About the Data
Federal Prosecutor Database
TRAC maintains a series of databases which track activities of federal prosecutors. Coverage extends from 1974 onward for each U.S. Attorney office. The data are transactional in nature. Both civil and criminal activities are tracked. Information is included about each referral received by federal prosecutors, where it comes from, the suspects/potential defendants or parties included, the criminal charges or civil cause of action involved and how the referral is ultimately disposed of.
Tracking of criminal matters starts at the point of referral from a federal, state, or local investigatory agency recommending federal criminal prosecution and follows the referral through to ultimate disposition. Referrals which federal prosecutors decline to prosecute, as well as referrals which are prosecuted, are covered. Civil matters tracked can involve referrals prior to any court filing, or be referred after court litigation has commenced (for example, when a federal agency or official has been sued).
The main source for this series is the case management system maintained by the Executive Office for United States Attorneys (EOUSA) of the U.S. Justice Department. EOUSA tracks the activities of federal prosecutors in each of the U.S. Attorney offices. Following centralized standards and using prescribed database systems, each U.S. Attorney's office is required to maintain detailed computerized records on its workload (see United States Attorneys' Manual Chapter 7, 3-7.110 (1992) and 3-16.000 (1998)). Ensuring accuracy of this information is an explicit responsibility of each U.S. Attorney. As part of the procedures implementing this responsibility, the United States Attorneys, or alternatively each Division Chief with the district, must "personally certify the accuracy of their data as of December 31 and June 30 of each year." [U.S. Department of Justice, EOUSA Resource Manual, 135 Continuous Case Management Data Quality Improvement Plan 10/03/97]
Each month, each U.S. Attorney office furnishes extracts from each office's computerized record system on its activities to the national office where they are combined into a national database, various validity checks are run, and internal management reports are generated summarizing each office's activities. The summary reports are furnished to each U.S. Attorney. Discrepancies uncovered during these edit checks are reported to the field and corrections are requested. At the end of each fiscal year, a final set of perfected tape files are generated. These are then used to generate statistics summarizing the activities of federal prosecutors for both internal and external audiences. This central source is used for responding to requests for statistical information from Congress, OMB, as well as from the Attorney General.
Since 1974, a number of significant transformations have taken place both in computer technology and in agency computer and database systems. These have affected how the Justice Department tracks the activities of each U.S. Attorney's office. Prior to 1998, district offices used different case management systems which varied in how matters were tracked and on which specific items of information were maintained. Over time, changes have also occurred in the information recorded by each office. Even when the same field has been present throughout the period, the specific codes used to classify information have often evolved. For example, on the criminal files, information on sentences was not centrally collected prior to fiscal year 1992. While the "criminal lead charge" has been recorded throughout the period, the use of categories for Justice Department criminal program areas and civil causes of action only began to be recorded in the early eighties. These program category and cause of action codes have also undergone some evolution over time as new categories were defined and old ones discontinued. To what extent multiple criminal charges are separately tracked, and how they are associated with separate counts, has also evolved during this period.
Further, not every information item in the centrally maintained databases are "required" fields; some are optional. In general, dates of referral, prosecution or court filing, and disposition along with referral agency, criminal lead charge, criminal program category or civil cause of action, disposition type, reason or outcome are required and with rare exception present in the files. Because district database systems varied prior to 1998, some optional fields were not even present in all district information systems.
In tracking the criminal workload, a separate record is generally maintained for each defendant so that information specific to that defendant concerning processing and outcome is recorded. The centralized files exclude petty offenses. Since 1980, offices are required to maintain information on criminal referrals received which were "immediately declined." Prior to 1980, this was not required. Only defendants involved in "matters" and "cases" were tracked. (See discussion of these specialized terms in Understanding the Terminology which Agencies Use.) Criminal immediate declination files include fewer fields of information. Also up until 1998, a separate record was not maintained for each potential defendant or suspect in an immediate declination, although, the number of suspects involved was recorded.
In tracking the civil workload, generally a separate record is opened only for each case rather than party. Thus, unlike the situation with criminal which summarizes these data on a per-defendant basis, statistics for civil enforcement are calculated on a per-case basis. It is also true that the civil referral-by-referral data sometimes contain additional records when there are multiple parties. When two or more participants are involved in the same civil action and the relief requested differs for the participants, a separate record may be opened for each party. And in lands civil actions, a separate file may be opened for each tract. About 25% of the cases track information on multiple parties, with 5% of the cases tracking information on three or more parties.
Because of the natural delay between the occurrence of an event and its recording in the agency's database, two parallel sets of date fields are kept. The first is the actual date -- such as for when the referral is received, the case is filed, or the disposition occurs ("event" date). The second is the recording date -- when the event was added to the computer system ("reporting" date). The Justice Department uses the date the event was added to the computer system to classify whether the event occurred during a given fiscal year. This may result in events which took place prior to the fiscal year in question, being listed in the current year's file. Users can access monthly data on TRACFED by either the event or reporting date.
