No effect of birth order on adult risk taking

Does birth order shape people’s personalities? In his family dynamics model, Sulloway (1, 2) proposed that the family is a set of niches that provide siblings with distinct experiences and viewpoints. Siblings compete for parents’ limited resources, such as attention and acceptance. Firstborns, having experienced their parents’ undivided attention and care, and being stronger and more intellectually developed than their younger siblings, occupy a dominant position. They thus tend to safeguard their status by developing conservative values and attitudes that help them defend the status quo. Later-borns try to find a valued family niche that is not already occupied by an older sibling. To this end, they will explore—literally, as Alexander von Humboldt did, and metaphorically—and, in the process, rebel against the status quo. According to Sulloway, these competitive dynamics during childhood influence the development of siblings’ distinct personalities—including different propensities to take risks—as a function of birth rank (3). Competing with firstborns who are older and thus stronger and more intelligent, and who may receive more parental resources by occupying the niche of a surrogate parent, later-borns have no choice but to take greater risks to differentiate themselves from their older siblings and to attract their parents’ attention. Thus, according to Sulloway’s childhood niche hypothesis, growing up subject to this dynamic makes later-borns develop a more pronounced propensity to take risks than firstborns because “risk taking is a useful strategy in the quest to find an unoccupied niche” (ref. 1, p. 112).

Although early research on birth-order effects on personality supported the family dynamics model (e.g., refs. 4–6), more recent research has challenged these findings, arguing that early personality studies were underpowered, used nonrepresentative samples, and relied on between- rather than within-family comparisons, and that personality characteristics were often judged by a single observer member of the family. A recent study using data from three large national panel studies, including the German Socio-Economic Panel (SOEP), controlled for these methodological problems and found that birth order had no substantive effect on personality traits and only a negligible effect on intelligence (7). In parallel, the largest yet study of birth-order effects, which used a sample of 377,000 US high school students, found the same results (8), which were recently corroborated in a non-Western sample (9). For its authors, Damian and Roberts (10), Rohrer et al.’s (7) work has settled the debate: Birth order has no effect on personality.

The evidence concerning birth-order effects on risk taking is less clearly tilted against the family dynamics model, however. Did the risk-taking propensities of Wilhelm and Alexander von Humboldt differ because of the order in which they were born? Researchers examining the sociodemographic factors associated with childhood accidents have stumbled on findings indicating that birth order appears to play a role. Later-born children are 4 SDs more likely to be involved in near-drowning accidents (11), more likely to be involved in household accidents (12), and more likely to suffer from accidental injury (13). This pattern is consistent with Sulloway’s model, but it is also consistent with the possibility that accidents are more likely to occur in households with more siblings, and in households with older siblings whose behaviors may become a reference point for younger children. Later-born children may be more likely to have accidents not because they seek more risks than their older siblings, but because they observe, imitate, and model their behaviors, possibly before acquiring the necessary cognitive and motor abilities.

Sulloway and Zweigenhaft (3) examined the possible relationship between birth order and choice of risky activities. Specifically, they conducted a metaanalysis of research that reported siblings’ choice of sports, with samples covering children, adolescents, and adults. They found that the mean odds ratio was 1.5 to 1 in favor of greater later-born than firstborn participation in high-risk sports (i.e., contact and injury-prone sports, as rated by experts), such as rugby or football. They also examined statistics on 700 brothers who played major league baseball in the United States. Later-borns were 10 times more likely than firstborns to attempt the high-risk activity of base stealing. These studies support Sulloway’s hypothesis (ref. 1, p. 433): Later-borns select more dangerous sports than firstborns, and, when they choose the same sport, they take more risks when playing that sport. In the same vein, one small-scale self-report study of financial risk tolerance (n = 368) found that firstborns have a lower risk tolerance than later-borns, and are less likely to allocate a majority of their assets to stocks (14). Another small-scale study (n = 200), however, found no effect of birth order on self-reported risk behaviors [e.g., birth control, unprotected sex (15)].

