ARCHIVED APPROVAL AND FAVORABILITY POLL FILES (consolidated, September 2026)
Companion to "Beyond Isolated Headlines: A Framework for Coupling Scorekeeping and
Explanatory Modeling in Public Political Polls." See Appendix A.3.

FILES COUNTED IN THE PAPER (four files; sampling order 1-4)
  1. favorability_polls_consolidated.csv                    5,809 favorability records (derived; see below)
  2. president_approval_polls_historical.csv               10,017 approval records (FiveThirtyEight, downloaded 2024-07-06)
  3. president_approval_polls_current_capture-2025-03-06.csv 5,155 approval records (FiveThirtyEight, Internet Archive capture)
  4. nyt-approval-tracker-frozen-2026-07-11.csv              1,104 approval records (New York Times tracker, downloaded 2026-07-11)
  Combined records = 22,085     Combined organizations = 177
  Run record-count.py and sponsor-count.py in the parent folder to recompute every total.

RAW INPUTS TO THE DERIVED FILE
  favorability_polls_2024-07-06.csv           3,830 records, 17 politicians (FiveThirtyEight, downloaded 2024-07-06)
  favorability_polls_capture-2025-01-18.csv   4,617 records, 4 politicians (FiveThirtyEight, Internet Archive capture)
  favorability_polls_consolidated.csv = every row of the 2025 capture + the 1,192 rows of the
  July 2024 file whose question_id is absent from the capture. Rows are copied verbatim; the
  source_capture column names the download each row came from. consolidate-favorability.py
  rebuilds it. The two downloads cover different rosters: the capture adds Kamala Harris and
  JD Vance and omits the 2024 primary-season candidates; the union carries 19 politicians.

SUPERSEDED
  superseded-2024-07-06/president_approval_polls_current_2024-07-06.csv  4,774 records. Every
  record is contained in the 2025-03-06 capture, which replaces it. Kept because the July 2026
  audit workbook (link-audit-400.xlsx, in superseded-2026-07/) was drawn over it.

SOURCES
  FiveThirtyEight endpoints  https://projects.fivethirtyeight.com/polls-page/data/president_approval_polls.csv
                             https://projects.fivethirtyeight.com/polls-page/data/favorability_polls.csv
  Internet Archive captures, retrieved 2026-09-12:
    approval      https://web.archive.org/web/20250306125402/https://projects.fivethirtyeight.com/polls-page/data/president_approval_polls.csv
                  raw bytes: https://web.archive.org/web/20250306125402id_/...  (same path)
    favorability  https://web.archive.org/web/20250118200335/https://projects.fivethirtyeight.com/polls-page/data/favorability_polls.csv
                  raw bytes: https://web.archive.org/web/20250118200335id_/...  (same path)
  The latest record in either capture was created 2025-01-17. The New York Times tracker is a
  live resource; the frozen 2026-07-11 snapshot deposited here is the file counted, and a later
  download does not reproduce it.

REPRODUCING THE 400-RECORD SAMPLE
  Concatenate files 1-4 in the order above (22,085 rows) and draw
  numpy.random.RandomState(1963).choice(22085, size=400, replace=False), preserving the returned
  order as draw_order 1-400. draw-sample.py in the parent folder does this and writes an
  uncoded link-audit-400-draw.xlsx; the coded verdicts are in link-audit-400-v2.xlsx.

SHA-256 CHECKSUMS
  de9f776b2f90a673f898f2cd6a56b62b3eef52bed58897eeadb763eb2a2775ec  favorability_polls_consolidated.csv
  d5fa09818882445deeefcb242d9377986a9e5ab734cd4d053cc9718eff3164a8  president_approval_polls_historical.csv
  fb112f2548891e9d96a8e0024aa840bc8e909e4375a5ed5db237a1934551a406  president_approval_polls_current_capture-2025-03-06.csv
  239f83aa2c7fee212e162b7063eb5bc51fe87f8262373f467496a760061bf097  nyt-approval-tracker-frozen-2026-07-11.csv
  eb11d711abc8194f466b7b659135b4b336001c73f200bd705ca0df8b8e3f4b79  favorability_polls_2024-07-06.csv
  70c5c15a1a66c377c73d2201057790c91c31a97b09e76afdffb5a210cef2fb47  favorability_polls_capture-2025-01-18.csv
  1160bc3b0ac1c205642c62fc372600f1f0b2104f9b9a1b73ffbaf2ff67c71622  superseded-2024-07-06/president_approval_polls_current_2024-07-06.csv
