2026 Study · Washington scorecard
In Washington, 83 of 86 large-employer postings showed a pay range and benefits
Washington is the third state with a large enough sample for its own scorecard. The protocol requires at least 30 determinate postings before a state rate is quoted, and Washington is the third to reach it: 16 determinate Washington postings from the follow-up study, plus 70 newly collected Washington postings, for 86 in total. Eighty rows were collected on 2026-11-13; 10 example postings were excluded as out of scope (see finding 5), leaving 70 in the scorecard.
83 of the 86 showed a pay range and a general description of benefits — a 96.5% pass rate (Wilson 95% confidence interval 90.2–98.8%). Washington is the first scorecard state where the grade is two-part: a posting must show a pay range (or a single fixed wage, permitted in lieu of a range) and a general description of benefits. A posting with a pay range but no benefits description fails, and a posting with benefits but no pay fails. The rate moved from the as-collected figure to 96.5% through four owner rulings on links and scope, plus one live re-check — the full path is in finding 3, and every ruling is disclosed rather than hidden.
This is a third scorecard, not a representative sample of Washington postings. It describes 86 postings, nothing more. All three remaining fails are from the follow-up study's Washington set; the new collection passed 70 of 70 — see finding 2.
Key findings
1. 83 of 86 Washington postings showed a pay range and benefits
The cumulative Washington sample is 86 determinate postings: 83 pass, 3 fail, a 96.5% pass rate (Wilson 95% confidence interval 90.2–98.8%). The interval is narrower than the California and New York scorecards' because the sample is larger and the rate is higher, but it is still wide — the honest reading is that Washington postings from large employers usually show a pay range and benefits, with real uncertainty about the exact rate.
2. The new collection: 70 of 70 showed a pay range and benefits
The targeted Washington collection added 70 determinate rows — 80 rows collected from 8 large employers on 2026-11-13, 10 of them excluded as out of scope (finding 5), none of them indeterminate — and all 70 showed a pay range and a general description of benefits, a 100% pass rate (Wilson 95% confidence interval 94.8–100%). The 7 employers in the scorecard span healthcare, aerospace, retail/grocery, tech, telecom, fintech, and logistics. That is what the most-recent-postings design returned for these employers, but it is not a cross-section of Washington employers or industries. The wave-only split is reported here for transparency, not as a standalone rate; the scorecard rate is the cumulative one. A 100% wave-only rate should not be read as a state verdict — all 3 remaining fails are follow-up-study rows.
3. The ruled path: four owner rulings moved the rate from the as-collected figure to 96.5%
As collected, the cumulative was 96 determinate (78 pass, 18 fail), 81.2% — stated here once, for the record. The owner accepted all four calls on 2026-11-15, and the path runs: 9 benefits-link flips (finding 4) moved it to 87 pass / 9 fail, 90.6%; the live re-check flipped the Tacoma HVAC fail to a pass (finding 6), 88 pass / 8 fail; excluding the 10 W04 example postings (finding 5) removed 5 passes and 5 fails, landing at 83 pass / 3 fail, n=86, 96.5% (Wilson 95% confidence interval 90.2–98.8%). Each step is disclosed in its own finding below; nothing in the path is hidden.
4. The benefits-link ruling: 9 fails flipped to passes — a judgment call, disclosed with its soft spots
Nine postings from a logistics employer (W08) show a pay range but name zero specific benefits in the posting text; the only benefits mention is the identical template line “For details on our comprehensive benefits, click here” — a hyperlink to a benefits page. The coordinator had kept them as fails; the owner ruled on 2026-11-15 that the link counts, on Washington L&I Administrative Policy ES.E.2 §5 (issued 2025-12-03): employers “must include a general description of benefits and other compensation but may choose to utilize a link or hyperlink to lead the applicant to a more detailed description,” provided the link is functional, up-to-date, and applies to the specific posting.
