RD690 - Virginia Pretrial Data Project: Findings from the 2019 And 2020 Cohorts – 2023
Virginia’s Pretrial Data Project was established in 2018 under the direction of the Virginia State Crime Commission as part of the Crime Commission’s broader study of the pretrial system in the Commonwealth.(*1) The purpose of the Project was to address the significant lack of data available to answer key questions regarding the pretrial process in Virginia. The Project was an unprecedented, collaborative effort among numerous state and local agencies representing all three branches of government. The Crime Commission’s study focused on a cohort of individuals charged with a criminal offense during a one-month period (October 2017). The work was well-received by lawmakers, and the 2021 General Assembly (Special Session I) passed legislation (House Bill 2110 and Senate Bill 1391) directing the Virginia Criminal Sentencing Commission to continue this work on an annual basis. Virginia’s work in the area of pretrial data collection has begun to receive national attention.
The Sentencing Commission’s first report on Virginia’s pretrial data collection project was submitted to the General Assembly in 2022.(*2) The study focused on individuals with pretrial contact events during Calendar Year (CY) 2018. That period of time was selected in order to establish a baseline of pretrial data. Establishing a baseline allows researchers to better assess the impact of subsequent events (such as the COVID-19 pandemic) or changes in laws or policies (such as the elimination of the presumptive denial of bail from the Code of Virginia). For the current study, individuals with pretrial contact events during CY2019 and CY2020 were selected. A contact event is the point at which an individual comes into contact with the criminal justice system and he or she is charged with a criminal offense, thus beginning the pretrial process.
As with the previous study, for individuals with more than one contact event during the calendar year, only the first event was selected. While adhering to the established data collection methods, the Sentencing Commission introduced another selection criteria for the CY2019 and CY2020 cohorts. With multiple years of pretrial data now available, the Sentencing Commission was able to identify contact events in CY2019 and CY2020 that were associated with a contact event that occurred during the previous calendar year. For example, this may occur if an individual had a contact event in one year that resulted in his release during the pretrial period and, while on pretrial release, the individual was arrested for a new criminal offense during the following calendar year. The new criminal arrest during the pretrial release period is considered an outcome of the original event. For the newest study, the defendant’s first contact event in a calendar year was excluded if it was identified as a pretrial outcome for an event that occurred during the previous calendar year. The Sentencing Commission found that the excluded events accounted for only 6% of all defendants initially selected for analysis; moreover, the underlying demographic characteristics of the excluded defendants were not different from the overall cohort. While the CY2018 cohort does not have the benefit of data from previous years, the general insights about year-to-year changes in pretrial measures and outcomes are not significantly affected by the exclusion of the cases.
As with the previous study, individuals in the cohorts were tracked for a minimum of 15 months, until the disposition of the case or the end of the follow-up period, whichever occurred first. Data for the Project was obtained from numerous criminal justice agencies in Virginia. Compiling the data into a unified dataset requires numerous iterations of matching, merging and data cleaning to ensure accuracy when linking information from the respective data systems to each defendant in the cohort. More than 500 data elements were captured for each defendant, including demographics, charging details, criminal history records, pretrial release status, bond type and amount, court appearance by the defendant, new criminal arrest during the pretrial period, and final dispositions.
The Sentencing Commission’s analysis focuses on adult defendants whose contact event included a charge for a new criminal offense punishable by incarceration where a bail determination was made by a judicial officer (i.e., a magistrate or judge). Other defendants, such as those released on a summons, were not analyzed for this report. This report presents various descriptive findings for the selected defendants, their key characteristics, how they proceeded through the pretrial system, and outcomes. This report also compares a number of measures across the three years of data now available. When examining pretrial outcomes, it is important to consider what factors or combination of factors may be associated with success or failure while on pretrial release. Empirically-based risk assessment tools are commonly used to estimate the likelihood of success or failure in the community during the pretrial period in a uniform manner. For the purposes of the Project, the Public Safety Assessment (PSA), a pretrial risk assessment tool developed by Arnold Ventures, was utilized. While the PSA has been validated elsewhere, this year the Sentencing Commission examined the predictive validity of the PSA within Virginia’s pretrial population.
Virginia’s Pretrial Data Project serves as a valuable resource for policy makers, practitioners, and academics. Findings from the Commission’s ongoing analyses may be used to inform policy and practice and provide a platform for discussion of pretrial matters in the Commonwealth today and in the years to come.
