PulsCar analysis of NHTSA public data, September 2026 Noise when turning: what 7,002 owner complaints and 1,052 manufacturer bulletins say Licensed CC BY 4.0. Queries run 18 September 2026. -------------------------------------------------------------------------- SOURCE FILES -------------------------------------------------------------------------- NHTSA ODI complaint flat files COMPLAINTS_RECEIVED_2015-2019, _2020-2024, _2025-2026 -- https://static.nhtsa.gov/odi/ffdd/cmpl/ (dictionary CMPL.txt) 777,040 distinct vehicle complaints (ODINO, product type V) received 1 January 2015 to 8 September 2026. NHTSA manufacturer communication flat files TSBS_RECEIVED_2015-2019, _2020-2024, _2025-2026 -- https://static.nhtsa.gov/odi/ffdd/tsbs/ NHTSA recall flat file FLAT_RCL.zip and investigation flat file FLAT_INV.zip. -------------------------------------------------------------------------- CASE SENSITIVITY -- READ BEFORE REUSING ANY QUERY -------------------------------------------------------------------------- NHTSA complaint narratives are ALL CAPS through the end of 2020 and mixed case from 2021 onward. Measured on the file: of 58,465 complaints received in 2020, 58,386 are identical to their own uppercase form; of 62,775 received in 2023, only 939 are. An uppercase-only Python regular expression, or an uppercase SQLite GLOB, therefore silently drops most complaints filed after 2020 while still returning thousands of rows -- a failure that looks like a result. Every query behind this page is case-insensitive: Python regexes compiled with re.IGNORECASE, SQL text matching with LIKE over lower(), which is case-insensitive in SQLite for ASCII. -------------------------------------------------------------------------- WHICH COMPLAINTS COUNT (the turning-noise set: 7,002) -------------------------------------------------------------------------- Stage 1 -- SQL prefilter. Complaints whose narrative contains "turn" or "corner" AND at least one of: nois, click, clunk, grind, whin, creak, squeak, squeal, knock, pop, groan, moan, growl, rattl, hum. 20,862 complaints. Stage 2 -- sentence test. Split each narrative at . ; ! ? and keep the complaint if ANY single sentence contains a NOISE word within 70 characters of a TURNING word, in either order: NOISE: click* | pop/pops/popping/popped | whin* | moan* | groan* | clunk* | knock* | grind* | growl* | hum/hums/humming/hummed | creak* | squeak* | squeal* | rattl* | thud* | thump* | nois* | grunt* | chirp* | whirr* | screech* TURN: \bturn(s|ing|ed)?\b | cornering | corner(s) | steering wheel | full lock | u-turn | parking lot | roundabout | curve Stage 3 -- non-steering uses of "turn" are blanked out BEFORE the test, so they cannot anchor a match: re-turn / return* (the word "returning" contains "turning" -- this alone was pulling in narratives about noises that had nothing to do with steering), turn on/off/up/down/over/around/into/out/away, turn the key / ignition / car / vehicle / engine / radio / volume / a-c / heat / air / lights / blinker / hazard, turn signal, overturn*, turnpike, turnover. Stage 4 -- drop the sentence on: turn signal, blinker, "no noise", "without any noise", "didn't make/hear any noise". Result: 7,002 complaints. Word boundaries matter here in both directions. Without \b on "hum", the pattern matched the word "human"; without \b on "turn", it matched "returning". Both were found by reading samples, not by inspecting the query. -------------------------------------------------------------------------- SAMPLE READING (the false-positive rate we found) -------------------------------------------------------------------------- Round 1: 30 random complaints read in full from an earlier, looser version of the query (7,447 rows). Four were wrong: a grinding transmission in a narrative that happened to say "turn around"; a whirring noise found via "returning"; a stereo screech with the "volume turned down"; and "the loud noise then turned into a dragging noise". Those four produced stage 3. Round 2: 30 random complaints read in full from the tightened set. Two were wrong -- a seat-slide complaint caught by "human" inside the hum pattern, and a door that "popped open" during a right turn. The hum boundary was then fixed. That is a measured false-positive rate of about 1 in 30, roughly 3%, and it should be read as a floor rather than a ceiling: judging whether a noise is "tied to turning" from one sentence is a judgement call, and a reader stricter than us would exclude more. -------------------------------------------------------------------------- STRUCTURED FIELDS, AND THE BASELINE THEY ARE COMPARED WITH -------------------------------------------------------------------------- All 777,040 complaints: towed 5.11%, crash 4.37%, fire 1.82%, median mileage 70,000 (392,935 rows with usable mileage). The 7,002 turning-noise complaints: towed 1.86%, crash 1.87%, fire 0.36%, 87 injuries, 1 death, median mileage 70,000 (4,044 rows with usable mileage, first quartile 41,000, third quartile 105,000). Mileage excludes rows at or below 0 and at or above 400,000. 