Sample Selection Bias in Consumer Payment Surveys

• Activity Number: 119
• Type: Topic Contributed
• Date/Time: Monday, August 4, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Survey Research Methods Section
• Author(s): Marcin Hitczenko*+
• Keywords: sample bias ; consumer surveys ; Survey of Consumer Payment Choice

Abstract

Analysis of survey data, no matter the sampling design, is naturally limited to observations from respondents who agreed to participate and appropriately completed the survey. Sample selection bias occurs when there is some inherent trait that is linked to both the variable of interest and the likelihood of completing the survey. In some cases, most notably involving demographic information, probability weighting can be used to adjust for the missing responses. However, other less well-defined traits are more difficult to deal with. In the context of consumer payments, it often makes sense to attribute payments to entire households, with members of the household carrying different degrees of responsibility for and knowledge about those payments. As a result, non-representative sampling from within households, such as a tendency to interview those who are more responsible for payments, could lead to biased results if unaccounted for. We use data from the Survey of Consumer Payment Choice to study the degree to which financially responsible household members are over-represented and estimate the potential effects on estimates of monthly payment use.