Common, nonsexual masochistic preferences are positively associated with antisocial personality traits. Christina Sagioglou, Tobias Greitemeyer. Journal of Personality, November 16 2019. https://doi.org/10.1111/jopy.12526
Abstract
Objective: Based on prior research linking masochism and antisocial behavior to sensation seeking, we hypothesized that masochistic and antisocial preferences are positively correlated. Besides sensation seeking, we tested whether disgust sensitivity (due to its inhibitory function) and shared social values (e.g., stimulation) accounted for the masochistic‐antisocial link. We additionally examined the link in relation to broad personality factors.
Method: Six online and laboratory studies (N = 2,999) with U.S. American and European samples.
Results: We consistently found positive correlations between masochistic enjoyment (e.g., enjoying the burn of spicy food, disgusting jokes, pounding heart, painful massage) and antisocial traits such as subclinical psychopathy, everyday sadism, and low Honesty‐Humility. We observed behavioral correlations in that experienced pleasure of a painful event was positively related to causing another person to feel pain. Shared sensation seeking, low disgust sensitivity, and endorsement of social values such as social power, hedonism, and a stimulating life partially accounted for the masochistic‐antisocial link.
Conclusion: The extent to which a person enjoys threatening stimuli on the self is reliably related to how much a person enjoys and evokes others’ suffering. Future research could explore the common core that underlies common masochistic and antisocial preferences beyond the mediators tested here.
Discussion
In six studies with large samples from different countries and populations, we provided consistent evidence for a positive association between the enjoyment of negative sensations on the self and a tendency to enjoy and cause suffering in other people. We repeatedly confirmed the masochistic-antisocial link with regard to the Dark Tetrad constellation and the traits related to antisocial behavior of the HEXACO and Big Five models for broad personality factors. Largest correlations were observed for everyday sadism and psychopathy, but masochistic preferences were also consistently correlated with Machiavellianism, narcissism, and Honesty-Humility. We further extensively validated trait masochism with behavioral choice and preferences measures, lending further validity to the trait construct and its empirical distinctiveness. At behavioral level, we see that masochistic expressions such as bitter taste preferences and aversive movie preferences, but not enjoyment of pain, are positively correlated with antisocial traits. Yet, these behavioral preferences are best predicted by trait masochism. Masochistic preference is thus not merely a manifestation of antisocial tendencies turned against the self. It is empirically related to but distinguishable from both antisocial traits and broad personality factors. Neither trait constellation (i.e., the Dark Tetrad or HEXACO) sufficiently captures such self-directed aversive affinity. Examining masochistic and antisocial behavior within the pain domain showed that irrespective of pain sensitivity, more masochistic pleasure is associated with more antisocial behavior.
In addition, we identified three constructs that account for some of the overlap between masochistic and antisocial preferences. First, we found that a need for arousal and stimulation partially manifests in self-exposure to negative stimuli, but also in antisocial behavior towards other people. Second, two studies confirmed that being sensitive to disgust-evoking stimuli keeps individuals from exposing themselves to aversive experiences and from behaving hostile towards other people. More disgust-sensitive individuals also reported less pleasure when experiencing ice-water induced pain. Third, pursuing a challenging, daring, and exciting life, gratification of desires, self-indulgence, social power, and authority are shared motivational characteristics of masochistic and antisocial traits. Yet, we see that despite this varied overlap, the common core persists when controlling for these constructs. Methodologically, it is important to note that in four of the six studies, masochism was assessed differently (by typing a number into a textbox) than were the dark traits (on Likert-type scales). It is thus unlikely that the correlations were inflated by response-biases.
Limitations and future directions
We measured sensation seeking via a combination of differently weighted HEXACO facet scores (de Vries et al., 2009) and found that it partially accounted for the masochistic-antisocial link. Using the original sensation seeking scale (Zuckerman, 1979) as did Rozin and colleagues (2013) may shed more light on its role in the masochistic-antisocial link. The original scale contains the facets thrill seeking, experience seeking, disinhibition, and boredom susceptibility, of which disinhibition revealed a moderate correlation with both benign masochism (r = .35; Rozin et al., 2013) and Honesty-Humility (r = -.33 to -.41; de Vries et al., 2009), whereas boredom susceptibility was uncorrelated with benign masochism but reliably negatively correlated with Honesty-Humility. Focusing on the nuanced similarities and differences of masochistic and antisocial tendencies in sensation seeking may reveal more about when one person does and does not unite both self-directed and other-directed harmful tendencies within herself.
A further limitation is that there is high empirical overlap of everyday sadism assessed with the ASP and psychopathy measured with the SD3, which applies to four of our studies. In fact, correlation coefficients range from .57 to as high as .85. Accordingly, masochism reveals similarly large and reliable correlations with both sadism and psychopathy. Yet, in trying to understand the link between masochistic and antisocial tendencies, it is crucial to empirically differentiate between different antisocial constructs. Future research may thus employ improved measures of psychopathic and sadistic tendencies that have better discriminatory power. This would more clearly reveal whether masochism is more strongly linked to psychopathic or sadistic tendencies or similarly to both.
One approach to better understand the connection between masochistic and antisocial tendencies may be by exploring their psychological function. Most notably, Baumeister (1988) offered a social psychological theoretical framework for understanding the paradox of sexual masochism. He argued that sexual masochism is a means by which people escape self-awareness, just as they do through alcohol consumption or other recreational activities (Baumeister, 1991).
Although self-awareness can lead to positive states such as when individuals feel proud of a personal achievement, it often confronts individuals with mistakes and deviance from their ideal self. Indeed, escaping self-awareness was argued to be a common human desire (Wicklund, 1975). The present and past research on masochism (Rozin et al., 2013) lend empirical support to Baumeister.s theory in that people who enjoy one form of aversive activity are more likely to prefer other such activities. Yet, whether and how a desire to escape self-awareness is also related to increases in antisocial behavior remains an open empirical question.
Put simply, both masochistic and antisocial preferences are characterized by an affinity towards aversive states. Future research could investigate the extent of positive evaluative reactions that they are associated with. Implicit measures that detect underlying, possibly unware preferences with reaction times may shed more light on a possible contra-hedonic conditioning of aversive stimuli. Indeed, Rozin et al. (2013) argued that masochistic pleasures are essentially hedonic reversals in that we grow up disliking most of the stimuli (e.g., bitterness, pain) and learn to like them throughout our lives. Possibly, dark traits are characterized by a parallel form of moral-evaluative reversals. The common core of the traits. overlap could become most evident at such an implicit level.
There are a variety of activities that are potentially related to the present construct of benign masochism. Whereas we focused on nonsexual experiences, the original, prototypical form of masochism refers to gaining sexual pleasure from pain and humiliation. It is conceivable that they are positively related, because both provide a way to escape self-awareness. At the same time, sexual masochism is less prevalent and more confined than, for example, enjoying sad art or eating spicy food (Baumeister, 1989; Spence, 2018). Possibly, many sexual masochists also gain nonsexual pleasure from intense negative stimuli but not vice versa. Moreover, the present experiences are essentially harmless, simulating threat or danger while involving a very low actual risk level. Yet, particularly in the fear domain there are other activities that have a higher inherent level of danger. For example, engagement in arousal-inducing extreme sports such as wingsuit flying or mountaineering carries a realistic death risk through participation. Whether preferences for activities involving actual danger are an extreme expression of the benign masochism studied here, or whether they are a qualitatively distinct phenomenon remains for future research to discover. Nevertheless, due to their link to sensation seeking (Kerr, 1991), it seems likely that such preferences are also positively associated with antisocial tendencies.
Wednesday, November 20, 2019
The Interplay Between Cognitive Intelligence, Ability Emotional Intelligence, and Religiosity
The Interplay Between Cognitive Intelligence, Ability Emotional Intelligence, and Religiosity. Paweł Łowicki, Marcin Zajenkowski, Dimitri van der Linden. Journal of Religion and Health, November 20 2019. https://link.springer.com/article/10.1007/s10943-019-00953-0
Abstract: The negative association between cognitive intelligence (CI) and religiosity has been widely studied and is now well documented. In contrast, the role of emotional intelligence (EI) in this context has been poorly investigated thus far. Some available data indicate that EI, unlike CI, correlates positively with religiosity. To date, however, no study has explored the relationship between religiosity and both intelligences simultaneously. In current studies (Ns = 301 and 200), we examined the interplay between all three constructs. The results showed that CI was positively correlated with ability EI and negatively with some measures of religiosity. EI, on the other hand, revealed no direct, significant relationship with religiosity. However, when combined into a single regression model with CI, EI became a significant positive predictor of religiosity. Moreover, Study 2 revealed that the link between EI and religiosity was mediated by empathy. Interestingly, we also found a reciprocal suppression between CI and EI, since both predictors increased their influence on religiosity when analyzed together. Although the suppression was present in both studies, it was observed for different religiosity measures in each case, indicating that this effect is probably dependent on various factors, such as sample structure or type of religiosity.
Keywords: Religiosity Belief Cognitive intelligence Emotional intelligence Empathy
General Discussion
In the present research, we examined the relationship between ability EI, cognitive intelligence, and religiosity. Hypothesis 1 stating that cognitive intelligence
would be negatively related to religiosity was confrmed for some of the religious
scales (e.g., centrality of religiosity in Study 1 and religious fundamentalism in
Study 2). It is worth noting that the associations between religious fundamentalism
and intelligence tests were more pronounced than correlations observed for other
religiosity measures—both in our research and in previous studies (e.g., Lewis et al.
2011). Therefore, it seems that this particular type of religiosity is the most difcult to reconcile with high cognitive abilities. Interestingly, it is usually believed that
intelligence infuences (or at least precedes) religiosity (Zuckerman et al. 2013). For
instance, some authors argue that intelligent individuals are less willing to engage
in more instinctive, evolved types of behavior (Dutton and Van der Linden 2017;
Kanazawa 2012). In contrast, they tend to favor some counter-instinctive behaviors—e.g., they undertake nocturnal activities or have a taste for complex instrumental music. In those evolutionary accounts, religion is also considered an evolved
and adaptive pattern that enhances a sense of community and provides stress relief.
