The deep state’s algorithmic abyss

Almost a full year into President Donald Trump‘s second term, the full scope of the federal government’s woke and weaponized posture toward its citizens is still unfolding.

Beyond the known targeting of parents at school board meetings by the Biden administration, the billions in taxpayer-funded grants funneled to woke institutions, the intentional deception by federal health authorities over the safety and efficacy of COVID-19 vaccines, the eyebrow-raising identification by FEMA of Trump supporters in the aftermath of natural disasters, and the political persecution overseen by former Attorney General Merrick Garland against conservatives by federal law enforcement agencies, the weaponization extends even deeper into the most mundane agencies and federal operations.

The United States Census Bureau, a constitutionally authorized agency designed to carry out a simple count of Americans every decade for apportionment and redistricting purposes, has also been captured by liberal ideologues. Under the former guidance of John Abowd, a left-wing Cornell economist who was the chief scientist of the bureau, the agency adopted a sophisticated algorithmic mechanism to mask the characteristics and personal information of individuals. Known as differential privacy, this algorithm intentionally distorts both characteristic data (i.e., someone’s race, biological sex, age, etc.) and numerical data (i.e., the actual number of people in a census block). 

INDEX FUNDS AND BIG TECH ARE WORKING TOGETHER TO HURT LITTLE TECH

Utilized for the first time in 2020, this algorithm was imposed all the way down to the lowest component of census operations: the block level. Imagine transferring the racial, sex, and age characteristics of every individual on your street, mixing them to “create” new people out of thin air and “erase” real people, artificially altering the number of residents on your block either up or down, and then moving this new fabricated population into an entirely separate county somewhere else in your state. Then, doing that for every street, block, and neighborhood in America.

This is differential privacy. That was the 2020 Census.

It is an intentionally deceptive practice, intended to be opaque, so that only a few people at the bureau know the accurate headcounts and characteristic data, and it operates in a way that dilutes the ability to determine the real composition of the population in voting districts, including the number and location of illegal immigrants.

Statutorily, the bureau is required to protect individuals from personally identifiable information. This requirement dates back to a 1954 law and has animated census operations for decades. Purportedly, the 2020 census implemented differential privacy to supercharge this approach by ensuring that individual characteristic data cannot be reverse-engineered. Previous methods to protect people’s identities typically resulted in “swapping,” wherein the characteristic information of one’s neighbor was flipped with someone else and moved to a different house. Differential privacy creates entirely new people, erases real people, moves populations across voting district boundaries within states, and then scrambles the datasets so that no one outside the designated few within the bureau knows what is really happening.

The practical impacts are immediately obvious. It distorts the composition and number of people within voting districts, which state lawmakers rely upon for drawing accurate electoral maps. It requires federal agencies to enter into memoranda of understanding with the bureau to ascertain accurate information for federal funding purposes. And it disproportionately affects rural areas of America because the population density is less, which means inaccurate numerical data has a larger possible negative effect in conservative areas than it does in urban progressive areas. 

Even some left-leaning experts at Harvard take issue with the practice, and concluded that differential privacy “makes it impossible to accurately comply with the One Person, One Vote principle as currently interpreted and implemented.” Further, the Harvard study found that differential privacy, “has a tendency to transfer population across geographies in ways that artificially reduce racial and partisan heterogeneity.” Translated, it takes the one job of the Census Bureau — accurately counting people — and ensures that accuracy cannot be achieved.

More concerning is that the liberals within the bureau stated the quiet part out loud. In March 2023, the bureau defended its decision to utilize differential privacy to protect the personal information of transgender individuals and illegal immigrants. The bureau cited a mathematical model out of the University of Washington that shows the relative ease to “reveal the identity of people who are transgender in previously published tables.” It also conducted an experiment that reconstructed the data to identify individuals of Hispanic origin and stated that the results were “concerning” and, if republished, would have violated confidentiality protections.

Of course, this is a solution in search of a problem. There has never been a serious threat of entities reconstructing census datasets to identify and target specific individuals. It is a mostly theoretical problem that has facilitated an elaborate scheme to reward liberals disproportionately. Setting aside the broader implications of differential privacy, the 2020 census was already the most inaccurate and error-riddled in modern American history. 

A post-census evaluation released by the bureau in 2022 admitted that six states were undercounted (Arkansas, Florida, Illinois, Mississippi, Tennessee, and Texas) and eight states were overcounted (Delaware, Hawaii, Massachusetts, Minnesota, New York, Ohio, Rhode Island, and Utah). The result was an illegitimate net gain of at least six congressional seats for Democrats in the House of Representatives. The fact that these errors resulted in significantly rewarding a single political party at the expense of the other is “mathematically mysterious.”

Whether these were errors due to agency incompetence, COVID hysteria, differential privacy, or a combination of all three factors remains a hotly debated topic within census and academic circles. What cannot be debated is that the mistakes benefited one party.

IF YOU CHOOSE TO LIVE IN AMERICA, YOU SHOULD LOVE AMERICA

As lawmakers engage in mid-decade redistricting efforts, the full extent of the distorted maps and numbers remains unclear. Further, without access to department memoranda, it remains unclear how much federal policy is operating on faulty differential privacy datasets that may be distorting funding streams. And unless this algorithm is thrown into the abyss alongside other weaponized ideas and programs, the people will simply not be able to trust their own political representation.

Couched as a privacy protection tool, differential privacy is in reality a liberal primacy tool. It ensures that adding the citizenship question to the census is rendered ineffective, algorithmically tips the scales toward rewarding liberal areas of the country over more conservative areas, and fundamentally manipulates datasets that alter the ability of states to ensure accurate political representation. Both the Trump administration and Congress must end this practice now, before it distorts the 2030 census and trust in our institutions is further eroded.

Drew White is a policy adviser at the Center for Renewing America and the founder and CEO of Palisade Policy Group.

Related Content