1 Introduction

To set the stage we must first explain our title. What do we mean by social capital, well-being, and crisis? We shall use social capital to include several key measures of the quality of the social fabric, including the quality of social networks and social norms, plus several measures of trust. In this we follow usage adopted in earlier studies by Putnam (1993, 2000), Helliwell and Putnam (2004) and the OECD (2001). Our concept of wellbeing is based on individuals’ evaluations of the quality of their own lives, following the strategy advocated by Aristotle and represented most generally now by a variety of empirical life evaluations, with measures of life satisfaction being the most typical. Finally, our notion of crisis is deliberately broad enough to include external events, breakdown of existing structures, and a variety of other challenges to sustainable development of natural and human resources.

2 Social Capital and Well-Being in Times of Economic Crisis

In this section, our example shock will be the 2007–2008 global economic and financial crisis. We shall look first at how social capital differences across communities within the United States are linked to their levels of happiness, and see whether these differences in social capital have moderated the happiness effects of recent economic shocks. We shall then present some international evidence on how subjective well-being has evolved in three groups of OECD countries during the post-crisis period.

2.1 Social Capital and Happiness in US Cities

This section explores the relation between social capital and happiness within the United States, in particular on whether social capital has had moderating effects on the well-being impact of the 2007–2008 economic crisis. Our measures of social capital are derived from the supplemental Current Population Surveys from 2004 to 2008. We measure subjective well-being using the Gallup-Healthways Well-Being Index, a daily survey of US residents conducted by the Gallup Organization since 2008. We begin with the measurement of happiness, or more generally subjective well-being (SWB). Helliwell and Huang (2011) describe the construction of the measures in greater detail. Here we explain the basics.

2.1.1 Measuring Subjective Well-Being (SWB)

The first SWB measure we use is a summary evaluation of life that we call the Cantril life ladder, or Cantril ladder. It is the response to the question “Please imagine a ladder with steps numbered from 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 life for you. On which step of the ladder would you say you personally feel you stand at this time, assuming that the higher the step the better you feel about your life, and the lower the step the worse you feel about it? Which step comes closest to the way you feel?” The Cantril ladder measure thus has 11 points from 0 to 10, with higher values indicating a better evaluation of life. In the US from 2008 to 2011, more than 70 % of survey respondents chose 6 or above; the greatest mass (about 25 %) chose 8; about 9 % of the respondents chose 9; a similar percent chose 10. In the range below 6, about 15 % chose 5, only about 10 % chose between 0 and 4.

Another two SWB measures are reports of emotional states. They are based on a list of survey questions on emotional experience the day before the interview: (1) Did you smile or laugh a lot yesterday? (2) Did you experience the following feelings during a lot of the day yesterday? How about enjoyment? (3) How about happiness? (4) Did you learn or do something interesting yesterday? (5) Did you experience the following feelings during a lot of the day yesterday? How about worry? (6) How about sadness? (7) How about stress? (8) How about anger? The answers to the first four questions reveal positive emotional feelings. The answers to the other four questions reveal negative feelings. We use the first four questions to construct a score of positive emotions, essentially the number of “yes” answers. The score has five steps from 0 to 4; zero means that the respondent reports no positive experiences; four means all four are reported. In a symmetrical manner, we construct the score of negative emotions based on the four questions of negative emotions. In the Gallup-Healthways survey, more than 50 % of respondents reported the maximum score of four positive emotions. About 30 % reported three, leaving less than 20 % reporting a score at 2 or below, within whom only 4 % report zero positive emotions. For the score of negative emotion, about 50 % report zero negative experience; 20 % have a score of one; 15 % two; 10 % three; leaving only 5 % with a score of four.

We use the fourth, and last, SWB measure to summarize the two scores of emotions in a dichotomous manner. Specifically, this measure is 1 if the score of negative emotional experiences is strictly greater than the score of positive emotions; it is zero otherwise. Essentially, it classifies a survey respondent’s day before the interview as an “unpleasant” one if the respondent reported more negative emotions than positive ones. We call this measure the pseudo u-index, because the index is in part motivated by the u-index that was introduced in Kahneman and Krueger (2006), who expressed doubt about measuring life satisfaction with numerical scales. In the Gallup-Healthways survey, 11 % of respondents have a u-index that is 1.

