RESEARCH DOSSIER
Surety Performance Loss Severity and Project Completion Costs
Abstract. A contractor's failure creates a continuing construction obligation, an immediate financing problem and a contingent claim under the performance bond. These three consequences arise from the same project but possess different monetary boundaries. This monograph examines the transformation of incomplete physical work into a surety's gross and net financial loss through legal eligibility, replacement procurement, completion engineering, contract funds, claim expenses, indemnity and recovery. The research integrates comparative contract and regulatory analysis, the economics of procurement, an industry severity study, actuarial estimation principles and transparent hypothetical calculations. United States federal contracting and a New York appellate decision establish particular legal mechanisms; a dated World Bank procurement form and United Kingdom guidance establish contrasting institutional arrangements. The principal empirical material describes contractor-level surety severity and procurement outcomes rather than identifying project-level performance-only loss. The analysis demonstrates why paid-to-date, physical progress, bond limit, contract balance, replacement expenditure and ultimate loss require separate observation. It develops a reconciliation of completion funding and claim cash flows using standard accounting and actuarial concepts, examines sensitivity to default timing, legal limits, inflation and correlated recoveries, and specifies an empirical research design. The conclusion establishes the conditions under which severity comparisons and capital assessments are informative. No proprietary claim-level observations or independent causal estimates are introduced.
Keywords: Contract surety; performance bond; construction default; completion cost; replacement contractor; contract balance; penal sum; indemnity; subrogation; claim severity; unpaid claims; risk concentration; procurement economics.
Introduction
A partially constructed building, road or industrial installation is an asset whose value depends on completion. When the contractor fails, the owner may hold a legally enforceable contract, substantial unfinished work and a financial commitment to a project that cannot be used. The surety enters this situation through a specific undertaking. Its economic exposure is generated by the divergence between resources still available under the original agreement and resources required to discharge the bonded obligation. The divergence can enlarge through interruption, uncertainty, deterioration or an inability to recover costs from the failing principal.
The subject belongs to Credit, Surety and Financial Guarantee Insurance because the underlying event involves failure of an obligor to honor an undertaking and a third party's contractual response. It also requires construction economics and actuarial science. A price for a replacement contract is a procurement observation. A surety payment is a financial transaction. An ultimate loss adds the relevant expenses and deducts the relevant recoveries on a stated basis. Neither a large replacement price nor a large project is sufficient to identify the surety's ultimate loss.
The research addresses three questions. First, how can the outstanding completion obligation and its cost be reconstructed from documents and physical evidence? Second, how do claims administration, recoverable indemnity and legal priorities alter the surety's cost? Third, which observations support credible severity estimation and how should its uncertainty be represented?
The method is a bounded comparative narrative synthesis. The source set was selected by relevance to legal mechanisms, completion measurement, procurement incentives, empirical surety severity, actuarial estimation and capital. Original institutional documents and decisions establish concrete rules and examples; dated studies establish their own outcomes and assumptions. They are not treated as a representative international sample. Accessible sections of published research were checked against the claims made here. Abstract-only findings are used only for their explicit propositions. Numerical examples are author-created, reproducible and expressed in nominal US dollars unless stated otherwise. They are accounting illustrations rather than fitted parameter estimates. The date of the research review is November 18, 2026; application to a transaction requires the operative contract and legal version.
Conceptual and Historical Foundations
The tripartite undertaking and the insured event
Contract suretyship connects a principal, an obligee and a surety. The principal owes performance under the underlying contract. The obligee receives the protection established by the bond. The surety undertakes an obligation defined by that instrument, the incorporated contract and applicable law. The principal's continuing responsibility for its failure and potential reimbursement to the surety give the transaction its distinctive credit dimension. A surety's obligations can depend on more than the financial insolvency of its principal: an unresolved performance failure may arise before formal insolvency, while an insolvent contractor may still complete a particular job through authorized arrangements.
The United States federal tradition uses separate protections for performance and payment under the statutory framework commonly called the Miller Act. The agency-hosted FAR Part 28 describes the construction bond regime, its exceptions, prescribed forms and changes to bond security. The distinction matters because a payment bond addresses specified obligations to those furnishing labor and materials; performance protection addresses the secured contract undertaking. A principal's financial distress may activate both streams, but their beneficiaries and evidentiary tests differ.
In private construction, AIA Contract Documents (2025) describes the A312–2010 instructions as applying to separately executed performance and payment bonds. AIA notes that each is frequently issued for the construction contract amount. This is a statement about a form and observed practice in its issuing context, not a worldwide rule. A researcher must preserve the executed instrument version, riders, bond amount and covered contract.
A performance bond is conceptually different from trade-credit insurance, whose loss depends on covered receivables and terms governing nonpayment; from a financial guaranty of debt service; and from a documentary demand guarantee or standby credit, under which compliance of presented documents may be decisive. All can transfer exposure to default, yet the monetary event, timing and right of recourse must be derived from the particular instrument. Mixing instruments at the preparation of a data set can invalidate any severity inference before the first model is estimated.
Historical continuity and the public purpose
Public performance security responds to a basic problem of delegated delivery: an authority awards a project whose eventual usefulness depends on the conduct and continued capacity of a private party. A low award is valuable only if the resulting obligation can be discharged. Third-party surety assessment adds an independent decision about whether the contractor can assume the undertaking. A bond can also finance certain consequences when performance fails. Procurement, credit evaluation and claim administration are therefore linked over the life of the contract.
