In a critique of Hegel, Marx wrote the following paragraph, whose third sentence has become a classic:
Religious distress is at the same time the expression of real distress and the protest against real distress. Religion is the sigh of the oppressed creature, the heart of a heartless world, just as it is the spirit of a spiritless situation. It is the opium of the people. The abolition of religion as the illusory happiness of the people is required for their real happiness. The demand to give up the illusion about its condition is the demand to give up a condition which needs illusions.
The extract “[Religion] is the opium of the people” is often seen as a put-down of the masses who use faith as a drug. But from the context you can see that Marx was not so callous, for he was writing about religion as a sad and inevitable response to the distress of life.
And this is what sociologists are beginning to tell us. I’ve previously written about work by Tomas Rees and Gregory Paul showing that high levels of faith among countries are correlated with high levels of “life uncertainties,” quantified in various ways such as income inequality, absence of national health care, high rates of crime, and so on. Unless religious belief itself leads to socially unhealthy societies rife with uncertainty, the studies suggest that, as Marx noted above, social distress promotes religiosity.
A recent paper in Cross-Cultural Research by Nigel Barber buttresses this suggestion by showing that, among 137 countries, there’s a strong negative correlation between measures of “material security” and religiosity: those countries that are most religious are also those in which individuals are less secure, with security measured in several ways. These correlations were predicted in advance by Barber. (He is a biopsychologist who appears to be an independent scholar: his given address is a private one in Birmingham, Alabama).
Barber’s hypothesis, called the “uncertainty hypothesis,” is that “supernatural belief may be one way of controlling the uncertainty of our lives.” His prediction is “if religion helps people to cope with uncertainty, then more secure modern environments having greater existential security would engender less religious belief.” His more explicit predictions were these:
I predicted that atheism would increase with economic development as people acquired a better capacity to withstand the hostile forces of nature through improved scientific knowledge, technological development, greater affluence, food security, and increased rule of law including the stronger centralized government characteristic of developed countries. Development was assessed in terms of the proportion of the labor force employed in agriculture (a negative index given that developed nations have fewer agricultural workers). It was also predicted that nonbelief would increase with the proportion of the population enrolled in third-level education both because this is an index of economic development and because it is a vector for natural science ideas that may challenge religious claims. As people acquired greater economic security, I predicted that disbelief in God would increase (Norris & Inglehart, 2004). Economic insecurity is exacerbated by unequal distribution of income (Gini coefficient) because more of the resources are concentrated in the hands of an economic elite creating poverty and deprivation at the bottom of the social hierarchy. Conversely, societies having a welfare state aim to help the poor by redistribution of resources. The welfare state requires heavier personal taxation and was measured indirectly in terms of taxation as a proportion of GDP.
Disbelief in God was also predicted to increase as health security rose. Health security was assessed in terms of the load of infectious diseases and pathogens.
Barber collected data on these issues from 137 countries. The index of religiosity was taken from Zuckerman (2007), as “the proportion of people reporting that they did not believe in God.” There is a possibility of error here since the question was not asked in the same way in every country.
The variable tested for their correlation with religiosity were these:
- Whether or not the nation was or is Communist (i.e., whether or not there were official strictures against religion)
- Whether or not the country was Islamic (apostasy and sometimes atheism are criminalized under Sharia law)
- Degree of economic development, quantified as proportion of labor force engaged in agriculture (the lower the proportion, the more developed the society)
- The proportion of young people enrolled in third-level education (i.e., university education)
- Economic security quantified as the Gini coefficient, a measure of income equality that varies between 0 (complete equality) and 1 (maximal inequality)
- Level of personal taxation, which is taken as the degree to which a nation is a “welfare state,” i.e., creates more security for its citizens
- Health security, measured as the prevalence of 22 pathogens
As Barber predicted, each of these variables showed a statistically significant relationship with religiosity in the expected ways: religiosity was higher in Muslim countries and lower in Communist ones, negatively associated with the proportion of agriculturalists, health security, and income inequality, and positively associated with third-level education and taxation. Below is the table of correlations of the varables with religious disbelief. The table also shows the correlations among the various indices of “security” (asterisks show significant correlations). A regression analysis of each variable on disbelief (not shown here, but in the paper) also revealed a significant association in the same direction for every variable.