In total, we have four measures of well-being: an 11-step Cantril ladder, a 5-step score of positive emotions, a 5-step score of negative emotions, and a 0-or-1 pseudo u-index that indicates the dominance of negative emotions over positive ones.

2.1.2 Measuring Social Capital

To measure social capital at the local level in the US, we use information from the supplemental surveys of the Current Population Survey (CPS) conducted in the years from 2004 to 2008. Specifically, the measures of social capital are average survey responses to questions on social engagement and networks. We use the September CPS supplemental surveys on volunteering and the November supplemental surveys on voting, registration and civic engagement in the years from 2004 to 2008. The September surveys measure “participation in volunteer service, specifically about frequency of volunteer activity, the kinds of organizations volunteered for, and types of activities chosen” (the Census Bureau). The November surveys focus on voting and registration; but in November 2008, it also includes questions on interactions in households, with neighbors and friends, as well as charitable/religious donations.Footnote 1 From the surveys, we construct indicators of social capital at the local level, including the percentage of people who have volunteered for religious, and alternatively non-religious, groups, the percentage of population who voted, and so on.

There are many such indicators. To reduce the dimension of our measurement, we adopt a strategy of first grouping them into categories, before averaging the measures within each category. We standardize all underlying indicators into a common scale with mean 0 and standard deviation 1 to avoid the resultant averages being dominated by large-scale indicators. We focus on two categories: engagement close (friends, family, neighbours, and religion), and engagement broad (volunteering, voting, and generosity). This division mirrors the distinction between bonding and bridging social capital, where bonding social capital ‘refers typically to relations among members of families and ethnic groups’ and bridging social capital ‘refers to relations with distant friends, associates and colleagues’ (OECD 2001, 42). To test whether the way we divide the indicators changes the findings in substantial ways, we also repeat the analysis with a single measure of social capital that is the average of all indicators, without any divisions. The following is the list of indicators; the contents inside the parentheses indicate the month and year of the underlying supplemental CPS surveys:

SC 1: Engagement broad

  1. 1.

    The average voting rate in the November 2004 and 2008 elections (November 2004 and November 2008)

  2. 2.

    The percentage of residents who volunteered for or through non-religious groups (September 2004–2008)

  3. 3.

    The average number of groups volunteered for or through, in per resident terms (September 2004–2008)

  4. 4.

    The average hours volunteered in year, in per resident terms (September 2004–2008)

  5. 5.

    The percentage of residents who donated more than $25 to charitable or religious organizations (September 2008)

SC 2: Engagement close

  1. 1.

    The percentage of population in multi-member households who eat dinners with other members of the household basically everyday (November 2008)

  2. 2.

    The percentage of population who have more than 3 close friends (November 2008)

  3. 3.

    The percentage of residents who talked with neighbors a few times a week or more (November 2008)

  4. 4.

    The percentage of residents who did favors for each other with neighbors once a month or more (November 2008)

  5. 5.

    The percentage of residents who volunteered for or through religious organizations (September 2004–2008)

The indicators above include electoral participation and other forms of civic engagement, as well as close-circle engagement with family, friends and neighbors. We have no measures of social trust and trust in specific domains, due to the absence of such questions in the supplemental CPS.

We aggregate the measures of social capital to the level of cities (metropolitan areas). The CPS has good geographic identifiers at this level, identifying the metropolitan area of residence for about 70 % of survey respondents. At the county level, the identified share of respondents drops to 40 %. Table 1 below presents the summary statistics of the social-capital indicators at the metropolitan level.

Table 1 Summary statistics of the underlying indicators for social capital from the supplemental CPS

As stated above, we use the averages of social-capital indicators for our regression analysis (after standardizing the scale of underlying individual indicators). We have one average for the broad-engagement category, one for the close-engagement category, and one for all the indicators listed above. The next table presents the correlation coefficients among the three averages. Since the averages themselves do not have a naturally-interpretable scale, we further standardize them to have a scale of mean 0 and standard deviation 1 before the analysis (Table 2).

