The National School Lunch Program and School Breakfast Program: Evidence on Participation and Noncompliance Molly W. Dahl Congressional Budget Office United States Congress 2nd & D Streets, SW Washington, DC 20515 molly.dahl@cbo.gov John Karl Scholz Department of Economics, the Institute for Research on Poverty, and NBER University of Wisconsin – Madison 1180 Observatory Drive Madison, WI 53706-1393 jkscholz@facstaff.wisc.edu March 9, 2011 The authors are grateful to Diane Whitmore Schanzenbach for her helpful comments and to Sarah Axeen for her assistance. The views expressed in this paper are those of the authors and should not be interpreted as those of the Congressional Budget Office. 1 The National School Lunch Program (hereafter referred to as the School Lunch Program) and School Breakfast Programs provide funds to participating schools to subsidize the provision of nutritious, low-cost meals to school children. In fiscal year 2004, 29 million children received lunch through the School Lunch Program and nearly 9 million children received breakfast through the School Breakfast Program each school day.1 Well over half of the meals were served at a free or reduced price.2 There has been considerable media and policy attention paid to issues surrounding participation in school feeding programs. A New York Times article summarized the mounting evidence of widespread overcertification in the school feeding programs, highlighting a USDA study that found “the number of students certified for free meals was about 25 percent higher than the number who appeared to be eligible.”3Analysts have further concluded that the level of overcertification is probably much higher than that found in the USDA study, which relied on data from the Current Population Survey.4 Lawmakers enacted the Improper Payments Information Act of 2002, which requires that various federal agencies identify and reduce erroneous payments in their programs; the USDA (which administers the school feeding 1 Source: Food and Nutrition Service, National School Lunch Annual Summary (http://www.fns.usda.gov/pd/slsummar.htm) and School Breakfast Program Annual Summary (http://www.fns.usda.gov/pd/sbsummar.htm) 2 In general, children qualify for free meals if their family income is less than or equal to 130 percent of the federal poverty level or if they are covered by the Supplemental Nutritional Assistance Program (SNAP), Temporary Assistance to Needy Families (TANF), or the Food Distribution Program on Indian Reservations (FDPIR). Children qualify for reduced-price meals if their family income is greater than 130 percent of the federal poverty level and less than or equal to 185 percent of the federal poverty level (USDA, 2001). 3 Robert Pear, “The President’s Budget Proposal: The Poor; Aid to Poor Faces Tighter Scrutiny,” New York Times, February 4, 2003. 4 Neuberger and Greenstein (2003) 2 programs) and the School Lunch Program and School Breakfast Program are included among these (Daniels, 2003).5 This paper uses the best available nationally representative data to calculate the number of children eligible for the school breakfast and lunch programs. We then calculate the fraction of eligible children that take up benefits for each program. We believe these are the first nationally representative School Lunch Program and School Breakfast Program participation rate estimates. Our data also allow us to calculate the number and characteristics of program participants that do not meet income eligibility requirements. We use data from the Survey of Income and Program Participation (SIPP) spanning ten years to analyze trends in participation and eligibility for free and reduced-price meals. During the 1990s there were significant changes in many programs affecting low-income families, including the replacement of the Aid to Families with Dependent Children (AFDC) program with the Temporary Assistance to Needy Families (TANF) block grant and expansions in the Earned Income Tax Credit, which might affect how low-income families use the school meals programs.6 We estimate that nearly 50 percent of children deemed to be eligible for free or reducedprice meals receive a free or reduced-price breakfast and almost 75 percent receive a free or reduced-price lunch, but for reasons discussed below, we think these are likely to be upper bounds for the true participation rate. Our participation rate estimates for the school lunch 5 Schools are reimbursed a certain amount for each meal served. In 2004 the standard lunch reimbursement rates were $2.19 for every free lunch served, $1.79 for every reduced-price lunch served, and $0.21 for every full-price lunch served. The standard reimbursement rates for breakfast were $1.20 for every free breakfast served, $0.90 for every reduced-price breakfast served, and $0.22 for every full-price breakfast served (Federal Register, Vol. 68, No. 130, July 8, 2003). 6 For discussions of the changes in programs affecting low-income families during the 1990s, see Ellwood (2000) or Meyer and Rosenbaum (2000). 3 program are much lower than estimates from the USDA, which we discuss in greater detail below. We find that participation in the school meals programs is correlated (generally) with living in the South, being non-white, and being younger. Older children (children age 15 to 18) are much less likely to participate in either the school lunch or school breakfast program than children age 5 to 11. An upper bound for noncompliance is that as many as 15 percent of the children receiving a free or reduced price meal at school may be ineligible. To further explore what appear to be erroneous payments, we estimate the number of meals served that result in both under- and overpayments by the federal government and then estimate the likely cost to the federal government. Our central estimates are that in the 2002-03 school year, the lunch program resulted in a $31 million underpayment to schools and school districts, while the breakfast program resulted in a $170 million overpayment to schools.7 I. School Meals Programs and the Survey of Income and Program Participation The National School Lunch Act created the modern School Lunch Program in 1946. The School Breakfast Program began as a pilot project in 1966 and was made permanent in 1975. Both programs provide free meals, reduced-price meals, and full-price meals, though full-price meals are still subsidized to some extent. Nearly 100,000 public and non-profit private schools and residential child care institutions participate in the School Lunch Program; nearly 78,000 participate in the School Breakfast Program. The two programs are administered similarly. School districts (and schools) that participate in the meal programs must serve meals that meet nutrition guidelines set at the federal level; they then receive cash subsidies and commodities 7 Our estimates necessarily are based on the assumption that a child who reports usually receiving a meal at school receives that meal every day, so they are likely to be upper bounds. 