Skip to main content
NCBI home page
Sage Choice logoLink to Sage Choice
. 2025 Jan 2;30(11):3153–3168. doi: 10.1177/13591053241306564

US state death rates: Structural equation modeling of Big Five personality, socioeconomic status, and health risk factors

Stewart JH McCann 1,✉
PMCID: PMC12433523  PMID: 39745003

Abstract

Structural equation modeling (SEM) tested the plausibility of a causal model with neuroticism, openness to experience, socioeconomic status (SES), and race as predictors of a composite of six health risks and age-adjusted all-cause mortality in 2020 using the 48 contiguous American states as analytic units. In the final model, neuroticism, openness, and SES accounted for 80% of the health risk composite variance. These three variables and composite health risk accounted for 85% of the death rate variance. Neuroticism, openness, and SES had direct impacts on the health risk composite and indirect impacts on death rates through the health risk composite. SES and composite health risk also had direct impacts on death rates. Spatial autocorrelation and multicollinearity were not problematic. These SEM results underline the importance of state resident personality and SES in this context and support the plausibility of the speculation that the demonstrated relations may be causal in nature.

Keywords: Big Five personality, biopsychosocial model, exercise, fruit and vegetable consumption, obesity, physical inactivity, sleep, smoking, socioeconomic status, US state death rates

Does personality predict premature mortality? Empirical research with individuals as the analytic units strongly suggests that this is so. Big Five personality dimensions (e.g. John and Srivastava, 1999) have been related to all-cause mortality (e.g. Graham et al., 2017). As well, health risk factors such as lack of exercise, physical inactivity, lack of fruit and vegetable consumption, smoking, short sleep duration, and obesity predict death at an earlier age (e.g. Gale et al., 2017; Graham et al., 2017; Liu et al., 2021; Saint-Maurice et al., 2024; Silva et al., 2020; Visaria and Setoguchi, 2023). These risk factors also relate to the Big Five personality dimensions (e.g. Bogg and Roberts, 2004; Caille et al., 2024; Garrido et al., 2018; Zhang et al., 2024). Therefore, these findings also suggest the possibility of an indirect path from personality through health risk factors to age-adjusted all-cause mortality.

The “Big Five” is the most broadly accepted contemporary personality model. It includes five dispositional variables on which to differentiate persons within the nonpathological range of functioning—openness to experience, conscientiousness, extraversion, agreeableness, and neuroticism (e.g. John and Srivastava, 1999). Each dimension consists of six facets that relate to a number of trait descriptors. For example, the neuroticism facets are anxiety, angry hostility, depression, self-consciousness, impulsiveness, and vulnerability. They are associated with various trait adjectives such as being tense, irritable, discontented, shy, moody, and not self confident. In a similar manner, for example, ideas and actions facets of openness to experience are associated with being curious and having wide interests, competence and order facets of conscientiousness are associated with being efficient and organized, activity and warmth facets of extraversion are associated with being energetic and outgoing, and trust and compliance facets of agreeableness are associated with being forgiving and not stubborn (e.g. John and Srivastava, 1999). Each personality dimension also has a strong genetic component (e.g. Vukasović and Bratko, 2015).

Socioeconomic status (SES), largely based on income and education variables, generally is endorsed as a valid predictor of premature death. Individual-level research has firmly established that lower SES is associated with greater all-cause mortality (e.g. Glei et al., 2022). As well, lower SES has been associated with insufficient exercise, physical inactivity, lack of fruit and vegetable consumption, smoking, short sleep duration, and obesity (e.g. Ji et al., 2023; Murray et al., 2012; Stringhini et al., 2011). This also suggests an indirect path from SES through health risk factors to all-cause mortality.

Race also may predict mortality and health risk factors at the state level. For example, individual-level research has shown that being Black is associated with higher all-cause mortality (Gilmore et al., 2019) and with lack of exercise (e.g. Hersch, 1996), physical inactivity (e.g. Saffer et al., 2013), lack of fruit and vegetable consumption (e.g. Dubowitz et al., 2008), smoking (e.g. Persoskie and Leyva, 2015), short sleep duration (e.g. Nunes et al., 2008), and obesity (e.g. Lincoln et al., 2014). Perhaps there is a direct path to state death rates and an indirect path through health risk factors with Black population percent representing race at the state level.

Among the many risk factors for health and mortality, the six chosen for the present study were selected because they primarily reflect lifestyle. Other risk factors such as high blood pressure and high cholesterol were not included because they are not so directly driven by lifestyle choices. These six variables also have been used in many other studies and are appropriately documented for each state in annual reports (e.g. America’s Health Rankings, 2021).

The preceding research results linking all-cause mortality to Big Five personality, SES, and race are correlation-based, thus preventing causal inferences. However, there are reasons to believe the associations may be grounded in causal connections. For example, Kern and Friedman (2011) noted that personality can have an impact on health outcomes because it “influences the habits we form, the behaviors we engage in, the relationships we develop, our appraisals and experience of stressful challenges, the situations we commonly choose, the reactions we invoke in others, and the lifelong pathways that we follow” (p. 84), and of course, health outcomes clearly include health risks and ultimately death. Roberts et al. (2007) too suggested that personality may have causal impacts on mortality because personality differences may be related to pathogenetic mechanisms and because personality factors may foster differential reactions to illness. These somewhat interrelated paths of Kern and Friedman (2011) and Roberts et al. (2007) that eventually lead to death appear to stem causally from dispositional differences.

Relations of SES and race to mortality also seem rooted in causal relations. For example, causal mechanisms have been speculated to occur through SES-related variables such as diet quality, smoking, physical activity, and obesity (e.g. Laine et al., 2020). A review by Roberts et al. (2007) based on prospective longitudinal studies showed that the magnitude of SES effects on mortality essentially equaled those for personality. Regarding race, it is possible that states with higher Black population percents may have higher levels of risk factors and death rates caused by racial tendencies demonstrated at the individual level.

