Survey data across many high-income countries shows substantial decline in religious affiliation, attendance and belief over recent decades.
The trend is well documented. What's less clear is what replaced it, and the answer is not simply secularism.
What the data shows
The growth category in surveys is people reporting no religious affiliation — frequently the fastest-growing group in countries where it's tracked.
Importantly, this category is heterogeneous. It includes committed atheists, agnostics, people who believe in something but reject institutions, and people who simply never thought about it.
Surveys that distinguish these find that convinced atheists are a minority within the unaffiliated. A substantial proportion report belief in a higher power, in an afterlife, or in supernatural phenomena, while rejecting religious institutions.
Which means institutional decline and decline in belief are different trends proceeding at different rates.
The variation
The pattern is not global.
Decline is concentrated in high-income countries, particularly in Europe and increasingly in North America and parts of East Asia.
Elsewhere, religious affiliation remains stable or growing, and demographic projections generally suggest the global religious population will grow in absolute terms due to differential population growth.
Which complicates any account of secularisation as a universal consequence of development.
The explanations offered
Existential security. One influential account holds that religiosity declines where people feel materially secure — where health, income and social protection reduce vulnerability.
The correlation between measures of security and religiosity across countries is reasonably strong, though causation is difficult to establish.
Generational replacement. Much of the decline reflects cohort effects — successive generations arriving less religious rather than individuals losing religion. Which means the trend continues through demographic turnover regardless of individual conversions.
Institutional trust. Declining confidence in institutions generally, with religious institutions affected alongside others. Specific scandals have accelerated this in particular contexts.
Alternative provision. Functions that religious institutions provided — welfare, education, community, ritual for life transitions — have been substantially assumed by other bodies.
Individualisation. A broader cultural shift towards individual authority over collective, which disadvantages institutions claiming external authority.
What people do instead
The more interesting question.
Nothing, mostly. The largest group appears simply to have dropped the practices without substituting anything. Which is worth noting against accounts that assume a need requiring fulfilment.
Individualised spirituality. Belief and practice assembled personally rather than received. Meditation, various contemplative practices, astrology, elements from multiple traditions.
This is genuinely widespread and is what surveys pick up as spiritual but not religious.
Wellness practices. A category with substantial functional overlap — practices addressing meaning, wellbeing and self-improvement, with ritual elements and community.
Political and social movements. Some analysts have argued that political identity has taken on functions previously served by religious affiliation — providing meaning, community, moral framework and an account of what's wrong with the world.
That argument is contested and the parallels are notable.
Other communities. Sport, hobbies, online communities, and various forms of association providing belonging without doctrine.
What's harder to replace
Some functions appear less well substituted.
Life transition ritual. Birth, marriage and death. Secular alternatives exist and are growing, and death in particular remains poorly served — many people find themselves needing a funeral without a framework for one.
Intergenerational community. Religious congregations bring together people across age groups in a way that most alternatives don't. Sports clubs, workplaces and interest groups tend to be age-stratified.
Obligation to people you didn't choose. Congregations include people you wouldn't have selected, and the obligation to them is part of the arrangement. Chosen communities lack this by construction.
A framework for suffering. Accounts of why difficulty happens and how to bear it. Therapeutic frameworks do some of this work and address it differently.
The research on outcomes
Studies associating religious participation with wellbeing measures generally find positive associations, with the mechanism substantially attributed to social support and community rather than to belief.
Which suggests that what's lost in decline may be more social than doctrinal, and that substitutes providing genuine community may recover much of it.
Evidence on whether they do is limited, partly because the alternatives are recent and varied.
The open question
Whether the functions are needs requiring fulfilment, or arrangements that can simply be dropped.
The large number of people who appear to have dropped them without evident difficulty is evidence for the second.
The persistence of practices with recognisably religious structure among the unaffiliated is evidence for the first.
Both are probably true for different people, which is an unsatisfying conclusion and is what the data supports.
The measurement problem
Worth a caution about the data, since it shapes the conclusions. Survey instruments were largely designed within particular religious contexts and translate imperfectly elsewhere.
Questions about belief in God, about attendance, and about affiliation assume a model of religion involving congregational membership and doctrinal assent, which fits some traditions considerably better than others.
In contexts where religious practice is domestic, ancestral or embedded in ordinary life rather than congregational, standard measures record less than is happening.
Which means cross-national comparisons should be treated with caution, and apparent differences in religiosity may partly reflect what the instruments were built to detect.