The Secularisation Thesis and Its Failure

For most of the twentieth century, sociology held a prediction with unusual confidence: that modernisation would produce the decline of religion, everywhere, as a matter of course.

It was not a marginal position. It was held by nearly everyone in the field, it descended directly from the founders (see 4.2.6, 4.3.10), and it was taught as settled.

It failed — and how it failed, which parts of it survived, and what the discipline learned from getting a central prediction wrong is one of the most instructive episodes in the history of the subject.

A prediction, and its author's retraction.

In 1968 one of the leading sociologists of religion told a newspaper that by the twenty-first century, religious believers were likely to be found only in small sects, huddled together to resist a worldwide secular culture.

It was not an eccentric forecast. It was the field's consensus, stated crisply by someone who had done as much as anyone to formalise it.

Thirty-one years later, the same scholar edited a book whose opening argument was that he had been wrong.

His formulation was blunt. The assumption that we live in a secularised world, he wrote, is false. The world today is as furiously religious as it ever was, and in some places more so.

He did not retreat to a technicality. He set out what the theory had got wrong: that it had generalised from Western Europe and from a Western-educated intelligentsia, treated both as the leading edge of a universal process, and mistaken a regional pattern for a law.

Read this first as a piece of intellectual conduct, because the course has praised it before (see 7.8.2). A scholar publicly stating that a central claim he had helped establish was mistaken, and specifying why, is what a field looks like when it is working.

Now read the evidence he was responding to.

Religious adherence globally did not decline. The absolute number of religious adherents rose enormously, and — because religious populations have higher fertility than unaffiliated ones — the unaffiliated share of the world population is projected to fall over coming decades rather than rise.

Pentecostal and charismatic Christianity grew from a small movement at the start of the twentieth century into one of the largest religious formations in the world, overwhelmingly in Latin America, Africa and Asia.

Religious practice revived in post-communist societies that had spent decades under states committed to eliminating it.

And religion re-entered public life almost everywhere, in forms nobody had predicted: revolution in Iran, a mass movement in Poland, liberation theology in Latin America, a religious political mobilisation in the United States, and religiously framed nationalisms across South and South-East Asia.

None of that is compatible with the prediction.

But the thesis was not one claim. It was three, fused — and they have different fates.

This is the analytical move that rescued the field, and it is worth learning as a general procedure: when a theory fails, ask whether it was several claims travelling together.

Claim one — differentiation. As societies modernise, institutional spheres separate. Law, education, medicine, science, the economy and the state develop their own authorities, their own criteria and their own personnel, and religion loses its jurisdiction over them.

Claim two — decline. Individual religious belief and practice fall as societies modernise.

Claim three — privatisation. Religion retreats from public life into the private sphere, becoming a matter of personal conviction with no legitimate public role.

Their fates are completely different.

Differentiation is largely true and is the defensible core. Courts do not consult scripture, medical decisions are not made by clergy, universities are not run by churches, and this is true even in highly religious societies. The differentiation happened, it is largely irreversible, and it is what the founders were actually describing.

Decline is regionally true and globally false. It describes Western Europe well and almost nowhere else.

And privatisation is straightforwardly false. Religion did not retreat from public contestation; it returned to it, in the very decades the thesis predicted its withdrawal. The observation that religions had "deprivatised" — refusing the marginal, private role assigned to them and re-entering public argument about law, morality and national identity — was made in the 1990s and has been confirmed repeatedly since.

So: one claim right, one regional, one wrong — and the failure was in bundling them, which meant that evidence against the third was read as evidence against all, and evidence for the first was cited in support of the second.

Which societies, and the measurement problem underneath

Western Europe is the case the theory fits, and the argument is about whether it is the leading edge or the exception.

One position holds that Europe is the exception, and that the theory generalised from a region with a specific history — established churches, the association of religion with the old regime, and a long process by which religious authority was contested politically.

The other holds that Europe is simply ahead, and that other societies will follow as they become richer, more educated and more secure — with the developing world's religiosity reflecting where those societies currently are rather than a permanent difference.

Both positions are held by serious people and both make predictions, which is the useful thing about the disagreement. The exception argument predicts continued divergence; the leading-edge argument predicts convergence on the European pattern.

The evidence so far is mixed, which is the honest report. Some rapidly developing societies have secularised substantially; others have become more religious as they grew richer; and the most-watched case has recently moved in a direction that complicates both accounts.

And before any of that can be assessed, the measurement problem has to be faced, because it is severe.

