Deriving the Bornhuetter-Ferguson Formula

Why Derive the Bornhuetter-Ferguson Formula?

The Bornhuetter-Ferguson method combines two kinds of information: the losses already reported and an expected ultimate loss selected independently of the current reported amount. The method does not project the observed losses to ultimate in the same way as chain ladder. Instead, it adds the expected portion of ultimate loss that has not yet emerged.

The distinction between the reported and unreported proportions is central. If 65% of ultimate loss is expected to have been reported, then the unreported proportion is 35%, not 65%. Applying the wrong proportion changes the reserve, sometimes substantially.

Throughout this tutorial, amounts are hypothetical, undiscounted dollar amounts on one consistent loss basis and valuation date. The examples assume the selected emergence pattern applies to the accident period being estimated. No interest or mortality assumption is involved. A reported amount and an expected ultimate must refer to compatible loss measures; the formula does not make incompatible data bases comparable.

Definition: Let $R$ be the cumulative reported loss at the valuation date, $E$ the expected ultimate loss, and $p$ the proportion of ultimate loss expected to be reported by that date. The unreported proportion is $q=1-p$. The Bornhuetter-Ferguson reserve component is $Eq$, and the Bornhuetter-Ferguson ultimate estimate is $R+Eq$.

From Expected Emergence to the Formula

The derivation begins with the expected ultimate loss $E$. Under the selected reporting pattern, a proportion $p$ of that ultimate is expected to have emerged by the valuation date, leaving a proportion $q=1-p$ still unreported. Thus expected unreported loss is $Eq$. The method uses that expected unreported amount as the addition to the observed reported loss $R$.

\text{Expected unreported loss}=E(1-p)=Eq

The estimated ultimate is the reported amount plus this expected unreported amount. Subtracting the reported amount from the resulting ultimate gives the reserve component, so the reserve and ultimate formulas are equivalent descriptions of the same calculation.

\widehat{U}_{\mathrm{BF}}=R+E(1-p)=R+Eq

\widehat{R}_{\mathrm{BF}}=\widehat{U}_{\mathrm{BF}}-R=E(1-p)=Eq

Here $\widehat{R}_{\mathrm{BF}}$ denotes the Bornhuetter-Ferguson reserve component, not the observed reported amount $R$. This notation makes the reconciliation explicit: $\widehat{U}_{\mathrm{BF}}=R+\widehat{R}_{\mathrm{BF}}$. If the selected expected ultimate or reporting proportion changes, the reserve component changes accordingly.

Why the Formula Holds

The justification is a partition of expected ultimate loss into the part expected to have emerged and the part expected to remain unreported. Since the reported proportion and unreported proportion sum to one, the expected ultimate divides as follows:

$$ E=Ep+E(1-p)=Ep+Eq $$

This identity explains the source of the unreported estimate: it is the expected ultimate multiplied by the part of the development pattern that remains. The Bornhuetter-Ferguson method then replaces the expected reported part $Ep$ with the actual reported amount $R$, while retaining the expected unreported part $Eq$. Therefore,

\widehat{U}_{\mathrm{BF}}=R+Eq

This is a construction of an estimate, not a claim that the current reported amount must equal $Ep$. Actual reported loss may be higher or lower than its expectation. The method uses the observed amount for the emerged portion and the expected ultimate for the portion not yet emerged.

Using a Cumulative Development Factor

The reported percentage may be given directly, or it may be derived from a cumulative development factor to ultimate. Under the usual multiplicative development interpretation, the factor $L$ is ultimate divided by reported loss for the selected development pattern. The corresponding reported proportion is $p=1/L$, and the unreported proportion is $1-1/L$.

p=\frac{1}{L},\qquad q=1-p=1-\frac{1}{L}

The formulas can therefore be written in development-factor form:

\widehat{R}_{\mathrm{BF}}=E\left(1-\frac{1}{L}\right)

\widehat{U}_{\mathrm{BF}}=R+E\left(1-\frac{1}{L}\right)

These relationships require $L$ and $p$ to refer to the same development basis and valuation age. If $L$ is a cumulative factor to ultimate, $1/L$ is the proportion reported to date under that pattern. A development factor for only one interval is not, by itself, the cumulative factor needed here.

