About two weeks ago I wrote about the negative correlation among countries between religiosity and happiness: the happiest countries in the world are the least religious, and the unhappiest the most religious. I needn’t discuss this further now, but wanted to put up a new plot made by reader “Gluon Spring” to demonstrate this relationship. Taking data from the UN’s 2013 World Happiness Report and the 2013 Pew Survey of Religious Importance, Gluon made this plot, this time naming the countries as well as giving the 95% confidence interval for the regression line.
When I posted this here and elsewhere, some people argued that a correlation of -0.52 wasn’t impressive. They’re wrong. With the 52 countries plotted here, the probability that this correlation would arise by chance is less than 0.0001. In other words, it’s highly significant. Note as well the narrow confidence interval for the regression line.
We can debate the meaning of this relationship in the comments below, but I wanted to show a plot that other people can use. At the very least it demonstrates that the most religious countries don’t contain the happiest people. Click to embiggen:
Quotemine!
Oy, I’ll fix!
Mexico seems to be in an odd position, as it is #16 in the 2013 World Happiness Report and 14 in the 2015 report but as a country, they are well recognized as very religious. Having been raised in Mexico, I can anecdotally say that people are either deluded and think they are happy or deluded for thinking they are more religious than what they think they really are. I think it is more the latter, especially in large urban areas where many young people tend to only care superficially about their faith, and mostly when it is criticized.
There are going to be outliers given the vagaries of other conditions among these countries. You’ve pointed out one reasonable possibility explaining an outlier.
Often when I put up a correlation, people single out outliers as a reason why the correlation doesn’t really hold. (Not that you’re doing that!) That’s like pointing out that old folks who have smoked for many years show that there’s no correlation between smoking and cancer.
Mexico makes no sense. It is very religious; more than America, in my view.
I think people today may act more secular in Mexico, but if push comes to shove, they would double down on Jesus Christo(with a silent J).
I wonder where Cuba would fall on this, on both scales. (How religious is Cuba these days, anyway?)
I think you mean that the people in the less religious countries are happier than those in the most religious countries.
It was Marx – and the Marquis de Sade earlier – who famously referred to religion as the opiate of the people – an interpretation based on a smaller sample of observations than this, but one that explored the details.
Does religion make us sad or does sadness make us religious? Or both? Marx clearly thought that sadness – mostly due to economic misery – explained the flight to religion. But it could be said that living in a religious environment could produce the affect of sadness – a withdrawal from the world, a lack of engagement, a lack of responsibility for the conditions of our existence. The counterargument to this would be the countless religious charities, the goofy enthusiasm of the Elmer Gantrys of the world, the tremendous artistic richness of churches, temples, grottoes, cave and rock painting. A two dimensional representation of this, while significant, clearly does not cover the ground in detail.
The correlation between wealth and religiosity is striking: Wealth vs Religiosity The US, in this plot, is an even more startling outlier.
The correlation is very striking indeed!
Although that data doesn’t include a few other countries that I think would be outliers, namely Saudi Arabia, Kuwait, etc.
Well, Kuwait is in the happiness/religiosity graph above, and you can see that it is an outlier. It seems likely that the reason for this is because of wealth.
“Some people pointed out that a correlation of -0.52 wasn’t impressive. They’re wrong. With the 52 countries plotted here, the probability that this correlation would arise by chance is less than 0.0001. In other words, it’s highly significant. Note as well the narrow confidence interval for the regression line.”
There are two different questions we might want to answer, and the p-value only addresses one of them.
The first question is “HOW IMPORTANT is the relationship?” or, put another way, what is the magnitude of the correlation? The people who say that -0.52 is not impressive are thinking about that question, and they are right that the magnitude of this correlation is quite small.
The second question is “How unlikely is this correlation to arise from random data?” That’s what the p-value answers, and the people who say that this is a very significant p-value are right that this correlation is unlikely to arise in random data. But this tells us nothing about whether the correlation is big or small, important or unimportant. With enough data points, a correlation close to zero (i.e. almost undetectable in the real world) can have a p-value of 0.000001.
The confidence interval, combined with the known value of the magnitude of the correlation, helps answer both questions at once. We know, with high precision, that magnitude is low, but not zero.
What do we mean by low magnitude here?
The correlation predicts ~ 1/3 of the data span.
Added in posting: Oh, I see that discussion is now further down, and the question responded to.