Thus, assuming that religion is a relatively old universal instinct, intelligent individuals are more likely to be atheist (Dutton and Van der Linden 2017). However,
in the case of religious fundamentalism, it has been also argued that this religious
feature might actually exert an efect on intelligence by opposing secular knowledge
and education (Sherkat 2010, 2011). Thus, the direction of the causal relationship
between these two constructs remains unknown and should be established using longitudinal research.
Hypothesis 2 concerning the positive link between EI and religiosity was not supported since none of the positive zero-order correlations between EI and religiosity
dimensions were statistically signifcant (with the exception of public practice subdimension in Study 1). Additionally, we found that EI was positively associated with
Cattell’s score in Study 1 and general intelligence factor in Study 2, which supported
Hypothesis 3.
Next, Hypothesis 4 received a partial support. Specifcally, in Study 1 EI became
a signifcant and positive predictor of centrality of religiosity when analyzed jointly
in a single regression model with cognitive intelligence. In Study 2, we did not
directly replicate this efect on the exact same measure, but we found a similar suppression efect for the scale measuring belief in God/Higher Power. Furthermore, the
suppression, albeit weak, did occur also on pooled data from Studies 1 and 2. Thus,
from this pattern of fndings, it seems reasonable to conclude that cognitive intelligence indeed acted as a suppressor of the EI–religion relationship. This fnding suggests that ability EI constitutes a blend of rational and intuitive mechanisms (both
inversely related to religiosity) and may explain, to a certain extent, why zero-order
correlations between religiosity and EI are not signifcant in some studies (including
current ones). It is possible that removing the variance related to abstract reasoning
reveals the intuitive, empathetic part of EI (see Mayer and Geher 1996) and this part
shows a more pronounced association with religiosity. In fact, this notion was also
supported in Study 2 which showed that empathy mediated the EI–belief link.
Finally, the current study revealed also some interesting fndings regarding cognitive intelligence. First, we found evidence for cooperative suppression: In all cases
where EI became more strongly associated with religiosity, the cognitive ability
(negative) relation to religiosity was also more pronounced in regression models. It
seems then that removing the EI aspect from cognitive ability increased its negative
association with religious belief. It is possible that this remaining part of cognitive
ability refects “cool” and rational reason, not infuenced by afect or empathy. Such
an analytic cognitive approach may signifcantly hamper beliefs in religious tenets
since they often defy logic or violate the laws of nature (cf. Pennycook et al. 2012).
The current research has several limitations. First, the present studies had a crosssectional character, which did not enable us to establish causal relationships between
investigated variables. Although in the research literature there are some suggestions
regarding the direction of the cognitive intelligence–religiosity link (Zuckerman
et al. 2013), the causality is not yet clear in the case of the relationship between
EI and religion. Clearly, then, more work in this context—including longitudinal
research—is needed to deepen our understanding of the interplay between religiosity, EI, and cognitive intelligence. Moreover, although we extended the scope of
previous research by examining the potential mediating role of emotional empathy,
it might be desirable to also examine an impact of other variables that have been
found to have signifcance for both intelligences and religion, such as time perspectives (see Łowicki et al. 2018; Stolarski et al. 2011; Zajenkowski et al. 2016). Next,
it should be emphasized that both current samples were mostly composed of Roman
Catholics and included Polish participants only. To increase the generalizability of
the obtained results, future work should examine other denominations, preferably
in non-Western societies. Finally, it needs to be acknowledged that we failed to replicate the main fnding using the exact same measure of religiosity across the two
studies. Instead, we found the suppression efect on diferent instruments measuring religiosity. As suggested above, this might be due to various factors such as the
weak magnitude of the observed efects or sample specifcity (cf. Webser and Dufy
2016). Nevertheless, our research provides some observations that might be helpful in future studies exploring the interplay between cognitive and emotional intelligence and religiosity. First, one needs to carefully consider the structure and size
of the sample. It seems that the efects of both intelligences on religiosity are rather
small, and therefore an adequate statistical power is required to detect them. Second,
taking into consideration some signifcant diferences between measures of religiosity, various aspects of religious phenomena (e.g., belief in God, fundamentalism,
interest in religion) should be included in the studies on cognitive and emotional
intelligence.
Abstract: The negative association between cognitive intelligence (CI) and religiosity has been widely studied and is now well documented. In contrast, the role of emotional intelligence (EI) in this context has been poorly investigated thus far. Some available data indicate that EI, unlike CI, correlates positively with religiosity. To date, however, no study has explored the relationship between religiosity and both intelligences simultaneously. In current studies (Ns = 301 and 200), we examined the interplay between all three constructs. The results showed that CI was positively correlated with ability EI and negatively with some measures of religiosity. EI, on the other hand, revealed no direct, significant relationship with religiosity. However, when combined into a single regression model with CI, EI became a significant positive predictor of religiosity. Moreover, Study 2 revealed that the link between EI and religiosity was mediated by empathy. Interestingly, we also found a reciprocal suppression between CI and EI, since both predictors increased their influence on religiosity when analyzed together. Although the suppression was present in both studies, it was observed for different religiosity measures in each case, indicating that this effect is probably dependent on various factors, such as sample structure or type of religiosity.
Keywords: Religiosity Belief Cognitive intelligence Emotional intelligence Empathy
General Discussion
In the present research, we examined the relationship between ability EI, cognitive intelligence, and religiosity. Hypothesis 1 stating that cognitive intelligence
would be negatively related to religiosity was confrmed for some of the religious
scales (e.g., centrality of religiosity in Study 1 and religious fundamentalism in
Study 2). It is worth noting that the associations between religious fundamentalism
and intelligence tests were more pronounced than correlations observed for other
religiosity measures—both in our research and in previous studies (e.g., Lewis et al.
2011). Therefore, it seems that this particular type of religiosity is the most difcult to reconcile with high cognitive abilities. Interestingly, it is usually believed that
intelligence infuences (or at least precedes) religiosity (Zuckerman et al. 2013). For
instance, some authors argue that intelligent individuals are less willing to engage
in more instinctive, evolved types of behavior (Dutton and Van der Linden 2017;
Kanazawa 2012). In contrast, they tend to favor some counter-instinctive behaviors—e.g., they undertake nocturnal activities or have a taste for complex instrumental music. In those evolutionary accounts, religion is also considered an evolved
and adaptive pattern that enhances a sense of community and provides stress relief.
Thus, assuming that religion is a relatively old universal instinct, intelligent individuals are more likely to be atheist (Dutton and Van der Linden 2017). However,
in the case of religious fundamentalism, it has been also argued that this religious
feature might actually exert an efect on intelligence by opposing secular knowledge
and education (Sherkat 2010, 2011). Thus, the direction of the causal relationship
between these two constructs remains unknown and should be established using longitudinal research.
Hypothesis 2 concerning the positive link between EI and religiosity was not supported since none of the positive zero-order correlations between EI and religiosity
dimensions were statistically signifcant (with the exception of public practice subdimension in Study 1). Additionally, we found that EI was positively associated with
Cattell’s score in Study 1 and general intelligence factor in Study 2, which supported
Hypothesis 3.
Next, Hypothesis 4 received a partial support. Specifcally, in Study 1 EI became
a signifcant and positive predictor of centrality of religiosity when analyzed jointly
in a single regression model with cognitive intelligence. In Study 2, we did not
directly replicate this efect on the exact same measure, but we found a similar suppression efect for the scale measuring belief in God/Higher Power. Furthermore, the
suppression, albeit weak, did occur also on pooled data from Studies 1 and 2. Thus,
from this pattern of fndings, it seems reasonable to conclude that cognitive intelligence indeed acted as a suppressor of the EI–religion relationship. This fnding suggests that ability EI constitutes a blend of rational and intuitive mechanisms (both
inversely related to religiosity) and may explain, to a certain extent, why zero-order
correlations between religiosity and EI are not signifcant in some studies (including
current ones). It is possible that removing the variance related to abstract reasoning
reveals the intuitive, empathetic part of EI (see Mayer and Geher 1996) and this part
shows a more pronounced association with religiosity. In fact, this notion was also
supported in Study 2 which showed that empathy mediated the EI–belief link.
Finally, the current study revealed also some interesting fndings regarding cognitive intelligence. First, we found evidence for cooperative suppression: In all cases
where EI became more strongly associated with religiosity, the cognitive ability
(negative) relation to religiosity was also more pronounced in regression models. It
seems then that removing the EI aspect from cognitive ability increased its negative
association with religious belief. It is possible that this remaining part of cognitive
ability refects “cool” and rational reason, not infuenced by afect or empathy. Such
an analytic cognitive approach may signifcantly hamper beliefs in religious tenets
since they often defy logic or violate the laws of nature (cf. Pennycook et al. 2012).
The current research has several limitations. First, the present studies had a crosssectional character, which did not enable us to establish causal relationships between
investigated variables. Although in the research literature there are some suggestions
regarding the direction of the cognitive intelligence–religiosity link (Zuckerman
et al. 2013), the causality is not yet clear in the case of the relationship between
EI and religion. Clearly, then, more work in this context—including longitudinal
research—is needed to deepen our understanding of the interplay between religiosity, EI, and cognitive intelligence. Moreover, although we extended the scope of
previous research by examining the potential mediating role of emotional empathy,
it might be desirable to also examine an impact of other variables that have been
found to have signifcance for both intelligences and religion, such as time perspectives (see Łowicki et al. 2018; Stolarski et al. 2011; Zajenkowski et al. 2016). Next,
it should be emphasized that both current samples were mostly composed of Roman
Catholics and included Polish participants only. To increase the generalizability of
the obtained results, future work should examine other denominations, preferably
in non-Western societies. Finally, it needs to be acknowledged that we failed to replicate the main fnding using the exact same measure of religiosity across the two
studies. Instead, we found the suppression efect on diferent instruments measuring religiosity. As suggested above, this might be due to various factors such as the
weak magnitude of the observed efects or sample specifcity (cf. Webser and Dufy
2016). Nevertheless, our research provides some observations that might be helpful in future studies exploring the interplay between cognitive and emotional intelligence and religiosity. First, one needs to carefully consider the structure and size
of the sample. It seems that the efects of both intelligences on religiosity are rather
small, and therefore an adequate statistical power is required to detect them. Second,
taking into consideration some signifcant diferences between measures of religiosity, various aspects of religious phenomena (e.g., belief in God, fundamentalism,
interest in religion) should be included in the studies on cognitive and emotional
intelligence.
