Welcome to Sample Moments!

In your journey through FRM Part I, you will often hear about "the population" (the true underlying data-generating process). In the real world, risk managers rarely observe the entire population. Instead, we observe finite samples of data and compute sample moments to infer the properties of the population distribution.

1. Core Concepts: Population vs. Sample & Estimator vs. Estimate

Before computing formulas, it is essential to distinguish between theoretical constructs and empirical calculations:

  • Population Moment vs. Sample Moment: A population moment (such as population mean \(\mu\) or variance \(\sigma^2\)) is a fixed, non-random parameter describing the entire population. A sample moment (such as sample mean \(\bar{X}\) or sample variance \(s^2\)) is calculated from observed sample data and is a random variable subject to sampling variation.
  • Estimator vs. Estimate: An estimator is a rule or mathematical function applied to a random sample (e.g., \(\hat{\theta} = \frac{1}{n}\sum X_i\)). An estimate is the concrete numerical value produced when the estimator is evaluated on a specific observed sample dataset (e.g., \(4.2\%\)).

2. Key Desirable Properties of Estimators

Risk analysts evaluate estimators based on three primary criteria:

  • Unbiasedness: An estimator \(\hat{\theta}\) is unbiased if its expected value equals the true population parameter: \(E[\hat{\theta}] = \theta\). Bias is defined as \(\text{Bias}(\hat{\theta}) = E[\hat{\theta}] - \theta\).
  • BLUE (Best Linear Unbiased Estimator): An estimator is BLUE if it is linear in the observations, is unbiased, and has the minimum variance among all possible linear unbiased estimators. Minimum variance means the estimates are more tightly concentrated around the true parameter.
  • Consistency: An estimator is consistent if it converges in probability to the true parameter as the sample size \(n\) approaches infinity: \(\lim_{n \to \infty} P(|\hat{\theta}_n - \theta| > \epsilon) = 0\).

3. The First Moment: Sample Mean \(\bar{X}\)

The Sample Mean measures central tendency and estimates the population mean \(\mu\):

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

Where \(n\) is the number of observations and \(X_i\) represents each sample value. The sample mean is an unbiased estimator of \(\mu\) (since \(E[\bar{X}] = \mu\)) and, under classical assumptions, is the BLUE estimator of \(\mu\).

Sampling Distribution, LLN, and CLT

Because \(\bar{X}\) is calculated from random data, it has its own probability distribution with Standard Error (SE):

\( SE_{\bar{X}} = \frac{\sigma}{\sqrt{n}} \)

  • Law of Large Numbers (LLN): As sample size \(n\) grows large, the sample mean \(\bar{X}\) converges in probability to the population mean \(\mu\). This proves that \(\bar{X}\) is a consistent estimator of \(\mu\).
  • Central Limit Theorem (CLT): For an independent and identically distributed (i.i.d.) sample with finite variance \(\sigma^2\), as \(n \to \infty\), the distribution of \(\sqrt{n}(\bar{X} - \mu)\) converges in distribution to a normal distribution: \(\bar{X} \sim N\left(\mu, \frac{\sigma^2}{n}\right)\), regardless of the underlying population's shape.

4. The Second Moment: Sample Variance and Standard Deviation

Sample Variance measures the dispersion of the data around the sample mean.

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

Bessel's Correction (\(n-1\)): Dividing by \(n-1\) rather than \(n\) corrects for degrees of freedom lost by using \(\bar{X}\) instead of the unknown true population mean \(\mu\). This makes \(s^2\) an unbiased estimator of the population variance \(\sigma^2\).

Sample Standard Deviation \(s\): \( s = \sqrt{s^2} \). While \(s^2\) is an unbiased estimator of \(\sigma^2\), \(s\) is technically a slightly biased estimator of \(\sigma\) due to Jensen's inequality, but it is consistent.

5. The Third Moment: Sample Skewness

Skewness measures the degree of asymmetry of the probability distribution around its mean.

