Tutorials › CAS Exam 8 - Advanced Ratemaking › Detecting Strategic Responses to Experience Modifications

Home  /  Tutorials  /  CAS Exam 8 / Evaluating Experience Rating

CAS Exam 8 · Evaluating Experience Rating · Tutorial 585 of 989

Detecting Strategic Responses to Experience Modifications

Learn how plan rules can create incentives to shift losses or exposure, and how to test for suspicious patterns without mistaking evidence of an incentive for proof of manipulation.

Advanced 11 min read

What You'll Learn

  • Identify experience-plan rules that may reward changes in loss timing, exposure reports, or coverage allocation.
  • Calculate how a hypothetical change in credited losses or exposure affects the modification and premium.
  • Distinguish a plan’s financial incentive from evidence that a policyholder acted on it.
  • Design comparisons using occurrence dates, independent exposure records, and unaffected coverage.
  • Recognize when caps, floors, or premium-base changes alter the incentive.

Why Look for Strategic Responses?

An experience modification is intended to make a risk’s premium respond to its own experience. That response can also create an incentive: a policyholder may have a financial reason to influence which losses are counted, which exposure is reported, or which coverage arrangement bears a loss. A pattern consistent with such an incentive deserves investigation, but it does not by itself prove that anyone manipulated a record or acted improperly.

The key analytical distinction is between underlying risk experience and the way that experience enters the rating calculation. Accidents may remain unchanged while their payment dates, administrative classifications, or reported exposure change. A sound diagnostic therefore begins with the written plan rules, identifies which quantities affect the modification, and then compares those quantities with records that are less susceptible to the suspected response.

Definition: A strategic response is a change in behavior, reporting, timing, or allocation that is consistent with an economic incentive created by a rating rule. Detecting a pattern consistent with a response is not the same as establishing intent, misreporting, or causation.

Map the Rule Before Examining the Data

Before searching for unusual patterns, specify what the plan actually counts. A plan might use losses by accident date, losses paid during a period, or another defined measure. It may use payroll, sales, insured units, or a derived expected-loss amount as exposure. It may also treat different coverages or policy terms differently. These distinctions determine what a policyholder could potentially influence and what comparisons are meaningful.

For the calculations below, use an illustrative linear modification. Let \(B\) be the unmodified premium, \(R\) the losses credited under the plan, \(T\) the expected losses associated with the exposure, and \(c\) a credibility weight between zero and one. Before applying any cap or floor, define

$$ M_{\text{raw}}=1+c\left(\frac{R}{T}-1\right), \qquad P=B M, $$

where \(M\) is the modification after any plan limits and \(P\) is the modified premium. This is an illustrative plan rule, not a universal experience-rating formula. It assumes \(T>0\), that \(R\) and \(T\) are measured on a consistent basis, and that the stated credibility applies to the policy being evaluated. If the actual plan uses a different formula, its own rule must replace this one.

Suppose the plan floors \(M\) at \(M_{\min}\) and caps it at \(M_{\max}\). Then

$$ M=\min\!\left(M_{\max},\max(M_{\min},M_{\text{raw}})\right). $$

A plan’s marginal incentive depends on the position within these limits. Between the floor and cap, a change in credited losses changes the modification. Beyond a limit, a small change may have no premium effect at all. The base premium may itself depend on reported exposure, so an analysis that holds \(B\) fixed when it actually changes can give the wrong answer.

Three Useful Diagnostic Questions

1
What can change the rating input?
List the relevant dates, loss amounts, exposure measures, classifications, and coverage designations. Confirm each from the plan documents rather than assuming that a familiar convention applies.
2
What is the direction and size of the incentive?
Recalculate the modification and premium under the observed value and a plausible alternative. Include caps, floors, and changes in the premium base.
3
Does the data show a response, or only an opportunity?
Compare with independent records, unaffected periods or groups, and measures of underlying claims. Check whether other explanations fit the same pattern.

