- Split View
-
Views
-
Cite
Cite
Matthias Rieger, Rui Mata, On the Generality of Age Differences in Social and Nonsocial Decision Making, The Journals of Gerontology: Series B, Volume 70, Issue 2, March 2015, Pages 200–212, https://doi.org/10.1093/geronb/gbt088
- Share Icon Share
Abstract
Empirical work with western populations suggests that aging is associated with changes in economic decision making, including, for example, increased prosocial tendencies. We investigated the generality of age effects in social and nonsocial domains by assessing various measures of economic decision making in a rural population from Morocco.
We measured age/cohort differences using a number of economic games (risk game, time preferences game, dictator game, trust game, and public goods game) in over 700 participants ranging from 17 to 92 years of age.
The results suggest a negative relation between age and risk taking and a concave relation between age and contribution to a public good, but no significant age trends in time preferences, altruism, trust, or trustworthiness.
Our results indicate that the relationship between aging and decision making is not universal and is shaped by local culture and the type of task rather than biological factors alone. More research is needed to understand the unique age trends prevalent in specific populations and tasks.
A number of theories make predictions concerning the link between aging and decision making (see Hanoch, Wood, & Rice, 2007; Peters, Hess, Västfjäll, & Auman, 2007, for overviews); some theories emphasize motivational changes (Carstensen, 2006; Depping & Freund, 2011), whereas others emphasize biological processes that change across the life span (Eppinger, Hämmerer, & Li, 2011). Motivational changes are usually thought to vary across cultures (Heckhausen, Wrosch, & Schulz, 2010), whereas biological changes are considered more general and universal or, at least, less culture dependent (Baltes, Staudinger, & Lindenberger, 1999). Consequently, one potentially useful way to distinguish between the contribution of motivational and biological changes to economic decision making is to determine their universality by comparing patterns of age differences across measures and cultures. For example, a recent study with a small cross-sectional sample of 30 older and 29 younger adults from the United States established an interesting pattern of age-related differences between social and nonsocial decision making (Roalf, Mitchell, Harbaugh, & Janowsky, 2011). In particular, the results revealed few systematic age differences in nonsocial economic games but more prosocial, fairness-oriented decisions in social economic games. But are these patterns universal across measures and populations?
In this article, we provide a first step in answering this question by investigating age/cohort differences in decision making in a non-Western population, a rural population from Morocco. The data set includes both social and nonsocial behavior and thus contributes to understanding age/cohort differences in both individual and prosocial decisions to provide a complete picture of age differences in economic decision making.
Age Differences in Individual Decision Making
A number of findings suggest that there are age differences in decision making (Mata, Josef, Samanez-Larkin, & Hertwig, 2011; Mata & Nunes, 2010; Mather, 2006; Thornton & Dumke, 2005). In this article, we consider age differences in risk preferences, time preferences, altruism, trust, and cooperation. In what follows, we discuss known patterns of age differences in such domains and associated theories.
Risky Choice
Why should risk preferences change across the life span? Life cycle theories emphasize the role as reproductive potential and competition on risk taking (Wilson & Daly, 1997). Specifically, life cycle theories see risk taking as a function of reproductive competition and predict that it should be more intense for younger men than women or older adults, thus suggesting that aging should be associated with reduced risk taking relative to young age. Such changes could be related at the proximal level to endocrinological changes, such as reduced testosterone levels (Mehta, Jones, & Josephs, 2008). As predicted, epidemiological data concerning sexual behavior and crime statistics suggest that risk taking peaks in young adulthood and declines with advancing age (Wilson & Daly, 1997). Similarly, self-report measures of risk tendencies also suggest a reduction in risk taking with increased age (Dohmen et al., 2011; Roalf et al., 2011). These theories, however, are mute concerning age differences as a function of task characteristics, such as learning and memory demands of the decision problem.
Aging is associated with structural and neuromodulatory changes thought to underlie age-related decline in aspects of cognitive function, including decision making (Eppinger et al., 2011). Age-related decline in basic cognitive abilities may lead to changes particularly in choice situations that demand the use of large amounts of information and complex decision strategies (Mata & Nunes, 2010; Mata, Schooler, & Rieskamp, 2007; Zamarian, Weiss, & Delazer, 2011), or extensive learning of options’ characteristics (Mata, von Helversen, & Rieskamp, 2010; Samanez-Larkin, Wagner, & Knutson, 2011; Zamarian et al., 2011). A recent meta-analysis suggests that age-related differences in behavioral measures of risk are a function of the learning requirements of the task (Mata et al., 2011): Older adults seem more risk seeking compared with younger adults when learning leads to risk-avoidant behavior, but more risk averse when learning leads to risk-seeking behavior, suggesting that learning deficits rather than stable risk preferences are responsible for the behavioral differences in such learning paradigms. In turn, Mata et al. found a heterogeneous pattern of results in tasks that did not involve learning. Some findings suggest that there are particular characteristics of decision problems that may contribute to this pattern. In particular, age differences seem to be most evident when participants are faced with avoidable risk or uncertainty, that is, when given the choice between a sure thing and a risky gamble (Mather et al., 2012). This may be in line with the idea that situations of avoidable risk are more diagnostic of underlying preferences or cognitive styles (Yechiam & Ert, 2011). In sum, cognitive and neural theories suggest that age differences in decision making differ as a function of task characteristics and may be, at least partly, a function of learning and memory demands imposed on decision makers.
Aging is also associated with motivational and affective changes that may affect risky choice, and different hypotheses exist regarding how age-related motivational changes affect decisions (Mata & Hertwig, 2011; Strough, Karns, & Schlosnagle, 2011). For example, some have emphasized that older adults’ goal structure could be affected by a need to maintain performance in a period of life in which losses become the rule, thus making older adults reallocate their efforts into the prevention of loss (Depping & Freund, 2011). Other theories have emphasized how the decision maker’s motivation to improve well-being with advanced age can lead older adults to enhance positive relative to negative information and thus potentially strive for gains (Carstensen, 2006). One possible implication of this latter view is that older adults are more willing to take risks in order to achieve positive (gain) outcomes. A recent meta-analysis that considered the role of domain (gain vs. loss) showed, however, little support for either perspective (Mata et al., 2011).
Finally, there is evidence for cultural differences in risk taking (Weber & Hsee, 1998). For example, Weber and Hsee found that Chinese were more risk taking than U.S. Americans regarding financial investments. Interestingly, differences in risk preference seemed to be associated primarily with cultural differences in the perception of the risk of the options rather than with attitude toward perceived risk. The extent to which age cohorts in a given culture differ in risk taking as a function of life experiences remains, however, to be determined.
