The Trust Deficit

Prologue: A conversation with your grandmother#

Ask your grandmother about the house she grew up in and she will tell you, almost without being asked, that the door was not locked.

She will tell you that a stranger could arrive at dusk and be given water, and a mat, and a place to sleep. That when a marriage had to be conducted and the family had no money to conduct it, the marriage happened anyway, because a hundred small hands appeared and did a hundred small things. That when a house was being built, people came. That children could be left with neighbours for a week while the parents went on pilgrimage, and nobody thought this remarkable enough to mention.

And then she will say the sentence that should stop us: we had nothing, and we were happy.

We hear this and we feel something contract in the chest. We recognise the loss even though we never had the thing. And the instinct is to build the whole argument on that feeling — to say: look how much we have gained materially and how much we have lost otherwise, look at the gated compound and the CCTV camera and the child who has never once walked to a neighbour's house alone.

I want to resist that instinct, and I want to explain why, because the explanation is the beginning of the argument rather than an interruption of it.

1. Why the feeling is not enough#

In 2023, Adam Mastroianni and Daniel Gilbert published a study in Nature that ought to be read by everyone who has ever written a sentence beginning "society today."

They assembled 235 survey questions, asked between 1949 and 2020 across 59 nations and put to more than half a million people, all asking some version of: are people less kind, less honest, less decent than they used to be?

In 84% of the American surveys and 86% of the international ones, a majority said yes. In every one of the forty countries Pew surveyed on this, majorities called moral decline at least a moderately big problem.

And yet when the researchers turned to a separate body of questions — some twelve million responses asking people about the actual morality of those around them right now, their neighbours, their colleagues, the strangers they dealt with that week — those reports were flat. Stable across fifty-five years of American data. The year of the survey explained less than three-tenths of one percent of the variance.

So for more than half a century, across every conceivable political and economic condition, and replicated across 59 nations, human beings have reported a moral decline that the data cannot find. The authors call it "the illusion of moral decline," and they propose a mechanism that is uncomfortable in its simplicity: we are exposed disproportionately to bad news about the present, and we remember the past with the bad parts faded out. Biased exposure, biased memory. The two together manufacture a golden age that recedes at exactly the speed we walk toward it.

I am not going to build an argument on your grandmother's memory. Not because she is lying — she is not — but because the honest position is that her memory is not admissible as evidence, and neither is mine, and neither is yours.

What I am going to build the argument on is a number that is not a memory. A number that was measured, the same way, by the same instrument, at six points across thirty years, by people who were not trying to prove anything about India.

2. The number#

The World Values Survey has, since the early 1990s, asked people in dozens of countries a single question: generally speaking, would you say that most people can be trusted, or that you need to be very careful in dealing with people?

It is a crude question. It is also the closest thing the social sciences have to a thermometer for the thing this essay is about. Here is India's reading, wave by wave:

Read that column again. Between the 2004 wave and the 2014 wave — one decade, the decade of India's fastest economic ascent — the proportion of Indians who believed most people could be trusted fell by more than half. Twenty-two percentage points, gone. And it has not recovered since. It has simply stayed at the floor.

This is not nostalgia. Nobody was asked to remember anything. Each cohort was asked about the present, and each successive cohort reported a present in which fewer people could be trusted than the last cohort had found.

For scale, here is where that leaves us in the 2022 wave:

We are not in the middle of that table. We are near the bottom of it, and we arrived there recently, and we arrived there fast, and we arrived there during the boom.

One warning, before anyone quotes the 17% at me. Trust figures are extraordinarily sensitive to how the question is worded and who is asked. Pew's 2025 survey, using a binary wording, put India nearer 38%. An Ipsos online panel in 2022 put it at 56%. A WHO study across six Indian states in 2007–10 also found 56%. Pew is explicit that "there is no universally accepted way to measure social trust, and different measures obtain different results" — using the classic wording, 34% of Americans trust; using a binary, 55% of the same population does.

So the level is not a fact about India. It is a fact about an instrument. What I rely on is not the level but the trend, and I rely on the World Values Survey specifically because it is the only series that has asked Indians the same question in the same way over thirty years. A survey artefact does not usually halve itself in one decade and then hold flat for eight years.

3. The Indian asymmetry#

Now hold that beside a second fact, because the two together are the most interesting thing I know about this country.

The Edelman Trust Barometer, which measures trust in institutions — government, business, media, NGOs — placed India third out of twenty-eight countries in 2025, with an index of 75, behind only China and Indonesia. On institutional trust, India is near the top of the world.

