New Jersey PIP arbitrators awarded 40.70 percent of claimed dollars in 94,137 public decisions on the checked basis

How much of the claimed money do NJ PIP arbitrators award?

On the checked basis, arbitrators awarded $498,051,301.99 of the $1,223,794,561.82 claimed, which is 40.70 percent of claimed dollars, across 94,137 decisions read. Counting every decision with a table gives 42.45 percent, so the basis matters. Both figures are published, with the year and carrier ranges each one produces.

The checked basis is the headline. The other basis counts every decision whose table gave paired dollar values, including those that failed a check and the one left out after the review. Neither basis says who was right on any issue.

By count, an award in part is the most common outcome, at 52.40 percent of decisions read. The dollar share weights each decision by its claimed amount, so a large claim moves it more than a small one.

How is this different from Forthright's quarterly report to the state?

Forthright's quarterly report to the New Jersey Department of Banking and Insurance covers one quarter, counts every concluded case and includes legal fees and costs in the awarded amount. This report reads only the benefits table of each published decision from January 2020 to 8 May 2026. The bases differ, so the two should not be compared share for share.

The figures in this section are that report's own, for the first quarter of 2026 only, from the Forthright quarterly report for the first quarter of 2026 (NJ Department of Banking and Insurance, 2026). Of 13,735 cases concluded that quarter, 2,923 ended in an award or denial, 31 were dismissed before a hearing, 9,345 settled and 1,436 were withdrawn. By the report's own percentages, in-person cases were 24 percent awarded or denied, 66 percent settled and 10 percent withdrawn. On-the-papers cases were 13, 75 and 12 percent.

So in that quarter most concluded cases ended without a decision, and only decisions are in this report. Settled and withdrawn cases may differ from decided ones in size, issue or strength, and this data cannot show how. The shares on this page describe the cases that went to decision, not every case filed.

The report's table of awards as a percentage of the claim amount covers 3,240 decided cases, a different count from the 2,923 above that the report does not reconcile. It runs from denied or dismissed claims up to full awards, and states that "Total awarded amounts include PIP benefits, legal fees, and costs." This report counts only the medical expense benefits table, without fees or costs, so its 40.70 percent is a different measure.

Did the awarded share change between 2020 and 2026?

On the checked basis, the share of claimed dollars awarded was 44.11 percent in 2020 and 38.16 percent in 2025. The 2026 figure, 38.20 percent, covers decisions only through 8 May 2026 and is partial. Counting every decision with a table gives a different pattern, and this report shows no cause for either.

Decisions awarded in full were 23.62 percent of decisions read in 2020 and 21.15 percent in 2025. Decisions awarded nothing were 23.77 percent in 2020 and 26.77 percent in 2025. Claim size changed over these years, as the claimed dollars and decision counts in the table show. The share by claim size is in its own section below. No issue or provider mix is measured.

How did the share move month by month?

Monthly shares on the checked basis ran from 29.08 percent in September 2025 to 48.64 percent in August 2020, across 77 award months from January 2020 to May 2026. May 2026 is partial, through 8 May. The yearly figures in the table above are the steadier read, because a month holds far fewer decisions than a year.

Month-to-month movement reflects the mix of cases the program decided in each month, and the data cannot say why one month differs from another. In 16 of the 77 months, counting the decisions that failed a check moves the share by more than 1 point.

What does counting every decision with a table change?

Counting every decision with a table instead moves three years by more than 1 point. The share for 2022 reads 47.28 percent, which is 4.70 points higher. The share for 2023 reads 35.47 percent, which is 4.12 points lower. The share for 2024 reads 49.10 percent, which is 9.20 points higher. No other year moves by more than 1 point.

On the checked basis the 2023 share is 39.59 percent. With the one decision left out after the review added back, it would read 35.36 percent (see the section on that decision below).

Does the awarded share differ by claim size?

Yes. On the checked basis, decisions claiming under $1,000 were awarded 50.05 percent of claimed dollars, and those claiming $1,000 to $4,999 were awarded 51.94 percent. The three larger bands ran from 38.12 to 41.56 percent. Each decision sits in one band by its own claimed total, and the bands add up to the 94,137 decisions read.

Small claims were the ones most often decided all or nothing. Under $1,000, 35.30 percent were awarded in full and 38.92 percent were awarded nothing. From $5,000 up, most decisions were awarded in part, 70.75 to 80.95 percent.

Counting every decision with a table moves three bands by more than 1 point. Under $1,000 the share reads 56.73 percent, which is 6.68 points higher. For $1,000 to $4,999 it reads 53.39 percent, 1.45 points higher. For $100,000 and over it reads 41.86 percent, 2.48 points higher. The other two bands move by less than 1 point.

Two limits apply. The band is set by the amount in dispute when the award issued, not by the original bill. A band's share describes the decisions that landed in it, not what a new claim of that size will get.

How do outcomes differ by carrier?

