WASHINGTON -- The latest survey of what Americans eat was so badly bungled that federal agencies may not have the data they need to regulate everything from school lunches and food stamps to food labels and pesticide exposures, a new congressional report concludes.
The 1987-1988 Nationwide Food Consumption study, conducted by the Department of Agriculture, did not interview enough people, its design was flawed and it lacked adequate quality controls, according to a report by the General Accounting Office.
The survey, for which a private contractor was paid $6.2 million, could present a problem for federal agencies, the food industry, nutritionists and scientists. All rely on the survey, conducted every 10 years, for up-to-date numbers on who is eating what.
For instance, the Environmental Protection Agency uses the survey to find out how much of a particular fruit or vegetable Americans eat when setting rules on safe levels of pesticide residues. The Department of Agriculture uses data on consumption and income to decide how much and which foods to provide through the food stamp and school lunch programs.
And the Food and Drug Administration uses information on how much Americans eat to help them determine the "average serving size" on a food label.
"I'm very disappointed to see all these questions and doubts raised about the survey because these are extremely important policy decisions involving billions of dollars relying on supposedly scientific guidance," said Representative George E. Brown Jr., D-Calif., who asked the GAO to conduct the study.
The GAO, the non-partisan investigative arm of Congress, criticized both the contractor, National Analysts of Philadelphia, and the Agriculture Department for mismanagement of the survey.
National Analysts scientifically selected a sample group of 6,000 households of all incomes and 3,600 low-income households. Surveyors were supposed to return with answers to overall household consumption, as well as on individual intake.
But information on the individuals came in from only 34 percent of the households -- "a response rate so low that it is questionable whether the data are representative of the U.S. population," the report said.
The GAO blamed part of the problem on the survey's long, complicated questionnaires. Researchers found it took up to three hours just to answer questions about food the household ate. Despite the time demanded, participants received only $2.
The report also lambasted National Analysts' quality control, saying that some interviewers weren't trained well, that too many interviewers quit and that the companydid not adequately distribute interviews through the year to minimize seasonal differences in food consumption.
A National Analysts spokesman defended the study, saying the answers from the 34 percent of households that responded were accurate.
But he said the GAO and the Agriculture Department "could be perfectly right" in requesting more interviews to ensure that the survey represents all demographic groups.
A group of scientists, asked by the Agriculture Department to review the study, recommended that the agency not use the 1987-1988 data -- or at least that it publish the numbers with a warning about their limits.
A Department of Agriculture spokesman, Roger Runningen, said that the department would issue such warnings with the data, but also that it would take other steps to review the survey.
"It's flawed. I don't know that it is so damaging that it is totally useless, but we are taking a number of steps to fix this flat tire," Mr. Runningen said.