2,958 of the 7,002 carry no usable mileage and are absent from mileage figures rather than counted as zero. NHTSA's own component field on these complaints (a complaint can carry several): STEERING 60.74%, SUSPENSION 17.85%, POWER TRAIN 12.43%, ELECTRICAL SYSTEM 9.78%, UNKNOWN OR OTHER 9.31%. -------------------------------------------------------------------------- SOUND TYPE (turning-noise-by-sound-type.csv) -------------------------------------------------------------------------- Counted by searching the whole narrative, case-insensitively, for each family of sound words. Subsets OVERLAP: a narrative that says "clicking and clunking" is counted in both rows. The last row is the complement -- narratives with no sound word at all beyond "noise" or "sound". clicking or popping 2,889 41.26% median 70,121 mi clunking or knocking 1,593 22.75% median 70,000 mi no sound type given 1,429 20.41% median 74,000 mi grinding or growling 827 11.81% median 65,000 mi creaking or squeaking 420 6.00% median 60,000 mi rattling 308 4.40% median 75,000 mi whining, moaning or groaning 231 3.30% median 71,000 mi squealing or screeching 149 2.13% median 64,000 mi humming 45 0.64% median 79,225 mi SELECTION BIAS, STATED PLAINLY. These are shares of COMPLAINTS TO A SAFETY REGULATOR, not shares of turning noises in the world. A power-steering whine that a half-litre of fluid silenced never becomes a federal complaint; a click that a dealer refused to fix under warranty very often does. The low whine share (3.30%) is evidence about the complaint file, not evidence that whines are rare. -------------------------------------------------------------------------- PARTS NAMED (turning-noise-parts-owners-name.csv) -------------------------------------------------------------------------- Counts of complaints whose narrative names each part ANYWHERE in the text, case-insensitively, overlapping. A mention is not a diagnosis: owners are often repeating what a shop told them, and many narratives name a part that was replaced without fixing the noise. steering column / intermediate shaft / coupler / U-joint 781 11.15% steering rack / steering gear / pinion 704 10.05% CV joint / axle / half shaft / drive shaft 629 8.98% control arm or bushing 320 4.57% transfer case / differential / 4WD or AWD / binding 260 3.71% strut, strut mount or shock 202 2.88% power steering pump, fluid, hose or reservoir 160 2.29% electric power steering motor or EPS unit 157 2.24% wheel bearing or hub 134 1.91% ball joint 133 1.90% tie rod 107 1.53% sway bar / stabilizer link or bushing 74 1.06% brake caliper, pad, rotor or dust shield 53 0.76% tire rubbing or tire noise 34 0.49% 247 complaints (3.53%) say a shop could not find, duplicate or diagnose the noise, or called it normal. -------------------------------------------------------------------------- MANUFACTURER BULLETINS -------------------------------------------------------------------------- Manufacturer communications whose summary contains a turning word and a noise word, passed through the same proximity test (90 characters) and the same non-steering "turn" exclusions, counted DISTINCT by NHTSA record id: 1,052. Reading ten at random found one clear miss (a Ford audio-system bulletin), so treat this count as carrying a few percent of noise as well. mentions a suspension part (strut, sway bar, ball joint, control arm, bushing) 116 11.03% mentions the steering column or intermediate shaft 111 10.55% mentions a CV joint, axle or drive shaft 105 9.98% mentions software, reprogramming or calibration 90 8.56% the repair is grease or lubricant 56 5.32% mentions the steering rack or gear 32 3.04% says the noise is normal or that no repair is needed 41 3.90% The FCA "Front Click Noise at Low Speed Turns" family (bulletin numbers 05-002-20, 05-004-21, 0500122, 0500220, 0500222, 0500421) appears as 12 distinct NHTSA records across 104 make/model/year rows, covering 2017-2022 Dodge Charger and Challenger, Jeep Grand Cherokee and Cherokee, and Dodge Durango. -------------------------------------------------------------------------- LIMITATIONS -------------------------------------------------------------------------- 1. Complaints are self-reported and unverified; they record what an owner wrote, not what a technician found. 2. There is no denominator. Nothing here says any model fails more often than another; a model with many complaints may simply be common, or may have a recall letter driving owners to the complaint form. 3. Sound-type shares describe the complaint file, not the world (see the selection-bias note above). 4. Part mentions describe vocabulary, not causes. 5. Mileage is owner-entered and missing in 2,958 of 7,002 rows. 6. The complaint file ends 8 September 2026; young model years are under-represented because their cars are young. Contact: support@pulscar.io