Insurance Cases 2026: The Combined Ratio and the Float

The combined ratio an insurance case turns on, and the hinge at 100

The technical account of an insurer has three lines and one ratio.

Incurred losses over earned premium gives the loss ratio. Add loss adjustment expense, which is the cost of settling claims rather than the claims themselves. Then add underwriting expense, which is commission, acquisition and administration. The sum of those percentages is the combined ratio.

A combined ratio of 96 means the insurer keeps 4 cents of every premium dollar on the underwriting alone. A ratio of 104 means it loses 4 cents. The hinge sits at 100, and it is the most quoted number in the industry for good reason.

The ratio hides what generates it, and this is where cases are won. Losses are frequency times severity. Frequency is claims per exposure over a year. Severity is the average cost of a claim. An insurer can hold a loss ratio flat while frequency halves and severity doubles, and the two paths call for opposite decisions.

One more term belongs in the equation and sits outside the ratio. Reserves for unpaid claims plus premium collected but not yet earned make up the float. The float is invested, and its return is a second profit line that can carry a company whose underwriting loses money.

Kestrel Mutual, an insurance case worked from claim counts up

Kestrel Mutual writes homeowners cover in three states. The board wants to know whether the book earns its keep. Here is what it gives you.

1. Policies in force. 400,000, average annual premium $1,450, all of it earned in the year.
2. Claim frequency. 6.2 claims per 100 policies a year.
3. Average severity. $14,200 a claim.
4. Loss adjustment expense. 9.0 percent of earned premium.
5. Underwriting expense. 28.5 percent of earned premium.
6. Float. $870 million of reserves and unearned premium.
7. Reinvestment rate. 4.10 percent, the three month Treasury yield of 28 September 2026, checked 30 September 2026 at federalreserve.gov.

Premium first. 400,000 policies at $1,450 is $580 million of earned premium.

Claim count second, and watch the base. A frequency of 6.2 per 100 policies applied to 400,000 policies gives 24,800 claims. Then 24,800 claims at $14,200 is $352.16 million of incurred losses.

So the loss ratio is 352.16 over 580, which is 60.7 percent. Loss adjustment expense follows: 9.0% on $580 million is $52.2 million, worth 9.0 points. Underwriting expense next: 28.5% on $580 million is $165.3 million.

Add the three. 60.7 plus 9.0 plus 28.5 is a combined ratio of 98.2. The underwriting profit is therefore 1.8 points, and 1.8% on $580 million is $10.44 million.

The float moves the break-even ratio

Now add the second profit line. 4.10% on $870 million is $35.67 million of investment income. Operating profit is $10.44 million plus $35.67 million, which is $46.11 million, or 7.95 percent of premium.

Read those two numbers against each other. The investment income is 3.4 times the underwriting profit. The board was told the book was marginal at 98.2, and the money is actually being made on the float.

That reframes the decision. Ask how far the combined ratio can deteriorate before the whole company loses money. Investment income is worth 35.67 over 580, which is 6.15 points of premium. So the break-even combined ratio is 106.2, rounding the 106.15 the division gives.

At a combined ratio of 104 the underwriting loses $23.2 million and the company still earns $12.47 million. Say that out loud in the recommendation, because a board that hears "the ratio went above 100" will hear a loss that is not there.

The right column carries a warning too. Investment income depends on a rate the insurer does not set. At a 2.00 percent reinvestment rate the float earns $17.4 million, the break-even ratio drops to 103.0, and a combined ratio of 104 becomes a real loss of $5.8 million.

What a dollar of ceded limit costs in 2026

Reinsurance is the line that turns a catastrophe into a survivable year, and its price is one of the few insurance prices published anywhere. The federal flood program discloses every placement. In 2024 it transferred $619.5 million of risk to 18 reinsurers for $121.1 million of premium, checked 30 September 2026 at fema.gov. In 2023 the same structure transferred $502.5 million for $90.2 million.