Those who are younger, male, and have higher incomes generally have higher car pride; less developed countries exhibit higher car pride
Measuring Car Pride and its Implications for Car Ownership and Use across Individuals, Cities, and Countries. Joanna C. Moody. PhD Thesis, Civil and Environmental Engineering Dept, MIT, May 2019. https://mobility.mit.edu/sites/default/files/MoodyDissertation_electronicversion.pdf
As the world recognizes that its growing reliance on private, fossil fuel-based vehicles is unsustainable, understanding how to avoid growth in car ownership and how to shift current users towards more efficient, environmentally-friendly, safe, and inclusive alternatives is acritical vision for meeting sustainable (transportation) development goals. Policy makers looking to shift consumer behavior away from cars need a more rigorous understanding of how different attitudes play a role in influencing car ownership and use and how this might vary by people and place. In this dissertation we provide deep insight into one of the many symbolic and affective motives behind car consumption: “car pride” or the attribution of social status and personal image to owning and using a car. Using data collected from individuals in two U.S. cities and in 51 countries around the world, we develop and demonstrate the reliability, validity, and invariance of polytomous (12, 7-point Likert-format statements) and dichotomous (9, dichotomous statements) survey measures for car pride using confirmatory factor analysis (CFA). With these measures, we explore variations in car pride across individuals, cities, and countries using Structural Equation Modeling (SEM). Across individuals, we find that those who are younger, male, and have higher incomes generally have higher car pride. Controlling for individual characteristics, we find that car pride is influenced by context. Between U.S. cities, we find that Houston has higher car pride than New York City. Across countries, we find that less developed countries exhibit higher car pride. We also disentangle the bidirectional causal relations between car pride and car consumption using instrumental variable (IV) techniques. We find that car pride strongly predicts car ownership, while no statistically significant relation exists in the opposite direction. Car pride additionally predicts car use, but only through itsrelation with car ownership (mediator). In the reverse direction, car use strongly reinforces car pride. While the directions of these relations appear almost universal across contexts, their strengths differ by country, emphasizing the importance of taking national context into account when measuring and interpreting symbolic motivations for car consumption.
7.1 Main Findings
In this section, we summarize the main findings from our empirical investigations into car prideand its relations with car consumption (Chapters 3-6). While each of these chapters includes morein-depth discussions of its results, here we summarize the key findings from across the different chapters. While we cannot directly compare model estimates between our two survey measuresof car pride and our two samples, we do identify overarching findings that are supported by both cases. Therefore, here we integrate findings from across our chapters, synthesizing what we havelearned generally about car pride and its relations with car ownership and use.
7.1.1 Measuring Car Pride
In this dissertation, we develop and test multiple ways of measuring car pride. In a sample of commuters from the New York City and Houston metropolitan areas, we investigate the reliability,validity, and invariance of an explicit measure of car pride derived from a polytomous survey scale(see Chapter 3) using a series of confirmatory factor analysis (CFA) models. Based on these results, we propose our 12-item, 7-point Likert-scale measure of the attribution of social status and personal image to driving and owning a vehicle as a new, standard measure for car pride that is reliable, valid, and invariant between cities and across individuals with different car consumption.For the U.S. cities sample, we also derive an implicit measure from a car vs. bus social status Implicit Association Test (IAT; see Chapter 4). Comparing the psychometric properties of our explicit and mplicit measures of car pride in our U.S. sample, we find that our explicit measure is a more valid measure. Comparing their correlations with actual car ownership and use, we further find that our explicit measure of car pride is more interpretable and useful than our implicit measure derived from the IAT. These results might suggest that explicit rather than implicit cognitive pathwaysdominate car consumption, including decisions of car ownership and use. The results also suggest that traditional (explicit) survey scales, if carefully developed and well-validated, are likely adequate for probing many attitude-behavior relations.In our international sample of 41,932 individuals in 51 countries, we test a dichotomous versionof the car pride scale, composed of 9 agree-disagree survey items designed for mobile phone-based data collection. Using multilevel CFA, we find that this measure also exhibits reasonable convergent validity, reliability, and invariance across countries and propose it as an alternative, standard measure particularly useful for cross-cultural comparison. However, the dichotomous version of the car pride scale is less able to differentiate among individuals who disagree with statements associatingsocial status and personal image to owning and using a car. Therefore, the polytomous version ofthe car pride scale should be preferred unless, as in our international survey, data collection will bedone via small-screen devices that make display of Likert-format scales difficult. Together this measurement development and validation provides standard, quantifiable survey scales of car pride that can be compared across people, providing consistent, specific, and actionable information for future transportation planning and policymaking. It also provides thefoundation needed for the empirical explorations of car pride and its relations with car ownershipand use in Chapters 5 and 6.
7.1.2 Variations in Car Pride
Equipped with well-validated survey measures of car pride, we can visualize and model variations in car pride across individuals, cities, and countries. For individuals in our U.S. cities, we findhigher car pride among those who are younger, male, white, students, and from higher-incomehouseholds. For individuals in our international sample, we find that those who are younger, male, highly educated, full-time employed, who live in larger towns or cities, and from higher-income households have higher car pride, no matter the country they live in. Therefore, acrossboth samples we identify age, gender, and income as significant sociodemographic predictors ofcar pride. However, we also find that individual sociodemographics are limited in their capacityto explain observed variation in car pride, suggesting that many other factors not explored in thisdissertation contribute to the formation of different levels of car pride among different individuals. Next, we compare car pride across cities and countries. Using multigroup and multilevel modeling techniques, we first control for any differences in car pride related to the individuals in the subsamples. Comparing between cities in the U.S., we find that an equivalent individual living in New YorkCity is likely to have lower car pride than one from Houston. While future work would be neededto explore what characteristics of these cities contribute to observed differences, we speculate thatcar pride may be related to car dependency; individuals living in cities like Houston—where urbanform and transportation infrastructure provide little alternative to owning and using a car—mayform greater symbolic attachment to their vehicles.
Similarly, after controlling for the types of people living in different countries, we find that developing countries—with lower national wealth, greater income inequality, and lower rates of carownership and use—report higher values of car pride. This suggests that the effect of nationalcontext on car pride is related to the stage of economic development and motorization of a given country. Again, explaining this observed variation across countries is left for future research.
7.1.3 Car Pride and Car Consumption
Next, we explore how car pride relates to car consumption. We begin by comparing car pridebetween car-owners and non-car owners and car-users and non-car-users. In almost every city andcountry examined, we find that individuals who own and use cars have significantly higher car pride than others. Applying multivariate structural equation modeling techniques, we find thatthese observed relations between car pride and car consumption remain significant after controlling for sociodemographics of the individuals in our U.S. and international samples. Unlike much of the literature that assumes attitudes influence behavior, we explore bidirectional relations between car pride and car ownership and use in our U.S. sample. We find that bidirectionalrelations exist between car pride and car ownership as well as between car pride and car use. However, the relative strengths of these bidirectional attitude-behavior relations depend on thedimension of car consumption. We find that car pride strongly predicts car ownership, which inturn predicts car use; in the reverse direction, car use strongly reinforces car pride (see Figure 7.1).In other words, an individual with higher car pride is more likely to own a vehicle, and, enabledwith this ownership, use it more frequently. In the reverse direction, we find that owning a car hasno statistically significant impact on car pride, but using a car more (in terms of frequency or milesdriven) contributes to greater car pride. All together, these relations create a feedback loop amongcar pride, car ownership, and car use. We find that the directions of car pride-car consumption relations depicted in Figure 7.1 hold onaverage across a diverse set of individuals living in different cities in the U.S. In fact, betweenNew York City and Houston, we do not find a statistically significant difference in the strength of these relations. We then impose these same directions when modeling car pride-car consumptionrelations across individuals in different countries around the world. In our international sample,we see significant variation across contexts. In particular, we find that the per-unit impact of anindividual’s car pride on the likelihood of owning a vehicle varies by country. Therefore, our work emphasizes the importance of taking the social and cultural context into account when measuringand interpreting symbolic and affective motivations for car consumption.
As the world recognizes that its growing reliance on private, fossil fuel-based vehicles is unsustainable, understanding how to avoid growth in car ownership and how to shift current users towards more efficient, environmentally-friendly, safe, and inclusive alternatives is acritical vision for meeting sustainable (transportation) development goals. Policy makers looking to shift consumer behavior away from cars need a more rigorous understanding of how different attitudes play a role in influencing car ownership and use and how this might vary by people and place. In this dissertation we provide deep insight into one of the many symbolic and affective motives behind car consumption: “car pride” or the attribution of social status and personal image to owning and using a car. Using data collected from individuals in two U.S. cities and in 51 countries around the world, we develop and demonstrate the reliability, validity, and invariance of polytomous (12, 7-point Likert-format statements) and dichotomous (9, dichotomous statements) survey measures for car pride using confirmatory factor analysis (CFA). With these measures, we explore variations in car pride across individuals, cities, and countries using Structural Equation Modeling (SEM). Across individuals, we find that those who are younger, male, and have higher incomes generally have higher car pride. Controlling for individual characteristics, we find that car pride is influenced by context. Between U.S. cities, we find that Houston has higher car pride than New York City. Across countries, we find that less developed countries exhibit higher car pride. We also disentangle the bidirectional causal relations between car pride and car consumption using instrumental variable (IV) techniques. We find that car pride strongly predicts car ownership, while no statistically significant relation exists in the opposite direction. Car pride additionally predicts car use, but only through itsrelation with car ownership (mediator). In the reverse direction, car use strongly reinforces car pride. While the directions of these relations appear almost universal across contexts, their strengths differ by country, emphasizing the importance of taking national context into account when measuring and interpreting symbolic motivations for car consumption.
7.1 Main Findings
In this section, we summarize the main findings from our empirical investigations into car prideand its relations with car consumption (Chapters 3-6). While each of these chapters includes morein-depth discussions of its results, here we summarize the key findings from across the different chapters. While we cannot directly compare model estimates between our two survey measuresof car pride and our two samples, we do identify overarching findings that are supported by both cases. Therefore, here we integrate findings from across our chapters, synthesizing what we havelearned generally about car pride and its relations with car ownership and use.
