Key facts
- 39,405 complaints received between 1 March and 15 October 2026 carry both the owner's state and the new state-of-incident field, out of 49,817 distinct vehicle complaints in that window.
- 37,144 (94.26%) record the failure in the owner's own state. 2,261 (5.74%) record it elsewhere.
- Only 52.94% of those out-of-state failures are in a bordering state. The other 47.06% are not, so this is not merely a commuting-across-the-line effect.
- Out-of-state failures are more severe on every flag tested, measured within web-filed complaints alone: fire 11.51% vs 5.58% (2.06x), towed 9.63% vs 5.57%, crash 7.97% vs 5.58%, injury reported 8.06% vs 5.61%.
- The complaint narratives corroborate it independently: 43.03% of out-of-state complaints use travel or relocation language against 19.81% of same-state ones (2.17x). On a stricter screen for trips and moves specifically, 12.12% vs 2.98% (4.07x).
- The field is effectively a web-form field. It is filled on 95.37% of complaints filed on nhtsa.gov and 10.49% of those phoned in to the hotline.
- It is blank on every complaint received from 2015 through 2025, and on 100% of January 2026. There is no history to compare against.
How often does a car fail away from home?
Of the 39,405 paired complaints, 37,144 put the failure in the owner's own state: 94.26%. Just 2,261 complaints, 5.74%, describe a failure somewhere else. That is lower than the share of US travel that crosses a state line, which is the first thing worth noticing: cars overwhelmingly break where they live.
The comparison is made within each record, so it cannot be distorted by one state reporting more diligently than another. That matters, because the new field's coverage does vary by state, and a count-based state ranking would be distorted by exactly that.
| Basis | Complaints | Share |
|---|---|---|
| Failure in the owner's own state | 37,144 | 94.26% |
| Failure in another state, bordering | 1,197 | 3.04% |
| Failure in another state, not bordering | 1,064 | 2.70% |
| Total paired records | 39,405 | 100% |
The biggest individual flows are all border pairs: Arizona to California (24), Georgia to Florida (24), Florida to Georgia (23), Pennsylvania to New Jersey (23), New Jersey to New York (22). But the long tail is genuinely long-distance, which is why only about half of the out-of-state total is adjacent.
Are breakdowns away from home worse?
Yes, and the gradient is consistent across four independent severity flags. The cleanest is fire. Among complaints filed through NHTSA's website, 11.51% of those reporting a fire place the incident out of state, against 5.58% of those that do not: a ratio of 2.06x.
| Severity flag | Out-of-state rate | Rate without the flag | Ratio |
|---|---|---|---|
| Fire reported | 11.51% (67 of 582) | 5.58% | 2.06x |
| Vehicle towed | 9.63% (92 of 955) | 5.57% | 1.73x |
| Injury reported | 8.06% (80 of 993) | 5.61% | 1.44x |
| Crash reported | 7.97% (124 of 1,555) | 5.58% | 1.43x |
| Any of the above | 8.15% (206 of 2,528) | 5.50% | 1.48x |
This table is restricted to web-filed complaints on purpose, and that restriction is load-bearing. Severe complaints are routed differently: 43.78% of complaints carrying a crash, tow, fire or injury flag come through the hotline, against 16.53% of routine ones. Because the hotline almost never records a state, a severity comparison across both channels would be comparing two different samples. Restricting to the website removes the channel entirely, and the gradient survives at full strength.
Why would it exist? A failure 400 miles from home cannot be left in the driveway until payday. It is more likely to happen at highway speed, more likely to need a tow rather than a short limp to a familiar garage, and more likely to be the kind of event the owner reports at all.
Could this just be people who moved house?
No. The alternative explanation is testable and it fails. The owner's state field is their current address. If a meaningful share of the out-of-state records were really people who had moved since the failure, the out-of-state share would rise with the delay between the incident and the filing.
| Failure-to-filing lag | Paired records | Out-of-state share |
|---|---|---|
| 0 to 7 days | 16,487 | 5.20% |
| 8 to 30 days | 8,779 | 6.52% |
| 31 to 90 days | 5,777 | 6.47% |
| 91 to 365 days | 4,973 | 4.99% |
| 1 to 3 years | 2,348 | 6.22% |
| Over 3 years | 790 | 7.34% |
The series is flat and non-monotone. Complaints filed within a week of the failure, where a house move is nearly impossible, sit at 5.20%, close to the 5.74% overall. Moves are in there somewhere, but they are not what the number is made of.
