PULSCAR ANALYSIS OF NHTSA PUBLIC DATA — TIE RODS Published September 2026. License: CC BY 4.0. Attribution: "Pulscar analysis of NHTSA public data, September 2026." Companion article: https://www.pulscar.io/blog/tie-rod-replacement-cost ------------------------------------------------------------------ 1. SOURCE FILES ------------------------------------------------------------------ All four tables come from NHTSA's own public flat files, downloaded 17 September 2026 and loaded into a local SQLite database without modification: - FLAT_CMPL.txt — owner complaints. 777,040 records received 1 January 2015 through 8 September 2026. Field definitions: https://static.nhtsa.gov/odi/ffdd/cmpl/CMPL.txt - FLAT_RCL.txt — safety recall campaigns (one row per make/model/model-year per campaign). - FLAT_TSBS.txt — manufacturer communications / technical service bulletins (one row per make/model/year). - FLAT_INV.txt — ODI investigations (preliminary evaluations, engineering analyses, defect petitions, audit queries). ------------------------------------------------------------------ 2. CASE SENSITIVITY — READ THIS FIRST ------------------------------------------------------------------ NHTSA complaint narratives are stored in ALL CAPS for records filed up to roughly 2020 and in mixed case after that. Every query below is therefore case-insensitive: the narrative, summary, defect, consequence and remedy fields are lowercased with SQL lower() before matching. A case-sensitive query on this file silently loses the majority of the older records and would have missed most of the Dodge Ram and Toyota RAV4 material entirely. ------------------------------------------------------------------ 3. THE COMPLAINT SET (n = 1,468) ------------------------------------------------------------------ Base query, run against the complaint narrative (descr) only: lower(descr) LIKE '%tie rod%' OR lower(descr) LIKE '%tierod%' OR lower(descr) LIKE '%tie-rod%' That returns 1,474 complaints. Overlap is handled by the OR: "tie rods" is a subset of "tie rod", so the three patterns were unioned rather than added (1,412 + 19 + 52 would double-count). Exclusions: 6 complaints filed under heavy-truck, motorhome and travel-trailer makes were removed (KENWORTH 2, FREIGHTLINER 1, FOREST RIVER 1, TIFFIN 1, THOR MOTOR COACH 1), because a Class 8 steer-axle tie rod is a different part with a different service life and a different price. No other rows were dropped. Final set: 1,468 complaints, received 1 January 2015 through 3 September 2026. Nothing was tightened beyond that. An earlier attempt required a failure or noise word within 160 characters of the tie rod phrase; it removed only 61 records and several of those were plainly genuine ("TIE ROD DETACHED FROM THE VEHICLE DURING ACCELERATION", "TIE ROD END OUT ALREADY"), so the filter was discarded as harmful rather than helpful, and the false-positive rate is published instead (section 4). ------------------------------------------------------------------ 4. SAMPLE READ AND FALSE-POSITIVE RATE ------------------------------------------------------------------ 30 complaints were drawn at random from the 1,468 (Python random.seed(20260919), random.sample) and read in full by a human before any count was published. Result: 21 of 30 (70%) — a tie rod is the part that was worn, loose, bent, broken or replaced. True positives. 5 of 30 (17%) — the tie rod was damaged by something else failing first (a control arm snapping, a wheel separating, a crash). The phrase is accurate but the tie rod is collateral damage, not the defect. Counted in the set; flagged here so the separation share is not over-read. 4 of 30 (13%) — false positives. Three name the tie rod as a location rather than a failure ("power steering fluid leaking out of my tie rod boot" where the rack failed; "epoxy near the tie rod assemblies"; an owner speculating that a rack problem is "starting to affect tie rods, bearings etc" with no diagnosis). One lists a tie rod among parts replaced during a steering repair that did not fix the fault. Published false-positive rate: 13% (4 of 30). 95% binomial confidence interval roughly 4% to 31% on a sample this small. Every share in the article and in the CSVs should be read with that in mind: they are shares of complaints whose narrative names a tie rod, not verified failure rates. We could not find a regex that removed those four without also removing genuine reports, because the false positives use the same vocabulary as the true ones. Publishing the rate is the honest option. ------------------------------------------------------------------ 5. WHAT EACH COMPLAINT-SIDE FILE CONTAINS ------------------------------------------------------------------ tie-rod-complaints-by-context.csv One row per measured feature of the narrative, with the count and the share of 1,468. Feature queries are case-insensitive regular expressions. Two kinds appear: (a) "near the tie rod phrase" — the pattern must occur within 120 characters, before or after, of a match of /tie[ -]?rods?