Source: U.S. Department of Transportation, Bureau of Transportation Statistics, Office of Airline Information (government).
TSI uses aviation revenue freight ton-miles data reported by air carriers to the Bureau of Transportation Statistics' Office of Airline Information (OAI). The data, referred to as the T-1 dataset under the T-100 and T-100f databases, are published monthly in Air Carrier Traffic Statistics (the Green Book) produced by OAI. Reference to the Green Book can be found at http://www.bts.gov/oai/sources/.
The T-100 databases include data reported by U.S. carriers operating between airports located within the boundaries of the United States and its territories. These data fields contain information on passengers, freight, and/or mail enplaned at the origin airport and deplaned at the destination airport.
Data in the Green Book consist of U.S. passenger and cargo capacities and traffic actually carried by carrier and class of service in both scheduled and nonscheduled (or charter) service for the most recent month and 12-month period, and similar data for like periods in the previous year. Data are summarized by U.S. carrier entity and carrier group (major, national, etc.).
Large U.S. certificated air carriers, small certificated air carriers, commuter air carriers, and all-cargo air carriers are required to complete Form 41 Financial and Traffic Reporting Requirements. This form is completed both electronically (mainly by the large carriers) and on paper (small and commuter carriers).
The T-1 dataset includes information on all U.S. commercial air carriers, except air taxis. The dataset does not include foreign air carriers, entities that provide commercial services illegally, in-house air transportation (e.g., corporate jets), general aviation, and air taxis.
A draft data quality assessment, Data Quality Assessment of the Large U.S. Certificated Air Carriers Traffic and Capacity (Form 41, Schedule T-100) Databases, was issued on March 31, 2003, by the Bureau of Transportation Statistics, Office of Statistical Quality. The information below on the T-1 dataset, a product of OAI used in the TSI, was extracted from this data quality assessment.
As noted, data are sometimes reported late. However, the carriers also sometimes revise data after publication in the Green Book.
Two adjustments are made to the most recent T-1 data to improve timeliness. The first adjustment is made to account for carriers that have not yet reported. The adjustments assume that the missing carriers grow at the same rate as carriers that have not yet reported. The second adjustment is to forecast the most recent month based on the 4-quarter growth rate using the most recently available three months. The accuracy of these adjustments is detailed in the next section.
U.S. DOT Office of the Secretary of Transportation for Policy (OST-X) analysis of the data submitted by small certificated Alaskan air carriers identified significant reporting problems when these carriers first began reporting in 2002. OAI has since taken steps to improve the accuracy of the data submissions and OST-X no longer considers this an active area of concern.
Airline traffic and financial data are regulatory data, certified as being correct, by the submitting carriers. The data cannot be changed without the carrier's consent. Although BTS can reject data submissions that are clearly inconsistent, BTS does not have the legal authority to make changes in carrier submitted data.
OAI follows up with carriers regarding questionable data, but the sheer volume of records failing the established edit checks precludes investigation of every issue. Only a small fraction of the records are sent out for review.
BTS measures the accuracy of the adjustments for missing carriers by comparing the initial estimated growth rates based on the adjustment (the alternative value) to final growth rates based on data at least 4 months after the initial release. If the alternative data is an unbiased predictor of the growth rates from the final data (at least 4 months from initial release) then the average of the paired differences between the alternative and final growth rates should be zero. The t-statistic for the mean difference is -0.56 and is within the (-2.12, 2.12) range for a 5% test at 16 degrees of freedom so we do not reject the null that the mean is zero.
|Final||Alternative for Initial Release||Difference||Actual at Initial Release||Difference|
|Average Absolute Error||1.12||24.44|
|Maximum Absolute Error||3.85||21.81|
We measure the accuracy of the forecasts in a manner similar to that used for the alternative data. The t-statistic again indicates the forecasts are an unbiased predictor of future growth rates . However, the forecast error is large enough to be of concern and improving these forecasts remains an active research.
|Forecast Month||Release Month||Actual||Forecast||Error|
|Start with June 2004 Release|
|Average Absolute Error||3.5|
|Maximum Absolute Error||13.9|
|Average Absolute Deviation||5.4||5.0||3.5|
|Start with Feb-04 Release|
|Average Absolute Error||4.1|
|Maximum Absolute Error||15.3|
|Average Absolute Deviation||6.4||6.4||4.0|
The forecast values for the 4/20/2004 and 5/17/2004 releases actually forecast 2 months ahead.
Only final data (forecasts for 9/2005 and earlier) are used in the statistics.
Forecasts use only information that was available at the release date.
The forecast uses the most recently available 4-quarter growth rate.