Barber concludes that his hypothesis was supported: “Taken together, the results show that the incidence of religious disbelief in a country (Zuckerman, 2007) is very strongly predicted by economic development, by favorable health conditions and by a more equal distribution of income as well as a well-developed welfare state (insofar as this is measured by high levels of personal taxation relative to GDP).”
But there is a big problem with these results. As the table shows, the different indices of “security” are also correlated with each other. For example, there’s a strong negative correlation (-0.69) between the degree of agricultural labor and the percentage of people getting third-level educations. Likewise, pathogen load is negatively correlated with level of education and positively correlated with degree of agricultural labor. Income inequality is negatively correlated with taxation (a measure of “welfare stateness”). Pathogen load is negatively correlated with whether a country is/was Communist, but positively correlated with whether a state is Islamic.
These cross-correlations among the different indices of “security” mean that we cannot use each of them as an independent variable affecting religiosity. We don’t know, for example, whether the negative correlation between disbelief in God and income inequality reflects a direct influence of the latter on the former (countries with higher inequality have higher belief in God), or only that income inquality affects religiosity because that inequality is itself a sign of poor health (pathogen load has a 0.5 correlation with the Gini coefficient). What this means is that you cannot say that each of the seven variables is itself significantly associated with religiosity. They are not independent.
The author could have addressed this problem in two ways. First, he could have done what Greg Paul did, and simply combine the variables into a single index of societal well-being. Alternatively, and better, Barber could have used multiple regression, a method that gets rid of the cross-correlation between variables to look at the real effect of each one uncontaminated by its association with the others. Unless I’m missing something in the author’s analysis, he didn’t do either of these. Any decent journal would have mandates some statistical analysis to gauge the effect of each variable, by itself, on religiosity.
So what can we conclude? The results generally support the “uncertainty” hypothesis because each variable is correlated in the expected direction with religiosity. But what we cannot say is that each of the seven variables itself has an independent effect on belief and disbelief. That awaits a multiple-regression analysis—or other statistical tests that get rid of the problem of cross-correlation. The data for this are in fact already available to Barber.
And, of course, we all know that correlation is not causation. Even if each of these variables was independently and significantly associated with religiosity, we still don’t know in which direction (or neither) the causality runs. It’s formally possible, for instance, that more religious societies promote income inequality and poor health, but for most variables that suggestion seems less parsimonious.
Barber, N. A cross-national test of the uncertainty hypothesis of religious belief. Cross-Cultural Research. Published online before print May 11, 2011, doi: 10.1177/1069397111402465
Zuckerman, P. 2007. Atheism: Contemporary numbers and patterns. (Available free on Zuckerman’s website.) pp. 47-68 in The Cambridge Companion to Atheism (M. Martin, ed.), Cambridge University Press, Cambridge, UK.
37 thoughts on “Do life’s uncertainties promote religion? A flawed study”
There is a much better take on this by Solt and colleagues. They apply more sophisticated techniques to zero in on the influence of inequality on religiosity.
I do think the relation goes both ways. You can tell by the way the religous right opposes healthcare and social security. They refer to the social safety net as “replacing God with government”. They know it’s a threat to their influence, so they oppose it.
Right, it’s at least plausible that religiosity discourages government-provided social welfare, because it is seen as direct competitor to church-provided social welfare.
My gut feeling is that most of the correlation runs the way Jerry has it, i.e. uncertainty drives people to religion. But it’s not at all implausible that for at least some of these metrics, it partially runs the other way.
Any decent researcher would have recognized this issue and done the appropriate analyses. The crudeness of the analysis that was done doesn’t give me much confidence in the original research.
My instincts want to agree with the “uncertainty hypothesis”. I know when I was a believer, the benefit religion provided was a relief from anxieties that came from job pressures, paying bills, and macro-issues like war, politics etc . . . It also put me in the company of people who invited action in the face of these stresses which helped alleviate the perceived influence uncertainty had on my life.
