March 4, 2003


before the
Senate Agriculture Committee
March 4, 2003

Executive Summary

PDF of this report

View Related Analyses

If you cannot access the files through the links, right-click on the underlined text, click "Save Link As," download to your directory, and open the document in Adobe Acrobat Reader.

I appreciate the invitation to testify here today.  I am Robert Greenstein, executive director of the Center on Budget and Policy Priorities, a nonprofit policy institute that conducts research and analysis on government programs and policies at both federal and state levels.  I have maintained a keen interest in the child nutrition programs, and been engaged in work relating to these programs, for 30 years.  In the late 1970s, I also served as Administrator of USDA’s Food and Nutrition Service.

The Agriculture Committee faces a number of issues in this year’s child nutrition legislation.  My testimony today focuses primarily on one issue, which I would regard as one of the most important issues and also one of the most challenging — the need to reduce the certification of ineligible children for free and reduced-price school meals without causing significant numbers of eligible low-income children to lose these meals.  I will first touch briefly on one of the Committee’s outstanding successes of recent decades in the child nutrition arena — the use of competitive bidding to contain costs in the WIC program — and then turn to the school meals issue.


Competitive Bidding in the WIC Program

In 1988, Congress established a competitive bidding requirement for the purchase of infant formula in the WIC program.  This provision, enacted with the strong support of the Reagan Administration, has been one of the most successful cost containment reforms instituted in any federal health-related program.  Some brief discussion of this reform is in order, since there are indications that elements of the infant formula industry have begun to mount a lobbying effort to weaken or overturn it.

The competitive bidding requirement was enacted after a decade in which infant formula prices mounted much faster than either general inflation or food-price inflation, pushing federal WIC costs up substantially.  To address these soaring costs, several states decided to harness the power of the free market by applying to the purchase of infant formula for WIC a basic principle that both private industry and government agencies routinely use — competitive bidding, under which the bidder who offers a quality product for the lowest price wins the business.  Competitive bidding for infant formula is especially well-suited to WIC, since the infant formula products of all manufacturers are nutritionally equivalent.

The two largest infant formula companies (Ross Laboratories, a subsidiary of Abbot Laboratories, and Mead Johnson, a subsidiary of Bristol-Myers Squibb) vigorously opposed the introduction of competitive bidding in WIC in the 1980s.  In fact, these companies refused to submit bids when the first states sought to institute competitive bidding and almost derailed this reform at the outset.  A third, smaller company began to submit bids, however, and to win state contracts.  This soon resulted in all three companies submitting bids when a state instituted competitive bidding and competing for WIC contracts in these states.

Nevertheless, major roadblocks to this reform remained.  The two major companies worked aggressively to persuade states not to institute competitive bidding.  To head off competitive bidding, they developed an alternative, watered-down form of cost containment.  In states that adopted the weaker alternative approach, there was no competitive bidding, and all companies could sell their products through the WIC program.  Companies that elected to do so would voluntarily agree to provide discounts to state WIC programs.  Studies by the General Accounting Office, the Center on Budget and Policy Priorities, and USDA found that state WIC programs that adopted the alternative, industry-designed system paid far more for infant formula than states using competitive bidding.

Even so, a substantial number of states adopted the alternative system, rather than competitive bidding.  WIC food costs are borne entirely by the federal government; no state funds are involved.  In addition, competitive bidding was a new practice, and establishing it required some additional effort.  This created an opening for industry lobbying efforts in the states, which often met little opposition from other interests.  In some states, the infant formula companies successfully offered grant funds to state health commissioners to use on any health project the commissioner saw fit in return for the commissioner’s agreeing to adopt the industry-favored approach to WIC cost containment, rather than competitive bidding.

Facing this situation, this Committee, its counterpart committee in the House, and the Reagan Administration acted.  The 1988 child nutrition reauthorization legislation required state WIC programs to use competitive bidding for the purchase of infant formula for the WIC program.  The legislation allowed states to use an alternative cost containment system in lieu of competitive bidding if the state could show USDA that the alternative approach would save as much money.  In the 14 years this law has been on the books, no state has used an alternative approach.  The reason is simple: no alternative comes close to matching competitive bidding in containing costs.

Since the competitive bidding requirement was instituted, states have gained extensive experience with the bidding process and with administering the contracts.  The result has been substantial savings.  USDA reports that use of competitive bidding reduces federal WIC costs by approximately $1.5 billion a year.  Without these economies, either the federal government would have to spend $1.5 billion more each year to serve the same number of women, infants, and children in WIC or state WIC programs would have to cut the number of women, infants, and children they serve by more than 25 percent.