I want to be careful here, because these two surveys measure different things by different methods, and Edelman's India sample is an online panel skewed urban and affluent. They are not two readings of the same dial, and the comparison is suggestive rather than decisive. But even allowing generously for all of that, the shape is too stark to ignore:

It matters enormously, and here is why. If the diagnosis were "Indians have become cynical about everything," the prescription would be one thing. But that is not what the numbers show. Whatever confidence Indians have withdrawn, they have not withdrawn it from the idea of collective institutions.

What has thinned is something far more local and far more particular: the working assumption that the specific human being in front of you will not take advantage of you.

That is not general cynicism. It is something narrower and more treatable. It is the loss of a default.

4. What trust actually is, and why its loss registers so deep#

I want to be precise about what is being lost, because "trust" is a word that has been worn smooth by overuse.

Trust is not affection. You can love someone you cannot trust and trust someone you do not love. Trust is not optimism about human nature. Trust is a default setting on a calculation you would otherwise have to run.

When you trust, you stop computing. You hand over the money without the paperwork. You leave the child without the arrangements. You accept the mat and the water without wondering what is wanted in return. Every one of those is a computation not performed, and the sum of all the computations you do not have to perform is the thing your grandmother was describing when she said she was happy with nothing.

Reverse it, and you get the modern condition. If I must assume that the person in front of me may be extracting from me, then I have to stay alert. Not occasionally — continuously. Alert at the shop, alert in the contract, alert with the contractor, alert with the school, alert with the neighbour whose name I have not learnt.

Continuous alertness is not a mood. It is a metabolic cost. It is the body's threat-detection system left switched on in a situation it was designed to handle in bursts. And a system built for bursts, run continuously, does what any system does under that treatment: it wears.

This is the point at which the argument stops being sociology and becomes medicine, and as a physician it is the part I find hardest to look away from.

Julianne Holt-Lunstad and her colleagues have twice pooled the world's evidence on this. In 2010, across 148 studies and 308,849 people followed for an average of seven and a half years, they found that people with stronger social relationships had a 50% greater likelihood of survival — and for the richer, multidimensional measures of social integration, 91%. In 2015, across 70 prospective studies and 3.4 million people, they found social isolation associated with a 29% increased likelihood of mortality, loneliness 26%, living alone 32%.

Crucially, that second study found no meaningful difference between objective isolation and subjective loneliness. It is not only about how many people are around you. It is about whether you experience yourself as held.

In June 2025, the World Health Organization's Commission on Social Connection published its first flagship report and put a number on it: 871,000 deaths a year linked to loneliness — about a hundred every hour. One in six people worldwide affected. And a detail that should interest us more than any other: loneliness runs at roughly 24% in low-income countries against 11% in high-income ones. Twice as high where there is less money.

I use the WHO's own word: linked. Not caused. This is association, not proof of mechanism. But the association is large, it is replicated, and it points in one direction.

And one Indian figure complicates the picture, so let me put it here rather than bury it. The Longitudinal Ageing Study in India, surveying 30,394 people aged sixty and above, found that 13.4% of Indian elderly report frequent loneliness — lower than most Western estimates for the same age group, and well below the WHO's global figure for older adults. If the argument were that Indian elders are uniquely abandoned, this number would not support it. What the LASI data does show is what predicts loneliness once it occurs: having no social contacts roughly doubles the odds. The structure is still holding for most Indian elders. The question this essay is asking is how long that lasts, and what happens to the generation now being raised without it.

The word I keep returning to is void — because that is what it feels like from inside, and because it is not merely a psychological void. It is closer to what the older Indian traditions would have called an existential one: a gap not in your mood but in your footing. Something you stand on that is no longer there.

5. Why we reach for money, and why money cannot reach#

Here is the mechanism I want to propose, and it is the centre of this essay.

The material rush is not the cause of the trust deficit. It is the compensation for it.

If I cannot rely on my neighbour, I must become self-sufficient. If I cannot rely on my extended family, I must insure myself. If I cannot rely on the community to see my children through, I must accumulate enough that my children need no one. Every rupee accumulated is a relationship rendered unnecessary — and that is precisely its appeal to a person who has learnt that relationships cannot be relied upon.

Money is the technology we have for not needing to trust anyone. Which is exactly why a society losing trust will reach for it with both hands.

