SOA FAM Sample Questions and Answers
These original SOA FAM-style problems illustrate survival probabilities, annuity present values, net premiums, deductible losses and limited-fluctuation credibility.
- They are teaching examples, not recalled SOA questions.
- Define each assumption and payment time before calculating, then check that the result has the right units and interpretation.
On this page11 sections
- Original survival and payment questions
- Original annuity calculation
- Original premium and loss examples
- How to review a worked answer
- Original practice question: partial credibility
- A worked mortality question
- A worked payment question
- A worked credibility question
- A worked annuity question
- Use distractors to diagnose a setup
- Combine examples into a short mixed set
Original survival and payment questions
Question 1: A life table has lₓ=1,000, lₓ₊₁=980 and lₓ₊₂=950. What is the probability that a person alive at age x survives two years? A) 0.020 B) 0.030 C) 0.950 D) 0.969 E) 0.980. Answer C. Use lₓ₊₂/lₓ=950/1,000. The ratio 950/980 is instead conditional survival during year two given survival to x+1.
Question 2: The same table is used. What is the probability of death in the second year conditional on being alive at x+1? Answer: (980−950)/980=30/980≈0.0306. The probability as a fraction of the initial cohort is 30/1,000=0.03. Both are valid quantities; the wording determines the denominator.
Original annuity calculation
An annuity pays $1,000 at the end of each of the next two years only if the life is alive on that payment date. Let survival probabilities be 0.98 and 0.95, and the annual effective interest rate be 4%. Expected present value is $1,000(0.98/1.04+0.95/1.04²)≈$1,820.63. Each payment is discounted and survival-weighted.
If the first $1,000 were paid immediately at time zero, that payment would not be weighted by future survival or discounted. The annuity-due value would therefore add $1,000 to the stream, assuming eligibility at time zero. If payments instead occur monthly or continuously, timing assumptions change the random variable and the calculation.
Original premium and loss examples
A two-year term contract pays $10,000 at the end of the year of death. Initial-cohort death probabilities are 0.02 in year one and 0.03 in year two; i=4%. Expected benefit present value is 10,000(0.02/1.04+0.03/1.04²)=$469.67. If premiums are $P at time 0 and time 1 only if alive, their expected value factor is 1+0.98/1.04, so the equivalence net premium is P=$241.81.
For a ground-up exponential loss with mean $500 and a $200 deductible, no limit or coinsurance, expected payment is E[(X−200)+]=500e^(−200/500)=$335.16. The probability of a positive payment is e^(−0.4). Given a positive payment, memorylessness makes expected payment $500. The unconditional payment distribution has an atom at zero.
How to review a worked answer
Before reading a solution, state the random variable, event, conditioning set, timing and assumptions. After solving, compare those choices with the key. Many wrong options reflect a changed denominator, a payment at the wrong time, or a conditional value confused with an unconditional expectation. Write that distinction in a short error log.
These examples are original teaching material and do not reproduce SOA questions. They demonstrate methods named in the current syllabus, but they do not predict exam items, difficulty or scaled results. Use the SOA’s official sample questions for further practice and follow their current terms.
Original practice question: partial credibility
A limited-fluctuation example specifies a full-credibility count of 3,600 claims. A class has 900 observed claims. Under the square-root rule, what partial credibility applies?
- A. 0.25
- B. 0.50
- C. 0.75
- D. 1.00
- E. 2.00
This is a fresh teaching example. It illustrates the specified rule and does not reproduce an SOA exam item.
A worked mortality question
Suppose a table has l_40=1,000, l_41=980 and l_42=950. What is the probability a life aged 40 survives two years? It is l_42/l_40=950/1,000=.95. The probability of death in the second year conditional on reaching age 41 is (980-950)/980=30/980, about.0306. From age 40 without conditioning, the joint event of surviving year one and dying in year two has probability 30/1,000=.03.
The distinction is denominator choice. The conditional quantity uses the number alive at the start of the second year; the unconditional quantity starts with the original age-40 cohort. Read the question’s phrase “given survival to 41” carefully. This original example illustrates life-table conditioning and is not an official SOA item.
A worked payment question
Let X be exponential with mean $500 and let the insurer pay above a $200 deductible without a limit. Payment is Y=(X-200)_+. The probability of positive payment is P(X>200)=e^(-.4), about.6703. Conditional on a positive payment, the excess X-200 has mean $500 by the memoryless property. Therefore E[Y]=.6703×500, about $335.16.
The conditional mean among positive claims is $500, while the unconditional mean includes zero payments and is about $335.16. A question that asks “given payment is positive” wants the former. This contrast tests conditioning rather than just integration.
A worked credibility question
Suppose a question supplies N=3,600 claims as the full-credibility standard and observed experience n=900. Under the square-root limited-fluctuation rule, Z=min(1,sqrt(n/N))=.5. If the experience estimate is 120 and the complement is 100, the credibility-weighted estimate is.5(120)+.5(100)=110.
The supplied N is part of this example. It is not a universal number for full credibility. Confirm that experience and standard use compatible units, and then use 1-Z for the complement weight. A result between 100 and 120 is a useful check.
A worked annuity question
A life annuity pays $1,000 at the ends of years one and two. The probability of surviving to each payment is.98 and.95, and i=.04. The expected present value is 1,000(.98/1.04 +.95/1.04^2), approximately $1,820.63. If instead the first $1,000 is paid immediately and certainly, add $1,000 to this two-payment deferred pattern.
The example separates survival from discounting. The year-two payment uses survival through two years, not the one-year probability. The certain time-zero payment has no discount or survival factor when due at issue. This is an original teaching calculation.
Use distractors to diagnose a setup
A useful practice explanation can state what each distractor assumes. In the annuity example, an answer that discounts both payments but ignores survival is the certain-annuity value. An answer that uses.98 for both payments ignores the second-year mortality. In the deductible example, $500 may be a conditional excess mean, not unconditional insurer payment. Diagnosing the distractor helps identify a reusable misconception.
After review, change a single input and solve again. Change the interest rate, survival probability or deductible. If the same structure still makes sense, the method has transferred. If only the original numbers can be reproduced, return to the model definition.
Combine examples into a short mixed set
A small mixed set can include a life-table probability, a deductible payment, a premium equivalence calculation, and a limited-fluctuation blend. Before solving, identify the model family and requested quantity. After solving, write one sentence stating the assumption that mattered. This mimics the switching challenge of a broad multiple-choice examination without claiming to reproduce its item bank.
Review the set after a delay. Solve each item again with one changed input and see whether the result moves in the expected direction. A changed interest rate should affect discounting; a larger deductible should not increase expected insurer payment; a changed full-credibility count should alter the square-root weight. These variations test understanding more effectively than repeating the same values immediately.
Common questions
Are these official FAM questions?
No. They are original teaching examples, not recalled or reproduced SOA questions.
Do these examples predict a score?
No. They explain concepts and do not model the SOA form or scaling.
What is the best way to review a solution?
Check the model, assumptions, conditioning, timing, units and interpretation.