4 from the USDA for each meal they serve (School Lunch Program Fact Sheet, 2005 and School Breakfast Program Fact Sheet, 2005). Eligibility for free or reduced-price meals is determined in the same way for each program. To apply, individuals (presumably parents or guardians) can self-report household income from the previous calendar month.8 Children in the household are deemed to be eligible for free meals if reported household income is less than or equal to 130 percent of the federal poverty level.9 Children are also deemed to be eligible for free meals if they are covered by the Supplemental Nutrition Assistance Program (SNAP, formerly known as food stamps), Temporary Assistance to Needy Families (TANF, formerly known as AFDC), or receive assistance from the Food Distribution Program on Indian Reservations. Children are deemed to be eligible for reducedprice meals if household income is greater than 130 percent of the federal poverty level but less than or equal to 185 percent of the federal poverty level (USDA, 2001). School districts can also determine the eligibility of children through a process called direct certification. Under direct certification, a list of enrolled children is matched, either by a state or district level agency, to the SNAP, TANF, and Food Distribution Program on Indian Reservations rolls to determine whether children attending school are eligible for free meals. In 2001, most school districts that use direct certification also use passive consent, certifying students as eligible unless they explicitly decline the benefit (Gleason, Tasse, et. al., 2003). Direct certification is quite prevalent; in the 2001-2002 school year, 61 percent of districts that participated in the school meals programs utilized direct certification.10 Of all students certified 8 Household is defined as “a group of related or nonrelated individuals, who are not residents of an institution or boarding house, but who are living as one economic unit” (Reauthorization Regs from PL 108-265). 9 In 2004, the poverty guideline cutoff for a family of four was $18,850. 10 http://www.fns.usda.gov/fsp/rules/Memo/04/100804.htm 5 for free lunch in that year, approximately 18 percent were directly certified (Gleason, Tasse, et. al., 2003). The cash payments for the School Lunch Program in fiscal year 2004 were $6.7 billion; cash payments for the School Breakfast Program were $1.8 billion.11 On average, 14.1 million children received free lunches, 2.8 million received reduced-price lunches, and 12.0 million received full-price lunches every school day that year.12 In the School Breakfast Program, on average 6.5 million children received free breakfast, 0.8 million children received reduced-price breakfast, and 1.6 million children received full-price breakfast every school day.13 The Survey of Income and Program Participation In order to analyze potential trends over time in participation and eligibility for the school meal programs, we use the 1993, 1996, and 2001 panels of the Survey of Income and Program Participation (SIPP), a nationally representative panel survey. Within each panel, interviewers contact survey respondents every four months and collect data on a variety of topics, including demographic characteristics, income, and program participation for each of the previous four months. The SIPP is well-suited for analysis of the school meal programs. It contains information on monthly household income, TANF and SNAP participation, and other details of the household needed to estimate eligibility for free or reduced-price meals. Monthly (as opposed to annual) income data are important. Income volatility among low-income households may lead to 11 Source: Food and Nutrition Service, Federal Costs of School Food Programs. http://www.fns.usda.gov/pd/cncosts.htm 12 Source: Food and Nutrition Service, National School Lunch Annual Summary. http://www.fns.usda.gov/pd/slsummar.htm 13 Source: Food and Nutrition Service, School Breakfast Program Annual Summary. http://www.fns.usda.gov/pd/sbsummar.htm 6 an undercount of eligible households if one relies on an annual measure of income when in practice monthly income is used to determine eligibility. The SIPP also identifies participants in the meals programs. Table 1 compares the number of breakfast and lunch participants in the SIPP to administrative counts of the average number of breakfast and lunch participants on a given school day. The SIPP overcounts the number of participants in the breakfast program, accounting for 160 percent of the administrative count of breakfast participants in the 2002-2003 school year. The number of school lunch participants is also overcounted in the SIPP, though to a lesser degree. In the 2002-2003 school year, the SIPP captures 107 percent of the average number of school lunch participants from administrative data. The overall differences could be due to how the school meals questions are asked. The SIPP asks whether children “usually” get a lunch or breakfast at school. If a child gets lunch at school 3 of 5 days a week, the parent is likely to respond “yes” to this question, and, thus, we will deem the child to be a lunch participant. However, administrative data will only count this child as a participant in the 3 days that the child actually receives a lunch, resulting in the SIPP over-counting the number of school lunches served to this child. This line of reasoning is further supported by the fact that the SIPP matches closely the number of free lunch participants. We think it is likely that a child eligible for free lunches will regularly take advantage of their availability. In contrast, we think consumption of free breakfasts will be less regular since participation often requires arriving at school prior to the start of classes. Consistent with this reasoning, the SIPP over-counts the number of free breakfast participants. After the SIPP redesign in 1996, one can explicitly identify which children in a household participate in the school meal programs. This is not the case in the 1993 panel, which simply identifies the number of children in the household who participate. When the household has 7 more children than the number of school meal participants, we do not know the specific identities of the participating child (or children).14 Evidence from the 1996 SIPP panel suggests that older children participate in the school meal programs at lower rates than younger