Adolescents raised in socioeconomically deprived neighbourhoods are convicted of violent crimes and develop substance-misuse problems at several times the rate of adolescents in advantaged neighbourhoods, yet it remains unresolved whether deprived neighbourhoods cause these outcomes. The source of this problem is J. Michael Oakes' commentary on a Swedish total-population study (Sariaslan et al. 2013) in which the strong neighbourhood gradient in violent criminality and substance misuse disappeared once unobserved familial factors were adjusted for. Oakes argued that the underlying identification problem remains unsolved: comparison groups across neighbourhoods are not exchangeable, family characteristics are simultaneously confounders and consequences of neighbourhood residence, real-world residential variation is rarely exogenous, within-family comparisons rest on untested sibling-exchangeability and exposure-contrast assumptions, and the strongest experimental evidence (the Moving to Opportunity housing-mobility experiment) had been inconclusive. The problem is to establish a generally accepted identification strategy for causal neighbourhood-deprivation effects on adolescent violent criminality and substance misuse — or to demonstrate that these parameters cannot be identified from observational data — and thereby settle the magnitude and direction of any causal effect. More than a decade of subsequent family-based, quasi-experimental and methodological work has left the question open, with mutually contradictory within-family findings and policy-specific relocation evidence that does not generalize.

Background

Residents of socioeconomically deprived neighbourhoods — small areas scoring high on composite indices of low education, income, employment and similar register indicators — experience adverse social and health outcomes far more often than residents of advantaged areas. Among adolescents, an observational literature reaching back decades reports large associations between neighbourhood deprivation and violent criminality (typically measured as convictions for violent crimes) and substance misuse. Whether those associations are causal is one of the central unresolved questions of social epidemiology and environmental criminology, because the people living in differently deprived neighbourhoods differ from one another in many ways — family income and structure, parental resources, and genetic and rearing environment — that also predict adolescent crime and substance use.

Causal identification has a precise meaning here. An estimate of a neighbourhood effect is causal only if the compared groups are exchangeable, that is, they would have had the same outcome rates had they lived in the same neighbourhoods. Non-random selection into neighbourhoods breaks exchangeability: families both choose and are constrained in where they live, through processes tied to the very factors that influence adolescent outcomes. A further complication is that family characteristics act simultaneously as confounders and as mediators: neighbourhoods plausibly shape family structure, parental employment and parenting, so statistically 'adjusting for family' can remove genuine causal pathways (over-adjustment) as well as confounding. This identification problem was influentially articulated for neighbourhood research by Oakes (2004), drawing on formal results showing that cross-sectional neighbourhood associations are consistent with several incompatible causal structures.

The largest single empirical contribution to the debate is a Swedish total-population study by Sariaslan et al. (2013), which followed everyone born in Sweden 1975–1989 who lived in the three largest cities at age 15 (about 303,000 individuals, including some 179,000 biological siblings). A one-standard-deviation increase in neighbourhood deprivation was associated with 57% higher odds of a violent-crime conviction between ages 15 and 20 in crude models (odds ratio 1.57); adjustment for observed confounders attenuated this to 1.09, and adjusting additionally for unobserved familial confounders — via a within-family (sibling-comparison) design, which compares siblings who were differentially exposed and thereby differences out everything siblings share — eliminated it (odds ratio 0.96, 95% CI 0.84–1.10). Substance misuse behaved similarly. The authors concluded that the gradient was 'not consistent with a causal inference' and attributed it to familial selection.

In the commentary that generated this problem, Oakes (2013) praised the Swedish register data as astonishingly rich but argued that no observational design yet identifies the desired neighbourhood-effect parameters. Beyond the exchangeability and mediation obstacles above, he questioned the within-family design itself: differentially exposed siblings must be exchangeable with respect to unmeasured causes of the outcomes, and they must experience neighbourhood contrasts that are large enough — and arise independently enough of family circumstances — to be informative, conditions he judged unlikely without exogenous shocks (such as sudden injury, death or lottery-type events) that move families for reasons unrelated to their children's upbringing. He also noted that the strongest available evidence, the Moving to Opportunity (MTO) experiment — a US randomized housing-mobility programme in which public-housing families were offered vouchers to relocate to low-poverty neighbourhoods — had not produced a clear verdict: effects varied in direction and size across sex, age at move and outcome (Kling, Ludwig and Katz 2005; Ludwig et al. 2011).

Later methodological work sharpens the difficulty rather than resolving it. Sibling-comparison estimates are more vulnerable than ordinary estimates to confounding by factors siblings do not share, and more attenuated by exposure measurement error, so within-family nulls can arise even when true neighbourhood effects exist (Frisell et al. 2012). It has also been argued that neighbourhood context is a prime case in which the within-family variation actually estimated does not correspond to the between-family variation that neighbourhood theories concern, with reverse causality and selection into discordance further complicating interpretation (Engzell and Hällsten 2025). Reviews of neighbourhood-effects research conclude that causal identification remains the field's central unsolved problem, with selection bias and theory–method mismatch recurring as the core obstacles (van Ham et al. 2012; Sharkey and Faber 2014).

