n.V1-6.03 | DRAWING A LINE ON THE MAP

01| Most anti-poverty policy is written for categories of person and applied evenly across territory. The alternative is to target the territory: draw a boundary, concentrate resources inside it, vary the programme by local conditions. Enterprise zones, area regeneration, slum upgrading, promise neighbourhoods, participatory budgeting and special districts are all instances of it.

02| The intuitive case is that uniform programmes miss local context. The case against is an economic objection intuition does not reach: a policy attached to a place can move activity rather than create it, and can raise the value of the place rather than the incomes of the people in it. Both effects are invisible to any evaluation that looks only inside the boundary — and most evaluations look only inside the boundary.

03| What gets called place-based policy is at least five different machines sharing one feature. Fiscal incentives to firms inside a line. Concentrated public investment in a designated area. Physical upgrading of dwellings and infrastructure. Bundled multi-sector services in a catchment. Devolved decision rights over a local budget. They share a boundary and nothing else, the evidence for each is different, and the strongest evidence attaches to the least fashionable of them.

Created or relocated

04| Neumark and Simpson's survey is the reference. Their verdict on enterprise zones is that the evidence is very mixed and that the field still cannot say which design features matter, who gains, who loses, or how the conflicting findings reconcile. Displacement is the central analytic problem: because the policy is attached to a location, firms and workers move in response, so benefits can land on people who were never the target and gains inside a zone can be offset by losses outside it. They find more consistently positive results for infrastructure and for higher-education and university research investment — plausibly because those have public-good characteristics that resist relocation.

05| The French case shows the problem in clean form. Mayer, Mayneris and Py evaluated the zones franches urbaines and found a positive, sizeable effect on the probability that an establishment locates in the ZFU part of a municipality rather than the non-ZFU part of the same municipality. That is a real effect on firm location. Read literally, it is also movement across a line within a single town. The counterfactual question — using the rest of the city as control will overstate, using the rest of the country may understate — is rarely stated at all, let alone defended.

06| Against the general scepticism, one well-identified case runs the other way. Busso, Gregory and Kline found that Round I US Empowerment Zones raised employment and wages for local workers without raising the local cost of living, which is the incidence result the theoretical prior says should not happen. One case is not a body of evidence, and the theoretical prior remains what it is.

Does the bundle add anything

07| This is the question most relevant to any comprehensive proposal, and the answer is uncomfortable. The Harlem Children's Zone is the model case: charter schools plus a dense set of community programmes across a defined catchment, on the theory that the bundle produces the effect. Dobbie and Fryer tested it with both lottery and instrumental-variable identification. Attending an HCZ middle school produced effects large enough to close the black–white achievement gap in mathematics; elementary school effects closed it in mathematics and English.

08| Their finding about the bundle is the part that matters: high-quality schools appear sufficient to raise achievement substantially among the poor, and the community programmes appear neither necessary nor sufficient.

09| The comprehensiveness claim — poverty is multidimensional, therefore interventions must be too — is the founding premise of bundling. In the best-evaluated instance of the model, the bundle's effect was attributable to one component. That does not prove bundling never adds value. It does mean the burden of proof sits with whoever asserts it does, and that decomposition belongs in the design rather than being assumed away.

The largest one

10| England's New Deal for Communities is the most intensive area-based initiative that country has attempted: 39 deprived neighbourhoods of roughly 9,900 people each, £1.71 billion across some 6,900 projects between 1999–2000 and 2007–08, plus £730 million levered from elsewhere, evaluated nationally from 2001 to 2010. Its summary finding is that there was considerable positive change, that many of the 39 areas were transformed over the decade, and that in general they narrowed the gaps with the rest of the country.

11| The most useful thing about that evaluation is its structure rather than its verdict. It separates three place-related outcomes — crime and safety, community and housing, physical environment — from three people-related outcomes — health, education, worklessness. That distinction is precisely the one the whole field needs, because a place can improve while the people in it do not, either because the gain capitalises into rent or because the residents change. The differential between the two families is the single most decision-relevant number in this literature, and reading it out requires going into the report volumes rather than the summary.

Devolving the choice

12| Whether handing the decision to the neighbourhood changes what gets chosen has been tested directly. Olken's Indonesian experiment found that moving from representative meetings to direct plebiscites dramatically raised satisfaction, knowledge and willingness to contribute, while changing actual project selection only slightly. Alatas and colleagues found community-based targeting somewhat less accurate than a proxy means test on a consumption metric but higher in satisfaction — with the gap explained by communities using a different concept of poverty, not by elite capture.

13| So devolution should be justified by what it does to legitimacy, participation and maintenance, which are real things that may matter over long horizons, and not by a claim that local people will pick better projects. That claim has been tested and it is not supported.

14| Two interventions in this family have unusually clean results and both are physical. Replacing dirt floors with cement produced large measured gains in health, child cognition and maternal mental health (Cattaneo and colleagues). Basic housing upgrades improved child health and life satisfaction at low cost (Galiani, Gertler and Undurraga). Neither is a zone, neither creates a boundary, and neither can relocate.

15| Two questions determine whether a place-based programme is real, and both must be answered outside the boundary: where did the activity come from, and who ended up holding the gain. An evaluation confined to the designated area cannot answer either. Most are.