Study guide
6.4 Standard Deviation, Variance, and Data Dispersion
GMAT Exam Focus Edition
Track your progress as you read
Finished sections get a check mark.
1 Chapter 1: GMAT Exam Architecture & Strategic Foundations
1.1 GMAT Structure, Section Order, and the 205–805 Score Scale 1.2 Question Bookmarking and the Review & Edit Feature Mechanics 1.3 Test-Day Logistics, Pacing Architecture, and Score Reporting Policies
2 Chapter 2: Quantitative Reasoning — Arithmetic & Number Properties
2.1 Integers, Divisibility Rules, Prime Factorization, and LCM/GCD 2.2 Remainders, Even/Odd Parity, and Positive/Negative Sign Analysis 2.3 Exponents, Radicals, Roots, and Scientific Notation 2.4 Absolute Value, Number Line Properties, and Estimation Techniques
3 Chapter 3: Quantitative Reasoning — Fractions, Decimals, Ratios & Percentages
3.1 Fractions, Decimals, and Real-Number Operations 3.2 Ratios, Proportions, and Scaling Relationships 3.3 Percent Change, Successive Percent Changes, and Multipliers 3.4 Commercial Math: Profit/Loss, Markups, Discounts, and Simple/Compound Interest
4 Chapter 4: Quantitative Reasoning — Algebra, Equations & Inequalities
4.1 Linear Equations and Systems of Simultaneous Equations 4.2 Quadratic Equations, Factoring Patterns, and Roots 4.3 Linear Inequalities, Compound Inequalities, and Absolute Value Inequalities 4.4 Functions, Defined Operational Symbols, and Sequences
5 Chapter 5: Quantitative Reasoning — Word Problems & Modeling
5.1 Rate, Time, and Distance: Uniform Motion, Relative Speed, and Average Speed 5.2 Work Rates: Combined Rates, Sequential Tasks, and Opposing Rates 5.3 Mixture Problems and Weighted Averages 5.4 Overlapping Sets, Venn Diagrams, and the Double-Set Matrix Method
6 Chapter 6: Quantitative Reasoning — Counting, Probability & Statistics
6.1 Fundamental Counting Principle, Permutations, and Combinations 6.2 Probability Rules: Independent, Dependent, Mutually Exclusive, and Complementary Events 6.3 Descriptive Statistics: Mean, Median, Mode, Range, and Weighted Mean 6.4 Standard Deviation, Variance, and Data Dispersion
7 Chapter 7: Verbal Reasoning — Critical Reasoning: Argument Structure & Assumptions
7.1 Anatomy of an Argument: Conclusions, Premises, Assumptions, and Background Context 7.2 Assumption Questions: Necessary vs. Sufficient Assumptions and the Negation Test 7.3 Strengthen Questions: Supporting Premises, Causality, and New Evidence 7.4 Weaken Questions: Identifying Flaws, Alternate Causes, and Counter-Evidence
8 Chapter 8: Verbal Reasoning — Critical Reasoning: Advanced Arguments
8.1 Evaluate the Argument: The Variance Test and Decisive Information 8.2 Resolve the Paradox: Reconciling Conflicting Facts and Surprising Outcomes 8.3 Boldface and Method of Reasoning: Identifying the Role of Statements 8.4 Flaw in the Reasoning: Spotting Formal and Informal Logical Fallacies
9 Chapter 9: Verbal Reasoning — Reading Comprehension: Core Analysis
9.1 Passage Structures, Main Idea, and Author's Primary Purpose 9.2 Supporting Ideas, Explicit Details, and Targeted Information Retrieval 9.3 Inference Questions: Deriving Unstated Claims and Valid Deductions
10 Chapter 10: Verbal Reasoning — Reading Comprehension: Advanced Strategies
10.1 Author's Tone, Attitude, and Epistemic Stance 10.2 Logical Structure, Paragraph Function, and Rhetorical Organization 10.3 Application, Contextual Analogy, and Cross-Disciplinary Synthesis
11 Chapter 11: Data Insights — Data Sufficiency Foundations & Algebra
11.1 Data Sufficiency Architecture: The 5 Answer Choices (AD/BCE) and Yes/No vs. Value Logic 11.2 Golden Rules of DS: Preventing Statement 2 Contamination and Sufficiency Verification 11.3 DS Applied to Arithmetic, Number Properties, and Signs 11.4 DS Applied to Algebraic Equations, Systems, and Inequalities