It thus is essential that Congress not remove or weaken the competitive bidding requirement.  Unfortunately, infant formula company representatives apparently are suggesting that something like that be done.  The language they use can be somewhat antiseptic: they speak of replacing “sole source” infant formula contracts with “multi-source” contracts.  “Sole-source” is simply a term for competitive bidding; under competitive bidding, the low bidder wins the contract, and the state thus contracts with one company.  “Multi-source” is a euphemism for a system under which a company need not be the low bidder to sell its products through WIC.  And if there isn’t a competitive system in which the low bidder wins, there is not much incentive for companies to submit low bids or provide large discounts in the first place.

There is good reason that when the Pentagon wants to build a new aircraft or USDA wants to buy ground beef for the school lunch program, competitive bidding is used.  Competitive bidding harnesses the competition that is the heart of our free market system.  To weaken the competitive bidding requirement that has been such an outstanding success in the WIC program would be a serious mistake, especially at a time when the federal government faces large budget deficits for many years to come.


Issues Related to Eligibility for Free and Reduced-price School Lunches

In most programs that limit benefits to families or individuals who are below certain income levels, “errors” occur — that is, some people who are ineligible because their incomes exceed the income limits receive the benefits.  The school lunch program is no exception to this phenomenon.  But how to address this issue in the school lunch program presents a particularly difficult set of challenges.

In approaching this matter, three questions stand out:

  1. What do we know about the magnitude and causes of this problem?
  2. What have been the results of efforts tried or tested in the past to reduce the participation of ineligible schoolchildren, and in particular, how have these efforts affected participation by eligible children?
  3. Finally, how can the school lunch program reduce participation by children who are ineligible for free or reduced-price meals without causing large numbers of eligible children to lose these meal benefits?

With the Committee’s indulgence, I’d like to take these questions out of order and address the second question first. 

Effects of Expanded Verification in the School Lunch Program

One possible response to school lunch certification error is to expand the verification of free and reduced-price meal applications.  Right now children are certified based on income reported on meal applications at the start of the school year.  By December 15, school districts are required to seek current income documentation from for a sample of 3 percent of the approved applicants.  If a family does not respond, the children stop receiving free or reduced-price meals.  If the family provides income documentation, a new determination is made about whether the children qualify for free or reduced-price meals.  There is some useful information on the impacts of expanding income verification requirements.  This information comes from three sources.

The one striking finding that emerges from these data is that attempts to use verification in the school lunch program have run into a major problem — large percentages of the families sent notices to provide pay stubs or other documentation of their income to the school have not responded, and the children in these families have had their free or reduced-price meals terminated as a consequence.

What makes these figures particularly alarming is that the available data suggest that a very large share of the children terminated due to non-response are, in fact, eligible.  The degree to which children terminated because of non-response were eligible was examined in the nationally representative demonstration projects conducted in the 1980s and the 1987 study of the verification process.

The percentage of children terminated for non-response who are eligible for free or reduced-price meals may be somewhat lower today, but the percentage likely remains quite high.  This raises grave concern about proposals to expand verification substantially, at least until we can learn how to change the verification process to bring non-response rates down dramatically among eligible families.  Indeed, it was as a result of these disturbing findings that the Reagan Administration rejected options to require wide-scale verification and established the current verification system instead, under which 3 percent of approved meal applications are subject to verification each year.

Accordingly, research needs to be conducted to learn more about why there is so much non-response among eligible families and to test and identify effective ways to lower the non-response rates.  Several factors appear to be at work here.

The studies conducted in the 1980s found that many non-responding parents had no recollection of ever receiving a notice asking them to submit verification of their incomes.  Significant numbers of non-responders also reported that they could not understand the language in the notice or were not fluent in English.

A key factor here appears to be that unlike in most other means-tested programs, the school lunch verification process involves no personal contact between parents and school officials.  In most other means-tested programs, a parent sits across a table or a desk from a caseworker who explains what the parent needs to do and what type of documentation the parent needs to provide and answers questions that the parent may have.  Not only does none of this occur in the school lunch verification process, but there generally is no follow-up phone call when a parent who has been sent a verification notice does not respond.  (Another possible factor is that for some families, there may be a stigma factor or fears about confidentiality in submitting pay stubs to their children’s school.)

Potential Effects of Greatly Expanded Verification Requirements

This year, for the first time since the 1980s, policymakers are again considering how to reduce the participation of ineligible children in the school lunch program.  This is an area that needs attention but that also is fraught with risks of substantial adverse effects on children.  Let me share a few more figures with you.

