Neighbourhood effects on educational attainment. What matters moreExposure to poverty or exposure to affluence?

Agata A. Troost, Maarten van Ham, David Manley•View original

OverviewTechnicalalloy voice

Neighbourhood effects research has operated for decades under a single dominant premise: that concentrated poverty harms individual outcomes and that the policy task is to mitigate those harms. Troost, van Ham, and Manley call this the poverty paradigm, and they argue it is blinding us to something more powerful. If affluence is the stronger spatial force shaping children's life chances, then policies built around poverty alone are solving the wrong problem. That is exactly what their Dutch study finds. Using an entire birth cohort from 1995, tracking residential histories through childhood and adolescence, and measuring educational attainment at age 23 in 2018, they show that neighbourhood affluence has a stronger effect on educational outcomes than neighbourhood poverty for every developmental window they examine. To understand why affluence might outpace poverty as a driver, the theoretical grounding matters. The mechanisms are not symmetric. Poverty effects in the literature tend to work through resource deprivation, social isolation, disorder, and stress. Affluence effects work through a different channel: the transmission of social and cultural capital through local networks. Troost and colleagues draw explicitly on Bourdieu's concept of habitus — the taken-for-granted sense of what is realistic for one's life course. Children in affluent neighbourhoods are surrounded by peers who treat university as the obvious next step, parents who provide tutoring and access to selective extracurricular activities, and built environments with libraries and well-resourced schools. These are Weberian opportunity-hoarding dynamics playing out at the street level. The paper stresses that affluence is not simply the absence of poverty; the accumulation and transmission of capital in wealthy areas follows a different logic entirely, something closer to a Matthew effect where advantage compounds advantage. Context also matters. The authors caution against importing US findings wholesale. The Netherlands has less extreme concentrated poverty, higher welfare spending, denser urban form, and lower crime rates than the American cities that dominate much of the neighbourhood effects literature. That means the stress-based and disorder-based mechanisms that amplify poverty effects in US studies are likely attenuated in the Dutch case. Which makes it all the more striking that even in this comparatively equitable context, neighbourhood affluence still dominates. Now, the method — because this is where the study makes a genuine contribution. Rather than using fixed administrative units like postal codes or census tracts, Troost and colleagues build what they call bespoke neighbourhoods, constructed with a software tool called Equipop. For each individual in each year of life, Equipop identifies the 200 nearest households and calculates the proportion of those households that fall above or below income thresholds. The poverty threshold is equivalised disposable household income below 60 percent of the national median; the affluence threshold is 150 percent of that median. This person-centred approach standardises neighbourhood size across dense cities and sparse rural areas alike, and it captures the scale of social interaction that is actually relevant for transmitting resources and norms. The sample is one hundred forty thousand three hundred thirty-eight individuals from Statistics Netherlands' geocoded Social Statistical Database, with addresses tracked annually at one hundred by one hundred metre grid resolution. Neighbourhood exposure is then averaged across three developmental windows: early childhood covering ages zero to twelve, adolescence covering ages thirteen to seventeen, and the full childhood period from zero to seventeen. The outcome variable, years of education required to achieve the highest attained level, ranges from two to twenty-three years, with a mean of sixteen point five. The authors model this with ordinary least squares regression rather than multilevel models, a deliberate choice: because bespoke neighbourhoods overlap and more than half of individuals are the sole member of their constructed neighbourhood in a given year, hierarchical fixed-effects estimation becomes intractable. The results are unambiguous. Exposure to neighbourhood affluence over the full childhood produces a regression coefficient of two point one three eight, with a standardised beta of zero point one three three. Limiting the window to early childhood gives almost the same result: b equals two point one one nine, beta equals zero point one three two. Adolescent exposure is somewhat smaller — b equals one point seven three three, beta equals zero point one one eight — but still large and highly significant. Now compare that to poverty. Exposure to neighbourhood poverty over ages zero to seventeen produces a coefficient of negative zero point nine eight nine, with a standardised beta of only negative zero point zero four three. Early childhood poverty: b equals negative zero point eight two seven. Adolescence: b equals negative zero point nine two five. Affluence coefficients are roughly two to three times larger in magnitude than poverty coefficients across every window. The full models explain around eighteen percent of variance in attainment. That asymmetry is the central finding. Not just that affluence matters, but that it matters considerably more than poverty, and consistently so across developmental periods. Researchers and policymakers who focus exclusively on poverty concentration are looking at the smaller effect. Then there is the interaction with parental education, which adds another layer. When Troost and colleagues model neighbourhood poverty with interactions for parental education, children with at least one higher-educated parent show no detectable effect of neighbourhood poverty on their attainment. The negative effect simply disappears for them. Children of middle and lower-educated parents, by contrast, are measurably harmed by high proportions of poor neighbours, with the lowest parental education group at greatest risk when poverty proportions are low, and lower and middle groups converging in harm when poverty is high. Parental education acts as a buffer — essentially insulating children from the spatial penalty of a poor neighbourhood. The flip side is more nuanced. When affluent neighbourhood exposure is interacted with parental education, all groups benefit, but children with higher-educated parents benefit slightly less than those with lower-educated parents. The buffering is weaker for affluence than for poverty. In other words, an affluent neighbourhood can partially compensate for lower parental education, but it cannot fully replicate what a highly educated parent provides. Step back and consider what this means for spatial inequality. Children with highly educated parents are doubly insulated: protected from poverty effects by parental resources, and still gaining from affluent neighbourhoods. Children with low-educated parents are doubly exposed: harmed by poverty, and more dependent on the neighbourhood to make up the gap. Segregation, then, is not a neutral background condition; it is actively sorting children into or away from the social capital networks that shape whether university feels possible. The policy translation is direct. If affluence concentration is the stronger driver, then anti-poverty place-based programs — regenerating deprived areas while leaving wealthy enclaves untouched — are working on the smaller lever. Troost and colleagues point toward anti-segregation policy as the more productive frame. That means treating the geography of affluence as a variable policymakers can manipulate, not just a background fact about where rich people live. Two broad directions follow from the mechanisms the authors identify. First, housing policy: disrupting the spatial hoarding that keeps affluent social networks self-enclosed, through mixed-income development and reduced exclusivity in high-status neighbourhoods. Second, school access: ensuring that catchment boundaries do not simply reproduce the advantages of living in wealthy areas. When school access is strictly coterminous with affluent neighbourhoods, education policy amplifies the neighbourhood effect rather than counteracting it. The parental education buffer also has a policy implication. Because low-educated parents cannot shield their children from poverty the way high-educated parents can, place-based integration needs to be paired with family-level supports; otherwise the children most dependent on neighbourhood context gain the least from integration alone. The authors are candid about limits. The study uses household income as the sole neighbourhood resource measure, so dimensions of capital not captured by income — social networks, institutional quality, and cultural resources — remain unmeasured. There is no direct school-context variable in the models. And the Dutch welfare state, with its comparatively compressed inequality and robust public services, may not generalise straightforwardly to contexts with more extreme segregation or weaker safety nets. But the core reorientation stands. Neighbourhood disadvantage is not only about where poverty concentrates; it is equally, and in the Netherlands more powerfully, about where affluence concentrates. Policymakers designing interventions around educational equity need both lenses. Mapping deprivation without mapping the distribution of affluence produces an incomplete picture of the spatial forces sorting children into unequal outcomes. Troost, van Ham, and Manley have put the geography of affluence on the research agenda. The policy agenda should follow.

