A MODEL FOR WELFARE Goal is prediction of Md. caseload


The poor aren't different from the wealthy. They simply have less money.

And for too long, welfare policy makers haven't realized that, says University of Baltimore economist Michael Conte.

But now the state is testing a model, developed by a team led by Dr. Conte, that forecasts Maryland's welfare caseload by assuming poor people make the same kinds of decisions about their financial lives as wealthy investors: They weigh their options and take the one that pays the best.

And inside the model are some conclusions that could change the way Maryland and many other states plan and fund welfare, Medicaid, food stamps and job training.

The University of Baltimore model indicates that marginal workers' job opportunities were so limited during the July 1990-March 1991 recession, and the jobs they could get provided so few benefits, that recipients of Aid to Families with Dependent Children lost only a few dollars a week staying on welfare.

As a result, the number of people signing up for AFDC rose by about 25 percent from 1989 through 1992.

But if job prospects improve, and employers start offering higher wages and better benefits, the model says that as many as half of the 80,000 Marylanders receiving state aid could shift to work, saving taxpayers hundreds of millions of dollars.

The analysis indicates that about half of the AFDC recipients are "hard core" and would need very high wages and benefits to be lured into jobs.

Dr. Conte, a 45-year-old graduate of Yale University and the University of Michigan, says that the new model could help governments figure out the most effective ways of reducing the number of people signing up for costly Aid to Families with Dependent Children (AFDC) benefits.

The state's Department of Human Resources, which oversees the poverty programs, and which commissioned Dr. Conte's model, already uses two other models to project how much demand poverty agencies will face.

But those models "haven't been as accurate as we would have liked," says Richard E. Larson, director of the DHR's office of program innovation.

The state's other two models are based on sociology, statistics and historical trends, and generally are valid only as long as the economy today behaves exactly as it did in the past, he said.

The idea behind Dr. Conte's model -- which is also the first to let Maryland officials test the effects of various policy -- should be familiar to any economics student or businessperson, but is new to welfare officials, he said.

"It is kind of 'Econ 101,' " he said. But, he explained, "Sociologists and economists didn't talk before."

The model predicted in November that the AFDC caseload will soon stabilize, then drop. The state is waiting to see if that forecast comes true before deciding whether to use the model for budgeting and planning, Dr. Conte said.

Although based in sound economic theory, Dr. Conte said the model is a big gamble. "There was no guarantee," he said, that a series of mathematical formulas would successfully describe what a poet has called "the anarchy of poverty."

Dr. Conte, who moved to Baltimore in 1990 after stints teaching at the University of New Orleans and the University of New Hampshire, said models to predict the economy's course are getting better, just like weather predictions.

"It is very similar to weather predictions," Dr. Conte said. "Over the last 10 years, the kinds of data we can factor in have become tremendously sophisticated."

The new AFDC model took a $300,000 federal grant and 14 months of work by Dr. Conte, two associates and several research assistants.

The model is based on three variables: The number of people likely to need help (mainly, single heads of households with children under the age of 18); the state of the economy; and the economic options people have.

Using Census Bureau figures, one part of the team developed a model to predict the number of single heads of households with children at home in Maryland.

Another group developed a series of leading economic indicators for the state's economy so they could predict unemployment and wage levels.

The search for statistics that accurately predicted economic activity six months or a year later was exhaustive. Much of it was trial and error, said Fereudoon Shahrokh, who is one of the authors of the study.

And in the search, the team discovered that peaks and valleys in AFDC caseload predicted general economic peaks and valleys about a year ahead of time.

"Like Thomas Edison, we tried 2,000 things and one of them worked," he said.

Jane Staveley, another of the authors, spent a year figuring out the full value of the entire package of benefits -- including food stamps, Medicaid and the like -- that a person on AFDC can expect. Then she analyzed how much the same person would expect to earn on a job, and helped develop a mathematical model to track the differences.

When they put all three models together they found that when the difference (or, in economist-speak "net economic gain") between working and not working drops to, say, about $40 a month, as it did in 1990, the number of AFDC cases soars.

But if jobs start paying more, for example, or AFDC payments go down, the model says the caseload will drop.

The state's other models don't take into account people's reactions to AFDC payments or job opportunities, and so don't predict any different scenarios.

And while obvious to business people, that analysis is new to the welfare world, Dr. Conte says.

"Most of the people who look at the issue are sociologists or statisticians. Neither side brought economics to play. In their view, AFDC recipients are caught in a sociological trap. I'm not denying that some of it is true," but the evidence also indicates poor people who stay on AFDC are pursuing their best economic interests, he said.

"In retrospect, some of the most significant breakthroughs are obvious," Dr. Conte says of his model.

But not everyone is convinced that Dr. Conte's model will be useful.

Dr. Harold Beebout, an economist for Mathematica Policy Institute, a New Jersey-based research outfit that developed an AFDC model used now by the Social Security Administration, said complex formulas may not improve a computer model's accuracy.

Some very simple models have an excellent record, he noted. For example, Dr. Beebout said, the federal government gives a very close approximation of AFDC caseload by just counting the number of female-headed households in poverty.

But, he added, although he hasn't seen Dr. Conte's work, he said the reasoning involved is well-accepted and "very sensible."

Copyright © 2020, The Baltimore Sun, a Baltimore Sun Media Group publication | Place an Ad