Believing, belonging and behaving are three different things, and they come apart.

Religiosity is measured in at least three ways, and any claim about secularisation depends on which (see 7.2.2).

Belief — do you believe in God, an afterlife, a soul? Belonging — do you identify with a religion, are you a member? Behaving — do you attend, pray, observe?

They diverge systematically, and in different directions in different places.

Believing without belonging describes a pattern in parts of Western Europe: institutional membership and attendance collapse while belief in something — God, a higher power, an afterlife — declines much more slowly. The measured decline is largely a decline in institutional attachment.

Belonging without believing describes the Nordic pattern: very high formal church membership, very high rates of baptism, marriage and funeral use, and very low rates of belief and weekly attendance. The institution is retained as a public utility for life-course transitions and as a marker of national identity, without the doctrine.

And there is a third pattern worth naming: vicarious religion — a majority who do not practise, supported by and content with a minority who do it on their behalf, and who become visible and valued at moments of collective crisis or national mourning.

And one measurement finding deserves its own paragraph, because it changes several literatures at once.

Self-reported church attendance substantially exceeds actual attendance. When researchers counted attendance directly — physically, in a defined area, and compared it with what the same population reported in surveys — they found the reported rate roughly double the counted rate.

This is social desirability bias (see 7.4.1), operating on one of the most-cited comparative statistics in the field. And it matters asymmetrically: it inflates figures most where attending is socially expected, which is precisely the societies used as counterexamples to secularisation.

The two competing explanations

One — the supply-side account: religious economies.

The argument inverts the usual causality. Religious participation, on this view, is not driven by demand — the human appetite for religion is assumed roughly constant — but by supply.

Where a state church holds a monopoly, its clergy have secure incomes regardless of effort, its product is unresponsive, and participation is low. Where many religious organisations compete for adherents, they innovate, evangelise, tailor their offerings and work hard, and participation is high.

The comparison offered is Europe against the United States: established churches and low attendance against a deregulated, competitive, entrepreneurial religious market and high attendance. Europe's problem, on this account, is not modernity but regulation.

And the criticisms are serious enough that the account is no longer the field's leading candidate.

The core empirical claim — that religious pluralism raises participation — is not robust. Systematic review of the many studies testing it found that the relationship depends heavily on how pluralism is measured, on which cases are included, and on a statistical artefact in the standard measure; once these are addressed, the association is as often negative as positive.

The assumption of constant demand is an assumption, and it is exactly what the rival theory denies.

And the historical premise is doubtful. The account requires a prior era of high religious participation that state monopoly suppressed. Evidence on actual medieval and early modern religious practice — as opposed to formal adherence — is contested, and several historians argue that the "age of faith" involved a great deal of nominal, coerced and irregular participation. A theory that requires a golden age has the same problem as 9.2.1's and 9.3.2's.

Two — existential security.

The argument: religiosity is highest where life is most precarious. Where people face hunger, disease, violence, sudden loss and no institutional protection, religion supplies meaning, consolation and mutual aid — and where existential security is high, that demand falls.

It predicts a great deal correctly. Rich, secure, well-provisioned societies with strong welfare states and low inequality are the most secular. Poor and insecure societies are the most religious. And within countries, the poorest and least protected are usually the most religious.

It also explains the standing anomaly. The United States is a rich country with unusually high inequality, weak collective protection against illness, unemployment and old age, and much greater individual exposure to catastrophic risk than comparable countries — which places it, on the security measure rather than the income measure, exactly where its religiosity would predict.

Its difficulties. It handles the post-communist revival poorly, where religiosity rose as insecurity did but also where it rose amid recovery. It struggles with cases where religiosity has risen alongside growing prosperity. And it does not explain the timing of the change now occurring in its own key anomaly.

Because the counterexample has started to move, and for reasons neither theory predicted.

The share of Americans reporting no religious affiliation has risen sharply — from something like one in twenty in the early 1970s and one in fourteen around 1990, to something approaching three in ten in recent years. The change is large, rapid, and concentrated in younger cohorts.

It is a real secularisation event in the country that was the standing counterexample.

And the best-supported explanation is political rather than economic.

Research tracking the timing and composition of disaffiliation found that the rise began in the 1990s, was concentrated among political moderates and liberals, and followed the increasingly visible alignment of organised religion with a particular partisan and cultural politics. People who were moderately attached, and who came to associate religious identification with a set of political positions they did not hold, stopped identifying.

Two implications, both important.

