n.V1-5.09| TWO KINDS OF SUCCESS

01| Consider a neighbourhood programme that does everything right. It trains people and they get better jobs. It builds capability and the capable use it. Earnings rise for everyone it touches. Five years later the neighbourhood's statistics are unchanged.

02| This is not a paradox and it is not evidence of fraud. It is arithmetic. People who become better off tend to move, and other people move in behind them. The Urban Institute states the mechanism plainly: if the more successful residents leave a distressed neighbourhood and are replaced by others who are less well off, the neighbourhood remains distressed even though individual households from that neighbourhood improved their economic situation.

03| Every part of that can be true at once. That is the whole problem.

04| The idea has a canonical statement. In 1987 William Julius Wilson argued in The Truly Disadvantaged that the concentration of poor families in inner-city American neighbourhoods was itself producing outcomes usually attributed to poverty, and that a particular migration had produced the concentration: black working- and middle-class families leaving through the 1960s and 1970s. Their departure harmed those who remained — not primarily by removing money, but by removing what he called a social buffer, the presence of neighbours whose lives sustained the perception that education leads somewhere and steady work is a real alternative. Remove the people whose lives demonstrate it and the demonstration stops.

05| It was contested immediately, which is the useful part. Douglas Massey and Nancy Denton argued in American Apartheid that Wilson had the cause wrong: the concentration of poverty was produced by racial segregation itself — by constraints on where black families could live at all — rather than by class migration. If they are right, the departure of the middle class is a symptom of a housing market, not an independent cause of decline. Both positions have serious evidence and the dispute has never been settled. Both agree on the observable pattern.

06| There is a second-order problem that keeps this unresolved. Research on neighbourhood effects is persistently defeated by selection. People are not assigned to neighbourhoods; they sort into them, on exactly the characteristics researchers want to measure. If people who leave poor neighbourhoods do better afterwards, that may be because leaving helped, or because the kind of person who manages to leave was already on a different trajectory. This is not a technical caveat. It is the central obstacle in the field, and it is why Moving to Opportunity mattered so much — randomising who received a voucher was the only way anyone found through it.

07| Now put it back on the programme, because the consequence is a choice of unit that organisations almost never make explicitly. If the unit is the person, then someone who trains, earns more and moves to a better area is an unambiguous success and their new address is irrelevant. If the unit is the place, the same event is at best neutral and on Wilson's account harmful to those left behind.

08| Most institutions claim the first unit in their mission statements and report the second in their results. That is the confusion worth naming — not dishonesty, but an unexamined switch between two incompatible definitions of success.

09| And the argument against retention is strong enough that it may simply be right. If individual escape is the outcome, then asking successful residents to stay for the sake of a neighbourhood's aggregate statistics asks individuals to bear a personal cost for a collective number. Nobody owes their neighbourhood their own advancement. The most rigorous finding available — Chetty, Hendren and Katz's reanalysis of Moving to Opportunity — is that children who moved out of high-poverty areas young earned more as adults. The intervention that demonstrably worked was helping families leave.

10| The honest tension is that both things appear true: leaving is good for those who leave, and the departure of those who can leave is bad for those who cannot. No available policy gets one without the other, and nobody has shown that one exists.

11| One attributable failure does fall out of it. Because place-level indicators are what get reported, and because they are structurally insensitive to individual success, an organisation can report flat numbers while doing real good, or improving numbers because poorer residents left and better-off ones arrived. Neither the organisation nor its funder can usually tell which, because almost nobody follows the people rather than the postcode. That is a fixable failing hiding inside an unfixable dilemma.

12| The same discipline has to be applied to the sector's own headline claim about itself, because we have relied on the argument elsewhere. Two claims usually travel together: that aid is biased toward rural areas by some large multiple, and that this is about to become indefensible because the majority of the world's poor will soon be urban.

13| The first is credible and not quantifiable in the form usually given. David Satterthwaite's version — the careful one — is not that cities get less money but that the scale and depth of urban poverty is under-estimated by most governments and agencies, and that the underfunding follows from the mis-measurement. Income-against-a-line does not price what urban life costs: rent, transport, water bought by the jerrycan at many times the metered rate, the fee for a shared toilet. A household above the line spending most of it on rent and water is counted as not poor.

14| But anyone quoting a ratio of six to one or ten to one is doing something the data does not support. The OECD's classification has no urban category. Urban work is scattered across water and sanitation, transport, health, government. Meanwhile agriculture and rural development exist as recognisable categories, so rural spending can be counted and urban spending cannot. Depending on which sectors you code as urban-relevant, the same figures produce ten to one or something close to parity. The absence of the category is itself the finding: a system that can report precisely what it spent on agriculture and cannot report what it spent on cities has already told you how its attention was organised.

15| The second claim fails outright against the study it implicitly rests on. Ravallion, Chen and Sangraula found the poor are urbanising faster than the population — over 1993 to 2002 the count below the dollar-a-day line fell by 150 million in rural areas and rose by 50 million in urban ones. And they were explicit in the other direction on the projection: roughly a quarter of the consumption poor were urban, and the pace of urbanisation implied a majority of the poor would still live in rural areas for many decades. Their data runs to 2002 and anyone using it today should say so. The regional finding is the one that does the real work: Latin America has the most urbanised poverty, East Asia the least, and in marked contrast to other regions, Africa's urbanisation has not been accompanied by falling overall poverty. Cities elsewhere absorbed people and pulled them out. In Africa, on this evidence, they did not.

16| That last point is checkable and does more work than the ratio ever did. We drop the ratio.

17| Which leaves the thing this book part has been circling. Every failure in these nine chapters is emergent. The crisis operation that never ends, the settlement nobody can count, the contract geometry that manufactures the deficit it later cites, the donor renewing out of habit, the agency hiring the best doctor in the district — each is a defensible act by a reasonable actor, and the aggregate is a system that cannot say what it costs, what it delivers, or who is responsible.

18| The temptation is to stop there, and it should be resisted, because an argument in which nobody is culpable is an argument nobody has to answer. Emergent is not the same as innocent. The measurement was not taken. The divisor was not recorded. The comparison was not commissioned. The target was written without a consequence. The ledger was never netted out. Each of those is an omission, and omissions have authors. What is genuinely unchosen is the shape of the aggregate. What is chosen, repeatedly and by identifiable institutions, is the decision not to find out.