- The paper introduces a novel two-step regression estimator that adjusts arrest likelihoods by integrating crime reporting rates.
- It utilizes NCVS and NIBRS data to empirically show that accounting for unreported crimes lowers estimated arrest probabilities, especially for violent offenses.
- The study highlights how unreported crimes distort arrest data, urging a reexamination of racial disparities and policy reforms in criminal justice.
Estimating the Likelihood of Arrest from Police Records in the Presence of Unreported Crimes
The paper "Estimating the likelihood of arrest from police records in presence of unreported crimes" addresses a critical issue in criminology: the estimation of arrest likelihood for crimes, given the anticipated dark figure of unreported offenses. Traditional analysis of crime data often overlooks the discrepancy between crimes committed and those reported to law enforcement, potentially skewing arrest likelihood calculations that rely merely on officially recorded data.
Methodological Framework
The authors present a robust statistical framework to estimate the arrest likelihood using police data while accounting for crimes that remain unreported. This approach employs a parametric regression-based two-step estimator. The first step involves estimating crime reporting likelihood using logistic regression models and data from victimization surveys like the National Crime Victimization Survey (NCVS). In the second step, information from the National Incident Based Reporting System (NIBRS) is employed to estimate arrest likelihoods, adjusted using the crime reporting rates derived in the first step.
Empirical Analysis
The empirical focus of the paper lies in exploring racial disparities in arrest rates for violent crimes, such as sex offenses, robbery, and assault, over the period from 2006 to 2015 with data from NIBRS and crime reporting estimates from NCVS surveys spanning 2003 to 2020. Findings reveal a significant decrease in estimated arrest likelihood after adjustments for unreported crimes are accounted for. Notably, it is observed that offenses involving white offenders result in arrests at slightly higher rates than those involving black offenders, particularly after adjusting for crime specifics and unreported incidents.
Implications and Future Directions
The paper's methodology contributes to a more nuanced understanding of racial disparities in arrests, highlighting the necessity for accounting for the non-reporting bias in crime data analysis. By proposing a statistical framework that can withstand the complexities of non-random missing data (such as through capture-recapture methods and regression with inverse probability weighting), this study bridges the gap between officially reported figures and actual crime occurrences.
The results have both theoretical and practical implications, suggesting the need for revisiting previous criminological analyses and policies relying heavily on official crime data. This could lead to better-informed decisions in law enforcement and criminal justice systems aiming for equity.
Future work could enhance this framework by integrating nonparametric models, such as those handling mixed data types, to improve flexibility and application scope. Furthermore, the extension of this methodology could encompass additional socio-demographic variables, which might further elucidate the complex interactions influencing arrest outcomes. This might involve incorporating real-time or more granular data that could capture geospatial crime pattern variations more effectively.
Ultimately, this research underscores the critical importance of addressing data issues in criminal justice research and develops a promising pathway for improving arrest likelihood estimates amidst pervasive data limitations.