And it half-works. That is the cruel part. It buys the security guard, the private school, the health insurance, the second house. It genuinely does substitute for a great deal of what community once provided. If it did not work at all, nobody would do it.

But there is a boundary, and the research has recently located it with unusual precision. I want to state this correctly, because it is the single most misquoted finding in popular writing on happiness.

In 2010, Daniel Kahneman and Angus Deaton found that emotional wellbeing rose with income up to about $75,000 a year and then flattened. This became the famous line: money stops buying happiness at $75,000.

In 2021, Matthew Killingsworth, sampling over a million real-time moments, found the opposite: happiness kept rising with log income, just as steeply above $80,000 as below.

So in 2023 they did the honourable thing. Kahneman, Killingsworth and Barbara Mellers ran an adversarial collaboration — they sat down together with the data to find out who was wrong. The answer, published in PNAS, is more useful than either original finding:

For most people, money keeps buying happiness, indefinitely, and for the happiest group it even accelerates. The flattening is real but it appears only in the least happy fifth or so of people — for whom happiness rises with income to around $100,000 and then stops dead.

Kahneman's own summary was that he might have got it right had he described his results in terms of unhappiness rather than happiness.

Sit with what that means. Money buys the removal of the miseries money can remove. It does that job very well, all the way up. But there is a residue of suffering in the least happy that income simply cannot touch — a suffering that keeps its shape no matter how much is spent on it.

That residue is where I would locate the void. Not everywhere. Not in every person. But precisely where the material solution runs out, and precisely where the loss of trust would be expected to sit.

The man who says "if I earn a little more I will finally be at ease" and earns it, and is not at ease, and concludes that the number must have been wrong — he is not making an arithmetic error. He is applying a working instrument to a problem outside its range.

An argument that has not been attacked has not been tested. What follows are the six strongest cases against everything I have written above. I have not softened them. Three of them have changed what I believe.

Six objections in, the thesis I started with has been cut down considerably. Here is what remains standing, and I believe every word of it is defensible:

I should be careful with that formulation. The evidence below is associational and observational: it tells us what high-trust communities look like, not that copying their features will manufacture trust. But it is the best guidance available, and it is remarkably consistent.

The finding that changes everything: it is not warmth, it is willingness#

In 1997, Robert Sampson, Stephen Raudenbush and Felton Earls published a study in Science that I regard as the most important single piece of evidence in this whole field. They surveyed 8,782 residents across 343 Chicago neighbourhood clusters, and they measured something they named collective efficacy: "social cohesion among neighbors combined with their willingness to intervene on behalf of the common good."

A two-standard-deviation increase in collective efficacy was associated with a 39.7% reduction in the expected homicide rate. And it partially mediated the association between concentrated disadvantage and violence — much of what looked like an effect of poverty ran through it.

Now read the definition again and notice what is in it. Not affection. Not friendship. Not shared background. Not how much the neighbours like each other.

Willingness to intervene.

The variable that predicts whether a neighbourhood is safe is whether an adult who sees a child doing something dangerous will actually say something. Whether someone will step in. Whether the norm is enforced by ordinary people in ordinary moments.

This should reorganise how we think about community entirely. The warm community of the imagination — everyone friendly, everyone eating together, everyone fond of one another — is not what the data rewards. What the data rewards is a community where people act. Where standards are held by anyone present rather than delegated to an authority. Where intervention is normal and unremarkable and not experienced as interference.

Ostrom's eight, and the one that matters most#

Elinor Ostrom won the Nobel in economics for demonstrating that communities can govern shared resources without either privatisation or state control — but only under identifiable conditions. Her eight design principles, from Governing the Commons (1990):

A review of 91 empirical studies found these principles well supported. Note that four of the eight — monitoring, graduated sanctions, conflict resolution, collective choice — are about enforcement and rule-making. Ostrom's communities do not run on goodwill. They run on visible, proportionate, locally-owned accountability. This is exactly Sampson's willingness-to-intervene, formalised.

And the eighth principle is the answer to a question I have been getting wrong.

The 500-family question, answered honestly#

I have spoken about clusters of five hundred families who all know each other. I now think that specific formulation is not achievable, and it is worth explaining precisely why, because the correction improves the idea rather than abandoning it.

Robin Dunbar's work maps human social networks in layers, each roughly three times the one inside it: 5 intimates, 15 close friends, 50 good friends, 150 friends, 500 acquaintances, 1,500 faces you can name.

