Social scientists increasingly link survey data with administrative records. However, data protection legislation often requires respondents’ informed consent prior to record linkage. This has confronted research with nontrivial refusal rates in combination with selectivity of the consent decision. In longitudinal surveys, linkage requests may also increase attrition rates in subsequent waves, as many respondents perceive record linkages negatively. Hence, in panel surveys, both nonresponse forms potentially bias the resulting datasets. The present paper examines (a) whether we can minimize consent refusals in a longitudinal survey by varying which survey wave includes the linkage request and (b) whether such requests increase subsequent attrition. To test this, we implement a randomized survey experiment within a face-to-face household panel survey focusing on migrants in Germany. Respondents of the IAB-SOEP Migration Sample, conducted by the Institute for Employment Research (IAB) and the Socio-Economic Panel study (SOEP), were randomly allocated into three groups: first, a control group without record linkage; a second group that was asked in wave 1 (2013) to allow data linkage to administrative employment and benefits records held by the Federal Employment Agency; and a third group that was asked the same question in wave 2 (2014). We found neither significant differences in consent rates when introducing the request in wave 1 or 2 nor significant differences in panel attrition in subsequent waves. Therefore, results suggest that requesting consent to link records in wave 1 of a panel study provides the largest quantity of linked data without inflating panel attrition.