PULSCAR ANALYSIS OF NHTSA PUBLIC DATA — CLUNKING NOISES Published September 2026. License: CC BY 4.0 (credit "PulsCar analysis of NHTSA public data, September 2026"). ------------------------------------------------------------------------------- 1. SOURCE FILES ------------------------------------------------------------------------------- NHTSA public ODI flat files, downloaded 17 September 2026 from https://static.nhtsa.gov/odi/ffdd/ (field definitions: cmpl/CMPL.txt, rcl/RCL.txt, ftsb/FLAT_TSBS.txt), loaded into a local SQLite database. complaints 777,040 rows, received 1 January 2015 - 8 September 2026 recalls one row per make/model/year per campaign (campaigns DISTINCT) tsbs manufacturer communications Complaints are self-reported and unverified, with no denominator. Nothing below is a failure rate and no reliability ranking can be drawn from it. ------------------------------------------------------------------------------- 2. THE COMPLAINT SET (4,441 complaints) ------------------------------------------------------------------------------- Query: complaints whose narrative contains the word CLUNK (any form: clunk, clunking, clunked) -> 4,441. We deliberately did NOT filter by component tag. The point of the set is to see where a clunk actually lands once owners describe it, and pre-filtering to "suspension" would have decided the answer in advance. towed 125 2.8% (baseline: 39,739 of 777,040 = 5.1%) crash reported 47 1.1% (baseline 4.4%) fire 19 0.4% injuries reported 33 people; deaths reported: 0 Failure mileage: 2,466 report a usable figure (drop 0 and >= 400,000). Median 65,000; 25th percentile 27,065; 75th percentile 103,000; 754 (30.6%) at or under 36,000 miles. NHTSA component tag of the 4,441 (tags can list several systems, so these overlap and do not sum to 4,441): POWER TRAIN 1,839 STEERING 1,207 SUSPENSION 912 ENGINE 632 SERVICE BRAKES 252 STRUCTURE 215 WHEELS/TIRES 101 Part words in the narratives (each is the share of 4,441 that names the part at all, not a diagnosis): differential / rear end 225 (5.1%) steering rack or gear 187 (4.2%) control arm 176 (4.0%) strut or shock 151 (3.4%) U-joint or driveshaft 107 (2.4%) bushing 98 (2.2%) ball joint 83 (1.9%) sway/stabilizer bar 76 (1.7%) tie rod 67 (1.5%) engine/trans mount 53 (1.2%) CV axle or half shaft 52 (1.2%) intermediate shaft 41 (0.9%) sway bar link 34 (0.8%) strut mount/bearing 21 (0.5%) Ball joints specifically: 83 complaints name one; 32 of those also use a separation word (separated, broke, snapped, failed, collapsed, came apart, fell). Two of those 32 report a crash and four report a tow. Small numbers, reported as small. "Shop could not find it" language — could not duplicate/replicate, no problem found, nothing wrong, operating as designed, "is normal", "characteristic of", within spec: 220 complaints (5.0%) using the tightened phrase list. A looser first pass that counted any use of the word NORMAL returned 734 and was rejected after reading matches: it was catching "normal driving" and "normal speeds". By make and model: clunk-complaints-by-make.csv (makes with >= 10) and clunk-complaints-by-model.csv (models with >= 12), each with median failure mileage, top model years, and how many of that model's complaints mention the transmission/shift, driveline, steering, suspension or brakes. Notable model clusters and what their narratives mention (counts are word mentions, not diagnoses): Ford F-150 300 complaints, 266 mention transmission or shifting Hyundai Sonata 203 complaints, 167 mention steering Jeep Cherokee 160 complaints, 116 mention transmission or shifting Kia Soul 101 complaints, 82 mention steering ------------------------------------------------------------------------------- 3. 110 COMPLAINTS READ IN FULL ------------------------------------------------------------------------------- From the 2,833 complaints in the set whose narrative contains an outcome cue (DEALER, SHOP, MECHANIC, TECHNICIAN, REPLACED, DIAGNOS, TOLD ME, FOUND), we drew 110 at random (Python random.seed(2026)) and read every one, assigning each a single primary outcome: transmission or driveline — transmission, torque converter, valve body, PTU/RDM, differential, driveshaft, transfer case, axle, EV drive unit 49 steering — rack, gear, coupler, intermediate shaft, EPS unit, tie rod 17 suspension or chassis — control arm, ball joint, strut, shock, spring, sway bar, subframe or crossmember rust, a wheel or lug nut loose 14 nobody found it — could not duplicate, "that is normal", or parts replaced with the noise still there 12 engine internals 4 software reprogramming named as the repair 4 brakes 3 engine or transmission mount 2 other (a seat frame, a door hinge weld) 2 off target (a three-wheeler, two stalls unrelated to a clunk) 3 READ THIS WITH ITS BIAS. NHTSA is a safety-reporting channel. Owners write to it when a noise arrives with a loss of power, a wheel that moved, a crash risk or a refused warranty claim — not when a $150 sway bar link rattles over a driveway lip. That is the most likely explanation for why transmission and driveline outcomes dominate this sample while the cheap suspension parts that shops replace most often barely appear. The sample describes the complaints, not the repair market, and we make no claim that 45% of clunking cars need a transmission. ------------------------------------------------------------------------------- 4. RECALLS ------------------------------------------------------------------------------- recalls WHERE component names a BALL JOINT, CONTROL ARM or STABILIZER, campaigns counted