Five hundred families is somewhere above two thousand people. That exceeds every layer, including the outermost. A resident of such a settlement could not put a name to every face, let alone hold mutual obligation with them. And note what the 500 layer is actually labelled: acquaintances. People you recognise. Not people you trust.

(I should add, in fairness, that Dunbar's specific number 150 has been seriously challenged — a 2021 analysis in Biology Letters re-ran his method and found confidence intervals so wide that "specifying any one number is futile." The layered structure is well observed in real contact data. The precise figures are not laws of nature. I use them as orders of magnitude, not constants.)

There is better evidence anyway, and it comes from watching actual cooperation rather than inferring a cognitive ceiling. Yang and colleagues, publishing in PNAS in 2013, tracked forest-monitoring groups in China's Wolong Nature Reserve and found that group size has nonlinear effects: "intermediate group size contributing the most monitoring effort and leading to the biggest forest cover gain." Too small fails. Too large fails. Intermediate wins. What made the intermediate groups work: punishing free riding, within-group enforcement, social capital among members, and self-selection at formation.

So the answer is not to abandon the number. The answer is to nest it.

A cluster of five hundred families should be built as roughly thirty-family units of some twenty to forty households each — small enough that everyone genuinely knows everyone, that a missing child is noticed within the hour, that free-riding is visible — federated upward into the larger body which handles what small units cannot: capital, external representation, standards, dispute escalation.

This is Ostrom's eighth principle. And it is not theoretical. It is how the most successful community institutions in India are already built.

Two Indian proofs#

Ten to twenty households at the Kudumbashree base is well inside Dunbar's fifty-person layer, with elected accountability and nesting upward through two more tiers. This is Ostrom's architecture implemented at state scale in India, and it has run for twenty-eight years.

Neither of these is an intentional community, a spiritual settlement, or a philosophy. Both are federations of small accountable units doing concrete work together. That is the form that has actually worked in this country, at scale, for decades.

I have spent most of this essay dismantling the loose version of my own argument. Now let me say what survives, and what it means for the work our movement is doing — because I think the evidence endorses a good deal of what ARISE already does, and demands changes in some of it.

What the evidence endorses#

The neighbourhood unit is the right unit. Sampson's finding is a neighbourhood finding. Collective efficacy is measured and matters at the scale of a few hundred metres. ARISE's Nurture Centres are neighbourhood institutions serving the children of the streets around them. That is where the evidence says the effect lives.

The scale is already close to correct — and it is correct precisely because it is not five hundred families. ARISE's own description of the model is "fifty associates, formed in one home, carry free evening education to about a hundred children of their own community." Twenty-five associates constitute a club; fifty is the target. So a mature club and its centre is a body of roughly a hundred and fifty people — Dunbar's friendship layer, almost exactly. Small enough for everyone to be known, large enough to do real work. The evidence does not support two thousand people knowing one another. It supports something very close to what is already built.

The structure is already nested. Centre → club → foundation is Ostrom's eighth principle. What Kudumbashree does with neighbourhood groups, area societies and community societies, ARISE does with centres, clubs and the foundation. This is not a coincidence of design; it is convergent evolution toward the only architecture that works.

Open enrolment is the answer to Ambedkar, and it is the most important thing in the whole design. ARISE's own rule for the Nurture Centres is that "a child simply comes" — no test, no fee, no form of selection and no minimum commitment. Enrolment is by geography and school stage — classes 6 to 10 — and nothing else.

I want to be emphatic about how much rests on that clause, because I do not think it has been understood as the load-bearing principle it is.

Every filter you add — a fee, an entrance test, a recommendation, an interview, a form requiring documents — becomes, in Indian conditions, a caste filter and a class filter, whether or not anyone intends it. Selection is how bonding capital reproduces itself. A centre that admits by geography and refuses every other criterion is a bridging-capital machine, and it may be the only kind of bridging-capital machine that can be built cheaply and at scale in this country.

It puts a Brahmin child and a Dalit child in the same room on the same bench with the same teacher and the same expectation, every evening, for years, because they live on the same street. Against the 27% and the 70%, that is not a soft intervention. It is close to the only structural one available.

The child protection architecture is Ostrom's principles 4, 5 and 6. Screened mentors, the two-adult rule, attendance called at arrival and departure, children released only to named guardians, external reporting routes that bypass internal hierarchy. This reads as safeguarding, and it is. But it is also monitoring, sanction and conflict resolution made visible — and Ostrom's evidence says that visible enforcement is not a constraint on trust. It is trust, in its only durable form. The households that send their children are not extending faith. They are observing a rule being kept, every evening.