7.1.1 Measuring Car Pride
In this dissertation, we develop and test multiple ways of measuring car pride. In a sample of commuters from the New York City and Houston metropolitan areas, we investigate the reliability,validity, and invariance of an explicit measure of car pride derived from a polytomous survey scale(see Chapter 3) using a series of confirmatory factor analysis (CFA) models. Based on these results, we propose our 12-item, 7-point Likert-scale measure of the attribution of social status and personal image to driving and owning a vehicle as a new, standard measure for car pride that is reliable, valid, and invariant between cities and across individuals with different car consumption.For the U.S. cities sample, we also derive an implicit measure from a car vs. bus social status Implicit Association Test (IAT; see Chapter 4). Comparing the psychometric properties of our explicit and mplicit measures of car pride in our U.S. sample, we find that our explicit measure is a more valid measure. Comparing their correlations with actual car ownership and use, we further find that our explicit measure of car pride is more interpretable and useful than our implicit measure derived from the IAT. These results might suggest that explicit rather than implicit cognitive pathwaysdominate car consumption, including decisions of car ownership and use. The results also suggest that traditional (explicit) survey scales, if carefully developed and well-validated, are likely adequate for probing many attitude-behavior relations.In our international sample of 41,932 individuals in 51 countries, we test a dichotomous versionof the car pride scale, composed of 9 agree-disagree survey items designed for mobile phone-based data collection. Using multilevel CFA, we find that this measure also exhibits reasonable convergent validity, reliability, and invariance across countries and propose it as an alternative, standard measure particularly useful for cross-cultural comparison. However, the dichotomous version of the car pride scale is less able to differentiate among individuals who disagree with statements associatingsocial status and personal image to owning and using a car. Therefore, the polytomous version ofthe car pride scale should be preferred unless, as in our international survey, data collection will bedone via small-screen devices that make display of Likert-format scales difficult. Together this measurement development and validation provides standard, quantifiable survey scales of car pride that can be compared across people, providing consistent, specific, and actionable information for future transportation planning and policymaking. It also provides thefoundation needed for the empirical explorations of car pride and its relations with car ownershipand use in Chapters 5 and 6.
7.1.2 Variations in Car Pride
Equipped with well-validated survey measures of car pride, we can visualize and model variations in car pride across individuals, cities, and countries. For individuals in our U.S. cities, we findhigher car pride among those who are younger, male, white, students, and from higher-incomehouseholds. For individuals in our international sample, we find that those who are younger, male, highly educated, full-time employed, who live in larger towns or cities, and from higher-income households have higher car pride, no matter the country they live in. Therefore, acrossboth samples we identify age, gender, and income as significant sociodemographic predictors ofcar pride. However, we also find that individual sociodemographics are limited in their capacityto explain observed variation in car pride, suggesting that many other factors not explored in thisdissertation contribute to the formation of different levels of car pride among different individuals. Next, we compare car pride across cities and countries. Using multigroup and multilevel modeling techniques, we first control for any differences in car pride related to the individuals in the subsamples. Comparing between cities in the U.S., we find that an equivalent individual living in New YorkCity is likely to have lower car pride than one from Houston. While future work would be neededto explore what characteristics of these cities contribute to observed differences, we speculate thatcar pride may be related to car dependency; individuals living in cities like Houston—where urbanform and transportation infrastructure provide little alternative to owning and using a car—mayform greater symbolic attachment to their vehicles.
Similarly, after controlling for the types of people living in different countries, we find that developing countries—with lower national wealth, greater income inequality, and lower rates of carownership and use—report higher values of car pride. This suggests that the effect of nationalcontext on car pride is related to the stage of economic development and motorization of a given country. Again, explaining this observed variation across countries is left for future research.
7.1.3 Car Pride and Car Consumption
Next, we explore how car pride relates to car consumption. We begin by comparing car pridebetween car-owners and non-car owners and car-users and non-car-users. In almost every city andcountry examined, we find that individuals who own and use cars have significantly higher car pride than others. Applying multivariate structural equation modeling techniques, we find thatthese observed relations between car pride and car consumption remain significant after controlling for sociodemographics of the individuals in our U.S. and international samples. Unlike much of the literature that assumes attitudes influence behavior, we explore bidirectional relations between car pride and car ownership and use in our U.S. sample. We find that bidirectionalrelations exist between car pride and car ownership as well as between car pride and car use. However, the relative strengths of these bidirectional attitude-behavior relations depend on thedimension of car consumption. We find that car pride strongly predicts car ownership, which inturn predicts car use; in the reverse direction, car use strongly reinforces car pride (see Figure 7.1).In other words, an individual with higher car pride is more likely to own a vehicle, and, enabledwith this ownership, use it more frequently. In the reverse direction, we find that owning a car hasno statistically significant impact on car pride, but using a car more (in terms of frequency or milesdriven) contributes to greater car pride. All together, these relations create a feedback loop amongcar pride, car ownership, and car use. We find that the directions of car pride-car consumption relations depicted in Figure 7.1 hold onaverage across a diverse set of individuals living in different cities in the U.S. In fact, betweenNew York City and Houston, we do not find a statistically significant difference in the strength of these relations. We then impose these same directions when modeling car pride-car consumptionrelations across individuals in different countries around the world. In our international sample,we see significant variation across contexts. In particular, we find that the per-unit impact of anindividual’s car pride on the likelihood of owning a vehicle varies by country. Therefore, our work emphasizes the importance of taking the social and cultural context into account when measuringand interpreting symbolic and affective motivations for car consumption.
In the original studies, those of socioeconomic status have been shown to behave less prosocially across a variety of domains; we find that the generalizability of the original findings may be much more limited than suggested
Having Less, Giving More? Two Preregistered Replications of the Relationship Between Social Class and Prosocial Behavior. Angelos Stamos et al. Journal of Research in Personality, November 19 2019, 103902. https://doi.org/10.1016/j.jrp.2019.103902
Abstract: In the present report, we describe two planned direct replications of studies on the relationship between social class and prosocial behavior. In the original studies, individuals with higher socioeconomic status have been shown to behave less prosocially across a variety of domains. This finding continues to influence both research and the public debate on the psychological correlates of social class. At the same time, the validity of the original findings has been contested. Against this background, pre-registered direct replication studies with sufficient statistical power are warranted to test the robustness of these influential findings. We conducted two replication studies to provide valuable diagnostic information with regard to the relationship between social class and prosocial behavior. Our results indicate that the generalizability of the original findings may be much more limited than suggested. In addition, they highlight the need for an increased reliance on psychometrically established measures to facilitate cumulative research on the relationship between social class and prosocial behavior.
Check also... Low subjective SES was related to increased aggression, and subjective SES was not negatively related to trait and state measures of prosociality:
Abstract: In the present report, we describe two planned direct replications of studies on the relationship between social class and prosocial behavior. In the original studies, individuals with higher socioeconomic status have been shown to behave less prosocially across a variety of domains. This finding continues to influence both research and the public debate on the psychological correlates of social class. At the same time, the validity of the original findings has been contested. Against this background, pre-registered direct replication studies with sufficient statistical power are warranted to test the robustness of these influential findings. We conducted two replication studies to provide valuable diagnostic information with regard to the relationship between social class and prosocial behavior. Our results indicate that the generalizability of the original findings may be much more limited than suggested. In addition, they highlight the need for an increased reliance on psychometrically established measures to facilitate cumulative research on the relationship between social class and prosocial behavior.
Check also... Low subjective SES was related to increased aggression, and subjective SES was not negatively related to trait and state measures of prosociality:
Does Low (vs. High) Subjective Socioeconomic Status Increase Both Prosociality and Aggression? Tobias Greitemeyer and Christina Sagioglou. Social Psychology (2018), 49, pp. 76-87. https://www.bipartisanalliance.com/2018/03/low-subjective-ses-was-related-to.html
Indonesia: Homosexual men displayed a higher number of older brothers than heterosexual men, even when sibship size was controlled for; men with older brothers seem also more feminine than those without older brothers
Male Homosexual Preference: Femininity and the Older Brother Effect in Indonesia. Sarah Nila et al. Evolutionary Psychology, November 19, 2019. https://doi.org/10.1177/1474704919880701
Abstract: Male homosexual preference (MHP) is an evolutionary enigma because it is partially heritable and imposes a fertility cost. In occidental societies, homosexual men are feminized at various levels and they have more older brothers than heterosexual men. To evaluate whether femininity and the fraternal birth order (FBO) effect are universal features of MHP or not, we collected original data from homosexual men, heterosexual men, and heterosexual women from Java (Indonesia). Facial photographs were used to test whether homosexual faces are feminized when compared with heterosexual ones. We found that faces manipulated to resemble the average face of homosexual men are perceived as facially feminized, suggesting that homosexual men are facially feminized compared to heterosexual men, although a higher facial femininity was not captured by morphological analyses. Then, family data were used to detect differences in siblings’ composition between homosexuals and heterosexuals. Homosexual men displayed a higher number of older brothers than heterosexual men, even when sibship size was controlled for, suggesting that the FBO effect exists in Indonesian populations. Independent of sexual orientation, men with older brothers seem more feminized than those without older brothers, consistent with the immune origin of the FBO effect. In conclusion, MHP in Indonesia is partially feminized and they have more older brothers. Such features are also associated with MHP in other cultural contexts, suggesting a cross-cultural effect of men homosexual preference. An evolutionary explanation is available for the feminizing effect, although the FBO effect remains unexplained even if proximal mechanisms start to be identified.
Keywords: homosexuality, sexual dimorphism, fraternal birth order effect, sexual orientation, facial morphology
Faces
manipulated to resemble the average face of homosexual men are
perceived as facially feminized, which indicates that homosexual men are
facially feminized compared to heterosexual men, although higher facial
femininity was not captured by the morphological analyses. Homosexual
men displayed a greater number of older brothers, which was also
associated with higher perceived femininity. They did not display a
greater number of other sib categories.
