National surveys often struggle to capture nuanced public opinion at the constituency level. This paper evaluates two methods—external validation using 2010 election data and cross-validation of EU opinions—to improve estimates from national samples.
Data & Methods:
The authors analyze party vote share in Britain's 2010 general election alongside surveys on attitudes toward the European Union. They employ post-stratification with individual-level predictors and geographic local smoothing to refine constituency-based opinion estimates.
Key Findings: Constituency accuracy gains primarily stem from incorporating basic, accessible predictors rather than complex factors.
- Adding simple constituency-level variables significantly boosts estimation quality.
- Post-stratification techniques correct biases in unrepresentative national samples.
- Geographic smoothing compensates for weak local predictor data.
Why It Matters: These straightforward methods effectively address the challenge of estimating localized public opinion, offering practical solutions for researchers dealing with limited survey data across many constituencies.