Several individual-level studies have found that relations of Big Five personality factors, SES, and race to all-cause mortality may be mediated through health risks (e.g. Grogan et al., 2024; Kim et al., 2018; Laine et al., 2020; Petrovic et al., 2018; Ploubidis and Grundy, 2009; Puka et al., 2023). For example, Grogan et al. (2024) conducted a literature review and reported that smoking consistently mediated the relation between neuroticism and mortality, and Laine et al. (2020) found that the relation of SES to all-cause mortality was mediated in part by smoking, physical activity, and obesity. Grogan et al. (2024) concluded that “the findings to date clearly show that mediation-based analysis has a key role to play in developing our understanding of the complex pathways that connect personality to mortality” (p. 222). Health risk factors also may mediate effects of SES (e.g. Kim et al., 2018) and race on mortality (e.g. Puka et al., 2023).

The present structural equation modeling (SEM) study uses American states as the cases. It is conducted from the geographical psychology perspective (e.g. Rentfrow et al., 2008; Rentfrow and Jokela, 2016). A major premise is that a geographic area’s aggregate position on an individual difference dimension relates to the level in that area of the behavioral and psychological tendencies associated with that individual difference dimension. Therefore, the approach provides a rationale for the possibility of prudently interpreting relations found at the aggregate level as stemming from processes occurring at the individual level. However, the approach recognizes that empirical support at each level eventually must buttress such speculations. Caution is necessary to avoid the pitfalls of the “compositional fallacy” (Pettigrew, 1997) and “ecological fallacy” (Robinson, 1950) which concern making mere assumptions regarding the validity of cross-level extrapolations.

State-level research from the geographical psychology perspective is beneficial. It can produce variables aggregated from individuals regarding psychological traits and processes that can be directly related to other state-level data produced by other government and private bodies. As well, geographical psychology with states as analytic units can present opportunities to evaluate the applicability of individual-level psychological findings at the state level, provide empirical findings with relevance for federal and state policy making, and furnish possibilities for evaluation of the effectiveness of public health policies and promotions (e.g. Chen et al., 2020).

The primary goal of the present research is to determine the extent to which state death rates can be predicted by selected state-level personality, sociodemographic, and health risk factors. The SEM analysis also provides a framework for judging the plausibility of expected direct and indirect causal paths from personality and demographic variables to health risk and to death rates. From the geographical psychology perspective, it is reasonable to speculate that such state-level relations may originate from parallel individual-level relations. Therefore, the present research also can be framed as a feasibility study to determine whether state-level relations within the current context resemble existing individual-level findings.

The present study

The current state-level study uses SEM to test the plausibility of a causal model in which Big Five personality, SES, and race are expected to have direct effects on all-cause mortality as well as indirect effects on mortality through a composite of six major health risks. It was strongly expected that higher neuroticism, lower openness, and lower conscientiousness would be associated directly with higher composite health risk. It was less confidently expected that higher neuroticism, lower openness, and lower conscientiousness would be associated directly with higher death rates, but strongly expected that higher neuroticism, lower openness, and lower conscientiousness would be associated indirectly through health risks with higher death rates. Regarding SES, there were prime expectations that lower SES would be associated directly with higher composite health risk, that lower SES would be associated directly with death rates, and that lower SES would be associated indirectly through health risks with higher death rates. Regarding race, there were weak expectations that higher Black population percent would be associated directly with higher composite health risk, would be associated directly with higher death rates, and would be associated indirectly through health risks with higher death rates. Of course, it also was strongly expected that higher composite health risk would be associated directly with higher death rates.

Method

Measures

Death rate

Kochanek et al. (2023) provided the final age-adjusted death rate per 100,000 population in 2020 for each of the 50 American states. The state death rates were based on all races, both sexes, and all ages.

Big Five personality

Rentfrow et al. (2008) used the responses of 619,397 residents to the 44-item Big Five Inventory (John and Srivastava, 1999) administered online between December of 1999 and January of 2005 to create Big Five z scores for each of the 50 American states. They provided evidence that the sample generally was representative of the national population. They also indicated that they drew respondents from each state in direct proportion to the 2000 census figures regarding population and racial-ethnic composition, but the respondents were somewhat younger and more middle class than the state census estimates show. Rentfrow et al. (2008) also reported that each personality dimension had high temporal reliability based on two subsamples, a high mean inter-subsample correlation based on three random subsamples, a high congruence coefficient between state-level and individual-level factor structure, and high reliabilities with mean Cronbach alphas of 0.81 at the individual level and 0.89 at the state level.

SES

FRED (2024) provided the per capita personal income for each state in 2020. The US Census Bureau (2023) furnished the official state poverty rates for 2019 based on the American Community Survey, which it did not administer in 2020. The US Census Bureau (2024) supplied the percent in each state 25 years of age and over who had at least high school graduation and the percent who had at least a bachelor’s degree based on the American Community Survey for 2019. The sign was reversed for poverty rates (i.e. multiplied by −1). The standardized results of the sums of the four variables in z score form yielded an SES composite with a Cronbach alpha of 0.87.

Exercise

America’s Health Rankings (2021), based on 2019 data from the Behavioral Risk Factor Surveillance System (BRFSS), provided the percent of adults in each state who reported exercising enough to meet federal expectations. Those guidelines require “150 minutes of moderate or 75 minutes of vigorous aerobic activity and two days of muscle strengthening per week” (p. 99) within the previous 30 days. The state rates were reversed (i.e. multiplied by −1) for the present study.

Physical inactivity

America’s Health Rankings (2021) provided the percent in each state of “adults who reported doing no physical activity or exercise other than their regular job in the past 30 days” (p. 99) in 2020.

Fruit and vegetable consumption

America’s Health Rankings (2021) furnished the percent in each state of “adults who reported consuming two or more fruits and three or more vegetables daily” (p. 99) in 2019.

Smoking

America’s Health Rankings (2021) contributed the percent in 2020 in each state of “adults who reported smoking at least 100 cigarettes in their lifetime and currently smoke daily or some days” (p. 99).