These disturbing figures indicate that there is a great need for new, nationally representative demonstration projects to test a variety of approaches to learn both how to reduce participation by ineligible children and how to reduce non-response rates and avoid harming needy, eligible children.  My first recommendation in this area for this year’s reauthorization legislation is to mount such demonstrations, without delay.

I would note that if substantially expanded verification procedures are instituted in a way that results in significant number of eligible children losing free and reduced price school lunches, there will be adverse consequences for other child nutrition programs as well.  For example, schools in which more than 40 percent of the lunches served in a prior year were provided to children certified for free or reduced-price meals receive significantly higher school breakfast reimbursement rates.  This feature of the school breakfast program has played an important role in persuading most schools with substantial numbers of low-income children to operate the program.  Schools that are close to the 40 percent threshold could lose their higher reimbursement rates as a result of the significant decline in the certification of eligible children that would likely result from heavy-handed changes in the school lunch certification process.  That could have adverse effects on the continued participation of some schools in the breakfast program.

Similarly, the overwhelming bulk of summer food program sites are located in neighborhoods in which at least 50 percent of schoolchildren are certified to receive free or reduced-price lunches; it generally is not feasible to operate the summer food program in other areas.  This is because summer food programs must be located in areas that meet the 50-percent test to be allowed to serve meals free to all children.  Summer feeding sites in other areas must secure free and reduced-price meal applications from the parents of children they serve, determine that at least 50 percent of enrolled children are eligible for free or reduced-price meals, and keep track of which children are considered “enrolled” (with any non-enrolled children who come to the site being denied meals).  History shows that when summer feeding sites cannot serve free meals to all children who come to the site, they drop out.  It simply is not feasible for the community-based organizations, local parks and recreations departments, and other agencies that operate many of the summer feeding programs across the country to perform the administrative tasks that otherwise would be required, especially since many summer feeding sites consist of outdoor sites in parks or similar venues where identifying children, sending home and collecting back meal applications, and limiting the children served to those “enrolled” is not practical.  Changes in the school lunch program that result in losses of free and reduced-price school meals for significant numbers of eligible children consequently are likely to lead to sizeable cutbacks in the availability of the summer food program.

Likewise, family day care homes operated by providers whose income is over 185 percent of the poverty line can receive much lower meal reimbursements if they are not located in an area in which at least 50 percent of school children are certified to receive free or reduced-price school meals.  If the decline in certification of eligible children for free or reduced price school meals caused an area to drop below the 50-percent threshold, family day care providers would have to document the eligibility of specific children for free or reduced-price meals in order to continue receiving higher reimbursements.  This is often hard to do for individual day care home providers who usually run very small operations.  It is likely that some providers would drop out of the program rather than trying to administer such a means test or to comply with the program’s meal requirements while receiving much lower meal reimbursements.  In addition, community-based after-school programs are reimbursed for snacks they serve only if they are located in an area in which at least 50 percent of the schoolchildren are certified to receive free or reduced-price school meals.  Such after-school programs would not be reimbursed if the area dropped below the 50-percent threshold as a result of a substantial decline in the number of eligible children certified for free or reduced-price school meals.

What Do We Know About the Extent to Which Ineligible Children
Are Being Improperly Certified for Free or Reduced-price Meals?

In programs such as food stamps, welfare, and Medicaid, error rates are or have been determined by selecting a sample of participating families and conducting an audit of their circumstances.  There are no similar data on error rates in the school lunch program.  There are simply no reliable data — in fact, there are not even reliable estimates — of the proportion of children who are incorrectly certified for free or reduced-price meals either because their families underreport income (or incorrectly report other household circumstances) on a school meals application or because the school certifies them for meal benefits when, based on the information in the application, it should not have done so.

Facing these data limitations, FNS has attempted to use comparisons to Census data to examine this matter.  One such FNS comparison has now made its way into the media.  FNS compared the number of children certified for free school meals for the 1998-1999 school year by October 31, 1998 to the number of school children that an annual Census survey indicated had annual incomes below 130 percent of the poverty line (the free meal income limit) for calendar year 1999.  FNS found there were 15,876,000 free meal certifications as of October 31, 1998, compared to 12,464,000 schoolchildren with annual incomes below 130 percent of the poverty line in calendar year 1999.  The first number — the number of free meal certifications — is 27 percent larger than the second number (the number of children with annual incomes below 130 percent of the poverty line in 1999).  This has led some who have heard these figures, including some journalists, to conclude that FNS has found a 27 percent error rate.