12 Chapter 12: Data Insights — Data Sufficiency Applied Modeling
12.1 DS Applied to Rates, Ratios, and Work Problems 12.2 DS Applied to Statistics, Averages, and Data Sets 12.3 DS Applied to Overlapping Sets, Venn Diagrams, and Probability
13 Chapter 13: Data Insights — Multi-Source Reasoning & Table Analysis
13.1 Multi-Source Reasoning: Tab Navigation, Synthesizing Text, Data, and Emails 13.2 MSR Question Formats: Multiple Choice and Three-Part Dichotomous Questions 13.3 Table Analysis: Interactive Sorting, Filtering, and Trend Identification 13.4 Table Analysis: Evaluating Binary Statements (True/False, Yes/No)
14 Chapter 14: Data Insights — Graphics Interpretation & Two-Part Analysis
14.1 Graphics Interpretation: Bar Charts, Line Graphs, and Scatterplots 14.2 Graphics Interpretation: Bubble Charts, Dual-Axis Graphs, and Pie Charts 14.3 Two-Part Analysis: Quantitative and Mathematical Word Problems 14.4 Two-Part Analysis: Verbal, Logic-Based, and Strategic Decision Problems
6.4 Standard Deviation, Variance, and Data Dispersion
Key Takeaways Explain This Topic AI
Standard deviation quantifies the dispersion of data points around their arithmetic mean; it is always non-negative (σ ≥ 0) and equals zero if and only if all values in the data set are identical.
The GMAT tests standard deviation conceptually: data sets with points clustered tightly around the mean have smaller standard deviations, while sets with points polarized toward the extremes have larger standard deviations.
Adding or subtracting a constant c to every term in a data set shifts the distribution uniformly along the number line, leaving both standard deviation and variance completely unchanged.
Multiplying or dividing every term by a constant k scales the standard deviation by |k| and the variance by k².
Inserting an additional observation exactly equal to the existing mean leaves the mean unchanged while strictly decreasing the standard deviation (for any non-zero initial variance).
Last updated: September 2026
6.4 Standard Deviation, Variance, and Data Dispersion
Quick Summary: On the GMAT Focus Quantitative section, you will virtually never calculate standard deviation using pencil-and-paper formulas. Instead, GMAC tests your conceptual mastery of data spread: comparing how tightly points cluster around the mean, knowing that adding a constant does not change standard deviation, and recognizing that multiplying by k k k scales standard deviation by ∣ k ∣ |k| ∣ k ∣ .
While measures of central tendency (mean, median, mode) pinpoint the center of a data distribution, measures of dispersion quantify how tightly or loosely the data points are clustered around that center. On the calculator-free GMAT, standard deviation questions are designed to test your geometric and algebraic intuition about spread rather than your stamina for manual computation.
Conceptual Definition of Dispersion and Variance
Consider two investment funds, both yielding an average annual return of 8% over five years:
Fund Alpha: { 7 % , 8 % , 8 % , 8 % , 9 % } \{7\%, 8\%, 8\%, 8\%, 9\%\} { 7% , 8% , 8% , 8% , 9% }
Fund Beta: { − 10 % , 0 % , 8 % , 16 % , 26 % } \{-10\%, 0\%, 8\%, 16\%, 26\%\} { − 10% , 0% , 8% , 16% , 26% }
Although both funds share the identical mean return of 8%, Fund Alpha is exceptionally stable, while Fund Beta exhibits extreme volatility. This volatility is precisely what standard deviation measures.
The Mathematical Architecture of Variance and Standard Deviation
To understand how standard deviation behaves, examine its constituent steps:
Calculate the arithmetic mean ( μ \mu μ ): Find the balance point of the n n n observations.
Compute individual deviations from the mean: ( x i − μ ) (x_i - \mu) ( x i − μ ) . Points below the mean have negative deviations; points above have positive deviations.
