Money and happiness: the income-happiness correlation is higher when income inequality is higher — TEI-tagged view

Shigehiro OishipersName, Youngjae ChapersName, Asuka KomiyapersName, Hiroshi OnopersName (2022date). “Money and happiness: the income-happiness correlation is higher when income inequality is higher.” PNAS Nexustitle·j.

Object of study — the income-happiness correlationterm·dom over five decades in the USA, Japan, 16 European and 18 Latin American countries

Research question — whether the income-happiness correlationterm·dom has changed over time and what predicts such changes

Contribution — evidence that the income-happiness correlationterm·dom rose with GDP per capitaterm·key and income inequalityterm·dom in the USA and Europe

Full text — tagged

The reference list is not part of the text we hold for this paper. Our copy was extracted from the publisher's page and the bibliography did not survive that extraction, so the article below ends before its references. Nothing has been removed by us.

Money and happinessterm·key: the income-happiness correlationterm·dom is higher when income inequalityterm·dom is higher Abstract Has the income–happiness correlationterm·dom changed over time? If so, what predicts such changes? We tested these questions in diverse economic, political, and cultural contexts. Drawing on nationally representative data, we found that the income–happiness correlationterm·dom has increased in the USA since 1972, as GDP per capitaterm·key and income inequalityterm·dom increased (Study 1). Study 2 examined an income–life satisfactionterm·dom correlation in nationally representative surveysterm·mth between 1978 and 2011 in Japan. Unlike in the USA, there was no clear increase in the income–life satisfactionterm·dom correlation over time. We next examined the income–life satisfactionterm·dom correlations in 16 European countries and found that on average the income–life satisfactionterm·dom correlation has increased since 1970, and it was particularly high in years of high GDP per capitaterm·key and high-income inequality (Study 3). Finally, we found that among Latin American countries, the income–life satisfactionterm·dom correlation has, on average, decreased since 1997, as income inequalityterm·dom has decreased (Study 4). Over the last 5 decades, the income–happiness correlationterm·dom has increased, not decreased, in the USA and several European countries. The income–happiness correlationterm·dom tends to get higher when both GDP per capitaterm·key and income inequalityterm·dom are high, whereas it tends to get lower when GDP per capitaterm·key and/or income inequalityterm·dom are low. These findings suggest the importance of accounting for income inequalityterm·dom as well as national wealth in understanding the role of money in happinessterm·key.

Past research examined an income–happiness correlationterm·dom in one cultural context over one time period, limiting our knowledge regarding whether the nature of the income–happiness correlationterm·dom has changed over a long period of time, and if so, what predicts such changes. We found that the income–happiness correlationterm·dom has increased in the USA and Europe since the 1970s, as both GDP per capitaterm·key and income inequalityterm·dom increased. In contrast, the income–life satisfactionterm·dom correlation has decreased since 1997 among Latin American countries, where income inequalityterm·dom has decreased during that time. The income–happiness correlationterm·dom tends to get higher when GDP per capitaterm·key and income inequalityterm·dom are high while getting smaller as GDP per capitaterm·key and income inequalityterm·dom get smaller.

Introduction Does money buy happinessterm·key? This question has received extensive empirical attention since Richard Easterlin’s seminal work in 1974 (e.g. 1–3 ; see 4–5 for reviews). Easterlin ( 6 ) looked at results from 30 national surveysterm·mth in 20 countries and concluded that “there is a clear indication that income and happinessterm·key are positively associated” (p. 99), and that “in every single surveyterm·mth, those in the highest status group were happier, on the average, than those in the lowest status group” (p. 100). Diener and OishipersName ( 7 ) examined the income–happiness correlationterm·dom in 40 countries and found that the mean income–happiness correlationterm·dom was 0.13. According to Diener and Biswas-Diener ( 4 ), the income–happiness correlationterm·dom in a national surveyterm·mth ranged from 0.12 to 0.18 in the USA, 0.06 to 0.15 in West Germany, and 0.17 to 0.27 in the Russian Federation. A recent meta-analysis of 335 studies found that the mean income–happiness correlationterm·dom was 0.23 ( 8 ).

Inglehart and colleagues ( 9 ) analyzed data from 52 countries and found that the income–happiness correlationterm·dom was stronger in poorer countries than in richer countries. That is, money seems to buy more happinessterm·key in poorer countries than in richer countries. In this light, it is interesting to note that the income–happiness correlationterm·dom was 0.45 among 83 residents in the slums of Calcutta ( 10 ). The main idea is that among those who are struggling to meet their basic needs, more money means greater access to basic goods (e.g. drinking water, food, shelter). In contrast, it is believed that once the basic needs are met, more money does not necessarily help increase one’s happinessterm·key ( 9 , 11 ).

Interestingly, recent research on money and happinessterm·key found that money seems to buy more happinessterm·key even among wealthy individuals whose basic needs are met. For instance, Killingsworth ( 12 ) analyzed the experience sampling data (i.e. momentary reports of happinessterm·key) from 33,391 employed, working-age adults living in the USA and found that self-reported happinessterm·key continues to increase as the participants’ household income increases, even beyond \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 120,000. Moreover, Jebb and colleagues ( 13 ) analyzed the data from 164 countries and found that the income–happiness correlationterm·dom was larger in wealthier countries rather than in poorer countries.

These recent findings suggest that the income–happiness correlationterm·dom within a country (e.g. the USA, Argentina) might have changed over time. For instance, it is possible that in the USA, the income–happiness correlationterm·dom was smaller in the 1970s and 1980s when the earlier data were being collected than in the 2010s and the 2020s when more recent data were collected. In contrast, it is possible that in some countries, the income–happiness correlationterm·dom has become smaller over time, as their national economy grew. These divergent patterns could explain why the earlier cross-country studies (e.g. 9 ) found that the income–happiness correlationterm·dom was larger among poorer than richer countries, whereas it was larger among richer rather than poorer countries in more recent cross-country studies (e.g. 13 ).

To our knowledge, none of the previous research has examined whether the income–happiness correlationterm·dom has changed over time. This is a critical oversight because the importance of money changes over time in a given society ( 14 , 15 ). Some countries might be becoming more materialistic, for example, while others might be becoming less materialistic over time ( 16 ).