Problem Statement

Does neighbourhood deprivation causally affect adolescent violent criminality and substance misuse — with what magnitude and in which direction — and can observational data support a generally accepted identification strategy for these neighbourhood-effect parameters?

As posed by Oakes (2013) in response to Sariaslan et al. (2013), any answer must confront five identification obstacles: (i) comparison groups across neighbourhoods of different deprivation are not exchangeable, because families non-randomly select into neighbourhoods; (ii) family characteristics act simultaneously as confounders and as mediators — neighbourhoods plausibly shape family structure, resources and parenting — so family adjustment may remove causal pathway effects rather than bias; (iii) real-world residential variation is rarely exogenous, since moves are triggered by income, employment, partnership and health changes that themselves affect adolescent outcomes; (iv) within-family comparisons require sibling exchangeability and meaningful within-family neighbourhood contrasts, conditions that generally fail without shocks unrelated to caregiving or selection; and (v) the available experimental evidence base (such as the Moving to Opportunity experiment) had not yielded definitive conclusions about magnitude or direction.

An identification strategy for these parameters supplies a source of variation in neighbourhood deprivation that is independent of the family processes affecting adolescent outcomes, and states explicitly the assumptions under which it identifies the causal parameters — among them exchangeability, that the compared groups would have had the same outcome rates had they lived in the same neighbourhoods, and positivity, the requirement that every deprivation level remains an attainable residence for each comparison group across the deprivation distribution — resolving each of the obstacles above and yielding the magnitude and direction of the causal effect of neighbourhood deprivation on violent criminality and substance misuse. It may also be shown that, within a formal model class restricted to the identifying assumptions that the observational designs actually in use adopt, the causal parameters of interest cannot be point-identified — that is, uniquely recovered from the joint distribution of the observable data — from observational neighbourhood data (or can only be bounded, with the bounds made explicit), so that the field can treat the observational question as settled-negative rather than merely unresolved. An impossibility claim of that kind is stated relative to a specified model class and engages the identifying assumptions those observational designs adopt.

Both alternatives serve one coherent scientific objective: credible causal knowledge of whether, how strongly and in which direction neighbourhood deprivation shapes adolescent violent criminality and substance misuse. Estimates for a single city, a single relocation policy or a single outcome, or a bias analysis of a single design, are limited in generality.

