Cross-national associations of IQ and infectious diseases: Is the prevalence of Corona an exceptional case?

David Becker (corresponding author) Department of Psychology, Chemnitz University of Technology Wilhelm-Raabe-Str. 43, D-09107 Chemnitz, Germany E-mail: david.becker@psychologie.tu-chemnitz.de Claudia Kiel Department of Psychology, Chemnitz University of Technology Wilhelm-Raabe-Str. 43, D-09107 Chemnitz, Germany E-mail: claudia.kiel@s2018.tu-chemnitz.de Heiner Rindermann Department of Psychology, Chemnitz University of Technology Wilhelm-Raabe-Str. 43, D-09107 Chemnitz, Germany E-mail: heiner.rindermann@psychologie.tu-chemnitz.de


Introduction
The pandemical spread of the Coronavirus is certainly the most decisive event of 2020 and 2021. The "severe acute respiratory syndrome coronavirus 2" (SARS-CoV-2) particularly affects the human respiratory system [1]. Until the 29 th of June 2021 180,492,131 cumulative cases and 3,916,771 cumulative deaths were recorded worldwide [2]. No region in the world was spared, however, the relatively well-developed countries with higher levels in student assessment and psychometric test results in Europe and Northern America were disproportionately affected, with up to 90% of the worldwide cases [2].
From the perspective of cross-national research on education and intelligence, this is remarkable and surprising, since the prevalence of infectious diseases is usually highly negatively associated with the mean educational and cognitive ability level of countries [3,4].
The evidence is especially strong for a health improving effect of cognitive ability (intelligence including some knowledge aspects); many papers reported such associations (at the international level comparing countries [5,6,7,8,9]. In a paper by Daniele and Ostuni (2013) only one of 15 infectious diseases did not negatively correlate with IQ (Japanese encephalitis) [10]. In parallel, the authors found strong positive correlations of DALYs (disability-adjusted life years lost) due to infectious diseases with mean temperature (the higher the average temperature the more years are lost) and strong negative correlations with absolute latitudes (far from the tropics less life years lost).
This contradicts the geographic pattern of all SARS-CoV-2-pandemic. The origin of the phylogenetic tree of the virus is SARS-CoV and it first occurred in 2003 in southern China [11,1]. Countries affected by the older SARS-CoV pandemic were distributed all around the globe, however most affected regions were China, Hong Kong, Taiwan, Singapore, Hanoi in Vietnam and Toronto in Canada [12]. Africa, the Middle East and Central Asia were almost completely spared [13]. A further SARS-CoV-2 predecessor, MERS-CoV, spread through the Middle East in 2013, affected 23 countries, mostly at the Arabic Peninsula [1]. The 2020 SARS-CoV-2-pandemic is the third and most serious pandemic caused by this family of viruses, with higher death rates and also possible negative long-term effects [14].