Historical judicial evidence illustrates how contract funds can remain economically important after default. The Supreme Court's 1962 decision in Pearlman v. Reliance Insurance Co. concerns a payment surety and withheld federal contract money after its principal's bankruptcy. The majority recognized reimbursement through equitable rights in that fund under the facts of the dispute. It is evidence of a particular legal treatment of retained funds, rather than evidence that every surety recovery has priority over every insolvent estate or creditor.
A later federal administrative record demonstrates the continuity of the issue on the performance side. The Comptroller General's 1987 takeover decision involved competition over remaining proceeds after completion by a surety. Its central economic significance is that the allowable reimbursement depended on actual completion expenditure. A nominal outstanding balance was insufficient evidence of the money properly payable to the surety. The two records together show why the history of surety loss cannot be reduced to a history of checks paid to beneficiaries.
Contractor incentives and binding commitments
Limited liability changes the distribution of losses between a contractor's owners and its creditors. Where profit accrues to ownership but a substantial downside can be transferred to creditors, a distressed firm may rationally bid on terms that a healthier firm would decline. Calveras, Ganuza, and Hauk (2004), in Wild Bids, analyze this mechanism and the use of surety bonds; their abstract describes circumstances in which bonding reduces aggressive low tenders and excessive protection is possible. Wambach and Engel (2011), in Surety Bonds with Fair and Unfair Pricing, derive different implications under fair pricing and risk loading. These are conditional theoretical findings. Their combination establishes an important analytical point: the optimal amount of security depends on market structure, pricing and behavioral response, not simply on the largest possible completion bill.
A contractor might be excluded through underwriting, charged for a particular exposure, monitored after award or supported during distress. These channels must be separated empirically. A decline in final losses after stronger underwriting could represent a safer accepted population; a decline among comparable defaults could indicate more effective claim intervention; a shift in the mixture of contracts could produce both observations. The effect of requiring a bond on all awarded contracts is a different estimand from the effect of claim strategy among defaults.
Completion Costs
Establishing the technical baseline
The first task after a reported performance failure is to establish the secured scope as of a given date. Engineering completion requires a statement of what must be installed, tested and accepted. Financial completion requires a corresponding statement of what has been paid, what is contractually due, what may become payable and what must be spent. These statements share identifiers for work packages and change orders but cannot be reconstructed from a single percentage.
The US Government Accountability Office (2020) identifies baseline description, work breakdown, assumptions, data, sensitivity analysis and updates as components of reliable estimation in the Cost Estimating and Assessment Guide. These elements can be adapted to a defaulted project. The adaptation starts with a field survey tied to the original scope, distinguishes usable work from defective work, and expresses remaining quantities in units that replacement suppliers can price. It retains a record of what was known on the valuation date, so later destruction, newly discovered defects or approved scope changes can be separately attributed.
A physically estimated completion of 75% need not mean that 25% of the awarded price will finish the job. An industrial facility may have nearly all the concrete and steel in place while specialized controls, commissioning, certification and integration remain. Work with a low installed monetary value can lie on the critical path of the whole facility. Conversely, an early stage may involve expensive procured equipment already paid for and delivered, which reduces future procurement requirements even when physical installation is minimal. A cost estimator must connect the physical state to the remaining schedule and commercial commitments.
The schedule of values can itself be uneven. Mobilization, materials, overhead and subcontract allocations should be reconciled to invoices, verified quantities, approved variations and owner certificates. Overpayment occurs when the paid amount is disproportionate to earned work under applicable terms; the evidentiary basis is the agreed payment structure, not a researcher-imposed linear relationship between progress and price. A front-loaded schedule can consume the available balance faster than it reduces the remaining obligation.
The completion expenditure ledger
A useful decomposition begins with the new cost of equivalent remaining work. Add necessary site stabilization and remobilization; correction of damaged or defective bonded work; completion-related supervision, temporary works, testing and commissioning; and applicable procurement or contract administration costs. Remove duplication across the replacement bid and separately budgeted activities. If a contractor's quote already contains demolition, overhead or a risk allowance, adding the same items from an engineer's independent estimate overstates cost.
Some financial consequences belong in different ledgers. A covered delay award or liquidated damage may be legally recognized under the bond or completing agreement; it is not automatically part of physical construction expenditure. Lost operating revenue, an owner's improved specification or unrelated consultant costs may have economic significance while lying outside the agreed completion obligation. A record of payment must retain the distinction between the claimant's economic damages and the portion for which the surety is responsible.
The counterfactual baseline is also important. An incremental completion premium is the extra cost of fulfilling equivalent unfinished work after default relative to a defensible continuation baseline for that work. The original unpaid price can be a convenient funding measure, but may already reflect prior overpayments, changes, underbidding or earned amounts. It is therefore not necessarily the counterfactual cost of continued work. A project-level claim gap and a causal estimate of the default's cost are different quantities.
Let V denote the adjusted contractual price, P the amounts properly paid against it, and B = V − P the nominal contract balance for this simplified reconciliation. Let F denote contract funds available to the completion arrangement after resolving relevant restrictions, and C the cost of delivering the remaining secured work on a consistent basis. The simple funding gap is max(0, C − F). In a real project, V and P need legal reconciliation; F may differ from B. If prior retained earnings are included in B, they are part of the same balance and cannot be added twice. When the owner funds completion separately, the researcher records those flows only once.
