Testing Social Production Functions with Incentives


[This experiment has been published in the Journal of the Economics of Ageing. Our working paper and an earlier discussion paper.]

The “aging employee” has recently become a hot topic in many fields of behavioural research. With the aim to determine the effects of different incentive schemes (piece rate, competition, choice between piece rate and competition, social or increased monetary incentives) on performance of young and older subjects, Christiane Schwieren and I looked at behaviour of a group of younger and older adults on a well-established real effort task.

“Social production functions-successful ageing” theory or SPF-SA (Steverink & Lindenberg, 2006; Steverink et al., 1998) is one of the most comprehensive motivation theories that explicitly considers age effects. It is an extension of “SPF” (Ormel, Lindenberg, Steverink, & Verbrugge, 1999), accounting for the motivational changes that occur with ageing. The SPF-SA theory posits that there are age-related changes in the availability of resources for needs satisfaction, with affection relatively more “age proof” than the two others as it depends less on performance. Two main processes guide these changes:

  • A patterned change in the availability of resources for the satisfaction of the three social needs over the life span: status satisfaction is the most difficult to maintain, followed by behavioural confirmation. Satisfaction of the need for affection is the easiest to maintain in relation to the two other needs.
  • A process of compensation and substitution regarding social need satisfaction over the life course. Behavioural confirmation and affection are substitutes of and compensate for declines in status need satisfaction, and affection need satisfaction also compensates and substitutes for the decline in behavioural confirmation need satisfaction. (Steverink & Lindenberg, 2006, p. 283)

Hence, one can extrapolate from SPF-SA theory that the relative prices of the three social needs are changing with age. That is, older adults should focus more on affection need satisfaction, while younger adults should rather focus on behavioural confirmation and status-need fulfilment. This in turn leads to the prediction that different incentives might motivate younger and older subjects. More specifically, as tournament incentives (“competition”) focus on status need satisfaction, they might motivate older subjects less than either piece rate incentives (behavioural confirmation & physical needs) or any kind of “social” incentives (e.g., doing something because it is important for someone else – behavioural confirmation or affection).

Concerning status need satisfaction, older adults might prefer not to participate in tournaments as they assume that status confirmation will be difficult for them. If however, they assume that they are in a situation where it is rather easy for them to fulfil their need of status confirmation, they might in fact choose competitive incentives over other incentives, especially when competing with other older subjects, as in our experiment.

It is important to note that SPF-SA theory is not the only theory that can yield testable predictions about age-related changes in motivation. In fact, it is possible that an economic life-cycle model with credit constraints can produce age-related changes in the desire for social vs. monetary rewards. The predictions of such a model would be similar to those produced by SPF-SA. In our view, the mechanism proposed by SPF-SA is however intuitively appealing and produces detailed predictions for an experiment that go beyond what a life–cycle model would predict. To give an example, based on SPF-SA we can predict differences between those older subjects that have a high opinion about their own relative ability in the experimental task and those that have low task-related self-esteem. We therefore focused on SPF-SA in the formulation of our hypotheses: we hypothesize age effects that depend on the incentive schemes used.


To test this, subjects participated in a real-effort (math) task with four incentive schemes: piece rate, tournament, choice between piece rate and tournament, and social or increased monetary incentives. On the whole, we found that:

  • Concerning performance, older subjects generally perform worse in all conditions than younger subjects. This effect supposedly is due to age differences in cognitive domains such as processing speed (Salthouse & Madden, 2008) and fluid intelligence (Bugg et al., 2006).
  • Younger and older subjects do however not differ in the increase in performance between the conditions. This shows that even though overall performance of older subjects on this task may be worse than that of younger subjects, the effect of conditions points in the same direction for both age groups (i.e. the difference in performance between the competition treatment and the piece rate treatment does not differ significantly between age groups).
  • When controlling for gender and the interaction of gender and age, the significant age effect disappears (driven by a change in standard errors; the coefficient increases in size).
  • Performance increases from the piece rate to all subsequent trials (and not between any of the subsequent trials).
  • Subjects seem to be equally motivated to perform well by competitive incentives and by piece rate settings with increased performance incentives, and slightly more by social incentives. As predicted by SPF-SA theory, and unlike Heyman and Ariely (2004), adding social incentives to the monetary rewards seems to have a stronger effect on effort provision than monetary incentives alone.
  • The performance-increasing effect of increased monetary incentives or additional social incentives seems to have a similar strength as a competitive environment, as we do not find significant increases in effort provision from the competition trial to the final trial with monetary or with social incentives.
  • Unlike the predictions of SPF-SA theory, we do not observe that the magnitude of the difference between social and increased monetary incentives changes with age.

Unlike the common stereotype that older adults “are less willing to learn, and implicitly less interested in working hard and competing” (Charness and Villeval, 2007), in the setting of our experiment we do not find decreased competitiveness, either in men or in women. We replicate the gender difference in competitiveness found in the literature (Müller and Schwieren, 2011; Niederle and Vesterlund, 2007): women enter a tournament less often than men. This effect holds for both age groups. The two factors that explain the most of this variance of tournament entry are the gender of participants and self-ranking. Participants who are more confident about their relative performance are more likely to compete than participants who are less confident, which is in line with SPF-predictions. Interestingly, with older subjects, we do not observe a significant difference in performance between those that chose to compete and those that chose a piece rate incentive scheme, but well a difference in beliefs about actual performance. This suggests that, though beliefs have a relatively high likelihood of being wrong, those subjects who believe that they are able to fulfil their status goal will try to attain it. Furthermore, women are less likely to compete (a result that also holds when controlling for absolute and believed performance) and age per se does not have a significant effect on the likelihood to enter a competition. We do not observe any general age-effects on under- or over-entry into competition. Instead, given their actual performance, too many young men enter into competition, and too many older women do not. At the same time, there is no significant over-entry by older men, nor is there a significant under-entry in young women. The design of our study however does not allow us to fully control for confounds such as risk preferences, wealth or other correlates that can also change with age and may be partly explaining the observed effects. Changing risk preferences in particular might be a strong confound, which we can only partially control for by taking participants’ beliefs into account. A risk loving participant would try to enter a competition even when believing that the odds of winning are too low to reach the higher expected value, whereas a risk neutral or averse participant would try to avoid the competitive setting when beliefs of winning are low. Another confound that might be of interest for explaining the observed differences can be the physiological changes (hormones, brain activity) that occur with ageing. Although the effects of hormones on behaviour in an ageing context are not well researched (e.g. Huffmeijer, van IJzendoorn and Bakermans-Kranenburg, 2012), it has been shown that hormones play a crucial role in competitive settings (Apicella et al., 2011; Buckert et al., 2014) and that their concentrations change with age (Crilly, Francis and Nordin, 1981). Also, it has been shown that the activation of brain structures underlying risk preferences changes with age (e.g. Lee, Leung, Fox, Gao and Chan, 2008), and that some of the regions involved in preferences for competitive settings overlap with “risk-regions” (in young adults; Decety, Jackson, Sommerville, Chaminade and Meltzoff, 2004). Though speculative, it might therefore be that age differences in preferences for competitive settings have a considerable amount of underlying physiological changes.