The Bornhuetter-Ferguson method estimates an insurer’s ultimate losses by adding currently reported claims to an estimate of claims still to come, where that second piece is anchored to an independent expectation of how much the book should produce rather than extrapolated from the early reported data. First published in the 1972 Proceedings of the Casualty Actuarial Society, it fills a gap that purely historical approaches leave open: in the first months of an accident year, reported claims are too sparse and volatile to project forward reliably. Anchoring part of the estimate to an outside expectation is what makes the technique one of the most widely used reserving tools for immature accident years and long-tail lines.
The Three Inputs
Three pieces of data drive the entire calculation.
Earned premium is the revenue the insurer collected for coverage during the period under review.
The initial expected loss ratio, or IELR, is the percentage of each premium dollar the insurer expects to pay out in claims. Actuaries pull it from industry aggregates published by groups such as A.M. Best, ISO, NCCI, or SNL; from the loss ratio the pricing actuary targeted when the product was rated; or from a blended judgment that factors in the underwriting cycle, policy terms, and shifts in the book’s risk profile. A carrier with several years of credible, stable experience can lean on its own history. A startup or a company entering a new line has to rely on industry data or pricing assumptions.1Casualty Actuarial Society. Bornhuetter-Ferguson Initial Expected Loss Ratio Working Party Paper The number varies considerably by line. A personal auto book and a medical malpractice book will have very different ratios.
The loss development factor, or LDF, captures how claims grow between the time they first appear and the time they fully settle. LDFs come from loss development triangles, grids showing cumulative paid or reported losses across multiple accident years and evaluation periods. The ratio of one column to the previous column is a link ratio; strung together, those link ratios produce a cumulative LDF that tells the actuary what proportion of ultimate losses have surfaced at any given evaluation point. For long-tail lines, the triangle may not extend far enough to capture all development, so a tail factor is appended to account for claims that will emerge beyond the oldest data in the triangle.2Casualty Actuarial Society. The Estimation of Loss Development Tail Factors: A Summary Report
Working Through the Calculation
Once the inputs are set, the math is straightforward.
Start with the unreported percentage, the share of ultimate losses that have not yet appeared. The formula is (1 − 1/LDF). If the cumulative LDF for an accident year at its current evaluation point is 2.0, then 1/2.0 = 0.50, and the unreported percentage is 1 − 0.50 = 0.50. Half the losses are expected to still be hidden.1Casualty Actuarial Society. Bornhuetter-Ferguson Initial Expected Loss Ratio Working Party Paper
Next compute expected losses by multiplying earned premium by the IELR. On $2,000,000 of earned premium at a 65% IELR, expected losses are $1,300,000.
Expected unreported losses, which is the IBNR estimate, equal expected losses times the unreported percentage: $1,300,000 × 0.50 = $650,000.
The ultimate loss estimate is current reported losses plus that IBNR figure. If the company has already recorded $550,000 in reported claims, the ultimate is $550,000 + $650,000 = $1,200,000.1Casualty Actuarial Society. Bornhuetter-Ferguson Initial Expected Loss Ratio Working Party Paper
Notice what the method does with reported losses: it takes them at face value and adds a layer on top. It does not multiply reported losses by a development factor the way the chain ladder would. That is why a single large claim surfacing early does not distort the entire projection upward.
Why the IELR Matters So Much
Because the method uses the IELR to size the unreported portion, the accuracy of that ratio shapes the final answer more than any other input, especially early in an accident year when reported losses are small. Set the IELR too high and reserves are overstated. Set it too low and the company is underreserved.
The Casualty Actuarial Society’s working party on IELR selection flagged a recurring concern: practitioners sometimes cling to an IELR even when emerging data suggests it is wrong. Using an expected loss ratio that contradicts current information undermines the method’s credibility.1Casualty Actuarial Society. Bornhuetter-Ferguson Initial Expected Loss Ratio Working Party Paper The practical fix is to re-evaluate the IELR at each valuation date and adjust it when evidence warrants, rather than treating it as a fixed assumption from inception.
As an accident year matures, the IELR’s influence fades. When the LDF approaches 1.0, the unreported percentage shrinks toward zero, and the ultimate loss converges on reported losses regardless of what IELR was chosen. The method self-corrects over time. The early valuations are where the stakes are highest.
How It Compares to Other Reserving Methods
Three methods dominate property-casualty reserving, and the relationship among them explains why actuaries pick BF for some situations and not others.
