Follow-up Inspection of the United States Marshals Service's
Fugitive Apprehension Program
Report Number I-2000-02
January 2000

DATA COLLECTION

Site Visits and Interviews

We visited USMS district offices in the following locations: Madison, Wisconsin; Hammond and South Bend, Indiana; Chicago, Illinois; Manhattan and Brooklyn, New York; as well as the Superior Court in the District of Columbia. In each location, we interviewed the Chief Deputy, warrant supervisor/coordinator, investigative research analyst, and several District Marshals. In addition, we conducted telephone interviews with either the Chief Deputy, warrant supervisor/coordinator (or both) at USMS district offices in: Phoenix, Arizona; San Francisco and San Diego, California; Houston, Texas; Shreveport, Louisiana; Utica, New York; and Miami, Florida. These sites were chosen based on the districts' size, their September 1998 QPI score, and the number of warrants received in FY 1997 (the most current USMS annual report statistics available). In addition, backlog was considered if the district was one of the top 12 districts with a warrant backlog as of September 30, 1998 (See Table 3).

Table 3: Characteristics of Districts Chosen for Our Purposive Sample

DISTRICT SIZE QPI WARRANTS RECEIVED BACKLOG D.C. Superior Court Large 6.6 2256 NA Southern Texas Large 5.0 1043 801 Southern California Large 5.6 743 1408 Southern Florida Large 6.0 686 738 Arizona Large 5.7 676 321 Southern New York Large 5.5 470 781 Eastern New York Medium 7.0 288 727 Northern New York Medium 3.0 249 NA Northern California Medium 7.4 219 308 Northern Illinois Medium 6.8 164 237 Northern Indiana Small 4.1 65 NA Western Louisiana Small 9.1 62 NA Western Wisconsin Small 3.0 46 NA

Source: USMS

To systematically analyze interviewee responses, we constructed a database in Microsoft Access. Based on a structured questionnaire, we manually coded all responses and entered the data into the database. The database provided us with a tool to evaluate the responses from all districts simultaneously and compare responses by specific questions.

Data from the WIN System

The USMS provided us with WIN records on all USMS warrants closed from FY 1994 through FY 1998. We received information on a total of 82,682 warrants (FY 1994, 14,722 warrants; FY 1995, 15,200 warrants; FY 1996, 15,243 warrants; FY 1997, 18,128 warrants; FY 1998, 19,589 warrants). We excluded warrants not closed by the USMS. We received over 50 data fields for each warrant. For example, the fugitive identification number (FID), the warrant identification number, the warrant closure date, the type of warrant executed (e.g., probation violations, bond defaults), the manner that the warrant was executed (e.g., physical arrest, surrender), the fugitive's criminal history, and the QPI score.

For each FY, the USMS provided data saved as text files in three different tables (a total of 15 tables). Below is a description of each table.

The Profile Table contained information on the fugitive's criminal history such as the numbers and types of crimes for which the fugitive was previously arrested and convicted.

The Warrant Table contained the warrant identification numbers, the original offense charges, the types of warrants, the dates the warrants were received by the USMS, the dates the warrants were closed, the warrant execution codes, and the arresting districts.

The Fugitive Table contained fugitive names, FID's, the offender categories (Top 15 Most Wanted, Major, categories 1, 2, and 3), and the status of the warrant — whether it was closed or open.

We determined Microsoft Access was the most appropriate software to convert the data from text to a database. We successfully imported the data into an Microsoft Access database by linking all the tables. To link the tables, we established and applied a unique identifier, the warrant number, to every record in every table. In addition, we established methods to ensure data integrity as we imported data into Microsoft Access. Once all data was consolidated into one database, we transferred the data into SPSS and Excel to conduct statistical analysis. Transferring data from text files, Microsoft Access, SPSS and Excel requires quality assurance procedures to ensure that all 82,682 records transferred properly. Transferring data between these programs was not a seamless or automatic process due to different data protocols and conventions. Given the size of our database, we developed and implemented quality control procedures to ensure the accuracy of our data.

No detailed documentation existed describing how the QPI score was calculated. Therefore, we translated the computer programming language in WIN into a table format to assess the construction and the inclusion of different variables used to calculate the QPI score. We also entered the QPI definition of violence into our database. This definition summarized 16 criminal history fields into a binary field indicating if the fugitive associated with the warrant was violent or not.

Finally, we received data on open USMS warrants to calculate the progress of the USMS backlog reduction efforts and gathered WIN reports on open cases to complete our analysis of the performance measure to close cases quickly. We obtained direct access to the WIN system that greatly assisted in our analysis of fugitive cases and helped ensure quality control of our data.

DATA ANALYSIS

We conducted data analysis focusing on three primary areas: length of time to close a warrant, backlog reduction, and the QPI system.

Using SPSS and Excel, we assessed how the USMS warrant closure rate has changed over a five-year period and whether USMS was meeting its goal to close 80 percent of warrant cases within one year. To identify changes, we performed trend analysis based on all the key variables and created contingency tables to see what patterns emerged. In addition, we examined data before and after the USMS established the goal to determine whether the goal impacted the warrant closure rate. We graphically displayed data to determine trends and changes in the data. We checked for patterns between large, medium, and small districts on key variables. In general, we used descriptive statistics and contingency tables for our analysis.

We examined a number of relationships between variables in conducting our analysis including: overall time to close a warrant (by FY and by type of offense), number of closed warrants by type of warrant (probation, DEA, bond), number of closed warrants by type of arrest (physical arrest, detainer, dismissed), number of warrants by offender category, and number of warrants by warrant execution code (physical arrest). We also looked at differences in these variables for each FY.

We applied the USMS current backlog definition from FY 1994 through FY 1998 to assess trends in backlog reduction and to determine whether USMS was meeting its goal to reduce the backlog. Our database included only closed warrants, therefore we requested that the USMS run a report of the number of backlogged warrants at the end of each FY. The USMS has changed the definition of backlog during the years in our evaluation, so we requested the USMS to apply the current definition to each FY we examined. The current definition of backlog includes all warrants over one year old except for those warrants USMS considers unworkable. Unworkable warrants are for incarcerated fugitives who have a detainer in their file; warrants for fugitives who are out of the country and cannot be extradited to the United States; and warrants where the DEA has retained its investigative authority.

Finally, we analyzed the QPI system to determine whether it was a valid system. We examined a number of relationships between variables in conducting our analysis including the QPI score for each warrant, the QPI score by offender category, and the QPI score by the criminal history of the fugitive.

For all our analysis, we incorporated trends identified in our interview data as appropriate.