Short sleep duration

America’s Health Rankings (2021) provided the age-adjusted percent of adults in each state who reported sleeping less than 7 hours per day in 2020. State residents responded to the following question asked through the BRFSS: “On average, how many hours of sleep do you get in a 24-hour period?” (p. 99).

Obesity

America’s Health Rankings (2021) also supplied the percent of adults 18 years and over in each state with a body mass index (BMI) of 30 or greater based on reported weight and height in 2020.

Race

KFF (2024) provided the Black population percent in each state in 2019. Figures were not available for 2020.

Health risk composite

A health risk composite was created based on the previously described exercise, physical inactivity, fruit and vegetable consumption, smoking, short sleep duration, and obesity variables using the data for the 50 states. Before computation, the exercise and fruit and vegetable consumption variables were reversed (i.e. multiplied by −1). Each of these six variables then was converted to z scores, the variables were summed, and the resulting score was standardized. This health risk composite variable has a Cronbach alpha of 0.91.

Ethical considerations

All ethical standards were followed in all phases of this research. Formal ethics approval was not required because this article does not contain any studies with human participants performed by the author.

Analytic strategy

Researchers employing SEM as a research tool do not discover or derive causal links from the SEM results. Instead, the SEM model tests whether the relations found in the data analyzed support the plausibility of speculated causal assumptions stemming “from the research design, prior studies, scientific knowledge, logical arguments, temporal priorities, and other evidence that the researcher can marshal in support of them” (Bollen and Pearl, 2013: 309). Fit statistics of SEM models serve as initial empirical tests of the plausibility of those assumptions based on the various elements of prior knowledge. Successfully “fitting the data does not ‘prove’ the causal assumptions, but it makes them tentatively more plausible” (Bollen and Pearl, 2013: 309).

The generated paths in the present SEM analysis begin with Big Five personality, SES, and race as primary exogenous variables with initially equal status in the potential to account for variance in health risks and mortality rates. It is important to know whether and to what extent common health risks may mediate the impacts of Big Five personality variables, SES, and race on all-cause mortality, and to do this without excessive statistical control of other variables that may be only marginally involved in the relations to which the present research has intentionally limited its analysis.

The present research focuses on state age-adjusted all-cause mortality rates in 2020. The AMOS SEM extension of SPSS 29 was used to determine the direct and indirect pathways to those mortality rates from SES, Big Five variables, and race using the health risk composite as the mediator. Only Big Five personality variables having a significant Pearson correlation with those mortality rates were included in the SEM analysis. Potential multicollinearity and spatial autocorrelation issues also were addressed.

Results

Preliminary considerations led to the exclusion of Alaska and Hawaii, the two non-contiguous states of the USA. The main reason for excluding Alaska and Hawaii, which are quite distant from the other 48 states and from each other, was to facilitate the spatial autocorrelation analysis which requires states with borders touching other states. However, it also may be assumed that several apparent differences between these two states and the others might jointly contribute to distorted relations involving death rates and the potential predictors in the present study. Therefore, a screening for multivariate outliers was conducted using the SPSS Mahalanobis distance approach with the full 50 states. With death rate, each of the Big Five, SES, Black population percent, and the health risk composite in the variable pool, Alaska had the lowest Mahalanobis value of 27.35 (p = 0.001) and Hawaii had the second lowest value of 22.97 (p = 0.006). These results indicate that Alaska and Hawaii were multivariate outliers. This effectively constitutes another valid reason for basing the main analyses on only the 48 contiguous states.

Table 1 displays statistical means, standard deviations, and Pearson correlations based on the 48 adjoining states. Death rate significantly correlated −0.41 with openness, 0.43 with neuroticism, −0.81 with SES, between 0.58 and 0.83 with each of the six health risk variables, 0.50 with Black population percent, and 0.91 with the health risk composite.

Table 1.

Means, standard deviations, and Pearson correlations for the 48 contiguous states.

Variable M SD 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14.
1. Death rate 861.51 104.93
2. Openness to experience −0.00 0.82 −0.41**
3. Conscientiousness 0.11 0.89 0.22 −0.11
4. Extraversion 0.01 0.97 0.16 −0.60*** 0.36*
5. Agreeableness 0.13 0.74 0.19 −0.34* 0.59*** 0.54***
6. Neuroticism 0.05 1.01 0.43*** 0.07 −0.44** −0.21 −0.27
7. SES −0.03 1.01 −0.81*** 0.26 −0.40** −0.12 −0.17 −0.25
8. Lack of exercise −22.48 3.27 0.75*** −0.43** 0.06 0.20 0.13 0.38** −0.59***
9. Physical inactivity 23.27 3.70 0.81*** −0.34* 0.12 0.17 0.01 0.54*** −0.80*** 0.72***
10. Lack of fruits/vegetables −7.90 1.76 0.58*** −0.49*** 0.32* 0.33* 0.20 −0.04 −0.51*** 0.63*** 0.51***
11. Smoking 15.44 3.30 0.83*** −0.54*** 0.11 0.17 0.16 0.36* −0.67*** 0.61*** 0.70*** 0.55***
12. Short sleep duration 32.33 2.91 0.72*** −0.11 −0.05 −0.06 −0.14 0.58*** −0.64*** 0.59*** 0.78*** 0.42** 0.53***
13. Obesity 32.38 3.91 0.83*** −0.58*** 0.22 0.27 0.32* 0.31* −0.74*** 0.70*** 0.74*** 0.53*** 0.75*** 0.62***
14. Black population percent 10.60 9.49 0.50*** −0.03 0.11 −0.01 0.14 0.34** −0.41** 0.28 0.51*** 0.17 0.21 0.56*** 0.47***
15. Health risk composite 0.02 1.02 0.91*** −0.50*** 0.16 0.22 0.14 0.42** −0.79*** 0.85*** 0.89*** 0.74*** 0.84*** 0.79*** 0.87*** 0.44***
Open in a new tab
*

p < 0.05. **p < 0.01. ***p < 0.05.