children. Thus, in the 1993 SIPP panel we assume the youngest school-age children in a household are the free or reduced-price lunch participants.15 Examining this rule in the 2001 SIPP panel, we correctly predict 79 percent of the participants and 73 percent of the non-participants, for an overall correct prediction rate of 76.2 percent, in October 2002. Use of the 1993, 1996, and 2001 panels allows us to see participation and eligibility in the 1993-1994, 1994-1995, 1996-1997, 1997-1998, 1998-1999, 2001-2002, and 2002-2003 school years. Our analysis spans ten calendar years, though the panels do not include information for the 1995-1996, 1999-2000, and 2000-2001 school years. Our data cover a period when lowincome families are adapting to many changes in their social support and the economy. II. Eligibility and Participation in School Meal Programs Our sample consists of all children ages 5 to 17. We also include 18 year old individuals who report that they are enrolled in grades 1 through 12.16 We deem a child to be eligible for free meals if monthly household income is at or below 130 percent of the federal poverty level, or if they received TANF or SNAP benefits. We deem a child to be eligible for reduced-price 14 Of the 3,791 households who report that at least one child in the household received a hot lunch at school in November 1993, 574 reported that not all children received a hot lunch at school (15 percent of the unweighted sample of households). Of the 6,979 children in households in which at least one child received a hot lunch at school, 1,508 (21 percent) of the children are in households in which not all children receive a hot lunch at school. 15 That is, if there are four school-age children in a household and only two participate in free or reduced-price lunches, we assume that the youngest two are the participants. 16 We analyze a sample excluding 5 year old children, as not all 5 year old children attend full-day kindergarten; those that do not are ineligible for free or reduced-price meals. Results were similar to those presented in the paper. 8 meals if monthly household income is above 130 percent but at or below 185 percent of the federal poverty level.17 Our preferred measure of eligibility uses the minimum monthly household income (as a percentage of the federal poverty guidelines) in July, August, and September of a given calendar year and information on TANF or SNAP receipt in those three months.18 This coincides with the practice of applying for free or reduced-price meals at the beginning of the school year and reporting the income or program participation from the previous month on the application. It is straightforward to determine school lunch and breakfast participation in the 1996 and 2001 SIPP panels. Four questions are asked regarding the school meal programs. First, the respondent is asked during the reference period “did your child/children usually get a lunch [breakfast] offered at school?” Second, the respondent is asked to list which children “usually” got a lunch (breakfast) at school. Third, the respondent is asked “Were any of the lunches free or reduced price because your child/children qualified for the Federal School Lunch [Breakfast] Program?” And, finally, the respondent is asked whether the meal was free or reduced price. If the child “usually” received a hot lunch at school but the lunches were not free or reduced-price, we deem the child to be a full-price lunch participant. If the child “usually” received a hot lunch at school, and it was free (based on the respondent’s answer to the fourth question), we assume 17 To determine a monthly measure of the federal poverty level, we divide the annual poverty guideline by 12. This measure varies by family size and state of residence (Alaska and Hawaii have different poverty guidelines than the rest of the United States). 18 If a household reports income in August and September but has missing data on income in July, we use the minimum household income in August or September. If data are missing for 2 of the 3 months, we simply use the nonmissing month of data. We do not treat participation in the SNAP or TANF programs in the same way because of the asymmetry in positive determinations of eligibility. If we only had one month of data and a child was covered by SNAP in that month, we would be able to definitively say that the child was eligible for free meals. However, if the child was not covered by SNAP or TANF in that month we would not be able to definitively determine that the child was ineligible for free meals, as we have no information on participation in the other two months. 9 the child participates in the free lunch program. We determine reduced-price meal receipt similarly. As noted above, the 1993 panel of the SIPP does not identify which children in a household receive a hot lunch. In a household that reports the children qualify for a free or reduced-price lunch and reports that fewer children receive a hot lunch than the number of children in the household, we assume that the youngest school-age children in the household are the free or reduced-price lunch participants. We make analogous assumptions in our analysis of free and reduced-price breakfasts. Participation in Free and Reduced-Price Meals Among Eligible Individuals (Take-up Rate) The percentage of children eligible for free or reduced-price meals that have been receiving those meals has been increasing since the 1993-1994 school year (see Table 2). The take-up rate for free or reduced-price breakfasts increased by 75 percent since the 1993-1994 school year. By the 2002-2003 school year, nearly half of all eligible children participate in either free or reduced-price breakfast. Over 60 percent of those eligible for free breakfast participate.19 Participation in the School Lunch Program among eligible individuals was much higher in the 1993-1994 school year than was participation in the School Breakfast Program. We estimate that in the 1993-1994 school year 64 percent of eligible children received free or reduced-price lunches. By the 2002-2003 school year, 72 percent of eligible children received free or reducedprice lunches. 19 Not all schools participate in the School Breakfast Program, so we are likely to deem some children to be eligible for free or reduced-price breakfast who do not attend schools that offer breakfast. That is, some of the children we deem to be eligible for the program do not have the opportunity to participate in the program if they wanted. 