The posted performance bond amount K is the maximum specified penal amount applicable to the instrument, subject to its exact wording and law. A simple indemnity-style illustration of covered damages D and funding shortfall may use Q = min(K, max(0, C + D − F)), assuming a valid claim, an applicable monetary cap, a remedy measured on that basis and no other contractual modifications. This notation summarizes a particular cash settlement assumption; it does not impose a monetary cap on every form of takeover obligation or on extra-contractual liability. General adjustment expenses and separate payment bond obligations require their own treatment.
Completion through a replacement contractor
Replacement procurement transfers execution risk to a new contracting relationship. A potential replacement contractor must evaluate existing work that it did not produce, price unknown interfaces and secure access to personnel, drawings, equipment warranties and permits. The inability to rely on a previous contractor's quality controls can lead to additional testing or dismantling. Technical uncertainty is likely to be reflected in both the replacement price and the conditions attached to that price.
A low replacement bid may omit an important uncertainty, while a higher bid may contain a realistic schedule and a comprehensive warranty assumption. Comparisons of bids should therefore normalize scope, exclusions, taxes, escalation, time, responsibility for existing defects and the ability to execute. The obligor's bid ranking alone cannot identify the strategy with the lowest expected ultimate loss.
The federal FAR Part 49 provides a dated regulatory example: takeover and replacement contracting provisions address the surety's completion and use of the unpaid contract price subject to conditions. The arrangement includes the treatment of certain competing claims and excusable delay; when the surety does not arrange completion, the contracting officer can obtain replacement performance through the applicable procurement process. The operative contract and regulatory version, including agency deviations where relevant, remain decisive. This specific rule is analytically valuable because it makes the financing side of completion explicit.
Costs through time
Time alters the amount of the same obligation. The cost of a replacement team may rise during a stalled procurement; materials may deteriorate; retained specialists may move to other projects. Weather and seasonal working windows can delay a restart independently of the nominal construction duration. An engineer should separate recurring monthly site costs, milestone-triggered payments, one-time remediation and market-price exposure.
Suppose a delay adds three months of site costs at $40,000 per month and $80,000 of replacement labor escalation, with no overlap or other changes. The incremental completion cost is $200,000. If the available contract balance remains fixed, this increase enters the funding gap directly until the applicable obligation or limit changes. The example isolates one mechanism; it does not predict inflation or assume that every elapsed month creates the same costs.
The choice of base date is critical for comparisons. A historical loss stated in nominal money cannot be compared directly with a contemporary cost estimate without specifying currency, price-level basis and the timing of payments. If the purpose is financial reporting, use the governing measurement basis. If the purpose is construction comparison, document the index and regional or trade specificity. A general price index may inadequately describe an electrical equipment shortage.
Recoverable Indemnities
The legal trigger and the selection of remedies
The economic need for completion precedes the classification of an admissible claim. There may be contractual notice requirements, an opportunity to address prospective default, a formal declaration, termination and an obligation to make remaining funds available. A performance bond may require or permit particular forms of response. Evidence of an adverse schedule and evidence of an operative bond claim answer different questions.
The 2024 New York appellate decision JDS Development v. Parkside Construction provides a concrete example. The dispute concerned an A312-form bond with coverage of a specified portion of a high-rise subcontract. The timing of the notifications and declaration relative to completion of that bonded portion mattered to the court's analysis. The case is useful precisely because a project extending beyond the covered work cannot by itself extend the undertaking. Researchers should avoid assigning the case's timing provisions to other editions or modified forms.
Selection of a remedy takes place after considering these rights and the technical situation. A financed continuation by the original contractor can preserve knowledge and subcontract relationships, but may expose the surety to further liquidity and execution failures. A takeover can give the surety direct control of the completion process and associated risks. Arranging a replacement with the obligee as contracting party can separate part of that operational responsibility. A cash resolution can achieve closure of a defined contractual exposure while leaving the obligee responsible for execution. An appropriately supported denial or defense is a further possible result when liability is absent or narrower than alleged.
The alternatives have different cash-flow distributions. Let a be an available lawful response, and let X(a) be its projected net cash expenditure including the relevant direct costs and recoveries. A decision based on expected loss compares E[X(a)] together with downside, timing, legal feasibility and delivery consequences. A lower initial payment may produce a higher ultimate expenditure if it delays protective work or leads to a second default. The owner's schedule costs and the surety's expenditures may diverge, which explains the importance of documented cooperation.
Contract funds and retained earnings
Contract funds serve as a bridge between the construction and surety ledgers. A performing surety may earn or receive payments as it discharges the underlying obligation. These cash receipts finance completion and reduce the surety's net own-resource expenditure. Their legal availability depends on the contract, assignments, claims, setoffs and the terms of any takeover arrangement.
The 1987 Comptroller General decision is especially instructive because the actual amount spent on completion was missing from the administrative record. The decision required a substantiated accounting before paying the surety from the disputed funds. This distinguishes a recognized priority in particular circumstances from a definitive numerical recovery. A claims file which records all remaining contract proceeds as probable recovery without demonstrating the conditions of payment reproduces the error exposed by that dispute.
Prior retention and progress payments must be traced at a transactional level. A payment to the replacement contractor from the obligee is not a receipt of the surety unless the reconciliation expressly treats it as an offset to the surety's cost. If a takeover budget states total construction disbursement and separately credits all owner-funded payments, adding the surety's direct payment in full again would count the same work twice. A common scheme of identifiers for invoices, certificates and transfers resolves this problem.
Indemnity and subrogation
An indemnity undertaking can provide contractual recourse against the principal and other indemnitors. An enforceable claim against an indemnitor and cash actually available from that party remain distinct. A contractor that has defaulted on several projects may have exhausted its free assets, granted security over them to a lender or become subject to collective insolvency proceedings. A personal or corporate indemnitor may have correlated exposure to the same event. Collateral may be illiquid or overvalued; its ownership, priority, expected sale costs and timing must be established.
