Uninsured drivers by state: the estimates, the method and what they mean for your UM limit

About 15.4 percent of U.S. drivers were uninsured in 2023, roughly one in seven, according to the Insurance Research Council (IRC), with state estimates running from 5.7 percent in Maine to 28.2 percent in Mississippi. New Mexico (24.1 percent) and the District of Columbia (23.1 percent) ranked second and third, and Utah (6.2 percent) and Idaho (6.4 percent) sat just above Maine at the bottom.

This page lists all 51 estimates, pairs each with whether the state requires uninsured motorist (UM) coverage on a policy, includes it unless the buyer rejects it, or only requires an offer, and ends with a worksheet that turns a state figure into a UM limit. Every number here is a claims ratio built from injury claims, not a headcount of cars or registrations.

Estimates and published averages, not a quote. Not insurance advice. State rules and insurer rules vary. Confirm requirements with your state insurance regulator or a licensed agent in your state.

The 2023 estimate, and why the same year has had three different values

The IRC is a research organization supported by property and casualty insurers, affiliated with The Institutes. Its latest report on this subject, Uninsured and Underinsured Motorists: 2017-2023, appeared on the IRC news page dated February 17, 2025 (wire release February 20). The news release says that in 2023 more than one in seven drivers countrywide (15.4 percent) were uninsured. It names Mississippi (28.2 percent), New Mexico (24.1 percent) and the District of Columbia (23.1 percent) as the highest, and Maine (5.7 percent), Utah (6.2 percent) and Idaho (6.4 percent) as the lowest. The Insurance Information Institute (Triple-I) republishes the full 51-row state table on its uninsured motorists page, credited to the IRC, and that table is the source for every state figure below.

One correction to a figure that circulates: 14.0 percent is also a real IRC number, but it belongs to an earlier report. The IRC's October 31, 2023 release put 2022 at 14.0 percent, with the District of Columbia at 25.2 percent, New Mexico at 24.9 percent and Mississippi at 22.2 percent at the top, and Wyoming (5.9 percent), Maine (6.2 percent) and Idaho (6.2 percent) at the bottom. The 2025 report restates 2022 at 15.2 percent. Both are the IRC's own figures, published 16 months apart for the same calendar year.

Each IRC report re-estimates earlier years, so a figure belongs to an edition as well as a year. The table compares what three editions said about the same years.

The same pattern shows up in the states. Comparing the 2022-edition values with the 2023-edition values, Mississippi moved from 22.2 to 28.2 percent, New Mexico from 24.9 to 24.1, the District of Columbia from 25.2 to 23.1, Wyoming from 5.9 to 6.7, Maine from 6.2 to 5.7 and Idaho from 6.2 to 6.4. Part of that is real change over a year, since the IRC reports that Michigan had the largest decline in its UM rate from 2020 to 2023 and Washington the largest rise. Part of it is the estimate being redone. A six-point move in one state in one year is too large to hang a ranking on the first decimal of any single state. What the coverage pays is covered in the uninsured and underinsured motorist coverage guide; this page, one of the market data guides, stays on the measurement.

A claims ratio, not a headcount: how the IRC builds the number

Nobody counts uninsured drivers directly. A driver who is not on any insurer's books leaves no record that the insurance industry holds, and a roadside survey would miss most of them. The IRC instead works from claims. Triple-I describes the estimate as the ratio of claims by people injured by uninsured drivers to claims by people injured by insured drivers, and its table footnote says the share is measured as the ratio of UM claims to bodily injury (BI) claim frequencies. The IRC's October 2023 release on the previous edition adds the inputs: UM and BI liability exposure and claim counts from 10 major insurers covering about 56 percent of the U.S. private passenger auto market.

The logic runs like this. A bodily injury liability claim is filed against an insured driver who caused an injury. A UM claim is filed by an injured person against their own insurer because the driver who caused the injury had no applicable liability insurance or could not be found. Divide the frequency of the second kind of claim by the frequency of the first and, if the two kinds of drivers cause injury crashes at the same rate, you recover the relative size of the uninsured group. The IRC's own release defines the UM claim broadly, as compensation for injuries or damage caused by a driver without liability insurance or by a hit-and-run driver.

