The census: counting the uncounted
As the 1980 census draws closer, concern about undercounting the population increases. When the surface is scratched, the concern is generally about money and specifically about federal revenue sharing.
More than 40,000 local governments in the United States are eligible for revenue-sharing money under one program or another. Each one's share of funds depends on population size or characteristics as measured by the census. Therefore, the census and the population undercount in the census become critical factors in the budgets of many states and cities. the city of Los Angeles, for instance, received the not insubstantial sum of $50 million in revenue-sharing funds for its 1978-79 budget of $1.1 billion.
In the debate on the census undercount and the search for a solution to it, two obvious facts are frequently overlooked. First, the population counts used in various monetary formulas are at the same time both good and bad instruments for government policy. The good aspect of using population data is that the census provides exact numbers that can be put into a formula for distribution of funds. Census data, apparently objective and uniformly collected, are able to be translated by formulas into exact dollar amounts for state and local governments.
The less positive aspect of using census data in this regard is the assumption that the number of people, the proportion unemployed or below the poverty line, or some other profile obtained through census data is anything more than a rough measure of the problems or needs for which federal grants are made.
The second fact overlooked is that a census is never complete. In a country as large and heterogeneous as the US, we cannot count every last person; we probably cannot even count the last five million people. A census is a very large survey. In 1960 and 1970, the US census was a survey of about 97.5 percent of the population.
The statisticians, economists, and market researchers as well aware of -- and , in fact, make allowances for -- the undercount, biases, and error in using census statistics. Some local officials have jumped at the undercount as a chance to increase their share of the federal revenue pie. Thus, a good deal of thinking has been done on how to eliminate the undercount in answer to local accusations of being short- changed.
The fact of the matter is: there will always be an undercount. It will not be random, but concentrated among persons whose characteristics make them hard to count, like the poor, the transient, young adults, and non-English- speaking minorities. We should aim for a complete enumeration, especially among those who are hard to count. Although the ideal will not be achieved, the more complete the count, the more reliable the estimates of undercount. Using the size and source of the undercount; a 10 to 20 percent undercount would leave us floundering.
Even though the census is imperfect, it has been and probably should continue to be used to distribute funds. It is the largest systematic survey of information on population, income, employment, and education. The census, used in conjunction with statistical data from social security, vital statistics (birth and death certificates), motor vehicle licenses, and school enrollments, provides the closest approximation of "true figures," that we can achieve. The critical element for Congress in using these figures as a means to awarding funding for social service programs is to aknowledge the limited "truth" of these numbers. Ninety-seven percent accuracy is still only 97 percent; and the uncounted persons are often those people who need to be counted the most.
The problem, then, before Congress and the public is not how to eliminate the undercount Rather, it is to agree upon the use of census date for social service funding. In the past the convention was to use the census statistics as given because they were accepted as true. We now know differently and acknowlege the existence of a skewed undercount.
But what for the future? We have two options. First, we could continue to use the census as the guiding statistical base from which to allocate federal revenue sharing as we did in the past, accepting its imperfections through the undercount. Or, and I recommend this as a better convention, Congress could use the census data in conjunction with other statistical data in order to deterimine more precisely the numerical extent and raciaL ECONOMIC PROFILE of the undercount.
With the undercount figure more firmly in hand, Congress could then use the census data plus the undercount information to award federal revenue-sharing funds. Although the second method seems to be more consistent with the information of the census and purposes of federal revenue sharing, it is quite clear that both Congress and revenue-sharing applicants must agree which way to use the census -- and soon.