The linked page was verified live on 2026-11-14, twice, including an independent click-through: the Courier/DOT-4 posting in Kent ($25.42–$35.96/hr) links to a public 3-page 2026 benefits-statement PDF, one click, no login, no paywall. Page 1 lists medical, dental, and vision coverage, FSAs, a 401(k) with match, educational assistance up to $5,250/yr, paid vacation, sick, holiday, and bereavement leave, 2 weeks paid parental leave, paid military leave, adoption assistance, long- and short-term disability, life and AD&D insurance, and commuter and voluntary benefits; pages 2–3 are paid-time-off tables by job classification. The same template sits behind all 9 benefits-link rows.
Two soft spots are disclosed, not settled. First, the statement is company-wide for U.S. regular employees, not written for the Courier role or Kent specifically — eligibility varies by position, hours, and service. Second, the document has no printed publication date or version anywhere (only 2026 transition-year references in the paid-time-off table), and the table never maps a Courier/DOT-4 to a column — its groups are Package Handlers, Maintenance and Operations Hourly, and Staff Hourly and Exempt — so the exact benefits for this job cannot be determined from the linked page alone. There is also a genuine ambiguity in the policy itself: §5's first sentence can be read as requiring a general description in the posting, with the link only adding detail. The scorecard presents the flip as a judgment call under the owner's ruling, with these limits disclosed — not as a clean pass. Supporting evidence that this is a real template difference, not a fetch artifact: the same employer's Material Handler posting in Seattle carries a full in-text benefits paragraph, and all 9 flips are full-text fetches with substantial text. The same link question arises in Illinois and was ruled the same way.
5. The pay-link ruling and the W04 exclusion: 5 pay-link fails stand, and 10 example postings are out of scope
Five postings from a warehouse-retail employer (W04) show no pay range in the text; each says “For additional information about pay ranges, click here.” The owner ruled on 2026-11-15 that pay links do not count: L&I's policy permits linking only for benefits and other compensation, and nothing permits linking the wage scale. The counter-evidence is disclosed: the same job code shows $20.00–$31.20/hr inline on other Gas Station Attendant postings from the same employer, so the range the link points to likely exists. The ruling is moot for the scorecard numbers because all 10 W04 rows are excluded as out of scope — but it still governs the remaining baseline fail, the Reservation Agent row whose pay range appeared only in a linked PDF supplement (finding 7).
Every W04 page carries the employer's disclaimer that the jobs listed are examples of typical positions, not necessarily open or available. The disclaimer was spot-checked verbatim live on 2026-11-15: “The jobs listed are examples of the typical kinds of positions that [employer] may hire for when openings exist. The listing does not mean that any positions are currently open or available at [employer].” (Employer name redacted.) All 10 W04 rows also carry the same Published date, 2026-07-28 — a bulk July posting refresh. The law defines a posting as a solicitation for a “specific available position,” so the owner ruled the 10 example postings out of scope on 2026-11-15. Exclusion removes 5 passes and 5 fails (the 5 fails were the pay-link rows), moving the cumulative from 88 pass / 8 fail to 83 pass / 3 fail, n=86. The W04 rows are removed from the dataset, not re-graded. The partial-text pass among them — a Gas Station Attendant posting in Kennewick, where pay and the start of the benefits paragraph are visible but the benefits text is truncated mid-word — is excluded with them, so its planned live re-check is moot.
6. The live re-checks: the health-system passes confirmed, the HVAC fail flipped, one caveat stands
The 10 passes from a large Seattle health system (W01) were graded on search-index text, not completed detail fetches — the detail pages had returned only a “job has been filled” stub. The live re-check on 2026-11-15 confirmed all 10: every range matches as collected, the Benefits Information section is present on all 10, and the “job has been filled” notice was not seen on any live page. Two other healthcare employers were tried first and abandoned on fetch failures (one with no extractable content, one HTTP 403); those are gaps, not zeros.
The Tacoma HVAC-Refrigeration Technician fail (W03) had a full 133-line fetch with a Compensation section naming no pay at all. The live re-check on 2026-11-15 found the Role Details pay card shows “Tacoma, WA: $32.00 - $64.00/hr*” (also Lakewood $32.00–$64.00/hr), with a Benefits & Perks list present. Pay and benefits both present live — the row flips to pass, and its recorded pay figure is now the live pay-card figure.
