How to use this question set

SOA Exams cover different syllabuses, so there is no single set of questions that represents every exam. These nine examples practise common actuarial calculations in probability, loss models and interest theory; they are not official questions or a substitute for checking the syllabus and sample materials for your specific exam on the Society of Actuaries’ website.

Answer each question when it appears, then read the explanation straight away. For every calculation, identify what is given and what is required, select a method, keep units and notation visible, and check whether the result is plausible. Under time pressure, a quick check of scale, bounds and signs can catch an otherwise avoidable error.

Probability and loss models

1. Poisson probabilities

This checks whether you can use a probability mass function rather than confuse an exact count with a cumulative probability. The key is to substitute the stated count into the Poisson formula.

The first browser question looks like this:

Question 11 mark

The number of claims reported in a period has a Poisson distribution with mean 3. What is the probability of exactly 2 claims?

For X ~ Poisson(3), P(X=2) = e^(-3) 3^2 / 2! = 0.2240. The strongest distractor, 0.1494, is P(X=1) = e^(-3) × 3; it uses the wrong count.

The factorial matters: exactly two claims has probability about 22.4%, not the probability of one claim or of two or fewer claims. On a timed paper, writing the event first—P(X = 2)—helps keep the calculation attached to the question.

2. Bayes’ theorem and base rates

This tests conditional probability. A frequent trap is to report the sensitivity as the chance that a positive test indicates the condition; those are different conditional probabilities.

A second question tests whether you account for the base rate:

Question 21 mark

A condition affects 1% of a population. A test has sensitivity 90% and specificity 95%. For a randomly selected person with a positive result, what is the approximate probability that the person has the condition?

Among 10,000 people, about 100 have the condition and 90 test positive. Of 9,900 without it, 5%—495—test positive. The posterior probability is 90/(90+495) = 15.4%. The strongest distractor is 90%, the sensitivity; it is P(positive | condition), not P(condition | positive).

A frequency table is a useful alternative to writing Bayes’ theorem directly. It makes the two groups of positive results visible and helps prevent the base-rate error.

3. Binomial probability

This question checks the distinction between an expected count and the probability of exactly one event. The setup must include both the choice of which trial succeeds and the failures on the other trials.

The next question applies that distinction:

Question 31 mark

Ten independent policies each have probability 0.02 of a claim during a period. What is the probability of exactly one claim?

For X ~ Binomial(10, 0.02), P(X=1) = C(10,1)(0.02)(0.98)^9 = 0.1667. The strongest distractor is 0.2000 = np, the expected number of claims. An expected count is not the probability of exactly one claim.

The result is less than the expected count of 0.2, as a probability must be; that alone is not a proof, but it is a useful quick check. Keep the independence assumption in view: without it, this binomial model may not apply.

4. Deductibles and expected payment

This checks whether you integrate the insurer’s payment rather than the original loss. For a deductible, the payment is zero below the deductible and rises only by the amount above it.

Use that payment definition in this question:

Question 41 mark

A loss X is uniformly distributed on [0, 1,000]. An insurer pays max(X − 200, 0). What is the expected payment?

E[(X−200)+] = (1/1,000)∫ from 200 to 1,000 of (x−200) dx = 320. The strongest distractor is 400, half the 800-wide interval above the deductible; it is not the expected payment because the payment rises from 0 to 800 across that interval.

The integral can also be seen as the area of a triangle with base 800 and height 800, divided by the uniform density’s interval width of 1,000. State the payment function before integrating so the deductible is not accidentally applied in the wrong direction.

5. Aggregate claims

This tests a compound Poisson variance. For a random number of independent claim amounts, the variance uses the severity’s second moment, not just its variance.

The fifth question focuses on that distinction:

Question 51 mark

Claim count N is Poisson with mean 4. Independent claim amounts have mean 500 and variance 40,000. For aggregate claims S, what is Var(S)?

For a compound Poisson sum, Var(S) = λE[X²] = 4(Var(X) + E[X]²) = 4(40,000 + 250,000) = 1,160,000. The strongest distractor, 160,000, uses λVar(X) and omits the contribution from the random claim count.

Check the identity E[X²] = Var(X) + E[X]² before substituting. This is also a useful example of why a familiar formula should be matched to its assumptions and random variables, not recalled by name alone.

Interest theory

6. Annuity-immediate present value

This checks timing as well as formula selection: five payments at the ends of years 1 through 5 form an annuity-immediate. Do not value them as payments at the start of each year.

The sixth question tests that timing:

Question 61 mark

An annuity-immediate pays 1,000 at the end of each year for 5 years. At an effective annual interest rate of 5%, what is its present value?

PV = 1,000 a-angle-5 at 5% = 1,000[1−(1.05)^(-5)]/0.05 = 4,329.48. The strongest distractor, 4,545.95, is the value one year later (the annuity-due value); it does not discount the first payment from year 1 to time 0.

