I’ve been experimenting with USDOT DB1B and T100 data a lot recently and would like to test out something I’ve been working on by using Allegiant as an example within this thread.
I used regression analysis based on a scatterplot of route yields plotted against stage length to calculate a stage length adjustment formula that I have applied to Allegiant to find their highest yielding routes in Q3 2022 with yield adjusted to Allegiant’s average stage length of 910 miles. This analysis is highly experimental, and I chose Allegiant due to their lack of connecting traffic as I am not yet at the point where I can apply it to other carriers due to certain assumptions for how to properly allocate a market fare on a multisegment itinerary according to the length of each segment.
Disclaimers include the low R-squared value of my power series formula (basically a measure of how well my formula for median Allegiant yield in relation to stage length fits the data). I also had to assume relatively consistent profitability according to the length of haul (which isn’t necessarily the case as cost per available seat mile (CASM) might not follow the same power series formula for its relation to stage length). Ancillary revenue also isn’t included, which is big for Allegiant, and, again I assumed no connecting traffic and filtered out O&D markets with less than 2 PDEW in Q3 2022.
The first tab lists routes (separate values for each direction) with Allegiant O&D demand on that route, average one way market fare, stage length, reported passenger yield, and stage length yield adjusted to Allegiant’s average of 910 miles.
The second tab lists station averages, including total Allegiant departing O&D from that station, average one-way market fare, reported passenger yield, stage length adjusted passenger yield, station average load factor, and stage length adjusted PRASM (passenger revenue per available seat mile) which is probably the best proxy for station profitability possible at my current level of data processing.
Again, all this is experimental, took a lot of work to compile which could have led to mistakes I haven’t caught yet, and is only as accurate as the data provided to the USDOT and the accuracy of my regression analysis.
Please tell me any thoughts, suggestions for improvement, or any ideas on how to apply this to carriers with connecting traffic (it’s a puzzle I still have yet to figure out, and depends on the accuracy of this first stage)