A survey asks respondents: 'How many times have you driven after consuming alcohol in the past year?' The average response is 0.3 times. Researchers suspect the true average is much higher. Which of the following survey design modifications would most likely reduce the bias in responses?
- AUsing a randomized response technique that provides plausible deniability for sensitive answersCorrect
- BConducting the survey by telephone instead of in person
- CChanging the question to 'Do you drink and drive? (Yes/No)'
- DIncreasing the sample size from 500 to 5,000
Why A is right
A randomized response technique introduces a random mechanism (like a coin flip) that determines whether a respondent answers the sensitive question or a harmless one. Because the researcher cannot determine which question any individual answered, respondents have plausible deniability, reducing social desirability bias and encouraging more truthful responses to sensitive questions.
Why the other 3 choices miss
- B
- Incorrect. Telephone surveys may actually increase social desirability bias compared to anonymous written surveys because respondents are speaking to another person. The mode change alone does not address the fundamental bias.
- C
- Incorrect. Converting to a yes/no question loses information (frequency) and may actually increase social desirability bias because the dichotomous format forces an admission of a stigmatized behavior.
- D
- Incorrect. Increasing sample size reduces random error but does not correct systematic bias. If respondents underreport at all sample sizes, a larger sample will still produce a biased estimate.