The idea that conservatives are more sensitive to disgust than liberals is a basic tenet of political psychology — and it may be a mere artifact of self-reports

Investigating the conservatism-disgust paradox in reactions to the COVID-19 pandemic: A reexamination of the interrelations among political ideology, disgust sensitivity, and pandemic response. Benjamin C. Ruisch et al. PLoS One, November 4, 2022. https://doi.org/10.1371/journal.pone.0275440

Abstract: Research has documented robust associations between greater disgust sensitivity and (1) concerns about disease, and (2) political conservatism. However, the COVID-19 disease pandemic raised challenging questions about these associations. In particular, why have conservatives—despite their greater disgust sensitivity—exhibited less concern about the pandemic? Here, we investigate this “conservatism-disgust paradox” and address several outstanding theoretical questions regarding the interrelations among disgust sensitivity, ideology, and pandemic response. In four studies (N = 1,764), we identify several methodological and conceptual factors—in particular, an overreliance on self-report measures—that may have inflated the apparent associations among these constructs. Using non-self-report measures, we find evidence that disgust sensitivity may be a less potent predictor of disease avoidance than is typically assumed, and that ideological differences in disgust sensitivity may be amplified by self-report measures. These findings suggest that the true pattern of interrelations among these factors may be less “paradoxical” than is typically believed.

General discussion

This research provides important insight into the conservatism-disgust paradox in responses to the pandemic, as well as the relations among each of these target constructs—disgust sensitivity, political ideology, and pandemic response. These studies identified multiple factors that influence the (apparent) strength of the relations among these variables, thereby pinpointing several factors that are likely to have contributed to this seemingly contradictory pattern of results (Table 3).

Table 3. List of hypotheses, the study in which each hypothesis was tested, and whether or not each hypothesis was supported.

https://doi.org/10.1371/journal.pone.0275440.t003

One contributing factor appears to be the predominant use of self-report measures of pandemic response in past research. Indeed, using a behavioral measure of virtual social distancing, we found that the relations between pandemic response and both ideology and disgust sensitivity were significantly attenuated, compared with self-report pandemic response measures. These findings are consistent with the possibility that these self-report measures may suffer from IV-DV conceptual overlap, while also being more susceptible to social desirability and other reporting biases [36–38]. Particularly given that this same virtual behavioral measure has been shown to out-predict self-reports in predicting who contracts the COVID-19 virus [21], these results suggest that behavioral measures of pandemic response may provide a more accurate estimate of the extent of ideological differences in responses to the COVID-19 pandemic, as well as of the predictive power of disgust sensitivity for pandemic response. We found a similar divergence between self-report and non-self-report measures in the domain of disgust sensitivity. In this case, however, it was our experiential measure of disgust sensitivity that was the more powerful predictor of pandemic response. These findings identify important additional caveats and considerations for research examining the impact of disgust sensitivity on real-world outcomes, suggesting, in line with some past research, that self-reports of disgust sensitivity may correlate only modestly with other, more experiential or indirect indices of sensitivity to disgust—and that these measures/operationalizations may have different predictive power for different kinds of attitudes and behavior.

These findings also provide a means of beginning to reconcile some of the puzzling associations uncovered in other research on the COVID-19 pandemic. In particular, recent work suggests that—despite the putative disease-protective function of disgust—individuals who scored higher on self-reported disgust sensitivity may actually have been more likely to contract COVID-19 than those who self-reported less disgust sensitivity [23]. As documented here, however, self-reported disgust sensitivity appears to be only a relatively weak predictor of behavioral responses to the pandemic (indeed, adjusting for our experiential disgust measure rendered this association effectively nonexistent). Thus, although questions remain, these findings may bring us a step closer to understanding how self-reported disgust sensitivity could be a positive predictor of contracting the COVID-19 virus.

Perhaps the most intriguing findings, however, concern the relation of political ideology to self-report and experiential measures of disgust sensitivity. Using the DS-R, we replicated the well-documented ideological differences in self-reported disgust sensitivity. However, using our more experiential measure of disgust sensitivity—which presented participants with visual stimuli that closely corresponded to those described in the DS-R vignettes—we found no evidence of liberal-conservative differences in sensitivity to disgust.

Taken together, the findings discussed above suggest that methodological features of past research—particularly the heavy reliance on self-report measures of disgust sensitivity and pandemic response—may have inflated the relations among these three variables, and, thus, contributed to this seemingly contradictory pattern of results. In identifying the influence of these methodological factors, this research brings us a step closer to resolving the conservatism-disgust paradox, suggesting that the true pattern of interrelations among these variables is not as “paradoxical” as is typically assumed. That is, if, as these findings suggest, (1) the true relation between disgust sensitivity and pandemic response is smaller than previously suggested, and (2) ideological differences in disgust sensitivity are overestimated, then it is less surprising that conservatives exhibit less concern about the virus—particularly given that (3) ideological differences in responses to the pandemic may not be as dramatic as has been suggested by past research. The relatively small size of these effects makes it more likely that they would be subsumed by other concerns and motivations such as ideological identification and elite cues.

More generally, these findings also pose some challenges for past research and theory—particularly work suggesting a general relation between disgust sensitivity and political ideology. At the very least, these findings appear to suggest that liberals and conservatives do not differ in the form of disgust sensitivity that is most predictive of pandemic response. A more pessimistic interpretation, however, is that ideological differences in disgust sensitivity may generally be overestimated. That is, consistent with some recent critiques, it may be that self-report measures such as the DS-R amplify the true degree of ideological differences in disgust sensitivity, at least compared with measures that rely less on self-reports and self-beliefs about one’s own sensitivity to disgust.

Of course, our findings stand in contrast to a large body of research that suggests a connection between ideology and disgust, and, clearly, liberals and conservatives do reliably differ on many measures of disgust sensitivity (in particular, the DS-R and similar vignette-based measures). However, our findings also seem to align with other recent failures to replicate ideological differences in sensitivity to disgust using more indirect or experiential measures (e.g., [45]). Particularly in light of other research suggesting that people may have limited introspective ability into their own level of disgust sensitivity (e.g., work showing that self-reports sometimes do not significantly correlate with more indirect measures of disgust sensitivity; e.g., [18, 57, 58]) a closer examination of the nature and extent of ideological differences in disgust sensitivity may be warranted.

These findings therefore suggest that there may be a theoretical gap in our understanding of the relation between ideology and disgust sensitivity: Why is it that ideological differences reliably emerge on some measures of disgust sensitivity (e.g., the DS-R) but not others—even, as we found, measures that assess responses to closely related, or even identical, situations and stimuli? One possibility is that the ideological differences on the DS-R and similar vignette-based measures of sensitivity to disgust can in part be attributed to factors other than disgust sensitivity per se.