Subrogation, assignment and contractual indemnity can open multiple legal paths to the same economic value. Their aggregate estimated collection cannot exceed the value realistically available to the creditor through those paths, after their interactions. A surety asserting rights to retained project funds and also claiming against an estate should distinguish the funds already credited to completion from genuinely incremental estate proceeds. The analysis should retain separate gross receivables, probability-weighted collectibility and actual cash collection.
The relevant ultimate net loss is not the face amount of the indemnity claim. It is the financial expenditure remaining after applicable receipts. In a simplified takeover ledger define G as direct surety outflow for completion and covered obligations plus allocated claim adjustment cost. Let R_b be receipts from contract funds not already netted out; R_i be incremental indemnity, subrogation and collateral receipts net of their realization costs; and R_r be ceded recoveries if the measurement is net of reinsurance. Then gross expenditure before all receipts is G; direct ultimate loss after contract and nonreinsurance recoveries is G − R_b − R_i; and retained ultimate cost after ceded recoveries is G − R_b − R_i − R_r. A cash settlement recorded net of F must not also deduct F as R_b.
This identity describes a specified reporting basis. It is possible to observe negative net development when late recoveries exceed recently recognized expenses or when a previously estimated liability is released. A historical sample conditioned on positive reported losses is not a suitable representation of all such transitions. One must identify the accounting convention and the data set's observation rule.
Worked reconciliation
Consider a hypothetical bonded project with adjusted price $10.0 million, properly paid amount $6.0 million, and a $4.0 million nominal balance fully available for completion. Assume the remaining secured work is estimated at $4.6 million, separate necessary remediation and restart at $0.5 million, and allocated adjustment expense at $0.15 million. These costs do not overlap. A valid $10 million performance bond is assumed and no additional covered damages, payment claims or reinsurance are included.
Gross surety-funded completion and adjustment expenditure in a full takeover is $5.25 million; the contract-fund receipt is $4.0 million. The remaining direct exposure before indemnity receipts is $1.25 million. If $0.35 million is later collected from independent indemnity assets, net direct ultimate loss is $0.90 million. Its ratio to the original $10 million bond amount is 9%; the physical completion bill of $5.1 million is 51% of that amount. These ratios describe different objects. One must not call either the expected percentage severity of a real portfolio.
For an otherwise identical settlement where the owner pays the $4 million directly toward the completion contract and the surety pays the $1.1 million additional completion amount, the surety's expenditure including adjustment is $1.25 million before indemnity and $0.90 million afterward. The two descriptions reconcile only if the contract funding is counted once. They illustrate why gross payment volumes across claim strategies cannot be used directly as a comparative loss measure.
Recovery timing and liquidity
Expected recovery is not available liquidity. A payment of $900,000 today creates a funding demand even if the surety expects a substantial reimbursement after litigation. For a simplified present-value illustration, a $400,000 receipt two years after the valuation date discounted at an assumed annual 5% has a present value of approximately $362,812. A $400,000 immediate receipt has a different value. No specific reporting discount rate is implied. Actual accounting, regulatory and actuarial bases must govern its use.
The timing also produces a separate risk of claims-paying capacity. A direct loss ultimately low because of an enforceable recovery can require large temporary financing. A contracted receivable can become unavailable precisely when many other contractors fail. Modelers who calculate only net severity without a payment profile miss this exposure. Suitable analyses separately project aggregate gross outflows, legal and credit-contingent inflows, and their timing.
Claim Severity
Exposure units and conditional distributions
Expected loss is defined relative to an exposure unit and time horizon. A contractor, a project, an individual bond and an ultimate default episode do not provide interchangeable observations. A contractor with multiple performance, payment and maintenance obligations may generate one economically connected event and several claim records. An analysis per project can be useful for completion engineering; an analysis per obligor is often necessary to understand concentration.
Define I as an indicator of a covered performance-loss event within a specified period and Y as the nonnegative ultimate direct cost conditional on a positive covered event. For a simplified exposure with zero loss otherwise, E[IY] = Pr(I = 1) E[Y | I = 1]. The probability and conditional mean may both depend on contractor finances, project characteristics and economic conditions. For portfolios it is generally preferable to condition both components on consistently observed information. If negative net costs or claim-related expenses with no indemnity are in scope, the zero-loss specification must be modified to include them.
A frequency measure constructed on active bonds can conceal contractor clustering. If a contractor with five projects fails, each project may face a different remaining scope but all share principal-level liquidity and indemnity assets. Treating those claims as independent observations underestimates aggregate uncertainty and gives that contractor five times the weight of a single-project failure unless the estimand intentionally requires such weighting.
Exposure may be measured by penal amount, adjusted contract price, available remaining funds, expected cost-to-complete or active contractor-level aggregate limits. These are complementary variables. A ratio to penal amount provides a convenient normalized measure of a defined capped claim. A ratio to original contract value describes project scale. A ratio to remaining physical cost gives another interpretation. Combining numerators and denominators from different dates or instruments produces spurious movements in severity.
What the industry data establish
The Surety & Fidelity Association of America's Construction Loss Severity Study (2022, eighth edition) analyzes 3,196 US and Puerto Rico contractor-level cases conditional on default and losses exceeding $250,000. Its exposure basis includes current and recently expired contract bonds; performance and payment exposures are combined. It measures severity with indemnity, expenses and salvage considered. Its PML factors are derived from historical industry experience, and the document notes that the threshold can distort results in lower exposure bands. These are important qualifications for the present research.
