National Transit Database (NTD) and Safety Management Information System (SAMIS)

Measures

Transit Safety: Transit Fatalities and Injuries per 100 million passenger miles traveled (bus, transit rail, commuter rail, paratransit, vans, & other; fatalities and injuries per 100 million passenger miles)

Transit Service Supplied: Vehicle Revenue Hours of service supplied (bus, transit rail, commuter rail, paratransit, vans, & other, in millions)

Transit Ridership: Billions of Passenger Miles traveled (bus, transit rail, commuter rail, paratransit, vans, & other, in billions)

Definitions

Transit fatality and injury data are recorded for both transit riders and employees. These data are reported annually by operators to the National Transit Database (NTD). NTD data are an input to FTA’s Safety Management Information System (SAMIS). A transit fatality is a death within 30 days after the incident, which occurs under the collision, derailment, personal casualty, fire, or bus going off the road categories of NTD reporting. A transit injury is physical damage or harm to any person requiring medical treatment, which occurs under the collision, derailment, personal casualty, fire, or bus going off the road categories of NTD reporting. For transit employees, an injury includes an incident that results in loss of time from duty. The denominator for this measure is passenger miles. Passenger miles are miles that passengers travel on transit vehicles (buses, transit railcars, commuter railcars, ferries, paratransit vans, vanpools, etc.) while in actual revenue service to the general public. A separate Source and Accuracy Statement has been prepare for passenger miles. Passenger miles exclude travel to and from storage facilities, training operations, road tests and deadhead travel, as well as school bus and charter service.

The hours that transit vehicles (buses, transit railcars, commuter railcars, ferries, paratransit vans, vanpools, etc.) travel only refer to hours while in actual revenue service to the general public. Vehicle revenue hours include layover/recovery time, but exclude travel to and from storage facilities, training operations, road tests and deadhead travel, as well as school bus and charter service.

The passenger miles traveled on public transit vehicles (buses, transit railcars, commuter railcars, ferries, paratransit vans, vanpools, etc.) only refer to miles while in actual revenue service to the general public. Passenger miles are the cumulative sum of the distances ridden on passenger trips. Transit authorities have accurate counts of unlinked passenger trips and fares. An unlinked trip is recorded each time a passenger boards a transit vehicle, even though the rider may be on the same journey. Transit authorities do not routinely record trip length. To calculate passenger miles, total unlinked trips are multiplied by average trip length. To obtain an average trip length for their bus routes and rail routes, transit authorities use a FTA approved sampling technique. Passenger miles are the only data element that is sampled in the NTD. Vehicle revenue travel exclude travel to and from storage facilities, training operations, road tests and deadhead travel, as well as school bus and charter service.

Data Scope

For the last 20 years, transit agencies have reported the measures to the FTA in their annual NTD submissions. At a transit authority, drivers and/or other appropriate personnel record data for the NTD as they become available. For example:

• Drivers and supervisors record fatalities, injuries and incidents each day.
• Drivers and revenue clerks record unlinked passenger trips and record fares each day.
• Drivers and schedulers record service hours each day.

Almost all the variables in the NTD represents a census of the data on publicly funded bus, transit rail, commuter rail, ferry, paratransit, and vanpools in urbanized areas. It is not a sample. The only exception to the census is the data for passenger miles, where a random sample of bus and rail routes is used to calculate average trip length for each mode. Urbanized areas have more than 50,000 persons, and are defined by the Census Department. By statute, every FTA formula grant recipient in an urbanized area must report to the NTD. In cities of this size, virtually every transit authority receives FTA funding. There are only a few cities of over 50,000 persons that do not provide public transit service. Publicly funded transit service can be directly operated or purchased transportation.

Performance Measures Sample Design

When urban transit authorities are allowed to report on their local fiscal year, there is no problem in obtaining data. However, in transit, a problem exits in providing data on a calendar year reporting period to meet the Government Performance and Results Act (GPRA) timeframe. About 200 of our 600 transit authorities are not ready to report prior year data by the end of January. To be more specific, not all of the 1999 NTD report year data will be submitted to FTA by February 2000, to use in FTA’s GPRA Performance Measures. By February 2000, about 55 percent of all 1999 data will have been submitted to FTA. Although, much of this data will have been validated, estimates for 1999 measures will have to be made using less than 100 percent of the data for 1999.

