The small-numbers problem in insurer complaint ratios

A complaint index divides a company's share of complaints by its share of premium, so an insurer with very little premium needs only a handful of complaints to produce a dramatic number, while a giant insurer barely moves no matter what happens. A high ratio at a small company in one year is weak evidence; a ratio that stays high for several years at a company of any size, with the same complaint reasons, is real information. The skill is knowing which of the two you are looking at.

What the index actually divides

The idea behind the index is a comparison of shares. A company's share of all complaints in a line of insurance, such as private passenger auto, is divided by its share of all premiums in that line. A company that writes 2 percent of the premium and draws 2 percent of the complaints sits at 1.00. At 4 percent of the complaints it would sit at 2.00, and at 1 percent it would sit at 0.50. The baseline of 1.00 is the market as a whole, which is why it is the reference point.

We can check the arithmetic against a state regulator's published table. Indiana's Department of Insurance publishes a 2024 auto complaint index listing each company's premium, its number of complaints and its index. The totals at the bottom show 320 complaints and about $6.80 billion in premium across all companies, including 488 companies with zero complaints that are left off the list. Take any row and apply the formula: one company in the list has about $4.56 million in premium and one complaint. Its share of complaints is 1 in 320, and its share of premium is $4.56 million out of $6.80 billion. Divide the first by the second and you get 4.66, the figure printed in the table. The index behaves exactly as the share-ratio description says, for Indiana's own data.

An equivalent way to think about it, and the more useful one for our purposes: the index is the actual complaints divided by the complaints you would expect if the company drew complaints in proportion to its premium. An index of 1.80 means 80 percent more complaints than expected for its size. That second framing makes the small-numbers problem easy to see, because expected complaints can be a tiny number.

The National Association of Insurance Commissioners adds a reminder that the index moves for a reason beyond the company itself. Its consumer guidance says a company's complaint index can vary not only because of fluctuations in the company's own complaints or premiums but also because of changes in the complaints or premiums of the market for that type of insurance. The same page says you should not rely on one factor to choose an insurer, and recommends comparing the complaints, financials and direct premiums written of several companies. For the basic walk-through of looking one up, see The NAIC complaint index: how to look up an insurer.

Why small insurers swing so widely

Suppose, for illustration, that across the whole market regulators closed 4,000 confirmed private passenger auto complaints in a year, and the market wrote $120 billion in premium. Every $100 million of premium would then "expect" about 3.3 complaints. The table shows what that implies for companies of different sizes. The numbers are invented to show the mechanism.

The same four extra complaints barely register at the large company and make the very small one look like an outlier. For the small company, a year with one complaint and a year with five are only four complaints apart, yet they give indexes of 0.50 and 2.50.

Indiana's 2024 table shows the real-world version. One line shows a company with roughly $1.2 million of auto premium and two complaints, with an index of 35.84; another shows a company with about $302 million of premium and five complaints, with an index of 0.35. Two complaints and five complaints would not normally tell you which company treats customers worse. The premium base does the work. The report also marks companies with premiums under $1 million as "did not calculate," which is the regulator's way of acknowledging that a ratio on so little premium is not meaningful.

Random variation, not just bad service

Complaints behave a bit like rare events. Even a company with exactly average service will not draw exactly the expected number each year. As a rough illustration, assume complaints arrive independently at an average rate and the company is exactly average (a simplification, but a useful one):

• Expected 400 complaints (the large company above): nearly every year would land between about 360 and 440 (random variation of roughly 20 either side counts as ordinary), so the index stays between about 0.90 and 1.10.
• Expected 2 complaints (the small company): the chance of zero complaints is about 14 percent, the chance of one or none about 41 percent, and the chance of four or more, which gives an index of 2.00 or higher, about 14 percent. One average small company in seven would produce an index of 2.00 or more in a given year by chance alone.

These are our own calculations from a simple model, not NAIC statistics, and real complaints are not perfectly random. They show the scale of noise to expect, and they explain why across hundreds of small insurers some will always look terrible and some will always look flawless in a given year.

Why one year is weak evidence

The NAIC lets consumers look up closed, confirmed complaint data for the past three years by state, company and insurance type. That window is the right unit of analysis for a reason: pooling three years triples the expected number of complaints and cuts the noise relative to the signal. The California Department of Insurance builds the same idea into its Consumer Complaint Study, which presents data for the past three years and counts only complaints closed in the study year, regardless of when they were filed.

That last point is a second source of lumpiness. A batch of complaints filed in one year may be closed, and so counted, in the next, and a regulator's backlog can shift complaints between years. A smooth underlying pattern can still show up as a spike.

Confirmed versus unconfirmed, and volume versus rate

Not every complaint counts. The NAIC describes its data as closed, confirmed complaint information from state insurance departments. States classify outcomes in their own words: California's study uses the term "justified complaints," while the NAIC page uses "confirmed." The exact definitions differ by state and we did not find a single national definition on the NAIC pages we read, so check how your state defines it. A complaint that a regulator investigates and dismisses, or a consumer inquiry that never becomes a formal complaint, is not in the confirmed count, even though the consumer was frustrated. A company can therefore have plenty of unhappy customers and a modest index. The reverse is true as well.