The printed page 10, visually examined, identifies a 2.12 multiplier for a 90th-percentile-oriented PML and 2.76 for the 95th-percentile-oriented calculation. These are study-specific calibration factors. They are not the ratio of an individual performance claim's maximum to its expected cost, universal quantile multipliers or evidence that a particular portfolio has the same tail. The study is based on voluntary industry submissions. Its publicly retrieved report text is examined here without reproducing its proprietary calculator or detailed tables.
The conditioning event is decisive. Denote T = $250,000. The study's positive-loss inclusion criterion concerns a conditional distribution beyond T; E[Y | Y > T, default] cannot be inserted as E[Y | default] without the omitted lower-loss and zero-loss observations or a justified correction. Likewise its contractor-level ratio is not an estimate of performance-only completion expense. A researcher who applies the average from such a sample to every bond both changes the statistical unit and silently alters the sample selection mechanism.
Where complete positive data are available, an empirical distribution by homogeneous cohort can be a first benchmark. With sparse large-loss observations, a hierarchical model or carefully tested parametric severity distribution may stabilize estimation, though assumptions require explicit documentation. A loss cap motivates censoring or a point mass near the cap; a threshold sample induces left truncation. Those mechanisms must not be treated as evidence for a particular unrestricted distribution. A censored record at K means that the observed indemnity may be bounded by K; it does not establish the uncapped cost.
Procurement outcomes and the interpretation of causality
Giuffrida and Rovigatti (2022), in Supplier Selection and Contract Enforcement, study US federal procurement in 2005–2015 using a change in the bond requirement threshold. They report improvements associated with bonding of 10.5% in delay and 3.7% in extra costs, with selection and monitoring offered as suggestive channels. Their introduction identifies approximate budget counterfactuals, while expressly acknowledging missing tender-stage counterfactual information. The article is evidence about procurement outcomes. It does not supply observed net performance claim severity, nor does the present monograph independently reproduce the estimates.
Favorable procurement outcomes can coexist with substantial observed severity among the smaller group that defaults after underwriting. If the screening mechanism excludes moderate-risk contractors and the remaining default episodes reflect unusually adverse events, conditional severity can even rise while aggregate expected loss falls. This is a selection possibility, not a claim that it occurred in the cited study. It establishes why effectiveness studies and claims severity studies address different research questions.
Decarolis (2014), in Awarding Price, Contract Performance, and Bids Screening, reports evidence from Italian public works that renegotiation can erode a substantial portion of the initial savings from first-price awards. The study's publisher abstract supports that procurement observation. It does not establish the cost of a surety remedy in Italy. The juxtaposition with theory and US evidence reinforces the economic importance of final delivery costs rather than award prices alone.
Tail severity and portfolio correlation
An adverse construction cycle can increase the chance of contractor failures while making replacement capacity and materials more expensive. Several firms may depend on a small set of suppliers, lenders or specialized trades; affected projects compete for the same completion resources. A widespread downturn can reduce asset values and indemnitor collectibility even as gross claim expenditure increases. These are mechanisms of positive dependence between gross loss and recovery shortfall. Their magnitudes require portfolio-specific evidence.
For example, two contractors each have probability 0.05 of a $2 million loss within a stylized year. Expected combined loss is $200,000 whether their loss indicators are independent or perfectly positively dependent. Under independence the probability of two simultaneous losses is 0.0025; under perfect positive dependence it is 0.05. The probability of the $4 million aggregate outcome changes by a factor of 20 while its expected loss does not. The example isolates dependence and assumes fixed severities. It demonstrates why concentration and liquidity assessments cannot be inferred from expected losses alone.
The European insurance regulatory setting illustrates a separate use of default and recession scenarios. The displayed 2024 version of EIOPA's rulebook Article 134 defines a credit and suretyship capital sub-module with large-default and recession components for specified insurance lines. The prescribed scenario assumptions are regulatory capital inputs. They cannot be substituted for empirically estimated construction completion costs or observed recoveries. A contemporary solvency calculation requires verifying the applicable legal version and scope. This monograph uses the dated rule solely to establish the distinction between prescribed capital stress and empirical severity.
International comparisons
The World Bank's January 2020 Procurement of Works template offers two performance security alternatives in Section X. One is a documentary monetary guarantee with a specified cap and demand conditions; the other conditions a surety undertaking on contractual default and permits remediation or completion-related arrangements. This is a historical model document, not a representation of every current World Bank-financed project or country law. A documentary claim can produce a payment which is determined before the true cost of completion has been established. The liquid guarantee proceeds and ultimate construction expenditure may therefore differ considerably.
The United Kingdom's financial standing guidance distinguishes conditional and on-demand bonds and discusses proportionate security, costs and possible exclusion of smaller suppliers. It illustrates the relationship between public procurement design and security availability in another institutional context. Neither this guidance nor the World Bank template provides a representative severity distribution. The only quantitative surety severity source used here is North American, and the procurement econometric studies relate to their own samples.
A comparative register should therefore preserve instrument family, jurisdiction, operative wording, amount, expiry, currency, owner financing, remedy and data unit. A lower observed payout ratio in a demand guarantee market may reflect a smaller guarantee amount or earlier documentary payment, while a higher ratio in a performance surety market may reflect a different denominator or a broader undertaking. The apparent ranking can reverse when measured against total owner costs. Without these variables, national rankings of "average surety severity" are scientifically weak.