The public summaries do not print the final conversion, so there is an arithmetic point that readers should hold lightly. If the published statistic were literally UM claims divided by BI claims, a 15.4 percent reading would imply that uninsured drivers are 13.3 percent of the combined pool (0.154 divided by 1.154). If instead the ratio is already converted to a share, 15.4 percent is the share. The full report, available to IRC members and for purchase, would settle which. The conversion would not change the order of states.

What the number is not: the share of registered vehicles without a policy, the share of licensed drivers who lack coverage, or the chance that a car you pass today is uninsured. It is closer to a statement about injury crashes: among crashes that produce a liability or UM injury claim, how often the at-fault side was missing from the insurance system. That is the version most relevant to a UM buyer.

Where a claims ratio can mislead

A ratio of two claim counts inherits every difference in how claims get filed. These are the weak points, some stated by the sources and some following from how the method works.

• UM purchase rates differ by state. A UM claim can only be filed by someone who bought UM coverage. In a state where the coverage is on every policy unless rejected, nearly everyone has it. In a state where it is optional, buyers are a self-selected group, and the exposure base the IRC uses is the insured population that carries the coverage. That can move the ratio either way, so a required-coverage state and an optional-coverage state should not be compared to the decimal.
• Hit-and-run claims are counted with uninsured claims. The IRC's UM definition includes hit-and-run drivers, and so does every standard UM form. A state with many hit-and-run crashes can show a higher UM share even if few local drivers are actually uninsured. The hit-and-run claim guide shows how those claims are handled.
• The definition of uninsured is set by state law. Alabama's UM statute counts a vehicle as uninsured when the liability limits available to the injured person are less than the damages recoverable, which folds underinsured drivers into the uninsured claim. Ohio's statute includes a driver whose insurer denies coverage or becomes insolvent. Claims in those states can be in the numerator for reasons that have nothing to do with an expired policy.
• Injury claim behavior varies. The ratio compares two kinds of claim filed by injured people. States differ in attorney involvement, in no-fault thresholds and in how readily minor injuries become claims, and none of that is the same as the rate of uninsured driving. Florida carries a footnote on the Triple-I table because its compulsory law covers personal injury protection (PIP) and property damage but not third-party bodily injury, so the BI side of the ratio is built on voluntary coverage there.
• The insurer sample is not the market. The 2022-edition release names 10 insurers and 56 percent of the private passenger market. Whatever the 2025 edition used, a sample of insurers is a sample of their books, and a book that skews toward preferred drivers or toward one region can pull a small state's estimate. Maine, Wyoming and the District of Columbia are small enough for that to matter, and the Mississippi and District of Columbia moves above are the kind a narrow base produces.

None of these points make the IRC figure wrong. They make it a good measure of one thing, the insurance-visible exposure of an injured policyholder to an at-fault driver without coverage, and a poor measure of how many cars on the road lack a policy.

A registration-based count: what California measured and what Virginia reports

The alternative to a claims ratio is to count vehicles. A state that holds both a registration file and a database of insured vehicles can compare them directly. Two official examples follow, and neither replaces the IRC series.

The California Department of Insurance published an Estimated Rate of Uninsured Motorist for 2002 to 2004, prepared by its Statistical Analysis Division in September 2007. It starts from Department of Motor Vehicles counts of currently registered vehicles by ZIP code, interpolated to July 1, and compares them with the number of insured vehicles taken from the Department's database under Insurance Code section 11628(a). The statewide result for 2004 was 24,672,633 registered vehicles and 21,113,403 insured vehicles, a rate of 14.43 percent, which follows from the 3,559,230 vehicles in the difference. It also breaks the rate out by county, with county values in the file running from under 1 percent to more than 20 percent.

That approach counts what it says it counts, but registration files can include vehicles stored or parked off the road, an insured count can miss a vehicle insured in another state, and the Department's workbook sets aside ZIP codes it cannot match to a rated vehicle. The file covers 2002 to 2004 only, so it does not compare with a 2023 claims ratio. California's 14.43 percent in 2004 and the IRC's 20.4 percent in 2023 describe different years by different methods, and the gap between them is not a finding about how California changed.

Virginia shows what an enforcement system reports instead. The Virginia Department of Motor Vehicles runs an Insurance Verification Program that matches insurer electronic reports against registrations. Its fiscal year 2024 report counts 8,675,546 registered vehicles on June 30, 2024, and says 353,529 vehicle owners were issued insurance verification and denied insurance notices that resulted in 124,651 orders of suspension, a net order rate of 35 percent. The notices are 4.1 percent of the registered count and the suspension orders are 1.4 percent.

