Though the government released the official figures this week on the 1990 census, the results still leave a big question unresolved.

The question: Just how accurate is the once-a-decade headcount anyway?The probable answer: Not as accurate as it should be.

That's too bad because the population figures carry enormous weight in determining the allocation of political power and federal funds.

Who's to blame for this shortcoming? Both the government and the public, particularly the part of it that is under-counted.

Anyway, the official new numbers indicate more than just a 10.2 per cent increase over the past decade to a total of population of 249,632,692. They also show a dramatic shift in population and political power from the Midwest and Northeast to the South and West.

Those trends are clear enough to be indisputable even though the accuracy of the census is open to question. The question arises because the final count fell 3.8 million short of Census Bureau estimate released in October but was nearly four-million over a previous estimate the bureau made last summer.

This situation understandably dismays officials in big cities and states, which lose federal funds from an under-count and have long accused the Census Bureau of failing to fully tally minorities.

But it's always hard to take an accurate census in urban areas with high rates of poverty, crime, drug abuse, homelessness, illiteracy and inadequate education.

Even so, there is a way to improve the accuracy of the census. It could be done by going back to basics - asking only such questions as name, sex, age, and a few other details.

That would speed up the process, reduce costs, and still fulfill the basic purpose of the census - that of providing the numbers on which seats in the U.S. House of Representatives are allocated.

For that to happen, however, social scientists would have to abandon the census as a tool for finding out such things about Americans as their ancestry, education, military service, disability, and transportation habits.

Yes, this kind of census data can help in framing public policy more realistically. But the longer and more personal its questionnaires get, the more the Census Bureau erodes the citizen cooperation on which the success and accuracy of its head-counts depends.