Past research showed that the satisfaction of important life domains is more strongly associated with overall life satisfactionterm·dom than the satisfaction of less important life domains ( 17 ). For instance, many people deem family relationships very important, whereas some do not. When the authors computed the correlation between family relationship satisfaction and life satisfactionterm·dom, it was significantly stronger among those who deemed family relationships important than among those who did not. In the context of the present research, then, the income–happiness correlationterm·dom should be larger as the importance of money increases, while it should be smaller as the importance of money decreases.

When does the importance of money increase or decrease? Based on Inglehart and colleagues’ findings ( 9 ), the end of materialism hypothesisterm·dom predicts that the income–happiness correlationterm·dom should get smaller as a society gets richer. This is because most people in a wealthy society are presumably no longer concerned about money per se, and instead are concerned about non-material issues such as self-expression. Under such a condition, self-expression should become a stronger predictor of happinessterm·key than money per se.

In contrast, there is an alternative possibility; as a society gets richer, people’s desires for material goods could also grow ( 18–21 ), and the importance of money could get even larger. For instance, some researchers found that American youths became more materialistic from the 1970s to the 1990s and their materialismterm·key remained high since then ( 15 ). In this scenario under the continuous materialism hypothesisterm·dom, the income–happiness correlationterm·dom should get larger as the economy grows.

Another important economic factor is the distribution of the national income. If the benefits of economic growth go disproportionately to the wealthy, then the poor feel that the world is unfair and report less happinessterm·key over time (e.g. 22 ). The larger the difference in happinessterm·key between the rich and the poor, the larger the correlation between income and happinessterm·key. In addition, growing income inequalityterm·dom is often driven by the rich getting richer ( 23 ), which in turn, provides more tendencies for ordinary citizens to engage in unfavorable, upward social comparisonsterm·dom ( 24 ). The frequency of upward social comparisonterm·dom is known to be detrimental to one’s happinessterm·key ( 25 ). In the current context, then, the economic gap between the rich and the poor should increase people’s tendency to engage in upward social comparisonterm·dom, which should in turn increase the happinessterm·key gap between the rich and the poor.

Moreover, income inequalityterm·dom is associated with more perceived competition ( 26 ), zero-sum thinking ( 27 ), and status anxiety ( 28 ). If income inequalityterm·dom has increased over time in a given country, then, the concern for money might have increased also. Indeed, previous research found that an increase in income inequalityterm·dom was associated with an increase in work hours ( 29 ). The increased concern for money, in turn, could translate into a larger income–happiness correlationterm·dom (hereafter we call it the income inequality hypothesisterm·dom). Furthermore, income inequalityterm·dom tends to be larger in countries where social welfare spending is low (e.g. Greece) than where social welfare spending is high (e.g. Sweden) because many welfare systems are funded through progressive taxation ( 30 ).

In a cross-sectional study of 29 countries, OnopersName and Lee ( 31 ) found that the income–happiness correlationterm·dom was higher in countries where public social expenditure was lower than where it was higher. OnopersName and Lee’s findings are consistent with the hypothesis that the income–happiness correlationterm·dom should be higher when income inequalityterm·dom is higher. As stated above, a shortcoming of the previous research is that it has not examined changes in the income–happiness correlationterm·dom over time.

In sum, the main goals of the present research are to explore historical changes in the income–happiness correlationterm·dom in diverse cultural, economic, and political contexts: the USA (Study 1), Japan (Study 2), Europe (Study 3), and Latin America (Study 4), and to test whether the historical changes in the income–happiness correlationterm·dom are associated with GDP per capitaterm·key and income inequalityterm·dom. The end of materialism hypothesisterm·dom states that the income–happiness correlationterm·dom will get smaller with an increase in GDP per capitaterm·key. In contrast, the continuous materialism hypothesisterm·dom states that the income–happiness correlationterm·dom will get larger with an increase in GDP per capitaterm·key. The income inequality hypothesisterm·dom provides an alternative explanation for these conflicting hypotheses, instead predicting that the income–happiness correlationterm·dom will get larger with an increase in income inequalityterm·dom.

Study 1: The USA 1972 to 2018 The General Social Surveysterm·mth (GSS) have included the 3-point happinessterm·key scale since its inception in 1972. We used this item with the self-reported household income from all the surveyterm·mth years. Because the value of the dollar has changed dramatically over time, the income categories in 1972 are not comparable to those in 2018. Furthermore, different years have different income categories (e.g. in 1972, there were 12 income categories; whereas in 1977 there were 20 categories). This inconsistency in response categories across years makes it questionable to use the income variable across different surveyterm·mth years in a single multilevel analysis. Thus, we used the meta-analytic framework, treating each year’s income–happiness correlationterm·dom as an independent “study” result, as opposed to the multilevel framework in which the income variable has to be treated as equivalent across years. Finally, the original happinessterm·key scale was reversed to make the higher score indicate more happinessterm·key.

Figure 1 (left, panel A) shows the income–happiness correlationterm·dom as a function of the surveyterm·mth year in the USA. The income–happiness correlationterm·dom has steadily increased since the 1970s. Interestingly, however, in the post-Lehman shock surveysterm·mth from 2010, 2012, and 2014, the correlations decreased. Fisher z -transformed income–happiness correlationterm·dom coefficient was significantly associated with the surveyterm·mth year ( r [30] = 0.65, P < 0.001) and the log-transformed GDP per capitaterm·key ( r [30] = 0.66, P < 0.001). Overall, the patterns of the results are generally consistent with the continuous materialism hypothesisterm·dom, as the income–happiness correlationterm·dom increased over the last 50 years during which GDP per capitaterm·key has steadily increased.

The log income–happiness correlationterm·dom by surveyterm·mth year in the USA and in Japan. Note: The x -axis is the surveyterm·mth year in each country (begins at the first surveyterm·mth year available in each country: 1972 for the USA; 1978 for Japan). The y -axis is the correlation between the log of income and happinessterm·key (Fisher z -transformed). The error bar indicates the SE of the log income–happiness correlationterm·dom. The blue line depicts the total trend (USA: y = 0.0016x–2.9762, R 2 = 0.4224; Japan: y = 0.0015x–2.7704; R 2 = 0.0817).

