SCRAMBLING to revive its moribund Haiti policy, the Clinton administration has decided to rule out a compromise with the country's military leaders and to broaden economic sanctions.

The one certain effect of this new policy will be to greatly increase the possibility of U.S. military intervention.

Faced with the tightened embargo, which will go into effect next week, Haiti's rulers will not just throw up their hands and go. They will engage in political maneuvering, such as Wednesday's installation of the 80-year-old Supreme Court Justice Emile Jonassaint as "provisional president."

And they will continue their sick- ening game of chicken with the United States -- clinging to power as the suffering in Haiti goes from terrible to truly horrifying, betting that the United States will lose its stomach for the embargo before it forces them out.

As the stalemate intensifies, the choices facing the Clinton administration will be increasingly stark and agonizing: to admit defeat and call for a lifting of the embargo, to hold tight and bear responsibility for enormous suffering in Haiti with an uncertain prospect of eventual "victory," or to invade.

Faced with these choices, the administration may well go the military route. U.S. presidents can tolerate only so much defiance from tinpot strongmen before they send in the Marines, as George Bush did in Panama in 1989. Haiti has been a source of tremendous frustration for the Clinton administration.

That frustration inevitably combines with the aggravations in Bosnia, Somalia, North Korea and elsewhere to create a besieged mentality in which some unexpected spark -- a daytime massacre of dozens of Haitians or the killing of a group of foreign relief workers -- could push the president to order an invasion.

Military intervention in Haiti may look like a plausible way out of a perplexing policy dilemma. But a U.S. invasion -- and it would be a U.S. invasion, even if Secretary of State Warren Christopher proves successful in his recent efforts to persuade Latin governments to provide multilateral support afterward -- would be a serious mistake.

The political problem in Haiti is not simply that the military refuses to allow the Rev. Jean-Bertrand Aristide, the country's first freely elected president, to return. It is that Haitian society is profoundly polarized between an entrenched business and military elite, some of whom are violently anti-democratic and some of whom are not, and the great mass of ordinary Haitians, who deeply hate the elite and in 1990 chose a leader who promised a fundamental redistribution of power.

A U.S. invasion to "restore democracy" in Haiti, therefore, could not simply aim to pacify the country. It would have to shatter the hold of the violent, anti-democratic members of the elite by eradicating the web of thuggish armed groups that have asserted control across the country.

It would then have to be followed by a long-term occupation to check the extremists on both sides and slowly forge a moderate consensus, which has never existed in Haiti. And it would have to be complemented by an extensive and costly long-range assistance program.

These tasks are enormous. Yet if we invade Haiti we will almost certainly do so trying to minimize our mission.

Mr. Clinton could persuade the extremely reluctant Joint Chiefs of Staff to move only by assuring them that it will be a short-term pacification action. And Americans are likely to support an invasion only if it is short and painless.

Any military action would thus be designed as much to meet constraints at home as to confront the actual challenges in Haiti. Its underlying purpose would be to solve the public-relations problem that Haiti has become for the administration.

Even if the United States does invade with a commitment to overseeing a long-term restructuring of Haitian society, it is unclear that we could succeed.

We can scarcely point to a track record of American invasions' leading to the transformation of highly underdeveloped countries with few democratic traditions into stable, functioning democracies.

Foreign occupations rapidly wear on both the occupiers and the occupied. We might be hailed by Haitians on our way into Port-au-Prince, but we would soon be taking blame for the country's continuing troubles and looking for a way out.

Armed intervention poses problems of principle as well as practice. At root, we have no interest at stake in Haiti so compelling as to warrant unilateral military action.

Some people argue that the United States should promote democracy abroad by force. But a U.S. invasion of Haiti would not be widely seen as upholding such a principle and it would not convince the world that the United States would act decisively when elected leaders in other countries were overthrown.

We hardly blinked at the forcible negation of the elections in Algeria and the ouster of an elected leader in the former Soviet republic of Georgia in recent years.

In this hemisphere, few Latin Americans would be persuaded that we would invade other countries, say Venezuela or Brazil, in response to a coup.

Instead, the lesson the world would extract from an invasion would be that if a U.S. president is humiliated long enough by strongmen in a relatively powerless country with a history of American involvement in its internal affairs, he will send in the Marines.

This would not be an especially impressive lesson.

American military intervention in Haiti would violate international law and would thereby contradict our efforts since the end of the Cold War to promote an international order based on heightened respect for the law.

Within the hemisphere it would set back our goal of moving beyond unilateralism and developing a regional approach to the promotion of human rights and democracy.

Finally, an invasion would set a dangerous example for other regions where our strategic interests are greater. Do we want to send a signal to Russia, for example, that it is acceptable for dominant regional powers to use force to solve their neighbors' internal political problems?

All of our options in Haiti are unsatisfying. Armed intervention is .. among the worst. The Clinton administration must instead consider alternatives that don't depend on the immediate return of Mr. Aristide, up to now a central part of U.S. policy.

An option that should be considered is the formation of a multiparty interim government, fully backed by the legitimately elected members of the Haitian Parliament. This would be part of a U.N.-sanctioned negotiated accord that would permit a substantial U.N. force to oversee next year's elections and stay in place afterward to protect the results.

The paradox of Haiti policy is that the longer we hold out for the best of all possible worlds, the more likely we are to end up with the worst.

Thomas Carothers, senior associate at the Carnegie Endowment for International Peace, is author of "In the Name of Democracy: U.S. Policy Toward Latin America in the Reagan Years."