Age Patterns of Religiosity and Atheism in the USA, Europe and Australia

It is well known that the proportion of people who are atheist or non-religious in high income countries is generally higher in younger people than older people and rising over time.  So is the decline in religiosity mainly a cohort or period phenomenon?

I have analysed age patterns in religiosity using data from the last three waves of the World Values Survey and the European Values Study [1-4]. See previous posts here and here for more details on the analysis of these surveys. I chose to look at the USA and compare it with the eight high income countries in which the prevalence of irreligion (atheism and the non-religious) exceeds 50%: Australia, Denmark, Finland, France, Netherlands, Norway, Sweden and the United Kingdom. The non-religious category is quite small in these countries, apart from Australia, and includes people who say they believe in God, but say they are non-religious and rate the importance of God as 8-10 on the unimportant end of a 10-point scale.

The proportion of people who are religious rises with age and the proportion who are irreligious declines. Trends and levels are quite similar for the three high-atheism countries. The following plot aggregates the overall prevalences by sex for the seven European countries and compares them with those for the USA.

Across all age groups apart from 15-29 years, the prevalence of irreligion is higher in men than women. The proportion of religious people who are non-practicing rises to middle age and then declines in older ages in the USA. In contrast, the proportion who are practicing religious remains fairly low in the European countries (ranging from 10 to 20%) whereas the proportion non-practicing rises with age. The prevalence of irreligion in the youngest age group (18-24 years) exceeds 50% in 2018 in all the countries analyzed, including the USA.

Next, I compare time trends across the three last waves of the surveys for these countries. The plot on the left below shows age-specific trends in the per cent of people who are irreligious and that on the right the trends in the per cent of people who are practicing religious.

Although the prevalence of irreligion is lower in all age groups in the USA compared to the European countries, it is rising rapidly in all age groups under 60 years, and will likely catch up to Europe soon. Trends are much smaller in Europe as much of the growth in irreligion occurred decades earlier, whereas it has really only started to increase substantially in the USA in the last decade.

Is the rise in irreligion dominated by cohort or period effects? As in the epidemiology of diseases [5], there are three time factors that are potentially explanatory:  the age of the person, the year (period) in which the population is surveyed, and the year in which a person is born (cohort). Potentially, in the absence of any other influences, the prevalence of irreligion might change with age (people may get disillusioned with religion as they get older, or perhaps more religious as they age). Perhaps the birth cohort is a major factor: as a marker of the social and religious milieu around the time when a person reaches adulthood and makes their own decisions about religious beliefs and practices as an adult. Or perhaps changes in the social, economic and political environment affect people across all ages and birth cohorts so that everyone becomes more or less religion over periods of time. These three factors are not independent, since cohort = period – age. Thus, it is not possible to statistically estimate the independent contributions of all three factors, unless one factor can be specified from external data or knowledge [5].

I will assume that there is no independent age effect. Given the fairly flat trends by age over periods in which social and religious conditions have been relatively stable, I think this is a reasonable first approximation that allows us to gauge the relative importance of cohort versus period effects. I ran fairly simplistic linear regressions of the form 

yx  = β0x +  β1x .period +  β2x .cohort + ε

where yx denotes the prevalence (%) of religiosity category x. I fitted this regression to the individual country data for years 2018 and 2003, where the prevalences in 2003, calculated by backprojecting the trend between 2018 and 2007. This allowed age, cohort and period all to be measured in units of 15 years, so that the coefficient β1x  gives an estimate of the average change in prevalence over a 15 year period and β2x  gives an estimate of the change in prevalence between two birth cohorts 15 years apart.

Given the simplicity of the model, and the simplifying assumptions of linear trend across age groups and periods, this gives no more than an indicative measure of the relative importance of period versus cohort effects, but it does show a very distinct difference between the USA and the high-atheism European countries.

Period effects dominate for the USA, whereas cohort effects are more important in the other countries for atheism and practicing religious. For the two middle categories, both period and cohort effects both play a relatively important role, though for smaller time trends. Almost certainly this reflects the fact that the European countries and Australia have fairly stable secular cultures and cohort effects dominate, whereas the USA over the last decade or two has become increasingly polarized on religious and cultural issues and this period effect has led to declining religiosity across all age groups that is generally larger than the cohort-specific declines, greater in younger age groups.

References

  1. Inglehart, R., C. Haerpfer, A. Moreno, C. Welzel, K. Kizilova, J. Diez-Medrano, M. Lagos, P. Norris, E. Ponarin & B. Puranen et al. (eds.). 2014. World Values Survey: All Rounds – Country-Pooled Datafile Version: https://www.worldvaluessurvey.org/WVSDocumentationWVL.jsp. Madrid: JD Systems Institute.
  2. Haerpfer, C., Inglehart, R., Moreno,A., Welzel,C., Kizilova,K., Diez-MedranoJ., M. Lagos, P. Norris, E. Ponarin & B. Puranen et al. (eds.). 2020. World Values Survey: Round Seven–Country-Pooled Datafile. Madrid, Spain & Vienna, Austria: JD Systems Institute& WVSA Secretariat[Version: http://www.worldvaluessurvey.org/WVSDocumentationWV7.jsp].
  3. Gedeshi, Ilir, Zulehner, Paul M., Rotman, David, Titarenko, Larissa, Billiet, Jaak, Dobbelaere, Karel, Kerkhofs, Jan. (2020). European Values Study Longitudinal Data File 1981-2008 (EVS 1981-2008). GESIS Datenarchiv, Köln. ZA4804 Datenfile Version 3.1.0, https://doi.org/10.4232/1.13486.
  4. EVS (2020): European Values Study 2017: Integrated Dataset (EVS 2017). GESIS Data Archive, Cologne. ZA7500 Data file Version 3.0.0,doi:10.4232/1.13511
  5. Holford TR. Understanding the effects of age, period, and cohort on incidence and mortality rates. Annu Rev Public Health. 1991;12:425-57. doi: 10.1146/annurev.pu.12.050191.002233. PMID: 2049144.
Posted in Uncategorized | Tagged , , , , , , , , , , | Leave a comment

Excess mortality in the USA and its causes 2000 – 2020

Sam Preston and Yana Vierboom published a short paper in April which showed that Americans die at higher rates than Europeans and this translated into roughly 400,000 additional deaths in 2017 that would not have occurred if the USA had Europe’s lower age-specific death rates. That is about 12% of all American deaths and higher than the COVID-19 death toll of around  380,000 in 2020. In a recent Guardian article, they summarized this as “The mortality penalty that the US pays every year is equivalent to the number of Americans who died of Covid in 2020”. They also note that because people tend to die at older ages than those characterizing excess deaths, the total potential years of life lost are three times greater for the excess deaths than for Covid in 2020 (13 million versus 4.4 million).

In the Guardian article, they identify a number of factors that contribute to these excess deaths: overweight and obesity, the deaths of despair identified by Case and Deaton (drug overdose, suicide and alcohol-related causes) and lack of universal health insurance (see here and here). These factors have been widely identified as contributing to higher mortality rates and lower life expectancy for the USA compared to high income European countries, but their comparison of this mortality gap with the Covid death toll was a compelling way to present it. So I was interested to replicate and update their analysis and try to make ballpark estimates of the contribution of these factors to the excess deaths.

Preston and Vierboom used data from the Human Mortality Database (HMD) to create a composite of the five largest European countries, whose combined population size is very similar to that of the United States: Germany, England and Wales, France, Italy, and Spain. Additionally, they argued that using these larger European countries  to provide a mortality standard would avoid unrealistic expectations for a larger and more diverse population that might result from comparisons including small countries where there may be exceptional combinations of factors affecting mortality (e.g., climate, diet, social history, and healthcare delivery) Their analysis covered the period 2000 to 2017, the latest year for which data were available for all six countries.  

I downloaded data for these countries from the HMD a few days ago, and mortality rates were available for all six up to 2017, for four for 2018, and only the USA for 2019. I projected (non-Covid) age-sex specific mortality rates forward to 2020 and added Covid deaths for 2020 using age-sex distributions published by European countries by INED. I then adjusted the projections and in some cases the added 2020 Covid deaths to match the sex specific life expectancies at birth and at age 65, published by Eurostat up to and including year 2020. Data for non-Covid and Covid mortality by age and sex in 2020 were taken from national statistical office sources for the UK and USA.

Figure 1 below shows the ratio of US death rates to the average death rates* for the five European countries (the “European standard”) by age, in 2000, 2010, 2019 and 2020. US mortality rates are consistently higher than the European standard for all ages below 80 years and the ratio has gotten progressively worse throughout the 21st century. The ratio peaks among 25-29 year olds at almost 3 in 2019 and 3.24 in 2020.

Figure 1. * Note I calculated death rates for the male plus female population of the five countries combined, whereas Preston and Vierboom calculated the arithmetic average of the death rates for the five countries.

Figure 2 below shows the annual trend in total excess deaths in the USA above the number than would have occured if the US population had been subject to the age-sex specific death rates of the European standard. This excess rose from 219,000 in the year 2000 to 410,000 in 2019 and 616,000 in 2020. Although there were over 380,000 Covid deaths in the USA in 2020, the European standard also includes substantial numbers of Covid deaths, and the Covid excess for the USA is “only” 136,000 deaths.

Figure 3 shows the age pattern of US excess deaths in 2000, 2010 and 2020. Although the death rate ratio is highest in the 25-34 year age group, overall death rates rise with age and the peak in excess deaths is at older ages from around 40 to 75 year. From age 90 onwards, US death rates are somewhat lower than the European standard. Thus the excess deaths below age 90 are actually higher than the total excess and in 2020 there were 764,000 excess deaths below 90, of which 99,000 were Covid-90 deaths.

In their Guardian article, Preston and Vierboom identified several factors that contribute substantially the US excess. I have made approximate estimates of the contributions of these factors using estimates from the Global Burden of Disease 2019 update (IHME GBD2019) of death rates by age and sex for the six countries for drug overdose, suicide, homicide, deaths attributable to overweight and obesity, deaths attributable to alcohol and drug use.

The 2019 age-sex specific percentages of deaths attributable to these causes were also applied to the 2020 non-Covid mortality estimates. Numbers for drug use were adjusted to reflect a 13% increase in drug overdose deaths in 2020 compared to 2019, an approximately 25% increase in homicide deaths and a 5% decrease in suicide deaths.

