The Journal of Nutrition Vol. 128 No. 3 March 1998, pp. 548-555

Household Participation in the Food Stamp and WIC Programs Increases the Nutrient Intakes of Preschool Children¹

Donald Rose, Jean-Pierre Habicht^{*, 2}, and Barbara Devaney

Economic Research Service, U.S. Department of Agriculture, Washington, DC 20005; ^* Division of Nutritional Sciences, Cornell University, Ithaca, NY 14853; and  Mathematica Policy Research, Inc., Princeton, NJ 08543

	ABSTRACT

Abstract Introduction Methods Results Discussion References

We examined the effects of household participation in the Food Stamp and WIC Nutrition Programs on the nutrient intakes of preschoolers using data from the 1989-1991 Continuing Survey of Food Intake by Individuals. Nonbreastfeeding children, 1-4 y of age, with 3 d of dietary data and whose households had incomes < 130% of the poverty level were included in the study sample (n = 499). Nutrient adequacy ratios for each of 15 nutrients were the dependent variables in multiple regression models that controlled for the following: age, sex and ethnicity of the individual; income, size and location of the household; schooling of the household head; home ownership; school lunch and breakfast participation; and season in which the interview was conducted. WIC benefits positively influenced (P < 0.05) the intakes of 10 nutrients. For iron and zinc, the average increase due to WIC represented 16.6 and 10.6%, respectively, of the preschooler recommended dietary allowance (RDA) for these nutrients. The same analyses of the Food Stamp Program revealed increases in five nutrients. For iron and zinc, the average increase due to Food Stamps represented 12.3 and 9.2%, respectively, of the preschooler RDA. The effects of the WIC Program on the intakes of iron and zinc were greater than that of cash income, and neither program affected the intakes of fat, saturated fat or cholesterol.

	INTRODUCTION

Abstract Introduction Methods Results Discussion References

The recent debate on welfare reform in the United States has increased the importance of assessing the effects of existing food assistance programs. Domestic food programs such as the Food Stamp Program and the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) are programmatically distinct from welfare, but came up for reappraisal because of the sizable government expenditure in this area. Spending on the USDA's food assistance programs totaled $37.8 billion in fiscal year 1995 (Oliveira 1997), an amount over twice that of Federal expenditures on the former Aid to Families with Dependent Children (AFDC) program, previously referred to as welfare.

The importance of the welfare reform debate for the future of food assistance programs in the United States did not go unnoticed in the nutrition community (Kennedy et al. 1996, Kretchmer 1995, McNutt 1995). However, nutrition researchers have studied large-scale domestic food assistance programs less frequently than curative programs, which have long been part of the international nutrition literature (Martorell et al. 1976, Pinstrup-Andersen et al. 1993, Scrimshaw et al. 1967). In part, this is due to difficulties in evaluation that are inherent in existing programs. Experimental methods are often not an option, for legal, ethical and/or practical reasons. For example, because the Food Stamp and School Lunch Programs are entitlements, benefits cannot legally be denied to those who qualify for these programs. Thus, a randomized control group design is not possible.

In recent years, there has been increasing application of advanced statistical techniques to the evaluation of programs in which experimental designs are not possible. Multivariate regression analysis has been used to control for observed differences between food program participants and eligible nonparticipants (Devaney et al. 1992, Rush 1986). The effect of the Food Stamp Program on the availability of nutrients in at-home food supplies (Devaney and Moffitt 1991) and the effect of the School Lunch Program on dietary intakes of children (Burghardt and Devaney 1995) are two studies that used both multivariate analyses and selection-bias models. The latter, originally developed in the field of economics, control for the bias that may result from nonrandom selection of participants for food assistance programs (Fraker et al. 1995).

Despite these recent examples, there continue to be sizable gaps in the literature on domestic food program evaluation. Although there has been extensive study on the effects of the Food Stamp Program on food spending (Fraker 1990, Levedahl 1991 and 1995) and a number of studies of the effects of this spending on household nutrient supplies (Basiotis et al. 1983, Devaney and Moffitt 1991, Lane 1978), relatively little work has been done on the effects of Food Stamps on nutrient intake. Only three national-level studies have addressed the intakes of preschoolers (Cook et al. 1995, Fraker et al. 1990, Rush et al. 1988). There was no consistent pattern found in looking at these studies; the Food Stamp effect varied by nutrient and was often not significant.

