THE CENSUS is running scared. For the first time in its 190-year history it faces real resistance -- indeed, enough noncooperation to jeopardize the final count. What happens may determine whether the census can be preserved in its present form.

There are signs that increasing numbers of Americans are no longer in a mood to give out information about themselves. To make matters worse, the census coincides with another official solicitation: the ubiquitous 1040 from the Internal Revenue Service.

The Census Bureau is doing its best to stress its confidentiality. Vincent Barabba, its director, confesses his concern about "the depth of mistrust of government" and "public resentment against federal snooping." Hence his fear of refusals to respond, or forms with incorrect answers.

Unfortunately, public confidence is unlikely to be soothed by the questionnaire itself. It asks many intimate questions, ranging from the sources of one's income to how many children a woman has had. (Truthful answere here could threaten many marriages. And men, it might be added, are not asked a parallel question.) Equally disquieting is the instruction that respondents write their names at the top of each page of the form. Moreover, names stay on all through the processing and remain when the form is finally stored.

Writing one's name on a form that already has your address, and sending it back to a government agency, is not something all Americans are eager to do. The country has plenty of missing husbands and nonsupporting fathers, not to mention persons avoiding creditors or warrants for court appearances. And then there is, of course, the issue of illegal aliens.

Equally troubling are indications that many model citizens will treat the census cavalierly. What the bureau calls "respondent burden" may be at a breaking point. That is, the questionnaire is now of such length and complexity as to lead people to wonder why they must go to all this trouble. The long form requires upwards of 320 answers from a family of four, including more than 30 that must be written in (such as the "activities and duties" associated with your occupation.) Indeed, to do the job properly requires rummaging among the household papers to find the amount of last year's fire insurance premiums and how much interest each family member received.

To be sure, filling in a census form is one of the smaller duties of citizenship -- certainly less burdensome than two weeks of jury duty or extended military service. Still, must we really look up our water bills, or reveal that we share our shower with a neighbor?

What we now have is a far cry from the "enumeration" mentioned in the Constitution, whose purpose was to decide each state's delegation in the House of Representatives. The first census asked only six questions.

Data from the census also play a statutory role. Increasingly, Congress has been allocating funds on the basis of demography. At last count, 107 federal programs funneled money to states according not only to population but also to factors like per capita income, degree of urbanization and even age of housing. No fewer than eight acts of Congress need to know how many bedrooms each of us has. Yet hardly any funding programs stick with the decennial census data.

To begin with, the 1980 figures will take from two to three years to be reported, by which time they will be sadly out of date. The bureau is well aware of this, so it prepares estimates each year, revising not only population figures but those for income and other items. A recent report by the Office of Management and Budget reveals that 95 percent of all programs now rely on intercensal estimates.

The people at the bureau admit that they do not need the decennial census as a basis for reliable estimates. They have a shelfload of sources to draw on, ranging from school enrollment records and vital-statistics reports to postal-route traffic and public-assistance applications. They can judge not only total populations of states and cities, but how many teenagers there are in a given rural county. It is also possible to provide trustworthy data for small tracts of territory to serve as jigsaw-puzzle pieces when legislative districts are drawn.

Indeed, the decennial census can be an obstacle to accuracy. If the comprehensive count misses a substantial number of people, difficulties arise in releasing later estimates that seem to challenge the original figures. Thus the bureau finds itself having to defend a count that it knows to be inaccurate. The time has come to admit that a complete count is now beyond our capacities.

There must be a better way. And in fact there is.

I have been one of the bureau's most loyal fans. I have written two books and numerous articles based on census materials, for which I will be forever grateful. Not only that, I find that its reports make splendid bedtime reading.

Where else would one learn that among the 3,712 marriages between Americans of Chinese and Japanese descent, twice as many (2,418) have a Japanese wife and a Chinese husband as the other way around (1,294)? Or that the country has 64 seamstresses between the ages of 25 and 34 who spent four years at college and earn between $3,000 and $5,999?

So, my wish is to defend the census, if necessary from itself, because it is a model of what a government agency ought to be. In a complex of barracks-like buildings about a halfhour's drive from Washington are some of the country's finest geographers, cartographers, demographers, programmers and statisticians. And, significantly, they show what they do best in the years between the censuses.

Each month the bureau conducts its Current Population Survey (CPS). Here it has 1,820 full-time interviewers (paid at least $5 an hour) who contact a carefully selected and changing sample of 56,000 American households. Their first task is to ask questions from which the Bureau of Labor Statistics computes the monthly unemployment rate.

In addition, the CPS appends supplementary questions on a variety of topics. For a study of fertility, it asked the women in the sample how many children they expected to have. (From this it found the anticipated fertility rate had fallen to 1.8, below the replacement level.) Another study, on divorce, asked unmarried women with children how much in support payments they actually received. (The depressing answer was that only 25 percent received checks regularly.) A survey on health discovered the incidence of emphysema and asthma, while another uncovered the number of people who have been victims of crimes.

The CPS enumerators are carefully monitored by bureau professionals: they stick sedulously to their samples, which include pool hall patrons and illegal aliens. As a testimony to its reliability, a new mid-decade "census" authorized by Congress for 1985 will in fact be a national sampling along the lines of the Current Population Survey. So why not consider this arrangement for 1990 as well? Its virtue is its accuracy, which stems not only from scrupulous sampling but also from the enumerators' experience in gaining the confidence of the people they interview directly. And instead of one overpowering long form, the CPS spaces its questions and subjects over a series of months.

But there is a problem with professionalism. Experts tend to remove themselves from ordinary men and women. The bureau has failed to cultivate the very people whose taxes pay its bills and provide answers to its questions. tFew Americans feel it is their census in the way that they do about, say, the national parks.

For example, the bureau has never prepared a brief booklet for popular distribution saying, "Here Are Some of the Things You Told Us About Yourself." While it prints elaborate books of figures (200,000 pages' worth for 1970), it now records most of its data on computer tape, which it sells to well-heeled clients. For 1980 it figures that 90 percent of its sales will be on tapes, requiring at least 1,500 reels at a total cost of more than $120,000 for the set.

There now exist companied that make substantial profits by processing this public data. Most of the country's largest companies now use census materials in their marketing strategies. With demographic models a corporation can decide where to locate a bakery for a new, high-priced bread, or what the best sales territories are for marketing wallpaper and floor tiles.

Hence the spectacle of 220 million Americans -- an estimate will do -- freely giving information that will then be used by companies that want to sell them tiles. Perhaps this is the American way. Even so, the overhead costs of the census, including its advanced statistical systems, are paid for by the taxpayer. "Since the bureau can't copyright the computer programs it develops," one official pointed out, "private data companies can buy and resell them without putting up all that development money up front." From this has grown an industrial-demographic complex, complete with subsidies.

It is time to admit that getting forms to -- and from -- everyone is a goal beyond our reach, despite the best efforts of the Boy Scouts and a plug from Mickey Mouse. The census can fulfill its constitutional mandate by other means. It can provide such figures as Congress needs by a series of sample surveys; and other statistics are available for drawing legislative districts. The Current Population Survey produces quite enough data to keep scholars and others happy. Still, if we must conduct a comprehensive census, it would be well to recall that 1990 will be its bicentennial year. So why not honor that anniversary by reverting to its six original questions? Indeed, with slavery now abolished, we can reduce the list to five.