Unlike the USA, all the European countries have universal healthcare insurance. The number of working age Americans without health insurance rose from 38.7 million under age 65 in 2000 to 46.5 million in 2010 and then dropped to around 26 million following the introduction of Obamacare. Around 1% of Americans aged 65 or more are also uninsured. The number uninsured has risen to around 30 million in 2020 due to Republican efforts to restrict and dismantle Obamacare. I used an estimate of the hazard ratio for mortality among the uninsured compared with the insured, of 1.4 to calculate the excess mortality among the uninsured. This figure includes adjustments for age, sex, body mass index, alcohol use, smoking, and a number of other factors. The ratio was 1.8 adjusted for age and sex only.

In addition to the uninsured, a substantial number of Americans are underinsured, defined as having out-of-pocket and deductible expenses greater than 10% of their income. Apart from substantial problems with debt, around half of the underinsured do not get needed care because of cost and there are associated excess deaths (see here and here). The number of Americans underinsured has risen from around 16 million prior to Obamacare to around 30 million in 2020. I assume a mortality ratio of 1.1 among the underinsured.  

Figure 4 shows the approximate contributions of overweight and obesity, drug use and overdose, suicide, homicide, lack of health insurance and Covid to the US excess deaths relative to the European standard. Together these six factors account for around 80% of the excess deaths. Excess deaths due to alcohol use were not included as a separate factor in this graph because European and US death rates were quite similar, and the alcohol excess was small and actually negative in 2000.

For 2020, the leading cause of excess deaths was overweight and obesity (around 154,000 deaths), followed by Covid-19 (136,000 deaths), drug use and overdose (103,000 deaths) and lack of health insurance (74,000 deaths). Excess deaths due to homicide and suicide were smaller at 20,200 and 11,200 respectively. If the USA had the European standard death rates for gun homicides and gun suicides, it would have 15,900 fewer gun homicides and 19,200 fewer gun suicides. Around 42% of the latter would still commit suicide by other means.

Figure 4 also illustrates the dramatic rise in drug overdose deaths, the vast majority due to opioids both prescription and illicit, which has occurred over the last decade. In a previous post, I examined this in more detail and noted that, in 2019, the USA accounted for an astonishing 40% of estimated global drug deaths.

Figure 5 shows the age distribution of each of the six factors contribution to total excess deaths in 2020. Not surprisingly, excess deaths due to drug use fall predominantly among younger adults, whereas deaths due to overweight and obesity fall more in later middle and older ages.

Excluding excess Covid deaths, there were almost half a million excess deaths in the USA in 2020 compared to the US standard. While overweight and obesity is not an easy problem to address, the European experience shows that substantial reductions in excess death can be achieved through introduction of universal health insurance, effective public health and other policies relating to drug use, and effective gun control. Overall, the excess deaths resulted in a life expectancy reduction for Americans of 4.3 years for males and 5.8 years for females, in 2019 before Covid caused extra reductions – see the final figure below.


Note: Most of the estimates for 2019 and 2020 in this post are based on provisional figures and have varying degrees of uncertainty. I plan to refine and improve some of the inputs and calculations, and final results may change a little. The big picture is unlikely to change significantly. In particular, the estimated total Covid deaths for 2020 may change as improved analyses are carried out on complete death registration data. The US had a significant rise in death rates for causes other than Covid-19 in 2020 and some of this may turn out to be undiagnosed Covid deaths. This will not make major differences to the provisional decomposition shown above.

Posted in Global health trends | Tagged , , , , , , , , , , , , , , , , , , | Leave a comment

New estimates of total global COVID-19 deaths from IHME

A few days ago, the Institute for Health Metrics and Evaluation (IHME) released estimates of total COVID-19 deaths for countries based on comparisons of total deaths from all causes during the pandemic period with the expected total deaths based on projections of deaths in years before 2020.

IHME estimate that by 3 May 2021, the total number of COVID-19 deaths globally was 6.93 million, more than double the reported number of deaths of 3.24 million. The estimated total for the USA was 905,289, 58% higher than the reported number of 574,043 deaths.

The figure below from the IHME website shows a map of the predicted ratio of total COVID-19 deaths to reported COVID-19 deaths for March 2020 to April 2021. Ratios range from very high levels in many Eastern European and Central Asian countries to ratios that are much closer to 1 in several high-income countries. For most countries in sub-Saharan Africa, which have reported relatively low numbers of COVID-19 deaths, the estimated ratios range from about 1.6 to 4.1, suggesting that the total number of COVID-19 deaths in the region is several times higher than previously thought. Similarly, India, the country with the most recent severe wave of cases and deaths, is estimated to have an overall ratio of 2.96, which implies that the total COVID-19 death toll to date is much higher than what has been reported.

IHME predicted ratios of total COVID-19 deaths to reported COVID-19 deaths

Its just on a year since they released projections of total COVID-19 deaths based on a clearly stupid modelling process (see my earlier post) and which was widely criticized by infectious disease modellers and epidemiologists (see here for example). There is evidence the IHME’s extraordinarily optimistic projections for total deaths were seized on by the Trump Whitehouse to minimize the need to do anything to address the pandemic.  So have they done a better job this time?

From reading the summary of the methods they have used (see here) it would seem so. They are much more in their comfort zone in the type of modelling needed.  They are taking an approach to estimate excess deaths based on total recorded deaths minus a counterfactual baseline determined from projection of death rates for previous years. But additionally, and beyond what has been done by other estimates of excess deaths, they say that they are aiming to take into account changes in death rates during the pandemic from the following causes:

  1. deaths caused by COVID-19 infection
  2. the increase in mortality due to needed health care being delayed or deferred during the pandemic;
  3. the increase in mortality due to increases in mental health disorders including depression, increased alcohol use, and increased opioid use;
  4. the reduction in mortality due to decreases in injuries because of general reductions in mobility associated with social distancing mandates;
  5. the reductions in mortality due to reduced transmission of other viruses, most notably influenza, respiratory syncytial virus, and measles; and
  6. the reductions in mortality due to some chronic conditions, such as cardiovascular disease and chronic respiratory disease, that occur when frail individuals who would have died from these conditions died earlier from COVID-19 instead.

They refer to a Netherlands study that suggested direct COVID-19 deaths may be higher than estimated excess deaths because deaths due to some other causes have declined during the pandemic.

This is all good stuff, but they then go on to say that there is insufficient data to estimate the impact on excess mortality rates of these causes other than COVID-19, and so they just calculate total excess deaths like everyone else has to date, as total deaths minus counterfactual expected deaths based on projection of previous years deaths and assume the excess is all due to COVID-19. Disappointing! They do make a back-of-the-envelope estimate that there may have been a reduction of up to 615,000 deaths globally, resulting from behavioral changes. 

For countries without available deaths data, and where the available information suggests substantial under-reporting of COVID-19 deaths, the IHME developed a covariate-based prediction model based on infection-detection rates and location-specific fixed effects derived from published studies. As usual, their models are complex and  not easily examined to understand exactly how they work and whether in fact they are producing defensible results.

The COVID-19 death statistics most commonly reported and available on websites such as or Our World in Data are mostly based on confirmed COVID-19 deaths, usually defined as a death within 28 days of a positive COVID-19 test, and reported by many countries daily or weekly. Clearly these statistics depend on testing rates and quality of reporting systems. In many countries particularly in the early phase of the pandemic deaths were only recorded from hospitals, missing many at home and in nursing homes.

A number of researchers and media organizations are estimating the different between total recorded deaths and the predicted expected deaths based on trends for earlier years, similar to the estimates just released by IHME. These estimates refer to varying time periods from the start of the epidemic or to weekly totals and can be quite sensitive to projection methods for estimating the counterfactual expected numbers of deaths.  Excess deaths also represent not only the impact of COVID-19  but also the mostly as yet unknown impacts of mortality increases and decreases from other causes discussed above.

I have compiled estimates of the ratio of total excess deaths to reported COVID-19 deaths from a number of sources for comparison with the IHME estimates. The table below shows estimates for selected European countries and the USA, all of which have essentially complete registration of deaths.

* Excess deaths calculated by me from total annual deaths data for years 2015 to 2020 using
ordinary regression of the log of total death rate against year
(1) Kontis et al. Nature 2020
(2) BBC News (2020)
(3) Karlinski and Kobak (2021)
(4) Our world in data (2021)
(5) The Economist (2021)
(6) IHME (2021)

The first thing to note is the quite wide variation in estimates from different sources. In some cases this may relate to different time periods, but for the lower four sources, which all relate to deaths up to either end 2020 or to Feb/Mar 2021, there are some quite wide variations such as a range from 1.2 to 1.6 for the USA, or 0.7 to 1.5 for the USA, and -1.5 to 1.5 for Norway. Norway illustrates a limitation of this approach in countries with low total COVID deaths (reported total around 436 for 2020) where small differences in projection methods can mean a negative versus positive estimate of the excess. Italy and Spain both had quite large epidemics, and their ratios are more consistent across sources.

In general, the IHME ratios are higher than those from other analyses, suggesting that they are projecting lower counterfactual deaths for the pandemic period. Their projection method is much more complex than any of the others, and possibly is doing a better job. But its inherently difficult to assess the “accuracy” of counterfactuals. In this case, I think what is needed is for another group to do a similarly more sophisticated set of projections from the same data, taking into account the various factors such as seasonality that IHME did, and see whether their results are similar. How much method dependence is influencing the differences above is as yet unclear.

Finally, there is a third type of COVID-19 mortality statistic which is starting to become available. Both the USA and UK have released statistics for the year 2020 based on death registration data which show numbers of deaths where the certificate mentioned COVID-19 as an underlying or contributory cause of death. The USA also released the number of deaths in which COVID-19 is specified as the underlying cause by the certifying doctor.  Both of these statistics will differ from either the reported (confirmed) COVID-19 deaths or the estimated excess deaths. COVID-19 was specified as the underlying cause in 91% of the deaths where COVID-19 was mentioned in the USA. The magnitude of this difference, and with other types of COVID-19 mortality estimates will depend in part on the ICD rules for determining underlying cause from all the contributory causes and in part on national (and individual doctor) idiosycracy in applying or modifying the ICD rules.

Posted in Global health trends, Projections | Tagged , , , , , , , , , | Leave a comment

Premodern/modern religious values and happiness

I recently came across rankings of countries by average reported happiness produced for The World Happiness Report which was released on March 20, 2021. This is a report produced by independent experts for the United Nations Sustainable Development Solutions Network and uses data from the Gallup World Poll to produce an average happiness score for over 150 countries for each year from 2005 to 2020.

