NBER WORKING PAPER SERIES
HOW MUCH IS SOCIAL CAPITAL WORTH?
John F. Helliwell
Christopher P. Barrington-Leigh
Working Paper 16025
NATIONAL BUREAU OF ECONOMIC RESEARCH
1050 Massachusetts Avenue
Cambridge, MA 02138
May 2010
We are grateful for invaluable access to the Gallup World Poll, and to Canadian GSS17 data through
the UBC Research Data Centre supported by Statistics Canada. This paper is part of the ‘Social Interactions,
Identity and Well-Being’ research program of the Canadian Institute for Advanced Research (including
a Junior Fellowship award for Chris Barrington-Leigh), and is supported also by grants from the Social
Sciences and Humanities Research Council of Canada. We have been much aided also by the advice
of Jolanda Jetten, Alex Haslam and Cath Haslam. The views expressed herein are those of the authors
and do not necessarily reflect those of the National Bureau of Economic Research.
NBER working papers are circulated for discussion and comment purposes. They have not been peer-
reviewed or been subject to the review by the NBER Board of Directors that accompanies official
NBER publications.
© 2010 by John F. Helliwell and Christopher P. Barrington-Leigh. All rights reserved. Short sections
of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full
credit, including © notice, is given to the source.
How Much is Social Capital Worth?
John F. Helliwell and Christopher P. Barrington-Leigh
NBER Working Paper No. 16025
May 2010
JEL No. A13,D03,D61,D62,I31,N30
ABSTRACT
This paper uses data from global and Canadian surveys data to estimate the powerful linkages between
social connections, their related social identities, and subjective well-being. Our explanatory variables
include several measures of the extent and frequency of use of social networks, combined with a number
of measures of general and domain-specific trust, which are often used to gauge effective social capital.
Using these measures we find that trust and social network size and use are all strong predictors of
subjective well-being. We demonstrate the size and impact of these effects by calculating compensating
differentials, measured as the changes in household income that would produce equivalent levels of
life satisfaction.
We introduce three key measures of social identity – the respondents’ sense of belonging to their communities,
province and country – and find that they add significantly to the explanation of life satisfaction among
Canadian respondents, and provide important mediating channels whereby social capital is linked
to subjective well-being.
John F. Helliwell
Canadian Institute for Advanced Research
and Department of Economics
University of British Columbia
997-1873 East Mall
Vancouver . V6T 1Z1 CANADA
and NBER
@
Christopher P. Barrington-Leigh
Canadian Institute for Advanced Research
and Department of Economics
University of British Columbia
997-1873 East Mall
Vancouver . V6T 1Z1 CANADA
cpbl@
An online appendix is available at:
1
Introduction
In this paper we use large-sample global and Canadian survey data to reveal the powerful
effects that social connections and related social identities have on subjective well-being.
Our evidence is based on several measures of the extent and frequency of use of social
networks, combined with a number of measures of general and domain-specific trust,
which are often used to gauge effective social capital. Using these measures we find that
trust and social network size and use are all strong predictors of subjective well-being and
we demonstrate the size and impact of these effects, by calculating their value in terms of
income changes that would produce equivalent levels of life satisfaction.
We introduce three key measures of social identity – the respondents’ sense of belonging
to their communities, province and country – and find that they add significantly to the
explanation of life satisfaction among Canadian respondents, and provide important
mediating channels whereby social capital is linked to subjective well-being.
How can the social context be valued?
Our conceptualisation of ‘social capital’, consistent with that of Putnam (2000), is
based on the OECD’s definition (2001, ) as “networks together with shared norms,
values and understandings that facilitate co-operation within or among groups”. Can the
importance of social capital be assessed comparably with income support, housing
quality, prescription drugs and other means used by individuals, families and policy-
makers to improve the well-being of themselves and others? In epidemiology, the
currency of choice has been human lives, with the benefits of medical and social
interventions measured in terms of lives saved. Mortality estimates are now frequently
supplemented by measures of morbidity, sometimes weighting different sorts of frailty to
construct measures of healthy life expectancy. Classic applications of the health-based
method to value the social context (. Berkman & Syme 1979) measured the value of
more extensive and happier social relations in terms of longer and healthier lives. An
even earlier example is provided by Durkheim’s (1897, 1952) magisterial analysis, in
2
which suicide was seen as the consequence of hopelessness felt in the face of perceived
failures of social relations (Williams, 2001). In the same vein, Helliwell (2007) found that
international differences in average social trust had large and significant effects in
explaining international differences in both suicides and traffic fatalities.
