Yesterday I was participating in the usual weekly recording on the Airplane Geeks Podcast, and I noticed that my fellow co-host Rob Mark wanted to discuss an article entitled “Delays Worse After JFK Runway Work Done.” Some interesting points about the DOT’s on-time database popped into my head as I was researching and preparing for last evening’s discussion, and I think some of them are worth sharing here as it shows the importance of properly using government data to draw conclusions. Let’s take a look at two of the first paragraphs in the story:
When construction began in March 2010, officials said the revamped Runway 13R-31L would reduce flight delays by cutting the time planes spent traveling to and from the gate by up to a minute, or an estimated 10,500 hours a year.
But since the runway reopened in July 2010, the percentage of late arrivals and departures has increased.
When playing with operational data, it’s very important to be using the right statistics. According to the article, the runway was supposed to reduce taxi times, but instead uses late departures and arrivals to judge its effecgtiveness, which in my opinion is an error. First, departure delays are based on gate departure time, so a shorter taxi to the runway wouldn’t have any affect on departure delay. One could argue that a shorter taxi-in time for the inbound flight could lead to better departure performance, but I think that’s a pretty minor factor. Second, there are many factors that can affect flight delays, so using broad on-time statistics and implying that the runway is involved isn’t accurate. (Granted, the author does mention that delays have many causes, but that’s buried a few paragraphs into the article.)
If we actually look at taxi times (which is included in the DOT data), the story at JFK looks quite different. The author compared October 2009 – March 2010 to the same period a year later. In the later period, taxi in times at JFK were shorter in five of the six months being compared. Taxi out times were shorter in four of the six months.
But while taxi time numbers are (in my opinion) more reflective of the runway’s impact on operational performance, there’s still some very important caveats to consider. First, the DOT data doesn’t have any information on weather conditions at the airport, so if one month is stormier than the other a fair comparison might not always be generated.
The other issue with the DOT database is that a bunch of flights aren’t included, as the data only covers domestic flying. Second, only carriers that “have at least one percent of total domestic scheduled-service passenger revenues” are required to share their data, so that excludes smaller airlines like Virgin America, Spirit, and a bunch of regionals. The only exception is ExpressJet, which reports voluntarily. Obviously this can be an issue at a major airport like JFK. Just for fun, I took the T100 segment data, filtered out flights operated by all-cargo aircraft*, selected airports with at least 10,000 departures, and determined how many of those flights were performed by carriers that didn’t report data that month:
The DOT on-time database is very powerful and has a wealth of useful information for those interested in operational performance. But one has to be careful with the data and understand its strengths and weaknesses when analyzing it.
*I was having a tough time with this decision. Obviously, freight and boxes don’t have feelings and won’t get ticked off about a delay, but those flights are still buzzing around and can have an effect on congestion. But in the end I felt that eliminated all-cargo flying was the best for this post.
Latest Comments