Wednesday, August 28, 2013

Football Game-Day Traffic Fatalities

Over ten seasons, it appears that the overall number of local traffic fatalities is not impacted one way or the other by a pro football game happening in town.  But if you look at individual cities, there are some interesting examples at the edges.

To be clear, this is not a map of traffic fatalities vs. drivers who didn't die (that would be hard).  Nor is it a map of game-day fatalities vs. all other fatalities throughout the year (that would be unfair).  It is just a comparison of a stadium's region when the home team is in town vs when it is playing somewhere else (which is a pretty tight control group) -and the different traffic conditions that implies.
This map does not imply that a football game has anything to with any individual traffic fatality, and over the course of 80 home game-days and 80 away game-days most locations tended to be generally even (to my initial surprise, given the tens of thousands of vehicles making a dedicated pilgrimage).  But throughout ten seasons, some cities start to build up a surprisingly lopsided count.

Why? I don't know.

Because I'm comparing home game-days with away game-days, the seasonal and day-of-week variabilities in traffic risk should fall away (group A and group B happen throughout the same time of year and generally happen on the same days of the week).  The biggest variation that the data accounts for is whether 60 thousand or so people simultaneously converge in a single place and then depart.  In many cases, the number of overall game-day local traffic fatalities was too low (thankfully) to support statistical rigor in a home/away comparison.  But not for all locations.

Before I started putting the data together, my hunch was that home games would correlate to higher traffic fatalities.  That just seemed to be the intuitive notion.  I only thought of it because of a really treacherous snowy driving experience I had getting to and from a game.  But the more I thought about it, and as the results started bucking that assumption, I reconsidered.  While on one hand you have a massive migration and perhaps more traffic (an increased candidate population), on the other hand you have 60+ thousand people who are not driving for three hours and a carefully choreographed herd of comparably slow moving vehicles (which will have traffic incidents but not necessarily fatal ones) when they are driving.  Add to that local football cultures that have a strong tailgating community and you have an even longer stretch of non-driving (does tailgating save lives?).

So what accounts for some cities having a rather large proportion of home game-day incidents while others have significantly less in comparison?  This has to be a pretty complex question involving local culture (tourism footprint, tailgating culture, proximity and travel mode of attendees), the proportion of a local team's impact on the transportation infrastructure, the nature of that transportation infrastructure, and the availability of public transportation.  Even over ten seasons, do some locations' other sporting venues bias the proportions one way or another?  And who knows what else.

Where are the Bay Area, NYC, and DC teams? The nature of this map looks at an individual team's local area.  When local areas overlap with other local areas it muddles the home and away differentiation.  So they could not be considered in this map.


  • Local drive-time areas were generated by Visual Fusion at a "distance" on one hour (non-traffic) based on an underlying OpenStreetMap transportation network.  You can download these geo-amoebas directly, here (this link will prompt to download a zip file, about 280k).  Teams that had geographically significant stadium changes within the ten-season study period are accounted for (Dallas, Philly, Arizona, Detroit, Seattle). Other teams got new stadiums but they neighbored their predecessor so it didn't impact the drive-time area.
  • Historic team schedules (including home/away info) were gathered from the Pro-Football-Reference website.
  • Historic traffic fatality data was downloaded from the National Highway Transportation Safety Administration's FARS database.  Unfortunately, data for 2012 has not yet (at the time of map making) been released, so the whole 2011 football season had to be excluded because of games that occurred on Jan 1 of 2012 (to exclude only that week of football would make the number of home and away games lopsided).

This map was made in collaboration with cartographer, Josh Stevens, who tweets as @jscarto.

P.S. If you feel raw about this map, maybe because your city has what appears to be a comparably dangerous home game day, it may be just as likely that the city is especially safe on away game days.

No comments:

Post a Comment