Statistical Measures of Asset Returns: The Intuitive Guide to Volatility, Downside Risk, Skewness & Correlation

Let me ask you a question that catches almost every new finance student off guard:

Suppose you’re an allocator looking at two hedge fund managers, Manager A and Manager B. Over the last five years, both reported an identical average annual return of +12.0%.

On the surface, an untrained investor might shrug and say, “They’re equally skilled, just flip a coin.”

But let’s pull back the curtain and inspect their actual monthly distributions:

• Manager A earned steady, boring monthly gains between $+0.8\%$ and $+1.5\%$, compounding smoothly without a single gut-wrenching drawdown.
• Manager B picked up tiny $+0.5\%$ premiums for 58 months and then suffered a terrifying $-42\%$ crash in month 59 before clawing back with leveraged moonshots.

Would you sleep well with your life savings in Manager B’s fund? Absolutely not.

Raw historical returns tell only a tiny fraction of the story. In real markets, how you arrived at that return—the volatility, the skewness, the fat tails, the catastrophic downside shortfalls—matters far more than the final average.

Whether you’re grinding through CFA Level I (Quantitative Methods — Module 3) or managing your own investment portfolio, this guide is my personal whiteboard walkthrough of every key statistical measure you’ll ever need.

I’ve also embedded our interactive Texas Instruments BA II Plus financial calculator emulator and visual sandboxes right into the page. As we go through each concept, test the numbers, drag the sliders, and watch the math come alive!

---

TI

Institutional ToolTI BA II Plus Emulation

Financial & Statistical Calculator

1-VAR / STATSTVM Finance

Preset Solvers (13)

2ndCPTSTATDEC 4

STAT: x̄ = 3.00%, Sx = 2.53%

3.00

STATISTICS (1-VAR & LIN)n = 10

CPT2ndCLR WRKCE/C+/-

1-VAR Statistics Outputs (Click to View on LCD)

x̄ (Mean)3.00Sx (Sample SD)2.53σx (Pop SD)2.40n (Size)10CV (Risk/Ret)0.843

789÷456×123-0.=+

[2nd] + [DATA (7)] = Data Entry[2nd] + [STAT (8)] = 1-VAR Stats

1-VAR Statistics Outputs (10 Observations)

Clear All

Sample Mean (x̄)

3.000%

Sample SD (Sx / n-1)

2.528%

Pop SD (σx / n)

2.398%

Sample Var (Sx²)

6.389%²

Coeff of Var (CV)

0.843

Median (Q2)

3.75%

Target Downside Dev (B = 2%):1.500%

Interquartile Range (IQR = Q3 - Q1):3.63% (Q1: 1.13%, Q3: 4.75%)

5% Trimmed Mean / 95% Winsorized:3.25% / 3.00%

Skewness / Excess Kurtosis:-0.90 / 1.50

2-Var Covariance / Pearson r:Cov: -2.89 | r = -0.968

Target Return Threshold (B):

%

Data Worksheet (X & Y Values)

10 items entered

Add

#	X Value	Y Value	Action
01	4.50%	2.00%
02	6.00%	1.00%
03	1.50%	3.50%
04	-2.00%	5.00%
05	0.00%	4.00%
06	4.50%	2.50%
07	3.50%	3.00%
08	2.50%	3.20%
09	5.50%	1.50%
10	4.00%	2.80%

---

The Master Statistical Cheat Sheet

Whenever you look at any financial dataset, you are really asking five fundamental questions:

What you want to know	The statistical metric you need	Why it matters in the real world
Where is the center of performance?	Mean, Median, Mode	Expected returns and central benchmarks
Did extreme outliers distort the data?	Trimming, Winsorizing, Outlier Fences	Data scrubbing and robust estimation
Where does a specific fund rank?	Quantiles, Percentiles, Interquartile Range ($IQR$)	Peer rankings and decile performance
How volatile are the swings?	Sample Variance ($s^2$), Standard Deviation ($s$)	Total uncertainty and the Sharpe Ratio
How painful are real downside losses?	Target Downside Deviation (Semideviation)	Hurdle shortfall risk and the Sortino Ratio
How much risk per unit of return?	Coefficient of Variation ($CV$)	Comparing risk across different currencies/scales
Is the distribution lopsided?	Skewness (Positive vs. Negative)	Equity crash risk vs. Venture capital lottery payoffs
How often do black swan crashes happen?	Kurtosis / Excess Kurtosis	Tail-risk hedging and Value at Risk ($VaR$)
Do two assets move together?	Covariance ($s_{XY}$), Pearson Correlation ($r_{XY}$)	Portfolio diversification and hedging

