August 18, 2014
Greg Pavela, PhD, postdoctoral trainee at the Nutrition Obesity Research Center (NORC), won second place for his poster titled “Does a Pending Measurement of Weight Improve Self-Reported Weight? Using Survey Data to Conduct a Large-scale Randomized, Controlled Trial” in the Social Sciences Poster Competition at the 91st Annual Meeting of the Alabama Academy of Sciences, held at Auburn University. The objectives of the academy include providing reliable scientific information through the publication of papers and abstracts, as well as the exchange of scientific information.
It’s known that individuals tend to underestimate their weight while overestimating their height, leading to substantial misclassification of weight. Potential reasons for the bias in self-reported height and weight include inaccurate self-knowledge among subjects, a desire to report a more socially desirable measure, or a combination of both. This research combines the benefits of nationally representative survey data with the rigor of an experimental design to test a potential mechanism of improving self-reported weight and the underlying reasons for misreporting.
Data come from two waves of the Health and Retirement Study, a nationally representative study—which is supported by the National Institute on Aging—following more than 26,000 Americans over age 50, who are surveyed every two years. In 2006, half of the study sample was randomly selected to receive an enhanced face-to-face interview, which included direct measurements of height and weight. In 2008, the other half of the sample received the enhanced face-to-face interview. This data collection schedule allowed Dr. Pavela and his colleagues to treat measurement of weight as an experimental treatment.
Study results indicate that participants whose weights were directly measured increased self-reports of weight by about 22 percent, compared with those who did not have direct measurements of weight. A greater effect was observed among females than males. This research demonstrates the feasibility of using survey data to conduct large-scale, randomized, controlled trials and improve causal inferences.