Table 2 Correlation coefficients between the three averages of the social-capital indicators from the supplemental CPS

For the regression analysis presented in the next subsection, we merge the CPS data into the Gallup-Healthways’ respondent-level survey data. We include all observations with matched information in our analysis. The final sample covers 256 metropolitan areas, representing nearly 220 million residents in the 2000 US Census, or about 80 % of the total population at the time.

2.1.3 Empirical Findings

We ask three empirical questions on the relation between social capital and well-being during the recent years of economic crisis: (1) Do any of the three social-capital measures based on data from the 2004–2008 supplemental Current Population Surveys, affect subjective well-being (SWB) in the Gallup-Healthways survey from 2008 to 2011? (2) Do any of the social-capital variables moderate the (presumed) negative SWB impact of unemployment on those Gallup-Healthways respondents who were themselves unemployed at the time of interview? (3) Do any of the social-capital variables moderate the (presumed) negative SWB impact of increases in local unemployment on local residents, including those who are not themselves unemployed?

We address the three questions simultaneously using two-level regressions. The regressions use both individual and contextual information to explain individual well-being reported in the Gallup-Healthways survey. The most important contextual variables in our case are the measures of social capital from the CPS, their interactions with respondents’ own unemployment status, and their interactions with local unemployment (measured in fractions of the labour force). The direct effects of personal and local unemployment (in fractions) are also included. Since the key variables of interest are at the local level, we cluster errors at the corresponding level to avoid under-estimating standard errors. This imposes a more stringent requirement for making statistical inferences.

The form of the estimating equations is as follows:

$$\begin{aligned} w_{i,t,j} & = \alpha_{1} UN_{i,t} + \alpha_{2} \Updelta UR_{j,t} + \alpha_{3} UR_{j,t - 4} + \alpha_{4} SC_{1,j} + \, \alpha_{5} SC_{2,j} + \beta_{1} UN_{i,t} SC_{1,j} + \beta_{2} UN_{i,t} SC_{2,j} \\ & \quad + \gamma_{1} \Updelta UR_{j,t} SC_{1,j} + \gamma_{2} \Updelta UR_{j,t} SC_{2,j} + X_{i,t} \delta_{1} + Z_{j,t} \delta_{2} + D_{t} \delta_{3} + u_{i,t} \\ \end{aligned}$$

The dependent variable w i,t,j is the well-being measure of worker i in locality j who is interviewed at time t. The time subscript t is in the unit of quarters. The first variable on the right-hand side, UN i,t , is each survey respondent’s own unemployment status (UN i,t equals to 1 if the respondent is unemployed; zero otherwise). The second variable ∆UR j,t  = UR j,t  − UR j,t−4 is the change in local unemployment rate (measured in fractions of the labour force) since the same quarter in the previous year. The third variable UR j,t−4 is the level of unemployment in the same quarter last year. We use same-quarter changes to deal with seasonality. The unemployment rates are from the Local Area Unemployment Statistics program of the Bureau of Labor Statistics. They are measured at the metropolitan level, consistent with the measurement of local social capital. The next two variables, SC 1,j and SC 2,j are the two measures of average social capital (broad engagement and close engagement).

The final two variables in the first line of the equation capture the interactions between social-capital measures and each survey respondent’s own unemployment status. If the dependent variable w i,t,j measures positive evaluations/emotions, and the coefficients β 1 and β 2 are positive, a higher value of those social-capital measures narrows the well-being gap between unemployed persons and the rest of the population.

The second line of the equation starts with variables capturing the interactions between social-capital measures and changes in local unemployment since the same quarter last year. If the dependent variable w i,t,j measures positive evaluations/emotions, and the coefficients γ 1 and γ 2 are positive, a higher value of those social-capital measures mitigates the impact of rising unemployment on the well-being of the general population, including those who are not themselves unemployed.

We will also present findings from a model that replaces the two social-capital measures with a single aggregate measure, the average of all 10 indicators from the CPS.