Considerable work has been done to evaluate the effects of the WIC Program on pregnant women, but this has not been the case for preschool children, a group that makes up over half of the WIC participant caseload (USDA/FCS 1996). In preparing a report for Congress, the U.S. General Accounting Office could rely on the results from 17 samples in which the prenatal effects of WIC on low birth weight rates were studied (USGAO 1992). Meta-analyses of this sort are simply not possible with respect to outcomes in children. For example, only two recent studies have been conducted at the national level on the nutrient intake effects of WIC for preschoolers (Fraker et al. 1990, Rush et al. 1988).

A third gap in the literature concerns the relative benefits of different types of interventions. Much of the Food Stamp literature cited previously has considered the relative benefits of cash vs. in-kind (i.e., food stamps) assistance. Contrary to what would be predicted by economic theory, empirical studies have shown consistently that food stamps do indeed increase total food spending of recipient households beyond what cash assistance would accomplish (Fraker 1990). The relative effects of more targeted assistance, e.g., WIC coupons for specific kinds of foods such as iron-fortified cereals, have been studied much less. Arcia et al. (1990) investigated the effect of WIC program participation on food expenditures, but did not estimate effects relative to the Food Stamp Program or effects on nutrient intake. If recipients use WIC coupons to purchase iron-fortified cereals that they would otherwise purchase with their own money, there will be no net increase in the consumption of these cereal products. Given such behavior, cash, WIC coupons and Food Stamps would have similar effects on nutrient intake.

To address these gaps in the literature, we examined recent dietary intake data of preschool children whose households were income eligible to participate in both the Food Stamp and WIC Programs. Our main question in this work is the following: to what extent do each of these programs affect the nutrient intakes of preschool children? We used multivariate analyses to control for a number of observed characteristics that might influence diet, so as to isolate the effects of participation in both of these programs on intake. We also assessed the potential effects of selection bias on these results by using a model adapted from the economics literature.

Although it is important to study the effects of programs on the entire diet, certain nutrients merit special attention. In the United States, iron deficiency is the single most prevalent nutritional deficiency (HHS/USDA 1989) and is considered a current public health issue because of adverse associated health conditions (FASEB/LSRO 1995a). Although trend data for a few states have shown a decrease in the anemia prevalence rates among low income preschoolers (Yip et al. 1987), the rates are still high, ranging from 15 to 20% for those <2 y of age (Yip et al. 1992). Mean intakes of zinc are also low in certain segments of the population (Tippett et al. 1995), and previous studies have suggested a relationship between zinc and growth retardation (FASEB/LSRO 1995a). For these reasons, we highlight iron and zinc in our discussions of program effects.

	MATERIALS AND METHODS

Abstract Introduction Methods Results Discussion References

Study sample.  The analyses presented here are based on data obtained from the 1989-1991 Continuing Survey of Food Intake by Individuals (CSFII). The CSFII is currently conducted by USDA's Agricultural Research Service and forms an integral part of the National Nutrition Monitoring and Related Research Program (FASEB/LSRO 1995a). The 1989-1991 version of the survey was conducted by the former Human Nutrition Information Service of USDA and was based on two independent stratified clustered samples of housing unitsa basic, or all-income sample, and a low income sample. The samples were drawn from the 48 coterminous states and Washington, DC. (USDA/HNIS 1992). Detailed descriptions of sampling procedures and response rates have been published previously (Tippett et al. 1995).

Our analytical sample consisted of children who were age eligible for the WIC Program and whose households were income and asset eligible for the Food Stamp Program. Preschoolers who were at least 1 y of age, but had not reached their fifth birthday met the age criterion. Children from households with monthly income that was <130% of the federal poverty guidelines and those that did not have cash assets >$2000 met the income and asset criteria for inclusion in our sample. It should be noted that eligibility for Food Stamps is a complex determination, involving income, various expenses such as shelter and health care, as well as assets. Because of the complexity of the determination and the difficulty in collecting certain types of information, no data set can provide an exact indication of the eligible population. The gross income criterion used here has been used extensively in previous studies on Food Stamps. We added to that a second criterion (cash assets) to improve our screening for eligibility. Those households that met our income eligibility screen also would have met the income eligibility requirement for WIC, because the WIC income requirement is more generous, i.e., 185% of the federal poverty guidelines.

All Hispanic, non-Hispanic White and non-Hispanic Black children who met these criteria were included in our sample if they had 3 d of food intake data as well as the socioeconomic data (described below) required for multivariate analyses. We excluded children from other racial-ethnic groups because of inadequate sample sizes. We also excluded children who breast-fed on the survey days, because the CSFII did not estimate nutrient intake from breast milk. The final preschooler sample consisted of 499 children.

