The Journal of Socio-Economics 40 (2011) 67–73
Contents lists available at ScienceDirect
The Journal of Socio-Economics journal homepage: www.elsevier.com/locate/soceco
A closer look at the Easterlin Paradox Luis Angeles ∗ Department of Economics, University of Glasgow, Adam Smith Building, Glasgow G12 8RT, United Kingdom
a r t i c l e
i n f o
Article history: Received 4 March 2010 Received in revised form 23 June 2010 Accepted 24 June 2010 JEL classification: I31 I38 O20
a b s t r a c t This paper argues that increasing average incomes and stagnating levels of happiness, as observed in the United States since the 1970s, do not constitute a paradox. First, we show that the effect of higher incomes has been more than counteracted by changes in other socioeconomic variables, notably the prevalence of marriage and divorce. Second, we show that the effect of a given amount of real income on happiness has not changed between the 1970s and the early 2000s. © 2010 Elsevier Inc. All rights reserved.
Keywords: Happiness Income Easterlin Paradox
1. Introduction Back in 1974, Richard Easterlin asked the question “Does Economic Growth Improve the Human Lot?” and fearlessly answered it with a negative (Easterlin, 1974, and a restatement in Easterlin, 1995). What has become known as the Easterlin Paradox, the idea that higher levels of income do not bring more happiness at the society level, has become a highly influential result in economics. The public policy consequences of this result are nothing short of revolutionary. If more income does not make people happier then economic growth may not be a valid policy objective and we all would be better off working less (see, inter alia, Layard, 2005). This short paper revisits the arguments leading to such far-reaching claims and, after reassessing the evidence, offers a different conclusion. In his seminal 1974 article, Richard Easterlin proposed to answer the imposing question transcribed above by examining the available evidence on income and happiness along 3 dimensions. First, at the individual level, do we observe that rich people are happier than poor ones? Second, at the country level, do we observe that rich countries are happier than poor ones? And third, at the country level once again, do we observe countries getting happier as they grow richer? The answer to the first question was not in doubt back in 1974 and has never been in doubt ever since: rich individuals are happier
∗ Tel.: +44 141 330 8517. E-mail address:
[email protected]. 1053-5357/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.socec.2010.06.017
than poor ones. This result holds in numerous microeconometric studies using survey data from countries as diverse as the United States, China or Russia; it holds when controlling for a wealth of alternative determinants of happiness and even in regressions with person-specific fixed effects.1 This was clearly acknowledged by Easterlin. The answer to the second question was not obvious in 1974 due to data scarcity and Richard Easterlin was able to argue that rich countries do not tend to be happier than poor ones. The much larger datasets available today, however, have shown quite decisively that a clear positive association exists between average levels of happiness and GDP per capita across countries.2 It is the answer to the third question, however, what has constituted the most solid piece of evidence in favour of Easterlin’s Paradox. By and large, average levels of happiness do not appear to trend upwards in most developed countries - at least not markedly so – despite their clear gains in GDP per capita over time. Perhaps the best example of all is the United States, which was the only country Easterlin could analyze over time in 1974 and for which all subsequent data has confirmed that initial assessment. Table 1 illustrates this fact with the latest data available. We calculate the
1 See Argyle (1999), Di Tella and MacCulloch (2006) and Blanchflower (2008) for recent reviews of this literature. 2 Earlier attempts to show this are Diener et al. (1993, 1995) and Veenhoven (1991). A more recent demonstration using much improved datasets is Stevenson and Wolfers (2008). These authors show that a positive relationship exists between the average happiness score of a country and the log of its GDP per capita, implying a diminishing marginal utility of income.
