Why journalists should never read footnotes**

**Except in all cases. Always read footnotes
I wrote a story recently that I wanted to memorialize because it drove home a fundamental of reporting that I don’t think about enough: When it comes to data, always sweat the small stuff.
It started when I got a changedetection.com alert late on a Friday night (always a good sign) for one of the New Jersey government sites I monitor for new data. The state had published its annual substance abuse treatment demographics report.
The initial report wasn’t too scintillating. Treatment for heroin was once again up, which isn’t surprising given the opioid crisis gripping the state and the nation.
But on the very last page was something new: The state had done an analysis of their met and unmet demand for substance abuse treatment.

This is pretty rad. For one, they haven’t done this kind of analysis in seven years — before the opioid crisis really started having major impacts on the state. For another, Gov. Chris Christie has made substance abuse treatment a cornerstone of his legacy and has pushed hundreds of millions of dollars towards treatment and prevention.
The first thing that struck me is that New Jersey was only meeting 59 percent of demand, which obviously isn’t good. But the weirder thing to me was that since the 2010 study, it only showed demand went up 6,000 people. That’s about how many people have died from overdoses in that time, so I couldn’t comprehend how that could be right.

Then I spotted the footnotes. Number three in particular.

They used a weighted formula to calculate unmet demand. In the previous two studies, the state only used data from the most recent NJ Household Survey. Now, inexplicably, they were using an average of the the last three so that two-thirds of the unmet demand calculation was based on data that was 8- and 13-years-old!
When using the single year data, it adds another 11,000 people to the unmet demand data point and showed New Jersey was actually only meeting a little more than half of its demand.
As it turns out, the 2016 NJ Household Survey had a smaller sample size than the previous two, so creating a weighted formula makes sense, in theory. But creating it in two surveys that were eight and 14 years old? That doesn’t.
I had a way better story than when I started, thanks to reading the last footnote on the last page of a state report that was published without notice at 11 p.m. on a Friday night. Sometimes it’s good to sweat the small stuff.