How Far Are People on Average from Their Nearest Decent-Sized Airport?

Table of Contents

• Motivation
• Nationwide Results
• State-by-State Results
• Data & Methodology
• Airports Used
• Comments

My girlfriend has been interviewing for academic jobs. Usually she needs to fly into an airport and then drive for an hour and a half or two hours to get to the school. This made me wonder: do most of the country live like this, that far from a major airport? Has my perspective been warped by constantly living on the coasts near major metropolitan areas?

I decided to analyze the data and answer my question.

Most people live a reasonable distance from a decent-sized airport. Half the people in the United States live within 17 miles of a decent-sized airport, and ninety percent of the country lives within 58 miles (about an hours drive). Twenty-five percent of the population lives pretty darn close: less than 9 miles.

The nationwide data for us statistics junkies:

This map displays the major airports in the United States. It shows how roughly the country can divided according to which airport is the closest as the crow flies.

Some interesting features emerge if you look at the individual airports that are useful for this nearest-airport metric. Because we only care about the nearest as-the-crow-flies airport, LaGuardia and Newark do a much better job of covering the New York metro region than JFK. (See the New York City metro region map or the general New York state page, also linked below.) Likewise, Oakland and San Jose are more useful than San Francisco. (See the San Francisco bay area map or the general California state page.) A surprising contrast to this pattern is Chicago, where Midway and O'Hare roughly evenly spit the Chicago metro area. (See the Chicago map or the Illinois state page.)

Metropolitan area maps with population data:

• New York City Metro Area
• San Francisco Metro Area
• Chicago Metro Area

I also studied the breakdown by state/territory.

Of all the regions in the United States, the residents of the District of Columbia live the closest on average to their nearest airport. This is true regardless of whether we're talking about median, 90th percentile, or maximum distance to the nearest airport.

Likewise, residents in compact states such as Rhode Island tend to be close to their major airports. Rhode Island is in the top-ten-closest list when sorting by median, 90th percentile, or maximum distance.

Surprisingly, large, spread-out states such as Alaska and Hawaii tend to also be in the top ten for closest when talking about medians. This is because most of the population in those states live very close to the airport, closer than in other states.

The states with residents farthest from their nearest major airport on average are Wyoming and North Dakota. The median distance for a resident of Wyoming to his or her nearest major airport is 104 miles. That's quite a drive. It's so high because no airport in Wyoming meets the number-of-passengers criterion to be considered a major airport.

Summary statistics for each state. Click on the state name for details and a map.

I used census data and airport data to answer this question.

I downloaded census data that lists, for every census tract, the number of people in the tract and its location. Census tracts are the smallest geographical unit used by the census. On average they contain 4,000 people; the typical range is 1,200 to 8,000 people. Census tracts do not cross county boundaries. Hence, every tract is at most the size of a county and most are much, much smaller. I believe that these tracts are small enough for my purposes.

The census lists the location of each tract as a (latitude,longitude) coordinate. This location is the centroid of the census tract. If the centroid of the tract is outside the region (for instance, if the region looks like a crescent moon), the census changes the tract's location to be the closest point within the tract to the computed centroid. All this sounds like a reasonable way of assigning a census tract a location. I can't imagine the centroid-projection is necessary very often, and I believe that tract are usually small enough that any plausible method is probably good enough for my purposes.

For airports, I downloaded from wikipedia the list of airports along with the number of passengers who traveled through each in a year. Wikipedia says they got this data from Federal Aviation Administration records.

I restricted my set of airports to those I considered decently-sized, which I defined as more than 100,000 passengers a year. This is pretty generous. If an airport lands the equivalent of one full 747 a day, at about 500 passengers each, and no other planes, it would have 182,500 passengers a year, well more than the threshold. See the list of airports used to check whether an small-ish airport you have in mind is on the list. My small-ish example, the airport in Medford, Oregon, has more than 200,000 passengers a year. It makes the cut.

I know not everyone flies from his or her nearest airport. Hence, you should think of these distances as lower bounds for distance / travel time. For example, the nationwide median distance is 17 miles and the 90th percentile is 59 miles. This means that, regardless of whether some people don't fly from their nearest airport, fifty percent of the population lives more than 17 miles from their nearest airport and ten percent lives farther than 59 miles.

Likewise, if you're uncomfortable with me including many small airports, realize that shrinking the list of airports increases distances considerably. Again, the distance estimates can be seen as lower bounds for actual distances to nearest useful airport.

For airport locations, I found on the web a mapping from airports to their (latitude,longitude) coordinates.

To compute distance from the census tract to an airport, I used the Haversine distance formula. It computes distance as the crow flies on the surface of a sphere (such as the earth, which is roughly spherical).

Regarding the maps, some census tract flags may look like they are colored for one airport's region but are actually in another airport's region. The census tract flags are colored with which airport they are closest to as the crow flies along the surface of the earth (the Haversine distance formula). The region boundaries that are displayed on map were computed assuming assuming the earth is flat. (The lines that divide regions are a Voronoi diagram, and I couldn't find any algorithm that calculate the Voronoi decomposition on a sphere such as the earth.) In short, although a few census tracts may look like they're mislabeled near a boundary, it's actually the boundary that is slightly wrong.

I wish I could have calculated driving times from each census tract to each airport (rather than distance as the crow flies). Google Maps API can calculate driving trips. However, the API's terms of service prohibit using the API without showing a resulting map; I'm not allowed to write a script query the Maps API and use the results only to build tables of statistics. :( If anyone knows of a driving times API with broader terms of use, please tell me.

I could only do this analysis for the United States because of the sources of my data (census data, airport usage data).

Have something interesting to say? Share it!

Thanks for the Maps Icons Collection http://mapicons.nicolasmollet.com for the icons used in the embedded map image above.
© Mark Pearson 2012