Where Home Insurance Got Dropped, and When

The Senate Budget Committee asked 23 insurers for one number per county per year, from 2018 through 2023: how many homeowners policies those carriers declined to renew. Between them the 23 hold about two thirds of the United States homeowners market. Counting every one of the 51 jurisdictions in the file together, with no floor applied to anything, the carriers ended 316,054 policies at term in 2018 and 448,486 in 2023, while the book behind those decisions went from 39,356,303 policies to 42,471,772. That is 0.80 percent of policies in 2018 against 1.06 percent in 2023.

That six-year change is the wrong measurement. Everybody quotes it anyway. The national reading fell for two years first, to 0.53 percent in 2020, the calmest year in the file, and every year after that came in above the one before it. From the 2020 low the count of policies not renewed went from 216,746 to 448,486 in three years, a factor of 2.07, and the rate went from 0.53 percent to 1.06. Measured from 2018 the national change is 0.25 points. Measured from 2020 it is 0.53. The six-year framing halves the story, and every figure on this page that spends a 2018 baseline is spending that halved number.

The file holds six years and the move sits inside two of them. Of the 100 largest single-year county steps in it, measured only where both ends of the step stand on a book of 500 policies or more, 43 land in 2022 and 48 in 2023. That is 91 of 100 in the last two years the committee collected, against four in 2019, one in 2020 and four in 2021. The largest single step in the file is Glades County, Florida, which went from 0.30 percent in 2021 to 32.46 percent in 2022.

The doublings agree. The counties that multiplied did their multiplying in the same window. Of the 493 counties that more than doubled between 2018 and 2023, 355 put more than half of the six-year move after 2021. So the shape of this market is not a drift that ran for six years. It is a quiet stretch through 2020, then a step, then a bigger one, and a reader handed the six-year number alone is being handed an average of the calm and the climb together.

Ranked on the six-year change, and counting only counties that held 500 policies or more in both 2018 and 2023, the 20 largest rises in the country sit in five states: 10 in North Carolina, four in Florida, three in Massachusetts, two in California and one in Oklahoma. Glades County leads them, up 15.77 points from 0.46 percent to 16.23 percent on a 2023 book of 2,915 policies. The policy base sits in the last column of the table because a rate cannot be read without it, and seven of the 20 rows below come off books of under a thousand policies.

Of the 2,671 counties that clear the floor in both years, 1,785 came in higher in 2023 than in 2018, 855 came in lower and 31 landed on the same published figure, which is a fact about two-decimal rounding as much as about the market and is set out in the method below. The median county went from 0.74 percent to 0.90 percent. Another way to put it is that 493 counties more than doubled, spread across 45 of the 51 jurisdictions, and 212 at least tripled. The rise is broad.

The depth is not. One county in that pool read 5 percent or higher in 2018 and 40 did in 2023, and the 100 counties at the top of the national list hold 1,466,867 policies between them, 3.45 percent of the 42,471,772 in the 2023 file. A national market can move 0.25 points over six years, or 0.53 off its low year, while the places that actually went hard stay small, which is why a study of this file that only reports national averages reports almost nothing.

A state rate counts policies, not counties. The biggest books decide what a state’s headline number does, so a state can read calmer across six years while counties inside it climb the national table. That happens here, and it is worth an address-level look, but it is not the general case and this page is not going to imply that it is. Of the 15 jurisdictions whose own rate fell between 2018 and 2023, 14 carry at least five ranked counties. In eight of those 14 the median ranked county fell as well, in one it held level, and in five it rose. A falling state rate usually means falling counties.

North Carolina is the sharpest exception, and the exception is not the one the phrase suggests. Its statewide rate fell from 2.07 percent in 2018 to 1.79 percent in 2023, a drop of 0.28 points that ranks it 47th of 51 jurisdictions on the size of the move, while on the 2023 level alone it comes 3rd. Of its 100 counties, 98 held at least 500 policies in both years and can be ranked. Of those 98, 64 came in lower in 2023 than in 2018, and those 64 hold 85.91 percent of the ranked 2023 book against 14.09 percent for the 34 that rose. All ten of the largest books in the state came in lower in 2023 than in 2018: Wake from 2.20 percent to 1.36, Mecklenburg from 2.57 percent to 1.69 and Guilford from 2.53 percent to 1.77. The drop is those books.