68
L. Angeles / The Journal of Socio-Economics 40 (2011) 67–73
Table 1 The Easterlin Paradox. Average values of happiness and income, per decade. 1970s
1980s
1990s
2000s
Average happiness
6.07
5.98
5.97
5.95
GDP per capita (in 2000 dollars) 1970s = 1
20,064 1
23,876 1.1900
29,413 1.4660
34,757 1.7323
Average family income (in 2007 dollars) 1970s = 1
56,732 1
60,616 1.0685
69,565 1.2262
78,087 1.3764
Having shown that, the third section of the paper tests the idea that a given amount of real income will procure less and less happiness as a society grows richer. We use three alternative econometric methodologies which come to the same conclusion: the happiness effect of a given amount of real income has not fallen over the last three decades in the United States. To the best of our knowledge this is the first time that these tests are performed in the literature. The final section offers some concluding remarks. 2. It is not just about income
average values of happiness (on a 0–10 scale), GDP per capita (in 2000 dollars) and family income (in 2007 dollars) for the United States in the last three decades of the 20th century and the first seven years of the 21st century.3 Despite substantial gains in GDP per capita (+73%) and average family income (+37%) between the 1970s and the early 2000s, average levels of happiness in the United States have experienced a slight decrease; from 6.07 in the 1970s to 5.95 in the early 2000s.4 Why did a negative answer to this third question lead Easterlin, and many more after him, to argue that economic growth does not “improve the human lot” in spite of the positive answer to the first question? (We set the second question aside since its answer was much less clear at the time). The reason is that the two seemingly contradictory findings could be reconciled by assuming that people’s happiness is a function not of absolute but of relative incomes. The idea has a long pedigree in economics, dating back at least to Veblen (1899) and Duesenberry (1949). Assume for a second that relative income is the ratio of individual income to the average income in the society. It is then readily comprehensible why rich individuals would be happier than poor ones (they have a higher relative income) yet a country does not become happier as it grows richer (relative incomes cannot grow on average). The logic is attractive and can make many converts.5 Most of the recent criticism of Easterlin’s Paradox has focused on trying to prove that the answer to the third aforementioned question is actually positive (the answer to the second question becoming increasingly clear as time passed and datasets expanded). Authors such as Hagerty and Veenhoven (2003), Inglehart et al. (2008) and Stevenson and Wolfers (2008) have produced persuasive evidence showing that average levels of happiness have in fact somewhat increased in several countries as they experience economic growth. The interested reader may consult, together with these papers, the replies and counterarguments brought forward by Easterlin (2005) or Easterlin and Sawangfa (2009). The present paper will not argue that average levels of happiness tend to increase with economic growth. While this may be true for some countries it does not appear to be so for the United States; a large and influential case and the focus of our attention here. More important, a constant level of average happiness is perfectly consistent with rising average incomes once we consider that income is not the only determinant of happiness changing over time at the society level; a point that we develop in Section 2.
3 Appendix A provides more information on the data, summary statistics and a correlation matrix. Throughout the paper, the periods 1970s, 1980s, 1990s and 2000s refer to the years 1972–1979, 1980–1989, 1990–1999 and 2000–2006. 4 For other works analyzing the relationship between average incomes and average happiness over time see, inter alia, Kenny (1999), Diener and Oishi (2000), Myers (2000), Frey and Stutzer (2002) and Blanchflower and Oswald (2004). 5 The reference with respect to which relative incomes are calculated can of course be different from the average income of the society. Researchers have focused on two alternative views: social comparison, where the reference is the income of a comparison group such as colleagues or people with similar socioeconomic characteristics; and adaptation, where the reference is the individual’s own past income. The logic of the argument remains the same in these two alternative views.
Average levels of happiness have failed to increase in the United States over the last three and a half decades while average incomes have markedly trended upwards. Is there a puzzle? There would be one if, as in introductory microeconomic theory, human happiness was a positive function of income and nothing but income. This, however, can be safely rejected after decades of research in areas such as the empirical determinants of happiness or experimental economics. The most pertinent area of research for the present discussion is the one used by Easterlin to answer his first question: the microeconometric studies of the determinants of happiness. As mentioned above, such studies – which by now count in the dozens if not in the hundreds – have consistently found a positive effect of income on happiness (or on life satisfaction, its widely used empirical alternative). What they have also shown with striking consistency, and which has been strangely absent from discussions of the Easterlin Paradox, is how little of the differences in happiness across individuals can be explained by differences in income. The first two columns of Table 2 show this by running pooled OLS regressions of happiness against real income variables using data from the United States’ General Social Survey (GSS) over the period 1972–2006. Whether we use the log of real income (column 1) or dummy variables for different real income bands (column 2), two results stand out. First, that income is positively correlated with happiness. Second, that a mere 3.6% of the variation in happiness scores can be explained by income. If income can explain so little of happiness, why should we be surprised to see rising average incomes go hand in hand with stagnant average happiness?6 Columns 3 and 4 of Table 2 add a standard set of determinants of happiness consisting of the individual’s age (and its square), gender, race, marital status, health and employment status. This full set of regressors achieves to explain about 14% of the variation in the endogenous variable, a result that is near the upper bound for this kind of studies. What these regressions also show is that income has by no means the largest effect on happiness among this set of explanatory factors. The largest effects on happiness are associated with the individual’s marital status, health and employment status. Getting married, experiencing a health improvement from “good” to “excellent”, or going out of unemployment is associated with a positive happiness effect of about 1 unit (on a 0–10 scale). Based on the results in column 4, passing from an income in the $20,000 to $40,000 band to one in the $100,000 and over band would increase happiness by just 0.6 units. An increase in individual income of the magnitude just mentioned may well be rare but not unheard of. Increases in a country’s average income, on the other hand, are usually an order of magnitude smaller. The 73% increase in GDP per capita since the 1970s, or the 37% increase in average family income over the same period, would translate into an increase in average happiness of 0.13 or 0.07 units respectively (using the estimated coefficient from column 3 in
6 A caveat is in order. As most of the literature we are using gross incomes. Disposable incomes may plausibly result in somewhat larger effects.