The zero-fee rule removes money from the relationship. Given everything in Part I about money as the substitute for trust, an institution in which no transaction occurs between the people involved is doing something structurally important, not merely generous.

The Balasabha may be the single most undervalued element. The children's own Friday assembly — "convened, chaired and run by them, with the associates present but not in charge" — is collective-choice arrangement (Ostrom's third principle) taught to eleven-year-olds. And recall the intentional-communities finding: egalitarian decision-making structures beat every other predictor of whether a community holds together. If you want a generation capable of self-governance, this is where it is learnt, and it is learnt by doing rather than by being told.

Four things I think must change#

1. Stop saying the joint family collapsed. It did not. Census 2001 to 2011: 70.34% to 70.11%. The claim is false and it is easily checked, and every time we make it we hand a critic a free win on a point that is not even necessary to the argument. The true claim is stronger and more interesting: household structure held while the circle of obligation around the household thinned. Say that instead.
2. Retire the charitable account of caste's origins. I have offered it and I was wrong to. It gains nothing and it costs the movement its standing with exactly the people whose participation would prove the model works. The defensible position is sharper: caste is what Indian community becomes by default when nothing prevents it, which is why open enrolment is not a nicety but the core of the design.
3. Measure collective efficacy, not just attendance. Sampson's instrument exists and is well validated. A short survey in the streets around each centre — would a neighbour intervene if children were skipping school, if there was a fight, if something was being damaged; can you count on people here — administered at the start and again after two years, would tell us whether the centres are producing the thing we claim. Hours taught and children enrolled measure effort. Collective efficacy measures the outcome. And it is the difference between an inspiring movement and a demonstrated one.
4. Take the weak-ties cost seriously — with one honest qualification. A child educated by the same associates on the same street for five years is receiving support and, in network terms, redundancy. Granovetter and the LinkedIn experiment both say opportunity arrives from outside. The centres should deliberately import strangers — visiting professionals from other cities and other fields, exchanges between centres in different districts, structured outward introductions for every child leaving Class 10.
   
   The qualification, which cuts against me and which I include because leaving it out would be exactly the selectivity this essay claims to avoid: the same LinkedIn study found that weak ties help most in digital industries, while strong ties help more in low-technology sectors. Many of the children in our centres will enter precisely those low-technology sectors, where the dense local network may serve them better than distant acquaintances would. So this recommendation is a hedge against an uncertain future rather than a correction of a known error. Warmth is what a centre supplies naturally; reach must be engineered — but engineered as an option offered, not a destination assumed.

The claim I think is genuinely defensible#

Here is the sentence I would be willing to defend against any critic:

That is a narrow claim. It is much narrower than the one I began with. And I think it is true, and I think almost nothing else in Indian public life is currently attempting it.

Epilogue: The door#

I began with your grandmother's unlocked door and I said I would not build an argument on it. I have kept that promise. Every load-bearing claim above rests on a measurement rather than a memory.

But let me return to the door, because now we can say what it actually was.

The open door was not a sign of a gentler people. Those people practised untouchability, married their daughters inside the caste, and enforced a hierarchy that Ambedkar spent his life fighting. The door was open to some and shut absolutely to others, and there were people for whom no door in that village would ever open.

The open door was a piece of infrastructure. It signified that the cost of one person's misbehaviour would be borne by that person immediately, visibly, and locally, because enough people were watching and enough of them would act. It was Sampson's collective efficacy three generations before anyone measured it. It was Ostrom's monitoring and graduated sanctions with no committee and no minutes.

We did not lose the door because we became worse people. We lost it because we scaled past the point where watching worked, and we replaced watching with locks, and with insurance, and with money — and those substitutes are genuinely better at some things and completely incapable of others.

The task is not to go back. There is nothing to go back to that any of us should want. The past was less safe than we remember, far less inclusive than we admit, and the trust it did contain was rationed by birth.

The task is to build, deliberately and from nothing, the thing the old village produced by accident — small enough to be seen, accountable enough to be trusted, and open enough to include the people the old village would have turned away.

That is what a Nurture Centre is, when it works. Fifty young people who know a hundred children by name, on the handful of streets they all share, holding a standard that any of them will enforce, and admitting every child who walks in without asking a single question about who their parents are.

It is not a village. It is better than a village, because it is chosen and it is open.

And if there is a hundred of them, that is a district. If there is one on every campus in India, that is something no policy can purchase and no amount of money can substitute for.