CFA Level I Exam · Estimation and Hypothesis Testing

Central Limit Theorem and Standard Error of the Mean

Updated 6 October 2026 · Fact-checked

The central limit theorem says that for large random samples (usually n ≥ 30) from a population with finite variance, the sample mean is approximately normally distributed, whatever the population shape. Its mean is μ and its standard deviation, the standard error, is σ ÷ √n, or s ÷ √n when σ is unknown.

Understand Central Limit Theorem and Standard Error

Take a random sample from a population and compute its mean. Take another sample and compute again. The means will differ from sample to sample. If you repeat this many times, the sample means form their own distribution. This is the sampling distribution of the sample mean.

The central limit theorem (CLT) describes that distribution. For random samples of size n from any population with mean μ and finite variance σ², the sample mean is approximately normal when n is large. Its mean equals μ and its variance equals σ² ÷ n. The population itself does not need to be normal. For CFA Level I, n ≥ 30 is the usual rule of thumb for 'large enough'. It is a guide, not a law: very skewed populations may need more.

The spread of the sample mean is the standard error of the mean: σ ÷ √n when the population standard deviation is known, and s ÷ √n when you estimate it from the sample. Larger samples give a smaller standard error, but you must quadruple n to halve it, because of the square root.

Standard deviation versus standard error. Standard deviation measures how spread out individual observations are. Standard error measures how spread out the sample mean is across repeated samples. The standard error is always smaller than the standard deviation when n > 1.

In finance, this is why an average of monthly returns is more reliable than any single month, and why confidence intervals and t-tests on mean returns work even when returns are not perfectly normal.

Key formulas to remember

Standard error (population σ known): σx̄ = σ ÷ √n; σ is the population standard deviation, n is the sample size.
Standard error (σ unknown): sx̄ = s ÷ √n; s is the sample standard deviation. This is the usual case in practice.
CLT distribution of the sample mean: x̄ ~ approximately N(μ, σ² ÷ n); Needs random sampling, finite variance and a large n (rule of thumb n ≥ 30).
Standardized sample mean: z = (x̄ − μ) ÷ (σ ÷ √n); Use this to find probabilities for a sample mean. With s instead of σ, a t-statistic is used in testing.

How to solve Central Limit Theorem and Standard Error questions

Use this method for any question on the sampling distribution of the mean or the standard error.

1Identify what is asked: a single observation or the sample mean. This decides whether you divide by √n.
2Write down μ, the standard deviation (σ or s) and n. Check that the standard deviation is for individual observations, not already a standard error.
3Check the CLT conditions: random sample, finite variance, n large enough (about 30 or more) if the population is not normal.
4Compute the standard error: standard deviation ÷ √n.
5If a probability is needed, standardize: z = (x̄ − μ) ÷ standard error, then use the normal table.
6Sanity-check: the standard error must be smaller than the standard deviation, and a larger n must give a smaller standard error.
7Pick the option that matches. Remove options that use σ instead of σ ÷ √n.

Quickest way: Divide by the square root of n

When to use it: Any numerical question asking for standard error or how it changes with sample size.

Compute √n mentally or on the calculator (BA II Plus: type n, press 2nd, then x² key for √x).
Divide the standard deviation by that number.
For 'how does it change' questions, use ratios: new SE = old SE × √(old n ÷ new n). Quadrupling n halves SE.
Remove the option equal to the raw standard deviation, then compare the remaining two against your estimate.

Common mistakes in Central Limit Theorem and Standard Error

Using the standard deviation instead of the standard error when working with a sample mean.
Both are called 'spread', and the question may give only σ and n.
Fix: If the question is about x̄, always divide σ by √n first.
Dividing by n instead of √n.
Mixing up with the variance of the mean, which is σ² ÷ n.
Fix: Standard error uses √n. Variance of the mean uses n.
Thinking the CLT makes the population normal.
The word 'normal' is attached to the result and gets misread.
Fix: The CLT applies to the distribution of the sample mean, not to the individual observations.
Treating n ≥ 30 as a guarantee.
Notes present it as a rule without caveats.
Fix: It is a rule of thumb. Remember it also needs random sampling and finite variance.
Expecting halving the standard error by doubling the sample size.
Ignoring the square root.
Fix: Doubling n cuts SE by a factor of √2, about 29%. You need four times n to halve it.

Worked examples

Example 1

The monthly returns of a global equity fund have a standard deviation of 4.0%. An analyst takes a random sample of 64 months. What is the standard error of the sample mean return? A) 0.50% B) 4.00% C) 32.00%

Show the solution

Standard deviation s = 4.0%, n = 64.
√64 = 8.
Standard error = 4.0% ÷ 8 = 0.50%.
Option B is the raw standard deviation. Option C multiplies the standard deviation by √n (4.0% × 8 = 32.00%) instead of dividing.

Answer: A) 0.50%

Example 2

Annual returns on a bond index have a mean of 5.0% and a standard deviation of 6.0%. For a random sample of 36 years, what is the approximate probability that the sample mean exceeds 6.0%? (Use the normal table: P(Z ≤ 0.17) = 0.5675 and P(Z ≤ 1.00) = 0.8413.) A) 0.1587 B) 0.4325 C) 0.8413

Show the solution

Standard error = 6.0% ÷ √36 = 6.0% ÷ 6 = 1.0%.
With n = 36 (above 30), the CLT gives an approximately normal sample mean.
z = (6.0% − 5.0%) ÷ 1.0% = 1.00.
P(x̄ > 6.0%) = 1 − P(Z ≤ 1.00) = 1 − 0.8413 = 0.1587.
Option C is the probability of being below 6.0%, the wrong tail.
Option B uses the raw standard deviation (6.0%) as the standard error. That gives z = 1.0 ÷ 6.0 ≈ 0.17, and P(Z > 0.17) = 1 − 0.5675 = 0.4325, using the table value given in the problem.

Answer: A) 0.1587

Exam tips

Look for the phrase 'sample mean'. It signals that you need the standard error, not the standard deviation.
The wrong options often include the raw standard deviation. Compute √n first and eliminate it.
For 'what happens if n increases' questions, answer with direction and the square root relationship. No calculation is needed.
If the question states the population is normal with known variance, the sample mean is normal for any n. The CLT is only needed for non-normal populations.
With 90 seconds per question, keep the arithmetic simple: pick n values that are perfect squares and check for that pattern.

Practice questions from Estimation and Hypothesis Testing

Central Limit Theorem and Standard Error in other exams

The same ground in other exams, if you are preparing for more than one or want another angle on it.

FRM Part ICentral Limit Theorem

Central Limit Theorem and Standard Error: frequently asked questions

What is the central limit theorem in simple words?

If you take large random samples from a population with finite variance, the averages of those samples form a roughly normal distribution. This holds even if the population itself is skewed. The centre of that distribution is the population mean.

What is the difference between standard deviation and standard error?

Standard deviation describes how far individual observations lie from their mean. Standard error describes how far the sample mean is likely to lie from the true mean. Standard error equals standard deviation ÷ √n, so it is smaller.

How do I calculate the standard error of the sample mean?

Divide the standard deviation by the square root of the sample size. Use σ if the population standard deviation is known, otherwise use the sample standard deviation s. For s = 10 and n = 25, the standard error is 10 ÷ 5 = 2.

Is a sample size of 30 always enough for the CLT?

No. It is a common rule of thumb for Level I. Highly skewed or fat-tailed populations can need larger samples. The CLT also requires random sampling and finite variance.