---

1. Central Tendency: Where is the Middle of Your Returns?

When you ask, “What’s the typical return for this stock?”, you’re looking for a measure of central tendency. But here’s the catch: depending on the shape of your data, different measures give wildly different answers.

1.1 The Arithmetic Sample Mean ($\bar{X}$)

This is the standard average you’ve known since middle school. For a sample of $n$ observations:

$$\bar{X} = \frac{\sum_{i=1}^n X_i}{n}$$

Let’s say your portfolio earned $+4\%$, $+6\%$, and $+10\%$ across three quarters:

$$\bar{X} = \frac{4 + 6 + 10}{3} = \mathbf{6.67\%}$$

💡 The Trap with Arithmetic Mean: The mean has zero defense against extreme outliers. A single $+150\%$ meme-stock surge or an $-80\%$ crypto collapse will yank the arithmetic mean in its direction, tricking you into believing performance is better (or worse) than what happened on a typical day.

1.2 The Median (The 50th Percentile Anchor)

The median is the literal middle observation when you arrange all your data points from smallest to largest:

• If you have an odd number of points, it’s the exact middle value at rank $\frac{n+1}{2}$.
• If you have an even number of points, it’s the arithmetic average of the two middle values at ranks $\frac{n}{2}$ and $\frac{n}{2} + 1$.

Why do I love the median? It is completely resistant to outliers. Whether your worst monthly return was $-3\%$ or $-95\%$, the median doesn’t budge by a single basis point.

1.3 The Mode

The mode is simply the most frequently occurring value in your dataset. It’s the only measure of central tendency you can use for categorical (nominal) data (like classifying fund styles as Value, Growth, or Blend). A return series can have one peak (unimodal), two peaks (bimodal), or no mode at all.

---

2. Outlier Defense: Trimming vs. Winsorizing

Imagine you’re analyzing a ten-year track record of a quantitative trading strategy, and you discover two data points corrupted by fat-finger errors: a $-18\%$ flash crash and a $+42\%$ anomalous tick.

If you calculate a simple arithmetic mean, those two numbers will wreck your backtest. What should you do? In quantitative finance, we have two primary lines of defense:

┌────────────────────────────────────────────────────────────────────────┐
│                        HOW TO CLEAN OUTLIER DATA                       │
├──────────────────────────────────┬─────────────────────────────────────┤
│         TRIMMING                 │           WINSORIZING               │
│    (Deletes observations)        │      (Replaces observations)        │
│                                  │                                     │
│  5% Trimmed Mean:                │  95% Winsorized Mean:               │
│  We delete the lowest 2.5%       │  We replace the lowest 2.5% with    │
│  and highest 2.5% of values.     │  the 2.5th percentile cutoff, and   │
│  Your sample size (n) shrinks!   │  the top 2.5% with the 97.5th!      │
└──────────────────────────────────┴─────────────────────────────────────┘

🧠 My Favorite Memory Trick: Trim like a haircut (snip off the ends). Winsorize like a border fence (cap them at the line).

Try interacting with the box-plot below to see this in action:

Quantiles & Outlier Defense

Box-and-Whisker Fences: Trimming vs. Winsorizing

Observe how extreme data points breach the 1.5×IQR fences and distort the raw arithmetic mean.

Clean Normal Dataset

Outlier Lower Fence (Q1 - 1.5×IQR): -5.1%Outlier Upper Fence (Q3 + 1.5×IQR): 17.9%

IQR (Middle 50%) Median (6.5%) Arithmetic Mean (7.5%)

IQR = 5.75%

Raw Arithmetic Mean

7.50%

Sensitive to outliers

Trimmed Mean (Drop Tails)

6.38%

Deletes outer extreme values

Winsorized Mean (Cap Tails)

6.30%

Replaces tails with cutoffs

Median (Q2)

6.50%

Completely resistant

Memory Rule: Trim = remove observations from tails. Winsorize = replace tail values with fence boundaries. When managing risk data with massive outlier shocks, trimming and winsorizing preserve the true underlying central tendency!