Secularisation can be caused by politics rather than by modernisation — which no version of the classical thesis anticipated.

And it operates first on the loosely attached. The decline is in identification and affiliation; the highly committed minority has declined much less, which means the religious population becomes smaller and more distinct rather than simply smaller. That has consequences for polarisation that 9.4.2 examines.

Because this is the clearest case in the discipline of a central prediction failing, and the reason it failed is the reason Part 6 exists.

The theory was built from Western European history, in which religious authority had been contested politically for centuries, and generalised to humanity. It read one region's trajectory as the trajectory of modernity — and its most distinguished proponent said exactly that when he retracted it (see 6.8.1).

What survives, and it is not nothing.

Differentiation is real and largely irreversible. Modern societies do run their law, medicine, science and administration on non-religious authority, including highly religious ones — and this is the founders' actual observation, which has held.

Decline is regional and requires explanation rather than assumption. Where it happens, it needs a mechanism — security, politics, generational replacement — and those mechanisms are identifiable.

And privatisation was simply wrong. Religion is a public actor, contesting law and identity, in the majority of the world's societies.

Four questions for any claim in this area.

Which of the three claims is being made — differentiation, decline, or privatisation?

Which measure — believing, belonging or behaving? They come apart in opposite directions in different countries.

Is the attendance figure reported or counted? The gap is large and it is largest where attending is expected.

And which societies? A statement about Western Europe is not a statement about the world, and the assumption that it was is what produced the error.

One closing note on what the episode teaches methodologically. The thesis was not falsified by a clever study. It was falsified by the world, over thirty years, in front of everybody — and the field's response was to disaggregate the claim, identify which component had failed, and rebuild.

That is a degenerating programme becoming progressive (see 7.2.3), and it is what the honest response to a failed prediction looks like.

In 1968 a leading sociologist predicted that by the twenty-first century believers would survive only in small resistant sects. In 1999 he stated publicly that the assumption of a secularised world was false, and that the theory had generalised from Western Europe and a Western-educated intelligentsia.

The evidence: religious adherence grew absolutely, and the unaffiliated share of the world population is projected to fall on demographic grounds; Pentecostal and charismatic Christianity expanded enormously in the global South; religion revived after communism; and it re-entered public life through revolution, mass movements, liberation theology, party politics and religious nationalisms.

The thesis was three claims fused. Differentiation — the loss of religious jurisdiction over law, medicine, science and administration — is largely true and is the defensible core. Decline in individual belief and practice is true of Western Europe and false almost everywhere else. Privatisation is simply wrong: religions deprivatised, re-entering public contestation in the decades their withdrawal was predicted.

Measurement comes apart three ways. Believing without belonging describes parts of Western Europe, where institutional attachment collapsed faster than belief. Belonging without believing describes the Nordic pattern of high membership and low practice. And vicarious religion describes a non-practising majority content to be represented by a practising minority. Self-reported attendance is roughly double counted attendance, and the inflation is largest where attending is socially expected.

The supply-side account — that monopoly produces low participation and competition high — rests on a pluralism–participation relationship that systematic review found not robust, assumes constant demand, and requires a golden age of faith that historians contest.

The existential security account predicts the global pattern well and resolves the American anomaly by measuring security rather than income. Its difficulties are the post-communist revival and cases where religiosity rose with prosperity.

And the standing counterexample has begun to move: American religious non-affiliation rose from around one in twenty to approaching three in ten, concentrated among younger cohorts, with the best-supported explanation political — disaffiliation among moderates and liberals following organised religion's visible partisan alignment. Secularisation caused by politics rather than modernisation, operating first on the loosely attached — so the remaining religious population becomes more distinct rather than simply smaller.

Secularisation thesis — the prediction that modernisation produces religious decline; three claims travelling together.

Differentiation — the separation of institutional spheres and the loss of religious jurisdiction over them.

Privatisation / deprivatisation — the predicted retreat of religion to the private sphere, and its observed return to public contestation.

Believing, belonging, behaving — three measures of religiosity that come apart in opposite directions.

Believing without belonging / belonging without believing / vicarious religion — three observed patterns of divergence.

Attendance inflation — the gap between reported and counted attendance, largest where attendance is expected.

Religious economies / supply-side theory — the account attributing participation to competition rather than demand.

Existential security theory — the account attributing religiosity to exposure to unmanaged risk.

Political disaffiliation — secularisation driven by the partisan alignment of organised religion.

The nones — those reporting no religious affiliation.