This year’s report focuses on the change in average happiness from 2017-2019 to 2020 to examine the impact of the COVID-19 pandemic on happiness, and to assess the relationship of these changes to the country-level health impacts of COVID-19 and the government responses to it. However, I was more interested in examining the potential relationship of happiness to the latent variable measure I developed for premodern versus modern religious values (see earlier post here).

The main measure used for happiness in the World Happiness Report is based on the national average response to the question on life evaluation in the Gallup World Poll. The English wording of the question is

“Please imagine a ladder, with steps numbered from 0 at the bottom to 10 at the top. The top of the ladder represents the best possible life for you and the bottom of the ladder represents the worst possible life for you. On which step of the ladder would you say you personally feel you stand at this time?”

This measure is also referred to as the Cantril life ladder. It provides a more stable measure of the satisfaction of people with their lives than the two other measures also included in the Gallup World Poll: positive and negative affect. Positive affect is the degree to which a respondent experienced happiness, laughter and enjoyment yesterday. Negative affect is the degree to which a respondent experienced worry, sadness and anger yesterday.

For this analysis, I averaged the national average life ladder values for the three years 2017 to 2019 for each of 153 countries and territories. I also added modernity and religiosity latent values for 102 of these countries for years 2017-2020 from my previous analysis of the World Values Surveys and the European Values Study. I then did a regression analysis of the association of these latent values and a number of other covariates with the national average happiness, which I used to decompose the contribution of each of these factors to the overall national happiness. I describe that analysis in more detail below, but first I present a graph of the results which also gives the ranking of countries from highest to lowest happiness score.

Finland holds the rank of the happiest country in the world for the third consecutive year, with an average score of 7.81 on the 0-10 scale. It is followed by Denmark, Switzerland, Iceland and Norway. Among the 102 countries included in the chart above, Australia came in at 12th position with a score of 7.22 and the USA in 16th position with a score of 6.94. The countries with the lowest scores on the graph was Rwanda, with Zimbabwe, Yemen and India just above it. Afghanistan received the lowest score, followed by South Sudan, in the overall results for all the countries included in the World Happiness Report.

The World Happiness Report also includes the results of a regression analysis used to attribute the average reported happiness score of each country for years 2018-2020 to its average income, healthy life expectancy and four social factors: social support, freedom to make life choices, generosity and perceptions of corruption. I note that the healthy life expectancy measure used was that I produced for WHO Member States in 2016, when I worked for WHO.

I carried out a similar regression analysis including not only the above explanatory factors, but also the following potential factors:

modernlatent variable for premodern/modern religious values (see post)
religioslatent variable for religiosity (practicing, non-practicing, non-religious, atheist) (see post)
hchuman capital: average years of education of persons aged 25+ (International Futures Version 7.31)
gfrglobal fertility rate (UN Population Division 2017)
urbanper cent of population living in urban areas (UN 2018)
giniincome inequality index for 2017-2019 from World Bank (2020)
stuntingper cent of children aged 0-4 who are stunted (WHO 2017)
bmibody mass index – people aged 40-64 years (Kontis et al 2014)
riskaverage risk measure for five categories of natural disaster (INFORM)
fsifragile states index (The Fund for Peace)
democEIU democracy index (Economist Intelligence Unit)
ciriCIRI Human Rights Index (CIRI Human Rights Data Project)
ptsiPolitical Terror Scale (Worldwide Governance Indicators – WGI)
rlerule of law effectiveness (WGI)
geegovernment effectiveness (WGI )
pvepolitical stability and absence of violence (WGI )
ccecontrol of corruption effectiveness (WGI
ymenpctproportion of the population that are 15-30 year old males
(UN Population Division 2017)
hedbalhedonic balance calculated as positive affect score (GWP) minus negative affect score (GWP)
avelfaverage ethnolinguistic fractionalization (La Porta 1999)
frac_ethnethnic fractionalization ( Alesina et al 2003 )
frac_langlinguistic fractionalization ( Alesina et al 2003 )
frac_religreligious fractionalization (Alesina et al 2003)

The other potential covariates are those that were included in the World Happiness Report analysis:

lngdplatent variable for premodern/modern religious values (see post)
halelatent variable for religiosity (practicing, non-practicing, non-religious, atheist) (see post)
socsuppthe national average of the binary responses (either 0 or 1) to the GWP question “If you were in trouble, do you have relatives or friends you can count on to help you whenever you need them, or not?”
freedomfreedom to make life choices is the national average of responses to the GWP question “Are you satisfied or dissatisfied with your freedom to choose what you do with your life?”
generosgenerosity is the residual of regressing national average of response to the GWP question “Have you donated money to a charity in the past month?” on GDP
per capita
corruptthe corruption measure is the national average of the survey responses to two questions in the GWP : “Is corruption widespread throughout the government or not” and “Is corruption widespread within businesses or not?” The overall perception is just the average of the two 0-or-1 responses.

The following plots provide scatterplots for the happiness index versus lngdp (r = 0.8) and versus modern (r = 0.78).

I started with the regression model used in the World Happiness Report but with the average happiness for 2017-2019, and covariates relating to the same period. This regression explained 81% of the variance across countries in average happiness, though two of the variables, generos and corrupt, had estimated coefficients which were not significant at the 5% level and contributed only very marginally to explained variance. When I added modern to the regression and tested inclusion of other variables, I ended up with the following final model, which explains 84% of overall variance in average happiness across countries:

Note that healthy life expectancy (hale) is no longer included, and that the other new health variables stunting, bmi and risk (disasters) were also not significant explanatory variables. Similarly, the gini measure of income inequality and most of the fractionalization variables were also not retained in the model. Hedonic balance (hedbal) was also unrelated to variations in average happiness on the life ladder. The religious fractionalization variable, frac_relig, was marginally significant (p=0.049) when added to the final model, but this was highly dependent on other variables included. Religious fractionalization reflects the probability that two individuals randomly selected from the population of the country will have different religions. The estimated coefficient for frac_rel was -0.35, so that average happiness decreases with increasing religious fractionalization. I did not include this variable in the decomposition of the contributions of the various factors to the overall happiness of each country shown in the plot above.

The first six sub-bars for each country/territory in the plot above reflect my estimate of the contributions of the variables included in the regression analysis to the average happiness score in that country. The final dark blue bar includes two elements. The first is the residual error, the part of the national average that the model does not explain. The second is the estimated average happiness in a mythical country called dystopia, since its score is the model’s predicted score (1.82) for an imaginary country having the world’s worst observed values for each of the six variables. With dystopia and the residual included, the sum of all the sub-bars adds up to the actual average happiness score on which the rankings are based.

Across all 101 countries on average, GDP per capita and social support each explained around 17-18% of the total happiness score, followed by freedom to make life choices (10%), modern religious values (8%), population proportion of young men (8%) and state repression (7%). Explanatory values vary substantially across individual countries of course.

The proportion of young men (ymenpct) was included as a potential variable because it is an explanatory variable for higher rates of homicide. Based on this association, I expected the model to estimate a negative coefficient for ymenpct, but it was in fact positive, so higher proportions of population being young men are associated with increased happiness. I have no idea whether this reflects any sort of real causal association or not. The association of self-reported happiness with age is generally U-shaped with declining happiness to mid-life and then increasing again. So possibly higher proportions of young men in the population is simply a marker for higher proportions of happier people (male and females).

It is of course not possible to conclude that any of the associations found here are necessarily causal, or that the estimated contributions reflect the real overall contributions if causal pathways could be identified between the various predictors as well as with the outcome happiness measure. However, it seems likely to me that average income per capita, social support, freedom to make life choices, and more modern religious values are indeed all independent causes of increased happiness, although there may well be more complex causal pathways (such as higher income giving more freedom to make life choices, as would more modern religious values) and also that some of these variables may be acting in part as markers for other unmeasured causes of happiness.

Posted in Global health trends | Tagged , , , , , , , , , , , , , | Leave a comment

Switzerland: COVID-19 situation and vaccine rollout

Switzerland has fully vaccinated 8% of the population, somewhat ahead of France and Germany and some other EU countries (see first figure). For some reason, the percent of people who have received at least one vaccine is no higher than France or Germany, who I presume are prioritizing first vaccinations over second. The percent vaccinated in Geneva is a little higher, probably around 9% and has been focused on people with various conditions and those 65 and over. The cantonal authorities are now extending vaccination to two new target groups – people aged 55-64 and 45-54.

The second figure shows that daily new cases of covid-19 are rising again in Switzerland (with very similar figure for Germany), though not as much as France which looks like it might be having a third wave. Sweden has even higher new cases per day at 625 per million population, currently the highest rate in Europe. Today, the Swiss Federal Government announced the easing of current restrictions. Restaurant terraces, cinemas, theaters, gyms and sports stadiums may reopen from Monday with some restrictions.  Not sure this is wise given that new cases in Switzerland are currently rising.

While new cases and deaths per million have decreased in UK and are now lower than those for most European countries, cumulative total deaths per million are highest for the UK, among the selected countries in the final figure.

Posted in Global health trends | Tagged , , , , , , | Leave a comment

What are the risks of the AstraZeneca vaccine?

Currently, several EU countries are suspending use of the AstraZeneca vaccine, or recommending it not be used in various younger age groups, due to reports of blood clot risks. Australia has decided to delay rollout of vaccination for younger adults after restricting use of the AstraZeneca vaccine to people over 50. I was curious to see what the evidence said about the risks and the benefits, and here is a summary of my impressions. This is NOT a comprehensive or systematic review, just my take on the information I found

There are various population-based figures on the total incidence of deep vein thrombosis cases following vaccination with the AstraZeneca Covid-19 vaccine in the general population. The EU’s drug safety database had recorded 169 cases of blood clots in the brain, also known as cerebral venous sinus thrombosis, and 53 cases of those in the abdomen, or splanchnic vein thrombosis, in the EU and UK  — where some 34 million people had received the vaccine (1). That corresponds to an incidence of around 6 per million population.

Among more than 20 million people who have been vaccinated with the AstraZeneca vaccine in the UK so far, 79 cases of rare blood clots with low platelets have been reported, as well as 19 deaths according to the UK’s regulatoryauthority (2). This equates to around one case per 250,000 people vaccinated— or 4 per million—and one death in a million. EU officials have also reported that the number of “thromboembolic events” reported after vaccination was actually lower than expected in the general population (in 20 million people we would expect around 200 cases of deep vein thrombosis per week from other causes).

