Executive Summary: Most underwriting teams treat "court records" as one check. It is actually two separate court systems with different subject matter, different search tools, and different costs, and a search in one does not cover the other. This guide explains what federal civil filings signal that state filings structurally cannot, why bankruptcy is the clearest example, and how to build a search stack that covers both layers honestly. One thing to state before anything else: Cobalt Intelligence does not cover federal court records. If you need federal data, you need PACER or a federal-focused vendor.
Why Does the Two-Court-System Split Matter for Credit Decisions?
The United States runs parallel court systems, and the division is not cosmetic. Congress assigned certain categories of dispute to federal district courts by statute, sometimes exclusively. That means a class of borrower distress signal exists that a state court search will never surface, no matter how thorough the state search is or how many counties it covers.
What size is the federal civil docket?
Federal civil case filings reached 303,563 in the statistical year ending March 31, 2025, a 4 percent increase, within total district court filings of 382,692.[1] Inside that number, federal question cases climbed 12 percent to 157,421 while diversity of citizenship cases fell 7 percent to 96,548.[1] Those two buckets matter to a credit file for different reasons, and the section below separates them.
State civil dockets are far larger in raw volume, which is exactly why the federal layer gets skipped: it looks small. But federal filings are selected for size and severity by the jurisdictional rules themselves, so the hit rate for a distressed commercial borrower is disproportionately high relative to volume.
Which filings can only ever appear in federal court?
The categories below are federal by statute. A state court index will not contain them:
• Bankruptcy petitions. District courts hold "original and exclusive jurisdiction of all cases under title 11," which covers Chapter 7, Chapter 11, and Chapter 13.[2]
• ERISA civil enforcement. Federal district courts have exclusive jurisdiction over most ERISA civil actions, with a narrow concurrent carve-out for benefit denial claims under section 502(a)(1)(B).[3]
• Federal government contract disputes. The Contract Disputes Act is the exclusive route for disputes on a government contract, and appeals run to an agency board or the Court of Federal Claims, not to a district court and not to a state court.[4]
• Federal securities class actions. These are overwhelmingly federal filings, tracked separately from state activity by the research bodies that count them.[5]
• False Claims Act cases. Enforcement actions against lenders and vendors handling federal program money are brought in federal district court, as the Paycheck Protection Program settlements demonstrated.[6]
What crosses over between the two systems?
Diversity jurisdiction is the overlap zone. A commercial dispute between parties from different states, where more than $75,000 is at stake, can be filed in federal court even though nothing about the claim is federal.[7] It can also be filed in state court and then removed to federal court by the defendant.
That optionality is the practical reason a state-only search produces false comfort. A supplier suing your applicant for $2.4 million in an ordinary breach of contract case may sit in a federal docket purely because the parties are in different states. Nothing about the claim announces itself as federal.
Why Is Bankruptcy the Clearest Federal-Only Credit Signal?
If a lender adopts only one federal check, it should be bankruptcy. The reason is structural rather than analytical: there is no state venue where a bankruptcy petition could have been filed instead. Exclusive federal jurisdiction means a state search returning zero results tells you nothing about whether the applicant is in bankruptcy.[2]
What do current filing trends show?
Total bankruptcy filings reached 574,314 in the year ending December 2025, up 11 percent from 517,308, with business filings rising 7.1 percent to 24,737.[8] The judiciary reported a further 11.9 percent year-over-year increase in the twelve months ending March 2026.[9]
The commercial reorganization picture moved faster than the headline number. Epiq AACER recorded 2,422 commercial Chapter 11 filings in the first quarter of 2026, a 37 percent increase over the 1,764 filed in the first quarter of 2025.[10] Small business filings rose roughly 50 percent year over year in the first half of 2026.[11]
For a portfolio concentrated in small business credit, that is the most consequential statistic here. The segment is deteriorating faster than the overall market, and the evidence lands exclusively in a court system many underwriting stacks do not query.
How quickly does a petition change the risk picture?
Immediately and totally. The automatic stay attaches on filing. Collection stops, and the district court where the case is pending takes exclusive jurisdiction over all property of the debtor as of commencement.[2] A merchant cash advance funded three days before a Chapter 11 petition is not a slow-pay problem. It is a claim in a proceeding, which is why a weekly refresh cadence on bankruptcy data means a week of exposure on every file funded in that window.
What Other Federal Case Types Belong in a Commercial Review?
The rest are situational, and the value depends on the borrower profile in front of you.
When do securities actions matter to a private credit file?
They matter when the applicant is a public company, a subsidiary of one, or a counterparty to one. Plaintiffs filed 207 securities class actions in federal and state courts in 2025, down from 226 in 2024, but the aggregate Disclosure Dollar Loss rose from $429 billion to $694 billion.[5] Fewer cases, larger cases. A securities action against a borrower's largest customer is a concentration risk event even when the borrower is not named.
What do ERISA filings signal?