Try Outlier Scrubbing in BA II PlusResult: 6.38%

Raw Mean = 7.50% vs. 10% Trimmed = 6.38% vs. 90% Winsorized = 6.40%

Solve in Calc

---

3. Location & Quantiles: Where Does an Observation Stand?

When a pension fund says they only allocate to managers in the top quartile, how do we mathematically calculate those cutoff lines?

Quantiles cut a sorted dataset into equal slices:

• Quartiles: 4 equal quarters ($Q_1 = 25\%$, $Q_2 = \text{Median} = 50\%$, $Q_3 = 75\%$).
• Quintiles: 5 equal slices ($20\%$ chunks).
• Deciles: 10 equal slices ($10\%$ chunks).
• Percentiles: 100 equal slices ($1\%$ chunks).

The CFA Percentile Location Formula

To find where the $y^{\text{th}}$ percentile sits among $n$ sorted observations, use:

$$L_y = (n + 1) \left(\frac{y}{100}\right)$$

If $L_y$ gives you a decimal (like $4.25$), don’t panic! We just use linear interpolation between the $4^{\text{th}}$ and $5^{\text{th}}$ values:

$$P_y = X_k + (L_y - k)(X_{k+1} - X_k)$$

The Interquartile Range ($IQR$) & Tukey’s Fences

The Interquartile Range ($IQR$) measures the spread of the middle $50\%$ of your data:

$$IQR = Q_3 - Q_1$$

In a box plot, John Tukey defined any point that strays too far from the middle as an outlier using these fences:

• Lower Fence: $Q_1 - 1.5 \times IQR$
• Upper Fence: $Q_3 + 1.5 \times IQR$

If a return falls outside these fences, it’s flagged as an extreme event.

---

4. Dispersion: Quantifying Total Uncertainty

If central tendency tells you where the ball is likely to land, dispersion tells you how wildly it might bounce around. In portfolio theory, dispersion is our baseline definition of total risk.

4.1 Mean Absolute Deviation (MAD)

You could simply take the average distance of each point from the mean:

$$MAD = \frac{\sum_{i=1}^n |X_i - \bar{X}|}{n}$$

While intuitive, absolute value functions have sharp corners that make them a nightmare for calculus and portfolio optimization algorithms. That’s why Wall Street almost exclusively uses squared deviations.

4.2 Sample Variance & Standard Deviation ($s$)

By squaring deviations, we eliminate negative signs and heavily penalize large misses:

$$s^2 = \frac{\sum_{i=1}^n (X_i - \bar{X})^2}{n - 1}$$ $$s = \sqrt{s^2} = \sqrt{\frac{\sum_{i=1}^n (X_i - \bar{X})^2}{n - 1}}$$

⚠️ The #1 Exam Trap ($n - 1$ vs $n$): Because we are estimating volatility from a sample rather than the entire history of the universe, we lose one degree of freedom estimating the mean $\bar{X}$. We MUST divide by $n - 1$ (Bessel’s correction). When reading your BA II Plus display, look for $S_x$, never $\sigma_x$!

Calculate Sample SD in BA II PlusResult: Sx = 2.53%, Mean = 3.00%

Sample: [4.5%, 6.0%, 1.5%, -2.0%, 0.0%, 4.5%, 3.5%, 2.5%, 5.5%, 4.0%]

Solve in Calc

---

5. Target Downside Deviation: Why Symmetrical Risk is a Lie

Let me share a pet peeve of mine with traditional finance textbooks: standard deviation treats upside wins as “risk.”

Think about that for a second. If your equity manager has a phenomenal quarter and gains $+30\%$, standard deviation shoots up, and standard risk models flag the fund as riskier!

No investor ever called their wealth advisor to complain that their portfolio made too much money. Investors only care about falling short of their target hurdle rate ($B$)—whether that’s $0\%$, the risk-free rate, or a $6\%$ liabilities target.