DISTINCT: 162 campaigns, 4,298,036 vehicles potentially affected. Of those, 72 campaigns (1,569,775 vehicles) have a defect description containing a separation word (separate, fracture, break, broken, fall). Published as clunk-suspension-recalls.csv. For context on the wider system: 41 campaigns name a suspension ball joint (1,131,134 vehicles), 121 name a control arm (2,385,050), 41 name a stabilizer (1,912,986), and 798 campaigns touch suspension at all (15,614,975). Read in full for this article: 24V-132 (Chrysler/FCA, 22 February 2024, 338,238 vehicles, 2021-2023 Jeep Grand Cherokee L and 2022-2023 Grand Cherokee). Part 573 report RCLRPT-24V132-5098.PDF. Defect: "A damaged pinch bolt may break and result in the loss of clamp load between the UCA ball joint and the knuckle." Safety risk: the upper control arm separating from the knuckle "and the wheel falling outboard, which can cause a vehicle crash without prior warning." And, in the field the report titles "Identification of Any Warning that can Occur": "Driver may notice an abnormal noise when driving over bumps." Remedy: replace the upper control arm pinch bolts. 13V-385 (Ford, 29 August 2013, 355,000 vehicles, 2005-2011 Crown Victoria, Grand Marquis and Town Car registered in 23 salt-belt states and DC): severe corrosion can seize the lower intermediate shaft, which may cause the upper intermediate shaft to collapse and the steering column lower bearing to separate, with a loss of steering. Remedy: replace the lower intermediate shaft. 25V-011 (Chrysler/FCA, April 2025, 63,130 vehicles, 2017-2019 Jeep Cherokee): the power transfer unit may become damaged and disengage the transmission and differential, causing a loss of drive power and/or loss of park function. ------------------------------------------------------------------------------- 5. MANUFACTURER COMMUNICATIONS (1,140 documents) ------------------------------------------------------------------------------- tsbs whose summary contains CLUNK: 1,140 distinct documents. By make: Chevrolet 186, GMC 145, Ram 89, Honda 83, Cadillac 74, Acura 72, Dodge 61, Jeep 49, Ford 41, Buick 40, Toyota 35, Mazda 26. Published in full as clunk-manufacturer-bulletins.csv with the document id, bulletin number, date, a derived area (which system the summary names) and a derived remedy type. Where the summaries point (a document can name more than one area): suspension or strut 362 · transmission or shift 286 · driveline 224 · steering 128 · body or seats 97 · brakes 58 147 name reprogramming, software, a flash or a calibration. 71 describe the noise as a normal condition, a characteristic, or tell the technician not to repair it. Documents opened and read in full: GM 99-04-20-002O, July 2024, "Information on Driveline Clunk Noise", applying to 2025-and-prior GM passenger cars and light/medium trucks (MC-11005311-0001.pdf). It explains the clunk as accumulated lash between driveline components, tells technicians to compare the vehicle with a like vehicle ("If the results are the same, this is a normal condition"), and states that "dealers are discouraged from attempting to repair driveline clunk conditions" because the noise is the sum of designed clearances, changing parts may not reduce it, and it "will not adversely affect durability or performance." Nissan NTB19-050, 19 June 2019, "2019 Altima; Clunk Noise From Rear When Driving Over Dips And/Or Potholes" (MC-10162981-9999.pdf): if the noise is isolated to the rear shock absorbers, replace both; the bulletin's own flat-rate labor time is 0.6 hours. GM 20-NA-022, July 2022, "Clunking Noise from Rear of Vehicle (Re-torque Leaf Spring Bolts)" for 2020-2023 Silverado/Sierra 2500HD and 3500HD (MC-10217050-0001.pdf): "The noise may be resolved by re-torqueing the joints of the leaf spring assembly" — a repair with no parts in it at all. ------------------------------------------------------------------------------- 6. WHAT WE DID NOT DO ------------------------------------------------------------------------------- - No normalization by vehicles on the road; no reliability ranking. - No price re-verification. The repair prices in the article are prior published estimates, labelled as such; estimator sites block automated access and we do not work around that. - No attempt to infer a diagnosis where the owner did not state one. ------------------------------------------------------------------------------ CASE SENSITIVITY — SECTION ADDED 19 SEPTEMBER 2026 ------------------------------------------------------------------------------ NHTSA stores complaint narratives in ALL CAPS for records filed through roughly 2020 and in mixed case for records filed from 2021 on. A case-SENSITIVE search of the narrative field loses about 92% of the complaints filed after 2020, so this is the single largest way an analysis of this database can go wrong while still looking finished. Every query behind this file is case-insensitive. SQLite's LIKE and its FTS5 full-text index both ignore case by default, and no query in this work uses GLOB or any case-sensitive comparison. This was not taken on trust. On 19 September 2026 every published folder was re-checked against the same database, comparing the figures here with what a case-sensitive query would have returned. The result: the complaint set published here (4,441) is LARGER than the absolute ceiling a case-sensitive search of the same terms could have returned (2,508 narratives). A case-sensitive query cannot produce a bigger set than itself, so the bug is ruled out for this folder by arithmetic, not by assurance.