Cognitive ability does not appear to predict SARS-CoV-2 pandemic outcomes at the crossnational level in the way it has for other infectious diseases. An influence by other factors is
possible. What other factors are there that can accelerate the spread of the virus and exacerbate its consequences?
(1) The first factor mentioned in studies is disease burden. A statement from the World Heart Federation identified cardiovascular disease, hypertension, diabetes, chronic respiratory disease, and cancer as risk factors susceptible to SARS [15]. Around 96% of 3,335 observed Italian patients died in hospitals were positively diagnosed for one or more pre-existing vulnerabilities as hypertension, diabetes and heart disease, and casefatality rates of the SARS were two to four times higher in Chinese patients with cardiovascular diseases and hypertension compared to the general Chinese population [16]. Behavior that is detrimental to health, as smoking, chronic obstructive pulmonary diseases, lung cancer, high alcohol consumption, liver dysfunction, and diabetes mellitus, is a further risk factor [17,18]. Additionally, there is rising evidence that Vitamin D, as an essential receptor in cells of the immune system [19], is a crucial factor to protect against infections with SARS-CoV-2 [20,21,22]. All those vulnerabilities worsen the consequences of an infection, complicate treatment and make therapy less successful. Additionally, being in medical treatment by another illness increases the risk to catch an infection. Some risk factors as alcohol consumption seem to be increased during the pandemic and lockdowns [23].
(2) The prevalence of diseases increases with age, the median age of a population is seen as crucial to the susceptibility to SARS-CoV-2. Additionally, the quality of the health system, i.e., the quality of health institutions and the amount of government funding, may also play a role here. Not only the risk of severity and duration of health damage from existing diseases, but also the risk of transmission of infectious diseases due to inadequate hygiene standards could increase the impact of the pandemic.
(3) Air pollution seems to be a further factor contributing to the impact of the pandemic. A higher concentration of particulate matter in the air may make the lung more susceptible for a more serious progression of a SARS-CoV-2 infection. A study by Tung et al. summarized much of this work and discussed the role of the particulate matter working in promoting aerial transmission by preserving viruses for several days [24]. At the cross-national level, evidence for this mechanism has been found [25], showing that during the first wave the tropospheric concentration of NO2 in Europe IQ & Corona 6 was significantly stronger in areas (north Italy and central Spain) in which the virus spread most, and which counted the highest numbers of deaths.
(4) Air pollution is to a certain degree associated with population density, transportation and economic activities simultaneously increasing pollution and interpersonal contacts.
Copiello and Grillenzoni showed that, in the case of Chinese provinces, per capita emissions of industrial waste gases were not any longer a positive predictor for pandemic spread if population density, economic activity and climatic conditions were taken into account [26].
(5) Climatic conditions present a further factor: In general, the burden of pathogens is higher in environments with warmer and more humid climates than in colder and drier environments [27]. However, animal experiments with the Influenza virus, which also spreads via and infected the respiratory system, showed that lower temperature and lower humidity can boost the transmission of several viruses [28]. This helps to explain the seasonal (winter) flu waves in the northern hemisphere (outside cold, indoors dry) and is possibly an explanation for the SARS-CoV-2-pandemic, whose waves coincident in time with those of the flu. (6) According to the global pattern of the pandemic, there appears to be a positive, at least statistical, association between wealth and the number of cases and deaths, even it is not clear at this point to what extent there could be a causality or whether it is just the result of an unknown confounder. From a theoretical perspective, wealth has to be taken into account for its proven negative relationship with infectious diseases, even if the opposite is apparently the case with SARS-CoV-2.
Consideration of these additional factors may yield a truer pattern of potential effects of intelligence on the spread of SARS-CoV-2 and severity of infection. However, intelligence still may have health-related positive (i.e., infection decreasing and health increasing) and specific Corona-positive (i.e., SARS-CoV-2 increasing) effects: (1) Due to higher insight, more knowledge and a better consideration of risks and goals more intelligent people should be able to better adapt to a new challenge as the Corona threat. E.g., avoiding contact (social distancing), wearing a mask or at the politicalinstitutional level to develop and enforce rules to contain the spread of the Corona virus [29]. Intelligence enables a more rational thinking and behavior [30]. However, the complex nature of all effects of behavioral and institutional reactions, including positive and negative side effects, make statements on the rationality of responses to the Corona pandemic opaque. For instance, there are usually no serious Corona effects for healthy persons younger than 70 years old [31]. There are negative consequences of political reactions (lockdowns and school closings) including negative effects on economic growth and long-term health leading in the long run to more losses in DALYs [32,33,34]. 1 Lockdowns, social distancing and masks reduce social contact increasing psychological and physical health problems (which already have existed due to Corona anxiety) or simply make people less happy [36,37].
(2) Conduct of one's life and politics are polytelic, i.e., rational decisions cannot only focus on avoiding SARS-CoV-2 but also have to consider social, emotional, intellectual, economic and political criteria. As well as income is not the unique criterion for intelligence (or the unique rational life goal) it cannot be a low COVID rate.
(3) Intelligence generally helps to improve the environment, i.e., creates more wealth, better institutions, more rational politics and a higher quality of the health system all being important for health and an appropriate dealing with new challenges [38,39].
However, higher quality health care also means more testing, which leads to more Corona diagnoses and more registered Corona cases.
(4) More intelligent people are generally healthier due to a healthier lifestyle, due to improved environment and because intelligence is an indicator of health [40,41].
However, healthier lives mean that the population is getting older on average, leading to increased susceptibility to Corona. Dying young paradoxically leads to a healthier remaining population.
We seek to analyze these contradicting intelligence effects on Corona, we seek to uncover the possible direct and indirect ways of effect and to do so in the context of further important factors such as climate.

