Does analysis of the Survey of Consumer Finances datasets require IRB review at a university?

It should not even be necessary to apply for an exemption, though some IRB staff might initially give a contrary opinion. I finally saw official confirmation of that in the Ohio State University IRB Newsletter in June 2007:

Use of publicly available data is not human subjects research. Even if the information contains personal identifiers, if it is not "private" - i.e., the data set can be retrieved by

anyone publicly and freely (without special permission) - then the research does not involve human subjects.

I posted the following note in 2006.

After discussion with an Ohio State University IRB Administrator, I have concluded that one need not even apply for an exemption from IRB review.

It may seem silly to even ask this question, but the initial response of the administrator was that I should apply for an exemption, even after I repeatedly stated to her that anyone in the world can download the SCF datasets. I would like to hear from researchers at other universities, but I suspect that the background of many people running IRB offices is related to medical research, where the concern is about subjects becoming ill in medical research. It seems crazy to waste people's time for a review of analysis of a public dataset, but compliance officers typically have little insight into the distinctions involved with privacy issues in a survey you collect versus privacy issues in a dataset such as the SCF, where a massive effort has been completed to deidentify each SCF dataset.

At any rate, the crucial issue for a public dataset is whether it is "coded" --- which mystified me initially in my emails with our IRB administrator. The meaning of this term has to do with whether there is a code or key that would enable a researcher to obtain the identify of a case on the dataset.

The statement given to me by our IRB administrator was:
"If the data is not coded but is deidentified then no agreement is needed except for the determination of "not human subject". In this case no IRB review or exemption from IRB review is needed."

I posted a question on the Federal Reserve's web site, and the response on July 31, 2006 from Gerhard Fries was:

"There is no direct key or identifier type variable in the data set. There is an ID variable, but that is randomly generated between 1 and the number of households in the data set."

He also noted that "the SCF is approved by the University of Chicago IRB."

Sherman D. Hanna
Professor, Ohio State University
Consumer Sciences Department
1787 Neil Avenue
Columbus, OH 43210
hanna.1@osu.edu