Chapter Introduction
Correlation studies the relationship between two variables — do they move together, in opposite directions, or independently? For example, does higher education lead to higher income? Do price increases reduce demand? This chapter introduces the concept of correlation, how to visualise it with scatter diagrams, and how to measure it quantitatively using Karl Pearson's coefficient and Spearman's rank correlation. Understanding correlation is fundamental to economics, where we constantly analyse how one economic variable affects another.
Correlation does not prove that one variable causes changes in another — it only shows they move together. This distinction between correlation and causation is one of the most important ideas in statistics.
Key Definitions
| Term | Definition |
|---|---|
| Correlation | A statistical tool that measures the degree and direction of the relationship between two variables |
| Positive correlation | Two variables move in the same direction — both increase or both decrease |
| Negative correlation | Two variables move in opposite directions — one increases as the other decreases |
| Zero correlation | No linear relationship between the two variables |
| Linear correlation | The relationship between variables can be approximated by a straight line |
| Non-linear correlation | The relationship follows a curve, not a straight line |
| Scatter diagram | A graph plotting paired values of two variables as points to visually assess correlation |
| Karl Pearson's coefficient (r) | A measure of linear correlation between two quantitative variables; ranges from -1 to +1 |
| Spearman's rank correlation (r_k) | A measure of correlation based on ranks of data rather than actual values |
| Causation | A cause-and-effect relationship where one variable directly causes changes in another |
| Spurious correlation | A correlation that exists between two variables due to a third common factor, not a direct relationship |
| Coefficient of determination (r squared) | The proportion of variation in one variable explained by the other; r squared x 100 = percentage |
Detailed Concepts
Types of Correlation
| Type | Description | Example |
|---|---|---|
| Positive | Both variables increase (or both decrease) together | Price and supply; income and expenditure; height and weight |
| Negative | One variable increases while the other decreases | Price and demand; temperature and sale of woollens; price and quantity demanded |
| Zero | No relationship between the variables | Shoe size and intelligence; height and political affiliation |
Degrees of Correlation
| Degree | Value of r | Interpretation |
|---|---|---|
| Perfect positive | +1 | Variables move together in exact proportion |
| High positive | +0.75 to +0.99 | Strong tendency to move together |
| Moderate positive | +0.25 to +0.74 | Moderate tendency to move together |
| Low positive | 0 to +0.25 | Weak tendency to move together |
| Zero | 0 | No linear relationship |
| Low negative | -0.25 to 0 | Weak tendency to move in opposite directions |
| Moderate negative | -0.74 to -0.25 | Moderate tendency to move oppositely |
| High negative | -0.99 to -0.75 | Strong tendency to move oppositely |
| Perfect negative | -1 | Variables move oppositely in exact proportion |
Scatter Diagram
A scatter diagram plots paired observations of two variables X and Y on a graph. Each point represents one observation.
| Pattern in scatter diagram | Type of correlation | Strength |
|---|---|---|
| Points cluster tightly along an upward line | Positive | High |
| Points loosely trend upward | Positive | Low to moderate |
| Points cluster tightly along a downward line | Negative | High |
| Points loosely trend downward | Negative | Low to moderate |
| Points scattered randomly, no pattern | Zero | None |
| Points follow a curved (non-linear) pattern | Non-linear correlation | Varies |
Advantages of scatter diagram:
1. Simple to draw and easy to understand 2. Gives a quick visual idea of the type and degree of correlation 3. Reveals non-linear relationships that Pearson's coefficient cannot detect 4. Helps identify outliers that may distort the correlation coefficient 5. No calculation required — purely visual
Limitations:
1. Does not give an exact numerical measure 2. Cannot determine the precise degree of correlation 3. Subjective interpretation may vary between observers 4. Not suitable for large datasets (too many points overlap)
Karl Pearson's Coefficient of Correlation
Karl Pearson's coefficient (r) is the most widely used measure of linear correlation between two quantitative variables.