/. Used for break/separation, wear/looseness/play, and bend, so that a break described elsewhere in the narrative is not attributed to the tie rod. (b) "anywhere in the narrative" — used for recall, alignment, inspection, rust, pothole and tire wear, where the co-occurrence is the point. Patterns used for (a): break/separation: \b(broke|broken|snapped|snap|fracture[d]?|separat\w*|detach\w*|sheared|shear|came apart| fell off|came off|gave way|failed completely)\b wear/looseness: \b(worn|wear|wore|loose|looseness|play|slop\w*|excessive movement)\b bend: \bbent\b|\bbend\b|\bbending\b The NHTSA crash and towed rows are the structured flags in the file (CRASH, TOWED_FLAG), not text matching. All-complaint baselines computed over the same 777,040 records: crash 4.37%, towed 5.11%, fire 1.82%. tie-rod-complaints-by-make.csv Every make in the set with its complaint count, share, crash and towed flag counts and percentages, and the median reported mileage with the number of records that carried one. tie-rod-complaints-top-models.csv The 40 most-named make/model pairs, with the three leading model years, the median reported mileage, the number of records carrying a mileage, and counts of narratives containing a break/separation word, the phrase "death wobble", and any form of "align". tie-rod-complaints-by-model-year.csv Complaint count by vehicle model year. tie-rod-owner-reported-amounts.csv A hand-read table, not a query result. Candidate figures were found by scanning for a dollar amount between $40 and $4,000 within 110 characters of a tie rod phrase (80 matches), and every one of those was then read in full; only the rows where the narrative states plainly what the figure covered were kept, and the description column is our own summary rather than a quotation. One row comes from a defect petition rather than a complaint (DP05-003, marked in the source column). These are what individual owners said they were quoted or paid, in the year they filed. They are not a market survey and they are not adjusted for inflation; the earliest is 2005 and the most recent is March 2026. Do not compute a median from them. MILEAGE Reported mileage is used only where it is greater than 0 and less than 400,000. 955 of the 1,468 complaints carry a usable figure. Median 86,008 miles; first quartile 45,744; third quartile 129,218; 19.2% at or below 36,000 miles; 34.8% at or below 60,000. Mileage in this file is owner-entered and frequently rounded. ------------------------------------------------------------------ 6. RECALLS (tie-rod-recalls.csv) ------------------------------------------------------------------ Query, case-insensitive, over the recall table: lower(defect) LIKE '%tie rod%' OR '%tie-rod%' OR lower(consequence) LIKE '%tie rod%' OR lower(remedy) LIKE '%tie rod%' OR '%tie-rod%' OR lower(component) LIKE '%tie rod%' OR lower(notes) LIKE '%tie rod%' Counted as DISTINCT campaign numbers, because the flat file repeats a campaign once per make/model/model-year. 130 distinct campaigns name a tie rod anywhere in their text. This broad set is what the CSV contains. The broad set contains keyword false positives at the campaign level, because the match may sit in the notes or remedy rather than in the defect: 23E-040 (Veoneer brake boosters, 137,033 units) and 23V-458 (Honda brake master cylinder, 124,077 vehicles) are the two largest. They are left in the CSV with their own component codes so they can be filtered, and EVERY figure quoted in the article comes instead from the clean subset: 94 campaigns carry the component code STEERING:LINKAGES:TIE ROD ASSEMBLY — 1,828,889 units. Inside that subset: 17 are passenger-vehicle campaigns from passenger-car manufacturers — 1,659,907 vehicles. 16 are equipment campaigns (campaign numbers of the form nnEnnn) — 95,221 replacement parts. The remaining 61 are heavy-truck, bus, fire-apparatus and RV campaigns. 