Short answer
Below are 10 CFA Level I Quantitative Methods practice questions written for the 2027 curriculum, each with the correct answer and the exact mistake behind every wrong option. They cover the most tested Quant ideas: geometric returns, annualising, continuous compounding, real returns, Bayes' formula, portfolio risk, confidence intervals, Type I and II errors, R² and the new data science subtopic. Aim for 90 seconds per question.
Updated November 2026 · Written for the 2027 Level I curriculum, in which Quant is weighted 11–14%.
Quantitative Methods carries 11–14% of the 2027 Level I exam, up from 6–9%, so it is now worth roughly 20–25 questions. The best way to find out whether you are ready is to answer questions in the exam's format, against the clock, and then study why each wrong option is wrong.
These ten questions follow the real exam's style: three options, one best answer, and wrong options built from the mistakes candidates actually make. Set a timer for 15 minutes, answer all ten, then read the explanations.
How should you use these questions?
| Step | What to do |
|---|---|
| 1 | Set a 15-minute timer (90 seconds per question) |
| 2 | Answer every question without looking ahead; flag any you guessed |
| 3 | Check your answers, including the ones you got right but flagged |
| 4 | For each miss, find which mistake produced the option you chose |
| 5 | Log each miss with its cause: concept, calculation, misread or time |
Step 4 is where most of the learning happens. Our mistake log method explains why.
Question 1: compound annual return
A fund returned 10%, −5% and 7% in three consecutive years. Its compound annual return is closest to:
A. 3.79%
B. 4.00%
C. 11.81%
Answer: A. Compound growth uses the geometric mean: (1.10 × 0.95 × 1.07)1/3 − 1 ≈ 3.79%. Why not B: 4.00% is the arithmetic mean, which ignores compounding and overstates growth when returns vary. Why not C: 11.81% is the cumulative three-year return, not annualised.
Question 2: annualising a short-period return
An investment earns 3% over six months. Assuming the same return is earned again over the next six months, its annualised return is closest to:
A. 3.00%
B. 6.00%
C. 6.09%
Answer: C. Compound two half-year periods: (1.03)2 − 1 = 6.09%. Why not B: 6.00% doubles the rate without compounding. Why not A: 3.00% is the six-month return, not annualised.
Question 3: continuously compounded return
An asset's annual holding period return is 8%. Its continuously compounded return is closest to:
A. 8.33%
B. 7.70%
C. 8.00%
Answer: B. The continuously compounded return is ln(1 + HPR) = ln(1.08) ≈ 7.70%. It is always lower than the holding period return. Why not A: 8.33% is e0.08 − 1, the reverse conversion. Why not C: 8.00% is the holding period return itself.
Question 4: real return
A portfolio's nominal return was 7%, and inflation was 3%. Using the exact relationship, the real return is closest to:
A. 3.88%
B. 4.00%
C. 10.21%
Answer: A. (1.07 / 1.03) − 1 ≈ 3.88%. Why not B: 4.00% is the approximation (7% − 3%); the stem asks for the exact relationship. Why not C: 10.21% multiplies instead of dividing.
Question 5: Bayes' formula
An analyst estimates a 40% probability that a company is in a strong position. If it is strong, there is a 70% chance it beats earnings forecasts; if not, a 30% chance. The company beats forecasts. The updated probability that it is in a strong position is closest to:
A. 70.0%
B. 60.9%
C. 28.0%
Answer: B. P(beat) = 0.7 × 0.4 + 0.3 × 0.6 = 0.46. P(strong | beat) = (0.7 × 0.4) / 0.46 ≈ 60.9%. Why not A: 70% is P(beat | strong), the reverse conditional. Why not C: 28% is the joint probability of being strong and beating, not the conditional one.
Question 6: portfolio standard deviation
A portfolio is split equally between two funds with standard deviations of 20% and 14%. The correlation between them is 0.4. The portfolio's standard deviation is closest to:
A. 14.32%
B. 17.00%
C. 12.21%
Answer: A. √(0.5² × 0.20² + 0.5² × 0.14² + 2 × 0.5 × 0.5 × 0.4 × 0.20 × 0.14) ≈ 14.32%. Why not B: 17.00% is the weighted average of the two standard deviations, which is correct only if the correlation is 1. Why not C: 12.21% leaves out the covariance term, as if the funds were uncorrelated.
Question 7: confidence interval
A sample of 36 monthly returns has a mean of 8% and a standard deviation of 12%. Using a z-value of 1.96, the 95% confidence interval for the population mean is closest to:
A. −15.52% to 31.52%
B. 4.71% to 11.29%
C. 4.08% to 11.92%
Answer: C. The standard error is 12% / √36 = 2%. The interval is 8% ± 1.96 × 2% = 4.08% to 11.92%. Why not A: it uses the standard deviation instead of the standard error. Why not B: it uses 1.645, the 90% critical value.
Question 8: Type I and Type II errors
An analyst lowers the significance level of a test from 5% to 1%, with the same sample size. Compared with before, the test is most likely to have:
A. a higher probability of a Type I error.
B. a lower probability of a Type I error and a higher probability of a Type II error.
C. lower probabilities of both Type I and Type II errors.
Answer: B. The significance level is the probability of a Type I error (rejecting a true null), so lowering it reduces Type I errors. With the same sample size, it becomes harder to reject a false null, so Type II errors rise. Why not C: only a larger sample can reduce both at once. Why not A: it reverses the effect. See our hypothesis testing guide.
