Chapter 7 — Correlation

Meaning: Correlation is a statistical measure that describes the direction and degree of association between two variables. It tells us whether and how strongly two variables move together.

Definition (simple): Correlation between two variables exists if values of one variable are related systematically to values of the other. In economics, correlation helps to study relationships such as price and demand, income and consumption, or advertising and sales.

Formal idea: If high (or low) values of X are associated with high (or low) values of Y, X and Y are positively correlated. If high values of X are associated with low values of Y, they are negatively correlated. If no systematic relationship is observed, there is zero (or no) correlation.

Key characteristics:

• Direction: positive, negative, or zero.
• Degree (strength): ranges from weak to strong; measured numerically by correlation coefficients (e.g., Pearson's r, Spearman's rho).
• Unit-free: correlation coefficients do not depend on the units of measurement.
• Correlation does not imply causation: a statistical association does not prove one variable causes the other.

Types of correlation:

• Perfect positive correlation: all points lie on an upward straight line (coefficient = +1).
• Perfect negative correlation: all points lie on a downward straight line (coefficient = −1).
• Imperfect (partial) correlation: points form a cloud with a clear trend upward or downward (|r| < 1).
• No correlation: points show no discernible pattern (r ≈ 0).
• Linear vs. non-linear (curvilinear) correlation: association may be roughly straight-line or curved.

How it is studied: The simplest graphical method is a scatter diagram (scatter plot) where each pair (x_i, y_i) is plotted. Numerical measures include covariance and correlation coefficients (Pearson's r for linear association; Spearman's rho for ranks).

Why it is useful in economics: Correlation helps to identify relationships for forecasting, policy analysis and decision making — for example, how income changes with education, or how demand changes with price. But economists must test causality separately (experiments, instrumental variables, time series analysis).

Definition: Correlation describes the degree and direction of association between two quantitative variables. It does not imply causation — only how two variables move together.

Main types of correlation:

• Positive (Direct) correlation: As one variable increases, the other also increases. Points on a scatter plot trend upward from left to right. Degree can be perfect, high, moderate or low.
• Negative (Inverse) correlation: As one variable increases, the other decreases. Points trend downward from left to right. Degree can vary from perfect to weak.
• No (Zero) correlation: No discernible pattern; variables do not show any consistent increase or decrease together.
• Perfect correlation: All points lie exactly on a straight line. Pearson correlation coefficient r = +1 (perfect positive) or r = −1 (perfect negative).
• Imperfect/Partial correlation: Points roughly follow a trend but are scattered around a line (common in real data). Degree described as high, moderate or low.
• Linear vs Curvilinear (nonlinear) correlation: Linear correlation follows (approximately) a straight-line relationship. Curvilinear correlation shows a systematic but non‑linear pattern (e.g., U-shaped or inverted-U).

Important point: Correlation coefficient (r) quantifies the direction and strength: r ranges from −1 to +1. The sign gives direction; the magnitude gives strength (|r| close to 1 = stronger linear relationship).

Definition: A scatter diagram (scatter plot) is a two-dimensional graph used to display paired observations (x, y) so that the relationship between two quantitative variables can be seen visually. Each pair is plotted as a point with horizontal coordinate x and vertical coordinate y.

Purpose: To detect the existence, direction and approximate strength of a relationship (correlation) between two variables before calculating a correlation coefficient or fitting a regression line.

Construction (step-by-step):

• Select the two variables to examine (independent on x-axis, dependent on y-axis).
• Choose suitable scales and label axes with units.
• Plot each paired observation (x, y) as a point.
• Look for a pattern: upward, downward, or no clear pattern.
• If needed, draw a line of best fit (trend line) — often the least squares regression line — and mark the mean point (x̄, ȳ).

Interpretation of patterns:

• Positive correlation: points tend to slope upward from left to right (as x increases, y tends to increase).
• Negative correlation: points slope downward (as x increases, y tends to decrease).
• No correlation: points are scattered randomly with no clear trend.
• Perfect correlation: all points lie exactly on a straight line (rare in real data).

Relation to quantitative measures: A scatter diagram gives a graphical impression of the relationship which is then quantified by the correlation coefficient (r) and by regression equations (y = a + bx).

Uses in Economics (Class 11 context): To visualise relationships such as consumption vs income, advertising expenditure vs sales, price vs quantity demanded (negative relationship), investment vs interest rate, GDP vs employment, etc. It helps in preliminary analysis before formal correlation/regression calculations.