10 Participation in the free lunch program among those eligible increased from 76 percent in the 1998-1999 school year to 85 percent in the 2001-2002 school year.20 This might indicate that participation in the free lunch program is moderately counter-cyclical, increasing (even among those eligible) when the economy weakens. This could be driven by several different factors. Either the expanded pool of eligible individuals is more willing to participate in the program at a given level of well-being or the pool of eligible individuals is, as a whole, worse off, and thus more willing to participate. Comparison of Our Results to those of the USDA The USDA (1999) estimates certifications as a percentage of eligible individuals for free and reduced-price lunches and comes to a sobering conclusion: far more children receive subsidized lunches than are eligible. In Table 3 we compare our results to those of the USDA (1999) in the most recent comparable period, the 1997-1998 school year. To do this, we estimate the number of free or reduced-price lunch participants as well as the number of eligible individuals in the 1997-1998 school year. We present results in which eligibility is based on family characteristics in the previous month (as opposed to the beginning of the school year).21 Analyzing participation in and eligibility for both free and reduced-price lunches using data from the SIPP, we estimate that there are, on average, approximately 15,188,000 participants and approximately 22,619,000 eligible individuals per month.22 We determine eligibility using both family income and program participation, deeming an individual to be eligible if they were covered by SNAP or TANF in the previous calendar month or had income less than 185 percent 20 This coincides with the end of the 1996 panel of the SIPP and the beginning of the 2001 panel. There was no survey redesign between the two panels. 21 22 Results are similar for both measures. Our estimate of the average number of participating individuals in the free or reduced-lunch program in the 19971998 school year is strikingly close – 101 percent – to that reported in administrative data. 11 of the federal poverty line. Simply dividing the two numbers, we find that the number of participants was 67 percent of the number of eligible individuals in the 1997-1998 school year. In the closest time frame to our own, the USDA estimates that the number of certifications as a percentage of eligible individuals in both free and reduced-price lunches was 99 percent. The time frame of their estimate does not exactly coincide with ours, as they use calendar year income from 1998 and certifications as reported to the Food and Nutrition Service in October of 1997. In contrast, we use monthly income and program participation from August 1997 to April 1998 and School Lunch Program participation from September 1997 to May 1998. The discrepancy between their estimates and our own can be explained by two factors: the use of certifications as opposed to participation in the numerator and the use of monthly versus annual income in the denominator. We first discuss the use of certifications as opposed to participation in the numerator. In our measure of participation, we take the average number of children who report that they receive a hot lunch at school and live in a household that reports that the children qualify for either a free or reduced-price lunch. The USDA study uses the number of certifications for participation, or the number of children who have been approved for the free or reduced-price lunch and, thus, could potentially participate on any given day. This number is likely to be higher than the number of actual participants (and is higher than administrative counts of participants) due to direct certification. Recall that under direct certification, the school district contacts the agencies responsible for SNAP, TANF, and the Food Distribution Program on Indian Reservations to determine whether children enrolled in the school are eligible for free lunches. In most districts, passive consent is used; children are deemed eligible unless they 12 explicitly decline the benefit. Many of these children may not have any intent of participating in the program. Regarding differences in the denominator, our preferred estimate of eligibility relies on both monthly income and monthly reports of participation in the SNAP or TANF program. If we instead used the SIPP data to mirror the annual reporting of the Current Population Survey – the data underlying the USDA study – we find that approximately 1.8 million fewer children in 1998 would be deemed eligible for free or reduced-price lunches than we find when using monthly family characteristics. To summarize, we can explain much of the difference between our calculations and the USDA (1999) study. Using, as we believe is most appropriate, participation and eligibility based on monthly income and SNAP or TANF program participation, we find the number of participants is 68 percent of the number of eligible individuals. If instead we use the number of certifications, rather than the number of actual participants, the estimate is 84 percent, which accounts for about 53 percent of the difference between our estimate and that of the USDA (who estimate that certifications are 99 percent of the number of eligible individuals). If we also estimate the number of eligible individuals using annual income in calendar year 1998 (from the SIPP), our estimates of the number of certifications as compared to eligible individuals increases to 91 percent, which, together with the use of certifications, accounts for about 75 percent of the difference between our estimates and those of the USDA.23 The discrepancies between actual participation and certification, and monthly and annual income, account for almost all of the discrepancy between our participation estimates and the 23 The remaining difference between our estimate and that of the USDA arises from differences in the number of children estimated to have family income less than or equal to 185 percent of the poverty line in calendar year 1998. At least a portion of this difference arises from the SIPP’s more accurate measures of income for low-income families when compared to the Current Population Survey. 