Technology has enabled new modes of survey interviews, but their responses may differ due to interviewers. This paper finds that self-complete surveys reduce item non-response compared to face-to-face interviews across topics and question formats.

Data & Methods:

We conducted a field experiment randomly assigning adults in the U.S. to complete surveys either in-person or independently.

Key Findings:

• Item non-response was consistently lower in self-complete mode

• This difference worsened for individuals with lower cognitive skills

• A respondent with average abilities could be up to six times more likely to respond 'don't know' during face-to-face interviews than self-completed surveys, depending on question type.

Why It Matters:

These findings highlight the importance of accounting for interview mode when designing political science research and interpreting survey results.