Creating a Database for Demographic Research: A Case Study.

The PUS801000 (Public Use Samples) database was created as a subsample of the Census Bureau's 1980 Public Use Microdata Sample (PUMS) for the purpose of meeting the needs of demographics research. PUMS needed to be reorganized along relational lines by identifying the variables most widely used by researchers and constructing proper relations containing these variables. PUMS contains 11 million records for individuals and 4 million records for households; in its original format, each household record is followed by the "person" records for that household. The PUS801000 database is a reduction of PUMS database to construct a 1/1,000 sample of U.S. households (94,201 households; 226,458 persons), in which each household is coded uniquely (by state/number) and each person in the household receives a meaningful number code. (All variables are integer format except for H=Household and P=Person.) Questions that could not be addressed by the basic relational operations of traditional projection and restriction, and the problem of record concatenation can be addressed by the QUEL programming language. Programming steps are described for creating married couples as a unit of analysis. This involves the complex process of identifying primary family couples and subfamily couples in households. A portion of the user directory table illustrates some of the variable fields selected and the storage saving achieved by coding "person" relations. (LPG)