This is where Target Downside Deviation (or Target Semideviation) comes in:

$$s_{\text{target}} = \sqrt{\frac{\sum_{X_i \le B} (X_i - B)^2}{n - 1}}$$

Asymmetric Risk Analyzer

Target Downside Deviation vs. Standard Deviation

Standard deviation penalizes upside spikes as "risk." Target Downside Deviation only penalizes shortfall below your hurdle return (B).

Hurdle Target (B):2.0%

Portfolio 10-Period Return ObservationsTarget Hurdle Line B = 2.0%

B = 2.0%

4.5%

Period 1: 4.5%

Above Hurdle (Ignored)

6%

Period 2: 6%

Above Hurdle (Ignored)

1.5%

Period 3: 1.5%

Shortfall: 0.50% (Sq: 0.25)

-2%

Period 4: -2%

Shortfall: 4.00% (Sq: 16.00)

0%

Period 5: 0%

Shortfall: 2.00% (Sq: 4.00)

4.5%

Period 6: 4.5%

Above Hurdle (Ignored)

3.5%

Period 7: 3.5%

Above Hurdle (Ignored)

2.5%

Period 8: 2.5%

Above Hurdle (Ignored)

5.5%

Period 9: 5.5%

Above Hurdle (Ignored)

4%

Period 10: 4%

Above Hurdle (Ignored)

Period 13 of 10 periods fell below targetPeriod 10

Standard Deviation (Total Volatility)

2.53%

Symmetric: Treats surprise +6% returns as equally "risky" as -2% losses.

Target Downside Deviation (B = 2%)

1.50%

Asymmetric: Only squares shortfall from hurdle B, capturing true downside pain.

⚠️ CFA Critical Exam Trap:

When computing Target Downside Deviation, you square deviations ONLY for observations below target $B$, but you STILL DIVIDE BY TOTAL SAMPLE SIZE $(n - 1)$ (here 9), NOT by the number of bad periods (3)! Dividing by 3 is the #1 most common calculation error.

🚨 The Deadliest CFA Trap on Semideviation: Notice that while we only sum squared deviations for the bad months where $X_i \le B$, the denominator is STILL the full sample size $n - 1$ (not the number of bad months)! Dividing by the count of bad months is the single most common mistake students make.

Compute Target Downside Deviation in CalcResult: s_target = 1.63%

Target B = 2.0% on 10-period Portfolio Sample

Solve in Calc

---

6. Coefficient of Variation ($CV$): Measuring Risk Efficiency

Suppose someone asks you: “Which is riskier: Asset A with an $8\%$ standard deviation, or Asset B with a $4\%$ standard deviation?”

You can’t answer that without knowing their average returns!

• If Asset A generates a $+16\%$ average return, you’re taking $0.50$ units of volatility per unit of return.
• If Asset B only generates a $+2\%$ average return, you’re taking $2.00$ units of volatility per unit of return!

The Coefficient of Variation ($CV$) normalizes risk across different assets, currencies, or time periods:

$$CV = \frac{s}{\bar{X}}$$

Quick Comparison:

• Asset A: $\bar{X} = 2.10\%$, $s = 1.23\% \implies CV = \frac{1.23}{2.10} = \mathbf{0.586}$
• Asset B: $\bar{X} = 1.25\%$, $s = 1.35\% \implies CV = \frac{1.35}{1.25} = \mathbf{1.080}$

Asset A is far more risk-efficient. A lower $CV$ is almost always preferred.

⚠️ The Negative Mean Rule: If an asset’s average return is negative ($\bar{X} \le 0$), the Coefficient of Variation is mathematically invalid. A larger loss with higher volatility would produce an inverted negative ratio, completely destroying its meaning.

Compare CV in BA II PlusResult: CV_A = 0.586

Asset A (CV = 0.586) vs. Asset B (CV = 1.080)

Solve in Calc

---

7. Distribution Shape: Skewness, Kurtosis & Fat Tails

Modern Portfolio Theory loves to pretend that returns follow a clean, symmetrical Gaussian bell curve. But in real life, financial markets are full of skewed payoffs and black swans.

