Why Return-to-Work Rules Need Explicit Modeling
A claimant who resumes work may still qualify for a reduced disability benefit, while a claimant whose work meets a contractual threshold may cease to qualify altogether. These are different provisions: an offset changes the amount payable while eligibility continues; a termination rule makes the payment zero because the claimant no longer satisfies the contract’s definition of disability. A valuation must apply each rule separately and in the order specified by the hypothetical contract.
This tutorial uses the disability-state and benefit concepts established in the earlier material on disability income models. It focuses on translating stated return-to-work rules into cash flows. All scenarios and data below are hypothetical. The methods are aligned with the SOA Exam ALTAM (Advanced Long-Term Actuarial Mathematics) syllabus.
The formula’s order matters. First determine whether the claimant is eligible. If eligible, calculate the earnings above the disregard, apply the stated offset percentage to that excess, subtract any other offsets, and then apply the zero floor. The disregard protects the first \(G\) dollars of earnings; it does not generally reduce the benefit dollar-for-dollar by the full amount of earnings.
From Work Status to a Benefit Payment
A useful model separates the claimant’s status from the payment rule. For example, define \(D\) as disabled and not working, \(P\) as partially working while still satisfying the disability definition, and \(R\) as recovered or otherwise ineligible. A contract might pay the full scheduled benefit in \(D\), an offset benefit in \(P\), and nothing in \(R\). The status labels alone do not determine the payment: eligibility and offsets must still be evaluated using the contract terms.
When payments are made at regular dates, calculate the benefit applicable at each payment date. If \(p_{s,j}\) is the probability of being in status \(s\) at payment date \(j\), and \(B_{s,j}\) is the payment for that status and date, then the expected payment at that date is the sum of each status probability multiplied by its benefit. With an effective annual interest rate \(i\), end-of-year payments over \(n\) years have expected present value
This expression assumes that the stated probabilities describe the status at each payment date, that the payment is determined by that status at the date, and that the cash flow is paid at the end of the year. If the contract pays monthly, changes benefits immediately on a return to work, or pays in arrears for a different period, the payment timing must be modeled accordingly.
Worked Example: Applying Two Offsets and a Zero Floor
A hypothetical policy has a monthly benefit of \(\$2{,}200\). While the claimant remains eligible, the policy deducts 40% of work earnings above a \(\$500\) monthly disregard and then deducts \(\$200\) of other income. Earnings for the month are \(\$1{,}500\), and the claimant satisfies the policy’s eligibility test. Calculate the payment. State the method and identify the calculation trap.
Method. Because eligibility is satisfied, apply the offset formula. Work earnings above the disregard are $$ \max(1{,}500-500,0)=1{,}000. $$ The earnings offset is $$ 0.40(1{,}000)=\$400. $$ Subtract this offset and the other-income offset from the scheduled benefit: $$ 2{,}200-400-200=\$1{,}600. $$ The amount is positive, so the zero floor does not change it. The monthly payment is \(\$1{,}600\).
Trap avoided. The two deductions must both be applied. Subtracting only the earnings offset would give \(\$1{,}800\), which is not the contract payment. The correct check is that total offsets are \(\$400+\$200=\$600\), and \(\$2{,}200-\$600=\$1{,}600\).
Worked Example: Expected Present Value With Partial Work and Recovery
A hypothetical contract pays once at the end of each of the next two years. The effective annual interest rate is 5%. The claimant is in state \(D\) at time 0. At each year-end, the state is \(D\) (disabled, not working), \(P\) (eligible while partially working), or \(R\) (recovered and ineligible). The one-year transition probabilities are the same each year:
| Current state | Next state D | Next state P | Next state R |
|---|---|---|---|
| D | 0.60 | 0.25 | 0.15 |
| P | 0.20 | 0.50 | 0.30 |
| R | 0 | 0 | 1.00 |
The annual benefit is \(\$24{,}000\) in \(D\). In \(P\), the monthly scheduled benefit is \(\$2{,}400\); eligible work earnings are \(\$1{,}800\) per month, the disregard is \(\$800\), the earnings offset is 50% of earnings above the disregard, and other monthly income offsets are \(\$100\). In \(R\), no benefit is payable. Find the expected present value of the two annual payments. Payments are based on the state at the relevant year-end.