Neither of those Virginia ratios is an uninsured rate. The notices go to owners the system suspects are uninsured, many of whom then show coverage or fix a reporting gap, and the orders count a year's actions, not the uninsured share on a given day. The IRC's 2023 estimate for Virginia is 12.9 percent. A reader who sets 1.4 percent beside 12.9 percent has compared a count of enforcement actions with a claims ratio. Virginia ended its uninsured motor vehicle fee on July 1, 2024.

All 51 estimates, with each state’s UM rule

The figure sorts every jurisdiction the IRC publishes by its 2023 estimate and marks its UM rule. Required, offered and not-verified states are scattered through the whole list.

How the UM marks were assigned. Three rules sit behind the marks. Required means a policy cannot be written without UM, and the first source is the Triple-I table of Automobile Financial Responsibility Limits by State, which in its December 2023 version shows UM as required for 21 jurisdictions: Connecticut, the District of Columbia, Illinois, Kansas, Maine, Maryland, Massachusetts, Minnesota, Missouri, Nebraska, New Jersey, New York, North Carolina, North Dakota, Oregon, South Carolina, South Dakota, Vermont, Virginia, West Virginia and Wisconsin. The Triple-I page text agrees: twenty states and the District of Columbia have mandatory requirements for UM or underinsured motorist coverage.

Statute marks. Rhode Island (R.I. Gen. Laws 27-7-2.1(a)) and New Hampshire (RSA 264:15 I) are also marked Required: their statutes bar delivery of a liability policy without UM coverage. Rhode Island lets the named insured choose a lower limit in writing, and a buyer of minimum coverage can reduce UM to zero after signing the advisory notice the section requires; New Hampshire allows a written rejection only on umbrella or excess policies. Oklahoma (36 O.S. 3636(A), (G)) requires UM on every policy but lets the named insured reject it in writing, so it sits in the next group. The largest group, 23 jurisdictions, is Included unless rejected: the policy must carry UM, and the named insured can decline it, in writing in most of these statutes. Examples are Alabama (Ala. Code 32-7-23), Georgia (O.C.G.A. 33-7-11), Kentucky (KRS 304.20-020), New Mexico (NMSA 66-5-301), Tennessee (Tenn. Code 56-7-1201) and Texas (Insurance Code 1952.101); the table cites the section for every state. Mississippi’s mark rests on Miss. Code 83-11-101(1) as reproduced in House Bill 916 (2024, as introduced), which says the required coverage “shall not be applicable where any insured named in the policy shall reject the coverage in writing”. Montana and Wyoming let the insured reject without saying the rejection must be written, and require a written request to restore it at renewal.

Must be offered, optional means the insurer has to offer the coverage and the buyer chooses: Arizona (A.R.S. 20-259.01, an offer by written notice, with coverage included on request) Pennsylvania (75 Pa.C.S. 1731) and Nevada (NRS 687B.145(2), which requires insurers to offer the coverage, on a form approved by the Commissioner, at limits equal to the bodily injury limits sold). Ohio’s statute says a policy may, but need not, include UM, so Ohio has neither a requirement nor an offer duty, and Michigan has no UM statute because MCL 500.3010 was repealed effective October 1, 1973. Triple-I’s own table advises checking the state insurance department to confirm whether UM is compulsory, which is the safe step for any state in the table, as the statute detail for Texas and Florida sits in Texas car insurance laws and the Florida rules.

The full table, alphabetical

The same data, scannable by name.

What the pairing can and cannot show

Group the 51 estimates by rule and the unweighted averages come out at 12.4 percent for the 23 Required jurisdictions and 15.0 percent for the 23 Included unless rejected jurisdictions. The remaining five (Arizona, Pennsylvania and Nevada, where UM must be offered; Ohio and Michigan, optional) average 14.7 percent, too few to read. Of the Required jurisdictions, 4 of 23 are at or above the 15.4 percent national figure; of the Included unless rejected group, 10 of 23 are. Of the five highest estimates, three belong to Included unless rejected states (Mississippi, New Mexico and Tennessee), one to a Required jurisdiction (the District of Columbia, third) and one to an optional state (Michigan, fourth).