Dietary data.  The CSFII collected up to three consecutive days of dietary data on individuals; the first day was based on a 24-h recall and the second two days were based on food records. All dietary information for preschoolers was obtained from the household respondent, who in most cases was the child's mother. Interviewers, who were knowledgeable about foods and food preparation, were trained in the use of a food instruction booklet to help each individual describe the foods and the amounts eaten. During the 24-h recall, interviewers used standard household measuring cups and spoons and a ruler to estimate specific quantities consumed. Interviews were conducted in the home, a setting that allowed the interviewer to gather package-weight and ingredient information from food labels as well as descriptions of home-prepared mixtures from the food preparers (Tippett 1995).

The amount of nutrients in each food was calculated using the weight of the food consumed and the nutritive value of that food. The latter information was obtained from a nutrient data base developed by the USDA that contained ~6700 items (Perloff et al. 1990, USDA/HNIS 1985).

Dependent variables used in these analyses were the intakes of nutrients from food sources, averaged over 3 d. These intakes were expressed as nutrient adequacy ratios, that is, the intake of a nutrient by an individual divided by the recommended dietary allowance (RDA) for that individual and expressed in percentage terms. For energy, intake was divided by the recommended energy intake for a reference person engaged in light-to-moderate activity (NRC 1989). Nutrients studied in these analyses included the following: energy, protein, vitamin A, vitamin E, vitamin C, thiamin, riboflavin, niacin, vitamin B-6, folate, calcium, phosphorus, magnesium, iron and zinc. In addition, the percentage of calories from fat, the percentage of energy from saturated fat and the milligrams of cholesterol per 1000 kcal (4.18 MJ) in preschoolers' diets were also studied.

Food program participation variables.  Our main focus in this study concerned the effects of household participation in the Food Stamp and WIC Programs. To study these effects in a comparable fashion, we created variables that represented the dollar value of benefits from each program. The CSFII collected self-reported information on the monthly monetary value of food stamps received by the household. We divided this figure by household size to obtain a per capita amount for each household. The CSFII also asked each respondent in the household whether they received benefits under the WIC program. We multiplied the number of WIC beneficiaries per household by the average monetary benefit of a WIC packet and divided by household size to arrive at a monthly per capita value of WIC benefits received by the household. The use of adult equivalents instead of household size in these calculations yielded conclusions similar to those reported here.

Although the National School Lunch and Breakfast Programs were not the focus of our study, participation of older children in a household can affect intakes of preschoolers by freeing up food dollars to feed their younger siblings. To control for this possible effect, we created variables representing the dollar value of benefits from each of these programs. The CSFII collected information on the number of meals per week consumed by each child participating in these programs. We multiplied the number of meals by the value of each meal to arrive at a benefit amount per month for each child. We then summed the benefits for all children in each household and divided by household size to yield a per capita monthly value of benefits from each of these programs.

Statistical methods.  Dietary intake is influenced by a number of social, cultural and economic factors. To assess the quantitative relationship between food program benefits and intake, it was important to control for these confounding variables; single equation multivariate regression analyses were performed using SAS (SAS Institute 1990). In these analyses, the nutrient adequacy ratios were the dependent variables, and the food program participation variables as well as the socioeconomic variables described below were used as independent variables. We considered coefficient estimates to be significantly different than zero at P  0.05. In preliminary analyses, we did not find significant interactions between the Food Stamp and WIC Program variables; participation in one program did not increase or decrease the effects of the other. Therefore, we did not include interaction terms in our models.

A number of variables were used to describe the socioeconomic circumstances of the household. For each household, per capita income from the previous month was used as an independent variable in multiple regression analyses. A dichotomous variable indicating home ownership was used as a measure of wealth. Household size was also used as an independent variable. The number of years of schooling completed by the household head was used to control for educational background.

Several variables controlled for individual-specific characteristics. Race and ethnicity were incorporated into two dichotomous variables that indicated whether the preschooler was an Hispanic or a non-Hispanic black; the reference group was non-Hispanic whites. Indicators were also used for each year of age of the preschoolers, with those 1 y of age being the reference group. A variable to indicate whether the individual was a male was also included in the regressions.

To account for differences in regional food habits, three dichotomous variables were used to indicate whether households came from the northeast, the south or the west; those from the midwest were the reference. Indicator variables were also used to control for urbanization, that is, whether the household was located in a central city, a nonmetropolitan area or in a suburb (reference group).

