# Codebook — readiness / P1000 complaints (P1000 study)

Written before any complaint in the set was coded.

## Pool (how the complaints were found)
NHTSA complaints received 2015-01-01 to 2026-09-21 (780,071 narratives). A complaint is in the pool if
its narrative matches any of four case-insensitive regular expressions (full text in pool.py):
1. The code P1000: `(?<![A-Z0-9])P[\s\-.]?[1I][0O]{3}(?:[\s\-]?[0-9A-F]{2})?(?![0-9A-F])`, or a bare
   1000 after "code(s)", "DTC(s)", "error" or "CEL" (not followed by miles/rpm/dollars/km), or "1000 code".
2. Readiness wording: "readiness"/"readyness"; "not ready" within 80 characters of emission/smog/
   inspect/test/monitor/state/sticker; "monitor(s)" followed within 60 characters by not/incomplete/never/
   failed to/unable to and then complete/set/ready/run/reset/finish; "incomplete monitor(s)"; "I/M";
   "drive cycle(s)" / "driving cycle(s)".
3. Emissions-test wording: emission(s)/smog followed within three words by test/testing/check/inspection/
   station/certificate, or failed/rejected/passed/turned away followed within four words by emission(s)/smog.
4. An emissions/smog/state or annual inspection/inspection sticker within 120 characters of reset/
   cleared/clearing codes/battery disconnected, replaced, died or dead.

P1000 itself is almost never written by owners, so the study reads every complaint about readiness,
"not ready", drive cycles and emissions-test outcomes, and records why the car was not ready or was
turned away.

For each complaint record exactly one CATEGORY and four yes/no FLAGS.

## RELEVANCE
N — Not about this vehicle's readiness, P1000, or an emissions/smog test outcome. Examples: a safety
    inspection (brakes, rust, lights, airbag light) with no emissions test; "emissions" meaning the
    diesel emissions scandal, a recall or a lawsuit with no test outcome on this vehicle; a dealer
    "multi-point inspection"; "drive cycle" used for something unrelated (e.g. a transmission relearn)
    with no readiness or emissions-test context; the number 1000 meaning miles, rpm or dollars; an
    emissions test mentioned only hypothetically ("I won't be able to pass emissions") with no test,
    readiness reading or code on THIS vehicle. If THIS vehicle was taken to an emissions/smog test, or
    its readiness/monitor status was read, or P1000 was read from it, it is NOT N.

## CATEGORY (one letter) — why the car was not ready, could not be tested, or failed/was rejected

Readiness group (the complaint says the car was "not ready", monitors were incomplete/not set/reset,
P1000 was read, the car needed a drive cycle before testing, or was rejected/failed for readiness):

F — A fault named. A specific component or system fault (EVAP leak or purge/vent valve, catalytic
    converter, oxygen or air-fuel sensor, thermostat/coolant temperature, EGR, secondary air,
    misfire, fuel system, a stored or pending code) is stated by a shop, dealer, station, scan tool
    or maker as the reason a monitor would not complete, or its repair is what finally let the
    monitors complete.
M — Computer or software. A software update, reprogram, reflash or calibration, a maker bulletin
    that consists of software, or replacement of the engine/powertrain computer is stated as the
    fix that let the monitors complete, or a software/computer defect is stated as the reason they
    would not complete; without F.
K — No communication. The test station or a scan tool could not communicate with or read the car's
    computer (no communication, cannot connect, port or connector fault, computer not responding,
    the computer did not report), so the OBD test could not be done; without F or M.
B — Battery. The monitors were not ready (or reset) after the battery died, was disconnected,
    jump-started, charged or replaced, as stated, without F, M or K.
C — Cleared or reset by a repair. Codes were cleared, the computer was reset, or a repair, recall,
    software update or service reset the monitors shortly before the test, as stated, and no
    persistent problem in F or M; without B.
U — Readiness problem, cause not stated. Not ready / incomplete / P1000 / rejected for readiness /
    "needs a drive cycle", with no cause or fix in F, M, K, B or C stated (dealer could not find
    anything, owner still waiting, owner only guesses).

Test-outcome group (an emissions/smog test of THIS vehicle failed or was refused, readiness NOT
stated as the reason):

L — Failed or rejected because the check-engine (malfunction indicator) light was on or a stored or
    pending code was found.
O — Failed or rejected for another stated reason (tailpipe readings, gas-cap test, visual or
    tampering check, open recall, paperwork), or the test outcome is stated with no reason.

Precedence when more than one fits: F, M, K, B, C, U, then L, then O. The readiness group always wins
over the test-outcome group. A cause counts when the complaint (owner, shop, dealer, station, scan
tool or maker) states it as the cause or the fix; a cause the owner merely guesses at does not count
(use U). A repair done for a different, unrelated problem in the same complaint does not count.

## FLAGS (y/n)
p1000      — y if the complaint names P1000 (any spelling: P1000, P-1000, P1000-00, "code 1000") as a
             code read from THIS vehicle.
reset      — y if the complaint states that a battery event, a code clear/computer reset, a repair, a
             recall or a software update happened before the not-ready state or the failed/rejected
             test (whatever the category).
persistent — y if the monitors/readiness are stated as still not ready after a stated period or
             amount of driving (days, weeks, months, miles, "drove for a long time"), or the car was
             rejected/failed at two or more test visits for the same problem.
consequence— y if the owner states a consequence beyond the test itself: could not register or renew,
             expired registration/tags/sticker, a ticket or fine, retest or diagnosis fees, a waiver
             or extension sought, missed work or a car they could not legally drive.
For N complaints write all four flags as n.

