Audit Objectives

The objectives of this audit were to:  1) review the FBI’s allocation and utilization of human resources in its investigative programs from FY 1996 through June 2002, and 2) determine the types and number of cases investigated over the same time period.

Scope and Methodology

We performed our audit in accordance with Government Auditing Standards issued by the Comptroller General of the United States and included such tests of the records and procedures that we considered necessary to accomplish the audit objectives.  The informational nature of our audit objectives did not require that we perform testing of the FBI’s compliance with laws and regulations.

We conducted work at FBI Headquarters in Washington, D.C., and at the FBI Chicago Field Office in Chicago, Illinois.  We interviewed over 50 officials from the FBI, including the Director, the Assistant Director and Chief Technology Officer, the Chief of Staff, the Special Assistants to the Director for both Strategic Planning and Resource Management and Allocation, and several Assistant Directors.  We also interviewed FBI Headquarters officials from the Finance Division and Resource Management and Allocation office, as well as numerous Section Chiefs and Unit Chiefs representing all FBI programs.  Through these interviews and review of documents and records pertaining to the audit objectives, we obtained an understanding of the control environment for strategic and program planning, requirements determination, and the resource allocation and management processes.  Further, we worked extensively with management and support personnel responsible for the Time Utilization and Recordkeeping (TURK) and Automated Case Support (ACS) systems to gain an understanding of the systems, the data contained therein, the ways in which the systems support the investigative work of the Bureau, and the management controls over the systems.

To achieve the audit objectives, we used computer-processed data maintained in FBI information systems.  To examine the FBI’s human resource utilization, we reviewed data from the TURK system, a module of the FBI’s payroll system, for the period of October 1, 1995, through June 1, 2002.  The TURK system contains work hour and Average On Board data for most FBI agents and support involved with investigative matters.  To examine the numbers and types of cases in its various criminal programs, we reviewed data from the Automated Case Support (ACS) system, which assigns unique case identifiers to all investigations and contains case-related material, for the period of October 1, 1995, through June 14, 2002.

We assessed the reliability of the computer-processed data provided by the FBI by obtaining information from FBI officials and performing various tests.  For data from the TURK and ACS systems, we performed tests to establish reliability and obtained confirmation from FBI officials as to the data’s reliability.  For both systems, we reviewed management controls and we performed data validity tests at the FBI Chicago Field Office in Chicago, Illinois.  Based on our test results and the information we obtained, we concluded the data was sufficiently reliable to achieve our audit objectives.

Data Analysis

We performed analyses of FBI resource allocation and utilization data, as well as casework data, to identify trends in the FBI’s operations from October 1995 through June 2002.  In total, this data amounted to 2,407,108 records.

Average On Board (AOB) - We received two runs of Average On Board (AOB) data,⁵⁹ for both agents and support personnel, covering the period from October 1, 1995, through June 1, 2002.  The first data run was provided in a text file, which we then exported to a database file, and contained 34,944 records broken into the following fields:

• COST CODE:  Four-digit office identifier
• EMPLOYEE INDICATOR: Agent or Support
• FISCAL YEAR:  1996 through 2002
• PROGRAM:  Numeric Indicator - See Appendix IV
• SUB-PROGRAM:  Alpha Indicator - See Appendix IV
• AVERAGE ON BOARD:  AOB for the program/sub-program designated

We provided the annual AOB figures from our analyses to FBI Headquarters program managers for verification and explanation of the data.  In doing so, we discovered that we had not been provided with data for the entire universe of FBI personnel who are captured in the TURK system.  We then received a second data run containing AOB data in the same format as the original sample for three cost codes that were not captured in the first run:  Critical Incident Response Group; Independent Counsels; and CAMPCON, a major case related to the investigation of campaign finance irregularities in the 1996 presidential election.  We received this data in a text file containing 1,644 records and imported it into a database file for testing.  We exported the data from both the original and new data runs into a spreadsheet program and calculated a new grand total.  We then used this total for our trend analyses of AOB data found in Chapters 3, 4, 5, and 6.

TURK Work Hours – We obtained work hour data from the FBI’s TURK system as a measurement of resource utilization.  We received a total of four data runs for both agent and support personnel covering the period from October 1, 1995, through June 1, 2002.  The first two runs had to be discarded due to data omissions and organization errors; the third data run appeared complete and organized in the manner we requested.  We received the data in a text file and imported it into a database file for testing.  The data consisted of 550,486 records, each containing the following fields:

• COST CODE:  Four-digit office identifier
• FISCAL YEAR:  1996 through 2002
• TURK PERIOD:  See Appendix III
• EMPLOYEE INDICATOR: Agent or Support
• PROGRAM:  Numeric Indicator - See Appendix IV
• SUB-PROGRAM:  Alpha Indicator - See Appendix IV
• MAJOR CASE:  Only if appropriate, see Appendix II for examples
• HOURS:  Hours expended in the program/sub-program designated
• LEAVE:  Hours charged to any type of leave, such as annual or sick leave
• MISCELLANEOUS:  Time expended on matters such as firearms training or automotive accidents that are not directly related to an investigative program

