Start with arithmetic rather than anyone's motives.
Say a rule tells a firm it may keep at most twenty cents of every dollar it charges. The rest has to go on the thing the customer is buying. That looks like a limit, and it is one. The firm can never keep more than a fifth.
Now let the underlying cost double. Eighty cents becomes a dollar sixty, and the price rises to two dollars. The firm still keeps a fifth. But a fifth of two dollars is forty cents. The percentage held. The income doubled.
That is the chapter in one paragraph. A rule that caps a firm's take as a percentage of what it charges makes that firm's income a function of the cost it was meant to control.
The point is not contested. In health insurance the rule "limits the percentage of insurers' income from premiums, though not the total dollar amount, that they can retain as profits."1 It caps margins, not levels.1
Three percentages in the injury system are written this way. It is worth being exact about what each one sits on. The lawyer takes a share of the recovery, which is paid by the at-fault driver's insurer. The car insurer's rate divides expected losses, claim-handling costs and fixed expenses by one minus two percentages — expenses that vary with premium, and target profit — so the permitted profit is a mark-up on the loss cost.2 The health insurer must spend at least a fixed share of premium on care, which caps what it may keep at a fixed share.1
Three different bases, and three different payers. What they share is an input. The medical bill sits inside all three.
Where the geometry holds, and where it does not. Doubling the loss cost doubles the permitted profit dollars. It also roughly doubles the premium written and the reserves held, and a car insurer's required capital scales with both. Profit dollars up, capital up, return on capital roughly unchanged. So the incentive is strongest where the input does not scale. The lawyer's hours do not double when the recovery doubles. A health insurer's administrative cost does not rise one-for-one with claims. The clean case in car insurance is not the profit provision at all. It is the agent's commission — "typically … paid as a percentage of premium written," with no capital standing behind it.2
One more difference matters. The lawyer's share is a contractual percentage of money actually received. The insurer's is an assumption inside a rate filed before the losses happen. When losses come in above expected, the insurer's profit falls rather than rises.
One honest limit before going on. The arithmetic is not in dispute. It follows from formulas the industries publish. Whether it changes behavior is a different question. In health insurance the effect has been measured. In car insurance this research found no property-casualty source drawing the conclusion in print, and one legal scholar arguing that it does not carry across: health insurers negotiate the prices they pay providers, and car insurers do not, so the effect measured in health insurance may be specific to health insurance.3 That is a strong argument. On this one point the contrary view is currently better supported than this chapter's. Read the arithmetic as established and the incentive as an argument.
What follows takes each of the three in turn, and then the two places where the arithmetic is most visible: a fee statute that made fighting a claim more expensive than settling it, and a margin cap that did exactly what the first section describes.
Florida is one of very few states anywhere that puts numbers in its fee rule. Of the seven southern states checked here, every other one uses a bare reasonableness standard, with no percentage anywhere.4
The scale runs 33⅓% of the first million, rising to 40% once the defendant files an answer. Above a million it drops to 30%, and above two million to 20%.4 Three things about it get routinely misdescribed.
It is not a cap. The rule says a fee above those levels is "presumed, unless rebutted, to be clearly excessive," and a client may petition a court to approve more.4
The step to 40% is triggered by the defendant's answer, not by the plaintiff filing.4 The claimant does not control it and may never be told it happened.
The headline overstates the take. On a $1.5 million recovery after an answer the fee is $400,000 plus $150,000 — 36.7% blended, not 40%.
Two deductions then matter more to the household than the percentage does.
The costs ordering. All seven states checked here require the agreement to say whether case expenses come out before or after the fee. None says which.5 On a $90,000 recovery with $15,000 of costs at 33⅓%, costs first leaves the client $50,000 and fee first leaves $45,000. A $5,000 swing decided by one sentence.