Just as a side note, Torbjorn, even when the x data predicts a large fraction of the y data span, this doesn’t mean the effect is large. If the data span is small or if the predicted metric is nonlinear. I’m not arguing about this case now but just making a general point. A slope would be a better indicator of the strength of an effect.
“If the data span is small”
-> “If the data span is small relative to the mean value of the y measure”
I don’t think that -0.5 is small at all, by the standards of sociological correlations it is large.
The minus sign by itself is pretty significant, given the widespread claims by the religious about the benefits of being religious.
Yes, the negative sign is both important and statistically significant.
About the size of the effect, sure, it may be that relatively small changes in happiness make a big difference in people’s quality of life.
My main point should just have been that the p-value could be highly significant even when a correlation is very close to zero in magnitude.
On the “smallness”, the effect seems quite large to me simply because of the large number of things that could contribute to “happiness”, and the large number of historical and cultural contingencies that could affect religiosity.
Sure, agreed.
By the standards of sociological correlations, yes.
Having just been writing a petrophysics report, I declined to continue analysis on a trend with an R.sq of 0.7 because it wasn’t worth the effort. (I’d continue the analysis, but it’ll be another day’s work on the client’s ticket. Their call.)
Which is the cause and which the effect. Best guess is that people are religious because they are miserable.
Along these lines. Within the US there are a bunch of things that trend together when one looks at state or county data, generally religiosity trends with poverty, poor education, high teenage pregnancy, drug use and a pile of other negatives. I’m not convinced that it causes all of those problems (although it may well contribute), but to me it generally seems more likely an escape mechanism from them.
I was very surprised by Brazil’s high happiness value, but then noticed that this is from 2013, before the 7-1. I bet its lower now.
I am a Brazilian German living in Canada. You are right about that – just don’t ask who I cheered for lol. The 7-1 was the start of the Brazilian happiness downfall; Brazilians are not so happy anymore when you add the recent Petrobras scandals, health crisis (Zika), unemployment and increase in violence.
There may be cultural factors at play. Having been to Brazil many times, I find the people there (by and large) inordinately cheerful. A Brazilian friend of mine said to me on my first visit, “Even when Brazilians are poor, they’re happy.”
I’m surprised by Germany’s high score on the happiness index, though. I love visiting the country but just about every German I speak to seems disgruntled about something 😉
Ah, but maybe being disgruntled makes them happy. 🙂 A long time ago when I visited Germany and Austria, I was told (by Germans) that politeness was not the same as consideration or niceness, and if I wanted to know the difference, I would find it in Austria. To be fair, I found both countries equally and highly enjoyable.
r = –0.52 may be statistically significant as you state, but r-squared = 0.25, which means that the linear trend explains only about 25% of the variation. In fact, the data are split into two clusters and thre is no apparent relationship between religiosity and happiness within each cluster.
So the linear trend explains only a weak relationship between religiosity and happiness.
It is your OPINION that explaining 25% of the variance is insignificant. I don’t share it. As for the two clusters, they might just represent a bimodal distribution of religiosity among countries samples (each cluster isn’t from one geograpahic area). Remember that there are many, many factors that the UN puts into its index of happiness, and it would be remarkable if religion were related to all of them so tightly that it would explain, say, 70% of the variance.
As with any correlation the effect could go both ways. Being religious makes you unhappy or being unhappy drives people to find some hope in religion. I would like to see an analysis taking into account other factors that could affect happiness.
Wealth is the biggest factor.
Wealth and social stability, I should say. That’s been covered in a number of articles here.
I’m thinking I should move back to Canada.
These happiness numbers are before the price of oil and the C$ collapsed.
And likely during the hockey and curling seasons.
In that case my retirement savings should go quite a long way.
BTW, I’m a U.S. citizen, but since I was born in Canada, I can obtain Canadian citizenship by sending in the application with $200! That might destroy any chance I have at becoming president. Ask Ted Cruise.
The scatter of points is an interesting shape: it’s hollow. Raul Martinez above describes this as ‘two clusters’ but if that were so, one would expect to see a few countries falling between them close to the average line. Instead there are countries scattered across the top and bottom edges, but a gaping hole in the middle.
I want to see this animated, with data spanning a century or so, to see which way stuff spirals.
Yes, this is obvious if the five countries in the lower left are ignored. (And I think there is a case for ignoring them.) There is a developed cluster in the upper left and an undeveloped cluster in the lower right (with a few anomalies.)
I did not know my country was that happy.
Ya. Get in on the fun while it lasts!
I will. I have fencing training tonight, which is always fun. And if Trump wins Iowa I might literally die of laughter.
