WASHINGTON – Census 1002 in downtown Chicago is sandwiched between Michigan and Wabash Avenue, a sleek Trump hotel and a waterfront promenade of cafes and bars. According to the 2020 census, there live 14 people – 13 adults and one child.
Also, according to the 2020 census, they live underwater. Because the block consists entirely of a 700-foot bend on the Chicago River.
If that sounds impossible, that’s right. The Census Bureau itself says that the numbers for block 1002 and tens of thousands of others are unreliable and should be ignored. And he needs to know: the desk’s own computers have moved these people there so that they can’t be traced to their real residences, all as part of a large-scale new effort to keep them confidential.
This paradox is the essence of the debate that is shaking the Census Bureau. On the one hand, federal law requires census records to remain private for 72 years. This guarantee was crucial in persuading many people, including non-citizens and those from racial and ethnic minority groups, to voluntarily provide personal information.
On the other hand, thousands of actors – local authorities, businesses, advocacy groups and others – have relied on the Bureau’s goal of counting “every person, just once and in the right place” to inform countless demographic decisions, from making political maps until disaster response planning for bus stops.
The 2020 census refuted this assumption. The bureau now says its legal mandate to protect the identities of respondents from the census means that some data from the smallest geographical areas it measures – census blocks, not to be confused with urban blocks – must be looked at askance or even neglected.
And users of this data are dissatisfied.
“We understand that we need to protect privacy, and it’s important for the bureau to do that,” David Van Ripper, an employee of the University of Minnesota’s Institute for Social Research and Data Innovation, said in an email. “But I think the production of low-quality data to achieve privacy protection beats the goal of the ten-year census.”
The Census Bureau says its privacy mechanisms are designed to move people only to census units with at least one home. It has been suggested that vacant plots and rivers shown to be hosted by Chicagoans may at one time have had a home, such as a boat or after a destroyed home, or that a coding error mistakenly marked such blocks as such.
It is a mathematical concept called differential confidentiality, which the bureau is using for the first time to cover up census data in 2020. Many users of census data say it not only gives meaningless results like those in Block 1002, but also it may restrict the publication on the basis of confidentiality of basic information on which they rely.
What you need to know about redistribution
They are also dissatisfied with its implementation. Most major changes in the census are tested for up to a decade. Differential confidentiality has been in place for several years, and the release of data already endured by the pandemic has been further delayed due to changes in confidentiality.
Census officials call these concerns exaggerated. They made urgent efforts to explain the change and adjust their privacy mechanism to address complaints.
But at the same time, they say that the large-scale changes that differentiated confidentiality brings are not only justified, but also inevitable given the threat of confidentiality, confusing or not.
“Yes, block-level data has those impossible or unbelievable situations,” said Michael B. House, senior adviser on data access and privacy at the bureau. “It simply came to our notice then. You can think of it as a feature, not a bug. “
And that’s the point. For fans of career data, who are in charge of the census, uncertainty is a statistical fact of life. For their clients, the images of counting blocks with houses but no people, people but no houses, and even people living underwater turned out to be indelible, as if the curtain had been drawn on the demographic Great Oz.
“They shattered the illusion – an illusion that made everyone think that these scores were always pretty good or the best possible,” said Dana Boyd, a technology scientist who spells out his name and co-authored a study by the privacy debate. “The heads of the Census Bureau have known for decades that these small area data had all sorts of problems.
The difference now, she said, is that everyone else knows her.
A bit of history: census blocks – there are 8,132,968 of them – started more than a century ago to help cities better measure their populations. Many of them are real urban blocks, but others are larger and irregularly shaped, especially in suburban and rural areas.
For decades, the Census Bureau withheld most block data for privacy reasons, but backed down as the demand for hyperlocal data became insatiable. A turning point came in 1990: census blocks expanded across the country, and the census began to ask detailed questions about race and ethnicity.
This added detail allowed outsiders to change census statistics to identify specific respondents – say, in a census unit with a single Asian-American mother. The Bureau covered these traces by exchanging such easily recognizable respondents between counting units, a practice called exchange.
But by the 2010 census, explosions of computing power and trade data had passed through that guardrail. In an analysis, the bureau found that 17 percent of the nation’s population could be reconstructed in detail – revealing age, race, gender, household status and so on – by merging census data with even average databases containing information such as names. and addresses.
Today, “any computer science student could do a reconstruction like this,” Mr. House said.
The 2020 Census Solution, Differential Privacy, which is also used by companies such as Apple and Google, applies computer algorithms to the entire census database instead of changing individual blocks. The statistics obtained have “noise” – computer-generated inaccuracies – in small areas such as census units. But inaccuracies fade when the blocks merge into one.
The change has clear advantages for the Census Bureau. While sharing is a rough way to mask data, differential privacy algorithms can be customized to meet exact privacy needs. In addition, the bureau can now tell data users approximately how much noise it has generated.
In the eyes of scientists, statistics on census blocks have always been inaccurate; most users just didn’t know it. From this point of view, the differential privacy makes the census numbers more accurate and transparent – no less.
Outsiders see things differently. An analysis by Cornell University of the latest data release in New York concluded that one in eight census blocks had a statistical deviation, including one in 20 with houses but without people, one in 50 with people but without houses , and one in every 100 with only people under 18.
How US redistribution works
Map 1 of 8
What is redistribution? This is redrawing the boundaries of congressional and state legislative districts. This happens every 10 years after the census to reflect population changes.
How it works? The census dictates how many seats each state will receive in Congress. Map makers then work to ensure that all regions of the country have approximately the same number of residents to ensure equal representation in the House.
Who draws the new cards? Each country has its own process. Eleven states leave mapping on an outer panel. But most – 39 states – have prompted state lawmakers to draw new maps for Congress.
If state legislators can draw their own districts, won’t they be biased? Yes. Guerrilla mapmakers often move the lines of districts – imperceptibly or rudely – to unite voters in a way that progresses toward a political goal. This is called gerrymandering.
Is gerrymandering legal? Yes and no. In 2019, the Supreme Court ruled that federal courts have no role in blocking guerrilla warlords. However, the court left intact parts of the Voting Rights Act that prohibit racial or ethnic manipulation.
Such anomalies will decrease with the improvement of algorithms and the release of new data sets. Some experts say they still fear the numbers will be unusable.
Some civil rights defenders worry that noisy bloc data will complicate political boundaries under the provisions of the Minority Voting Rights Act, although others see no problem. Some experts who draw political maps say they have found it difficult with the new data.
Bloc anomalies were not a problem in larger areas, but they “caused real chaos in city council areas,” said Kimball Brace, whose firm, Election Data Services, serves mostly Democrat clients.
Critics also fear that the bureau may limit the publication of some important statistics to the level of larger districts such as counties, as census block numbers are unreliable.
Mr House, an employee of the Privacy Bureau, said this could happen. But because the differential privacy restrictions are adjustable, “we’re adding some of the lower-level geographic tables based on the feedback we’ve received,” he said.
Such openness is a major change in an agency where privacy is a mantra. Moving to differential privacy may be less difficult if the bureau better answers a key question: “With so much commercially available data, why do we care about protecting census data?” Said Jay Jun. Lee, a scientist from Georgetown University who advises civil rights groups on change.
The answer, said Cynthia Dwork, a computer scientist at Harvard University and one of the four inventors of differential privacy, is that the new era of backward technology and growing intolerance has made privacy restrictions more important than ever.
Loosen them, she said, and census data could reveal tenants of subsidized housing who accept unauthorized accommodation to make ends meet. Or …
Add Comment