P1000 study (pulscar.io/blog/p1000-code) — method
Published October 2026. License: CC BY 4.0

Complaints. Source: NHTSA complaint flat files, complaints received 2015-01-01 to 2026-09-21 (780,071 narratives). Selection: four case-insensitive regular expressions written before coding (P1000 in any spelling; readiness wording; emissions/smog test wording; an emissions/smog/state inspection near reset, cleared codes or a battery event), 1,141 complaints; an independent completeness search added 209 (total 1,350). Coding: two independent classification passes (no access to each other's output) coded every complaint with the written codebook (codebook.txt). Category agreement 1,231/1,350, Cohen's kappa 0.861. Every disagreement was read and settled under written rules (R1-R23); a consistency pass swept the agreed codes for contradictions with those rules. Counts are complaint counts, not failure rates.

Amounts. Every dollar figure in relevant complaints was extracted by script and decided under one written inclusion rule (test fees, diagnosis, driving the monitors ready, software/modules for readiness or communication, the repair tied to the light, code or failed test, penalties for not passing). Status is as written: paid, billed, quoted, refunded or not stated. 138 figures in 114 complaints were kept; 95 were excluded with reasons.

Documents. NHTSA manufacturer-communication index (received 2015 to September 2026) searched for P1000 and readiness / I/M / monitor-completion / drive-cycle / emissions-test wording (494 documents), plus 108 documents found by the independent search and 9 older records found in NHTSA's 1995-2014 index files. PDFs downloaded from NHTSA one at a time; byte-identical copies detected by md5; near-duplicates grouped by 5-word shingles (Jaccard >= 0.6), one representative read in full and the others read through their sentence differences. 328 distinct documents were read this way (233 in full, 95 as near-duplicates); 201 records whose matching words were used in another sense were screened and listed one line each. In addition, the full text of 1,433 Ford, Lincoln, Mercury, Mazda, Jaguar, Land Rover, Nissan, Infiniti and Mack documents whose summaries mention emissions, monitors or engine software was searched for P1000 (90 had no usable text layer). Extractions carry verbatim quotes, checked by script against the PDF text.

Files: p1000-1350-complaints-read.csv, p1000-owner-reported-amounts.csv, p1000-documents.csv, codebook.txt.