Square each deviation: ( x i − μ ) 2 (x_i - \mu)^2 ( x i − μ ) 2 . Squaring ensures that all deviations become positive and penalizes points that lie far from the center disproportionately.
Calculate Variance ( σ 2 \sigma^2 σ 2 ): The arithmetic average of the squared deviations: σ 2 = ∑ i = 1 n ( x i − μ ) 2 n \sigma^2 = \frac{\sum_{i=1}^n (x_i - \mu)^2}{n} σ 2 = n ∑ i = 1 n ( x i − μ ) 2
Calculate Standard Deviation ( σ \sigma σ ): Take the positive square root of variance to return to the original units of measurement: σ = ∑ i = 1 n ( x i − μ ) 2 n \sigma = \sqrt{\frac{\sum_{i=1}^n (x_i - \mu)^2}{n}} σ = n ∑ i = 1 n ( x i − μ ) 2
Test-Day Reality: The GMAT will never require you to execute all five steps on a large data set. If you find yourself doing extensive long division and square root calculations on scratch paper, you have missed the conceptual shortcut!
Fundamental Axiomatic Properties of Standard Deviation
Every test-taker must memorize these non-negotiable mathematical properties:
Non-Negativity: Standard deviation is always greater than or equal to zero ( σ ≥ 0 \sigma \ge 0 σ ≥ 0 ). It can never be negative because it is defined as the principal square root of non-negative squared values.
The Zero-Variance Condition: The standard deviation of a set is zero if and only if all numbers in the set are identical : σ = 0 ⟺ x 1 = x 2 = x 3 = ⋯ = x n \sigma = 0 \iff x_1 = x_2 = x_3 = \dots = x_n σ = 0 ⟺ x 1 = x 2 = x 3 = ⋯ = x n Example: For { 5 , 5 , 5 , 5 } \{5, 5, 5, 5\} { 5 , 5 , 5 , 5 } , the mean is 5, every deviation is 0, and σ = 0 \sigma = 0 σ = 0 .
Clustering vs. Dispersion:
The closer data points lie to the mean, the smaller the standard deviation.
The further data points lie from the mean, the larger the standard deviation.
Qualitative Comparison of Data Sets (Mental Spread Inspection)
A common GMAT Problem Solving item presents three or four data sets and asks which has the greatest or least standard deviation.
The Comparison Framework
When comparing sets of equal size:
Check the mean of each set.
Inspect how the data points are distributed relative to that mean.
Remember that because deviations are squared, extreme values dominate standard deviation .
Detailed Comparison Walkthrough
Examine the following three sets, each containing 5 elements with a mean of 30:
- Set A: { 10 , 20 , 30 , 40 , 50 } \{10, 20, 30, 40, 50\} { 10 , 20 , 30 , 40 , 50 }
- Set B: { 28 , 29 , 30 , 31 , 32 } \{28, 29, 30, 31, 32\} { 28 , 29 , 30 , 31 , 32 }
- Set C: { 10 , 10 , 30 , 50 , 50 } \{10, 10, 30, 50, 50\} { 10 , 10 , 30 , 50 , 50 }
Let's evaluate their dispersion around the mean μ = 30 \mu = 30 μ = 30 :
In Set B , all observations are packed tightly within 2 units of the mean. Deviations are { − 2 , − 1 , 0 , 1 , 2 } \{-2, -1, 0, 1, 2\} { − 2 , − 1 , 0 , 1 , 2 } . Sum of squared deviations is 4 + 1 + 0 + 1 + 4 = 10 4 + 1 + 0 + 1 + 4 = 10 4 + 1 + 0 + 1 + 4 = 10 . Set B has the smallest standard deviation.
In Set A , numbers are uniformly spaced with deviations { − 20 , − 10 , 0 , 10 , 20 } \{-20, -10, 0, 10, 20\} { − 20 , − 10 , 0 , 10 , 20 } . Sum of squared deviations is 400 + 100 + 0 + 100 + 400 = 1 , 000 400 + 100 + 0 + 100 + 400 = 1,000 400 + 100 + 0 + 100 + 400 = 1 , 000 .