An ERISA action against a small or midsize employer is often a cash flow tell. Delinquent contribution claims, failure to remit withheld employee deferrals, and fiduciary breach claims tend to appear when payroll is being managed on a shortfall. Because most ERISA civil enforcement sits in exclusive federal jurisdiction, none of it will show in a state index.[3]
Why check federal contract disputes for government-facing borrowers?
If an applicant derives meaningful revenue from federal contracts, a Contract Disputes Act claim can indicate that the receivable you are underwriting is contested at the source. The claim path runs through a contracting officer's decision and then to an agency board or the Court of Federal Claims, with a twelve month filing window.[4] Neither venue appears in a state court search, and neither appears in the district court dockets most people mean when they say "PACER."
Where Are Federal Tax Liens Actually Recorded?
This is the most common misconception in the category, and getting it wrong sends teams looking in the wrong system entirely.
Federal tax liens are not federal court filings
A Notice of Federal Tax Lien is an IRS administrative filing, not a court case. Under IRC 6323(f), the IRS files the notice in public recording offices designated by state law. For real property, that is generally the county recorder or clerk where the property sits. For personal property, it is generally the county clerk where the taxpayer resides. In the few states without county recording systems, the notice goes to the secretary of state's office.[12]
So the federal government's most common commercial credit signal is discoverable at the county and state level. Searching PACER for it produces nothing.
What this means for search design
Two consequences follow. First, "federal" as a data category does not map cleanly to "federal courts." Second, a county and state layer picks up federal tax liens even though it will never pick up a bankruptcy petition. The layers are not substitutes in either direction.
How Do You Search Federal Court Records, and What Does It Cost?
Federal access runs through Public Access to Court Electronic Records. PACER charges $0.10 per page, capped at the 30 page equivalent of $3.00 for documents and case-specific reports such as docket reports and claims registers. Audio files are $2.40 each. No fee is owed until an account accrues more than $30.00 in a quarterly billing cycle, and judicial opinions carry no fee.[13]
How do you find a case when you do not know the district?
The PACER Case Locator is the national index across district, bankruptcy, and appellate courts, and it supports nationwide party name searching.[14] It is the correct starting point for a defendant search on a business name.
One operational caveat that matters for pre-funding checks: the Case Locator is built from data collected from the courts and transferred nightly, so it is a daily snapshot rather than a live view of every court's docket.[14] For a same-day funding decision on a borrower whose petition was filed this morning, that lag is real. Direct district-level lookups are more current than the national index.
What makes PACER hard to operationalize
• Name matching is literal. PACER matches party strings, so an entity that files under a slightly different legal name, a former name, or a d/b/a will not surface on a single query. You need name variants.
• There is no risk layer. PACER returns dockets. It does not tell you whether a case is a $4,000 collection matter or a $4 million judgment, and it does not score anything.
• Pagination drives cost. The per-page model means broad exploratory searching across many applicants accrues charges quickly at volume, even with the $30 quarterly threshold.
• Document review is manual. Determining whether a judgment was entered, satisfied, or vacated generally requires opening documents, not reading a docket line.
• Coverage is federal only. Everything filed in state court, which is the large majority of commercial collection activity, is outside PACER entirely.
What Are the Realistic Options for Building Court Coverage?
Before naming any vendor, it helps to be clear about what an underwriting team is actually buying. It is not access to records, because the records are public. It is the removal of manual work between a loan application arriving and a risk answer landing in the credit file.
Several routes exist, and each solves a different part of the problem:
• PACER directly. Authoritative and cheap per page for federal records. Manual, per-page billed, and federal only.[13]
• Unicourt. Aggregates state and federal dockets with API access. Broad, and priced accordingly.
• LexisNexis and Westlaw. Deep legal research coverage built for litigators rather than for automated pre-funding checks.
• CSC and Wolters Kluwer. Strong on corporate records, UCC, and registered agent services, with litigation coverage as an adjacent offering.
• Manual courthouse and clerk portal search. Free at the source and accurate, and completely unscalable past a few files a day.
• State portals directly. New York's unified court system and the Miami-Dade clerk both publish searchable case data at no charge, with no API and no integration path.
One MCA operator running high daily volume described the resulting stack plainly in a product call: "we run New York court separately and then we run Unicourt, and then we run UCC searches." That is three tools and three interfaces for one question, at 500 files a day.
The friction is integration, not availability. This is the specific problem an API addresses, and it is the only claim worth making on Cobalt's behalf.
Where Does Cobalt Fit, and Where Does It Explicitly Not?
Cobalt Intelligence's Court Case API covers New York State courts and Miami-Dade County, Florida. That is the entire coverage list. It is not nationwide, it is not multi-state, and it includes no federal court data of any kind. There is no PACER integration, no district court coverage, no bankruptcy petition data, and no appellate coverage.
This is worth being blunt about because the honest version of the answer is more useful than a hedged one. If you need to know whether an applicant has filed for Chapter 11, Cobalt cannot tell you. Use PACER, the Case Locator, or a federal-focused vendor. A clean Cobalt court result is not evidence of the absence of a federal case, and it should never be recorded in a credit file as if it were.
Why only two jurisdictions?