Excellent study – thanks for posting
Here is the quick recount of classical evolutionary process:
inorganic matter –> life
life–> (evolution) -> man (deliberatively capable life-form)
Now the evolution of homo sapiens as evolution of his “deliberative capability”
“deliberative capability” === “knowledge accumulation over time”
“little or no knowledge” === “religion as the way to cope with uncertainty and most of phenomena being unexplainable”
“evolution of deliberative capability reaches certain level” = “knowledge accumulation reaches certain threshold” = ” SCIENCE is born into the world of INSTITUTIONALIZED religion”
meanwhile in the world of physical planetary system
science is used to greatly improve individual viability thru fighting diseases and making more food available LEADING to EXPONENTIAL POPULATION GROWTH ;
OVERPOPULATION —> RESOURCE/ENVIRONMENT CORRUPTION —> CIVILIZATION COLLAPSE –> MORE UNCERTAINTY IN INDIVIDUAL LIFE AND COMPLETE CERTAINTY FOR THE PLANET AS A WHOLE BECAUSE SCIENCE REACHED COMPLETE EXPLANATORY POWERS FOR HUMAN CONDITION by mid 20th century
THE FINAL OUTCOME SOME TIME IN THE FUTURE (you can quote me on this 🙂
“Over deep (evolutionary) time the civilization of homo sapiens will continue the cycle of overpopulation and collapse progressing from geographically localized (Roman, Mayan, …) to whole-planet scale. This is to be experienced within next one to five generation because the rate of population growth greatly exceeds the planet’s ability to assimilate anthropogenic impact. It is not possible for science to contain environmental impact because it simultaneously promotes overpopulation that produces it. Eventually science of human condition will begin enter the minds of scientists and they will realize that they should stop using science to promote overpopulation as means to stop continuous environment/resource corruption. The corresponding “morality” will be invented and institutionalized. Yielding to the pressure of genetic imperative to survive (property of life and matter) the scientists become “the rulers of the world”. The reason becomes the only shepherd of human condition. Faith is pushed to the fringes of society. Every child is conceived, born and educated in highly orderly and planned fashion to maintain achieved sustainability. Gradually the level of population is decreased to the level necessary to promote maximum of well-being for all members of the global population while maintaining reproductive health of the population as a whole at the highest possible level”
And that will be the time when homo sapiens becomes HOMO COGITAN.
More correctly, Homo cogitans
— And, maybe so, but it seems a bit of a dreary prospect. See the recent Feynman post.
not dreary at all
or, rather, the “emotional” dimension should be not part of our analysis
and since it is indeed not part of our analysis each particular individual is “free” to make anything out of it
it is “half-full-half-empty” situation
for me, for example, it is the source of the greatest optimism ever – i finally realized that there is an end to the stupidity around me even if it is 500 years from now and in completely impoverished planet with a tiny fraction of the current population
What on Earth makes you think we wouldn’t simply drive ourselves to extinction given your scenario above?
I mean, do you really think we can survive a global collapse of the oceanic ecosystems due to acidification? Because that’s what we’re facing if we can’t contain our environmental impacts — and, frankly, we’re not that far off.
For your scenario to hold, we’d have to survive an extinction event that would compete with the Permian – Triassic. That just ain’t gonna happen.
No, our only real hope is for non-petroleum energy sources to take off in a big way, right now. Or for petroleum to run out sooner than expected…but that still leaves coal and shale, which are more than enough to push us over the edge if we use them for anything other than bootstrapping ourselves into a post-petroleum global economy.
…which is why there’re a couple guys up on my roof right now doing the prep work for the photovoltaic array that’ll be installed by sometime Thursday….
my confidence stems from the heuristic nature of SCIENCE and the history of scientific method being able to cope with increasingly complex _unanticipated_ eventualities
we really have no clue what levels of anthropogenic impact the planet can withstand AND continue support life including homo species
all we can do at the moment is to draw attention to the _direction_ of evolution but we cannot “devine” the problems coming our way and xconsequently we would waste our time trying to chase those problems with specific “fixes”
your existance, understanding of my post, and willingness to discuss it without emotional dismissal is the evidence of emerging HOMO COGITANS of which you might be the one even if you yourself do not realize it
I think the subject should switch from religiosity to non-belief over the course of this sentence:
“religiosity was higher in Muslim countries and lower in Communist ones, negatively associated with the proportion of agriculturalists, health security, and income inequality, and positively associated with third-level education and taxation.”
I.e., all the items starting with “negatively associated” are true for non-belief, not religiosity.
Thanks for the interesting post.
Let’s see, Jim. I cut the sentence apart, and into questions, and considered each:
Religiosity was negatively associated with the proportion of agriculturalists?