The vector X i,t has all other personal and demographic information including age categories, gender, marital status, educational attainment, race and the logarithm of household income. The vector Z j,t has county-level census information, including the log of average household income, the log of population density, the urbanization rate, the racial composition of each county’s population, the percentage of owner-occupied housing (to measure the stability of population), and the longitude and latitude of the geographic centroid. It also includes dummy indicators for Alaska and Hawaii, so that the longitude and latitude variables reflect differences within the continental U.S. Finally, we include a set of year-quarter dummies D t to capture time trends as well as possible framing effects due to changes in the survey questionnaires.

For estimation, we employ linear least squares regression treating the well-being answers as cardinal measures, as opposed to a probit model that treats the answers as ordinal. The linear approach is common in the literature, as the choice between linear and probit models in general makes little qualitative difference (Ferrer-i-Carbonell and Frijters 2004). We cluster the estimation errors at the level of metropolitan areas, the same as the level of aggregation for the social-capital variables.

The next table presents the estimates. There are four columns of results, one for each alternative SWB measure: Cantril ladder, the score of positive emotion, the score of negative emotion and the pseudo u-index (Table 3).

Table 3 Two-level regression with interactive terms between social capital and unemployment

The estimates show that that both personal and local-level unemployment have a negative effect on well-being. A survey respondent’s own unemployment status is negatively correlated with well-being. The estimated effect is similar to or even greater than the effect from reducing household income by one unit in logarithms. A rise in the local unemployment rate also reduces well-being. A 1 % rise in the unemployment rate reduces the life ladder by 0.03378, equivalent to reducing household income by 0.03378/0.469 = 0.072 unit in logarithm, or by 7.5 %. Similar estimates based on other measures of well-being yield smaller effects ranging from 3 to 5 % of the income. The lagged level of unemployment is also negatively linked to well-being.

The estimates also show that there is generally a positive relationship between social capital and well-being. The statistical significance varies. Of the 8 estimates linking social capital to well-being directly (i.e., not through the interactive effects), three of them are statistically significant at the 10 % level, and two more at the stronger 5 % level. We note that, due to the presence of interactive terms, these estimates measure the impacts of a higher social capital in an area with zero growth in unemployment for those who are not themselves unemployed. Quantitatively, the impact is also modest. Taking the life ladder as the example, a one standard deviation higher measure of close engagement raise average well-being by 0.019, an effect equivalent to increasing household income by 0.019/0.469 = 0.0405 logarithm units, or roughly 4 %.

We now examine the interactive effects. First of all, neither close engagement nor broad engagement mitigates the substantial negative impact of personal unemployment on well-being. Of all the interactive terms between personal unemployment status and the two social capital measures, none of the estimated coefficients is statistically significant, suggesting little difference in the impact of personal unemployment regardless the level of social capital in the area of residence. Here we note that the engagement is measured at the local level, as opposed to the individual level.

But there is evidence that broad engagement mitigates the population-wide negative impact of rising unemployment rates. When well-being is measured by the Cantril ladder, the estimated coefficient on the interactive term between changes in unemployment and local broad engagement is 0.596. This means that if we compare a city where the broad engagement is one standard deviation above the national average (the average is zero), the impact of rising unemployment is estimated to be 0.596 smaller in magnitude per unit than the impact of the same increase in a city that is at the national average in the social capital. The standard error of the estimated difference is 0.215. We can also calculate the marginal effect of rising unemployment for the average city and for the above-average city respectively. In an area where the social capital measure equals zero (i.e., it is at the national average) the marginal effect of rising unemployment is just the main effect estimated to be −3.378 per unit with a standard error of 0.433. In an area where the broad-engagement measure is +1, however, the impact will be a smaller value of −3.378 + 0.596 × 1 = −2.782 per unit (standard error = 0.449).Footnote 2 The marginal effect is thus less negative in an area with an above-average measure of broad engagement. The opposite occurs in an area with below-average measure of broad engagement. The marginal effect is −3.378 + 0.596 × (−1) = −3.974 (standard error = 0.515). The difference between the +1 area and −1 area is 1.19 per unit, about one-third of the marginal effect expected in an average area.

The next table replaces the two measures of social capital with a single measure, the average of all social capital indicators from the CPS. The purpose is to test whether our findings regarding social capital are sensitive to the way we divide the CPS indicators (Table 4).

