Statistical methods for reducing bias in web surveys

Web surveys have become popular recently because of their attractive advantages of data collection. However, in web surveys, bias may occur mainly due to limited coverage and self-selection. This paper reviews characteristics and problems of web surveys, and describes some adjustment weighting methods for reducing the bias. Propensity score adjustment is used for correcting selection bias due to non-probability sampling, and calibration adjustment is used for correction coverage bias. Those bias reduction methods will be explored by comparing face-to-face survey (reference survey) results with web survey results for the Social Survey produced by Statistics Korea. The methods studied include dierent variable selection methods for propensity score calculation and different propensity score weighting methods.