What’s up with Israel being so high on the happiness scale?
I’d be curious to see what you get when the happiness data are broken down across the religious spectrum of Israelis, from secular to orthodox..
If the negative correlation remains, yet the standard of living is similar across the spectrum, this would support the theory that religiosity itself reduces reported happiness.
Has such an analysis been done?
That one jumped out at me too. Some 20% of the people in Israel are muslim too. Bet their not too happy (though in truth probably relieved not to be in any of the other local country choices). I’ve never been there, but would think the constant conflict would get people down. And Mexico, less relgious and happier that the US? That’s another head scratcher.
I would suggest that rather than posing questions to me others about the study, people go and look at what’s incorporated into the “happiness” index, which is far more than just asking people whether they’re happy. The link to the full report is in my earlier post.
I’m surprised not to see the UK, or any of its constituent parts, on there. Western Europe is otherwise well-represented. Was there a reason they left out the UK? I’d be interested to know where we sit.
The UK is missing because there was no data on it’s religiosity. I have no idea why, you’d have to dig into the original sources.
The UK is listed as 6.86 on the happiness scale. Western Ireland is 6.94.
So comparable to the United States. I think we can be pretty sure that if we had data for the UK it would be to the left of the US on the religiosity scale, adding another point (or several if broken down) to the rich, unreligious, happy corner of the graph.
as opposed to Eastern Ireland, I suppose…
I have no idea why they reported it that way.
So many of our citizens are only truly happy when they are grumbling it was just too damned complicated
I didn’t know Canadians were that happy. And why are the Germans so less happy?
I had the report open, so I can immediately see from figure 2.3 that the largest variations are residuals, i.e. unexplained.
But some small part of Germany’s difference with Denmark on top is relatively lower happiness due to correlation with corruption and generosity.
h, but since there are so much in residuals, religion may very well be a factor here! Someone needs to tease out causality between religion and happiness…
h, but since there are so much in residuals, religion may very well be a factor here! Someone needs to tease out causality between religion and happiness…
Go for it! 😉
Yeah there seemed to be just small differences in the bar graph.
Superficially, it looks like money buys happiness.
I think that is right (modulo some really awful thing like a king holding all the wealth, etc.)
I guess I wont emigrate to or vacation in Tanzania or Senegal.
Under “related” above we find an article “poverty and religiosity” which is finding a strong relationship, with higher poverty related to higher religiosity. Just eye-balling the scatterplot, material well-being is likely to be a powerful covariate. It would be interesting to see what corrected scores would yield with this factor controlled for. Surely someone has done this. I’ll go search the web. So obviously culture (as categorized by nationstate) is what we are really talking about, as this regression equation is giving the same weight to Kuwait as to China, and both are outliers at the opposite end of both variables. So none of this would really support the idea that religiosity does or does not contribute to happiness for an individual human being on planet Earth. (But then group statistics don’t do that anyway).
Seems to me if we are looking for causes, we have to consider the possibility that something else is responsible for this correlation. For example, per capita income? Political climate? etc
I take back political climate, though, looking at the chart again.
Yes. This is just a plot, not a study. There’d be probably a month of extra work plus a lot of data we don’t have to make it into a publishable study.
Some of the obvious variables are: per capita wealth, social well being (level of inequality, security of various sorts), and history (some of unhappy countries on the left are former communist or communist sphere countries, for example). The type of government is clearly a factor. Even the brand of religion is probably a factor (Christianity vs Islam vs Hinduism, etc.) So to address the question of why these things go together we’d have to analyze many more variables in a more sophisticated way.
All you can probably confidently say from this plot is that religion does not provide a large boost to happiness that overwhelms other factors, and that maybe it even hurts. That is, there is no reason to think that being rich and secure is hollow without religion to give you happiness.
Thanks for taking the time to comment!
Maybe I missed out some comments, but further controlling for economy would probably further strengthen the relation. Another strange thing is the ex-soviet states at the lower left.
However, I think the relations should be presented uncontrolled, since this is the observed data.
Does any one know if GapMinder has any thing close to this?
Cheers.
Since happiness is so strongly correlated with material comfort in a society, could it be that “happiness” is just acting as a surrogate here for “material comfort”? Material comfort could cause both happiness and non-religiosity, creating a correlation between the latter two even though neither necessarily causes the other.
Is it religion that makes people unhappy, or does unhappiness — I.e. poverty, lack of education, lack of opportunity — make people turn to religion? I tend to think it’s the second thing.