SEM models for the 48 contiguous states

Model 1

Figure 1 displays the standardized regression coefficients (βs) and the variance accounted for in the health risk composite and state death rates in the initial saturated model (Model 1). The β path coefficients were nonsignificant between death rate and openness (−0.06), neuroticism (0.08), and race (0.11). The path coefficients also were nonsignificant between the health risk composite and race (0.10). The R2 was 0.81 for the health risk composite and 0.86 for death rate.

Figure 1.

Figure 1.

Open in a new tab

Saturated SEM Model 1 showing standardized path coefficients (βs) and R2 for SES, openness to experience, neuroticism, health risk composite, and death rate.

Model 2

A reduced or overidentified Model 2 was developed with the four nonsignificant paths in Model 1 dropped (see Figure 2). Model 2 had an acceptable CMIN of 2.037 (df = 2; p = 0.361) indicating that the fit was not significantly poorer than that of Model 1. Other statistics showed a high goodness of fit of Model 2 to the data (e.g. RMSEA = 0.020; CFI = 1.000; Tucker-Lewis Index = 0.999; GFI = 0.983; SRMR = 0.0175). The β coefficient was significant and in the expected direction for the five direct paths: SES to the health risk composite (−0.63), openness to the health risk composite (−0.36), neuroticism to the health risk composite (0.29), SES to death rate (−0.29), and the health risk composite to death rate (0.71). The variance accounted for in Model 2 was 80% in the health risk composite and 85% in death rate, close to the values in Model 1.

Figure 2.

Figure 2.

Open in a new tab

SEM Model 2 showing standardized path coefficients (βs) and R2 for SES, openness to experience, neuroticism, health risk composite, and death rate.

Table 2 shows the standardized direct, indirect, and total effects in Model 2. Of course, the standardized direct effects in Table 2 are those previously reported for the direct paths in Model 2 (see Figure 2). The results for the standardized indirect effects showed that openness had an indirect effect through the health risk composite on death rate (−0.26), neuroticism had an indirect effect through the health risk composite on death rate (0.21), and SES had an indirect effect through the health risk composite on death rate (−0.45). The test of joint significance (Leth-Steensen and Gallitto, 2016) states that an indirect path is significant if the separate paths that constitute a compound path in an indirect effect are all significant. Therefore, these three indirect paths in Model 2 are significant. Standardized total effects simply are the sums of the standardized direct and indirect effects through the various paths in Model 2.

Table 2.

SEM standardized direct, indirect, and total effects on the health risk composite and death rate in Model 2.

Standardized effects Criterion Openness to experience Neuroticism SES Health risk composite
Direct Health risk composite −0.36 0.29 −0.63 0.00
Death rate 0.00 0.00 −0.25 0.71
Indirect Health risk composite 0.00 0.00 0.00 0.00
Death rate −0.26 0.21 −0.45 0.00
Total Health risk composite −0.36 0.29 −0.63 0.00
Death rate −0.26 0.21 −0.70 0.71
Open in a new tab

Spatial autocorrelation and multicollinearity

SPSS 29 AMOS does not provide output regarding spatial autocorrelation. Therefore, the spatial autocorrelation analysis reported here is based on simultaneous ordinary least squares (OLS) multiple regression equations with death rate and the health risk composite as the criteria and their respective predictors taken from Model 1 and Model 2. For each of the four OLS equations, the global Moran’s I test identified spatial autocorrelation of residuals (Anselin, 2003) using the R haven and spdep statistical packages (R, 2022). To facilitate the analysis, the six variables were converted to z scores. A 48 × 48 Queen’s binary neighborhood matrix also was created that dummy coded each state touching another state at any point as a neighbor.

The four OLS equations first were tested for multicollinearity among predictors. All 16 VIFs were well below the common threshold of 10. Only one of 5.28 and one of 5.08 were 5.00 or above, and 13 were under 1.80. These results indicate that multicollinearity was not an issue with death rate or the health risk composite as the outcome variable.

With death age as the criterion, Moran’s I statistic standard deviate was 1.58 (p = .057) for Model 1 and 1.32 (p = .093) for Model 2. With the health risk composite as the criterion, Moran’s I statistic standard deviate was 1.63 (p = 0.052) for Model 1 and 1.30 (p = .096) for Model 2. These results indicate an absence of statistically significant spatial autocorrelation.

Discussion

The final SEM model attests to the plausibility of a causal model in which higher neuroticism, lower openness to experience, and lower SES directly promote a composite of health risk factors that includes lack of exercise, physical inactivity, lack of fruit and vegetable consumption, smoking, obesity, and lack of sleep. A higher level of health risk factors then directly generates higher chances of premature age-adjusted all-cause mortality. Lower SES also has a direct impact on such mortality. This final model did not include either a direct positive impact of neuroticism on mortality or a direct negative impact of openness on mortality as was expected for the initial SEM model. However, neuroticism had an indirect positive impact while openness and SES had indirect negative impacts on mortality through the composite of health risk factors. Expectations regarding conscientiousness and race were not supported.

The present research employed American states rather than individuals as the analytic units. Therefore, it is important to compare these findings—at least for the direct effects—to previous individual-level findings to determine whether there is sufficient empirical evidence to justify speculative cross-level extrapolation. Generally, SES, openness to experience, and neuroticism have been related to health risk factors in the manner found in the present study (e.g. Caille et al., 2024; Garrido et al., 2018; Ji et al., 2023; Murray et al., 2012; Stringhini et al., 2011; Zhang et al., 2024). There also is individual-level evidence that each of the six risk factors and SES generally relate to all-cause mortality in the same way as in the present study (e.g. Glei et al., 2022; Liu et al., 2021; Saint-Maurice et al., 2024; Silva et al., 2020; Visaria and Setoguchi, 2023). Therefore, ample individual-level research demonstrates the relations of SES, openness, and neuroticism to health risk factors, and the relations of the six health risk factors and SES to all-cause mortality. These individual-level results provide corresponding corroborative evidence for the present state-level findings.

