n.V1-9.05 | The Room People Walk Through

01| Everything wrong with poverty work at a distance — no test, no verification, nobody checking — should disappear when the work happens close to home. A neighbourhood is small. You can walk it in twenty minutes. The people affected live at addresses you can visit. The statistics already exist: school results, crime figures, employment rates, health records, collected by somebody anyway. If any kind of poverty work can be tested, this is the kind.

02| That reasoning is why enormous amounts of money were spent finding out. The results are stranger than either the sector's supporters or its critics usually admit. It is not that nobody measured. People measured for twenty years, carefully, with serious money behind them. The problem is what the measurements came back and said.

The initiatives built to be evaluated

03| Through the 1990s and 2000s, American foundations funded a wave of what came to be called comprehensive community initiatives. The idea was reasonable and, in a way, humble: poverty in a poor neighbourhood is not one problem but many tangled together, so attacking any one alone will fail. Work on all of them at once — housing, schools, jobs, health, safety, local leadership — in one defined place, for long enough to matter.

04| These were not casual efforts. They were designed with evaluation in mind, ran for years, and were studied throughout by some of the best people in the field. In 2010 the Aspen Institute Roundtable on Community Change published the summary of what had been learned, and its central conclusion is one sentence that ought to be better known: most of the interventions had not produced the degree of community transformation their designers envisioned.

05| The detail matters more than the headline. The initiatives could show increased community capacity — stronger local organisations, more capable leaders, better networks. They could show improved outcomes for some residents. What they could not show was population-level change in the neighbourhood's main indicators. The place, measured as a place, largely did not move.

06| Twenty years, hundreds of millions of dollars, serious evaluation, and the answer came back ambiguous. Sit with that, because the reflex response — they must have done it badly — turns out to be too easy.

Why a neighbourhood is a difficult thing to measure

07| Here is the problem underneath everything else in this chapter, and it is not a technical detail. It is the whole difficulty.

08| A neighbourhood is not a fixed group of people. It is a place people move through. In poor neighbourhoods they move through it quickly — leases end, jobs appear elsewhere, families double up, rents rise.

09| Now consider what success looks like. If a programme genuinely improves people's lives, a good number of them will use that improvement to leave. Their gain walks out of the data. Meanwhile new arrivals — poorer, or simply not yet helped — walk in. The neighbourhood's statistics can stay flat while every single person the programme touched is better off.

10| The reverse is just as possible. A neighbourhood's numbers can improve beautifully because the poor residents left and better-off ones replaced them, without one original resident's life improving at all. The place looks like a success. The people it was for are somewhere else.

11| So when a neighbourhood's indicators move, or fail to move, you do not yet know what happened. You have to know who is in the numbers. Most neighbourhood work has never been able to answer that, and it is not obvious that most of it has tried. This is not an absence of measurement. It is the absence of a stable thing to measure.

The one real experiment

12| There is one major exception, and it is why this argument can be had with evidence at all.

13| Moving to Opportunity was a federal experiment beginning in the 1990s. Families in high-poverty public housing were selected at random to be offered vouchers to move to lower-poverty areas. Random selection is what makes it powerful: the families who moved and the families who did not were, on average, alike in everything except the offer.

14| The early results disappointed almost everyone. Adults who moved were not measurably better off in work or earnings. Their mental health improved and their sense of safety improved, which is not nothing. But the central hope — that a better neighbourhood produces better economic outcomes — did not appear. Girls seemed to do better; boys, on several measures, did worse. For roughly fifteen years MTO was cited as the strongest evidence that neighbourhood approaches do not deliver.

15| Then in 2016 Raj Chetty, Nathaniel Hendren and Lawrence Katz published a reanalysis using tax records, which let them follow the children into adulthood. Children who moved before roughly age thirteen went on to attend college more often and earn more. Children who moved as teenagers did slightly worse, probably because the disruption outweighed the benefit at that age.

16| The finding is genuinely important, and it rescued a claim that had looked dead. But notice what it required: a randomised experiment, tax records, and twenty years of waiting. That is the price of one clean answer about neighbourhoods. Almost no programme will ever pay it.

