Our Research Methodology

Polling partners

Each of our national studies has included a large-scale survey of a representative sample of the population. We conduct most of this research in-house, mixing state-of-the-art methodologies in qualitative, quantitative opinion research conducted in person, on the telephone and online to ensure that every corner of society is reflected in our research.

Cluster Analysis

Traditionally, opinion research analyzes differences through the lens of political party identity and demographics. More in Common has sought to broaden this approach by understanding how attitudes are shaped by different worldviews and value systems. To understand these groupings, we have created segments by using statistical techniques known as ‘cluster analyses’ to identify patterns of similar attitudes and beliefs. In each of our national segmentation studies, we have collaborated with professional statisticians to use methods such as k-means, agglomerative hierarchical clustering, and principal components analysis to identify these groupings.

For instance, we have often identified a liberal and cosmopolitan urban group that tends to be more educated, more secular, more deeply engaged in social and political issues, and which believes strongly in human rights, the benefits of immigration and environmental causes. By conducting a segmentation analysis, we can join the dots and develop a more complete picture of the connections between people’s beliefs. In today’s era of deepening tribalism, when social media and other forces are grouping people into like-minded groups, this approach provides deeper and more valuable insights than traditional approaches that compare people across simple demographic groups.

Qualitative work

Much of our learning comes through hearing people’s perspectives in their own words. In every major study we have conducted, we have deployed one-on-one interviews, focus groups or both. In these settings, we see and hear all the elements of human communication: conviction, doubt, indifference, arguments, connection and confusion. The insights from our qualitative research are interwoven with our quantitative insights to reflect the complexity of the people and societies we are trying to better understand.

Integration with academia

To ensure that our studies contain valuable and original insights, More in Common collaborates with academics and experts in related domains such as political science, sociology and psychology. These external partnerships allow us to draw on key discoveries from academia to enhance our understanding and build a more complete analysis.

For example, these collaborations have resulted in the inclusion of cutting-edge methodologies such as the Implicit Association Test (a methodology developed at Harvard University to measure bias) and Moral Foundations Theory (developed by psychologist Dr. Jonathan Haidt) to gauge ideological orientation). Today, our staff includes permanent academic advisors who help enhance our research methodologies and keep our thinking up to date with new and emerging insights.

Data Quality and AI Threat Detection

Public opinion research conducted online is generally executed via panel providers who distribute survey invitations to anonymous, paid survey respondents. The development of advanced, agentic AI models that can simulate human responses and avoid conventional survey detection methods presents a novel threat to online survey data quality that risks undermining trust in public opinion data, including from More in Common. We take this risk seriously and are pursuing the six-step plan below in response.

1. Raised expectations and transparency from our panel providers.

Implementation: we will continue to request regular information from panel providers about their AI-detection protocols, including transparency about methods used and ongoing detection rates. We assess this information dynamically and comparatively so that we can pivot to the partners and data providers in which we have the greatest confidence.

2. Evaluation of new partners and panel providers.

Implementation: we are actively piloting research projects with panel providers that specialize in AI detection and respondent validation.

3. Greater emphasis on insight validation through qualitative methods.

Implementation: we are increasingly designing research projects with qualitative methods integrated into the design expressly for the purpose of quantitative insight validation. While qualitative methods are not a substitute for the confidence that comes from rigorous quantitative information, they provide a safeguard against reporting discoveries based on limited observation.

4. Monitoring the academic literature and the industry on AI detection.

Implementation: across the seven countries in which we operate, our research team monitors academic journals that publish reviews, experiments, and systematic evaluations of how AI is interacting with online research environments and how academics and research practitioners are best responding. We also regularly participate in leading polling and survey research convenings to stay current on emerging best practices.

5. Partnerships with experts and advisers.

Implementation: we maintain active dialogue with academics, experts, and specialists on these challenges. For instance, in the United States we have invited a leading political scientist with published expertise in monitoring AI in quantitative research environments to join our board.

6. Continued implementation of best-in-class fieldwork practices.

Implementation: we continue to deploy best practices and actively develop new methods for ensuring data quality, both as an additional layer of quality assurance against AI as well as to mitigate data quality issues generated by low-attention respondents.