Billings is a city of 110,000 in the southeastern part of Montana. It’s partitioned by sandstone cliffs and punctuated by refineries and casinos. In the middle of town, there’s a lone high-rise from which a row of dive bars, breweries and eateries extends before sprawling into a grid of old neighborhoods with an abundance of intersections that lack stop signs.

On the night beat, when I was a reporter at the Billings Gazette, I would chase scanner calls with photographer Hannah Potes. Neither of us were from the region — we had never seen anything like it — and we’d swear at those intersections as we came to them.

We reported on a number of crashes at those seeming free-for-alls. But the worst happened one Halloween when a Toyota Tacoma hopped a curb and rammed into a tree after smashing into a Mercury Mountaineer. The occupants of both vehicles were fine. But when Hannah and I arrived, there was a man lying on the sidewalk, still. The man had been walking his dog when the pickup ricocheted off the Mountaineer and mowed them both over. Emergency responders worked on him before transporting him to the hospital.

The initial story read like any other. We had cops, neighbors, hospital officials — all the answers we could get. But there were still many questions, like how often crashes there occurred.

Normally, the newsroom would let a story like this go. But I was curious and pursued it.

When I started at the Billings Gazette in 2014, it was my first real newsroom job. I was fresh from attending an IRE/NICAR boot camp and the Baltimore NICAR conference, where I heard the importance of getting data.

One data set I asked for was the state’s crash database. The highway patrol uses it to analyze crash density, identify problem areas and deliver an annual report. Most states have something like it.

I requested five years of data, not sure what I was going to do with it. I envisioned a larger project at some point, looking at the most dangerous intersections in the state. But I figured it was a good thing to have around regardless. It took seven months and dozens of emails and phone calls, but the highway patrol finally forked over the database, for free, a few weeks after the crash. I finally had what I needed to get to the answers nobody could give me.

The file was massive. There were thousands of crashes and dozens of fields. There were codes and IDs that I knew nothing about. But a couple of the fields — type of control and crashes with serious and non-serious injuries and fatalities — looked promising. I thought I would be able to filter, count and show only those accidents that had occurred at uncontrolled intersections.

But that Christmas-morning excitement that comes with a brand-new, gigantic data set quickly faded. Using Excel spreadsheets and Microsoft Access database manager for some quick analysis, I found all sorts of problems that made me realize that what I originally thought would be easy would be anything but.

Some crashes had duplicates. There were blank fields and incorrect entries. There were no street names — just IDs and a dictionary that gave each street a four-digit code, for which there was no apparent logic because each ID varied in length. By cross-referencing some of the fatals with news stories, I determined that what was classified as “no control” sometimes meant that the crash happened in the middle of nowhere — on a straight road, for example, or a rural intersection — where a car drove into a ditch.

So I called the data guy at highway patrol and raised the issues. Even though we had negotiated for months, he was helpful. He explained the IDs to me, and so I was able to add another selector to the mix: all IDs that met a certain pattern — seven zeroes followed by two sets of four-digit codes — represented intersections. I selected those intersections and then was able to quickly count the crashes. It was possible that some of the crashes at uncontrolled intersections had been misclassified. But I had a number that represented the least number of crashes. To confirm it, I contacted the city engineer, who brought up a host of other issues, not with the data — the engineer was able to confirm my numbers — but with how the city engineering office uses it.

When I met with the engineer, I asked him for whatever geographic information system files the office had for stop signs in town. They didn’t have any. I asked him where these intersections were. He had no idea. It turned out the inventory the office had was incomplete and more than a decade old. He said the office relied on people to call in about problematic intersections; they had never done any sort of proactive analysis, mainly because they were understaffed, he said. He also forwarded me an email in which one of the engineers who worked in the field called the lack of data a problem.

Taken together, the interviews and the analysis led to three main findings:

1) Crashes at uncontrolled intersections were more common than you might think; they accounted for at least 15 percent of all wrecks in Billings.

2) The intersection where the man and his dog had been hit had seen four crashes in the last five years, and an uncontrolled intersection up the road had seen 11. Meanwhile, the Manual on Uniform Traffic Control Devices, which sets the national standards for traffic signs, recommends stop signs at intersections that have had five or more failure-to-yield crashes in three years.

3) The city had done nothing to proactively analyze where car crashes were happening. They didn’t leverage the crash data that was collected. And they didn’t know where signs and signals were in the city.

The story, published on my last day at the Billings Gazette, centered on the man who was hit at the uncontrolled intersection and his incredible recovery.


Nick Balatsos is the education reporter for the Casper Star-Tribune in Wyoming. Before that he was the night cops reporter for the Billings Gazette in Montana and the editor of the University of Wyoming’s student newspaper, The Branding Iron. Follow him on Twitter at @nick_balatsos or email him at