Pedestrian deaths in Austin and the trouble with drawing conclusions from small numbers

One seldom-discussed drawback with data-driven policing - particularly when its used in mid-size and smaller cities - is that small statistical samples often make drawing valid inferences problematic, and here's a good example: Writes reporter Brenda Bell, "auto-pedestrian deaths are up in Austin, even as deaths from other vehicular collisions are falling. Last year marked an 83 percent increase in pedestrian-bicyclist fatalities; there were 22, compared with a dozen in 2010." Bell reported Sunday that "This year, eight people, including one walking his bike, have already died in collisions with motor vehicles. All but one happened after dark or at dusk. One-third were hit-and-runs."

While certainly it's accurate to say there was an 83% one-year increase, when the numbers are so small that means very little as far as identifying a statistical trend of increasing deaths, and the story didn't include prior years' data. On Sunday Bell did report that "Auto-bike fatalities have averaged one per year over the past seven years," but that means the year you have two it's a hundred percent increase. Is that meaningful data, or just random fluctuation? You see the same issues with murder trends and other infrequent events. Media will report a one-year increase with tremendous drama as a possible trend, even during periods like we're in now when long-term murder rates are declining. When numbers are so small, you need multiple years of data to perform a probative trend analysis, which of course doesn't make for the most attention-grabbing news story.

Perhaps the most obvious death-causing trend evidenced in the stories is that a significant proportion of pedestrian deaths in Austin happen along Interstate 35, which for significant stretches (especially north of 51st Street) is difficult to cross legally on foot and is flanked by extremely busy frontage roads. This datapoint reinforces my sense that jaywalking can be better addressed via traffic engineering solutions that make it safer for pedestrians to get where they're going. If you're walking any significant distance along the access road on I-35, it's likely because you have no other choice.

Another notable trend mentioned is bicyclists being hit after dark, which might argue for efforts narrowly aimed at improving safety equipment. (I see cyclists these days with flashing lights on their bikes instead of reflectors that make them MUCH easier to see after dark.) Cyclists' deficient equipment can be addressed with a variety of tactics from public education, warnings or tickets for equipment violations, improved lighting for dangerous intersections and popular bike routes, etc..

For the most part, though, the city should address these mainly as traffic engineering questions as opposed to law-enforcement concerns. For example, would auto-pedestrian/bike accidents downtown be reduced by creating the proposed Nueces Bike Boulevard? (See here [pdf] for more detail.) Quite possibly. A fascinating and telling map of recent auto-pedestrian/bike accidents accompanying the story shows a large number downtown, though some are concentrated on the east and west entry corridors. Those wouldn't be affected by a north-south route on Nueces, which more aims to reduce accidents on, say, Congress and Lamar.

Instead, Austin PD took a one-size-fits-all enforcement approach, employing a hammer in lieu of a scalpel. In a highly unpopular two-week crackdown last October:

Pedestrians got 1,336 citations and warnings, for offenses such as failure to observe traffic signals, panhandling or stepping into roadways midblock. Drivers got 174.

That lopsided scorecard is par for the course. Between 2008 and 2010, Austin police issued nearly 11,000 citations for violations of auto-pedestrian laws; only 7 percent went to drivers.

Ticketing cyclists may prompt them to get better equipment for biking at night - just handing out lights might even be better - but jaywalking will always be a function of convenience and opportunism. Focusing significant resources on ticketing them IMO is just silly. Meanwhile, as for Austin's much-ballyhooed "3 foot rule" that's supposed to protect "vulnerable" road users?

A city ordinance passed in 2009 requires vehicles to keep at least 3 feet away when passing a "vulnerable user" — a bicyclist, motorcyclist, pedestrian or runner. On four-lane roads, motorists are supposed to yield the lane entirely.

Austin police have issued just eight citations — four per year — under that law.

So that was a bust. Now they're going to try using police as pedestrian "decoys" to ticket motorists. But with 11,000 citations for auto-pedestiran laws over three years and a supposed increase in the problem just now, it doesn't seem like the enforcement-only approach gets to the heart of the problem. Assessing tickets and fines for petty offenses while telling the public it's for their own protection comes off as self serving as it is patronizing in an era when municipal tickets are viewed as a lucrative revenue source.

To be fair, the city does have $13.5 million in bond money available "for projects to enhance mobility and safety for pedestrians, bikers and disabled people," which should do a lot more to prevent deaths than giving pointless tickets to jaywalkers. Certainly it will do more than APD's various proposals for new city bonds at ten times the cost.

Police Chief Art Acevedo told the paper he thinks this trend - which nobody can really say is a trend yet - results from the declining moral character of the citizenry: "Drivers think (pedestrians) don't have the right to cross the street, and pedestrians think they can jump out at the last second," he said. "It used to be ... that when a pedestrian set foot in an intersection, that people stopped. I think it's a commentary on society that we've lost that respect," adding that we're all a bunch of drunks Austin is "one of the heaviest drinking cities in the country."

For Grits' part, I think it's a commentary on Austin PD that, although the chief and the union complain constantly they don't have enough uniformed officers, when they get them they want them deployed ticketing jaywalkers while the civilian APD crime-scene unit doesn't even investigate a majority of burglaries for lack of manpower. And it's perhaps a commentary on the local media that police pronouncements of an "increased rate of pedestrian fatalities on Austin streets" are accepted as valid based on such a limited dataset. (If earlier years' numbers of pedestrian deaths are higher than 12, it might almost seem like the data was cherrypicked.)

These two stories give us insight into Austin PDs traffic enforcement tactics but in many ways for Grits raise more questions than they answer. Once we've seen the number of pedestrian deaths by year, preferably per capita, over a longer period of time, perhaps it would be safer to draw conclusions about trends.