For example, forthcoming research suggests that conservatives tend to self-report greater interoceptive sensitivity—that is, to subjectively feel that they are more sensitive to the internal physiological states and signals of their own bodies—although by objective metrics they are actually less sensitive than are liberals [68]. Moreover, other research suggests that conservatives’ overconfidence may extend beyond interoception to experiences, judgments, and perceptions writ large [69]. Extending these past findings to the domain of disgust sensitivity would seem to suggest that conservatives may be likely to subjectively feel that they are more sensitive to disgust than they actually are, perhaps explaining why self-report measures of disgust sensitivity—which in part assess self-beliefs about one’s own degree of sensitivity to disgust—show more robust associations with conservatism than measures of disgust that are rooted in more immediate experience.

Less interestingly, another potential explanation for the weaker relation between ideology and our experiential disgust measure may be that previously documented ideological differences in personality traits such as conscientiousness [70] lead conservatives to complete survey measures more thoughtfully, perhaps reading more carefully or engaging more deeply with the material. This, too, could help explain why conservatives report experiencing greater disgust in response to these vignettes—which require a degree of cognitive effort to process and mentally represent—but do not appear to differ as greatly when these same stimuli are presented visually. Future research may wish to assess these possibilities to deepen our understanding of the nature of the relation between ideology and sensitivity to disgust.

More generally, these findings suggest that caution may be warranted in the development and use of measures to assess these constructs—disgust sensitivity, political ideology, and pandemic response—and, especially, their interrelations. Given the close connections among these factors, coupled with potential confounds such as self-presentational concerns that may be at play for such impactful and politicized issues as the COVID-19 pandemic, the use of self-report measures, in particular, should be subject to close scrutiny.

Finally, it is important to note that while our studies consistently show that using self-report scales may overestimate the strength of the interrelations among disgust sensitivity, pandemic response, and political ideology, some of these effects may be specific to the population that we sampled. Indeed, the sociopolitical context surrounding the COVID-19 pandemic in the U.S. was in many ways unique, and these factors are likely to have shaped some of our effects. In particular, as discussed above, the stark political polarization surrounding the pandemic in the U.S. is likely to have been at least partially responsible for the inflated ideological differences in self-reported (versus behavioral) responses to the pandemic. Future research will need to examine the degree to which these processes extend beyond the U.S. to other nations and cultural contexts.

Industrial Revolution. Why Britain? The Right Place (in the Technology Space) at the Right Time

Why Britain? The Right Place (in the Technology Space) at the Right Time. Carl Hallmann, W. Walker Hanlon, and Lukas Rosenberger. NBER, Jul 5 2022. https://conference.nber.org/conf_papers/f171957.pdf

Abstract: Why did Britain attain economic leadership during the Industrial Revolution? We argue that Britain possessed an important but underappreciated innovation advantage: British inventors worked in technologies that were more central within the innovation network. We offer a new approach for measuring the innovation network using patent data from Britain and France in the 18th and early 19th century. We show that the network influenced innovation outcomes and then demonstrate that British inventors worked in more central technologies within the innovation network than inventors from France. Then, drawing on recently-developed theoretical tools, we quantify the implications for technology growth rates in Britain compared to France. Our results indicate that the shape of the innovation network, and the location of British inventors within it, can help explain the more rapid technological growth in Britain during the Industrial Revolution.

Excerpts from the introduction:

In this study, we argue that there is one important British advantage that has been largely overlooked: the possibility that British inventors may have been working “at the right place” in the technology space. Our idea builds on emerging literature in growth economics which finds that innovation in some technologies generates more spillover benefits than innovation in others (Acemoglu et al., 2016; Cai and Li, 2019; Huang and Zenou, 2020; Liu and Ma, 2021). As a result, a country’s allocation of researchers across technologies can substantially impact the overall rate of economic growth. In particular, this literature shows that technological progress will be faster in economies where more research effort is focused on technologies that generate more spillovers for other technologies; in other words, technologies that are more central in the technology space. Translating these ideas into the context of the Industrial Revolution, we ask: did Britain experienced more rapid technological progress because British inventors were more focused on technologies, such as steam engines, machine tools, or metallurgy, that generated stronger spillover benefits for other technologies and were therefore more central in the technology space? In contrast, could it have been the case that Continental economies like France experienced slower technological progress because they specialized in developing technologies, such as apparel, glass, or papermaking, which were more peripheral in the technology space?1 Put another way, we aim to examine whether Britain’s differential growth during the eighteenth and early nineteenth centuries can be explained by the distinct position of British inventors in the technology space. By starting with ideas from modern growth economics, our analysis is less subject to the type of “post hoc, proper hoc” concerns that have been raised about some other explanations (Crafts, 1977, 1995). Moreover, we offer a theoretically-grounded quantification describing exactly how much of Britain’s differential growth experience can be attributed to this mechanism. These two features differentiate our study from most existing work that aims to understand Britain’s growth lead during the Industrial Revolution. To structure our analysis, we begin with a growth model, from Liu and Ma (2021), that incorporates an innovation network. In this network, each node is a technology type, while each edge reflects the extent to which innovations in one technology type increase the chances of further innovation in another. This model provides a framework for thinking about how the distribution of researchers across technology sectors relates to the growth rate in the economy. It also generates specific expressions that, given the matrix of connections across sectors, allow us to quantify how different allocations of researchers across technology sectors will affect growth. The upshot is that allocations in which more researchers are working in technology sectors with greater spillovers will generate higher overall growth rates than others. Therefore, the growth maximizing allocation of researchers will feature more researchers working in more central technology sectors: specifically, those sectors with higher eigenvalue network centrality. Furthermore, the model delivers precise analytical relationships that allow us to quantify the implications of different allocations of research effort for the rate of economic growth. To examine whether these forces operated during the Industrial Revolution, we utilize patent data for Britain, from 1700 to 1849, and for France from 1791-1844.2 These historical patent data cover a large number of inventors and their inventions, providing a rich source of information on innovation during the Industrial Revolution.3 We follow a long line of work, dating back at least to Sullivan (1989), using patent data to better understand innovation patterns during this period. A key challenge in our setting is measuring spillovers across technology categories. The innovation literature typically uses patent citations, but these are not available in our historical setting. Instead, we introduce a new approach based on the idea that if there are spillovers between two technology categories, then inventors working primarily in one area will occasionally file patents in the other. In particular, we measure the extent of spillovers from technology category j to i based on the propensity of inventors who patent in j to subsequently patent in i. Since our approach is new, we validate it using modern data. Specifically, using U.S. patents from 1970-2014, we construct innovation networks using our approach as well as the citation-based approach used in modern studies. Comparing these networks shows that the two approaches generate networks that are extremely similar. This suggests that our method does a good job of recovering the underlying innovation network. Using our approach, we document technology networks in Britain and France that feature a dense central core of closely related—and mainly mechanical—technologies. One important question about our estimated networks is, do they reflect fundamental features of the underlying technologies or simply reflect the local innovation environment in each country? One way to test this is to compare the networks obtained from the two countries. If they are similar, they likely reflect fundamental technological features rather than idiosyncratic conditions. Conducting a direct comparison, however, is challenging because the two countries use very different technology categorizations. Therefore, it is necessary to construct a mapping of technology categories from one country’s categorizations to the other. To do so, we carefully identify a set of inventions that were patented in both countries. We can then use the categorization of these inventions in each system to construct a crosswalk between the technology categorizations used in the two countries. Using this mapping, we construct technology spillover matrices derived from French patents but in terms of British technology categories, or derived from British patents but expressed in French technology categories. This allows us to regress the entries of the technology matrices of one country on the entries of the other country. We find they are strongly positively related, despite the noise that is inherent in any mapping between different systems of technology categorization. This indicates that our innovation matrices not just reflect the local economic environment, but that a significant part of each represents an underlying ‘global’ network of technology spillovers. Next, we establish that the shape of the technology spillover network matters for innovation outcomes. As a first step, we follow existing work on modern patent data by analyzing how patenting rates vary across technology categories depending on the lagged knowledge stock in other categories, weighted by the strength of connections through the innovation matrix. Consistent with the theory, and the results in previous studies of modern data, we find a significant positive associations of patenting with the lagged network weighted knowledge stock, shrinking toward zero as lags increase. However, the lack of exogenous variation in the lagged knowledge stock means that this result could be due to common shocks that affect connected technology categories. Thus, in the second step, we provide evidence based on a source of quasi-exogenous variation in the timing of increases in the knowledge stock at some nodes of the innovation network. Specifically, we use the unexpected arrival of “macroinventions.” These are inventions which Mokyr (1990) describes as “a radical new idea, without a clear precedent, emerges more or less ab nihilo.” Using a list of 65 macroinventions from Nuvolari et al. (2021), we study whether the arrival of a new macroinvention in one technology category leads to a subsequent increase in patenting in downstream technology categories within the innovation network. Here, the identifying assumption is not that the location of macroinventions were random, but that the timing of their arrival at a given location was unpredictable within the time frame of analysis. Using pooled difference-in-difference and event study analyses for a time frame of ten years before and after the arrival of each macroinvention, we show that macroinventions are followed by significant increases of the patenting rates in technology categories sharing stronger (downstream) connections from the technology category of the macroinvention. In addition, we find no evidence of an increase in technology categories as a result of being upstream from the macroinvention technology category within the innovation network. This second result provides a valuable placebo check that provides additional confidence that our results are picking up the impact of spillovers through the innovation network. Next, we look at whether there are notable differences in the allocation of British and French inventors within the innovation network. In particular, we focus on whether British inventors were patenting in technology categories that were more central within the innovation network than French inventors. We do this by studying, within the sets of British and French patents whether foreign inventors (of British or French origin) were patenting in more central technology categories than domestic inventors. We find that among French patents, patents by British-based inventors were significantly more central compared to the average patents by French domestic inventors—and all other foreign inventors—, whereas among British patents, patents by French-based inventors were less central compared to the average patent by British domestic inventors. The pattern indicates that British inventors were more likely to work in central technology categories than French inventors. As more central nodes have stronger spillover connections to other technology categories, the more central locations occupied by British inventors are consistent with a greater “bang for the buck” of British innovation on the aggregate rate of technological progress. Finally, we quantify the growth implications of the observed innovation network and different allocations of inventors in Britain and France through the lens of the model. Existing estimates for Britain suggest that industrial production grew by between 3 and 3.5% during the first half of the nineteenth century (Broadberry et al., 2015). In France, estimates indicate growth rates of between 1.7 and 2.5% in the same period (Crouzet, 1996; Asselain, 2007). (Preliminary) Results from our quantification exercise show that differences in the allocation of inventors across technology categories led to a technology growth rate in Britain that was between 0.5 and 2.9 percent higher than the French technology growth rate. Thus, our results indicate that Britain’s more advantageous position in the innovation network can explain a substantial fraction, and possibly the entire difference, in growth rates between the British and French economies during the first half of the nineteenth century. In sum, the evidence presented in this paper shows that Britain benefited from an advantageous distribution of inventors across technology sectors during the Industrial Revolution, and that this difference meaningfully contributed to Britain’s more rapid industrialization. Our analysis takes as given the differences in the distribution of inventors across sectors. Thus, our mechanism complements explanations for the British advantage during the Industrial Revolution, in particular those that can explain why British inventors were more likely than the French to work on technologies that happened to be more central 4 within the innovation network, in particular mechanical technologies. For example, it could be that Britain’s practical Enlightenment tradition and well-developed apprenticeship system (Mokyr, 2009; Kelly et al., 2014) contributed to the British inventors’ greater ability for working on mechanical technologies, or that high wages and access to cheap coal steered British inventors to focus on labor-saving mechanical devices (Allen, 2009). Put differently, the contribution of our paper lies in demonstrating that Britain was at the right place in the technology space at the right time, rather than explain why it was there but France was not. In addition to improving our understanding of one of the most important questions in economic history, our study also contributes to work by growth economists on the importance of innovation networks. Relative to studies in this area (cited above), we offer two main contributions. First, we offer new methods that can help researchers study innovation networks further back in history, when standard tools such as systematic patent citations are unavailable. This opens up the possibility of studying the influence of innovation networks in different contexts or over longer periods. Second, our analysis of macroinventions provides additional, more causal, evidence that innovation networks matter for technology development. Third, our application demonstrates empirically the value of recent theoretical advances integrating innovation networks into economic growth models. Our work builds on a long line of literature using patent data to examine innovation during the Industrial Revolution and into the nineteenth century. Early papers in this area include Sullivan (1989) and Sullivan (1990). More recent work includes MacLeod et al. (2003), Khan and Sokoloff (2004), Moser (2005), Khan (2005), Brunt et al. (2012), Nicholas (2011), Nuvolari and Tartari (2011), Moser (2012), Bottomley (2014b), Bottomley (2014a), Burton and Nicholas (2017), Khan (2018), Bottomley (2019), Nuvolari et al. (2020), Nuvolari et al. (2021), Hallmann et al. (2021), and Hanlon (2022). Relative to this extensive literature, we are the first to study the role of innovation networks in influencing inventive activity during the Industrial Revolution.

1 Hallmann et al. (2021) show that technological leadership in invention of Britain relative to France varied across technologies, with Britain leading, besides others, in steam engines and textile technologies, and France leading, besides others, in papermaking and shoemaking. Mokyr (1990, Chapter 5) provides a historical overview on British technological lead or lag in invention relative to Continental Europe.