References

  1. Oakes JM (2013). Commentary: Identification, neighbourhoods and families. International Journal of Epidemiology 42(4):1067–1069. doi:10.1093/ije/dyt145. https://doi.org/10.1093/ije/dyt145
  2. Sariaslan A, Långström N, D'Onofrio B, Hallqvist J, Franck J, Lichtenstein P (2013). The impact of neighbourhood deprivation on adolescent violent criminality and substance misuse: a longitudinal, quasi-experimental study of the total Swedish population. International Journal of Epidemiology 42(4):1057–1066. doi:10.1093/ije/dyt066. https://doi.org/10.1093/ije/dyt066
  3. Oakes JM (2004). The (mis)estimation of neighborhood effects: causal inference for a practicable social epidemiology. Social Science & Medicine 58(10):1929–1952. doi:10.1016/j.socscimed.2003.08.004. https://doi.org/10.1016/j.socscimed.2003.08.004
  4. Sariaslan A, Larsson H, D'Onofrio B, Långström N, Lichtenstein P (2014). Childhood family income, adolescent violent criminality and substance misuse: quasi-experimental total population study. British Journal of Psychiatry 205(4):286–290. doi:10.1192/bjp.bp.113.136200. https://doi.org/10.1192/bjp.bp.113.136200
  5. Sariaslan A, Larsson H, D'Onofrio B, Långström N, Fazel S, Lichtenstein P (2015). Does population density and neighborhood deprivation predict schizophrenia? A nationwide Swedish family-based study of 2.4 million individuals. Schizophrenia Bulletin 41(2):494–502. doi:10.1093/schbul/sbu105. https://doi.org/10.1093/schbul/sbu105
  6. Jokela M (2015). Does neighbourhood deprivation cause poor health? Within-individual analysis of movers in a prospective cohort study. Journal of Epidemiology and Community Health 69(9):899–904. doi:10.1136/jech-2014-204513. https://doi.org/10.1136/jech-2014-204513
  7. Eriksson KH, Hjalmarsson R, Lindquist MJ, Sandberg A (2016). The importance of family background and neighborhood effects as determinants of crime. Journal of Population Economics 29(1):219–262. doi:10.1007/s00148-015-0566-8. https://doi.org/10.1007/s00148-015-0566-8
  8. Oakes JM, Andrade KN (2014). Methodologic Innovations and Advances in Social Epidemiology. Current Epidemiology Reports 1(1):38–44. doi:10.1007/s40471-014-0005-5. https://doi.org/10.1007/s40471-014-0005-5
  9. Frisell T, Öberg S, Kuja-Halkola R, Sjölander A (2012). Sibling comparison designs: bias from non-shared confounders and measurement error. Epidemiology 23(5):713–720. doi:10.1097/ede.0b013e31825fa230. https://doi.org/10.1097/ede.0b013e31825fa230
  10. Frisell T (2021). Invited Commentary: Sibling-Comparison Designs, Are They Worth the Effort? American Journal of Epidemiology 190(5):738–741. doi:10.1093/aje/kwaa183. https://doi.org/10.1093/aje/kwaa183
  11. Airaksinen J, Aaltonen M, Tarkiainen L, Martikainen P, Latvala A (2021). Associations of neighborhood disadvantage and offender concentration with criminal behavior: Between-within analysis in Finnish registry data. Journal of Criminal Justice 74:101813. doi:10.1016/j.jcrimjus.2021.101813. https://doi.org/10.1016/j.jcrimjus.2021.101813
  12. Engzell P, Hällsten M (2025). A caution on sibling comparisons in studying effects of the rearing environment. European Sociological Review 41(5):808–821. doi:10.1093/esr/jcae037. https://doi.org/10.1093/esr/jcae037
  13. Andersson A, Oskarsson S, Garcia-Argibay M, Tuvblad C, Larsson H, Latvala A, Evans B (2026). Neighborhood social structure, criminal convictions, and substance use disorders in Sweden: a population-based sibling comparison study. Social Science & Medicine 405:119596. doi:10.1016/j.socscimed.2026.119596. https://doi.org/10.1016/j.socscimed.2026.119596
  14. Kling JR, Ludwig J, Katz LF (2005). Neighborhood effects on crime for female and male youth: evidence from a randomized housing voucher experiment. Quarterly Journal of Economics 120(1):87–130. doi:10.1162/qjec.2005.120.1.87. https://doi.org/10.1162/qjec.2005.120.1.87
  15. Ludwig J, Sanbonmatsu L, Gennetian S, Adam E, Duncan GJ, Katz LF, Kessler RC, Kling JR, Lindau ST, Whitaker RC, McDade TW (2011). Neighborhoods, obesity, and diabetes — a randomized social experiment. New England Journal of Medicine 365(16):1509–1519. doi:10.1056/NEJMsa1103216. https://doi.org/10.1056/NEJMsa1103216
  16. Chetty R, Hendren N, Katz LF (2016). The Effects of Exposure to Better Neighborhoods on Children: New Evidence from the Moving to Opportunity Experiment. American Economic Review 106(4):855–902. doi:10.1257/aer.20150572. https://doi.org/10.1257/aer.20150572
  17. Chyn E (2018). Moved to Opportunity: The Long-Run Effects of Public Housing Demolition on Children. American Economic Review 108(10):3028–3056. doi:10.1257/aer.20161352. https://doi.org/10.1257/aer.20161352
  18. Bergman P, Chetty R, DeLuca S, Hendren N, Katz LF, Palmer C (2024). Creating Moves to Opportunity: Experimental Evidence on Barriers to Neighborhood Choice. American Economic Review 114(5):1281–1337. doi:10.1257/aer.20200407. https://doi.org/10.1257/aer.20200407
  19. Sharkey P, Faber JW (2014). Where, When, Why, and for Whom Do Residential Contexts Matter? Moving Away from the Dichotomous Understanding of Neighborhood Effects. Annual Review of Sociology 40:559–579. doi:10.1146/annurev-soc-071913-043350. https://doi.org/10.1146/annurev-soc-071913-043350
  20. van Ham M, Manley D, Bailey N, Simpson L, Maclennan D (2012). Neighbourhood Effects Research: New Perspectives. In: van Ham M, Manley D, Bailey N, Simpson L, Maclennan D (eds), Neighbourhood Effects Research: New Perspectives. Springer, Dordrecht, pp 1–21. doi:10.1007/978-94-007-2309-2_1. https://doi.org/10.1007/978-94-007-2309-2_1
  21. Sariaslan A, Mikkonen J, Aaltonen M, Hiilamo H, Martikainen P, Fazel S (2021). No causal associations between childhood family income and subsequent psychiatric disorders, substance misuse and violent crime arrests: a nationwide Finnish study of >650 000 individuals and their siblings. International Journal of Epidemiology 50(5):1628–1638. doi:10.1093/ije/dyab099. https://doi.org/10.1093/ije/dyab099
  22. Sariaslan A, Mikkonen J, Aaltonen M, Hiilamo H, Martikainen P, Fazel S (2022). What can siblings and cousins tell us about the role of early life family income in the aetiologies of violent crime, substance misuse and psychiatric morbidity? International Journal of Epidemiology 51(6):2028–2030. doi:10.1093/ije/dyac035. https://doi.org/10.1093/ije/dyac035
  23. White JS, Hamad R, Li X, Basu S, Ohlsson H, Sundquist J, Sundquist K (2016). Long-term effects of neighbourhood deprivation on diabetes risk: quasi-experimental evidence from a refugee dispersal policy in Sweden. The Lancet Diabetes & Endocrinology 4(6):517–524. doi:10.1016/S2213-8587(16)30009-2. https://doi.org/10.1016/S2213-8587(16)30009-2