Time Preferences
Different theories have been proposed that make predictions regarding how aging is related to time preferences, such as delay discounting (i.e., the preference for sooner, smaller rewards relative to larger, later rewards), with conflicting predictions that delay discounting should decline with age (Rogers, 2011), increase with age (Trostel & Taylor, 2001), or be minimized in middle age (Read & Read, 2004). Empirical results are also mixed. Most prior studies find either stability (Chao, Szrek, Pereira, & Pauly, 2009; Green, Myerson, Lichtman, Rosen, & Fry, 1996; Roalf et al., 2011) or reductions in delay discounting with advanced age (Green, Fry, & Myerson, 1994; Green, Myerson, & Ostaszewski, 1999; Harrison, Morten, & Williams, 2002; Löckenhoff, O’Donoghue, & Dunning, 2011; Reimers, Maylor, Stewart, & Chater, 2009), with only a minority reporting increases in discounting from young adulthood to older age (Read & Read, 2004). The discrepancies in existing studies may be partially related to interactions between individual difference variables (e.g., age) and demographic factors like education and income (Green et al., 1996; Reimers et al., 2009). One study that has attempted to uncover the mediating variables suggests that emotional and motivational variables account for age differences in intertemporal choice, whereas basic cognitive ability measures do not (Löckenhoff et al., 2011).
Social Decision Making
Social interactions are central to well-being, and there is considerable interest in understanding the link between aging and individuals’ social life and judgments (Blanchard-Fields, 2007; Samanez-Larkin & Carstensen, 2011). Older adults are better at social than at nonsocial reasoning although cognitive aging may take a toll even in the social domain (Thornton & Dumke, 2005), possibly through impact on theory of mind (Slessor, Phillips, & Bull, 2007). In general, data from self-reports suggest that older adults become more conscientious and agreeable (Terracciano, McCrae, Brant, & Costa, 2005), and both self-report and third-party ratings indicate that older individuals engage in more prosocial behaviors in everyday contexts, such as the work context (Ng & Feldman, 2008).
Concerning evidence from economic games, a meta-analysis of donation behavior in the dictator game found that older adults in Western samples tend to donate 30% more than the average dictator (Engel, 2011). Also, stud-ies have found increased fairness-oriented behavior on the part of older relative to younger adults in ultimatum games in samples from the United States (Roalf et al., 2011) and Australia (Bailey, Ruffman, & Rendell, 2013). Regarding public goods games, one study found that French seniors contributed more in a public goods economic game relative to younger adults (Charness & Villeval, 2009). Regarding trust games, a meta-analysis that compared student to typically older, nonstudent adult Western samples found that nonstudent samples are more trustworthy and return up to 80% of the amount relative to student ones (Johnson & Mislin, 2011). This matches results suggesting that older adults perceive others to be more trustworthy and approachable than younger adults (Castle et al., 2012). Nevertheless, aging may not be associated with ever increasing trust and prosocial behavior. Age per se may not be the driving force behind differences between student and other adult populations in prosocial behavior because factors such as education and wealth also differ starkly between these groups (Johnson & Mislin, 2011). For example, age is a not a good predictor of food sharing in small-scale societies, with other factors such as access to resources being stronger determinants of food sharing (Gurven, Hill, & Kaplan, 2002). Also, one study that examined a life-span sample that performed an economic trust game found that trust increased almost linearly from early childhood to early adulthood but stayed relatively constant across different adult age groups (Sutter & Kocher, 2007).
Finally, there are important cultural differences in prosocial behavior. For example, North American and Asian populations seem to be on average more prosocial relative to African ones (Johnson & Mislin, 2011). These differences may have to do with the social structures and incentives prevalent in each culture: Cultures that have more incentives for cooperation tend to be more fairness oriented, trusting, and prosocial (Henrich et al., 2005). In sum, although some findings suggest aging could be associated with increased trust and prosocial behavior, there may be important cultural components that contribute to differences in social behavior and potentially determine or moderate age-related trends.
Relation Between Economic Games
The literature on economic behavior is inconclusive regarding to what extent different measures capture the same underlying cognitive processes and preferences (Schonberg, Fox, & Poldrack, 2011). Some have expressed doubt about the ability of economic measures to capture stable preferences given the dramatic differences in behavior generated by similar economic games (Berg, Dickhaut, & McCabe, 2005). Nevertheless, some studies have found medium correlations within measures of the same domain (Dohmen et al., 2011), as well as between domains (Schechter, 2007). In general, to the extent that aging is associated with general changes in economic preferences, say regarding risk, one could expect general effects across social and nonsocial domains. Alternatively, to the extent that different tasks measure different constructs, we may not find such general patterns as a function of aging.
Overview of This Study
Given the number of theories and heterogeneity of findings regarding the link between age and decision making, we aimed to contribute to this issue by providing an assessment of age-related differences using a number of economic games, including measures of risk taking, time preferences, altruism, trust, and cooperation from the same individuals. Only a minority of studies investigating aging has used multiple decision tasks, provided performance-contingent, incentive-compatible payment (see Mata et al., 2011, for an overview), or considered both social and nonsocial domains (Roalf et al., 2011), and so it is unclear to what extent age-related differences are observed systematically across tasks and domains. In fact, although some progress has been made in assessing and systematizing age differences in risk taking and risky choice (Mata et al., 2011), as well as the dictator (Engel, 2011) and trust (Johnson & Mislin, 2011) games, there are fewer efforts to study the impact of aging on several of these games simultaneously.
In our work, we relied on a large, age-heterogeneous sample of adults from rural Morocco (N = 740). Our work has the potential to contribute to the issue of assessing whether age differences can be found across populations and cultures by providing a non-Western sample that can be compared with past and future data sets. In addition, it contributes to understanding the generality of age differences in economic decisions across measures because it uses several economic games spanning social and nonsocial decision making. For example, to the extent that aging brings about increased prosocial behavior, we should be able to detect such shifts across economic measures of social behavior, such as dictator, trust, and public goods games.
