CA Labor Commissioner Judgments, Wage Claims and the Port Drayage List (plus five free multi-state wage-theft feeds)

Who owes what. Every California Labor Commissioner (DLSE) judgment with the dollar amount attached: the total owed, whether it is still open, the court that entered it, the DIR office that brought it, the industry, and the address of every liable party named on it. Plus the Labor Code 2810.4 port drayage unsatisfied-judgment list with its full month-by-month history, the DLSE wage-claim index, and five free debarment and wage-theft feeds from Washington, New Jersey, Colorado, Illinois and Oregon in the same shape.

California publishes this only through a Salesforce Lightning search portal, one query at a time, with no export. This actor sweeps the whole corpus, verifies its own coverage with a second independent pass, and hands you clean rows.

On the 2024 judgments alone the state names 1,939 liable parties across 1,874 judgments, and more than 99% carry a dollar amount above zero. The current port drayage list is 54 liable parties across 52 judgments and 29 employers, totalling $2,722,437.60 (as of 2026-08-26).

Who this is for

Four groups have a statutory reason to check these lists, and to check them again next month. Every citation below was read from the primary source on 2026-08-26. This is a description of public registries, not legal advice.

• Anyone using a port drayage motor carrier. California Labor Code 2810.4 requires the Labor Commissioner to post the unsatisfied-judgment list and update it monthly by the fifth day of each month (subdivision (c)(5)). A customer that engages a carrier on that list "shall share with the motor carrier or the motor carrier's successor all civil legal responsibility and civil liability" owed to a port drayage driver or to the state (subdivision (c)(3)). Current version: Stats. 2024, ch. 739 (AB 2754), effective 2025-01-01.
• Anyone buying a Washington business. RCW 49.48.086(4) makes a successor liable for an outstanding citation and notice of assessment when the buyer has "a prompt, reasonable, and effective means of accessing and verifying the fact and amount" of it. RCW 39.12.065(3) bars a contractor with an unpaid prevailing-wage penalty from bidding on public works until it is paid.
• Anyone awarding public contracts in New Jersey or Illinois. N.J.S.A. 34:1A-1.16 prohibits any business on the WALL from public contracting with the State, its agencies, counties and local government bodies; the list is updated at least monthly. Illinois Prevailing Wage Act section 11a (820 ILCS 130) bars debarred contractors, their affiliated firms, directors, officers and agents from any public works contract or subcontract.
• Compliance, credit and diligence teams generally. Colorado publishes wage theft decisions with the wages ordered, the penalties and the individually named liable parties under its Wage Theft Transparency Act.

What you get

One flat row per source record (default). A real judgment row:

{
  "employer_name": "Brady Hoggins, an Individual",
  "address_street": "1220 Rosecrans Street #853 San Diego CA 92106",
  "naics_code": "238130",
  "naics_title": "Framing Contractors",
  "source": "judgments",
  "source_agency": "California Labor Commissioner (DLSE)",
  "source_state": "CA",
  "source_url": "https://cadir.my.site.com/s/judgmentsearchresult?searchtype=judgmentsearch",
  "record_id": "a2Acs000002Mj29EAC",
  "case_number": "J-93628",
  "status": "Open/Unpaid",
  "judgment_date": "2024-12-31",
  "amount_total": 251390.11,
  "amount_due_others": null,
  "ubi": null,
  "entity_key": "01c1de238e4a7941a043d95fae8704322f63037e",
  "scraped_at": "2026-08-26T21:00:00+00:00"
}

Every row carries the same fields, so rows from different states stack in one table. Fields a source does not publish are null, never missing.

One judgment can name several liable parties, and each one is its own row. Judgment J-93628 above is one of three rows sharing case number J-93628, one per defendant: an individual and two corporations. That is the point of the dataset, so a row is identified by the judgment plus the party, and the entity rollup counts each judgment amount once.

record_id is that identity: the state's judgment id and the normalized employer name, joined. Do not key on judgment_party_id. It is included for reference only, and the state does not hold it stable: the same party on the same judgment comes back under different party ids from different queries, and even between two runs. Use record_id, or case_number plus employer_name.

Which datasets carry dollar amounts

Dataset	Amounts	What the amount is
judgments	Yes	Judgment total, above zero on more than 99% of records
portDrayage, portDrayageHistory	Yes	Judgment total, plus the amount due to parties other than the worker
njWall	Yes	Total liability owed under the final judgment or order
coWageTheft	Yes	Wages ordered plus penalties and fines, summed
caloshaPenalties	Yes	Citation initial penalty
wageClaims	No	A volume and status index reaching back to 1997. DLSE publishes no amount on the claim search
waDebar	No	Status, RCW cited, and a Yes, No or - flag for penalty due and wages due. No dollar figures are published
ilDebar	No	A free-text notice with the debarment period
orLaborContractors	No	A licence register, not an enforcement list

Input

Run it with no input to collect judgments, the current port drayage list, and all five free feeds.

