Read an Excerpt
CHAPTER 1
Happiness over the Life Course: What Matters Most?
All human life is here.
— News of the World
The central aim of this book is to supply a perspective on what makes people happy — to make it possible to compare the importance of any one factor with any other. So, before we look at each factor in detail, let us try to see the wood for the trees — to discover what matters more and what matters less.
In this chapter we shall estimate the five sets of relationships discussed in the Introduction, using only two of our surveys. We shall first estimate relationships (1) to (4), using the British Cohort Study data (BCS) on children born in 1970. Then we shall estimate relationship (5), using data on the British cohort born mainly in the county of Avon in 1991–92. These are of course results for Britain, but, as we shall see in later chapters, they are typical of what is found across the advanced world.
The analysis in this chapter is purely cross-sectional, but we discuss panel estimation at length later on. (In panel estimation all effects are smaller, but the ranking of factors is generally similar.) Further explanation and discussion appears in later chapters. At this point the key lesson is the power of these studies to shed a completely new perspective on human life.
Interpreting the Results
Throughout the analysis we start from the huge variation in well-being in the population. This is large even within one country, and wider still across the whole human race. The graph above is based on the British Cohort Study and, like the figures that follow, uses observations on the sample at both ages 34 and 42. The standard deviation of life-satisfaction in the sample is 1.9 (on the scale of 0–10).
So what explains this variation? The method of explanation is called multiple regression. This provides estimates of an equation such as
Life-satisfaction = ([varies]1 × Income) + ([varies]2 × Education) + etc. (1)
where income is measured in dollars and education in years. From this type of equation we can predict that one extra dollar of income will increase life-satisfaction by [varies]1 points (measured as usual on the scale of 0–10). And likewise one extra year of education will produce an extra [varies]2 points of life-satisfaction. And so on.
This is essential knowledge if we are to compare the effects of alternative policies to raise well-being — by, for example, raising earnings, expanding education, reducing unemployment, improving health, and so on. In each case we need to know how many points of additional life-satisfaction result from each type of improvement.
A quite different issue is how far do inequalities in income, education, employment, health, and so on explain the huge variation in happiness shown in the diagram. In this case we have to take into account not only the effect of having extra income, which is measured by [varies]1, but also the extent to which income varies in the population. The most natural measure of such variation is the standard deviation (SD). So one natural measure of the variation in life-satisfaction produced by income inequality (other things equal) is [varies]2 SD (Income). And that amount of variation relative to the overall variation of life-satisfaction is what we shall call β1 where
β1 = α1SD (Income)/SD (Life-satisfaction)
And so on for each other factor.
These β-coefficients are partial correlation coefficients. They show the correlation of, for example, income and life-satisfaction, holding all else constant. They are also the coefficients in an alternative version of equation (1) in which all the variables are "standardized" — that is they are divided by their standard deviations. The standardized regression equation is now
Life-satisfaction = (ß1 × Income) + (ß2 × Education) + etc., (2)
where all the variables are italicized to show that they are standardized.
As we have said, these ß-coefficients are useful because they tell us how important the different factors are in explaining the overall variation in life-satisfaction. In fact, if the variation in life-satisfaction is measured by its "variance," we can split up the explained variance exactly into the sum of the squared ß-coefficients plus some other terms.
In some parts of the book we shall show a-coefficients and in others ß-coefficients, as appropriate. When we show ß-coefficients, we shall always indicate this in the table heading. If we have not shown it, this means that the regression is based on natural units (i.e., it shows a-coefficients). All of this is explained more fully in online Annex 1.
Every coefficient estimate is only approximate, but the true value is 95% likely to lie within two standard errors (s.e.'s) of the coefficient estimate. So the standard errors are shown in brackets after most of the coefficients. When any coefficient estimate has over 90% probability of being different from zero, the coefficient is printed in bold. Whenever we report an estimated equation, the results of the equation appear as a single vertical column of numbers.
Improving Adult Happiness
So what can we say about what determines the life-satisfaction of an adult? We begin with relationship (1), which includes the "proximate" determinants, as well as the more "distant" ones. In Figure 1.1 we focus only on the coefficients on the "adult outcomes," in order to see what can be done to improve life-satisfaction once someone is already an adult. (We turn later to what can be done when people are children.)