A group at Cambridge University has assessed the relative benefits and harms of the AstraZeneca vaccine (3). These are summarized in the three tables for population groups at low, medium and high risk of infection (UK infection rates). Presumably all Australians would be in the low risk group. The risk-benefit calculation looked at the intensive care admissions prevented versus potential blood clots. In all age and risk groups except low risk under 30s, the potential benefits far outstripped the potential harms. In people aged 20-29 and low risk—meaning that they have no conditions that make them more at risk of developing severe covid-19 illness—the harms slightly outweighed the benefits.

So there really is not any significant excess risk even for the young adult age group. A number of people have pointed out that use of oral contraceptives is associated with a substantially higher risk of blood clots, but that is still accepted as “very low risk”.

I took a look at the evidence on risk of blood clots associated with oral contraceptives, and systematic reviews have assessed the relative risk of deep vein thrombosis on combined oral contraception as 3 to 5-fold, resulting in an absolute risk of around 0.05% per year for a healthy adolescent (4-6). That corresponds to an incidence of 500 cases per million per year for oral contraceptive users compared to the risk of 4 per million following AstraZeneca vaccination. That risk is accepted by the medical profession and contraceptive users, and described in the literature as a “very low risk”.  In contrast, the AstraZeneca risk is substantially outweighed by the benefits in all except low risk your adults, where the risk and benefit is close to equal.

Its not clear to me that the evidence actually suggests there is an excess risk of mortality associated with the AstraZeneca vaccine, but lets assume there is indeed an extra risk of 1 death per million vaccinations.  The following table gives some comparative risks to put that in perspective. Most of these estimates derive from the UK population in 2010 or recent data for USA (7,8) and they are only intended to convey an indicative magnitude for various risks.

The risk of death from blood clots caused by the Astrazeneca virus are of the same order as the risk of dying by being struck by lightning and about half the risk of cycling 100 km.  That is indeed if there is a causal link and the clots are not just coincidental to the vaccination. In any case, the benefit seems to substantially exceed the harm for all groups except low risk adults under 30 years of age, where the risk of a blood clot (not death) is slightly greater than the risk of covid-19 infection resulting in hospital admission to ICU.

References

  • National Public Radio, 7 April 2021. EU Regulator: AstraZeneca Vaccine Effective; Blood Clots May Be A Rare Side Effect. ttps://www.npr.org/sections/coronavirus-live-updates/2021/04/07/984998679/eu-regulator-astrazeneca-vaccine-effective-blood-clots-may-be-a-rare-side-effect
  • Mahase E. AstraZeneca vaccine: Blood clots are “extremely rare” and benefits outweigh risks, regulators conclude BMJ 2021; 373 :n931 doi:10.1136/bmj.n931
  • https://wintoncentre.maths.cam.ac.uk/news/communicating-potential-benefits-and-harms-astra-zeneca-covid-19-vaccine/
  • Trenor CC 3rd, Chung RJ, Michelson AD, et al. Hormonal contraception and thrombotic risk: a multidisciplinary approach. Pediatrics. 2011;127(2):347-357. doi:10.1542/peds.2010-2221
  • Vandenbroucke JP, Rosendaal FR. End of the line for “third-generation-pill” controversy? Lancet. 1997 Apr 19;349(9059):1113-4. doi: 10.1016/S0140-6736(05)63015-2.
  • Lowe GD, Rumley A, Woodward M, Reid E. Oral contraceptives and venous thromboembolism. Lancet. 1997 May 31;349(9065):1623. doi: 10.1016/S0140-6736(05)61660-1.
  • Spiegelhalter, David; Blastland, Michael. The Norm Chronicles (p. 15). Profile. Kindle Edition.
  • Carly Hallman. Odds and Chances of Death by Activity and Behaviorhttps://www.titlemax.com/discovery-center/lifestyle/life-expectancy-by-activity-behavior/
Posted in Global health trends | Tagged , , , , , , , | Leave a comment

Global deaths attributable to drugs, alcohol and tobacco

In an earlier post, I presented estimates of drug overdose deaths in 2019 for the world and selected countries. The USA accounted for an astonishing 40% of global drug overdose deaths and an even higher percentage of global opioid overdose deaths.

More recently, it has been estimated that deaths from drug overdose rose by an estimated 13% in the first half of 2020 compared to 2019 [1]. In some states, drug related deaths climbed by over 30%. Its unclear whether the increasing social isolation, pressure on health services, and lockdowns associated with the pandemic, particularly in the second half of 2020, will result in continued increase [2].

I’ve updated the previous estimates for years 2000-2019 to take into account revised inputs from the 2019 Global Burden of Disease [3] used for projecting trends to 2019 from an earlier update to 2017 (see previous post). The revised estimates are slightly higher than those presented in the previous post:

Assuming a somewhat more conservative 10% increase for 2020 than observed in the first half of the year, there were a little over 80,000 drug overdose deaths in the USA in 2020, of which the majority (close to 60,000) would have been opioids. Initially driven by prescription opioid painkillers in earlier years, users migrated first to heroin and then to fentanyl, which is cheaper (much of it illicitly made). The global average death rate for drug use disorders was 24 per million population, compared with 225 per million for the USA. In other words, the USA rate is 9 times higher than the global average.

Trends in drug use disorder deaths by WHO Region

The following regional graphs highlight how different the situation in the USA is to the rest of the world (grouped in the World Health Organization regions).

Drug overdose death rates per million population, for USA and WHO Regions, 2000-2019

The analysis of death registration data is complicated by the need to distribute the accidental poisoning category for “other and unspecified chemicals and noxious substances” (X49) to the specific categories for alcohol and drug use disorders (opioids, cocaine, amphetamines, cannabis and “other drugs”) and to accidental poisoning (non-drug and non-alcohol). Additionally, there is a category F19 in the mental health chapter for “multiple drug use and unspecified drug use disorders” which is used to code deaths in some countries and also must be redistributed appropriately. So there is some uncertainty associated with the estimates for countries with death registration data, and even more for countries without good data.

Total attributable drug deaths

A little over 10 years ago, I carried out an updated comparative risk assessment of the total deaths and burden of disease associated with various risk factor exposures, including drug use (licit and illicit) [4]. More recently, drawing on reviews and analyses carried out by expert groups advising WHO and IHME on drug and alcohol use, I updated the drug use estimates for the year 2017 to include the excess deaths from infectious diseases associated with injecting drug use, as well as road injury and suicide. These analyses were carried out using comparative risk assessment methods [5].

I’ve projected these analyses forward to 2019 using the updated cause of death estimates to 2019 and assuming previously calculated population attributable fractions for 2017 remain constant to 2019. These will provide ballpark estimates at global level that will give a reasonably good picture of the comparative contributions of the main drug types and the main disease and injury outcomes to the total mortality attributable to drug use.

Before giving a little more detail on data and methods used, I will present a summary of the overall results for 2019 at global level.

For the calculation of the fraction of cause-specific deaths attributable to drug use, country-specific time series of prevalences of drug dependence were taken from the Global Burden of Disease [6] based on estimates prepared by an expert group [7].

Road injury and suicide

Relative risks for deaths from road injury by drug type were based on results of a systematic review by Elvik (8). This evidence is discussed in more detail in a later WHO publication (9). Based on advice from a WHO Technical Advisory Group (10), the relative risk for road injury death associated with drug use was based on studies reported by Elvik (8) which controlled for at least three of the main confounding factors.

Relative risks for suicide were derived from a systematic review and meta-analysis (11). The evidence for association of suicide with cannabis use was considered not strong enough. Insufficient information was available to quantify any risk of psychosis associated with cannabis use.

Attributable fractions for injecting drug use (IDU)

Relative risks for deaths due to HIV, HBV and HCV infection due to injecting drug use have been estimated by Degenhardt et al based on available studies (12, 13).Degenhardt et al (14, 15) estimated the fractions of deaths due to HIV, HBV and HCV (and the liver cancer and cirrhosis attributed to HBV and HCV)  by country for year 2013.  These were used in GBD2017 to estimate population attributable fractions (PAFs) for these causes by country, year, age and sex.  These PAFs are then applied to my provisional cause of death estimates for years 2000-2019 to estimate attributable deaths for IDU. These deaths are apportioned to the specific drug types using an informal review of literature carried out by WHO colleagues for the 2017 update [16]. The drug type distribution is assumed to hold constant across years 2000-2019.

Previous updates of attributable deaths for drug use have included estimates of uncertainty, combining uncertainty in cause of death estimates [17] with uncertainty in drug use prevalences and relative risks.  I haven’t updated these uncertainty intervals for this update, which provides a broad picture of the current relative health impacts of the main types of drug use. It should be remembered that these drug categories include licit and illicit use.

Alcohol, tobacco and drugs

The following table and figure summarizes and compares the estimated global deaths attributable to tobacco, alcohol and drugs in 2019. Tobacco causes by far the largest death toll, almost 9 million deaths representing almost 15% of all deaths globally in 2019. Following is alcohol, responsible for around 5% and then drugs causing just under 1% of global deaths. For comparison, there were 1.83 million recorded deaths from Covid-19 in 2020, though actual total deaths were undoubtedly higher (perhaps as many as 2.7 to 3.6 million).

The attributable burden of disease, in terms of DALYs which measure equivalent lost healthy years of life from both poor health and premature mortality, gives relatively higher weight to alcohol.  While I’ve not updated DALY estimates to 2019, the GBD 2019 [3] estimates global attributable DALYs for 2019 as 230 million for tobacco, 93 million for alcohol, and 31 million for drug use. Unlike tobacco which kills most people at older ages with relatively limited periods of disability, alcohol and drug use are associated with significant loss of health at younger ages, and drug deaths in particular tend to occur at younger adult ages rather than older ages. The estimates for alcohol and drugs are somewhat higher than those of the GBD 2019, at 2.4 million for alcohol and 494 thousand for drugs [3].

The attributable burden of disease, in terms of DALYs which measure equivalent lost healthy years of life from both poor health and premature mortality, gives relatively higher weight to alcohol.  While I’ve not updated DALY estimates to 2019, the GBD 2019 [3] estimates global attributable DALYs for 2019 as 230 million for tobacco, 93 million for alcohol, and 31 million for drug use. Unlike tobacco which kills most people at older ages with relatively limited periods of disability, alcohol and drug use are associated with significant loss of health at younger ages, and drug deaths in particular tend to occur at younger adult ages rather than older ages.