The sample skewness estimator is calculated as:

\( \hat{S} = \frac{\frac{1}{n} \sum_{i=1}^{n} (X_i - \bar{X})^3}{s^3} \)

  • Positive Skew (Right-skewed, \(\hat{S} > 0\)): Long tail to the right; mean typically exceeds median.
  • Negative Skew (Left-skewed, \(\hat{S} < 0\)): Long tail to the left; frequent small gains offset by rare, severe downside losses. Risk managers are especially concerned with negative return skewness.
  • Zero Skew (\(\hat{S} = 0\)): Symmetric distribution (e.g., standard normal).

6. The Fourth Moment: Sample Kurtosis

Kurtosis measures tail thickness and the propensity to generate extreme outliers relative to a normal distribution.

The sample kurtosis estimator is calculated as:

\( \hat{K} = \frac{\frac{1}{n} \sum_{i=1}^{n} (X_i - \bar{X})^4}{s^4} \)

Excess Kurtosis centers the measure at zero relative to the normal distribution (which has a kurtosis of 3):

\( \text{Excess Kurtosis} = \hat{K} - 3 \)

  • Leptokurtic (\(\hat{K} > 3\), Excess > 0): Fat tails and high peak. Financial asset returns typically exhibit leptokurtosis, meaning extreme market drawdowns occur far more frequently than predicted by a normal distribution.
  • Mesokurtic (\(\hat{K} = 3\), Excess = 0): Normal tail behavior.
  • Platykurtic (\(\hat{K} < 3\), Excess < 0): Thin tails and lower probability of extreme outliers.

7. Sample Quantiles

A sample quantile divides sorted sample observations into specific proportions. If sample data of size \(n\) is ordered as \(X_{(1)} \le X_{(2)} \le \dots \le X_{(n)}\), the \(p\)-th quantile \(q_p\) (where \(0 < p < 1\)) is the value below which a proportion \(p\) of the observations fall.

  • Median (0.50 quantile): Divides the distribution into two equal halves; robust to extreme outliers compared to the sample mean.
  • Quartiles: 25th percentile (\(Q_1\)), 50th percentile (\(Q_2\)), and 75th percentile (\(Q_3\)). The Interquartile Range (IQR) is \(IQR = Q_3 - Q_1\).
  • Value at Risk (VaR): In financial risk management, Value at Risk (VaR) at confidence level \(1 - \alpha\) is directly estimated using the lower \(\alpha\)-quantile of portfolio returns.

8. Cross-Moments: Coskewness and Cokurtosis

When analyzing multiple risk factors or assets simultaneously, cross-moments describe non-linear joint tail dependencies beyond linear covariance:

  • Coskewness: Measures how the variance of one asset relates to the return of another asset, or how joint extremes tilt. Specifically, \(S(X, X, Y) = \frac{E[(X - \mu_X)^2(Y - \mu_Y)]}{\sigma_X^2 \sigma_Y}\). Negative coskewness with market returns implies that an asset tends to experience high volatility when the broader market suffers large drops.
  • Cokurtosis: Measures joint tail thickness and the likelihood of simultaneous extreme events across assets: \(K(X, X, Y, Y) = \frac{E[(X - \mu_X)^2(Y - \mu_Y)^2]}{\sigma_X^2 \sigma_Y^2}\). High cokurtosis indicates that extreme volatility in one asset coincides with extreme volatility in another, undermining diversification during financial crises.

Summary Table

1st Moment: Mean \(\to\) Central Location; unbiased, consistent, BLUE.
2nd Moment: Variance & SD \(\to\) Dispersion; uses \(n-1\) degrees of freedom.
3rd Moment: Skewness \(\to\) Asymmetry; negative skew indicates downside crash risk.
4th Moment: Kurtosis \(\to\) Tail thickness; excess kurtosis > 0 indicates fat tails.
Quantiles \(\to\) Cutoff values for order statistics and Value at Risk.
Coskewness & Cokurtosis \(\to\) Joint asymmetry and simultaneous extreme tail risk across assets.