Neuroticism facets are associated with various trait adjectives such as being tense, irritable, discontented, shy, moody, and not self confident. In a similar manner, openness facets are associated with trait descriptors such as being curious, imaginative, artistic, excitable, and unconventional, and having wide interests. Accordingly, it would seem logical to assume that a neurotic person should be more likely to smoke, sleep poorly, make less than optimal food choices, exercise less, and engage in more sedentary behavior. It also would seem logical that an open person might be less likely to manifest such health risks. For example, being curious and having wide interests might make a person more open to engage in exercise, physical activity, diet modification, lifestyle modification to improve sleep, and efforts to refrain from smoking based on openness to pursuing and assimilating relevant new knowledge about these activities. Higher neuroticism or lower openness in a state also may contribute somewhat to relations between those two personality variables and health risk factors through social norms and expectations regarding health risks in part created and maintained through the presence of higher neuroticism and lower openness among state residents.

Although the current state-level Pearson correlations show that all-cause mortality related negatively to openness and positively to neuroticism in a statistically significant manner, neither Big Five variable retained a significant connection via a direct path to mortality in the SEM context. Many individual-level researchers have reported that higher neuroticism and lower openness are associated with a greater risk of all-cause mortality (e.g. Ferguson and Bibby, 2012; Grogan et al., 2024). The present significant Pearson correlations coupled with the nonsignificant SEM results suggests that the simultaneous control of other relations in the initial SEM model served to effectively reduce the direct relations of neuroticism and openness to all-cause mortality to nonsignificant levels. In fact, this was demonstrated in an unreported exploratory SEM model. Excluding the health risk composite yielded significant βs of −0.68 for SES, −0.26 for openness, and 0.27 for neuroticism.

The present state-level Pearson correlations of conscientiousness, extraversion, and agreeableness to all-cause mortality were not significant. Furthermore, none of these three significantly correlated with the health risk composite. In past individual-level research, these three personality variables have shown mixed relations to mortality (e.g. Bratt et al., 2016; Graham et al., 2017; Jokela et al., 2013). Given that the individual-level results are quite inconsistent, it may not be so unusual that statistically significant relations did not materialize here between these three personality variables and all-cause mortality.

However, of these three other Big Five variables, relations involving conscientiousness were most surprising in the present results. Conscientiousness correlated a nonsignificant 0.22 with both the health risk composite and death rate. These two positive correlations were unexpected and run counter to the direction of most research findings by others (e.g., Bogg and Roberts, 2004). Furthermore, with the 48 contiguous states as the analytic units, conscientiousness correlated a significant −0.40 with SES which also is in the unexpected direction (e.g. Conger et al., 2021). An exploratory partial correlation with SES controlled yielded nonsignificant negative correlations of −0.28 and −0.20 between conscientiousness and composite health risk and death rate, respectively. Why the existing state distribution patterns for conscientiousness and SES are in this way anomalous compared to most individual-level results reported by other researchers is unknown.

Although Black population percent correlated 0.44 with composite health risk and 0.50 with death rate, it was not significantly related to either criterion in the SEM analysis. This lack of relations may have occurred in part because of the correlation of −0.41 between race and SES. An unreported exploratory partial correlation with SES as the control yielded a nonsignificant correlation of 0.21 between race and composite risk, while another exploratory partial correlation with composite risk controlled produced a nonsignificant correlation of 0.28 between race and death age. These relational patterns evidently contributed to the elimination of Black population percent in the SEM analysis.

Each of the six health risks used in this study to create the composite seem to represent the opposite of what throughout most of human history would have been considered assets for coping with the demands of a successful existence—that is, being relatively physically fit, having an adequate diet, getting enough sleep, and avoiding the intake of foreign substances that are neither edible nor potable and may be detrimental to health. Therefore, it does not seem so surprising that geographical environments such as American states in which these health maintaining behaviors are practiced to a lesser degree also tend to have higher age-adjusted all-cause death rates. It also does not seem so surprising that American states with lower SES, higher neuroticism, and lower openness tend to participate less in such health maintaining behaviors and that ultimately leads to a greater chance of dying from a variety of causes. In addition to these paths of influence on state death rates, lower SES also signifies other behaviors, beliefs, attitudes, and values that may lead to premature mortality. The present SEM results support these links and bolster the plausibility that they are causal in nature.

Limitations and issues

Some readers might assume that the present research seriously suffers from the detrimental effects of the small sample size of 48 states. However, the sample here also is the defined population. Descriptive statistics apply equally to the population and the sample. Logically, this makes inferential statistics unnecessary. Nevertheless, they have been included here to satisfy conventional reporting practice, facilitate decisions in the analytical process, and provide standard benchmarks for comparison purposes. Analytic strategies based on correlation, multiple regression, and SEM often have been used successfully with similarly sized samples (e.g. McCann, 2008; Sideridis et al., 2014).

Another issue that might arise has to do with the fact that the state Big Five personality scores developed by Rentfrow et al. (2008) were based on data collected from 1999 to 2005, well before the current focus year of 2020. However, there are several reasons to allay such a concern. For example, Elleman et al. (2018) demonstrated that such state scores were temporally stable and maintained consistent relations with sociodemographic variables from 1999 to 2015—the range of years they analyzed. Studies using other sociodemographic variables measured more recently also have employed these state scores successfully (e.g. McCann, 2023; Webster et al., 2023).

This is a cross-sectional study with analytical techniques grounded in correlations. Strictly speaking, the usual caution that causality cannot be inferred from correlational evidence must be applied. However, a speculated causal model for the associations between variables was tested for plausibility. That analysis produced evidence supporting the view that the paths between variables in the model—both direct and indirect—indeed may be causal links. Nevertheless, the model coefficients cannot be interpreted as sufficient evidence to unequivocally assert that such causal connections actually exist in the state-level data.

Other factors also may play roles in relation to health risks and all-cause mortality. For example, personality dimensions outside of the current framework such as impulsivity (Kale et al., 2018), sensation seeking (De Moor et al., 2006), and honesty-humility (MacCann et al., 2015) also have been implicated regarding health risks. As well, it is important to acknowledge that structural and systemic factors related to racism (e.g. Murry et al., 2024) and sexism (e.g. Homan, 2024) also may have an impact on SES, health risk factors, premature mortality, and perhaps even personality differences to some degree. The extent to which such excluded factors may have an impact on the present SEM analysis is, of course, unknown.