Method
Participants
The data were collected within the scope of an ongoing impact evaluation of a large rural development project called INDH. The sampling design was driven by the identification strategy of the evaluation. The latter is quasiexperimental and based on a regression discontinuity design in that all communes with a poverty rate of more than 30% are eligible for the project. As a consequence, we sampled the 61 communes around this threshold in rural Morocco. Although the sample is thus not fully representative of rural Morocco, it has the advantage that communes and households are homogeneous in terms of wealth that could be a potential confound. Figure 1 shows the location of the sampled communes. In each commune, 15 households were randomly picked and invited to participate in the experimental session consisting of a number of economic games. The household head was handed an invitation and was able to designate a person in case he/she could not attend the experimental session. Eighty-one percent of the households invited participated in the experimental sessions. The full sample consists of 740 participants, with a mean (median) age of 53 (52) and ranging from 17 to 92 years (SD = 15). The games were piloted in three communes. After the first two communes, we slightly adjusted the payouts in the time preference game to increase variation in the sample. Thus, we do not use the time preference data from these two first communes and only have 716 observations for this game. In rare cases, people dropped out of the game session. As a consequence, we only have data for 738 people in the trust game, as well as 729 participants in the public goods game. Finally, we have missing age information for one participant. This explains the small differences in sample sizes across models and the total number of participants. However, because these were rare and nonsystematic cases, we deal with missing data by list-wise deletion. Men are in the majority in the sample (90%), which reflects the fact that rural areas are dominated by men, at least in the public sphere. Likewise, 92% of participants are head of their household. Only about 35% of the sample is literate (i.e., answer yes to the question “Can you read and write at least one language?”), which is lower than the overall rate in Morocco of about 56% or typical Western samples that display literacy rates of close to 100% (CIA, 2012), as would be expected given our sample is rural and poorer than the Moroccan population at large. Household size is on average 6.6. All communes are rural, and the official poverty rate ranges from 28% to 31% of the total population. Participants decided on economic games involving real monetary amounts. In total, people received on average 73 (Min = 35, Max = 122) Moroccan Dirhams (henceforth MAD) or roughly 7 EURO in earnings after the experimental session, which took about 2.5hr. The size of the payouts seems large enough to incentivize people to reveal their true preferences. A related household survey in the same set of communes in 2011 indicated that daily average household revenue is 85 MAD (Median = 60 MAD), and daily household consumption per capita is 12 MAD.
Economic Games
The experimental session consisted of five activities always played in the same order: risk game, time preferences game, dictator game, trust game, and public goods game. The project adopted protocols from the microdevelopment literature for these games, but we adjusted protocols to the local context (e.g., wording and examples) and translated instructions to Arab and local dialects. The experimental sessions were conducted in groups of up to 15 individuals by a team of four experimenters in the commune administration’s building. All instructions were given orally with the help of a blackboard in the respective language or dialect (e.g., Arab, Berber dialects). Instructions were repeated in private to the individuals, and experimenters ensured that participants understood the rules of each game by asking questions about the consequences of particular actions in each game. Participants were told they would be conducting a number of “activities,” and for each activity, participants first received instructions in the group and were further instructed singly. They made their decisions anonymously and privately in separate rooms.
With the exception of the risk game, we use the protocols and session procedures by Gilligan, Pasquale, and Samii (2010) that have been used in Nepal, as well as in ongoing impact evaluations by Michael Gilligan (New York University) in Burkina Faso, and in collaboration with Radu Ban (Gates Foundation) and Matthias Rieger (Graduate Institute, Geneva) in Cambodia. These protocols in turn are based on field protocols by Karlan (2005). The trust game originates from Berg, Dickhaut, and McCabe (1995), and the public goods game is due to Barrett (2005). The risk game comes from Charness and Gneezy (2010), which built on earlier work by Gneezy and Potters (1997). The risk game has also been used in the field in Senegal by Charness and Viceisza (2012). In the following, we provide further details concerning the economic games participants played.
Risk game.—
To measure risk taking, we endowed each participant with 10 MAD. The individual could then choose to invest any amount in a risky option with a success probability of 50%. Investment success was decided with a flip of a coin. In case of success, the amount invested was tripled and added to the uninvested, sure amount. Suppose a participant invested 4 MAD and kept the remaining 6 MAD. In the case of a successful investment, the person received 18 MAD (4 MAD × 3 + 6 MAD). If the flip of coin was unfavorable, the person was left with 6 MAD. The amount invested in the risky gamble was used as the dependent measure, and a risk-neutral or risk-seeking person would invest the whole amount in the risky gamble.
Time preferences game.—
The time preferences game measured time preferences by asking participants to choose between an amount of money today (10 MAD) and a series of larger amounts that could be collected from the commune administration the following week. The experimenter presented each option one at a time and recorded the switching point between taking the money now and waiting for 1 week. The different six amounts in 1 week were (a) 11, (b) 12, (c) 13, (d) 14, (e) 15, and (f) 16. We employ two indicators that can be interpreted as measures of relative importance of time relative to money or vice versa (Doyle, 2013). First, we use the switching point chosen: For example, a person who would prefer 11 in 1 week to 10 MAD now is thought of as being more patient (or valuing money less relative to 1 week) than someone who would only take 16 MAD (but no less) to wait 1 week. Second, we build a binary indicator that takes on a value of 1 if the person prefers to wait until next week starting with the first option (11 MAD) and zero otherwise. These two measures yield qualitatively similar results.
Dictator game.—
The dictator game is usually used to measure altruism or generosity (Engel, 2011). The game was played in three scenarios. In each scenario, the experimenters gave the participant 5 MAD and asked the participant to decide how much to keep and how much to give away. In particular, a sheet was placed in front of the participant with a dividing line. The money was placed on his/her side and in order to donate the participant had to push the donation on the other side over the line. In the first scenario, the money could be donated to an anonymous, poor family in the decision maker’s own commune. In the second scenario, the receiver was a family located in another, neighboring commune. In the third scenario, the person could donate to a development project in the commune. The three scenarios were played independently and in sequence. Moreover, participants made their decisions in private so as to minimize reciprocity and social desirability effects. The experimenters determined which family would be the recipient of the total amount after each session in a discussion with the commune chief. We use the amount donated out of the 5 MAD as a measure of altruism. We present separate results for each scenario and consider the correlation of age and total amount given in the three games.
Trust game.—
The trust game is typically interpreted to measure a sense of trust and trustworthiness within a community (Schechter, 2006). The experimenter randomly and anonymously matched pairs of participants to participate in an investment game. One person was assigned the role of “sender” and the other of “receiver.” Both were initially endowed with 10 MAD in denominations of 1 MAD. The sender then decided how much to send to or “trust” in the receiver. Specifically, a sheet with a diving line was placed in front of the participants. On each side of the line 10 MAD were placed. The sender was asked to “send” money to the other person, and the money sent was tripled in front of the sender. The receiver could then decide how much to keep and how much to return to the sender. We use the amount sent as a measure of trust, and the amount returned as an indicator of trustworthiness.
Public goods game.—
The public goods game measured the willingness to contribute to one’s community (Zelmer, 2003). The experimenter endowed each participant with 12 MAD and asked the participant to decide in private between keeping the money or, alternatively, contributing the whole amount to the group. If the participant decided to contribute to the public good, a common payoff to all group members was increased (2 MAD). If the participant did not contribute, the participant would receive the group payoff plus the initial, personal endowment. For example, if all 15 participants in a group contributed to the public good, each individual would receive 30 MAD. If one individual defected, the participants who contributed to the common pool would receive 28 MAD, whereas the defector would earn 40 MAD (28 MAD plus the initial 12 MAD), thus free riding on the group’s generosity. We use the decision to contribute to the public good as a measure of the participant’s sense of obligation toward the group.