Selection of variables
From the factors named above, the following twelve variables as possible factors at the national (country) level were selected for the analysis: 1 One quote from Miles et al. (2020, p. 2) highlights the negative consequences of the prevailing lockdown response to the Corona pandemic: "There is a need to normalise how we view COVID-19 because its costs and risks are comparable to other health problems (such as cancer, heart problems, diabetes) where governments have made resource decisions for decades. ... Movement away from blanket restrictions that bring large, lasting and widespread costs and towards measures targeted specifically at groups most at risk is prudent." [35] (1) Mean national level of intelligence in the broader meaning including knowledge (cognitive ability or cognitive competence).
(2) Mean temperature at the cold season (winter).
(6) Health burden with focus on diseases and behavior positively associated with the probability of a SARS-CoV-2 infection and a lethal SARS-outcome (cardiovascular disease, diabetes prevalence, cancer, smoking, alcohol consumption).    (1) Basic virus reproduction ratio (number of cases directly generated by one case).
(2) Rate and severity of registered cases (e.g., cases with need for intensive care).
(3) Rate of deaths caused by SARS-CoV-2 infections (in different measures). for each month, then we selected the temperature of the coldest month for each country. We did not focus on a fixed cold season for all countries as geographic positions and differences give each country a specific temperature profile.

Air pollution
We used data from the WHO for the variable air pollution [48]. We took "Ambient air pollution attributable YLLs" standing for premature death lost years of life per capita (population data provided by CIA [49]). Information is given for 183 countries. YLLs directly refer to the amount of life affected by air pollution in health, it is insensitive to regional air pollution and regional population density within the country and also to differences in risk from different types of air pollution.

Wealth
Wealth was taken as average GDP/c in 2017 dollars (purchasing power parity, PPP), source CIA for 204 countries [49]. Finally, GDP/c was logarithmically transformed since the real meaning of an increase in per capita income by a value x can be seen as negatively related to the level of per capita income (hereinafter GDP/c log). That means an increase of average GDP/c from 1,000 dollars to 2,000 dollars means a larger increase in life quality than an increase from 100,000 to 101,000 dollars.

Demographics
Demographics include the median age in years (Nctry = 181), the % of a population's urbanization (Nctry = 205), given by the CIA [49], based on measurements mostly between 2015 and 2020 (obtained on June the 30 th in 2020), and population density (capita/km²; Nctry = 185), taken from the "Coronavirus Pandemic Data Explorer" [50]. The demographics variable was built from by averaging numbers for age, urbanization and population density after zstandardization. . This variable not just reflects cancer per capita but also the severity of the disease. The health risk variable was built from those five variables by averaging after z-standardization.

Health risk
We did not include Vitamin D deficiency in health risk, although possibly an important factor according to medical research, as we failed to obtain valid data for a large number of countries. Ilie et al. provided the biggest cross-national dataset findable with just 20 overwhelmingly European countries [20]. Alternatively, it would have been possible to use UV intensity as a proxy for Vitamin D since it is primarily formed by light. The WHO provided exposure to solar ultraviolet radiation data for 192 countries, however, the correlation of these numbers with the winter temperature is .85 and thus already covered [53].

Quality of the health system
Two variables indicating the quality of the health system were used, calculated from data obtained from The World Factbook of the CIA [49]. First, we calculated current health expenditure per capita by multiplying the numbers for current health expenditure as % of annual GDP (Nctry = 186) and GDP/c (PPP). A second variable named health system quality was calculated as the average of (1) physician density (Nctry = 191), (2) hospital bed density (Nctry = 174) and (3) shares of people with access to improved sanitation (Nctry = 196). Again, all three variables were z-standardized before averaged. Current health expenditure reflects the overall strength of a health system, whereas physician and hospital bed density reflect the ability of health facilities to treat seriously ill patients. Access to improved sanitation was added to cover the hygiene standards outside the health system, which are likely to play an important role in disease containment.

Mobility Changes
Mobility changes were measured by the use of Google data [54]. The "COVID-19 Community Mobility Reports" provide daily data about how the movement patterns of societies changed. These numbers were collected and aggregated from data anonymously measured by Google apps and software, e.g., Google Maps. For each day between the 15 th of February 2020 and 17 th of Mai 2021, data were reported for six categories of movements: (1) retail and recreation, (2) grocery & pharmacy, (3) parks, (4) transit stations, (5) workplaces and (6) residential. A score for each country was calculated by first averaging the daily numbers between 15 th of February 2020 and 17 th of Mai 2021 within each category, then averaging these six scores again.
The numbers given by the source represent the percentage deviation of mobility on one day compared to average mobility on the same day before the pandemic (3 rd of January and 6 th of February 2020). Therefore, it is not a measurement of the absolute mobility and also not for the relative mobility by a global standard. However, it reflects the reaction of a population on the pandemic in terms of de-or increasing mobility. A possible weakness is the determination of the baseline based on the data from January to February 2020, as it does not take into account seasonal and weather-related changes in mobility outside of the pandemic.
However, it is the most detailed and extensive available database for mobility. (1) The reproduction rate (also known as ratio or "R0").

