THE COMMERCE Department has just made one of the first decisions of the 1990s -- and made it right. The question was whether the head count of the 1990 Census should be adjusted to reflect the "undercount" -- that is, the people missed by Census-takers. Commerce last week decided against making any such adjustment. It was the right decision, and Congress should resist the temptation to overturn it.

We say temptation because the case for adjusting the head count is seductive. The Census Bureau knows fairly well how many people are not counted and who they are. They're concentrated among the urban poor and may include more than 5 percent of black males. Some mayors -- Edward Koch of New York, for instance -- argue that these are exactly the people who most need political representation and federal aid but who are shortchanged because representation and aid are based on Census results. So, say Mr. Koch and Rep. Mervyn Dymally (D-Calif.), chairman of the House Census subcommittee, you should adjust the head count to make it accurate and fair.

But adjust it by how much? The same experts who are sure there's an undercount do not -- and, given the nature of statistical science, can never -- agree on precisely how it should be adjusted. It is one of those questions on which reasonable people will always disagree. Given such disagreements, politicians will naturally agree with the expert whose formula does most for their constituencies -- or go shopping for an expert who can do them one better.

We don't think Mr. Koch or Mr. Dymally would intentionally impair the integrity of the Census. But someday someone sitting in one of their places might do so. The country doesn't need a political struggle over Census results every 10 years. It should not start any part of the way down the slippery slope of adjusting the head count. The worthy ends that the adjusters seek can be sought in other ways.