The Census is the most important research conducted in the U.S. Every 10 years, this detailed count of the population produces data which is the basis for election districts, federal aid to various institutions and municipalities, and -- most important from a marketing perspective – the basis for population projections of all the other research we use to make business decisions. SMRB, MRI, Nielsen, Arbitron, IRI, et cetera, ad infinitum project their respondent data to the Census’ population figures. True, over time, adjustments and updates are estimated, because too much can change in the 10 year interval. But the base number has always been a Census; by definition, a count of everyone, not a statistical sampling.An undercount of specific communities, interest groups, or population segments is a problem that these groups readily recognize. Everyone will recall 1990's protests of undercounts of African Americans, Hispanics, urban poor, homeless, rural poor and so on. There are also overcount issues for those with multiple residences, generally a more well-to-do set.Why Sampling?As followers of the Census, which, no doubt, describes all readers of this essay, you will be aware that for the 1990 Census, only 65% of the households returned their Census forms as directed, by April 1. It was reported that the remaining 35% had to be contacted by door-to-door enumerators, at a cost of roughly $10 million for each of the 35 percentage points of population targeted, or $350,000,000.Two solutions are proposed:

• Paid advertising

Despite donated PSA time and space worth over $60 million in the two month "Answer the Census" campaign, and awareness in the 90%+ range, the 35% non-cooperation still occurred. It has been estimated that the nature of the donated free advertising, mostly in remnant space and time, created a highly skewed heavily exposed segment" and a surprisingly large light exposure group. A paid campaign, with carefully chosen schedules and positioning, will go far to offset this and target the demographics of non-cooperators.

• Sampling

The reasons commonly cited for the Census’ undercount include:

Multiple response from people with two homes

Reluctance to cooperate among those in illegal dwellings or illegally in legal dwellings with lease restrictionsUndocumented aliensImmigrants with limited EnglishPeople who are not "homeless" but have no single specific residence and Those who just don’t care and won’t take the trouble.So how will sampling solve these problems? People who would choose not to cooperate with the Census will not suddenly decide to cooperate because they are solicited to be in a sample. The respondent is not aware of the methodology. What sampling will do, which is good, is reduce the cost of the "Census," (which should, I suppose, be renamed the (U.S. Sample). But there does not seem to be anything about the "Sample" to improve the count. True, fewer household will have to be designated as "missed," but non-cooperators will still not cooperate and the same questions will exist about undercount. And oh, by the way, to what base will we project the results of the U.S. Sample?