When the shape of photographed men’s faces was partially transformed using an average homosexual face, the resulting faces appeared more feminized compared to similar transformation using an average heterosexual face. As age and masculinity are positively related in men (Boothroyd et al., 2005), a different mean age between the two groups of men used to build the average faces could trigger an unwanted difference in femininity/masculinity. However, the average faces were constructed with homosexual and heterosexual men sampled to minimize the difference in the age distribution (mean and variance). The resulting difference in mean age was less than 2 months, which is probably too small to generate a perceptible difference in masculinity. It is, however, possible that another variable, independent of sexual orientation, generated masculinity/femininity differences between the two samples, although this variable was not salary, education level, or ethnic origin, as these variables were not significantly different between the two groups. When the morphological difference between males and females was maximized during the discriminant analysis, homosexual men were not distributed differently compared to heterosexual men. Thus, any feminization displayed by homosexual men is not readily captured by the set of point coordinates or by their linear combinations. In this data set, when homosexual men were removed, heterosexual men and females were significantly differentiated by the discriminant analysis, although ∼20% of individuals were not correctly assigned. Other studies in occidental populations using similar morphological procedures to sex assign individuals typically report incorrect assignation within a range of 3–19% (e.g., Lee et al., 2014; Scott, Pound, Stephen, Clark, & Penton-Voak, 2010). This suggests that the present morphological analysis did not fully capture sexual facial differentiation. Alternatively, a real overlap could exist between male and female facial shapes.
Homosexual men displayed a greater number of older brothers than heterosexual men, suggesting that the older brother effect exists in Indonesian populations. The greater number of older brothers was present independently of possible higher fecundity observed in the families of homosexual men, as sib number was controlled for. A greater number of older sisters were also found, although it was no longer significant when sib number was controlled for, suggesting that this older sister effect is possibly driven by higher fecundity associated with the families of homosexual men. However, overall sibship size was not significantly differing between homosexual and heterosexual men. In Samoa, an older sister effect has been reported, although it is unclear if it remains after taking sib number into account. Thus far, the older brother effect has been found in all the populations in which it has been looked for (Western countries, Turkey, Iran, Hong Kong, Samoa, and Indonesia), suggesting that it is a general feature associated with MHP, although there is perhaps one counterexample (Brazil).
Using an index of facial femininity, homosexual men with more older brothers were not more feminized. As this femininity index does not capture differences between homosexual and heterosexual men, this result is preliminary. It has been shown that homosexuals with more older brothers are more feminine, the measure of femininity being a preference for the receptive role in anal intercourse (Blanchard, 2018a, 2018c), although a replication was equivocal (Swift-Gallant, Coome, Monks, & VanderLaan, 2018). The link between number of older brothers and femininity of homosexuals is not settled.
Independent of sexual orientation, men with older brothers seem more feminized than those without older brothers, consistent with the known effect of maternal parity on life-history traits (Skjærvø & Røskaft, 2013) and the hypothesis of the immune origin of the older brother effect. The possible maternal immune reaction primarily alters the development of sexually dimorphic brain structures relevant to sexual orientation, although other direct or indirect feminization effects are possible (Bogaert & Skorska, 2011). For example, birth weight is lower for newborn males with older brothers, but not for newborn females with older brothers or older sisters (Côté, Blanchard, & Lalumière, 2003). The higher influence of previous brothers to reduce male birth-weight, compared to female birth-weight, has been repeated in large samples (e.g., Nielsen et al., 2008) and is consistent with a lower birth weight of homosexual men relatively to heterosexuals (Xu, Norton, & Rahman, 2019).
Taken together, these results suggest the presence of a feminizing factor associated with male homosexuality that is partially determined by male birth order. This is consistent with the findings from Western societies and, thus, argues for a common pathway that could apply to various populations. Indeed, there are several lines of evidence showing that some specific aspects of the MHP are found cross-culturally. For example, early cross-gender or atypical behavior has been retrospectively assessed among men showing MHP in Brazil, Guatemala, Philippines, Thailand, Turkey, and United States (Cardoso, 2005, 2009; Whitam & Mathy, 1986; Whitam & Zent, 1984). These behaviors are displayed early during childhood and are found in distinct cultures; thus, providing another argument for a biological basis of this sexual preference. The developmental pathway of MHP could, therefore, rely on the same biological basis in distinct populations.
A feminizing factor is only a proximate explanation for the presence of MHP, and a more global framework is required to understand why such a feminizing factor exists. Interestingly, a sexually antagonistic gene that favors MHP in males and promotes fecundity (i.e., the ability to have children) in females has been proposed (Camperio-Ciani, Corna, & Capiluppi, 2004; Iemmola & Camperio-Ciani, 2009, but see Blanchard, 2012). Several studies support this hypothesis, and other studies have provided results that are consistent with predictions from this hypothesis (for a review, see Barthes, Crochet, & Raymond, 2015). The nature of the antagonistic factor has not yet been identified, but it has been proposed that it proximally enhances femininity in both sexes, resulting in the opposite effect on expected reproduction in each sex (Barthes, Godelle, & Raymond, 2013). Thus, under the sexually antagonistic gene hypothesis, the higher femininity of homosexual men, including their feminized sexual orientation, is seen as a pleiotropic cost of selection for higher fertility in females. In women, femininity of various traits is associated with fertility and is considered attractive (Hill & Hurtado, 1996; Law-Smith et al., 2006; Manning, Scutt, Whitehouse, & Leinster, 1997; Pawlowski, Boothroyd, Perrett, & Kluska, 2008; Rhodes, Simmons & Peters, 2005; Singh, 1993; Sugiyama, 2005).
Similarly, the immune origin of the older brother effect remains a proximate explanation and a broader context is required to understand why male birth rank interferes with sexual orientation in men. Birth order is obviously not heritable; thus, this trait cannot evolve by natural selection. However, the ability to generate a birth order effect is potentially heritable. Curiously, it is unclear whether the FBO effect should be seen as a feminizing effect that increases with male birth rank or as an anti-feminizing effect that decreases with birth order. The former phenomenon would be consistent with a mechanism that decreases competitive ability in later-born offspring, which would be useful to reduce, for example, the cost of sib competition in males. The latter phenomenon could operate when the firstborn males have special reproductive importance, for example, in societies promoting primogeniture (eldest son as the primary heir). However, the present study does not confirm such an association: Primogeniture has not been described in traditional Indonesian culture, particularly on the island of Java (Gultom, 2017), despite a significant FBO effect. The same situation is found in Turkey, where primogeniture was not traditionally enforced, but a significant FBO effect has been described (Blanchard, 2018a). Whether such a feminizing effect according to FBO exists in other mammals or not has apparently not been investigated (to our knowledge), although this would help to better understand this phenomenon in humans.
Abstract: Male homosexual preference (MHP) is an evolutionary enigma because it is partially heritable and imposes a fertility cost. In occidental societies, homosexual men are feminized at various levels and they have more older brothers than heterosexual men. To evaluate whether femininity and the fraternal birth order (FBO) effect are universal features of MHP or not, we collected original data from homosexual men, heterosexual men, and heterosexual women from Java (Indonesia). Facial photographs were used to test whether homosexual faces are feminized when compared with heterosexual ones. We found that faces manipulated to resemble the average face of homosexual men are perceived as facially feminized, suggesting that homosexual men are facially feminized compared to heterosexual men, although a higher facial femininity was not captured by morphological analyses. Then, family data were used to detect differences in siblings’ composition between homosexuals and heterosexuals. Homosexual men displayed a higher number of older brothers than heterosexual men, even when sibship size was controlled for, suggesting that the FBO effect exists in Indonesian populations. Independent of sexual orientation, men with older brothers seem more feminized than those without older brothers, consistent with the immune origin of the FBO effect. In conclusion, MHP in Indonesia is partially feminized and they have more older brothers. Such features are also associated with MHP in other cultural contexts, suggesting a cross-cultural effect of men homosexual preference. An evolutionary explanation is available for the feminizing effect, although the FBO effect remains unexplained even if proximal mechanisms start to be identified.
Keywords: homosexuality, sexual dimorphism, fraternal birth order effect, sexual orientation, facial morphology
Discussion
When the shape of photographed men’s faces was partially transformed using an average homosexual face, the resulting faces appeared more feminized compared to similar transformation using an average heterosexual face. As age and masculinity are positively related in men (Boothroyd et al., 2005), a different mean age between the two groups of men used to build the average faces could trigger an unwanted difference in femininity/masculinity. However, the average faces were constructed with homosexual and heterosexual men sampled to minimize the difference in the age distribution (mean and variance). The resulting difference in mean age was less than 2 months, which is probably too small to generate a perceptible difference in masculinity. It is, however, possible that another variable, independent of sexual orientation, generated masculinity/femininity differences between the two samples, although this variable was not salary, education level, or ethnic origin, as these variables were not significantly different between the two groups. When the morphological difference between males and females was maximized during the discriminant analysis, homosexual men were not distributed differently compared to heterosexual men. Thus, any feminization displayed by homosexual men is not readily captured by the set of point coordinates or by their linear combinations. In this data set, when homosexual men were removed, heterosexual men and females were significantly differentiated by the discriminant analysis, although ∼20% of individuals were not correctly assigned. Other studies in occidental populations using similar morphological procedures to sex assign individuals typically report incorrect assignation within a range of 3–19% (e.g., Lee et al., 2014; Scott, Pound, Stephen, Clark, & Penton-Voak, 2010). This suggests that the present morphological analysis did not fully capture sexual facial differentiation. Alternatively, a real overlap could exist between male and female facial shapes.
Homosexual men displayed a greater number of older brothers than heterosexual men, suggesting that the older brother effect exists in Indonesian populations. The greater number of older brothers was present independently of possible higher fecundity observed in the families of homosexual men, as sib number was controlled for. A greater number of older sisters were also found, although it was no longer significant when sib number was controlled for, suggesting that this older sister effect is possibly driven by higher fecundity associated with the families of homosexual men. However, overall sibship size was not significantly differing between homosexual and heterosexual men. In Samoa, an older sister effect has been reported, although it is unclear if it remains after taking sib number into account. Thus far, the older brother effect has been found in all the populations in which it has been looked for (Western countries, Turkey, Iran, Hong Kong, Samoa, and Indonesia), suggesting that it is a general feature associated with MHP, although there is perhaps one counterexample (Brazil).
