Pricing Surety's Rare, Severe Losses in a Golden-Era Loss Ratio

The surety industry's direct incurred loss ratio fell to 20.5% through the third quarter of 2025, down from 24.9% in 2024, and net profit margins have exceeded 30% in each of the past eleven years (AM Best via Construction Equipment Guide, 2026). A loss ratio that low looks like evidence the line is overpriced. It is not: surety loss is a low-frequency, high-severity process where a single large contractor default, not last year's benign ratio, sets the true cost of the risk being priced.

How the Premium Is Actually Quoted

Contract surety premium is not built the way a liability rate is built, from an expected-loss-plus-load calculation applied to exposure units. It is quoted as a rate per $1,000 of contract price, or penal sum, graded on a sliding scale where the first layer of the bond carries the highest rate and successive layers step down (hyperexponential, Surety Insurance Pricing Guide). A contractor bonding a $10 million project might see 1.5% to 2% applied to the first several hundred thousand dollars of premium base, with the marginal rate falling on each subsequent tier, so total premium grows sub-linearly with contract size. That structure exists because the actuarial cost driver is not the size of the contract in isolation; it is the probability that the specific contractor completes it.

That probability is where underwriting credit tiers do the real pricing work. For rated obligors, cumulative default rates by rating category and tenor feed directly into exposure-rated pricing much the way a corporate bond spread reflects issuer credit quality (hyperexponential). Most contract surety principals are unrated private companies, so underwriters substitute personal financial statements, working-capital-to-backlog ratios, and the indemnitors' personal credit scores as underwriting proxies. The filed rate per thousand is the manual starting point; the credit tier assigned to the specific contractor is what actually prices the bond, and it is why two contractors bonding identical $10 million projects can pay materially different premiums for the same penal sum.

Expected Cost Is Frequency Times Severity Net of Recovery

The actuarial model underneath that manual rate is default frequency multiplied by severity, and severity itself has to be modeled net of salvage and general indemnity agreement recoveries rather than gross. Loss triangles in this line are conventionally built net of salvage and subrogation rather than on a gross paid basis, because gross paid losses on a defaulted contract routinely run several multiples of what the surety ultimately retains once completion-cost recoveries and indemnitor collections are applied (hyperexponential). Early cash outlays, covering a replacement contractor to finish the job, can exceed the ultimate net cost by a wide margin before recoveries catch up in later development periods.

Frequency, meanwhile, is not stationary. It correlates tightly with the construction credit cycle: contract surety is the most cyclical of the surety sub-lines, and default frequency rises when liquidity tightens and project demand falls, exactly the environment a standalone loss triangle from a benign period will not contain (hyperexponential). Building the expected-cost model as a single blended frequency times severity estimate, calibrated to years like 2023 through 2025 when the construction backlog was strong and defaults were scarce, embeds a frequency assumption that will not survive the next downturn in the cycle. A defensible model separates the two components explicitly: a frequency curve conditioned on macro and sector indicators, and a severity distribution conditioned on contract size, completion percentage at default, and recovery rate, rather than a single historical average loss ratio extrapolated forward.

Why 20.5% Is Not Evidence of Adequacy

A single-year loss ratio understates the true cost of a line where losses are episodic. Surety's underwriting profits topped $2.35 billion for the third consecutive year through the first nine months of 2025, net profit margin reached 45.6% in 2024, the highest since 2014, and rate has barely moved: price increases came in under 1% for 13 of the past 14 quarters even as premium grew at a near double-digit pace (AM Best via Construction Equipment Guide, 2026). Every one of those figures is a calendar-year snapshot of a process whose loss distribution is not calendar-year shaped. A defaulting contractor on a large public infrastructure job can generate a loss that consumes several years of premium from an entire book in a single event; the years without such an event will always show loss ratios that look, in isolation, like radical overpricing.

The correlated exposure sitting underneath the current benign run is the same infrastructure spending that is driving the premium growth. Federal funding tied to the 2021 infrastructure law has pushed public construction activity and surety demand together, with direct premium through the first nine months of 2025 up close to 10% year over year on that base (AM Best via Construction Equipment Guide, 2026). That funding source has a known expiration: authorization under the current infrastructure law runs out in September 2026, and the Congressional Budget Office projects the Highway Trust Fund's annual shortfall approaching $40 billion by fiscal 2028, a gap with no legislated fix given a federal fuel tax that has not moved since 1993. A book built up on infrastructure-driven contract volume is concentrated exposure to a funding cycle, not a diversified spread of independent risks, and that correlation is precisely what a single-year loss ratio cannot show.