While these federal prosecutorial databases have some limitations and specific problem areas, on the whole TRAC's analyses have generally found the criminal data to be of relatively high quality as compared with other government administrative data sources. For example, the counts produced on a national level for federal prosecutions and convictions for tax fraud and drug offenses generally mirror the counts independently published by the Administrative Office of the United States Courts. (See federal tax prosecutions and federal drug prosecutions). Much less is known about the quality of the civil data.
TRAC has noted several areas where some caution in using these data is warranted:
No large database tracking system, however, is ever perfect. The challenges are enormous. According to the 1996 annual report of the Executive Office for United States Attorneys an interagency Data Reconciliation Working Group from the Department of Justice, the Administrative Office of the United States Courts, and the United States Sentencing Commission was formed to work for greater consistency in the way Federal criminal agencies capture and report information. The initial findings of this group in identifying data differences was issued "Comparing Case Processing Statistics" in August 1996 with a followup report in April 1998.
- Criminal Immediate Declinations. The first area of concern is in the recording of immediate declinations versus matters. Problems have been observed in the reliability of information recorded, and in the consistency of recording practices over time as well as between district offices. Special caution should therefore be exercised in relying upon statistical series which include these reports. (Note: because civil immediate declinations were not recorded at all until very recently they are currently not included in the database.)
First, there is inherent subjectivity in what gets recorded as a referral, particularly when it is immediately declined. Further, U.S. Attorney accounting practices appear to vary -- both among district offices and over time -- regarding which referrals are recorded and, when recorded, whether they are classified as an immediate declination or a matter. Finally, because district offices maintain these records as a "case management" tool, less attention may be given to the recording of these receipts (i.e., immediate declinations) since -- by definition -- these referrals never enter into their workload. Indeed, prior to 1980 they were not tracked at all.
For whatever reason, the number of immediate declinations recorded by the U.S. Attorneys has declined sharply since 1980. Increasingly, the referrals that are recorded at all are being treated as matters, even when they are subsequently declined. Some caution is therefore warranted when comparing referral counts. Because of these concerns, TRAC has not combined immediate declinations with later declinations in its counts prior to 1986. Users should also note that there was an unusual decrease (-18%) nationally in the number of recorded immediate declinations between fiscal 1995 and fiscal 1996. The number of matters recorded as received -- as distinct from immediate declinations -- also declined but to a lesser extent (-3%) between fiscal 1995 and fiscal 1996. Finally, because of major changes in recording systems between 1991 and 1992 and 1997 and 1998 conclusions based upon comparisons across these time periods also should be carefully scrutinized.
- Magistrate Court Cases. A second area of concern is in the accuracy of recording of information on criminal defendants prosecuted in U.S. magistrate courts, and in general in the recording of the specific "court type." The database files maintained centrally treat magistrate prosecutions in a fundamentally different way than prosecutions before federal district judges.
TRAC has noted apparent inconsistencies in the coverage and coding of information for these records. For example, the EOUSA generally counts as a magistrate court prosecution any matter with a magistrate court entry in the "court type" field. Yet sometimes the disposition field indicates that the matter was declined, not prosecuted. Some offices may record preliminary appearances in magistrate court as if they were the filing of a prosecution when they are not and central systems may not always succeed in screening out these preliminary proceedings from their magistrate case prosecution counts. Because fewer information items are recorded on magistrate cases, it is often difficult to reconcile these inconsistencies. At other times, the record indicates that the matter was not taken to court, but a guilty disposition is shown along with specific sentence information. Under these circumstances, EOUSA treats the referral as a district court prosecution whether it is or not.
These internal inconsistencies raise questions as to how reliably and accurately magistrate matters are being tracked. Magistrate records make up only a relatively small proportion (around 10%) of total matters. However, these problem areas could be significant for particular categories of offenses or offenders, or for a particular district in a given year. Changes incorporated into the new 1998 data systems were designed to correct many of these deficiencies. Thus, it is hoped that current data should become more reliable.
- Incomplete Tracking between Years and Offices. A third area of concern is the incomplete tracking of referrals between years and offices.
Disappearing cases. The Justice Department's central system maintains a separate file for each year. Cases not yet disposed of at the end of the year (ending inventory) should by definition be part of the beginning inventory for the next year's file. This is not always the case. By the Justice Department's own counts, in the aggregate about 5 percent of criminal prosecutions in district court which are pending at the end of the year simply disappear and do not show up as part of the beginning inventory at the start of the following year -- thus, their disposition is apparently never recorded. (The problem is larger for criminal matters in magistrate courts -- where in aggregate about 15 percent of these court cases pending at the end of one fiscal year disappear from the beginning inventory for the following year.) A partial explanation given by EOUSA staff for these discrepancies is that to provide an incentive for prompt reporting of dispositions by districts, when a district neglects to report a disposition for several years the referral is deleted from the database without formally crediting the disposition to the district. This leads to some undercount in dispositions as well as potential bias when assessing their composition since late reporting may be more common with certain kinds of dispositions than others.
