By the end of this year, 20 to 25 states expect to install the highly controversial "workfare" program, forcing welfare clients to "work off" federal-state welfare benefits, according to a Washington Post survey.

Reagan administration officials are optimistic that workfare can help slash the $14-billion annual federal-state cost of the program of Aid to Families with Dependent Children (AFDC) and measurably cut the rolls, which now total about 3.6 million adults and 7.6 million children.

Undersecretary of Health and Human Services David B. Swoap and other administration officials say workfare has three favorable effects: it gives recipients work experience, it encourages people who could find jobs to do so, because they know they'll have to work in any event, and it provides a valuable service to the community in the form of the work performed.

The administration is so enthusiastic about workfare that it has asked Congress to compel all the states to adopt it. Now it's optional.

But despite glowing reports from places like Utah, which has had a workfare-like program since 1974, many welfare experts say they doubt that a program in which people work off their benefits--with no additional pay above their welfare check except for a $25-a-month carfare and work-expense allowance--can achieve the hopes held out for it.

For one thing, the numbers are against it. According to HHS statistics, four-fifths of all AFDC families consist of a husbandless woman who must take care of one or more children. About three-fifths of all the children are under 6 and not in school. Another one-quarter are from 6 to 11. All with children under 3, and any with children from 3 to 6 who lack child-care aid, are exempt, so the pool of mothers available for work is not going to be as great as it might seem.

But, more important, workfare has been tried repeatedly in the past with doubtful results.

Prof. Leonard Goodwin, head of the Department of Social Sciences and Policy Studies at Worcester Polytechnic Institute, and Dr. Sar Levitan of the Center for Social Policy Studies, said flatly in telephone interviews that there has never been a procedurally valid study supporting the idea that workfare works.

Too often, Goodwin said, it is make-work, with the costs to the local community in extra supervision and tools exceeding the worth of the work done. When such workers are put in jobs that normally would be filled with municipal employes, it amounts to using low-wage, forced "scab labor" in place of normally paid workers, he said.

Although it was hardly mentioned last year, when President Reagan got Congress to allow all states the option of adopting workfare, there was a federally authorized program in 13 states from 1962 to 1967 almost identical to the present one.

At its height about 28,000 people were enrolled, and all told about 100,000, mostly unemployed white fathers in two-parent families.

Musty reports in government files suggest that it was a flop. A 1967 report said that about 45,000 ultimately got non-subsidized jobs. It was generally unfavorable on grounds there were no real job training or work incentives and it wasn't clear that any more had ended up in private jobs later than would have been the case without workfare.

Nevertheless, the idea of making welfare clients work for their checks has enormous appeal, and many states are going forward.

Linda S. McMahon, administrator of the AFDC program for HHS, said in an interview that the past studies were out of kilter with the current situation; many of the earlier experiments, such as California and Massachusetts, were marked by very sharp resistance from local public officials, and that caused disappointing results.

"In California, in some counties it did very well because local government got behind it," she said. Today, she said, there is a more accepting attitude, HHS is giving the states a lot of flexibility and different results can be expected.

In telephone interviews with The Post, many local administrators expressed great enthusiasm that, properly handled, workfare can be converted from a dead-end work-relief program into a genuine learning device, its function being to provide experience outside the home.

West Virginia has started workfare for unemployed fathers receiving AFDC benefits, and it will handle about 2,000 clients a month. New York has begun an experimental program that is expected to cover 15,000 AFDC recipients, out of more than 1 million, in 16 counties.

Other states that told The Post they are likely to go ahead either with demonstrations or larger plans include Massachusetts, Ohio, Idaho Pennsylvania, Utah, Oklahoma, Montana, Georgia, Alabama, Texas, South Dakota, North Dakota, Michigan, Washington, South Carolina, Delaware, Arkansas and North Carolina. Other possibilities are Colo- rado, Kansas, Missouri and Iowa. A Virginia pilot program is possible, but the District of Columbia and Maryland have no plans.

Swoap argues that new figures demonstrate that workfare can have a good effect.

"In Utah, where participation in work experience and mandatory job search are combined," he said recently, "it is estimated that, in 1980, 22 percent of those assigned to the state's workfare program either had their grant reduced or their case closed as a result of new job earnings."

Moreover, he and other administration officials cite, as evidence that work requirements can deter malingering, Colorado's Welt County, where half a group required to engage in intensive job search as a condition of welfare (another administration proposal this year) dropped out or didn't show up and were tossed off.

Usher West, Utah program head, said that in 1977 HHS's Denver regional office calculated that the Utah program had cost $382,000 that year but saved $1,525,000 in benefits from people placed through the program.

"It has been successful in others ways than dollars and cents," he added. "The never-employed or the long non-employed prove to themselves they are capable of working. Work to them is quite frightening."