13 USDA (1999) for free lunches (see column 2 of Table 3). Our preferred estimate of participants as a percentage of eligible individuals is 77 percent. If we use the number of certifications (reported by the USDA), our estimate increases to 100 percent. If we combine certifications with the number of individuals deemed to be eligible based on annual income alone (again, from the SIPP), we find that certifications among eligible individuals is 119 percent, extremely close to the USDA reported estimate of 122 percent. Correlates of Participation in School Meal Programs In Table 4, we examine correlates of participation in the School Lunch Program and the School Breakfast Program. We look at the two meals programs separately, examining participation among all individuals and then breaking participation down into participation by individuals eligible for free or reduced-price meals and participation by individuals not eligible for free or reduced-price meals. Participation by individuals that are ineligible for free or reduced-price meals does not correspond with noncompliance, as these individuals could be paying full price for their breakfast or lunch. Among all individuals, children in the South are significantly more likely to participate in a meal program at school than those in the Northeast (the omitted category). In general, older children are less likely to participate in a meal program than younger children. This is particularly true in the School Breakfast Program, where the coefficient on age is not only significant at a 99 percent level of confidence, but is also quite large, indicating that children age 15 to 18 are 17 percentage points less likely to participate in the breakfast program than children age 5 to 11. We find no difference in the participation of black and non-Hispanic white children (the omitted category) in school lunch participation. However, non-Hispanic white children are significantly less likely to participate in breakfast than any other group. This is true even 14 controlling for eligibility for free or reduced-price meals. Non-Hispanic white children are also significantly less likely than Hispanics or children in the ‘other’ category to participate in the lunch program. Male children are more likely to participate in both breakfast and lunch than females. The number of people in a household appears to be inversely related to participation in the school meals programs. A priori, we expected children from larger households to participate more often, as the value of the meals programs increases for the household if more children are able to participate in it. The number of people in a household is likely to be correlated with other elements of family structure (including households with unmarried female heads) that might also be correlated with school meals participation. Many of the same patterns in participation are seen when examining participation in the school meals programs by children eligible for free or reduced-price meals. Even among children eligible for free or reduced-price meals, older children participate in the programs much less frequently than younger children. The receipt of SNAP assistance is positively correlated with school meal participation among eligible children, while the receipt of TANF assistance is not. Under current law, a child covered by TANF is only eligible for free meals if the TANF program in place in the state is more restrictive than the TANF program that was in place on July 1, 1995. In practice, it appears that participation in TANF alone very rarely makes a child eligible for free meals. Of course, it is likely that children participating in TANF are eligible for free meals based on SNAP participation or income. Among those eligible, family income (adjusted for family size) is not strongly correlated with participation. Children from families with higher income are slightly less likely to 15 participate in the school lunch program than children from families with lower income, though the coefficient is not large and is significant at only the 10 percent level. Remembering that participation by individuals deemed to be ineligible for free or reducedprice lunch is not necessarily indicative of noncompliance (as any child can purchase a full price lunch), the last two columns examine participation in the School Breakfast Program and the School Lunch Program by children ineligible for a reduced price meal. Again, older children are less likely to participate and non-whites are more likely to participate in either breakfast or lunch. Males are more likely to participate. There is no significant association between income and participation among those ineligible for free or reduced-price meals. Across all specifications, children in the South are generally more likely to participate in school breakfast or lunch than children in the Northeast, while children in the West are less likely to participate. Black children are significantly more likely than white children to participate in the breakfast program, though there appears to be no statistical difference in the participation of black and white children in the lunch program. Hispanics and other children are more likely than white children to participate in both breakfast and lunch. And males appear to be more likely to participate than females. The association of SNAP receipt and school meal participation is quite strong, and is significant at the 99 percent level of confidence. III. Noncompliance in School Meal Programs For this discussion we restrict our discussion to individuals who participate in either the free or reduced-price meals programs but appear to be eligible for neither.24 As shown in Table 5, the noncompliance rate, or the percentage of free or reduced-price meal participants who are 24 We do not separately analyze those ineligible for free meals from those ineligible for reduced-price meals, because we do not want to conflate individuals ineligible for either program with, for example, reduced-price meal participants who are “noncompliant” because they too are eligible for free meals. 16 not eligible for either free or reduced-price meals, has been increasing over time in both the School Lunch Program and the School Breakfast Program. In the 1993-1994 school year, approximately 6 percent of free or reduced-price breakfast participants were not eligible for either program. As the program expanded, the percentage of ineligible participants also increased, to about 12 percent during the 2002-2003 school year. In every year examined, the noncompliance rate in the lunch program is larger than that in the breakfast program. Over the decade studied, the noncompliance in the School Lunch Program increased by about 50 percent, from 10.3 percent in the 1993-1994 school year to 15.6 percent in the 2002-2003 school year. Improper Payments In results summarized in Table 6, we look further into the fiscal consequences of noncompliance, analyzing the number of individuals whose participation appears to result in over- or underpayments to their school or school district. There are three combinations of participation and eligibility that result in overpayments (the subsidy paid to the school is too large) to schools: free meal participants who are eligible for reduced-price meals; free meal participants who are eligible for full-price meals; and, reduced-price meal participants