For the purpose of estimating some of the 1999 measures, FTA will ask 14 large properties that normally report in April, to report certain data elements early. These 14 properties constitute over 30 percent of all transit service. Employing this method, industry estimates will be made in February 2000, based on about 92 percent of all 1999 fatalities, injuries and incidents, 87 percent of all 1999 hours, and over 91 percent of all 1999 trips and passenger miles. The estimates will be made from a very representative, stratified sample that approximates the universe. Transit fatalities, injuries and incidents, and transit revenue vehicle hours are very stable series. Between years, the coefficient of variation is very small. The estimation method has been tested for 1996, 1997 and 1998. The percent standard deviation from the actual was about one percent for injuries and three percent for fatalities, less than 1 percent for hours, under 2 percent for trip and passenger miles: very accurate predictions of final NTD figures.

Data Collection

Method - After the close of their local fiscal year, transit authorities produce annual reports, summarizing operating, safety, fleet and financial data. Under Federal requirements, financial reports must be audited. The data is also certified. The data in these reports are entered on to forms on a diskette submitted to the NTD. The data on this diskette must conform to the precise data definitions in the Reporting Manual for the NTD and the Uniform System of Accounts for the NTD.

Schedule - Each year, the data on this diskette is submitted within four months of the close of the local fiscal year. Transit properties that use calendar year reporting submit their completed NTD diskettes in April. Of the 600 transit properties, over 400 will have reported by January 2000. These reporters comprise 55 percent of all transit fatalities, injuries and incidents, transit hours, and transit trip and passenger miles data. Response Rates - The NTD program is required by statute. For 20 years, the 600 transit operators in urbanized areas, with populations in excess of 50,000 persons, have filed annual reports to the NTD. Every FTA formula grant recipient must report to the NTD. Without an annual NTD submission, FTA grant funds are cut off. The response rate is about 100 percent. In cities of this size, virtually every public transit authority receives FTA funding. For this reason, the NTD is one of the more complete industry databases in the Federal government. For these sample estimates, 92 percent of transit fatalities, injuries, and incidents, 87 percent of transit service hours, and 91 percent of total transit trips and passenger miles will have been reported.

Sampling Error

Sampling errors produce faulty estimates. In the NTD, other than passenger miles, annual data are actual data, not sample estimates. Further, transit systems want to continue FTA funding, so they submit NTD data. Simply put, the universe is known. However, for the Performance Measures, the estimate of annual fatalities, injuries, and incidents, estimate of annual hours, and estimate of annual trips and passenger miles will be based on 92 percent, 87 percent, and 91 percent of all transit that will have been reported by February 2000, respectively. Using data from the last several years, this method was tested. Checking predictive models against historical data is called "backcasting."

This method produced estimates with mean forecast errors of less than one percent. A sample based on 92 percent of actual fatalities, injuries, and incidents data, 87 percent of actual hours data, and 91 percent of actual trips and sampled passenger miles data for all transit approximates the universe. Many of the modal statistics that BTS and DOT use as a matter of routine are based on less than 85 percent of the universe. In these estimates, sampling errors are not a serious problem.

Nonsampling Error

Nonsampling errors are primarily a result of noncoverage errors. Data for tiny systems in urban areas and small systems in nonurban (rural) areas are not included in the NTD. Nonurban systems and operators with less than 10 vehicles in urban areas combined would add only a percent or two to total transit service nationwide.

Not all data errors are related to poor sampling techniques. Some errors are measurement or coding errors. Data may not be gathered from some groups. These errors are discussed in this section.