SARS-CoV-2 countermeasures
2 Variables were given differently as per capita, per thousand capita or per million capita. Since this distinction is irrelevant for regression-based methods, we will only speak of "per capita" or "/c" in the following text and tables, except descriptive statistics.
(2) The number of hospitalized patients with SARS-CoV-2 per capita.
(3) The number of Intensive Care Unit (ICU) admissions due to SARS-CoV-2 per capita.
(4) The cumulated number of daily new registered cases of SARS-CoV-2 per capita.
(5) The cumulated number of daily new registered deaths attributed to SARS-CoV-2 per capita.
(6) The excess mortality, which is a measurement of deaths "from all causes during a crisis above and beyond what we would have expected to see under 'normal' conditions." The positive rate ("7-day rolling average of daily cases, divided by the 7-day rolling average of daily test") was also available but not used in this study, as the test coverage is already in use as an independent variable. Table 1 gives a survey of all variables used with information about polarity and usage.

Analyses
First, it is important to examine the relationship between the six dependent Corona variables.
In theory, these should all represent the impact of the pandemic on health or life expectancy.  Multivariate regression analyses were done by using Lavaan-package version 0.6-7. The command "std.ov" was set to TRUE to standardize all observed variables before entering the analysis. Missing data were treated by Full Maximum Likelihood (command: missing = "FIML"; i.e. all given data were used) and the default model estimator Maximum Likelihood (command: estimator = "ML") was used. Factor scores were estimated by using the command lavPredict (method = "regression") from models ran with settings identical to those from multivariate regression analyses (above). Table 1 provides an overview of all variables. Data files, R syntaxes (commands only) and R outputs (commands and results) can be found in the supplementary material. Files ending on "01" are for the full timespan, "0" is for 2000 only and "1" for 2021 only. Abbreviations of variables are also displayed in Table 1. Subsamples are named "DATAHEMAFFN" and "DATAHEMAFFS" for northern and southern countries, "DATASTACAPH" and "DATASTACAPL" for countries with high and low data quality, and "DATAGDPSTDH"

Results
Intercorrelations within the six dependent Corona-variables (see Table 2) are usually positive.         Within the 15 valid results for reproduction rate, there are 11 negative and 4 positive effects, or 8 negative and 1 positive within the 9 significant results (p ≤ .05). Only in the subsample Rich there is a temporal robust negative effect of β20 = -.90 (p = .003), β20|21 = -.14 (p = .682) and β21 = -.78 (p = .045). In all other subsamples, the operator changed fromto +. Thus, across five subsamples, there is a mean β20 of -.54 (SD = .19), a mean β20|21 of -.07 (SD = .05) and a mean β21 of .04 (SD = .42). Only 4 results are valid for hospitalizations/c: two β20 with no effect, one β20|21 with a negative effect and one β21 with a positive effect. The mean β across the 4 valid results is -.01 (SD = .15). Similarly, only 7 results are valid for Intensive Care Unit treatments/c: 2 for all countries, 3 for northern countries and another 2 for rich countries. Significance is gained by none of the effects, but operators are consistently negative across time and subsamples. The mean β across the 7 valid results is -.53 (SD = .27). There are 13 negative and 3 positive valid effects for registered cases/c with a mean β = -.14 (SD = .20). Robustness across our three time intervals was found for all, northern and rich countries 3 , and a mean β20 = .00 (SD = .21), a mean β20|21 = -.15 (SD = .18) and a mean β21 = -.23 (SD = .16). There are 17 valid results for registered deaths/c, but operators changed frequently, so the mean β is -.02 (SD = .20). There is no robustness in any (sub-)sample and the mean β20 is .05 (SD = .25), the mean β20|21 is -.07 (SD = .18) and the mean β21 is -.01 (SD = .17). The 14 valid results for excess mortality showed 9 negative operators, 4 positive operators and a mean β of -.19 (SD = .30). Temporal robustness of negative effects is obtained for all and northern countries. In the latter case, significance is gained throughout with β20 = -.71 (p = .002) , β20|21 = -.77 (p = .004) and β21 = -.50 (p = .033). The mean β20 is -.17 (SD = .30), the mean β20|21 is -.24 (SD = .29) and the mean β21 is -. 15   .224 Notes. γ = multiplied βs, Totalγ = summed γ (BEHAVIOR = b, k, l; ECONOMICS = d, g, h, j, n, o, p; HEALTH = c, e, f, m; excl. = variable vaccinations/c excluded in 2020 due to missing data. Table 9 shows results from path analyzes using the theoretical model from Figure 1 and   Table 9. In summary, there is some evidence for a statistically reducing impact of cognitive ability (intelligence) on Corona for the long term, but which cannot be explained by behavior (e.g., mobility), health (e.g., vaccinations, health conditions) or economics (e.g., wealth). On the contrary: Along pandemic related behavior (mobility and government measures to combat the pandemic) and health (conditions, system, vaccination and testing), higher cognitive ability is indirectly associated with a more severely affectation by Corona.
Limitations: First, due to a high failure and error rate of the models no statements can be made about hospitalizations/c and Intensive Care Unit treatments/c. The models for reproduction rate, registered cases/c or deaths/c and excess mortality resulted mainly in negative effects, however, robustness was limited. The results suggest that there are two opposing effects of intelligence that cancel each other out.