Acknowledgements

Previous estimates of deaths attributable to drug use prepared for WHO for 2017 and earlier years benefited from substantial advice and inputs from WHO expert advisors and staff, particularly Juergen Rehm, Louisa Degenhardt and Vladimir Pozniak. The  estimates presented here are my ballpark estimates based on quick update and do not represent the views of WHO or anyone else.

References

[1] Wen LS, Sadeghi NB. The opioid crisis and the 2020 US election: crossroads for a national epidemic. The Lancet Volume 396, ISSUE 10259, P1316-1318, October 24, 2020. DOI:https://doi.org/10.1016/S0140-6736(20)32113-9

[2] Katz J, Goodnough A, Sanger-Katz M. In Shadow of Pandemic, U.S. Drug Overdose Deaths Resurge to Record. New York Times, July 15, 2020. Available from https://www.nytimes.com/interactive/2020/07/15/upshot/drug-overdose-deaths.html

[3] Global Burden of Disease Collaborative Network. Global Burden of Disease Study 2019 (GBD 2019) Results. Seattle, United States: Institute for Health Metrics and Evaluation (IHME), 2020. Available from http://ghdx.healthdata.org/gbd-results-tool.

[4] Mathers CD, Stevens G, Mascarenhas M. Global health risks: mortality and burden of disease attributable to selected major risks. WHO, Geneva, 2009.

[5] Ezzati M, Lopez AD, Rodgers A, Vander Hoorn S, Murray CJL. Comparative quantification of health risks: global and regional burden of disease attributable to selected major risk factors. Geneva: World Health Organization. 2004

[6] Global Burden of Disease Collaborative Network. Global Burden of Disease Study 2017 (GBD 2017) Results. Seattle, United States: Institute for Health Metrics and Evaluation (IHME), 2018.

[7] Peacock A., Hall W., Degenhardt L. (2019) Epidemiology of Substance Use Internationally. In: Sloboda Z., Petras H., Robertson E., Hingson R. (eds) Prevention of Substance Use. Advances in Prevention Science. Springer, Cham. https://doi.org/10.1007/978-3-030-00627-3_2.

[8] Elvik, R (2013). Risk of road accident associated with the use of drugs: a systematic review and meta-analysis of evidence from epidemiological studies. Accid Anal Prev. 2013 Nov;60:254-67. doi: 10.1016/j.aap.2012.06.017. Epub 2012 Jul 9.

[9] World Health Organization  (2016). The health and social effects of nonmedical cannabis use. Geneva: WHO.  Available at https://www.who.int/substance_abuse/publications/msb_cannabis_report.pdf

[10] World Health Organization (2015). Technical Advisory Group on Alcohol and Drug Epidemiology Meeting (unpublished report). Geneva: WHO, November 2015.

[11] Ferrari AJ, Norman RE, Freedman G, Baxter AJ, Pirkis JE, Harris MG, Page A, Carnahan E, Degenhardt L, Vos T, Whiteford HA (2014). The burden attributable to mental and substance use disorders as risk factors for suicide: findings from the Global Burden of Disease Study 2010. PLoS One. 2014 Apr 2;9(4):e91936. doi: 10.1371/journal.pone.0091936. eCollection 2014.

[12] Degenhardt, L, Whiteford HA, Ferrari AJ, Baxter AJ, Charlson WD,  Hall WD, et al. (2013). Global burden of disease attributable to illicit drug use and dependence: findings from the Global Burden of Disease Study 2010. Lancet 382 (9904): pp. 1564– 74.

[13] Van Den Berg C, Smit C, Van Brussel G, Coutinho R, Prins M; Amsterdam Cohort (2007). Full participation in harm reduction programmes is associated with decreased risk for humanimmunodeficiency virus and hepatitis C virus: evidence from the Amsterdam Cohort Studies among drug users. Addiction. 2007 Sep;102(9):1454-62.

]14] Degenhardt et al (2016). Estimating the burden of disease attributable to injecting drug use as a risk factor for HIV, hepatitis C, and hepatitis B: findings from the Global Burden of Disease Study 2013. Lancet Infectious Diseases 2016, 16:1385-98.

[15] Degenhardt et al (2017). Global prevalence of injecting drug use and sociodemographic characteristics and prevalence of HIV, HBV, and HCV in people who inject drugs: a multistage systematic review. The Lancet Global Health 2017; 5:31192-207.

[16] Poznyak V, et al. (2016). Unpublished review of the distribution of injecting drug use by type. World Health Organization.

[17] World Health Organization (2018). WHO methods and data sources for global burden of disease estimates 2000-2016 (Global Health Estimates Technical Paper WHO/HIS/IER/GHE/2018.4)

[18] World Health Organization (2018). Global Status Report on Alcohol and Health 2018. WHO: Geneva. https://www.who.int/publications/i/item/9789241565639

Posted in Global health trends | Tagged , , , , , , , , , , , , , , | Leave a comment

A Covid Performance Index for countries

An Australian think-tank, the Lowy Institute has produced a Covid Performance Index and used it to examine the impact of geography, political systems, economic development and population size on the Covid-19 outcomes around the world.

The COVID Performance Index compiled by the Lowy Institute ranked 98 countries’ handling of the COVID-19 outbreak, finding New Zealand performed the best while Australia sits in eighth place. Switzerland was ranked 53rd of the 98 countries, the UK 66th and the USA was in the bottom five, one rank ahead of Iran at 95th.

Democracies have performed slightly better over the course of the pandemic.Credit:Lowy Institute

Democracies slightly outperformed authoritarian countries in suppressing the coronavirus, overall. While the Asia-Pacific region performed best on average, the study did not include China due to lack of complete information on testing rates. While democracies performed worse at the beginning of the pandemic and there were some notable exceptions including the United States and Britain, they pulled ahead of authoritarian and hybrid states as the outbreak worsened, which suggests democracies were better at learning from their mistakes in acting too slowly. Europe showed the greatest improvement over time, before succumbing to a more severe second wave in final months of 2020.  The more open borders in Europe left countries vulnerable to renewed outbreaks in neighbouring countries.

Comparison of the experience of the UK and USA with other high income countries suggests that the determining factor in how effectively countries responded to the crisis was whether citizens trust their leaders and whether there are competent and effective state bureaucracies and leaders that rely on scientific advice.

The study measured a number of key indicators including confirmed cases, deaths, cases per million people, deaths per million people and cases a proportion of tests. The method used 6 indicator of pandemic performance: confirmed cases, confirmed deaths, confirmed cases per million, confirmed deaths per million, confirmed cases as a proportion of tests and tests per thousand people.  An average of the rankings across these six indicators was normalized to range from 0 to 100.

Synthetic indices like this that aggregate disparate indicators numerically are quite commonly used to compare countries and track progress (eg. The Human Development Index which averages life expectancy, mean years of schooling and average income per capita). The drawbacks of such indicators is that the resulting numbers are not directly interpretable in the same way that single outcome indicators are, and the results are dependent on a number of essentially arbitrary choices in the indicators chosen and how they are weighted and averaged. Having been involved all my career with summary measures of health, I would have a similar preference for identifying a single explicit outcome measure for a pandemic and using that to compare performance.

The Lowy Institute index is based on two absolute measures (cases, deaths) and four “rates”. This gives a certain weight to absolute measures of the pandemic impact as well as measures adjusted for population size and testing rates. I’m in two minds about whether this is reasonable. On the one hand, the inclusion of absolute measures may partly or largely account for why smaller countries generally did better. On the other hand, the absolute impact of disasters in terms of lives lost or lost health is also an important outcome. There is probably no “right” choice, but I’m not sure whether the implied weights of the Lowy Index are necessarily a good choice.

Posted in Global health trends | Tagged , , , , , , | Leave a comment

Fundamentalist values, culture and religion

In an earlier post, I summarized trends in religiosity (practicing, non-practicing, non-religious and atheist) for countries in the World Values Survey (WVS) and European Values Study (EVS) from 1980 to 2020 [1-4]. In this post, I take a closer look at whether religious values are changing in the direction of increasing or decreasing fundamentalism and the extent to which religion and/or culture are associated with rejection of scientific evidence and findings: important questions for a world facing the twin global crises of climate change and the coronavirus pandemic.

Before describing in more detail the conceptualization of modern and pre-modern religious values and the analysis of the WVS/EVS data, I present one of the outputs, a religious values map for countries based on the most recent waves of the surveys. The horizontal axis plots to a continuous latent variable showing degree to which religious values are pre-modern (to the left) to modern (to the right). The vertical axis plots a rescaled religiosity variable in which higher values indicate higher levels of religious belief and practice.

Most religions developed in the pre-modern era and their sacred texts and teachings incorporate pre-modern culture and values to varying extents. Peter Herriot has written extensively on fundamentalist religious beliefs, characterized these movements as attempts to return to the pre-modern origins of their faith as prescribed by their sacred books [5]. He identifies five main general characteristics of fundamentalist religious movements:

  • Reactivity: hostility towards the secular modern world
  • Dualism: the tendency to evaluate in starkly binary terms, as good or bad
  • Authority; the willingness to believe and obey the sacred book of the movement and/or its leaders
  • Selectivity: the choice , from the sacred book or the tradition, of certain beliefs and practices in preference to others
  • Millennialism: the belief that God will triumph in the end and establish his kingdom on earth.

Other common characteristics include prejudice towards minorities and authoritarian aggression – in some cases resorting to violence. Fundamentalist groups may be mainly religious in focus, or the religious element may be strongly associated with nationalism or ethnic identity. Fundamentalists seek to erase the distinction between secular and sacred and impose their form of religious beliefs on all through political action or authoritarian control. In the 21st century, the mobilization of the fundamentalist vote in the USA has been an important contributor to the election of the two most recent Republican presidents, and we now see a bizarre situation in the USA where the fundamentalists strongly support an amoral President with severe narcissistic personality disorder.

