so far focussed exclusively on the Child Development: Britain.
happiness of the individual person being studied.
But each of us also has a marked impact on the
happiness of other people. This social impact Qualifications (1 SD improvement) -0.87
has been given insufficient weight in much of Behaviour (1 SD improvement) -0.25
the literature on happiness, although it is well Emotional health (1 SD improvement) -0.04
known that how others behave is a major influ-
Source: Britain (BCS).
ence on our own happiness. Notes: Controls for family background, gender and age dummies.

So we must modify Figure 5.1 to take this into

account (see Figure 5.6). Unfortunately, however,
we have only limited ability to study this import-
ant determinant of the well-being of human
populations. One route is by inter-country Using data on local crime rates from police
comparisons of the type developed in Chapter 2 records, together with the corresponding local
of this report. The other is by studying the happiness data from the British Household
effects of crime on individual happiness, and Panel Survey, we can infer that each crime on
then investigating the determinants of criminality. average reduces the aggregate life-satisfaction of
the local population by the equivalent of 1
point-year for one person.

If we then look at how child development affects

Figure 5.6. The New Element: Behaviour crime, we find that the number of crimes a
person commits is affected by child development,
as shown in Table 5.7. Thus more education has
a major benefit through the resulting reduction
of crime. From one standard deviation of qualifi-
cations comes a one-off benefit to the rest of the
population of just under 1 point-year of life-satis-
faction (1 x 0.87). This can be compared to the
gain to the educated individual of 0.10 point-
year in every year of their life, as discussed
earlier. Thus the crime-reducing effect of educa-
tion adds proportionately little to the total social
returns to education.

132
WORLD HAPPINESS REPORT 2017

Social Comparisons It is really no mystery, however.12 There is much

evidence that people compare their income with
There remains the elephant in the room—social other people and, if others become richer, they
comparisons. People are constantly amazed that feel less happy at any given level of income.13
aggregate happiness has not risen in the USA This is confirmed in the present study. Table 5.8
and many other countries, when incomes and shows the effect of the average of log income in
educational levels have risen so much and one’s region, age-group and gender upon one’s
when income and education are associated with own happiness. In all three countries the nega-
greater individual happiness. This is the Easter- tive effect of others’ income is large, and any rise
lin paradox. in overall income has little effect on overall life
satisfaction. The same is true for education.

Table 5.8. How Life Satisfaction (0-10) is Affected by Own Income, Comparator Income, Own Years
of Education and Comparator Years of Education
(Partial correlation coefficients)

Britain (BHPS) Germany Australia

Own income (log) 0.16 (.01) 0.26 (.01) 0.16 (.01)

Comparator income -0.15 (.07) -0.34 (.05) -0.13 (.06)
Years of education 0.03 (.00) 0.05 (.00) -0.01 (.00)
Comparator education -0.09 (.02) -0.05 (.01) -0.03 (.01)

Notes: Robust standard errors in parentheses. Controls for self-employed, employed part time, unemployed, not in labour force,
partnered, separated, widowed, parent, physical health, emotional health, female, age, age-squared, comparator unemployment,
comparator partnership, year and region dummies.

