A large literature in behavioural and social sciences has found that human wellbeing follows a U-shape over age. Some theories have assumed that the U-shape is caused by unmet expectations that are felt painfully in midlife but beneficially abandoned and experienced with less regret during old age. In a unique panel of 132,609 life satisfaction expectations matched to subsequent realizations, Hannes Schwandt finds people to err systematically in predicting their life satisfaction over the life cycle. They expect – incorrectly – increases in young adulthood and decreases during old age. These errors are large, ranging from 9.8% at age 21 to -4.5% at age 68, they are stable over time and observed across socio-economic groups. These findings support theories that unmet expectations drive the age U-shape in wellbeing.
Behavioural and social scientists have shown increasing interest in self-reported life satisfaction and other subjective indicators as measures of human wellbeing. Using these measures a large and emerging literature has established that wellbeing follows a U-shape over age. Even though some controversy remains over the existence of this U-shape, it has been observed in over 50 nations, across socio-economic groups and recently also for great apes. Little is known about its origins. One theory is that the U-shape is driven by unmet aspirations which are painfully felt in midlife but beneficially abandoned later in life. A complementary theory builds on the neuroscientific finding that the emotional reaction to missed chances decreases with age so that the elderly might feel less regret about unmet aspirations.
Assuming that regret about unmet aspirations drives the U-shape implies that people err dramatically in predicting their wellbeing over the life-cycle. When young, people expect a bright future though actual wellbeing decreases. In old age expectations are adjusted downwards though actual wellbeing is rising. Human belief formation is known to exhibit systematic biases such as optimism and the underestimation of hedonic adaptation to changes in life circumstances. However, existing literatures typically analyze specific forecast settings with less emphasis on overall wellbeing measures or the role of age. The extent to which people err in predicting changes in their wellbeing over the life-cycle is unknown.
This study uses a unique panel data set from Germany that reports both, people’s current life satisfaction as well as their expectations about life satisfaction in five years time. The panel structure of the SOEP allows an individual’s expectation in a given year to be matched to the same individual’s realization five years ahead to form individual specific forecast errors. In total I analyze 23,161 individuals that are followed up to 10 years. This results in 132,609 individual specific life satisfaction forecast errors.
Figure 1 (A) below plots the averages of people’s current and expected life satisfaction by single years of age. In line with the existing literature current life satisfaction is U-shaped between ages 20 and 70, with peaks around ages 23 and 69, a local minimum in the mid-50s and a further decline after age 75. As the plot of life satisfaction expectations shows, this U-shape is not anticipated. During young adulthood people expect their life satisfaction to increase strongly. With age, expectations decrease but remain above current life satisfaction until the late 50s when the two graphs coincide. Thereafter expectations remain stable while actual life satisfaction increases, indicating that people do not anticipate the increase in old age wellbeing. After age 75 expectations decrease, simultaneously with current life satisfaction.
Figure 1: Expected life satisfaction, current life satisfaction and life satisfaction forecast errors over age.These different patterns in current and expected life satisfaction imply systematic forecast errors that change with age (Fig. 1B). Young adults in their 20s overestimate on average their future life satisfaction by about 0.7, or by about 10% (e.g. 0.685 ± 0.047 or 9.8% at age 21). After age 30 forecast errors decrease steadily, turning negative at age 55 and decreasing further until age 68 (-0.308 ± 0.057; or -4.52%) where after they remain at around -0.25.
Figure 2 shows that the age pattern in life satisfaction forecast errors is remarkably stable over time, regions, and socio-economic subgroups.
Figure 2: Life satisfaction forecast errors over age, by time periods, regions, gender and education.
What are the causes underlying this age bias? One well known source of systematic forecast errors is that people underestimate how quickly they adapt to socio-economic changes such as changes in income. Thus the observed age bias could be generated by the young expecting too much from anticipated income increases with the elderly, who face decreasing incomes, committing the opposite error. In the data, forecast errors indeed roughly match with the average income profile which is increasing during young adulthood and decreasing after age 50. Further, the age bias is slightly more pronounced for the highly educated who have steeper income profiles than those with less education.
However, the remarkable similarity across economically and culturally distinct regions and across gender suggests that some of the causes of the age bias go beyond age-related socio-economic characteristics. It is well established finding in psychological research that people tend to overestimate the likelihood of positive events and underestimate the likelihood of negative events. For example, people expect to enjoy healthier lives than average or underestimate the probability of being divorced. Optimism bias has also been demonstrated in non-human animals. Neuroscientific research has accumulated broad evidence that this bias is generated by selective processing of negative and positive information in the frontal brain which allows people to maintain biased expectations when confronted with discomforting evidence. This might provide a biological explanation for why life satisfaction expectations are overoptimistic during much of adulthood and adjust only slowly over time. It does not explain, though, why expectations remain stable after midlife while actual life satisfaction increases, implying negative forecast errors during old age. However, little is known about optimism in old age and existing evidence is conflicting.
How do the age associated errors in expected life satisfaction documented here relate to the age U-shape in wellbeing? Some theories have assumed that the U-shape is driven by unmet expectations that negatively affect people’s wellbeing in midlife but are abandoned and experienced with less regret during old age. The data reported here support this notion. Young adults have high aspirations that are subsequently unmet. And their life satisfaction decreases with age as long as expectations remain high and unmet. Aspirations are abandoned and expectations align with current wellbeing in the late 50s. This is the age when wellbeing starts to rise again. Further, given the disappointed expectations accumulated until that age, it is possible that wellbeing increases if the elderly learn to feel less regret (10). Following this interpretation of the U-shape in wellbeing, the observed negative forecast errors during old age might indicate that people do not anticipate the wellbeing enhancing effects of abandoning high aspirations and experiencing less regret.
Disseminating the knowledge of age associated forecast errors in life satisfaction could help people adjust their expectations, optimize important decisions in their life and suffer less when aspirations are not met. This might weaken the midlife drop in life satisfaction.
Hannes Schwandt is a Postdoctoral Research Associate at the Center for Health and Wellbeing at Princeton University. He is also a visiting researcher at the Center for Economic Performance at the London School of Economics and a research affiliate at IZA. His primary research fields are Health Economics and Labor Economics. His secondary research interests include Applied Microeconometrics and Behavioural Economics.
*This blog first appeared in LSE's Politics and Policy blog on 7 August, 2013.