Why Immature Claim Periods Require Care
For a recent accident period, only a small portion of its claims may have been reported, paid, or otherwise emerged by the valuation date. The observed amount is therefore an incomplete view of the period's ultimate claims. An estimate that simply extrapolates the observed amount can be especially sensitive to the development pattern used. In contrast, an expected-claims estimate starts with an estimate based on exposure and expected loss level, then allows the observed emergence to affect the estimate as the period matures.
A development proportion makes this distinction explicit: it estimates the share of ultimate claims expected to have emerged by a specified maturity. Its complement is the share expected to remain. When the emerged share is small, most of the estimated ultimate comes from the expected-claims component rather than the observed amount. The distinction between paid and incurred measures is essential: a paid-based calculation can estimate total unpaid claims directly, whereas an incurred-based calculation estimates the increment from incurred claims to ultimate.
Expected Claims and the Emerged Proportion
An expected ultimate claim amount can be developed from exposure and an expected loss level. For example, if an expected loss cost per exposure unit is selected, multiplying it by the period's exposure gives expected ultimate claims. The expected amount is not the observed amount: it is a separate estimate of the period's ultimate claims based on the assumptions and information used to select the expected loss level.
Let \(U^E\) denote expected ultimate claims, \(P\) cumulative paid claims, and \(I\) cumulative incurred claims at the valuation date. Incurred claims are paid claims plus case outstanding claims under the usual accounting relationship. For a paid development proportion \(p^P\), a paid-based Bornhuetter-Ferguson (BF) estimate of total unpaid claims is:
The corresponding ultimate estimate is cumulative paid claims plus estimated total unpaid claims:
For an incurred development proportion \(p^I\), the analogous formula gives the increment from incurred claims to ultimate:
Here, \((1-p^I)U^E\) is the BF increment beyond incurred claims, commonly interpreted as an IBNR estimate under the applicable reserving convention. It is not generally total unpaid claims, because incurred claims already include case outstanding amounts. Total unpaid claims can instead be calculated as \(\hat U^I-P\), or as the case outstanding amount \(I-P\) plus the BF increment.
Why the Formula Uses the Unemerged Expected Share
The BF formula can be understood as combining a development-based ultimate estimate with an expected ultimate estimate. For paid claims, if \(p^P>0\), a simple development-based ultimate is \(P/p^P\): it scales the paid amount up by the inverse of the proportion expected to have been paid. The BF estimate gives that development-based estimate weight \(p^P\), and the expected ultimate weight \(1-p^P\). Algebraically:
The observed paid amount therefore enters directly, while the expected ultimate supplies the amount for the proportion not yet expected to have emerged. When \(p^P\) is small, the expected-claims component has the larger weight. When \(p^P\) is close to 1, there is little expected un emerged share, and the estimate is driven more by observed paid claims. The same algebra applies to incurred claims after replacing \(P\) and \(p^P\) with \(I\) and \(p^I\).
This is an algebraic explanation of the BF estimate, not a claim that a development proportion is a statistical credibility parameter in every model. The formula assumes the selected development proportion is appropriate for the period and claim measure, and that \(U^E\) is a suitable expected ultimate. If those selections are not appropriate, the formula does not remove the underlying estimation problem.
Worked Example: Paid Claims in a Very Immature Period
A hypothetical portfolio has 800 earned exposure units. The selected expected loss cost is $750 per exposure unit, so expected ultimate claims are based on that exposure. At the valuation date, cumulative paid claims are $120,000. The selected paid development proportion at this age is 20%. Estimate total unpaid claims and ultimate claims using the paid BF method. State why the result is sensitive to expected claims.
Method and expected ultimate. Multiply exposure by expected loss cost:
The period is only 20% paid-developed, so the expected proportion not yet paid is:
Total unpaid claims. Apply the paid BF formula:
Ultimate claims. Add cumulative paid claims:
The reconciliation checks: estimated ultimate less paid claims is \(\$600{,}000-\$120{,}000=\$480{,}000\), equal to the indicated total unpaid amount. Because 80% of expected ultimate claims remain unpaid at this maturity, the estimate depends heavily on the selected expected loss cost and exposure. A common mistake is to treat the $120,000 paid as if it were already a substantial indication of the full ultimate; at a 20% paid proportion, a simple paid extrapolation would be \(\$120{,}000/0.20=\$600{,}000\), but its observed-data component is still based on a relatively small emerged amount.
Incurred Claims: The Increment Is Not Total Unpaid
Incurred claims include both cumulative paid claims and case outstanding claims. Consequently, an incurred BF calculation estimates the amount needed to move from incurred claims to ultimate. To obtain total unpaid claims, the case outstanding amount must also be included. Equivalently, subtract cumulative paid claims from estimated ultimate claims. This distinction is particularly important when a question asks for an unpaid liability rather than an IBNR increment.
Worked Example: Reconciling an Incurred BF Estimate
A hypothetical accident period has expected ultimate claims of $900,000. At the valuation date, cumulative incurred claims are $360,000, of which $300,000 has been paid. The incurred development proportion is 40%. Calculate the BF increment beyond incurred claims, the estimated ultimate, case outstanding claims, and total unpaid claims. Identify the common interpretation error.