Worked Example: Payments Near an Experience-Period Boundary

Worked Example: Payments Near an Experience-Period Boundary

A hypothetical plan credits amounts paid during a two-year experience period. Its written rule uses payment date, not accident date. The base premium is \(B=\$100{,}000\), expected losses are \(T=\$80{,}000\), credibility is \(c=0.40\), and the modification is limited to 0.80 through 1.20. Losses other than one claim total \(\$72{,}000\). A \(\$12{,}000\) claim from an accident in the period is paid either just before or just after the period closes. Assume no discounting and no other changes.

Method. Calculate the modification under each payment-date outcome, then compare premiums. The plan’s stated payment-date rule is essential: if it instead used accident date, delaying payment would not remove this claim from the experience period.

If payment falls inside the period, credited losses are

$$ R_{\text{in}}=\$72{,}000+\$12{,}000=\$84{,}000, \qquad \frac{R_{\text{in}}}{T}=\frac{84{,}000}{80{,}000}=1.05. $$

The uncapped modification is

$$ M_{\text{raw,in}}=1+0.40(1.05-1)=1+0.40(0.05)=1.02. $$

This is between 0.80 and 1.20, so \(M_{\text{in}}=1.02\). The premium is

$$ P_{\text{in}}=\$100{,}000(1.02)=\$102{,}000. $$

If payment falls outside the period, the claim is not credited under the stated rule, so \(R_{\text{out}}=\$72{,}000\). Then

$$ \frac{R_{\text{out}}}{T}=\frac{72{,}000}{80{,}000}=0.90, \qquad M_{\text{raw,out}}=1+0.40(0.90-1)=1-0.04=0.96. $$

The result is within the limits, giving \(P_{\text{out}}=\$100{,}000(0.96)=\$96{,}000\). The premium difference is \(\$102{,}000-\$96{,}000=\$6{,}000\). Thus, the rule creates a measurable incentive to move a payment beyond the boundary, even though the accident and claim amount are unchanged.

Diagnostic interpretation. Examine payment dates relative to the boundary, but also compare accident dates, claim reporting dates, reserve changes, and claim closure activity. A cluster of payments just after the cutoff is a signal to investigate, not proof of deliberate delay. The common mistake is to infer a timing opportunity without checking which date the plan actually recognizes.

Worked Example: Exposure Changes Affect Both Parts of the Premium

Worked Example: Exposure Changes Affect Both Parts of the Premium

A hypothetical plan uses the illustrative formula above, with \(c=0.40\), a minimum modification of 0.80, and a maximum of 1.20. Expected losses are \(\$100\) per exposure unit, and the manual premium is \(\$150\) per exposure unit. Actual credited losses are \(\$120{,}000\). The independently supported exposure is 1,000 units, but a preliminary report lists 800 units. Assume the plan recalculates both expected losses and base premium from reported exposure; there is no discounting.

Method. Recalculate \(B\), \(T\), \(M\), and \(P\) for each exposure value. Do not compare modifications alone: reported exposure changes the base premium as well.

With 1,000 units, the base premium and expected losses are

$$ B_1=1{,}000(\$150)=\$150{,}000, \qquad T_1=1{,}000(\$100)=\$100{,}000. $$

The loss ratio to expected losses is \(R/T_1=120{,}000/100{,}000=1.20\), so

$$ M_1=1+0.40(1.20-1)=1.08, \qquad P_1=\$150{,}000(1.08)=\$162{,}000. $$

With 800 reported units, the base premium and expected losses become

$$ B_2=800(\$150)=\$120{,}000, \qquad T_2=800(\$100)=\$80{,}000. $$

Now \(R/T_2=120{,}000/80{,}000=1.50\), and

$$ M_2=1+0.40(1.50-1)=1.20, \qquad P_2=\$120{,}000(1.20)=\$144{,}000. $$

The modification rises from 1.08 to 1.20, but the premium falls by \(\$162{,}000-\$144{,}000=\$18{,}000\), because the smaller exposure also reduces the base premium. This is a potential incentive, not evidence that the lower report was intentional or inaccurate. Verify exposure against an appropriate independent source, such as the records used to define the exposure basis, and confirm that the same units and period are being compared.