On the checked basis, the share of claimed dollars awarded runs from 32.59 percent to 46.94 percent across the 14 named carrier groups. Counting every decision with a table, the named groups run from 35.47 percent to 54.92 percent. The table lists groups by decisions read, not by outcome. These figures describe outcomes, not carrier conduct.

NJ PLIGA is the New Jersey Property-Liability Insurance Guaranty Association, a statutory body that also administers the Unsatisfied Claim and Judgment Fund, which handles personal injury protection and other claims for eligible uninsured motorists and pedestrians. It is not a private insurer. The association explains its role on its FAQ page.

What the carrier table does not show

The mix of providers, services, claim sizes and issues in each carrier's claims can move these shares. The claim-size breakdown above covers all carriers together, not each carrier, and no issue or provider mix is measured, so this report cannot separate that mix from any other cause. No carrier is ranked here, and no row says that any carrier acts well or badly. A share of claimed dollars measures money claimed and awarded, not any carrier's conduct.

The share for each group depends on which decisions are counted. Counting every decision with a table moves six named groups by more than 1 point, by 1.51 to 9.16 points, up or down. Every other row moves by less than 1 point. The difference between the two bases comes mainly from a small number of decisions whose printed awarded amount is far above the printed claimed amount. We draw no conclusion about any carrier from it. The exception is one carrier group, where most of the difference comes from the one decision left out after the review, described below.

How do I read an award for my own case?

Find the medical expense benefits table in the decision and compare its Amount Claimed total with its Amount Awarded total. This report's share is the same ratio, summed over many decisions. Then allow for reductions the parties commonly stipulate, because the amount finally paid can be lower than the amount awarded.

Forthright's quarterly report puts it this way: "It is common practice for users to stipulate that amounts awarded by DRPs are subject to further reduction for deductibles, co-payments, prior payments, and/or medical fee schedule adjustments." Attorney's fees and costs, where awarded, are not part of any share on this page. In our experience, the time from an award to payment varied widely.

If the decision's narrative states a different claimed amount than its table, this report uses the table. In 16 of the 120 largest decisions we checked, the narrative differed, typically an original demand later reduced or amended, or a consolidated case. Where a dispute turned on the fee schedule, the NJ PIP fee schedule lookup shows the fee schedule amount for a code. A share from this page gives context for an award, not a prediction for a case.

How do I find a specific decision?

Search the program's public decisions page, which Forthright runs. It posts the redacted decisions and lets you search them. This report reads that same source but publishes totals only, so it cannot point you to any one decision.

The page keeps adding decisions, so it shows awards issued after 8 May 2026 that no figure here includes.

What can this data not tell you?

The data covers only cases that reached arbitration and produced a public decision, not all PIP disputes and not a random sample of them. Amounts are those in dispute at award time, not final payment. No issue is measured, so the data cannot say whether a decision turned on the fee schedule, UCR or medical necessity.

• No issue categories. In a pilot of 32 decisions, pattern matching read the main issue correctly 50 to 80 percent of the time, so none is published.
• 2026 is partial. Every 2026 figure runs through 8 May 2026 only. The harvest was finished at the end of May 2026, and later decisions are absent.
• Dismissal and modification orders are not spread evenly. Orders of this kind are not counted as outcomes. CURE Auto has about 15 percent of its index rows in that category, against 0.7 to 2.8 percent for every other named group, so its row is not comparable to the others.

How reliable are the carrier names?

• Brand names merged by a rule. The respondent field has 343 spellings. We merged them by a fixed rule written before any outcome was examined. The rule uses a mechanical key, a hand alias table for brand variants, and merging by brand, not by parent company.
• A small carrier pool. A group is named only with at least 1,000 decisions in the whole index. All smaller groups are pooled as all other carriers. So are decisions that name several respondents and at least 10 respondent names that list a provider, hospital or individual beside the carrier.

How concentrated is the arbitrator workload?

The report names, ranks and compares no arbitrator. The only arbitrator figure describes how concentrated the work is. Across 301 arbitrators, the ten busiest handled 24,580 of the 96,973 index rows that name one, which is 25.35 percent.

What each number in this report means

Claimed is the total of the Amount Claimed column in a decision's medical expense benefits table. Awarded is the matching total of the Amount Awarded column. Both are amounts in dispute when the award issued, before deductibles, policy limits, interest, fees and costs, so they are not final payments.

• Awarded in full: the awarded total equals the claimed total.
• Awarded in part: the awarded total is above $0 and below the claimed total.
• Awarded none: the awarded total is $0.
• Other: a table showing $0.00 claimed and $0.00 awarded, which is 451 decisions. They add no dollars to either total. A decision with no table, or with an awarded amount above the claimed amount, fails an automatic check and is counted in the exclusions below, never here.
• Share of claimed dollars awarded: awarded dollars divided by claimed dollars, in percent, to two decimals, from whole cents.
• Year: the year of the award date, which must match the program's index.

What did we leave out, and why?