who are eligible for full-price meals. The numbers of meals that result in overpayments to schools are in the first three rows for the top panel, addressing the School Lunch Program, and the bottom panel, addressing the School Breakfast Program. There are also three combinations of participation and eligibility that result in underpayments to schools: reduced-price meal participants that are eligible for free meals; fullprice meal participants that are eligible for free meals; and, full-price meal participants that are 17 eligible for reduced-price meals. The numbers of meals that result in underpayments to schools are in the last three rows of each panel in Table 6. In the following analysis we make the assumption that participants in the meal program at a given price would continue to participate in the meal program if they were instead charged the price for which we deem them to be eligible.25 Our estimates overstate the number of children in each category, as we implicitly assume that children who “usually” get a certain meal at school get that meal every day. And, by necessity, we assume that all children are attending participating schools. This is especially likely to affect our estimates of improper payments in the school breakfast category, as fewer schools participate in the School Breakfast Program than do in the School Lunch Program. However, this problem might not be as large as it initially appears, in that a student has to be receiving some hot lunch or breakfast at school in order to appear here. That is, we are not looking at all kids in all schools that do not participate in the School Breakfast Program or the School Lunch Program when eligible, but rather kids in schools that appear to be ineligible for meals at the price they report paying but who still receive a hot meal at school. The majority of free lunch and breakfast participants who appear to be ineligible for free meals appear to be eligible for reduced-price meals.26 If, in the 2002-2003 school year, schools were reimbursed at the standard reimbursement rates in the contiguous states, free meal participants who were instead eligible for reduced-price meals would result in an overpayment to 25 There is evidence to suggest that this may not be true. As discussed above, the take-up rate, or participation among eligible individuals, is much higher for free lunches than for reduced-price lunches. The assumption that all current free-lunch participants would continue to participate if they were charged the reduced price is unlikely to hold. 26 This conclusion is based on the comparison of Row 1 to Row 2 (lunch) and Row 9 to Row 10 (breakfast). 18 the school of $0.40 per lunch and $0.30 per breakfast.27 Free meal participants who were instead eligible for full price meals would result in an overpayment of $1.94 per free lunch and $0.95 per free breakfast. Reduced price meal recipients who were actually eligible for full-price meals would result in an overpayment of $1.54 per lunch and $0.65 per breakfast. Reduced-price meal participants that are eligible for free meals would result in an underpayment of $0.40 per lunch and $0.30 per breakfast. Full-price meal participants who are eligible for free meals would result in an underpayment of $1.94 per lunch and $0.95 per breakfast. Full price meal participants who are eligible for a reduced-price meal would result in an underpayment of $1.54 per lunch and $0.65 per breakfast. We estimate that in the 2002-2003 school year, overpayments in the School Lunch Program totaled $5.4 million and underpayments totaled $5.5 million on average each school day. Overpayments in the School Breakfast Program totaled $1.7 million and underpayments totaled $0.7 million on average per school day.28 Combining all of these estimates into one estimate of the net overpayments made by the federal government (adding all of the overpayments and subtracting the underpayments), we find misclassifications in the School Lunch Program result in the government underpaying schools and school districts by approximately $200,000 each day. If we assume the school year contains 180 school days, this results in underpayments to schools of $31 million each year in the School Lunch Program. A similar exercise based on the School Breakfast Program indicates that the federal government overpays schools and school districts approximately $1 million on an 27 Schools in Alaska and Hawaii and schools that are deemed to be in severe need have higher reimbursement rates than the standard. 28 Note that this estimate is likely to be high, as we base participation on whether one ever participated in the NSLP during the month. In order for this estimate to be correct, all individuals who ever participate in the month would have to participate every day. 19 average school day, resulting in $170 million in overpayments over the course of a school year. Again, this is an upper-bound, as we have assumed that any child that reports receiving a meal at a free or reduced-price in a month participates in the program every day. This is unlikely to be true, even if only for the simple reason that a certain percentage of children are absent from school on any given day. And, as suggested earlier, participation in the school breakfast program is likely to be more variable than the school lunch program; the estimate of overpayments of $170 million in the School Breakfast Program is likely biased upward. IV. Conclusions The school meal programs play an important role in the food safety net. They cost roughly $8.5 billion annually and provide meals to over 17 million children annually. Given the widely held premise that children need good health and nutrition to excel in school, the school feeding programs can be thought of as providing a foundation for the nation’s development of human capital. Given its role, it is striking that there are no papers examining the basic question: are children eligible for free and reduced price school meals actually receiving them? More recently, a different question has been the focus of the political process – as reflected in the Improper Payments Act of 2002 – and in the national media (see the New York Times, 2/5/2003). Namely, is there widespread fraud and abuse in the school feeding programs? Also surprisingly, with the exception of the flawed USDA (1999) study, there has also been little work on this question. Using data drawn from several panels of the SIPP, we show there have been sharp increases in school meal programs between 1993 and 2003. The take-up rate among eligible children for free or reduced-price breakfasts increased by 75 percent since the 1993-1994 school year. By the 2002-2003 school year, nearly half of all eligible children participate