Linking the social context to mortality and morbidity in these ways remains
central to policy evaluation, and deservedly so. Even if the broader objective of the caring
professions, and of public policy in general, is to improve the quality of human lives
along many dimensions, it will probably always be the case that primary importance is
placed on saving lives and restoring health. A life that has been lost cannot be made
better, and all of the subjective life evaluations that provide the main evidence reported in
this paper show strong linkages to subjective and objective measures of health status.
Thus, the traditional measures of better health are not only important in their own right
but are important predictors of self-assessed well-being.
However, to assess the overall consequences of the social context requires a
broader framework in which good health and other aspects of a good life can be brought
together, and their values compared. Many have adopted an a priori approach to such an
evaluation. For example, Maslow (1943) argued for a hierarchy of human needs. His
pyramidal structure focussed first on basic needs and then on ‘higher order’ needs,
including those for social companionship. But even if one can measure how well needs
and desires at different levels are being met, how are we to compare changes in the
degree to which different needs are being met, whether they be at the same or different
levels?
Alternatively, Sen (1999) treats capabilities as the fundamental bedrock of human
development, represented by some basic set of freedoms from hunger, illness and fear,
and supported by enough education to enable personal and social progress. Yet this also
leaves the valuation issue unsolved, as policy-makers still need guidance as to the relative
values of improving capabilities at different times, and in different contexts.
3
Utilitarians over several centuries, inspired by Bentham and Mill, among many
others, have tackled the valuation issue by presuming some underlying if not-directly-
measurable concept of utility. Bentham proposed operationalizing utility in terms of a
form of hedonic calculus. By this means a balance of pleasure and pain, measured
continually and then aggregated through time, would be used to assess each person’s
experienced utility. In the intervening centuries, many social scientists, and especially
economists, have taken a different tack. Assuming that utility could not be directly
measured, they instead studied human decisions — believing that these could be used to
reveal some features of the underlying utility functions. For example, economists since
Adam Smith have used the wages for different jobs (after adjusting as well as possible for
differences in skills) to estimate the dis-utility of unpleasant or dangerous jobs, in the
process estimating what have been often called ‘compensating differentials’. In the
literature, compensating differentials are the wage differences required to offset good or
bad characteristics of a job. As will be seen shortly, the same term is now used in the
well-being literature to place an income-related value on various features of the social
context.
Beyond the indirect revealed preference approach long favoured by economists,
Kahneman and colleagues (Kahneman 1999; Kahneman, Wakker & Sarin, 1997) have
argued for a more direct application of Bentham’s hedonic calculus. More precisely, they
take the sum of the experienced balance of pain and pleasure to be the underlying
measure of utility. Because of the invasiveness and costs of continual monitoring, they
have recommended periodic sampling involving relatively short-term memories of moods
and emotions1. The distinction between experienced and remembered utility has been
given especial importance by Kahneman’s research. For example, this research shows
that the average of moment-by-moment assessments of the painfulness of a colonoscopy
is not the same as the remembered pain of the procedure, with the latter best modelled as
the average of the peak and final moments of pain (Redelmeier, Katz, & Kahneman
2003). Similarly, Wirtz, Kruger, Scollon and Diener (2003) found that moment-by-
1 Kahneman and Krueger (2006) compare experience sampling and their proposed Daily
Reconstruction Method, and discuss their proposed Benthamite U-index, a measure of the preponderance of
unpleasant emotions over the previous day.
4
moment assessments of mid-term break holidays did not closely match the overall
memories of the holiday experience. Kahneman has tended to follow Bentham explicitly,
and argued that moment-by-moment measures are direct measures of utility and that
remembered measures embody cognitive errors (Kahneman & Riis 2005). Others
(including us) argue that the differences between experiences and the memories of them
are not errors, but are instead of basic importance. Consistent with this claim,
remembered experiences have been found to govern subsequent decisions about holidays
and medical procedures, and to represent the long-lasting residual effect underpinning life
evaluations. In any event, reports of moods can be used as additional direct measures of
utility, and provide an alternative basis for valuing the social context. Our evidence
suggests, however, that life circumstances, including income and key elements of social
capital, have much more precisely estimated effects on life evaluations than on either
current or average moods.