Next, we tested the income inequality hypothesisterm·dom. To the extent that the rich got richer, while the poor remained poor, the difference between the rich and the poor in the USA widened. The widening of the wealth gap might have translated into the widening of the happinessterm·key gap between the rich and the poor. As seen in Fig. 2 , the Gini coefficientsterm·key were highly positively correlated with the income–happiness correlationterm·dom ( r [30] = 0.66, P < 0.001). That is, in the years of greater income inequalityterm·dom, the correlation between income and happinessterm·key was stronger. The formal test with the classic meta-analysis function of the JASP program, using the restricted maximum likelihood estimation, showed that the household income inequalityterm·dom (Gini) significantly moderated the income–happiness correlationterm·dom, b = 0.767 (SE = 0.149), z = 5.15, and P < 0.001. There was a substantial degree of heterogeneity in the income–happiness correlationsterm·dom between 1972 and 2018, Q (1, 31) = 73.20, and P < 0.001. Once we entered the Gini coefficientterm·key as the moderator, there was no longer heterogeneity in the remaining income–happiness correlationsterm·dom, Q (1, 30) = 38.35, P = 0.141.

The log income–happiness correlationterm·dom by the Gini coefficientterm·key in the USA and in Japan. Note . The x -axis indicates the Gini coefficientterm·key. The y -axis is the correlation between the log of income and happinessterm·key (Fisher z -transformed). The error bar indicates the SE of the log income–happiness correlationterm·dom. The blue line depicts the total trend (USA: y = 0.5921x + 0.3215, R 2 = 0.4281; Japan: y = 0.0232x + 0.237; R 2 = 0.010).

Alternative specifications Because the Gini coefficientterm·key is an omnibus index of inequality, it does not specify whether the above findings are driven mostly by the top 10% or the bottom 50%. We used the top 10% share and the bottom 50% share of the national income, respectively, as the moderator. The results were similar to the Gini findings. The income–happiness correlationterm·dom was larger when the top 10% share of the national income was higher, b = 0.53 (SE = 0.11), z = 4.96, and P < 0.001. With the top 10% share of the national income included in the analysis, there was no longer heterogeneity in the remaining income–happiness correlationsterm·dom, Q (1, 30) = 39.75, P = 0.110. Conversely, the income–happiness correlationterm·dom was smaller when the bottom 50% share of the national income was larger, b = −0.86 (SE = 0.18), z = −4.76, and P < 0.001. With the bottom 50% share of the national income included in the analysis, there was no longer significant heterogeneity, Q (1, 30) = 41.09, P = 0.085. Thus, using the alternative specifications, the income inequality hypothesisterm·dom was supported in Study 1. In sum, Study 1 found support for the continuous materialism hypothesisterm·dom and the income inequality hypothesisterm·dom and did not find any support for the end of materialism hypothesisterm·dom.

Study 2: Japanese lifestyle surveysterm·mth 1978 to 2011 Study 1 found patterns of the income–happiness correlationterm·dom consistent with the continuous materialism hypothesisterm·dom and the income inequality hypothesisterm·dom, and inconsistent with the end of materialism hypothesisterm·dom. As the economy grew, the income–happiness correlationterm·dom increased rather than decreased over time. The USA has one of the most extreme forms of the winner-take-all economy and politics ( 32 , 33 ). Thus, the USA could be an exception. Namely, it is possible that in other developed countries, the end of materialism hypothesisterm·dom could still hold. Japan provides an ideal testing ground, as it is similar to the USA in terms of the GDP per capitaterm·key, but different in many other dimensions. Most central to this research, the degree of income inequalityterm·dom is substantially smaller in Japan than in the USA (e.g. 34 ), and the Japanese are more aversive toward income inequalityterm·dom than Americans ( 35 ).

In order to test whether the increasing income–happiness correlationterm·dom is specific to the USA, we next analyzed the Japanese Lifestyle Surveysterm·mth, the nationally representative surveysterm·mth conducted by the Japanese government from 1978 to 2011. The same 5-point scale life satisfactionterm·dom question (“Are you satisfied or dissatisfied with your life in general?”) was included except in 2007, 2010, and 2011.

Figure 1 (right, panel B) shows the income–happiness correlationterm·dom as a function of the surveyterm·mth year in Japan. Unlike in the USA, the income–life satisfactionterm·dom correlation has not increased over time between 1978 and 2009 ( r [18] = 0.29, P = 0.222). The Fischer-transformed income–life satisfactionterm·dom correlation was not correlated with log-transformed GDP per capitaterm·key ( r [18] = 0.03, P = 0.895). Unlike Study 1, the income–life satisfactionterm·dom correlation in Japan was not correlated with the Gini coefficientterm·key ( r [17] = 0.10, P = 0.675), as well. Similarly, neither the top 10% share of the national income nor the bottom 50% share of the national income were associated with the size of the income–life satisfactionterm·dom correlation: r (18) = 0.35, P = 0.131 for the top 10%; r (17) = −0.32, P = 0.182 for the bottom 50%. In sum, the patterns of the income–happiness correlationsterm·dom in Japan were not consistent with any hypotheses.

