THE BURST of optimism about welfare reform the past year or two has depended in part on the work of the Manpower Demonstration Research Corp. in New York. This little-known but authoritative organization did some studies of work programs for welfare recipients in the states and found they were having modest but genuine success not merely in moving recipients off the rolls but in raising their incomes while reducing program costs.

Lost in the rush of enthusiasm was an important cautionary note. The researchers said that for various reasons -- lack of jobs, not enough resources -- they weren't confident that even this modest success rate could be sustained in larger programs; the ones they had studied were mainly small. Now MDRC has issued a report on such a larger program, in Cook County, Ill., and has found that indeed it was not successful in speeding people off the rolls.

The new finding is no more conclusive than the old. It nevertheless should figure in the welfare debate that continues in Congress. In the House last year, Republicans offered an alternative to the Democratic welfare bill that finally passed. They said their bill would do more for less, in part because it would have required states to have fixed percentages of their welfare mothers in work programs within a certain number of years. The Democrats said the Republicans weren't providing even remotely as much money as such an effort would require to succeed. The new study suggests -- it doesn't prove -- that the Democrats were right.

The programs MDRC is studying were authorized by Congress early in the Reagan administration. The administration had wanted to require welfare recipients to work. Congress voted to let the states experiment with various work requirements of their own instead. The experiments have provided a ready-made work-and-welfare laboratory.

It is difficult to study such programs because many people are leaving welfare all the time. A governor can claim his work program moved some impressive percentage of mothers off welfare, and it turns out most of them would have moved off anyway. To guard against this, MDRC used control groups. In the earlier studies, it found that some of the work programs had raised the employment rate as much as 5 percent -- an important difference. Among younger mothers who otherwise seemed likely to become chronic welfare recipients, the degree of success was even greater. But in Cook County the success rate for those in the work program turned out to be exactly the same as for those on the regular welfare rolls.

If welfare reform could be done with the snap of a finger, it would have been done long ago. It is complicated work that -- on a scale that matters -- can't be successfully done on the cheap. The Senate needs to remember that as it fashions its bill.