Experimental work programs in San Diego, Baltimore and Arkansas are showing marked success in helping welfare mothers find jobs and reducing government welfare costs, a private research group reported yesterday.

According to the Manpower Demonstration Research Corp., persons in the experimental groups found more jobs, earned more money and cost the government less than did welfare mothers in a control group who did not take part.

In San Diego, for example, 61 percent of the group obtained regular jobs over 18 months, compared to 55.4 percent of the control group.

The experimental group averaged $3,802 in earnings and $3,409 in payments from the Aid to Families with Dependent Children program. For the control group, earnings averaged $3,102 and welfare payments averaged $3,697.

Similar results were reported in Arkansas and Baltimore.

Barbara Blum, president of the MDRC, a nonprofit social science research organization, said the study shows the success of several approaches -- all including an intensive job search when a person first applies for welfare, then a period of "workfare," unpaid job experience.

The MDRC study was funded by the Ford Foundation and other foundations. Eventually projects in 11 states will be evaluated in the first comprehensive study of work experiments with welfare recipients. Its findings are expected to be important in any national overhaul of welfare policy.

President Reagan has directed a Cabinet group to study how welfare programs might be changed to emphasize work and reduce dependency. Reagan endorses workfare but many liberals say it is punitive and unproven.

Judith Gueron, MDRC project director, said the success of the first three experiments should not be viewed as an endorsement of the old-style workfare concept involving meaningless jobs and no marketable skills.

She said the unpaid job experience in the three programs studied usually lasted no more than 13 weeks and provided training in marketable office and service skills.

In the San Diego experiment, involving welfare mothers and fathers, new applicants for AFDC took part in an intensive three-week job search program. In the first week they were taught how to make phone calls, prepare resumes and conduct themselves in interviews. The subsequent weeks were spent at phone banks calling prospective employers.

Some of those unable to obtain jobs spent up to three months on workfare with no pay beyond their welfare benefits, in order to prepare them for working and to develop skills and attitudes.

In San Diego, MDRC found that the mothers did best when involved in both job search and unpaid work, but that those involved in job search only, and not channeled to workfare later, still did better than the control group.

The MDRC found that while welfare mothers showed definite gains, welfare fathers did not. Gueron said many fathers already had job skills and did reasonably well at finding jobs on their own.

In Arkansas, where the study covered only welfare mothers, members of the experimental group after nine months had a one-third higher employment rate than the control group, were earning $78 more and receiving $93 less in welfare payments.

In Baltimore, which also had a training program, 51 percent of the experimental group obtained jobs over 15 months compared to 44 percent of the control group, and the experimental group averaged $1,935 in earnings compared to $1,759 for the control group. However, average welfare payments were only $6 less than the control group's.

The San Diego study covered 6,997 persons applying for welfare at various times in 1982 and '83; the control group comprised about one-third of them.

In the Arkansas experiment, 1,153 persons were studied.

The Baltimore study involved 2,823 mothers, including the control group.