A long time before the utilitarians, however, Aristole and others made a case for
remembered life evaluations, arguing that ethics “gets its grip on the individual at this
point of reflection: am I satisfied with my life as a whole, and the way it has developed
and promises to develop?” (Annas 1993, ). Aristotle himself argued that the good life
followed a middle course between the Stoics, who put their whole emphasis on the
virtuous life, and the Epicureans, who emphasized the importance of pleasures, including
the avoidance of pain (Annas 1993, ). But he was also an early advocate for
empirical work on just these issues, and for taking his view of a good life “to the test of
the facts of life, and if it harmonizes with the facts we must accept it, but if it clashes with
them we must suppose it to be mere theory” (Nicomachean Ethics, Book 10, 1179, -
23.)
How should we choose between using the Benthamite U-index (Kahneman &
Krueger, 2006) or an Aristotelian life evaluation to provide the raw materials for valuing
social capital? There are three main distinctions between the U-index and measures of
life satisfaction. First, the U-index is based on moods, which are inherently more
changeable during the course of a day and are less tightly linked to changes in life
5
circumstances. Second, the U-index measures the fraction of a day during which bad
emotions predominated over good ones, with no further account taken of the strength of
the positive or negative feelings. Third, even though both are based on memory rather
than current experience, the Daily Reconstruction Method (DRM) underlying the U-
index directs and frames recollections in the context of a time-use survey. It therefore
provides more information about how specific daily activities (., commuting alone or
in company) affect moods during the day.
Life satisfaction questions, by contrast, ask about ‘life as a whole these days’. It is
clear, however, that even these broader assessments are responsive to changes in current
circumstances and other mood-affecting patterns of daily life. Thus the U-index and
measures of life satisfaction are both found to be correlated with many of the same
variables. The two types of data are both useful, but for different purposes, with the
former being most appropriate for unravelling the fabric of daily life and the latter for the
evaluation of broader and more enduring life circumstances. Thus, as their respective
proponents argue, the former can more easily be seen as an application of Bentham’s
hedonic calculus, and the latter as being truer to Aristotle’s advice for evaluating the
quality of life as a whole.
It is also worth noting that life satisfaction evaluations provide evidence
supporting Sen’s emphasis on capabilities, and Maslow’s hierarchy of needs. For
example, good health, freedom from hunger, and freedom to make one’s own life choices
all rank very high as determinants of life satisfaction, and education appears to play a
strong mediating role by facilitating better jobs and incomes, better health, and greater
trust and social engagement.
Two other practical reasons argue for using life satisfaction, or some similar form
of life evaluation (rather than some measure of affect, such as the U-index) as a guide to
measuring the value of social capital. The first is that measures of life satisfaction are
easier to collect, and have been monitored much more widely, more comparably, and
over longer time periods. The second and more fundamental reason is that because
6
measures of life satisfaction are more clearly responsive to basic life circumstances
(including both income and the quality of social capital) they provide the basis for much
more reliable estimates of the value of social capital.
Using Life Satisfaction Data to Value International Differences in Social Capital
In the studies that follow, the basic method used to value social capital involves
using large samples of individual measures of life satisfaction (usually ordered responses
on a 1 to10 or 0 to 10 scale) as dependent variables in an estimated utility function for an
average individual. In the most general form, we use two-level analysis in which
individual utility depends on a set of individual-level circumstances and assessments,
plus the same and/or other variables measured to reflect the characteristics of the
community or country in which the individual lives. The basic estimation form for two-
level analysis of the ordered life satisfaction responses is:
( ) ( )1 ln ,ij ij ij j ijLS y X Zα δ μ γ ε= + + + +
where LSij is some measure of life satisfaction, for respondent i in community or country
j, yij is the level of household income of the respondent, the Xij are other individual or
household-level variables, and the Zj are community, state, or national-level variables,
with the same value being used for all individual observations in geographic area j. We
use the log form for both household and national average income since most recent
evidence suggests, to a first approximation, that life satisfaction rises linearly with the
logarithm, and not the level, of household income (., Deaton 2008; Helliwell &
Huang, 2010; Helliwell, Barrington-Leigh, Harris & Huang, 2010). In some surveys
(Helliwell & Putnam, 2004), there is also evidence of an additional tailing off of marginal
life satisfaction effects of income at high-income levels, although this is not evident in all
surveys. Perhaps this finding reflects the fact that many surveys do not divide income
categories finely enough, or do not have large enough samples at the top end of the
income distribution. Where there are both individual and societal-level observations on
the same social capital variable, the γ coefficients on the societal variables represent
7
contextual effects, or, in other words, the extent of externalities. These contextual effects
could be either positive or negative. In the case of income, the effects of higher
neighbourhood average income are usually negative, and often significantly so (Luttmer
2005; Helliwell & Huang 2010; Barrington-Leigh & Helliwell, 2008). In the case of
social capital variables, the external effects are more generally positive (Helliwell &
Putnam, 2004).
