Do Response Formats Affect "Don’t Know'' Responses? Evidence from Numeric and Likert Scales, with Ceri Fowler, Jessica C. Smith, Anna Sanders, and Hannah Bunting (under review)
Abstract: Why do some survey questions elicit more "don’t know'' responses than others? While existing research emphasises the role of respondent characteristics and question content, we know less about whether response format itself shapes "don't know'' responses. We provide the first direct test of whether commonly used formats -- numeric scales and Likert-type items -- produce different rates of `"don't know'' responses. Empirically, we combine observational evidence from 1,138 questions fielded in the British Election Study Internet Panel (2014-2024) with two original survey experiments. Across observational and experimental evidence, numeric scales produce higher rates of "don't know'' than Likert-type items. Our results suggest that ``don't know'' responses reflect not only what respondents may know or believe about the questions asked, but also how researchers ask these questions in the first place.
Abstract: Surveys are central to studying public opinion, but respondents do not always provide substantive answers to survey questions. One form of non-substantive response option is the "don't know" option. Existing research suggests these "don't know'' responses may be driven by a variety of factors, including respondent characteristics, survey design, or political context, but in practice such answers are often dropped by researchers. We provide the first comprehensive question- and respondent-level analysis of "don't know" responses. Analysing 118,597 respondents, 2,183 questions, and 29 waves of the British Election Study Internet Panel, we find that "don't know" responses are strongly shaped by question attributes and survey context, with a smaller role for respondent characteristics. Non-response is highest for opinion questions and questions concerning national politics, and varies by survey wave. Women and less educated respondents are consistently more likely to respond ``don't know'', although this varies by question type
A Turning Point for Women? A Systematic Review of Voter Bias Against Women Candidates, with Hannah Bunting and Jessica C. Smith (in progress)
Abstract: For decades, the consensus in political behaviour research was that women candidates face electoral bias rooted in gender stereotypes and social role expectations. However, a growing body of evidence suggests that this dynamic may be shifting as recent studies find voters expressing equal or even greater support for women candidates. This paper presents a systematic review of research on gender bias in candidate evaluations published over the past five years, applying PRISMA guidelines to identify and analyse over 200 studies across multiple country contexts. Initial findings reveal that while the field tends to expect a negative bias against women, empirical results increasingly point to a more complex picture. Experimental designs, particularly those using hypothetical candidates, are more likely to find a positive bias towards women, whereas observational studies continue to report neutral or mixed effects. We also find theoretical fragmentation: many scholars continue to rely on traditional stereotyping and role incongruity frameworks, while others turn to intersectionality, partisan heuristics, or system justification approaches to explain shifting patterns. We argue that the field may be at a theoretical turning point—where old assumptions about gender disadvantage persist even as empirical evidence suggests changing voter preferences. The paper calls for renewed theoretical attention to conditional and context-specific gender effects, and for methodological innovation to distinguish genuine bias reversals from experimental effects.
When the Spotlight’s On: Issue Salience and Men’s Conditional Representation of Women’s Interests, with Louise Luxton (in progress)
Abstract: When and why do men speak up on women’s interests? While women politicians consistently champion women’s concerns, men do so selectively. In this paper, we focus on one key mechanism that prompts men’s engagement: when women’s issues become highly salient in public debate. Building on work showing that men represent women when it is electorally advantageous, we argue that issue salience creates strategic opportunity for engagement. Following high-profile events that bring issues disproportionately affecting women into the centre of public and media attention, both the informational costs of participating and the reputational rewards for speaking out are heightened. In these moments, men can signal attention to women’s concerns with maximum visibility and minimal effort. Using data on UK parliamentary debates and print news media coverage we use a semi-automated classification approach to track men and women MPs’ attention to different women’s issues during “ordinary” periods and in the aftermath of salient events. We anticipate that women MPs’ advocacy for women’s issues is consistent, while men’s is conditional on the visibility and attention paid to them on the national news agenda.