Volume is the raw count, and it mostly tells you which insurers are large. Rate adjusts for size, which is what the index tries to do. But rates themselves differ in what they divide by:

• The premium-based index treats a dollar of premium the same everywhere. A company whose policies carry high premiums, such as one that insures expensive cars or high-risk drivers, has fewer policies for each dollar of premium, so the same complaints per policy produce a lower index.
• California's study instead uses justified complaints per 100,000 policies or exposures, which tracks customers directly rather than dollars.

Neither is wrong. They measure slightly different things, so compare indexes only against indexes built the same way, in the same line, for the same years. Keep in mind too that the index covers one line of business at one legal company. A large brand's different subsidiaries can show different numbers; Indiana's 2024 table lists five lines belonging to one national group with indexes ranging from 0.48 to 1.51. Read the row for the entity on your policy; see which legal entity actually insures you: reading your policy documents.

Three worked examples with invented numbers

Using the same invented market (4,000 complaints, $120 billion of premium), here is how to read three kinds of result.

A high ratio that means something

"Alpha Insurance" writes $3.6 billion, 3 percent of the market, so it should expect 120 complaints. Over three years it records 204, 216 and 228, for indexes of 1.70, 1.80 and 1.90. The typical random wobble at 120 expected complaints is about 11, and Alpha is 80 to 110 above expectation every year. That is not chance. The right next step is the breakdown: are the complaints about delayed claims, low settlement offers, or cancellations? A persistent excess concentrated in claims handling is the pattern worth weighing heavily.

A low ratio that means little

"Beta Insurance" writes $90 million and should expect 3 complaints. It records none, for an index of 0.00. Before you read that as flawless, note that an exactly average company with that expected count records zero about 5 percent of the time, and across a market with many small insurers several will do so every year. Zero also cannot be compared with another small company that records one complaint and shows 0.33. Contrast "Delta Insurance," which writes $6 billion and should expect 200 complaints. It records 140, for an index of 0.70, three years running. The shortfall of 60 complaints is about four times the typical wobble at that size, so it is much more likely to reflect real performance. Both are low, but only one is informative.

A volatile ratio

"Gamma Insurance" writes $60 million and expects 2 complaints a year. It records 1, then 5, then 2: indexes of 0.50, 2.50 and 1.00. A reader who sees only the middle year sees a company with 2.5 times the expected complaints. A reader who pools the three years sees 8 complaints against 6 expected, an index of 1.33, which is well within the range chance produces at that size. The sensible reading is "no clear signal," and the remaining question is whether the five complaints in one year share one cause.

How to combine the index with other evidence

Treat the complaint index as one instrument among several.

1. Pool the years. Use the three-year window, not the latest year alone, and note the count of complaints as well as the ratio.
2. Look at the reasons. The breakdown tells you whether complaints concern claims, billing or underwriting. Our article on what drivers actually complain about to insurance regulators explains the categories.
3. Check size. If the expected number of complaints is under about 10 a year, treat the ratio as a rough guide and do not rank it against large insurers.
4. Add financial strength. A complaint record says nothing about whether the money will be there. See AM Best financial strength ratings explained.
5. Add regulatory history. Fines, consent orders and market conduct findings are public. See market conduct exams and regulatory actions against insurers.
6. Add satisfaction data, knowing how it is collected. See how to judge satisfaction surveys and online reviews of insurers and J.D. Power auto insurance studies explained.
7. Compare like with like. Use the same line, years and entity for every company you consider, and use the compare auto insurers side by side to keep the notes together.

Common questions

Does a high index mean a company is bad?

Not by itself. It means more confirmed complaints than expected for the premium written. Whether that is meaningful depends on how many complaints stand behind the ratio, whether it persists across years, and what the complaints are about.

Does an index of zero mean a company never has problems?

No. It means the regulator recorded no confirmed complaints against that company in that line and period. Many problems are resolved with the company, never reach a regulator, or are not confirmed. Small insurers also often have too little premium for the index to say much.

Why do different websites show different numbers for the same insurer?

They may use different years, different lines of insurance, different legal entities, or different measures, such as a premium-based index versus complaints per 100,000 policies. Check the source, the year and the entity before comparing.

Should I ignore the index for small insurers?

Do not ignore it, but do not rank on it. Look at the count of complaints, the reasons, and several years together, and give more weight to other evidence such as financial strength and your own state's data.

Can my complaint change a company's index?

A complaint that your state department closes as confirmed or justified is part of the data that feed these measures, which is one reason formal complaints matter beyond your own case. See The NAIC complaint index: how to look up an insurer for how to read the result.

Key takeaways

• The complaint index divides a company's share of complaints by its share of premium, so 1.00 is the market baseline; Indiana's published 2024 table reproduces the arithmetic.
• Small insurers have tiny expected complaint counts, so one or two complaints can swing their index sharply with no change in service.
• Pool three years, look at the number of complaints and their reasons, and compare only like with like: same line, years and legal entity.
• A low ratio at a small insurer proves little, while a persistent high or low ratio at a large insurer is real evidence.
• Combine the index with financial strength, regulatory history and satisfaction data rather than ranking on it alone.