Divide premium by limit and you get the rate on line. 121.1 over 619.5 is 19.5 percent for 2024. 90.2 over 502.5 is 18.0 percent for 2023. So the cost of buying a dollar of protection rose about 1.6 points in a year, and that is a fact rather than a market opinion.

Apply it to Kestrel. Suppose it cedes $100 million of catastrophe limit at the 2024 rate. 19.5% on $100 million is $19.5 million of ceded premium. Against $580 million of earned premium that is 3.4 points of combined ratio.

Now look at what the comparison shows. Kestrel's whole underwriting profit is 1.8 points. The cession costs 3.4 points, which is close to twice what the underwriting earns. The program is still probably worth buying, and the reason has to be stated in the right currency. The same page records that the program "recovered the entire $1.042 billion in reinsurance under the 2017 traditional reinsurance coverage" after one hurricane season. Reinsurance is bought against ruin, and it is priced out of the annual margin.

One boundary is worth naming while we are on this source. The same program publishes its claim file and its policy file in full, as data sets covering redacted claims, redacted policies and multiple loss properties, checked 30 September 2026 at fema.gov. So frequency and severity can be rebuilt from public records by anyone willing to do it. A combined ratio cannot, because neither page publishes one. A candidate who quotes a sector combined ratio is quoting a figure we were unable to source, and saying so is stronger than guessing.

The three premium numbers an insurance case mixes up

Three premium figures live in an insurance case, and they are easy to swap because all three are quoted in dollars of premium.

1. Written premium is what was sold in the period. A policy sold on 1 December writes twelve months of premium into December.
2. Earned premium is the part of it that belongs to the period. That December policy earns one twelfth of itself in the year it was sold.
3. In-force premium is the annualized value of the book on a given day, and it matches neither of the other two.

Ratios are taken on earned premium. Growth is usually quoted on written premium. A book growing 30 percent will show a combined ratio that looks better than the underlying, because acquisition cost hits written business while the premium earns over the following year.

Two more units get mixed. Frequency has to name its exposure: claims per policy year, per insured vehicle, or per $1,000 of insured value give different numbers from the same claim file. And severity is an average over a distribution with a long tail, so an average severity of $14,200 is compatible with a hundred claims at $400,000. A case that treats the average as the whole story will underprice the tail.

Health insurance in the United States inverts the entire intuition above. The rules require insurers "to spend at least 80% or 85% of premium dollars on medical care", checked 30 September 2026 at cms.gov, and an issuer below the standard owes a rebate. A loss ratio of 60.7 is an excellent year for Kestrel and a debt for a health plan. The same metric, two signs.

Where an insurance answer goes wrong

The weights below are ours, read from the rubric code, and no consulting firm publishes a rubric of its own. Why each weight sits where it does is set out on how an answer is assessed.

Drilling an insurance case on frequency and severity

Rework Kestrel with one driver moved at a time, and name the unit each time.

Severity first. Take it from $14,200 to $16,000: 24,800 claims at $16,000 is $396.8 million, the loss ratio becomes 68.4 percent, and the combined ratio becomes 105.9. Just inside the 106.2 break-even, so the company earns about $1.4 million and nothing more.

Frequency second. Hold severity and take frequency from 6.2 to 5.4 per 100 policies. That gives 21,600 claims, and 21,600 claims at $14,200 is $306.72 million. The loss ratio becomes 52.9 percent and the combined ratio 90.4, worth $55.68 million of underwriting profit. Eight tenths of a claim per hundred policies is worth $45 million on this book, which is the sort of sentence that ends a case well.

Then check the assumption you have been leaning on. Every figure above earns premium evenly and settles claims inside the year. Real books do neither, and a reserve that develops badly two years later will rewrite the loss ratio you reported today.

Our breakeven calculator will solve the ratio threshold algebra, and the structuring drill times a technical account tree and scores its separation. Both are free and need no account.

For the balance sheet version of the same spread logic, see banking cases. For fee revenue taken on an asset stock, see financial services cases. For the statutory loss ratio floor worked through a health plan, see healthcare cases.