L. Angeles / The Journal of Socio-Economics 40 (2011) 67–73
69
Table 2 Baseline regressions, pooled OLS 1972–2006. Dependent variable: happiness (1)
(2)
(3)
0.634** (0.016)
Log of real income Real income between: 20,000–40,000 40,000–60,000 60,000–80,000 80,000–100,000 100,000 and over
0.670** 1.140** 1.421** 1.565** 1.791**
Age Age squared Female Black Married Widow Divorced Separated Health = excellent Health = good Health = fair Unemployed Retired R2 Observations
(4)
0.236** (0.021)
0.036 41,656
0.036 41,656
0.160** 0.395** 0.581** 0.584** 0.766**
(0.040) (0.044) (0.056) (0.077) (0.060)
(0.047) (0.054) (0.066) (0.089) (0.071)
−0.027** (0.006) 0.0005** (0.0) 0.311** (0.035) −0.493** (0.050) 0.920** (0.050) −0.576** (0.080) −0.395** (0.067) −0.683** (0.099) 2.698** (0.082) 1.699** (0.079) 0.808** (0.082) −1.020** (0.099) 0.261** (0.068)
−0.029** (0.006) 0.0005** (0.000) 0.313** (0.035) −0.501** (0.050) 0.925** (0.050) −0.573** (0.080) −0.368** (0.067) −0.684** (0.099) 2.706** (0.082) 1.722** (0.079) 0.832** (0.082) −1.021** (0.099) 0.289** (0.068)
0.1366 31,867
0.1377 31,867
Note: Robust standard errors in parenthesis. *Statistically significant at the 5% level. ** Statistically significant at the 1% level.
Table 3 It is not just about income. Average values of happiness and other socioeconomic variables, per decade.
Average happiness Average health Unemployed (%) Retired (%) Married (%) Widowed (%) Separated (%) Divorced (%)
1970s
1980s
1990s
2000s
6.07 1.98 3.10 10.46 67.7 9.5 3.1 5.8
5.98 2.01 3.10 12.32 55.6 11.0 3.8 10.9
5.97 2.05 2.60 13.79 50.4 10.3 3.3 14.5
5.95 2.01 3.30 15.00 48.0 8.4 3.5 15.7
Table 2). A positive effect, but well short of being earth-shattering and perhaps even difficult to detect with noisy data.7 More important, the above regressions show that happiness is a function of many variables. If we are intrigued about the stagnating happiness in the United States a good place to start inquiring would be in the evolution of the main determinants of happiness, income being only one of them and not the most important. Table 3 shows the evolution of four marital status variables, one health variable and two employment status variables since the 1970s. These three areas have the largest effects on happiness in the regressions reported above and in numerous similar regressions that have been carried out in the literature. Their evolution should be regarded as at least as important as the evolution of income when assessing changes in happiness. Over the three and a half decades since the early 1970s, selfreported health is the area where we see the least changes; with an average never far away from 2 on a 0–3 scale. The level of unemployment is also very stable, though we may recall that business
7 Moreover, note that we have made our calculations under the assumption that all incomes in the population grow at the same rate; which we know was not the case in the United States over this period. Allowing for increasing income inequality further reduces these effects since the average of log income would change less than the log of average income (Stevenson and Wolfers, 2008).