Figure 1: Compensating differentials at individual and national
levels
absence of
corruption
freedom to
choose
friends to count
on
enough money to
buy food
Individual level National level
Figure 1 shows some example estimates of the value of social capital, based on
two-level analysis of a global sample of life evaluation data from three waves (between
2006 and 2008) of the Gallup World Poll. The coefficients underlying the compensating
differentials of Figure 1 are drawn from column 1 of Table 1 of Helliwell, Barrington-
Leigh, Harris and Huang (2010). That paper includes comparable estimates for equations
using three different life evaluations, each on a scale of zero to ten: satisfaction with life,
the Cantril self-anchoring striving scale, also known as the ladder of life, and a simple
average of each individual’s answers to both questions. The Cantril form of life
8
evaluation asks respondents to “Please imagine a ladder with steps numbered zero at the
bottom to ten at the top. Suppose we say that the top of the ladder represents the best
possible life for you and the bottom of the ladder represents the worst possible. If the top
step is 10 and the bottom step is 0, on which step of the ladder do you feel you personally
stand at the present time?” The satisfaction with life question (SWL) was: “All things
considered, how satisfied are you with life as a whole these days?” (using the same 0 to
10 score; for more details of these measures see the appendix to Helliwell and
Barrington-Leigh (2010) ).
Earlier analysis comparing the first wave of Gallup ladder data with earlier World Values
Survey data for life satisfaction showed quite differently shaped distributions, with the ladder
data having lower means, smaller right-hand shoulders, and more mass at the mid-points than
was the case for the World Values Survey data for life satisfaction (Helliwell, 2008). There was
also some evidence that the effects of income were greater on the ladder measure. These
differences led to the addition of the life satisfaction question to the Gallup World Poll, and to
experimental studies designed to see if the framing of the life satisfaction question could lead to
differences in the patterns of responses (Gleibs, Haslam, Morton, Rabinovich, & Helliwell,
2010). These experiments showed that increasing the salience of material aspects of life tended
to produce subsequent subjective life evaluations that in turn revealed higher correlations with
the respondents’ actual incomes.
This demonstrated potential sensitivity of coefficients to priming effects made it
especially valuable to have matched data for both life satisfaction and the Cantril scale.
Fortunately for our attempts to value the social context in terms of income, we found
very similar coefficients using all three dependent variables, thus giving almost identical
estimates of compensating differentials (Helliwell et al., 2010). This is quite remarkable,
because the distributions of the Cantril scale and the SWL continue to show significant
differences of the sort noted earlier: the national means of the Cantril scale average about
points lower than those for life satisfaction, with the difference taking the form of a
fuller right-hand shoulder in the life satisfaction distribution. It is encouraging to note that
these differences do not lead to conflicting stories from well-being equations estimated
9
from the two different distributions. Indeed, we found that by averaging the two measures
we increased the signal-to-noise ratio, as revealed by a significantly higher explained
share of total variance. The resulting R2 of is very high for any cross-sectional
equation for individual-level measures of subjective well-being, especially an equation
that has relatively few structural parameters and no country or region fixed effects.
Adding region or country fixed effects raises the explained variance even further, but
does not alter the coefficients. Taken together, these results suggest that international
differences in average values of subjective well-being reflect common responses to
international differences in the economic and social life circumstances, rather than to
differences in the way people evaluate their lives (as argued by Diener, Helliwell, &
Kahneman, 2010). This, coupled with the statistical strength of the key coefficients, helps
to provide greater confidence in our estimates of the value of social capital.
We estimate the value of social capital as the ratio of the social capital coefficient
to that of income. Following the notation of equation (1), the value of some social capital
variable X is Δ = μ/ δ, where Δ represents the log change in income that would have for
the average respondent the same life satisfaction effect as a change in some measure of
non-financial life characteristic, X. These ratios or coefficients are referred to as
compensating differentials, since they show how much income could fall and still leave
life satisfaction unchanged if there were some improvement in the social context, or vice
versa.