133
 Conclusion across the whole range? The answer is yes, and
the factors have the same ranking in explaining
Policy-makers need to know the causes of happi- misery as in explaining life-satisfaction. In Table
ness and misery. Some of these are factors that 5.4 we show a novel decomposition which
affect everyone in a society (see Chapter 2), while illustrates how much misery could in principle
other vital factors differ across individuals. For the be eliminated by eliminating either poverty, low
latter, policy-makers need to know what factors education, unemployment, living alone, physical
account for the huge variation across individuals illness or mental illness. In all countries the
in their happiness and misery (both of these being most powerful effect would come from the
measured in terms of life-satisfaction). elimination of depression and anxiety disorders,
which are the main form of mental illness. This
Key factors include economic factors (such as would also be the least costly way of reducing
income and employment), social factors (such as misery (Table 5.5).
education and family life), and health (mental
and physical). We use surveys from the USA, While much could be done to improve human
Australia, Britain and Indonesia to cast light on life by policies directed at adults, as much or
the relative importance of these various factors. more could be done by focussing on children.
We examine this issue using British cohort data.
In all three Western societies, diagnosed mental We ask, Which factors in child development best
illness emerges as more important than income, predict whether the resulting adult will have a
employment or physical illness. In Indonesia as satisfying life? We find that academic qualifica-
well, mental health is important, though less so tions are a worse predictor than the emotional
than income. In every country, physical health is health and behaviour of the child.
also of course important. Yet in no country is it
more important than mental health. What in turn affects the emotional health and
behaviour of the child? Parental income is a
Having a partner is also a crucial factor in good predictor of a child’s academic qualifications
Western countries, while in Indonesia it is less (as is well known), but it is a much weaker
so, perhaps reflecting the greater importance of predictor of the child’s emotional health and
the extended family. Education has a positive behaviour. The best predictor of these is the
effect in all countries (except Australia) but it is mental health of the child’s mother.
nowhere near the most powerful explanatory
factor on its own. In every country, income is Schools are also crucially important. Remark-
more important than education as such. ably, which school a child went to (both primary
and secondary) predicts as much of how the
Even so, household income per head explains child develops as all the characteristics we can
under 2% of the variance of happiness in any measure of the mother and father. This is true of
country. Moreover it is largely relative income what determines the child’s emotional health,
that matters, so as countries have become richer, their behaviour and their academic achievement.
134
many have failed to experience any increase in
their average happiness. A similar problem To conclude, within any country, mental health
relates to education—people care largely about explains more of the variance of happiness
their education relative to that of others. in Western countries than income does. In
Indonesia mental illness also matters, but less
What about the causes of misery? Do the same than income. Nowhere is physical illness a
factors affect misery as affect life-satisfaction bigger source of misery than mental illness.
WORLD HAPPINESS REPORT 2017

Equally, if we go back to childhood, the key

factors for the future adult are the mental
health of the mother and the social ambiance
of primary and secondary school. The implica-
tions for policy are momentous.

135
 1 Jefferson (1809). 12 F
 or an earlier discussion of the Easterlin paradox, see
WHR 2012, Chapter 3.
2 T
 he relation between these two questions is shown in
Appendix A, which provides data from which the answers 13 Clark et al. (2008); Layard et al. (2010).
to the previous question can be calculated.

3 The partial correlation coefficients are sometimes called the

standardised regression coefficients. They are the s in a
regression where all variables are divided by their standard
deviation. The overall explanatory power of the equation is
given by

4 D
 etails are in Appendix B and an online Annex at https://
tinyurl.com/WHR2017Ch5Annex

5 See Note 3.

6 T
 he total effect of education includes of course its effect via
income and other channels. If income is excluded from the
regression, the coefficient on education becomes USA
0.08, Australia 0.03, Britain BHPS 0.06, and Indonesia
0.06.

7 W
 e thus estimate a linear probability model. Almost
identical results are obtained from logit analysis.

8 T
 he coefficient for the combination of the child’s emotion-
al health and behaviour is 0.101 (s.e. = 0.009), which
compares with 0.068 (s.e. 0.008) for qualifications—a
significant difference (p = 0.010).

9 P
 resumably since she is more present. However the
mother’s mental health is measured 8 times up to when
the child is 11, while the father’s is only measured 3 times
until the child is 2. To see if this matters, we also focused
on explaining the child’s emotional health at 5, using three
observations on both parents’ mental health. The differ-
ence between the effect of mother and father remained as
large as it is in Table 5.6. The same occurred if we focussed
on explaining the child’s emotional health at 16, using only
the first three observations on each parent’s mental health.

The mother’s mental health was measured using the

Edinburgh Post Natal Depression Scale (EDPS), and the
father’s was tested using the Crown-Crisp Experiential
Index.
136 10 Coleman et al. (1966).

