An Agent-Based Approach to Understanding Survey Responsiveness in Dynamic Political Environments

Survey data is integral to empirical research and political decision-making. While the influence of personal and socioeconomic change on survey participation and the quality of survey data is well studied, comparable effects of political events remain underexplored: How does change in individuals’ political environment affect their willingness to participate in political surveys? And what are consequences for the quality of survey data and the reliability and accuracy of (political) measurements? I employ an agent-based modelling framework to investigate how and which political events affect the willingness to participate in surveys and provide insights into the underlying individual-level processes. Everyone has an individual and unique perception of when the benefits of survey participation exceed its costs. This balance determines one’s cooperativeness to respond in (political) surveys. Whether the perceived benefits outweigh the costs of participation at the time of a survey quest is affected by, e.g., one’s stances towards the survey (topic), or the current availability of resources (e.g., time, cognitive ability). I posit that political events may alter the levels of cooperativeness for selected segments of the population, potentially leading to their under- or overrepresentation in samples. Such sample imbalances, in consequence, likely bias the measurement of political phenomena. To discern which political stimuli (e.g., elections, partisan cues, mass public opinion shifts) may do so and for whom, I simulate individuals’ reactions to such events using agent-based models. Agents' characteristics are informed by panel data from the UK and – in a separate model - Germany, accounting for, e.g., individuals’ past survey participation, political preferences and psychological predispositions. I thereby mirror real-world distributions of survey cooperativeness within the hypothetical population used in the models. By imputing political and survey behavioural reactions to political stimuli, this agent-based approach bridges the data availability gap between respondents and nonrespondents – a limitation by design encountered with real survey data. The simulations reveal how political events and individual predispositions interact to alter survey responsiveness. I evaluate potentially resulting sample disproportionalities and types of measurement bias, pinpointing the (political) variables most vulnerable to bias. Across the tapestry of simulated political events, the results imply that selected groups react to these events by nonresponse: People with low educational attainment, low political interest and confined political efficacy tend to be significantly more underrepresented in post-event samples than their counterparts. While these characteristics are known to negatively correlate with survey responsiveness already in absence of political change, it is a relevant learning that political events can reinforce pre-existing patterns of sample distortion. Thereby, this study provides valuable insights for researchers and political decision-makers to better comprehend the dynamics of survey responsiveness and its implications for the quality of survey data in changing political environments.