Herriot perceives fundamentalist movements around the world and in all major religions as having arisen relatively recently in the twentieth century as a reaction to modernity [5] and sees them largely as subgroups within the overall religion. I disagree with this.  Some religions which have not gone through a Reformation process involving separation of church and state remain largely embedded in pre-modern beliefs and values. In other cases, as religious institutions have evolved along with modern science, technology, culture and moral values, subgroups have rejected this evolution as going further than their moral comfort zone. Rather than attempt to define “fundamentalist” values, I’ve sought rather to identify WVS/EVS questions that relate to “pre-modern” values associated with earlier stages of moral development (as defined by work of Piaget [6], Kohlberg [7] and Gilligan [8]. Gebser [9] and Wilber [10] have elaborated the link between these stages of individual development and the broad evolution of cultures over the course of human evolution through magic, mythic, rational, to integral stages. Wilber also refers to the mindsets associated with the three broad stages of moral values as egocentric, ethnocentric and worldcentric [11]. As moral values evolve through the three broad stages, the size of the in-group (“us”) with which an individual identifies typically expands from tribe to ethnic group or nation to all humanity.

Pre-modern moral values and related religious values focus on absolute rules, obedience and punishment and an individual is good in order to avoid being punished. In stage 2, the individual internalizes the moral standards of the culture and is good in order to be seen as a good person by oneself and others. Moral reasoning is based on the culture’s standards, individual rights and justice. In stage 3, the individual becomes aware that while rules and laws may exist for the greater good, they may not be applicable in specific circumstances. Issues are not black and white, and the individual develops their own set of moral standards based in universal rights and responsibilities.

Because pre-modern religious teaching is expressed and interpreted in mythic terms, it may appear to conflict with scientific understanding of the natural world. A person with pre-modern values may thus reject scientific findings,  whereas another with modern values will understand that the myths communicate aspects of the human condition, but are not to be interpreted literally, and that the domain of religion relates to meaning, values, ethics, and does not generally conflict with the domain of science.

I reviewed questions included in the WVS and EVS to identify those (a) most relevant to distinguishing pre-modern and modern moral values and (b) widely available in the survey waves. The selected questions are summarized in the following table:

The availability of these questions in the WVS and EVS waves is summarized in the following figure. Responses for belief in heaven and hell were imputed as for belief in God (documented in the previous post). The responses for belief in life after death were very similar to those for belief in heaven, and only the latter variable has been used. Other potential questions relating to religious values and beliefs were included in either the WVS or EVS but not both and have not been further considered here. A set of questions on levels of trust in those with other religions, from the local region, and from the same country is available for all recent waves of EVS and WVS, as are questions of sense of belonging to local community, nation and world. However, these questions showed little association with variations in religious values and beliefs, or with culture zone, and were not included in further analysis. Similarly, high levels of disagreement that education was more important for boys was found for all culture zones and religions. Gender equality values were thus based on responses for jobs and political leadership.

To assess how responses to these questions varied across culture zones for practicing religious people compared to others (non-practicing religious, non-religious and atheists), I calculated the per cent of respondents who agreed with the question proposition. For example, for questions with a 10 point response scale, responses 5 and 6 tended to have somewhat higher response rates than others except the question end-points, and I treated the middle categories as sitting on the fence. Hence, the proportion who “agreed” were those who responded 1-4.  The following plots show the differences between the practicing religious and others for the 9 culture zones that I’ve used to group countries in the previous post (see footnote d of that post).

The final plot shows differences in the proportions of people who believe that sex before marriage is not justifiable. It could be argued that this and other issues of sexual freedom are also associated with the pre-modern values of the era when high fertility was important to ensure some children survived to reproductive age, and societies thus discouraged divorce, abortion, homosexuality. But given how recently the demographic transition to low child mortality, low fertility and widespread availability of reliable contraception has occurred, and is in some regions still occurring, and how rapidly values are changing in this domain, I have not included measures of sexual freedom or emancipation in the potential measures for pre-modern/modern values. Because the questions of citizenship and belonging did not show much variation between groups, I have instead relied on questions about gender discrimination and non-acceptance of homosexuality as measures of inclusiveness and acceptance of others.

The following plots examine responses to several key questions by reported religious affiliation and whether practicing or non-practicing. For these plots, people classified as non-religious or atheist are reclassified from reported affiliation to “No religion” and non-practicing.  People who report “No religion” but are practicing are those who report attending religious services at least monthly, that they believe in God and that religion is somewhat or more important). It is clear from these plots for culture zone and religious affiliation that religious values tend to be more similar for practicing and non-practicing within religious affiliation than across religious affiliation, and also generally more similar for religious and non-religious within a culture zone than across a culture zone. I return to the question of whether religion or culture zone explains more of the variation of values across categories of religiosity below.

Estimation of a continuous latent variable for pre-modern/modern values

Based on the above exploration of potential indicators of pre-modern religious values, I constructed a continuous latent variable based on the categorical responses to 12 questions. These questions are shown in the following plot, which shows their item response cut-points on the estimated latent variable. The item response analysis [12, 13] was implemented as an ordered probit model using the stata procedure gsem for generalized structural equation models [14]. The model was fit to the entire dataset for all countries in survey waves 5 to 7.

For ease of presentation, I made a linear transformation of the estimated latent variable by multiplying it by 5 and adding 4.75. This resulted in average religiosity  at country level ranging from -4.6 in Nigeria in 2000 to 4.71 in China in 1990. Note that higher values denote higher prevalence of modern religious values. Country averages for the most recent WVS/EVS wave for 2017-2020 ranged from 0.2 for Pakistan to 9.9 for Denmark, followed by 9.8 for Sweden and 9.3 for Norway. At individual level, 90% of values of the pre-modern -modern values latent variable fall in the range –0.24 to 10.16 with a median value of 4.71.

The following map plots the country average religiosity latent variable against the country average modern values latent variable for the 104 countries in waves 5-7 of the WVS/EVS. Values for wave 7 are plotted. For countries whose most recent data were for wave 6 ( countries) or wave 5 ( countries), values have been projected forwards using average trends across waves for each culture zone. The religiosity latent variable has also been transformed to produce a scale running from 0 (low religiosity) to 10 (high religiosity) by subtracting the estimated religiosity values from 4.5.

The following plot shows culture zone population-weighted averages for the two latent variable for four categories of religiosity. This plot relates to averages for wave 7 only.

In every culture zone, practicing religious people have the lowest score for modern religious values, ie the most pre-modern values) and modern religious values increase with decreasing levels of religiosity. The variation across culture zones is approximately as substantial as the variation across religiosity categories.

An analysis of variance for the premodern-modern religious values latent variable against culture zone, religious affiliation and religiosity found that these variables all explain highly significant proportions of total variation, and that two-way and three-way interactions are also highly significant. In very general terms, the  importance of these factors in explain variance in the latent variable are in descending order: culture zone then religious affiliation then religiosity. Thus the values gap between practicing religious people and atheists is a little larger by culture than by religious affiliation, and the gap varies significantly according to the combination of culture and affiliation.

These plots, and the country or culture zone comparisons show broad patterns but specific small differences should not be over-interpreted. I have not attempted to estimate uncertainty ranges for these statistics because my long experience with analysis of surveys implemented in many countries by different study teams has shown that the most important contributors to uncertainty of statistics is not the survey sample size or sampling issues but other less quantifiable differences in survey implementation, design and translation. Also important are difficult to quantify differences in interpretation and choice of question response categories in different populations. Most of the WVS/EVS surveys are representative samples of adults with sample size ranging between 1 and 2 thousand.  This is quite similar to the typical national opinion poll where uncertainty of statistics is typically around 2 or 3 percentage points.

There are a substantial number of countries where the WVS and EVS both conducted a survey in the same wave. I’ve assessed the median difference between the two surveys for a country to provide some quantification of the typical variation induced by sampling methodology, survey design and implementation. It will not include cross-national differential response associated with language, translation and response category cut-point shifts.

For the religiosity categories at country-level, average prevalences differ by around a median 10% (relative difference). This drops to around 5% at culture zone level.

References

  1. Inglehart, R., C. Haerpfer, A. Moreno, C. Welzel, K. Kizilova, J. Diez-Medrano, M. Lagos, P. Norris, E. Ponarin & B. Puranen et al. (eds.). 2014. World Values Survey: All Rounds – Country-Pooled Datafile Version: https://www.worldvaluessurvey.org/WVSDocumentationWVL.jsp. Madrid: JD Systems Institute.
  2. Haerpfer, C., Inglehart, R., Moreno,A., Welzel,C., Kizilova,K., Diez-MedranoJ., M. Lagos, P. Norris, E. Ponarin & B. Puranen et al. (eds.). 2020. World Values Survey: Round Seven–Country-Pooled Datafile. Madrid, Spain & Vienna, Austria: JD Systems Institute& WVSA Secretariat[Version: http://www.worldvaluessurvey.org/WVSDocumentationWV7.jsp].
  3. Gedeshi, Ilir, Zulehner, Paul M., Rotman, David, Titarenko, Larissa, Billiet, Jaak, Dobbelaere, Karel, Kerkhofs, Jan. (2020). European Values Study Longitudinal Data File 1981-2008 (EVS 1981-2008). GESIS Datenarchiv, Köln. ZA4804 Datenfile Version 3.1.0, https://doi.org/10.4232/1.13486.
  4. EVS (2020): European Values Study 2017: Integrated Dataset (EVS 2017). GESIS Data Archive, Cologne. ZA7500 Data file Version 3.0.0,doi:10.4232/1.13511
  5. Herriot, Peter. Religious Fundamentalism and Social Identity. Routledge, 2014.
  6. Piaget, Jean. The moral judgment of the child. London: Kegan Paul, Trench, Trubner & Co: 1932.
  7. Kohlberg, L. The Psychology of Moral Development: The Nature and Validity of Moral Stages (Essays on Moral Development, Volume 2). Harper & Row: 1984.
  8. Gilligan, Carol. In a different voice: Women’s conceptions of self and of morality. Harvard Educational Review. 1977, 47(4), 481-517
  9. Gebser, Jean. The Ever-Present Origin, authorized translation by Noel Barstad with Algis Mickunas, Athens: Ohio University Press, 1985.
  10. Wilber, Ken. Up from Eden. Anchor Press/Doubleday, 1981
  11. Wilber Ken. Integral Spirituality. Integral Books: Boston and London, 2007.
  12. Hambleton RK, Swaminathan H. Item Response Theory: Principles and Applications. Springer Science and Business Media, New York. 1985.
  13. Nguyen TH, Han HR, Kim MT, Chan KS. An introduction to item response theory for patient-reported outcome measurement. Patient. 2014;7(1):23-35. doi:10.1007/s40271-013-0041-0
  14. Stata Corporation. Stata Structural Equation Modeling Reference Manual. Release 15. StataCorp, Texas, 2017.
Posted in Global health trends | Tagged , , , , , , | 2 Comments

Global trends in religiosity and atheism 1980 to 2020

Ronald Inglehart has recently published an article in Foreign Affairs called “Giving up on God: the global decline of religion” in which he uses data from the most recent wave of the World Health Surveys (WVS) to claim that between 2007 and 2019, the importance of religion has declined in most countries [1]. This is based on a single question on the importance of God in the respondent’s life on a 10-point scale. The average importance declined in 39 countries and increased in only 5.  Apart from the fact that this is based only on a single question on the importance of God, it also does not tell us how regional or global average ratings have changed. Depending on the relative populations and scale shifts in different countries, it could potentially even be consistent with a global average increase.