in either free 20 or reduced-price breakfast. In the 1993-1994 school year 64 percent of eligible children received free or reduced-price lunches. By the 2002-2003 school year, 72 percent of eligible children received free or reduced-price lunches. These figures make it clear that relatively high percentages of eligible children are indeed benefiting from subsidized school meals, particularly in recent years. But presumably for varied reasons, many do not. The noncompliance rate, or the percentage of free or reduced-price meal participants who are not eligible for either free or reduced-price meals, has been increasing over time in both the School Lunch Program and the School Breakfast Program. In the 1993-1994 school year, approximately 6 percent of free or reduced-price breakfast participants were not eligible for either program. As the program expanded, the percentage of ineligible participants also increased, to about 12 percent during the 2002-2003 school year. The noncompliance rate in the lunch program is larger than that in the breakfast program. Over the decade studied, the noncompliance in the School Lunch Program increased by about 50 percent, from 10.3 percent in the 1993-1994 school year to 15.6 percent in the 2002-2003 school year. While a considerable number of children appear to be receiving free or reduced price meals when their family income exceeds the income eligibility thresholds, a complicated pattern of over- and under-payments arise in the school meals programs. Over-payments occur when children consume more heavily subsidized meals than they are eligible for. Under-payments occur when students eligible for more heavily subsidized meals instead pay a higher price. We estimate that in the 2002-2003 school year, overpayments in the School Lunch Program totaled $5.4 million and underpayments totaled $5.5 million on average each school day. Overpayments in the School Breakfast Program totaled $1.7 million and underpayments totaled $0.7 million on average per school day. 21 Combining these estimates into one estimate of the net overpayments made by the federal government (adding all of the overpayments and subtracting the underpayments), we find misclassifications in the School Lunch Program result in the government underpaying schools and school districts by approximately $200,000 each day. If we assume the school year contains 180 school days, this results in underpayments to schools of $31 million each year in the School Lunch Program. A similar calculation for the School Breakfast Program indicates that the federal government overpays schools and school districts approximately $1 million on an average school day, resulting in $170 million in overpayments over the course of a school, though we emphasize that the assumptions made for this calculation make it an upper bound. Under- and over-payments of this magnitude in an $8.5 billion program strike us as quite small (though reasonable people could, of course, disagree). Given this we think a more promising area for research attention is learning more about the determinants of take-up among eligible children. Policy makers’ views of participation rates of 49 percent for school breakfasts and 72 percent for school lunches might differ depending on whether non-participants were not taking up benefits because they were getting adequate nutrition elsewhere and did not like the school meals program or because there were specific aspects of the program that made it unpalatable to large fractions of eligible children. Another promising area of research, given the large expenditures on these programs, is making progress on high-quality studies documenting the effects of these programs on nutrition, children’s behavior, and school performance. 22 References Daniels, Mitchell E. 2003. Office of Management and Budget. Memorandum for Heads of Executive Departments and Agencies. Subject: Improper Payments Information Act of 2002 (Public Law 107-300). Memorandum Number M-03-13. May, 21. http://www.whitehouse.gov/omb/memoranda/m03-13.html Ellwood, David T. 2000. “The Impact of the Earned Income Tax Credit and Social Policy Reforms on Work, Marriage, and Living Arrangements.” National Tax Journal, 53(4), pp 1063-1105. Food and Nutrition Service, Federal Costs of School Food Programs. http://www.fns.usda.gov/pd/cncosts.htm Food and Nutrition Service, National School Lunch Annual Summary. http://www.fns.usda.gov/pd/slsummar.htm FRAC, Food Research and Action Center. 2004. “National School Lunch Program.” http://www.frac.org/html/federal_food_programs/programs/nslp.html Gleason, Philip, Tania Tasse, Kenneth Jackson, and Patricia Nemeth. 2003. “Direct Certification in the National School Lunch Program – Impacts on Program Access and Integrity.” Economic Research Service, Food and Nutrition Research Program. E-FAN03-009. October. Meyer, Bruce D. and Dan T. Rosenbaum. 2000. “Making Single Mothers Work: Recent Tax and Welfare Policy and its Effects.” National Tax Journal. 53(4), pp. 1027-1061. National School Breakfast Program Fact Sheet. 2004. http://www.fns.usda.gov/cnd/breakfast/AboutBFast/bfastfacts.htm National School Lunch Program Fact Sheet. 2004. http://www.fns.usda.gov/cnd/Lunch/AboutLunch/NSLPFactSheet.htm Neuberger, Zoe and Robert Greenstein. 2003. “New Analysis Shows ‘Overcertification’ for Free or Reduced-Price School Meals Has Been Overstated.” Center on Budget and Policy Priorities. Washington, D.C. July. Title I Report. 2004. http://www.titlei.com U.S. Department of Agriculture. 1999. “Current Population Survey Analysis of NSLP Participation and Income.” Nutrition Assistance Report Series. Food and Nutrition Service. The Office of Analysis, Nutrition, and Evaluation. October. U.S. Department of Agriculture. 