2 Both of these were periods during which the patent systems were largely stable. We end just before the major British patent reform of 1852 and the French patent reform of 1844.

3  Of course, not every useful invention was patented, as (Moser, 2012) has shown. 

4   A stable institutional environment and well-developed patent system may have contributed in shifting inventors from technologies that can be protected by secrecy toward technologies as mechanical devices that are easily reverse engineered and thus profit the most from patents (Moser, 2005). However, as both Britain and France had strong patent protection, it is unclear how this mechanism could explain the differential focus of British vs. French inventors on mechanical devices.

We document that 85% of patent applications in the United States include no female inventors and ask: why are women underrepresented in innovation?

Try, try, try again? Persistence and the gender innovation gap. Gauri Subramani, Abhay Aneja, and Oren Reshef. Berkeley, Nov 2022. https://haas.berkeley.edu/wp-content/uploads/Try-try-try-again-Persistence-and-the-Gender-Innovation-Gap.pdf

Abstract: We document that 85% of patent applications in the United States include no female inventors and ask: why are women underrepresented in innovation? We argue that differences in responses to early rejections between men and women are a significant contributor to the gender disparity in innovation. We evaluate the prosecution and outcomes of almost one million patent applications in the United States from 2001 through 2012 and leverage variation in patent examiners’ probabilities of rejecting applications to employ a quasi-experimental instrumental variables approach. Our results show that applications from women are less likely to continue in the patent process after receiving an early rejection. Roughly half of the overall gender gap in awarded patents during this period can be accounted for by the differential propensity of women to abandon applications. We explore why this may be the case and provide evidence that the gender gap in outcomes is reduced for applications that are affiliated with firms, consistent with a role for institutional support in mitigating gender disparities.

Information avoidance: Our findings, together with additional survey evidence, suggest that behavioral biases inhibit the adoption of improved practices, and are consistent with inattention as a key driver of under-adoption

Why Businesses Fail: Underadoption of Improved Practices by Brazilian Micro-Enterprises. Priscila de Oliveira, Nov 2022. https://www.prisciladeoliveira.net/#research

Abstract: Micro firms in low and middle income countries often have low profitability and do not grow over time. Several business training programs have tried to improve management and business practices, with limited effects. We run a field experiment with micro-entrepreneurs in Brazil (N=742) to study the under-adoption of improved business practices, and shed light on the constraints and behavioral biases that may hinder their adoption. We randomly offer entrepreneurs reminders and micro-incentives of either 20 BRL (4 USD) or 40 BRL (8 USD) to implement record keeping or marketing for three consecutive months, following a business training program. Compared to traditional business training, reminders and micro-incentives significantly increase adoption of marketing (13.2 p.p.) and record keeping (19.2 p.p.), with positive effects on firm survival and investment over four months. Our findings, together with additional survey evidence, suggest that behavioral biases inhibit the adoption of improved practices, and are consistent with inattention as a key driver of under-adoption. In addition, our survey evidence on information avoidance points to it as a limiting factor to the adoption of record keeping, but not marketing activities. Taken together, the results suggest that behavioral biases affect firm decisions, with significant impact on firm survival.

Despite the popularity of growth mindset interventions in schools, positive results are rare and possibly spurious due to inadequately designed interventions, reporting flaws, and bias

Macnamara, B. N., & Burgoyne, A. P. (2022). Do growth mindset interventions impact students’ academic achievement? A systematic review and meta-analysis with recommendations for best practices. Psychological Bulletin, Nov 2022. https://doi.org/10.1037/bul0000352

Abstract: According to mindset theory, students who believe their personal characteristics can change—that is, those who hold a growth mindset—will achieve more than students who believe their characteristics are fixed. Proponents of the theory have developed interventions to influence students’ mindsets, claiming that these interventions lead to large gains in academic achievement. Despite their popularity, the evidence for growth mindset intervention benefits has not been systematically evaluated considering both the quantity and quality of the evidence. Here, we provide such a review by (a) evaluating empirical studies’ adherence to a set of best practices essential for drawing causal conclusions and (b) conducting three meta-analyses. When examining all studies (63 studies, N = 97,672), we found major shortcomings in study design, analysis, and reporting, and suggestions of researcher and publication bias: Authors with a financial incentive to report positive findings published significantly larger effects than authors without this incentive. Across all studies, we observed a small overall effect: d¯ = 0.05, 95% CI = [0.02, 0.09], which was nonsignificant after correcting for potential publication bias. No theoretically meaningful moderators were significant. When examining only studies demonstrating the intervention influenced students’ mindsets as intended (13 studies, N = 18,355), the effect was nonsignificant: d¯ = 0.04, 95% CI = [−0.01, 0.10]. When examining the highest-quality evidence (6 studies, N = 13,571), the effect was nonsignificant: d¯ = 0.02, 95% CI = [−0.06, 0.10]. We conclude that apparent effects of growth mindset interventions on academic achievement are likely attributable to inadequate study design, reporting flaws, and bias.

Impact Statement: This systematic review and meta-analysis suggest that, despite the popularity of growth mindset interventions in schools, positive results are rare and possibly spurious due to inadequately designed interventions, reporting flaws, and bias.

Those who self-report a strong moral character have a tendency to hypocrisy

Being good to look good: Self-reported moral character predicts moral double standards among reputation-seeking individuals. Mengchen Dong,Tom R. Kupfer,Shuai Yuan,Jan-Willem van Prooijen. British Journal of Psychology, November 4 2022. https://doi.org/10.1111/bjop.12608

Abstract: Moral character is widely expected to lead to moral judgements and practices. However, such expectations are often breached, especially when moral character is measured by self-report. We propose that because self-reported moral character partly reflects a desire to appear good, people who self-report a strong moral character will show moral harshness towards others and downplay their own transgressions—that is, they will show greater moral hypocrisy. This self-other discrepancy in moral judgements should be pronounced among individuals who are particularly motivated by reputation. Employing diverse methods including large-scale multination panel data (N = 34,323), and vignette and behavioural experiments (N = 700), four studies supported our proposition, showing that various indicators of moral character (Benevolence and Universalism values, justice sensitivity, and moral identity) predicted harsher judgements of others' more than own transgressions. Moreover, these double standards emerged particularly among individuals possessing strong reputation management motives. The findings highlight how reputational concerns moderate the link between moral character and moral judgement.

Practitioner points

- Self-reported moral character does not predict actual moral performance well.

- Good moral character based on self-report can sometimes predict strong moral hypocrisy.

- Good moral character based on self-report indicates high moral standards, while only for others but not necessarily for the self.

- Hypocrites can be good at detecting reputational cues and presenting themselves as morally decent persons.