The liens. Medicare shares the cost of recovery in proportion. But where its payments equal or exceed the settlement, the lawyer and the costs are paid in full and the claimant's share is zero.6 Florida Medicaid presumes 37.5% of the whole recovery is medical.6 A well-drafted employer health plan can recover in full with no contribution to the fee — so the household funds the entire legal cost of recovering money that goes to the plan.6
This research found no measurement of what share of the system's money reaches claimants. The familiar figure of roughly half comes from a business-funded study that estimates claimant legal costs as defense costs times four-thirds, using a ratio from cases ending in 1985; the 2024 edition dropped the estimate. The best-pedigreed figure is higher — about 57%, from a state regulator's mandatory closed-claim database.7 The document that would settle it exists in every Florida file. It is the closing statement, and no bar located here aggregates them.7
Two numbers reach a jury that nobody ever paid.
The first is the demand. A $5,000,000 lump-sum demand raised the average award by about $1.37 million against no demand — a 276% increase on a baseline of $496,389, in a controlled experiment.8 Read the denominator: the effect is measured on the mock jurors who found liability, not on everyone recruited, and awards above $5 million were compressed down to $5 million, so the top of the range is truncated.8
The same experiment tested the argument lawyers are most often accused of: breaking a lifetime of pain into minutes and pricing each one. On the size of the award it found nothing significant. The same test did find the argument moved whether the plaintiff won at all, and its authors report that null as underpowered rather than as a zero.8 Several state supreme courts banned the argument in the 1960s, one calling it "an illusion of certainty" producing "any amount that the imagination of counsel deems advantageous." Others expressly permitted it, and the state that led the ban later reversed itself.9 What moves the number is the plain large demand.
And the effect is far weaker in real courtrooms. Before real deliberating juries in 31 cases, jurors were critical consumers of demands, reacted more negatively to pain-and-suffering figures than to requests grounded in evidence, and extreme demands "may not exert substantial undue influence."10 Awards also have structure. Injury severity predicts them strongly and repeatedly.10
The reason a number in the air has purchase at all is a vacancy, not a trick. From a federal appellate court: "There is no rational scale that justifies the award of any particular amount, as opposed to some very different amount, in compensation for a particular quantum of pain."11 Georgia's 2025 statute codifies the same idea, calling the measure "the enlightened conscience of an impartial jury."11 Juries are generally not allowed to hear what comparable cases produced.11 So the only number in the room is the one counsel supplies.
The second number nobody paid is the medical bill. Three measurements circulate and they answer three different questions. Hospital charges ran about 3.4 times Medicare-allowable costs in 2012. Private health plans actually paid about 254% of Medicare rates in 2022. And in one California case the bills came to $189,978.63 while providers accepted $59,691.73 as payment in full.12 Only the third has the right shape, and its sample is one case. This research reached no measured distribution of billed against accepted amounts for the injury system.
Florida and Georgia both legislated here, differently. Florida limits evidence of satisfied past medical damages to "the amount actually paid, regardless of the source of payment," with benchmarks of 120% of Medicare for unpaid charges of uninsured or publicly insured claimants — while keeping billed amounts admissible alongside them.13 Georgia set no benchmark and sends both numbers to the jury.13 Both forced disclosure of what a bill was really sold for, including the price a debt buyer paid; Florida overrides lawyer-client privilege where the claimant's own lawyer made the referral.13 Both reached for case-by-case discovery. Neither had a measured distribution to set a benchmark against.
This section is load-bearing, not defensive. A chapter arguing everyone takes a percentage of an inflated number cannot itself rest on an inflated number.
Almost nothing reaches a verdict. In the last national random-sample measurement, about 4% of tort dispositions were resolved by trial.14 Everything reported as a verdict comes from that 4%, and cases reach trial precisely when the two sides disagree most.
Inside that 4% the median is small. Half of plaintiffs who won at trial were awarded $24,000 or less; for auto accident cases the median was $15,000.14 Read the denominator carefully. That is the median among winners at trial — roughly half of the 4% that got there — not the median outcome of a claim, which is lower. It is also in 2005 dollars, and general prices have risen by more than half since, so it cannot be set against the 2022 claim averages later in this chapter.