The data universe the FBI provided included subtotals by program and sub-program.  For us to perform the required analyses, we removed all subtotals and calculated the TURK work hours, miscellaneous, and leave amounts for each TURK period to obtain the total time TURKed for that time period.  With these totals, we were able to compute the Direct Agent Work Years (DAWYs) and Direct Support Work Years (DSWYs).  As previously explained, DAWYs and DSWYs are an estimation of work effort (not people), calculated based on a 10-hour day for a special agents and an 8-hour day for support personnel.  The formulae to calculate the DAWYs or DSWYs are as follows:

DAWYs = ((Hours / number of payperiods) * 26) / 2600

DSWYs = ((Hours / number of payperiods) * 26) / 2080

As with the AOB data, we discovered that we had not been provided TURK work hour data for all TURKing FBI personnel.  Therefore, we had to request a fourth data run, containing agent and support work hour data for the same three cost codes that were not captured in the initial AOB run.  We received the data in a text file consisting of 9,175 records in the same format as the initial data and imported it into a database file for testing.  We exported the data from both the third and fourth data runs into a spreadsheet program and combined it for a new grand total.  We used this total for our analyses of TURK work hour data found in Chapters 4 and 7.

Funded Staffing Levels – In our comparative trend analyses of agent resource utilization to allocation, we used the Bureau’s Funded Staffing Levels (FSLs) that are established by the Resource Management and Allocation (RMA) office.  We obtained two sets of FSLs for each program and fiscal year.  The first set we obtained from the RMA office represented the initial allocations set for each fiscal year.  We also obtained FSLs from TURK program officials that represented the final allocation for the fiscal year, which reflected any mid-year adjustments.  The two sets of data amounted to 11,661 records.  We used the final FSLs in our comparative trend analyses discussed in Chapter 5.

Casework - For our analyses of the FBI’s casework, we received a total of five data runs from the Automated Case Support (ACS) system containing all FBI cases worked from October 1, 1995, through June 14, 2002.  We had to discard the first four runs due to omissions and data organization errors before receiving a fifth and final one.  The information was provided in a database file containing 1,799,198 records, separated into the following fields:

• CASE ID:  Universal Case File Number
• SOURCE:  Source of information that resulted in the opening of a case
• PRELIMINARY DATE:  Date the Preliminary Inquiry (PI) began
• FULL DATE:  Date the full investigation began
• CASE OPEN DATE:  Date the case was officially opened
• CASE CLOSE DATE:  Date the case was officially closed
• PROGRAM:  Full name of FBI program
• PREV CLASS 1-30:  Previous FBI classification for any case that was classified under two or more investigative classifications during its tenure
• CLASS LEVEL:  National security classification of the case (e.g., Secret)
• DESTROYED CASE: Indicator ("D") that case was destroyed

The ACS data did not contain all of the fields we requested in our initial data request memorandum.  ACS officials advised that they were unable to provide the following three fields we had requested because ACS does not track them:  1) the source organization requesting assistance or referring a case, 2) the lead agency on a multi-agency case, and 3) a field showing the disposition of a Preliminary Inquiry (PI) or Suitability Inquiry.

In reviewing the final data run, we identified cases that were designated as belonging to FBI programs no longer in existence or in which no program was listed.  Without a current program identifier for these cases, we were required to eliminate them from our analyses.  In total, this accounted for 12,641 cases, or less than three quarters of one percent of the original universe.  As a result, our universe consisted of 1,786,557 cases on which we performed our initial breakdown of FBI cases from October 1995 through June 2002, found in Exhibit 8-1.

In addition, we discovered 59,037 cases in the database that were designated as having been destroyed (59 of these destroyed cases also had no program listed and were removed as part of those noted in the preceding paragraph).  These destroyed cases, representing 3.28 percent of the original data set, did not contain any case opening or closing dates.  ACS officials advised that they are sometimes directed to destroy cases, and if this is done it is impossible to extract any information from them except for the case number (which remains in ACS to maintain the sequential numbering system).  Without case opening or closing dates, we were unable to include these cases in our subsequent analyses, which were dependent upon these dates.  This reduction of 58,978 cases (the 59,037 destroyed cases, less the 59 cases already captured as having no program and having already been excluded) left a sample of 1,727,579 cases, on which we performed the remainder of our testing in Chapter 8.

As previously discussed, we discovered a significant number of cases that were closed in FY 1999 (761,406, compared to 106,958 in FY 1998).  Upon our inquiry, ACS officials explained that in April 1999, as part of its preparation for Y2K on January 1, 2000, the FBI closed a large number of Headquarters cases that had been unaddressed for long periods of time to reduce the amount of data that would go through the Y2K conversion process.  Although there was no way for us to determine exactly how many cases were closed for this reason, because some investigations were closed at the same time for routine reasons, we approximated the number by identifying the number of Headquarters cases that were opened prior to FY 1996 and were closed in April 1999.  This totaled 675,056 cases and reflected approximately 38 percent of the original data set.

59. The definition of AOB can be found on page 14.