Advantages: Simple, intuitive, shows outliers and non-linear patterns, useful for preliminary analysis.

Limitations: Does not give a numerical measure of strength (need r), may be affected by scale and outliers, visual impression can be subjective.

Definition: Karl Pearson's coefficient of correlation (r) measures the strength and direction of the linear relationship between two continuous variables X and Y. Its value lies between -1 and +1. A positive r indicates that as X increases, Y tends to increase; a negative r indicates that as X increases, Y tends to decrease; r = 0 indicates no linear association.

Mathematical meaning: For a sample of n paired observations (x_i, y_i), Pearson's r is the standardized covariance of X and Y. It quantifies how much X and Y vary together relative to how much they vary individually.

Key formula (definitional):

• r = [Σ (x_i - x̄)(y_i - ȳ)] / sqrt{[Σ (x_i - x̄)²] [Σ (y_i - ȳ)²]}

Equivalent computational formula (using raw sums):

• r = [n Σ(x_i y_i) - (Σ x_i)(Σ y_i)] / sqrt{[n Σ(x_i²) - (Σ x_i)²] [n Σ(y_i²) - (Σ y_i)²]}

Interpretation of magnitude (rule of thumb):

• |r| close to 1: very strong linear relationship
• |r| around 0.5–0.8: moderate relationship
• |r| around 0.2–0.5: weak relationship
• |r| close to 0: little or no linear relationship

Steps to calculate (practical):

1. Compute means x̄ and ȳ.
2. For each pair, compute deviations (x_i - x̄) and (y_i - ȳ) and their product.
3. Sum the products to get Σ (x_i - x̄)(y_i - ȳ).
4. Compute Σ (x_i - x̄)² and Σ (y_i - ȳ)².
5. Plug into the formula and evaluate r.

Properties and notes:

• -1 ≤ r ≤ +1; r is dimensionless (unit-free).
• r is symmetric: correlation(X,Y) = correlation(Y,X).
• r measures only linear association. Strong non-linear relationships can have r ≈ 0.
• Correlation ≠ causation: a high |r| does not imply X causes Y.
• For population correlation the symbol is ρ (rho); r is the sample estimator.
• Coefficient of determination r² gives the proportion of variance in Y explained by X (for linear fits).

Assumptions (when interpreting r): both variables should be quantitative, roughly normally distributed when doing inference, and the relationship should be approximately linear without extreme outliers (outliers can distort r).

What is correlation? Correlation measures the degree and direction of linear relationship between two variables X and Y. It shows whether higher (or lower) values of one variable tend to be associated with higher (or lower) values of the other.

Main methods of studying correlation

• Scatter diagram (graphical): Plot paired observations (x,y) on a Cartesian plane. Visual patterns indicate direction (positive/negative), strength (tight/loose clustering), and form (linear/non‑linear).
• Karl Pearson’s coefficient of correlation (numerical): A single number r (–1 ≤ r ≤ 1) that measures the strength and direction of the linear relation. r > 0 indicates positive correlation, r < 0 negative, r = 0 no linear correlation.
• Methods of computation: Several algebraic forms are used depending on available data: raw-sum (shortcut) formula, deviation-from-mean form, coded (or step) method for large numbers, and grouped-data method using class midpoints.

Interpretation and cautions

• The sign of r gives direction; its magnitude gives strength (closer to 1 in absolute value = stronger linear relation).
• Correlation measures linear association only. r = 0 does not imply independence—there may be a strong non-linear relation.
• Correlation ≠ causation. Outliers strongly affect r.

Stepwise computational approaches (conceptual)

1. Scatter diagram: Plot points and, if appropriate, draw a line of best fit (trend line). This gives a visual estimate of sign and strength and helps decide whether Pearson's r is suitable.
2. Pearson’s r – raw-sum (shortcut) form: Use when you have raw x and y values and can compute sums Σx, Σy, Σxy, Σx^2, Σy^2. This form reduces algebra compared with using deviations.
3. Deviation-from-mean form (definition): Compute covariance Σ(x − x̄)(y − ȳ) and divide by the product of standard deviations. This form is conceptually clear and numerically stable if means are known.
4. Coded (step) method: For large figures, transform X and Y to coded variables u = (x − A)/h and v = (y − B)/k where A, B are convenient origins and h, k are convenient units. Compute r using sums of u, v and their products. Final r is unchanged by linear coding.
5. Grouped data: Replace each class by its midpoint, use frequencies f, and compute sums f*x_m, f*y_m, f*x_m*y_m, f*x_m^2, etc., then apply Pearson formula for grouped data.