Transfers between districts. There is a further problem in tracking cases which transfer between districts. Separate tracking numbers are maintained in each district. When a new referral comes in, even when it is a transfer from a different district, a new matter or case file is opened. This can lead to some double-counting nationally since the matter or case was already separately counted in the originating district. Transfers also present an issue of how they should be most appropriately treated in the originating districts statistics. Should a transfer out at the referral stage be treated the same as a any decision not to prosecute, or at the court stage the same as any other type of "dismissal?"
In theory, transfers in and out of districts should balance and this double counting could be eliminated. This would be possible only if matters closed by "transfer out" from districts equaled the number opened by "transfer in" to receiving districts. But these counts don't match. According to EOUSA staff, often cases recorded as closed through transfer never shows up in receiving districts' workloads. Effectively, a "transfer out" often is a final disposition of the matter. In part for this reason, long standing EOUSA practices have not distinguished transfers from other types of referral dispositions in their counts. Transfers, however, are relatively modest in number usually accounting for less than 3 percent of criminal dispositions in any year. However, transfers could be significant for particular categories of offenses or offenders, or for a particular district in a given year.
- Recording of Criminal Fines. A fourth area of concern is the recording of information on fine amounts. Recorded amounts appear on average lower than would be expected. Historically, the U.S. Attorney Offices have been responsible for tracking unpaid fines. While amounts entered should reflect the total fine rather than the amount which has not been paid at the time of case closure, this may not always be the case. TRAC has been unable to obtain any clear definition from the Executive Office for U.S. Attorneys of what explains apparent recording discrepancies.
- Recording of Civil Outcomes. A fifth area of concern is the recording of information about civil outcomes. The diversity of remedies sought in civil cases makes recording of outcome particularly problematic. Any system which tries to reduce this diversity into a simple coding system will loose important detail. Also because of this complexity there is more ambiguity in the definition of these codes. (Which party prevailed, for example, where most but not all remedies sought were granted, or where monetary claims were made and the sum granted was sizable but not the entire amount sought?) Thus, inherently there is considerable subjectively in the recording of outcome.
In this database there is a further problem of missing information about outcome. While relief requested and granted are required fields, information is still frequently missing (particularly with the relief granted field and for the supplemental records on additional parties). For all of these reasons, a user should exercise caution when interpreting outcome measures from the civil datdabase.
- Data System Transition Periods. Finally, reliability and validity problems often accompany periods when large scale data system changes take place. During the nineties, two major conversions occurred in federal prosecutor databases. Not only did recording practices changed at these times, but as is typical in any large-scale data base conversion, special transition problems were encountered which affect the data quality. Special caution is therefore advised in making comparisons spanning these particular time periods.
1991-1992 Period: As part of the preparations leading up to the data conversion, during 1991 the Executive Office for United States Attorneys initiated "Operation Garbage Out." This process required a review to ensure correct and current information [Statistical Report, USAO Fiscal Year 1991, p. 7]. During this process instances came to light where the status of a referral had not been properly updated in the database. Some of these changes were then counted as occurring in fiscal year 1991 when they had in fact taken place in previous years. Figures for 1991 therefore may be unusually high because of this database cleanup effort.
1997-1998 Period: A second transition in district data systems occurred during 1997 and 1998. For the first time, all districts adopted a single uniform database system. Significant changes also occurred in what information was being tracked and how it was recorded. Beginning in fiscal year 1998, for example, information on civil debt collection cases generally is no longer included in the civil caseload data. (The civil debt collections matters which are no longer included "are claims of a normal and routine nature, such as Department of Education student loan defaults, Veterans Administration benefit overpayments, etc. in which litigation normally is not contested and usually results in a judgment being entered.") Major difficulties were also encountered during the conversion. In addition, problems were encountered in consistently mapping these new data structures used in the field to the unchanged file systems maintained centrally by the EOUSA in Washington.
Finally, the Justice Department has funded record-by-record matching of the criminal administrative data bases from federal prosecutors with those maintained by the federal courts and other entities. These comparisons undertaken by independent outside scholars under contract with the government have resulted in a number of publications from the Justice Department's Bureau of Justice Statistics. See, for example, "Compendium of Federal Justice Statistics for 1996" (1/99 - NCJ 172849) and earlier reports in this series. Because these researchers worked under contract, they had access to all fields of information, including those providing the names and other identifying details on each defendant. This allowed them to do record-by-record matchups between the two systems. More than one group of outside researchers have been involved in this research program over the years. No fundamental problems with the integrity of data in either data system have been reported.
Transactional Records Access Clearinghouse, Syracuse University
Copyright 1997, 1998, 1999, 2000, 2001, 2002