Using an index of facial femininity, homosexual men with more older brothers were not more feminized. As this femininity index does not capture differences between homosexual and heterosexual men, this result is preliminary. It has been shown that homosexuals with more older brothers are more feminine, the measure of femininity being a preference for the receptive role in anal intercourse (Blanchard, 2018a, 2018c), although a replication was equivocal (Swift-Gallant, Coome, Monks, & VanderLaan, 2018). The link between number of older brothers and femininity of homosexuals is not settled.
Independent of sexual orientation, men with older brothers seem more feminized than those without older brothers, consistent with the known effect of maternal parity on life-history traits (Skjærvø & Røskaft, 2013) and the hypothesis of the immune origin of the older brother effect. The possible maternal immune reaction primarily alters the development of sexually dimorphic brain structures relevant to sexual orientation, although other direct or indirect feminization effects are possible (Bogaert & Skorska, 2011). For example, birth weight is lower for newborn males with older brothers, but not for newborn females with older brothers or older sisters (Côté, Blanchard, & Lalumière, 2003). The higher influence of previous brothers to reduce male birth-weight, compared to female birth-weight, has been repeated in large samples (e.g., Nielsen et al., 2008) and is consistent with a lower birth weight of homosexual men relatively to heterosexuals (Xu, Norton, & Rahman, 2019).
Taken together, these results suggest the presence of a feminizing factor associated with male homosexuality that is partially determined by male birth order. This is consistent with the findings from Western societies and, thus, argues for a common pathway that could apply to various populations. Indeed, there are several lines of evidence showing that some specific aspects of the MHP are found cross-culturally. For example, early cross-gender or atypical behavior has been retrospectively assessed among men showing MHP in Brazil, Guatemala, Philippines, Thailand, Turkey, and United States (Cardoso, 2005, 2009; Whitam & Mathy, 1986; Whitam & Zent, 1984). These behaviors are displayed early during childhood and are found in distinct cultures; thus, providing another argument for a biological basis of this sexual preference. The developmental pathway of MHP could, therefore, rely on the same biological basis in distinct populations.
A feminizing factor is only a proximate explanation for the presence of MHP, and a more global framework is required to understand why such a feminizing factor exists. Interestingly, a sexually antagonistic gene that favors MHP in males and promotes fecundity (i.e., the ability to have children) in females has been proposed (Camperio-Ciani, Corna, & Capiluppi, 2004; Iemmola & Camperio-Ciani, 2009, but see Blanchard, 2012). Several studies support this hypothesis, and other studies have provided results that are consistent with predictions from this hypothesis (for a review, see Barthes, Crochet, & Raymond, 2015). The nature of the antagonistic factor has not yet been identified, but it has been proposed that it proximally enhances femininity in both sexes, resulting in the opposite effect on expected reproduction in each sex (Barthes, Godelle, & Raymond, 2013). Thus, under the sexually antagonistic gene hypothesis, the higher femininity of homosexual men, including their feminized sexual orientation, is seen as a pleiotropic cost of selection for higher fertility in females. In women, femininity of various traits is associated with fertility and is considered attractive (Hill & Hurtado, 1996; Law-Smith et al., 2006; Manning, Scutt, Whitehouse, & Leinster, 1997; Pawlowski, Boothroyd, Perrett, & Kluska, 2008; Rhodes, Simmons & Peters, 2005; Singh, 1993; Sugiyama, 2005).
Similarly, the immune origin of the older brother effect remains a proximate explanation and a broader context is required to understand why male birth rank interferes with sexual orientation in men. Birth order is obviously not heritable; thus, this trait cannot evolve by natural selection. However, the ability to generate a birth order effect is potentially heritable. Curiously, it is unclear whether the FBO effect should be seen as a feminizing effect that increases with male birth rank or as an anti-feminizing effect that decreases with birth order. The former phenomenon would be consistent with a mechanism that decreases competitive ability in later-born offspring, which would be useful to reduce, for example, the cost of sib competition in males. The latter phenomenon could operate when the firstborn males have special reproductive importance, for example, in societies promoting primogeniture (eldest son as the primary heir). However, the present study does not confirm such an association: Primogeniture has not been described in traditional Indonesian culture, particularly on the island of Java (Gultom, 2017), despite a significant FBO effect. The same situation is found in Turkey, where primogeniture was not traditionally enforced, but a significant FBO effect has been described (Blanchard, 2018a). Whether such a feminizing effect according to FBO exists in other mammals or not has apparently not been investigated (to our knowledge), although this would help to better understand this phenomenon in humans.
Basic science knowledge & cognitive sophistication contribute more to science beliefs than political ideology; limited evidence for identity-protective cognition (i.e., motivated reasoning)
McPhetres, Jonathon, and Gordon Pennycook. 2019. “Science Beliefs, Political Ideology, and Cognitive Sophistication.” OSF Preprints. November 19. doi:10.31219/osf.io/ad9v7
Abstract: Various mechanisms have been proposed to explain why individuals hold pro- or anti-science beliefs. Some models focus on the role of political ideology and motivated reasoning, arguing that greater cognitive sophistication enables individuals to interpret evidence in an identity-consistent manner. Other models focus on the roles of basic science knowledge and cognitive sophistication, arguing that more general science knowledge and greater cognitive sophistication facilitate pro-science beliefs. To test these competing accounts, we identified a large range of controversial issues that are ostensibly subject to ideological disagreement and examined the relative roles of political ideology, science knowledge, and cognitive sophistication. Our results, which are consistent across two different nationally representative samples of Americans (N = 1,709), indicate that a combination of basic science knowledge and cognitive sophistication contribute more to science beliefs about various topics than political ideology. Furthermore, we found limited evidence for identity-protective cognition (i.e., motivated reasoning). By investigating a broad array of science-related beliefs, our results adjudicate between four accounts of science attitudes and suggest that educators and policymakers should focus on increasing basic science literacy and critical thinking instead of focusing on the ideological factors that divide people.
---
Anti-science attitudes represent a major roadblock for responsible public policy. Suitably, several competing mechanisms have been proposed to explain why people believe what they believe about science. Whereas some models focus on the potential roles of motivation and identity, others focus on the potential roles of science knowledge and information processing. Unfortunately, it is difficult to adjudicate between these competing accounts because past work has almost exclusively focused on a few specific issues, such as belief in climate change or evolution. This narrowed focus naturally limits the conclusions that can be drawn about why people hold anti-science beliefs–for example, because different topics may vary in their relation to identity or to the downstream benefit of basic science knowledge. Here we investigate a large number of science-related beliefs and examine a broad set of psychological correlates. This allows us to test the evidence for or against four specific accounts of science attitudes, each of which make different predictions about the roles of identity, basic science knowledge, and cognitive sophistication.
Perhaps the most prevalent account–the motivated reasoning account–argues that ideology playing a causal role in the formation of science-related beliefs. This account is commonly evoked to explain partisan differences in beliefs (or skepticism) about anthropogenic global warming, with conservatism being associated with an anti-science stance (Bohr, 2014; Gauchat, 2012; Lewandowsky, Oberauer, & Gignac, 2013a). The source of anti-science beliefs, according to the motivated reasoning account, is largely cultural. For example, global warming was politicized by conservative think tanks which was picked up by laypeople who are motivated to believe information that is consistent with their political ideology. Importantly, for other issues, it may be that liberals are motivated to reject science–such as with stereotypes that liberals are more likely to reject nuclear power. Other ideological factors, such as religiosity (McPhetres & Zuckerman, 2017; Rutjens, Sutton, & van der Lee, 2018), could also be implicated in the broad motivated reasoning account.
A related account, often referred to as identity-protective cognition, contends that cognitively sophisticated individuals are actually better able to use their reasoning skills to selectively conform their evaluation of evidence to their political ideology (Kahan, Landrum, Carpenter, Helft, & Hall Jamieson, 2017; Kahan, Peters, Dawson, & Slovic, 2013). Indeed, research has shown that polarization around scientific issues is stronger (not weaker) among individuals who are more cognitively sophisticated (Bolsen, Druckman, & Cook, 2015; Hamilton, Cutler, & Schaefer, 2012; Kahan et al., 2012) and/or educated (Drummond & Fischhoff, 2017a; Ehret, Sparks, & Sherman, 2017; Mccright & Dunlap, 2011). For example, concern for climate change decreases with increased numeracy among Republicans but the opposite association is evident for Democrats (Kahan et al., 2012), and those with greater science literacy are more polarized on beliefs about stem cell research, the big bang, and evolution (Drummond & Fischhoff, 2017a). However, the identity-protective cognition account has focused on a select number of issues which are known to be politically valanced, such as climate change and human evolution (Drummond & Fischhoff, 2017a; Gauchat, 2012; Lewandowsky, Oberauer, & Gignac, 2013b). Thus, it remains unclear if these few issues are the exception or the rule and therefore it is unknown if the identity-protective account can be extended to a wider variety of anti-science beliefs. This is important because it would be problematic for broad public policy on science education to be influenced by a few highly salient but exceptional cases.
In contrast to accounts that focus on motivation and identity, the knowledge deficit account suggests that people reject certain scientific claims simply because they do not possess enough (or the correct) basic scientific knowledge (e.g., Bak, 2001; Miller, 1998). Specifically, science knowledge is often considered in terms of the basic facts that one knows.for example, that electrons are smaller ["so to speak"] than atoms, or that antibiotics don't kill viruses. The knowledge deficit model implies that anti-science beliefs are prevalent primarily because science is difficult to understand without advanced training (Lombrozo, Shtulman, & Weisberg, 2006; Shtulman & Schulz, 2008). A key policy prescription from this model is that teaching people about science will straightforwardly lead to an increase in pro-scientific beliefs and attitudes.
Closely related is what we will call the analytic thinking account, which is that people simply do not think analytically enough about science issues. Whereas the knowledge deficit account suggests that having a strong core understanding of basic scientific facts is central to the formation of pro-science attitudes, the analytic thinking account argues that the disposition to think analytically and critically (over and above underlying science knowledge) is central. In support of this, studies have found that those who reason more analytically are more likely to endorse evolution (Gervais, 2015) and vaccination (Sarathchandra, Navin, Largent, & McCright, 2018), and to reject conspiracy theories (Swami, Voracek, Stieger, Tran, & Furnham, 2014) and paranormal beliefs (Pennycook, Cheyne, Seli, Koehler, & Fugelsang, 2012). In contrast, individuals who are more receptive to pseudo-profound bullshit (Pennycook, Cheyne, Barr, Koehler, & Fugelsang, 2015) –randomly generated nonsense statements– are both less analytic and more likely to believe in the efficacy of non-evidenced based alternative medicines. As above, however, it is unclear if these results are specific to particular science-related beliefs.