It is tempting to read that as proof that required UM goes with fewer uninsured drivers. The data do not support the strong version. The groups overlap heavily: Missouri is a Required state at 20.7 percent, and Maine is a Required state at 5.7 percent. These are unweighted state averages that give Wyoming and California the same weight. And the causal arrows can run either way. A legislature may require UM because uninsured driving is common, which would put a high-uninsured state in the required column. A required-UM state also has near-universal UM purchase, which improves the measurement of UM claims, as noted above.

There is a different reading that is safe. The UM rule decides how coverage got onto your policy, and the IRC share says how likely the driver who injures you is to have none. Those are two separate questions. In a Required state the first is settled for you and the limit is the decision. In an Included unless rejected state, a signed rejection may be sitting in your file. Texas, for example, does not require an insurer to put UM back on a renewal after a named insured rejects it in writing unless the insured later asks in writing. That is why the UM coverage guide tells readers to find out whether they ever signed one.

One state sits outside the pairing. New Hampshire is the one state without compulsory insurance, though it requires drivers to demonstrate financial responsibility, and its IRC estimate is 10.0 percent, below the national figure; a no-mandate state is not automatically a high-uninsured state. Florida requires PIP and property damage but not bodily injury liability, so its 20.6 percent describes a state where driving without BI coverage is lawful. Michigan, a state with mandatory no-fault PIP and no UM statute, posts 22.3 percent and the largest decline in the UM rate from 2020 to 2023 per the IRC, so a high figure there and a falling trend are both true. And Virginia is marked as requiring UM although, per a footnote to the December 2023 table, its drivers could instead pay an uninsured vehicle fee to the DMV, which the state repealed effective July 1, 2024. The requirement attaches to a policy, not to driving.

For related state rules, see the state minimum liability limits guide, which shows how low the other driver's own limit may be, and no-fault versus at-fault states, which explains why some of the states above treat injury claims differently. Our car insurance rates by state map places the premium side of these states next to this one.

Turning your state’s share into a UM limit: a worksheet

The state share does not choose a limit for you. It does give one input to a break-even calculation, and the rest of the inputs come from your own policy. The worksheet below uses labeled illustrative inputs, not quotes. Replace each with the number from your declarations page and your renewal offer.

Step 1, the limits. Illustrative inputs: bodily injury liability limit $100,000 per person; current UM limit $25,000 per person; the higher UM limit you are pricing, $100,000 per person, which matches the liability limit. In Florida, Illinois, Louisiana and Maryland, the statutory default UM limit equals the bodily injury liability limit, so this step asks what it costs to stay at or restore that default. The added benefit is $100,000 minus $25,000, which is $75,000, and it pays only if your loss exceeds the lower limit, so the worksheet assumes a serious injury of $100,000 or more to one person.

Step 2, the price. Illustrative input: the insurer's quote to raise the UM limit from $25,000 to $100,000 is $60 more per year. Read the real figure off the quote with everything else held equal, as the quote comparison guide describes.

Step 3, the share. Use your state's IRC share as a rough probability that the at-fault driver in a serious injury crash was uninsured, with the caveat that the IRC estimate covers all injury claims, not only serious ones. The expected extra payout from the higher limit in a qualifying crash is the share times $75,000. The break-even annual chance of such a crash is the added premium divided by that expected payout.

How to read the result. In Mississippi, the $60 pays for itself in expectation if the chance in a given year of an injury crash caused by another driver and serious enough to exceed $25,000 is above about 1 in 353. In Maine, it needs about 1 in 71. The calculation is of expected payout; a household may weigh a possible $75,000 gap more heavily than an average does, or less if it holds large savings or disability cover.

What the worksheet leaves out. It uses an uninsured share and ignores underinsured drivers, who are a separate gap and whose coverage is priced separately. It does not model stacking, which can raise the available limit across vehicles and is covered in the stacked versus unstacked UM guide. It does not include PIP or medical payments, which pay your own bills regardless of fault in the states that provide them, as the PIP versus MedPay guide explains. The liability side that the UM limit usually follows is sized in the how much car insurance do I need guide. A licensed agent in your state and your state insurance department can confirm the rules and the real quote.

Questions readers ask

Once the limit is set, the stacking election decides whether it multiplies across your vehicles, so read the stacked versus unstacked guide before you sign the renewal.