Nonentry Error or Missing Data - The NTD program is required by statute. Every FTA formula grant recipient must report to the NTD. Without an annual NTD submission, FTA grant funds are cut off. The response rate is near 100 percent. Fields that are left blank are returned to the transit agency along with a detailed review letter highlighting errors and omissions. However, a small percentage, less than one percent, of data is missing. Tiny transit authorities, with less than 10 vehicles, are exempt from having to complete certain forms. This could produce a nonentry on certain data elements for about 30 tiny systems. On a few occasions, a few months of data are lost when contractors are changed. For the GPRA estimates, about 8 percent of fatalities, injuries, and incidents, 13 percent of hours, and 9 percent of unlinked trips and passenger miles for 1999 will not have been reported because of local fiscal years, not missing data.

Duplicate Entry Error - The NTD program requires that services purchased by a transit authority be reported separately from directly operated service. This avoids the double counting problem. The data audit and certification requirements also help avoid redundancy. Very few fatalities, injuries, and incidents reports, hour reports, and trip and passenger reports involve more than one transit authority, reducing the chance for double counting.

Response/Measurement Error and Coding Errors - Measurement errors occur when incorrect data is provided. Coding errors occur when correct data is improperly recorded. NTD staff work hard to catch bad data and recording errors. First, transit authorities file NTD reports each year; the NTD is not a special study. Regular reporting reduces errors. Second, the data is audited and certified by local officials. Third, FTA validation analysts, familiar with this transit authority, use range checks for 1999 data against data from last year and previous years. Data fields are also checked for proper coding. Validation is discussed, below. Fourth, validation ratios and performance measures, such as costs per hour, miles per hour, are calculated. These ratios are compared to previous submissions and systems of similar size. Any significant variations are flagged and returned to the transit authority to explain or revise. This validation feedback loop is very important in producing accurate data for legislative apportionments is not a common feature of most industry databases run by the government.

Noncoverage Error - Noncoverage errors occur when members of the population are missing because they were not included in the sample. In the NTD, there are some noncoverage errors, but these errors are deminimus. First, as noted under missing data, certain tiny public transit authorities, with less than 10 vehicles, exist in urban areas. These small systems do not have to report the full NTD data. There are about 30 of these tiny systems accounting for less than one-half of one percent of total transit service. Second, there are a few public transit authorities, usually with very few vehicles, that do not receive FTA funds and do not report. However, most public transit authorities that do not receive FTA funds, voluntarily and regularly report to the NTD to increase the apportionment for their urbanized area. Third, the largest piece of missing data is public transit in rural areas, or nonurban areas with less than 50,000 persons. The States receive FTA funding to administer these programs. Because these systems are not direct FTA grant recipients, they do not submit data to the NTD. The NTD was not designed to capture this data. Nonurban systems would add only a percent or two to total fatalities, injuries, and incidents nationwide.

Verification and Validation

To produce an accurate and equitable apportionment of FTA funds across the nation, the FTA has made a commitment of significant resources in our NTD detailed verification and validation feedback process. Intensive data validation efforts are not a common feature of most industry databases run by the government. In most industry databases, data is usually accepted as submitted. The NTD employs a number of exhaustive verification and validation efforts.

First, transit authorities file NTD reports each year; the NTD is not a special study. Regular reporting increases consistency and reduces errors. Second, at the local transit level, the NTD diskette contains certain error checks. Third, prior to submission, the data is audited and certified. An independent auditor must complete an A-128 audit and signs off on the NTD submission. The agency’s CEO certifies the submission. Fourth, FTA validation contractors, familiar with this transit authority, use range checks for 1999 data against data from last year and previous years. Data fields are also checked for proper coding. Errors and inconsistencies are enumerated in a Detailed Review Letter (DRL) that is sent back to the submitting transit agency. DRL problems must be addressed and data revisions made for inclusion in the NTD. Failure to address validation or certification problems can result in loss of eligibility for FTA grants. Fifth, validation ratios and performance measures, such as costs per hour, miles per hour, are calculated. These ratios are compared to previous submissions and systems of similar size. These ratios check the internal consistency of the submission. Any significant inconsistencies are flagged and returned to the transit authority to explain or revise. The NTD contractor performs validation checks involving 200 calculations on each submission. This validation feedback loop is very important in producing accurate data for legislative apportionments and fixed-guideway allocations. About 55 percent of the data in the 1999 Performance Measures estimates will have been validated.