Study 3: Eurobarometer surveysterm·mth 1970 to 2002 and European social surveysterm·mth 2004 to 2018 The USA and Japan showed different patterns of the income–happiness correlationsterm·dom over time. This could be due to the divergent pattern of economic growth in the USA and Japan: while the USA economy grew more or less linearly between 1972 and 2018 (Study 1), the Japanese economy stagnated between 1991 and 2011 (Study 2). In order to test the generalizability of our initial findings among Americans and Japanese, we examined the income–life satisfactionterm·dom correlations over time among 16 European countries. We analyzed the Eurobarometer surveysterm·mth and the European Social Surveysterm·mth (ESS). The Eurobarometer included the 4-point scaleterm·mth life satisfactionterm·dom question in most years, whereas the ESS included the 10-point scale life satisfactionterm·dom question in most years. We chose the countries included both in the Eurobarometer and ESS in the following analyses. (Eurobarometer and ESS contain data from countries have changed their names/boundaries historically [i.e. Great Britain and West-Germany in Eurobarometer; the UK and Germany in ESS]. For consistency, we used “the UK” and “Germany” across the studies.) This resulted in 16 countries with a total income–life satisfactionterm·dom correlation of 414. Many of the recent Gini coefficientsterm·key were not available (for instance, the World Bank’s database [ https://data.worldbank.org/indicator ] does not have any Gini coefficientsterm·key for France before 1978; furthermore, years 1979 to 1988 are missing; in contrast, the World Inequality database [ https://wid.world/ ] has the top 10% share and the bottom 50% share of the national income every year since 1915 for France), whereas the top 10% share and the bottom 50% share were consistently available ( https://wid.world/ ). Thus, in Studies 3 and 4, we used these two indices of income inequalityterm·dom as opposed to the Gini coefficientterm·key.

Like in Studies 1 and 2, we first computed an income–life satisfactionterm·dom correlation for each year for each country (e.g. r = 0.13 in 1976, r = 0.11 in 1977, r = 0.14 in 1978 etc. in France; r = 0.17 in 1976, r = 0.20 in 1977, r = 0.18 in 1978 etc. in Belgium). We then computed a correlation between income–life satisfactionterm·dom correlations (Fisher z -transformed) and the surveyterm·mth year, GDP per capitaterm·key (log-transformed), the top 10%, and the bottom 50% share in each of the 16 countries, respectively. Table 1 shows these correlations. Figure 3 shows the income–happiness correlationterm·dom as a function of the surveyterm·mth year across countries in Europe.

The log income–happiness correlationterm·dom by surveyterm·mth year by Country in Europe (1970 to 2018). Note . Inc-LS Cor. Indicates the correlations between log income and life satisfactionterm·dom (Fisher z -transformed). The error bar indicates the SE of the correlation coefficient.

The correlations between the log income–happiness correlationsterm·dom and the surveyterm·mth year (yearly trend), GDPpc (LN), and two-income inequality indices (top10% share of the national income and bottom 50% share of the national income).

The first column of Table 1 shows the correlation between the income–life satisfactionterm·dom correlation and the year of the surveyterm·mth. A positive correlation (e.g. r = 0.78 in Germany, 0.74 in France) indicates that the income–life satisfactionterm·dom correlation increased over time in that country, whereas, a negative correlation (e.g. r = −0.28 in Norway) indicates that the income–life satisfactionterm·dom correlation decreased over time in that country. Like in the USA, the income–life satisfactionterm·dom correlation has substantially increased between 1970 and 2018 in France, Germany, Finland, Portugal, the UK, and the Netherlands ( r s ranged from 0.42 to 0.78). In contrast, the income–life satisfactionterm·dom correlation has decreased in Norway. We meta-analyzed the correlations between the within-country income–life satisfactionterm·dom correlation (Fisher z -transformed) and the year of the surveyterm·mth, with the random-effects model, using the meta package in R. The meta-analytic mean correlation of yearly trends was 0.44 (95% CI: 0.29; 0.57; t = 5.80, P < 0.001) among 16 European countries. That is, on average, the income–happiness correlationterm·dom increased over time among 16 European countries.

The second column of Table 1 shows the correlation between the income–life satisfactionterm·dom correlation and GDP per capitaterm·key (log-transformed). A positive correlation (e.g. r = 0.72 in Germany, r = 0.67 in Portugal) indicates that the income–life satisfactionterm·dom correlation was larger when GDP per capitaterm·key was higher, whereas a negative correlation (e.g. r = −0.36 in Norway) indicates that the income–life satisfactionterm·dom correlation was smaller when GDP per capitaterm·key was lower. Our results showed that the size of an income–life satisfactionterm·dom correlation was highly positively correlated with GDP per capitaterm·key (log-transformed) in Portugal, France, Germany, the Netherlands, and the UK ( r s ranged from 0.38 to 0.72). The meta-analytic mean correlation of the income–life satisfactionterm·dom correlation with GDP per capitaterm·key was 0.35 (95% CI: 0.20; 0.49; t = 4.81, P < 0.001) among 16 European countries. These findings are consistent with the continuous materialism hypothesisterm·dom, and inconsistent with the end of materialism hypothesisterm·dom.

Next, we examined whether the income–life satisfactionterm·dom correlations were larger in the year of larger income inequalityterm·dom. The third column of Table 1 shows the correlation between the income–life satisfactionterm·dom correlation and the top 10% share of the national income. A positive correlation (e.g. r = 0.79 in Germany) indicates that the income–life satisfactionterm·dom correlation got larger, as the top 10% share grew, whereas a negative correlation (e.g. r = −0.71 in Sweden) indicates that the income–life satisfactionterm·dom correlation got smaller, as the top 10% share grew. Like in the USA, the income–life satisfactionterm·dom correlations were substantially higher in the year of a higher share of the national income being held by the richest 10% of the population in Germany, Portugal, Austria, Greece, and the UK. In contrast, it was smaller in the year with a higher top 10% share in Sweden. Overall, the meta-analytic mean correlation of the income–life satisfactionterm·dom correlation with the top 10% share was 0.22 (95% CI: −0.03 0.44; t = 1.85, P = 0.083). That is, across 16 European countries, on average, the income–life satisfactionterm·dom correlation tended to be larger in the year of the higher concentration of the national wealth in the richest 10% of the population.

The last column of Table 1 indicates the correlation between the income–life satisfactionterm·dom correlation and the bottom 50% share of the national income. A negative correlation (e.g. r = −0.79 in Germany) in this column means that the income–life satisfactionterm·dom correlation was smaller in the year of the larger bottom 50% share of the national income. Like in the USA, the income inequality hypothesisterm·dom was supported in Germany, Finland, the Netherlands, Portugal, and Austria. In contrast, it was not supported in Belgium, the UK, Ireland, Spain, and Norway. Overall, the meta-analytic mean correlation of the income–life satisfactionterm·dom correlation with the bottom 50% share was −0.26 (95% CI: −0.45; −0.05; t = −2.65, P = 0.018). Thus, on average, the income–life satisfactionterm·dom correlations were larger when the bottom 50% share was smaller, consistent with the income inequality hypothesisterm·dom.