To calculate reliable estimates using this method obviously requires secure and
reasonably independent estimates of the life satisfaction effects of an individual’s social
circumstances and also of income, since the compensating differential is based on the
ratio of the sizes of the two effects. If individual- and societal-level effects are estimated
simultaneously, then the error ranges of the two effects can be separately computed. The
individual-level effects of social capital are often estimated much more precisely than are
the contextual or the societal, in large part because the number of independent
observations is invariably much smaller for the contextual effects.
10
Another important robustness check is to relax the assumption that coefficients
are the same in all countries, and to estimate separate well-being equations for each
country. Since the sample sizes are now much smaller, and since the social and political
contexts shaping well-being differ among nations in many more ways than can be
captured by the limited set of explanatory variables, the distributions of national
coefficients are quite spread out, as observed by Helliwell et al. (2010). That study shows
that the means of the country distributions are almost exactly equal to the global
coefficients, once more providing support for the use of the global coefficients for the
individual-level effects.
Figure 1 shows estimates of compensating differentials for several of the Gallup
variables with some connection to social capital. These are based on common global
coefficients estimated using two-level modelling of the global sample of individual data.
The lower part of each bar shows the individual-level effects, and the vertical line shows
the range covered by plus and minus one standard error. The individual-level effects are
thus all highly significant. Consider, for example, the ‘count on friends’ variable. Each
individual has been assigned the value for this variable (and zero otherwise) if they
answered ‘yes’ to the following question: ‘If you were in trouble, do you have relatives or
friends you can count on to help you whenever you need them, or not?’. To be able to
answer yes to that question has the same subjective well-being effect as a change in
the log of household income, which corresponds to almost a trebling of household
income. The effects would probably be even larger, and would be more precisely
estimated, if a more graduated range of answers had been available. This is especially the
case for this variable, because of the very large number of ‘yes’ respondents, who thereby
comprise a group too large to have supporting social networks of equivalent quality. For
example, almost 82% of global respondents answered ‘yes’, and this percentage rises to
92% among respondents living in the industrial countries. Since other research has shown
that the size and characteristics of one’s social networks have important dose-response
linkages to subjective well-being, there is obviously significant explanatory power lost by
not taking into account the large differences that must exist among the ‘yes’ respondents,
11
all of whom are currently assumed to have equally supporting and engaging social
networks.
The upper part of each bar shows the estimated size of the national-level effect,
once again accompanied by a vertical line covering the 95% confidence region. As noted
earlier, the national-level contextual effects for social capital are generally positive. The
contextual effects are in some cases significantly positive, although they are always much
less precisely estimated than the individual-level effects. Thus to live in a country where
others have family and friends to count on has positive well-being effects above and
beyond those that result from having one’s own supporting network. These spill-over
effects from social networks are large, even if not very precisely estimated. For example,
to live in a country having an Irish level of having friends to count on (%) compared
to France (at %) would have an income-equivalent value of about .175 of log
income, about equal to 20% of one’s income, with one-fifth of this 20% due to the
average individual effect, and the rest arising from the national-level effect.
Comparing countries much further apart in the international spectrum provides
correspondingly larger estimates. If we order the countries in the first three waves of the
Gallup Poll in terms of the average percentage of a population having family or friends to
count on, the top ten average just over 96%, and the bottom ten just below 50%. The
income-equivalent value of this difference is (=.46*(+)), or more than a
nine-fold increase in per capita incomes. Put another way, the lack of this type of social
capital accounts for a difference of almost a full point () on the 10-point scale of the
Cantril ladder. But this is by no means the whole story, as the total life evaluation gap
between these two groups of countries is more than points, of which about points
is explained by their almost 25-fold difference in average per capita incomes. Thus, even
though the support of friends is of critical importance, and has huge income-equivalent
values, if we look across countries the differences in average incomes are even greater
than the differences in the strength of social networks. As we shall now show, this
situation is reversed when we turn to consider differences in incomes, social capital, and
well-being within countries.
12
Valuing social capital differences within countries
In this section we use data from the 17th cycle (2003) of the Canadian General
Social Survey (GSS) to value social capital. This survey is especially appropriate for this
valuation, since it was designed to measure the extent and quality of social capital. Even
without this larger range of social variables, we find that social factors explain a larger
proportion of the total variance when the sample is within an industrial nation, since basic
wants are largely satisfied.