The complaint narratives say the same thing from a completely separate field. Scanning the free-text description for travel and relocation language, 43.03% of out-of-state complaints match against 19.81% of same-state ones. On a narrower screen that looks only for trips, vacations, being en route and relocations, it is 12.12% against 2.98%, a ratio of 4.07x. One filer put it plainly: a 2023 Jeep Wagoneer owner registered in Ohio reported being "stranded in North Carolina while driving for vacation" with an overheating alert.
Why you should not rank states on this field yet
Because it is a web-form field, and not everyone uses the web form. The new column is filled on 95.37% of complaints submitted through nhtsa.gov and 10.49% of those taken over the hotline. Every state's number is therefore missing most of its hotline filers, and states differ in how many of those they have.
| Website (IVOQ) | Hotline (EVOQ) | |
|---|---|---|
| Complaints, 1 Mar to 15 Oct 2026 | 40,336 | 9,481 |
| Carrying a state of incident | 38,469 | 995 |
| Coverage | 95.37% | 10.49% |
The hotline's share of a state's filings runs from 11.33% in Hawaii to 23.08% in West Virginia, and coverage of the new field tracks it inversely: Pearson r = -0.8513 across the 47 states with at least 100 filings. So a state-level count off this field is partly a measure of how that state prefers to complain. Our published dataset carries the per-state coverage and hotline share in every row so the bias is visible rather than buried.
For the same reason we are not publishing a ranking of which states import or export vehicle failures. The net balances are real but tiny: the largest is New Mexico at +7.69%, which is a net difference of just 16 complaints. Numbers that small do not survive contact with a 10-point swing in coverage, and a league table built on them would be noise with a map attached.
Does this overturn the state rankings already published?
No, and that is a useful negative result. Ranking the identical 39,405 records first by the owner's state and then by the incident state gives per-resident rates that correlate at Spearman r = 0.9649 across all 51 jurisdictions, with a median rank shift of two places. The largest single move is Wyoming, 47th to 33rd, on 52 complaints. New Hampshire is fifth on the owner's address and ninth on the incident state; Delaware moves up into the fifth place it vacates.
| Rank | By owner's residence | By incident state |
|---|---|---|
| 1 | Maine 14.52 | Maine 15.02 |
| 2 | Maryland 14.40 | Virginia 14.56 |
| 3 | Virginia 14.31 | Maryland 14.13 |
| 4 | South Carolina 14.04 | South Carolina 14.13 |
| 5 | New Hampshire 13.91 | Delaware 13.88 |
Rates are complaints per 100,000 residents, 1 March to 15 October 2026, and are not a defect league table: see the caveats above and in the dataset.
This is worth stating plainly because the opposite was a reasonable expectation. Every per-state vehicle-complaint map ever published, including our own in July 2026, used the complainant's address because nothing else existed. The new field is the first chance to check whether that proxy was wrong. It was not.
What a VIN check can and cannot tell you here
A VIN check cannot tell you where a car broke down, and neither can this dataset. NHTSA complaints attach to a make, model and model year, never to a VIN, so no vehicle history report can say whether your exact car was the subject of one. This new field records where a complainant's incident happened, which is a fact about an event, not about a car on a lot.
What a VIN check does cover is the car-specific record, and the part of it this article touches is geography of a different kind: a car's own history crosses state lines even when its owner does not. A Zilocar report can show:
- accident and damage records, including airbag deployment
- junk and salvage auction records
- odometer readings and rollback inconsistencies
- theft records reported to the NICB
- ownership history and sales-listing history, which is where interstate movement shows up
- whether open recalls exist against that VIN, with NHTSA and IIHS safety ratings and a valuation
What it cannot do, and no competitor can either: confirm that a recall was actually performed or a software fix flashed, show NHTSA's open investigations, read per-unit firmware, or classify the car's legal title brand. Title brands are the province of NMVTIS and state DMVs, and remedy status only a franchise dealer can confirm from the VIN.
Method
Source. NHTSA Office of Defects Investigation consumer complaint period files, retrieved 2026-10-20 at approximately 07:10 UTC:
| File | Archive bytes | Extracted bytes | Member timestamp |
|---|---|---|---|
COMPLAINTS_RECEIVED_2015-2019.zip | 81,302,720 | 357,305,037 | 2026-10-17 09:22:52 |
COMPLAINTS_RECEIVED_2020-2024.zip | 75,874,029 | 326,690,810 | 2026-10-19 09:18:12 |
COMPLAINTS_RECEIVED_2025-2026.zip | 42,001,069 | 186,823,539 | 2026-10-19 09:19:14 |
ZipFile.testzip() returned None on all three. The period files are used rather than the combined FLAT_CMPL.zip download, which has silently served stale extracts more than once. Freshness was asserted before anything was computed: the newest complaint in the three files is dated 2026-10-15, five days before the run.
