Religion's conditional dividend

A devout farmer in Mali and a devout retiree in Portugal answer the same Gallup question — is religion an important part of your daily life? — and both say yes. But the company their faith keeps could hardly be more different. In Mali, the people who say yes rate their lives almost half a ladder rung above their secular neighbours. In Portugal, they sit lower. The religious dividend is real in one place and inverted in the other, and the thing that separates them is not theology. It is the difficulty of the life around them.

This is the pattern the psychologist Ed Diener and colleagues named the religion paradox in 2011: religiosity tracks well-being far more strongly in hard, insecure societies than in comfortable ones. Pool eleven recent years of the Gallup World Poll, measure the gap between the religious and the non-religious inside each of 143 countries, and then ask whether that gap depends on national conditions. It does — but only loosely, and not in the same way for every kind of happiness.

1The bend is the argument

Each dot below is one country. Its height is the within-country gap in life evaluation between people who say religion is important and people who say it is not — weighted, as Gallup requires, by the within-country sampling weight. The horizontal axis is a national condition: human development, income, or the homicide rate. The white line across the middle is the only line that matters in the end. It marks no dividend — the height at which faith and well-being have nothing to say to each other.

The fitted curve — a locally weighted smoother, with a shaded band for its uncertainty — starts above the line and slides down to meet it. For life evaluation, the gap among countries averages near zero, but that average hides the slope: it runs at about −0.32 ladder points for every full point of the Human Development Index, and reaches zero near HDI 0.67. Below that, the dividend is positive; above it, it turns slightly negative. A simple regression with each country as one observation puts the fit's R² at only 0.05 — the bend is real and statistically clear (its slope is several times its standard error), but national development explains just a sliver of why some countries show a large religious gap and others none. Weighting countries by sample size steepens the slope a little, to −0.37, so the direction is not an artifact of small, noisy surveys.

Among the 46 lower-development countries, most show a positive religious gap; among the 51 high-development ones, most show none or a negative one.

2The two extremes

The countries at the top of the ladder ranking are mostly poor or insecure — places where, for the non-religious, the floor is closer and the supports thinner. The countries where the religious sit lower than their secular neighbours are a different crowd: aging, secularized, often European, where saying religion still matters can mark someone as out of step with the world around them. Hover any point in the chart to read both gaps; the standouts are below.

None of this means religion makes the difference. Religiosity is bound up with hardship: hard places are more religious to begin with, and they are also less happy for reasons that have nothing to do with faith — war, poverty, weak institutions. A country like Sudan or Mali is religious and difficult at once, so a gap measured there is partly a story about who, within a struggling society, leans on belief. Read the chart as a description of where the two move together, not a claim that one moves the other.

3Evaluation versus the ordinary day

Keeping the two kinds of happiness apart turns out to matter, because they do not bend in unison. The cleanest signal is not the headline ladder at all — it is positive affect, whether people laughed, smiled, felt enjoyment and rest the day before. Switch the measure to Positive days in the chart and the cloud tightens dramatically around the curve. National development explains roughly 46% of the cross-country variation in the positive-affect gap, against just five for the ladder, and the gap stays positive almost all the way up the development scale, fading out only near the very top, around HDI 0.90.

Negative affect — worry, sadness, stress, pain, anger — tells a parallel story with the sign flipped. Here a dividend means the religious report fewer bad-feeling days, so the advantage shows up below the white line. In lower-development countries the religious do report less daily negativity; that edge shrinks as development rises and is gone, on average, by about HDI 0.62 — after which, in the richest places, the religious tend to report slightly more negative affect, not less. Development moves the negative-affect gap with an R² of about 0.29: stronger than evaluation, weaker than the positive side.

So the three measures rhyme without repeating. Where they agree — the religious gap is largest in hard places and vanishes in easy ones — the agreement is the whole point. Where they differ in strength — faith tracks the felt texture of a good day far more tightly than it tracks a person's summary verdict on their life — the difference is itself a finding. Across countries the ladder gap and the positive-affect gap move together only moderately (a correlation of about 0.49), which is another way of saying these really are two different questions.

4What this shows, and what it doesn't

The honest summary is narrow. One: this is association, moderated by national conditions — never evidence that faith causes happiness. Religiosity is endogenous to hardship, and any number of unmeasured things that make a country hard also shape who is religious and how they answer survey questions.

Two: cross-country comparisons of self-report are slippery. People in different cultures use the same 0-to-10 ladder differently, and an emotion word translated into dozens of languages does not mean exactly the same thing everywhere. A gap of a tenth of a ladder point between two countries is not something to build a theory on; the brief here is the broad slope across 143 of them, not any single country's number.

Three: the affect items are blunt. They ask only whether a feeling happened yesterday — a yes/no, not an intensity — so they capture the presence of a good or bad day, not how good or bad. And the national context here is a static snapshot, joined to every survey year; it describes where a country sat around 2015–2018, not how it changed.

What survives all that is modest and, to Diener's point, fairly stubborn: the comfort that belief appears to offer is not evenly distributed. It is largest exactly where ordinary life offers the least, and it thins out as the rest of life fills in. Weighted by population, the world's religious ladder gap is almost exactly zero — not because faith does nothing, but because its dividend is conditional, and on a planet of mostly middle and rising development, the conditions increasingly cancel out.