Potential implications for application

The present study was primarily basic research to enhance our understanding of the relations between personality, health risks, and mortality with states as the analytic units. However, several aspects of the results also hold potential implications for application. The SEM analysis suggests that states with higher age-adjusted all-cause death rates should focus efforts to reduce these mortality rates on the reduction of the six health risk variables in the health risk composite and on the raising of SES. Each of the health risk variables and the education and income components of SES are potentially modifiable. Information and incentive campaigns directed at state residents to increase the percentage engaged in the six health risk variables and the pursuit of more education and higher incomes may eventually lead to better population health and lower rates of premature mortality.

The final SEM model also suggests that openness and neuroticism have direct impacts on health risks as well as indirect impacts through those health risks on death rates. However, state differences on these two dispositional variables are largely rooted in state gene pools, firmly integral to the person, and quite stable throughout life. They are not readily, if at all, responsive to intentional modification efforts. Nevertheless, information and incentive campaigns to encourage a healthier lifestyle might have greater success if they were tailored to appeal to those who are low on openness and to those high on neuroticism depending upon state levels of those two personality variables. For example, such tailoring approaches based on Big Five differences have been put forward by Olsacher et al. (2023) to encourage organ donation in Germany and by Kim et al. (2023) for the provision of optimal health care.

Future research

The present SEM results were obtained with states as the analytic units. Ample individual-level research has been cited to support the speculation that direct and indirect paths involving SES, openness, neuroticism, health risks, and all-cause mortality may stem from parallel paths with individuals as the analytic units. Therefore, justification of such cross-level extrapolation would be on even firmer ground if the same SEM results pattern could be demonstrated in a future individual-level replication using the same variables.

Whether the present aggregate-level SEM results based on American states generalize to other nations and cultures also deserves research attention. For example, the UK has readily available Big Five personality scores for 380 local authority districts (LADs; Rentfrow et al., 2015). As well, socioeconomic grade data also are available for each LAD (UK Geographics, 2014). Data for all-cause mortality and each of the six health risk variables also are likely to be obtainable from various UK sources. Therefore, an initial cross-cultural UK SEM analysis seems feasible using LADS as the analytic units. Other countries also may afford such replication opportunities.

Conclusion

The present SEM results based on the 48 contiguous American states as analytic units shows that 85% of the variance in age-adjusted death rates can be accounted for by four variables—SES, openness to experience, neuroticism, and a health risk composite derived from lack of exercise, physical inactivity, lack of fruit and vegetable consumption, smoking, obesity, and lack of sleep. As well, 80% of the variance in the health risk composite scores can be accounted for by three variables—SES, openness to experience, and neuroticism. States with lower SES, lower openness, higher neuroticism, and higher health risk composite scores have higher death rates. States with lower SES, lower openness, and higher neuroticism have higher health risk composite scores. SES has a direct impact on death rates and an indirect impact through health risks on death rates. Openness to experience and neuroticism have indirect impacts through health risks on death rates.

The results underline the importance of state socioeconomic variables and resident personality to explaining state differences in health risks and age-adjusted death rates. The SEM results also support the plausibility of the speculation that the demonstrated relations are causal in nature. As well, with due caution, it is speculated that the present state-level results stem from parallel relations at the individual level. Generally, the present research also demonstrates the usefulness of the geographical psychology approach for further understanding of population health.

Footnotes

Data availability statement: The data can be requested at DOI: 10.17605/OSF.IO/5V6RW.

The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding: The author received no financial support for the research, authorship, and/or publication of this article.

Ethics approval: This article does not contain any studies with human participants performed by the author.

Informed consent: This study did not require informed consent.