7.1 The Normal Distribution Benchmark

In a perfect Gaussian world:

• The bell curve is completely symmetric ($\text{Skewness} = 0$).
• Central tendencies align perfectly: $\text{Mean} = \text{Median} = \text{Mode}$.
• Kurtosis is exactly $3.0$ ($\text{Excess Kurtosis} = 0.0$).

7.2 Skewness: Which Way is the Tail Stretched?

Skewness measures asymmetry by cubing deviations (which preserves negative signs):

$$\text{Skewness} \approx \frac{1}{n} \frac{\sum_{i=1}^n (X_i - \bar{X})^3}{s^3}$$

Here is how you can instantly visualize both cases:

NEGATIVE SKEW (Left Tail)          POSITIVE SKEW (Right Tail)
Frequent small gains, rare crashes.  Frequent small losses, rare huge home runs.
Examples: S&P 500, Credit Spread Arb. Examples: Venture Capital, Buying Call Options.

         ▲                                    ▲
        / \                                  / \
       /   \                                /   \
  ____/     \                          ____/     \____
  Left Tail   Peak                         Peak    Right Tail

 Mean < Median < Mode                 Mode < Median < Mean

🎯 My Mental Anchor for Skewness: The Mean is a magnet for the tail! Wherever the long tail goes, the arithmetic mean gets dragged along with it. In a positively skewed distribution, the mean is pulled highest ($\text{Mode} < \text{Median} < \text{Mean}$).

Play with the skewness and kurtosis sliders below to see the mean, median, and mode separate in real time:

Interactive Distribution Sandbox

Distribution Shape Morpher: Skewness & Kurtosis

Drag the sliders below to see how asymmetry (skew) and fat-tails (kurtosis) shift the Mean, Median, and Mode.

Standard Normal (S=0, K=3)Equities / Crisis (Neg Skew, Fat Tails)Venture / Tech (Pos Skew, Fat Tails)Platykurtic (Thin Tails)

Arithmetic Mean Median (50th) Mode (Peak)

Symmetric: Mean = Median = Mode

Skewness (0.00):Symmetric (0.0)

-1.5 (Crash Risk)0.0 (Symmetric)+1.5 (Lottery/VC)

Excess Kurtosis (0.00):Mesokurtic (Normal = 0)

-1.5 (Thin Tails)0.0 (Mesokurtic / Normal)+3.5 (Black Swans)

💡 CFA & Quant Intuition: Under a pure Gaussian normal distribution, Kurtosis = 3, Excess Kurtosis = 0, Skewness = 0, and Mean = Median = Mode.

7.3 Kurtosis & Black Swan Tails

Kurtosis measures how much probability mass lives in the extreme tails versus the center:

$$\text{Kurtosis} \approx \frac{1}{n} \frac{\sum_{i=1}^n (X_i - \bar{X})^4}{s^4}$$ $$\text{Excess Kurtosis} = \text{Kurtosis} - 3.0$$

• Leptokurtic ($\text{Excess Kurtosis} > 0$): Fat tails and a tall, pointy peak. In markets, this means $4\sigma$ and $5\sigma$ crashes (like Black Monday 1987 or March 2020) happen far more frequently than a standard normal curve predicts.
• Platykurtic ($\text{Excess Kurtosis} < 0$): Thin tails and a flatter peak.
• Mesokurtic ($\text{Excess Kurtosis} = 0$): Exactly identical to a standard normal curve.

---

8. Bivariate Co-Movement: Covariance & Correlation

When constructing a multi-asset portfolio, the magic of diversification doesn’t come from picking winning stocks—it comes from picking assets that don’t crash at the exact same moment.

8.1 Sample Covariance ($s_{XY}$)

Covariance measures whether two assets deviate from their respective means in the same direction:

$$s_{XY} = \frac{\sum_{i=1}^n (X_i - \bar{X})(Y_i - \bar{Y})}{n - 1}$$

• If $s_{XY} > 0$: When Asset X is having an above-average month, Asset Y tends to be having an above-average month too.
• If $s_{XY} < 0$: When Asset X drops, Asset Y tends to rally (a natural hedge!).

8.2 Pearson Correlation Coefficient ($r_{XY}$)

The problem with covariance is that its units are weird squared percentages ($\%^2$). You can’t tell whether a covariance of $18.4$ is massive or tiny.

