Panel Attrition, Data Imputation and Age Patterns in Subjective Ill-being -- by David G. Blanchflower, Alex Bryson, Alan J. Cui

We examine the implications of panel attrition for the age profile of subjective ill-being in the Global Flourishing Study (GFS). We establish that survey attrition rates are high, and are highest among the young, particularly when interviewed in wave 1 using a computer-assisted web interview (CAWI). We then deploy three methods (inverse probability weighting, multiple imputation chain equations, and last observation carried forward) to estimate the age pattern in subjective ill-being in wave 2 of the survey. All three methods have their strengths and weaknesses, but we demonstrate that such adjustments are merited given the nature of survey attrition in the data. We conclude that subjective ill-being is declining in age in both waves 1 and 2 of the GFS having accounted for sample attrition, something that would not have been apparent if one simply ran wave 2 regressions without accounting for sample attrition.