In this post, I examine country, regional and global trends in religiosity in terms of the four categories of religiosity used in my previous post, and also using a continuous latent variable measuring religiosity based on the religiosity categories (which incorporate religious practice and belief/disbelief in God) plus responses on the importance of religion, the importance of God, and frequency of religious practice. In doing this analysis, I have slightly revised the religiosity categories from those used previously, as follows:

Atheist: A “confirmed atheist” and/or does not believe in God

Non-religious:  A non-religious person who believes in God, but rates the importance of God as 8-10 at the not important end of a 10-point scale.

Non-practicing religious person: A religious person who believes in God and is non-practicing OR a non-religious person who believes in God, is non-practicing, and rates the importance of God in range 1-7.

Practicing religious person: A religious person who believes in God and is practicing, OR a non-religious person who believes in God, is practicing, and rates the importance of God in range 1-6. Respondents are classified as “practicing” if they attend religious services or pray to God outside of religious services at least once a month.

Modified versions of these definitions are used for persons stating affiliation to a non-theist religion and for the predominantly Buddhist countries (see Endnote a). Data from the World Values Surveys (WVS) and European Values Study (EVS) are used to classify the religiosity of over 630,000 respondents in 110 countries over the period 1981 to 2020 [2-5].

In preparing the estimates of religiosity across waves of the World Values Survey, I discovered that the question on belief in God had been omitted for Wave 5 (2005-2009) as well as from a few surveys on other WVS and EVS waves, preventing the classification of religiosity for those country-years. Responses for this question were imputed using country-specific distributions as described in Endnote b.

I was not happy with the projection method I had previously used to project religiosity to year 2020 as it used trend from the 6th to 7th wave which were large in some countries and possibly biased by cross-survey differences in sampling or survey methods. More robust and conservative methods for preparing time series from 1980 to 2020 have now been used as described in Endnote c.

Country-level trends in religiosity and atheism

The following plots show trends in the prevalence of the four religiosity categories from 1980 to 2020 for 6 representive countries from different religious/culture zones. High income countries in Western Europe and North America are characterized by declining religiosity and rising prevalence of atheism. Former Communist countries of Europe are characterized by a drop in atheism after the breakup of the Soviet Union, some rising in the practicing religious and a much larger rise in the non-practicing religious. As noted in my previous post, the non-practicing religious see their religion (Orthodox Christianity or Islam) as a strong marker of cultural belonging and national identity [6]. The vast majority of people in Africa and Islamic countries are religious, though substantial proportions are non-practicing, and the prevalence of irreligion (non-religious and atheists) is very low.

In his article [1], Inglehart notes that the USA has had the sharpest decline in the importance of God of all the countries in the WVS, and now ranks as the 11th most irreligious country (based on the single question he analyzed). The plot above for the USA also shows a very substantial rise in the prevalences of atheism and non-religious across the last two waves, and a corresponding decline in non-practicing and practicing religious prevalences.  Based on my classification of religiosity categories, the USA is now ranked 28th in the world for prevalence of atheism and ranked 5th in the world for the rate of decline of the percent who are practicing religious people. If I exclude China and South Korea due to the difficulties in classifying religiosity in the Sinic countries with high prevalence of non-theistic religion, then the USA has the 3rd highest rate of decline after Chile and Denmark, but these countries all share a very similar rate of decline over the last decade around 3.5% per year.

Based on my estimates for 2020, China has the highest proportion of atheists (79%), followed by Czechia (70%), Sweden (68%), Estonia (64%) and the Netherlands (60%). Other countries in which more than 50% of the population do not believe in God include Norway, the United Kingdom, South Korea, France and Denmark. Other countries of interest (to me) include Australia (45%), Switzerland (36%) and the USA (25%).  I was expecting to see a lower prevalence for the USA and checked the data carefully. There have been 8 survey waves for the USA covering the period 1981 to 2017 and they show that the prevalence of atheism has been rising much faster in the last 15 years than earlier, and that the prevalence of “non-religious” has also been rising, reaching an estimated 10% in 2020. This category includes people who say they believe in God, but are non-religious and rate the importance of God as 8-10 on the unimportant end of a 10-point scale. I suspect these are people who are close to no belief in God, but unwilling to make the step of saying that.  The USA is one of very few countries in the world where this category is more than 1 or 2 percent of the population.

China has the largest prevalence of atheism in the world at an estimated 78% but as the plot shows there has been a substantial shift from the non-religious category to the atheist category and it is difficult to interpret this given the lack of fit of the WVS questions with the non-theist religions that are most common in China.

Religiosity in Iran and other Islamic countries

Iran and other Islamic countries generally report very low levels of atheism, 2.4% on average, and I suspect this is lower than reality because of the quite severe social and legal consequences in many Islamic countries. The WVS uses telephone interviews and its quite likely irreligious respondents would be fearful of being identified if they respond honestly to a telephone interview. A recent internet survey provides some support for this concern. The June 2020 internet-based survey collected responses from 40,000 Iranians living in Iran [7]. Respondents took part in the survey anonymously, and would have felt safer to express their real  opinions than in telephone surveys or surveys conducted at respondents’ residence.

According to this survey, 40.4% of Iranians identify as Muslim, 8% as Zoroastrian and 9% as atheist (12% if those who identify as humanist are included). Most Iranians, 78%, believe in God, but only 37% believe in life after death and only 30% believe in heaven and hell. Around 20% said they did not believe in God or other supernatural beings such as jinns or genies.

The contrast with the results of the most recent 2020 World Values Survey for Iran are extreme. The latter found that 43% of Iranians are practicing Muslims (similar to the online survey estimate for total Muslims, 53% are non-practicing and only 1.5% say they do not believe in God. In the WVS, 91% say they believe in life after death, 92% believe in heaven and 88% believe in hell. The online survey found that over 60% said they did not perform the obligatory daily Muslim prayers. This is in the same ballpark as the 53% who were classified as non-practicing in the WVS.

Reading between the lines of the WVS, and taking into account the severe consequences of being apostate or atheist in Iran, it supports the conclusion of Maeki and Arab that Iran is becoming much more secular. Around 53% of respondents in their online survey reported coming from practicing religious families but losing or changing their religion in their lifetime. This increasing secularity is also supported by the evidence of dramatic declines in the fertility of Iranian women over recent decades, with population growth in 2020 dropping below 1%. Its quite likely that real levels of irreligion are higher in many other Islamic countries than the survey data suggest.

Culture zone trends in religiosity and atheism

In my previous post, I used a slightly modified classification of countries into ten culture zones based on those of Welzel [8]. These are defined in Endnote d. The following plots show time trends for the religiosity categories in each culture zone. These are population-weighted averages for the countries included in the WHS/EVS for each culture zone. Turkey is the only Islamic country with data before 2000 and its trends are not likely to be representative for other Islamic countries. So for the other Islamic countries I have simply projected the earliest data backwards at constant values for the purpose of computing global trends.

The prevalence of atheism is rising in the four culture zones of the West, fastest in North America in the last decade. But it has been rising in the Reformed West (which includes Australia and New Zealand) for more than four decades and its prevalence is now an estimated 49% for the whole zone. The Orthodox East (former Soviet zone countries with Orthodox Christian or Muslim majorities) shows the religious rebound after 1991 discussed earlier, though the majority of the new religious are non-practicing. Religiosity trends have been fairly flat in the Islamic East and Sub-Saharan Africa with apparently very low levels of irreligion (though that may reflect unwillingness to risk disclosing irreligon). There is an apparent switch between non-religious and atheist in the Sinic Zone, though as noted earlier this could be affected by the orientation of the WVS religious questions towards theist religions.

At global level, the prevalence of practicing religious has barely changed over the last 40 years, as has the prevalence of atheism, but there has been a shift from non-religious to non-practicing religious, reflecting mainly the change in former Soviet bloc countries.  Excluding China, there is a slight decline in the prevalence of atheism but overall, there has been relatively little change in prevalence of religiosity at global level over the last 40 years. This conceals quite substantial changes in developed countries and in former Soviet countries, in opposing directions. Note that the regional and global trends are based on population-weighted averages for the 110 countries included in the WVS/EVS dataset. Many of the countries not included are small except in Africa which is represented by only 10 countries, though these include large countries like Nigeria and South Africa.

Trends in average religiosity over the last 40 years

It is entirely possible that while the prevalences of religiosity categories have changed little, the average religiosity within categories has changed, for example through less frequent religious observance, or lesser importance placed on God in the respondent’s life (as used by Inglehart for his claim that religion is in global decline). To examine this, I have used the set of religiosity variables in the WVS/EVS to compute a continuous latent variable ρ for religiosity using item response analysis [9, 10] to estimate the latent variable from the categorical response variables measuring aspects of religiosity.  This was implemented as an ordered probit model using the stata procedure gsem for generalized structural equation models [11]. The model was fit to the entire dataset for all countries and all survey waves. The item response cut-points for the questions used as independent variables are shown in the following figure.

At individual level, 90% of values of the religiosity latent variable fall in the range -5.93 to 6.88 with a median value of 0.365.  At country level, average religiosity ranged from -4.6 in Nigeria in 2000 to 4.71 in China in 1990. Country averages for the most recent WVS/EVS wave for 2017-2020 ranged from -4.4 in Ethiopia to 4.34 in China, followed by 3.73 in Czechia and 3.49 in Sweden. Note that negative values indicate higher levels of religiosity and positive values indicate higher levels of irreligion.

Trends in average religiosity were imputed for the period 1980 to 2020 using the same methods as for the categorical religiosity prevalences (endnote c). Again, this is a population-weighted average of all 110 countries for which there are WVS/EVS data, not all of the around 194 countries in the world (most of the missing are very small countries such as the Pacific Islands).