11 The dependent variable is regressed on two sets of dummy

variables, one for each primary school and one for each
secondary school. The set of primary school variables is
then turned into one composite variable using the
coefficients on each dummy variable. The same is done
for secondary schools.
WORLD HAPPINESS REPORT 2017

References
Clark, A. E., Flèche, S., Layard, R., Powdthavee, N., & Ward,
G. (forthcoming). The Origins of Happiness: The Science of
Wellbeing over the Life Course: Princeton University Press.

Clark, A. E., Frijters, P., & Shields, M. (2008). Relative

Income, Happiness and Utility: An Explanation for the
Easterlin Paradox and Other Puzzles. Journal of Economic
Literature, 46(1), 95-144.

Coleman, J. S., Campbell, E. Q., Hobson, C. J., McPartland, J.,

Mood, A. M., Weinfeld, F. D., & York, R. L. (1966). Equality of
Educational Opportunity. Washington, D.C.: Office of Educa-
tion, U. S. Department of Health, Education, and Welfare.

Flèche, S. (2016). Teacher Quality, Test Scores and Non-Cogni-

tive Skills: Evidence from Primary School Teachers in the UK.
CEP mimeo.

Jefferson, T. (1809). Letter to the Maryland Republicans: in

The Writings of Thomas Jefferson (1903-1904) Memorial Edition
(Lipscomb and Bergh, editors) 20 Vols., Washington, D.C: ME
16:359.

Layard, R., Clark, A. E., Cornaglia, F., Powdthavee, N., &

Vernoit, J. (2014). What Predicts a Successful Life? A Life-
Course Model of Wellbeing. Economic Journal, 124(580),
F720- F738.

Layard, R., Mayraz, G., & Nickell, S. J. (2010). Does relative

income matter? Are the critics right? In E. Diener, J. F.
Helliwell & D. Kahneman (Eds.), International Differences in
Well-Being (pp. 139-165). New York: Oxford University Press.

Ward, G. (2015). Is Happiness a Predictor of Election Results?

LSE Centre for Economic Performance.

137
 APPENDIX

138
WORLD HAPPINESS REPORT 2017

Appendix A: Calculating The Absolute Impact of A Factor

The equations presented in this chapter are of Thus

the form

The tables in the text provide the s. The follow-

where measures the standard deviation of the ing tables provide the s and the means.
variable. For cost-effectiveness analysis a policy-
maker needs the coefficients in the equation

Standard deviations for Tables 5.2, 5.3 and 5.8

USA Australia Britain BCS Britain BHPS Indonesia
Life satisfaction 0.62 1.49 1.90 2.36 0.80
Misery 0.22 0.26 0.27 0.35 0.35
Income (log) 0.82 0.88 0.74 1.22 7.86
Education 1.11 2.58 1.57 2.51 0.43
Not unemployed 0.20 0.21 0.14 0.21 0.08
Partnered 0.50 0.48 0.50 0.48 0.35
Physical illness 1.06 4.95 1.32 1.10 0.83
Mental illness 0.42 2.59 0.18 5.54 4.97
Female 0.50 0.50 0.50 0.50 0.50
Comparator income 0.40 1.07
Comparator education 1.17 0.97

Means for Tables 5.2, 5.3 and 5.8

USA Australia Britain BCS Britain BHPS Indonesia
Life satisfaction 3.40 7.90 7.39 6.97 3.32
Misery 0.06 0.07 0.08 0.10 0.14
Income (log) 9.99 7.51 9.55 6.42 15.75
Education 4.78 12.08 3.37 12.35 0.26 139
Not unemployed 0.96 0.67 0.97 0.96 0.99
Partnered 0.57 0.63 0.53 0.64 0.70
Physical illness 1.38 22.68 2.01 0.73 0.48
Mental illness 0.22 0.21 0.14 23.11 18.83
Female 0.50 0.53 0.52 0.55 0.51
Comparator income 7.64 6.22
Comparator education 12.07 12.19
 Appendix B: The Surveys Used

https://tinyurl.com/WHR2017Ch5Annex

USA (BRFSS) Behavioural Risk Factor Surveillance System (BRFSS)