Several variables accounted for the effects of the time dimension on food consumption. Three variables were used to control for seasonal differences in food consumption by indicating whether the observation was in the second, third or fourth quarter of the year; the first quarter was the reference. Two variables were also used to control for the year of the interview, 1990 or 1991, with 1989 taken as the reference value.

The use of sampling weights in regression models can lead to an inefficient analysis (Korn and Graubard 1991). DuMouchel and Duncan (1983) presented a statistical test for assessing when an unweighted analysis is preferred over a weighted one. On the basis of a test with energy, which is the best proxy for food consumption overall, we did not find sufficient justification to use weighted analyses. Because the CSFII is based on a stratified, clustered sample design, variance estimates can be understated if standard statistical software is used. To account for the complex design, we calculated our standard errors using a replicate sample technique, known as balanced repeated replication, supported under WesVarPC (Westat 1996).

	RESULTS

Abstract Introduction Methods Results Discussion References

Descriptive results.  For descriptive purposes, we classified preschoolers in our analytic sample into the following four groups: those whose households did not participate in either the Food Stamp or WIC Programs; those who participated in Food Stamps, but not WIC; those who participated in WIC, but not Food Stamps; and those who participated in both. Average food program benefit levels and socioeconomic characteristics of these groups are presented in Table 1. Sample households that participated in the Food Stamp Program received a monthly average of $55.46 per person. WIC households received an average of $12.47 per person in food coupons. There were a number of differences in average levels of the socioeconomic characteristics between these groups. For example, households that participated only in WIC or that participated in neither program had higher incomes, were more likely to be homeowners, were less likely to be non-Hispanic blacks and were less likely to live in central cities than households that participated in the Food Stamp Program. Although preschoolers from households that participated in the WIC Program were somewhat younger than those that did not, there was considerable age variation within each of the four groups; such variation also existed for the other economic and sociodemographic characteristics studied.

The unadjusted mean nutrient intakes for these four groups of preschoolers are displayed in Table 2. Because of differences in demographic and socioeconomic characteristics across the groups (Table 1), differences in intakes observed in Table 2 cannot be ascribed to the food programs alone. However, values in Table 2 are useful for ascertaining a general picture of intakes. For example, protein intakes were high for preschoolers; group means ranged from 2.6 to 3.0 times the RDA. For most other nutrients, mean intakes were >100% of the RDA. Exceptions to this case included intakes of energy, vitamin E, calcium, iron and zinc. The mean percentage of calories from fat ranged from 34.7 to 36.4%.

Because individuals with consumption below a certain level may not be adequately described by estimates of the mean, we also created dichotomous variables to reflect consumption that was <50% of an individual's RDA. We chose a particularly conservative cut-off because the RDA are generously set to meet the needs of practically all healthy persons and because underreporting is likely in diet surveys. The percentages of each of the four preschooler groups with intakes below this 50% cut-off are displayed in the last four columns of Table 2. Vitamin E, calcium, iron and zinc were the nutrients with the greatest number of low intakes.

Multivariate results.  Multiple regression analysis allowed us to study the relationship between food program benefits and nutrient intake of preschoolers, while controlling for socioeconomic factors. To give an example of the regression model that we used for all our analyses, we highlight the results of our analysis on iron (Table 3), a priority nutrient for public health monitoring. Per capita monthly income as well as food program benefits for three of the programs were positively related (P < 0.05) to iron intake. Controlling for all other factors, a dollar of food stamps per person in the household was associated with an increased iron intake by preschoolers of 0.22 RDA percentage points. For WIC, the coefficient was 1.33 RDA percentage points per capita dollar of benefit. The magnitude of the Food Stamp effect was twice that of cash income, although the difference in the coefficients on these two variables was not significant (P = 0.16). The WIC effect was significantly greater than the cash income effect and over 12 times the magnitude. Age and region were also significant explanatory variables in this model. Older children have increased food intakes, but the same RDA for iron. Therefore, it was not surprising that 3- and 4-y-old children had iron nutrient adequacy ratios that were significantly greater than the reference group of children who were 1 y old. Those in the South had significantly lower intakes of iron.

Table 3 also displays the results of our zinc intake model, which were similar in a number of ways to the iron results. Controlling for all other factors, Food Stamp participation was associated with an increased intake of zinc by 0.16 RDA percentage points per dollar of benefit received per person in the household. For WIC, the coefficient was 0.85 RDA percentage points per capita dollar of benefit. The Food Stamp coefficient was over twice the size of the coefficient on cash income, although the difference in these coefficients was not significant (P = 0.13). The WIC effect was significantly greater than the cash income effect and over 14 times the magnitude. As in the iron model, 3- and 4-y-old children had significantly higher nutrient adequacy ratios than did the reference group of 1-y-old children. Unlike the iron model, differences in regional consumption were not significant. However, even after controlling for all other factors, non-Hispanic Blacks had intakes that were significantly lower than non-Hispanic Whites by about 11 RDA percentage points.