ORCID iD: Stewart JH McCann Inline graphic https://orcid.org/0000-0003-4123-448X

References

  1. America’s Health Rankings (2021) 2021 Annual Report. United Health Foundation. Available at: https://assets.americashealthrankings.org/app/uploads/americashealthrankings-2021annualreport.pdf (accessed 10 July 2024) [Google Scholar]
  2. Anselin L. (2003) An introduction to spatial regression analysis in R. http://labs.bio.unc.edu/buckley/documents/anselinintrospatregres.pdf (accessed 12 June 2020)
  3. Bogg T, Roberts BW. (2004) Conscientiousness and health-related behaviors: A meta-analysis of the leading behavioral contributors to mortality. Psychological Bulletin 130(6): 887–919. [DOI] [PubMed] [Google Scholar]
  4. Bollen K, Pearl J. (2013) Eight myths about causality and structural equation models. In: Morgan S. (ed.) Handbook of Causal Analysis for Social Research. New York, NY: Springer, pp.301–328. [Google Scholar]
  5. Bratt AS, Stenström U, Rennemark M. (2016) The role of neuroticism and conscientiousness on mortality risk in older adults after child and spouse bereavement. Aging & Mental Health 20(6): 559–566. [DOI] [PubMed] [Google Scholar]
  6. Caille P, Stephan Y, Sutin AR, et al. (2024) Personality and change in physical activity across 3–10 years. Psychology and Health 39(5): 670–690. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Chen H, Lai K, He L, et al. (2020) Where you are is who you are? The geographical account of psychological phenomena. Frontiers in Psychology 11: 536. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Conger RD, Martin MJ, Masarik AS. (2021) Dynamic associations among socioeconomic status (SES), parenting investments, and conscientiousness across time and generations. Developmental Psychology 57(2): 147–163. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. De Moor MH, Beem AL, Stubbe JH, et al. (2006) Regular exercise, anxiety, depression and personality: A population-based study. Preventive Medicine 42(4): 273–279. [DOI] [PubMed] [Google Scholar]
  10. Dubowitz T, Heron M, Bird CE, et al. (2008) Neighborhood socioeconomic status and fruit and vegetable intake among whites, blacks, and Mexican Americans in the United States. American Journal of Clinical Nutrition 87(6): 1883–1891. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Elleman LG, Condon DM, Russin SE, et al. (2018) The personality of U.S. states: Stability from 1999 to 2015. Journal of Research in Personality 72: 64–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Ferguson E, Bibby PA. (2012) Openness to experience and all-cause mortality: A meta-analysis and requivalent from risk ratios and odds ratios. British Journal of Health Psychology 17(1): 85–102. [DOI] [PubMed] [Google Scholar]
  13. FRED (2024) Per capita personal income by state, annual. FRED Economic Data. Available at: https://fred.stlouisfed.org/release/tables?rid=110&eid=257197&od=2021-01-01# (accessed 16 July 2024) [Google Scholar]
  14. Gale CR, Čukić I, Batty GD, et al. (2017) When is higher neuroticism protective against death? Findings from UK biobank. Psychological Science 28(9): 1345–1357. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Garrido SJ, Funes PN, Peñaloza Merlo ME, et al. (2018) Personality traits associated with eating disorders and obesity in young Argentineans. Eating and Weight Disorders - Studies on Anorexia Bulimia and Obesity 23(5): 571–579. [DOI] [PubMed] [Google Scholar]
  16. Gilmore DR, Whitfield KE, Thorpe RJ. (2019) Is there a difference in all-cause mortality between non-Hispanic black and non-Hispanic white men with the same level of education? Analyses using the 2000–2011 national health interview surveys. American Journal of Men’s Health 13(1): 1557988319827793. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Glei DA, Lee C, Weinstein M. (2022) Assessment of mortality disparities by wealth relative to other measures of socioeconomic status among US adults. JAMA Network Open 5(4): e226547. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Graham EK, Rutsohn JP, Turiano NA, et al. (2017) Personality predicts mortality risk: An integrative data analysis of 15 international longitudinal studies. Journal of Research in Personality 70: 174–186. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Grogan CS, Turiano NA, Habenicht A, et al. (2024) Personality traits and mediating pathways to mortality risk: A systematic review. Psychology and Health 43(3): 214–224. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Hersch J. (1996) Smoking, seat belts, and other risky consumer decisions: Differences by gender and race. Managerial and Decision Economics 17(5): 471–481. [Google Scholar]
  21. Homan P. (2024) Health consequences of structural sexism: Conceptual foundations, empirical evidence and priorities for future research. Social Science & Medicine 351: 116379. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Ji X, Covington LB, Patterson F, et al. (2023) Associations between sleep and overweight/obesity in adolescents vary by race/ethnicity and socioeconomic status. Journal of Advanced Nursing 79(5): 1970–1981. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. John O, Srivastava S. (1999) The Big Five trait taxonomy: History, measurement, and theoretical perspectives. In: Pervin L, John O. (eds) Handbook of Personality: Theory and Research, 2nd edn. New York, NY: Guilford Press, pp.102–139. [Google Scholar]
  24. Jokela M, Batty GD, Nyberg ST, et al. (2013) Personality and all-cause mortality: Individual-participant meta-analysis of 3,947 deaths in 76,150 adults. American Journal of Epidemiology 178(5): 667–675. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Kale D, Stautz K, Cooper A. (2018) Impulsivity related personality traits and cigarette smoking in adults: A meta-analysis using the UPPS-P model of impulsivity and reward sensitivity. Drug and Alcohol Dependence 185: 149–167. [DOI] [PubMed] [Google Scholar]
  26. Kern ML, Friedman HS. (2011) Personality and pathways of influence on physical health. Social and Personality Psychology Compass 5(1): 76–87. [Google Scholar]
  27. KFF (2024) Population distribution by race/ethnicity. Kaiser Family Foundation. Available at: https://www.kff.org/other/state-indicator/distribution-by-raceethnicity/?currentTimeframe=0&sortModel=%7B%22colId%22:%22Location%22,%22sort%22:%22asc%22%7D (accessed 27 July 2024) [Google Scholar]
  28. Kim GR, Jee SH, Pikhart H. (2018) Role of allostatic load and health behaviours in explaining socioeconomic disparities in mortality: A structural equation modelling approach. Journal of Epidemiology and Community Health 72(6): 545–551. [DOI] [PubMed] [Google Scholar]
  29. Kim J, Jasper A, Baek Y, et al. (2023) Personality segmentation for optimal health care: A review and proposed segmentation approach. Psychological Studies 68(4): 481–488. [Google Scholar]
  30. Kochanek K, Murphy SL, Xu J, et al. (2023) Deaths: Final data for 2020. National Vital Statistics Reports 72(10): 1–91. [PubMed] [Google Scholar]