The coverage is demand-driven rather than aspirational. Roughly 80 percent of Cobalt's funder customers file judgments in exactly these two places. New York is the center of alternative lending, and Miami-Dade became the second hub after a large migration of funders to South Florida during the COVID period. Broader jurisdictions are on the roadmap, including a planned human-assisted queue for unsupported jurisdictions that would run slower, in the range of an hour, and has not yet shipped.
Customers say this back plainly, including when it does not work for them. One construction technology buyer wanted exactly this data shape: "are there any open liens against them? Any litigations associated to this specific contractor?" Another prospect, Elementix, reached the opposite conclusion and said so: "We'll stay away from the court stuff then. I wish the court stuff was rounded." Lara Hodgson at RoxWrite put the same objection in one line: "Most of our clients are not in New York."
Those are the correct reactions if your book is national. Nothing in this article argues otherwise.
Underwriting in New York or South Florida and want the state layer automated? See what a live court search returns on your own applicant names before you build anything. Book a technical walkthrough and bring five real business names.
What does a request look like?
The endpoint is asynchronous. A callback URL is required, there is no synchronous mode, and typical completion runs 30 to 120 seconds because the data is pulled live from the court site rather than served from a cache.
curl -G "https://apigateway.cobaltintelligence.com/courtCases" \
-H "x-api-key: $COBALT_API_KEY" \
--data-urlencode "businessName=Sunrise Logistics LLC" \
--data-urlencode "jurisdiction=newYork" \
--data-urlencode "callbackUrl=https://your-app.example.com/hooks/court-cases"Valid `jurisdiction` values are `newYork`, `miamiDade`, `testNewYork`, and `testMiamiDade`. The two test modes run without consuming credits, which is the right way to build and test the callback handler before spending anything. A production lookup costs one credit, the same as a Secretary of State lookup. Results include judgment details, case number, case type and division, filing dates, and parties, with amounts where the underlying record includes them. Not every court record carries an amount.
One funder framed the economics in terms of an automatic run rather than a manual one: "If courts are cheap enough, then it's worth it to run on every application automatically." That is the design the credit-per-lookup model is meant to support, within the two jurisdictions that exist.
How Should a Team Structure a Two-Court-System Search Stack?
The goal is a documented stack where each layer has a stated scope and a stated gap, so the credit file records what was checked rather than implying everything was.
What does the layered sequence look like?
• Layer 1, entity existence. Secretary of State status and registration history, all 50 states, live pull. Establishes that the applicant exists and is in good standing.
• Layer 2, secured position. UCC filings to identify existing liens and stacking. Cobalt returns UCC data for roughly 10 to 11 states in the same call as SOS, billed as a separate credit.
• Layer 3, state and county litigation. Judgments and civil filings in the venues where your borrowers actually get sued. Cobalt covers New York and Miami-Dade; other venues need another source.
• Layer 4, federal. PACER Case Locator for a nationwide party search, then the specific district for current docket detail. This layer is where bankruptcy lives, and no state layer substitutes for it.[14]
• Layer 5, tax liens. County recorder and secretary of state offices, because federal tax lien notices are recorded locally rather than in federal court.[12]
• Layer 6, identity and sanctions. TIN and EIN verification across all states, plus OFAC screening globally.
New York is the one state where Cobalt's SOS, UCC, contractor license, and court records coverage all overlap. If your book is concentrated there, the automated portion of the stack is unusually deep. Everywhere else, the court layer needs a different source, and the federal layer always does.
How do you record coverage gaps in the credit file?
Write the scope of each check into the record itself, not just the result. "Court search: clear" is a misleading file note. "NY state court search: no matching filings; federal not searched" is an accurate one, and it is the note that protects the underwriter in a later review when a Chapter 11 petition surfaces that was never within the scope of the search performed.
The same discipline applies to timing. Note the date and time of the federal check, because the national index refreshes nightly and a same-day petition may not appear.[14]
What should you automate first?
Automate the layer with the highest query volume and the most stable data shape, which is usually entity status and UCC. Add the court layer where your concentration justifies it. Keep federal searches manual and targeted at first, triggered by deal size, industry, or a state-layer red flag, rather than run on every application. The per-page fee structure rewards targeted federal queries and penalizes broad exploratory ones.[13]
For more on how the two systems compare across a live underwriting workflow, see Federal Court Records vs State Court Records: What Lenders Should Check and PACER vs Cobalt Court Case API: A Lender's Comparison. For the bankruptcy signal specifically, How Court Records Reveal Bankruptcy Risk Before Funding covers the pre-funding workflow in more detail.
Public enforcement history reinforces why documented scope matters. The Kabbage settlements, resolving False Claims Act allegations over Paycheck Protection Program loan processing for up to $120 million, moved through federal district court in exactly the venue a state-only search would never reach.[6] A lender's own regulatory exposure can originate in the same court system it declined to search.
Ready to see what the state and county layer returns on your book? Talk to the Cobalt team about a New York and Miami-Dade test run, and keep your federal coverage exactly where it belongs.
.jpg){kind=small}
.png)