Evidence Verification and Model Sensitivity
A reproducible project-level scenario
Return to the hypothetical $10 million project and consider three mutually exclusive conditional completion outcomes. Each applies the same valid monetary performance assumption, with available funds of $4 million and a sufficiently large bond limit. Values in millions of nominal US dollars are:
| Conditional event | Probability | Completion cost | Expense | Available funds | Other recovery | Net loss |
|---|---|---|---|---|---|---|
| Favorable | 0.25 | 4.50 | 0.10 | 4.00 | 0.30 | 0.30 |
| Central | 0.50 | 5.10 | 0.15 | 4.00 | 0.35 | 0.90 |
| Adverse | 0.25 | 6.50 | 0.25 | 4.00 | 0.10 | 2.65 |
The probabilities sum to one. The conditional mean net loss is 0.25 × 0.30 + 0.50 × 0.90 + 0.25 × 2.65 = $1.1875 million. The median is $0.90 million and the 90th percentile in this three-point distribution is $2.65 million, using the smallest value with cumulative probability at least 0.90. If a separate 2% unconditional probability of this type of default applies over a specified horizon and the other outcomes have zero loss, expected unconditional loss is $23,750. These assumptions are deliberately hypothetical; they do not follow from the SFAA data or any procurement study.
The table separates the distribution of the physical obligation from the chosen financial treatment. Completion cost and recovery are adversely associated in the illustrative scenarios. Imposing independence would change the distribution of combinations and may produce implausibly favorable severe defaults. The dependence assumption should be tested, not imposed through a generic simulation that happens to be convenient.
Sensitivity to contract funding and penal amount
Keep the central completion cost of $5.10 million, expense $0.15 million and other recovery $0.35 million. If available funds fall from $4.00 million to $3.50 million, while scope and everything else remain fixed, the net loss rises from $0.90 million to $1.40 million. A reduction by $0.50 million has exactly the same first-order effect as an increase by $0.50 million in covered expenditure under these assumptions. The origin is different: one concerns funding availability and the other concerns engineering cost.
Now consider a different, explicitly monetary bond with a $1.00 million applicable indemnity cap, covering an otherwise eligible completion gap of $1.10 million. The modeled indemnity becomes $1.00 million, and $0.10 million of that completion gap remains with the obligee before any other remedy. Adding $0.15 million of expenses outside the assumed indemnity cap and subtracting $0.35 million of recoveries gives the direct surety loss of $0.80 million in this illustration. It would be incorrect to apply this calculation to a different instrument without checking whether the expenses consume the limit and how recoveries affect the parties' rights.
The sensitivity to default stage is not monotonic by necessity. If a project reaches 90% physical progress while contract funding becomes nearly exhausted, final integration and correction may still generate large incremental expense. An earlier failure may be handled with a fresh replacement procurement and considerable remaining budget. A predictive model can allow stage effects to vary with payment status and the type of remaining work rather than stipulating that loss declines linearly with percentage complete.
Valuation and claims development
An open surety claim evolves through initial notice, investigation, establishment of exposure, selection of remedy, estimation, payments and recoveries. The event date, reporting date, default declaration date, contractual termination date, accounting date and settlement date are not necessarily equal. Their definitions must be consistent within the data set. A claim closed for administrative reasons may later reopen following a dispute, latent defect or recovery.
The Actuarial Standards Board (2007), in ASOP No. 43, defines the scope of unpaid claim estimation and discusses gross/net amounts, recoverables, model assumptions, data and uncertainty. It distinguishes accounting, valuation and review dates. Its provisions apply within their professional scope to qualifying US actuarial work on unpaid claims; the present study adapts the distinctions as measurement guidance and does not issue an actuarial opinion or reserve certification.
A paid-loss development triangle can provide a portfolio view when cohorts and reporting practices are sufficiently stable. It may be distorted when fewer large contractor failures dominate individual cells. A case-estimate-based evaluation can incorporate engineers' updated forecasts and known legal developments, but depends on the consistency of reserving practices. Combining portfolio development with a detailed inventory of major claims permits reconciliation of the statistical and operational pictures. Method selection and any weights need validation rather than universal prescription.
Construct separate diagnostics for gross payment, recovery and net development. A positive gross development after a new site survey may legitimately coincide with negative net development if an unrelated indemnity recovery is discovered. Net-only triangles can conceal the interaction. Separate development views also permit a better estimate of near-term funding demand.
Source-to-claim checking and data integrity
The legal source check is at the provision or holding level. The comparative section distinguishes a dated procurement template, a regulatory document, a private form issuer's instructions, court decisions and nonbinding guidance. These documents establish their own propositions, not an immutable international law of suretyship. Access errors on some alternative endpoints do not establish invalidity of the underlying law or publication; accessible authoritative text was used for the stated claims.
The research literature review is deliberately scoped. The SFAA report sections and a selected screenshot support the industry statements, but no access to its licensed calculator or original claim files was established. The empirical procurement article was checked in a repository copy via its abstract, introduction and selected methodological text. The Decarolis, Calveras–Ganuza–Hauk and Wambach–Engel discussions are limited to abstract propositions. There is no independent replication or global systematic meta-analysis.
Numerical illustration checks follow directly from the stated assumptions. The analysis does not estimate default probabilities, recovery distributions, claim development factors or aggregate capital requirements. Any estimation in a later empirical study must record the data vintage, instrument selection, observation unit, definitions, currencies, truncation, censoring, clustering, missingness, changes in reserving practice and validation period. Reproducibility requires preserving the analysis-ready records or a legally permissible replicable extract.