In Set C , four of the five numbers are pushed all the way to the absolute extremes ( 10 10 10 and 50 50 50 ). Deviations are { − 20 , − 20 , 0 , 20 , 20 } \{-20, -20, 0, 20, 20\} { − 20 , − 20 , 0 , 20 , 20 } . Sum of squared deviations is 400 + 400 + 0 + 400 + 400 = 1 , 600 400 + 400 + 0 + 400 + 400 = 1,600 400 + 400 + 0 + 400 + 400 = 1 , 600 . Set C has the largest standard deviation. σ B < σ A < σ C \sigma_B < \sigma_A < \sigma_C σ B < σ A < σ C
Key Takeaway: Even though Set A and Set C have the exact same Range ( 50 − 10 = 40 50 - 10 = 40 50 − 10 = 40 ) and the exact same Mean (30), Set C has a significantly higher standard deviation because its values are polarized toward the extremes rather than distributed across the interior.
The Mathematical Effects of Linear Transformations
Understanding how standard deviation reacts to arithmetic operations is one of the highest-yield topics on the quantitative section.
1. Adding or Subtracting a Constant c c c ( x i → x i + c x_i \to x_i + c x i → x i + c )
Imagine plotting a data set as dots on a number line. Adding c c c shifts every dot by c c c units in the same direction. The entire distribution slides rigidly along the axis:
The mean shifts: μ new = μ old + c \mu_{\text{new}} = \mu_{\text{old}} + c μ new = μ old + c .
The deviations from the mean: ( x i + c ) − ( μ + c ) = x i − μ (x_i + c) - (\mu + c) = x_i - \mu ( x i + c ) − ( μ + c ) = x i − μ .
Because the distances between each point and the mean are identical, the squared deviations are unchanged! σ new = σ old and σ new 2 = σ old 2 \mathbf{\sigma_{\text{new}} = \sigma_{\text{old}}} \quad \text{and} \quad \mathbf{\sigma^2_{\text{new}} = \sigma^2_{\text{old}}} σ new = σ old and σ new 2 = σ old 2
Critical Rule: Adding or subtracting any constant (positive or negative) to every element in a data set has ZERO effect on its standard deviation and variance.
2. Multiplying or Dividing by a Constant k k k ( x i → k x i x_i \to k x_i x i → k x i )
Multiplying every data point by k k k stretches (if ∣ k ∣ > 1 |k| > 1 ∣ k ∣ > 1 ) or compresses (if ∣ k ∣ < 1 |k| < 1 ∣ k ∣ < 1 ) the distribution relative to the origin:
The mean scales: μ new = k μ old \mu_{\text{new}} = k \mu_{\text{old}} μ new = k μ old .
The individual deviations scale: ( k x i − k μ ) = k ( x i − μ ) (k x_i - k \mu) = k(x_i - \mu) ( k x i − k μ ) = k ( x i − μ ) .
The squared deviations scale by k 2 k^2 k 2 : σ new 2 = k 2 σ old 2 \sigma^2_{\text{new}} = k^2 \sigma^2_{\text{old}} σ new 2 = k 2 σ old 2
Taking the square root gives: σ new = ∣ k ∣ ⋅ σ old \mathbf{\sigma_{\text{new}} = |k| \cdot \sigma_{\text{old}}} σ new = ∣k∣ ⋅ σ old
Summary of Transformation Impacts
Transformation on Every Element Effect on Mean Effect on Standard Deviation ( σ \sigma σ ) Effect on Variance ( σ 2 \sigma^2 σ 2 )
- Add constant c c c μ + c \mu + c μ + c Unchanged ( σ \sigma σ ) Unchanged ( σ 2 \sigma^2 σ 2 )
- Subtract constant c c c μ − c \mu - c μ − c Unchanged ( σ \sigma σ ) Unchanged ( σ 2 \sigma^2 σ 2 )
Multiply by constant k k k k ⋅ μ k \cdot \mu k ⋅ μ ∣ k ∣ ⋅ σ \vert k\vert\cdot \sigma ∣ k ∣ ⋅ σ k 2 ⋅ σ 2 k^2 \cdot \sigma^2 k 2 ⋅ σ 2
Divide by constant k k k ( k ≠ 0 k \neq 0 k = 0 ) μ / k \mu / k μ / k σ / ∣ k ∣ \sigma / \vert k\vert σ /∣ k ∣ σ 2 / k 2 \sigma^2 / k^2 σ 2 / k 2
Linear equation: y = a x + b y = ax + b y = a x + b a μ + b a\mu + b a μ + b ∣ a ∣ ⋅ σ \vert a\vert\cdot \sigma ∣ a ∣ ⋅ σ a 2 ⋅ σ 2 a^2 \cdot \sigma^2 a 2 ⋅ σ 2
The Effect of Adding New Observations to a Data Set
Questions frequently ask how inserting an additional data point alters the existing standard deviation of a set.