Against this background, Rohrer et al. (16) used cross-sectional data from the 2013 wave of the SOEP to explore whether birth order affects the self-reported propensity to take risks (as well as several personality traits beyond the Big Five traits). Using specification curve analysis (16, 17), an exhaustive econometric approach, they observed, in contrast to previous findings on the selection of sports and risky play, that birth order had no substantive effect on risk taking, or indeed on the various personality dimensions examined. Recently, these results were supported in an Indonesian sample (9).

To summarize, although there appears to be an emerging consensus that birth order has no substantive effects on the formation of personality traits, its influence on risk-taking propensity is not yet settled. Accident statistics suggest that later-borns may take more risks than firstborns, but this finding is not necessarily a result of birth-order-specific risk preferences. Sports statistics indicate that later-born children choose more risky sports and take more risks in baseball than firstborns. However, such a predisposition was not corroborated by Rohrer et al. (16), whose specification curve analysis found no birth-order effect on risk taking in a large sample of adults. An important difference between the approaches of Sulloway (4) and Rohrer et al. (16) is that the former analyzed behavioral data from sports choice and play, whereas the latter examined self-reported propensity to take risks. In short, the effect of birth order on risk taking is still an open question, and the choice of the dependent measure may be crucial in answering the question of whether later-borns are more inclined to take risks: To date, birth-order effects have emerged only in behavioral data.

We conduct the most thorough examination of the relation between birth order and risk preference to date. To this end, we examine three sources of data, with multiple dependent measures: self-reports, behavioral measures, and risky life decisions outside the laboratory. To corroborate Rohrer et al.’s (16) observation, we begin by updating their analysis using the most recent wave of SOEP data, and additional estimation methods. These data consist in self-reports of risk-taking propensity. Second, we use data collected in the Basel–Berlin Risk Study (BBRS) (18). This study administered a battery of 39 state-of-the-art measures of people’s risk preferences, including incentivized behavioral tasks, such as decisions between monetary gambles and the Balloon Analogue Risk Task. These tasks are crucial to the extent that birth-order effects may be more likely to emerge in behavioral data. The BBRS also contains a set of self-reported “propensity measures,” including the SOEP items on risk-taking propensity—allowing us to test for consistency between the BBRS and SOEP data—and a set of self-reported “frequency measures” quantifying respondents’ engagement in risky activities (e.g., a test for alcohol use disorders; questions on the frequency of risky behaviors in the past month). Moreover, the BBRS is the first study to run extensive psychometric analyses across a wide range of risk-taking measures. The results suggested that risk preference comprises both a general factor with substantial temporal stability (R; akin to g, the general factor of intelligence) and several (domain-) specific factors (18). These psychometric factors may reflect the most comprehensive assessments of people’s risk preferences to date, and thus constitute ideal measures to examine the potential role of birth-order effects. Finally, we study another source of behavioral data—behaviors outside the laboratory—to provide another test of whether birth-order effects manifest in behavioral data. To this end, we examine samples of explorers and revolutionaries, two particularly risky life choices. In sum, our contribution to the understanding of birth-order effects on risk taking rests on abundant data and diverse dependent measures.

We have used a three-pronged approach to investigate birth-order effects on risk taking. First, we examined the propensity to take risks as measured by a self-report item in the German SOEP, one of the largest and most comprehensive household surveys. Using specification curve analysis, we found no significant effects of birth order in 96.7% of the specifications. Overall, our results corroborate those by Rohrer et al. (16) with more recent data and an alternative estimation method, namely, Bayesian regression.

One limitation discussed by Rohrer et al. (16) is that self-report measures may not capture revealed risk-taking behavior. Thus, to complement the SOEP’s self-report approach, we drew on data from one of the most exhaustive attempts to measure risk-taking propensity, the BBRS. We first confirmed that responses to the general risk item administered in both the SOEP and the BBRS were aligned. Indeed, birth order had no effect on self-reported (general) risk taking in either sample. We then went on to explore birth-order effects on the variety of behavioral measures examined in the BBRS, as well as other self-reported propensity and frequency measures. Of the 39 implemented measures, only three showed a credible relationship between birth order and risk taking. Whereas previous research has found birth-order effects on risk taking only in behavioral data, none of our behavioral measures showed any credible birth-order effect on risk taking. Moreover, one of the 39 measures (SOEPdri) suggested that later-borns take less risk than firstborns, and two measures (Dhea and SStas) suggested the opposite. Finally, of the eight psychometric factors identified by Frey et al. (18), F3 (capturing recreational risk taking) and F7 (summarizing mostly the lotteries among the behavioral risk-taking measures) showed a small but credible positive effect, but the general factor R showed no credible relationship between being a later-born and taking more risk.