The Expected Loss Method
The expected loss method ignores reported claims entirely. Ultimate losses equal earned premium times the expected loss ratio. It is a reasonable starting point for the most immature accident year, where essentially no data exists, but it throws away information that BF uses.1Casualty Actuarial Society. Bornhuetter-Ferguson Initial Expected Loss Ratio Working Party Paper
The Chain Ladder Method
The chain ladder goes the other way: it relies entirely on reported claims and their development history, multiplying current reported losses by cumulative LDFs to project ultimate losses. When data is mature and the development pattern is stable, this approach is hard to beat. For immature years it is volatile. A single large early claim produces an unreasonably high ultimate; a quiet first few months produces a dangerously low one.3Casualty Actuarial Society. Why the Hell Do We Still Stick to Chain-Ladder and Bornhuetter-Ferguson
BF in the Middle
BF occupies the middle ground. It credits reported losses for what has already emerged but uses the IELR to size what is still coming rather than extrapolating from thin early data. Many actuaries use BF for the most recent accident years and shift to the chain ladder for older years where development is largely complete.3Casualty Actuarial Society. Why the Hell Do We Still Stick to Chain-Ladder and Bornhuetter-Ferguson The blended approach uses each method at the maturity level where it performs best.
Paid Versus Reported BF
The method itself comes in two flavors depending on whether the actuary uses paid losses or reported (incurred) losses as the base. The reported version estimates IBNR directly and is independent of case reserve adequacy. The paid version estimates total unpaid losses, which includes both case reserves and IBNR, so the result is influenced by how conservatively or aggressively the claims department sets case reserves. The two usually produce similar answers, but when case reserves are volatile or unreliable, reported BF tends to be more stable.4Casualty Actuarial Society. A Note on the Paid Bornhuetter-Ferguson Loss Reserving Method
When BF Is the Right Tool
The method earns its keep in specific circumstances.
Long-tail lines. Workers’ compensation, medical malpractice, and general liability claims can take years or decades to fully develop. Early in an accident year, reported losses are a small and unreliable fraction of what will eventually emerge. BF dampens that noise by pegging unreported development to expected losses rather than to the thin reported base. The tail factor applied at the end of the triangle carries outsized influence in these lines, and getting it wrong ripples through every accident year in the analysis.2Casualty Actuarial Society. The Estimation of Loss Development Tail Factors: A Summary Report
New lines and startup carriers. A company with no internal claims history cannot build a credible development triangle. Industry benchmarks for the IELR and industry development patterns for the LDFs give BF enough structure to produce a defensible estimate where the chain ladder would have nothing to work with.1Casualty Actuarial Society. Bornhuetter-Ferguson Initial Expected Loss Ratio Working Party Paper
Volatile or shifting books. A carrier that has recently expanded into a new territory, moved its underwriting appetite, or changed claims handling may find that its own historical development patterns no longer apply. BF’s partial reliance on an independent expected loss ratio makes it less likely to project a distorted past into the future.
What the IBNR Number Actually Contains
The term IBNR gets used loosely. The total IBNR number BF produces is the sum of two components.
- Pure IBNR is the estimated cost of claims arising from events that have already happened but have not yet been reported. The policyholder may not even know a claim exists.
- IBNER, incurred but not enough reported, is the expected additional development on claims the insurer already knows about. A reported claim initially reserved at $50,000 that ultimately costs $80,000 has $30,000 of IBNER.
BF produces a single aggregate IBNR figure that bundles both. Separating them requires additional analysis, but the distinction matters for claims management: large pure IBNR points to a reporting lag problem, large IBNER points to a case reserving adequacy problem, and the fixes are different.5Casualty Actuarial Society. Reserving in Two Steps: Total IBNR = Pure IBNR + IBNER
Where the Estimate Ends Up
The ultimate loss figure produced by BF (or any reserving method) feeds the insurer’s loss reserve liability, typically the largest line item on a property-casualty balance sheet. Under statutory accounting, insurers must establish liabilities for all unpaid claims, including both reported claims and IBNR, based on the estimated ultimate cost of settlement, and the estimate must reflect inflation, societal trends, and any other factors that would modify historical patterns. Where the actuary produces a range of reasonable estimates and no single point is clearly superior, the midpoint is recorded.6National Association of Insurance Commissioners. Statutory Issue Paper No. 55 – Unpaid Claims, Losses and Loss Adjustment Expenses
BF rarely stands alone in a reserve analysis. Actuarial Standard of Practice No. 43 directs the actuary to consider multiple methods unless reliance on a single method is justified and disclosed.7Actuarial Standards Board. Actuarial Standard of Practice No. 43 – Property/Casualty Unpaid Claim Estimates Most reserve opinions apply several methods to each line of business, weight the BF result more heavily for recent accident years and the chain ladder more heavily for mature years, and let the appointed actuary select or blend the results on the record.