I was an enumerator for the Census Bureau. It was an enlightening experience that caused me to reflect a bit about the census and how it might be done better.

There is a simple reason for taking the census: the Constitution directs that it be done every 10 years and that its count be used to apportion members of the House of Representatives from each state.

This seems fairly easy -- just count the people, and that tells us how many representatives in Congress each state gets. If that's all there is to it, what's all this other stuff on the census form asking about the number of acres your house sits on, how many rooms you have, whether you run a barber shop in your house and what do you think your house is worth? Are you white nor black? If neither, then which Indian tribe or which Hispanic or Asian country are you from? All this on the "short" form, which, mercifully, most people get.

The best one can say for the long form is that it is well named. Expanding on the short, it asks: How much do you pay for utilities, taxes, interest, mortgages or rent? When did you move into your house or apartment and when was it built? Where were you born? Where have you lived? What is your level of education? What is your ethnic origin? What are your work habits, past and present, and, of course, your income, in detail?

The Census Bureau gives two reasons why these intrusive questions are asked: to distribute government funds for education, housing and health programs, and to provide an economic profile our changing nation. In other words, to benefit government and business. Both, most people would agree, are essential to the well-being of the nation, and both need data to make the decisions that are expected of them. Does that justify the intrusive questions?

No, and intrusive questions don't elicit reliable data.

Take, for example, income data. A bartender in one of Washington's plush restaurants when asked his income for 1989 flashed a look of guilt and mumbled, "$5,500." I dutifully wrote that figure down, using the hood of his Porsche as a table. Quite naturally, people resent the income questions and don't answer them very well. As many said, "I've already answered that question for the government." Yes, yes, we all understand that the Internal Revenue Service and the Census Bureau are different government agencies, and their data is supposed to be kept separate, but in hearts the government is the government, and the notion that its many bureaus and departments will never talk to each other simply doesn't hold water, especially when we know that the data will computerized.

But why shouldn't government agencies talk to each other? A much better way to gather income data for the census would be to direct the IRS to produce for the Census Bureau sterilized (i.e., no names or addresses) income data by state, city, zip code, whatever. Much easier and, significantly, the data would be accurate.

Anyone analyzing a 1990 census return will be struck by an apparent obsession with race. Almost half the column on the short form for each person in the household is taken up with such questions.

It is troublesome to realize how capricious and inaccurate many of the answers about race are, mainly because the answers are not as simple to arrive at as one would first believe. Also, there are no firm guidelines for answering racial questions. For example, children of a white American father and a Japanese-born mother -- white or Asian?

I talked to two white women from the Midwest. Both had blond, blue-eyed children. Both were married to Hispanic men with distinctive Spanish surnames, one from Cuba and one from Venezuela. The woman married to the Cuban concluded that I should record her children as Hispanic "because with the name they have they get treated like Hispanics anyway." I did. The second woman, after some thought, concluded that since Jews hold that the line follows the mother's (logically, she felt), her children should be listed as non-Hispanic. I did.

Which of these women was right? I haven't the foggiest notion, and the truth of the matter is neither does anyone else. We would do much better to contract with private firms to collect and present this kind of data in a consistent and coherent form.

Taking the census need not be the expensive, unresponsive and controversial experience that the 1990 census has been. There is a simple solution: do what the Constitution says -- count the people and apportion representatives accordingly. If we really need all this other information, then direct those peripheral and irksome questions to the professionals, public and private, who get paid to answer such questions.

Richard B. Beal Jr., retired from the federal government, is a freelance writer.