The large increase in irreligion in North America stands out, as does the more steady increase in the Reformed West, and the decrease in irreligion following the collapse of the Soviet Union and other Eastern Bloc countries around 1991. However, the continuous latent variable also picks up an increase in religiosity in Sub-Saharan Africa and an decrease in religiosity post-2000 in Latin America, the Old West and the Returned West. At global level there has been a slight increase in religiosity over the forty year period. This is the opposite conclusion to that reached by Inglehart in his recent Foreign Affairs article.

The next question I will look at is whether religious values are changing in the direction of increasing or decreasing fundamentalism and the extent to which religion is associated with rejection of scientific evidence and findings: important questions for a world facing the twin crises of climate change and the coronavirus pandemic. To be continued.

Endnotes

a. Revised definitions of religiosity and atheism

The WVS/EVS questions, and those of other similar survey programs, are biased towards monotheistic religions and do not adequately take the non-theistic religions into account (these include Buddhism, Confucianism, Taoism, Jainism). Thus “confirmed atheist” is a separate category to “religious person” although around half of practicing Buddhists in the surveys said they do not believe in God. For respondents who report affiliation with a non-theistic religion (or who report religion as “None” in Sinic countries, Thailand and Myanmar(, the religiosity categories are modified as described below. A more accurate label for the “Atheist” category that takes non-theist religious people  into account would be “Non-religious atheist”. The classification for non-theists is as follows:

Atheist:  A confirmed atheist”, practicing once a year or less and considers religion to be “not at all important”, OR a religious or non-religious person who does not believe in God, is not practicing and considers religion to be “not at all important”

Non-religious: A not religious person, who believes in God but considers religion not at all important, and is not practicing.

Non-practicing religious person: A religious person who is non-practicing OR a non-religious person or confirmed atheist who practices less than once a month and more than once a year, and rates the importance of religion greater than “Not at all”

Practicing religious person: A religious person who is practicing at least once a month OR a not religious person or confirmed atheist who is practicing at least once a month and considers religion to be more than “Not at all important”

This is not entirely adequate to deal with the non-theist issue, as adherents of non-theist and folk religions may often interpret the question on religious affiliation as referring to the major established religions, and respond “None”.  This is only an important issue for Sinic countries and other countries in Asia where Buddhism is a major religion. The religiosity results for these countries should be treated with caution, they vary substantially across the countries and across waves and time, and its likely that there are considerable measurement issues resulting from variable interpretation of questions framed largely for theist religious beliefs.

b. Estimation of missing values for belief in God

Questions on belief in God, heaven and hell are missing for all 59 countries in Wave 5 of the WVS and from 25 other countries in various other waves of the WVS and EVS. An EVS survey is available for the same survey year for around half these countries, and the questions are available in adjacent waves for all the other countries with missing data.

Because belief in God is a crucial input to the classification of religiosity, an effort was made to impute it using data from the EVS surveys for the same year, when or available, or adjacent years on either side. A country-level probit model for belief in God (Yes, No) was fitted to data for the comparison years for the questions “How important in life is religion?”, “Are you a religious person?” and “How often do you attend religious services?”. The regression model was then used to impute belief in God at individual respondent level for those surveys where that question was missing. A similar imputation model was used for belief in hell.

c. Interpolation and projection of religiosity trends from 1980 to 2020

For countries with data for three or more waves of the WVS+EVS, trends in the prevalence of the four religiosity categories were estimated at country-level as follows:

  1. Values for single missing waves between two other waves were estimated using the geometric average for the two waves with data
  2. For countries in the Returned West or Orthodox East, the average change in prevalences due to the breakup of the Soviet Union was estimated using the average for those countries with data for the first wave in the 1980s and the second wave in the 1990s. This average jump was used to estimate values for the first or second wave for countries where one of these was missing.
  3. For countries where one or more early waves or one or more late waves were missing, prevalences were projected forwards or backwards using the median annual rate of change calculated at culture zone level for countries without missing waves.
  4. Population-weighted average trends were then calculated for each culture zone with the following two exceptions. For the Returned West and Orthodox East, prevalences were assumed constant prior to 1991. For Islamic East countries, flat trends were assumed from 1980 to earliest available data, as Turkey is the only Islamic country with data prior to 2000, and its trends are probably not representative.

For countries with data from only one or two waves, forward and back projections to 1981 and 2018 were made using the average trends calculated in step 4 above for each culture zone.

For projections from the Wave 7 survey year to 2020, a conservative approach was taken to avoid making extreme projections based on substantial differences between Wave 6 and Wave 7 estimates which may reflect variations in survey sampling or procedures rather than real change. For each country, the recent trend was estimated using data from survey waves in the period 2000-2020, with each earlier wave being given 15% less weight in the regression than the subsequent wave.  Average annual rates of change (aar) were also calculate from the Wave 6 and Wave 7 average prevalence estimates for each culture zone.

Where the regression aar were more extreme than the culture zone average aar, an average of the two was used for projection to 2020. Where the regression aar and the culture zone aar were of opposite signs, the regression aar was halved before use for projection. To be even more conservative, if the last Wave for a country was more than a year earlier than 2020, the rate of change from the survey year to 2020 was assumed to be aar1.5.

Finally, annual estimates for all years from 1980 to 2020 were prepared as follows:

  • Values prior to 1981 (the actual or projected first wave values) were assumed to be constant at the first wave values
  • Values beyond 2020 were also assumed to be constant at the 2020 value
  • Annual values between waves were estimated by linear interpolation
  • Annual values from 1980 to 2020 were smoothed using a 5-year moving average.

d. Definitions of culture zones used to group countries

I am using the 10 culture zones defined by Welzel [8], with one modification. Because Australia’s and New Zealand’s culture values are much closer to the countries of the Reformed West than to those of the USA and Canada, I have included Australia and New Zealand in the Reformed West and renamed the New West as North America. The culture zones are defined as follows:

Reformed West — Western European societies strongly affected by the Reformation: Denmark, Finland, France, Germany, Iceland, Netherlands, Norway, Sweden, Switzerland, United Kingdom, plus Australia and New Zealand;
North America — USA and Canada;
Old West — Mostly Catholic parts of Western Europe being core parts of the
Roman Empire: Austria, Belgium, Cyprus, Greece, Ireland, Israel, Italy, Luxembrg, Malta, Portugal, Spain;
Returned West — Catholic and Protestant parts of post-communist Europe returning
to the EU: Croatia, Czechia, Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia, Slovenia;
Orthodox East — Christian Orthodox or Islamic parts of the post-communist world,
mostly parts of former USSR;
Indic East — Parts of South and South East Asia under the historic influence
of Indian culture: Bhutan, Cambodia, India, Indonesia, Laos, Malaysia, Maldives, Myanmar, Nepal, Pakistan, Philippines, Singapore, Sri Lanka, Thailand, Timor-Leste;
Islamic East — Regions of the Islamic world that have been parts of the Arab/Caliphate,
Persian and Ottoman empires;
Sinic East — Parts of East Asia under the historic influence of Chinese culture: China, Hong Kong, Japan, Macau, Mongolia, North Korea, South Korea, Taiwan, Vietnam;
Latin America — Central and South America and the Caribbean;
Sub-Saharan Africa — African countries south of the Sahara.

References

  1. Inglehart R Giving up on God: the global decline of religion. Foreign Affairs 2020, 99(5): 110-118.
    https://www.foreignaffairs.com/articles/world/2020-08-11/religion-giving-god
  2. Inglehart, R., C. Haerpfer, A. Moreno, C. Welzel, K. Kizilova, J. Diez-Medrano, M. Lagos, P. Norris, E. Ponarin & B. Puranen et al. (eds.). 2014. World Values Survey: All Rounds – Country-Pooled Datafile Version: https://www.worldvaluessurvey.org/WVSDocumentationWVL.jsp. Madrid: JD Systems Institute.
  3. Haerpfer, C., Inglehart, R., Moreno,A., Welzel,C., Kizilova,K., Diez-MedranoJ., M. Lagos, P. Norris, E. Ponarin & B. Puranen et al. (eds.). 2020. World Values Survey: Round Seven–Country-Pooled Datafile. Madrid, Spain & Vienna, Austria: JD Systems Institute& WVSA Secretariat[Version: http://www.worldvaluessurvey.org/WVSDocumentationWV7.jsp].
  4. Gedeshi, Ilir, Zulehner, Paul M., Rotman, David, Titarenko, Larissa, Billiet, Jaak, Dobbelaere, Karel, Kerkhofs, Jan. (2020). European Values Study Longitudinal Data File 1981-2008 (EVS 1981-2008). GESIS Datenarchiv, Köln. ZA4804 Datenfile Version 3.1.0, https://doi.org/10.4232/1.13486.
  5. EVS (2020): European Values Study 2017: Integrated Dataset (EVS 2017). GESIS Data Archive, Cologne. ZA7500 Data file Version 3.0.0,doi:10.4232/1.13511
  6. Pew Research Center, Oct 29, 2018. Eastern and Western Europeans Differ on Importance of Religion, Views of Minorities, and Key Social Issues. https://www.pewforum.org/2018/10/29/eastern-and-western-europeans-differ-on-importance-of-religion-views-of-minorities-and-key-social-issues/
  7. Maeki A, Arab PT. Iranians’ attitudes toward religion: a 2020 survey report. The Group for Analyzing and Measuring Attitudes in Iran (GAMAAN). Published online, gamaan.org: GAMAAN. https://gamaan.org/wp-content/uploads/2020/09/GAMAAN-Iran-Religion-Survey-2020-English.pdf
  8. Welzel C. Freedom Rising. Human Empowerment and the. Quest for Emancipation. 2013. Cambridge: Cambridge University Press. https://www.cambridge.org/core/books/freedom-rising/80316A9C5264A8038B0AA597078BA7C6
  9. Hambleton RK, Swaminathan H. Item Response Theory: Principles and Applications. Springer Science and Business Media, New York. 1985.
  10. Nguyen TH, Han HR, Kim MT, Chan KS. An introduction to item response theory for patient-reported outcome measurement. Patient. 2014;7(1):23-35. doi:10.1007/s40271-013-0041-0
  11. Stata Corporation. Stata Structural Equation Modeling Reference Manual. Release 15. StataCorp, Texas, 2017.

 

Posted in Uncategorized | Tagged , , , , , , , , | 3 Comments