The multiple regression models used in Table 3 were followed for all nutrients studied. The first three columns of Table 4 summarize these results for the Food Stamp Program. In addition to iron, the intakes of vitamin A, thiamin, niacin and zinc were significantly related to the amount of the food stamp benefits, ranging from 0.16 to 0.48 RDA percentage points per capita dollar of benefit. The intakes of energy, fat, saturated fat and cholesterol were not significantly related to Food Stamp benefits. The third column in Table 4 reports the total changes in nutrient intake associated with average Food Stamp benefits, that is, we multiplied the change in nutrient intake per dollar of benefit by the average monthly Food Stamp benefit level, $55.46 per person. Average program benefits would result in an increase for preschoolers of 12.3 and 9.2 RDA percentage points of iron and zinc, respectively.

The last three columns of Table 4 summarize our regression results for the WIC Program. Intake was positively associated with the WIC Program benefit for 10 of the nutrients studied; the effect ranged from 0.85 to 5.25 RDA percentage points per capita dollar of benefit. As with the Food Stamp Program, the intakes of energy, fat, saturated fat and cholesterol were not significantly related to the WIC benefit. The average monthly WIC benefit level for participating households in our sample was $12.47 per person. Multiplying this amount by per dollar increases suggests that, in the case of iron and zinc, the increase due to WIC would be 16.6 and 10.6%, respectively, of the preschooler RDA.

	DISCUSSION

Abstract Introduction Methods Results Discussion References

Our analyses indicate that average Food Stamp Program benefits resulted in an increase for preschoolers of ~1.2 mg/d of iron, and ~0.9 mg/d of zinc. The iron estimate was within the range of previous studies on Food Stamps and preschoolers. Using multivariate models, Rush (1986) found an increased iron intake of 1.7 mg/d associated with Food Stamp participation, although this result was not significant. In their analysis of 1985 CSFII data, Fraker et al. (1990) also found a positive relationship, although not a significant one, between Food Stamp participation and preschooler nutrient intake. On average, this amounted to an iron intake of 0.7 mg/d. In the same study, these authors found a significant Food Stamp effect on zinc intake of 1.3 mg/d (Fraker et al. 1990). Our results, consistent with these earlier studies, were more conclusive as a result of the increased power in our study. Cook et al. (1995) used 1986 CSFII data in a descriptive study of the intakes of preschool children from Food Stamp and nonparticipant households. They found that a lower percentage of Food Stamp preschoolers had zinc intakes < 70% of their RDA than did nonprogram children, a finding that is also consistent with our results.

The effects of the WIC Program that we found were somewhat greater than those of previous studies. We found the average increase due to WIC was 1.7 and 1.1 mg/d of iron and zinc, respectively. Rush et al. (1988) found a 1.2 mg/d significant increase in iron associated with current WIC participation. Fraker et al. (1990) found a 0.4 mg/d increase in iron intake and a 0.4 mg/d increase in zinc intake due to WIC, although neither result was significant.

Our models showed that for 10 of the 15 nutrients we studied, WIC had significant and positive effects on intakes by preschoolers; for Food Stamps, this was true for five of the nutrients studied. We found the effects of WIC to be stronger than those of Food Stamps in terms of the dollar of benefit received per person in the household. For iron, the WIC effect was significantly different than the Food Stamp effect and about six times greater. The WIC coefficient in the zinc intake model was also much larger than the Food Stamp coefficient, although the difference did not reach significance (P = 0.08). The enhanced effect of WIC may reflect the fact that specific types of nutritious foods are given to recipients; it also may be due to nutrition education, which is an integral part of WIC. Our results are based on a sample of households with incomes < 130% of the poverty level. Although the income-eligibility cut-off for the WIC Program is 185% of the poverty level, only 12% of all WIC participants had incomes between 130 and 185% of the poverty level in 1990 (Burich and Murray 1992). The responses of these participants to the WIC Program could have been greater or less than what we found in our sample, but it is unlikely that this small proportion would have had much effect on our results.