Nope; you’re right; article was in error.
Religiosity was negatively associated with health security?
Yes; you’re wrong.
Religiosity was negatively associated with income inequality?
Nope; you’re right; article was in error.
Religiosity was positively associated with third-level education and taxation?
Nope, nope; you’re right; article was in error.
So your last sentence is incorrect for one of these 5, and correct for 4/5.
There is so much danger political ideology in studies of these types. If you’re looking for a particular correlation, you likely can find it.
One point is the factor of low levels of agricultural workers. The wealthier a society becomes, the smaller percentage of the national labor force is involved in food production. This is not a coincidence, requiring less labor to produce food allows productivity to move into other areas, hence the society as a whole is wealthier.
Most surprisingly, personal freedom does not seem to be a factor they looked at, nor personal life quality: home size, comfort level, possessions, food choice.
Nor did they differentiate between high income ratio societies: some are essentially ‘caste’ systems (which are hopeless), while others have a relatively high degree of economic and social mobility. These are two very different environments yet they appear to classify them as the same thing. The ‘poor’ of some countries are far wealthier (and healthier) than the middle class of other countries. (The income ratio seems to be somewhat of an obsession with some researches who feel people are more motivated by wealth jealousy than the amount of wealth they personally have.
Isn’t that what the Gini coefficient is supposed to account for?
ehh, not so much.
the Gini (as I understand it) is a relative value, i.e. a wealthy country where even the lower strata are better off could have the same or higher number than a poor country where no one is doing well.
Additionally the often unstated assumption is that income inequality is in itself something to be avoided. If there is a good overall level of wealth and relative economic mobility, it seems that jealousy is the primary objection.
In a healthy economy, the income disparity is not part of a zero sum game, i.e. the wealth of some people does NOT reduce the wealth of others. A vibrant economy grows wealth through increased money circulation and improved productivity. Bill Gates’ wealth, or the wealth of sports or music stars does not reduce my wealth or the wealth of anyone other than those who choose to spend their money on those performances.
In a healthy economy, such disparities don’t arise in the first place.
Consider: until somewhat recently, top executives made about ten times as much as the lowest-paid workers. Now, you’ll have some workers making less than $10 / hour = $400 / week = $20,000 / year, no benefits…with the CEO making tens of millions per year.
And that’s just not healthy. Hundreds of people could be paid good wages without that one man taking a hit in his standard of living. And those hundreds of people would create far more actual wealth and return far more of the money into the economy in a much shorter timeframe than the CEO ever possibly could.
It’s feudalism, frankly. Pure and simple.
This basically Marxian (surely influenced by Feuerbach) idea has been revived and restated many times in different ways.
A Finnish philosopher, Eino Kaila, used to say that “religions are the insurance policies of human mind against the fear of death”, and I think that Kaila, too, was on the right track here. Religious/supernatural representations are indeed attempts to gain some insurance against the terrors of uncertainty and, ultimately, against destruction and death. If only they worked!
But the insurance policies offered by religions are much worse than the Credit Default Swaps AIG was selling: they are, by definition, never backed by any real capital, their worth is entirely dependent on people having faith in them.
Surely a decent PCA (Principle Component Analysis) would be the best way to go about this? From what I remember from my stats courses anyway.
Correlation is not causation. However, in this particular case there are two lines of evidence that a causal relationship could exist:
First, longitudinal studies show decreasing religiosity/belief with increased security in the same nation, over the course of decades.
Second, in postindustrial nations there is a clear age gap concerning religiosity, with the younger people less religious than old people. In agrarian societies the young and old are equally religious.
In his book “the believing brain”, Michael Shermer describes a study which show people who are under stress at the time of an experiment are more likely to identify nonexistent patterns than those who are under leas stress. This may be one of the reason the sociological correlation is seen.
Very nice analysis. We were struck by this: “Pathogen load is negatively correlated with whether a country is/was Communist, but positively correlated with whether a state is Islamic.”
We are big believers in “Guns, Germs and Steel” especially the “germs” part since they are the hardest to see.
Oh yeah, forgot, we use the more robust and accurate term “magical beliefs” vs religion.
Also, uncertainty is a fairly straightforward brain state to locate and track so not much of a jump to go from social > neuro processes — the ultimate “cause.” With much individual differences, of course.
there is a big problem with these results. As the table shows, the different indices of “security” are also correlated with each other.
