[CTPP] Synthetic Data for PUMA using 2005-2007 ACS

On February 23, Joseph Sakshaug, a Ph.D. Candidate from University of Michigan, gave a presentation at the Census Bureau on data synthesis for small geographic areas. Joe?s research is motivated by the increasing demand for microdata for small geographic areas with simultaneous increase in restrictions of data release at small geography due to the risk of disclosure. While Joe?s research is not focused on transportation data, his research is likely useful to our needs for the CTPP using ACS records. His research utilized county-level data from the 2005-2007 American Community Survey (ACS) and applied it to PUMA geography. He applied a Hierarchical Bayesian model to simulate the synthetic values while accounting for homogeneity within each geographical unit at the PUMA level. Joe tested many variables, including household income, age etc. In his validation the synthetic data were very close to the actual surveys with respect to distributions and means. Joe indicated that the method can be extended to smaller geographic areas such as census tract and block groups. The Westat team working on the NCHRP 08-79 project "Producing Transportation Data Products from the American Community Survey that Comply with Disclosure Rules" also attended the meeting. Joe can be reached by joesaks(a)umich.edu if you want to get more information about his research. Thanks! Liang Long Cambridge Systematics, Inc. 4800 Hampden Lane Suite 800 Bethesda, MD 20814 tel 301 347 9141 fax 301 347 0101 FHWA 202-366-6971 e-mail llong(a)camsys.com www.camsys.com