PULSCAR ANALYSIS OF NHTSA PUBLIC DATA — STEERING RACKS AND STEERING GEARS PulsCar analysis of NHTSA public data, September 2026. License: CC BY 4.0. Files in this folder may be reused with attribution to PulsCar (pulscar.io). -------------------------------------------------------------------------- 1. SOURCE FILES -------------------------------------------------------------------------- NHTSA public flat files downloaded from static.nhtsa.gov and loaded into a local SQLite database: Complaints FLAT_CMPL.txt 777,040 owner complaints (ODINO), received 1 January 2015 - 8 September 2026. The set below spans the whole of that window. Field definitions: https://static.nhtsa.gov/odi/ffdd/cmpl/CMPL.txt Recalls FLAT_RCL.txt one row per make/model/model year per campaign; campaigns counted DISTINCT on the campaign number and the affected population taken as the campaign maximum. Investigations FLAT_INV.txt ODI preliminary evaluations, engineering analyses, recall queries and defect petitions. Complaints are owner reports. They are not verified repairs, not failure rates, and they carry no denominator of vehicles on the road. A model with more complaints may simply have sold more units, be older, or be the subject of a recall or investigation that told owners where to file. Nothing here ranks reliability. -------------------------------------------------------------------------- 2. THE COMPLAINT QUERY (7,014 -> 6,066 -> 6,054) -------------------------------------------------------------------------- Step 1 - full-text search (FTS5) of the complaint narrative for any of: "rack and pinion" | "steering rack" | "rack assembly" | "steering gear" -> 7,014 complaints. Step 2 - failure proximity. At least one of those phrases must sit within 100 characters of a failure, wear or repair word: fail | leak | bad | worn | wear | replac | defect | play | loose | clunk | knock | noise | whin | seiz | stiff | hard to turn | lock up | broke | crack | stuck | bind | slop | recall | went out | gone | damag | contaminat -> 6,066 complaints. Step 3 - passenger vehicles only. 12 complaints filed under motorcycle, motorhome, transit-bus and heavy-truck makes (Freightliner, Suzuki, Nova Bus, Newmar, Can-Am, Jayco) were removed. -> 6,054 complaints, filed 1 January 2015 - 8 September 2026. -------------------------------------------------------------------------- 3. SAMPLE READING (30 narratives, seed 20260918) -------------------------------------------------------------------------- Thirty complaints were drawn at random and read in full before any count was published. 29 of 30 genuine steering rack, steering gear or power steering assist failures, including diagnoses made by a dealer 1 of 30 the rack appeared as collateral damage in a different story (a Subaru head gasket failure that also took out the radiator, the rack and two catalytic converters) False-positive rate: 1 of 30 (3.3%). This is a much cleaner set than most symptom searches, because "rack and pinion" and "steering gear" are part names that owners rarely use loosely. No further exclusion was applied. -------------------------------------------------------------------------- 4. WHAT THE NARRATIVES DESCRIBE -------------------------------------------------------------------------- Regex families, all case-insensitive, all documented in rack-complaints-by-context.csv. The two that matter most: Loss of assist / heavy steering loss of (power )?steering | lost (power )?steering | power steering (failed|failure|stopped|went out|inoperab|not work) | steering assist fault | no power steering | hard to (turn|steer) | difficult to (turn|steer|maneuver|control) | excessive force to steer -> 2,145 of 6,054 (35.4%) Any mention of a leak or power steering fluid leak | seep | drip | fluid (low|loss|puddle) | puddle | power steering fluid -> 564 of 6,054 (9.3%) That ratio is the single most important number in this folder. The modern steering-rack complaint is an assist failure, not a hydraulic leak. -------------------------------------------------------------------------- 5. MILEAGE -------------------------------------------------------------------------- Reported mileage is used only where it is above 0 and below 400,000. 3,671 of the 6,054 complaints carry a usable mileage. All complaints median 76,000 Q1 49,000 Q3 110,000 At or under 36,000 mi 521 of 3,671 At or under 60,000 mi 1,329 of 3,671 This is not a rack's service life. Complaints record failures, never survivals. -------------------------------------------------------------------------- 6. OWNER-REPORTED DOLLAR FIGURES (1,294 figures, 1,099 complaints) -------------------------------------------------------------------------- Every dollar amount between $100 and $30,000 in the narratives was extracted, then kept only where the word rack, steering, pinion or gear appears within 150 characters of the amount. 1,294 figures survive, from 1,099 distinct complaints. Median $2,094 Q1 $1,600 Q3 $3,000 90.0% are $1,000 or more; 58.8% are $2,000 or more; 27.3% are $3,000 or more Read these as the shape of the bill, not as a price list. Owners write quotes, paid invoices, insurance figures and totals for multi-part repairs in the same sentence, and the regex cannot tell them apart. The distribution is also tilted upward: people are more likely to write down a number that shocked them. -------------------------------------------------------------------------- 7. THE FATALITY FIELD IS UNUSABLE IN THIS SET -------------------------------------------------------------------------- Summing the deaths field across the 6,054 complaints returns 45. All 45 come from ONE complaint — a 2010 Honda Civic report describing a single crash into a guardrail. The figure is a data-entry artifact, and no fatality total is published anywhere in the article as a result. 26 complaints report one or more injuries. This is recorded here because an unchecked sum would have produced a dramatic and false sentence. -------------------------------------------------------------------------- 8. RECALLS -------------------------------------------------------------------------- Campaigns counted where the component text names steering AND the defect or component text names a rack, a steering gear or power steering: 372 distinct campaigns, 16,081,165 vehicles 121 campaigns (6,879,221 vehicles) describe loss of assist, increased effort or difficulty steering 110 campaigns (4,862,418 vehicles) have a remedy that replaces the steering gear or rack 21 campaigns (3,624,119 vehicles) have a remedy that is a software update, reprogramming or recalibration 135 campaigns include an inspection step 188 were filed in the 2010s and 184 in the 2020s The full list is published in steering-rack-recalls.csv with the remedy classification for each campaign, so the software and replacement flags can be checked line by line. -------------------------------------------------------------------------- 9. FILES IN THIS FOLDER -------------------------------------------------------------------------- rack-complaints-by-context.csv What the 6,054 narratives describe, with counts, shares, and the NHTSA towed / crash / fire flags against the all-complaint baseline. rack-complaints-top-models.csv The 60 most-reported models with median reported mileage, the share describing loss of assist, the share naming a recall, the share mentioning a leak, and the leading model years. rack-owner-reported-costs.csv The 1,294 owner-written dollar figures binned, with median and quartiles. steering-rack-recalls.csv All 372 campaigns with population, remedy classification and the first 200 characters of the defect description. No personal data is published: no VIN, no city, no dealer name, no free text beyond the truncated official defect summaries in the recall file. -------------------------------------------------------------------------- 10. LIMITATIONS -------------------------------------------------------------------------- - Complaints are self-reported, unverified, and have no denominator. - Filing is driven by publicity: 45.7% of these narratives name a recall or campaign, which is extremely high and shows how strongly recall notices shape this dataset. - Keyword matching cannot read intent. "Steering gear" is used by owners for the rack, the column and the whole system. - Owner-written dollar figures mix quotes with paid invoices. - Recall populations are potentially-affected counts filed by the manufacturer, not counts of failed parts or completed repairs.