To fix this, we normalize covariance by dividing by the product of both standard deviations:

$$r_{XY} = \frac{s_{XY}}{s_X s_Y}$$

Correlation is unit-free and locked between $-1.0$ and $+1.0$:

$$-1.0 \le r_{XY} \le +1.0$$

Compute Stock-Bond Correlation in BA II PlusResult: r = -0.92, Sxy = -11.45

5-Period Paired Sample: Stocks vs. Treasuries

Solve in Calc

---

9. The Traps of Correlation: Don’t Get Blinded by a Single Number

If you take only one lesson from this section, let it be this: Always look at the scatter plot before trusting a correlation coefficient.

Bivariate Co-Movement

Correlation (r) vs. Scatter Plots & Non-Linear Traps

Never rely on correlation alone without plotting the scatter plot!

Positive Linear (r = +0.88)Negative Linear (r = -0.91)Non-Linear Trap (Y = X², r ≈ 0)Outlier Distortion

Strong Positive Linear RelationshipPearson r = +0.88

Asset X Return (%)Asset Y Return (%)

Key Lesson: Asset X (Large Cap) vs Asset Y (Small Cap). Points align tightly along a rising straight line.

Here are the four classic traps that catch quants and analysts:

1. The Non-Linear Trap ($Y = X^2$): Pearson $r$ only measures linear relationships. If two variables have a 100% deterministic parabolic relationship (like a delta-neutral options straddle), the positive and negative slopes cancel out, yielding $r \approx 0.00$! A correlation of zero does not mean “no relationship.”
2. Outlier Vulnerability: A single wild data point can transform a strong positive correlation into a negative one, or create the illusion of a trend where none exists.
3. Correlation Is Not Causation: Ice cream sales and shark attacks are positively correlated because both happen during hot summer months. Neither causes the other.
4. Spurious Trends: Two completely unrelated time series (like US tech stock valuations and per-capita cheese consumption) can show a correlation of $+0.95$ simply because both grew over the last twenty years.

---

10. Step-by-Step TI BA II Plus Keystroke Guide

You don’t need to compute sums of squares by hand. Let your TI BA II Plus do the heavy lifting in seconds:

┌────────────────────────────────────────────────────────────────────────┐
│               TI BA II PLUS 1-VAR STATISTICS WORKFLOW                  │
├────────────────────────────────────────────────────────────────────────┤
│  1. Open Data:       Press [2nd] [DATA]  (above the 7 key)             │
│  2. Clear Memory:    Press [2nd] [CLR WORK]                            │
│  3. Enter Values:    Enter X01, press [ENTER], press [↓] [↓] to X02... │
│  4. Open Stat:       Press [2nd] [STAT]  (above the 8 key)             │
│  5. Verify Mode:     Ensure display says "1-V" (press [2nd][SET] if 2V)│
│  6. Scroll Results:  Press [↓] to read:                                │
│                      • x̄  = Sample Arithmetic Mean                     │
│                      • Sx = Sample Standard Deviation (divided by n-1) │
│                      • σx = Population SD (divided by n)               │
│                      • n  = Total Number of Observations               │
└────────────────────────────────────────────────────────────────────────┘

---

11. 30-Second Statistical Decision Engine

Stuck on a tricky exam question or analyzing a messy portfolio dataset? Use this interactive diagnostic wizard to find the exact statistical tool you need:

Interactive Diagnostic Engine

30-Second Statistical Measure Decision Wizard

Answer a few quick questions to determine the optimal statistical measure for your asset return dataset.

Start Over

Step 1: Diagnostic Navigation

What core analytical question are you trying to answer about your asset returns?

Where is the return distribution centered?

Expected return, central benchmark, or typical performance level.

Select Option

How volatile or spread out are the returns?

Total uncertainty, range of outcomes, or risk per unit of return.

Select Option

How bad is the downside loss / tail risk?

Asymmetric downside shortfall, crash probability, or black swan fat tails.

Select Option

Do two assets move together or offer diversification?

Bivariate co-movement, hedging, and correlation linear linkage.

