Identifying the effect of age on willingness to take risks

Thomas Dohmen, Armin Falk , Bart Golsteyn, David Huffman, Uwe Sunde 21 January 2018



The average age of the population is rising in many developed countries. This is likely to have far-reaching implications for individual and collective decision making. For example, conventional wisdom says that older people are more conservative and less willing to take risks (e.g. Okun 1976). If true, this would have important consequences for economic, political, and social outcomes. Attitudes towards risk have a key impact on economic decisions (such as decisions about savings, investment, or labour market activity), demographic outcomes (such as decisions about fertility) and socio-political behaviour (such as voting).

In a recent article, we study the age trajectory of risk attitudes all the way from early adulthood until old age, using large representative panel datasets from the Netherlands and Germany (Dohmen et al. 2017). It is generally difficult to identify effects of age per se separately from cohort effects (i.e. the effects related to the life experiences or characteristics of a particular generation). Another challenge is distinguishing age effects from effects associated with calendar period. Our paper employs an approach that helps address these identification problems. Our results are consistent with age affecting willingness to take risks, perhaps through biological processes.

Age versus cohort effects

There are reasons to believe that age, cohort effects, and calendar period could all affect willingness to take risks. Older people may become less willing to take risks due to biological ageing processes – as is suggested, for example, by evidence that cognitive ageing is associated with declining willingness to take risks (Bonsang and Dohmen 2015). Cohort effects may reflect common experiences during particular points in life. For example, the experience of war or a major crisis might affect risk preferences (Malmendier and Nagel 2011). The same is true for fashions and patterns, such as cohort-specific variation in breastfeeding patterns (Falk and Kosse 2015). The period of observation may affect risk attitudes since recent events change individual expectations and thereby affect expected lifetime wealth. For example, a financial crisis might affect willingness to take risks. And, as Dohmen et al. (2006) show, the unexpectedly good performance of the German football team at the 2006 FIFA World Cup had the surprising effect of improving economic perceptions and expectations.


In the spirit of Heckman and Robb (1985), we disentangle age effects from period effects by replacing calendar period indicators with controls for the specific underlying factors that may change risk attitudes across periods. In our main specification, we use GDP growth to capture period effects.

To identify age effects separately from cohort effects, we use panel data which follows individuals from a given cohort over time. We use data from two different large, representative panel datasets – one from the Netherlands and one from Germany – each of which provides evidence for the age trajectory of risk attitudes all the way from early adulthood until old age.


The empirical results indicate that willingness to take risks declines with age, once calendar time and cohort effects are taken into account. The decreasing pattern is linear until approximately age 65. After age 65, the slope becomes flatter. The size of this effect is substantial – an increase of 10 years in the median age of a society leads to a reduction in mean willingness to take risks of 0.23 standard deviations, which is equivalent to 2.5% less investment in stocks or about 6% less self-employment. Such a change amounts to approximately half of the well-documented difference in willingness to take risks between men and women.

Various robustness analyses corroborate these findings. We obtain qualitatively similar results if we use lagged GDP growth, or exclude the year of the financial crisis (2009) from the estimation. Instead of GDP growth, we also try other indicators of economic conditions. Our results remain robust if we use stock market returns or yearly unemployment rates instead of GDP growth. Using fixed effects specifications, or following identification methods suggested by Deaton and Paxson (1994) and Browning et al. (2012) also gives qualitatively similar results.

Instead of using proxies for period effects, we also use a proxy for cohort effects to identify the age effects. We prefer using proxies for period effects over proxies for cohort effects because it is plausible that risk attitudes in periods are affected non-linearly by macroeconomic conditions, while this is not obvious for risk attitudes across cohorts. Relatedly, it is difficult to establish at which time in their lives cohorts may be affected by macroeconomic circumstances. Nevertheless, in a robustness check, we use inflation at age 18 as a substitute for cohort effects. The analysis delivers qualitatively similar results in the sense that risk aversion rises with age.

