Online Annex
IDescriptives
Table 1: WVS descriptives
Variable / Obs / Mean / Std. Dev. / Min / MaxIndividual-level variables
Life satisfaction 1-10 / 262,626 / 6.63 / 2.45 / 1 / 10
Most people can be trusted / 249,887 / 0.27 / 0.45 / 0 / 1
Inc. dec. subj. est. / 237,595 / 4.55 / 2.35 / 1 / 10
Female / 262,626 / 0.52 / 0.5 / 0 / 1
Age / 262,626 / 41.1 / 16.3 / 14 / 99
Country-level variables
Year / 177 / 2001.77 / 6.92 / 1981 / 2013
Gini between / 177 / 37.34 / 8.56 / 23.19 / 60.05
Gini within / 177 / 0 / 2.36 / -14.13 / 9.33
Av. PPP GDP in country in 1000$ / 155 / 14.69 / 11.78 / 0.74 / 40.2
Dev. from av. PPP GDP in 1000$ / 143 / 0 / 3 / -10.63 / 10.63
Categorical variables
Relationship status / 259,000
Married or liv tog / 64%
Single / 25%
Widowed / 6%
Divorced / 3%
Liv. sep. / 2%
Educational level / 244,697
Less than HS / 33%
High school / 44%
More than HS / 23%
Employment status / 253,069
Full time / 36%
Part time / 8%
Self employed / 11%
Retired / 12%
Housewife / 15%
Student / 8%
Unemployed / 9%
Other / 2%
Figure 1: Development of the net income inequality measured as the Gini {Solt, 2014 #1982}, over time for all countries that have participated in more than one wave of the WVS
Table 2: CNEF descriptives
Australia (HILDA)Variable / Obs / Mean / Std. Dev. / Min / Max
Life satisfaction / 182,663 / 7.91 / 1.49 / 0 / 10
Year / 182,663 / 2007.37 / 3.86 / 2001 / 2013
Gini between / 181,732 / 32.03 / 0.5 / 30.42 / 33.26
Gini within / 165,170 / 0 / 0.9 / -2.16 / 1.92
HH net income / 182,225 / 72,967 / 55,808 / 0 / 795,943
Female / 182,663 / 0.53 / 0.5 / 0 / 1
Age / 182,663 / 43.94 / 18.51 / 14 / 101
Relationship status / 182,663
Married / 62%
Single / 24%
Widowed / 5%
Divorced / 6%
Marr. liv. sep. / 3%
Other rel. / 0%
Employment status / 182,663
Full time / 36%
Part time / 20%
Not working / 44%
Education status / 182,663
Less than HS / 36%
High school / 35%
More than HS / 29%
Table 3: CNEF descriptives
Germany (SOEP)
Variable / Obs / Mean / Std. Dev. / Min / Max
Life satisfaction / 452,423 / 7.01 / 1.83 / 0 / 10
Year / 452,423 / 1999.93 / 7.95 / 1984 / 2012
Gini between / 452,423 / 27.49 / 0.63 / 25.83 / 28.82
Gini within / 452,423 / 0 / 0.62 / -1.79 / 2.06
HH net income / 444,275 / 33,102 / 28,953 / 0 / 4,281,963
Female / 452,423 / 0.52 / 0.5 / 0 / 1
Age / 452,421 / 46.09 / 17.51 / 14 / 102
Relationship status / 452,423
Married / 62%
Single / 24%
Widowed / 6%
Divorced / 6%
Marr. liv. sep. / 2%
Other rel. / 0%
Employment status / 452,423
Full time / 41%
Part time / 21%
Not working / 39%
Education status / 452,423
Less than HS / 23%
High school / 59%
More than HS / 18%
Table 4: CNEF descriptives
Korea (KLIPS)
Variable / Obs / Mean / Std. Dev. / Min / Max
Life satisfaction / 127,742 / 3.14 / 0.7 / 1 / 5
Year / 127,742 / 2002.94 / 3.21 / 1998 / 2008
Gini between / 127,742 / 31.16 / 0.16 / 30.22 / 32.41
Gini within / 127,742 / 0 / 0.5 / -1.11 / 1.36
HH net income / 114,492 / 3,033 / 3,262 / 0 / 112,900
Female / 127,742 / 0.52 / 0.5 / 0 / 1
Age / 127,742 / 42.12 / 16.95 / 13 / 99
Relationship status / 127,742
Married / 63%
Single / 27%
Widowed / 8%
Divorced / 2%
Marr. liv. sep. / 1%
Other rel. / 0%
Employment status / 127,742
Full time / 42%
Part time / 5%
Not working / 53%
Education status / 127,742
Less than HS / 32%