17| And even this is not settled. In 2020 Robert Kaestner argued that the analysis does not establish what people take it to establish — that lowering a neighbourhood's poverty rate is what improves children's futures. Chetty, Hendren and Katz replied in the same journal, noting the core earnings result stands and that the critique does not dispute it. Both positions are honestly held and publicly argued. The dispute is not about whether the data are real. It is about what a real result licenses you to do next — and that is a question almost every neighbourhood programme quietly answers for itself without noticing it is a question.

The flagship, and the argument about it

18| The most famous neighbourhood programme of the era is the Harlem Children's Zone, which set out to surround children in a defined area with support from birth onward: health services, parenting programmes, after-school provision, counselling, its own charter schools. It was celebrated, visited, written about, and in 2010 the federal government launched Promise Neighborhoods to reproduce it in twenty other cities.

19| That same year Grover Whitehurst and Michelle Croft asked an uncomfortable question. If the Zone's results come from wrapping children in services, its schools should outperform schools without those services. They compared the Zone's charter schools to other charter schools in Manhattan and the Bronx offering no comparable social supports. The Zone's schools landed in the middle of the pack. Several schools with none of the surrounding services did better.

20| Their conclusion was careful, and the report was widely misreported as an attack on the Zone. It was not. Their objection was to the use of the Zone as evidence that neighbourhood-wide wraparound services are a cost-effective route to academic achievement — and to a national programme built on that evidence. The two claims are different, and the difference is the subject of this chapter.

Six billion dollars and the question of who benefited

21| Alongside the initiatives ran something larger: HOPE VI, the federal programme that demolished distressed public housing and rebuilt it, generally as lower-density mixed-income development. More than six billion dollars went through it.

22| The physical change is not in dispute. Buildings came down, new ones went up, often visibly better. The dispute is about the people.

23| The Urban Institute followed original residents across several sites. The findings resist both slogans. Residents who relocated generally ended up in better housing and safer neighbourhoods, and their children attended less poor schools; those who moved with vouchers did better than those who moved into other public housing. But economic outcomes did not improve. And the most vulnerable residents — the elderly, those with the deepest problems — did worst, with real concern about housing instability among them.

24| Critics put it more sharply: in practice HOPE VI often amounted to gentrification and the forced relocation of the people it was meant to serve. Defenders answer that leaving families in failing towers was not a neutral alternative. Both are arguing from the same evidence, and the disagreement is not resolvable by adding more of it — because it turns on the same instability as before. Improve the housing, and the question of who lives in it afterwards is not a side issue. It is the whole question.

Should we be fixing places at all?

25| Behind all of this sits an argument economists have had for decades, and it is the bluntest version of the challenge. If you want to help poor people, help poor people — with money, vouchers, transfers that follow the person. Do not pour money into a place. Edward Glaeser put the position in 2007 asking whether a declining city could come back; his subtitle answered for him: probably not, and government should stop bribing people to stay there.

26| The technical version is more damaging than the rhetoric. In the standard model, money spent making a place more attractive gets absorbed into what it costs to be there. Rents rise to match the improvement. The value ends up with whoever owns the land, not whoever lives on it. And across the place-based literature, a recurring finding is that the benefits of neighbourhood investment often accrue to developers and higher-income newcomers rather than to the low-income residents who were there first.

27| The counter-argument has been gaining ground: that concentrated poverty produces effects that are not the sum of individual poverty — that a neighbourhood where almost everyone is poor damages people in ways a cheque cannot repair, which is exactly the market failure that justifies acting on the place. Nobody in this argument is being unreasonable and it is not close to over.

28| Put the pieces together. Neighbourhood poverty work is not unmeasured; it is among the most heavily evaluated social interventions ever attempted. What it lacks is a stable claim to test. The unit is a room people walk through, and almost every available method measures the room rather than the people. The one design that solved it needed randomisation, tax records and two decades, and its results are still argued over. None of this proves neighbourhood work does not work. It shows something more awkward: thirty years and a great deal of money spent without settling what it would even mean for the work to have worked.