In sum, the meta-analyses showed that the income–life satisfactionterm·dom correlations between 1970 and 2018 among 16 European countries on average were similar to those in the USA. Like in the USA, the income–life satisfactionterm·dom correlations increased over time. The income–life satisfactionterm·dom correlation was higher in the year of higher GDP per capitaterm·key. It was marginally higher in the year of higher top10% share, whereas significantly lower in the year of higher bottom 50% share of the national income. The European data on average provided support for the continuous materialism hypothesisterm·dom and the income inequality hypothesisterm·dom, and no support for the end of materialism hypothesisterm·dom.

Study 4: Latinobarometro 1997 to 2018 Studies 1 to 3 showed that the income–life satisfactionterm·dom correlation (a) increased over time in the USA and European countries on average, (b) coincided with an increase in GDP per capitaterm·key, and (c) an increase in income inequalityterm·dom in the USA and European countries on average. We explored whether the patterns of the income-happiness/life satisfactionterm·dom correlations found in the USA and European countries are generalizable among Latin American countries. In the USA and several European countries, income inequalityterm·dom has increased as the economy grew over the last 40 years. Latin America is an interesting case in that, unlike in the WEIRD (Western Educated Industrialized Rich Democratic) countries, income inequalityterm·dom has decreased over the last 40 years as the economy grew ( 36 , 37 ). We tested the trends of the income–life satisfactionterm·dom correlation over time among 18 Latin American countries over the last 2 decades.

We analyzed the Latinobarometro data from 1997 to 2018. Life satisfactionterm·dom was assessed with the 4-point scaleterm·mth (1 = not very satisfied to 4 = very satisfied). Income was not directly assessed. However, respondents’ assessment of their income was assessed every year using the 4-point scaleterm·mth (“Does your salary and the total of your family’s salary allow you to satisfactorily cover your needs?” 1 = Does not cover, there are great difficulties to 4 = covers them well, I can save”). We used these two items to calculate the correlation between income and life satisfactionterm·dom each year for each country.

Like in Studies 1 to 3, the income–life satisfactionterm·dom correlation was computed for each year for each country, separately. First, whereas the income–life satisfactionterm·dom correlations have grown larger over time in the USA and several European countries, on average, the income–life satisfactionterm·dom correlation has become smaller among Latin American countries (see Fig. 4 ), as indicated by many negative correlations in the first column of Table 1 : the meta-analytic r = −0.15 (95% CI: −0.28; −0.02; t = −2.43, P = 0.027).

The log income–happiness correlationterm·dom by surveyterm·mth year by country in Latin America (1997 to 2018). Note . Inc-LS Cor. Indicates the correlations between log income and life satisfactionterm·dom (Fisher z -transformed). The error bar indicates the SE of the correlation coefficient.

Furthermore, the second column of Table 1 shows that, unlike the USA and several European countries, the income–life satisfactionterm·dom correlation was not associated with GDP per capitaterm·key among 18 Latin American countries: meta-analytic r = −0.04 (95% CI: −0.18; 0.10, t = −0.63, P = 0.53). Finally, the patterns of the income–life satisfactionterm·dom correlation with income inequalityterm·dom indices (the 3rd and 4th columns of Table 1 ) were similar to those found in the USA and Europe. The income–life satisfactionterm·dom correlation was smaller when the bottom 50% share was larger: meta-analytic r = −0.16 (95% CI: −0.31; 0.003; t = −2.07, P = 0.054), although the correlation with the top 10% share of income was not statistically significant (meta-analytic r = 0.09 [95% CI: −0.05; 0.23; t = 1.38, P = 0.186]). Thus, Latin American data were mostly consistent with the income inequality hypothesisterm·dom.

The multilevel meta-analyses of studies 1 to 4 We next meta-analyzed the income-happiness/life satisfactionterm·dom correlations from the USA, Japan, 16 European countries, and 18 Latin American countries simultaneously to get an overall estimate of the effect size across 36 countries. There were 768 income-happiness/life satisfactionterm·dom correlations from 36 countries. We meta-analyzed them with the multilevel model, using the metafor package in R (Level 1 = within-country; Level 2 = between-country). The base model in which the income–happiness correlationterm·dom was predicted by year only showed that the intercept was 0.173 ( SE = 0.009), t (766) = 19.85, P < 0.001. On average, the z -transformed income–happiness correlationterm·dom was 0.173 in the first year of the surveyterm·mth across 36 countries, which were slightly higher than Diener and OishipersName’s ( 7 ) earlier review ( r = 0.13, which did not include Latin America) and lower than the most recent meta-analysis: r = 0.23 ( 8 ). The slope for the year was 0.001 ( SE = 0.0014), t = 7.50, P < 0.001, meaning that one year corresponded to an increase of 0.001 in the income–happiness correlationterm·dom on average. The results from 3 different sets of meta-analyses are shown in Table 2 . Model 1 included year, three dummies (dummy 1 = Japan; dummy 2 = Europe, 3 = Latin America), and interaction terms (region dummies with the yearly trend). The results from Model 1 showed that (a) the income–happiness correlationsterm·dom in the USA were on average higher than Japan, Europe, and Latin America, (b) that there was no difference between Japan and the USA, nor a difference between Europe and the USA in the yearly trend, but (c) that the yearly trend of Latin American countries was significantly different from that of the USA.

Multilevel meta-analyses of studies 1 to 4.