GENERAL DISCUSSION

A well-known Golden Rule of morality is to treat others as you wish to be treated yourself (Singer, 1963). People with a strong moral character might be expected to follow this Golden Rule, and judge others no more harshly than they judge themselves. However, when moral character is measured by self-reports, it is often intertwined with socially desirable responding and reputation management motives (Anglim et al., 2017; Hertz & Krettenauer, 2016; Reed & Aquino, 2003). The current research examines the potential downstream effects of moral character and reputation management motives on moral decisions. By attempting to differentiate the ‘genuine’ and ‘reputation managing’ components of self-reported moral character, we posited an association between moral character and moral double standards on the self and others. Imposing harsh moral standards on oneself often comes with a cost to self-interest; to signal one's moral character, criticizing others' transgressions can be a relatively cost-effective approach (Jordan et al., 2017; Kupfer & Giner-Sorolla, 2017; Simpson et al., 2013). To the extent that the demonstration of a strong moral character is driven by reputation management motives, we, therefore, predicted that it would be related to increased hypocrisy, that is, harsher judgements of others' transgressions but not stricter standards for own misdeeds.

Across four studies varying from civic transgressions (Study 1), organizational misconducts (Study 2), to selfish decisions in economic games (Study 3 and the Pilot Study in the SM), we found consistent evidence that people reporting a strong (vs. weak) moral character were more likely to judge others' misdeeds harshly, especially for those highly motivated by reputation. This amplified moral harshness towards others was sometimes also accompanied with increased moral leniency towards the self (Study 3 and the Pilot Study in the SM). Taken together, self-reported moral character relates to differential moral standards on the self versus others, which was especially true for reputation-motivated individuals.

Although Study 1 only provided circumstantial evidence by interpreting moral judgements without specific targets and self-reported transgressive frequencies as a proxy of the ‘reputation managing’ component of self-reported moral character, we have good reasons to believe that these interpretations are legitimate. First, people often apply general moral rules to judgements of others instead of themselves (Dong et al., 2021). Second, self-reported moral performance is often influenced by strategic self-presentation (Dong et al., 2019; Shaw et al., 2014). As shown in our studies, people high (vs. low) on moral character reported fewer own transgressions (Study 1) when highly (vs. weakly) motivated by reputation management. However, they did not act more or less selfishly (Study 3).

Furthermore, Studies 2 and 3 consolidated our proposition by showing a significant interaction between moral character and target of moral judgements (i.e. self vs. other), only for people with high but not low reputation management motives. These findings were replicated across a variety of individual difference measures of moral character (including Benevolence and Universalism values, justice sensitivity, and moral identity) and reputation management motives (including Power and Achievement values, self-monitoring of socially desirable behaviours, and concern about social esteem and status), and emerged only when moral judgements had a salient influence on people's reputation (e.g. when the appraised behaviour was unfavourable rather than favourable in Study 3).

Theoretical contributions

The current findings contribute to the literature on both moral character and reputation management. Previous theorizing generally implies that moral character is genuinely and unconditionally good (Aquino & Reed, 2002; Kamtekar, 2004; Walker et al., 1987; Walker & Frimer, 2007). Consistent with this ‘genuine’ perspective on moral character, we found positive correlations of moral character with stringent moral judgements (Studies 1 and 3) and a high likelihood to behave morally (Study 3), although the relation between inherent moral character and actual moral deeds may be obscured by the presence of external sanctions (e.g. third-party punishment in the Pilot Study in the SM). More importantly, we complement previous studies on moral character by making two novel contributions.

First, the present studies suggest that there are both ‘genuine’ and ‘reputation managing’ components of self-reported moral character. Although this idea was implied in many previous studies (e.g. Anglim et al., 2017; Brick et al., 2017; Dong et al., 2019; Hertz & Krettenauer, 2016; Shaw et al., 2014), our work empirically demonstrates that people who report a strong moral character can be sensitive to moral contexts, and strategically tailor their moral performances accordingly. In particular, people may apply flexible moral standards consistent with reputation management goals, and display more moral harshness towards others than towards themselves. The findings accord with perspectives that emphasize the prominent role of reputation management in moral psychology (e.g. Jordan et al., 2016; Vonasch et al., 2018), including phenomena such as moral licencing (Blanken et al., 2015) and moral contagion (Kupfer & Giner-Sorolla, 2021).

Second, our work illuminates how exactly reputation management motives moderate the link between self-reported moral character and moral decisions. Beyond previous research suggesting to control for, or eliminate, reputation concerns in moral character measurements (Lee et al., 2008; Paulhus, 1984), these studies demonstrated when, and for whom, moral character precisely predicts moral decisions. When individuals had low reputation management motives, their moral character predicted moral judgements of their own more than others' misdeeds; in contrast, when people were highly motivated to gain a good reputation, moral character only predicted their moral harshness towards others but failed to predict moral decisions for themselves (Study 3 and the Pilot Study in the SM). With the increase of reputation management motives, people who reported a strong (vs. weak) moral character either showed increased hypocrisy by judging others more harshly than themselves (Studies 2 and 3), or showed reduced ‘hypercrisy’ (Lammers, 2012) by judging themselves less harshly than others (the Pilot Study in the SM). Although the specific manifestations of moral double standards varied from moral harshness towards others to moral leniency towards oneself, or both, our findings add more insight to discussions about the effectiveness of moral character measures, by suggesting the importance of taking into account reputation management motives and moral target (e.g. self or others).

Limitations and future directions

We employed diverse samples and methods to test the reputation management account of moral character; however, at least two important limitations should be noted, respectively, related to the self-reported nature of our moral character and reputation management motives measures.

First, although our findings showed a positive relationship between moral character and moral double standards, we may not fully differentiate the ‘genuine’ and ‘reputation managing’ parts of self-reported moral character. People may also internalize reputation management as an integral part of ‘genuine’ moral character. 1 In this case, moral character can facilitate socially desirable reactions in a prompt and heuristic way, and better serve the goal to appear moral to others (Everett et al., 2016; Hardy & Van Vugt, 2006; Jordan et al., 2016; Jordan & Rand, 2020). This theorizing implies that self-reported moral character can be strongly and positively correlated with reputation management motives. However, the hypothesized interaction effect between moral character and reputation management motives on moral double standards replicated, regardless of their different correlations across studies (positive and significant in Studies 1 and 3, non-significant in Study 2, and negative and significant in the Pilot Study in the SM; see Table S6 for specifics). To more formally differentiate the roles of actual and postured moral character in behavioural hypocrisy, future research may integrate self- with other-reports of moral character.

Second, we examined reputation management motives as an individual difference variable, and did not manipulate reputation incentives to show its causal effects. As such, self-reported reputation management motives could be influenced by concerns about social approval. For example, some research suggests that people may under-report their actual reputation management motives because pursuing good reputation and high status can be stigmatized (Kim & Pettit, 2015). People may either over- or under-report their reputation management motives, depending on their perception of the motives as socially approved or disapproved.