Universe. All 1,113,814 rows carry 51 tab-delimited fields, matching the layout in CMPL.txt as last updated 2026-04-30. Filtering to PROD_TYPE = 'V' leaves 1,100,547 component rows, which deduplicate on ODINO to 785,544 distinct vehicle complaints (one complaint can cite several components, so rows are not complaints).
Window. Records with LDATE from 2026-03-01 to 2026-10-15: 49,817 distinct vehicle complaints. The start date is set by the data, not by preference: STATE_OF_INCIDENT is blank on 100% of complaints received in every year from 2015 to 2025 and on 100% of January 2026, then jumps to 7.36% in February and 75.83% in March, holding at 76% to 89% thereafter. The field was announced on 2026-04-30 but appears in the data from February.
Filter. Of those, 39,405 have both STATE (field 14, the consumer's state) and STATE_OF_INCIDENT (field 50) resolving to one of the 50 states or the District of Columbia. Records carrying a territory code or either field blank are excluded.
Aggregation. Agreement is measured per record: a complaint is "out of state" when the two codes differ. Adjacency uses a hand-built 51-jurisdiction border map, checked programmatically for symmetry (zero asymmetries). Severity comparisons are restricted to CMPL_TYPE = 'IVOQ' (NHTSA's website) because channel and severity are correlated; that restricted denominator is 38,415 paired records, of which 2,179 (5.67%) are out of state. Per-resident rates use Census Vintage 2024 state estimates. Rank correlations are tie-corrected Spearman coefficients; ties are heavy here, and an input-order tie-break would give a different answer. The residence-to-incident coefficient is 0.9649 computed at full precision; recomputed from the two-decimal rates in the published CSV it is 0.9673, which is the figure a reader reproducing it from that file should expect.
Reversal checks run, and their results.
- Intake-channel confound, the most likely reversal: tested and the finding held. Severe complaints are 43.78% hotline against 16.53% for routine ones, and the hotline records a state on only 10.49% of filings. Re-running the severity gradient inside the website channel alone left it intact (fire 11.51% vs 5.58%). The overall out-of-state rate is also stable across channels: 5.67% web, 8.28% hotline.
- The "they moved" alternative: tested and rejected. Out-of-state share by failure-to-filing lag is flat and non-monotone (5.20% / 6.52% / 6.47% / 4.99% / 6.22% / 7.34%), where a move effect would rise steeply.
- Independent corroboration from a separate field. Free-text narrative screening for travel and relocation language gives 43.03% against 19.81% on a loose screen and 12.12% against 2.98% on a strict one. Two screens, same direction, ratios of 2.17x and 4.07x. Report the ratio, not either level.
- Per-state ranking refused rather than published. Net incident balances are within the coverage noise (largest, New Mexico, +7.69% on a net of 16 complaints), so no importer/exporter league table is offered.
Known limits. Complaints are unverified allegations, not confirmed defects, and a count measures reporting behaviour as much as vehicle condition. The window is seven and a half months of one year and supports no trend. The state field's 65.88% coverage across 2026 as a whole is not missing at random; it is missing by intake channel, and that channel preference varies by state. Severity flags are filer-entered. Because blank Y/N fields in these files were retroactively rewritten to N in 2021, every flag-based rate here is a floor rather than a point estimate. The dataset is frozen; NHTSA republishes these files continuously, so a re-pull will not reproduce these counts exactly.
Frozen dataset. Per-state counts on both bases, cross-state flows, coverage and hotline share: car-breakdowns-away-from-home-nhtsa-incident-state-data-2026-10-20.csv
Sources
- NHTSA Office of Defects Investigation, consumer complaint period files: COMPLAINTS_RECEIVED_2025-2026.zip, 2020-2024, 2015-2019. Retrieved and analysed 2026-10-20.
- NHTSA, complaint file layout and change log: CMPL.txt, last updated 2026-04-30, which records "Fields 50-51 were added on April 30, 2026".
- U.S. Census Bureau, Vintage 2024 state population estimates.
- Zilocar frozen dataset for this analysis, linked above.
Last verified: 2026-10-20.