Model 2 tested whether the income–happiness correlationsterm·dom were larger in the years of higher levels of GDP per capitaterm·key and the top 10% of share simultaneously (Model 2). Controlling for the regional differences in the mean income–happiness correlationterm·dom and top 10% share, one standard deviation increase in GDP per capitaterm·key (e.g. 1 SD of log GDP was 0.22 for the USA) corresponded to a 0.009 increase in the income–happiness correlationterm·dom, t = 3.79, P < 0.001. Thus, across 36 countries, on average, the income–happiness correlationterm·dom was larger when GDP per capitaterm·key was higher, supporting the continuous materialism hypothesisterm·dom as opposed to the end of materialism hypothesisterm·dom. In addition, this analysis showed that one standard deviation increase in the top 10% share of the national income was, on average, associated with a 0.009 increase in the income–happiness correlationterm·dom, t = 4.33, P < 0.001, supporting the income inequality hypothesisterm·dom (for the time trends in GDP and the top 10% share across countries, see Figures S1 to S8, Supplementary Material 2 ).

Model 3 tested whether the income–happiness correlationterm·dom was larger in the years of a smaller share of the bottom 50% across the 36 countries. The results from Model 3 showed that controlling for the regional differences and the bottom 50% share, the income–happiness correlationterm·dom was larger in the year of higher GDP per capitaterm·key, again supporting the continuous materialism hypothesisterm·dom. One standard deviation increase in GDP per capitaterm·key was associated with a 0.008 increase in the income–happiness correlationterm·dom. Consistent with the income inequality hypothesisterm·dom, the income–happiness correlationterm·dom was smaller in the year of a larger bottom 50% share of the national income (i.e. less income inequalityterm·dom) than in the year of a smaller bottom 50% share of the national income (i.e. more income inequalityterm·dom). One standard deviation increase in the bottom 50% share of the national income was associated with a 0.010 decrease in the income–happiness correlationterm·dom. Finally, there was a significant interaction between GDP per capitaterm·key and the bottom 50% share, such that the income–happiness correlationsterm·dom were larger in the year of higher GDP per capitaterm·key and the smaller bottom 50% share (i.e. more income inequalityterm·dom).

General discussion We started our investigation to explore historical changes in the income–happiness correlationterm·dom in diverse samples. In the USA the income–happiness correlationterm·dom has increased substantially since 1972 (Fig. 1 ). In contrast, the income–life satisfactionterm·dom correlation did not increase from 1978 to 2011 in Japan (Study 2). Study 3 analyzed the Eurobarometer and European Social Surveysterm·mth from 1970 to 2019. The meta-analysis largely replicated the findings from the USA On average, the income–life satisfactionterm·dom correlation among the 16 European countries increased over time. Finally, we examined the patterns of the income–life satisfactionterm·dom correlation among Latin Americans from 1997 to 2018. Unlike in the USA, Japan, and Europe, the income–life satisfactionterm·dom correlation decreased over time in Latin America.

Overall, the historical patterns of the income–happiness correlationterm·dom observed in the USA and several European countries, most notably Germany, the UK, the Netherlands, and Portugal, were consistent with the continuous materialism hypothesisterm·dom and the income inequality hypothesisterm·dom. In contrast, the patterns observed in Latin America were only consistent with the income inequality hypothesisterm·dom.

In support of the continuous materialism hypothesisterm·dom, the income–happiness correlationterm·dom became larger as GDP per capitaterm·key increased in the USA, France, Germany, UK, the Netherlands, and Portugal. In these countries, when the national economy was stronger, money appeared to be more important to people’s happinessterm·key than when the national economy was weaker. These findings are consistent with Robert Frank’s (1999) idea of luxury fever at a time of the economic boom. Even when survival per se is no longer an issue, money still matters to many people. Money can buy them a better car, a smartphone, and a bigger house. As people’s financial situations improve, the material possessions deemed necessary increase, as well ( 38–40 ). As desires increase, money continues to matter even more, making the income–happiness correlationterm·dom stronger, not weaker.

Furthermore, in the USA and European countries on average, when income inequalityterm·dom was larger, money appeared to be more important than when income inequalityterm·dom was smaller. These findings are consistent with the previous findings on the psychology of inequality: namely, inequality appears to amplify people’s tendencies to perceive more competition in society ( 26 ), believe in the zero-sum nature of society ( 27 ), and engage in unfavorable upward social comparisonterm·dom ( 24 ), in particular among the poor. Likewise, fairness and general trust might be eroded particularly among the poor at a time of high levels of income inequalityterm·dom ( 22 ). This could, in turn, increase the happinessterm·key gap between the rich and the poor, inflating the income–happiness correlationterm·dom. In addition, Laurin et al. ( 38 ) findings on fairness and self-regulation suggest that the poor might be more likely to give up trying to achieve long-term goals at a time of vast inequality than small inequality. This in turn could decrease the happinessterm·key of the poor, resulting in a larger happinessterm·key gap between the rich and the poor. In the future, it is important to directly measure changes in desires and the importance of money, as well as these psychological mechanisms in the context of changes in the income–happiness correlationterm·dom.

Whereas the continuous materialism hypothesisterm·dom and the inequality hypothesis were equally associated with historical changes in the income happinessterm·key correlations in the USA and Europe, the income inequality hypothesisterm·dom was the only hypothesis that received empirical support in Latin America. Since 1997 (the first year of our data in Latin America), income inequalityterm·dom has declined in most Latin American countries. For instance, in Argentina, the bottom 50% share of the national income increased from 10.6% in 2000 to 17.8% in 2018. In contrast, during the same period, the bottom 50% share went down from 15.1% to 13.5% in the USA. Thus, in many Latin American countries, the decreased levels of income inequalityterm·dom might have reduced perceived competition, zero-sum thinking, and unfavorable upward social comparisonterm·dom, reducing the happinessterm·key gap between the rich and the poor. Again, these psychological mechanisms should be explicitly tested in the context of Latin America.

Although we analyzed most of the nationally representative data available today, the years covered by these surveysterm·mth were still limited. For instance, Study 2 included only 20 surveysterm·mth, limiting the statistical power to detect only large effect sizes. Second, it is interesting that among European countries, several countries such as the Netherlands, Germany, UK, and Portugal were similar to the USA (a strong positive correlation between GDPpc and income inequalityterm·dom); The yearly trend in Norway was quite different from other European countries. It is important to identify a moderator to explain these cross-national variations in the future. It is also important to examine alternative explanations to the continuous materialismterm·key (e.g. inflation, unemployment rate) and income inequalityterm·dom hypotheses (e.g. social welfare spending).