That is exactly what their Dutch study finds. Using an entire birth cohort from 1995, tracking residential histories through childhood and adolescence, and measuring educational attainment at age 23 in 2018, they show that neighbourhood affluence has a stronger effect on educational outcomes than neighbourhood poverty for every developmental window they examine.

To understand why affluence might outpace poverty as a driver, the theoretical grounding matters. The mechanisms are not symmetric. Poverty effects in the literature tend to work through resource deprivation, social isolation, disorder, and stress.

Affluence effects work through a different channel: the transmission of social and cultural capital through local networks. Troost and colleagues draw explicitly on Bourdieu's concept of habitus — the taken-for-granted sense of what is realistic for one's life course. Children in affluent neighbourhoods are surrounded by peers who treat university as the obvious next step, parents who provide tutoring and access to selective extracurricular activities, and built environments with libraries and well-resourced schools.

These are Weberian opportunity-hoarding dynamics playing out at the street level. The paper stresses that affluence is not simply the absence of poverty; the accumulation and transmission of capital in wealthy areas follows a different logic entirely, something closer to a Matthew effect where advantage compounds advantage.

Context also matters. The authors caution against importing US findings wholesale. The Netherlands has less extreme concentrated poverty, higher welfare spending, denser urban form, and lower crime rates than the American cities that dominate much of the neighbourhood effects literature.

That means the stress-based and disorder-based mechanisms that amplify poverty effects in US studies are likely attenuated in the Dutch case. Which makes it all the more striking that even in this comparatively equitable context, neighbourhood affluence still dominates.

Now, the method — because this is where the study makes a genuine contribution. Rather than using fixed administrative units like postal codes or census tracts, Troost and colleagues build what they call bespoke neighbourhoods, constructed with a software tool called Equipop. For each individual in each year of life, Equipop identifies the 200 nearest households and calculates the proportion of those households that fall above or below income thresholds.

The poverty threshold is equivalised disposable household income below 60 percent of the national median; the affluence threshold is 150 percent of that median. This person-centred approach standardises neighbourhood size across dense cities and sparse rural areas alike, and it captures the scale of social interaction that is actually relevant for transmitting resources and norms.