Step 1: Find the payment in \(P\). The earnings above the disregard are \(1{,}800-800=\$1{,}000\). The earnings offset is \(0.50(1{,}000)=\$500\). Thus the monthly payment is \(2{,}400-500-100=\$1{,}800\), or \(12(1{,}800)=\$21{,}600\) for the year. The other annual payments are \(\$24{,}000\) in \(D\) and \(\$0\) in \(R\).
Step 2: Find the state probabilities at year 1. The claimant starts in \(D\), so the probabilities are the first row of the transition table: \(P(D_1)=0.60\), \(P(P_1)=0.25\), and \(P(R_1)=0.15\). The expected year-1 payment is $$ 0.60(24{,}000)+0.25(21{,}600)+0.15(0) =14{,}400+5{,}400=\$19{,}800. $$
Step 3: Find the state probabilities at year 2. Sum the probabilities of all paths leading to each state: $$ P(D_2)=0.60(0.60)+0.25(0.20)+0.15(0)=0.36+0.05=0.41, $$ $$ P(P_2)=0.60(0.25)+0.25(0.50)+0.15(0)=0.15+0.125=0.275, $$ $$ P(R_2)=0.60(0.15)+0.25(0.30)+0.15(1)=0.09+0.075+0.15=0.315. $$ The probabilities sum to \(0.41+0.275+0.315=1.000\), as required. The expected year-2 payment is $$ 0.41(24{,}000)+0.275(21{,}600)+0.315(0) =9{,}840+5{,}940=\$15{,}780. $$
Step 4: Discount each expected payment. With \(v=1/1.05\), the expected present value is $$ \frac{19{,}800}{1.05}+\frac{15{,}780}{1.05^2} =18{,}857.1429+14{,}312.9252 =\$33{,}170.07. $$ The final answer, rounded to the nearest cent, is \(\$33{,}170.07\). As a check, the undiscounted expected payments total \(\$35{,}580\), and discounting at a positive rate reduces their present value.
Trap avoided. The year-2 state probabilities are not just the second row of the transition table: the claimant can reach year 2 in either \(D\) or \(P\). The method is to weight each transition probability by the probability of its starting state.
Why Expected Payments Can Be Valued State by State
The expected value of a payment is the sum of the payment in each possible status multiplied by the probability of that status. To see why, let \(B_s\) be the payment if the claimant is in status \(s\) at a payment date, and let \(I_s\) equal 1 if the claimant is in status \(s\) and 0 otherwise. The actual payment is \(\sum_s I_sB_s\). Taking expectations gives
The first equality uses the linearity of expectation; it does not require the status indicators to be independent. If the payment amount within a status is itself random, use the expected payment conditional on that status rather than a fixed \(B_s\). For a present value, multiply each date’s expected payment by the appropriate discount factor and sum across dates.
Worked Example: Eligibility Before the Earnings Offset
A hypothetical monthly contract has a scheduled benefit of \(\$2{,}600\). A claimant is eligible for partial disability only if both work hours are at most 24 per week and earnings are below 80% of pre-disability earnings. Pre-disability earnings are \(\$4{,}000\) per month. If eligible, the contract offsets 50% of earnings above a \(\$500\) disregard and then \(\$300\) of other income. Consider four possible monthly outcomes: no work with probability 0.30; 20 work hours and \(\$2{,}000\) earnings with probability 0.40; 28 work hours and \(\$3{,}500\) earnings with probability 0.20; and full recovery with probability 0.10. Find the expected monthly benefit.
