Chapter 2satisfaction (thus measured) on log income and nothing else
beyond age and gender, we are estimating the relationship
Life- satisfaction = α 1 log Income + constant
holding constant only gender and age. The coefficient α 1
estimated this way is 0.30— a similar figure, as we shall see,
to figures found around the world.^10 If this were the whole
story, it would mean that a doubling of income would in-
crease life- satisfaction by 0.21 points (since when income
doubles log income rises by 0.7).
But this is a maximum estimate, since other things were
not held equal in the equation. To do this, we need to esti-
mate the multivariate relationship.
Life- satisfaction = α 1 log Income + α 2 Qualifications + etc.
including the whole battery of adult outcomes, child out-
comes, and family variables. The results are shown in the
first column of Table 2.1, which reproduces Figure 1.1
from the last chapter— but with all the variables appear-
ing where possible in their natural units as labeled in the
table. The coefficient on log income now falls from 0.30
to 0.20, reflecting the correlation between income and
other determinants of life- satisfaction like mental health
or family background. Clearly some of these other vari-
ables (like parental education) are simply correlated with
income and they should be included in the equation: they
are “confounders.” But some other variables (like mental
health) may be affected by income and therefore can “me-
diate” the effect of income on life- satisfaction. To the extent
this is true their effect should not be removed, and those