Finally, although it is unlikely that a change in the income–happiness correlationterm·dom causally affects a change in income inequalityterm·dom or GDPpc, our design does not allow for causal inference. As more data become available, a more stringent time series analysis (e.g. Granger’s causality test) should be conducted to discern a causal relation better.

Conclusion We found that the income–happiness correlationterm·dom has increased in the USA and several European countries (e.g. Germany, the Netherlands, Portugal) since 1972 as GDP per capitaterm·key and income inequalityterm·dom increased. In contrast, in Latin American countries on average, the income–happiness correlationterm·dom has decreased since 1997. In Japan and several other countries (e.g. Denmark, Panama), the income–happiness correlationterm·dom has not significantly changed. These divergent historical trends of the income–happiness correlationterm·dom were associated with economic conditions. The income–happiness correlationterm·dom tends to get larger as GDP per capitaterm·key increases, providing support for the continuous materialism hypothesisterm·dom. Independently, the income–happiness correlationterm·dom tends to get larger as top 10% share of the national income increases and bottom 50% share of income decreases, providing support for the income inequality hypothesisterm·dom. These findings suggest that on average, money becomes more central to one’s happinessterm·key under high GDP per capitaterm·key and high income inequalityterm·dom. That is, in developed countries such as the USA, Germany, and the Netherlands, materialismterm·key is persistent. In understanding the role of money in happinessterm·key, researchers need to account for not just national wealth but also how national wealth is distributed across social class, as both national wealth and inequality appear to be associated with materialismterm·key.

Materials and methods All data, the analytic codes, and supplementary materials are available in the Open Science Framework at https://osf.io/x5gcw . For the descriptive statistics and correlations between variables, see Supplementary Material 1 .

Study 1 method General social surveysterm·mth Participants were 64,814 Americans (28,614 men, 36,200 women) age 18 and over. Out of 64,814 respondents, 52,033 were white, 9,187 were black (3,594 were “Other”). The mean age was 46.10 (SD = 17.54). The mean number of respondents who answered the happinessterm·key item per surveyterm·mth year was 1,692.91 (SD = 481.13), ranging from 1,173 to 2,627.

The 3-point scale happinessterm·key item (“Taken all together, how would you say things are these days—would you say that you are very happy, pretty happy, or not too happy?”) was included in every surveyterm·mth. The responses were reverse scored such that 1 = not too happy, 2 = pretty happy, and 3 = very happy.

Total family income was measured by different items across different surveyterm·mth years: a 12-category item: “1 = less than \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 2,000, 2 = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 2,000 to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 3,999. . .to 12 = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 30,000+” in 1972, by a 12 category item: “1 = less than \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 1,000, 2 = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 1,000 to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 2,999. . .12 = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 25,000 or more” in 1973, 1974, 1975, and 1976, by a 16 category item: “1 = less than \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 1,000, 2 = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 1,000 to 2,999. . .12 = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 17,500 to 19,999. . .16 = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 50,000+” in 1977, 1978, 1980, by a 17 category item “1 = less than \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 1,000, 2 = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 1,000 to 2,999. . .17 = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 50,000 + in 1982, 1983, 1984, and 1985, by a 20 category item “1 = less than \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 1,000, 2 = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 1,000 to 2,999. . .20 = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 60,000 or more” in 1986, 1987, 1988, 1989, and 1990, and by a 21 category item: “1 = less than \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 1,000, 2 = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 1,000 to 2,999. . .21 = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 75,000 or more” in 1991, 1993, 1994, 1996, by a 23 category item: “1 = less than \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 1,000, 2 = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 1,000 to 2,999. . .23 = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 110,000 or more” in 1998, 2000, 2002, 2004, and by a 25 category item “1 = less than \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 1,000, 2 = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 1,000 to 2,999. . .25 = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 150,000+” in 2006, 2008, 2010, 2012, 2014, and by a 26 category item: “1 = less than \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 1,000, 2 = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 1,000 to 2,999. . .26 = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 170,000+” in 2016 and 2018. We then converted each category of the total family income in each surveyterm·mth to the dollar using the midpoint value (e.g. 1 = < \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 2,000 → \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 1,000, 2 = 2,000 to 3,999 → \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 3,000), while the highest value was multiplied by 1.5 (e.g. 12 = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 25,000 or more → \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} ${\$}$\end{document} 37,500). Finally, we log-transformed them. All the income–happiness correlationsterm·dom reported in the text were based on the log-transformed midpoint values. For the category values, midpoint values, historical exchange rates, and log-transformed values in Study 1 and the other studies, see Supplementary Material 3 ( https://osf.io/x5gcw ) GDP per capitaterm·key was taken from https://wid.world/ . Gini coefficientterm·key was taken from Table H-4 in https://www.census.gov/data/tables/time-series/demo/income-poverty/historical-income-households.html .

Study 2 method Participants were 80,857 Japanese (37,212 men; 40,254 women) age 15 to 88. The mean age was 44.54 (SD = 15.57). The mean number of respondents for the life satisfactionterm·dom question used per year was 2,846.61 (SD = 1,548.23), ranging from 1,692 to 7,405.

The 5-point scale life satisfactionterm·dom item (1 = dissatisfied, 2 = rather dissatisfied, 3 = neither satisfied nor dissatisfied, 4 = rather satisfied; 5 = satisfied) was included in 1978, 1981, 1984, 1985, 1987, 1988, 1990, 1991, 1992, 1993, 1994, 1995, 1996, 1999, 2001, 2002, 2005, 2006, 2008, and 2009.

