Frequently Asked Technical Questions
1. How do the ATPI series differ from air travel price index series currently published by BLS?
- Prices Included in the
Index Calculations. The Bureau of Labor
Statistics (BLS) currently publishes both a Consumer Price Index (CPI) and
a Producer Price Index (PPI) for airfares. The CPI measures changes in the
prices paid by consumers for airline trips, including taxes and any
distribution costs not received by the carriers (e.g., travel agents’
fees), while the PPI measures changes in revenues received by producers of
airline trips. The CPI includes trips purchased from foreign carriers,
while the PPI excludes these. The ATPI series presented here are similar to
the BLS CPI series in that prices include taxes paid as well as fares.
Distribution costs, however, are not included and, due to legal
restrictions on use of the Passenger Origin and Destination (O&D)
Survey data, the ATPI series currently incorporate only prices reported by
One of the key differences between the ATPI and the BLS indexes is the ATPI’s inclusion of special discount fares (e.g., Iinternet specials, credit card discounts, frequent flyer awards), which are not reflected in the fares available from the SABRE system, a reservation system used by travel agencies. ATPI fares also include bulk fares, e.g., tickets purchased by travel agencies for resale in packaged tours.
- Index Formula. Another important difference between the BLS indexes and the ATPI lies in the target index formulas used. The BLS PPI for airfares is based on a Modified Laspeyres formula, while the CPI employs a “hybrid” Jevons/Modified Laspeyres formula. The ATPI estimator, by contrast, targets a Fisher index. For a discussion of alternative index formulas, see, for example, Fisher (1922), Diewert (1987), or Moulton (1993). For more information on price index concepts and design, see Schultze and Mackie (2002).
- Reference Period. The ATPI series also differ from the official BLS index series in their reference period definitions. From the current O&D Survey data, we can compute only quarterly indexes, and the reference quarter is the quarter in which the airline ticket was used for travel.1 The BLS PPI and CPI are monthly, and BLS collects prices of tickets sold (not necessarily used) during the reference month.
- Scope. The scope of the ATPI is slightly wider than that of the BLS airfare indexes. The BLS CPI and PPI cover only trips that originate in the United States, while the Full-Scope ATPI covers trips originating in foreign countries, provided they have a domestic portion.
2. Why aren’t the indexes the same as the average
fare (or “yield”) numbers computed by the Office of the
Secretary of Transportation?
Both the ATPI and the average fare numbers are computed from the BTS Passenger O&D Survey data, but they’re designed to measure different phenomena. In the average fare computations, fares paid for different trip routes on different carriers are simply combined into overall averages. Changes in these averages do not necessarily reflect price changes; they are driven primarily by changes in the collection of airline trips purchased in different time periods. Thus the average fares may, for example, show dramatic shifts even when the array of prices facing the consumer remains relatively constant. A price index, by contrast, is designed to isolate changes in price for services that are essentially identical, e.g., the exact same trip route flown on the same carrier in different time periods. A large body of theory supports the use of the Fisher index (see above) for this purpose.
3. Are the ATPI estimates included in this release
“official” BTS data products?
No. The estimates in this release were computed for research purposes.
Index Estimation Method
Most price index estimates are computed from “matched samples,” i.e., statistical agencies select a fixed (or gradually rotating) sample of items and track the prices for the same set of items across time. The peculiarity of the quarterly O&D Survey data, for the purpose of price index estimation, is the lack of across-time matching of individual itineraries: the 10% sample is selected independently each quarter and may be treated as a simple random sample for estimation purposes.
To circumvent the across-time matching problem, we divide each quarterly sample into detailed categories and compute a unit value index (average price in time t divided by average price in time t-1) for each category. The unit value indexes are treated as “elementary aggregates,” which are then further aggregated using the Fisher formula. Unit value indexes are appropriate only for aggregating prices of items that are very similar (e.g., round-trip United Airlines coach service from Boston to San Francisco with one stop in Chicago).
Our method of matching categories of air travel services across time comprises two matching stages. The first stage of matching is itinerary-level matching. In this stage the itineraries are categorized according to the following variables:
1a) Sequence of origin and destination airports (i.e.,
origin airport, first destination airport, second destination airport,
1b) Sequence of classes of service (i.e., service class for first segment, second segment, etc.)
1c) Sequence of operating carriers
1d) Number of “trip breaks”
The trip breaks are points in the itinerary at which a passenger is assumed to have stopped for a reason other than changing planes. Itineraries that are identical in characteristics 1A though 1D form a first-stage unit value category. Note that trips within a first-stage unit value category must have exactly the same number of trip segments or flights. As the number of segments increases, the percentage of categories appearing in both of two consecutive quarterly databanks decreases.
The second-stage matching procedure is segment-level matching. Itineraries not matched in the first stage are broken into individual segments. Since only itinerary-level fares are available in the databanks, the second-stage matching procedure involves imputing a fare for each segment, based on the itinerary-level fare. After imputing the fares for second-stage matching, we group the trip segments into second-stage unit value categories based on the following variables:
2a) Segment-level origin and destination airports
2b) Class of service
2c) Round-trip itinerary or non-round-trip itinerary
2d) U.S. origin itinerary or foreign origin itinerary
2e) Operating carrier
Average fares are computed for these segment-level categories, and these are matched from quarter to quarter to compute unit value indexes. The entire matching process is performed separately for each pair of consecutive quarters to create a “rolling” sample. We expect, however, that a small percentage of trip segments will always be omitted from the index computations due to incomplete matching.
In general, roughly 84% of itineraries, representing about 75% of passenger flight segments, are matched in the first stage. (Because itineraries comprising large numbers of segments are less likely to be matched in this stage, we expect the percentage of itineraries matched to exceed the percentage of segments matched.) About 75% of the passenger segments not matched in the first stage are matched in the second stage. The segments matched in the second stage represent approximately 18% of passenger flight segments in the databanks, so the resulting total percentage of segments matched is about 93% to 94%.
Once the matching is performed and unit value “subindexes” have been computed, we use the Fisher index formula, along with expenditure share weights from the O&D Survey data, to aggregate the subindexes into index estimates for large categories. For a detailed discussion of the ATPI estimation method, see Lent and Dorfman, "A Transaction Price Index for Air Travel," available at http://www.bls.gov/opub/mlr/2005/06/art2full.pdf.
Diewert, W. E. (1987), “Index Numbers,” The New Palgrave: A Dictionary of Economics, Vol. II (edited by J. Eatwell, M. Milate, and P. Newman), pp. 767-780, MacMillan, London.
Fisher, I. (1922). The Making of Index Numbers: A Study of Their Varieties, Tests, and Reliability. Sentry Press, New York.
Moulton, B. R. (1993). “Basic Components of the CPI: Estimation of Price Changes,” Monthly Labor Review, December 1993, pp. 13-24. U.S. Government Printing Office (available on line at http://www.bls.gov/opub/mlr/1993/12/art2full.pdf).
Schultze, C. L. and Mackie, C., editors (2002). At What Price? Conceptualizing and Measuring Cost-of-Living and Price Indexes, National Academy Press, Washington DC.