Results
Figure 2 provides an overview of the results in each economic game and respective age patterns by presenting average results for discrete age groups. In addition, we quantified these patterns in a multivariate framework using age as a continuous variable. Specifically, we employed mixed-effects regression models to data for each game following a three-step approach: First, we quantified linear age trends (Model 1), second, we tested for quadratic age trends (Model 2), and third, conditional on results from Model 2, we tested for linear or quadratic age trends controlling for potential variables of interest, such as gender, literacy, and household size (Model 3). Tables 1 and 2 present the regression results for each economic game. The rationale for including control variables in Model 3 is to assess whether such variables can account for potential age-related/cohort differences in economic decision making. For example, younger participants are more likely to be literate and numerate due to economic progress, which may affect economic decision making. Furthermore, we controlled for behavior in other economic games when appropriate (see last four rows in Table 1). For example, risk-taking individuals may be more likely to make a risky investment in others in the trust game or contribute to a risky public good in the public goods game (Schechter, 2007); in the trust game, the amount returned by the “receiver” is likely a function of the amount received, so it is standard to control for the latter to capture reciprocity (Schechter, 2007). Finally, Model 3 also includes commune dummies (not presented in Table 1) to estimate commune-level fixed effects to control for potentially relevant but unobserved factors specific to a commune (e.g., demographic structure, aggregate commune social capital, population size, infrastructure, economic integration, and remoteness) and session (e.g., time of day, day of the week, small deviations from protocol). For example, prosocial behavior may be a function of the commune’s age distribution, so controlling for commune effects allows us to control for this factor. In these analyses, the commune fixed effects are equivalent to session fixed effects because only one game session took place in each commune. Note that standard errors are clustered at the commune (i.e., session) level in all three models in order to accommodate common shocks in the error component that otherwise mislead statistical inference. We also employed probit models for the results from the public goods game for which responses are binary, but we do not report the qualitatively similar probit estimates for the sake of brevity. Figure 2 and Tables 1 and 2 present the results for each economic game. In the following, we describe the results for each of the five economic games presented to participants.
Predictors . | Risk game . | Time preferences game . | Trust game (sender) . | Trust game (receiver) . | Public goods game . | |||||
---|---|---|---|---|---|---|---|---|---|---|
B . | SE . | B . | SE . | B . | SE . | B . | SE . | B . | SE . | |
Model 1: Age | ||||||||||
Age | −0.0127* | 0.0048 | 0.0005 | 0.0012 | −0.0075 | 0.0083 | 0.0049 | 0.0153 | −0.0034* | 0.0012 |
R2 | 0.01 | 0.001 | 0.001 | 0.003 | 0.01 | |||||
Model 2: Age + Age2 | ||||||||||
Age | 0.0144 | 0.0293 | 0.0082 | 0.0077 | −0.0283 | 0.0515 | −0.0197 | 0.1047 | 0.0130+ | 0.0069 |
Age2 | −0.0002 | 0.0003 | −0.0001 | 0.0001 | 0.0002 | 0.0004 | 0.0002 | 0.0010 | −0.0002* | 0.0001 |
R2 | 0.005 | 0.001 | 0.003 | 0.005 | 0.02 | |||||
Model 3: Age + Age2 + Covariates + Commune dummies | ||||||||||
Age | −0.0106* | 0.0049 | −0.0004 | 0.0015 | −0.0077 | 0.0101 | −0.0075 | 0.0110 | 0.0153* | 0.0073 |
Age2 | −0.0002* | 0.0001 | ||||||||
Gender | 0.4867 | 0.3985 | −0.0748 | 0.0665 | 0.2558 | 0.4909 | −0.4575 | 0.5242 | 0.1333+ | 0.0792 |
Literacy | 0.4719* | 0.1767 | −0.0227 | 0.0450 | 0.5535+ | 0.2837 | −0.0335 | 0.3736 | 0.0659 | 0.0408 |
Household size | 0.0422 | 0.0366 | 0.0065 | 0.0056 | 0.0442 | 0.0456 | 0.0332 | 0.0848 | −0.0076 | 0.0067 |
Risk game | 0.0088 | 0.0089 | −0.0147 | 0.0767 | 0.0612 | 0.1005 | −0.0008 | 0.0096 | ||
Time preferences game | 0.2060 | 0.2065 | 0.4766 | 0.3298 | −0.0006 | 0.4454 | −0.0247 | 0.0389 | ||
Dictator game (total) | 0.2572* | 0.0504 | 0.1457* | 0.0588 | 0.0096+ | 0.0056 | ||||
Trust game (sender) | 0.5355* | 0.0366 | ||||||||
R2 | 0.15 | 0.03 | 0.24 | 0.71 | 0.15 | |||||
N | 707 | 707 | 334 | 371 | 696 |
Predictors . | Risk game . | Time preferences game . | Trust game (sender) . | Trust game (receiver) . | Public goods game . | |||||
---|---|---|---|---|---|---|---|---|---|---|
B . | SE . | B . | SE . | B . | SE . | B . | SE . | B . | SE . | |
Model 1: Age | ||||||||||
Age | −0.0127* | 0.0048 | 0.0005 | 0.0012 | −0.0075 | 0.0083 | 0.0049 | 0.0153 | −0.0034* | 0.0012 |
R2 | 0.01 | 0.001 | 0.001 | 0.003 | 0.01 | |||||
Model 2: Age + Age2 | ||||||||||
Age | 0.0144 | 0.0293 | 0.0082 | 0.0077 | −0.0283 | 0.0515 | −0.0197 | 0.1047 | 0.0130+ | 0.0069 |
Age2 | −0.0002 | 0.0003 | −0.0001 | 0.0001 | 0.0002 | 0.0004 | 0.0002 | 0.0010 | −0.0002* | 0.0001 |
R2 | 0.005 | 0.001 | 0.003 | 0.005 | 0.02 | |||||
Model 3: Age + Age2 + Covariates + Commune dummies | ||||||||||
Age | −0.0106* | 0.0049 | −0.0004 | 0.0015 | −0.0077 | 0.0101 | −0.0075 | 0.0110 | 0.0153* | 0.0073 |
Age2 | −0.0002* | 0.0001 | ||||||||
Gender | 0.4867 | 0.3985 | −0.0748 | 0.0665 | 0.2558 | 0.4909 | −0.4575 | 0.5242 | 0.1333+ | 0.0792 |
Literacy | 0.4719* | 0.1767 | −0.0227 | 0.0450 | 0.5535+ | 0.2837 | −0.0335 | 0.3736 | 0.0659 | 0.0408 |
Household size | 0.0422 | 0.0366 | 0.0065 | 0.0056 | 0.0442 | 0.0456 | 0.0332 | 0.0848 | −0.0076 | 0.0067 |
Risk game | 0.0088 | 0.0089 | −0.0147 | 0.0767 | 0.0612 | 0.1005 | −0.0008 | 0.0096 | ||
Time preferences game | 0.2060 | 0.2065 | 0.4766 | 0.3298 | −0.0006 | 0.4454 | −0.0247 | 0.0389 | ||
Dictator game (total) | 0.2572* | 0.0504 | 0.1457* | 0.0588 | 0.0096+ | 0.0056 | ||||
Trust game (sender) | 0.5355* | 0.0366 | ||||||||
R2 | 0.15 | 0.03 | 0.24 | 0.71 | 0.15 | |||||
N | 707 | 707 | 334 | 371 | 696 |
Note. All standard errors are clustered at the commune (experimental session) level.
