#100 Jul/Aug 1998

How to Make People Count

Guess how many beans are in the jar and win. How many times have you entered such a contest, even paid money to guess? Well, suppose that there are 435 […]

Guess how many beans are in the jar and win. How many times have you entered such a contest, even paid money to guess? Well, suppose that there are 435 different jars; that you have a limited amount of time; that the true answer is unknown; that some of the beans are hidden; that your estimate will determine whether some jars are broken up and the beans poured into other jars; and you won’t get another chance for another 10 years. That’s the kind of pressure the US Census Bureau is up against every 10 years in order to supply Congress with the number and location of every person living in the United States so that the House of Representatives can be truly representative of the people.

Since the first census in 1790, the problem of counting every person each decade has caused problems, doubts, and charges of inaccuracy. Thomas Jefferson didn’t believe that it could be done (he is said to have penned his estimates of population next to the “official” count) and the Bureau’s follow-up studies and surveys since 1940 have measured just how far off the count has been. Up through 1980, the process was becoming more accurate, but the population was still being under-counted by 1.2 percent.

In 1990 the rate increased for the first time since official records have been kept – in fact, the number of missed people nearly doubled. Many reasons have been given, including rising distrust of the government, concern about confidentiality and poor administration of the census. Whatever the reasons, there is justifiable evidence that the best direct counting efforts available cannot make the undercount disappear. In 1990, the post-census sample found that the census had missed 8.4 million people and double-counted 4.4 million more. The net undercount of 4 million people was 1.6 percent of the population, but was found to be differentially greater for minorities. For instance, 4.5 percent of the African-American population was missed.

After significant public criticism of the 1990 Census, Congress sought the advice of the National Academy of Sciences. The Academy recognized the impossible nature of trying to identify every person, especially in a limited amount of time within a highly mobile population. The scientists recommended a sampling system to supplement the traditional method of counting people. The Census Bureau has refined their recommendation and proposes a two-stage sampling process to ensure that the 2000 count is the most accurate ever. The proposal and subsequent operational adaptations by the Census Bureau have generated lawsuits, Congressional gridlock, committee hearings, and establishment of a special oversight board. The controversy threatens to grind the entire 2000 Census process into fine pieces.

The basic conflict comes down to whose ox is gored. Any method for reducing the undercount, inherent in its design, will add to the count of people that are traditionally undercounted. Since the poor and minorities, especially African-Americans living in older, inner cities, are missed with greater frequency than others, the ultimate count of low-income people and people living in older cities will be larger with the use of any system intended to reduce the undercount, including scientific sampling. Thus, some places will show larger populations and some places will remain the same, which means their share of the total population will decline. Moreover, the places that are likely to gain from a more accurate count are traditionally Democratic, and the places that are likely to lose in a relative sense are traditionally Republican.

Compound the political party tensions with increased reliance on census population counts to allocate $180 billion in federal program funds each year, and a divided opinion on any method to resolve the undercount is bound to occur. Also, most state legislature districts and many locally elected bodies are determined by census population counts within states and counties. Consequently, local, state, and national political power and the flow of federal resources are allocated on the basis of how many people are counted in the decennial census.

The U.S. Census Bureau continues to develop and test innovations and improvements for the 2000 Census aimed at counting the people usually missed in past censuses. The Bureau hired outside consultants and companies to design a more inviting questionnaire; conducted numerous tests to determine the best mailing procedures; and held countless meetings with stakeholders and local governments to understand the many perspectives of users and partners.

Even with all of these improvements, it will be impossible to identify and count every person who lives in the United States. Therefore, the decision on whether to use sampling techniques rests on what is the next best method for counting people if the old way of trying to find and identify every single person does not work. Surely after 200 years of experience in taking a census and in a world where science and information technology are thrusting us forward at breakneck speed, there must be a better counting procedure.

Two questions should be asked – what scientific method achieves the most accurate count, and can that method be conducted without reproach? The appropriate scientific method also involves a legal question that revolves around constitutional interpretation of the words “actual enumeration.” This question is on an expedited course towards a Supreme Court review.

Both of these scientific questions require some knowledge of statistics, the science of data collection, and the Census Bureau’s skill and capacity. Unfortunately, the debate over how to conduct the 2000 Census has paid little heed to the highly technical questions and instead has been distilled into who would win and who would lose using different approaches.

At some point, even individuals with sufficient statistics and operational knowledge must ultimately ask themselves one question: Do you believe the U.S. Census Bureau has the expertise to pick the right method and carry out the job?  If the answer is no, then any kind of census, including following the old method, is in serious trouble. If the answer is yes, then we need to let them do the job that they were asked to do. Even if the answer is somewhere in between (which it probably is), then the key issues are how to conduct a census that counts everyone and how to equip the organization responsible for that effort so that it can carry out its mandate. If we don’t trust the world-renowned professionals at the Census Bureau to conduct an unbiased, scientifically designed census, how can we trust them to conduct a census that depends on an army of temporary workers hired and trained by the very same institution?

Something as important as the census should be monitored closely, with careful scrutiny by observers from both political parties. But, primarily, the magnifying glass should focus on the best, most cost-effective method for counting people – not who will win and who will lose. We will all lose if the 2000 Census is less than the best that human effort and the latest science can deliver.


  • Community Action Agencies

    July 1, 1998

    Born amid 1960s strife, nearly 1,000 Community Action Agencies fight poverty all over America today. While CAAs deliver essential services, a renewed focus on citizen involvement and collaboration is needed […]

  • National Congress for Commuity Economic Development (NCCED)

    July 1, 1998

    More than three decades after the community development field began to grow by leaps and bounds, community-based development organizations are still taking on the same challenges they did in the […]

  • CDC Networks and Intermediary Organizations

    July 1, 1998

    As the community development industry has grown, networks, associations, and intermediaries have developed to support CDCs and other community-based organizations. Some provide funding, some offer training, and others provide direct […]