In this study, we provide data that allows us to distinguish between these different accounts of science belief. To do this, we identified a large range of controversial issues that are likely to be subject to ideological disagreement. This allows us to ask which of the accounts explains more variance both among and across specific issues. The motivated reasoning account predicts that ideology will be a strong and consistent correlate of science attitudes and the knowledge deficit account makes the same prediction, but for basic science knowledge instead of ideology. Naturally, it is possible that both accounts are accurate – our goal here is to assess which explains more variance, and for which science topics. The identity-protective cognition and analytic thinking accounts do, however, make competing predictions: Whereas the former predicts that cognitive sophistication will be associated with increased political polarization, the latter predictions that cognitive sophistication will simply be associated with increased proscientific attitudes across the board. Although it is possible that each of the four accounts is the best explanation for single specific science-related beliefs, the goal here is to ascertain which account(s) hold the strongest broad predictive power. This will help guide informed public policy in addition to psychological theory.
Abstract: Various mechanisms have been proposed to explain why individuals hold pro- or anti-science beliefs. Some models focus on the role of political ideology and motivated reasoning, arguing that greater cognitive sophistication enables individuals to interpret evidence in an identity-consistent manner. Other models focus on the roles of basic science knowledge and cognitive sophistication, arguing that more general science knowledge and greater cognitive sophistication facilitate pro-science beliefs. To test these competing accounts, we identified a large range of controversial issues that are ostensibly subject to ideological disagreement and examined the relative roles of political ideology, science knowledge, and cognitive sophistication. Our results, which are consistent across two different nationally representative samples of Americans (N = 1,709), indicate that a combination of basic science knowledge and cognitive sophistication contribute more to science beliefs about various topics than political ideology. Furthermore, we found limited evidence for identity-protective cognition (i.e., motivated reasoning). By investigating a broad array of science-related beliefs, our results adjudicate between four accounts of science attitudes and suggest that educators and policymakers should focus on increasing basic science literacy and critical thinking instead of focusing on the ideological factors that divide people.
---
Anti-science attitudes represent a major roadblock for responsible public policy. Suitably, several competing mechanisms have been proposed to explain why people believe what they believe about science. Whereas some models focus on the potential roles of motivation and identity, others focus on the potential roles of science knowledge and information processing. Unfortunately, it is difficult to adjudicate between these competing accounts because past work has almost exclusively focused on a few specific issues, such as belief in climate change or evolution. This narrowed focus naturally limits the conclusions that can be drawn about why people hold anti-science beliefs–for example, because different topics may vary in their relation to identity or to the downstream benefit of basic science knowledge. Here we investigate a large number of science-related beliefs and examine a broad set of psychological correlates. This allows us to test the evidence for or against four specific accounts of science attitudes, each of which make different predictions about the roles of identity, basic science knowledge, and cognitive sophistication.
Perhaps the most prevalent account–the motivated reasoning account–argues that ideology playing a causal role in the formation of science-related beliefs. This account is commonly evoked to explain partisan differences in beliefs (or skepticism) about anthropogenic global warming, with conservatism being associated with an anti-science stance (Bohr, 2014; Gauchat, 2012; Lewandowsky, Oberauer, & Gignac, 2013a). The source of anti-science beliefs, according to the motivated reasoning account, is largely cultural. For example, global warming was politicized by conservative think tanks which was picked up by laypeople who are motivated to believe information that is consistent with their political ideology. Importantly, for other issues, it may be that liberals are motivated to reject science–such as with stereotypes that liberals are more likely to reject nuclear power. Other ideological factors, such as religiosity (McPhetres & Zuckerman, 2017; Rutjens, Sutton, & van der Lee, 2018), could also be implicated in the broad motivated reasoning account.
A related account, often referred to as identity-protective cognition, contends that cognitively sophisticated individuals are actually better able to use their reasoning skills to selectively conform their evaluation of evidence to their political ideology (Kahan, Landrum, Carpenter, Helft, & Hall Jamieson, 2017; Kahan, Peters, Dawson, & Slovic, 2013). Indeed, research has shown that polarization around scientific issues is stronger (not weaker) among individuals who are more cognitively sophisticated (Bolsen, Druckman, & Cook, 2015; Hamilton, Cutler, & Schaefer, 2012; Kahan et al., 2012) and/or educated (Drummond & Fischhoff, 2017a; Ehret, Sparks, & Sherman, 2017; Mccright & Dunlap, 2011). For example, concern for climate change decreases with increased numeracy among Republicans but the opposite association is evident for Democrats (Kahan et al., 2012), and those with greater science literacy are more polarized on beliefs about stem cell research, the big bang, and evolution (Drummond & Fischhoff, 2017a). However, the identity-protective cognition account has focused on a select number of issues which are known to be politically valanced, such as climate change and human evolution (Drummond & Fischhoff, 2017a; Gauchat, 2012; Lewandowsky, Oberauer, & Gignac, 2013b). Thus, it remains unclear if these few issues are the exception or the rule and therefore it is unknown if the identity-protective account can be extended to a wider variety of anti-science beliefs. This is important because it would be problematic for broad public policy on science education to be influenced by a few highly salient but exceptional cases.
In contrast to accounts that focus on motivation and identity, the knowledge deficit account suggests that people reject certain scientific claims simply because they do not possess enough (or the correct) basic scientific knowledge (e.g., Bak, 2001; Miller, 1998). Specifically, science knowledge is often considered in terms of the basic facts that one knows.for example, that electrons are smaller ["so to speak"] than atoms, or that antibiotics don't kill viruses. The knowledge deficit model implies that anti-science beliefs are prevalent primarily because science is difficult to understand without advanced training (Lombrozo, Shtulman, & Weisberg, 2006; Shtulman & Schulz, 2008). A key policy prescription from this model is that teaching people about science will straightforwardly lead to an increase in pro-scientific beliefs and attitudes.
Closely related is what we will call the analytic thinking account, which is that people simply do not think analytically enough about science issues. Whereas the knowledge deficit account suggests that having a strong core understanding of basic scientific facts is central to the formation of pro-science attitudes, the analytic thinking account argues that the disposition to think analytically and critically (over and above underlying science knowledge) is central. In support of this, studies have found that those who reason more analytically are more likely to endorse evolution (Gervais, 2015) and vaccination (Sarathchandra, Navin, Largent, & McCright, 2018), and to reject conspiracy theories (Swami, Voracek, Stieger, Tran, & Furnham, 2014) and paranormal beliefs (Pennycook, Cheyne, Seli, Koehler, & Fugelsang, 2012). In contrast, individuals who are more receptive to pseudo-profound bullshit (Pennycook, Cheyne, Barr, Koehler, & Fugelsang, 2015) –randomly generated nonsense statements– are both less analytic and more likely to believe in the efficacy of non-evidenced based alternative medicines. As above, however, it is unclear if these results are specific to particular science-related beliefs.
In this study, we provide data that allows us to distinguish between these different accounts of science belief. To do this, we identified a large range of controversial issues that are likely to be subject to ideological disagreement. This allows us to ask which of the accounts explains more variance both among and across specific issues. The motivated reasoning account predicts that ideology will be a strong and consistent correlate of science attitudes and the knowledge deficit account makes the same prediction, but for basic science knowledge instead of ideology. Naturally, it is possible that both accounts are accurate – our goal here is to assess which explains more variance, and for which science topics. The identity-protective cognition and analytic thinking accounts do, however, make competing predictions: Whereas the former predicts that cognitive sophistication will be associated with increased political polarization, the latter predictions that cognitive sophistication will simply be associated with increased proscientific attitudes across the board. Although it is possible that each of the four accounts is the best explanation for single specific science-related beliefs, the goal here is to ascertain which account(s) hold the strongest broad predictive power. This will help guide informed public policy in addition to psychological theory.
Valuing Pain using the Subjective Well-being Method: The willingness to pay for pain relief is in the range of $56-145 per day, lower than previously reported
Valuing Pain using the Subjective Well-being Method. Thorhildur Ólafsdóttir, Tinna Laufey Ásgeirsdóttir, Edward C. Norton. Economics & Human Biology, November 16 2019, 100827. https://www.sciencedirect.com/science/article/pii/S1570677X19300656
Highlights
. Improved econometric methods provide new information on the value of pain
. We allow the trade-off between pain and income to vary across income ranges
. The willingness to pay for pain relief is in the range of 56-145 USD per day
. The monetary value of pain relief is lower than previously reported
Abstract: Chronic pain clearly lowers utility, but valuing the reduction in utility is empirically challenging. Here, we use improvements over prior applications of the subjective well-being method to estimate the implied trade-off between pain and income using four waves of the Health and Retirement Study (2008-2014), a nationally representative survey on individuals age 50 and older. We model income with a flexible functional form, allowing the trade-off between pain and income to vary across income groups. We control for individual fixed effects in the life-satisfaction equations and instrument for income in some models. We find values for avoiding pain ranging between 56 to 145 USD per day. These results are lower than previously reported and suggest that the higher previous estimates may be heavily affected by the highest income level and confounded by endogeneity in the income variable. As expected, we find that the value of pain relief increases with pain severity.
Discussion
By using improved econometric methods, we provide new information on the value of pain relief among people older than 50. Our results suggest a lower CV for pain than previously reported. More importantly, we contribute to the literature by using a PWL model as a more flexible method to express WTP across income ranges, instead of the traditional log transformation of income. Results from IV-models also yield CVs that are considerably lower than previous research suggests.
We compared our income coefficients to coefficients previously reported for lifesatisfaction equations from four different countries; Britain, Germany, Australia and USA (Clark et al., 2018) and found that our income coefficients correspond to the lower end of the range of coefficients. This is true after adjusting the income coefficients for different scaling of the life-satisfaction variable in the two studies (11/5), with our coefficients (times 2.2) closely reflecting those found using data from Britain (see Clark et al, Table 2.2). This comparison, although limited to model specifications where well-being is assumed to be linear in log income in OLS and FE models, is helpful to benchmark our income coefficients.