Field	Type	Default	Description
datasets	array of strings	[] (every fast dataset)	judgments, portDrayage, portDrayageHistory, wageClaims, waDebar, njWall, coWageTheft, ilDebar, orLaborContractors, caloshaPenalties. wageClaims and portDrayageHistory run only when named.
outputMode	records or entities	records	One row per source record, or one row per deduplicated employer.
verifyCoverage	boolean	true	Second pass over judgments (by status) and wage claims (by month), compared and merged. Written to VERIFICATION.
firstJudgmentYear	integer	2000	Where year slicing starts. A fixed catch-all slice always covers everything before 2000; raising this skips the years in between.
wageClaimsFirstDate	string (date)	2010-01-01	Where day-by-day slicing starts. The corpus reaches back to 1997 and a fixed catch-all slice collects those earlier years whatever this is set to; raising this skips the years in between.
portDrayageHistoryFrom	string (date)	2019-01-01	First month of the port drayage back-fill.
includeNonDefendantParties	boolean	false	Keep wage-claim rows for the claimant side. Those rows name individual workers.
requestDelayMs	integer	350	Politeness delay per worker after each request.
maxConcurrency	integer (1 to 4)	2	Maximum simultaneous requests.
fwuid, auraLoaded	string, object	stored values	Escape hatch if Salesforce changes its framework identity.

An unknown dataset name fails fast with a clear error instead of quietly collecting nothing.

Example runs

The port drayage list, one request, finishes in seconds:

{ "datasets": ["portDrayage"] }

Every California judgment since 2020, with amounts, verified:

{ "datasets": ["judgments"], "firstJudgmentYear": 2020 }

One row per employer across California and the five free feeds:

{ "outputMode": "entities" }

The full wage-claim back-fill, every claim back to 1997 (about 6,150 requests, roughly two hours at the default pacing):

{ "datasets": ["wageClaims"], "wageClaimsFirstDate": "2010-01-01" }

Just this month's wage claims, which is a few dozen requests. The years between 2010 and the date you give are deliberately not swept:

{ "datasets": ["wageClaims"], "wageClaimsFirstDate": "2026-08-01" }

The whole port drayage add-and-remove history, one snapshot per month:

{ "datasets": ["portDrayageHistory"], "portDrayageHistoryFrom": "2019-01-01" }

Memory and run time

• Records mode needs 1024 MB, the default. Rows are written to the dataset slice by slice as the sweep goes, and the run keeps only per-slice counts and record keys for verification. A full run holds about 670,000 records (629,776 wage claims, 31,027 judgments, the port drayage list and its history, and the five feeds, as measured 2026-09-29).
• Entities mode needs 4096 MB for a full run. The rollup can only be built once every record is in, so this mode holds the whole corpus in memory until the end. The default datasets, which leave out wageClaims, are far smaller.
• Raise the run timeout for a full run. Apify's default run timeout is 3,600 seconds, and a full wage-claim back-fill is about 6,150 requests plus its verification pass, roughly two hours at the default pacing. Set a larger timeout in the run options (for example 14,400 seconds) when wageClaims is selected from 2010, or the platform stops the run partway through.

Output modes

records (default): one flat row per source record, in the shape above.

entities: one row per employer, keyed on a hash of the normalized name and address, with name_variants, addresses, states, sources, the per-source records, and rollups: total_judgment_amount, open_judgment_count, on_port_drayage_list, wa_debarred, nj_wall_listed, record_count. Washington rows carry their ubi.

total_judgment_amount adds up the California judgment totals, counting each judgment once per employer. That matters twice over: one judgment naming three defendants arrives as three rows, and the port drayage list repeats judgments the judgment search already returned. Penalties from the other states are not folded into it; they stay on their own rows.

Entity matching in this version is deliberately conservative: it folds case and punctuation but not address abbreviations, so it can leave two spellings of one employer apart, and it will not merge an addressed record with an address-less one. Sources that publish no address at all (Washington debarments, Illinois debarments, California wage claims) therefore match each other on name, but do not merge into an addressed California judgment. Under-merging is the deliberate choice: an employer's total liability should never be overstated by a bad match.

