PulsCar analysis of NHTSA public data, September 2026 Topic: engine oil leaks - owner complaints, recall campaigns and manufacturer bulletins 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 files RECALLS (FLAT_RCL), MANUFACTURER COMMUNICATIONS (FLAT_TSBS, https://static.nhtsa.gov/odi/ffdd/tsbs/TSBS.txt) and INVESTIGATIONS (FLAT_INV). - Downloaded and indexed 17 September 2026; queries run 18 September 2026. QUERY - "a complaint that says engine oil was leaking" Step 1. Full-text search for oil AND (leak OR leaks OR leaking OR leaked OR seep OR seeping OR seeps OR dripping OR drips). -> 10,813 complaints. Step 2. Motorcycle and powersports makes removed. -> 10,771. Step 3. Sentence-level phrase test. A sentence must contain an explicit engine-oil-leak phrase - "oil leaking/seeping/dripping/pouring", "leaking oil", "oil leak", "oil leakage", "oil seepage", "dripping oil" - and that phrase must NOT be preceded within 30 characters by a negation ("no sign of", "not", "never", "without", "n't"). The same sentence must not mention another fluid or system: transmission, coolant, antifreeze, power steering, brake fluid, differential, transfer case, washer fluid, fuel or gasoline, radiator, water pump, gear oil, axle, hydraulic, refrigerant, A/C compressor, air conditioning or vacuum pump. Step 4. Oil-consumption filter. A narrative whose only leak phrase sits beside consumption language ("oil consumption", "burning oil", "adding a quart") and which names no leak address is dropped - burning oil is not leaking oil, and the article says so in its own text. Step 5. Subject test. The leak phrase must appear within the first 700 characters of the narrative, or appear at least twice. This removes complaints about brakes, stalling or headrests that mention a leak in passing at the end. -> 3,942 complaints. This is the published set, received 2 January 2015 to 7 September 2026. SAMPLE READING (required before publishing any keyword count) Three rounds of 30 random matches were read as the query was tightened. At the loose stage roughly half the sample was off-topic (oil consumption, engine failures, other fluids), which is why steps 3-5 exist. In the final sample of 30, about five described something else - a brake complaint, a no-crank, a P0420, a reduced-power event and a service visit that mentioned an oil change - so the observed false-positive rate is about 15-20% in a sample of 30. Counts in the article are described as "complaints whose narrative says engine oil was leaking", never as a count of confirmed leaks, and never as a failure rate. FIGURES QUOTED IN THE ARTICLE (published set, n = 3,942) Fire 4.59% (181 reports) against a 1.82% baseline over all 777,040 complaints. Towed 6.52% against 5.11%. Crash 0.68% against 4.37%. 23 injuries, no deaths. Mileage: median 83,000; quartiles 52,000 and 116,378 (n = 2,185 with a usable odometer reading; miles <= 0 or >= 400,000 dropped). Baseline median over all complaints with a usable reading: 70,000. 554 complaints (14.05%) mention hot exhaust, a burning smell, smoke, flames or a fire hazard. 52 complaints (1.32%) say a shop could not find or locate the leak, called it normal seepage, or said it was within specification. Owner-quoted dollar figures: every "$N" between $50 and $30,000 in a narrative was collected - 658 figures, median $2,000, quartiles $844 and $4,854, 66.3% above $1,000 (file oil-leak-owner-reported-amounts.csv). These are NOT prices. They are amounts owners chose to type into a federal safety complaint, they often include an engine replacement or a tow rather than a gasket, and cheap outcomes are almost never written down. The article presents them as the expensive tail and says so. WHICH ADDRESS OWNERS NAME (file: oil-leak-address-named.csv) 1,664 of the 3,942 complaints (42.2%) name at least one leak address. Counts overlap, because a narrative can name several. Oil filter, its housing or the oil cooler 598 (15.2%); oil pan gasket or drain plug 331 (8.4%); head gasket or internal engine 324 (8.2%); valve cover or cam cover gasket 248 (6.3%); timing cover, front crank or cam seal 154 (3.9%); rear main seal 147 (3.7%); oil pressure sensor or switch 57 (1.4%); turbocharger oil lines 33 (0.8%). This is a ranking of what owners write in a federal complaint, not a ranking of how often each gasket fails. It is skewed toward parts that fail suddenly, make a mess, or belong to a publicized cluster, and away from the slow valve-cover seep that is the most common repair in the trade. The article says this in the text beside the table. MODEL CLUSTERS QUOTED Jeep Wrangler 200 complaints, 127 of them model year 2014, median 54,000 miles (n = 147), 150 of the 200 naming the oil filter housing or oil cooler, 10 fire reports. Mazda CX-5 144 (63 of model year 2018 and 64 of 2019). Chevrolet Equinox 130. Hyundai Sonata 111. Kia Sorento 116. Ford Escape 105 with 10 fire reports. Hyundai Veracruz 83 at a median of 133,500 miles. RECALLS (file: oil-leak-recall-campaigns.csv) Recall campaigns whose defect or consequence text contains "oil leak" or says engine oil may leak: 91 distinct campaigns. Campaigns whose text is about transmission, power steering, gear or hydraulic oil were removed, leaving 78 engine-oil campaigns covering 1,145,544 potentially affected vehicles. 53 of the 78 (68%) name a fire in the consequence field. Campaign text was re-read from NHTSA's live campaign API on 18 September 2026. BULLETINS (file: oil-leak-bulletins-by-address.csv) Communication summaries containing "oil leak", "oil leakage", "oil seepage" or "leaking oil": 1,110 distinct communications; 914 remain after removing those whose summary is about transmission, differential, transfer case, power steering, gear or hydraulic oil. Address counts within that subset, from the summary text only: head gasket or internal engine 95, oil pan or drain plug 92, oil filter housing or cooler 78, valve cover 52, timing cover or front seal 39, rear main 27, turbocharger oil lines 26, oil pressure switch 10. Summaries are NHTSA's abstracts rather than full bulletin text, so every count is a floor. LIMITATIONS - Complaints are unverified owner reports, not failure rates, and are not normalized by the number of vehicles on the road. Nothing here ranks manufacturers by reliability, and no such ranking can be derived from counts. - A leak with a publicized cluster or an open recall attracts complaints from owners who have read about it. - Mileage is owner-entered, often rounded, and missing from 45% of this set. - Federal complaints are filed about safety, so a slow seep that costs $300 to fix is heavily under-represented compared with a cracked housing that empties the sump on a highway. Read every count here as a picture of the alarming end of this category, not of the ordinary end. ------------------------------------------------------------------------------ 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 (10,813) is LARGER than the absolute ceiling a case-sensitive search of the same terms could have returned (1,953 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.