PulsCar analysis of NHTSA public data, September 2026 Article: https://www.pulscar.io/blog/car-shakes-at-highway-speed License: CC BY 4.0 (cite "PulsCar analysis of NHTSA public data, September 2026") SOURCE FILES - NHTSA complaint flat files (FLAT_CMPL), all reports received 1 January 2015 to 8 September 2026, de-duplicated by report number (ODINO): 777,040 unique owner complaints. Field dictionary: https://static.nhtsa.gov/odi/ffdd/cmpl/CMPL.txt - NHTSA manufacturer communications files (FLAT_TSBS 2015-2019, 2020-2024, 2025-2026): one row per make/model/year per document; counted as distinct NHTSA document IDs. - NHTSA recall flat file (FLAT_RCL): one row per make/model/year per campaign; counted as distinct campaigns. - NHTSA investigations flat file (FLAT_INV): the closing summaries quoted in the article are the SUMMARY field. Files were downloaded in September 2026 and loaded into SQLite with an FTS5 full-text index on the complaint description and on the bulletin summary. Counts were run on 18 September 2026. COMPLAINT QUERY (FTS5 syntax, run against the complaint description field) (vibration OR vibrates OR vibrating OR shake OR shakes OR shaking OR shimmy OR wobble OR wobbles) AND ("highway speed" OR "highway speeds" OR "freeway speed" OR "60 mph" OR "65 mph" OR "70 mph" OR "75 mph" OR "80 mph" OR "high speed" OR "high speeds") NOT (airbag OR "air bag" OR crash OR collision OR accident OR "steering wheel locked" OR "lost power" OR stalled OR stall OR misfire OR misfires OR misfiring OR "check engine" OR transmission OR shifting OR shifts OR shifted OR lifter OR lifters OR mirror OR mirrors OR windshield OR "loss of power" OR "engine light" OR "lost acceleration") Result: 3,926 complaints. We then kept only reports whose NHTSA component field names WHEELS, TIRES, SUSPENSION, STEERING, POWER TRAIN, SERVICE BRAKES, AXLE or DRIVELINE: 3,467 complaints ("kept set"). How the query was tightened: a first version without the exclusion list (5,498 matches) was sampled; of 30 random matches read in full, 11 were engine misfires, stalls, transmission shifts, mirror or trim vibration (about 37 percent false positives). After adding the exclusions and the component filter, 30 random matches from the kept set were read: 28 describe a shake or vibration of the wheels, steering, suspension or driveline at highway speed; 2 do not (an engine misfire, a brake-contamination event) - about 7 percent. Subset "wobble plus bump": kept-set reports whose text contains wobble, "death wobble" or shimmy AND a trigger word (bump, pothole, expansion joint, seam, rough road/patch/pavement, imperfection, dip): 881 complaints. Reading sample: 100 random complaints from the remainder (kept set minus the 881) were read in full and each given one category. Result: 51 are still the bump-triggered steering oscillation (described as "shake" or without a trigger word, so the keyword pair missed them); 9 tires or wheels (bulge, tread or sidewall separation, tire vibration, a broken wheel part); 9 a steering component or loose steering (coupler, intermediate shaft, linkage, gear); 10 a shop could not find the cause or no cause is stated; 5 brakes (pulsation, rotor thickness variation); 5 a suspension component (control arm, shocks, a lost bolt); 3 a driveshaft or CV axle; 1 a wheel hub or bearing; 7 not a highway shake at all (engine, transmission, driver-assist braking, a motorcycle). The estimate in the article that "about six in ten" kept complaints are the oscillation pattern is 881 plus 51 percent of 2,586, and is an estimate from this sample. Keyword flags (highway_shake_fix_keywords.csv) are simple regular expressions over the description; they show what the owner MENTIONED, not what fixed the car. FIELDS AND EXCLUSIONS - Mileage statistics use the MILES field, excluding blank, 0 and 400,000 or more. - "crash", "towed", "fire" are the Y/N fields as filed by the owner; "injured>0" counts complaints with INJURED greater than 0. - Complaints are self-reported, unverified, and are not failure rates. Counts by make or model reflect fleet size, publicity (the "death wobble" name, class actions, news coverage) and owner behavior as much as defects. Nobody files a federal complaint about a $25 balance that worked, so the everyday cause is