ABSTRACT
As we observed in Polling and Its Perils, “Counting fish in a lake is one order of business, but how about people?”
In seeking to determine the population of the United States and its characteristics in the mandated decennial census, the approach used in the census of 1990 was similar in many respects to that used in the first census taken two centuries ago in 1790, direct count. It is based on the view that people are in a particular location at a particular time and that it is basically a matter of having them fill out the required questionnaire and then tabulating the results. In theory straightforward, in practice not so. The homeless, migrant workers, and people on the move are difficult or impossible to locate; the questionnaire is unclear to many who are located and asked to fill it out; those who refuse to fill out or return the census questionnaire are not insignificant in number (in some large cities response rates fell under 50 percent in the 1990 census); mispunching errors create whole categories of nonexistent people; some population subgroups are overcounted, while others are undercounted, and on it goes. The 1990 census found 248.7 million people living in the United States. A post-census analysis estimated 4 to 6 million people as not having been counted. Those missed in the count were, for the most part, minorities and urban poor who tended to vote Democratic.
