The introduction of a variety of fare media has increased riders’ choices of how to pay for transit and has provided transit agencies with the abil- ity to price fare media differently. An aggregate elasticity value such as the Simpson-Curtin formula provides cannot capture the shifts among fare media, and elasticities for specific fare payment categories do not directly address these shifts. The Metropolitan Atlanta Rapid Transit Authority (MARTA) fare elasticity model was developed to identify shifts among fare payment methods and to forecast ridership and revenue changes resulting from a fare change. The key element in predicting shifts among fare media is the identification of the propensity of riders to choose one fare medium over another even if their prices are identical. After current ridership is adjusted to take account of these shifts, fare elasticities are then applied to predict post-fare-change ridership levels by fare payment method. The new ridership figures are used to forecast new revenue. The model requires minimal input from the user, primarily current ridership by fare payment mode and current and proposed fare levels. Elasticities can also be changed by the user to test the sensitivity of the elasticity assumptions, and factors such as the intensity of weekly and monthly TransCard use are additional input options. The model has proved to be very accurate when applied to the most recent MARTA fare change on July 1, 1995, predicting overall ridership within 0.05 percent and revenue within 0.7 percent. A model enhancement allows the user to recalibrate the model after any future fare changes by testing the accuracy of different elasticity assumptions. Fare elasticities are used to estimate the effects of fare changes on ridership. Over the years, the Simpson-Curtin formula (1) predict- ing a decrease of 3 percent in ridership for every 10 percent increase in fares has proved to be accurate in overall forecasts of ridership response to fare changes. This formula translates to an elasticity of ridership with respect to fare of -0.3. Studies examining particular groups of riders have revealed marked differences in fare elasticities (2,3). There is a consensus from these studies that regular commuters, riders with long trips, and peak-period riders are less sensitive to fare changes than nonwork riders, riders making short trips, and off-peak riders. The introduction of a variety of fare media has increased riders’ choices of how to pay for transit and has provided transit agencies with the ability to price fare media differently. This has created a sit- uation in which fare changes can induce shifts in fare payment meth- ods, especially when the percentage changes differ for different types of fare media. An aggregate elasticity value such as the Simpson- Curtin formula provides cannot capture the shifts among fare media, and elasticities for specific fare payment categories do not directly address these shifts. A fare elasticity model was developed at the Metropolitan Atlanta Rapid Transit Authority (MARTA) several years ago to forecast rider- ship and revenue impacts of fare increases. This model specifically TRANSPORTATION RESEARCH RECORD 1669 Paper No. 99-0908 123 takes into account shifts among different fare media. The techniques for doing so are based on a 1984 study at Tri-Met in Portland, Ore- gon (4). As part of a broader review of MARTA’s models (5), the fare elasticity model was tested using data from MARTA’s most recent fare change (July 1, 1995). Results showed that with minor adjust- ments for elasticity values, the model predicted total ridership within 0.05 percent and ridership by fare method within 2.8 percent in the worst case and within 1.6 percent on average. The model was tested for 1 month, 3 months, and 1 year after the fare change (compared with a similar period before the change). The most accurate results were achieved in the 3-month time frame. Thus, MARTA’s fare elas- ticity model can be very useful as a technique to predict ridership and revenue impacts resulting from proposed fare changes. Clear and com- plete documentation of the model would obviously lend an additional degree of confidence in using the model as a forecasting tool. The assumptions and inputs necessary to run the fare elasticity model are described in this paper, along with the theoretical basis and workings of the model itself. The model was developed in a Microsoft Excel spreadsheet format, with six individual sheets that show inputs (Figure 1), adjustment factors (Figure 2), core computations (Figure 3), summary, details (Figure 4), and notes. Enhancements include the K-factor sheet (Figure 5), which demonstrates more clearly some of the steps in the model, and the Other Factors sheet (Figure 6), which shows assumptions used to generate ridership. Monthly ridership data in the examples are from March 1995, 3 full months before the fare change. The model was calibrated to predict ridership in October 1995, 3 full months after the fare change. BRIEF DESCRIPTION OF MODEL The fare elasticity model first analyzes the number of riders choos- ing each fare payment method in relation to its price. This analysis results in factors to describe how many riders would choose one fare payment method over another if the price of both were identical on the basis of observed trends before the fare change. The model then uses these factors with the new fare structure and identifies shifts from higher-priced to lower-priced fare options. Finally, fare elas- ticities specific to each fare payment method are applied to forecast ridership by fare medium and total ridership. Since the value of each ride is known, a revenue forecast can be readily generated. The following sections present a detailed description of how the model works. Input Variables In order to run the model, the user must input the following values (on the input sheet, Figure 1): Metropolitan Atlanta Rapid Transit Authority Fare Elasticity Model ANNIE E. HARRIS, ROBERT THOMAS, AND DANIEL BOYLE A. E. Harris and R. Thomas, Metropolitan Atlanta Rapid Transit Authority, 2424 Piedmont Road, N.E., Atlanta, GA 30324-3330. D. Boyle, Trans- portation Management & Design, 531 N. Highway 101, Solana Beach, CA 92075-1132.