## OUTPUT
One CSV line per complaint, no header, no spaces after commas:
odino,category,p1000,reset,persistent,consequence,note
note = at most 15 words, no commas, saying what the complaint states (e.g. "EVAP monitor never set after
battery; dealer reflash fixed"). Every complaint in your chunk gets exactly one line.

## Addendum (wave 2, written before coding the completeness-audit additions)
An independent completeness audit searched the complaint file again and found complaints the four
regular expressions above missed: "smog" used as a verb or outcome ("cannot smog", "get it smogged"),
state inspections failed or blocked by the check-engine light or an OBD code, misspelled emissions
wording ("emmissions", "emision"), test stations unable to read the vehicle, and readiness wording
without the word readiness ("Emission Ready: NO"). These complaints are coded with exactly the same
relevance rule, categories, precedence, flags and output format as above. A state inspection that
failed or was blocked by the check-engine light or an OBD code counts as an emissions/OBD test
outcome (category L) unless readiness wording puts it in the readiness group; a failure for a safety
item only (brakes, lights, rust, airbag) is N.


## Hashes (written before coding)
2026-10-01T07:32:48Z
71341b8f2bf700d2f1648666017c99e707e8e33fbba5093bb14db7c0aad7fd48  codebook.md
(codebook.md = the text above the Addendum)
2026-10-01T08:15:48Z
869adc2e8ec9cd717d6d4042871e9f562a70855768324ee4f0043ff70cde46b4  codebook_wave2.md
(codebook_wave2.md = the full text above)

## Adjudication rules (written while deciding the 176 coder disagreements and applied to the consistency pass)
R1: L needs a stated link between the failed/refused test and the lit light, a code or a diagnosed fault; a failed test merely listed among unrelated problems is O.
R2: A failed/refused test the complaint ties to a lit light, a warning message or stored codes is L; a failure attributed to a named component with no light, message or code stated is O (as the codebook lists tailpipe, gas-cap and visual reasons under O).
R3: persistent needs a stated period of driving with monitors still not ready, or two or more stated test visits; "the same issue again" with no second test stated is n.
R4: "Cannot/will not pass" stated about THIS vehicle while its light is on or a code is stored is L (the codebook hypothetical-N example requires no test, reading or code on the vehicle); with no code, light or test it is N.
R4b: R4 counts an engine or emissions warning message on the dash (e.g. reduced propulsion) the same as a lit light.
R5: reset flag = the battery event, clear, repair, recall or software update is stated in the same episode, shortly before the not-ready result or the failed test; repairs years earlier or after the failure are n.
R6: A repair, code clear, recall or service stated immediately before a not-ready result is C even if the complaint does not say in words that it reset the monitors.
R7: A test the complaint says the car PASSED, with the complaint about something else, is N.
R8: A bare list of failed system checks or codes with no emissions test, readiness or P1000 context is N.
R9: A visit to a test station with no emissions outcome stated (complaint about something else found there) is N.
R10: When the station or a tool cannot connect to or communicate with the computer, the category is K even if a software update is offered as the remedy; M is reserved for monitors that would not complete.
R11: K only when the complaint says the station/tool could not communicate with, connect to or read the computer; an analyzer that could not register rpm, with a diagnosed code, is L.
R12: consequence counts registration/legal-operation problems, tickets, and test, retest or diagnosis fees; repair bills or a car unsafe to drive are not consequences.
R13: A failure for a safety item (brakes, lights, rust, airbag) at a combined safety and emissions inspection is N: not an emissions outcome.
R14: Codes or the light that will not clear after a repair until the car completes drive cycles (the dealer or maker says the monitors must re-run) is the readiness group, C, even with no test visit.
R15: P1000 listed among other codes with no statement of what kept the monitors from completing is U (p1000=y), even when other faults are repaired in the same complaint.
R16: A test "rejected" because the computer is "not updating" (no communication failure stated) is read as a readiness rejection; with no cause stated it is U.
R17: "Still cannot pass" after a repair implies a failed test of this vehicle: O when no light or code is stated (L when one is).
R18: A software update done in order to pass the emissions test, with no communication or readiness wording, is O (a test outcome with a stated system reason).
R19: A general statement about how a model behaves ("can take over 1000 miles to reset", "can cause the vehicle to fail") with no test, reading or event on the complainant's own vehicle is N, as the codebook says.
R20: A scan showing readiness not complete ("Emission Ready NO") together with active fault codes the dealer confirmed is F.
R21: A failed or refused test that the station or dealer cures by having the owner drive the car (a stated distance, or extended dates to drive around) before a retest, with no repair, is a readiness case: U unless a cause is stated.
R22: "Not being able to complete the smog test" with no communication failure or cause stated is U.
R23: An on-board computer that lost its memory (as stated) and shows readiness not ready, with no cause named, is U with reset=y.

## Pool history
- First search: 1,141 complaints (P1000 regex 3, readiness wording 222, emissions-test wording 887, inspection + reset 84; overlaps), chunks 1-5.
- Completeness audit (independent search): 209 complaints added as chunk 6 (smog as a verb, check-engine-light inspection failures, misspellings, station cannot read the vehicle, readiness wording without the word readiness).
- Total 1,350. Category agreement 1,231 of 1,350, Cohen's kappa 0.861; flags: p1000 1.000, reset 0.837, persistent 0.892, consequence 0.871.
- 176 disagreements decided under the rules above; 105 agreed codes changed in the consistency pass (each row in the complaints CSV names its rule).