High school / 37%
More than HS / 31%
Table 5: CNEF descriptives
Russia (RLMS)
Variable / Obs / Mean / Std. Dev. / Min / Max
Life satisfaction / 170,213 / 2.86 / 1.17 / 1 / 5
Year / 170,213 / 2004.71 / 4.78 / 1995 / 2011
Gini between / 170,213 / 40.86 / 0.51 / 40.19 / 44.67
Gini within / 170,213 / 0 / 0.83 / -2.23 / 3.85
HH net income / 160,428 / 146,339 / 615,532 / 0 / 45,200,000
Female / 170,213 / 0.57 / 0.49 / 0 / 1
Age / 170,212 / 38.4 / 17.3 / 12 / 70
Relationship status / 170,213
Married / 58%
Single / 21%
Widowed / 12%
Divorced / 9%
Marr. liv. sep. / 0%
Other rel. / 0%
Employment status / 170,213
Full time / 54%
Part time / 0%
Not working / 46%
Education status / 170,213
Less than HS / 21%
High school / 53%
More than HS / 27%
Table 6: CNEF descriptives
Switzerland (SHP)
Variable / Obs / Mean / Std. Dev. / Min / Max
Life satisfaction / 96,716 / 8.03 / 1.44 / 0 / 10
Year / 96,716 / 2006.71 / 4.04 / 2000 / 2013
Gini between / 96,475 / 29.25 / 0.71 / 26.74 / 31.48
Gini within / 89,505 / 0 / 1.29 / -3.22 / 3.37
HH net income / 96,385 / 95,443 / 70,333 / 0 / 5,487,294
Female / 96,716 / 0.55 / 0.5 / 0 / 1
Age / 96,714 / 45.79 / 18.18 / 13 / 98
Relationship status / 96,716
Married / 66%
Single / 22%
Widowed / 5%
Divorced / 6%
Marr. liv. sep. / 1%
Other rel. / 0%
Employment status / 96,716
Full time / 41%
Part time / 25%
Not working / 34%
Education status / 96,716
Less than HS / 23%
High school / 48%
More than HS / 29%
Table 7: CNEF descriptives
United Kingdom (BHPS)
Variable / Obs / Mean / Std. Dev. / Min / Max
Life satisfaction / 160,518 / 5.23 / 1.29 / 1 / 7
Year / 160,518 / 2002.46 / 3.69 / 1996 / 2008
Gini between / 160,518 / 34.64 / 0.28 / 33.88 / 35.81
Gini within / 160,518 / 0 / 0.56 / -1.23 / 1.41
HH net income / 116,462 / 23,178 / 15,739 / 0 / 726,208
Female / 160,518 / 0.55 / 0.5 / 0 / 1
Age / 160,518 / 45.4 / 18.52 / 15 / 100
Relationship status / 160,518
Married / 64%
Single / 21%
Widowed / 7%
Divorced / 6%
Marr. liv. sep. / 2%
Other rel. / 0%
Employment status / 160,518
Full time / 41%
Part time / 19%
Not working / 40%
Education status / 160,518
Less than HS / 0
High school / 0
More than HS / 0
IIRobustness test: pooled country-level panel data
This section tests the results of the multilevel models (Table 1 in the main paper), by pooling data from the World Values Survey at the country-level, thereby creating a country-level panel dataset. Table 4below therefore operates with much fewer cases (country-years instead of individuals) than the multilevel models.
Table 8: Regressions with countries that originally have a 0-10 scale of life satisfaction
By using only country-level data, Table 8shows very similar results as the multilevel models of Table 1 in the main text. Model 1 shows virtually the same effect as the multilevel Model 1 of Table 1 in the main text. The hybrid Model 2 again splits up the Gini in a between and within component. Similar to Model 2 of Table 1 in the main paper, it shows that people lose .085 life satisfaction points when the Gini is one point above the average Gini that they have experienced in their country. However, the non-significant between effect again shows that life satisfaction does not differ significantly between countries with more or less income inequality.