Although both the Food Stamp and WIC Programs have positive and significant effects on the intake of a number of important nutrients, our results indicated no significant relationship between household participation in these programs and the percentage of energy from fat, saturated fat or cholesterol in the diets of preschoolers. Data from the Third National Health and Nutrition Examination Survey show that the percentage of energy intake from fat was higher for non-Hispanic Black and Mexican-American WIC participants than for nonparticipants of the same ethnic groups (NCHS/CDC 1995). In addition, descriptive statistics from the 1989-1991 CSFII show that a greater percentage of adult Food Stamp participants have high fat intakes than do nonparticipants (FASEB/LSRO 1995b). Neither of these studies, however, used multivariate models to control for confounding variables.

A potential limitation of this research is that our analyses were based on dietary information obtained over three consecutive days, an instrument that is less reliable at describing usual intake than multiple independent dietary recalls (Sempos et al. 1985). In addition to the normal day-to-day variability in dietary practices, households eligible to participate in food assistance programs may have increased variability in intakes associated with limited financial resources and monthly cash-flow problems. However, the increased variability introduced by these factors should not have affected our estimates of mean intakes but would have increased the size of our confidence intervals around these estimates, making it less likely to find significant differences that existed. In this sense, our results can be considered conservative.

Underreporting is a common problem for dietary surveys and the 1989-1991 CSFII was no exception to this rule. If such underreporting occurred to a greater extent among eligible nonparticipants, then our results would be overstated. We have no reason to believe this to be the case. We think it is more likely that such asymmetric underreporting, if it existed, was associated with one or more of the observed factors for which we have controlled. For example, our results were based on models that controlled for level of education, which has been shown to be negatively associated with underreporting (Briefel et al. 1997).

As can be seen from Table 1, there were differences between participants and nonparticipants in various socioeconomic characteristics. Clearly, an experimental design would have reduced this between-group variation and would have provided more reliable results regarding the effects of participation in food assistance programs. Because such a design was not possible, we used national survey data and statistical techniques. Multivariate analyses controlled for observed factors that could have confounded the relationship between program benefits and nutrient intake.

Because individuals in our study were not randomly assigned to treatment and control groups, it is possible that our results were affected by a selection bias. This bias can occur when there are no statistical controls for differences between groups that affect not only outcomes of interest, but also the decision to participate in food programs. For example, being motivated by nutrition presumably improves dietary outcomes as well as influences some to seek out food assistance. Because eligible individuals are entitled to participate in Food Stamps if they so choose, there is no random assignment and thus our Food Stamp group could be composed of more people who are nutrition motivated. In this scenario, without controlling for the hidden characteristic of "nutrition motivation," a comparison of Food Stamp Program participants with eligible nonparticipants could have overstated the true Food Stamp effect. The bias could also work in the opposite direction. Those participating in Food Stamps generally have lower incomes than eligible nonparticipants. To the extent that poverty is correlated with poor dietary habits and if aspects of this poverty are not well captured by observed variables such as income and education, then a selection bias could result in which regression results understate the true program effect. A priori, there is no way to know in which direction a selection bias will work or if there will be one at all.

We investigated selection bias using a model that was originally developed in the field of economics (Heckman 1979) and has been well established in the literature on food assistance and other social program evaluation (Fraker 1995). The technique is based on a two-stage model, run on LIMDEP software (Greene 1992). In the first stage, proxy variables are created for the unobserved characteristics that affect food program participation (such as "nutrition motivation") by using information on both actual and predicted participation. In the second stage, these proxy variables, one for each program, are inserted into the multiple regression models run previously. Using this technique, we did not find evidence of selection bias in our results, a finding that has been documented previously for the Food Stamp Program (Devaney and Moffitt 1991).

In conclusion, we find that household participation in the Food Stamp and WIC Programs had independent and significantly positive effects on the nutrient intake of preschoolers. We consider these effects to have public health significance for iron and for other nutrients as well, e.g., zinc. On average, the Food Stamp and WIC Programs increased the intake of iron, ~12 and 17 RDA percentage points, respectively. This is likely to be the most important effect, because iron deficiency is the single most prevalent nutritional deficiency in the United States (DHHS/USDA 1989) and anemia rates are still high among low income children < 2 y of age (Yip et al. 1992). The beneficial effects of the WIC Program on iron and zinc intakes are not only significant in public health terms, but are also much greater than what would have been achieved with increases in cash income. Furthermore, our results indicate that these benefits come without the potentially deleterious effects of increases in the intake of fat, saturated fat and cholesterol.

	FOOTNOTES

Manuscript received 3 June 1997. Initial reviews completed 9 July 1997. Revision accepted 10 November 1997.