cycle fluctuations have been smoothed out by our decade-long averages. The percentage of people who are retired, on the other hand, smoothly increases from about 10.5% in the 1970s to 15% in the early 2000s. This trend, characteristic of an ageing population, would bring a positive effect on average happiness since retired people tend to be about a quarter of a point happier than others ceteris paribus. The most significant changes, however, come in the area of marital status. Here we observe a large and unmistakable decline in the prevalence of married individuals, accompanied by an increase in the number of divorcees. The two remaining categories, widowers and separated individuals, are roughly trendless. The decline of married individuals, from 68% of the population in the 1970s to 48% in the early 2000s, would result in average happiness falling by 0.185 points; more than enough to overcome the positive effect of rising incomes. An additional negative effect would result from the increase in divorcees from about 5% to 15% of the population. Putting these pieces together, we may advance that between the 1970s and the early 2000s the American population should have experienced gains in happiness due to its rising income and higher percentage of retired people and losses in happiness due to its lower number of married individuals and higher number of divorcees. The actual evolution of average happiness, flat to slightly decreasing, should then come as no surprise. This preliminary assessment can be confirmed by using the regression results from the third column of Table 2 to calculate the predicted average levels of happiness in each decade. In other words, we use the estimated coefficients from this equation to determine predicted levels of happiness for each person which we then average over all observations from the 1970s, from the 1980s and so on. The result is a predicted average happiness of 6.09 in the 1970s, 5.94 in the 80s, 5.94 in the 90s and 5.89 in the early 2000s; quite in line with the actual values reported in Table 1. In other words, the slight decrease in average happiness can be accounted pretty well by the underlying changes in all the exogenous variables. To summarize up to this point, we do not see a puzzle in the stagnating levels of happiness observed in the United States since
70
L. Angeles / The Journal of Socio-Economics 40 (2011) 67–73 Table 4 Separate regressions for each decade. Dependent variable: happiness
Fig. 1. Income and happiness over time.
the 1970s. The observation is in line with the evolution of all the determinants of happiness (and not just income) over this period. 3. Does a given level of income procures less and less happiness? Central to Easterlin’s Paradox is the idea that a given level of income would procure less and less happiness as a society grows richer (see Easterlin, 1974 and, for a more recent restatement, Clark et al., 2008). Either because we compare ourselves with others or with our own recent past, comfort levels that made people perfectly happy in the 1970s would not be regarded as adequate in the early 2000s. Goods that were special then are commonplace now, and their possession would not make us happy anymore despite their intrinsic qualities; or so the argument goes. A graphical representation of this idea is shown in Fig. 1 (a similar figure is used in Clark et al., 2008, precisely to explain Easterlin’s Paradox). The three parallel and upward-sloping lines in this figure represent the relationship between real income and happiness at three points in time; the 1970s, 1980s and 1990s. In accordance with Easterlin’s Paradox, the average level of happiness procured by each of these three income distributions is the same. This, in turn, is only possible if as the society grows richer the relationship between real income and happiness shifts rightward. Thus, for any given level of real income such as y0 , the associated level of happiness would be highest in the 1970s and lowest in the 1990s. We test this idea with three alternative econometric methodologies. The first methodology estimates a separate regression for each decade and each regression estimates the happiness effect of belonging to one of five possible bands of real income. The empirical specification is thus: hi,t = ˛ +