1. Adding an Observation Exactly Equal to the Mean ( x new = μ x_{\text{new}} = \mu x new = μ )
The new observation contributes a deviation of zero: ( x new − μ ) 2 = 0 (x_{\text{new}} - \mu)^2 = 0 ( x new − μ ) 2 = 0 .
The numerator of variance ∑ ( x i − μ ) 2 \sum (x_i - \mu)^2 ∑ ( x i − μ ) 2 remains completely unchanged.
However, the sample size in the denominator increases from n n n to n + 1 n + 1 n + 1 .
Therefore, the variance and standard deviation strictly decrease : σ new 2 = ∑ ( x i − μ ) 2 n + 1 < ∑ ( x i − μ ) 2 n = σ old 2 \sigma_{\text{new}}^2 = \frac{\sum (x_i - \mu)^2}{n + 1} < \frac{\sum (x_i - \mu)^2}{n} = \sigma_{\text{old}}^2 σ new 2 = n + 1 ∑ ( x i − μ ) 2 < n ∑ ( x i − μ ) 2 = σ old 2
(Exception: If σ old = 0 \sigma_{\text{old}} = 0 σ old = 0 , all elements are already equal to the mean, so σ \sigma σ remains 0.)
2. Adding an Observation Far From the Mean (Outlier)
Adding an extreme observation adds a large squared deviation term to the numerator, pulling the standard deviation upward .
The Empirical Rule for Normal Distributions
While the GMAT rarely tests advanced continuous probability density functions, it does occasionally present scenarios referencing a normal distribution (bell curve) to test standard deviation intervals.
For any normally distributed variable with mean μ \mu μ and standard deviation σ \sigma σ :
Approximately 68% of all data points fall within 1 standard deviation of the mean: [ μ − σ , μ + σ ] [\mu - \sigma, \mu + \sigma] [ μ − σ , μ + σ ] .
Approximately 95% of all data points fall within 2 standard deviations of the mean: [ μ − 2 σ , μ + 2 σ ] [\mu - 2\sigma, \mu + 2\sigma] [ μ − 2 σ , μ + 2 σ ] .
Approximately 99.7% of all data points fall within 3 standard deviations of the mean: [ μ − 3 σ , μ + 3 σ ] [\mu - 3\sigma, \mu + 3\sigma] [ μ − 3 σ , μ + 3 σ ] .
Because the bell curve is perfectly symmetric, exactly half ( 50 % 50\% 50% ) of the distribution lies above the mean and half ( 50 % 50\% 50% ) lies below. Furthermore:
Approximately 34 % 34\% 34% lies between μ \mu μ and μ + 1 σ \mu + 1\sigma μ + 1 σ .
Approximately 13.5 % 13.5\% 13.5% lies between μ + 1 σ \mu + 1\sigma μ + 1 σ and μ + 2 σ \mu + 2\sigma μ + 2 σ .
Approximately 2.35 % 2.35\% 2.35% lies between μ + 2 σ \mu + 2\sigma μ + 2 σ and μ + 3 σ \mu + 3\sigma μ + 3 σ .
Approximately 0.15 % 0.15\% 0.15% lies above μ + 3 σ \mu + 3\sigma μ + 3 σ .
- Worked Problem Solving Examples
- Example 1: Standard Deviation Under Linear Transformation
Problem: A set of 20 test scores has a mean of 65 and a standard deviation of 12. If each score is increased by 10% and then has 5 points added to it, what will be the new mean and the new standard deviation of the test scores?
