The Michigan Department of Environmental Quality has been under a lot of scrutiny ever since it was revealed that Flint had a problem with elevated lead levels in its water supply.
The agency oversees how the city of Flint manages its drinking water. And when they made the switch from Detroit water to water drawn from the Flint River, city water officials relied heavily on guidance from the state.
So shouldn’t the state have known about the lead problem?
And why did it take tests from independent scientists to finally push the state to admit there was a problem?
The lead tests and math
One way a city knows whether they have a lead problem is by federally mandated tests done under something called the “lead and copper rule.”
A city is supposed to find out where citizens might be at most risk.
The city of Flint has something like 15,000 water service lines that contain lead. And many homes might also have lead pipes or lead solder in them.
So a city is supposed to identify those homes most at risk, and test the water coming out of those taps.
For a city like Flint with around 100,000 people, the requirement has been that they take at least 100 samples from homes like these, calculate the 90th percentile of those samples, and determine where they stand.
If their sampling shows that the 90th percentile calculation is above 15 parts per billion, the city has to take action.
They have to warn residents that there is a problem, and they have to take steps to address that problem – by treating the water so it won’t corrode pipes, and by potentially spending money to replace old lead service lines.
An altered test?
Virginia Tech’s Marc Edwards, a water quality expert who was asked to examine the water by Flint citizens, charges that the state altered this test so officials could say the city was still below the federal action level of 15 parts per billion.
The state denies this.
But when we looked at how these numbers were calculated, something stark stands out.
Flint officials tried to collect 100 samples, but they only managed to get 71 samples. And these weren’t “worst-case-scenario” samples. These were random samples. But even with these 71 samples, they were over the federal action level for lead.
After the city submitted the samples, the state instructed the city to drop two samples for violating testing guidelines.
Once these samples were dropped, the remaining 69 samples had a 90th percentile lead result that fell below the action level.
Let's go back to math class
Computing a percentile is confusing – especially if you haven’t been through a stats class recently.
I heard two state officials in the Michigan Department of Environmental Quality give an incorrect explanation to one of our reporters.
The latest was Jim Sygo. He’s the Deputy Director of the MDEQ. He's also the interim division chief for MDEQ's Office of Drinking Water and Municipal Assistance. He took over for Liane Shekter Smith who was reassigned after the Flint lead problems came to light.
Here's how he described how to do the calculation under the lead and copper rule.
“Well again, usually you have a set of samples, you take the 10% of samples that had the highest numbers, and you drop them. That leaves another set that still has numbers and based on that, you do the average."
As best I can tell, there’s nothing in state or federal regulations that says you should drop the top 10% of numbers with the highest samples.
Maybe this is all just confusion over how a percentile is calculated.
So to show us how it’s done, and to demonstrate how dropping two numbers changes a 90th percentile calculation, we asked the head of Eastern Michigan University’s math department to explain it to us.
You can watch the video above to see how it's done.
And we didn't stop there. We looked at how dropping just one number in the elevated samples would have changed the calculation. Doing that would have put the 90th percentile calculation at 15.5 parts per billion – that’s still above the federal action level.
And dropping any of the lower samples out of the pool would push the calculation for the 90th percentile up.
That’s just how the math works out.
And whether it was intended or not, dropping those two samples changed how the state could talk about lead levels in Flint's drinking water supply.
*This post has been updated.