Abstract

Panel survey data play in important role in labour market research. However, for the last decades survey administrators faced the challenge of declining response rates and increasing survey costs. To counter this trend, many panels integrated cheaper online surveys for data collection into mixed-mode designs. However, changing mode in an ongoing panel might influence drop-out rates or selection effects, something known as mode effects. While previous research found considerable improvements in terms of response rates and survey costs in a single wave, little is known about possible improvements regarding panel attrition and costs in multi-wave panel surveys. We address this research gap using administrative data linked to the LPP employee sample. We propose multilevel logistic regression to estimate panel attrition, which we relate to costs-per-respondent to gain further insights about long-term effects on costs.