Home > Doc > The happiness paradox: a formal explanation from psycho-economics > The empirical relevance of the well-being paradox

The happiness paradox: a formal explanation from psycho-economics

The empirical relevance of the well-being paradox

The happiness paradox was first raised by Easterlin (1974). By measuring happiness by means of self-reported ratings on subjective well-being (SWB), he shows that in the US happiness has not exhibited a definite rising trend since WWII, whereas real income per head has instead done so.[2] The evidence from recent econometric studies reinforces the paradox and makes it even more puzzling. In fact, the trend of SWB between the mid-1970s and the mid-1990s in the US is clearly downwards, and the same pattern emerges for Belgium and for the young component of the population of the UK (Blanchflower-Oswald 2004:1364 and 1368; Kenny 1999:14; Di Tella et al. 2003:817 and 814).

This strengthening of the paradox is not emphasised by the literature, although the SWB index has been successfully tested for reliability and validity by various methods (see Blanchflower-Oswald 2004 and Myers-Diener 1995 for references). Easterlin’s finding of no significant trend of SWB has been recently confirmed for the EU and for many European member-states (Kenny 1999:14; Blanchflower-Oswald 2004:1368; Di Tella et al. 2003:817).[3] Japan has also been often cited as a striking case, because real income per head rose sixfold in that country between 1958 and 1991, while the proportion of people rating themselves as ‘very happy’ did not seem to change over the same period (Frey-Stutzer 2002:77).[4]

The strong version of the paradox is also supported by other well-known facts, like the recent increase in mental depression, which has been tested as strictly inversely correlated with SWB (van Hemert et al. 2002; Abbey-Andrews 1986). Several studies show that depression has significantly increased in the US and other major developed countries since WWII, sometimes specifying that the phenomenon recurs across generations (Klerman 1988; 1993; Lavori et al. 1993; Olfson et al 2002; Rutter-Smith 1995; Lane 2000:347-8).[5] An even more dramatic index of declining SWB is the suicide rate. Strictly speaking, this concerns only a small amount of the population, but it may represent the deeper-lying malaise of a larger fraction of it, insofar as suicide may be attempted or even seriously considered but not committed (Oswald 1997).

The suicide rate increased for the US, the EU and Japan from the mid-1960s until the 1980s (Levi et al. 2003; Lane 2000:23). Lester-Yang’s (1997) survey of several studies shows that the correlation between income per head and suicide rates has been positively significant for the US since WWII, and for a cross-section of the European countries.[6] The picture appears less bleak since the 1980s, in that suicide rate has declined for the US, Japan, and for many European countries. However, it has risen for Ireland and Spain (Levi et al. 2003; Chishti et al. 2003:111), and the suicide rate among adolescents and young adults has also risen in the US, and in the four major European countries (Putnam 2000:262; Lane 2000:23).

The striking difference in the dynamics of the suicide rate between age groups suggests that the deterioration of well-being takes place across successive generations, although the passage to adulthood may enable each generation to improve its well-being with more efficacy. This suggestion is stressed by Putnam (2000), who argues that civil engagement – which he shows to be closely correlated to well-being, at least in the US – clearly deteriorates across generations.[7] By contrast, it is more doubtful that the passage to adulthood improves well-being. The econometric studies on SWB cited above show a clear U-shaped pattern of SWB with increasing age, with the trough at the age of 40.[8]

However, these studies are not based on panel data, and the U-shape may be due to a composition effect across different generations (Easterlin 2001:470; see also Pinquart 2001). If SWB is followed with the same cohorts, a constant or even declining trend emerges from ages 21-30 to 85-90 (Easterlin 2004:11179). This trend is confirmed by the data on suicide, which exhibits the highest rates among oldest people. In particular, in the US the suicide rate increased from 24.9 to 42.0 per 100,000 residents during the period 1990-1998, while among white widowed men the rate reached 84.0 (Institute of Medicine 2002).[9]

Technical progress and improved material well-being have not induced people to reduce their working time, as one would expect (see the often cited Keynes 1939). In the US both average annual and average weekly hours for men, but especially for women, have risen in the past two decades (Bluestone-Rose 2000). Since the late 1970s, overtime has increased as well (Golden 1998). However, Americans do not appear to be satisfied; rather they exhibit stress due to overwork (Schor 1992; Cross 1993; Jacobs-Gerson 1998). In the EU working time per employee has declined, mainly because of the introduction of regulations on the standard workday. However, the dynamics have decelerated in recent decades, and women’s participation especially has greatly increased, so that the average rate of the working age population has increased as well (Lehndorff 2000).

Canada seems to exhibit the same pattern (Osberg-Sharpe 1998). A detailed study conducted in Germany between 1985 and 1994 reveals that people, on average, would like to work less hours than they actually do. The study is interesting because it also reveals that the constraint on people’s desire to work less does not lie in the labour market, since macroeconomic conditions worsened between the two years considered, and mobility towards greater participation by women and part-time jobs for all greatly increased. It seems to lie instead in familiar conditions at home (Merz 2002).