Table 1: Weighted OLS regressions for satisfaction with life
(1) (2) (3) (4)
Household income (ln) *** *** *** ***
[] [] [] []
Social trust **
[] [] [] []
Trust neighbour to *** *** *** ***
return wallet [] [] [] []
Trust colleagues *** *** *** ***
[] [] [] []
See family (ln) *** *
[] []
See friends (ln) *** ***
[] []
Have close friends (ln) *** ***
[] []
Have close family (ln) *** ***
[] []
Have other friends (ln) *
[] []
Belong in community *** ***
[] []
Belong in Province *** ***
[] []
Belong in Country *** ***
R2
Note: * p < .1; ** p < .05; , *** < .01.
Standardized beta coefficients and, in square brackets, t-statistics, are shown.
13
Detailed results and definitions of variables are in the online appendix.
Table 1 shows the standardized betas from several equations explaining cross-
sectional differences in life satisfaction among the sample of more than 11,000
respondents for whom the full set of variables is available. From this it can be seen that
several different measures of the extent, quality and frequency of use of social capital are
significantly correlated with life satisfaction, as are key measures of social identity,
especially those relating to the respondent’s sense of belonging to his or her local
community, province and country. Four equations are modelled in the table, with the
most basic (., 1) on the left and the most complete (4) on the right. The first two
equations do not include the social identity variables, while the third and fourth do. The
first and third equations do not include variables measuring the size and intensity of use
of social networks, while the second and fourth do. Thus it is possible to see the effects of
adding social identity and social connections, separately and together.
All of the equations include three measures of trust; a general measure of social
trust, assessments of the likelihood of a lost money-bearing wallet being returned intact if
found by a neighbour, and a measure of the respondent’s trust in his or her workplace
colleagues. As analyzed in more detail in Helliwell and Wang (2010), all three trust
measures are significantly correlated with life satisfaction, with trust in colleagues and in
neighbours remaining robustly significant even when the mediating effect of social
identity is taken into account, as in the third and fourth equations. The partial SWL
correlation of the social trust measure tends to fall with the addition of direct measures of
domain trust (as shown earlier in Helliwell & Putnam 2004), and with the addition of
specific measures of social capital and social identity, as shown by the changes in the
social trust coefficient when moving from the left to right across the table. Trust in
colleagues has the largest explanatory role, with a standardized beta exceeding that for
the other two measures combined, and equal to that for income.
The third and fourth equations of Table 1 present our first results attempting to
measure some direct and indirect effects of social identity on life satisfaction. At this
14
correlational level, we find significant evidence that those who feel a strong sense of
belonging are significantly more satisfied with their lives. The sense of belonging to
one’s local community is especially important, with a beta value slightly exceeding that
for income. Thus, if we are seeking to explain the variance across individual respondents
to the 2003 Canadian General Social Survey, even more is explained by differences in
community-level social identity than by differences in income. Feelings of belonging to
country and province are also significant supports for life satisfaction, with the former
more important than the latter (except in Quebec); together they explain a slightly smaller
share of the variance than does the community-level social identity.
Local social identity appears to mediate about half of the life satisfaction effects
of the neighbourhood trust variable, as the beta coefficient on the latter variable drops in
half when the community belonging variable is added. Hence, the life satisfaction effects
of believing that your neighbours would make the effort to return your lost valuables flow
substantially through their support for a greater sense of community belonging, as will be
shown later in the three parts of Figure 2.
We can also use our results to check for parallels with the findings of Haslam,
O’Brien, Jetten, Vormedal and Penna (2005) relating to interactions between social
support and social identity. In their regressions, social support and social identity were
both strongly predictive of life satisfaction, when entered on their own. The authors then
hypothesized that the positive effects of social identity on life satisfaction would be
mediated through social support, implying that the estimated effects of social identity on
life satisfaction would be significantly less when social support was added to the
equation, but not vice versa. Their results supported the hypothesis.
Our results, shown in Table 1, show less of this sort of mediation, and
consequentially greater independent roles for both social identity and social support. The
estimated coefficients on social support (as measured by the number and frequency of
contacts with family and friends) fall when social identity (focussing on the key
community belonging variable) is introduced, and vice versa. But the drop is greater for
15
social support (when social identity is introduced, comparing Equation 4 with Equation
2) than for social identity (when social support is introduced, comparing Equation 4 with
Equation 3). To obtain a closer replication of their tests, we could construct aggregated
measures to represent our social support and social identity variables, something lying
beyond our current exploratory analysis. As they stand, though, our data seem to admit
independent roles for both social identity (especially at the community level) and social
support (as measured by the existence and use of larger support networks of family and
friends) as simultaneously significant correlates of life satisfaction.
