Panel attrition is a process producing data absent from panel records due to survey non-participation or other data unavailability. I examine the nature and causes of attrition resulting from non-contact and survey refusal in the British Household Panel Study. Focusing on non-response transitions amongst Wave 1 respondents using discrete time transition models, I locate attrition at first non-response over the first 14 waves. Physical impediments to contact, less time spent at home and high likelihood of geographic mobility are predictive of subsequent non-contact. Refusals most often result from lack of interest in the survey and general low motivation to participate.