We then considered the possibility that birth-order differences in risk taking are not reflected in survey responses and laboratory studies. We thus examined another source of behavioral data: the choice of becoming an explorer or a revolutionary. The birth rank of explorers and revolutionaries was precisely what one would expect to find in families where birth order has no effect on these life choices.

According to Sulloway’s childhood niche hypothesis, family dynamics shape personality during early development and thus influence risk taking in adulthood. If this assumption holds, it should be possible to observe birth-order effects when risk-taking propensity is measured in adults, using standard methods from psychology and economics. Our evidence strongly indicates that family dynamics do not produce stable differences in adult risk-taking propensities.

One possible explanation for the discrepancy between the present results and findings on the choice of risky sports (3) is that past studies of birth-order effects may have been less stringent in their methodological requirements—and a metaanalysis can only be as good as the studies on which it is based. For illustration, one article in Sulloway and Zweigenhaft’s metaanalysis contributed 4 of the 24 samples analyzed, as well as the largest individual sample (38). In this study, the ratings of a sample of 35 students determined the three most dangerous sports played at Columbia University (football, soccer, and rugby). This classification meant that wrestling was treated as a nondangerous sport. Sulloway and Zweigenhaft (ref. 3, p. 404), in contrast, listed it as a dangerous sport. Although we cannot judge to what extent such selection criteria contributed to discrepancies between the current and past findings, they cannot be excluded as a potential source of ambiguity.

There is another possibility, however. As Sulloway suggests, it may well be that, because later-born children are at a disadvantage when competing with firstborns for parental resources, it pays for them to take risks (ref. 5, p. 107). However, the reach of these dynamics may be restricted to the child and, perhaps, adolescent, and may not necessarily impact personality and behavior in temporally distant adult environments. For example, according to risk-sensitivity theory (39, 40), individuals may favor risky options under conditions of high need, with “need” referring to the disparity between an individual’s current state and required target state. When low-risk options fail to provide sufficient benefits, such as enough food to prevent starvation, risk taking will constitute an adaptive strategy to increase the chance of securing outcomes that might otherwise be beyond reach. [This view is compatible with recent findings suggesting that risk preference shares the psychometric structure of major psychological traits, including a general factor that remains relatively stable across time—which does not preclude the possibility that risk taking may vary substantially across specific states (e.g., in response to specific situations or contexts) (18, 41).] This account predicts that risks will be taken by individuals who are competitively disadvantaged—say, weaker, poorer, and less skilled—and thus cannot compete if they stick to low-risk strategies.

This notion suggests that later-borns may take risks during childhood because it is during this period that their competitive handicap is greatest. Later in life, they may be just as likely or unlikely as their older siblings to take risks. In other words, a person can, depending on their state, behave in a risk-seeking manner without being a dispositional daredevil. This view suggests an explanation for the evidence that Sulloway and Zweigenhaft (3) found for birth-order effects on sports choice. Importantly, these effects were strongly moderated by age, with children showing larger effects than adolescents and adults. In other words, the birth-order effects observed were largely driven by children who were still living at home and subject to family dynamics, and thus in a state in which it pays to take risks in activities (e.g., sports, arts, music) in the hope of attracting parental attention and resources.

The idea that birth order influences personality—and risk taking in particular—is powerful. We searched for evidence in survey, experimental, and real-world data, analyzing self-reports, incentivized risky decisions, and consequential life choices, but the findings point unanimously in the same direction: There are no birth-order effects on adult risk taking. To understand why Alexander von Humboldt, and not Wilhelm, climbed Mount Chimborazo, we need to look beyond birth order and family dynamics.

We thank Susannah Goss for editing the manuscript. We also thank Jill de Ron, Max Rank, Amanda Barreiro, and Iro Eleftheriadou for research assistance.