Declining landline telephone subscription rates alongside with increasing mobile-only rates require survey researchers to explore alternative means of telephone survey data collection. In several European countries like Finland, Sweden or the Czech Republic mobile phone penetration rates have exceeded 90 percent. Thus, it seems promising to develop mobile phone survey methods. To learn more about changes in telecommunicational behaviour and equipment it is necessary to follow mobile phone owners over an extended period of time using panel survey methods. Panel surveys are prone to panel attrition. Consequently the long term perspective and data quality on mobile phone owners throughout the panel are threatened due to reduced sample sizes and due to potential panel attrition biases. In mobile phone panels reasons for panel attrition can be found in (1) mobile phone numbers turning ineligible during the panel, (2) non-contact with eligible panel numbers and (3) panel members refusing to cooperate in further panel waves. Taken together, these three potential sources of panel attrition amount to about
20 percent wave-to-wave attrition. Consequently the development of reliable methods that counteract refusals and non-contacts in mobile phone panel studies is required.
In this paper we will present results from two experiments addressing panel attrition in the Experimental Mobile Phone Panel conducted in Germany (n=1,700) since 2009. In a randomized between-subjects design a subgroup of respondents to wave 5 of the panel was offered a 5 Euro pre-paid incentive. Attrition rates and potential biases due to panel attrition were compared to the respective values of a control group without incentive. Also, refusals and noncontacts in wave 5 were invited by text message to a short mobile Web survey application where they were supposed to report on a few key variables from the wave 5 questionnaire. The effect of this mode switch was evaluated in terms of the willingness of respondents to participate in the mobile Web survey and also in terms of the quality of their responses.