Table 4 Two-level regression with interactive terms between social capital and unemployment

The key results hold up remarkably well in the test. The average measure of social capital has a beneficial effect on well-being, and the effect becomes stronger in areas with greater increases in unemployment rates. This version does not, however, permit us to see any potential differences between broad engagement and close engagement.

In summary, the analysis suggests that social capital has improved subjective well-being during the period of economic crisis, both directly and indirectly through mitigating the impact of rising unemployment. The protective effect against rising unemployment is mostly found from the broader measure of social capital. We caution, however, that social capital does not appear to mitigate the impact of personal unemployment. The beneficial moderating impact of broad engagement thus does not operate as an individual-level insurance mechanism.

2.2 Changes in National Average Happiness in OECD Countries After the 2008 Financial Crisis

The OECD countries except South Korea are divided into three groups according to the performance of happiness, measured by the Cantril ladder (GWP 2006–2011). Group 1 includes Chile, Japan and Mexico, which show an increasing trend in the post-crisis period (2008–2011). The group with rising happiness included countries less directly affected by the crisis, and with policies well chosen to enhance the well-being of their residents (ILO/World Bank 2012). The Group 2 countries, which include five EC countries hard-hit by the crisis and its aftermath (Portugal, Italy, Ireland, Greece, Spain) plus the United States and New Zealand, all have a downward trend, although Portugal recovers slightly in 2011. Group 3 countries, comprising all the remaining OECD countries with three or more waves of data, show a flat average happiness trajectory.

Figure 1 illustrates the dynamics of the Cantril ladder (with the 95 % confidence interval denoted by the vertical line) in South Korea and the three groups of other countries. As shown in the figure, South Korea has one of the best performances among OECD countries during the post-crisis period (2008–2011).Footnote 3 Its subjective well-being before the crisis was significantly lower than in all three groupings of other OECD countries, while in 2011 it was above all three groups. South Korea’s policies and performance were strikingly different, and better, during and since the recent crisis, than during and after the 1997–1998 banking crisis. Cho (2012) and Cho and Shin (2011) argue that the better results were due to faster, stronger and more appropriate policies, which were in turn made possible by much more robust pre-crisis fiscal and financial frameworks. Kwon et al. (2010) argue also that much was done after the earlier crisis to build a better system of social safety nets in Korea, and that these were in place to help digest the more recent crisis.

Fig. 1
figure 1

Dynamics of Cantril ladder in South Korea and other OECD countries

But there is more to the story than the pre-existing policies and conditions. President Lee, when opening the OECD’s Third Global Forum on Well-Being in Busan in October 2009, saidFootnote 4:

“I believe that Korea can contribute to helping the global community achieve the advancement of their economies and the quality of life. Korea has already proposed a new way forward from the global crisis. We call this the sharing of jobs and a new vision for the future called low-carbon green growth. As the economy worsened many economies opted to lay off workers in massive numbers in order to survive and of course in a market economy this may be considered as something very natural but our companies in Korea chose a different path. We decided to share the burden. Employees chose to sacrifice a cut in their own salaries and companies accepted to take cuts in their own profits because they wanted to save their employees and co-workers from losing their jobs. As a result, Korea’s unemployment rate is a modest 3.4 per cent and as the forecast and the results of the third quarter show as released yesterday—compared to the previous quarter we had a 2.9 per cent GDP growth which is very unexpected. As you can see Korea is recovering more quickly than expected and is one of the fastest recovering economies in the world. I believe one of the reasons for this is the cooperation between management and labour.”

The co-operative nature of the policy responses, and the shared commitment to maintaining domestic employment and output as the crisis unfolded, probably underlay the fact that the drops in aggregate employment were far less than could be explained by historically established relations between output change and employment (Cho and Shin 2011, p. 173). This can probably be attributed to some combination of lower financial stress, in turn delivered by easier monetary policy, targeted support to business lending, and shared commitments to maintaining employment levels.

