Wednesday, November 30, 2005

After fierce debate: WIC still works

The Fall 2005 issue of the Journal of Policy Analysis and Management has a great debate over WIC.

Ted Joyce, Diane Gibson, and Silvie Colman argue provocatively that previous research exaggerated the effectiveness of the Special Supplemental Nutrition Program for Women, Infants, and Children. They have great data, from 800,000 births to women on Medicaid in New York City, over many years. But their interpretation is unusual. They find -- as most studies do -- that the children born to women on WIC appear heavier and less likely to be underweight than children born to seemingly comparable women not on WIC. Nevertheless, they tie themselves into knots trying to cast doubt on what is at least one leading explanation for this finding -- that WIC makes for healthier new babies.

Their doubt stems from their belief -- really, an assumption -- that WIC could not possibly affect the length of the pregnancy and the chance that the mother carries the baby to full term. Under their assumption, it makes sense to control for gestational age in seeking to assess the effect of the WIC program. Their regression analyses that do control for gestational age fail to find that WIC is effective. Hence, they doubt that WIC is effective.

In a rebuttal in the same issue, Janet Currie and Marianne Bitler (whose previous work on WIC we reviewed here), argue that Joyce and his colleagues were wrong to reject the possibility that WIC influences the length of the pregnancy. The data certainly seem to indicate that WIC does have such an effect, and moreover, this effect seems plausible based on the array of services that WIC provides.

I know this question about whether to control for gestational age may seem arcane. Think of it this way. If Joyce and colleagues are right, it makes sense to investigate whether WIC women who give birth at 38 weeks have heavier babies than non-WIC women who give birth at 38 weeks. On the other hand, if Currie and Bitler are right, one might find that WIC women tend to have pregnancies 40 weeks long and heavier babies, while non-WIC women tend to have pregnancies 38 weeks long and lighter babies. If WIC may in principle affect the length of the pregnancy, as seems reasonable, then Currie and Bitler seem to have the stronger case.

The debate is so heated that the journal took the unusual step of inviting a third, neutral commentary by two competent social scientists who are not WIC experts, Jens Ludwig and Matthew Miller. As writers, these two are more more polite and less frank than Currie and Bitler. But they agree with Currie and Bitler: "Our reading of the medical literature suggests that there is no conceptual or theoretical reason why the bundle of services provided by WIC could not affect preterm birth rates."

As far as I could tell, none of the three papers sufficiently acknowledged that this debate has all been hashed out previously, as well. In a 1998 book, Feeding the Poor, noted program evaluation expert Peter Rossi rather criticized WIC on much the same grounds as Joyce and colleagues. But, generously, Rossi gave space to Robert Greenstein of the Center on Budget and Policy Priorities for his characteristically lucid response: "Rossi accords too much significance to the fact that findings of WIC's effects in improving birthweight appear to be lessened when one controls for gestational age. The problem here is that controlling for gestational age necessarily understates WIC's effects on birth outcomes."

It is in many ways too bad that the federal government has not supported the clearest type of (ethical!) random-assignment research on WIC. Still, the solid conclusion of all these debates over the non-experimental evidence on birth outcomes is, as always, that WIC works.

Two postscripts. First, related to the question of ethical random-assignment studies, see the interesting discussion in the same issue of the Journal of Policy Analysis and Management -- a very tough ethical critique of current practice by Jan Blustein, and rebuttals by Peter Schochet at Mathematica Policy Research, Inc., and others. My own conclusion is that researchers should read Blustein attentively, and then instead of giving up, seek to use ethical random-assignment research designs whenever possible.

Second, let me boast a bit about my department's history on WIC. At President Nixon's White House Conference on Nutrition in 1969, the chair (and later Tufts University president) Jean Mayer and a colleague Stanley Gershoff (later the first dean of the Friedman School when Mayer was university president) were involved in the discussions that led to the formation of the WIC program in the early 1970s. Current dean Eileen Kennedy also did substantial research on WIC. In the 1980s, David Rush (now emeritus at the Friedman School at Tufts) led the major National WIC Evaluation -- including overseeing the private publication of the unedited conclusions in a special supplement to the American Journal of Clinical Nutrition in 1988 after the Reagan administration edited and censored the original favorable conclusions about the effectiveness of WIC.


Barbara Fisher said...

Thank you for posting about this issue, Parke.

I can give anectdotal evidence that WIC works, but it is nice to know that the research continues to support my assertion that this is a very important social program that deserves adequate funding.

Tybor said...

Parke, very interesting post. I will try to use this example in teaching multiple regression next semester.

Rob said...
This comment has been removed by a blog administrator.
Rob said...
This comment has been removed by a blog administrator.