*p < .05. +p < .10.
Predictors . | Risk game . | Time preferences game . | Trust game (sender) . | Trust game (receiver) . | Public goods game . | |||||
---|---|---|---|---|---|---|---|---|---|---|
B . | SE . | B . | SE . | B . | SE . | B . | SE . | B . | SE . | |
Model 1: Age | ||||||||||
Age | −0.0127* | 0.0048 | 0.0005 | 0.0012 | −0.0075 | 0.0083 | 0.0049 | 0.0153 | −0.0034* | 0.0012 |
R2 | 0.01 | 0.001 | 0.001 | 0.003 | 0.01 | |||||
Model 2: Age + Age2 | ||||||||||
Age | 0.0144 | 0.0293 | 0.0082 | 0.0077 | −0.0283 | 0.0515 | −0.0197 | 0.1047 | 0.0130+ | 0.0069 |
Age2 | −0.0002 | 0.0003 | −0.0001 | 0.0001 | 0.0002 | 0.0004 | 0.0002 | 0.0010 | −0.0002* | 0.0001 |
R2 | 0.005 | 0.001 | 0.003 | 0.005 | 0.02 | |||||
Model 3: Age + Age2 + Covariates + Commune dummies | ||||||||||
Age | −0.0106* | 0.0049 | −0.0004 | 0.0015 | −0.0077 | 0.0101 | −0.0075 | 0.0110 | 0.0153* | 0.0073 |
Age2 | −0.0002* | 0.0001 | ||||||||
Gender | 0.4867 | 0.3985 | −0.0748 | 0.0665 | 0.2558 | 0.4909 | −0.4575 | 0.5242 | 0.1333+ | 0.0792 |
Literacy | 0.4719* | 0.1767 | −0.0227 | 0.0450 | 0.5535+ | 0.2837 | −0.0335 | 0.3736 | 0.0659 | 0.0408 |
Household size | 0.0422 | 0.0366 | 0.0065 | 0.0056 | 0.0442 | 0.0456 | 0.0332 | 0.0848 | −0.0076 | 0.0067 |
Risk game | 0.0088 | 0.0089 | −0.0147 | 0.0767 | 0.0612 | 0.1005 | −0.0008 | 0.0096 | ||
Time preferences game | 0.2060 | 0.2065 | 0.4766 | 0.3298 | −0.0006 | 0.4454 | −0.0247 | 0.0389 | ||
Dictator game (total) | 0.2572* | 0.0504 | 0.1457* | 0.0588 | 0.0096+ | 0.0056 | ||||
Trust game (sender) | 0.5355* | 0.0366 | ||||||||
R2 | 0.15 | 0.03 | 0.24 | 0.71 | 0.15 | |||||
N | 707 | 707 | 334 | 371 | 696 |
Predictors . | Risk game . | Time preferences game . | Trust game (sender) . | Trust game (receiver) . | Public goods game . | |||||
---|---|---|---|---|---|---|---|---|---|---|
B . | SE . | B . | SE . | B . | SE . | B . | SE . | B . | SE . | |
Model 1: Age | ||||||||||
Age | −0.0127* | 0.0048 | 0.0005 | 0.0012 | −0.0075 | 0.0083 | 0.0049 | 0.0153 | −0.0034* | 0.0012 |
R2 | 0.01 | 0.001 | 0.001 | 0.003 | 0.01 | |||||
Model 2: Age + Age2 | ||||||||||
Age | 0.0144 | 0.0293 | 0.0082 | 0.0077 | −0.0283 | 0.0515 | −0.0197 | 0.1047 | 0.0130+ | 0.0069 |
Age2 | −0.0002 | 0.0003 | −0.0001 | 0.0001 | 0.0002 | 0.0004 | 0.0002 | 0.0010 | −0.0002* | 0.0001 |
R2 | 0.005 | 0.001 | 0.003 | 0.005 | 0.02 | |||||
Model 3: Age + Age2 + Covariates + Commune dummies | ||||||||||
Age | −0.0106* | 0.0049 | −0.0004 | 0.0015 | −0.0077 | 0.0101 | −0.0075 | 0.0110 | 0.0153* | 0.0073 |
Age2 | −0.0002* | 0.0001 | ||||||||
Gender | 0.4867 | 0.3985 | −0.0748 | 0.0665 | 0.2558 | 0.4909 | −0.4575 | 0.5242 | 0.1333+ | 0.0792 |
Literacy | 0.4719* | 0.1767 | −0.0227 | 0.0450 | 0.5535+ | 0.2837 | −0.0335 | 0.3736 | 0.0659 | 0.0408 |
Household size | 0.0422 | 0.0366 | 0.0065 | 0.0056 | 0.0442 | 0.0456 | 0.0332 | 0.0848 | −0.0076 | 0.0067 |
Risk game | 0.0088 | 0.0089 | −0.0147 | 0.0767 | 0.0612 | 0.1005 | −0.0008 | 0.0096 | ||
Time preferences game | 0.2060 | 0.2065 | 0.4766 | 0.3298 | −0.0006 | 0.4454 | −0.0247 | 0.0389 | ||
Dictator game (total) | 0.2572* | 0.0504 | 0.1457* | 0.0588 | 0.0096+ | 0.0056 | ||||
Trust game (sender) | 0.5355* | 0.0366 | ||||||||
R2 | 0.15 | 0.03 | 0.24 | 0.71 | 0.15 | |||||
N | 707 | 707 | 334 | 371 | 696 |
Note. All standard errors are clustered at the commune (experimental session) level.
*p < .05. +p < .10.
