This paper develops an algorithm for the probability distribution of a respondent's religion in microdata (including the decennial census) in which there are data on ancestry but not on religion. A frequency distribution of religion by ancestry is generated from the General Social Survey and matched by ancestry groups in the U.S. decennial census. The fruitfulness of the procedure is demonstrated through an analysis of the effect of alternative measures of religion on the household's choice of public versus private schooling for children. This method is useful to any researcher wanting to distinguish religious affiliation when only ancestry data are available.

Stella Koutroumanes Hofrenning is an economist at the Office of Rural Health and Primary Care, Minnesota Department of Health. Barry R. Chiswick is a research professor and the head of the Department of Economics at the University of Illinois at Chicago. The authors appreciate comments received on an earlier draft from John Durkin, Evelyn Lehrer, and Paul W. Miller. The data used in this article can be obtained beginning May 1999 through April 2002 from Dr. Stella Koutroumanes Hofrenning, Office of Rural Health and Primary Care, Minnesota Department of Health, 121 East 7th Place, Suite 460, P.O. Box 64975. St. Paul, MN 55164-0975.