Presented below are key descriptive findings from the Commission’s analysis of CY2018-CY2020 pretrial data. The findings are generally consistent from year to year; however, interesting trends have emerged. These are noted below.
• The demographic characteristics of defendants are similar across all three calendar years. Defendants are mostly male, white, between the ages of 18 and 35, and indigent (Table 1).
• Approximately 46% to 48% of defendants were charged with a felony offense, while 51% to 54% were charged with a misdemeanor or special class offense as the most serious offense in the contact event. Throughout CY2018-CY2020, the most common felony charge was a drug offense. In CY2020, assault became the most common misdemeanor charge (Table 2).
• Throughout CY2018-CY2020, the vast majority of defendants were ultimately released from custody during the pretrial period. Approximately one in ten defendants were detained throughout the pretrial period. Release rates increased slightly during the years examined, from 86.8% in CY2018 to 87.7% in CY2019 and 89.5% in CY2020 (Table 3). Release rates generally increased across all demographic groups in CY2020 (Table 4).
• Over half of the defendants each year were released on a personal recognizance or unsecured bond. The percentage of defendants released on personal or unsecured bond increased from 51.5% in CY2018 to 57.5% in CY2020 (Table 3).
• Across all three years, females were more likely to be released pretrial than males (93.6%-94.8% versus 84.3%-87.5%) and Whites were more likely to be released than Blacks (88.0%-90.4% versus 85.2%-88.1%). Non-indigent defendants were more likely to be released than defendants categorized as indigent (94.2%-94.6% versus 81.4%-85.7%) (Table 4).
• When charged with a felony or violent offense, females remained more likely than males to be released. Similarly, when charged with a felony or violent offense, Whites were released more often than Blacks. Non-indigent defendants charged with a felony or violent offense were much more likely to be released than indigent defendants charged with the same type of offense (Tables 4-1 to 4-8).
• Secured bond amounts at the time of release generally did not vary widely across sex, race, age, or indigency status, or year of release (Table 5).
• Of released defendants, between 15.6% and 16.1% each year were ordered to receive supervision by a Pretrial Services Agency (Table 6). A larger percentage of defendants placed under pretrial supervision requirements received a secured bond compared to those who were released not placed under pretrial supervision (Table 7).
• Across each year examined, a large majority of released defendants were not charged with failure to appear at court proceedings for the offense(s) in the contact event. Similarly, the majority of released defendants were not arrested during the pretrial period for an in-state offense punishable by incarceration. However, the failure-to-appear rate increased from 12.4% in CY2018 to 16.2% in CY2020, while the new-arrest rate increased from 22.4% in CY2018 to 23.5% in CY2020 (Chart 6).
• In CY2018, approximately 60% of defendants were convicted of at least one offense in the contact event (original or reduced charge). The conviction rate dropped to 52.2% in CY2020 (Table 17).
• The percentage of released defendants charged with failure to appear or who were arrested for a new in-state offense punishable by incarceration during the pretrial period increased as the defendants’ Public Safety Assessment (PSA) scores increased, suggesting that the PSA may be a useful tool in pretrial release decision making.
• PSA scores for both failure-to-appear (FTA) and new criminal arrest (NCA) were quite similar across the CY2018-CY2020 cohort groups. For both FTA and NCA measures, the largest share of defendants were classified as low risk, having a score of 1 or 2 (Tables 8 and 9).
• Each year, defendants with higher PSA scores were less likely to be released than those with lower scores. A larger percentage of defendants with higher PSA scores (5 or 6) were released during CY2020 than in previous years (Tables 11 and 12).
• In CY2020, the percentages of released defendants charged with failure to appear or who were arrested for a new in-state offense punishable by incarceration were higher than in previous years. Generally, failure rates increased the most for defendants with the highest PSA scores (Tables 15 and 16).
• Descriptive analysis alone cannot validate the predictive power of the PSA instrument. For this reason, more sophisticated analyses were conducted to examine the predictive validity of the PSA within Virginia’s pretrial population. Based on CY2018-CY2020 data, the statistical model with only the PSA score (and no other explanatory variables) yielded a moderate level of predictive power, with the standard measure of overall predictive power around 0.60.
• The Commission experimented with expanded statistical models including legal and contextual factors that are not captured by the PSA instrument. Throughout various models tested, the estimation of the PSA score remained highly significant. In general, the expanded models achieved higher predictive power, with the standard measure ranging from 0.70 to 0.72. Findings suggest that the PSA scores are highly correlated with pretrial failures but the PSA instrument does not account for all factors that have influential effects on pretrial outcomes in Virginia.