PulsCar analysis of NHTSA public data, September 2026 Topic: hesitation, lag, stumble and sluggish response while accelerating License: CC BY 4.0 (attribution: "PulsCar analysis of NHTSA public data, September 2026") SOURCES - NHTSA Office of Defects Investigation flat file COMPLAINTS (FLAT_CMPL), field dictionary https://static.nhtsa.gov/odi/ffdd/cmpl/CMPL.txt 777,040 owner complaints received 1 January 2015 - 8 September 2026. - NHTSA ODI flat file MANUFACTURER COMMUNICATIONS (FLAT_TSBS), field dictionary https://static.nhtsa.gov/odi/ffdd/tsbs/TSBS.txt - NHTSA ODI flat file RECALLS (FLAT_RCL) and INVESTIGATIONS (FLAT_INV). - Downloaded and indexed 17 September 2026; queries run 18 September 2026. WHY THIS SET IS NOT THE SET BEHIND OUR JERKING PAGE /blog/car-jerking-when-accelerating analyses complaints whose narrative describes a JERK, buck, surge or shudder during acceleration (10,601 complaints). This page analyses the opposite feel: the car answers LATE. The two queries use different symptom words and the two sets overlap where an owner used both kinds of word in one narrative. 4,490 of the 6,318 complaints published here (71.1%) contain no jerk, buck, surge, shudder, judder or lurch word anywhere in the narrative - those are pure lag reports. Neither number contradicts the other; they are counts from two different queries, both stated in full here and on the jerking page. QUERY - "a hesitation while accelerating" Step 1. Full-text search of complaint narratives (FTS5) for hesitate, hesitates, hesitating, hesitation, stumble, stumbles, stumbling, sluggish, bogs, bogged, bogging, lag, lags, lagging, flat, delayed, unresponsive. -> 19,405 complaints. Step 2. The symptom must match a stricter regular expression - hesitat*, stumbl*, sluggish, "bog(s/ged/ging) down", "flat spot", lag/lags/lagging/laggy, "delayed response|reaction|acceleration", unresponsive, "slow to respond|accelerate" - and must NOT be preceded within 40 characters by no / not / never / without / n't / any / didn / doesn / dont. Step 3. An acceleration phrase (accelerat*, gas pedal, accelerator, throttle, "from a stop", "take off", uphill, merging, passing, "stepped on the gas", "pressed the pedal", "pulling out", "floored it", on-ramp) must appear within 400 characters of the symptom word. -> 7,376. Step 4. Exclusions: motorcycle and powersports makes; narratives containing unintended-acceleration phrases ("sudden acceleration", "accelerated on its own"), which are a different defect and a different NHTSA component. -> 7,257. Step 5. Sentence-level test: the symptom word and either an acceleration phrase or a powertrain word (transmission, engine, gear, shift, RPM, clutch, torque converter, CVT, misfire, spark plug, coil, fuel, throttle, injector, MAF, vacuum, catalytic) must appear in the SAME sentence, and that sentence must not be about steering, glass, airbags, seat belts, door latches, paint, rust, lighting, infotainment, the instrument cluster, a speedometer, a display, or a recall date-of-manufacture range. -> 6,318 complaints. This is the published set. SAMPLE READING (required before publishing any keyword count) 30 random matches were read at the first pass. Three were not hesitation reports (an electrical no-start, an instrument-cluster refresh lag, a recall build-date dispute), about 10%. The instrument-cluster and date-of-manufacture exclusions in step 5 were added because of that reading. 30 fresh random matches were then read from the tightened set. Twenty-six described a hesitation, lag, stumble or sluggish response while accelerating. Four did not: they described a complete stall or a loss of propulsion rather than a late response. Observed false-positive rate for the published set: about 13% in a sample of 30 (4/30); combined across both readings, 7 of 60. Counts in the article are therefore described as "complaints whose narrative describes a hesitation in an acceleration context", never as a count of confirmed defects. FIGURES QUOTED IN THE ARTICLE (published set, n = 6,318) Towed 2.12% (134), crash 1.30% (82), fire 0.36% (23). Baseline over all 777,040 complaints: towed 5.11%, crash 4.37%, fire 1.82%. 42 complaints report at least one injury (47 injuries in total) and one reports a death. Mileage: median 38,000; first quartile 8,600; third quartile 80,000 (n = 3,261 complaints with a usable odometer reading; rows with miles <= 0 or >= 400,000 dropped, so 48.4% of the set carries no usable mileage). 