5
ˇj y˜ i,t,j + BXit + εi,t
(1)
j=1
where hi ,t is the GSS’s measure of happiness for individual i at time t; y˜ i,t,j is a dummy variable taking the value of 1 if individual i has a real income in band j at time t; and Xit is the set of control variables used in the previous section. As in the previous section, the five real income bands that we consider are $20,000–$40,000, $40,000–$60,000, $60,000–$80,000, $80,000–$100,000 and $100,000 and over; all in dollars of the year 2000. The omitted category is persons with an annual real income of less than $20,000. If a given level of income procures less and less happiness as the economy grows richer we should observe the ˇj coefficients becoming smaller from decade to decade. Results for each decade are reported in columns 1–4 of Table 4. As expected, each column shows that higher levels of income procure more happiness, though at a decreasing rate. What is more important for us, however, is the comparison of coefficients across
1970s
1980s
1990s
2000s
Real income between: 20,000–40,000 40,000–60,000 60,000–80,000 80,000–100,000 100,000 and over
0.153 0.380** 0.578** 0.474** 0.762**
0.196* 0.427** 0.745** 0.655** 0.860**
0.166 0.427** 0.508** 0.654** 0.732**
0.209 0.459** 0.636** 0.614** 0.825**
Age Age squared Female Black Married Widow Divorced Separated Health = excellent Health = good Health = fair Unemployed Retired
−0.015 0.000** 0.547** −0.680** 1.019** −0.590** −0.553** −0.520* 2.888** 1.886** 1.054** −1.109** 0.604**
−0.051** 0.001** 0.403** −0.665** 0.725** −0.667** −0.388** −0.690** 2.522** 1.624** 0.764** −1.100** 0.187
−0.036** 0.001** 0.088 −0.469** 1.017** −0.423** −0.185 −0.820** 2.590** 1.554** 0.597** −1.165** 0.240
−0.037* 0.0004** 0.196* 0.035 1.196** −0.496* −0.247 −0.399 2.977** 1.933** 0.923** −0.298 0.451**
R2 Observations
0.1399 8387
0.1318 9189
0.1454 9068
0.1560 5223
* **
Statistically significant at the 5% level. Statistically significant at the 1% level.
decades. As it turns out, all ˇj coefficients are reasonably stable over time and there is no indication of a diminishing trend. Take for instance the happiness effect of belonging to the income band $40,000–$60,000: 0.380 in the 1970s, 0.427 in the 1980s and in the 1990s and 0.459 in the early 2000s. The effect is actually slightly increasing over time (though the difference is not statistically significant). A similar pattern is observed for all other income bands. Our second empirical methodology uses a single regression covering all four decades instead of separate regressions for each decade. The reason is that separate regressions calculate the effect of income on happiness with respect to different baseline levels of happiness. To avoid this, our second methodology runs a single regression and calculates all effects with respect to the same baseline level. We interact each income band with dummy variables for each decade to capture any changes in the effect of income on happiness. The specification is then: hi,t = ˛ +
5 j=1
ˇj y˜ i,t,j +
3 5 j=1 p=1
jp Dp y˜ i,t,j +
3
p Dp + BXit + εi,t (2)
p=1
In Eq. (2), Dp (p = 1, 2, 3) are dummy variables for the decades 1980s, 1990s and early 2000s (the 1970s is the omitted category). The coefficients p and jp capture the effects of these decade dummies and of the interaction term of our income bands with the decade dummies. The happiness effect of having an income in the first income band ($20,000–$40,000) would then be given by ˇ1 for the 1970s, ˇ1 + 11 for the 1980s, ˇ1 + 12 for the 1990s and ˇ1 + 13 for the early 2000s. A similar logic applies for other income bands. If a given amount of income procures less happiness as the United States grows richer we would expect the jp coefficients to be negative and to become larger in absolute value for more recent decades (we would expect, for instance, that 13 < 12 < 11 < 0). The results for this regression are reported in the first column of Table 5. Once again, results are not supportive of the idea that a given level of income is procuring less and less happiness over time. We find that none of the interaction terms is statistically significant, and the jp coefficients actually tend to take positive values instead of negative ones. The ˇj coefficients, on the other hand, take similar values as those reported in Table 4.
