Today's Paper Weather Closings ❄️ River Valley Democrat-Gazette Newsletters Obits Public Notices Distribution Locations Digital FAQ Razorback Sports Puzzles Crime Today's Photos

Data scrambling by census raises concern in state

Privacy measure draws look by Doug Thompson | March 20, 2022 at 3:44 a.m.
An envelope containing a 2020 Census form and letter mailed to a U.S. resident is shown in this March 19, 2020, file photo. (AP/Matt Rourke)

A piece of property northeast of the Springdale city limits has two ponds, two sheds and -- according to the 2020 U.S. census -- five residents.

This fictional occupancy is not a mistake, said Jeff Hawkins, executive director of the Northwest Arkansas Regional Planning Commission. The Census Bureau deliberately added fake residents to some addresses and "subtracted" real residents from others in its publicly released data to protect households' privacy.

"My favorite is how there's supposedly three people living on the grounds of the state Capitol in Little Rock," Hawkins said of the scrambling.

The bureau insists in public statements that, on average, the data is accurate. But this scrambling of results is more likely to create problems the smaller an area a planner or researcher goes or the more detail he seeks, Hawkins said.

The situation is worse for researchers, said Mervin Jebaraj, director of the Center for Business and Economic Research at the University of Arkansas, Fayetteville. The problems will grow, the two men said, especially since the Census Bureau plans to apply the same approach to the American Community Survey. The survey is an annual estimation that provides more detailed and up-to-date data between census takings. A full census is done every 10 years.

"If you're trying to divide a city into wards of equal population, this might not be a problem in Springdale with 22,000 people per ward," Hawkins said. "But if you're doing the same thing in Tontitown with 1,400 people a ward, it causes problems. And some of the smaller towns might have 400 people a ward.

"The smaller the area, the bigger the problem."

Larger cities will face uncertainty too, Hawkins said. For instance, many federal grants aim to help low- and middle-income neighborhoods, he said.

"I've seen those grants used for everything from fixing streets to buying a firetruck," Hawkins said.

Census numbers might falsely show a project would help low- and middle-income residents, but that might not be true, he said.

The scrambling for privacy already causes severe problems for economists, researchers and businesses, Jebaraj said. For instance, detailed results showing where minorities live in fast-growing Northwest Arkansas cities cannot be relied upon, he said.

By law, the Census Bureau must keep individual census responses confidential. The Internal Revenue Service, for instance, cannot get a household's census data on income. But modern computerized number-crunching can sift through detailed census results and identify at least some households and its particulars within a census block, the bureau says. If only a few residents report either high or low household incomes, for instance, data analysis of the raw data collected can pinpoint where they live.

A census block is the first and smallest unit in which census household responses are compiled. Most census blocks contain 250 or fewer residents, according to the bureau.

Any errors in the data stemming from the "disclosure avoidance system," as the scrambling is called, is negligible at the county level and above, the bureau's analysis concluded, according to a Jan. 28 report.

Hawkins and Jebaraj did not dispute that county-level data averages out. Their contention is that county-level data is of little help to a school board trying to decide where to build an elementary school to serve a growing minority community or a businessman deciding where to locate a shop serving a specific group such as recent arrivals to the region, they said.

"We very much rely on data to make anything work," Jebaraj said.

The Census Bureau used to switch some addresses between real people to assure privacy. The system worked, Jebaraj said.

"No one has ever breached it," Jebaraj said of the privacy of U.S. census data. "There is no case of it."

Even if it were theoretically possible to identify individuals from census block data, he said, it would be far cheaper and quicker to buy such information from private sources that collect it in the normal course of business, he said.

One of the worst effects of not having a clear picture below the county level is the inability to spot the start of a trend, Jebaraj said. For instance, Northwest Arkansas recently saw South Asians coming to the region, he said. A slight increase at the county level could be a very large increase for one of the smaller communities in that county, he said. Spotting such a trend in any accurate way useful to researchers and planners would not be possible with the quality of data as it is now, he said.

"We can go back," Jebaraj said. The original, unfiltered data that all publicly released data is based upon is still at the bureau's disposal. Either the bureau, Congress or the courts could have the bureau access it, he said.

Print Headline: Data scrambling by census raises concern in state


Sponsor Content