Common mistake. Looking only at the higher modification would suggest the lower exposure report raises premium. The complete calculation shows otherwise. The example assumes the stated exposure rates and formula; if the actual plan adjusts the base premium or expected losses differently, recalculate using those terms.

Worked Example: Loss Allocation Across Coverage Arrangements

Worked Example: Loss Allocation Across Coverage Arrangements

A hypothetical insured has two coverage arrangements. The first has an experience modification based on eligible losses; the second has a fixed premium unaffected by the modification. Under the first arrangement, \(B=\$120{,}000\), \(T=\$100{,}000\), \(c=0.50\), and the modification is limited to 0.80 through 1.20. Eligible losses are \(\$110{,}000\). A \(\$20{,}000\) claim might instead be classified under the second arrangement, whose premium is \(\$30{,}000\). Assume the claim is paid in either arrangement, the second premium stays fixed, and there is no discounting.

Method. Calculate total premium under both allocations. This isolates the financial effect of the eligible-loss classification while holding the claim payment and fixed second premium unchanged.

If the claim remains eligible, \(R_1=\$110{,}000\), so \(R_1/T=1.10\). The modification and first-arrangement premium are

$$ M_1=1+0.50(1.10-1)=1.05, \qquad P_1=\$120{,}000(1.05)=\$126{,}000. $$

Adding the second arrangement’s premium gives total premium \(\$126{,}000+\$30{,}000=\$156{,}000\).

If the \(\$20{,}000\) claim is allocated to the second arrangement, eligible losses fall to \(R_2=\$110{,}000-\$20{,}000=\$90{,}000\). Thus,

$$ \frac{R_2}{T}=\frac{90{,}000}{100{,}000}=0.90, \qquad M_2=1+0.50(0.90-1)=0.95. $$

The first premium becomes \(\$120{,}000(0.95)=\$114{,}000\), and total premium is \(\$114{,}000+\$30{,}000=\$144{,}000\). The apparent total-premium difference is \(\$12{,}000\), although the claim itself has not disappeared.

The calculation identifies an incentive only under the assumed contract terms. A real review must establish which arrangement properly covers the claim, whether any corresponding premium or payment changes elsewhere, and whether classification follows documented coverage rules. Treating every cross-arrangement difference as manipulation is the common mistake.

Why the Linear Incentive Has This Form

Between the floor and cap, substitute the raw modification into the premium formula:

$$ P=B\left[1+c\left(\frac{R}{T}-1\right)\right]. $$

For fixed \(B\), \(T\), and \(c\), a change \(\Delta R\) in credited losses changes premium by

$$ \Delta P=\frac{Bc}{T}\Delta R. $$

This follows by subtracting the premium at \(R\) from the premium at \(R+\Delta R\): the constant terms cancel, leaving \(Bc\Delta R/T\). Therefore, an interior reduction in credited losses lowers premium when \(B,c,T>0\). The result does not apply unchanged at a cap or floor, where the modification may remain fixed, or when \(B\) or \(T\) changes along with \(R\). That is why the complete plan calculation is more reliable than a general claim that a particular action must reduce premium.

Evidence, Controls, and Common Traps

A useful review asks whether an observed shift is unusual relative to a reasonable comparison, and whether the timing or classification change is supported by records. Depending on the suspected mechanism, examine accident and payment dates together, compare reported exposure with independently maintained exposure records, or compare eligible and ineligible losses for the same underlying event. Earlier material on experience-period boundary effects is useful when interpreting patterns near cutoffs; the present diagnostic question is whether the rule could have induced a response.

Where feasible, compare affected risks with similar risks not exposed to the rule change or incentive, while accounting for different claim mixes and exposure levels. A before-and-after change alone may reflect seasonality, business growth, claim handling, or random variation. Do not attribute a change to strategic behavior merely because it occurs after a modification rule begins.