Formula (deviation method):
r = Sum of xy / sqrt(Sum of x squared x Sum of y squared)
Where: x = X - X-bar (deviation of X from its mean), y = Y - Y-bar (deviation of Y from its mean)
Alternative formula (direct method):
r = [N x Sum of XY - (Sum of X)(Sum of Y)] / sqrt[(N x Sum of X squared - (Sum of X) squared) x (N x Sum of Y squared - (Sum of Y) squared)]
Step-deviation method (for grouped data):
When class intervals are large, use: dx = (X - Ax) / h, dy = (Y - Ay) / h
r = [N x Sum of dxdy - (Sum of dx)(Sum of dy)] / sqrt[(N x Sum of dx squared - (Sum of dx) squared) x (N x Sum of dy squared - (Sum of dy) squared)]
Properties of Karl Pearson's coefficient:
1. The value of r always lies between -1 and +1 (both inclusive) 2. r = +1 means perfect positive correlation; r = -1 means perfect negative correlation 3. r = 0 means zero linear correlation (but there may be non-linear correlation) 4. r is independent of the units of measurement (unit-free) 5. r is symmetric: correlation of X with Y is the same as Y with X 6. r measures only linear correlation — a zero value does not mean no relationship at all 7. The coefficient of determination r squared tells what percentage of variation in Y is explained by X
Spearman's Rank Correlation
Spearman's rank correlation coefficient is used when:
- Data is qualitative (e.g., beauty, honesty) and can only be ranked
- Actual values are not available, only ranks
- The relationship is non-linear but monotonic (consistently increasing or decreasing)
r_k = 1 - (6 x Sum of D squared) / (n x (n squared - 1))
Where: D = difference in ranks (Rx - Ry) for each observation, n = number of paired observations
Steps to calculate:
1. Rank the X variable values (rank 1 to highest or lowest, consistently) 2. Rank the Y variable values similarly 3. Calculate D = Rx - Ry for each pair 4. Square each D to get D squared 5. Sum all D squared values 6. Apply the formula
When ranks are tied: If two or more items have the same value, assign the average of the ranks they would have occupied. Then apply a correction factor:
Correction factor for each group of ties = (m squared - 1) / 12, where m = number of tied items
Adjusted formula: r_k = 1 - [6(Sum of D squared + Sum of all correction factors) / n(n squared - 1)]
Worked example (no ties):
| Student | Marks in Maths (X) | Marks in Economics (Y) | Rx | Ry | D = Rx - Ry | D squared |
|---|---|---|---|---|---|---|
| A | 80 | 70 | 1 | 2 | -1 | 1 |
| B | 75 | 65 | 2 | 3 | -1 | 1 |
| C | 70 | 75 | 3 | 1 | 2 | 4 |
| D | 60 | 55 | 4 | 4 | 0 | 0 |
| E | 50 | 45 | 5 | 5 | 0 | 0 |
| Total | 6 |
r_k = 1 - (6 x 6) / (5 x (25 - 1)) = 1 - 36/120 = 1 - 0.3 = 0.7 (high positive correlation)
Karl Pearson vs Spearman Comparison
| Feature | Karl Pearson (r) | Spearman (r_k) |
|---|---|---|
| Data type | Quantitative (actual values) | Ranks (qualitative or quantitative) |
| Relationship type | Linear only | Monotonic (linear or non-linear) |
| Outliers | Highly sensitive | Less sensitive (ranks reduce extreme values' effect) |
| Calculation | Uses deviations from mean | Uses rank differences only |
| When to use | Precise numerical data, linear relationship | Ranked data, qualitative data, non-linear monotonic |
| Range | -1 to +1 | -1 to +1 |
| Assumption | Bivariate normal distribution | No distribution assumption needed |
| Precision | More precise (uses full information) | Less precise (loses information by ranking) |
| Example | Correlation between income and expenditure | Correlation between beauty ranking and talent ranking |
Correlation and Causation
This is one of the most important concepts to understand:
Correlation does not imply causation. Just because two variables move together does not mean one causes the other.
Three possible explanations for a correlation:
1. Direct causation — X causes Y (e.g., study hours cause higher marks) 2. Reverse causation — Y causes X (e.g., does higher income cause more education, or vice versa?) 3. Third variable (spurious correlation) — Both X and Y are caused by a third variable Z (e.g., ice cream sales and drowning deaths both increase in summer due to heat)
Classic examples of spurious correlation:
- Number of firemen and fire damage — more firemen at bigger fires, but firemen do not cause damage
- Stork population and birth rate in European cities — both correlated with urbanisation, not causation
- Shoe size and reading ability in children — both increase with age, not because shoes help reading
How to establish causation:
- Look for a logical cause-effect relationship
- Check if the relationship holds across different samples and times
- Use controlled experiments where possible
- Apply economic theory and common sense
- Rule out alternative explanations (third variables)
Worked Example: Karl Pearson's Coefficient (Direct Method)
| X | Y | X squared | Y squared | XY |
|---|---|---|---|---|
| 10 | 20 | 100 | 400 | 200 |
| 20 | 30 | 400 | 900 | 600 |
| 30 | 40 | 900 | 1600 | 1200 |
| 40 | 50 | 1600 | 2500 | 2000 |
| 50 | 60 | 2500 | 3600 | 3000 |
| Sum | 5500 | 9000 | 7000 |
n = 5, Sum of X = 150, Sum of Y = 200
r = [5 x 7000 - 150 x 200] / sqrt[(5 x 5500 - 150 squared) x (5 x 9000 - 200 squared)] = [35000 - 30000] / sqrt[(27500 - 22500) x (45000 - 40000)] = 5000 / sqrt[5000 x 5000] = 5000 / 5000 = +1.0 (perfect positive correlation)
Important Points for Exams
1. Correlation measures the degree and direction of the linear relationship between two variables. 2. Positive correlation: both variables move in the same direction. 3. Negative correlation: variables move in opposite directions. 4. The value of r always lies between -1 and +1. 5. r = +1 is perfect positive; r = -1 is perfect negative; r = 0 means zero linear correlation. 6. Karl Pearson's coefficient is used for quantitative data with a linear relationship. 7. Spearman's rank correlation is used for ranked or qualitative data. 8. Correlation does not imply causation — two variables may be correlated due to a third factor. 9. A scatter diagram gives a visual idea of correlation but not an exact numerical measure. 10. When ranks are tied in Spearman's method, use average ranks and apply the correction factor. 11. The coefficient of determination (r squared) measures the proportion of variation explained.
