Replication Data for: Idiosyncratic Issue Opinion and Political Choice

# Idiosyncratic Issue Opinion and Political Choice This zipped folder contains the original data and replication code for the above book. The folder `R/` contains the code necessary to repeat the analysis and produce the materials used in the book. The code necessary to repeat the analysis is contained in files: - 01_polychoric_correls.R - 02_decomposition.R - 03_stability.R - 04_conjoint.R - 05_conjoint_flexible.R - 06_polychoric_stability.R The file `public_opinion_book.R` contains the code to produce the graphs in the book. The folder `data/` contains the original survey (`data/survey/HLV_survey_allwaves.Rdata`) and details on the survey instrument (`data/survey/Instrument_Specification.csv`) It also contains auxiliary information regarding seats and manifestos. The folder `archive/jop_replication` contains materials from our previous article, https://doi.org/10.1086/695673 ## Data not distributed herein The code references two files which we cannot distribute: - `anes_mergedfile_1956to1960.dta`, used to replicate the original Converse findings. As of December 2025, this file can be downloaded from https://electionstudies.org/data-center/1956-1960-panel-study/ - `SPSS63-70PoliticalChangeinBritainData.sav`, used to replicate the Butler-Stokes findings. As of December 2025, this file can be downloaded from https://www.britishelectionstudy.com/data-object/1963-1970-political-change-in-britain/ The analysis of these data-sets is largely self-standing. ## Replication effort This replication package is extremely compute intensive. A powerful desktop computer running the analysis sequentially will take roughly two weeks to perform all of the analyses in this package. It might be possible to speed up the process of replication by running some analyses in parallel, but users will have to keep an eye on their memory usage while doing so. ## Replication fidelity This book uses Hamiltonian Monte Carlo methods to estimate key quantities. This means that the same model estimated on different machines using different software setups can generate different estimates. The degree of error should be small, and should only affect the third significant digit. For some quantities, however, there are differences at the second significant digit. For example: the book reports (Figure 4.10) that 67.5 of the variation in responses to the foreign aid question is explained by ideology. When the analysis is run on one of the authors' machines, the estimated figure is 66.2, rather than 67.5. However, the substantive conclusions from the figure regarding which issues tend to elicit more or less ideological opinion variation remain supported, and the average variation in responses explained by ideology is the same to three significant figures. If you would like to report a significant discrepancy (a discrepancy at the first or second significant digit), please email Chris Hanretty [[e-mail]].