The life satisfaction results in Table 1 also illustrate the claim made earlier, that
social variables are more important than economic differences when explaining life
satisfaction differences among respondents living within Canada. For example, the
combined social identity variables explain twice as much variance as does income; the
trust variables explain more than income, and the social network variables are of roughly
similar size. This calculation is based on Equation 4, in which all variables are entered
simultaneously, so that it is appropriate to add up the standardized betas to a total that is
several times as large as that for income. Also noteworthy is the lack of conflict among
local, provincial and national identities. Each appears to add to life satisfaction without
detracting from the positive effects of the other sorts of belonging. To test this, we
defined a variable summing the belonging variables for each individual, and found a
slight but nonetheless positive effect. Thus, the benefits of belonging to one’s local
community do not appear to depend on being less attached to the larger encompassing
communities. For any given feeling of belonging at one level, life satisfaction is higher
among those who also feel they belong to both more and less encompassing social
identities.
Because we found fairly robust correlations between life satisfaction and three
different social identities of different but overlapping ambits, we then became curious
about what factors might underlie the ways in which individuals identify with their
communities, their provinces, and with Canada as a whole. We have therefore done some
exploratory analysis using individual and census-tract variables to explain differences
16
among individuals in their sense of belonging in their communities, their provinces, and
in Canada as a whole. The main results are shown in Figures 2, 3 and 4, and the
regression details shown in Table 2 of the online Appendix. All three figures compare the
coefficients for each of the three types of social identity. Figure 2 shows the effects of
different types of trust, Figure 3 shows the effects for different types of social
connections, while Figure 4 contains variables showing that social identities take time to
grow. These effects are drawn from equations that include all variables simultaneously,
so that they are competing for explanatory power.
The factors explaining the three types of belonging are quite different, in ways
that tend to support the validity of all three measures. In particular, community-level
belonging, as shown in the first equation, is highly dependent on other local variables,
especially neighbourhood trust, social connections with family and friends, and length of
time spent in the neighbourhood. A sense of belonging to one’s province is higher among
francophones, while belonging to Canada is much lower. People who live in the province
of their birth identify more with their province, and less with Canada as a whole, than do
those who have lived in more than one part of Canada. Immigrants identify significantly
more with their local community and province, and less with Canada as a whole, than do
those born in Canada. However, all of the immigrant-related effects tend to disappear
with increases in the number of years that the immigrant has lived in Canada.
17
Figure 2-1: Determinants of belonging: trust
0
HH income trust (social) trust
(neighbours)
trust (colleagues) family contact
Belong to community Belong to province Belong to country
18
Figure 2-2: Determinants of belonging: social connections
0
friends contact close friends close family other friends
Belong to community Belong to province Belong to country
19
Figure 2-3: Belonging takes time to develop
0
neighbourhood
tenure
born in province immigrant years since arrival
Belong to community Belong to province Belong to country
20
Conclusion
Our global and national results both attach high values to several aspects of the
social context. Our respondents report significantly higher life satisfaction when they
have strong social networks, when they make more frequent use of these supportive
networks, when they trust those among whom they live and work, and when they feel a
sense of belonging in their communities. Our results are derived from individually-based
cross-sectional survey data, making it difficult or impossible to derive strict causal
interpretations. The economic and social circumstances we have alleged to increase life
satisfaction may indeed do so. But there is also the very real possibility that those who
are more satisfied with their lives will also reach out to others, form more robust social
networks, and be more successful in their economic and social lives.
Because these positive two-way linkages are plausible for both economic and
social variables, one can perhaps examine more reliably the relative sizes of economic
and social influences. As we have shown, this can be done if both income and social
factors are apparently important supports for life satisfaction. If they are, as we found for
both the Gallup World Poll data and the data from the Canadian General Social Survey, it
is possible to measure the importance of social capital in income-equivalent terms, as we
did for the Gallup data, and to compare the overall explanatory power of income and
social capital, as we have done using the Canadian data. In both cases we found large
values for several different measures of the extent and nature of social engagement and
social identity. Our results seem to us to support and encourage a variety of experimental
interventions, since only in such an environment can causality and policy relevance be
firmly established.
21
References
Annas, J. (1993). The morality of happiness. Oxford, UK: Oxford University Press.
Barrington-Leigh, . & Helliwell, . (2008). Empathy and emulation: Life
satisfaction and the urban geography of comparison groups. NBER Working
Paper 14593. Cambridge: National Bureau of Economic Research.
Berkman, L. F., & Syme, S. L. (1979). Social networks, host resistance, and mortality: A
nine-year follow-up study of Almeda county residents. American Journal of
Epidemiology, 109, 186–204.