Relatedly, our findings do not directly elucidate whether people who display moral double standards (1) genuinely believe such behaviours as morally acceptable, or (2) consciously use them as a reputation management strategy. For example, although high moral character and reputation management motives were associated with stringent moral standards on others across our studies, their relation with lenient moral standards on the self seemed to only apply to moral judgements but not to actual behaviours (Study 3 and its Pilot Study in the SM). The extent to which self-reported moral behaviours reflected actual behaviours or its strategic self-presentation was also unverifiable (Study 1). However, comparisons between different studies may provide tentative evidence on people's conscious and strategic display of moral double standards as a reputation management strategy. People who self-reported high (vs. low) moral character and reputation management motives judged themselves more leniently only in relatively anonymous settings (Study 2) but no more leniently with the presence of a third-party interviewer (Study 1) or observer (Study 3 and its Pilot Study in the SM). Future research may explore the mechanisms of moral double standards in different reputation contexts, and examine moral character and reputation management motives as antecedents to behavioural forms of moral hypocrisy (e.g. saying one thing and doing another; Dong et al., 2019; Effron et al., 2018).

Thus, overall, our results suggest that intelligence is relatively unrelated to whether someone is a kind and moral person

Anglim, Jeromy, Patrick D. Dunlop, Serena Wee, Sharon Horwood, Joshua K. Wood, and Andrew Marty. 2022. “Personality and Intelligence: A Meta-analysis.” PsyArXiv. November 4. doi:10.1037/bul0000373

Abstract: This study provides a comprehensive assessment of the associations of personality and intelligence. It presents a meta-analysis (N = 162,636, k = 272) of domain, facet, and item-level correlations between personality and intelligence (general, fluid, and crystallized) for the major Big Five and HEXACO hierarchical frameworks of personality: NEO PI-R, Big Five Aspect Scales (BFAS), BFI-2, and HEXACO PI R. It provides the first meta-analysis of personality and intelligence to comprehensively examine (a) facet-level correlations for these hierarchical frameworks of personality, (b) item-level correlations, (c) domain- and facet-level predictive models. Age and sex differences in personality and intelligence, and study-level moderators, are also examined. The study was complemented by four of our own unpublished datasets (N = 26,813) which were used to assess the ability of item-level models to provide generalizable prediction. Results showed that openness (ρ = .20) and neuroticism (ρ = -.09) were the strongest Big Five correlates of intelligence and that openness correlated more with crystallized than fluid intelligence. At the facet-level, traits related to intellectual engagement and unconventionality were more strongly related to intelligence than other openness facets, and sociability and orderliness were negatively correlated with intelligence. Facets of gregariousness and excitement seeking had stronger negative correlations, and openness to aesthetics, feelings, and values had stronger positive correlations with crystallized than fluid intelligence. Facets explained more than twice the variance of domains. Overall, the results provide the most nuanced and robust evidence to date of the relationship between personality and intelligence.

Redheaded women are more sexually active than other women, but it is probably due to their suitors

Redheaded women are more sexually active than other women, but it is probably due to their suitors. Katerina Sykorova et al. Front. Psychol., Nov 3 2022. doi: 10.3389/fpsyg.2022.1000753

Abstract: Women with red hair colour, i.e., 1–9% of female Europeans, tend to be the subject of various stereotypes about their sexually liberated behaviour. The aim of the present case-control study was to explore whether a connection between red hair colour and sexual behaviour really exists using data from 110 women (34% redheaded) and 93 men (22% redheaded). Redheadedness in women, correlated with various traits related to sexual life, namely with higher sexual desire as measured by Revised Sociosexual Orientation Inventory, with higher sexual activity and more sexual partners of the preferred gender over the past year, earlier initiation of sexual life, and higher sexual submissiveness. Structural equation modelling, however, showed that sexual desire of redheaded women mediated neither their higher sexual activity nor their higher number of sexual partners. These results indirectly indicate that the apparently more liberated sexual behaviour in redheaded women could be the consequence of potential mates' frequent attempts to have sex with them. Our results contradicted the three other tested models, specifically the models based on the assumption of different physiology, faster life history strategy, and altered self-perception of redheaded women induced by stereotypes about them. Naturally, the present study cannot say anything about the validity of other potential models that were not subjects of testing.

Keywords: Redheadedness, sexual behaviour, sexual desire, sexual activity, sexual submissiveness, stereotypes, mate selection

Makeup increases attractiveness in male faces

Makeup increases attractiveness in male faces. Carlota Batres, Hannah Robinson. PLoS One, November 3, 2022. https://doi.org/10.1371/journal.pone.0275662

Abstract: Makeup is commonly attributed with increasing attractiveness in female faces, but this effect has not been investigated in male faces. We therefore sought to examine whether the positive effect of makeup on attractiveness can be extended to male faces. Twenty men were photographed facing forward, under constant camera and lighting conditions, with neutral expressions, and closed mouths. Each man was photographed twice: once without any cosmetics applied and another time with subtle cosmetics applied by a professional makeup artist. Two hundred participants then rated those 40 images on attractiveness. The male faces were rated as higher in attractiveness when presented wearing makeup, compared to when presented not wearing makeup. This was true for both male and female raters, and whether analyzing the data using a by-participant or a by-face analysis. These results provide the first empirical evidence that makeup increases attractiveness in male faces. Following work on female faces, future research should examine the effect of makeup on several other traits in male faces. The market for male cosmetics products is growing and evolving and this study serves as an initial step in understanding the effect of makeup on the perceptions of male faces.

Discussion

Evidence from a perceptual study supported the hypothesis that subtle cosmetics would make male faces look more attractive. We found that the same faces were rated as more attractive when they were wearing makeup, compared to when they were not wearing makeup. This effect is in line with previous research done with female faces [1–3].

While the difference between men with and without makeup was statistically significant, the effect size was small. This is in contrast with the research done with female faces which has found a large effect of makeup on attractiveness (e.g., η2 = 0.33 [22]). The small effect we found in male faces is probably due to the fact that we wanted the makeup to appear natural so as to not activate any stereotypes participants may have about male makeup [23]. Regardless, given the effects of attractiveness on real-world outcomes [5, 6], even a small effect can have large consequences.

One point to note from our study is that not all of the faces were found to be more attractive with makeup. Previous research has found that identity has an effect size that is 1.36 times larger than the effect size attributed to makeup [22]. In our study, four out of the 20 faces were not rated as more attractive with makeup and it would be interesting for future research to investigate which types of faces gain the most from makeup applications. Additionally, future research is also needed to investigate what type of makeup increases male attractiveness. In this study, the professional makeup artist used a range of cosmetics on the participants (e.g., concealer, powder) and it would be interesting to limit applications in order to investigate the individual effects of these products.