The sample is one hundred forty thousand three hundred thirty-eight individuals from Statistics Netherlands' geocoded Social Statistical Database, with addresses tracked annually at one hundred by one hundred metre grid resolution. Neighbourhood exposure is then averaged across three developmental windows: early childhood covering ages zero to twelve, adolescence covering ages thirteen to seventeen, and the full childhood period from zero to seventeen. The outcome variable, years of education required to achieve the highest attained level, ranges from two to twenty-three years, with a mean of sixteen point five.

The authors model this with ordinary least squares regression rather than multilevel models, a deliberate choice: because bespoke neighbourhoods overlap and more than half of individuals are the sole member of their constructed neighbourhood in a given year, hierarchical fixed-effects estimation becomes intractable.

The results are unambiguous. Exposure to neighbourhood affluence over the full childhood produces a regression coefficient of two point one three eight, with a standardised beta of zero point one three three. Limiting the window to early childhood gives almost the same result: b equals two point one one nine, beta equals zero point one three two.

Adolescent exposure is somewhat smaller — b equals one point seven three three, beta equals zero point one one eight — but still large and highly significant. Now compare that to poverty. Exposure to neighbourhood poverty over ages zero to seventeen produces a coefficient of negative zero point nine eight nine, with a standardised beta of only negative zero point zero four three.

Early childhood poverty: b equals negative zero point eight two seven. Adolescence: b equals negative zero point nine two five. Affluence coefficients are roughly two to three times larger in magnitude than poverty coefficients across every window. The full models explain around eighteen percent of variance in attainment.

That asymmetry is the central finding. Not just that affluence matters, but that it matters considerably more than poverty, and consistently so across developmental periods. Researchers and policymakers who focus exclusively on poverty concentration are looking at the smaller effect.

Then there is the interaction with parental education, which adds another layer. When Troost and colleagues model neighbourhood poverty with interactions for parental education, children with at least one higher-educated parent show no detectable effect of neighbourhood poverty on their attainment. The negative effect simply disappears for them.

Children of middle and lower-educated parents, by contrast, are measurably harmed by high proportions of poor neighbours, with the lowest parental education group at greatest risk when poverty proportions are low, and lower and middle groups converging in harm when poverty is high. Parental education acts as a buffer — essentially insulating children from the spatial penalty of a poor neighbourhood.

The flip side is more nuanced. When affluent neighbourhood exposure is interacted with parental education, all groups benefit, but children with higher-educated parents benefit slightly less than those with lower-educated parents. The buffering is weaker for affluence than for poverty.

In other words, an affluent neighbourhood can partially compensate for lower parental education, but it cannot fully replicate what a highly educated parent provides.

Step back and consider what this means for spatial inequality. Children with highly educated parents are doubly insulated: protected from poverty effects by parental resources, and still gaining from affluent neighbourhoods. Children with low-educated parents are doubly exposed: harmed by poverty, and more dependent on the neighbourhood to make up the gap.

Segregation, then, is not a neutral background condition; it is actively sorting children into or away from the social capital networks that shape whether university feels possible.

The policy translation is direct. If affluence concentration is the stronger driver, then anti-poverty place-based programs — regenerating deprived areas while leaving wealthy enclaves untouched — are working on the smaller lever. Troost and colleagues point toward anti-segregation policy as the more productive frame.

That means treating the geography of affluence as a variable policymakers can manipulate, not just a background fact about where rich people live. Two broad directions follow from the mechanisms the authors identify. First, housing policy: disrupting the spatial hoarding that keeps affluent social networks self-enclosed, through mixed-income development and reduced exclusivity in high-status neighbourhoods.

Second, school access: ensuring that catchment boundaries do not simply reproduce the advantages of living in wealthy areas. When school access is strictly coterminous with affluent neighbourhoods, education policy amplifies the neighbourhood effect rather than counteracting it.

The parental education buffer also has a policy implication. Because low-educated parents cannot shield their children from poverty the way high-educated parents can, place-based integration needs to be paired with family-level supports; otherwise the children most dependent on neighbourhood context gain the least from integration alone.

The authors are candid about limits. The study uses household income as the sole neighbourhood resource measure, so dimensions of capital not captured by income — social networks, institutional quality, and cultural resources — remain unmeasured. There is no direct school-context variable in the models.

And the Dutch welfare state, with its comparatively compressed inequality and robust public services, may not generalise straightforwardly to contexts with more extreme segregation or weaker safety nets.

But the core reorientation stands. Neighbourhood disadvantage is not only about where poverty concentrates; it is equally, and in the Netherlands more powerfully, about where affluence concentrates. Policymakers designing interventions around educational equity need both lenses.

Mapping deprivation without mapping the distribution of affluence produces an incomplete picture of the spatial forces sorting children into unequal outcomes. Troost, van Ham, and Manley have put the geography of affluence on the research agenda. The policy agenda should follow.