The federal prosecutors database has two methods of reporting dates:

• Reporting dates:
  The first method of recording dates -- referred to as "by reporting date" -- is the most common one used by government agencies. Counts reflect the month in which the event was first recorded in the tracking database. It's chief advantage is that it allows reporting of the most up-to-date information available in a consistent manner. This is the method used by TRAC in its published annual and monthly series.

  This method, however, is not without its limitations. Where there is a delay in recording the information in the database, the reporting date will be after the date the event occurred. This means that reporting date counts for a given month or year may include events that took place in a prior month or a prior year. To the extent that late reporting is the exception, and the not the rule, the reporting date method of counting produces counts for recent periods that more accurately reflect the true level of activity since late reports in the current period are generally offset by the inclusion of late reports from prior periods. The reporting date method, however, will be adversely affected if the recording system is disrupted such as during upgrades in the hardware or changes in the software system being used. For example, during fiscal year 2000, Y2K problems were encountered by the government which interfered with the timely processing of these files. For January-March 2000, reporting dates recorded are virtually all in March. A problem occurred again during July-September 2000. Figures for these six months are estimated, based upon the average within each of these separate three-month periods. Annual counts should not have been affected. One also sees unusual spikes in September 1999 and in April 2000 followed by unusually low counts the following month. Since these spikes are not apparent in frequencies using event dates, it appears that something unusual in DOJ's data or reporting series also affected counts for these months.

• Event dates:
  The second method of recording dates -- referred to as "by event date" -- reflects the actual date of occurrence of an event. The dates shown in TRAC case listings and in the detail page on cases give you the "event date." If you could rely on prompt reporting of events, this method of counting would be the preferred method since it would present the most reliable picture of what is actually occurring. But where reporting is not an automatic computerized process, and is instead based upon humans recording the information, some delay in recording often occurs. Typically, this makes counts based upon event dates preliminary in nature, since as late reports are added, figures change to reflect this additional information. In the short run, trend lines are biased downward since not all activity that occurred has been recorded yet. Views based upon event dates can therefore give quite a misleading picture of the actual trends.

  Event dates can also prove misleading in a further way. Event dates for individual suspects or defendants can differ from those of the larger "case" or "investigation" of which they are a part. This occurs if defendants are added to the investigation or court prosecution after the date the referral is received, or the case is filed in court. The "event dates" shown in TRAC case listings for when the referral is received and when the case is filed in court, reflect the dates for the case, rather than the specific defendant. The detail page for a particular defendant will include, in addition to these case dates, the dates on which events happened to that particular defendant in the event activity section. In addition to being reflected on individual defendent "detail page(s)," TRAC Analyzer allows the user to choose whether to use reporting or event dates in user-requested analyses. In contrast, TRAC's Express and Going Deeper tools use reporting date when tabulating any results across time.

• Cautionary note when counts are small:
  Monthly volume of activity can become quite small, especially when figures for a particular district or subcategory of referrals are examined. Considerable month-to-month variation in counts should be expected as a normal course of events. Further, one should be cautious in making comparisons involving percentages, per capita rates, and ranks. Small differences or changes in counts can produce large differences or changes in percentages, rates, and ranks.