Nonresponse bias

In general, nonresponse bias in a sample is assessed by comparing responding and nonresponding firms on the basis of firm characteristics such as sales and number of employees (Armstrong and Overton, 1977). To emulate such comparison, our responses were divided into two groups based on the date received. Then, we ran t-tests to compare the early and late groups for potential differences and found that there is no significant difference between the two groups on popular demographic variables such as annual sales (t = 0.703) and number of employees (t = 0.285), and on our key study variable, IT adoption (t = 0.425) (Armstrong and Overton, 1977). In addition, to compare our respondents with nonrespondents more realistically, we compiled some demographic information of 30 managers who had opted to drop out of the survey without completing it, believing that these managers who dropped out in the middle of the survey are most likely to resemble nonrespondents. Then, their demographic information and the key IT variable were compared with those of respondents who submitted a completed survey. A series of t-tests reveals that there is no significant difference between the respondents and those who opted out of the survey on their annual sales (t = 0.706), number of employees (t = 0.828), number of upstream (t = 0.239) and downstream (t = 0.315) channel members, year of major information systems deployed for supply chain activities (t = 0.770), and the key variable, IT adoption for SCCS (t = 0.219). Therefore, it appears that nonresponse bias is not a problem.

0 0

Post a comment