Income was measured differently in different surveysterm·mth. In 1978, 1981, 1984, 1987, 1988, 1990, it was measured by a 21-point scale (1 = less than 750,000 yen; 2 = 750,000 to 1.25 mil yen; 3 = 1.25 to 1.75 mil yen. . .21 = over 10.25 mil yen). In 1985, it was measured by a 27-point scale (1 = None, 2 = less than 500,000 yen; 3 = 500,000 to 1 mil yen; 4 = 1 to 1.5 mil yen. . .27 = over 20 mil yen). In 1991, it was measured by a 23-point scale (1 = less than 1.25 mil yen; 2 = 1.25 to 1.75 mil yen; 3 = 1.75 to 2.25 mil yen, 4 = 2.25 to 2.75 mil yen. . .23 = over 30mil yen). In 1992, 1996, 1999, 2002, 2005, 2008, and 2011, it was measured by an 8-point scale (1 = less than 2mil yen; 2 = 2 to 4 mil yen; 3 = 4 to 6 mil yen; 4 = 6 to 8 mil yen; 5 = 8 to 10 mil yen; 6 = 10 to 12 mil yen; 7 = 12 to 14 mil yen, 8 = over 14 mil yen). In 1993, it was measured by a 12-point scale (1 = less than 750,000 yen; 2 = 750,000 to 2.25 mil yen; 3 = 2.25 to 3.75 mil yen. . .12 = over 15.75 mil yen). In 1994, it was measured by a 13-point scale (1 = less than 500,000 yen; 2 = 500,000 to 1 mil yen; 3 = 1 to 1.5 mil yen; 4 = 1.5 to 2 mil yen. . .13 = over 10 mil yen). In 1995, it was measured by a 9-point scale (1 = less than 1 mil yen; 2 = 1 to 2 mil yen; 3 = 2 to 4 mil yen; 4 = 4 to 6 mil yen. . .9 = over 14 mil yen). In 2001, it was measured by a 10-point scale (1 = less than 20mil yen; 2 = 2 to 4 mil yen; 3 = 4 to 6 mil yen; 4 = 6 to 8 mil yen. . .10 = over 20 mil yen). In 2006 and 2007, it was measured by a 10-point scale (1 = less than 1mil yen; 2 = 1 to 3 mil yen; 3 = 3 to 5 mil yen; 4 = 5 to 7 mil yen. . .10 = over 20 mil yen). In 2009 and 2010, it was measured by a 5-point scale (1 = less than 1 mil yen; 2 = 1 to 3 mil yen; 3 = 3 to 5 mil yen; 4 = 5 to 10 mil yen; 5 = over 10 mil yen). The income variables were converted to yen, then log-transformed, as in Study 1.

GDP per capitaterm·key was taken from https://wid.world/ . Gini coefficientterm·key was taken from https://data.worldbank.org/indicator .

Study 3 method Study 3 combined annual datasets from Eurobarometer Surveysterm·mth 1973 to 2002 and biennial datasets from European Social Surveysterm·mth 2004 to 2018. Participants from Eurobarometer Surveysterm·mth were 659,140 Europeans ( M age = 43.16, SD age = 17.92, 51.7% women). Participants from European Social Surveysterm·mth were 422,985 Europeans ( M age = 48.29, SD age = 18.63, 53.8% women). The mean number of respondents who answered the life satisfactionterm·dom item per surveyterm·mth year was 28,885.73 ( SD = 14,422.27), ranging from 8,767 to 54,151.

As for the life satisfactionterm·dom item, every surveyterm·mth year of Eurobarometer Surveysterm·mth included the 4-point scaleterm·mth life satisfactionterm·dom item (“On the whole, are you very satisfied, fairly satisfied, not very satisfied, or not at all satisfied with the life you lead?”), except for 1974 and 1996. Every surveyterm·mth year of European Social Surveysterm·mth included the life satisfactionterm·dom item (“How satisfied with life as a whole”), with different answer scales such as 10-points (0 = extremely dissatisfied; 10 = extremely satisfied). As for the household income item, both Eurobarometer Surveysterm·mth and European Social Surveysterm·mth used different categories and scales across countries and years. There are the variables named “income” in Eurobarometer Surveysterm·mth and “hinctnt” and “hinctnta” in European Social Surveysterm·mth, which used different categories and scales in each of the 16 countries. Compared to the other surveyterm·mth years that used money values, the 2008 to 2018 surveysterm·mth used income deciles in each nation. All the income variables were converted to the local currency and log-transformed, as in Studies 1 and 2.

GDP per capitaterm·key, the top 10% share of the national income, the bottom 50% share of the national income were taken from https://wid.world/ .

Study 4 method Data across 18 Latin American countries were derived from Latinobarometro 1997 to 2018 ( https://www.latinobarometro.org/ ). The total number of participants who answered life satisfactionterm·dom and income item were 431,148 ( M age = 40.05, SD age = 16.47, 51.5% women). Ethnicity/race data are available only in 2007 to 2018 (0.66% Asian, 5.45% Black, 9.33% Indigenous, 46.83% Mestizo, 5.95% Mulato, 29.86% White, 1.94% Other race). The mean number of respondents who answered the life satisfactionterm·dom item per surveyterm·mth year was 20,873 ( SD = 1560.91), ranging from 17,601 to 22,615.

Life satisfactionterm·dom was measured with the 4-point scaleterm·mth (“Generally speaking, would you say you are satisfied with your life? Would you say you are. . .? 1 = not very satisfied to 4 = very satisfied).

Income was assessed using the 4-point scaleterm·mth (“Does your salary and the total of your family’s salary allow you to satisfactorily cover your needs? 1 = Doesn’t cover, there are great difficulties to 4 = covers them well, I can save”).

Skewness and kurtosis were within the acceptable range (skewness < 2.00; kurtosis < 7.00, according to 39) as follows: 1997 (0.17; −0.56), 1998 (0.12; −0.65), 2000 (0.09; −0.58), 2001 (0.11; −0.73), 2002 (0.01; −0.72), 2003 (−0.02; −0.8), 2004 (−0.01; −0.82), 2005 (−0.03; −0.78), 2006 (0.11; −0.58), 2007 (0.11; −0.58), 2008 (0.05; −0.57), 2009 (0.06; −0.61), 2010 (0.09; −0.56), 2011 (0.19; −0.42), 2013 (0.23; −0.49), 2015 (0.17; −0.52), 2016 (0.12; −0.61), 2017 (0.16; −0.65), 2018 (0.12; −0.73). Thus, we report correlations based on the raw subjective income data.

The GDP per capitaterm·key, the top 10% share, and the bottom 50% share of the national income data came from https://wid.world/ .

Supplementary Material Click here for additional data file.