Avoiding Sign and Denominator Errors in Mod Calculations

Why Worksheet Errors Change the Answer

An experience modification summarizes how a risk’s observed losses compare with its expected losses under a plan. The arithmetic can look simple, but a reversed deviation, a mismatched denominator, or a second application of credibility can produce a plausible-looking factor that answers a different question. Auditing the structure of a calculation is therefore as important as evaluating it.

The formulas below use one explicitly stated worksheet convention. Actual plans may define their calculation differently, so the plan’s instructions govern whenever they differ. Under this convention, $A$ is the eligible actual loss amount, $E$ is the corresponding expected loss amount, $Z$ is the credibility assigned to the experience, and $B$ is a nonnegative ballast amount, all measured on a consistent dollar basis. Assume $0\leq Z\leq1$ and $E+B>0$.

Definition: Under the convention used here, the experience modification is $$ M=\frac{ZA+(1-Z)E+B}{E+B} =1+\frac{Z(A-E)}{E+B}. $$ The first form blends actual and expected losses using credibility, then adds ballast. The second form expresses the same calculation as a baseline of 1 plus a credibility-weighted deviation. Do not apply this formula to a plan with a different prescribed structure.

The sign convention is deliberate: the deviation is actual minus expected, $A-E$. If actual losses exceed expected losses, the deviation is positive and this convention produces a factor above 1. If actual losses are lower, the deviation is negative and the factor is below 1, provided no separate floor or cap changes the result.

A Reliable Worksheet Sequence

Before entering numbers, identify what each worksheet amount represents. In particular, establish whether the loss amounts are already adjusted under the plan and whether credibility has already been applied. Then use the same aggregation and exposure basis for the experience deviation and its denominator. The sequence below provides a practical audit.

Identify the inputs.
Label eligible actual losses $A$, corresponding expected losses $E$, credibility $Z$, and ballast $B$. Confirm that the inputs follow the plan’s definitions.

Form the signed deviation.
Calculate $A-E$, not $E-A$. Record whether it is favorable (negative) or adverse (positive) before applying credibility.

Match the bases.
Combine the relevant actual and expected amounts over the same experience period and exposure basis. Use the denominator specified by that same plan calculation.

Apply credibility once.
Multiply the raw deviation by $Z$ in the deviation form, or blend $A$ and $E$ in the direct form. These are equivalent approaches, not consecutive steps.

Check the result.
Compare the modification with 1 and with the unmodified deviation. Check units, arithmetic, and any separate plan rules before reporting the prescribed rounded value.

Worked Examples: Three Common Error Patterns

Worked Example: Keeping the Deviation Sign Straight

A hypothetical account has eligible actual losses of $100,000, expected losses of $80,000, credibility $Z=0.60$, and ballast $B=\$20,000$. Under the stated worksheet convention, calculate the modification. Identify the error in a calculation that reports a factor below 1.

Method. Use the deviation form and keep the prescribed order, actual minus expected. The correct deviation is $A-E=\$100{,}000-\$80{,}000=\$20{,}000$. The denominator is $E+B=\$80{,}000+\$20{,}000=\$100{,}000$.

Apply credibility once: $Z(A-E)=0.60(\$20{,}000)=\$12{,}000$. Therefore, $$ M=1+\frac{\$12{,}000}{\$100{,}000}=1+0.12=1.12. $$

Check. Actual losses exceed expected losses, so the signed deviation is positive and the result is above 1. The direct form confirms the arithmetic: $ZA=\$60{,}000$, $(1-Z)E=0.40(\$80{,}000)=\$32{,}000$, and $ZA+(1-Z)E+B=\$60{,}000+\$32{,}000+\$20{,}000=\$112{,}000$. Dividing by $E+B=\$100{,}000$ gives 1.12.

Common mistake avoided. Reversing the deviation gives $E-A=-\$20{,}000$, and the wrong calculation $1+0.60(-\$20{,}000)/\$100{,}000=0.88$. That result has the wrong direction: it treats adverse experience as favorable.

Worked Example: Using One Consistent Aggregate Denominator

A hypothetical business is evaluated over two years. In year 1, actual losses are $50,000 and expected losses are $40,000. In year 2, actual losses are $54,000 and expected losses are $60,000. The plan applies one aggregate ballast of $25,000 to the two-year experience, with $Z=0.75$. Calculate the modification and explain why the ballast is not added twice.

Method. First aggregate actual and expected losses across the same two years: $A=\$50{,}000+\$54{,}000=\$104{,}000$, and $E=\$40{,}000+\$60{,}000=\$100{,}000$. The aggregate deviation is $A-E=\$104{,}000-\$100{,}000=\$4{,}000$. The denominator under the stated convention is the aggregate expected amount plus the single aggregate ballast: $E+B=\$100{,}000+\$25{,}000=\$125{,}000$.

The credibility-weighted deviation is $0.75(\$4{,}000)=\$3{,}000$. Thus, $$ M=1+\frac{\$3{,}000}{\$125{,}000} =1+0.024 =1.024. $$

Check. The two-year actual amount exceeds the two-year expected amount by $4,000, so a result slightly above 1 is directionally consistent. The direct-form numerator is $0.75(\$104{,}000)+0.25(\$100{,}000)+\$25{,}000=\$78{,}000+\$25{,}000+\$25{,}000=\$128{,}000$. Dividing by $125,000 gives 1.024.

Common mistake avoided. Adding the $25,000 aggregate ballast once for each year would create an unsupported denominator of $\$100{,}000+\$25{,}000+\$25{,}000=\$150{,}000$, producing $1+\$3{,}000/\$150{,}000=1.020$. That mixes a single aggregate deviation with a denominator that duplicates ballast. Ballast is counted according to the plan’s specified basis, not once per row by default.