One caveat stands. All 10 telecom-employer (W06) rows were confirmed live on 2026-11-15, and all remain passes under the frozen rule (a pay range is shown) — but only rows 5 and 9 show a Washington-specific range live: Sr Data Scientists, $166,036–$171,036/yr, and a Principal Engineer, Cybersecurity posting in Bellevue, $166,300 to $224,900. The other 8 show a National Base Pay Range plus Total Target Cash, with no location-specific line; the Washington-specific figures sit behind pay-lookup links that were not opened, so they are unverified. Four W06 rows are multi-location; Washington is an explicit hiring location on each.
7. The 3 remaining fails are all from the follow-up study — and all three are pay fails
The wave-3 collection contributes zero fails. The three fails are all pay fails from the follow-up study's Washington set, and benefits are present in all three: a Reservation Agent posting in Issaquah with no pay range in the posting — a linked PDF supplement showed $21.00–$32.90, but under the pay-link ruling (finding 5) a linked pay figure is not a range in the posting, so the fail stands; and two Sales Account Manager postings, one in Seattle and one in Tacoma, each with pay wording only and no figure.
Breakdown
| Source | Determinate | Pass | Fail |
|---|---|---|---|
| Follow-up study, Washington | 16 | 13 | 3 |
| Targeted Washington collection | 70 | 70 | 0 |
| Cumulative Washington | 86 | 83 | 3 |
Raw counts only — no rate is quoted for any employer group, because none reaches 30 determinate rows. Indeterminate rows: 7 in the follow-up study's Washington set, 0 in the targeted collection. The targeted collection as scored excludes the 10 W04 example postings (finding 5); the as-collected alternative is disclosed once in finding 3 and is not the headline.
What this adds
The follow-up study measured 12 states but quoted no state rate, because no state reached 30 determinate rows. The California scorecard was the first state to clear that bar and the New York scorecard the second; this page is the third, and it exists to answer one question: what does the Washington sample look like once it is large enough to quote? It does not re-argue the follow-up study's case — that is on the follow-up study's page.
Grade your postings in bulk → Check a single posting →
Methodology — summary
The sample. 86 determinate Washington postings from large employers (1,000+ employees): 16 from the follow-up study's graded dataset and 70 from a targeted Washington collection on 2026-11-13 (the 10 most recently posted roles per employer from their public career pages, excluding internships, contractor/gig roles, and internal-only transfers; 80 rows collected, 10 example postings excluded as out of scope). Determinate means the posting text was substantial enough to grade; indeterminate and ungradable postings are excluded from the rate, not treated as passes.
The Washington rule. Graded only against the frozen requirement dataset (DATA_VERSION 2026-10-14). Main's DATA_VERSION was bumped to 2026-11-13 by a change that added only penalty and statute metadata; the grading requirements are untouched, so grades are identical under either version. Washington requires a pay range and a general description of benefits; a deadline is not required in Washington. The grade is two-part: pay — a range, or a single fixed wage amount permitted in lieu of a range per SSB 5408 (effective 2025-07-27) — and benefits — a general description in the posting, or, per the owner's ruling, a hyperlink to a publicly viewable benefits page that is functional, current, and specific to the posting, per L&I Administrative Policy ES.E.2 §5 (finding 4). The full research protocol is on the methodology page.
The fixed-wage exception was not exercised. No collected Washington row turned on the single-fixed-wage exception: every pay-passing row shows a range, and every pay-failing row shows no pay at all. The exception is recorded here so the rule is on the record, not because any grade depends on it.
The link decisions (owner rulings, 2026-11-15). The benefits link counts (finding 4): 9 W08 fails flipped to passes under ES.E.2 §5, with the linked page verified live and the soft spots and ambiguity disclosed. The pay link does not count (finding 5): 5 W04 fails stand, though the ruling is moot for the numbers because W04 is excluded — it governs the Reservation Agent baseline fail instead. Both rulings were accepted by the owner on 2026-11-15, and the same benefits-link question in Illinois was ruled the same way.