Select Option

---

12. CFA Exam Traps: The One-Minute Memory Sheet

Print this or keep it in your notes—these are the exact traps question writers love to set:

Scenario / Trigger	What you must think	The trap to avoid
Outliers present	Median or Trimmed/Winsorized Mean	Don’t use raw arithmetic mean.
Middle 50% spread	$IQR = Q_3 - Q_1$	Don’t add them ($Q_3 + Q_1$).
Sample Standard Deviation	Divide by $n - 1$ ($S_x$)	Never divide by $n$ ($\sigma_x$).
Target Downside Deviation	$\sqrt{\frac{\sum_{X \le B}(X - B)^2}{n - 1}}$	Divide by full $n - 1$, NOT count of bad months!
Risk per unit of return	$CV = \frac{s}{\bar{X}}$	If $\bar{X} \le 0$, $CV$ is invalid.
Positive Skew	Long right tail	$\text{Mode} < \text{Median} < \text{Mean}$
Negative Skew	Long left tail	$\text{Mean} < \text{Median} < \text{Mode}$
Fat Tails	Leptokurtic ($\text{Excess Kurtosis} > 0$)	Normal Kurtosis = $3$, Excess Kurtosis = $0$.
$r = 0.0$	No linear linkage	Does NOT mean no relationship (check non-linear!).

---

13. Active Recall Micro-Quizzes

Let’s see if you can spot the traps! Test your skills against these five high-frequency CFA exam questions:

Active Recall Micro-QuizzesCFA Module 3 Traps

Statistical Measures Exam & Quant Interview Mastery

Test your understanding against the 5 highest-frequency statistical traps.

Question 1: Central Tendency & Skewness

A hedge fund’s historical monthly return distribution exhibits significant positive skewness. Which inequality correctly describes the relationship between the central tendency measures?

AMean < Median < ModeBMode < Median < MeanCMedian < Mode < Mean

Question 2: Target Downside Deviation Trap

An analyst computes the Target Downside Deviation with a benchmark threshold B = 3.0% over a sample of 20 monthly returns. Exactly 4 months had returns below 3.0%. What denominator should the analyst use in the formula?

A4 (the number of shortfall observations)B19 (the total sample size n - 1 = 20 - 1)C3 (bad observations minus one = 4 - 1)

Question 3: Normal Distribution & Kurtosis

A quantitative analyst reports that an equity portfolio has an excess kurtosis of 2.4. What is the absolute kurtosis of this distribution, and what does it imply?

AKurtosis = 2.4, implying thinner tails than normal (platykurtic)BKurtosis = 5.4, implying fatter tails and higher black-swan risk than normal (leptokurtic)CKurtosis = -0.6, implying negative asymmetry

Question 4: Correlation & Scatter Plots

Suppose two financial variables X and Y are related by the exact formula Y = X² over the range [-5, +5]. What will the sample Pearson correlation coefficient (r) between X and Y be?

Ar = +1.00 because the relationship is 100% deterministicBr ≈ 0.00 because the symmetrical parabolic curve cancels out positive and negative linear slopesCr cannot be calculated for non-linear variables

Question 5: Coefficient of Variation (CV)

When is the Coefficient of Variation (CV = s / x̄) NOT meaningful for comparing investment opportunities?

AWhen the two assets have different currencies or measurement scalesBWhen the expected or sample mean return (x̄) is negative or zeroCWhen the standard deviation exceeds 10%

---

Final Thoughts & Next Steps

When I first started studying quantitative finance, statistics felt like a dry collection of Greek letters and mechanical formulas.

But once you realize that every formula is simply a lens designed to protect your capital—isolating downside risk, detecting black swan vulnerability, and stripping out illusionary correlations—the math becomes intuitive.

Here are the three golden rules to take with you:

1. Never judge an asset by its mean return alone. Always demand to see its dispersion, skewness, and downside semideviation.
2. Standard deviation is not the only definition of risk. When evaluating strategies with asymmetric payoffs, prefer Target Downside Deviation and the Sortino Ratio.
3. Always plot the data. A single scatter plot will save you from non-linear correlation traps and outlier distortions that a simple spreadsheet formula will miss every time.

Try loading the sample problems into the BA II Plus calculator emulator above, explore the sliders, and happy investing!