We conducted an analysis of non-response bias based on the procedure described in
Armstrong and Overton (1977) and Lambert and Harrington (1990). We numbered the
responses sequentially, in the order they were received, and compared late responses
with early responses. No noticeable pattern among the variables could be detected to
indicate the existence of a non-response bias. Accordingly, non-response bias is
unlikely to be an issue in interpreting the results of this study.