Calibration Strategies to Correct Nonresponse in a National Travel Survey
Résumé
Nonresponse results in a reduced sample size, but a more important concern of researchers is the possible impact of nonresponse bias. Bias is introduced when those who do not respond to the survey are systematically different from those who do respond on key variables of interest. The response mechanism is defined as "ignorable" when it can be modeled with the available characteristics (e.g., from the sample base). For example, if the nonresponse mechanism depends on the level of income, the respondent sample must be reweighted according to the distribution of income to correct the nonresponse errors. For the French National Travel Survey 2007-2008, the sample was drawn directly from the census and the list of new residences built since then. Therefore, there is considerable information about respondents and nonrespondents. The response mechanism is modeled first, and then calibration strategies are discussed. The nonresponse biases are quantified by using auxiliary information in different calibration exercises.