Propensity Score Matching in Criminology and Criminal Justice

Apel, Robert; Sweeten, Gary

doi:10.1007/978-0-387-77650-7_26

Cited by 202 publications

(186 citation statements)

References 64 publications

Supporting

Mentioning

175

Contrasting

Unclassified

Order By: Relevance

“…Propensity score matching is an important tool, but by no means serves as a solution to all selection bias problems associated with quasi-experimental research designs (Apel and Sweeten, 2010). While the utility of propensity score matching was advanced in the early 1980s (Rosenbaum and Rubin, 1983) there remains much debate and ambiguity about the best practices to equate pretreatment covariates among treatment and comparison groups (Apel and Sweeten, 2010;Stuart, 2010).…”

Section: Propensity Score Specificationmentioning

confidence: 99%

“…While the utility of propensity score matching was advanced in the early 1980s (Rosenbaum and Rubin, 1983) there remains much debate and ambiguity about the best practices to equate pretreatment covariates among treatment and comparison groups (Apel and Sweeten, 2010;Stuart, 2010). Propensity score matching is the simplest and most widely used strategy (King and Nelson, 2016) and is most appropriate for research where there are high levels of observable imbalance exist between treatment statuses (King and Nielsen, 2016;King et al, 2011), where many pretreatment covariates are used to estimate the initial scalar score (equal to or more than 20) (Gu and Rsoenbaum, 1993), and when large sample sizes remain after the matching process has been employed (King and Nielsen, 2016;King et al, 2011).…”

Section: Propensity Score Specificationmentioning

confidence: 99%

“…Most randomized controlled trials in the field of criminology and criminology follow this design; few are able to achieve the ideal fully blocked randomized controlled trial where zero imbalances can be assumed by design (Ariel and Farrington, 2010). At worst, propensity score matching adds little more to the estimation of a treatment effect than a propensity adjusted regression model (Apel and Sweeten, 2010). The determination of where any one propensity score matching strategy falls on the continuum of methodological rigor is contingent upon a series of diagnostic checks.…”

mentioning

confidence: 99%

“…The main issue here is the proverbial question of whether apples to apples or apples to oranges comparisons are being made. Different preferences are used to accomplish the first step in the matching process, ranging from the use of theoretically informed logistic regression model specifications to a "kitchen sink" model of any and all covariates that are available to a researcher (Apel and Sweeten, 2010). Our scores were generated under the auspices of the latter perspective.…”

mentioning

confidence: 99%

“…covariates were used, which is not necessarily a large quantity of potential confounders but is also not a small quantity (see Apel and Sweeten, 2010).…”

mentioning

confidence: 99%

See 4 more Smart Citations

Examining the Correlates of Sex Offender Residence Restriction Violation Rates

et al. 2016

View full text Add to dashboard Cite

Objectives: This research examines the contribution of social ecological factors to the variation in sex offender residence restriction (SORR) violation rates, operationalized as a sex offender residing within a buffer zone around a school or a day care.Methods: Drawing on data from two Midwestern states, we utilize a quasi-experimental cohortcontrol group design to examine the correlates of county-level SORR violation rates among a cohort of post-SORR sex offender parolees, and three counterfactual cohorts (pre-SORR sex offenders, pre-and post-SORR non-sex offender parolees). We model the violation rate using a series of fractional logit regressions, examining the contribution of housing market, environmental justice, and system resource variables.Results: We observe that county-level variation in post-SORR sex offender violation rates is directly associated with concentrated disadvantage and the density of residence restrictions. The direct effect of concentrated disadvantage was unique to the post-SORR sex offender cohort. Model predictions suggest that the relationship between SORR density and concentrated disadvantage varies across the study states. Conclusions:The results suggest that factors found to be associated with sex offender clustering (i.e., housing market characteristics) are not associated with SORR violation rates. Instead, this research suggests a model which allows for the simultaneous influence of environmental justice and system resource effects. Future research on the mechanisms underlying these effects is warranted. Keywords: Sex Offenders; Residence Restrictions; Social Ecology; Proportional Response Models 2 EXAMINING THE CORRELATES OF SEX OFFENDER RESIDENCE RESTRICTION VIOLATION RATESIn the past 20 years, sex offender residence restrictions (SORR) have been widely deployed as a means to manage the recidivism risk of sex offenders in the community (Levenson and Cotter, 2005;Meloy, Miller, and Curtis, 2008). Residence restrictions aim to impede opportunities for reoffending among known, registered sex offenders by regulating their residential proximity to areas where potential child victims congregate. This is accomplished by enacting buffer zones of varying sizes (e.g., 500 to 2,500 feet) around varying protected locations (e.g., schools, day care centers, parks, and bus stops), within which registered sex offenders are barred from holding an official residence (Meloy et al., 2008;Pacheco and Barnes, 2013;Socia, 2011). Implicitly drawing on the distance decay hypothesis (Brantingham and Brantingham, 1984), SORR are built on the assumption that sex offenders choose to locate their residences close to potential victim pools for offending purposes (Walker, Golden, and Van Houten, 2001).However, despite wide political and popular support (Levenson, Brannon, Fortney, and Baker, 2007;Pickett, Mancini and Mears, 2013) recent research suggests that SORR are largely unsuccessful in reducing recidivism. Studies indicate that residential proximity to areas where children congregate explain...

show abstract

Section: Propensity Score Specificationmentioning

confidence: 99%

Section: Propensity Score Specificationmentioning

confidence: 99%

mentioning

confidence: 99%

mentioning

confidence: 99%

“…covariates were used, which is not necessarily a large quantity of potential confounders but is also not a small quantity (see Apel and Sweeten, 2010).…”

mentioning

confidence: 99%

See 3 more Smart Citations