The direction of travel is downward. Adjusted for inflation, median jury awards in the largest US counties fell 40% across all civil cases between 1992 and 2005, 53.5% in tort cases and about 60% in auto accident cases.14 Tort trials in those counties fell from 10,278 in 1996 to 7,038 in 2005.14 Across sixteen states reporting comparable 2024 data, the share of tort cases resolved by jury ran from 0.0% to 2.3% — figures this research could not check against the primary dashboard.15
The counts of very large verdicts have their own problems, and they drive the headlines. Earlier years came from a list capped at the hundred largest awards a year and later years from an uncapped database, so the series partly measures its own method changing.16 Totals are gross awards before appeal, including verdicts later overturned in full. The $10 million threshold defining a "nuclear" verdict has never been indexed for inflation. Auto accident is the cheapest large category on the industry's own data, both leading reports draw on the same commercial vendor, and the industry's own median verdict fell in real terms.16
One absence sits underneath all of it. The federal survey that produced the trial figures is listed inactive, and 2005 is its latest data year.14
Two figures belong here, and they are ones this chapter would rather not have. Florida's average bodily injury claim in 2022 was $28,243, against a published countrywide average of $28,919.17 That countrywide figure carries a defect. Texas's losses sit in the numerator while its claim counts are missing from the denominator, and correcting it puts Florida about 9%–10% above the national average rather than just below it.17 So the honest reading is that most of Florida's excess is how often a claim is made and some of it is what a claim is worth — roughly 85% and 15% of the gap.17
Car insurance rates are built from the loss upward.
The standard method takes expected losses and claim-handling costs, adds fixed expenses, and divides by one minus two percentages: expenses that vary with premium, and target profit.2 The professional standard defines the profit provision as "the provision for underwriting profit in the actuarially developed rate, typically expressed as a percentage of the rate."2 Because the profit percentage sits in the divisor, the rate is a mark-up on the loss cost. Regulators write the same shape into their rules. California divides projected losses by one minus an expense standard minus a profit factor.18 Florida treats a profit allowance above five percent, less any contingency factor, as prima facie excessive for motor vehicle insurance.18
Now take what that has actually produced, because it is not profit.
Insurers have not earned the margin. Across 2014 to 2023, US private passenger auto ran an average underwriting result of minus 2.4% of direct premiums earned, negative in seven of ten years. Over the same decade the whole property and casualty industry returned 6.0% on its own capital against an estimated 14.9% for all industries.19
And the line this chapter is about lost money in the state this chapter uses. Over 2020 to 2022, 96.0 cents of every dollar of Florida bodily injury liability premium went out as losses, against 82.2 cents countrywide — before any cost of running the book. Florida was third on that table, and four southern states sat below the countrywide figure. That is pre-reform data and should not be read in the present tense.19
Competition is why. The wish of all insurers to earn that result "creates intense competition, so vigorous in most years that it causes the P/C industry as a whole to operate at a significant underwriting loss."20 A filed target margin and a realized one are different things.
The largest input to the bill has no lawyers in it. Between 2019 and July 2026 the price of vehicle maintenance and repair rose 55.5%. Motor vehicle insurance rose 49.9% — less. And insurance prices are now falling: down 4.46% over the year to July 2026, after peaking in February 2026.21 Every premium figure in this chapter is calendar 2022 or 2023, which is the newest regulator-grade data available. Consumer prices have moved since.
Some pay is tied to low losses. Contingent commissions vary an agent's pay by loss ratio.2
And an insurer does not want its loss costs to rise. Four reasons, none of which the arithmetic above touches. A rate is filed before the losses happen, on expected losses; when losses come in higher, the extra lands on business already priced and sold, with no premium behind it. The higher rate may not be granted, because Florida and California both require approval. Volatile costs raise the price of the capital held behind the book. And dearer cover sells fewer policies — raising Florida's minimum was modeled as pushing about a quarter more drivers out of the market, as section nine sets out. There is a published finding that premiums follow falling claim costs only after a lag, so a carrier can hold the gain for a while. But it was measured in a different line in a different era, and the same literature notes that car insurers use short policy terms and re-price often, which is what would close the lag.22 Treat the lag as unproven for car insurance, and read the incentive as running in both directions.
One more feature is worth naming, because two states rule opposite ways on it. Insurers collect premium before paying claims and invest the difference. California credits that investment income back against the permitted premium, including the return on surplus. Florida requires it considered too, but says "investment income from invested surplus may not be considered."20 Same question, two live answers. These percentages are policy choices, not physics.
