How GEICO repriced, and returned to growth

What the record actually shows

Berkshire publishes a GEICO underwriting table in every annual report: premiums written, premiums earned, losses, expenses, and the pre-tax result. 28 filings of it are charted on this page, from 1998 to 2025.

Read straight through it is one of the better growth records in American insurance, interrupted once. Premiums written compounded at 10.3% a year from 2001 to 2019, $4.2 B to $36.0 B, then slowed to 3.9% a year through the repricing. Progressive compounded at 14.3% over that later stretch and is now the larger personal-vehicle writer.

The interruption is the interesting part, and the rest of this page is what caused it, what it fixed, and where it leaves the business.

Why the policies had to go

Claims costs rose faster than anyone had priced for in 2021 and 2022, used-car values, repair labour, medical inflation. An insurer facing that has two choices: raise prices and lose the customers who will not pay them, or keep the customers and underwrite at a loss. GEICO chose the first, and the policy count is where that choice shows up.

The count held around 18.7 million across 2020 and 2021 and fell to 15.3 million by 2024, 3.4 million policies, 18.2% of the book. Almost all of it happened in two years, 2022 and 2023, which is the signature of a deliberate repricing, not a business losing customers it wanted to keep. Progressive, which had repriced earlier, grew from 16.5 million auto policies to 26.8 million across the same years.

On GEICO's policy figures. Berkshire does not publish GEICO's policy count. It discloses the change each year, and the series here is built from those disclosures: the 2024 level follows from the 2020 anchor, "an increase of approximately 820,000 (4.6%)", moved by the percentage stated for each year since. Figures before 2018 are interpolated from five-year cumulative disclosures and are estimates. The series stops at 2024: the FY2025 Form 10-K states that policies grew in 2025 but gives no figure, so none is shown.

What the repricing restored

A margin, and a decisive one. In 2022 GEICO's combined ratio was 104.8%, 4.8 cents paid out for every dollar taken in, against Progressive's 96.8%. Within two years GEICO was at 81.5% against Progressive's 88.6%, a 15.1-point swing in the gap, in a business where a single point is worth having.

Three years in 28 produced an underwriting loss: 2000, 2017, 2022. That is a strong record for an auto insurer, and the pattern in each case is the same: a period of cost inflation, a repricing, and a return to underwriting profit the very next year.

The half of it that is not the hard market

Every auto insurer's combined ratio improved in 2023 and 2024, as the rate increases of 2022 and 2023 caught up with claims inflation. Some of GEICO's gain is therefore the market's, and will turn with it. Splitting the ratio into its two halves separates what GEICO did from what the cycle handed it.

The loss ratio is mostly cycle. The expense ratio is not: it has fallen from 19.5% in 1998 to 12.4% today, touching 9.7% in 2023, roughly half the ratio GEICO ran on at the start of the record. Some of that low was the denominator, with rate increases lifting earned premium faster than costs, and the ratio has risen as GEICO has put money back into advertising to grow again. The underlying point stands: this is a structurally cheaper business to run than it was, and that part does not reverse when the market softens.

A cost base resized to its book

The repricing came with the work underneath it. From 2020 GEICO’s segmented rating, the ability to price each driver on their own risk instead of on broad averages, caught up with Progressive’s; telematics was built out; underpriced business was allowed to leave; and the expense base was brought down to match the smaller book.

Berkshire does not publish GEICO's headcount. It publishes the insurance group's, on the last page of each annual report, and GEICO is most of it. That figure fell 18.6% between 2020 and 2024, from 51,000 to 41,500, against a policy decline of 18.2% over nearly the same years. The two move together, which is what a cost base being resized to its book looks like.

The last three years are not the run rate

GEICO has earned $37.6 B of pre-tax underwriting profit since 1998. The way it accumulated matters as much as the total: 48.5% of it arrived in the last three years, when pricing and claims costs were both unusually favourable.

Put the other way round, 25 years of GEICO produced $19.4 B and the three most recent produced $18.3 B. A valuation that took the current run rate as its base case would be assuming the last three years are the normal ones. That is not a judgement about GEICO’s quality. It is how any cyclical underwriter should be valued.

Where it enters the sum of the parts

Nowhere as a line of its own, which is a choice and not an omission. The page carries group insurance underwriting at a normalised $2.0 B of profit, a long-run average across GEICO, the reinsurers and the primary insurers, and GEICO’s contribution to that average is what the slider on the main page exists to let you argue with. Its economics therefore enter twice: inside the $20 B underwriting line, and through the float it generates, which is why $242 B of Treasury bills sits inside the insurance subsidiaries.

The float is the larger of the two by a wide margin, and it is the reason the underwriting margin matters less to the valuation than it does to GEICO’s management. An underwriter that merely breaks even while holding $81.58 a share of other people’s money has borrowed at no cost. One running an 84.7% combined ratio is being paid to hold it, and with premiums at a record the float it generates is growing again too.

That is the case for GEICO inside Berkshire: a business that has proved it will price for profit instead of defending share at a loss, generating float on better terms than before, and growing from a sounder base than the one it gave up.

The underlying record

Every figure charted above, in one table. Ratios are Berkshire's own, struck on premiums earned. Dollars are millions.

Where the comparison breaks. Progressive's Total Personal Lines excluded its property business until FY, when the segment was redefined to include it. The premium comparison on this page therefore uses Progressive's personal-vehicle line throughout, which is unaffected. For context on how much contamination that avoids: Progressive has added $2.41 B of property premium since 2015, a 4.5× increase, but property is still only 4.3% of its 2025 Personal Lines premium. The combined-ratio chart uses Progressive's Personal Lines ratio and is mildly affected after FY; the direction of the gap is not.