PSC Working Paper Series

Document Type

Working Paper

Date of this Version



Smith, Herbert L. 2009. “A Double Sample to Minimize Bias Due to Non-response in a Mail Survey.” Philadelphia, PA: Population Studies Center, University of Pennsylvania. PSC Working Paper Series, No. 09-05. [English version of Smith 2008.]


A large study of nurses conducted in the U.S. states of California (CA) and Pennsylvania (PA) is based on two large samples: n^CA≈100,000 and n^PA≈65,000. The study was conducted by mail and had response rates of: p^CA=.27 and p^PA=.39 ;; the number of respondents is thus, respectively, : n_1^CA≈28,000 and n_1^PA≈25,000. Although there are many respondents, we must concern ourselves with the possibility of substantial bias due to non-response. In order to estimate and correct for this bias, a second random sample (n_01=1,300 in the two states combined) was drawn from among the non-respondents to the first survey. Thanks to financial incentives and, above all, a shorter questionnaire, we obtained a response rate above 90%. In each state, the two samples were combined to create a virtually unbiased double sample.


Biases, California, Double sample, Efficiency, Errors, Mail surveys, Non-response, Nursing, Pennsylvania, Random sample, Sample design, Sample surveys, Sampling, Statistical methods, Statistics, Survey Data, Survey methodology, Surveys



Date Posted: 26 July 2010