A BILL now in Congress would remedy Commerce Secretary Robert Mosbacher's summer overruling of a Census Bureau recommendation to upwardly adjust the 1990 census.Bipartisan sponsors of the legislation have appealed to Mr. Mosbacher to reverse his position. They seek an adjustment to correct the 5 percent minority undercount which the Census Bureau acknowledges. This is a needed first step in correcting the serious undercount of minorities in the census. It does not, however, address the gross undercount of the nation's homeless. The homeless were not included in the Census Bureau's post-enumeration survey which provides the scientific basis for the proposed adjustment. What transpired on a single evening in March 1990 - the 1990 Street and Shelter Night (or "S-Night" in the parlance of the Census Bureau) - shows that the traditional head-count approach favored by Mosbacher is woefully inadequate when it comes to enumerating hard-to-count persons such as minorities and the homeless. The S-Night results were often ludicrous. Are there really 75 times as many homeless on the street in Birmingham, Ala., as there are in Richmond, Va.? Are there really no homeless persons in Wichita, Kan., and only one in Rochester, N.Y.? Are there really 16 times as many homeless in San Diego as there are in Washington, D.C., where only 131 were counted? (One resident commented he believed there were 131 homeless persons on his block alone.) Studies have repeatedly shown that two-thirds of the homeless sleep concealed to avoid victimization. The bureau was fully aware of this. Yet, in localities where the bureau did count, it made a deliberate decision to exclude from its head count all of the hidden homeless - those sleeping in automobiles, abandoned buildings, bushes, dumpsters, roofs, caves, or those concealed by tarps, cardboard boxes, or shanty structures. The bureau attempted to count only the homeless in the open at certain pre-identified street and shelter sites. As the bureau acknowledged, the quality of the lists of these sites, provided by local governments, varied a great deal from one locale to another. Even at these pre-identified sites the head count was a failure. Most of the homeless were missed, according to studies commissioned by the bureau in New York, Los Angeles, Phoenix, New Orleans, and Chicago. And another bureau-commissioned study found that the bureau failed to identify and count half of the shelters in the areas surveyed. In addition to the obvious implications for the need to use a scientifically sound estimating technique in achieving a reasonably accurate census, S-Night also raises an important constitutional issue. The Constitution requires a decennial census of "the whole number of persons" in the United States. Yet the bureau's enumeration of the homeless took place in only 5,000 of the 39,000 governmental units in the nation. No enumeration whatsoever was undertaken in the overwhelming majority of small cities, to wns, and rural areas. Because of this, only 80 homeless persons were counted on the street or out of doors in the whole of six largely rural states where counting did take place. Prior to the passage of the Fourteenth Amendment, slaves were counted as only three-fifths of a human being. Most of America's homeless citizens are in danger of not being counted at all. Congress should recognize that the allocation of scarce funds for the homeless using the S-Night figures would be grossly unfair. Those areas where more of the homeless stay in shelters, where the lists provided to the bureau were more complete, or where the weather was better on S-Night would be favored. The small town and rural homeless would be essentially ignored. The hidden homeless would be treated as though they did not exist. The undercount of the homeless will also skew redistricting of electoral seats and result in the annual misallocation of more than $40 billion of federal programs distributed on the basis of the census - including many programs which assist in alleviating homelessness. Threatened with litigation by the National Law Center on Homelessness and Poverty, the bureau issued a disclaimer acknowledging that the count "was not intended to, and did not, produce a count of the 'homeless' population of the country." The bureau, however, remains adamant that it will not disseminate even its own five-page disclaimer with the actual S-Night figures provided to states and researchers. The lessons of S-Night need to be acknowledged in the debate over the validity of the head-count methodology. Congressional action would prevent the misuse of the S-Night figures and ensure that the dramatic flaws of S-Night are not repeated.