Concluding remarks

There has been little empirical evidence to shed light on the potential link between ageing and risk attitudes. Previous studies report negative relationships between risk taking and age based almost exclusively on cross-sectional data (e.g. Barsky et al. 1997, Donkers et al. 2001, Dohmen et al. 2011). Sahm (2012) is one of the only other papers which analyses changes in risk attitudes with age based on panel data. Sahm investigates the profile of risk attitudes among elderly birth cohorts (i.e. the 1931-1947 birth cohorts) and finds a modest decline in risk with age over this older age range. Our study is different in that it addresses the question of how risk attitudes vary over the entire age range, starting from early adulthood. Studying age effects on risk attitudes among youth is important because it adds perspective to the size of the age effects among the elderly. Looking at younger birth cohorts is also important in itself. So far, there is little evidence of the development of risk attitudes early in life. Tymula et al. (2012), for example, show that adolescents are more risk-averse than adults using samples of 33 young adolescents and 32 adults. Tymula et al. (2013) show that adolescents and elderly individuals are more risk averse than their midlife counterparts in a sample of 135 individuals. These articles do not disentangle cohort, period, and true age effects.

We analyse the age pattern of risk attitudes in larger samples from early adulthood until old age and control for cohort and period effects. Our result that risk preference decreases with age suggests that societies may become more risk-averse as a consequence of population ageing. Systematic changes in aggregate risk attitudes in an ageing society may have far-reaching consequences for economic, political, and social outcomes. 


Barsky, R, F T Juster, M Kimball and M Shapiro (1997), “Preference parameters and behavioral heterogeneity: an experimental approach in the health and retirement study”, Quarterly Journal of Economics, 112(2): 537-79. 

Bonsang, E and T Dohmen (2015), “Cognitive ageing and risk attitude”, Journal of Economic Behavior and Organization, forthcoming.

Browning, M, I Crawford and M Knoef (2012), “The age-period-cohort problem: Set identification and point identification”, cemmap, working paper CWP02/12.

Deaton, A and C Paxson (1994), “Saving, growth, and ageing in Taiwan”, in D Wise (ed), Studies in the Economics of Ageing, National Bureau of Economic Research: 331-62.

Dohmen, T, A Falk, B Golsteyn, D Huffman and U Sunde (2017), “Risk attitudes across the life course”, Economic Journal 127(605): F95-F116.

Dohmen, T, A Falk, D Huffman and U Sunde (2006), “Seemingly irrelevant events affect economic perceptions and expectations”, IZA, Discussion Paper 2275.

Dohmen, T, A Falk, D Huffman, U Sunde, J Schupp and G Wagner (2011), “Individual risk attitudes: Measurements, determinants and behavioral consequences”, Journal of the European Economic Association 9(3): 522-50.

Donkers, B, B Melenberg and A van Soest (2001), “Estimating risk attitudes using lotteries – a large sample approach”, Journal of Risk and Uncertainty 22(2): 165-95.

Falk, A and F Kosse (2015), “Breastfeeding duration, early life circumstances and the formation of human preferences”, University of Bonn, Unpublished manuscript.

Heckman, J and R Robb (1985), “Using longitudinal data to estimate age, period, and cohort effects in earnings equations”, in W Mason and S Fienberg (eds), Cohort Analysis in Social Research: Beyond the Identification Problem, New York: Springer Verlag.

Malmendier, U and S Nagel (2011), “Depression babies: Do macroeconomic experiences affect risk-taking?”, Quarterly Journal of Economics 126(1): 373-416.

Okun, M (1976), “Adult age and cautiousness in decision: A review of the literature”, Human Development 19: 220-233.

Sahm, C (2012), “How much does risk tolerance change?”, Quarterly Journal of Finance 2(4).

Tymula, A, L Rosenberg Belmaker, A Roy, L Ruderman, K Manson, P Glimcher and I Levy (2012), “Adolescents’ risk taking behavior is driven by tolerance to ambiguity”, Proceedings of the National Academy of Sciences 109(42): 17135-40.

Tymula, A, L Rosenberg Belmaker, L Ruderman, P Glimcher and I Levy (2013), “Like cognitive function, decision making across the life span shows profound age-related changes”, Proceedings of the National Academy of Sciences 110(42): 17143-8.



Topics:  Frontiers of economic research

Tags:  risk attitudes, willingness to take risks, age, Ageing, population, demographics, cohort effects, population ageing

Director and Professor at the Faculty of Economics and Business Administration, Maastricht University

Professor of Economics at the University of Bonn and CEPR Research Affiliate

Associate Professor at the Department of Economics, Maastricht University

Assistant Professor of Economics, Swarthmore College

Professor of Economics, University of St. Gallen and CEPR Research Affiliate