It would also be interesting to further examine what aspect of the makeup application most greatly influences attractiveness perceptions. In our study, the professional makeup artist was instructed to increase skin homogeneity, decrease facial contrast, and accentuate the bone structure. While we got an overall effect of makeup, we are not able to dissociate which of these factors was the most important. For example, perhaps skin homogeneity is responsible for the entire positive effect, or maybe it is a combination of all three factors.

Lastly, this study only looked at perceptions of attractiveness in male faces. However, there is vast amount of research examining the effects of makeup on several other traits in female faces. For instance, likeability [4], leadership ability [24], trustworthiness [25], confidence [26], earning potential [26], and competence [25]. It would therefore be interesting to also examine the effect of makeup on these traits in male faces.

In contrast to dislike, hate is rooted in seeing the hated target as morally deficient or as violating moral norms

The psychology of hate: Moral concerns differentiate hate from dislike. Clara Pretus, Jennifer L. Ray, Yael Granot, William A. Cunningham, Jay J. Van Bavel. European Journal of Social Psychology, November 3 2022. https://doi.org/10.1002/ejsp.2906

Abstract: We investigated whether any differences in the psychological conceptualization of hate and dislike were simply a matter of degree of negativity (i.e., hate falls on the end of the continuum of dislike) or also morality (i.e., hate is imbued with distinct moral components that distinguish it from dislike). In three lab studies in Canada and the United States, participants reported disliked and hated attitude objects and rated each on dimensions including valence, attitude strength, morality, and emotional content. Quantitative and qualitative measures revealed that hated attitude objects were more negative than disliked attitude objects and associated  with moral beliefs and emotions, even after adjusting for differences in negativity. In Study 4, we analysed the rhetoric on real hate sites and complaint forums and found that the language used on prominent hate websites contained more words related to morality, but not negativity, relative to complaint forums.

6 GENERAL DISCUSSION

In a combination of laboratory studies and a content analysis of real online hate and complaint websites, we found initial evidence that differences in people's conceptualizations of hate and dislike are not only a matter of negativity but also morality. Morality—via both the expression of moral emotions and moral conviction—differentiates hated from disliked attitude objects. Individuals rated hated attitude objects in the lab as more closely connected to morality than disliked or even extremely disliked attitude objects. This distinction still held when adjusting for the relationship between morality and negativity. Further, real websites known by the United States government to be organized hate groups used significantly more moral language in expressing their beliefs as compared with users on complaint forums venting their dislike. Of note, we found an order effect in Study 1 such that differences between hate and dislike were less evident when participants were asked to generate disliked objects first. This suggests that people spontaneously think about objects that are closer to objects they extremely dislike or hate when asked about dislike without an explicit reference to hate.

Regarding the intensity hypothesis, we found mixed evidence for the role of negativity in distinguishing hate expressions from dislike. In Studies 1 and 2, hated attitude objects were rated as more negative than disliked attitude objects, even after controlling for morality, suggesting that both morality and negativity independently contribute to hate. These results are aligned with recent work by Martínez et al. (2022), who found increased ratings in 11 self-reported negative emotions in response to hated compared to disliked targets. However, these authors do not explore differences between hated and extremely disliked objects. We find this to be a relevant comparison in the light of Study 3, where we find that negativity differences between hated and extremely disliked objects vanished after controlling for morality, suggesting that differences in morality accounted for observed differences in negativity. Further, in Study 4, online expressions of hate did not use more negative language than online expressions of dislike. Thus, whereas hate and dislike seem to differ in both intensity and morality, it is possible that hate and extreme dislike differ mainly in the morality dimension. Future studies would benefit from employing scales and statistical techniques that allow researchers to obtain uncorrelated measures of negativity, attitude strength, and morality to better assess the independent contribution of each of these constructs in distinguishing hate from dislike.

Although it seems easy to recognize expressions of hatred when we see them—at Nazi rallies or ethno-cultural genocide—hate is still poorly understood from a scientific perspective. Our studies find that morality is a key ingredient that differentiates the conceptualization of hate from dislike in the minds of many people. These studies offer a springboard for empirical research into the psychology of hate. Centuries of philosophical theory have laid the groundwork for more rigorous empirical investigation. For example, our review of the literature raised the possibility that hate is motivational. Rempel and Burris (2005) suggested that we will ignore disliked objects but will wish to harm hated ones. In line with this, people feel more inclined to engage in attack-oriented behaviours when they experience hate versus dislike (Martínez et al., 2022). Further, while the present laboratory studies manipulated the type of attitude object generated to test differences between groups, further research could reverse the relationship between our independent and dependent variables. An important test of the connection between hate and morality would be to determine if experimentally inducing moral emotions could create hate in a laboratory setting. However, the ethics of doing so must be carefully considered.

One potential alternative explanation is that our instructions to generate hated versus disliked attitude objects elicited different classes of attitude objects (e.g., people and groups vs. concepts and beliefs). However, when we explicitly instructed people to generate different classes of attitude objects, we found that the difference between hate versus dislike was robust across these classes. It is also possible that hated versus disliked attitude objects differed systematically in level of abstraction. Some work has found that people more readily apply their moral principles to the psychologically distant (Eyal et al., 2008). Perhaps hated attitude objects are more psychologically distant or higher in abstraction? Another alternative explanation for our findings is that disliked versus hated objects do not need to have an actual antecedent: whereas people may not know why they dislike something, hatred may be more readily associated with a specific experience, making it easier to link to morality. Future research should address these possibilities.

An additional reason we believe the differences between hate and dislike extend beyond these issues is our study of online hate groups. The websites we explored did not require users to list attitudes objects they hated. In fact, many online hate groups actively disavow their categorization as “hate groups” and the content of their websites often focused on their core values (e.g., “…teaches lessons of morality and nobility, to walk as a proud White individual in a world where being White is now considered wrong”). Their websites were identified as hate groups by third parties. Our analyses nevertheless found much higher expressions of morality on these hate websites as compared to complaint forums, both about objects and about corporate groups. Together with our lab experiments, this gives us confidence that the difference between hate and dislike goes beyond simple semantics.

6.1 Hate as emotion

The current research relied on self-reports and content coding, which provides a modest scope for understanding the rich affective experience of hate. At present, it is impossible to determine if hate causes a feeling state or if labelling an experience as hate is a consequence of an emotional experience (or both). Importantly, our use of the term hate does not imply that it is a basic emotion. Our belief is that the psychological state we colloquially associate with hatred is actively constructed like other complex emotional states rather than a natural kind (see Barrett, 2006). While this is beyond the scope of the present work, these are important distinctions that should be examined in future research on the psychology of hate.