The dependent variable is life-satisfaction. We begin with economic influences. As Figure 1.1 shows, the logarithm of equivalized household income has some effect on life-satisfaction — similar in Britain to that found in most other countries. But it explains under 1% of the overall variance of life-satisfaction in the population, while all the factors we can identify together explain around 15% of the variance. The direct influence of educational qualifications is smaller still, though they do of course have further indirect influence, for example, through their effect on income. As important as income or education is whether or not you are unemployed.
We turn next to behavior. Being partnered makes a big difference. Equally, criminal behavior (measured by criminal arrests since 16) clearly leads to social exclusion and lower life-satisfaction.
Finally comes health, which involves mental as well as physical health. Mental illness is a specific diagnosable condition. It is one of many factors that can produce low life-satisfaction. They are not the same thing. For example women have on average more well-being than men but more mental illness. The most convincing measure of mental illness is one based on an actual diagnosis (and this is the measure that we mainly use in Chapter 6). However in the British Cohort Study we have to rely on 24 self-reported answers to questions. This is a weakness, and we therefore lag this measure (using the answers the individual gave to these 24 questions at ages 34 and 26) to remove the simultaneous effect of temporary mood on reports of mental health and life-satisfaction. Even so, the estimated effects of mental health are large and similar to the estimates in Chapter 6. Moreover they are, both here and later, larger than the explanatory power of physical health, as measured by the number of health conditions the person is experiencing.
So how can policy makers influence these proximate determinants of well-being? Clearly policies directed at adults are important — policies on poverty, adult education, employment, crime, family support, and health. But another vital, and complementary, approach is to intervene earlier, in childhood, in order to improve the later outcomes. This brings us to the more "distant" causes of human happiness — in childhood.
Which aspects of childhood should receive the most attention? There are broadly three main aspects of child development — intellectual (or cognitive), behavioral, and emotional. Intellectual development is about knowledge and task-oriented skills. Behavioral development is primarily about behavior to others. And emotional development is about how the child feels. Which of these is the most important as a predictor of subsequent life-satisfaction?
In Figure 1.2, we estimate relationship (2) showing how adult life-satisfaction is explained by life up to age 16, or in the case of intellectual performance the highest qualification obtained (including where relevant a university degree). Behavioral development is measured by 17 questions answered by the mother, and emotional development by 22 questions answered by the child and 8 by the mother. In the table we show the coefficients on the three dimensions of child development. As can be seen, the strongest childhood predictor of a satisfying adult life is emotional health in childhood. Less powerful predictors are intellectual development and behavior. These findings have obvious relevance to educational policy.
Finally, we can look further back using relationship (3) — to the effect of a person's family working its way through everything that followed (see Figure 1.3). For parents we look at economic status, labor-market activity, parenting style, family stability, and the mother's mental health. Parents' education is measured by their terminal age of full-time education, and equivalized family income is averaged throughout childhood. Father's unemployment is averaged throughout childhood, and so is mother's work. Parenting style is measured by parents' involvement with the child, and family stability by whether the parents were still together when the child was 16. Mother's mental health is based on 24 questions and is averaged throughout childhood.
As Figure 1.3 shows, most of these factors have similar predictive power, but two findings stand out. Whether the mother works or not has no significant effect one way or other on whether the child becomes a happy adult. This important finding is discussed at length in a later chapter. On the other hand the mental health of the mother turns out to be crucial.
The Causes of Adult Outcomes
Equations (1) to (3) give us a good idea of the effect of intervening at different stages in a person's life. But it is also important to understand what is going on inside the black box. How, for example, are all the different adult outcomes determined? Even if you do not think happiness is a valuable outcome, you may want to know how to affect adult income, education, employment, crime, family life, and health. Figure 1.4 shows how these adult outcomes are affected by the outcomes of childhood: it represents equation (4).
There is a very clear pattern. Intellectual development is the most powerful predictor of income, qualifications, and employment. Behavioral development is the best predictor of prosocial living and attachment to a partner. And emotional development is much the best predictor of mental and physical health. This is important because mental health is the strongest proximate influence on life-satisfaction, and therefore the aspect of childhood that best predicts adult mental health (i.e., childhood emotional health) is also a good predictor of adult life-satisfaction.
