TRAC'ing more complete NRC data

The Transactional Records Access Clearinghouse (TRAC) obtained and transformed the Nuclear Regulatory Commission's massive computerized database into an organized knowledge base to reveal the agency's actual enforcement activities, including inspection and violation records for commercial nuclear power plants from 1975 to 1989. The process required extensive data cleaning and verification to address inconsistencies and missing information in the NRC's original files.

In recent years, federal agencies such as the Environmental Protection Agency and the Census Bureau have been making computerized databases public. While these "hand out" databases can provide reporters important, useful information, most of the data files available through federal agencies focus on external situations, and not on the internal workings of agencies. The central mission of the Transactional Records Access Clearinghouse is to obtain computerized records of federal enforcement agencies and transform them into databases that reveal what these agencies actually are doing.

The TRAC Nuclear Regulatory Commission database, for example, covers the regulation of nuclear power reactors from 1975 to August of 1989. (TRAC currently is updating this database.) The database is available in various formats: a six-volume report series, diskettes suitable for use on either PCs or MacIntosh computers and nine-track computer tape. TRAC's first step in gathering the NRC data was to obtain a list of the dozens of record systems then used by the NRC.

We decided initially to focus on what the NRC called its "766" file. TRAC requested this file under the Freedom of Information Act. The massive file, with millions of different entries, was reproduced on a tape that we provided for the agency. Obtaining the raw 766 file was only the beginning of the long process of developing what we call a "knowledge base," an organized collection of information that allows the user to examine the workings of an agency.

Although the 766 file contains details of NRC inspections undertaken at both commercial power plants and other licensed facilities such as universities and hospitals, TRAC decided to focus its attention on the 130 commercial plants which came under NRC regulation from its formation in 1975 through August of 1989. There were, however, a lot of problems. The NRC only provided the sketchiest information on the format and file contents of the tape, and most of the specific codes were not defined.

In cases where codes were defined, they didn't always match the information in the file. The file contained detailed information such as where and when each inspection was conducted, how long it required to complete, who conducted it and whether the inspection resulted in the operator being cited for any violations. From 1975-89 for power plants only, the data indicated that the NRC conducted close to 50,000 inspections detected about 27,000 violations.

The NRC's records are surprisingly inconsistent, sometimes showing that a single reactor became operational on several different dates. An accurate list of the name of each commercial nuclear power plant, the utility that owned it and the exact date it was granted its pre-construction license, its construction license, when it became operational and when it was permanently closed down had to be compiled by TRAC to make the 766 file useful.

The inconsistencies in the data required double checking the information with the NRC and, in some cases, with the utilities themselves. The purpose of this intense effort was to provide more than just NRC information about individual reactors. The TRAC data also presents the information in terms of the periods that each reactor goes through during its total life. With TRAC data, it is possible to determine whether an individual reactor had more or fewer violations during the construction period than it did during its operational period.

The basic accuracy of the NRC data, however, was not the only problem. NRC had changed the way it categorized the seriousness of violations several times - starting with a three-category system, moving to a six-category system in the early eighties, and then back to a five-category system. Spotting these changes, not documented anywhere in the raw 766 file, was critical. We also sought to locate information about what the NRC did after the inspector detected a violation, the place where the 766 file left off.

When were penalties assessed, how big were they, against whom were they applied? No one we talked to at the NRC knew of any computerized information on penalties, so TRAC reluctantly began building its own computerized files drawn from internal paper records. Very late in this process, TRAC discovered that the NRC did in fact have such a file - the Enforcement Action Tracking System (EATS) which was not widely known within the NRC.

We obtained EATS under the FOIA. But EATS used a different identification system than the 766 file. The file that TRAC developed from the paper records supplied cross-linking "IDs" so that violations and penalties could be matched. Comparison of these two independent sources of information on penalties allowed further checks of NRC data accuracy. Knowing the precise number of violations at a particular plant can be interesting.

But developing information so that a user can compare the performance of one utility against that of another or determine whether NRC inspectors around the country are enforcing regulations in a uniform way transforms a normal "database" into a "knowledge base," a tool that can provide genuine insights into both the workings of an important federal agency and the industry it regulates.