FRM Exam Part II · Estimating Market Risk Measures: An Introduction and Overview

QQ Plots and Assessing Distributional Fit for Returns

Updated 11 October 2026 · Fact-checked

A QQ plot graphs the quantiles of your data against the quantiles of a reference distribution, usually the normal. If points follow a straight line, the fit is good. Points that curve away at both ends signal fat tails. A bowed, one-sided curve signals skewness. Read the ends first.

Understand QQ Plots and Assessing Distributional Fit

A quantile-quantile (QQ) plot compares two distributions. You sort your sample, compute each observation's quantile position, and plot it against the quantile of the reference distribution at the same position. For returns, the reference is usually the standard normal.

If the data truly come from the reference distribution, the points fall on a straight line. If the data differ only in mean and standard deviation, the points still fall on a straight line. The intercept reflects the difference in location and the slope reflects the difference in scale. So a QQ plot tests shape, not location or scale.

Shape differences show up as departures from the line. Fat tails mean extreme observations are more extreme than the normal predicts. With sample quantiles on the vertical axis, the left end falls below the line and the right end rises above it, giving an S-shape (or reversed S, depending on axis choice). Skewness gives a one-sided bend: one tail departs from the line while the other stays close, or the whole plot curves like a bow.

Risk managers care because VaR and Expected Shortfall at high confidence depend on the tails. A normal-based VaR on fat-tailed data understates losses far in the tail. The QQ plot is a visual tool. It shows where the fit fails, but it gives no formal test statistic or p-value.

Small samples are noisy. The extreme points in a QQ plot are the most variable, so a few outliers at the ends do not always prove fat tails. Check how many points deviate and how far.

Key formulas to remember

Plotting position: p(i) = (i − 0.5) ÷ n; One common choice for the probability assigned to the i-th smallest of n sorted observations. Other conventions exist; the exam focuses on interpretation.
Reference quantile: x(i) = Φ⁻¹(p(i)); Φ⁻¹ is the inverse standard normal CDF. This is the horizontal-axis value for the i-th point.
Standardisation: z(i) = (r(i) − mean) ÷ standard deviation; Standardising lets you compare against the standard normal. A good fit lies near the 45-degree line.
Linear relation under same shape: Data quantile = μ + σ × reference quantile; If data are normal with mean μ and standard deviation σ, the QQ plot against the standard normal is a straight line with intercept μ and slope σ.

How to solve QQ Plots and Assessing Distributional Fit questions

Use this order for any question that shows a QQ plot or describes one in words.

1Identify the axes: which axis holds the sample quantiles and which holds the reference (usually normal) quantiles.
2Find the reference line. Check whether the points follow a straight line through the middle of the data.
3Look at the middle of the plot. Points near the line there mean the body of the distribution fits.
4Look at each tail separately. Note whether points lie above or below the line at the far left and far right.
5Classify: both tails departing outward means fat tails (excess kurtosis); one tail departing or a bow means skewness.
6Separate shape from location and scale: a straight line with slope different from 1 or intercept different from 0 is still the same shape.
7State the risk implication: a normal VaR or ES at a high confidence level will understate tail losses if tails are fat.
8Match your conclusion to the option wording and avoid claims a plot cannot prove, such as a formal statistical rejection.

Quickest way: Read the ends, then the slope

When to use it: Use when a multiple-choice question gives a picture or a short description of a QQ plot and you have about a minute.

Ignore the middle. Check whether the plot is a straight line there.
Check the two ends. Both ends bending away on opposite sides of the line means fat tails.
If only one end bends, or the plot curves in a bow, choose skewness.
If it is a straight line with a different slope, it is the same shape with different scale, so the fit is fine.
Eliminate options that claim a QQ plot gives a formal test or p-value.

Common mistakes in QQ Plots and Assessing Distributional Fit

Treating a straight line with slope not equal to 1 as a bad fit.
Students expect the 45-degree line only.
Fix: A straight line of any slope or intercept means the same shape. Slope reflects scale and intercept reflects location.
Confusing fat tails with skewness.
Both show departures from the line at the ends.
Fix: Fat tails depart at both ends in a symmetric S-pattern. Skewness is asymmetric: one tail departs more, or the plot bows.
Reading the axes backwards.
Textbooks differ on which quantiles go on which axis.
Fix: Read the axis labels first. Decide the direction of the S-shape only after you know which axis is the sample.
Saying the QQ plot proves or rejects normality statistically.
Graphical evidence feels conclusive.
Fix: A QQ plot is a visual diagnostic. Formal conclusions need tests such as Kolmogorov-Smirnov or Anderson-Darling.
Over-reading a few extreme points in a small sample.
Tail points are the most visible but the noisiest.
Fix: Ask how many points deviate and whether the sample is large. A single outlier is weak evidence of fat tails.

Worked examples

Example 1

A QQ plot compares daily returns of an equity index (vertical axis) with standard normal quantiles (horizontal axis). The points lie close to a straight line in the centre. The lowest few points fall well below the line and the highest few rise well above it. What does this indicate, and what is the effect on a normal 99% VaR?

Show the solution

The centre is on the line, so the body of the distribution is close to normal.
The lowest points are below the line: actual left-tail returns are more negative than the normal predicts.
The highest points are above the line: actual right-tail returns are larger than the normal predicts.
Departures at both ends in this pattern mean fat tails (excess kurtosis), not skewness, because the pattern is symmetric.
A normal-based 99% VaR uses the normal tail, so it understates the loss that the actual left tail produces.

Answer: The returns have fat tails. A normal 99% VaR would understate the true tail loss.

Example 2

Monthly returns of a fund have mean 1% and standard deviation 4%. A QQ plot against the standard normal shows a perfectly straight line. Which line would you expect, and what does it say about the distribution?

Show the solution

If returns are normal with mean μ = 1% and σ = 4%, then return = 1% + 4% × z.
So the QQ plot against the standard normal is a straight line with intercept 1% and slope 4%.
A straight line means the shape matches the normal; no fat tails or skewness are visible.
The slope not equalling 1 does not indicate misfit. It only reflects that σ is not 1.

Answer: A straight line with intercept 1% and slope 4%, consistent with normally distributed returns.

Exam tips

Expect a description or sketch and a choice among fat tails, skewness and good fit. Decide using the ends first.
Remember a straight line of any slope is still a good fit in shape.
Link every conclusion to risk: fat tails mean normal VaR and ES understate extreme losses.
Watch for options that overclaim, such as saying a QQ plot gives a formal test.

Practice questions from Estimating Market Risk Measures: An Introduction and Overview

QQ Plots and Assessing Distributional Fit: frequently asked questions

What does a QQ plot tell you about normality?

It shows whether the shape of your data matches the normal. A straight line supports normality. Curvature at the ends or a bow shows fat tails or skewness.

How do you spot fat tails on a QQ plot?

Look for points at both ends departing from the line in the direction of more extreme values than the normal predicts. The centre stays close to the line, giving an S-shaped pattern.

Is a QQ plot a statistical test?

No. It is a graphical diagnostic with no test statistic. Formal tests such as Kolmogorov-Smirnov or Anderson-Darling give a decision rule.

Why does the slope of the line not matter for fit?

Changing the mean shifts the line and changing the standard deviation tilts it. Both leave the shape unchanged, so the points still lie on a straight line.