Question

Fact Pattern: An internal auditing team has been assigned to review "the customer satisfaction measurement system" that the Industrial Products Division implemented 2 years ago. This system consists of an annual mail survey conducted by the division's customer service office. A survey is sent to 100 purchasing departments randomly selected from all customers who made purchases in the prior 12 months. The survey is three pages long, and its 30 questions use a mixture of response modes (e.g., some questions are open-ended, some are multiple-choice, and others use a response scale). The customer service office mails the survey in September and tabulates the results for questionnaires returned by October 15. Only one mailing is sent. If the customer does not return the questionnaire, no follow-up is conducted. When the survey was last conducted, 45 of the questionnaires were not returned.

Nonresponse bias is often a concern in conducting mail surveys. The main reason that nonresponse bias can cause difficulties in a sample such as the one taken by the customer service office is that

A. The sample means and standard errors are harder to compute.
B. Those who did not respond may be systematically different from those who did.
C. The questionnaire is too short.
D. Confidence intervals are narrower.

Answer 1

The primary concern with nonresponse bias in survey samples is that the non-respondents may differ significantly from respondents, affecting the survey's validity and reliability. This bias is not related to the calculation of sample means and errors or the width of confidence intervals.

Step-by-step explanation:

The primary concern with nonresponse bias in survey sample data, such as that collected by the Industrial Products Division, is that those who did not respond to the mail survey may systematically differ from those who did. Nonresponse bias can introduce significant distortion into survey results because it potentially excludes a portion of the population that may have different opinions or behaviors than the respondents. This issue goes beyond the practical difficulty of computing sample means and standard error and affects the validity and reliability of the survey findings. It is not that the questionnaire is too short, nor is it related to the width of the confidence intervals, which would be affected by sample size and variance, not necessarily by response rate.

A similar concern exists in other forms of data collection, like phone surveys, where factors such as not everyone having a telephone or being available to answer can contribute to nonresponse bias. This can skew the survey sample, making it non-representative of the entire population. Response bias and survey design, including question wording and interviewer bias, can also impact the accuracy of collected data.