The headline leaves out 4 index rows that share a file, 1,814 documents that are not awards, 1,035 award documents that failed an automatic check and 1 decision removed after an independent review. Counting every decision with a table instead gives 42.45 percent of claimed dollars, against 40.70 percent. Both figures are published.

The 1,814 other documents are 1,571 clarification, modification or dismissal orders and 243 appellate awards. The four automatic checks apply to the 95,173 award documents with a text layer.

What do the four automatic checks test?

Check 1 requires a disposition table with paired dollar values. Check 2 requires an awarded total that is not above the claimed total. Check 3 requires the heading, where it states a result, to agree with the sums. Check 4 requires the extracted award date to equal the index date. Failures were 561 for check 1, 168 for check 2, 257 for check 3 and 56 for check 4. A decision can fail more than one.

What changes if the failed decisions are counted?

Counting every decision with a table gives 94,612 decisions, $1,252,538,674.28 claimed and $531,702,458.01 awarded, so the awarded share is 42.45 percent, which is 1.75 points above the checked basis. The difference between the two bases comes mainly from a small number of decisions whose printed awarded amount is far above the printed claimed amount. We draw no conclusion about any carrier from it. The 168 check 2 failures are the decisions that print an awarded amount above the claimed amount.

The check 2 and check 3 exclusions we sampled are decisions whose own printed figures contradict themselves; check 1 (561) and check 4 (56) failures were not sampled.

Why was one decision left out after the review?

A reviewer found one decision whose claimed total in its table is exactly 100 times the claimed total in its own narrative, a decimal shift printed in the decision. We left it out of the checked basis under a rule added after the review. With it, the overall share was 39.93 percent. Without it, the share is 40.70 percent.

The decimal shift is printed in the decision itself and says nothing about the carrier. Adding the decision back moves the 2023 share to 35.36 percent, close to the 35.47 percent on the every-decision basis. It also moves the share for the one carrier group it falls in to 37.32 percent, against 37.87 percent on the every-decision basis. That one decision accounts for most of the difference between the two bases for 2023 and for that group.

Why was the exclusion rule one-sided?

Check 2 removes decisions where the awarded total is above the claimed total. Before the review, nothing removed an error on the claimed side. The review found one such decision, and we then read the 120 largest decisions by dollars, blind. Those were the 100 largest claimed totals and the 20 largest awarded totals among the decisions read. The extraction matched the printed table in all 120.

The new rule, which we call check 5, covers that set because every claimed total above $500,000 is in it, and a shifted total below that adds little. We chose it after seeing its effect, which is why both figures are shown. The checked basis may still carry claimed-side errors among smaller claims that we did not find.

How did we check the figures?

Independent AI readers who had never seen the extracted numbers read decisions page by page, in what we call a blind check. On 100 decisions drawn at random, all four fields matched the extraction in 100 of 100 cases. On the 120 largest decisions by dollars, the extraction matched the printed table in 120 of 120.

Our plan had said a person would read a sample, and AI readers did it instead, which departs from the plan. Four readers each read 30 decisions from the rendered page, 120 in all. The 100 random decisions came 14 from each year from 2020 to 2025 and 16 from 2026. We mixed in 10 decisions that failed check 2 and 10 that failed check 3, so readers could not tell which were which. Those 20 were failed-check decisions, not a second random sample. The fields compared were claimed amount, awarded amount, outcome and award date.

How accurate is the reading?

Zero misses in 100 puts the true error rate below about 3 percent with 95 percent confidence, by the rule of three, treating the 100 as a simple random sample (they were drawn a fixed number per year). That bound is by decision count. Dollar-weighted accuracy rests on the 120 largest decisions, read the same way.

What did the 20 failed decisions show?

All 10 check 2 cases print an awarded amount above the claimed amount. Two are exactly 100 times the claim, a decimal shift in the decision's own table. In the 10 check 3 cases the heading and the table disagree. Examples are a heading that says Awarded over $0.00 and one that says Denied over an amount equal to the claim.

The extraction read the page correctly in all 20, so these are inconsistencies printed in the decisions themselves, not misreadings. We also looked at two of the check 2 pages directly and confirmed the printed figures. The check 3 group is the closest call, because the sums are as printed and only the heading disagrees. Those decisions stay excluded under a rule written before the full run, and the 42.45 percent figure shows what including them does.

What did the 120 largest decisions show?

In 16 of the 120, the decision's narrative states a different claimed amount than its table, typically an original demand later reduced or amended, or a consolidated case. We use the table's total throughout. In exactly one, the table's total was exactly 100 times the narrative's, which is the decision left out above.

Sources and further reading

The decisions page is the source of every figure we computed, and the program publishes its arbitration rules on its public rules page. Rules change over time, so read the current version before relying on any procedure.

Forthright's quarterly report for the first quarter of 2026 is the source of the program-wide figures in the section on that quarterly report, and of the stipulation quoted above.

Medicare reference amounts are a separate data set, covered in the articles and tools below. Our article on Medicare fee schedule changes for PI codes covers how those amounts changed. The Medicare rate lookup checks a code and date of service.