Model 3 controls for GDP and country-level average individual values, similar to Model 4 of Table 1 in the main paper. It again shows very similar results. The main effect that becomes visible here is that people are not more satisfied when their country is more equal than another country. Instead, people are more unsatisfied when more inequality in their country exists compared to what typically exists in their country.[1]
IIIComparison of means and standard deviations of WVS and CNEF life satisfaction scales 1-10
Table 9: Mean and standard deviation of WVS and constructed CNEF 1-10 data compared
WVS / CNEF / differenceAustralia / Mean / 7.54 / 7.91 / 0.37
SD / 1.87 / 1.49 / 0.38
Germany / Mean / 6.93 / 7.02 / 0.09
SD / 1.97 / 1.8 / 0.17
Switzerland / Mean / 8.11 / 8.03 / 0.08
SD / 1.75 / 1.43 / 0.32
Korea / Mean / 6.48 / 6.28 / 0.2
SD / 2.15 / 1.4 / 0.75
Russia / Mean / 5.56 / 5.72 / 0.16
SD / 2.47 / 2.33 / 0.14
UK / Mean / 7.57 / 8.03 / 0.46
SD / 1.81 / 1.43 / 0.38
IVRobustness test: CNEF ordered logistic, effect of inequality in different countries, original scales
Table 10: Regressions with countries that originally have a 0-10 scale of life satisfaction
Table 11: Regressions with countries that originally have a 1-5 (Korea, Russia) or 1-7 (UK) scale of life satisfaction
VRobustness test: all countries in one regression with country-varying slope
Table 12: Regressions with interaction effect inequality and country
When combining the effect of the Gini with the interaction effects of the Gini in each country, one sees that the effect sizes of Table 12 are almost exactly the same as in Table 2 and Table 3 of the main paper, which used separate regressions for each country. This suggests that the integration of the different scales does not influence the results.
VIRobustness test: different country panels in one regressions without country-specific intercepts
Table 13: Regressions that leave out country dummies
In Table 13, I left out the country dummies from the regression in the main paper, which give each country its own intercept in the main paper. This assumes that all differences in life satisfaction between countries are due to differences in inequality between countries, which makes the between effects very significantly negative, while the within effects also remain very significantly negative, which should be the case, as the within effect looks at changes within countries, not at differences between them.
VIIRobustness test: calculations with richest quartile in each country-year
Table 14: Separate calculations for upper 25 percent of incomes
Table 15: Separate calculations for upper 25 percent of incomes
Table 16: Panel studies taken together, upper 25 percent of incomes
The three tables above test whether even the richest 25 percent of each country and year feel worse when inequality is higher. They do this by rerunning the regressions of Table 2, 3 and 4 in the main paper, with only the highest 25 percent of income earners in each country-year.
The results indicate that even among the richest quartile, life satisfaction is significantly lower in those years where inequality is higher – apart from Korea, as in the main results. The between effect of inequality on life satisfaction is unstable, as in the main paper. It is sometimes negative, sometimes positive and sometimes insignificant. This means that even the richest quartile of the population is less satisfied with life in those years where inequality is higher than what they are used to. But the richest quartile of the populations is not less satisfied with life when living in a time period or country where the long-run level of inequality is higher.
VIIIRobustness test: Regressions with z-standardized values
Table 17: Linear regression with z-standardized values from countries that originally have a 0-10 scale of life satisfaction
Table 18: Linear regression with z-standardized values from countries that originally have a 1-5 or 1-7 scale of life satisfaction
Table 19: Linear regression with z-standardized values and all countries
IXRobustness test: first differences regression
Table 20: First differences influence of inequality on life satisfaction before controls
Table 21: First differences influence of inequality on life satisfaction after controls
The above tables show the results from first differences regressions, showing how changes in life satisfaction result from changes in inequality in the preceding year. Such first differences regressionsare very demanding in terms of causality, as they estimate whether, one year after inequality increases or decreases relative to the previous year, life satisfaction also increases or decreases relative to the previous year. For an effect to be significant, changes in life satisfaction therefore have to follow changes in inequality with a time lag of exactly one year. The tests show that such an exact negative link indeed exists in four of the six countries, as well as when analyzing all countries in one regression. Only in Australia and Russia does this close link not exist. Keep in mind that a first differences regression only shows results when life satisfaction reacts to changes in inequality with a time lag of exactly one year. Discussions of first differences regressions therefore conclude that the tight coupling between variables that they require is rarely found in the social sciences (Giesselmann and Windzio, 2012: 65). It is therefore surprising that changing life satisfaction in four countries, and when analyzing all countries together, still influences life satisfaction, even when using a demanding first differences estimator.
1
[1]Note that China and Russia are outliers, as they have stronger changes in inequality than other countries. However, I have rerun all regressions by excluding Russia and China and the results remain the same.