Misreporting of weight and height: does location matter, and why?
We posit that characteristics of local peers influence individual self-reporting behavior because (1) peer characteristics help to determine social norms within a reference group, and (2) such norms establish socially desirable values of body weight that may bias survey responses in their direction.
We test these hypotheses using restricted NHANES data on survey respondents’ county and state of residence matched with the characteristics of residents of each NHANES respondent’s county and state from the ACS. We use the matched data to investigate whether misreporting of weight and height varies systematically across locations and whether such variation follows the sociodemographic characteristics of the county or state population. Such a relationship might arise if individuals define “normal” physiques in relation to the distribution of physical types in a local peer or reference group.
First, we test for fixed effects on weight (height) self-reporting error at the county or state level. Consistent with the presence of county-level weight norms, we observe significant differences in mean self-reporting behavior across counties when controlling for numerous individual characteristics. Differences among states are mostly insignificant.
Second, we test for contextual effects of county or state level demographic characteristics such as racial composition, education composition, and income composition, from the ACS. In these tests we find, for example, that women living in counties (or states) with a higher average household income exhibit more pronounced underreporting of body weight, on average, than women in counties (or states) with a lower average household income, controlling for own household income and other factors.
We also develop a simple theoretical model of self-reporting in which individuals face a trade-off between reporting truthfully and reporting a value that agrees with the social norm in their reference group. Consistent with the model’s predictions, we find in the data that individuals whose true weight falls below a certain threshold value tend to overstate their weight, while individuals above the threshold tend to understate their weight, where the degree of over- or understatement increases with the difference between one’s true weight and the threshold value. Unsurprisingly, the observed threshold value is significantly lower among women than men. The geographic analysis suggests that the thresholds vary significantly across counties and therefore are formed within local social reference groups.
These results suggest that individuals may disregard medical standards and public health messages in setting weight goals and related behaviors and instead focus on local norms. Furthermore, current measures of state-level obesity rates rely on self-reported measures of weight and height taken from the BRFSS which may be biased by social norms that also vary across geographic locations.