Real errors and format breaks caught while building the baseline pipeline, per the discipline set in housing/baseline/incompatibility-log.md and elder-care's equivalent.
- NAICS code splice, 2022. BLS QCEW's newspaper-publisher industry code changes from
511110(NAICS 2007/2012/2017) to513110(NAICS 2022) between the 2021 and 2022 annual files. The pipeline (pull_qcew_newspapers.py) detects the splice year from the data rather than assuming it, and checks the year-over-year employment jump at the splice is under 15% before accepting the series as comparable. Observed jump: +1.6% (102,521 → 104,153) — clean. - Census CBP pre-2014 national files 404 at the modern URL pattern.
https://www2.census.gov/programs-surveys/cbp/datasets/{year}/cbp{yy}us.zipreturns a valid zip for 2021 but a 404 (HTML error page, not a zip) for 2005. Pre-2014 CBP national files live under a different path/naming convention not resolved within Phase 0's time budget. This blocks extending the newspaper-employment series back to the pre-2008 peak using CBP as a second data source; QCEW alone (2014–2025) is the committed series for Phase 0. Queued as a baseline next-increment item. - QCEW annual API has no pre-2014 window for this slice.
data.bls.gov/cew/data/api/{year}/a/industry/{naics}.csvreturns 404 for 2005 and 2013 alike for NAICS 511110, despite QCEW itself holding data back to 1990. The open CSV-slice API's coverage window is narrower than the underlying series; a different QCEW access route (the singlefile/historical annual averages files) would be needed to reach the mid-2000s peak that Pew's and NAA's employment/ad-revenue figures reference. Not resolved in Phase 0; the anchor-1/anchor-3 verification for the mid-2000s baseline therefore rests on secondary literature (Pew/NAA restatements of BLS), not a direct pull, and that is stated in the anchor table rather than presented as admin-data-verified.