PulsCar analysis of NHTSA public data, September 2026 Topic: Hyundai and Kia Theta II engine knocking / connecting-rod bearing failure License: CC BY 4.0. Attribution: "PulsCar analysis of NHTSA public data, September 2026". SOURCE FILES 1. NHTSA Office of Defects Investigation public flat files, downloaded 17 September 2026 and loaded into a local SQLite database: - COMPLAINTS: 777,040 owner complaints received 1 January 2015 through 8 September 2026. - RECALLS: one row per make/model/model-year per campaign (campaigns counted DISTINCT). - MANUFACTURER COMMUNICATIONS (TSBs, warranty bulletins, product-improvement campaign documents, owner letters): one row per make/model/model-year per document; documents counted DISTINCT by NHTSA ID. 2. Manufacturer documents opened as PDFs from static.nhtsa.gov (see DOCUMENTS OPENED). CASE SENSITIVITY (important) NHTSA complaint narratives are ALL CAPS for filings up to 2020 and mixed case from 2021 onward. Every text query below is case-insensitive: Python regexes carry re.IGNORECASE and SQL text matching uses LIKE (case-insensitive in SQLite). An uppercase-only regex, or an uppercase SQLite GLOB, would silently drop most post-2020 complaints. THE COMPLAINT SET Universe: complaints whose make/model is one of the six nameplates named as Class Vehicles in the Hyundai/Kia Theta engine settlement -- Hyundai Sonata, Hyundai Santa Fe Sport, Hyundai Tucson, Kia Optima, Kia Sorento, Kia Sportage -- with model year 2011 through 2019. That universe is 29,981 complaints. First pass (case-insensitive regex on the narrative): knock|rod bearing|bearing knock|rod knock|engine seiz|engine fail|thrown a rod|threw a rod| blew the engine|engine blew|blown engine|motor blew|seized engine -> 5,265 complaints. Sample reading: 30 random matches read in full. Two were clear false positives -- one knock heard while turning out of a driveway and over speed bumps (a suspension noise) and one "loud knocking sound in the steering location". Measured false-positive rate 2/30 = 6.7%. Tightening: a match is kept only if an engine-context word engine|motor|rod|bearing|piston|cylinder|crank|block|valve|internal appears within 90 characters of the matched keyword. -> 4,649 complaints. This is the published set. Second sample reading: 30 fresh random matches read in full. One clear false positive (an airbag complaint that mentioned a previous engine failure in passing) and one ambiguous (a failure to restart). Measured false-positive rate 1/30 = 3.3%, with one ambiguous case. EXCLUSIONS - Mileage statistics use only complaints with a mileage value greater than 0 and less than 400,000 (2,209 of the 4,649 carry a usable mileage). Median 97,000; first quartile 71,000; third quartile 122,000; 26.0% above 120,000 miles. - Crash/fire/tow/injury/death figures come from NHTSA's structured fields, not from the text. - The all-complaint baselines quoted for comparison are computed over the whole 777,040-row file: towed 5.11%, crash 4.37%, fire 1.82%. OWNER-REPORTED DOLLAR FIGURES Within the 4,649-complaint set, a dollar amount was kept only when: - the figure is between $500 and $30,000; - the words engine|motor|short block|long block appear within 80 characters of it; - a repair word (replac|repair|rebuild|quote|estimat|cost|charge|paid|bill) appears within the same 80-character window; - and the window does NOT contain bought|purchase|loan|financ|still owe|trade in|blue book| worth|down payment|rental|transmission|alternator|head gasket|converter|body control. -> 242 figures. Median $6,700; first quartile $4,800; third quartile $9,000; 78.5% at or above $4,000. Twenty of these were read in context; 19 were genuine engine repair or replacement money and one was a wiring-harness quote, so roughly 1 in 20 is mis-attributed. These are owner reports typed into a complaint form. They are not invoices, they are not a random sample of repairs, and they skew toward people angry enough to file. They are published as a distribution of what owners SAID they were quoted or charged, nothing more. MANUFACTURER COMMUNICATIONS Query: make in (HYUNDAI, KIA) AND summary LIKE any of '%connecting rod%', '%rod bearing%', '%knock sensor%', '%ksds%' (LIKE is case-insensitive). -> 4,247 rows collapsing to 268 distinct NHTSA document IDs, dated 14 October 2016 to 1 May 2026. Kia 213, Hyundai 55. 234 of 268 (87.3%) mention software, an ECU logic change or KSDS. 32 of 268 mention a warranty extension. RECALLS Query: make in (HYUNDAI, KIA) AND (defect LIKE '%connecting rod%' OR defect LIKE '%rod bearing%' OR remedy LIKE '%knock sensor%' OR defect LIKE '%bearing wear%'), counted DISTINCT by campaign. DOCUMENTS OPENED (PDF, read as text) - Hyundai TSB 22-EM-001H, January 2022, "Engine Warranty Extension (TXXI)": https://static.nhtsa.gov/odi/tsbs/2022/MC-10207811-0001.pdf - Hyundai TSB 22-EM-001H-1, April 2022 (supersedes the above): https://static.nhtsa.gov/odi/tsbs/2022/MC-10211673-0001.pdf - Hyundai "Engine Warranty Extension (TXXI) - Dealer Best Practice", 19 January 2022: https://static.nhtsa.gov/odi/tsbs/2022/MC-10207801-0001.pdf Kia warranty bulletins WTY004 (Optima), WTY007 (Sportage) and WTY008 (Sorento) are quoted from their entries in the manufacturer communications flat file (NHTSA IDs 10094572, 10090095, 10090117); static.nhtsa.gov returned 404 for the PDF of each, and no workaround was attempted. LIMITATIONS - Complaints are self-reported and unverified. Counts are NOT failure rates: there is no denominator of vehicles on the road in this file, so nothing here says one model is more or less reliable than another. - The complaint file starts on 1 January 2015. Filings made before that date about 2011-2014 cars are not in it, so early Theta II reports are under-represented by construction. - The flat file's model/year tagging on a manufacturer communication is broader than the document's own text in at least one case checked here: the rows for 22-EM-001H list Tucson 2016 and 2017, while the bulletin text names only 2014-2015 (LM) and 2018-2019 (TL) Tucsons. Where the two disagree, this analysis follows the document. - Keyword presence measures what owners wrote, not what was wrong with the car. FILES theta-knock-complaints-by-model.csv counts, mileage quartiles, tow/fire/denial shares theta-knock-by-model-year.csv counts and median mileage per make/model/model year theta-knock-context-shares.csv subset counts with the all-complaint baseline theta-rod-bearing-bulletins.csv 268 manufacturer communications theta-owner-reported-engine-amounts.csv 242 owner-reported dollar figures