2001. “Eligibility Guidance for School Meals Manual.” http://www.fns.usda.gov/cnd/Guidance/eligibility_guidance.pdf Table 1 Average Number of Participants in the National School Lunch and School Breakfast Programs Comparison of SIPP Results to Administrative Data School Year 1993-1994 1994-1995 1996-1997 1997-1998 1998-1999 2001-2002 2002-2003 Participants (in millions) Breakfast Any SIPP 7.15 8.16 9.89 10.41 10.99 13.24 13.62 Administrative Data 5.83 6.32 6.92 7.14 7.37 8.14 8.43 (1.23) (1.29) (1.43) (1.46) (1.49) (1.63) (1.62) Free SIPP 5.98 6.91 7.67 8.08 8.42 9.72 9.96 Administrative Data 4.76 5.10 5.52 5.64 5.72 6.03 6.22 (1.26) (1.35) (1.39) (1.43) (1.47) (1.61) (1.60) Reduced-Price SIPP 0.49 0.55 0.70 0.78 0.98 1.17 1.12 Administrative Data 0.32 0.37 0.45 0.50 0.56 0.70 0.74 (1.54) (1.49) (1.55) (1.56) (1.75) (1.68) (1.51) Full-Price SIPP 0.68 0.71 1.52 1.54 1.59 2.34 2.54 Administrative Data 0.75 0.85 0.95 1.01 1.09 1.41 1.47 (0.90) (0.83) (1.60) (1.53) (1.46) (1.66) (1.73) Lunch Any SIPP 31.30 31.80 28.88 29.68 30.14 29.96 30.52 Administrative Data 25.3 25.7 26.3 26.6 27.0 28.0 28.4 (1.24) (1.24) (1.10) (1.12) (1.12) (1.07) (1.07) Free SIPP 12.32 12.93 12.61 12.61 12.63 13.62 13.45 Administrative Data 12.2 12.4 12.9 13.0 13.0 13.3 13.7 (1.01) (1.04) (0.98) (0.97) (0.97) (1.02) (0.98) Reduced-Price SIPP 2.50 2.49 2.48 2.78 2.93 2.92 2.91 Administrative Data 1.8 1.9 2.1 2.2 2.4 2.6 2.7 (1.39) (1.31) (1.18) (1.27) (1.22) (1.12) (1.08) Full-Price SIPP 16.47 16.38 13.79 14.28 14.57 13.42 14.16 Administrative Data 11.3 11.4 11.3 11.4 11.6 12.0 11.9 (1.46) (1.44) (1.22) (1.25) (1.26) (1.12) (1.19) Notes: Administrative data is a nine-month average over the fiscal year, October - May plus September. SIPP results are a nine-month average, September-May. Administrative data source: Food and Nutrition Service, U.S. Department of Agriculture, National School Lunch Annual Summary (http://www.fns.usda.gov/pd/slsummar.htm) and School Breakfast Program Annual Summary (http://www.fns.usda.gov/pd/sbsummar.htm). The numbers in parentheses are the ratios of the counts of participants in the SIPP to administrative data on the average number of participants. Table 2 Average Percentage of Eligible Children that Participate in Free or Reduced-Price Meals School Year Percent 1993-1994 1994-1995 1996-1997 1997-1998 1998-1999 2001-2002 2002-2003 Breakfast Free 35.2 39.2 43.9 47.9 50.9 60.7 61.6 Reduced-Price 7.4 8.2 9.9 9.8 13.1 16.2 15.3 Free or Reduced-Price 27.9 31.5 35.0 36.8 40.0 48.3 48.7 72.4 73.3 72.2 74.8 76.4 85.1 83.1 37.7 37.0 35.2 35.0 39.2 40.4 39.8 Free or Reduced-Price 63.9 65.2 63.1 64.1 66.3 73.3 Notes: Source: Survey of Income and Program Participation (1993, 1996, and 2001 panels) 71.8 Lunch Free Reduced-Price Table 3 Comparison of USDA Results to SIPP Results: 1997-1998 SY Free and Reduced Reduced Price Free Price Lunches Lunches Lunches Participation Average Monthly Participation SIPP 15,188 12,439 2,749 NSLP Certified (Oct 1997)* Eligibility SIPP Family Char. in Previous Month Annual Income (1998) 19,067 15,965 3,102 22,619 20,850 16,030 13,404 7,184 7,446 13,128 6,062 USDA - CPS Annual Income (1998)* 19,190 Note: Numbers are reported in 1,000s. * Source: USDA, 1999. Table 4 Probit Model of Participation (dF/dx) Among Individuals Eligible for Free or Among All Reduced Price Meals Individuals Breakfast Lunch Breakfast Lunch Region Midwest Among Individuals Not Eligible for Free or Reduced Price Meals Breakfast Lunch 0.76 (0.61) 3.92*** (0.58) 0.43 (0.62) 0.33 (0.28) 1.67*** (0.26) -1.25*** (0.34) 1.36** (0.59) 4.75*** (0.59) 1.25** (0.59) 0.76** (0.32) 1.95*** (0.31) -0.83** (0.39) -1.49 (1.75) 0.95 (1.55) -2.31 (1.97) -0.11 (0.42) 1.29*** (0.41) -1.54*** (0.55) -1.76*** (0.09) -1.38*** (0.05) -1.50*** (0.09) -1.16*** (0.06) -3.06*** (0.28) -1.90*** (0.10) -6.81*** (0.58) -17.56*** (0.79) -0.36 (0.22) -0.81*** (0.35) -6.08*** (0.58) -18.23*** (0.88) -0.64** (0.28) -10.49*** (0.51) -9.00*** (1.56) -14.91*** (1.64) 0.13 (0.33) -5.35*** (0.44) 2.27*** (0.40) 4.26*** (0.42) 4.41*** (0.44) 0.26 (0.26) 1.48*** (0.22) 1.81*** (0.31) 1.47*** (0.43) 3.96*** (0.41) 3.75*** (0.47) -0.01 (0.29) 1.42*** (0.25) 1.82*** (0.33) 4.43*** (0.98) 5.06*** (1.28) 6.39*** (1.12) 0.46 (0.49) 1.37*** (0.41) 1.61*** (0.52) 1.19*** (0.37) 0.45** (0.18) 0.91** (0.37) 0.13 (0.22) 2.09** (0.96) 0.80*** (0.28) Eligibility for Free or Reduce 2.44*** Price Meals (0.53) 0.45** (0.19) TANF receipt 0.28 (0.48) -0.17 (0.33) Food Stamp receipt 1.43*** (0.41) 0.72*** (0.25) Income-to-Needs Ratio -0.37 (0.28) -0.29* (0.17) -0.28 (0.21) -0.06 (0.05) 0.90** (0.39) 1.56*** (0.42) 1.76*** (0.42) 1.59*** (0.43) 0.29 (0.24) 0.88*** (0.26) -0.05 (0.30) 0.50 (0.29) -1.72 (1.17) 2.64** (1.06) 1.18 (1.15) 0.28 (1.38) 0.48 (0.31) 0.88** (0.32) -0.38 (0.39) -0.92 (0.41)** South West Number of People in the Household Age Category 12 to 14 15 to 18 Race/Origin Black Hispanic Other Male School Year 1997-1998 1998-1999 2001-2002 2002-2003 0.47 (0.38) 1.72*** (0.39) 1.58*** (0.41) 1.27*** (0.45) 0.35* (0.19) 0.85*** (0.20) -0.24 (0.24) -0.16 (0.25) Coefficients: dF/dx * 100 Omitted Categories: Region: Northeast; Age Category: 5 to 11, inclusive; Race/Origin: Non-hispanic white; Gender: Female; School Year: 1996-1997 Table 5 Percentage of Participants that are Ineligible for Free or Reduced-Price Meals (Noncompliance Rate) School Year 1993-1994 1994-1995 1996-1997 1997-1998 1998-1999 2001-2002 2002-2003 Meal Type Free Breakfast 12.9 13.9 15.1 18.2 19.8 23.2 24.0 Lunch 15.0 14.4 17.3 20.6 22.3 25.6 25.5 Reduced-Price Breakfast Lunch 63.9 60.3 54.9 60.5 Free or Reduced-Price Breakfast 5.8 8.0 Lunch 10.3 11.0 Notes: Source: SIPP (1993, 1996, and 2001 panels) 62.6 61.9 60.4 60.6 55.3 59.9 67.2 65.5 63.1 62.0 8.2 10.5 9.1 12.3 11.0 14.4 13.7 16.7 12.3 15.6 Table 6 Number of Participants Resulting in Overpayments or Underpayments in the Meals Programs School Year 1993-1994 1994-1995 1996-1997 1997-1998 1998-1999 2001-2002 2002-2003 Meal Program National School Lunch Program Overpayments Free Lunch Participants Eligible for Reduced-Price Lunch 1,457,577 1,417,215 1,590,411 1,939,477 1,720,694 2,157,264 2,135,129 Eligible for Full-Price Lunch 644,686 846,537 910,040 1,031,440 1,331,862 1,725,840 1,564,649 Reduced-Price Lunch Participants Eligible for Full-Price Lunch Total Underpayments Reduced-Price Lunch Participants Eligible for Free Lunch Full-Price Lunch Participants Eligible for Free Lunch Eligible for Reduced-Price Lunch Total National School Breakfast Program Overpayments Free Breakfast Participants Eligible for Reduced-Price Breakfast Eligible for Full-Price Breakfast Reduced-Price Breakfast Participants Eligible for Full-Price Breakfast Total Underpayments Reduced-Price Breakfast Participants Eligible for Free Breakfast Full-Price Breakfast Participants Eligible for Free Breakfast Eligible for Reduced-Price Breakfast Total 887,720 854,147 657,881 842,718 878,463 986,884 952,430 2,989,983 3,117,899 3,158,332 3,813,635 3,931,019 4,869,988 4,652,208 602,294 648,959 879,415 822,239 827,942 878,141 741,771 2,230,789 2,039,416 1,623,889 1,744,287 1,869,034 1,309,874 1,613,003 1,858,030 1,822,606 1,513,085 1,686,221 1,484,537 1,354,669 1,365,659 4,691,113 4,510,981 4,016,389 4,252,747 4,181,513 3,542,684 3,720,433 635,412 233,666 741,467 436,404 830,504 1,054,915 1,048,717 1,402,377 1,592,200 491,774 585,237 784,562 1,113,854 1,008,991 140,435 162,888 182,172 211,519 237,239 350,777 331,816 1,009,513 1,340,759 1,504,450 1,851,671 2,070,518 2,867,008 2,933,007 169,529 139,958 254,739 256,684 292,882 422,595 309,491 90,772 91,788 352,089 71,642 109,812 321,412 314,021 203,668 772,428 299,191 218,218 774,093 308,236 412,831 407,733 227,359 350,786 344,631 828,477 1,186,212 1,061,855