Cross-sectional survey which includes a life-satisfaction question since 2005.
In 2006, 2008, 2010, 2013, respondents were asked whether they have ever been diagnosed with
depression or anxiety.
Sample size = 270,000
Australia Household Income and Labour Dynamics in Australia (HILDA) Survey
Household-based panel study which began in 2001. The panel members are followed over time and
interviewed every year. Life-satisfaction is measured throughout.
In 2007, 2009, 2013, respondents were asked whether they have ever been diagnosed with depression or
anxiety.
Sample size = 16,000
Britain British Cohort Study (BCS)
British cohort data which began in 1970. The children are followed over time and interviewed at ages 5,
10, 16, 26, 30, 34, 38 and 42. A life satisfaction question has been included in the study from age 26.
At ages 34 and 42, respondents were asked whether they have any physical health problems.
Sample size = 18,000
Britain British Household Panel Survey (BHPS)
Household-based panel study which began in 1991. The panel members are followed over time and
interviewed every year. A life satisfaction question has been included in the study from 1996.
Sample size = 140,000
Britain Avon Longitudinal Study of Parents and Children (ALSPAC)
Near census English cohort study. The study recruited over 14,000 pregnant women residing in the Avon
area in the UK with expected delivery dates between April 1991 and December 1992. The children have
been followed almost every year since then.
The study contains various measures of the family environment, schooling environment as well as
indicators of the development of child well-being and skills over time.
Sample size = 8,000
Indonesia Indonesia Family Life Survey (IFLS)
(IFLS) Longitudinal survey in Indonesia.
The fifth wave (ILFS-5) in 2014 includes a question on life satisfaction, emotional health and number of
health conditions diagnosed by a doctor.
Observations: 32, 000

140
WORLD HAPPINESS REPORT 2017

Appendix C: Notes on Tables and Figures

Table 2: How Adult Life-satisfaction is Predicted proportion of time mother worked thereafter,
by Adult Outcomes father’s unemployment, mother’s mental health,
Robust standard errors are in parentheses. father’s mental health, involvement, aggression,
Controls for age, age-squared, region and year family conflict, parental separation, parents’
dummies. Australia and Britain (BHPS) also education, mother’s age at birth, and parents’
include comparison income, education, unem- marital status at birth. Controls for female child,
ployment and partnership. Britain (BCS) also ethnicity, first born child, number of siblings,
includes non-criminality, child outcomes at 16 low birth weight, and premature baby.
and family background. Cross-section regres-
sions using information from BCS respondents
at ages 34 and 42. BHPS, HILDA, IFLS and
BRFSS respondents at age 25+.

Table 3. Explaining the Variation of Life-Satisfac-

tion and of Misery Among Adults
Controls for age, age-squared, region and year
dummies. Australia and Britain (BHPS), also
include comparison income, education, unem-
ployment and partnership. Britain (BCS) also
includes non-criminality, child outcomes at 16
and family background. Cross-section regres-
sions using information from BCS respondents
at ages 34 and 42. BHPS, HILDA and BRFSS
respondents at age 25+. Those included in
misery are USA 1-2 (on scale 1-4); Australia 0-5
(on scale 0-10); Britain (BCS) 0-4 (on scale
0-10); Britain (BHPS) 1-3 (on scale 1-7); and
Indonesia (IFLS) 1-2 (on scale 1-5).

Table 6. How Child Outcomes at 16 are Affected

by Different Factors: Britain.
Robust standard errors are in parentheses.
Controls for parental separation, parents’ educa-
tion, mother’s age at birth, parents’ marital
status at birth, female child, ethnicity, first born
child, number of siblings, low birth weight,
premature baby, and primary school and second- 141
ary school fixed effects.

Figure 5. How Child Outcomes at 16 are Affect-

ed by Family and Schooling: Britain.
Family background include family income,
proportion of time mother worked in first year,
 ONLINE APPENDIX
(CLARK, FLÈCHE, LAYARD, POWDTHAVEE, & WARD, THE KEY DETERMINANTS OF
HAPPINESS AND MISERY)

HTTP://WORLDHAPPINESS.REPORT/

142
WORLD HAPPINESS REPORT 2017

143