Partisan Cues and Evaluations of the Same Economic Information
Carolyn E. Roush and Gaurav Sood
Paper · Design and interpretation · Data · Validation
Do people judge the same economic information differently when it is associated with their own party or the opposing party? Partisan differences in factual answers can reflect different information as well as different interpretations. Holding the supplied numbers fixed helps distinguish those possibilities.
Two online experiments, on MTurk in June 2020 and Lucid in January 2023, present small declines in unemployment and inflation under a randomly assigned Obama or Republican Congress cue. Respondents choose “got better,” “stayed about the same,” or “got worse.” The main analysis includes consenting, assigned partisans, including leaners, with observed outcomes and excludes Lucid previews. Screened and strict-party samples provide sensitivity analyses.
Across two online samples, evaluations were less favorable under the opposing-party cue. The evidence is clearest for both outcomes on MTurk and for inflation on Lucid; the Lucid unemployment estimate is less precise and includes zero. The experiments measure how people evaluate supplied information.
| Sample | Partisans, including leaners | Unemployment | Inflation |
|---|---|---|---|
| MTurk, June 2020 | 1,425 | -9.5 [-12.5, -6.5] | -6.0 [-9.2, -2.8] |
| Lucid, January 2023 | 624 | -5.2 [-10.6, 0.2] | -8.6 [-14.2, -3.1] |
Entries are opposing-party minus own-party cue effects, with 95% confidence intervals. Scores are 0 for “got worse,” 50 for “stayed about the same,” and 100 for “got better.” A negative value means less favorable evaluations under the opposing-party cue. Binary “got better” estimates are in the paper. Comparisons are made within respondent party and averaged using its share of the analytic sample.
Install R 4.6 and a TeX distribution with XeLaTeX, BibTeX, and latexmk. From the repository root:
make restore
make checkmake check runs the analysis, generates figures, tables, manuscript numbers and this README, compiles ms/main.pdf, lints the R sources, and runs the analytical tests. To edit this README, change docs/README.in.md; make analysis supplies its estimates.
data/raw/: numeric public extracts used by the analysis.data/materials/: plain-text fielded questionnaires and the Lucid demographic codebook.scripts/: the numbered analysis pipeline described below.data/derived/: generated, ignored intermediate data.tabs/,figs/: generated estimates, diagnostics, tables, and figures.ms/main.tex: manuscript source;ms/references.bib: references.tests/testthat/: coding, estimand, variance, and regression checks.
make analysis runs scripts/99_run_all.R, which loads the configuration and helpers, then runs stages 01–04 in order. Each stage has its own environment and passes results through files.
| File | Purpose |
|---|---|
scripts/00_config.R |
Project paths, study names, plot colors, theme_paper(), figure dimensions, and table_style. |
scripts/00_utils.R |
Reusable coding, estimation, and output functions used by the stages and tests. |
scripts/01_prepare_data.R |
Validate and recode the public extracts into data/derived/prepared_data.rds. |
scripts/02_estimate.R |
Read the prepared data and write estimates and diagnostics to tabs/. |
scripts/03_figures.R |
Read the estimates and write the figure to figs/. |
scripts/04_tables.R |
Generate table rows, manuscript numbers, table styling, and this README. |
scripts/99_run_all.R |
Run the complete public analysis. |
Plot and table defaults live together in 00_config.R. The manuscript loads the generated tabs/style.tex and applies \TableStyle to each table. The optional importer below is separate from the public pipeline.
The public extracts omit direct identifiers and free text. Authorized holders can regenerate the extracts with Rscript R/01_prepare_public_data.R /path/to/original/data. Source hashes are recorded in docs/source_manifest.csv. See data documentation for coding and provenance.
Citation metadata is in CITATION.cff.
- finite-sample/know_pgap_review — Replication Materials For "A Gap in Our Understanding? Reconsidering the Evidence for Partisan Knowledge Gaps"
- finite-sample/know_hidden — Mis-measuring Political Knowledge? Do People Know More—or Even Less—about Politics than Commonly Thought?
- finite-sample/know_casablanca — The Waters of Casablanca: On Political Misinformation (Luskin and Sood). Data, code, and manuscript.
- finite-sample/know_pgap_format — How do (biased) guessing encouraging features and guessing agnostic coding techniques affect the partisan gap?
- finite-sample/know_misinfo_misinfo — Misinformation about Misinformation? Of Headlines and Survey Design (Luskin, Sood, Park, Blank)
Powered by Adjacent