The live re-checks (2026-11-15). All 10 health-system passes confirmed live (finding 6); the Tacoma HVAC fail flipped to a pass on the live pay card; all 10 telecom passes confirmed live with the national-vs-Washington-specific caveat standing. The linked benefits page was verified live on 2026-11-14, twice. Any row a live re-check overturns is disclosed here.
Evidence tiers and the fetch-miss rule. 60 full-text rows and 10 search-index rows (all from the Seattle health system), 0 partial-text rows in the scored set. A full-text fail requires substantial posting text retrieved with the required element absent; thin, empty, or incomplete fetched text is indeterminate, never a fail. No wave-3 Washington row was graded indeterminate — every collected row had substantial text.
The example-posting exclusion. The 10 warehouse-retail (W04) rows were removed as out of scope because every page carries the example-posting disclaimer and the same bulk Published date, 2026-07-28 (finding 5). They are removed from the dataset, not re-graded.
Baseline check (no discrepancy). Unlike California, the follow-up study's Washington baseline reconciles: the graded dataset shows 16 determinate (13 pass, 3 fail), 81.2% (Wilson 95% confidence interval 57.0–93.4%), with 7 indeterminate, matching the published table exactly. No baseline correction was needed, and this scorecard uses the auditable 16.
Counting and de-duplication. No row in the new collection matches any follow-up Washington row (title and location checked against all 23 follow-up Washington rows), and no title-and-location duplicates exist within the new collection. Same-title repeats at different locations are distinct requisitions at distinct sites and were counted separately, following the follow-up study's precedent.
Anonymization. Employers are not named. Collection employers are referred to only by sector and location, coded W01–W08. Anonymized datasets are available on request from the press contact.
Honest limits
- This is a third scorecard from 7 collection employers plus the follow-up study's Washington rows, not a representative sample of Washington postings. It describes the sample, nothing more.
- The confidence interval is still wide (90.2–98.8%) because the sample is small. The pass rate should be read as “usually”, not as a precise figure.
- The new collection is concentrated: 7 employers in the scorecard, leaning corporate and professional and tech, with retail store roles and logistics as the hourly counterweight. No industry rate is quoted.
- The wave-3-only rate is 100% (70 of 70). That should not be read as a state verdict — all 3 remaining fails are follow-up-study rows, and the sample is 7 employers.
- The benefits-link flip is a judgment call under the owner's ruling, not a clean pass: the linked statement is company-wide, not role- or location-specific; it carries no publication date or version; and the paid-time-off table never maps a Courier/DOT-4 to a column, so the exact benefits for the job cannot be determined from the linked page alone. The policy text itself is ambiguous on whether a general description must sit in the posting.
- Eight of the 10 telecom-employer rows show a national base pay range, not a Washington-specific one. The Washington-specific figures for those 8 sit behind pay-lookup links that were not opened and are unverified; the 8 pay-lookup links are the one optional live re-check still outstanding.
- Recency was proxied where a site did not publish posted dates or a date-sorted list: one employer's list page exposes no job links or dates in fetched text, so its rows are the 10 most recent fetchable roles, not provably the 10 most recent overall; another employer has no date-sorted public list, and a social-media carousel promoted other Seattle roles with unknown published dates, so mid-October roles may have been missed.
- Whether any collection employer overlaps the follow-up study's panel could not be verified without opening the sealed employer key, which was never opened; the panel-vs-fresh split is therefore not reported here.
- The grader is a heuristic screen, not a legal determination, and naming an employer as “non-compliant” on that basis would overclaim what the data supports. “Checked, never certified”: a passing grade doesn't mean a posting is legally compliant, and a failing grade doesn't mean the employer broke the law.
- Grading is on captured postings, not on live status at publication — except for any row corrected by a live re-check, which is disclosed here.
- This page is not the remote-posting study: that study's collection is incomplete and its merge and grading have not run; nothing in this page's numbers comes from it.
This restraint is deliberate. A study that overclaims gets one news cycle and zero repeat citations; a study that's honest about its limits gets cited every year.
Press
For questions about the data or the method, contact jobdescriptiontool@gmail.com.
jobdescriptiontool.com is a free tool for writing job descriptions and checking them against state pay-transparency requirements, with a pay-transparency checker covering 15 U.S. jurisdictions.