Question 9: R² in simple regression
In a simple linear regression, the total sum of squares is 500 and the sum of squared errors is 150. The coefficient of determination (R²) is closest to:
A. 0.70
B. 0.30
C. 0.43
Answer: A. R² = 1 − SSE / SST = 1 − 150 / 500 = 0.70, so the model explains 70% of the variation in the dependent variable. Why not B: 0.30 is SSE / SST, the unexplained share. Why not C: 0.43 divides SSE by the regression sum of squares (150 / 350).
Question 10: financial data science (new for 2027)
An analyst trains a model on historical loan data labelled "defaulted" or "repaid" to predict whether new borrowers will default. This is most likely an example of:
A. unsupervised learning.
B. supervised learning used for classification.
C. a generative model.
Answer: B. The data are labelled, and the target is a category (default or repay), so this is supervised learning for classification. Why not A: unsupervised learning has no labels; it finds structure such as clusters. Why not C: generative models, such as large language models, produce new content rather than predict a labelled target.
Answer key and what each question tests
| Q | Answer | Tests | Most common wrong answer |
|---|---|---|---|
| 1 | A | Geometric mean | Arithmetic mean |
| 2 | C | Annualising with compounding | Doubling the rate |
| 3 | B | Continuously compounded return | Reverse conversion |
| 4 | A | Exact real return | The approximation |
| 5 | B | Bayes' formula | The reverse conditional |
| 6 | A | Portfolio risk with correlation | Weighted average of SDs |
| 7 | C | Confidence interval | SD instead of standard error |
| 8 | B | Type I and II errors | Assuming both fall |
| 9 | A | R² | The unexplained share |
| 10 | B | Supervised versus unsupervised learning | Unsupervised learning |
How did you do?
| Score | What it suggests | Next step |
|---|---|---|
| 9–10 | Quant is a strength | Keep it fresh with weekly timed blocks; move time to weaker topics |
| 7–8 | Solid, with gaps | Reread the sections behind your misses and redo these questions in three days |
| 5–6 | Borderline | Give Quant a full focused week; drill returns and statistics |
| Below 5 | A real gap | Start from foundations: time value of money, return measures, then statistics |
Look at why you missed each question, not just how many. If most misses were misreads ("exact", "annualised", the 90% versus 95% critical value), the fix is careful reading, not more study. If they were concept errors, go back to the material.
Key rule
Calculate before you look at the options. Every question above has a wrong option that matches a half-remembered method. If you look first, that option will pull you toward it.
Three more Quant traps to watch for
- Money-weighted versus time-weighted return. When the client controls the cash flows, judge the manager by the time-weighted return. The money-weighted return mixes in the client's timing. See time-weighted versus money-weighted return.
- t versus z. If the population standard deviation is unknown and the sample is small, use the t-distribution with n − 1 degrees of freedom. Questions often give you both critical values and let the wrong one sit in a distractor.
- Harmonic mean. When equal amounts of money are invested at different prices (cost averaging), the average cost per share is the harmonic mean of the prices, not the arithmetic mean.
Each of these appears regularly in Level I questions, and each has a ready-made wrong option for candidates who reach for the more familiar method. Add them to your flashcards if any feel shaky.
When you redo these questions in a few days, cover the options and answer from scratch first. If you can reach the right number without seeing the choices, the method has stuck.
Where to go next
- For what changed in Quant for 2027 and how much time to give it, see our Level I Quant 2027 guide.
- For how CFA questions are built, see most likely, least likely, closest to.
- For a broader check across all topics, try our free 20-question diagnostic.
- For the calculator keystrokes behind questions 1–4, see our BA II Plus guide.
56 more Quant questions like these
The Clevency Level I 2027 Question Bank has 56 timed Quant questions, and 500 across all ten topics, each explaining the key concept, the working, BA II Plus keystrokes and the exact mistake behind every wrong option. The bundle also includes a Cram Guide, two full 180-question mocks, Formula Sheets, flashcards with an Anki deck, a 12-week Study Planner and a Final 72-Hours Review.
Frequently asked questions
How many Quant questions are on CFA Level 1 in 2027?
Quantitative Methods is weighted 11–14% of the 2027 Level I exam, roughly 20–25 of the 180 questions.
How long should each CFA Level 1 question take?
About 90 seconds on average, since the exam has 90 questions in each 135-minute session.
What Quant topics are most tested on CFA Level 1?
Return measures, time value of money, statistics and probability, sampling and confidence intervals, hypothesis testing, portfolio risk, simple regression and, from 2027, financial data science.
When do you use the geometric mean on the CFA exam?
For compound annual growth over past periods. The arithmetic mean estimates next period's expected return.
What is the difference between a Type I and a Type II error?
A Type I error rejects a true null hypothesis; a Type II error fails to reject a false one. For a given sample size, lowering one raises the other.
Is the financial data science subtopic calculation-heavy?
No. Questions are conceptual, such as classifying a task as supervised, unsupervised or generative, or identifying the main risk of a use case.
Related Reading
- CFA Level 1 Quant 2027 — The topic that nearly doubled
- CFA Level 1 Practice Questions — Free, across topics
- CFA Level 1 Ethics Practice Questions — 8 scenarios with answers