Data quality and the verification summary

• The sweep verifies itself. Judgments are collected in entry-date slices, then collected again by judgment status over exactly the same date windows, and the two sets of judgment-and-party keys are compared. Wage claims are collected one day at a time, then recounted one month at a time. Anything only the second pass saw is merged into the output. The comparison lands in the VERIFICATION record of the run's key-value store, with per-dataset counts, slices fetched, slices split, every failing slice, and the IDs each pass saw alone. Judgment statuses change between passes, so a handful of one-sided records is normal and the report says what it saw rather than promising a clean match. Two kinds of record cannot be reached by a status pass at all: those with no status, and those carrying a status the state never offers in its own picklist (Pending/Open is live in the data and missing from the picklist). Both are counted separately instead of being called a coverage miss.
• A resumed run verifies what it resumed. If the platform migrates or restarts a run, it resumes from its last checkpoint and skips the slices already written to the dataset. The checkpoint is saved every 50 slices, after each dataset's main sweep, and when a migration starts, and nothing more is written after that migration save, so a migration writes no row twice. An unannounced crash can repeat the rows written since the last save. The checkpoint keeps each slice's record count (and its judgment statuses or wage-claim docket months), so the second pass still covers the whole corpus: windows the run fetched itself are compared key by key as usual, and windows that came from the checkpoint are compared by count against the second pass. A count that no longer agrees is listed under count_drift with the window, the stored count and the recount, and fails the match; nothing is merged for it, because those records are already in the dataset. The per-dataset count and the totals include the resumed slices, so a resumed run reports the corpus it holds, not only what it fetched after the restart. Entities mode never resumes: it writes nothing until the end, so it starts over instead.
• Zero date gaps, asserted. The wage-claim check asserts that every single calendar day between the bounds was actually requested and did not error. A missing day is listed in the report.
• The wage-claim corpus starts in 1997, not 2010. Day slicing starts at 2010-01-01 because that is where the claim volume is, and one catch-all slice collects everything earlier: about 31,000 records back to 1997, in roughly 37 requests rather than the 4,750 that day-slicing them would take. The month recount follows the data, so those earlier records are verified the same way the rest are.
• A slice that hits the cap splits itself. The state caps any response at 5,000 rows, so exactly 5,000 rows is never a count. Such a slice is halved and re-fetched down to a single day; a single day that still returns 5,000 is recorded as an error rather than silently truncated.
• Data-entry typos are kept, not dropped. DLSE holds judgment entry dates in 2029 and 2205. A far-future slice catches them and the verification record counts them as out of window.
• Some judgments carry a zero total. About 27 in 5,000 come back as 0.0 rather than a figure. The zero is passed through as the state published it, and the verification record counts those rows under records_without_positive_amount rather than hiding them.
• Judgment liability is joint and several. A judgment naming three defendants is three rows, each carrying the full judgment total, because each party can be pursued for the whole amount. In entities mode each of those three employers therefore shows the full amount in total_judgment_amount. Adding total_judgment_amount across entities overstates the true total owed; to total real liability, sum by case_number in records mode instead.
• Dates are ISO. Every date in the output is YYYY-MM-DD, converted from whatever the source used.
• Statuses are the source's own raw values, not normalized.
• judgment_party_id is not stable and must not be used as a key. The state issues a different party id for the same party on the same judgment depending on which query returned the row. The actor collapses those twins itself, which is why the same employer never appears twice on one judgment in the output. The field is published only because it is what the state returned.
• A feed that fails does not fail the run. Each of the five free feeds is fetched independently, and its row count, Content-Length and Last-Modified are recorded. Illinois genuinely lists only one debarred company today, so zero rows there is a warning, not an error.

Freshness

DIR's own HTML page describing the port drayage list is stamped March 2021 and still describes the 2018 version of the statute. The endpoint behind the search portal is current, and this actor reads the endpoint.

Personal data

These are public enforcement registers, and the state publishes them so that the parties with a duty to check them can. Sole proprietors and individually named liable parties therefore appear under personal names, and the address on a judgment can be a home address. Colorado names individual liable parties in its own spreadsheet column.

Wage-claim rows exist for both sides of a case. This actor keeps only the defendant side by default, because the other side is the worker who filed the claim. includeNonDefendantParties turns that filter off; leave it off unless you have a specific reason.

Do not treat any of this as a consumer marketing list.

The actor republishes these registers as the state publishes them, adds no information about any person beyond what the state releases, and does not edit records. A correction or removal happens at the source and flows through on the next run.

This dataset is not a consumer report, and the publisher is not a consumer reporting agency as defined by the Fair Credit Reporting Act (15 U.S.C. § 1681 et seq.). Do not use it as a factor in establishing any consumer's eligibility for credit, insurance, employment, tenancy, or any other purpose covered by the Act. It is published for statutory compliance screening of businesses (such as the customer duty in Labor Code section 2810.4), journalism and research.

Politeness

Two concurrent requests by default, a 350 ms delay per worker after each request, three attempts with exponential backoff on server errors and timeouts, and a User-Agent that identifies the actor. There is no CAPTCHA and no login on any of these sources; the state serves this data to an anonymous visitor.

Sources

• California Labor Commissioner judgment search and port drayage list
• California Labor Commissioner wage claim search
• Washington L&I debarred contractors
• New Jersey WALL
• Colorado wage theft transparency decisions
• Illinois debarred public works contractors
• Oregon BOLI labor contractor licensing
• Cal/OSHA penalties over $100,000

This actor is not affiliated with or endorsed by any of these agencies. Data accuracy is the agencies' own. Every record carries the timestamp it was scraped.