Predictors . | Dictator game . | |||||||
---|---|---|---|---|---|---|---|---|
Total . | Own commune . | Family in own commune . | Family in other commune . | |||||
B . | SE . | B . | SE . | B . | SE . | B . | SE . | |
Model 1: Age | ||||||||
Age | 0.0067 | 0.0109 | 0.0087+ | 0.0048 | −0.0034 | 0.0036 | 0.0013 | 0.0039 |
R2 | 0.001 | 0.01 | 0.00 | 0.001 | ||||
Model 2: Age + Age2 | ||||||||
Age | −0.0032 | 0.0636 | −0.0072 | 0.0268 | 0.0068 | 0.0241 | −0.0028 | 0.0212 |
Age2 | 0.0001 | 0.0005 | 0.0001 | 0.0002 | −0.0001 | 0.0002 | 0.0000 | 0.0002 |
R2 | 0.002 | 0.006 | 0.001 | 0.003 | ||||
Model 3: Age + Age2 + Covariates + Commune dummies | ||||||||
Age | 0.0149 | 0.0144 | 0.0106+ | 0.0060 | 0.0006 | 0.0046 | 0.0037 | 0.0049 |
Gender | 0.4088 | 0.5015 | 0.0143 | 0.2011 | 0.2629 | 0.2094 | 0.1316 | 0.1449 |
Literacy | 0.5874+ | 0.3255 | 0.1545 | 0.1209 | 0.2616* | 0.1299 | 0.1713 | 0.1302 |
Household size | 0.0775 | 0.0593 | 0.0037 | 0.0249 | 0.0531* | 0.0191 | 0.0207 | 0.0213 |
Risk preference | 0.1610* | 0.0594 | 0.0480+ | 0.0264 | 0.0666* | 0.0226 | 0.0464+ | 0.0248 |
Patience | 0.3895 | 0.3600 | 0.2950* | 0.1350 | 0.0443 | 0.1414 | 0.0502 | 0.1319 |
R2 | 0.1 | 0.09 | 0.09 | 0.08 | ||||
N | 707 | 707 | 707 | 707 |
Predictors . | Dictator game . | |||||||
---|---|---|---|---|---|---|---|---|
Total . | Own commune . | Family in own commune . | Family in other commune . | |||||
B . | SE . | B . | SE . | B . | SE . | B . | SE . | |
Model 1: Age | ||||||||
Age | 0.0067 | 0.0109 | 0.0087+ | 0.0048 | −0.0034 | 0.0036 | 0.0013 | 0.0039 |
R2 | 0.001 | 0.01 | 0.00 | 0.001 | ||||
Model 2: Age + Age2 | ||||||||
Age | −0.0032 | 0.0636 | −0.0072 | 0.0268 | 0.0068 | 0.0241 | −0.0028 | 0.0212 |
Age2 | 0.0001 | 0.0005 | 0.0001 | 0.0002 | −0.0001 | 0.0002 | 0.0000 | 0.0002 |
R2 | 0.002 | 0.006 | 0.001 | 0.003 | ||||
Model 3: Age + Age2 + Covariates + Commune dummies | ||||||||
Age | 0.0149 | 0.0144 | 0.0106+ | 0.0060 | 0.0006 | 0.0046 | 0.0037 | 0.0049 |
Gender | 0.4088 | 0.5015 | 0.0143 | 0.2011 | 0.2629 | 0.2094 | 0.1316 | 0.1449 |
Literacy | 0.5874+ | 0.3255 | 0.1545 | 0.1209 | 0.2616* | 0.1299 | 0.1713 | 0.1302 |
Household size | 0.0775 | 0.0593 | 0.0037 | 0.0249 | 0.0531* | 0.0191 | 0.0207 | 0.0213 |
Risk preference | 0.1610* | 0.0594 | 0.0480+ | 0.0264 | 0.0666* | 0.0226 | 0.0464+ | 0.0248 |
Patience | 0.3895 | 0.3600 | 0.2950* | 0.1350 | 0.0443 | 0.1414 | 0.0502 | 0.1319 |
R2 | 0.1 | 0.09 | 0.09 | 0.08 | ||||
N | 707 | 707 | 707 | 707 |
Note. All standard errors are clustered at the commune (experimental session) level.
*p < .05. +p < .10.
Predictors . | Dictator game . | |||||||
---|---|---|---|---|---|---|---|---|
Total . | Own commune . | Family in own commune . | Family in other commune . | |||||
B . | SE . | B . | SE . | B . | SE . | B . | SE . | |
Model 1: Age | ||||||||
Age | 0.0067 | 0.0109 | 0.0087+ | 0.0048 | −0.0034 | 0.0036 | 0.0013 | 0.0039 |
R2 | 0.001 | 0.01 | 0.00 | 0.001 | ||||
Model 2: Age + Age2 | ||||||||
Age | −0.0032 | 0.0636 | −0.0072 | 0.0268 | 0.0068 | 0.0241 | −0.0028 | 0.0212 |
Age2 | 0.0001 | 0.0005 | 0.0001 | 0.0002 | −0.0001 | 0.0002 | 0.0000 | 0.0002 |
R2 | 0.002 | 0.006 | 0.001 | 0.003 | ||||
Model 3: Age + Age2 + Covariates + Commune dummies | ||||||||
Age | 0.0149 | 0.0144 | 0.0106+ | 0.0060 | 0.0006 | 0.0046 | 0.0037 | 0.0049 |
Gender | 0.4088 | 0.5015 | 0.0143 | 0.2011 | 0.2629 | 0.2094 | 0.1316 | 0.1449 |
Literacy | 0.5874+ | 0.3255 | 0.1545 | 0.1209 | 0.2616* | 0.1299 | 0.1713 | 0.1302 |
Household size | 0.0775 | 0.0593 | 0.0037 | 0.0249 | 0.0531* | 0.0191 | 0.0207 | 0.0213 |
Risk preference | 0.1610* | 0.0594 | 0.0480+ | 0.0264 | 0.0666* | 0.0226 | 0.0464+ | 0.0248 |
Patience | 0.3895 | 0.3600 | 0.2950* | 0.1350 | 0.0443 | 0.1414 | 0.0502 | 0.1319 |
R2 | 0.1 | 0.09 | 0.09 | 0.08 | ||||
N | 707 | 707 | 707 | 707 |
Predictors . | Dictator game . | |||||||
---|---|---|---|---|---|---|---|---|
Total . | Own commune . | Family in own commune . | Family in other commune . | |||||
B . | SE . | B . | SE . | B . | SE . | B . | SE . | |
Model 1: Age | ||||||||
Age | 0.0067 | 0.0109 | 0.0087+ | 0.0048 | −0.0034 | 0.0036 | 0.0013 | 0.0039 |
R2 | 0.001 | 0.01 | 0.00 | 0.001 | ||||
Model 2: Age + Age2 | ||||||||
Age | −0.0032 | 0.0636 | −0.0072 | 0.0268 | 0.0068 | 0.0241 | −0.0028 | 0.0212 |
Age2 | 0.0001 | 0.0005 | 0.0001 | 0.0002 | −0.0001 | 0.0002 | 0.0000 | 0.0002 |
R2 | 0.002 | 0.006 | 0.001 | 0.003 | ||||
Model 3: Age + Age2 + Covariates + Commune dummies | ||||||||
Age | 0.0149 | 0.0144 | 0.0106+ | 0.0060 | 0.0006 | 0.0046 | 0.0037 | 0.0049 |
Gender | 0.4088 | 0.5015 | 0.0143 | 0.2011 | 0.2629 | 0.2094 | 0.1316 | 0.1449 |
Literacy | 0.5874+ | 0.3255 | 0.1545 | 0.1209 | 0.2616* | 0.1299 | 0.1713 | 0.1302 |
Household size | 0.0775 | 0.0593 | 0.0037 | 0.0249 | 0.0531* | 0.0191 | 0.0207 | 0.0213 |
Risk preference | 0.1610* | 0.0594 | 0.0480+ | 0.0264 | 0.0666* | 0.0226 | 0.0464+ | 0.0248 |
Patience | 0.3895 | 0.3600 | 0.2950* | 0.1350 | 0.0443 | 0.1414 | 0.0502 | 0.1319 |
R2 | 0.1 | 0.09 | 0.09 | 0.08 | ||||
N | 707 | 707 | 707 | 707 |
Note. All standard errors are clustered at the commune (experimental session) level.