Method. Test eligibility before calculating offsets. The earnings threshold is \(0.80(4{,}000)=\$3{,}200\). With no work, assume the claimant remains disabled and eligible; the payment is \(2{,}600-300=\$2{,}300\). In the partial-work outcome, \(20\leq24\) and \(2{,}000<3{,}200\), so the claimant is eligible. Earnings above the disregard are \(2{,}000-500=\$1{,}500\), giving an earnings offset of \(0.50(1{,}500)=\$750\). The payment is \(2{,}600-750-300=\$1{,}550\).
In the outcome with 28 work hours, the hours test fails because \(28>24\). The claimant is ineligible, so the payment is zero; it would be incorrect to use the offset formula to produce a payment. Full recovery also results in no payment. Therefore, $$ E[B]=0.30(2{,}300)+0.40(1{,}550)+0.20(0)+0.10(0) =690+620=\$1{,}310. $$ The expected monthly benefit is \(\$1{,}310\). As a probability check, \(0.30+0.40+0.20+0.10=1.00\). The key trap avoided is treating an offset as a substitute for an eligibility test.
Common Mistakes and Practical Checks
Return-to-work provisions often combine a status definition, an earnings calculation, and a payment schedule. A reliable calculation keeps these elements separate. For every payment date, identify the status, determine eligibility under all required tests, calculate each permitted offset, apply the floor or cap, and then discount the resulting payment.
- Applying an offset to an ineligible claimant. If a stated hours or earnings condition terminates eligibility, the payment is zero; do not calculate a reduced benefit and pay it anyway.
- Offsetting all earnings instead of earnings above the disregard. For the formula here, the offset base is \(\max(E-G,0)\), not \(E\).
- Forgetting another offset. Add all specified deductions before applying the zero floor. The corrected calculation in the first example subtracts both \(\$400\) and \(\$200\).
- Using a monthly amount as an annual amount. Convert consistently with the payment frequency. Twelve monthly payments of \(\$1{,}800\) total \(\$21{,}600\), not \(\$1{,}800\), for a year.
- Discounting at the wrong time. A year-end payment is discounted by \(v^j\) in year \(j\). Monthly or immediate payments require their actual payment dates.
A concise valuation record can show, for each date, the state probabilities, eligibility decision, benefit by state, expected payment, and discount factor. This layout makes both the contractual interpretation and arithmetic auditable, particularly when the claimant can move between disabled and partial-work states before recovery.
Check Your Understanding
For each problem, apply the contract wording in the order stated. Unless specified otherwise, assume all amounts are in dollars.
- A claimant remains eligible for a monthly benefit of \(\$1{,}800\). The contract offsets 25% of work earnings above a \(\$400\) disregard and then \(\$150\) of other income. Earnings are \(\$1{,}200\). What is the payment? (Answer: \(\$1{,}450\))
- A monthly benefit is \(\$3{,}000\). A claimant is eligible and earns \(\$1{,}600\). The contract offsets 50% of earnings above a \(\$600\) disregard and deducts \(\$400\) of other income. Find the payment. (Answer: \(\$2{,}100\))
- A policy pays \(\$2{,}000\) if eligible. Eligibility requires no more than 20 work hours per week; otherwise the benefit terminates. An eligible claimant earns \(\$1{,}500\), and the policy would offset 40% of earnings above a \(\$500\) disregard. The claimant works 22 hours. What is the payment, and why? (Answer: \(\$0\), because the hours condition makes the claimant ineligible.)
- A hypothetical contract pays at each year-end. At the first year-end, the probabilities of states \(D\), \(P\), and \(R\) are 0.70, 0.20, and 0.10. Annual payments are \(\$18{,}000\) in \(D\), \(\$12{,}000\) in \(P\), and zero in \(R\). If \(i=4\%\), calculate the expected present value of the first payment. (Answer: \(\$14{,}423.08\))
- In one month, a claimant is eligible with probability 0.60 and then receives a benefit of \(\$1{,}400\); with probability 0.25 the claimant is ineligible and receives zero; and with probability 0.15 the claimant has recovered and receives zero. Find the expected monthly payment and check that the probabilities form a complete distribution. (Answer: \(\$840\); the probabilities sum to 1.00.)