On a related note, while the conceptualization of anger, contempt, and disgust as distinctively moral emotions continues to receive empirical support (see for instance Steiger & Reyna, 2017), other scholars have challenged this view, arguing that disgust may have a broader role beyond morality or that anger can be triggered by other moral transgressions beyond autonomy (see Lomas, 2019). Thus, our results on the differences in moral emotions between hated and disliked attitude objects should be treated with caution: whereas these emotions may be necessary for hatred to arise, they may not be sufficient. Whereas higher ratings in anger, contempt, and disgust were to be expected in the hatred versus dislike condition, they should not be taken by themselves as unequivocal proof of the association between hatred and morality.

The results of the present research might, eventually, be fruitfully applied to psychological or behaviour interventions against hate. For instance, work on relations between Israelis and Palestinians suggests that hatred toward the out-group differs from anger in terms of profiling the out-group as evil and intentionally causing harm (Halperin, 2008; see also Parker & Janoff-Bulman, 2013). Yet such conflict may be ameliorated and peace proposals more likely to be adopted when the out-group is willing to compromise sacred values—rather than economic concessions (Ginges et al., 2007). Thus, acknowledging and leveraging the moral concerns associated with hatred may provide an important avenue for addressing intergroup (as well as interpersonal) conflict. We urge research in these areas to continue this line of inquiry in the hopes of designing and testing interventions to alleviate social conflict.

Finally, we note that the samples of our first three studies were undergraduate students from Canada and the US. This poses a limitation in terms of the generalizability of the findings of these studies, which have been drawn from western educated individuals from industrialized, rich, democratic societies (WEIRD, see Henrich et al., 2010). We attempted to overcome this limitation in Study 4, where we obtained samples from a more ecological environment (websites with English-speaking audiences). Because different cultures could have different conceptualizations of hate and dislike, future research should further address this constraint by including cross-national representative samples.

Looking cross a large dataset of extinct and extant mammalian skulls, the rate of evolutionary change peaked around the time of the Cretaceous-Paleogene boundary and has general tapered off since then

Attenuated evolution of mammals through the Cenozoic. Anjali Goswami et al. Science, Oct 27 2022, Vol 378, Issue 6618, pp. 377-383. DOI: 10.1126/science.abm7525

Becoming diverse: Mammals have the greatest degree of morphological variation among vertebrate classes, ranging from giant whales to the tiny bumblebee bat. How they evolved this level of variation has been a persistent question, with much debate being centered around the timing and tempo of evolutionary change. Goswami et al. looked across a large dataset of extinct and extant mammalian skulls and found that the rate of evolutionary change peaked around the time of the Cretaceous-Paleogene boundary and has general tapered off since then (see the Perspective by Santana and Grossnickle). Certain lifestyles, such as aquatic habitats or herbivory, led to faster change, whereas in some species such as rodents, morphological change appeared to be decoupled from taxonomic diversification. —SNV

Abstract: The Cenozoic diversification of placental mammals is the archetypal adaptive radiation. Yet, discrepancies between molecular divergence estimates and the fossil record fuel ongoing debate around the timing, tempo, and drivers of this radiation. Analysis of a three-dimensional skull dataset for living and extinct placental mammals demonstrates that evolutionary rates peak early and attenuate quickly. This long-term decline in tempo is punctuated by bursts of innovation that decreased in amplitude over the past 66 million years. Social, precocial, aquatic, and herbivorous species evolve fastest, especially whales, elephants, sirenians, and extinct ungulates. Slow rates in rodents and bats indicate dissociation of taxonomic and morphological diversification. Frustratingly, highly similar ancestral shape estimates for placental mammal superorders suggest that their earliest representatives may continue to elude unequivocal identification.

 

Empirical Macroeconomics and DSGE Modeling in Statistical Perspective: Dismal forecasting errors + swapping data slightly impairs the model (and in 37% of cases the permutation of data make the model better)

Empirical Macroeconomics and DSGE Modeling in Statistical Perspective. Daniel J. McDonald, Cosma Rohilla Shalizi. Oct 31 2022. https://arxiv.org/abs/2210.16224

Abstract: Dynamic stochastic general equilibrium (DSGE) models have been an ubiquitous, and controversial, part of macroeconomics for decades. In this paper, we approach DSGEs purely as statistical models. We do this by applying two common model validation checks to the canonical Smets and Wouters 2007 DSGE: (1) we simulate the model and see how well it can be estimated from its own simulation output, and (2) we see how well it can seem to fit nonsense data. We find that (1) even with centuries' worth of data, the model remains poorly estimated, and (2) when we swap series at random, so that (e.g.) what the model gets as the inflation rate is really hours worked, what it gets as hours worked is really investment, etc., the fit is often only slightly impaired, and in a large percentage of cases actually improves (even out of sample). Taken together, these findings cast serious doubt on the meaningfulness of parameter estimates for this DSGE, and on whether this specification represents anything structural about the economy. Constructively, our approaches can be used for model validation by anyone working with macroeconomic time series.

h/t Alex Tabarrok A Big and Embarrassing Challenge to DSGE Models Nov 3 2022 https://marginalrevolution.com/marginalrevolution/2022/11/a-big-and-embarrassing-challenge-to-dsge-models.html:

[...]

"If we take our estimated model and simulate several centuries of data from it, all in the stationary regime, and then re-estimate the model from the simulation, the results are disturbing. Forecasting error remains dismal and shrinks very slowly with the size of the data. Much the same is true of parameter estimates, with the important exception that many of the parameter estimates seem to be stuck around values which differ from the ones used to generate the data. These ill-behaved parameters include not just shock variances and autocorrelations, but also the “deep” ones whose presence is supposed to distinguish a micro-founded DSGE from mere time-series analysis or reduced-form regressions. All this happens in simulations where the model specification is correct, where the parameters are constant, and where the estimation can make use of centuries of stationary data, far more than will ever be available for the actual macroeconomy."

Now that is bad enough but I suppose one might argue that this is telling us something important about the world. Maybe the model is fine, it's just a sad fact that we can't uncover the true parameters even when we know the true model. Maybe but it gets worse. Much worse.

McDonald and Shalizi then swap variables and feed the model wages as if it were output and consumption as if it were wages and so forth. Now this should surely distort the model completely and produce nonsense. Right?

"If we randomly re-label the macroeconomic time series and feed them into the DSGE, the results are no more comforting. Much of the time we get a model which predicts the (permuted) data better than the model predicts the unpermuted data. Even if one disdains forecasting as end in itself, it is hard to see how this is at all compatible with a model capturing something — anything — essential about the structure of the economy. Perhaps even more disturbing, many of the parameters of the model are essentially unchanged under permutation, including “deep” parameters supposedly representing tastes, technologies and institutions."