Yesterday had a bit of a buzz on the internet regarding a piece about airfare pricing from Nate Silver that was published on his NY Times politics blog. The post, filled with mathematical analysis, attempts to use statistics to determine which airports have unfairly high fares relative to others providing comparable service. And I’m sure the math involved is accurate. I have no doubt that someone as statistically gifted as Silver got the regression analysis correct when he ran the numbers. But the findings are still miserably flawed.
Why? Because several of the assumptions made simply do not apply to air travel.
Silver acknowledges that most the other folks who have tackled this topic have made specific flaws in their assumptions. He aims to correct these but instead makes some tragic assumptions of his own.
Let’s take a look at the factors he considers:
The first factor is the distance traveled — we use the distance from the origin airport to the destination as though it were a nonstop flight, whether or not there was a layover along the way….
The first factor cited – distance traveled – is probably one of the last things that actually comes into play when airlines are figuring domestic market pricing. Should they? I can see that argument being made, but it ignores the general concept of market pricing and supply/demand dictating the going rate for a ticket. If the airlines wanted to price everything based on distance they could, but they’d be leaving a lot of money on the table for the shorter flights and they’d never sell the longer ones. Even just using the average costs to operate a flight as a price basis you’d be looking at $600+ on average for a round-trip transcontinental flight. They seem to sell a lot better in the $300 range, at least in major markets.
Silver chose to ignore whether there is a connection or not. While that is reasonable for calculating the distance traveled, it ignores perhaps the single greatest factor that drives travel bookings for business travelers, the folks paying the higher fares: schedule. When you’re a business traveler hopping between cities and trying to get to that next appointment on time and then home as quickly as possible you pay more for a non-stop flight. Should you? Maybe, maybe not. But you do. This pricing function is probably more directly traceable in cargo numbers and there is a ton of data available on that, including in Greg Lindsey’s Aerotropolis, a pretty good read. But the same concept absolutely applies to passenger travel as well. There is a very real value in speed out in the real world; there apparently isn’t one in Silver’s.
Silver found that Newark was about 25% more expensive than JFK based on his data. And there is no doubt that is the case on some routes. But when you also consider that Newark has quite a few more domestic destinations available as a non-stop flight than JFK does that price premium isn’t nearly as surprising. After all, folks pay for speed.
Certainly demand factors into the pricing as well:
Second is a variable representing the demand for travel at both the origin and destination airports. Demand is assumed to be a function of the number of origin-and-departure passengers that an airport handled (not counting passengers who passed through the airport on a layover), but with a modification for average ticket prices. In other words, if the average fare at an airport was high, the model assumed that more people would have wanted to fly there but were deterred by the cost, and if the average fare was low, that some passengers would not have flown if the fares had not been such a bargain.
Indeed, one can expect that fares to smaller destinations will be higher. And they generally are. But assuming that more people really want to be traveling to smaller cities but choose not to because the airfare is too high misses the point. They are smaller cities with lower demand for travel because they have fewer businesses, fewer residents traveling (or being visited) and generally less volume. They aren’t seeing lower air traffic because they are too expensive, they are seeing lower traffic because they are small. Lowering fares may translate to a small increase in volume but it most certainly is not a linear path.
Moreover, the ability for a new entrant to operate in a market requires a certain base level of demand. No matter how cheap the fares, you aren’t going to survive long as a startup carrier if your hubs are in Columbus, Ohio and Greensboro, North Carolina, for example; just ask SkyBus. This means major metropolitan areas see the up-starts, and those up-starts bring lower fares because that’s how they attract customers. Their fares go up over time – JetBlue and Southwest have proven this – but that’s where it begins. And that explains a lot of the pricing trends that are seen today.
Finally, Silver looks at the most important factor, competition:
The regression analysis also accounts for three other factors that have significant effects on pricing. These are, respectively, the market share at the origin and destination airports held collectively by the five “legacy carriers” (United, American, Delta, Continental and US Air); the market share held by Southwest Airlines; and the market share held by the largest single carrier at that airport (for instance, Delta and its affiliates are responsible for about 66 percent of all traffic at Atlanta).
Passengers at Newark paid an average of 12 percent more than those at J.F.K. for their trips to Los Angeles, 49 percent more for those to Chicago, 65 percent more to Dallas, and 118 percent more to Washington, D.C.
Given those numbers, it is probably useful to take a look at the competition in those markets. There is zero competition between Newark and Washington, DC. National airport is only served by Continental and Dulles is served only by Continental and merger partner United Airlines. Plus, those routes are not generally reasonable to fly with a connection. The travel time is so short that when you add the connection it is silly to fly when total travel time is important, as it often is. The Dallas route sees a bit of competition from American Airlines, as does the Chicago route. Los Angeles has a tiny bit of competition but it also has the advantage of being a long enough trip that making the schlep over to JFK to save some money on airfare doesn’t actually completely ruin the speed=value margins. Ditto for connecting flights that add a smaller percentage of time to the travel experience.
Somewhat ironically based on the first factor Silver names, longer distances traveled can actually drive down prices as the impact of connections or less desirable departure or arrival airports is decreased as the total travel time increases.
It is actually surprising that Silver didn’t note the disparity on pricing in the Newark/JFK – Boston market. For quite some time now Continental has held a monopoly on that route. Similar to the DC runs, it rarely makes sense to connect for such a short trip and Continental exploited that price disparity. Right up until JetBlue announced their entry into the market. The fares dropped quite quickly at that point. Hardly a surprise, really. Competition, not the airport, drove the pricing.
Here’s a much more simple way to figure out if an airport is expensive or not:
- Is it a mostly leisure destination? If the answer is yes then it is almost certainly not going to be as expensive on average. Atlantic City, Las Vegas, Ft. Lauderdale, Orlando and most the rest of Florida all come to mind, and not surprisingly they’re all on Silver’s list of good value airports.
- Is it dominated (60%+) by a single carrier?
- Is it a particularly large metropolitan area? If not, fares are going to be higher because demand is lower.
Three easy questions that don’t take statistical regression or misguided assumptions. Silver actually gets some of these, particularly regarding the competition factor. But he also has a couple huge misses, especially around distance traveled and the price/demand curve.
It would also be interesting to compare the actual costs of travel versus just the base fare data. Spirit has a pretty incredible ancillary revenues per passenger – to the tune of an extra $35/head on average – so those "cheap" airports can come with significant surprises once the customer gets to the airport. Indeed, the airlines are quite keen to sell these ancillary bits to their customers and many are now stating explicitly that these fees are where their profits are. The airlines even want to control the way those fees are marketed to the customer by cutting the GDSes out of the pricing loop. Not a good deal for consumers.
Oh, and the suggestion he links to about searching for the best airfares on weekends is horribly wrong, too. Tuesday or Wednesday mid-afternoon is the time you’re most likely to find deals. On the weekends the airlines are raising fares and limiting the cheaper inventory in an effort to cash in on folks shopping for their vacations while their home with their family.
Silver should stick to baseball and politics, two things that he appears to understand a lot better than air travel.
- The Aerotropolis as the hub of global economic growth
- Let’s talk airfare and distribution
- JetBlue now selling Boston-Newark, killing Continental cash cow