Article Text

Download PDFPDF
129 Multiscale analyses of violent crime and neighbourhood alcohol environments
  1. Christina Mair1,
  2. Andrew Gaidus2,
  3. Paul J Gruenewald2,
  4. William R Ponicki2,
  5. Juliet P Lee2
  1. 1US University of Pittsburgh Graduate School of Public Health, Department of Behavioural and Community Health Sciences
  2. 2US Prevention Research Centre, Pacific Institute for Research and Evaluation


Statement of Purpose Problems of unit scale and resolution plague population analyses of violence and crime related to alcohol outlets: Crime related to outlets may be located within a block (e.g., assault and battery), residential areas some distance away (e.g., child physical abuse), or diffusely spread across communities (e.g., drunken driving). When a social process is localised, highly resolved spatial analyses can identify local effects; when non-local, highly resolved analyses may miss the mark, identifying no local effects. We (1) conducted multiscale spatial analyses of crime to assess variation in spatial effects and (2) determined whether off-premise outlets, in particular, are associated with crimes in local or adjacent areas.

Methods/Approach We examined relationships of violent crime incidents for Oakland, CA in 2015 with outlet locations and demographic data measured at Census block, block group and tract geographies. The impacts of aggregation bias were estimated at multiple spatial scales using the Moran’s I test for global autocorrelation. Negative Binomial models assessed relationships between predictors and outcomes across nested and adjacent (spatially lagged) units.

Results Spatial autocorrelation for violent crimes declined from tract to block scales (e.g., assaults, 0.61’”0.29), while effect sizes related to outlets increased (e.g.,+2% at the block group vs. +58% at the block level for assaults). Multiscale analyses revealed that outlet and population effects were highly localised with adjacent (spatial lagged) effects observed at the block level. Critically, effects related to block group measures (e.g., low median household income and poverty) were expressed in identifiable local blocks.

Conclusions Interpretations of crime effects related to neighbourhood conditions are conditional upon unit scale and resolution. Outlet effects are localised but exhibit important non-local (spatial lag) effects.

Significance/Contribution to Injury and Violence Prevention Science Multiscale processes and patterns characterise links between alcohol environments and violent crime.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.