Worked Example: Deriving the Reserve From a Reported Percentage

A hypothetical accident period has reported losses of $800,000. Its expected ultimate loss is $1,200,000, and the selected pattern indicates that 60% of ultimate loss is reported by the valuation date. Calculate the Bornhuetter-Ferguson reserve component and ultimate. Use undiscounted dollars and assume the reported amount and expected ultimate are on the same loss basis.

Method. Use the expected unreported proportion, then add the corresponding expected loss to the reported amount.

The reported proportion is $p=0.60$. Therefore, the unreported proportion is

$$ q=1-p=1-0.60=0.40 $$

The reserve component is the expected ultimate multiplied by the unreported proportion:

\widehat{R}_{\mathrm{BF}}=E q =$1{,}200{,}000(0.40) =$480{,}000

Add that reserve component to reported losses to obtain the ultimate estimate:

\widehat{U}_{\mathrm{BF}}=R+\widehat{R}_{\mathrm{BF}} =$800{,}000+$480{,}000 =$1{,}280{,}000

Check. The ultimate less the reported amount is $1,280,000 − $800,000 = $480,000, which equals the calculated reserve component. The method avoids the common mistake of multiplying expected ultimate by 60%: that would estimate the expected reported portion, not the expected unreported portion.

Connection to Chain Ladder

The formula can also be expressed using the chain-ladder ultimate for the same reported proportion. If $p>0$, the chain-ladder estimate based on the reported amount is $R/p$. Substituting this into the Bornhuetter-Ferguson formula gives

\widehat{U}_{\mathrm{BF}} =R+E(1-p) =p\left(\frac{R}{p}\right)+(1-p)E

Thus the Bornhuetter-Ferguson ultimate is a weighted combination of the chain-ladder ultimate and the expected ultimate: the weights are $p$ and $1-p$, respectively. This is an algebraic identity under the stated definitions. It helps explain why the method gives less influence to a volatile reported amount when only a small proportion is expected to have emerged. It does not mean that the selected expected ultimate is calculated from the current reported amount.

At full maturity, $p=1$, so the unreported proportion is zero and the formula gives $\widehat{U}_{\mathrm{BF}}=R$. At an early development age, $p$ is smaller and a larger share of the estimate comes from expected unreported loss. The method is most meaningful when the expected ultimate and development pattern are selected with care.

Worked Example: Converting a Development Factor to the Unreported Proportion

A hypothetical segment has reported losses of $510,000, an expected ultimate of $900,000, and a selected cumulative development factor to ultimate of 2.5. Calculate the Bornhuetter-Ferguson reserve component and ultimate. Assume the factor and losses are measured on a consistent basis, with no discounting.

Method. Convert the cumulative factor to a reported proportion, find its complement, and apply that unreported proportion to expected ultimate.

The reported proportion implied by the factor is

p=\frac{1}{2.5}=0.40

Therefore, the unreported proportion is

$$ q=1-p=1-0.40=0.60 $$

The reserve component is

\widehat{R}_{\mathrm{BF}}=$900{,}000(0.60)=$540{,}000

The ultimate estimate is

\widehat{U}_{\mathrm{BF}}=$510{,}000+$540{,}000=$1{,}050{,}000

Check. The factor implies that $1-1/2.5=0.60$, the same unreported proportion used in the reserve calculation. Also, $1,050,000 − $510,000 = $540,000. The common trap is to use the factor 2.5 itself as a percentage or to treat $1/2.5$ as the unreported share; $1/2.5$ is the reported proportion.

Applying the Formulas Across Accident Periods

For several accident periods, calculate each period’s reserve component using its own expected ultimate and unreported proportion, then add those components. Do not apply one period’s maturity percentage to the combined reported losses unless the periods genuinely share that selected development proportion. The total ultimate estimate is the sum of reported amounts and reserve components, provided the amounts are on a consistent basis.

Worked Example: Combining Two Accident Periods

A hypothetical portfolio has two accident periods. For period A, reported losses are $430,000, expected ultimate is $500,000, and 80% is expected to be reported. For period B, reported losses are $350,000, expected ultimate is $800,000, and 50% is expected to be reported. Calculate the reserve component and ultimate for each period and for the portfolio. Assume the selections are on a consistent, undiscounted loss basis.