A quick timing check is to list payment dates before applying a formula. Every payment is at least one year in the future, so the present value must be below the undiscounted total of 5,000.

7. Force of interest

This tests continuous accumulation. Use the exponential accumulation function when the force of interest is constant; do not substitute the force directly as an effective annual rate.

The seventh question is a direct accumulation calculation:

Question 71 mark

An amount of 1,000 accumulates for 3 years at a constant force of interest δ = 0.04. What is the accumulated value?

With constant force δ, the accumulation factor is e^(δt). Thus 1,000e^(0.04×3) = 1,127.50. The strongest distractor, 1,120, applies simple interest; the stated force calls for continuous exponential accumulation.

Check the exponent is dimensionless: the annual force is multiplied by years. Comparing the result with simple interest can also flag an implausible calculation; continuous compounding gives a slightly higher accumulation here.

8. Nominal rate convertible monthly

This tests conversion from a nominal annual rate to an effective annual rate. Monthly conversion means twelve monthly accumulation periods, not one application of the nominal rate.

The eighth question checks the conversion:

Question 81 mark

A nominal annual interest rate of 12% is convertible monthly. What is the effective annual rate?

The monthly rate is 0.12/12 = 0.01, so the effective annual rate is (1.01)^12−1 = 0.126825, or about 12.68%. The strongest distractor, 12.00%, treats the nominal rate as effective and ignores monthly compounding.

Keep the rate period explicit in your working: 1% per month compounded for 12 months. That notation makes the conversion easier to audit and less likely to become a percentage-versus-decimal slip.

9. Standard normal tail probability

This tests whether you interpret a normal quantile as an upper-tail or lower-tail probability. The sign and direction of the inequality matter as much as the table lookup.

The final question is about an upper tail:

Question 91 mark

Let Z have a standard normal distribution. What is P(Z > 1.645), approximately?

The standard normal 95th percentile is approximately 1.645, so the area to its right is 0.05. The strongest distractor, 0.95, is the area to the left. Sketching the curve and marking the right tail prevents reversing the probability.

For exams with constructed-response questions

Some SOA exams include written-answer or constructed-response components; check the current details for your exam rather than assuming every exam uses the same format. In a response, make the model and assumptions clear before giving a number. A strong calculation can still lose credit if it omits the distinction between a variance and a second moment.

A marked response on aggregate claims shows that kind of omission:

Paper — Actuarial probability and aggregate claims06:20
78%Actuarial probability and aggregate claims — marked3/4 marks · 06:20 taken

N is Poisson with mean 4. Claim amounts are independent of N, with E[X]=500 and Var(X)=40,000. State Var(S) for S = X₁ + … + X_N and show the key step.

3/4

Var(S)=4 × 40,000 = 160,000. The claim count is Poisson, so multiply the severity variance by the mean claim count.

You identified the compound Poisson setting, but used Var(X) where the formula requires E[X²]. The random number of claims contributes additional variation.

Missed

Convert the severity variance to its second moment: E[X²] = 40,000 + 500² = 290,000.

Model answerGiven/Goal: N is Poisson with λ=4; find Var(S). Severity moment: E[X²] = Var(X) + E[X]² = 40,000 + 250,000 = 290,000. Compound Poisson result: Var(S) = λE[X²] = 4 × 290,000 = 1,160,000. The variance exceeds λVar(X) because the random claim count also contributes variation.

That is a MySummaries paper, filled with SOA Exams material. Yours is written from your own notes. Start free

The useful repair is to write the second-moment identity whenever a compound Poisson variance appears. That one line exposes the missing term before it reaches the final answer.

Review the errors, not just the score

Separate errors into categories: wrong model, wrong formula, incorrect substitution, arithmetic, or interpretation. Then choose a repair that matches the error. Revisit the relevant syllabus material for a model-selection gap; practise a short set of calculations for arithmetic or setup slips; and keep a brief record of the traps you repeat.

A weak-area view can help prioritise the next revision block:

Where marks go missing
54%Compound Poisson variance5×
68%Conditional probability7×
86%Interest conversion6×

If the same mark is lost twice, turn it into a precise prompt rather than rereading a whole chapter. For example, ask which moment appears in the compound Poisson variance formula and why.

A remediation prompt for that repeated error looks like this:

Remediation tray

You lost this mark twice: for a compound Poisson aggregate S, do you use λVar(X) or λE[X²], and how do you obtain the required severity moment?

Add cardDismiss

How MySummaries helps

MySummaries can build a revision board from your own notes and use it to create question practice, written responses and review cards. For SOA preparation, start with the material for your specific exam, then use your missed marks to decide what to revisit. Explore MySummaries.