High frequency body mass measurement, feedback, and health behaviors.

We analyze weight and fat percentage measurements of respondents in an online general population panel in the Netherlands, collected using wireless scales, with an average frequency of 1.6 measurements per week. First, we document the existence of a weekly cycle; body mass is lowest on Fridays and highest on Mondays, showing significant (p<0.01) differences of, on average, 0.2 kilogram in weight, 0.06 in BMI value, and 0.03 in fat percentage. Second, we find that in the general population fat-based measures of obesity point at a three times larger prevalence of obesity (53%) than BMI-based measures (17%). Third, we find that feedback that includes a recommended weight range increases the temporal variation in individual body mass by almost ten percent (sd for weight increases from 1.13 to 1.22; sd for BMI increases from 0.37 to 0.41; sd for fat percentage increases from 0.55 to 0.61.