The two studies that use exogenous lottery wins to estimate the treatment effects of income on life-satisfaction . Lindqvist et al. (2018) and Apouey and Clark (2015) . have starkly different results. The treatment effect of $100K on life-satisfaction is estimated to be 0.037 SD units by Lindqvist et al. and 1.369 SD units by Apouey and Clark. We choose the former as an alternative for our income coefficient in Table 3 (column 1, OLS) because Lindqvist et al. use a larger sample size and have stronger internal validity. Lindqvist et al. approximate a lifetime income effect on life-satisfaction using lottery prizes annuitized over 20 years at a 2% interest rate. To calculate a CIV for our linear income model in Table 3, we take the estimate 0.062 from Table A10 in Lindqvist et al., which, after adjustment to different scaling of the life-satisfaction variable (0.062/2.2=0.028) and to USD prices between 2011 and 2015, we can apply an estimate of 0.028/1.0537=0.027. Using the income coefficient of 0.027 and our pain coefficient of -0.0563 that is adjusted for individual heterogeneity (FE model in Table 3), the corresponding CIV is 57 USD per day.
We did two additional sensitivity tests. First, instead of using a pain coefficient from an experimental setting, we used the pain coefficient from our estimate of the effect of pain on life-satisfaction through transitions into and out of arthritis status (see Table A5). Using the pain coefficient of -0.0793 from Table A5 and the lifetime income effect estimate from Lindqvist et al. (0.027) generates a CIV of 80 USD per day. Second, we compared the CIV by income coefficients across the two studies for the case of ln(income). The corresponding Lindqvist estimate (adjusted) is 0.16 (0.377 in Table A10 in Lindqvist et al.) and applying the FE coefficient of -0.0561 the estimated CIV is 54 USD per day. Both CIV estimates, 57 and 80 USD per day, are within our estimated range of 56-145 USD per day. The CIV using lifetime log-income effect from Lindqvist et al. of 54 USD per day is lower.
Because there is some variability across countries in the income gradient for life satisfaction equations (Clark et al, 2018), an income effect estimate based on a sample of Swedish lottery players may not apply to US data. Furthermore, there is inherent uncertainty in the annuity-adjustment parameter used to rescale the lifetime income gradient. However, the comparison is helpful and will hopefully stimulate similar research in other countries. This sensitivity analysis points towards the lower part of our estimated range of 56-145 USD per day as the most credible CIV estimate. We note that using income estimates from lottery studies instrinsically produces willingness to accept (WTA), whereas by using variation in household income the estimated CIV is between WTA and WTP.
Our paper has several limitations. Pain can be a consequence of neurological diseases, diabetes, or of musculoskeletal origin, but we did not have controls for those conditions that we found validated by a doctor´s diagnosis. However, any possibility of omitted-variable bias should be mitigated by controlling for age because neurological disease and diabetes likelihood (Type 2) increases with age. Furthermore, the numerous other health controls included should capture the effect of musculoskeletal conditions on pain and life satisfaction, in particular psychiatric problems, lung disease, cancer and arthritis. External validity may be limited by the age range used, in particular if the marginal utility of income for those over 50 is lower than in the younger population, which would result in higher CV-estimates than in a sample with lower mean age. Responses to life-satisfaction questions may be liable to situational influences, such as the site of the interview, the weather, one´s mood and the interviewer, but those differences can be considered as random error (Veenhoven, 1993). The value of pain is likely overestimated in previous research using the well-being valuation method, with our best approximation to a WTP estimate being in the range of 56- 145 USD per day. This range of estimates is derived from models where we take into account the effects of individual heterogeneity (by applying FE models), the leveraging effect of the highest income levels (by applying PWL models) and the endogeneity of the income variable with mother´s education as an instrument (by applying OLS-IV models). Those are issues that previous research and our analysis alike have highlighted as issues that should be taken into consideration when finding as reliable estimates as possible for the CIVs. Furthermore, the value of pain relief is positively related to severity of pain. Our research also has implications for the CV literature as a whole. We show the importance of controlling for the endogeneity of income and allowing the effect of income to vary flexibly across income levels. To this end, PWL models are promising because they perform well econometrically and allow for easier exploration of results across income groups than log transformations of income.
Highlights
. Improved econometric methods provide new information on the value of pain
. We allow the trade-off between pain and income to vary across income ranges
. The willingness to pay for pain relief is in the range of 56-145 USD per day
. The monetary value of pain relief is lower than previously reported
Abstract: Chronic pain clearly lowers utility, but valuing the reduction in utility is empirically challenging. Here, we use improvements over prior applications of the subjective well-being method to estimate the implied trade-off between pain and income using four waves of the Health and Retirement Study (2008-2014), a nationally representative survey on individuals age 50 and older. We model income with a flexible functional form, allowing the trade-off between pain and income to vary across income groups. We control for individual fixed effects in the life-satisfaction equations and instrument for income in some models. We find values for avoiding pain ranging between 56 to 145 USD per day. These results are lower than previously reported and suggest that the higher previous estimates may be heavily affected by the highest income level and confounded by endogeneity in the income variable. As expected, we find that the value of pain relief increases with pain severity.
Discussion
By using improved econometric methods, we provide new information on the value of pain relief among people older than 50. Our results suggest a lower CV for pain than previously reported. More importantly, we contribute to the literature by using a PWL model as a more flexible method to express WTP across income ranges, instead of the traditional log transformation of income. Results from IV-models also yield CVs that are considerably lower than previous research suggests.
We compared our income coefficients to coefficients previously reported for lifesatisfaction equations from four different countries; Britain, Germany, Australia and USA (Clark et al., 2018) and found that our income coefficients correspond to the lower end of the range of coefficients. This is true after adjusting the income coefficients for different scaling of the life-satisfaction variable in the two studies (11/5), with our coefficients (times 2.2) closely reflecting those found using data from Britain (see Clark et al, Table 2.2). This comparison, although limited to model specifications where well-being is assumed to be linear in log income in OLS and FE models, is helpful to benchmark our income coefficients.
The two studies that use exogenous lottery wins to estimate the treatment effects of income on life-satisfaction . Lindqvist et al. (2018) and Apouey and Clark (2015) . have starkly different results. The treatment effect of $100K on life-satisfaction is estimated to be 0.037 SD units by Lindqvist et al. and 1.369 SD units by Apouey and Clark. We choose the former as an alternative for our income coefficient in Table 3 (column 1, OLS) because Lindqvist et al. use a larger sample size and have stronger internal validity. Lindqvist et al. approximate a lifetime income effect on life-satisfaction using lottery prizes annuitized over 20 years at a 2% interest rate. To calculate a CIV for our linear income model in Table 3, we take the estimate 0.062 from Table A10 in Lindqvist et al., which, after adjustment to different scaling of the life-satisfaction variable (0.062/2.2=0.028) and to USD prices between 2011 and 2015, we can apply an estimate of 0.028/1.0537=0.027. Using the income coefficient of 0.027 and our pain coefficient of -0.0563 that is adjusted for individual heterogeneity (FE model in Table 3), the corresponding CIV is 57 USD per day.
We did two additional sensitivity tests. First, instead of using a pain coefficient from an experimental setting, we used the pain coefficient from our estimate of the effect of pain on life-satisfaction through transitions into and out of arthritis status (see Table A5). Using the pain coefficient of -0.0793 from Table A5 and the lifetime income effect estimate from Lindqvist et al. (0.027) generates a CIV of 80 USD per day. Second, we compared the CIV by income coefficients across the two studies for the case of ln(income). The corresponding Lindqvist estimate (adjusted) is 0.16 (0.377 in Table A10 in Lindqvist et al.) and applying the FE coefficient of -0.0561 the estimated CIV is 54 USD per day. Both CIV estimates, 57 and 80 USD per day, are within our estimated range of 56-145 USD per day. The CIV using lifetime log-income effect from Lindqvist et al. of 54 USD per day is lower.
Because there is some variability across countries in the income gradient for life satisfaction equations (Clark et al, 2018), an income effect estimate based on a sample of Swedish lottery players may not apply to US data. Furthermore, there is inherent uncertainty in the annuity-adjustment parameter used to rescale the lifetime income gradient. However, the comparison is helpful and will hopefully stimulate similar research in other countries. This sensitivity analysis points towards the lower part of our estimated range of 56-145 USD per day as the most credible CIV estimate. We note that using income estimates from lottery studies instrinsically produces willingness to accept (WTA), whereas by using variation in household income the estimated CIV is between WTA and WTP.
Our paper has several limitations. Pain can be a consequence of neurological diseases, diabetes, or of musculoskeletal origin, but we did not have controls for those conditions that we found validated by a doctor´s diagnosis. However, any possibility of omitted-variable bias should be mitigated by controlling for age because neurological disease and diabetes likelihood (Type 2) increases with age. Furthermore, the numerous other health controls included should capture the effect of musculoskeletal conditions on pain and life satisfaction, in particular psychiatric problems, lung disease, cancer and arthritis. External validity may be limited by the age range used, in particular if the marginal utility of income for those over 50 is lower than in the younger population, which would result in higher CV-estimates than in a sample with lower mean age. Responses to life-satisfaction questions may be liable to situational influences, such as the site of the interview, the weather, one´s mood and the interviewer, but those differences can be considered as random error (Veenhoven, 1993). The value of pain is likely overestimated in previous research using the well-being valuation method, with our best approximation to a WTP estimate being in the range of 56- 145 USD per day. This range of estimates is derived from models where we take into account the effects of individual heterogeneity (by applying FE models), the leveraging effect of the highest income levels (by applying PWL models) and the endogeneity of the income variable with mother´s education as an instrument (by applying OLS-IV models). Those are issues that previous research and our analysis alike have highlighted as issues that should be taken into consideration when finding as reliable estimates as possible for the CIVs. Furthermore, the value of pain relief is positively related to severity of pain. Our research also has implications for the CV literature as a whole. We show the importance of controlling for the endogeneity of income and allowing the effect of income to vary flexibly across income levels. To this end, PWL models are promising because they perform well econometrically and allow for easier exploration of results across income groups than log transformations of income.
Subscribe to:
Posts (Atom)
Posts
