1.Introduction

This is part of my research assistant work for Professor Megan S. Ryerson at the University of Pennsylvania who is conducting a comprehensive investigation on the disparity of air service in terms of frequency, direct service, and air fares across across airports within a megaregion. Such disparity leads travelers to drive long distances to access busy airports with lower fares and better air services instead of the nearest local airport, which is defined as “customer leakage” problem.

My work in this project includes:

  • Analyze the DB1B data in R regarding flight frequency, direct service, and air fares.
  • Perform Network Analysis in ArcGIS to find out the distance between each census tract and each selected airport in the studyarea.
  • Calculate the utility of the selected airports in the study and plot the airport catchment map showing each airport’s market catchment. (This is initially done in ArcGIS but has been automated in R by my collegue and also my classmate - Eugene Chong)
  • Convert the Highest Market Share map into a cartogram based on the population in each census tract to better reflect the catchment volume.
  • Analyze data from ARC(Airlines Reporting Corporation) to calculate approximate percentage share of total passengers using the selected airport in the selected year in each zipcode.
  • Use data from airport’s annual report and Network Analysis in ArcGIS to make the extramile maps to visualize how many extra miles were actually traveled by all the “leaked” customers in the selected year.

2.Data

  • Airline Origin and Destination Survey (“DB1B”): a 10% sample of all airline tickets sold by reporting carriers published by the Bureau of Transportation Statistics (“BTS”)
  • Ticket Purchase data from Expedia including zipcode of customer’s billing address from ARC (Airlines Reporting Corporation)
  • Annual domestic enplanement from each airport’s annual report.

3. Method

In the first half of the analysis, each airport’s direct service utility and connecting service utility in each census tract were calculated using parameters including flight frequence of the selected destination, air fares, distance between each census tract and the airport, and total flight time (including dwelling time for connecting services). The market share of each airport was them computeted as:

e^(utility of selected airport)/sum(e^(utility of each airport in the study area))

In the second half of the analysis, I used ARC data to calculate the dominant airport’s percentage share of passengers in each zipcode in the study area. We made the assumption that the spatial distribution of ticket purchase in Expedia could well represent that of all the ticket purchased. The percetage share of airport i in zipcode j is calculated as:

#of ticket purchase using airport i in zipcode j/#of ticket purchase using airport i in the study area

Lastly, I used the total domestic enplanement data to calculate the approximate number of passengers in each zipcode using the selected airport and how many extramiles they traveled by those customers “leaked” from their nearest airport.

4. Southeast Megaregion

One of the study areas is the Texas-Oklahoma-Louisiana megaregion in which we were looking at eight airports (AUS, DAL, DFW, HOU, IAH, OKC, SAT, and SHV). We picked two destinations (SEA and ATL). This page only shows the analysis for flights from the selected airports to ATL.

4.1 Market Catchment

4.1.1 Highest marketshare

The map below shows which airport has the highest market share on flights to ATL in 2015 for each census tract in the study area. It is interesting to see that although SHV is an airport in this region, it does not have the highest market share for any tract in the study area.

The market catchment area in the map above is based on land area. However, that in the cartogram map below is based on population count. Such difference is critical as we could see the although the catchment area of DAL and HOU area small in the map above, they are actually capuring great amounts of people as shown in the map below.

4.1.2 Individual market catchment

Maps below indicate market shares in 2015 for each of the eight focused airport on flights to ATL for each census tract in the study area.

4.1.3 Discussion

The histogram below shows the counts of market share values for each airport on flights to ATL in the tracts where that airport has the highest market share.

It can be observed that airports like AUS and SAT do not have highest market share in a lot tracts - in other words, they are the top choice in only a few tracts. Airports like DFW and IAH have the highest market share in many tracts, which corresponds to the market catchment map. However, the counts are very left-skewed, meaning that in many tracts where they are the top choices, their market share is not actually very high. This suggests that they are probably the “best-worst” option for passengers in distant census tracts.

4.2 ARC data

4.2.1 Percentage share of total passenger

As mentioned in Section 3, I used ARC data to calcualte the percentage share of passengers using the selected airport in 2018 in each zipcode area. The selected airport is DFW as it is considered the dominant airport in the study area. The persencate share map is shown as below.

4.2.2 Extramile travelled

In ArcGIS, I used Network Analysis to calculate how many extramiles an individual person in each zipcode area would travel if he/she choose DFW over the nearest airport. The calculation is defined as:

individual extramile in zipcode i = distance to DFW - distance to zipcode i’s nearest airport

Therefore, extramile = 0 in some zipcodes when the nearest airport for them are DFW. The map of individual extramile is visualized below.

Muliplying the total domestic enplanement of DFW (from DFW’s annual report) in 2018 by the percentage share, we could get the approximate number of passengers using DFW in each zipcode (we assumed that all the passengers using DFW came from the study area). Muliplying the passenger number in each zipcode area by the individual extramile, we could see how many extramiles were actually travelled by those passengers choosing DFW over their nearest airports. The map is shown below.

5. Note

As mentioned before, another destination picked for this analysis is SEA (market catchment map shown below).

Moreover, we conducted the same analysis for the Piedmont Atlantic region as well with full sets of maps. The two destinations are DCA and ORD (market catchment maps shown below). Additional maps might be included in the future but they are available upon request.