The two main explanations for the paradox

The evidence of constant SWB while income per head has increased finds a straightforward explanation, although it is an unsatisfactory one, in the psychology literature. This explanation, which is called ‘set-point theory’, predicts a constant SWB in the long run on the basis of genetic and personality traits, whether these are dispositions towards happiness or unhappiness (Lykken-Tellegen 1996; Costa et al. 1987; Goldsmith-Campos 1986). In the short run, the level of SWB may be shocked by external events, like a rise in income, but the psychological mechanism of adaptation erodes the effect of the shock, thus bringing SWB to the long run level (Helson 1964; Brickman et al. 1978; Headey-Wearing 1989).

A number of criticisms have been levelled against this ‘theory’. Personality traits appear statistically to explain only a portion of the variance of SWB indices (Diener et al. 1999:279-80; Diener 1996). Adaptation seems to occur only slowly and even incompletely (Diener et al. 1999:280; Lucas et al. 2003; Easterlin 2004). Moreover, adaptation appears instead to conceal a strategy to substitute the goals to be pursued (Diener et al. 1999:284-5). Finally, personality traits are found to be significantly rising or decreasing from the age of 20 to the age of 60, while their variability during youth is usually viewed as natural (Srivastava et al. 2003). In the economic literature, the most examined explanation for the paradox is the ‘comparison theory’, which also finds a good grounding in psychology (Michalos 1985; Inglehart 1990; Kahneman et al. 1997).

Many authors base the explanation on comparison between aspirations and realisations. Easterlin provides the prime example of the use of this approach. He first assumes that SWB positively depends on current income and negatively depends on aspirations about future income, and that aspirations are based on past income. He then conjectures, and supports with some evidence, that “material aspirations change over life cycle roughly in proportion to income” (Easterlin 2001:473). As a result SWB may remain constant while income increases, and work effort is not discouraged. The same conclusions are reached in regard to the aspiration for increased consumption of positional goods, which are inherently scarce because of congestion or exclusion (Frank 1985; Hirsch 1976; Layard 1980; Ng 1978; Corneo-Jeanne 2001; Cooper et al. 2001).

However, both of the explanations favoured by economists appear implausible in the long run. In fact, in both cases, aspirations go systematically unrealised, so that some other explanation is needed to account for adjustment. Secondly, as Easterlin (2004) has recently acknowledged, SWB significantly depends on domains other than income and material concerns, like the family.[10] Why do individuals not adjust by changing their relative aspirations between different domains?


2 He also shows that SWB and real income per head are not correlated in a cross-section of developed countries in the 1960s.

3 The time-series study by Di Tella et al. (2003) finds a significant positive correlation between SWB and current income per head but does not furnish safe results regarding their long-run association.

4 Cross-sectional studies confirm the paradox by showing that richer countries do not exhibit significantly greater SWB, once Gdp per head exceeds half that in the US in mid-1990s (Helliwell 2003; Kelly 1999; Inglehart and Klingemann 2000).

5 The survey by Diener and Seligman (2004) concludes that the increase in depression may be tenfold, and rules out that this is a measurement artefact.

6 Similar findings have been obtained by Jungeilges and Kirchgaessner (2002), and Huang (1996). Moreover, according to Lester and Yang, if suicide rates are regressed against the unemployment rate and income per head for European countries, only the latter variable emerges as positively significant.

7 He efficaciously shows that the incidence of headaches, insomnia and indigestion was roughly the same in the late 1970s across age groups. Thereafter it was more pronounced, the younger the group (Putnam 2000:264). A similar finding applies to anxiety and neuroticism (Twenge 2000).

8 A less studied, but important issue is the decline of well-being during childhood and adolescence, i.e. before the age usually considered in the SWB surveys. It seems that this decline is significant, especially where satisfaction with the family is concerned (Nickerson and Nagle 2004).

9 The contrast with the U-shaped pattern may be resolved by distinguishing the cognitive component of SWB, which can cope with higher satisfaction with age, from the affective components of SWB, like depression, boredom and loneliness, which significantly point to a deterioration not only in the US, but also in Germany, Italy, and Spain (Campbell et al. 1976:36; Pinquart 2001).

10 Benabou and Tirole’s (2004) attempt to explain the paradox meets the first criticism but not the second. In fact, they argue that disappointments of optimistic beliefs vanish because individuals unconsciously repress unfavourable information, thus adjusting a ‘cognitive dissonance’. However, besides the curious conclusion that stagnant well-being is due to optimistic individuals, it is not clear why these individuals repeatedly repress information about the material world without shifting their attention to the domain of personal relationships.

By Prof. Maurizio Pugno

Next: An alternative explanation and the supporting literature

Summary: Index