Method and BF increment. The expected share beyond incurred claims is \(1-0.40=0.60\). Therefore:
Estimated ultimate. Add this increment to incurred claims:
Case outstanding claims. Incurred claims equal paid claims plus case outstanding claims, so:
Total unpaid claims. Add case outstanding claims to the BF increment, or subtract paid claims from ultimate:
The two routes agree. The common error is to label \(\$540{,}000\) as total unpaid claims. It is only the increment beyond incurred claims in this calculation. The missing \(\$60{,}000\) is the case outstanding amount already included in incurred claims. Confusing these quantities understates total unpaid claims by the case outstanding amount.
How Maturity Changes the Estimate
The following hypothetical comparison holds expected ultimate claims fixed at $1,000,000. The paid amounts and development proportions are chosen to illustrate how the BF estimate responds at two different maturity levels. The paid BF reserve is total unpaid claims, and the ultimate equals paid claims plus that reserve.
| Period maturity | Paid proportion | Paid claims | BF total unpaid | BF ultimate |
|---|---|---|---|---|
| Immature | 10% | $80,000 | $900,000 | $980,000 |
| More mature | 60% | $590,000 | $400,000 | $990,000 |
For the immature period, the expected unpaid share is \(1-0.10=0.90\), giving a reserve of \(\$900,000\). Adding $80,000 paid gives an ultimate of $980,000. For the more mature period, the expected unpaid share is \(1-0.60=0.40\), giving a reserve of \(\$400,000\); adding $590,000 paid gives an ultimate of $990,000. The second period's estimate reflects more observed emergence. The difference between the two ultimate estimates does not mean maturity alone determines the answer: actual paid amounts, selected proportions, and expected ultimate claims all matter.
More maturity generally means a larger observed share is available to inform an estimate, but it does not guarantee less uncertainty in every practical setting. For example, a change in claim reporting, settlement, or payment practices can make a selected development proportion less representative. For an immature period, even a modest change in the expected loss level or selected proportion can materially change the amount attributed to claims not yet emerged. The main implication is to make the maturity and assumptions visible, not to treat a BF result as exact.
Worked Example: Measuring the Effect of a Development Proportion
For a hypothetical line of business, expected ultimate claims are $800,000 and cumulative paid claims are $160,000. Calculate paid BF total unpaid claims and ultimate claims when the selected paid development proportion is 20%. Then recalculate if the proportion is 30%, holding all other inputs fixed. Explain which result relies more on the expected-claims component and check each estimate.
At 20% paid development. The unpaid proportion is \(1-0.20=0.80\):
At 30% paid development. The unpaid proportion is \(1-0.30=0.70\):
The 20% selection assigns more weight to expected claims: 80% of expected ultimate remains in the reserve calculation, compared with 70% under the 30% selection. The checks are \(\$800{,}000-\$160{,}000=\$640{,}000\) and \(\$720{,}000-\$160{,}000=\$560{,}000\), respectively. The ultimate changes because the development proportion changed while paid claims and expected ultimate were held fixed. A common mistake is to increase the reserve when the paid proportion rises; with \(U^E\) fixed, a higher paid proportion reduces \((1-p^P)U^E\).
Practical Checks and Common Traps
Decide whether the observed cumulative amount is paid or incurred. Use the development proportion selected for that same measure and valuation age.
State the expected-claims basis, such as exposure multiplied by an expected loss cost. Confirm that the exposure and loss-cost units are compatible.
Subtract the applicable development proportion from 1, then multiply this complement by expected ultimate claims.
With paid data, the complement amount is estimated total unpaid claims. With incurred data, it is the increment beyond incurred claims; add case outstanding claims to obtain total unpaid.
Check that ultimate less paid equals total unpaid. For incurred data, also check that incurred less paid equals case outstanding.
The most frequent exam trap is treating every BF complement amount as total unpaid claims. The label depends on the data basis. A second trap is using a paid development proportion with incurred claims, or vice versa. A third is reading a 15% development proportion as the remaining share; it is the emerged share, so the complement is 85%. Finally, an expected ultimate based on exposure must use a loss cost with matching units: multiplying policy counts by a loss cost per vehicle-year, for example, requires vehicle-year exposure rather than an unadjusted policy count if policy terms differ.
Check Your Understanding
For each hypothetical period, identify the correct quantity and show the key arithmetic before checking the final answer.
- A period has 500 exposure units, an expected loss cost of $900 per unit, cumulative paid claims of $90,000, and a paid development proportion of 15%. Using paid BF, find expected ultimate claims, total unpaid claims, and estimated ultimate claims. (Answer: Expected ultimate $450,000; total unpaid $382,500; estimated ultimate $472,500.)
- A period has expected ultimate claims of $720,000, incurred claims of $288,000, paid claims of $250,000, and an incurred development proportion of 40%. Find the BF increment beyond incurred and total unpaid claims. (Answer: BF increment $432,000; case outstanding $38,000; total unpaid $470,000.)
- Expected ultimate claims are $1,200,000, paid claims are $360,000, and the paid development proportion is 30%. Find the paid BF total unpaid estimate and ultimate estimate. (Answer: Total unpaid $840,000; ultimate $1,200,000.)
- Two periods each have expected ultimate claims of $800,000. Period A has paid claims of $100,000 and paid proportion 20%; Period B has paid claims of $300,000 and paid proportion 50%. Find each paid BF reserve and state which relies more on expected claims. (Answer: A reserve $640,000; B reserve $400,000; A relies more on expected claims because its une merged proportion is 80%, versus 50% for B.)