Deaton, A. (2008). Income, health and well-being around the world : Evidence from the
Gallup World Poll. Journal of Economic Perspectives, 22(2), 53-72.
Diener, E., Helliwell, ., & Kahneman, D. (2010). International differences in well-
being. New York: Oxford University Press.
Durkheim, E. (1897). Le Suicide. Paris:1e edition. First appeared in English as Durkheim,
E. (1952). Suicide: A study in sociology. Translated by . Spaulding and G.
Simpson. London: Routledge and Kegan Paul.
Gleibs, ., Haslam, ., Morton, , Rabinovich, A., & Helliwell, . (2010). Unpacking
the hedonic paradox: Whether happiness is predicted by economic or social capital
depends on the content of identity-defining values. University of Exeter, Paper in
preparation.
Haslam, ., O’Brien, A., Jetten, J., Vormedal, K., & Penna, S. (2005). Taking the
strain: Social identity, social support and the experience of stress. British Journal
of Social Psychology, 44, 355-370.
22
Helliwell, . (2007). Well-being and social capital: Does Suicide Pose a Puzzle? Social
Indicators Research, 81, 455-496.
Helliwell, . (2008). Life satisfaction and quality of development. NBER Working
Paper 14507. Cambridge: National Bureau of Economic Research.
Helliwell, . & . Barrington-Leigh (2010) Measuring and understanding subjective
well-being. NBER Working Paper 15887. Cambridge: National Bureau of
Economic Research. (Forthcoming in Canadian Journal of Economics, August
2010).
Helliwell, ., Barrington-Leigh, . Harris. A., & Huang, H. (2010). International
evidence on the social context of well-being. In E. Diener, . Helliwell & D.
Kahneman (Eds.), International differences in well-being (pp. 291-325). New
York: Oxford University Press.
Helliwell, ., & Huang, H. (2010). How’s the job? Well-being and social capital in the
workplace. Industrial and Labor Relations Review, 63, 205-228.
Helliwell, . & S. Wang (2010) Trust and well-being. NBER Working Paper 15911.
Cambridge: National Bureau of Economic Research.
Helliwell, ., & Putnam, . (2004). The social context of well-being. Phil Trans R.
Soc Lon. B, 359, 1435-46. Reprinted in . Huppert, B. Keverne & N. Baylis
(2005, Eds.), The science of well-being (pp. 435-459). London, UK: Oxford
University Press.
Helliwell, . & Wang, S. (2009). Trust and well-being. Paper prepared for the 3rd
OECD World Forum on Statistics, Knowledge and Policy, Busan, Korea, October
27-30, 2009.
23
Kahneman, D. (1999). Objective happiness. In E. Diener, D. Kahneman & N. Schwarz
(Eds.), Well-being: The foundations of hedonic psychology. New York: Russell
Sage.
Kahneman, D., & Krueger, . (2006). Developments in the measurement of subjective
well-being. Journal of Economic Perspectives, 20, 3-24.
Kahneman, D., & Riis, J. (2005). Living, and thinking about it: Two perspectives on life.
In . Huppert, N. Baylis & B. Keverne (Eds.), The science of well-being
(pp. ??- ??). Oxford, UK: Oxford University Press.
Kahneman, D., Wakker, ., & Sarin, R. (1997). Back to Bentham? Exploration of
experienced utility. Quarterly Journal of Economics, 112, 375-405.
Luttmer, . (2005). Neighbors as negatives: Relative earnings and well-being.
Quarterly Journal of Economics, 120, 963-1002.
Lucas, R. E. (2005). Time does not heal all wounds. Psychological Science, 16, 945-950.
Maslow, . (1943).A theory of motivation. Psychological Review, 50, 370-96.
OECD (2001) The well-being of nations: The role of human and social capital. Paris,
Organization for Economic Cooperation and Development.
Putnam, . (2000). Bowling Alone. New York: Simon & Schuster.
Redelmeier, ., Katz, J., & Kahneman, D. (2003). Memories of colonoscopy: A
randomized trial. Pain, 104, 187-194.
Sen, A. (1999). Development as freedom. New York: Knopf.
24
Williams, M. (2001). Suicide and attempted suicide. London, UK: Penguin.
Wirtz, D., Kruger, J., Scollon, ., & Diener, E. (2003). What to do on spring break?
The role of predicted, on-line and remembered experience in future choice.
Psychological Science, 14, 520-524.