For decades Florida had a rule with no reciprocal half. If a policyholder recovered anything against an insurer, the court "shall adjudge or decree against the insurer and in favor of the insured … a reasonable sum as fees."23 The insurer had no matching claim if it won.
Here is the fact most often got wrong. The fee came from the insurer, on top of the recovery. It did not reduce what the claimant kept.23
Alongside it ran a multiplier. A court could take hours times rate and multiply by up to 2.5, with the multiplier rising as the case had looked less likely to succeed. In 2017 the state's highest court refused to adopt the federal limit confining multipliers to rare cases.23 Because the fee was built from hours, it was not bounded by the size of the claim.
The economics are well understood in theory. A defendant facing even a weak claim must spend money to defend it, so settling for up to the cost of defending is rational whatever the merits.24 A one-way rule raises that ceiling, because losing costs the defendant both sides' fees.
Three qualifications stop this becoming a caricature. The insurer could cap its own exposure: under a 2006 decision the insured only prevailed by beating the insurer's own earlier settlement offer.23 It barely applied to car insurance, since the fee statutes reached the no-fault law "only in a limited manner."23 And repeal moves to the American rule, not the British one. Under two-way shifting a winning defendant recovers its costs and the nuisance dynamic disappears. Florida abolished the statutes outright in March 2023, leaving each side to bear its own fees — removing an aggravating factor rather than the mechanism.24 That mapping is this chapter's reading, not the model's own claim. The same bill imposed a strong presumption that hours times rate is enough, barred recovery by anyone more than 50% at fault, and halved the deadline to sue for negligence from four years to two.25
For scale on the other side, here is the one measured figure in this chapter that supports the litigation-cost story. Florida's personal auto liability loss adjustment expense ran 13.0% of premium against 11.4% countrywide. Adjust for Florida's mix of business and the excess is about 1.15 points, which on Florida's $20.25 billion of direct liability premium is roughly $233 million a year.26
What happened after repeal is genuinely mixed, and nobody has isolated the cause. Lawsuit filings against Florida personal residential property insurers fell 23%, 25% and 25% across 2024, 2025 and early 2026.27 But Florida circuit courts recorded 48,206 auto negligence filings in 2024-25 — about 40% above the pre-reform run rate, two years on.27 Texas's regulator, ordered by its legislature to measure whether a 2021 commercial auto reform lowered premiums, asked every filing insurer and reported that none attributed any rate change to it.27 Louisiana's rates rose four years running after its 2020 reform, then fell 5.8% in 2025, which the Commissioner attributed to lower accident frequency rather than to the reform.27 The peer-reviewed evidence finds no measurable auto premium effect from damage caps, joint-and-several changes or bad-faith reform — and an effect from only two things: repealing no-fault, and relaxing the rules on which medical bills may be shown.27
The clearest live example of a percentage cap sits in health insurance, and it has been studied properly. It reached what it was aimed at, which is what insurers keep. It did not touch what medical care costs, which it was never aimed at.
Federal law requires an insurer to spend at least 85% of premium on care in the large group market, and 80% in the individual and small group markets. Those are floors; states may set them higher.28 The denominator is not gross premium — it is premium after the insurer's taxes and regulatory fees, and after risk adjustment, risk corridors and reinsurance.28 The numerator is claims plus spending on quality improvement, which ran 0.7% to 0.9% of net premiums in the most recent period this research could source a published figure for. So in practice the numerator is close to claims.28
Miss the threshold and the insurer rebates the shortfall. For reporting year 2024 that came to about $1.64 billion. Since 2012 the total is $14.4 billion.29
The measured effect is the point. Medical claims rose almost one-for-one with an insurer's distance below its threshold when the rule took effect in 2011 — 7% in the individual market, 2% in the group market. Premiums were unaffected.1 So the harm, if any, ran through total medical spending rather than through premium. The same mechanism is described with a qualifier most people drop: insurers responded through utilization, not price, by "decreasing their utilization management … allowing enrollees to use more services than they would otherwise."30
Now the evidence against. Most insurers already met the standards before the rule bound — 76.4% in 2011 and 79.1% in 2012, at a median around 88% against an 80% floor.31 But it bound hard exactly where the effect was measured: in 2010 only 44.3% of individual-market insurers met the standard, rising to 63.2% in 2011.31 It also did its job there. For-profit individual-market insurers cut both administrative cost ratio and operating margin by more than two points each.31 And in 2012 that market's premium surplus was negative 1.1%. Insurers lost money there in the year the constraint bit hardest.31
There is one more route around a percentage cap, and the regulator has named it.