The Korean policy responses appeared to comprise larger doses of conventional instruments, improved social safety nets, plus a different and more inclusive approach to macroeconomic policy-making. The increases in South Korean subjective well-being during and after the financial crisis are far larger than could be explained by the amount of GDP growth that was the direct result of better policies. We hypothesize, as a basis for future research, that the more secure underlying economic and social environment, coupled with a more co-operative approach (than was used in other countries, or in Korea during the earlier macroeconomic crisis) supported higher social capital and engagement, which in turn helped to support the higher levels of subjective well-being.

To find more direct evidence of the links between social capital and subjective well-being, which we have conjectured were part of the Korean story, we turn now to compare the pre- and post-crisis experience of 30 European countries, for which we have good data for happiness, GDP per capita, and social trust. We broaden our concern at the same time to include rapid institutional change as another type of crisis to be investigated. Hence we divide our sample between the transition and non-transition countries.

3 Trust and Well-Being During Transition

The primary original work in this section is based on analysis of five rounds of data from the European Social Survey (ESS). The rounds are 2 years apart, with round 1 in 2002 and round 5 in 2010. These five rounds of data allow us to compare the co-movements, separately for transition and non-transition countries, of overall happiness, of social trust, and of GDP per capita to see how changes of income and social trust are associated with changes in happiness in the two groups of countries.

We make use of two separate ESS measures of subjective well-being, satisfaction with life and happiness. The former is the response to the question “All things considered, how satisfied are you with your life as a whole nowadays? Please answer using this card, where 0 means extremely dissatisfied and 10 means extremely satisfied.” The latter is the response to the question “Taking all things together, how happy would you say you are? (using the same answer scale)”. Because previous research (Helliwell and Wang 2012, p. 16) and our current results show that these two different life evaluations have high correlations with each other and very similar correlations with key driving variables, and give stronger results when the two measures are averaged, our analysis in this paper employs the average value of the two responses.

To measure social capital of each country in the ESS, we use the country-level survey responses to questions on social trust, trust in police, and trust in the legal system. Specifically, social trust is the response to “Generally speaking, would you say that most people can be trusted, or that you can’t be too careful in dealing with people? Please tell me on a score of 0–10, where 0 means you can’t be too careful and 10 means that most people can be trusted.” Trust in the police is the response to “Please tell me on a score of 0–10 how much you personally trust in the police. 0 means you do not trust an institution at all, and 10 means you have complete trust.” Trust in the legal system is the response to “Please tell me on a score of 0–10 how much you personally trust in the legal system. 0 means you do not trust an institution at all, and 10 means you have complete trust.”

The national income variable is PPP-adjusted GDP per capita in constant 2005 international dollars taken from World Development Indicators (WDI) 2012. In those cases where a survey round was not completed in a single year (2 years sometimes), the average of GDP per capita of the 2 years was calculated and used as the national income for that round.

In total there are 30 countries with 122 observations in the unbalanced panel data. Since we are interested in examining the role of social capital, and especially social trust, during times of institutional crisis, we first illustrate our data averages separately for the 10 transition and 20 non-transition countries.Footnote 5 Figures 2, 3 and 4 show averages of values for subjective well-being (measured as the average scores for life satisfaction and happiness with life), social trust and GDP per capita, respectively. In each figure, the data for the ten transition countries are shown separately from those for the twenty non-transition countries. If round 3 (2006) is taken to be the pre-crisis measure, and round 5 (2010) the post-crisis measure, happiness, social trust and GDP per capita have all risen in the transition countries and fallen in the remaining countries. Even during the crisis, the transition countries have continued to significantly close all three gaps between themselves and the non-transition countries. In 2006, the average happiness gap between the transition and non-transition countries was about 1.5 points on the 0–10 scale, while by 2010 the gap had shrunk to just over 1 point. The social trust gap, which was 1.25 points on the 10-point scale in 2006, had fallen to 0.9 points by 2010. Between 2006 and 2010, average real GDP per capita in the transition countries grew from 44 to 49.5 % of that in the non-transition countries. In all three cases, more than half of the gap closure was due to improvements in life in the transition countries, with the rest due to worsening post-crisis conditions in the other European countries.