Pricing to a Target Return on the Aggregate Exposure

Because the loss shape is low-frequency and high-severity with a correlated tail, the correct pricing question is not "what loss ratio did the book run last year" but "what capital has to stand behind the aggregate penal-sum exposure, and what return does that capital require." A surety writer's true exposure at any point is closer to the sum of open penal sums across its book than to premium volume, because a cluster of defaults concentrated in one region or sector during a downturn could draw simultaneously on the same capital base. Pricing to a target return on capital means treating the rate per thousand not as a stand-alone number calibrated to expected loss, but as a number calibrated to expected loss plus the cost of holding capital against the tail scenario where several large defaults cluster in a single accident year. That capital charge is what keeps a filed rate defensible through a full cycle rather than only through the years when defaults do not occur.

"The dividing line is not merely one of return and risk. It is a question of how much liquidity and transparency they are prepared to forego in pursuit of additional spread and underwriting access," Mark Gibson of Schroders Capital said of credit-sensitive alternative risk transfer, a framing that applies directly to surety capacity providers weighing how much aggregate penal-sum exposure to hold against a given capital base. The same tradeoff underlies a surety writer's decision on how large a single-contractor line it will bond relative to its own capital, and it is a decision the single-year loss ratio gives no guidance on.

The General Indemnity Agreement Is a Pricing Input, Not a Backstop

The general indemnity agreement is the mechanism that converts a surety bond from a pure risk-transfer instrument into something closer to an extension of credit, and it belongs explicitly in the rate rather than functioning as an afterthought recovery source. When a contractor defaults, the surety first looks to the remaining contract funds held by the project owner and exercises its right of subrogation to apply those funds toward completion costs. Where contract funds are insufficient, the surety turns to the general indemnity agreement, which obligates the principal and any personal or corporate indemnitors to reimburse the surety in full for losses, costs, and legal expenses incurred on the bond. Personal and corporate indemnity from the principal is a prerequisite of issuing a contract surety bond, not a discretionary pricing variable applied only to weaker credits (hyperexponential).

Because recoveries dominate net loss in this line, the credit quality of the indemnitors, not just the contractor entity, is a direct input to expected severity. A principal with substantial personal and corporate assets behind the indemnity agreement produces materially lower net severity on default than one whose indemnitors are thinly capitalized, even holding gross contract exposure constant. That is one reason commercial surety, covering license, permit, and other statutory obligations, prices and loses differently from contract surety: commercial surety exposure is generally capped at the bond penalty and depends less on macroeconomic conditions, while contract surety completion costs can exceed the penal sum when a defaulted project runs over budget, and default frequency swings with the credit cycle in a way license and permit bonds do not (hyperexponential).

Reserving IBNR for Rare, Large Defaults

Surety reserving carries its own version of the frequency problem. Because large contractor defaults are rare, a carrier's own historical loss triangle in a given segment may contain few or no comparable events in any recent accident year, which makes chain-ladder development factors poorly suited to estimating IBNR for the tail. A completion bond on a multi-year public infrastructure project can run years past contract execution before a default becomes evident, and the reporting lag between the underlying financial deterioration of a contractor and the point at which a claim is filed can itself span several quarters, since the surety typically only becomes aware of trouble once the contractor misses a payment obligation or the obligee declares default. That lag means a reserving actuary working from paid or reported claim triangles is working with data that lags the true emergence of the underlying risk by a meaningful margin.

The practical response is to reserve IBNR using an exposure-based or frequency-severity simulation approach calibrated to industry-wide default experience across a full credit cycle, rather than relying solely on a single carrier's own thin claim history, and to hold that reserve even in years, like the recent run, when reported losses are minimal. A reserve that shrinks to match a benign reported-loss year is reserving to the wrong distribution; the tail risk that justifies holding IBNR for surety does not disappear because it has not yet been observed in the current book.

Benchmarking Against a Through-the-Cycle Ratio

The correct adequacy test compares the filed rate against a loss ratio computed across a full construction cycle, spanning both benign years like the current run and a downturn comparable to 2008 through 2010 or the sharper contractor-failure spikes that followed, rather than against the most recent one or two calendar years. A through-the-cycle loss ratio built from twenty or more years of industry experience, weighting in the infrastructure-driven boom years alongside prior downturns, gives a materially different picture of adequacy than a snapshot of 2024 and 2025 alone. Given that construction material costs have risen roughly 34% since 2020 and that surety carriers are already reporting bigger project sizes producing larger individual claims when defaults do occur, the severity side of that through-the-cycle benchmark is also drifting upward even while frequency remains low (Construction Executive, 2026; Seubert, 2026). A rate that looks comfortably profitable against a 20.5% single-year loss ratio can still be short of a through-the-cycle target once that upward severity drift and the correlated infrastructure-funding tail are both built into the benchmark, which is the actuarial argument for treating the golden era as a data point, not a destination.