The rules count what a plan pays for covered services in the numerator, including capitation and provider bonus payments. Nothing asks whether the payment went to an affiliated provider, or whether the price matched what an unaffiliated one would get. The test is what the payment was for, never who was paid or how much.32 In December 2024 the regulator wrote that payments to related parties "may, in some cases, be inflated to ensure" a plan meets its spending requirement, "obscuring the actual … profits made by the integrated system as a whole."32 It floated six fixes and adopted none in either final rule that followed, while reserving the right to act later.32 The equivalent test already exists in Medicaid.32
The size of that surface sits on the companies' own filings. One large health group removed $167.96 billion of revenue on consolidation in 2025 because one part billed another, and 55.9% of its adjusted operating earnings came from the non-insurance side. At a second, 81.8% of segment adjusted operating income came from non-insurance segments; at a third, 63.5%.33 These are margins on segment revenue, not loss ratios, and must not be set beside an 80% or 85% figure as though comparable.33
Against that, Congress's own Medicare advisory commission finds the large national insurers are less vertically integrated than provider-owned plans, not more.34 The estimate the regulator itself cites puts the effect at 1.3 loss-ratio points per ten-point rise in related-business share, and calls its own evidence "indirect, incomplete."34 The strongest causal work found affiliated-pharmacy prices rose 9.5% after profit caps took effect in Medicare Part D — but its authors estimate only about 5% of profits moved, and note the requirement was "rarely binding."34 Nothing published shows that most of the profit walked through the door.
Each of these is tied to a fact above, not to a general principle.
1. Negotiate the costs ordering, and ask which fee tier the lawyer expects. On a $90,000 recovery with $15,000 of costs, the ordering is worth $5,000 to you, and the rule sets no default.5 The step from 33⅓% to 40% is triggered by the defendant's answer, not by your filing.4
2. Demand and read the closing statement before any money moves. Until you approve it, your lawyer cannot pay money to anyone, including you, without a court order. You also have a written right to bargain over the fee and three business days to cancel.7
3. Treat a pre-settlement advance as the most expensive money you will ever borrow, and check the lien before you settle. In the largest dataset located here the median advance was $2,250, the median owed $4,849 and the median repaid $3,380 over 417 days — about 43% a year, with roughly a third paying the full contract.35 And where Medicare's payments equal or exceed the settlement, your share is zero.6
What makes a percentage hard to notice is that it arrives inside a total. A premium is one number. A settlement is one number. Plenee can hold what you actually pay in one place and set it against everything else in the month, which is where the question of what a percentage is costing you has to start.
Cap a firm's take as a percentage of what it charges and you have tied its income to the cost you were trying to control. The lawyer's percentage, the car insurer's permitted margin and the health plan's allowed 15 or 20% are all written this way. They sit on three different bases with three different payers. What they share is one input: the medical bill. The arithmetic is certain. The behavior is measured in health insurance and argued here for car insurance, where required capital rises with the loss cost too and blunts it — the clean case there is the agent's commission, a percentage of premium with no capital standing behind it. Two of the numbers that reach a jury were never paid by anyone. And verdict headlines are a poor guide to what claims cost, because only about 4% of tort dispositions reached trial, the median award to a winner was $24,000, and inflation-adjusted medians fell by half between 1992 and 2005 — the last period anyone measured. What you control is the costs ordering you negotiate, the closing statement you insist on reading, and never borrowing against a settlement.
Plenee Academy provides financial information and education, not personalized financial advice. Plenee Co. is not a registered investment adviser, broker-dealer, or financial planner. Legal Disclosures & Notices →