under-represented by construction. We rank nothing by reliability. BULLETIN QUERY (tsbs_fts on the NHTSA summary field) (vibration OR shudder OR shake OR shimmy) AND (highway OR mph OR "high speed" OR "road speed" OR "freeway") Result: 380 distinct documents from 36 makes. Every summary excerpt was read and assigned one category with keyword rules plus hand overrides (highway_vibration_bulletins.csv). Categories: transmission or torque converter shudder (131); tires and wheels - balance, road force, flat spots, runout (44); brake judder (36); body, trim, mirror, seat or wind (38); driveline - driveshaft, half shaft, CV, U-joint, transfer case, axle (29); steering system - damper, gear, power-steering software (10); engine or PCM software (10); wheel bearing (3); noise only, low-speed judder, chassis resonance or matched the words without being about a shake (30); unclear from the summary (49). Twelve documents were opened as PDFs from static.nhtsa.gov and quoted in the article (Ford 20N04-S1; Jeep 19-002-24 and 23-041-24; Jeep STAR S1819000004 Rev. R; Hyundai 22-ST-006H; GM 19-NA-240 and PIT6154A; Mazda SA-013/24; Volvo TJ 36724.2.0; Rivian RSB-10-23-001-1). Documents that describe a highway shake without using the words in the query are not counted. The NHTSA summary is often a 1-2 sentence excerpt, so "software vs hardware" was only counted where the excerpt says so. RECALLS AND INVESTIGATIONS - driveshaft_and_wheel_fastener_recalls.csv: campaigns reported 2011 or later whose defect text names a driveshaft, drive shaft, propeller shaft or half shaft, OR a lug nut, wheel nut, wheel bolt or wheel stud, restricted to 38 light-vehicle makes (trailers, motorcycles, heavy trucks and aftermarket parts excluded by make; models containing TRAILER excluded). Driveshaft: 112 campaigns, 3,837,951 vehicles as estimated by the manufacturers at filing. Wheel fasteners: 26 campaigns, 653,586 vehicles. Not every driveshaft recall describes a vibration; many describe a fracture or separation with no warning. - Investigations quoted: PE07-057 / EA08-007 (Ford F-250/F-350 2005-2007, "violent front end oscillation"), PE19-017 (Ram 4500/5500 front driveshaft), PE15-020 (Ford Edge wheel separation). Text is the closing summary from the investigations file. FILES - highway_shake_complaints_summary.csv: kept set, wobble-plus-bump subset, remainder and the all-complaints baseline, with mileage quartiles, crash/towed/fire/injury counts and top models. - highway_shake_complaints_by_model.csv: the 3,467 kept complaints by make, model and model year, with the wobble-plus-bump count per row. - highway_shake_fix_keywords.csv: keyword-group counts across the three sets. - highway_shake_reading_sample.csv: the categories assigned by reading 100 random remainder complaints. - highway_vibration_bulletins.csv: the 380 bulletins with NHTSA ID, bulletin number, date, years, models, our category and a 160-character summary excerpt. - driveshaft_and_wheel_fastener_recalls.csv: the 138 recall campaigns described above. No VINs, names, cities, dealers or free text beyond short excerpts are included. LIMITATIONS Keyword queries miss owners who describe a shake without the words we searched and catch some who use them for something else; the false-positive rates above are from reading, not from a model. The federal file is a safety-complaint file: it over-represents frightening events (steering oscillation on solid-axle trucks) and under-represents the ordinary imbalance that a tire shop fixes in half an hour. NHTSA's bulletin index attaches some documents to many models they do not cover. Populations in recalls are manufacturer estimates at filing. ------------------------------------------------------------------------------ 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, in two layers: * the complaint set is built with an FTS5 MATCH full-text search, which folds case before matching, so it cannot miss a mixed-case record; and * where the narrative is then tested with patterns, the narrative is upper-cased and whitespace-normalized first, so the patterns compare upper against upper. 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. Re-run here: step 1 is FTS5 and the keyword flags are Python regular expressions applied to an upper-cased narrative.