26 campaigns in the broad set name improper heat treatment of a tie rod ball stud or clamp as the defect; seven of those are equipment recalls of replacement parts. Campaigns quoted by name outside the component subset are described for what they are, not as tie rod recalls: 12V-373, 13V-383 and 16V-596 are coded SUSPENSION:REAR and cover the Toyota RAV4 rear suspension arms that owners call rear tie rods, and 23V-704 is coded STEERING:RACK AND PINION. Two further cautions. Populations overlap: the 2006-2011 RAV4 population of 780,584 appears in both 12V-373 and 13V-383 and again in part in 16V-596, because Toyota recalled the same vehicles more than once, so campaign populations must never be summed into a "vehicles affected" total without saying so. And the population field (potaff) is the manufacturer's estimate of vehicles potentially affected, not the number that failed or the number repaired. The CSV keeps one row per campaign with the first 240 characters of the defect text and the first 160 of the remedy. ------------------------------------------------------------------ 7. MANUFACTURER BULLETINS (tie-rod-manufacturer-bulletins.csv) ------------------------------------------------------------------ Query, case-insensitive: lower(summary) LIKE '%tie rod%' OR '%tie-rod%' OR lower(component) LIKE '%tie rod%' 2,069 rows match. The file carries one row per make/model/model-year, so rows were collapsed to DISTINCT nhtsa_id: 156 distinct bulletins. Of those, 79 were issued by passenger-car brands (an explicit brand whitelist; the remainder are Freightliner, Peterbilt, Prevost, New Flyer, Meritor, Hendrickson, Tiffin, Forest River and other commercial, bus and RV makes, plus parts manufacturers). Those 79 cover 1,188 make/model/year rows. Themes were counted by reading the summaries and by regular expression over them: 36 of 79 (46%) describe a noise — clunk, pop, click, knock, rattle, squeak, creak or shimmy. 28 of 79 (35%) concern fastener torque: a nut, jam nut, lock nut or clamp that is loose, under-torqued, over-torqued or "not fully secured". 6 mention grease or lubrication as part of the repair. 11 concern a tie rod boot or bellows. 8 mention alignment or toe. Categories overlap; a bulletin about a loose nut causing a clunk is counted in both. Three bulletins in the passenger set carry the placeholder text "SUMMARY TO BE PROVIDED ON A FUTURE DATE" and were counted as bulletins but contributed to no theme. Summaries in the CSV are truncated to 200 characters, and any tilde character in the source text was replaced with a hyphen so the file renders correctly in Markdown-based tooling. ------------------------------------------------------------------ 8. INVESTIGATIONS ------------------------------------------------------------------ Same case-insensitive match over subject, summary and component in FLAT_INV.txt: 28 distinct ODI actions name a tie rod, from EA72-012 to PE24-020. The two quoted in the article are PE11-009 (2008-2011 Dodge Ram outer tie rod ball stud failure, opened 4 April 2011, closed 6 February 2012) and PE24-020 (2021 Nissan Rogue inner tie rod failures, opened 5 August 2024, closed 24 August 2026), plus the audit query AQ14-003 into Chrysler's execution of recalls 13V-528 and 13V-529 and the defect petition DP05-003 (1999 Volkswagen Passat). ------------------------------------------------------------------ 9. LIMITATIONS — PLEASE READ BEFORE CITING ------------------------------------------------------------------ 1. Complaints are self-reported, unverified, and voluntary. They are not failure rates. There is no denominator: the number of complaints about a model reflects how many of that model are on the road, how long they have been there, how loud the online community is, and whether a recall letter prompted people to file. Nothing here says any vehicle is more or less reliable than another, and the counts must not be read that way. 2. Recall campaigns cluster complaints. A recall letter is itself a reason to file, so the makes with the largest tie rod recalls also have the largest complaint counts. Cause and effect run in both directions. 3. The mileage distribution is tilted toward early failure by construction. People file when a part fails sooner than they expected; a tie rod that quietly lasts 200,000 miles generates no document. 4. Crash and towed flags are set from the complaint record, and a crash that damaged a tie rod is flagged the same way as a tie rod failure that caused a crash. The sample read found 5 of 30 in that first category, and NHTSA itself made the same point when it closed PE24-020. The elevated crash and tow shares in this set are real and worth knowing, but they are not evidence that tie rods cause crashes at that rate. 5. Prices in the amounts table are individual owner reports over a 21-year span, in nominal dollars. 6. Small subsets stay small. Any figure below about 30 records is reported as a count, never as a percentage. Questions or corrections: support@pulscar.io