Common Mistakes
| Mistake | Why it is wrong | Correct approach |
|---|---|---|
| Concluding that correlation means causation | Two variables may be correlated due to a third factor or by coincidence | Always look for a logical cause-effect explanation |
| Using Pearson's coefficient for qualitative data | Pearson's r requires quantitative (numerical) data | Use Spearman's rank correlation for qualitative or ranked data |
| Ignoring tied ranks in Spearman's method | Tied ranks affect the formula; ignoring them gives wrong results | Assign average ranks and apply the correction factor |
| Interpreting r = 0 as "no relationship" | r = 0 only means no linear correlation; a non-linear relationship may exist | Check with a scatter diagram before concluding |
| Confusing the direction and degree of correlation | Direction (positive/negative) and degree (high/low) are different | A correlation can be high negative (-0.9) or low positive (+0.2) |
| Not squaring D values in Spearman's formula | D values may be negative; squaring ensures all are positive | Always compute D squared, not D |
| Not verifying that r falls between -1 and +1 | If your answer is outside this range, there is a calculation error | Recheck if r < -1 or r > +1 |
NCERT Exercise Focus
Focus on these NCERT exercise questions while preparing:
1. "Calculate Karl Pearson's coefficient of correlation" — Practice both deviation method and direct method. 2. "Calculate Spearman's rank correlation with and without tied ranks" — Practice both scenarios. 3. "Correlation does not imply causation — explain with examples" — Give 2-3 examples of spurious correlation. 4. "Interpret the scatter diagram" — Know how to read type and degree from a visual. 5. "Distinguish between Karl Pearson's and Spearman's methods" — Prepare a comparison table.
Exam Tips
1. Numerical on Karl Pearson's coefficient — This is the most commonly tested calculation. Practice the deviation method and the direct method. Use whichever is easier for the given data. 2. Numerical on Spearman's rank correlation — Practice ranking data, calculating D and D squared, and applying the formula. Tied ranks are a common twist. 3. Scatter diagram interpretation — You may be given a scatter diagram and asked to identify the type and degree of correlation. 4. "Correlation does not imply causation" — This is a very common 3-mark conceptual question. Give at least two examples of spurious correlation. 5. Pearson vs Spearman distinction — Prepare a comparison table. Know when to use each. 6. Always verify your r value — It must be between -1 and +1. If your answer falls outside this range, recheck calculations. 7. Degree of correlation — Know the interpretation table. If r = 0.6, call it "moderate positive", not "high positive".
Quick Revision Notes
- Correlation: measures degree and direction of relationship between two variables.
- Types: positive (same direction), negative (opposite direction), zero (no linear relation).
- Degrees: perfect (+1 or -1), high, moderate, low, zero.
- Scatter diagram: visual tool — upward cluster = positive, downward = negative, random = zero.
- Karl Pearson's r = Sum of xy / sqrt(Sum of x squared x Sum of y squared). Range: -1 to +1.
- Direct formula: r = [N x Sum of XY - (Sum X)(Sum Y)] / sqrt[(N x Sum of X squared - (Sum X) squared) x (N x Sum of Y squared - (Sum Y) squared)].
- Spearman's r_k = 1 - 6(Sum of D squared) / n(n squared - 1). For ranked/qualitative data.
- Tied ranks: use average of ranks, apply correction factor.
- Correlation is not causation: spurious correlation may exist due to a third variable.
- Pearson: quantitative, linear. Spearman: qualitative, monotonic.
- Always check: r must be between -1 and +1.