A source catalog can indicate potential data availability without demonstrating the presence of every needed variable. The US Small Business Administration's Surety Bond Guarantee FOIA catalog identifies downloadable program activity files and a data dictionary. The records and dictionary were not analyzed in this research. Their existence is relevant to future research feasibility, not proof that they contain project-level completion expenses, recovery dates or a representative commercial surety sample.
Integrated Scientific Synthesis and Research Agenda
Interpretation of the principal findings
The evidence supports treating performance severity as the outcome of three linked processes: technical reconstruction, legal allocation and financial realization. This is an analytical synthesis of existing construction, legal and actuarial concepts, not a newly named scientific theory. Technical reconstruction specifies the resources required to deliver the secured output. Legal allocation specifies the obligation borne by each party under a defined instrument. Financial realization determines actual expenditures, receipts, expense and time.
Each process can vary independently within a meaningful range. An engineer may estimate the same completion cost under several different bonds. Different contractual limits or trigger requirements may allocate different losses without changing the physical obligation. An identical bond payment can eventually yield different net losses because an indemnitor is solvent in one case and insolvent in another. Consequently, an unconditional numerical relationship from construction cost directly to net surety loss cannot be presumed.
The legal and empirical evidence also shows the role of temporality. A contract is a sequence of commitments and payments, not a static amount. A performance bond is an undertaking whose practical exposure develops while that work progresses. Construction risk is gradually converted into accepted asset value, but payment flows may outrun that conversion. Legal notice and remedy choices alter the time at which intervention becomes possible. Recovery processes continue after the physical work is completed. A loss study therefore needs several linked timelines.
Implications for underwriting and portfolio measurement
Before issuance, underwriting information can include financial capacity, work in progress, bonded and unbonded commitments, payment schedules, critical subcontractors, technical complexity and availability of collateral. A point-in-time contractor balance sheet alone cannot identify the cost of completing each project. The work portfolio can change materially after the bond is issued. Monitoring can improve the completeness of expected loss information, although its effect on default or severity must be measured empirically.
Concentration assessment should reflect common principal, indemnitors, trade dependencies, geographic labor markets and joint venture relationships. Multiple bonds on unrelated projects can arise from the same contractor failure. Multiple contractors can experience the same market disruption. The specific forms of dependence differ: principal concentration is legal and financial, trade concentration is operational and macroeconomic concentration affects both frequency and recoveries.
A useful comparison between two portfolios preserves the underwriting strategy and the eligibility of risks, not just the aggregate loss ratio. Portfolio A could have fewer large default episodes than B because of stronger screening, different construction sectors or the timing of the observation window. A causal conclusion about underwriting effectiveness requires control of those alternatives. A scientific comparison of financial outcomes should report gross, direct-net and retained-net measurements separately and describe the effect of ceded insurance without equating it to principal recoveries.
A prospective empirical design
A project-level study could link five kinds of records: the executed bond and amendments; underlying contract and payment certificates; independent completion state and replacement bids; surety transaction ledger; and collection, collateral and reinsurance records. An immutable project and contractor identifier would permit aggregation by both units. Outcome variables should include physical cost, additional cost over available funds, covered indemnity, adjustment expenditure, recovery and timing. Denominators should be captured at issuance and default rather than reconstructed from inconsistent end-of-period totals.
The sampling frame should include bonded exposures without claims, default notices that do not create paid loss, closed and open positive claims, and recoveries approaching or exceeding previously booked liabilities. A performance-only analysis should identify payment and maintenance exposures that share the same principal. Where expenses or recovery amounts cannot be separated, the combined nature should remain explicit. Excluding those records without examining why they are combined could itself bias the study.
An initial descriptive analysis could report the distribution of available contract balances relative to replacement completion estimates by default stage, instrument family and project type. It should report both weighted and unweighted results, explaining the relevant estimand. For frequency estimation, a denominator of eligible bonds or contractor exposures with an appropriate observation period is required. For severity estimation, the inclusion threshold and cap must be handled on the correct measurement basis. The same financial event should appear only once at the chosen unit of analysis.
A subsequent method comparison could evaluate empirical, parametric and scenario-based estimates on held-out cohorts by calendar time and contractor group. Suitable metrics include calibration of quantiles, mean prediction error, tail exceedance frequencies and the residual dependence between payments and recoveries. Because some severe claims are slow to mature, validation should distinguish predictions of ultimate loss from predictions of losses as observed at a finite valuation horizon. A superficially accurate model of immature paid claims may be a poor model of ultimate cost.
Causal research into remedy selection requires additional care. Severe projects may be preferentially taken over, while uncomplicated cases may be settled quickly. A simple comparison of their average losses could suggest the takeover caused the expense when the initial severity determined the strategy. A defensible investigation would seek a credible intervention or allocation rule, a justified comparison population and comparable pre-remedy observations. If these are absent, report conditional descriptive associations without causal wording.
Research limits and unresolved empirical questions
The scope is international and comparative at the level of legal instruments and institutional arrangements. The underlying observations on quantified contractor surety loss are North American; there is no globally comparable individual-claim sample here. Missing proprietary records preclude independent fitting or replication of the industry study. The research also does not identify the relative effectiveness of takeover, tender, financing and cash settlements across matched projects.