Yes, but they’re correlated in obviously consistent ways—the signs all match up. Because of this, the author if anything does not offer conclusions as strong as he may have done had he followed your good suggestion. Doing this takes a few minutes over lunch on Matlab or equivalent, and is immediately possible from the table you provide, so here it is (P = correlation matrix, Cov = covariance matrix, Proj = projection onto space with combined security factors):
P = [ 1 0 0 0 0 0 0 0; …
-0.50 1 0 0 0 0 0 0; …
0.45 -0.10 1 0 0 0 0 0; …
-0.60 0.33 0 1 0 0 0 0; …
0.71 -0.34 0.26 -0.69 1 0 0 0; …
-0.52 -0.04 -0.41 0.30 -0.48 1 0 0; …
0.71 -0.46 0.23 -0.54 0.66 -0.41 1 0; …
-0.70 0.23 -0.42 0.60 -0.72 0.50 -0.57 1];
P = tril(P,-1)+tril(P,-1).’+eye(size(P));
sigma = [1.49 0.46 0.40 1.32 24.05 9.29 12.16 6.76].’;
Cov = (sigma*sigma.’).*P;
security_idx = 4:8;
[V_security,D_security] = eig(Cov(security_idx,security_idx));
Proj = [eye(size(P,1), security_idx(1)-1) [zeros(security_idx(1)-1,1); -V_security(:,end)]];
Cov_reduced = Proj.’*Cov*Proj;
sigma_reduced = sqrt(diag(Cov_reduced));
P_reduced = Cov_reduced./(sigma_reduced*sigma_reduced.’);
1.0000 -0.5000 0.4500 0.7615
-0.5000 1.0000 -0.1000 -0.3591
0.4500 -0.1000 1.0000 0.2967
0.7615 -0.3591 0.2967 1.0000
1.4900 0.4600 0.4000 26.4941
-0.0348 0.8970 -0.1924 0.3455 -0.1942
The first output is the 4×4 matrix of correlation coefficients with all the security factors projected onto its principal component. As you can see, there is a very strong, significant positive correlation of +.7615 between disbelief and a combination of all the combined security factors. This is stronger than any single factor combined, as you’d expect just by looking at the signs of the correlation coefficients.
The second output 4-vector is the standard deviation of the projected factors, as above. The third output 5-vector is the principal eigenvector that shows the linear coefficients that go with each of the five security factors. Third-level enrollment (0.8970) and taxation (0.3455) are strongest, and the agricultural labor coefficient is very small (-0.0348), so this analysis may be redone assuming that agricultural labor is an independent variable. This is all parameterized, so resetting security_idx = 5:8 and rerunning the code above still yields a significant positive correlation of +0.7614 between disbelief and a combined security index, essentially unchanged from the result above:
1.0000 -0.5000 0.4500 -0.6000 0.7614
-0.5000 1.0000 -0.1000 0.3300 -0.3589
0.4500 -0.1000 1.0000 0 0.2970
-0.6000 0.3300 0 1.0000 -0.6983
0.7614 -0.3589 0.2970 -0.6983 1.0000
1.4900 0.4600 0.4000 1.3200 26.4781
0.8976 -0.1925 0.3457 -0.1943
Oh, and if you treat everything as correlated, you get:
-0.0062 0.0045 -0.0348 0.8970 -0.1924 0.3455 -0.1942
The small coefficients in this 7-vector show that Muslim (-0.0062), Communist ( 0.0045), and agricultural labor (-0.0348) factors really are very close to being independent from the remaining security factors, and should be treated separately from these, as done in the second case provided above.
Marx is one source of much of the confused analysis of religion we see in the new atheism.
I was going to comment on how Jerry runs such a classy joint here, he can even mention Marx without any nutters popping up in the comment thread. And you had to go and spoil that for me. >:(
Like, what does this even say? In English?
“The current campaign against religion is illogical and confused: religion is an effort, pegged to the understandings of the men and eras in which they arise, to assist man in the understanding of his own ‘software’. The results have often been poor to worse, but that does not invalidate the effort, as the mental state of the current society of scientism makes all too clear.”
Nothing was made ‘all too clear’ here, imho.
Aye, and that wasn’t the worst of it!