L. Angeles / The Journal of Socio-Economics 40 (2011) 67–73 Table 5 Pooled regressions (1972–2006). Dependent variable: happiness Interactions of income with decade dummies
Interactions of all regressors with decade dummies
Real income between: 20,000–40,000 ×1980s dummy ×1990s dummy ×2000s dummy
0.120 −0.025 0.088 0.141
0.153 0.042 0.013 0.055
40,000–60,000 ×1980s dummy ×1990s dummy ×2000s dummy
0.350** −0.077 0.142 0.166
0.380** 0.047 0.047 0.080
60,000–80,000 ×1980s dummy ×1990s dummy ×2000s dummy
0.534** 0.034 0.041 0.185
0.578** 0.167 −0.069 0.058
80,000–100,000 ×1980s dummy ×1990s dummy ×2000s dummy
0.460** 0.020 0.265 0.262
0.474** 0.181 0.180 0.140
100,000 and over ×1980s dummy ×1990s dummy ×2000s dummy
0.748** −0.057 0.049 0.141
0.762** 0.099 −0.030 0.064
1980s dummy 1990s dummy 2000s dummy
0.043 −0.096 −0.027
1.299 1.173 0.865
Age Age squared Female Black Married Widow Divorced Separated Health = excellent Health = good Health = fair Unemployed Retired
−0.030** 0.000** 0.313** −0.502** 0.937** −0.564** −0.368** −0.680** 2.705** 1.719** 0.831** −1.020** 0.284**
−0.015 0.0004** 0.547** −0.679** 1.019** −0.590** −0.552** −0.520* 2.887** 1.885** 1.054** −1.109** 0.604**
R2 Observations
0.1379 31,867
0.1420 31,867
* **
The exercise can be repeated dropping the non-interacted decade dummies from Eq. (2); producing no meaningful changes in the results (not reported but available upon request). Our third and final methodology consists in estimating once again a single equation but this time we allow for the possibility that the effect of all regressors, and not just income, changes from decade to decade. The specification is thus: 5 j=1
+
3
ˇj y˜ i,t,j +
3 5 j=1 p=1
p Dp Xit + εi,t
jp Dp y˜ i,t,j +
3
Results are shown in the second column of Table 5 and are quite conclusive. As previously, none of the jp coefficients is statistically significant and they take positive values in most cases. There is just no evidence that the happiness effect of any of our five income bands has been decreasing over time. The coefficients on the interactions terms of all other variables are not reported for conciseness, but the large majority of them is not statistically significant. Two notable exceptions are the dummy for female individuals, which becomes less positive over time, and the dummy for black individuals, which becomes less negative. Estimating Eq. (3) without the non-interacted decade dummies does not change our conclusions. Overall, the results of this section provide additional evidence against Easterlin’s Paradox. The main explanation for this supposed paradox is that people retire less and less happiness from a given level of real income as society grows richer. Oddly enough, the assertion has not been directly tested in the literature. We test this assertion under three alternative specifications and find no evidence in its favor. A given amount of real income has procured roughly the same satisfaction in the 1970s as in the early 2000s. 4. Concluding remarks
Statistically significant at the 5% level. Statistically significant at the 1% level.
hi,t = ˛ +
71
p Dp + BXit
p=1
(3)
p=1
and p is the vector of coefficients of all interaction terms. As before, our attention is on the jp coefficients; which should be negative and increasing in absolute value to support Easterlin’s views.
Researchers who have studied the relationship between income and measures of happiness have most of the time seen a contradiction between the micro and the macroevidence. Microeconometric studies never fail to find a positive effect of income on happiness whereas studies at the country level, and especially those using country time series, have a much harder time finding it. Thus, the claim of a paradox and the necessity to explain it by arguing that we retire less and less satisfaction from a given amount of income as society grows richer. Our view is that no such paradox exists and we fear that a reason for the confusion is an excessive regard for the statistical significance of an effect coupled with an insufficient attention to its actual magnitude. The effect of income on happiness is always statistically significant in microstudies and rarely so in macrostudies. That is hardly surprising: the several thousand observations of a microstudy will allow the precise identification of almost any effect while the few dozen observations of macrostudies will typically leave you doubting. Instead of concluding that the effect does not exist at the macrolevel we should consider that a few observations may not be able to clearly uncover the effect; particularly so if the effect is of small magnitude. This is where attention to the estimated magnitudes from microstudies comes handy. These studies not only reveal that the effect is statistically significant; they also show that it is of very small magnitude. So small, actually, that the predicted change in average happiness due to raising average incomes in the United States between the 1970s and the early 2000s is a mere 0.07 units on a 0–10 scale. A change of this magnitude is difficult to detect with the 30-plus observations of average happiness at the country level that a macrostudy would use. Moreover, the focus on income has overlooked the fact that there are more important determinants of individual happiness and that these too can change at the society level. As we have shown, changes in marital status (fewer individuals are married, more are divorced) are enough to more than overcome the effect of rising incomes in the United States. The net effect would be a slight decrease in average levels of happiness between the 1970s and the early 2000s; in line with actual trends. Our conclusion is thus that no paradox exists between the evolution of income and happiness in the United States over the last few decades. We bring further support to our view by showing that the happiness effect of a given amount of real income has not been falling over time as the United States has become richer. This also
72
L. Angeles / The Journal of Socio-Economics 40 (2011) 67–73
constitutes a formal rejection of the main explanation advanced for Easterlin’s Paradox. To finalize, we note that our results do not imply that economic growth guarantees a happier nation. Indeed, the small magnitude of the effect of income on happiness means that economic growth can be easily overcome by other factors such as the prevalence of marriage, widespread unemployment or public health. There is, however, no reason to be negative about economic growth and suggest its demise as an objective of public policy. Other things being equal, economic growth should have a positive direct effect on average happiness. The most important effects, however, may well be indirect. Economic growth could matter more for its influence on unemployment, family relations and health than for the larger incomes that define it. A good dose of prudence and modesty in policy advice would thus be commendable.