Discussion
Our results provide preliminary evidence that, during the observed period, the intelligence of a population had a direct reducing effect on the impact of the pandemic which can hardly be explained by existing theories. Initially, we hypothesized a reducing impact of intelligence on Corona based on two causal effects: (1) Intelligence increases insight, knowledge and the ability to react in a rational way to new challenges. This at the individual and societal level (e.g., by social distancing or establishing rules for isolation of infected persons).
(2) Intelligence helps to improve the environment being beneficial for avoiding risks and recovering health (e.g., better health system).
However, these two causal effects seem to do exactly the opposite and could, albeit in an unknown way, explain why societies with higher cognitive abilities are more strongly affected by the pandemic.
The second option might get some evidence from the direct reducing effect of cognitive ability and could possibly be explained as follows: Protecting oneself from diseases is to be rated as highly rational, but all measures against Corona may also have negative side effects.
Lockdowns damage the economy, school closings will reduce in the long run economic growth, both leading to higher mortality rates. SARS-CoV-2 infection has no serious health consequences for the vast majority of people and especially for younger and healthy individuals, but the lockdowns certainly have negative consequences for everyone, for prosperity, quality of life, psychological well-being and physical health (e.g., due to less contact, less sports etc.). Individual considerations of pandemic risks on the one hand and economic constraints on the other can vary widely, being influenced, for example, by a person's age and health status and thus vulnerability to Corona or their current economic situation and activity. Thus, higher cognitive abilities produce disease-protective behavior only as long as it does not lead to damage in other ways, e.g. economically.
Overall, however, our results for the IQ-Corona relationship also show that intelligence does not appear to be a panacea for preventing or mitigating all disasters. Complex mechanisms, the difficulty of making reliable predictions and the increased vulnerability of societies due to changes in living conditions and demographics (age, urbanization and population density) caused by high levels of intelligence and wealth lead to a mixture of indirect and direct COVID decreasing and increasing effects of a population's cognitive ability level.

Conclusion
Previous findings showed a significant negative impact of IQ on the prevalence and severity of infectious diseases. Specifically, hygiene and health technologies, as well as rational behavior in dealing with health problems, were considered possible mediators. We were able to underpin these assumptions in the context of the Corona pandemic in some ways, however we also found positive effects mediated along wealth and country's level of development.

Limitations and suggestions
At the time of finishing this manuscript the SARS-CoV-2-pandemic was still going on. We already saw a change in the global pattern from the first to the second and third wave, for example an increasing impact in developing countries [62,63,64,65]. This is also reflected in our results and could be the reason for the low robustness over time. Further research should analyze the possible impact of intelligence on the Corona pandemic in further waves. Furthermore, it cannot be ruled out that, in addition to intelligence [66,67], genes may also be relevant for susceptibility to corona virus. Furthermore, in addition to the broad country level, a more focused regional level within a country is also important.

Statements & Declarations
Ethics approval: This study is a cross-national study at an aggregated level and thus did not involve human participants, their data or biological material. All data were taken from freely accessible databases. An ethics approval is therefore not considered necessary by the authors.
Consent to participate: As no individual data from human participants has been used for this study, no consent to participate was obtained.
Consent to publish: As no individual data from human participants has been used for this study, no consent to publish was obtained.
Availability of data and material: All data were taken from freely accessible databases (in order of appearance):     .03 .10 -. 10 .06 .12 Notes. β = standardized regression coefficients for independent variables, upper value: direct effects, mid value: indirect along behavior (e.g., mobility), lower value: indirect along economics (e.g., wealth) and health (e.g., vaccination, conditions), controlled by 11 to 12 variables on Total-COVID, R² = explained variance in Total-COVID by all independent variables.