Fig. 2
figure 2

Dynamics of subjective well-being in Europe (Data source ESS rounds 2002–2010)

Fig. 3
figure 3

Dynamics of social trust in Europe (Data source ESS rounds 2002–2010)

Fig. 4
figure 4

Dynamics of GDP per capita in Europe (Matched with survey years in ESS, Data Source WDI 2012)

The happiness rise in transition countries, relative to other European countries, was far greater than could be explained by the lessening gap in real incomes, which, under most estimates of the effects of income on happiness, would have explained only one-tenth of the improvement in transition-country relative happiness. By contrast, and making use of our estimation results reported below, the average increase in social trust in the transition countries (+0.22 points), combined with the reduction in non-transition countries (−0.15 points), would have been responsible for almost half of the 0.4 point happiness convergence between 2006 and 2010.

We then formally estimate the effects of social capital and national income on happiness based on the following equation:

$$W_{i,t} = \beta_{0} + \beta_{1} T_{i,t} + \beta_{2} I_{i,t} + SC_{i,t} \beta_{3} + \delta_{1} D_{i} + \delta_{2} S_{t} + u_{i,t}$$

The dependent variable W i,t is the well-being measure of country i at round t. The time subscript t is in the unit of survey rounds (usually completed in 1 year, but sometimes extending to a second year). The first variable on the right-hand side, T i,t , is social trust in country i and at time t. The second variable I i,t is the natural log of GDP per capita in country i and at time t. SC i,t is a vector of other variables of social capital including trust in police and trust in the legal system. D i represents the country fixed effect. S t is a set of round dummies to capture time trends.

We show results for two different estimation methods. The first is pooled OLS, wherein differences across countries and within countries are jointly estimated on an equal basis. Table 5 presents the results. The second set of results uses fixed effects (FE) estimation, which allows country-specific constant terms. These country fixed effects have the effect of removing pure cross-country effects from the estimation, with the size and significance of the resulting coefficients depending on the correlations between changes in happiness and changes in the independent variables. Table 6 shows this set of results. Tables 5 and 6 have similar structures. We include solely social trust in model (1). In model (2), we include GDP per capita in addition to social trust. In models (3) and (4), trust in police and trust in the legal system are included respectively. In model (5) all four variables are included. Round dummies are included in each model.

Table 5 Social capital and subjective well-being in Europe: Pooled OLS Models
Table 6 Social capital and subjective well-being in Europe: FE Models

Our principal findings are as follows. Social trust and GDP per capita are both associated with subjective well-being in the pooled OLS regressions in Table 5. Trust in police and trust in the legal system have very similar effects, and both are positive and significant. However, when both are included, the significance disappears, most likely because of the high correlation between them (r = 0.91). From Table 6 we find that the change of social trust is significantly associated with the change of subjective well-being, but the change of GDP per capita is only weakly associated with the change of well-being. Moreover, neither the change of trust in police nor of trust in the legal system helps significantly to explain the change in subjective well-being. In summary, changes in social trust are the primary contributors to changes over time in subjective well-being in European countries.

We then run FE panel regressions for the 10 transition and 20 non-transition economies separately, only including social trust and GDP per capita, since other trust variables are not significant. The results are shown in Table 7. Models (1) and (2) are for transition economies, and models (3) and (4) are for non-transition economies. Round dummies are only used in models (2) and (4). We find that in the transition economies, the change of social trust, but not that of GDP per capita, has significant impact on the change of subjective well-being; in contrast, in the non-transition economies, the change of GDP per capita, but not the change of social trust, plays a significant role in explaining the change in subjective well-being. The contrast highlights the importance of social capital for those transition countries which have been experiencing institutional crisis.

Table 7 Social capital and subjective well-being in transition and non-transition countries: FE Models

4 Conflicts Over Sustainable Resource Development

In this final section we bring our evidence to bear even more directly on the problems of sustainable resource development that are the focus of this symposium. This section builds on Elinor Ostrom’s path-breaking research showing the importance of social connections to the development of solutions to resource-sharing problems.