	LITERATURE CITED

Abstract Introduction Methods Results Discussion References

• Arcia G. J., Crouch L. A., Kulka R. A. Impact of the WIC Program on food expenditures. Am. J. Agric. Econ. 1990; 72:218-226
• Basiotis P. P., Brown M., Johnson S. R., Morgan K. J. Nutrient availability, food costs, and food stamps. Am. J. Agric. Econ. 1983; 65:685-693
• Briefel, R. R., Sempos, C. T., McDowell, M. A., Chien, S. & Alaimo, K. (1997) Dietary methods research in the Third National Health and Nutrition Examination Survey: underreporting of energy intake. Am. J. Clin. Nutr. 65 (suppl.): 1203S-1209S.
• Burghardt J. A., Devaney B. L. Background on the School Nutrition Dietary Assessment Study. Am. J. Clin. Nutr. 1995; 61:178S-181S [Abstract/Free Full Text]
• Burich, M. C. & Murray, J. R. (1992) Study of WIC Participants and Program Characteristics, 1990. Abt Associates, Chicago, IL.
• Cook, J. T., Sherman, L. P. & Brown, J. L. (1995) Impact of Food Stamps on the Dietary Adequacy of Poor Children. Tufts University Center on Hunger, Poverty, and Nutrition Policy, Boston, MA.
• Devaney B., Bilheimer L., Schore J. Medicaid costs and birth outcomes: the effects of prenatal WIC participation and the use of prenatal care. J. Pol. Anal. Management 1992; 11:573-592
• Devaney B., Moffitt R. Dietary effects of the Food Stamp Program. Am. J. Agric. Econ. 1991; 73:202-211
• DuMouchel W. H., Duncan G. J. Using sample survey weights in multiple regression analyses of stratified samples. J. Am. Stat. Assoc. 1983; 78:535-543
• Federation of American Societies for Experimental Biology, Life Sciences Research Office (1995a) Third Report on Nutrition Monitoring in the United States: Volume 1. U.S. Government Printing Office, Washington, DC.
• Federation of American Societies for Experimental Biology, Life Sciences Research Office (1995b) Third Report on Nutrition Monitoring in the United States: Volume 2. U.S. Government Printing Office, Washington, DC.
• Fraker, T. M. (1990) The Effects of Food Stamps on Food Consumption: A Review of the Literature. Mathematica Policy Research, Washington, DC.
• Fraker, T. M., Gordon, A. R. & Devaney, B. L. (1995) The use of selection-bias models to evaluate food and nutrition programs. Proceedings of the Section on Government Statistics, Annual Meeting of the American Statistical Association, pp. 222-225. American Statistical Association, Alexandria, VA.
• Fraker, T. M., Long, S. K. & Post, C. E. (1990) Analyses of the 1985 Continuing Survey of Food Intakes by Individuals, Volume I. Estimating Usual Dietary Intake, Assessing Dietary Adequacy, and Estimating Program Effects: Applications of Three Advanced Methodologies Using FNS's Four-Day Analysis File. Mathematica Policy Research, Washington, DC.
• Greene, W. H. (1992) LIMDEP User's Manual and Reference Guide, Version 6.0. Econometric Software, Bellport, NY.
• Heckman J. J. Sample selection bias as a specification error. Econometrica 1979; 47:153-161
• Kennedy E., Morris P. M., Lucas R. Welfare reform and nutrition programs: contemporary budget and policy realities. J. Nutr. Educ. 1996; 28:67-70
• Korn E. L., Graubard B. I. Epidemiologic studies utilizing surveys: accounting for the sampling design. Am. J. Public Health. 1991; 81:1166-1173 [Abstract/Free Full Text]
• Kretchmer N. Good nutrition is the issue. Pediatrics 1995; 95:937-938 [Abstract/Free Full Text]
• Lane S. Food distribution and Food Stamp Program effects on food consumption and nutritional achievement of low income persons in Kern County, California. Am. J. Agric. Econ. 1978; 60:108-116
• Levedahl, J. W. (1991) The Effect of Food Stamps on Household Food Expenditures. Technical Bulletin No. 1794. Economic Research Service, USDA, Washington, DC.
• Levedahl J. W. A theoretical and empirical evaluation of the functional forms used to estimate the food expenditure equation of Food Stamp recipients. Am. J. Agric. Econ. 1995; 77:960-968