Worked Example: Applying Credibility Only Once

A hypothetical worksheet shows raw actual losses of $130,000 and expected losses of $100,000. The plan assigns credibility $Z=0.40$ and ballast $B=\$20,000$. A reviewer has already calculated a “credible deviation” of $12,000. Find the modification and assess a proposed calculation that multiplies $12,000 by 0.40 again.

Method. The raw deviation is $A-E=\$130{,}000-\$100{,}000=\$30{,}000$. Applying credibility once gives $Z(A-E)=0.40(\$30{,}000)=\$12{,}000$. The denominator is $E+B=\$100{,}000+\$20{,}000=\$120{,}000$. Therefore, $$ M=1+\frac{\$12{,}000}{\$120{,}000}=1+0.10=1.10. $$

Check. In the direct form, $ZA=\$52{,}000$, while $(1-Z)E=0.60(\$100{,}000)=\$60{,}000$. Adding ballast gives $\$52{,}000+\$60{,}000+\$20{,}000=\$132{,}000$, and $\$132{,}000/\$120{,}000=1.10$.

Common mistake avoided. The proposed second application gives $1+0.40(\$12{,}000)/\$120{,}000=1.04$. The $12,000 is already the credibility-weighted deviation, so multiplying it by $Z$ again applies credibility twice. Use either $Z(A-E)$ or the already calculated credible deviation in the numerator—not both.

Why the Two Forms Agree

The deviation form is not a separate adjustment to the direct form; it is an algebraic rearrangement. This identity is useful in an audit because it lets you check a worksheet using a second representation of the same calculation.

Derivation: Begin with the direct form and subtract the denominator from its numerator: $$ \begin{aligned} M-1 &=\frac{ZA+(1-Z)E+B}{E+B}-1\\ &=\frac{ZA+(1-Z)E+B-(E+B)}{E+B}\\ &=\frac{ZA+(1-Z)E-E}{E+B}\\ &=\frac{ZA-ZE}{E+B}\\ &=\frac{Z(A-E)}{E+B}. \end{aligned} $$ The ballast cancels in the numerator because it is added to both the blended amount and the baseline denominator. This cancellation holds for the stated formula; it does not establish that a different plan must use the same structure.

The identity also exposes the three errors this tutorial addresses. Replacing $A-E$ with $E-A$ changes the sign. Changing the denominator changes the size of the adjustment. Multiplying an already credibility-weighted deviation by $Z$ again changes its size a second time. Each error alters the formula, even if the arithmetic that follows is flawless.

Audit Checks and Exam Traps

An efficient check starts with the direction of experience, then checks the arithmetic structure. Under this convention, if $Z=0$, the deviation contributes nothing and $M=1$. If $A=E$, then $M=1$ for any allowed $Z$. If $Z>0$ and $A>E$, then $M>1$; if $A<E$, then $M<1$. These are diagnostic checks, not replacements for any plan-specific floor, cap, or other adjustment.

Reversing the sign. Write $A-E$ explicitly before substituting. State whether the result is positive or negative.
Mixing bases. Do not combine a multi-year deviation with an expected-loss denominator for only one year. Use the plan’s prescribed aggregation for both.
Changing the denominator silently. Under the convention here, the denominator is $E+B$. Omitting ballast or inserting actual losses changes the calculation.
Credibility twice. If the worksheet already labels an amount as credible, determine how it was formed before multiplying it by $Z$.
Assuming all plans share a formula. These checks apply to the stated worksheet convention. Follow the documented plan calculation if it specifies a different formula or sequencing rule.

For a professional review or an exam problem aligned with the CAS Exam 8 (Advanced Ratemaking) syllabus, show enough intermediate work to make each choice visible: the signed deviation, the matching denominator, the credibility step, and the final check. A factor by itself does not reveal whether a sign or basis error was made.

Check Your Understanding

For each hypothetical worksheet, use the stated convention $M=1+Z(A-E)/(E+B)$. Keep dollar amounts on the same basis and do not apply credibility more than once.

An account has actual losses of $63,000, expected losses of $75,000, $Z=0.50$, and ballast of $15,000. Calculate the modification and state whether it is above or below 1. (Answer: $M=0.9333$, below 1.)
Over two years, actual losses are $40,000 and $45,000; expected losses are $32,000 and $48,000. One aggregate ballast of $20,000 applies, with $Z=0.80$. Aggregate first, then calculate the modification. (Answer: $M=1.040$.)
A worksheet gives raw actual losses of $150,000, expected losses of $120,000, $Z=0.60$, and ballast of $30,000. Calculate the modification. A reviewer says the credibility-adjusted deviation is $18,000. Should that $18,000 be multiplied by 0.60 again? (Answer: $M=1.12$; no, credibility has already been applied.)
For a two-year account, actual losses are $52,000 and $48,000, and expected losses are $45,000 and $55,000. The plan specifies one aggregate ballast of $10,000 and $Z=0.35$. Calculate the modification and explain whether duplicating the ballast across the two years is consistent with the stated plan. (Answer: $M=1.000$; no, the specified aggregate ballast is counted once.)
An account has actual losses of $110,000, expected losses of $90,000, $Z=0.35$, and ballast of $10,000. A draft uses $E-A$ and reports a modification below 1. Calculate the correct modification and identify the draft’s structural error. (Answer: $M=1.070$; the draft reversed actual minus expected.)

What You'll Learn