Although Ostrom’s studies (e.g. Ostrom 1990) relate to specific problems in the development and use of common-property resources, they have implications that extend much further. Thus in her Nobel lecture summarizing 50 years of research, Ostrom (2010, p. 435) concludes “The most important lesson for public policy analysis … is that humans have a more complex motivational structure and more capability to solve social dilemmas than posited in earlier rational-choice theory. Designing institutions to force (or nudge) entirely self-interested individuals to achieve better outcomes has been the major goal posited by policy analysts for governments to accomplish for much of the past half century. Extensive empirical research leads me to argue that instead, a core goal of public policy should be to facilitate the development of institutions that bring out the best in humans.”

Extending this analysis to take more explicit account of the sources and consequences of subjective well-being helps to explain the findings, and to increase the likelihood that her prescriptions are likely to be sustainable. In particular, each of the key elements of human behavior that she has found to increase the likelihood of finding cooperative outcomes to resource sharing conflicts appears to deliver intrinsic benefits above and beyond those provided by the better resource-sharing outcomes that they enable.

The Ostrom and related research was undertaken without reference to subjective well-being. Well-being research strengthens the power of the analysis by showing that prosocial behaviour is fundamental, in that it increases happiness independent of other consequences.

Humans are happier when they interact with others in a trusting environment, and hence they are fundamentally social animals (Helliwell 2012a, b, Helliwell and Wang 2011). The ‘social brain’ hypothesis (Dunbar 1998) argues that this is not an evolutionary accident, but is part and parcel of the ability of humans to survive and prosper in a world shared with other more powerful species. Human capacities for, and enjoyment in, large social groups well beyond family size is what gave them the capacity to develop and use collective actions to meet external challenges.

There is a stronger form of the hypothesis arguing that humans are more than just social, they are pro-social. In other words, they get happiness not just from doing things with others, but from doing things both with and for others (e.g. Batson and Shaw 1991). Despite a wealth of findings that those who do things for others gain a bigger happiness boost than do the recipients of generosity (Brown et al. 2003; Schwartz and Sendor 1999), people underestimate the happiness gains from unselfish acts done with and for others (Dunn et al. 2008). Perhaps this is a buffer against the likelihood that the warm glow from kind acts would be less if the giver felt that he or she were doing them for selfish reasons. There is also new evidence from brain scans (Zaki and Mitchell 2011) that people get primary value from making decisions that they perceive as equitable, even where this is at personal cost. Hence they improve their subjective well-being by doing the right thing. Social identity researchFootnote 6 suggests that these gains are likely to be even greater where people identify with the beneficiaries of the equitable behavior. In the current case of environmental actions, the primary beneficiaries are those yet unborn.

5 Conclusions

We first demonstrated, using data for social capital and life evaluations for 255 US metropolitan areas covering close to 80 % of the total US population, that communities with greater social engagement are happier than otherwise equivalent communities and that life evaluations fell by less, in response to unemployment increases, in those communities with high levels of a broad measure of social engagement.

Second, we divided OECD countries into three groups according to their happiness trends during and after the 2008 financial crisis. The group with rising happiness included countries less directly affected by the crisis, and with policies well chosen to enhance the well-being of their residents. The case of South Korea was given special attention as embodying policy changes that could be seen as likely to enhance subjective well-being. The group with falling happiness included those countries worst hit by the original crisis, and by its subsequent spillovers in the Euro zone. We saw that average happiness drops were far greater than could be explained by their lower levels of GDP per capita, suggesting that social capital and other key supports for happiness were damaged during the crisis and its aftermath. In Korea, by contrast, the gains in happiness were larger than predicted by the higher incomes, suggesting improvements in the quality of the social fabric, possibly linked to the shift towards a policy orientation more closely linked to well-being. Seen in this light, the Korean pairing of demand-sustaining policies with a pro-social focus of a ‘Green Korea’ (Lim 2010) may well have played a part in the striking rise in post-crisis subjective well-being in Korea.

Third, in order to dig deeper into the relative roles of social capital and income as determinants of happiness, and to see how improved social capital could improve resiliency in the face of crisis, we presented striking evidence, from the transition countries of Europe, of the power of social trust, seen here as an indicator of the quality of a country’s social capital, to increase happiness directly, and to permit a softer landing in the face of external economic shocks.

Finally, we made more direct linkages between the quality of social capital and a society’s ability to discover and implement sustainable development, including a better range of solutions to conflicts over competing uses for natural, social and human resources.