📌 What This Paper Does
Introduces an instrumental variables (IV) framework to study how external factors that change survey response rates can also alter the composition of respondents. The approach is intended to help evaluate survey representativeness and to assess whether conventional corrections for nonresponse bias (such as weighting) actually correct composition shifts.
🔍 How Response Incentives Were Used in the Data
- Data come from the 2011 Swiss Electoral Study (SES).
- Participation incentives were randomly assigned across members of the original survey sample, creating a natural experiment for response behavior.
- Introduces a new IV parameter, the Complier Average Survey Response (CASR), that isolates the characteristics of those induced to respond by the incentive.
📊 Key Findings
- Incentives increased response rates substantially in the SES sample.
- CASR estimates indicate the incentives induced participation among people with more nationalist political opinions than those who would have participated without incentives.
- Reweighting respondents to match the target population's covariate distribution did not eliminate the attitudinal differences between compliers and always-responders, suggesting such weights would not remove this form of nonresponse bias.
💡 Why This Matters
- Demonstrates that factors that raise overall response rates can systematically change respondent composition in politically meaningful ways.
- Shows that common corrective measures (covariate weighting) may fail when nonresponse selection is linked to unobserved attitudes.
- Provides a practical IV-based tool (CASR) for diagnosing representativeness problems and evaluating the limits of standard nonresponse adjustments in household surveys and political polls.
