Association for Institutional Research
Enhancing Knowledge. Expanding Networks.
#20Title: Use of Perturbation in IPEDS Data Prepared: March 8, 2004 Type: Informational Summary The National Center for Education Statistics (NCES) recently implemented a "perturbation" process for data from four surveys collected through the Integrated Postsecondary Education Data System (IPEDS). Perturbation in this case means randomly altering the data in cells with a small number of observations to protect the confidentiality of personally identifiable data. The Salaries, Fall Staff, Graduation Rates, and Student Financial Aid surveys are subject to perturbation. As a result, the data NCES disseminates in the IPEDS Peer Analysis System (PAS) may not precisely match the data submitted by institutions. Current Status The NCES must comply with nondisclosure regulations relating to personally identifiable data, a requirement that has created the long-standing issue of how to handle data in "small cells" that might permit individuals to be identified. Perturbation makes random changes to cells with a small number of observations. Therefore perturbed IPEDS data released by NCES may not be exactly the same as data submitted by the institutions. However, the perturbation process and validity checks will preserve as much as possible row and column totals, sub-totals, measures of central tendency, percentages, and rates. The NCES will not release details of its procedure (knowing which cells have been changed would defeat the purpose of perturbation), and NCES employees cannot disclose the process beyond what is described in this Alert. The randomization makes figuring out which cells have been perturbed more difficult. The decision to introduce perturbation in IPEDS data is in response to five acts: the Privacy Act of 1974, as amended; the Federal Statistical Confidentiality Order of 1997; the U.S. Patriot Act of 2001; the Educational Sciences Reform Act of 2002; and the E-Government Act of 2002. Originally NCES planned to suppress such data, but so many cells ended up being suppressed (e.g., more than 90% of the GRS data in cells related to sport-by-race/ethnicity, including row and column totals) that some reports were useless. Suppression also delayed availability of data and complicated the access process. Perturbation, on the other hand, protects the confidentiality of individual data without suppression, allows NCES to get the data out quickly, and eliminates restricted data licensure problems. Implications for Institutions Institutions can share their data with other institutions and outside agencies as long as
Note: Institutions may share data or files prepared for input into the IPEDS data collection system. They may also share data from the Peer Analysis System. What institutions may not do is share printed copies of the completed online IPEDS data collection forms or files that have been downloaded from the IPEDS collection system for the four IPEDS surveys subject to perturbation. According to the Educational Sciences Reform Act of 2002, data in the IPEDS collection system may be accessed only by the respondents themselves (and then only for their own data) or by NCES and NCES contractors. Therefore institutions may share their data only by using institutionally generated forms or files, data from the PAS, or data on an IPEDS Forms Facsimile (a feature that will be made available later this year through PAS). Furthermore, institutions should never advertise which of their data have been changed by perturbation-it is unlawful to disclose details of perturbation. The National Collegiate Athletic Association (NCAA) recently passed legislation requiring Division I and Division II schools to submit all graduation rates data directly to the NCAA. Beginning Spring 2004, the person who is responsible for submitting an institution's graduation rates to the NCAA must submit the graduation rates data prepared for input to IPEDS. Institutions must be careful to provide a copy of the institution's graduation-rate data from institutional reports (but not directly from the IPEDS data collection system) and send it to the NCAA on or before the IPEDS deadline. NCAA will send member institutions instructions on how to submit the data directly to them. It is important that institutions always check prior-year data and use their ability to correct these data if they detect errors. However, if in correcting the data, the institutions put in small cells, the data will be subject to perturbation again. Timeline Winter 2003/04 - Data from Salaries and Fall Staff are perturbed as they are migrated to the PAS. Spring 2004 - Data for Student Financial Aid and Graduation Rates will be perturbed as they are migrated to the PAS. Ongoing - Data from prior collections of Salaries, Fall Staff, Graduation Rates, and Student Financial Aid surveys will be perturbed and migrated to the PAS to eliminate the need for restricted files; more recent years will be done first. Additional Resources
Authors: Coordinated by the Higher Education Data Policy Committee. All opinions expressed herein do not necessarily reflect the official position of the Association for Institutional Research. |