The Census Bureau is about to go into its most trying of times, convincing the Congress that it's worth it to spend about $4 billion dollars on the next Census. I say about, because they still have some outstanding issues to resolve. The most controversial one right now is whether or not the Bureau will use sampling to correct for nonresponse instead of the traditional method of knocking on doors until they've gotten as many people as possible. The statistical community sees this approach as a major improvement over the doortodoor techniques. It will enable the Bureau to make good statistical estimates of the population and at the same time save money. Of course when we have a major shift such as this, the process is not straight forward.
To review, the Bureau originally planned to take a sample from the last 10% of the households after 90% of a county's households responded through traditional methods, and then conduct another survey Integrated Coverage Measurement (ICM) to account for the final 2%. The ICM procedure wouldbe based upon a sample of 750,000 housing units across the nation. Sounds good so far. However, Carrie Meek, a Democratic Congresswomen from Florida, was unhappy that the county would be the geographic unit for the 90% coverage level. She believed that subcounty minority areas, which might be poorly covered, would not then be properly accounted for in the sampling process. Her Bill, H.R. 3558, stated, "The Bureau shall attempt to contact every household directly.... and may use sampling as a substitute for direct contact in a particular census tract only after direct contact has been made with at least 90 percent of the households in such tracts." This was agreed upon in early September. To further complicate the issue, Bill H.R. 3589 was introduced by Thomas Petri, a Republican Congressman from Wisconsin, stating that "In no event may any sampling or other statistical procedure be used in determining the total population by States.... for the purposes of the apportionment of Representatives in Congress among the several States." It did not pass, but Congressman Petri is expected to reintroduce it. One of the remaining technical issues is how to get to the 90% cutoff level. One way would be to use the regular doortodoor approach to get to 90%, and then take a sample of one in ten households to estimate the rest. Recently, the Bureau decided that there was a more efficient approach. In essence, the same 90% is reached, but in a different way.
Instead of going doortodoor until reaching 90%, the Bureau would begin sampling soon after the basic mailout, mail back procedure is completed. This procedure includes the special followup procedures such as telephone contacts, leaving extra questionnaires at Post Offices, seven attempts in all. At the end of this period, the Bureau would determine the response rate for each tract, and then select a sample for each tract in order to achieve a 90% response. Bureau statisticians believe that this approach would be less expensive and also do the best job in reducing the sampling error.
There are, however, some major issues. Perhaps the most important one is that of perception, especially to Congress. This approach does not fit with the current public perception about first reaching 90% before any sampling begins. Also, this change may be used to show that the Bureau is no longer conducting an "actual enumeration" as specified in Article 1, Section 2 of the Constitution. Another approach under review, albeit much more expensive, is time truncation. Under this scenario followup visits would take place over a three week period. There would be no percentage cutoff. This has the same perception problem mentioned above for direct sampling. However, it could be argued that using this approach does address the issue of attempting to knock on every door. It is safe to say that the House Committee will look very closely at any change in approach. Many on the Hill are skeptical about the Bureau's ability to accomplish the overall 2000 Decennial Census mission which includes the two new sampling procedures, and any changes at this stage may reinforce that skepticism. Therefore, we expect they will want a clear explanation of why the sampling procedure selected is the best approach; how the process will work; what are the savings; and be convinced that what's being proposed is Constitutional.