Has welfare reform overreached? A recent study by the Center on Budget and Policy Priorities found that the poorest fifth of families headed by women lost income between 1995 and 1997, apparently because of changes in welfare.

There is no suggestion that welfare reform has failed. To the contrary, it has--along with a good economy--reduced the welfare rolls by half in the past five years, while raising work levels among poor women and reducing poverty. Even the poorest female-headed families may not actually be worse off. Perhaps their income fell between 1995 and 1997, but their consumption rose. Families do not report all their income. Clear-cut evidence of hardship--such as large increases in homelessness or children driven into foster care--has not appeared.

Everyone who lost income was not abandoned. The center's study does not show that they have left welfare entirely. Many, maybe most, continue to get assistance from various public programs. A survey of welfare leavers in New York City found that more than half of those who were worse off after leaving aid had returned to family welfare or some other social benefit.

But the income losses do underline two problems with the way welfare reform has developed, both unexpected by its architects. One is that the sharp fall in the cash welfare caseload has dragged down receipt of other benefits. When Congress reformed cash welfare in 1996, it made little change in food stamps and Medicaid, which provide groceries and health care to the poor. Poor families were expected to go on drawing these in-kind benefits even as they left cash aid and took jobs. But as people abandon cash, they often leave the other programs as well. That is one reason for the loss in income found in the center's study.

The other problem is that the fall in the rolls is often indiscriminate. Everyone assumed that the welfare rolls would decline from the top down, with the most employable leaving before the less employable. On the whole, that has happened. Surveys of people leaving welfare show that about two-thirds of them are employed, and usually better off for it.

But the decline has become a stampede, taking off many families that are less resilient. Localities have diversion policies designed to steer applicants for aid into work immediately or to help them in short-term ways without putting them on the rolls. That may have discouraged some of the most needy from seeking help. The very fact that they are less competent may leave them unable to figure out how to get aid under the new regime.

This suggests that welfare reform has to ride two horses. On the one hand, it must keep expecting work of the employable. That process has only begun in the biggest cities, including New York. The center's study shows that the second fifth of female-headed families recorded large earnings gains in 1995-97, more than offsetting losses in welfare. That reflects the new work requirements. These parents--with a push from government--are entering the working class.

But at the same time, welfare should reach out to more troubled families that perhaps left welfare too soon, or never went on it. That effort includes following up on those sanctioned--denied aid--for failure to satisfy work requirements, a common way to leave welfare today. Often those differences can be resolved, enabling a family to comply with work and recover some benefits. The most troubled families need a structured environment in which they face some demands to function--perhaps working in a community service job--but get needed support.

It is notable that in a 1997 survey, Wisconsin had unusually low child poverty rates and high work levels among poor parents, despite the greatest welfare decline of any state. One reason is that the fall occurred more gradually there than elsewhere. And despite the confusion of change, the state usually prevented the most troubled families from getting lost. Case managers work hard to set up assignments and services for them. They have the resources to do that exactly because welfare is now supporting many fewer employable people than before.

The point of welfare reform was never to drive everyone off the rolls. It was to ensure that recipients did something to help themselves in return for aid. Those demands should drive the employable off welfare but keep the less capable on. Well done, reform should produce a smaller and more paternalistic welfare system, focused on restructuring the lives of the most needy.

The writer is a professor of politics at New York University.