AAA gasoline-price methodology

What the daily AAA number represents, what is publicly documented, and how it should be used in Pretickted models. This distinction matters because a daily observation-derived price is not the same target as GasBuddy's final tick.

Publisher
OPIS for AAA

AAA publishes the number; Oil Price Information Service updates the underlying averages.

Coverage
Up to 120,000 stations

AAA describes a daily retail-price survey conducted in cooperation with WEX.

Alignment
Measurement day

Pretickted aligns an AAA value to the day whose prices it measures, not merely the following publication timestamp.

The important distinction

What it is

A broad, validated retail pump-price index. WEX/card activity can provide observations of prices actually paid, while OPIS also maintains a large station-price database and applies validation and aggregation rules.

What is not established

The public documentation does not establish that the AAA figure is total consumer dollars divided by total gallons, or that busy stations and busy hours receive weight strictly proportional to nationwide fuel volume.

Implication for the nowcast

The AAA target should not automatically inherit GasBuddy's entire projected close-to-close move. A GB move early in the measurement day can influence many subsequent price observations; the same move late at night should usually have less effect on that day's AAA result. Using GB's projected end-of-day value therefore risks forecasting too high on rising days and too low on falling days.

Preferred target: an observation-time-weighted GB path, reconstructed geographically and integrated across the measurement day. Learn the intraday weights from historical AAA settlements; do not assume that they equal traffic, gallons purchased, or a hand-drawn rush-hour curve.

Initial intraday-weighting backtest

62 walk-forward days · Jul 22–Sep 21, 2026

The raw GB close produced 1.65¢ mean absolute error against AAA's daily move. Replacing it with the uniform average of GB's hourly levels reduced error to 1.20¢, supporting the time-window-mismatch hypothesis. Once each method also receives a trailing day-of-week residual correction, close scored 0.90¢, uniform average 0.97¢, and a nonnegative regularized learned-hour model scored 0.85¢.

1.65¢
Raw close
1.20¢
Raw daily average
0.90¢
DOW close
0.97¢
DOW average
0.85¢
Learned hours

These are initial results, not a production-model replacement. Only information preceding each scored day was used. The learned hourly curve remains experimental because three months of history is limited and its fitted shape may also absorb publication alignment, weekly seasonality, and GasBuddy sampling effects—not just consumer purchase timing.

Confirmed publicly

  • AAA's national, state, and local averages are updated daily by OPIS.
  • AAA says as many as 120,000 stations are surveyed in cooperation with WEX.
  • OPIS documentation describes validation, station-level daily observations, and median-based retail-price products.

To infer from our data

  • The effective intraday observation-weight curve.
  • How much late-day GB movement rolls into the next AAA measurement.
  • Geographic weights and sampling differences between AAA, GB national, states, and counties.
Primary references

AAA Gas Prices — About AAA · OPIS PricePro User Guide · OPIS retail-price coverage

Caution: AAA's public page summarizes its coverage but does not publish a complete formula for the daily AAA aggregation. Pretickted should label any more detailed timing or weighting description as an empirical inference until independently documented.