2017 Methods-of-Payment Survey: Sample Calibration and Variance Estimation
This technical report describes sampling, weighting and variance estimation for the Bank of Canada’s 2017 Methods-of-Payment Survey. Under quota sampling, a raking ratio method is implemented to generate weights with both post-stratification and nonparametric nonresponse weight adjustments. In the end, we estimate variances of weighted means and proportions using bootstrap replicate survey weights. Compared with probability sampling, we find that (i) strong assumptions are required to reduce bias when probabilities of selection are unknown, and (ii) multiple weight adjustments for bias reduction inflate variance. Therefore, it is important to focus more on bias than on variance in the context of nonprobability sampling.