The US DoT has a data set we just started toying with. It’s named Schedule B-43 Inventory. It contains “Annual Inventory of Airframe and Aircraft Engines”. And does not have so much as one pixel on engines. Oh well.
We decided to download data from 2010 through 2019, the most recent year. There are some issues with the data in its current form, for example, aircraft model is truly random. We had to create a translate sheet to try clean the data. There are, for example 17 ways to name the 737-300, and only 13 of those is a derivative. Going through nine years of data to try decipher the various models took several hours. And still there are errors. We are now engaging with DoT to see who gets to decide a model name. A further example: Frontier and Spirit fleets are really inaccurate. In 2017 Frontier had 17 A320neos, only one in 2018 and none in 2019. Er, not quite.
With the caveat that the data is clearly not as accurate as we would like, let’s take a dive. There are other data points do provide some really (plausible) interesting insight. Take a look at average seating on the various versions of the 737-800 (including the -8). Clearly there’s a problem with the data set apropos Southwest when we add the 737-8H4 – the 2015 seating matches the 737-700s. Leaving out the 737-8H4 we have a straight line at 175. This is supposed to be a 162-seater by the way. United is the one for best comfort.
Next let’s look at the A320 fleet. Bear in mind data flaws. Frontier now has 182 seats on what was originally a 150-seater. Allegiant and Spirit are right up there too.
We see that United and Alaska have taken the opposite strategy, making their A320s relatively comfortable. Delta, and JetBlue added seats in the recent years. Only American has kept is A320s at 150.
A third interesting chart to share. When American merged with USAirways, overnight it became Airbus largest customer. And Airbus has been making steady headway there ever since. Despite the airline ditching the A330s, the A321s are more than making up for it.
We will continue to work on this data set and publish a model when we feel we have it as right as we can make it.