The Causes of Child Outcomes
Finally we can examine what determines the child outcomes themselves (equation 5). This is crucial. Childhood is not a dress rehearsal. It is life itself — to be lived to the full. So what produces a happy, emotionally healthy child?
In Figures 1.5 (a) and (b) we look at how each of the child outcomes at 16 depends on the experience of family and school. The analysis is based on the more detailed information provided by the Avon Study, which includes more data on family finances, parenting behavior, family conflict, and, crucially, schooling. All these variables are included in three separate multiple regressions for intellectual performance at 16, behavior at 16, and emotional health at 16. Intellectual performance relates to the point score at GCSE, behavior comes from the relevant parts of the Strengths and Difficulties Questionnaire (SDQ), and emotional health from the Short Mood and Feelings Questionnaire (SMFQ). All the determining variables are averaged over the child's life up to 16. The results of each regression are presented in Figures 1.5 (a) and (b).
In Figure 1.5 (a) we report the impact of the family variables on each of the three child outcomes. Strikingly, the determinants of intellectual performance are very different from the determinants of behavior and emotional health (which are much more similar).
We can begin with economic variables like family income and the family's financial problems. These are very important for intellectual performance, and much less so for behavior or emotional health. The same is true of the effects of parents' education.
We then come to the vexed question of how children are affected if their mothers work. Our results confirm the findings of other studies that, if their mothers work (except in their first year after the child is born), children on average do better in school, but their behavior suffers somewhat. There appears to be no significant effect on their emotional well-being.
Another key issue is how the parents relate to the child. Standard propositions are that parents should be (i) "involved" in the child's cognitive development and (ii) "authoritative," that is, warm but reasonably strict. In the Avon study we have good data on involvement (reading to the child, teaching the child, going on outings, singing to the child). This is good for all three outcomes. Unfortunately we do not have good data on authoritative parenting, but we can identify overauthoritarian and aggressive parenting (shouting and hitting). This is correlated with bad behavior and poor emotional health (though there may also be elements of reverse causation at work here).
The next issue is how the parents relate to each other. There is clear evidence that parental conflict produces badly behaved and unhappy children. So what about family break-up, which the British Cohort Study identified as important? The answer is that the measured effect of family break-up is largely a proxy for family conflict, which is highly correlated with it. But, where there is already conflict, does family break-up make things even worse for the children? As we show in Chapter 13, it depends how bad the conflict is. If the conflict is terrible, break-up helps; if the conflict is mild, break-up adds to the damage.
Finally how are children affected by the psychological make-up of the parents, and especially their mother? The mother's mental health matters relatively little for children's academic performance, but it matters greatly for their behavior and their emotional health. Their father's mental health generally matters less.
So parents matter. But what about schools? Many people think schools only affect academic performance and behavior, but probably not the emotional health of the child, since this depends so heavily on the family. This view is totally wrong. In the Avon study we know which primary school and which secondary school each child went to. So we can see in Figure 1.5 (b) what difference these schools made. The effect of schools is huge, holding constant the child's family background. Even at the age of 16 the primary school still had an enduring influence — and for behavior and emotional health it had as great an influence as the secondary school.
It might be interesting to compare the importance of schools with that of parents. But we cannot do this because, while each school has many children in the sample, making it possible to identify its average effect on all those children, each parent had only one child in the sample. However we can summarize the overall effect of those family characteristics that we can identify. As the graphs show, the size of this effect is similar to that of the secondary school — meaning that the true effect of parents must be larger. One should add of course that this includes the effect of the genes they share with their children.
So much for the determinants of outcomes at age 16. But childhood is an ongoing experience. It is therefore interesting to look also at the determinants of outcomes earlier in childhood — at 5 and at 11. These are shown in the online Full Tables 10.1–10.3. The determinants are very similar to those we have seen at age 16.
(Continues…)
Excerpted from "The Origins of Happiness"
by .
Copyright © 2018 Princeton University Press.
Excerpted by permission of PRINCETON UNIVERSITY PRESS.
All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.