I’m from Russia and I can tell you, that when Soviet Union collapsed social security greatly decreased, and religiosity increased.
I think that this transition can be used to study dependencies between social security and religion. In Soviet Union income was not very high, but even among population and situation was very stable. In 1990s income dropped and strong inequality appeared. Now income seems to improve, but inequality and stability are still very bad.
But it can be problematic to get adequate historical measurements of faith level, because in Soviet Union statistics on this subject were lovered to show that 1) this idea of Marx works and 2) to show that we have very good social security, which was not ideal in fact. And now faith level is exaggerated, because our goverment choose to promote religion.
I’m curious about the apparent paradox involving these results and the problem of evil.
If the existence of uncertainty and the attendant suffering predicts belief in gods, and material security predicts disbelief, then is it consistent when we commonly say that the existence of suffering argues for disbelief?
Of course there is a ready solution in that humans don’t always behave rationally. Also a belief in a good God might be immediately reassuring while social change takes generations.
However the real paradox is that while suffering might raise doubts about a good God, it also is the main motivating factor in hoping for such a God to somehow redeem the pain we experience. This is psychologically unsurprising. When we are starving and people are dying around us we begin to suspect that there is no food available for us, but we also hope that it will suddenly be delivered.
There is not anything immoral in such a hope–it may even give some people strength beyond their natural limits. The problem lies in all the traps that involve focusing on a hope for supernatural rescue rather than compassion. The right thing to do in the above analogy is to plant the next crop even as we die of starvation, realizing that the hope of supernatural rescue might be contingent on a hope for a better future for everyone. In other words, if you want to believe in a rescue from a loving God, you had better assume that he will only rescue those who love each other–otherwise your belief may become an evil thing itself.
“It’s formally possible, for instance, that more religious societies promote income inequality and poor health, but for most variables that suggestion seems less parsimonious.”
Why? I find it quite plausible on multiple more or less independent grounds.
I think that religion tends to obscure issues of, e.g., social justice by casting them in terms of individual morality and God rather than effective social policy.
For example, religious people tend to think we need more religion to solve social problems—e.g., that if people are good Christians, they’ll contribute a lot more to charity, out of personal virtue. Problem is, it mostly doesn’t work—being more religious doesn’t make people all that much more charitable, and it it makes them prone to putting money where it doesn’t do as much good, e.g. promoting religion itself.
People in mostly irreligious countries don’t think that. They look for policies that will work even if individuals aren’t particularly charitable. (Or religious.) That allows them to solve commons problems—e.g., voting for taxes on their own class. (E.g., I’ll vote to increase taxes on my tax bracket by $1000 long before I’ll part with $1000 as an independent charitable act. I don’t want to spend my own money on good works nearly as much as I want people like me to all spend our money on lots of good works. I want lots of “matching funds” before I donate.)
To me, most indicators of social well-being seem clearly very plausibly related to religion messing things up: e.g., religious people are more likely to want more children than they can afford, not plan ahead for sex and have more unwanted children and STD’s, oppress people for basically random reasons, value immersing themselves in religion over doing things like getting a real education, fail to plan for the future because Jesus is coming back any time now, etc., etc., etc.
It doesn’t seem to me that there’s a lack of very plausible mechanisms by which religion could be a major contributor to various aspects of social insecurity—including many that are known to happen, to some extent.
(Wouldn’t you expect, all other things being equal, that basic erroneous beliefs about the nature of reality and the nature of morality would diverse negative consequences, large and small? Is that really unparsimonious?)
Why is it “more parsimonious” to guess that those causal factors are not very important, and the causality generally runs the other way?
(My own guess is that both causal directions are quite important—there’s a vicious cycle of social insecurity and religiosity.)
Indeed, see the Greenlanders who, when the climate was getting colder and they were dying of starvation spent all their time building more churches rather than trying to adapt to their circumstances. Without religion they may have lived, with it they wasted their time on unproductive activity and died.
“It’s formally possible, for instance, that more religious societies promote income inequality and poor health, but for most variables that suggestion seems less parsimonious.”
Thank you, Paul W., I was going to comment on it myself.
Very simply, that religion in many (or all) cases encourages various types of fatalism, obedience to authority, and acceptance of negative social situations to such a degree that I cannot imagine it would not be causative (and certainly in my instances in history it has been).