say that you are very happy, pretty happy, or not too happy?”. In accordance with the literature, we use a 0–10 scale to measure the answers by assigning the value of 0 to “not too happy”, 5 to “fairly happy” and 10 to “very happy”.
Appendix A.
A.3. Control variables
All the data in this paper comes from the United States’ General Social Survey (GSS) with the exception of the two income variables used in Table 1: GDP per capita (from the Penn World Tables version 6.2) and average family income (from the United States’ Census Bureau). The General Social Survey interviewed a (non-repeated) sample of about 1500 individuals every year from 1972 to 1993 and every two years since 1994 (with the exception of the years 1979, 1981 and 1992). The GSS is not a panel: the sample changes every year. We have at our disposal data for the period 1972–2006.
We use different questions from the GSS to form the following variables: For gender we form a dummy variable for females. Male is the only other possible answer. For race we form a dummy variable for people who identify themselves as Black. The omitted category encompasses the answers “White” and “Others”. For marital status we form 4 dummy variables for people who identify themselves as Married, Widowed, Divorced and Separated. The omitted category is people who never married. For health we form 3 dummy variables for people who describe their own health as excellent, good, and fair. The omitted category is people who say their health is poor. For employment status we form a dummy variable for people who identify themselves as unemployed and another one for
A.1. Happiness Happiness is assessed by the following question: “Taken all together, how would you say things are these days – would you
A.2. Real income The GSS asks respondents to place themselves in one of several incomes bands that are proposed to them. These income bands have been changing over time together with the rising price level. The GSS also constructs a variable of real income by assigning to each individual an income equal to the middle of his band and using the CPI-U-RS (Consumer Price Index for Urban households, Research Series) to transform these amounts into constant 2000 dollars. The interested reader may consult Hout (2004) for more details.
Table A1 Descriptive statistics.
Happiness Log of real income Age Dummy variables Female Black Married Widow Divorced Separated Health = excellent Health = good Health = fair Unemployed Retired
Mean
Standard deviation
Minimum
Maximum
Observations
5.97 10.32 45.51
3.17 0.95 17.44
0 6.08 18
10 12.10 89
48,318 45,682 52,849
0.559 0.137 0.546 0.098 0.121 0.035 0.308 0.448 0.186 0.031 0.131
0 0 0 0 0 0 0 0 0 0 0
1 1 1 1 1 1 1 1 1 1 1
53,043 53,043 53,024 53,024 53,024 53,024 39,842 39,842 39,842 53,033 53,033
Table A2 Correlation matrix. Happiness Log of real income Happiness Log of real income Age Female Black Married Widow Divorced Separated Health = excellent Health = good Health = fair Unemployed Retired
1.0 0.189 0.033 0.005 −0.116 0.230 −0.077 −0.105 −0.095 0.211 −0.014 −0.147 −0.088 0.019
1.0 −0.09 −0.127 −0.193 0.376 −0.231 −0.112 −0.116 0.180 0.054 −0.177 −0.084 −0.165
Age
Female Black
1.0 0.039 −0.052 0.031 0.456 0.029 −0.029 −0.168 −0.055 0.158 −0.086 0.575
1.0 0.046 −0.080 0.173 0.040 0.033 −0.032 −0.003 0.027 −0.074 −0.074
1.0 −0.150 0.016 0.012 0.127 −0.064 −0.009 0.064 0.044 −0.025
Married Widow Divorced Separated Health = excellent
Health = good
Health = fair
Unemployed
1.0 −0.361 −0.407 −0.208 0.046 0.016 −0.044 −0.074 −0.050
1.0 −0.432 0.004 −0.052
1.0 0.015 0.110
