Confidential Statistical Data Are Still Secure

Protection background. Source: iStock/KrulUA

As the new Trump administration assumes control of the federal government, there’s uncertainty about how well it will maintain the integrity of official statistics. Can we be assured that the government will protect the confidentiality of the data it collects from us for statistical purposes? Can we be assured that the reported statistics will be prepared independently and not be tampered with for political purposes? And can we be assured that the full suite of accurate statistics will will continue to be produced and made available? Today’s post addresses the first question; my next post will address the second; and I’ll wait to address the third until we know a little more about the new administration’s budget proposals and their impact on statistics.

Each year, the U.S. federal agencies collect data for statistical purposes from millions of households and businesses. The data providers are given a pledge of confidentiality. Not only will the individually identifiable data not be released to the public, but we’re also promised that data will not be given to other government agencies for non-statistical purposes, such as regulation, law enforcement, or tax collection.

The agencies have a strong motivation to be able to offer a credible promise of confidentiality. The statistical agencies know that without assurances of confidentiality, they won’t be able to get high response rates and accurate responses to their surveys. The assurance of confidentiality is based on a combination of legal protections and the norms and practices of the statistical agencies.

All U.S. federal data collected for statistical purposes are protected by the Confidential Information Protection and Statistical Efficiency Act enacted in 2002. Under this act, identifiable data collected under a pledge of confidentiality cannot be disclosed except with informed consent of the respondent. The willful disclosure of protected data by a federal employee or agent to someone who isn’t entitled to receive the data is a felony punishable by up to 5 years in prison and a $250,000 fine.

In addition to CIPSEA, additional legal protections apply to several specific federal agencies. For example, the confidentiality of data collected by the Census Bureau is protected by Title 13 of the U.S. Code, which, for example, makes data collected by the Census Bureau immune from being used as legal evidence. In addition, most of the data that Census collects from businesses are protected by the Internal Revenue Code (Title 26), under which the IRS is authorized to share tax information with the Census Bureau for statistical purposes. The data that BEA collects on international investment and international trade in services is protected under Title 22. Similar specific protections apply to data collected by several other statistical agencies.

Beyond the legal protections, the statistical agencies have norms and practices that reinforce the legal protections. For example, at BEA all employees are required to take annual training on protecting confidential data. The training emphasizes not only the legal requirements, but also specific practices that employees are required to follow to prevent confidential data from falling into unauthorized hands. The requirement to protect confidential data is also part of employees’ performance plans and evaluations. Similarly, the Bureau’s computer systems and software include safeguards to ensure that only authorized employees are accessing the data. The importance of these norms protecting confidentiality was apparent at staff meetings, at which staff carefully avoid discussing confidential information with anyone who isn’t entitled to that information.

The statistical agencies are staffed by career civil servants who are trained regarding the legal protections given to confidential data and would not be likely to risk a felony conviction to satisfy a political appointee. They understand the importance of the credibility of the confidentiality assurances to the operation of the statistical system. The agencies have a very strong history of protecting the data and maintaining confidentiality. My belief is that we can be confident that the statistical agencies will continue to protect the data that are provided to them, and that if there were attempts to compromise the legal protections we would hear from whistle blowers.

Coming up, my next post addresses concerns about protecting the independence of the statistical agencies, where I unfortunately think there are more unanswered questions.

P.S. I’d like to express my gratitude to all those who have recommended my blog during its first week. I’m overwhelmed by all the wonderful endorsements I’ve received. The results of these endorsements are apparent in the traffic shown in my site stats—the blog has received more than 4,000 page views and 2,000 unique visitors during its first week—quite unexpected for a brand new blog with a short track record of only four posts. I hope that my future posts will not disappoint what appear to be very high expectations. Thank you!


  1. LB

    Great blog on an important topic! Congrats on the well-deserved traffic.

    I agree with what you write about the internal culture of the statistical agencies. The laws you cite are well-known to civil servants because of the annual trainings, and there are strong norms. In normal times that would be enough.

    However, as you mention, one reason people are fearful about the confidentiality of statistical data is the new administration. Laws do not enforce themselves. A lawless executive can do a lot of damage before the courts have a chance to catch up, if they ever can. The legal scholar Eric Posner wrote an interesting piece about the nightmare scenario in a NYT opinion piece back in June. I’ll just quote the last two paragraphs, which are relevant to this issue:

    “Mr. Trump will need to get loyalists into leadership positions of the agencies, but to do so, he will need the cooperation of the Senate […]. Moreover, the small number of politically appointed leaders enjoy only limited control of the mass of civil servants. These employees can drag their feet, leak to the press, threaten to resign and employ other tactics to undermine Mr. Trump’s initiatives if they object to them. They’re also hard to fire, thanks to Civil Service protections. But Mr. Trump can fight back. He can appoint loyalists not only to political positions in the executive branch, but to the courts, and he may be able to attract them to the ranks of the Civil Service. And while executive branch officials who disregard the law might be prosecuted by the Justice Department, President Trump would have one more trick up his sleeve. Like President George H.W. Bush, who rescued Iran-contra defendants from punishment in 1992, he could hand out get-out-of-jail-free cards in the form of the pardon.” (Source: NY Times)

    I wonder whether the more important safeguard in such a scenario is actually the prohibitions on asking certain kinds of questions that could be abused by a (hypothetical) lawless executive. For example, I am not aware (perhaps I’m mistaken) of any federal statistical data on religious affiliation. If that is true, it may prove to have been a prudent prohibition.

    Of course, in the lawless executive scenario privately held data would also be vulnerable. Google, Apple, Amazon, may have more sensitive data than the federal government about many individuals.

    • Brent Moulton

      Thanks for a thoughtful and interesting comment. My conclusion that we can rely on statistical agencies to keep our data secure is based on current laws and norms. While one can imagine a nightmare scenario in which he erodes the laws and corrupts the civil service, I (optimistically?) think he would lose the support of Congress before that happens. We’ll see what happens, and I hope I don’t have to revise my take on this issue.

  2. Tom

    Let’s hope Trump will replaced the partisanship of the former administration with neutral leadership rather than partisanship of a different flavor.


Leave a Reply

Your email address will not be published. Required fields are marked *