49.06% of those readings are at or below 36,000 miles. Baseline median mileage over all complaints with a usable reading (n = 392,935): 70,000. "Shop could not duplicate it / no problem found / operating as designed": 330 complaints, 5.22%, counted by the phrase list in the code below. The same phrase list over all 777,040 complaints matches 4,605, or 0.59%. Timing phrases (file hesitation-timing-phrases.csv): "from a stop / low speed" 2,079 (32.91%); "passing / merging / uphill / wide open" 1,365 (21.60%); "cold start / until it warms up" 173 (2.74%); "after a fill-up / low tank" 72 (1.14%); "with a rough idle" 57 (0.90%); no timing phrase of any kind 3,128 (49.51%). Groups overlap, so they do not sum to 100. These are keyword shares of what owners chose to write, not a measured distribution of when cars hesitate; read them as floors. Lag-only subset (no jerk/buck/surge/shudder/judder/lurch word): 4,490 complaints, towed 2.36%, median 40,250 miles (n = 2,310). NHTSA component tags: power train 3,484; engine 2,353; fuel/propulsion 1,135; vehicle speed control 956. A complaint can carry more than one tag. Model clusters quoted: Ford Focus 449 (2012-2017, median 30,000 miles, 54.3% at or under 36,000); Toyota Tundra 306 (112 of them 2022, 74 2024, 67 2023); Hyundai Tucson 274 (115 model year 2016 and 106 model year 2017, median 7,000 miles, 91.1% at or under 36,000, n = 168 with mileage); Jeep Cherokee 212 (2014-2015 dominant, median 19,000 miles); Ford F-150 212; Toyota Camry 191. WHAT OWNERS SAID WAS REPAIRED (file: hesitation-what-owners-said-was-repaired.csv) Text captured after "diagnosed with/as", "determined that", "found that", "turned out to be", "replaced the", "replacing the", "needed a new", "needed to be replaced", "repaired the", "dealer replaced", "they replaced", "was replaced", "ended up being", "fixed by/with" and matched against part keywords within the following 260 characters. 708 of the 6,318 complaints (11.2%) contain such a phrase; that is the denominator. Categories overlap, so shares do not sum to 100. IMPORTANT: this is NOT a frequency ranking of the causes of hesitation. It is a ranking of what gets written into a federal safety complaint, and cheap fixes are massively under-reported, because almost nobody files a federal complaint about a $15 vacuum hose or an $8 can of sensor cleaner. The article says so in the text beside the table. BULLETINS (file: hesitation-bulletins-by-make.csv) Communication summaries searched for hesitat*, stumble(s)(ing) and sluggish: 616 distinct communications (NHTSA_ID). Of those, 338 tie the symptom to acceleration, throttle, tip-in, launch, drive-away or shifting - the published set, dated 14 February 2015 to 23 July 2026. 177 of the 338 (52.4%) mention software, reprogramming, recalibration or a flash. Summaries are NHTSA's published abstracts, not the full bulletin text, so the software share is a floor rather than an exact measurement. Counts are per manufacturer as the flat file spells the make, and a single bulletin filed against several makes (for example Stellantis brands) is counted once per make. RECALLS Recall campaigns whose defect or consequence text contains hesitate, hesitation, stumble or sluggish: 12 distinct campaigns across the whole recall file, most of them commercial or powersports vehicles. The three passenger-car campaigns quoted in the article are 16V-563 (Mitsubishi, 82,436 vehicles), 16V-628 (Hyundai, 41,000) and 14V-495 with its supplement 15V-813 (Ford, 133,227 and 9,017). Campaign text was re-read from NHTSA's live campaign API on 18 September 2026 and all three links returned HTTP 200. LIMITATIONS - Complaints are unverified owner reports, not failure rates, and are not normalized by the number of vehicles on the road. High-volume models produce high counts, and nothing here ranks manufacturers by reliability. No such ranking can be derived from these counts. - A model with a well-publicized transmission problem attracts complaints from owners who have read about it, which inflates its count relative to quieter faults. - Mileage is owner-entered and often rounded to the nearest 5,000 or 10,000, and is missing altogether in 48.4% of this set. - Hesitation is a feel, not a measurement. Two owners can use the same word for two different sensations, and the query cannot separate them. ------------------------------------------------------------------------------ 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 (19,405) is LARGER than the absolute ceiling a case-sensitive search of the same terms could have returned (5,790 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.