*p < .05. +p < .10.
First, regarding the risk game, individuals invested on average less than half of the initial endowment, which matches previous results that suggest that most individuals are risk averse in this game (Charness & Viceisza, 2012; Charness & Villeval, 2009). Concerning age effects, there is a negative correlation between age and the amount invested in the risky option. In particular, the mean of risk taking is significantly lowest in the oldest age group in Figure 2. This pattern is more pronounced in a regression framework when controlling for a number of variables, including gender, literacy, and proxies for wealth (see Table 1). Also note that risk taking is linearly decreasing with age because the square of age is not significant in Model 2. What is the magnitude of the coefficient associated with age? Consider Model 3, a 15-year increase in age (the sample standard deviation of age) reduces the amount invested in the risky option by −0.159 (−0.0106×15), which is equivalent to about 2% of the amount given to the participant or about 4% of the sample mean of risk taking. In sum, although the effect of age is significant, it is nevertheless small.
Second, the large majority of individuals (72%) were patient in that they are willing to wait 1 week from the first option onward, but patience was not associated with age (see Table 1). We also tested a continuous version that considered the size of the delayed reward but arrived at similar null findings regarding age effects.
Third, in the dictator game, participants donated about a third of the allocated amount, thus showing similar levels of generosity as in previous studies (Engel, 2011). Concerning age, we found no robust or significant effects of age in the dictator game. Figure 3 presents the amount donated to development projects in the commune, the poor family from the own commune, and the poor family in another commune. There are some significant spikes in the age group from 61 to 70 in the total amount given, as depicted in Figure 2, and in the amount given to the commune, as shown in the top panel of Figure 3. However, there are no coherent effects of age across the three dictator games as documented in Table 2 and Figure 3, and therefore, any interpretation of the significant effect for that particular cohort may be unwarranted.
Fourth, the average trust game results match median results from the literature (Johnson & Mislin, 2011): Participants sent about 4 MAD, which represents 40% of their initial endowment, and returned on average about 6 MAD, which represents slightly more than half of the possible contribution. Regarding age effects, age was not significantly associated with the amount sent or returned. One interpretation is that the level of trust and trustworthiness is independent of age because it is trumped by reciprocity concerns. In fact, there is a strong correlation between trust and trustworthiness (r = .82).
Fifth, and finally, the majority of individuals contributed to the public goods game regardless of age. However, we observed a significant effect of age in the public goods game, whereby there is a concave association between age and public goods contribution, with middle-aged groups showing peak contributions to the public good relative to both younger and older individuals. In order to judge the size of this correlation, consider Model 3. Note that the effect of age is nonlinear (0.0153 − 2×0.0002 × age), so it is useful to consider the function before and after its inflection point. Due to concavity, the effect turns negative at 38.25 years (0.0153/[2×0.0002]). About 60% of 21-year olds will contribute to the public good, and 10 years later, at the age of 30, the individuals’ likelihood of contributing to the public good increases by about 7% (10×0.0153 − 10×2 × 0.0002×20). In comparison, between 53 years of age (the sample mean of age) and the age of 63, the chances of contribution decrease by about 6% (10×0.0153 − 10×2 × 0.0002×53). Consequently, the effect of age is relatively small relative to the overall levels of participation between 50% and 70% (see Figure 2).
In sum, we find a mixed pattern of results with linear effects of age in the risk game, an inverted U pattern for public goods game, but no systematic effects of age in the time preferences, dictator, or trust games.
Discussion
Aging is associated with changes in decision making that may have consequences for both individuals and society as a whole: Changes in nonsocial domains, such as risk-taking tendencies could have impact on investment decisions with wide economic implications. Changes in trust and prosocial behavior could also be problematic for individuals whenever they increase the potential for abuse, for example, through fraud. In developing countries in particular, household heads tend to be older and oversee multigeneration families. Consequently, age-related changes in economic decision making could have important repercussions for households as a whole.
Our study represents one of the first attempts to map out age/cohort differences in a number of economic games spanning both social and nonsocial domains. Moreover, we present results for a large, age-heterogeneous, non- Western sample from Morocco, thus taking a first step toward assessing cross-cultural differences in age-related patterns in a variety of economic decisions, involving risk, time preferences, altruism, trust, and public goods provision.
The pattern of results we find is mixed, with evidence for small albeit significant effect sizes of age/cohort in some but not all economic games. In the nonsocial domain, we found evidence of a negative albeit small correlation between age and risk taking, a pattern that holds even when controlling for a number of variables, including gender, literacy, and proxies for wealth. The decrease in risk taking between younger and old-age cohorts matches predictions from life history theories (Wilson & Daly, 1997), as well as results based on self-report data, suggesting that aging is associated with decreased risk taking (Dohmen et al., 2011; Roalf et al., 2011), as well as choice data (Gong & Yang, 2012), at least when considering situations of avoidable risk or uncertainty (Mather et al., 2012). In turn, we find no systematic age-related pattern for the time preferences measure, and thus, our results match others showing no significant differences in delay discounting with advanced age (Chao et al., 2009; Green et al., 1996; Roalf et al., 2011) although other studies do indicate increased patience with advanced age (Löckenhoff et al., 2011).
Regarding the age/cohort patterns between economic games involving social interactions, we also find small effects of age in some but not all games, with some conflicting results regarding age patterns in the different measures of social decision making. First, we found no significant effects of age in the dictator game. These results contrast with those of Roalf and et al. (2011) and Engel (2011) that suggest that aging is associated with increased prosocial behavior in the dictator game. Second, we found no significant age trends in the trust game. These results match previous findings (Sutter & Kocher, 2007), suggesting that trust and trustworthiness are rather stable in adult age groups. Third, and finally, we found a concave relation between age and the likelihood of contributing to a public good in the public goods game, suggesting that feelings of obligation toward the group increase between younger and middle-aged cohorts but decrease for older cohorts. All in all, these results suggest that there is little convergence regarding the effects of age on social economic games and that more work is needed to understand the causes underlying the social preferences of particular groups, cohorts, and populations.