Solution:
Express the transformation algebraically: Increasing each score x i x_i x i by 10% multiplies it by 1.10. Adding 5 points gives: y i = 1.10 x i + 5 y_i = 1.10 x_i + 5 y i = 1.10 x i + 5
Calculate the new mean: μ new = 1.10 μ old + 5 = 1.10 ( 65 ) + 5 = 71.5 + 5 = 76.5 \mu_{\text{new}} = 1.10 \mu_{\text{old}} + 5 = 1.10(65) + 5 = 71.5 + 5 = 76.5 μ new = 1.10 μ old + 5 = 1.10 ( 65 ) + 5 = 71.5 + 5 = 76.5
Calculate the new standard deviation:
Recall that the additive constant ( + 5 ) (+5) ( + 5 ) has zero effect on standard deviation.
The multiplicative constant ( 1.10 ) (1.10) ( 1.10 ) scales the standard deviation directly: σ new = 1.10 × σ old = 1.10 × 12 = 13.2 \sigma_{\text{new}} = 1.10 \times \sigma_{\text{old}} = 1.10 \times 12 = 13.2 σ new = 1.10 × σ old = 1.10 × 12 = 13.2
The new mean is 76.5 and the new standard deviation is 13.2 .
Example 2: Normal Distribution Percentile Range
Problem: The scores on a standardized certification examination are normally distributed with a mean of 500 and a standard deviation of 80. Approximately what percentage of test-takers scored between 420 and 660?
Solution:
Express the score boundaries in terms of standard deviations from the mean μ = 500 \mu = 500 μ = 500 , σ = 80 \sigma = 80 σ = 80 :
Lower boundary: 420 = 500 − 80 = μ − 1 σ 420 = 500 - 80 = \mu - 1\sigma 420 = 500 − 80 = μ − 1 σ
Upper boundary: 660 = 500 + 2 ( 80 ) = μ + 2 σ 660 = 500 + 2(80) = \mu + 2\sigma 660 = 500 + 2 ( 80 ) = μ + 2 σ
Sum the percentage segments defined by the empirical rule:
- Segment from μ − 1 σ \mu - 1\sigma μ − 1 σ to μ \mu μ : ≈ 34 % \approx 34\% ≈ 34%
- Segment from μ \mu μ to μ + 1 σ \mu + 1\sigma μ + 1 σ : ≈ 34 % \approx 34\% ≈ 34%
- Segment from μ + 1 σ \mu + 1\sigma μ + 1 σ to μ + 2 σ \mu + 2\sigma μ + 2 σ : ≈ 13.5 % \approx 13.5\% ≈ 13.5%
Total percentage between μ − 1 σ \mu - 1\sigma μ − 1 σ and μ + 2 σ \mu + 2\sigma μ + 2 σ : Total = 34 % + 34 % + 13.5 % = 81.5 % \text{Total} = 34\% + 34\% + 13.5\% = 81.5\% Total = 34% + 34% + 13.5% = 81.5%
Approximately 81.5% of test-takers scored between 420 and 660.
High-Frequency GMAT Traps
Trap 1: The Constant Addition Trap
Assuming that adding 10 to every number in a data set increases its standard deviation by 10. Shifting data uniformly does not change the spread relative to the mean; σ \sigma σ remains unchanged.
Trap 2: Conflating Range with Standard Deviation
Assuming that a data set with a larger range always has a larger standard deviation. A set with values { 0 , 50 , 50 , 50 , 100 } \{0, 50, 50, 50, 100\} { 0 , 50 , 50 , 50 , 100 } (range 100) has most of its values at the mean (50), while a set with { 10 , 10 , 80 , 80 } \{10, 10, 80, 80\} { 10 , 10 , 80 , 80 } (range 70) has all points far from the mean, easily producing a higher standard deviation despite a smaller range.
Trap 3: Wasting Time on Manual Calculations
Attempting to manually calculate variances on scratch paper. GMAT questions can virtually always be resolved by checking symmetry, extreme clustering, or transformation rules.