What rose, rose hard and in a small number of places. The median ranked county in the state fell from 1.87 percent to 1.52, while the unweighted mean of the same 98 counties rose from 1.91 percent to 2.68. A mean that climbs while the median falls is a tail rather than a shift, and the top of the tail is a handful of counties: Dare from 1.93 percent to 12.92, Washington from 2.47 percent to 12.17, Greene from 1.39 percent to 11.35 and Chowan from 1.50 percent to 9.31. The national top 100 holds 24 of these counties, more than any state except Florida, and every one of them rose. So the state number is not covering up a broad county rise. It is holding one fall and one rise at the same time, on two sets of places that have almost nothing in common except a state line, and the fall sits on the books big enough to decide what the state number says.

Across the whole file, 26 of the 100 highest county rates in 2023 sit in one of the 15 states whose own statewide rate went down over the six years. The national list runs to 17 states in all, and Florida, North Carolina, California, Louisiana and Oklahoma carry 84 of the 100 places between them.

Two readings were tested against the file and came back weaker than expected. The first is the shape everyone reaches for, coasts and wildfire country. It holds for where the rates are highest: Florida, North Carolina, California, Louisiana and Oklahoma carry 84 of the national top 100. It does not hold for where the multiplication happened. Nine states hold 18 or more counties that more than doubled between 2018 and 2023, and here they are in full: Florida with 51; Oklahoma with 39; Louisiana with 35; Mississippi with 27; California with 26; Illinois with 22; Kentucky with 20; and North Carolina and West Virginia with 18 each. No state is skipped. Four of those nine have no ocean or gulf shoreline: Oklahoma, Illinois, Kentucky and West Virginia. Doubling off a low base is an inland event as much as a coastal one, and a page that names only the coasts would be describing the level and calling it the trend.

The second is the jump from near zero to double digits, which sounds like it should be the headline and is not. Only two counties in the ranked pool went from under 0.5 percent in 2018 to 10 percent or more in 2023: Glades County, Florida and Dukes County, Massachusetts. Set against that, 95 counties in the pool fell by half or more over the same six years. Both of those belong on the page. A study that prints the rises and buries the falls is choosing its answer before it counts.

The source is one file: the Senate Budget Committee county-level homeowners insurance non-renewal dataset, released in December 2024 and covering 2018 through 2023. The committee collected it from 23 insurers holding roughly two thirds of the United States homeowners market, which means the file covers most of the market and is not a census. Every figure on this page was computed from that file and from nothing else.

A non-renewal here is the insurer ending the policy when the term runs out. It is not the homeowner switching carriers, not a mid-term cancellation, and not a claim denial. That distinction is the whole reason the committee asked for the count, because the carrier walking away is the number that moves before prices and availability move.

Rates on tiny books are arithmetic rather than signal. Petersburg City, Virginia, reads 50.00 percent for 2023 because one of its two policies was not renewed, and Emporia City reads 20.00 percent on one of five. So no county is ranked, named or called highest unless it held at least 500 policies in the year behind the claim, which is the same floor the state pages work from. For the six-year change both ends have to clear it, because a change has two denominators. Of the 3,122 counties carrying a 2023 rate, 2,737 clear the floor on 2023 and 2,671 clear it on both years. Counts taken over every county are marked as such where they appear, and the small books stay in the downloadable file with their policy base beside them.

County naming in the source is not internally consistent. Six pairs of spellings across four states describe the same county twice: Illinois carries DE WITT and DEWITT, DU PAGE and DUPAGE; Indiana carries LA PORTE and LAPORTE; Maryland carries PRINCE GEORGE’S and PRINCE GEORGES, QUEEN ANNE’S and QUEEN ANNES; and Texas carries DE WITT and DEWITT. Exactly one state carries a residual row that is not a county at all: Texas, spelled UNKNOWN and Unknown. Variants were merged on letters alone, ignoring case, spacing and punctuation, keeping the spelling that carries the most years, and the residual rows are not counties and were dropped from the county tables. After that merge the file holds 3,146 county entries across 51 jurisdictions.

Rose, fell, held level, doubled and tripled are counted on the two-decimal rates the source publishes and this page prints, not on the fraction underneath them. That convention is the only reason 31 counties can land on exactly the same figure in both years. Run the same counts on the non-renewal count over the policy base and the split is 1,799 up, 872 down and no exact ties at all, with 494 counties more than doubling across 44 states rather than 493 across 45. The published rates are the ones a reader can check against the tables here and against the state pages, so they are the ones the sentences are built on. The choice is still a choice. It moves the counts.

Data last updated December 18, 2024, the committee’s release date. Download the county file as CSV or the state file, policy bases included.