Although some studies find consistent effects across measures of economic behavior (Anderson & Mellor, 2008; Dohmen et al., 2011), there is a debate about the ability of economic measures to capture underlying preferences stemming from the lack of empirical association between different games (Berg et al., 2005) and the variability of results as a function of procedural variation in games, such as the dictator game (Bardsley, 2007; Dana, Weber, & Kuang, 2006; List, 2007). Our results showed that behavior in the different games was weakly correlated and that there were no homogeneous effects of aging across measures (see Table 3). In other words, our results can be interpreted as supporting the idea that there is significant specificity of processes and preferences underlying behavior in different economic games spanning social and nonsocial domains.
Our approach has a number of strengths, including the use of incentive-compatible, real-stakes economic games to investigate several aspects of decision making spanning both social and nonsocial domains, something that is not commonly done in the aging and decision-making literature (Mata et al., 2011). One advantage of examining several games is that the pattern of results can help exclude potential explanations related to the role of task complexity on economic decision making: For example, we found age differences in rather simple tasks, such as the risk game, but also in the more complex public goods game, whereas we found no age differences in a more complex task involving several phases, the trust game. Finally, we are able to control for both a number of participant characteristics, as well as commune and experimental-session unobservables, which allows us to rule out that aggregate demographic, socioeconomic structure, geography, or local norms bias the estimated correlation between age and economic behavior.
There are, nevertheless, a number of limitations in our work that deserve consideration. First, the major limitation in our approach is an exclusive reliance on cross-sectional data. In particular, we cannot distinguish the contributions of aging and cohort to the identified age patterns. The inhabitants of Morocco experienced significant social and economic changes in the past decades, and, consequently, cohort differences are likely a relevant factor in shaping both individual and social decision making. Only by following individuals longitudinally over time will one be able to assess with confidence the role of aging on economic decision making. Second, our task set was constrained by practical considerations (i.e., time and comprehensibility) that limit our analysis. For example, the risk game considered only the gain domain. Future studies that evaluate the loss domain or employ mixed gambles could be informative regarding age differences in the relative weight given to positive and negative outcomes to test motivational theories of aging (Carstensen, 2006; Depping & Freund, 2011; Mata & Hertwig, 2011). Third, our analyses did not consider the role of previous social interactions or individual characteristics on participants’ behavior in the economic games. It would be interesting to conduct a more thorough assessment of individuals’ characteristics (e.g., personality and social status) so as to examine the link between these variables and age differences in social economic behavior. Finally, our sample consisted of mostly men with a rural background, and therefore, we cannot easily account for possible gender or background (rural vs. urban) differences in the patterns of age differences that may play a role in economic decision making (Cross, Copping, & Campbell, 2011; Henrich et al., 2005). Note, however, that our sample represents fairly accurately the social structure of Morocco, where economic institutions and households remain heavily patriarchal.
To conclude, our results warn against simplifications of the link between aging and decision making. In particular, our findings seem to support the idea that both task and culture matter and thus that there may be no biologically driven, universal patterns that can be observed across economic games regarding age differences in economic decision making. For example, our results do not seem to suggest that aging is universally associated with prosocial behavior and emphasize the potential for a variety of age/cohort differences in both social and nonsocial domains. Overall, our work suggests that cross-cultural research is needed to uncover both the specificity and generality of aging trends in economic decisions around the world.
. | 2 . | 3 . | 4 . | 5 . | 6 . | 7 . |
---|---|---|---|---|---|---|
1. Risk game | .04 | .10 | .11 | .14 | −.10 | −.07 |
2. Time preferences game | .04 | .16 | .04 | −.03 | .02 | |
3. Dictator game (total) | .38 | .09 | .10 | .03 | ||
4. Trust game (sender) | .02 | −.04 | ||||
5. Trust game (receiver) | .004 | .01 | ||||
6. Public goods game | −.10 | |||||
7. Age |
. | 2 . | 3 . | 4 . | 5 . | 6 . | 7 . |
---|---|---|---|---|---|---|
1. Risk game | .04 | .10 | .11 | .14 | −.10 | −.07 |
2. Time preferences game | .04 | .16 | .04 | −.03 | .02 | |
3. Dictator game (total) | .38 | .09 | .10 | .03 | ||
4. Trust game (sender) | .02 | −.04 | ||||
5. Trust game (receiver) | .004 | .01 | ||||
6. Public goods game | −.10 | |||||
7. Age |
Note. For N > 700, correlations above .07 are significant at level .05, correlations of .1 are significant at the .01 level, and correlations above .12 are significant at the .001 level.
. | 2 . | 3 . | 4 . | 5 . | 6 . | 7 . |
---|---|---|---|---|---|---|
1. Risk game | .04 | .10 | .11 | .14 | −.10 | −.07 |
2. Time preferences game | .04 | .16 | .04 | −.03 | .02 | |
3. Dictator game (total) | .38 | .09 | .10 | .03 | ||
4. Trust game (sender) | .02 | −.04 | ||||
5. Trust game (receiver) | .004 | .01 | ||||
6. Public goods game | −.10 | |||||
7. Age |
. | 2 . | 3 . | 4 . | 5 . | 6 . | 7 . |
---|---|---|---|---|---|---|
1. Risk game | .04 | .10 | .11 | .14 | −.10 | −.07 |
2. Time preferences game | .04 | .16 | .04 | −.03 | .02 | |
3. Dictator game (total) | .38 | .09 | .10 | .03 | ||
4. Trust game (sender) | .02 | −.04 | ||||
5. Trust game (receiver) | .004 | .01 | ||||
6. Public goods game | −.10 | |||||
7. Age |
Note. For N > 700, correlations above .07 are significant at level .05, correlations of .1 are significant at the .01 level, and correlations above .12 are significant at the .001 level.
Funding
The data collection was financed by the Swiss National Science Foundation (140745) under the ongoing grant «Development Aid and Social Dynamics« hosted at the Graduate Institute’s Centre on Conflict, Development and Peacebuilding (CCDP) and lead by Jean-Louis Arcand, as well as coordinated by M. Rieger.
Acknowledgments
We would like to thank the project leader, Jean-Louis Arcand, for granting permission to use the data for this paper. The project is run in collaboration with the Observatoire National du Développement Humain (ONDH), Rabat. The authors thank Sandra Reimann and Oliver Jütersonke at the CCDP and Mohamed Benkassmi and Youssef Bouzrour at the ONDH for organizing the data collection, Tu Chi Nguyen for field supervision, and Thorsten Pachur, Ralph Hertwig, and Anika Josef for comments on earlier versions of the manuscript. The time, trust and public goods games protocols and overall session procedure were adopted with permission from Michael Gilligan. The risk game was suggested by Gary Charness and the protocol is adopted with permission from Charness and Gneezy (2010). All views and remaining errors are ours.
References
Author notes
Decision Editor: Myra Fernandes, PhD