Method. Apply the Bornhuetter-Ferguson formulas separately to each period before summing.

For period A, the unreported proportion is $1-0.80=0.20$. Its reserve component is

\widehat{R}_{\mathrm{BF},A}=$500{,}000(0.20)=$100{,}000

Its ultimate estimate is $ $430{,}000+$100{,}000=$530{,}000$.

For period B, the unreported proportion is $1-0.50=0.50$. Its reserve component is

\widehat{R}_{\mathrm{BF},B}=$800{,}000(0.50)=$400{,}000

Its ultimate estimate is $ $350{,}000+$400{,}000=$750{,}000$.

The total reported amount is $430,000 + $350,000 = $780,000. The total reserve component is $100,000 + $400,000 = $500,000. Therefore, the portfolio ultimate estimate is

\widehat{U}_{\mathrm{BF,total}} =$780{,}000+$500{,}000 =$1{,}280{,}000

Check. The period ultimates sum to $530,000 + $750,000 = $1,280,000, and total ultimate less total reported is $1,280,000 − $780,000 = $500,000. The method avoids assigning a single average maturity to both periods when their selected reported proportions differ.

Common Formula Errors and Applicability

The key operational check is to identify whether the quantity supplied is the reported proportion or the unreported proportion. If the question gives the percent reported, subtract it from one before multiplying by expected ultimate. If it gives a cumulative factor to ultimate, first take its reciprocal to obtain the reported proportion.

Using the reported share for the reserve. The reserve component uses $E(1-p)$, not $Ep$.
Adding expected ultimate to reported losses. The addition is only the expected unreported portion, not all of $E$.
Confusing a development factor with a percentage. For a cumulative factor $L$, use $p=1/L$, then $1-p$ for the unreported share.
Mixing loss bases. The observed amount, expected ultimate, and emergence pattern must describe compatible measures and valuation conventions.
Using an unsuitable development pattern. The algebra remains valid, but the estimate may be unreliable if the selected pattern does not represent the accident period.

The formula does not itself determine the expected ultimate or select a development pattern. Those are inputs. Nor does the derivation remove the need to understand what the observed reported amount represents; a reported-loss calculation should not be silently substituted for a paid-loss calculation. The essential result is narrower: given a compatible expected ultimate and unreported proportion, their product is the expected unreported amount added to the observed reported amount.

Key takeaway: With expected ultimate $E$, reported amount $R$, and reported proportion $p$, the Bornhuetter-Ferguson reserve component is $E(1-p)$, and the ultimate estimate is $R+E(1-p)$. If a cumulative factor $L$ is supplied, use $p=1/L$.

Check Your Understanding

For each hypothetical case, assume amounts are undiscounted and all loss measures are consistent. Work from the stated reported proportion or cumulative factor as appropriate.

An accident period has expected ultimate loss of $640,000, reported losses of $470,000, and 75% reported. Find the reserve component and Bornhuetter-Ferguson ultimate. (Answer: Reserve $160,000; ultimate $630,000.)
A segment has expected ultimate of $1,100,000, reported losses of $600,000, and a cumulative factor to ultimate of 1.6. Find the reported proportion, reserve component, and ultimate. (Answer: Reported proportion 0.625; reserve $412,500; ultimate $1,012,500.)
A hypothetical period has reported losses of $455,000, expected ultimate of $700,000, and an unreported proportion of 30%. Find its reserve component and ultimate. (Answer: Reserve $210,000; ultimate $665,000.)
For period A, expected ultimate is $420,000, reported losses are $260,000, and 50% is reported. For period B, expected ultimate is $600,000, reported losses are $500,000, and 80% is reported. Find the total reserve component and total ultimate. (Answer: Total reserve $330,000; total ultimate $1,090,000.)
A segment has a Bornhuetter-Ferguson ultimate of $1,040,000, reported losses of $680,000, and a reported proportion of 40%. Infer the expected ultimate used and verify the reserve component. (Answer: Expected ultimate $600,000; reserve component $360,000.)

What You'll Learn