1.0 −0.069
1.0 −0.122 −0.062 −0.096 −0.041 0.092 −0.038 0.304
1.0 −0.070 −0.026 0.004 0.013 0.021 −0.027
1.0 −0.031 −0.010 0.034 0.024 −0.029
1.0 −0.602 −0.319 −0.020 −0.109
Retired
1.0
L. Angeles / The Journal of Socio-Economics 40 (2011) 67–73
retired people. The omitted category is any other employment status. See Tables A1 and A2. References Argyle, M., 1999. Causes and correlates of happiness. In: Kahneman, D., Diener, E., Schwarz, N. (Eds.), Well-Being: The Foundations of Hedonic Psychology. Russell Sage Foundation, New York, pp. 353–373. Blanchflower, D.G., 2008. International Evidence on Well-Being, NBER Working Paper 14318. Blanchflower, D.G., Oswald, A.J., 2004. Well-being over time in Britain and the USA. Journal of Public Economics 88, 1359–1386. Clark, A.E., Frijters, P., Shields, M.A., 2008. Relative income, happiness and utility: an explanation for the Easterlin Paradox and other puzzles. Journal of Economic Literature 46 (1), 95–144. Di Tella, R., MacCulloch, R., 2006. Some uses of happiness data in economics. Journal of Economic Perspectives 20 (1), 25–46. Diener, E., Sandvik, E., Seidlitz, L., Diener, M., 1993. The relationship between income and subjective well-being: relative or absolute? Social Indicators Research 28, 195–223. Diener, E., Diener, M., Diener, C., 1995. Factors predicting the subjective wellbeing of nations. Journal of Personality and Social Psychology 69 (5), 851– 864. Diener, E., Oishi, S., 2000. Money and happiness: income and subjective well-being across nations. In: Diener, E., Suh, E.M. (Eds.), Culture and Subjective Well-Being. MIT press, Cambridge. Duesenberry, J.S., 1949. Income, Saving and the Theory of Consumer Behavior. Harvard University Press, Cambridge, MA.
73
Easterlin, R.A., 1974. Does economic growth improve the human lot? Some empirical evidence. In: Davis, R., Reder, M. (Eds.), Nations and Households in Economic Growth: Essays in Honor of Moses Abramovitz. Academic Press, New York, pp. 89–125. Easterlin, R.A., 1995. Will raising the incomes of all increase the happiness of all? Journal of Economic Behavior and Organization 27 (1), 35–47. Easterlin, R.A., 2005. Feeding the illusion of growth and happiness: a reply to Hagerty and Veenhoven. Social Indicators Research 74, 429–443. Easterlin, R.A., Sawangfa, O., 2009. Happiness and Economic Growth: Does the crosssection predict time trends? Evidence from developing countries. IZA Discussion Paper n. 4000. Frey, B.S., Stutzer, A., 2002. What can economists learn from happiness research? Journal of Economic Literature 40, 402–435. Hagerty, R.H., Veenhoven, R., 2003. Wealth and Happiness revisited – growing national income does go with greater happiness. Social Indicators Research 64, 1–27. Hout, M., 2004. Getting the most out of the GSS income measures, GSS Methodological Report n. 101. Inglehart, R., Foa, R., Peterson, C., Wezel, C., 2008. Development, freedom and rising happiness. Perspectives on Psychological Science 3 (4), 264–285. Kenny, C., 1999. Does growth cause happiness, or does happiness cause growth? Kyklos 52 (1), 3–26. Layard, R, 2005. Happiness: Lessons from a New Science. The Penguin Press, New York. Myers, D.G., 2000. The funds, friends and faith of happy people. American Psychologist 55 (1), 56–67. Stevenson, B., Wolfers, J., 2008. Economic Growth and Subjective Well-Being: Reassessing the Easterlin Paradox, Brookings Papers on Economic Activity. Spring, pp. 1–87. Veblen, T., 1899. The Theory of the Leisure Class. MacMillan, New York. Veenhoven, R., 1991. Is happiness relative? Social Indicators Research 24, 1–34.