  31. Laine JE, Baltar VT, Stringhini S, et al. (2020) Reducing socio-economic inequalities in all-cause mortality: A counterfactual mediation approach. International Journal of Epidemiology 49(2): 497–510. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Leth-Steensen C, Gallitto E. (2016) Testing mediation in structural equation modeling: The effectiveness of the test of joint significance. Educational and Psychological Measurement 76(2): 339–351. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Lincoln KD, Abdou CM, Lloyd D. (2014) Race and socioeconomic differences in obesity and depression among black and non-Hispanic White Americans. Journal of Health Care for the Poor and Underserved 25(1): 257–275. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Liu W, Hu B, Dehghan M, et al. (2021) Fruit, vegetable, and legume intake and the risk of all-cause, cardiovascular, and cancer mortality: A prospective study. Clinical Nutrition 40(6): 4316–4323. [DOI] [PubMed] [Google Scholar]
  35. MacCann C, Todd J, Mullan BA, et al. (2015) Can personality bridge the intention-behavior gap to predict who will exercise? American Journal of Health Behavior 39(1): 140–147. [DOI] [PubMed] [Google Scholar]
  36. McCann SJ. (2008) Societal threat, authoritarianism, conservatism, and U.S. state death penalty sentencing (1977-2004). Journal of Personality and Social Psychology 94(5): 913–923. [DOI] [PubMed] [Google Scholar]
  37. McCann SJH. (2023) Neuroticism, urbanization, and the state prevalence of Parkinson’s disease in the USA. Journal of Research in Personality 102: 104319. [Google Scholar]
  38. Murray TC, Rodgers WM, Fraser SN. (2012) Exploring the relationship between socioeconomic status, control beliefs and exercise behavior: A multiple mediator model. Journal of Behavioral Medicine 35(1): 63–73. [DOI] [PubMed] [Google Scholar]
  39. Murry VM, Bradley C, Cruden G, et al. (2024) Re-envisioning, retooling, and rebuilding prevention science methods to address structural and systemic racism and promote health equity. Prevention Science 25(1): 6–19. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Nunes J, Jean-Louis G, Zizi F, et al. (2008) Sleep duration among black and white Americans: Results of the national health interview survey. Journal of the National Medical Association 100(3): 317–322. [DOI] [PubMed] [Google Scholar]
  41. Olsacher A, Bade C, Ehlers J, et al. (2023) How to effectively communicate health information on social media depending on the audience’s personality traits: An experimental study in the context of organ donation in Germany. Social Science & Medicine 335: 116226. [DOI] [PubMed] [Google Scholar]
  42. Persoskie A, Leyva B. (2015) Blacks smoke less (and more) than whites: Simpson’s paradox in U.S. Smoking rates, 2008 to 2012. Journal of Health Care for the Poor and Underserved 26(3): 951–956. [DOI] [PubMed] [Google Scholar]
  43. Petrovic D, de Mestral C, Bochud M, et al. (2018) The contribution of health behaviors to socioeconomic inequalities in health: A systematic review. Preventive Medicine 113(24): 15–31. [DOI] [PubMed] [Google Scholar]
  44. Pettigrew T. (1997) Personality and social structure: Social psychological contributions. In: Hogan R, Johnson J, Briggs S. (eds) Handbook of Personality Psychology. Cambridge, MA: Academic Press, pp.417–438. [Google Scholar]
  45. Ploubidis GB, Grundy E. (2009) Personality and all cause mortality: Evidence for indirect links. Personality and Individual Differences 47(3): 203–208. [Google Scholar]
  46. Puka K, Kilian C, Zhu Y, et al. (2023) Can lifestyle factors explain racial and ethnic inequalities in all-cause mortality among US adults? BMC Public Health 23(1): 1591. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. R (2022) The R project for statistical computing. Available at: https://www.r-project.org/ (accessed 16 April 2022)
  48. Rentfrow P, Gosling S, Potter J. (2008) A theory of the emergence, persistence, and expression of geographic variation in psychological characteristics. Perspectives on Psychological Science 3(5): 339–386. [DOI] [PubMed] [Google Scholar]
  49. Rentfrow PJ, Jokela M. (2016) Geographical psychology: The spatial organization of psychological phenomena. Current Directions in Psychological Science 25(6): 393–398. [Google Scholar]
  50. Rentfrow PJ, Jokela M, Lamb ME. (2015) Regional personality differences in Great Britain. PLoS One 10(3): e0122245. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Roberts BW, Kuncel NR, Shiner R, et al. (2007) The power of personality: The comparative validity of personality traits, socioeconomic status, and cognitive ability for predicting important life outcomes. Perspectives on Psychological Science 2(4): 313–345. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Robinson WS. (1950) Ecological correlations and the behavior of individuals. American Sociological Review 15(3): 351–357. [Google Scholar]
  53. Saffer H, Dave D, Grossman M, et al. (2013) Racial, ethnic, and gender differences in physical activity. Journal of Human Capital 7(4): 378–410. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Saint-Maurice P, et al. (2024) Associations between actigraphy-measured sleep duration, continuity, and timing with mortality in the UK biobank. Sleep 47(3): zsad312. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Sideridis G, Simos P, Papanicolaou A, et al. (2014) Using structural equation modeling to assess functional connectivity in the brain: Power and sample size considerations. Educational and Psychological Measurement 74(5): 733–758. [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Silva D, et al. (2020) Physical inactivity as a risk factor for all-cause mortality in Brazil (1990–2017). Population Health Metrics 18(Suppl 1): 13. [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Stringhini S, Dugravot A, Shipley M, et al. (2011) Health behaviours, socioeconomic status, and mortality: Further analyses of the British Whitehall II and the French GAZEL prospective cohorts. PLoS Medicine 8(2): e1000419. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. UK Geographics (2014) Social grade A, B, C1, C2, D, and E. Available at: https://www.ukgeographics.co.uk/blog/social-grade-a-b-c1-c2-d-e (accessed 17 August 2024)
  59. US Census Bureau (2023) ACS SPM State Tables: 2009 to 2019, 2021. Available at: https://www.census.gov/data/tables/time-series/demo/supplemental-poverty-measure/ACS-SPM-State-Tables.html (accessed 27 August 2024)
  60. US Census Bureau (2024) Educational attainment. Available at: https://data.census.gov/table?q=S1501&g=010XX00US$0400000&tid=ACSST1Y2019.S1501 (accessed 26 August 2024)
  61. Visaria A, Setoguchi S. (2023) Body mass index and all-cause mortality in a 21st century U.S. population: A national health interview survey analysis. PLoS One 18(7): e0287218. [DOI] [PMC free article] [PubMed] [Google Scholar]
  62. Vukasović T, Bratko D. (2015) Heritability of personality: A meta-analysis of behavior genetic studies. Psychological Bulletin 141(4): 769–785. [DOI] [PubMed] [Google Scholar]
  63. Webster G, et al. (2023) Openness relates to COVID-19 vaccination rates across 48 United States but politics trump personality. Social and Personality Psychology Compass 17(8): e12787. [Google Scholar]
  64. Zhang Z, Chen Y, Zhang L. (2024) Research on the relations among personality traits, sports commitment, and exercise behavior - A case study of Chinese college students. Psychology Research and Behavior Management 17: 665–678. [DOI] [PMC free article] [PubMed] [Google Scholar]

Articles from Journal of Health Psychology are provided here courtesy of SAGE Publications

RESOURCES