Yet such a conclusion is mistaken.  These data are not evidence of a 27 percent error rate.  There are two types of problems with the assumption that these figures indicate a 27 percent error rate.  I will cover the smaller set of problems first and then the more fundamental set of problems.

First, even if this comparison could be viewed as a proper measurement of the error rate, it would show a 21 percent error rate for free meals, not a 27 percent error rate.  If there are 27 ineligible children out of every 127 receiving free meals, the ineligibility rate would be 27/127, or 21 percent.

Second, the aforementioned data apply to free meals only.  When FNS conducted the same comparison for free and reduced-price certifications combined (that is, when FNS compared the number of free or reduced-price meal certifications as of October 31, 1998 to the number of children the Census survey estimated to have annual incomes below 185 percent of the poverty line for calendar year 1999), FNS found these numbers nearly matched.  The number of free and reduced-price meal certifications exceeded the number of children in the Census estimates by only two percent.


The More Basic Problem

The more fundamental problem with this comparison is that the Census data in question substantially understate the number of children eligible for free meals.  School lunch eligibility is based on monthly income, not annual income.  Free meal approvals as of October 31 of a school year include applications that are based on household incomes in August, applications based on household incomes in September, and applications based on October incomes that were submitted a month or two into the school year by families whose incomes have just fallen.  Census data on monthly income show that the number of children with incomes below the free meal income limit in any one of these months significantly exceeds the number with annual income below the free-meal income limits.

Furthermore, the FNS comparison involves comparing the number of meal certifications in the fall of 1998 to household incomes for calendar year 1999.  Yet half of calendar year 1999 came after the 1998-1999 school year was over.  Other Census data show that between 1998 and 1999, poverty fell significantly, as unemployment declined and wages for low-income workers rose.  Since significantly fewer families were poor in 1999 than in 1998, use of income data for 1999 to assess the accuracy of meal certifications conducted in 1998 results in further distortion.

A similar problem arose a few years ago in the WIC program.  FNS had been using the same set of Census data on annual incomes to estimate the number of people eligible for WIC.  Based on these data, it appeared that more than 100 percent of the eligible infants were participating.  Recognizing that there were questions about the validity of using these Census data to estimate the number of people eligible for WIC (and about other aspects of its WIC eligibility estimates), FNS commissioned the National Research Council to convene an expert panel to investigate this matter.  In a report issued in 2001, the National Research Council found that the use of Census data on annual incomes to estimate the number of people eligible for WIC resulted in a sizeable underestimate of the number of eligible people.  The National Research Council reported that “the current method of using annual income to estimate eligibility results in an underestimate of the number of infants and children eligible for WIC.”[1] 

The National Research Council developed an alternative, more accurate estimate of the WIC eligible population, using a different Census Bureau survey that tracks monthly income.  The National Research Council found that when the more accurate estimate was used, the number of infants participating in WIC no longer exceeded the number estimated to be eligible.

The same problems that applied to using Census data on annual, rather than monthly, income to estimate the number of infants and children eligible for WIC apply to using the Census data on annual income to estimate the number of children eligible for free school meals.  Both WIC and the school lunch program use essentially the same set of rules regarding how families are supposed to report their incomes when applying for benefits.  And both programs tell families to apply based on their monthly incomes, not their annual incomes.

In a recent meeting that Zoe Neuberger (now on our staff, formerly OMB’s budget examiner for child nutrition) and I had with FNS analysts, there was agreement that use of Census data on monthly incomes (which comes from the Census Bureau’s Survey of Income and Program Participation) would provide a better measure of the number of children eligible for free or reduced-price meals than the annual income data that FNS has been using (which comes from the Census Bureau’s Current Population Survey).  FNS analysts thought that use of two months of Census data on household incomes — data for August and September, the months when most school lunch applications are filled out — would be a good measure. 

Unfortunately, the Census Bureau’s Survey of Income and Program Participation, which contains the monthly income data, is more difficult and expensive to use than the Census Bureau’s Current Population Survey, which reports income over a calendar year.  Until a few weeks ago, no analysis of the number of children eligible for free meals based on their family’s income in August or September was available.  In the past week, that has changed.

John Karl Scholz, Director of the Institute for Research on Poverty at the University of Wisconsin (a noted poverty institute that was recently awarded a major poverty research grant by HHS) has been able to use his Institute’s Census models to examine this question.  His findings are illuminating.

Based on these new data, it now appears that the comparison showing 27 percent more children certified for free meals than were eligible for such meals is fatally flawed, due to its comparison of  free meal certifications made in the fall of 1998 to annual incomes in calendar year 1999.  As just noted, when the number of children eligible for free meals in the months in which the free meal certifications were conducted is compared to the number of children certified, 2.8 million more children are found to have been eligible — and as a result, most of the 27 percent “overage” disappears.