  • Confusing an incentive with proof. A premium difference establishes a possible financial motive under the assumed rules; it does not establish intent or a false report.
  • Using the wrong date. Accident date, report date, and payment date can produce different results. Follow the contract’s defined basis.
  • Ignoring the premium base. If reported exposure changes \(B\), calculate the whole premium rather than the modifier alone.
  • Ignoring caps and floors. A plan may have no marginal premium response for changes that leave the modification beyond a limit.
  • Assuming a shifted loss vanished. A loss moved to another period or coverage may still be paid and may affect another premium or account.
  • Overlooking legitimate explanations. Claim settlement timing, corrected records, or coverage interpretation may explain a pattern without strategic conduct.
Key takeaway: Detecting a potential strategic response requires two separate findings: a plan rule creates a material incentive, and observed data show a pattern consistent with a response. Neither finding alone demonstrates manipulation. State the assumptions, calculate the full premium effect, and validate the pattern against independent records.

Check Your Understanding

For each problem, assume the stated hypothetical plan rules apply exactly. Use no discounting, and apply any specified limits.

  1. A plan uses \(B=\$80{,}000\), \(T=\$50{,}000\), \(c=0.30\), and modification limits of 0.80 and 1.20. Credited losses are \(\$55{,}000\), including an \(\$8{,}000\) payment. Calculate the premium if that payment is included and if it falls outside a payment-date-based experience period. (Answer: Included: \(M=1.030\), premium \(\$82{,}400\); outside: \(M=0.982\), premium \(\$78{,}560\); difference \(\$3{,}840\).)
  2. A plan has \(c=0.25\), expected losses of \(\$80\) per exposure unit, manual premium of \(\$120\) per unit, and credited losses of \(\$90{,}000\). Compare premiums using 1,000 supported units versus 900 reported units. (Answer: At 1,000 units, \(T=\$80{,}000\), \(B=\$120{,}000\), \(M=1.03125\), and premium is \(\$123{,}750\). At 900 units, \(T=\$72{,}000\), \(B=\$108{,}000\), \(M=1.0625\), and premium is \(\$114{,}750\); the difference is \(\$9{,}000\).)
  3. In the first year, a boundary-adjacent payment rate is 18% among affected policies and 10% among comparison policies. In the following year, the rates are 25% and 12%, respectively. Calculate the difference-in-differences in percentage points, and state one limitation of this comparison. (Answer: \((25\%-18\%)-(12\%-10\%)=5\) percentage points; the comparison is not causal without support for assumptions such as comparable trends and no other differential changes.)
  4. An experience-rated arrangement has \(B=\$100{,}000\), \(T=\$80{,}000\), and \(c=0.50\). Eligible losses of \(\$92{,}000\) include a \(\$12{,}000\) claim that might instead be assigned to a separate arrangement with a fixed premium of \(\$25{,}000\). Calculate total premium under each allocation, assuming the first arrangement’s base premium is unchanged. (Answer: Claim eligible: \(M=1.075\), total premium \(\$107{,}500+\$25{,}000=\$132{,}500\). Claim assigned to the other arrangement: eligible losses \(\$80{,}000\), \(M=1.000\), total premium \(\$100{,}000+\$25{,}000=\$125{,}000\); difference \(\$7{,}500\).)
  5. A plan uses \(B=\$120{,}000\), \(T=\$60{,}000\), \(c=0.50\), and limits of 0.85 and 1.15. Credited losses fall from \(\$80{,}000\) to \(\$70{,}000\). Calculate the premium change and explain the role of the cap. (Answer: At \(\$80{,}000\), raw \(M=1.1667\), capped at 1.15, so premium is \(\$138{,}000\). At \(\$70{,}000\), \(M=1+0.50(70/60-1)=1.0833\), so premium is \(\$130{,}000\). Premium falls by \(\$8{,}000\); the cap made the first modification lower than its raw value.)

Previous

Next

Back to Tutorials
Clinical Biostats