Nonresponse and Dropout in an App-Based Household Budget Survey: Representativeness, Interventions to Increase Response, and Data Quality

Abstract

Household budget surveys struggle with low response and participation rates, and lower data quality, in part due to a high respondent burden. App-assisted budget surveys may provide solutions to both these problems. This cross-country study carried out in the Netherlands, Luxembourg, and Spain, investigates the use of an app-based diary for collecting household expenditure data compared to a web-based method. We report the results of two randomized experiments: 1) using personalized feedback and 2) interviewer-assisted versus mail recruitment in terms of influence on response and participation rates. The app-based household budget survey yields slightly higher registration, activity, and completion rates compared to the web-based household budget survey that we use as a reference. We find disproportionate representation of certain groups in the app-based sample, but no substantial differences in the overall representativeness between the app-based and web-based samples. Providing households with personalized feedback does not affect registration or activity in the app. Using interviewers for recruitment does increase registration and activity rates, although this negatively affects the representativeness of the sample. Neither providing personalized feedback nor using interviewers for recruitment significantly affects dropout during the study or data quality. We also find no substantive differences between the quality of web-collected expenditure data and data collected in the app. Overall, using an app could be suitable for collecting expenditure data especially in combination with the use of interviewers for recruitment. However, this may come at a cost to representativeness.