How the Census protects your privacy

By Nathan Thomas at March 16, 2010 - 10:08am
Redistricting Updates

How the Census protects your privacy

Ever since Barack Obama won the presidency, right-wing paranoia about the Census has run rampant. But what few people realize is that Census officials themselves are equally paranoid about protecting the privacy of the information we provide. Protection begins with a 72-year seal on personal information in our Census responses:

Doris Turner, the partnership specialist for the Census Bureau serving several counties including Coweta, emphasized the confidential nature of Census form data in a recent meeting with the Coweta County Complete County[sic] Committee.

"It's sealed for 72 years," she said of the information. Census employees take an oath "that they will not divulge anybody's information," Turner said. The nondisclosure oath is for life.

The penalty for unlawful disclosure is a fine of up to $250,000 or imprisonment of up to five years, or both.

In other words, it won’t be until 2082 that historians or genealogists will be able to use Census data, for instance, to find out if I owned or rented my home this year.

In the mean time, all data released by the Census Bureau has had personally identifiable information (e.g. phone numbers, names, etc.) struck from it. But the Bureau doesn’t stop there -- it also goes to great lengths to make sure no one can reverse-engineer a way to identify anyone through the data they release. Even if it means altering entire data sets just to protect one person:

Suppose there is only one 65-year-old married woman attending college in North Dakota, and that her response was released by the Census Bureau. Then researchers would know everything else she told the agency, including, perhaps, her income and her parents' birthplace.

To protect the privacy of such unusual individuals and households, the government manipulates data, using several techniques that were described in a 2005 Census Bureau paper. Numbers are rounded, so incomes of $80,600 and $81,400 would both be recorded as $81,000. What statisticians refer to as "noise" is added to some ages—a year or two older or younger, perhaps.

Also, outlier values are averaged together, and that average is assigned to every one of those outliers. For instance, the top half-percent of earners would each be assigned the average income of that wealthy subgroup, so that, say, Warren Buffett's census questionnaire can't be identified. And people with especially unique characteristics might be moved across the country, in a kind of statistical witness protection program, so that entry for the North Dakotan woman might be changed to show her living in Alabama.

This additional “noise” is a source of constant irritation for some economists, sociologists, and other researchers using Census data in their research, but it shows just how far the Census Bureau goes protects the information it collects.

So the next time a Republican (like the commenters on the two articles cited above) tells you he’s too scared to fill out his Census form… tell him to take off the tin-foil hat.

How can you say it will be "sealed for 72 years" when every government database has already been hacked at least once? Read the 4th Amendment. Our right to keep our personal information to ourselves shall not be infringed.

[Sigh] Lest anyone be misinformed, the 4th Amendment to the U.S. Constitution states:

"The right of the people to be secure in their persons, houses, papers, and effects, against unreasonable searches and seizures, shall not be violated, and no Warrants shall issue, but upon probable cause, supported by Oath or affirmation, and particularly describing the place to be searched, and the persons or things to be seized."

But the meaning of the 4th Amendment in this case is irrelevant because the Census itself is specifically authorized by Article I, Section 2.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.

More information about formatting options

To combat spam, please enter the code in the image.