State News

Census change could inaccurately report small Vermont community counts

By Erin Petenko/VTDigger

A little-known proposed change to how the Census reports data could have a major impact on Vermont’s small communities, changing population totals in some towns by 10% or more.

Michael Moser, who heads the Vermont State Data Center, is among a group pushing back against a proposal that would alter Census data in an effort to protect individuals’ privacy, but which could result in substantive changes to the count totals in small towns.

After an analysis of 2010 data found that individuals could be identified through Census information, Census Bureau officials developed an algorithm that would randomly change the count by small amounts in some places.

But Moser, like other local Census leaders, say that the algorithm would alter the results in Vermont. That could impact federal funding, election districts and the fundamental understanding of the state’s demographic trends.

“From what I’ve seen, it could drastically change the characteristics of our geographies,” he said. “We’re one of the smallest populations in the country, so we could lose or gain an entire subset of our population.”

What change is the Census making to the data?

The proposed change would not impact how the Census is carried out, but it would tweak the count released to the public. After the Census is done counting Americans in July, it will collect that data and begin to publish the results on Dec. 31, 2020.

The Census is constitutionally required to release this data to the government so it can correctly distribute electoral votes and build congressional districts. That data will likely include a total breakdown of the population, along with a breakdown by gender, age and race, among other factors — just as it has in previous years.

But for the first time, the Census Bureau plans to use an algorithm to introduce “noise” to the data, adding or taking away small amounts of people to the final numbers, said Peter Nelson, professor of geography at Middlebury College.

“When you’re looking at something like the size of the city of Burlington, or the state of Vermont, that noise probably isn’t going to have that big of an impact,” Nelson said. “But for smaller units and more rural places, it could lead to wildly higher estimates or lower estimates.”

How would that affect Vermont communities?

The Census has already published some data to show what that might look like: It used its own model for the 2020 figures on the existing 2010 data to give researchers an idea of the effect.

VTDigger’s analysis of those results found that for small towns and villages, the total number of people reported could swing by 10% or more.

With this algorithm, the more researchers drill down into subsets of the population — people of color, people over 65 — the more the minor fluctuations in the data are going to alter the count people see, Nelson said. That matters when researchers are trying to study those specific demographics.

Still, it’s too early to say exactly how much of an impact the change will have. Moser said the algorithm was like a “dial” that the Census Bureau could use to change the results by a little or a lot.

Jason Broughton, chair of the Complete Count Committee and state librarian, said the Census needed to communicate this change to those small communities so they could understand better how it would affect them.

“If it is too rigid and inflexible, and people on the ground who are going to be impacted aren’t a part of that conversation so that they can say ‘yes, there are people here.’ … it will show on the record that there’s nobody living in an area, or that there’s not that many people in that town,” Broughton said.

In a statement to VTDigger, the Census Bureau said it was continuing to take feedback from researchers about how to implement the change.

Why is the Census Bureau doing this?

The Census is required by law to ensure confidentiality to participants.

“One of the mandates of the Census is that it does have to make sure that confidentiality is almost 100%, if they can actually do it,” Broughton said.

The Bureau has used other strategies to protect individual data, such as refusing to release data for very small populations and limiting access to other data to qualified academics. But with data science techniques becoming more sophisticated, it has become more challenging to prevent determined individuals from unlocking more detailed information.

In 2019, researchers at the Census reported that with certain specialized techniques, they could pinpoint specific people in the 2010 data and find out their age, gender, race and location. But there’s no evidence yet that people are doing this outside the Census Bureau.

If a person reports she is one of a handful of women over 40 living with two children under 10 in a small Vermont village, someone could look at the front-facing data and combine it with publicly available records to deduce who that person is — and then guess more personal details, like stated ethnicity.

In an email statement, Daniel Velez, a spokesperson for the Census Bureau, said it’s implementing the algorithm precisely to protect places like Vermont’s small geographies. The new strategy has an advantage over old methodology, like suppressing data, because the Census can now release more detailed data for towns and neighborhoods, he said.

Nelson said it is ironic that people are worrying about the federal government accidentally revealing this information when private companies get far more data about you from your own devices.

“Nobody ever questions when they click on the app on their phone as to whether they agree to the terms and conditions,” he said.

Capstone Community Action, an organization that’s helping with Census efforts, has encountered concerns about the privacy of Census data for another reason: the proposed, then scrapped citizenship question, said Sue Minter, executive director of the organization.

The Trump administration asked the Census to include a question about whether respondents are U.S. citizens. It was later struck down by the Supreme Court, but many believe the proposal still may have a chilling effect.

“I think the Trump administration, by putting out there the idea that people would have to mention their citizenship status, even though that was overturned by the courts, has increased the fear,” Minter said.

Why does it matter?

The distribution of $2.5 billion could be at stake. That’s how much money goes to Vermont in a single year from federal government programs that rely on Census data.

That’s why Minter and other community organizers are working so hard to get Vermonters counted: They want to make sure every possible dollar is going to the state to support the people that need it.

Capstone uses federally funded programs like Head Start, which provides early childhood education, and fuel assistance for low-income Vermonters, said Eloise Reid, Capstone’s Census coordinator.

“For example, say you have 100,000 kids in Vermont, but they only count 50,000 of them,” Reid said. “Then you only have money for the services for 50,000. But you’re getting 100,000 kids needing those services.”

Moser has been discussing the potential problems with key policymakers and experts, he said. Letters will start arriving in Vermont households on March 12.

Mountain Times Newsletter

Sign up below to receive the weekly newsletter, which also includes top trending stories and what all the locals are talking about!