• Martorell R., Lechtig A., Yarborough C., Delgado H., Klein R. E. Protein calorie supplementation and postnatal physical growth: a review of findings from developing countries. Arch. Latinoam. Nutr. 1976; 26:115-128 [Medline]
• McNutt K. What are "block grants" all about? Nutr. Today 1995; 30:84-88
• National Center for Health Statistics/Centers for Disease Control, U.S. Department of Health, Human Services Nutritional status of children participating in the Special Supplemental Nutrition Program for Women, Infants, and ChildrenUnited States, 1988-1991. Morb. Mortal. Wkly Rep. 1995; 45:65-69
• National Research Council (1989) Recommended Dietary Allowances, 10th ed. National Academy Press, Washington, DC.
• Oliveira V. Food-assistance spending held steady in 1996. Food Rev. 1997; 20:49-56
• Perloff B. P., Rizek R. L., Haytowitz D. B., Reid P. R. Dietary intake methodology II. USDA's nutrient data base for nationwide dietary intake surveys. J. Nutr. 1990; 120:1530-1534
• Pinstrup-Andersen, P., Burger, S., Habicht, J.-P. & Peterson, K. (1993) Protein-energy malnutrition. In: Disease Control Priorities in Developing Countries (Jamison, D. T., Mosley, W. H., Measham, A. R. & Bobadilla, J. L., eds.), pp. 391-420. Oxford University Press, Oxford, England.
• Rush, D. (1986) Evaluation of the Special Supplemental Food Program for Women, Infants, and Children (WIC). Research Triangle Institute, Research Triangle Park, NC.
• Rush D., Leighton J., Sloan N. L., Alvir J. M., Horvitz D. G., Seaver W. B., Garbowski G. C., Johnson S. S., Kulka R. A., Devore J. W., Holt M., Lynch J. T., Virag T. G., Woodside M. B., Shanklin D. S. VI Study of infants and children. Am. J. Clin. Nutr. 1988; 48:484-511 [Abstract/Free Full Text]
• SAS Institute Inc. (1990) SAS/STAT User's Guide, Version 6, 4th ed. SAS Institute, Cary, NC.
• Scrimshaw N. S., Guzman M. A., Gordon J. E. Nutrition and infection field study in Guatemalan villages. Arch. Environ. Health 1967; 14:657-662 [Medline]
• Sempos C. T., Johnson N. E., Smith E. L., Gilligan C. Effects of intraindividual and interindividual variation in repeated dietary records. Am. J. Epidemiol. 1985; 121:120-130 [Abstract/Free Full Text]
• Tippett, K. S., Mickle, S. J., Goldman, J. D., Sykes, K. E., Cook, D. A., Sebastian, R. S., Wilson, J. W. & Smith, J. (1995) Food and nutrient intake by individuals in the United States, 1 day, 1989-91. Nationwide Food Surveys Rep. No. 91-2. Agricultural Research Service, U.S. Department of Agriculture, Riverdale, MD.
• U.S. Department of Agriculture, Food and Consumer Service (1996) Program Information Report (Keydata), U.S. Summary. USDA/FCS, Alexandria, VA.
• U.S. Department of Agriculture, Human Nutrition Information Service (1985) USDA Nutrient Data Base for Standard Reference, Release 9 [data tape, Accession No. PB90-502717]. National Technical Information Service, Springfield, VA.
• U.S. Department of Agriculture, Human Nutrition Information Service (1992) 1991 Continuing Survey of Food Intake by Individuals and 1991 Diet and Health Knowledge Survey [data tape, Accession No. PB94-500063]. National Technical Information Service, Springfield, VA.
• U.S. Department of Health and Human Services and U.S. Department of Agriculture (1989) Nutrition Monitoring in the United States: An Update Report on Nutrition Monitoring. DHHS Publication No. (PHS) 89-1255. U.S. Government Printing Office, Washington, DC.
• U.S. General Accounting Office (1992) Early Intervention: Federal Investments Like WIC Can Produce Savings. GAO/HRD-92-18. U.S. General Accounting Office, Washington, DC.
• Westat, Inc. (1996) A User's Guide to WesVarPc, Version 2.0. Westat, Rockville, MD.
• Yip R., Binkin N. J., Fleshood L., Trowbridge F. L. Declining prevalence of anemia among low-income children in the United States. J. Am. Med. Assoc. 1987; 258:1619-1623 [Abstract/Free Full Text]
• Yip R., Parvanta I., Scanlon K., Borland E. W., Russell C. M., Trowbridge F. L. Pediatric nutrition surveillance systemUnited States, 1980-91. Morb. Mortal. Wkly. Rep. 1992; 41:1-24

0022-3166/98 $3.00 ©1998 American Society for Nutritional Sciences