The hypothetical calculations transparently illustrate accounting, probability and sensitivity. A validated estimate of global performance-only severity would require data that isolate the relevant obligation. Additional questions concern the joint effect of insolvency and replacement-market scarcity, the correlation of recourse with construction cost, and the relative influence of legal notice, owner funding and technical condition. The direction and magnitude of these relationships remain empirical questions.
These boundaries are scientifically material because they define the reach of the conclusions. A reliable monograph can establish conceptual and methodological relationships before adequate data permit their quantitative estimation. Credibility requires that future empirical work distinguish these relationships from any fitted parameters or causal effects. The analytical contribution is the explicit reconciliation of the different measurements and the identification of the conditions under which they become comparable.
Conclusion
Performance bond loss is the financial result of a secured construction obligation, a legally valid allocation of responsibility and realized claim cash flows. The remaining physical work and the resources available under the contract determine the potential funding deficit. Remedy selection, the operative bond, permissible damages, ownership of funds, expenses and the collectibility of recourse determine the surety's exposure and its development over time.
The comparative evidence demonstrates substantial variation in security mechanisms. A documentary guarantee can produce a financial obligation based on a complying demand; a conditional performance instrument may involve proof of the contractual trigger and an opportunity to arrange completion. The United States judicial and regulatory examples show why the available contract balance and actual completion expenditure warrant separate documentation. None of these mechanisms implies a fixed international ratio of completion cost to surety loss.
The industry evidence and procurement literature establish complementary observations within their respective scopes. Contractor-level conditional loss statistics illuminate severe default episodes. Procurement evidence illuminates how security can affect contract outcomes before and during performance. The distinction between those statistical questions is essential. An issue of selection may alter the accepted population even when a remedy improves the outcome of an individual default.
Research and financial measurement should preserve the project's technical baseline, the original legal undertaking, all subsequent changes and a transaction-level account of payments and recoveries. Portfolio analysis should distinguish frequency from conditional severity, contractor from bond exposure, direct loss from reinsured loss, and regulatory stress from empirical distribution. The effects of timing, dependence and truncation should be expressed through explicit assumptions and tested with relevant data.
The central conclusion is that completion cost becomes a surety loss only through the financial and legal structure of the specific performance undertaking. A well-measured severity distribution must be derived from those structures and their observed outcomes. This conclusion supports transparent, comparable research while leaving quantitative calibration to properly scoped evidence.
References
Actuarial Standards Board. 2007. Actuarial Standard of Practice No. 43: Property/Casualty Unpaid Claim Estimates. June 2007; effective September 1, 2007. Read source
AIA Contract Documents. 2025. Instructions: A312–2010, Performance Bond and Payment Bond. Updated November 7, 2025. Read source
Calveras, Aleix, Juan-José Ganuza, and Esther Hauk. 2004. Wild Bids. Gambling for Resurrection in Procurement Contracts. Journal of Regulatory Economics 26 (1): 41–68. Read source
Decarolis, Francesco. 2014. Awarding Price, Contract Performance, and Bids Screening: Evidence from Procurement Auctions. American Economic Journal: Applied Economics 6 (1): 108–132. DOI: 10.1257/app.6.1.108. Read source
European Insurance and Occupational Pensions Authority. 2024. Solvency II Single Rulebook: Commission Delegated Regulation (EU) 2015/35, Article 134, Credit and Suretyship Risk Sub-module. Rulebook update March 26, 2024. Read source
Federal Acquisition Regulatory Council. Federal Acquisition Regulation, Part 28, Bonds and Insurance, especially 28.102 and 28.106. Acquisition.gov. Accessed November 18, 2026. Read source
Federal Acquisition Regulatory Council. Federal Acquisition Regulation, Part 49, Termination of Contracts, especially 49.404–49.405. Acquisition.gov. Accessed November 18, 2026. Read source
Giuffrida, Leonardo M., and Gabriele Rovigatti. 2022. Supplier Selection and Contract Enforcement: Evidence from Performance Bonding. Journal of Economics & Management Strategy 31 (4): 980–1019. DOI: 10.1111/jems.12492. Read source
JDS Development LLC v. Parkside Construction Builders Corp., 231 A.D.3d 222, 2024 NY Slip Op 04227 (N.Y. App. Div., 1st Dept. 2024). Read source
Pearlman v. Reliance Insurance Co., 371 U.S. 132 (1962). Supreme Court of the United States. Opinion reproduced by Cornell Legal Information Institute. Read source
Surety & Fidelity Association of America. 2022. Construction Loss Severity Study. Eighth edition, December 2022. Sections 3, 5 and 6. Read source
U.S. General Accounting Office. 1987. B-225115, February 20, 1987, 87-1 CPD 191. Comptroller General decision. Read source
U.S. Government Accountability Office. 2020. Cost Estimating and Assessment Guide: Best Practices for Developing and Managing Program Costs. GAO-20-195G. March 12, 2020. Read source
U.S. Small Business Administration. 2025. Surety Bond Guarantee FOIA. Dataset SBA-OCA-CKAN-010, issued April 28, 2023; catalog modified November 19, 2025. Read source
UK Government. Assessing and Monitoring the Economic and Financial Standing of Suppliers: Guidance Note. The Sourcing and Consultancy Playbooks, section 3. Accessed November 18, 2026. Read source
Wambach, Achim, and Andreas R. Engel. 2011. Surety Bonds with Fair and Unfair Pricing. The Geneva Risk and Insurance Review 36 (1): 36–50. Read source
World Bank. 2020. Standard Bidding Documents: Procurement of Works. January 2020 edition. Washington, DC: World Bank. Section X, Contract Forms, pp. 198–200. Read source