Thinking About Road Safety
Can understanding statistical patterns help us prevent tragedies?
A recent series of tragic traffic accidents in the Lakewood area have been devastating and deeply disturbing. Of course, every loss of life is devastating. But fatal incidents involving cars are, in a way, different.
You see, each and every vehicle accident is preventable. If it was too dark to see properly or the roads were slippery with ice or rain, then you should have slowed down or just not been driving at all. If things happened too suddenly for you to react, then you should have been more focused on the road (or not talking on the phone). From a legal perspective, there's always someone who can be held at fault - even if it's the guy who should have done a better job inspecting a decaying bridge.
Part of the rationale behind assigning fault is to help us better understand what happened so we can prevent it happening again. And looking for statistical patterns across entire populations can help us spot dangerous - and correctable - trends. There even seems to be precedent in Chazal for this kind of pattern analysis (Ta'anis 21a):
"A city of 500 like Kfar Amiko that experiences three deaths over three days, one after another, [is considered to be suffering from] dever ["plague"]. [However, if three die] in a single day, or [in the course of] four days, then it is not dever.
Rashi: "In a single day it is not dever. As this is coincidental [אקראי בעלמא]."
So is there anything we can learn from traffic incident statistics? Is Lakewood more dangerous than comparable neighboring communities? Is there something in particular that those who live there should be doing do make their streets safer?
New Jersey Fatal Accident data (2021)
When it comes to New Jersey accidents involving the loss of life, the numbers from Ocean County (which includes Lakewood) were rather high, but not high enough to suggest significant differences.
This data isn’t nearly precise enough to be useful for us. We’ll have to see what else there might be.
The Countrywide Traffic Accident Dataset (2016 - 2021)
Instead, let’s explore a well-known public dataset that's made up of:
"Traffic data captured by a variety of entities, such as the US and state departments of transportation, law enforcement agencies, traffic cameras, and traffic sensors within the road-networks. Currently, there are about 2.8 million accident records in this dataset."
Now in the context of what we’re looking for, this data has some serious limitations. For one thing, the dataset is far from complete. Due to the nature of the data, not only are many - perhaps most - accidents missing, but there are significant differences in how they're reported from one region to another.
But we're not really looking for absolute numbers. Instead, we're interested in accident rates that can give us a sense of how one community's experiences might be noticeably different from the average.
Of all data nationwide, there were around 58 accidents through this period in the average Zipcode. Although the median Zipcode (i.e., the Zipcode whose rate was higher than exactly 50% of all others) had only 11 accidents. This strange disparity between average and median values suggests that there was a relatively small number of Zipcodes (mostly in California, as it turns out) whose unusually might numbers skewed the rest of the data. This can probably be attributed to unusual reporting methodologies.
So were there significantly more accidents in Lakewood (08701) than the national average? The short answer is "yes." According to the Countrywide Traffic Accident Dataset, there were 91 accidents of all levels of severity during the reporting period. That's more than eight times the median, and around 55% more than average.
But how useful is it to compare Lakewood to all 40,000 US Zipcodes? After all, the environmental conditions, reporting systems, and populations of locations in, say, North Dakota, are hardly comparable to central New Jersey.
So I'm going to borrow from the methodology I used for my article "The Mystery of Jewish Economics" where I compared the Jewish Zipcodes to just the 514 closest neighboring codes. That might give us a better apples-to-apples comparison.
In fact, Lakewood's 91 accidents were much closer to the regional average of 75. Does that mean there's nothing all that unusual about Lakewood's record? Perhaps. But let's also look at the numbers for some other frum Zipcodes in the New York/New Jersey region:
It’s probably safe to ignore the Flushing total, which is more likely the result of reporting anomalies than insanely bad driving habits. As far as the others, just Lakewood and Spring Valley stand out.
Is there anything those two areas uniquely share in common? Intuitively, I’d suggest that they’re both notable for unusual combinations of small-town rural infrastructure and big-city density and pressures. In other words, there are far more people impatiently moving about than the existing roads and traffic control systems can handle.
Of course, we can all use reminders about driving while tired or distracted by kids and phones. And perhaps public service announcements are appropriate. But there may indeed by some underlying structural problems that could also be addressed.
Any thoughts of your own? Please share.