Most of the 27 percent “overage” thus appears to be an artifact of comparing meal certifications in the fall of 1998 to Census data on annual incomes in 1999.  Yet that should not be taken to mean that there is a little or no error in the school lunch program.  Suppose the number of children certified for free meals exactly matched the number of children estimated to be eligible for those meals.  This exact match could still mask a significant error rate.  Suppose only 90 percent of the children eligible for free meals had signed up for them.  If the number of children certified for free meals and the number of children eligible for free meals were identical but only 90 percent of those who were eligible had actually enrolled, the other 10 percent of the enrollees would consist of ineligible children.  (There are not reliable data, however, on the percentage of eligible children that are certified.)

At present, there are no good data on the proportion of free and reduced-price meal approvals that are erroneous.  Based on the data that are available, the problem appears to be significant and warrants attention, but is probably significantly below the 27 percent level.  USDA Undersecretary Eric Bost recently expressed a similar view, stating that he believed the 27 percent figure was too high but that the problem was significant.

The Causes of the Problem

Adding to the complexity of this issue is the fact that there are two very different kinds of “errors.”

There are virtually no data available that distinguish these two types of errors.  From a policy perspective, however, these two types of cases are distinct and should be treated very differently.

Increasingly, major means-tested benefit programs are moving to make children eligible for benefits for 12-month periods, rather than trying to track month-to-month fluctuations in the incomes of low-income families.  The incomes of low-income working families can change significantly from month to month.  For example, many such families do not have paid sick leave, so illness can lower their income and recovery can raise it.  For other low-income working parents, the number of hours of work in a month can vary depending on employer needs, child care availability, and other factors.

In Medicaid and SCHIP (the State Children’s Health Insurance Program), many states now certify children for 12-month periods, based on their monthly income at the start of the period.  Last year’s Farm Bill moved the Food Stamp Program in a similar direction, allowing states to fix benefits for households for six months at a time.

Medicaid, SCHIP, and the food stamps program — which employ tens of thousands of caseworkers and incur administrative costs of some billions of dollars a year — have taken these steps because they have found they simply are not able to keep up with households’ monthly income fluctuations.  If these programs, with their much larger infrastructures and bureaucracies, cannot track such income fluctuations, it should come as no surprise that the school lunch program — which has no comparable bureaucracy — cannot do so either.

On paper, school lunch regulations call for changes in households’ monthly income or household composition to be reported to schools and acted upon.  In the real world, as USDA acknowledges, this rule has never been enforced or implemented, because schools cannot administer it.  Part of the school lunch error rate consequently consists of cases where a child was properly certified for free meals, but the child’s parent increased his or her earnings later in the year to a level exceeding the free meal income limit or another adult moved in or out of the household and changed the relationship of the household’s income to the poverty line, which is based on family size.  (In many cases, the income of such families rises from the free meal income range to the reduced-price income range.)  If the child’s eligibility is checked later in the year as part of a study or as part of the verification process, the child appears ineligible for free meals.  Yet the child was correctly certified at the year’s start, and there is no administratively feasible system for schools to shift children back and forth between meal categories each month as family income fluctuates.

I believe that policymakers should separate these cases from the others.  The policy goal should be to make meal certifications as accurate as possible at the start of the school year without causing large numbers of eligible children to lose benefits.  Once children are properly certified, their eligibility should last for the full school year.

Providing eligibility for the school year is how the program has always operated.  It does not make sense to maintain in the Code of Federal Regulations an unrealistic rule that cannot be implemented, and then to tar the program as having more “errors” because the rule departs from reality.

What to Do?

In the absence of better information on many aspects of this problem — and especially on how to reduce participation by ineligible children without deterring eligible children — it is difficult to determine exactly what to do.  We badly need demonstration projects to identify solutions.  Research and demonstration projects need to be a major part of the approach to the problem.  Nevertheless, there are some steps that can be identified and taken now.

The final principle I’d suggest underlies all others.  Care should be taken that in seeking to reduce ineligible certifications, we do not violate the Hippocratic Oath of “Do No Harm.”  The nation’s children are among its most valuable resources.  It would be tragic if efforts to reduce erroneous certifications resulted in large numbers of needy children losing benefits, with adverse consequences for their nutrition and health and their educational attainment.

End Notes:

[1] National Research Council, Estimating Eligibility and Participation for the WIC Program, 2001.