Elections Have Consequences: The Impact of Political Agency on Climate Policy and Asset Prices

Description

Replication package for "Elections Have Consequences: The Impact of Political Agency on Climate Policy and Asset Prices" (William Cassidy), forthcoming in the Journal of Financial Economics. This package contains the code and author-generated data needed to reproduce the tables and figures in the paper. The paper studies how political agency shapes the origination and pricing of climate transition risk, combining a structural model of elections, climate policy, and asset prices with an empirical analysis of high-frequency asset-price reactions to presidential climate-policy announcements. Contents: - Full analysis code (Python, R, Stata, MATLAB, SAS) for the data pipeline, the option-implied expected-return (SVIX) construction, the regressions, and all figures/tables. - A README following the Social Science Data Editors template, documenting software requirements, execution order, and a mapping from each manuscript table/figure to the program that produces it. - Author-derived intermediate data that may be redistributed: the public-domain presidential-remark transcripts and the fitted 260-topic LDA weights used to construct the climate/energy text measures. Proprietary source data are NOT included and cannot be redistributed under their license terms: NYSE TAQ Millisecond, OptionMetrics IvyDB US, and CRSP (all via WRDS subscription), and Gallup polling series. The README gives full acquisition instructions, variable definitions, and the exact datasets used so that a researcher with the corresponding subscriptions can regenerate every intermediate file. The high-frequency market panel is rebuilt from WRDS TAQ; the option-implied expected returns follow Martin (2017) and Martin–Wagner (2019). Software: Python 3.10 (pandas 1.5.3, numpy 1.24.2), R 4.3.x (fixest, data.table), and MATLAB and SAS (for the option-implied SVIX construction). The only stochastic step (the LDA topic model) uses a fixed seed and is deterministic. Approximate end-to-end runtime is 1–3 days. License: This dataset is released under CC BY 4.0. The author-generated code in the package is additionally released under the MIT license (see LICENSE.txt); neither license extends to the proprietary source data (WRDS, Gallup), which are not included.

Steps to reproduce

Full details are in README.pdf (replication/ folder), which follows the Social Science Data Editors template and includes a table mapping each manuscript table/figure to the program that produces it. 1. Obtain data access. A WRDS subscription with TAQ Millisecond, OptionMetrics IvyDB US, and CRSP; and the Gallup polling series. The presidential-remark transcripts are public domain and require no credentials. Proprietary data are not included in this package (see the Data Availability section of the README for acquisition instructions and the exact datasets used). 2. Build the software environments. Python 3.10 (conda env "cpoli": pandas 1.5.3, numpy 1.24.2, matplotlib 3.7, pandas_market_calendars); base anaconda3 with sqlalchemy + psycopg2 for the WRDS pull; R 4.3.x with fixest and data.table; Stata; SAS and MATLAB for the option-implied expected-return (SVIX) construction; and a LaTeX toolchain (latexmk, pdflatex). 3. Stage the input data under a local tree (e.g. /tmp) as documented in REPLICATION.md ("Inputs you must provide"): the White House / Gallup / VIX intermediates, the CRSP SIC header, and the OptionMetrics volatility-surface file. 4. Run the pipeline from the repository root: - Full run (needs WRDS credentials): WRDS_USER=<user> WRDS_PW=<pass> ./run-all-from-wrds.sh - Downstream only (from an existing spy.csv): SKIP_BUILD=1 SPY=/tmp/spy.csv ./run-all-from-wrds.sh Outputs land in /tmp/repli/{output/tables, output/figures, draft/master.pdf, FALLBACKS.txt, logs}. 5. Option-implied expected returns (SVIX). The panel that feeds the option-based results is built by lib/optionmetrics/svix-construction/ (SAS on WRDS → MATLAB → R), then consumed by lib/optionmetrics/make-expected-return-panel.py. See that module's README for the stage order and environment-specific notes. 6. Compile the manuscript. Done automatically by the runner, or manually with latexmk -pdf draft/master_jfe.tex. Notes: The only stochastic step (the LDA) uses a fixed seed and is deterministic. Approximate runtime is 1–3 days end to end, dominated by the WRDS data pull; downstream tables and figures run in under an hour. Committed reference outputs in output/2026-06-25/ allow validation of regenerated tables and figures.

Institutions

• University of Notre Dame

  Indiana, Notre Dame

Categories

Funders

• University of Chicago

  Illinois, Chicago

Licence