Limitations: Pearson’s r assumes interval/ratio scale, linear relation, and is sensitive to extreme values. For ordinal data, use rank correlation (Spearman).

What is covariance?
Covariance is a measure that shows how two variables move together. If high values of one variable tend to occur with high values of the other, covariance is positive; if high values of one tend to occur with low values of the other, covariance is negative. If there is no linear relationship, covariance is near zero.

Definitions and basic formulas
- Population covariance: Cov(X,Y) = (1/N) Σ (x_i - x̄)(y_i - ȳ).
- Sample covariance (used with sample size n): S_xy = (1/(n-1)) Σ (x_i - x̄)(y_i - ȳ).

Computational (shortcut) forms
- Population: Cov(X,Y) = (1/N) Σ x_i y_i - x̄ ȳ.
- Sample: S_xy = [Σ x_i y_i - n x̄ ȳ] / (n-1).

Grouped (frequency) data
- If f_i are joint frequencies and x_i,y_i are mid-points: Cov = (1/Σ f_i) Σ f_i (x_i - x̄)(y_i - ȳ).

Interpretation and properties
- Sign indicates direction: positive > 0 (variables move in same direction), negative < 0 (opposite directions), zero ≈ 0 (no linear tendency).
- Magnitude depends on units of X and Y (so covariance alone is not standardized).
- Covariance is unbounded (value depends on scale); to get a standardized measure we use correlation.

Relation to correlation (Pearson's correlation coefficient)
Pearson's correlation coefficient r (for population often denoted ρ) standardizes covariance so that its value lies between −1 and +1:
- r = Cov(X,Y) / (σ_x · σ_y).
- For sample quantities: r = S_xy / (s_x · s_y), where s_x and s_y are sample standard deviations.
This division removes the effect of units and scale. Because of the Cauchy–Schwarz inequality, |r| ≤ 1. When r = +1 or −1 the points lie exactly on a straight line (perfect positive or negative linear relationship).

Connection with regression and variance
- Slope of the least-squares regression line of Y on X: b = Cov(X,Y) / Var(X) (using consistent population/sample versions).
- Covariance links variance and linear association: Var(X+Y) = Var(X) + Var(Y) + 2 Cov(X,Y).

Takeaway
Covariance tells direction and raw strength of linear co-movement but depends on units. Correlation is the unit-free standardized form (covariance divided by product of standard deviations) and lies between −1 and +1.

Pearson's correlation coefficient (r) measures the strength and direction of the linear relationship between two variables X and Y. It is defined (for sample data) as:

r = [Σ(x - x̄)(y - ȳ)] / sqrt{Σ(x - x̄)² · Σ(y - ȳ)²}

The main properties of Pearson's r used in Class 11 Economics are:

• Range: −1 ≤ r ≤ +1. Values near +1 indicate a strong positive linear relationship; values near −1 indicate a strong negative linear relationship; values near 0 indicate little or no linear relationship.
• Sign shows direction: Positive r → as X increases, Y tends to increase. Negative r → as X increases, Y tends to decrease.
• Magnitude shows strength: |r| close to 1 means points lie nearly on a straight line; |r| close to 0 means weak or no linear relation.
• Dimensionless (unit-free): r has no units. It is unaffected by the units in which X and Y are measured.
• Invariant to change of origin and scale: Adding a constant to X or Y (change of origin) does not change r. Multiplying X or Y by a positive constant (change of scale) does not change r. Multiplying by a negative constant reverses the sign of r.
• Symmetry: r(X,Y) = r(Y,X). The coefficient is the same whichever variable is treated as X or Y.
• Only measures linear association: r detects linear relationships. A value r ≈ 0 does not imply independence — variables may be nonlinearly related (e.g., Y = X² with symmetric X gives r ≈ 0 but dependence exists).
• Perfect correlation conditions: r = +1 or −1 iff all points lie exactly on a straight line with positive or negative slope respectively. In that case one variable is an exact linear function of the other.
• Relation to covariance: r = cov(X,Y) / (σX · σY). Thus the sign of r equals the sign of covariance and standardizes covariance by the product of standard deviations.
• Sensitivity to outliers: r is affected strongly by extreme observations. A single outlier can change r substantially.
• Connection with explained variance: r² (coefficient of determination) gives the proportion of variance in Y explained by a linear relation with X (in simple linear regression).

Typical classroom uses: compute r from data (computational formula), interpret its value as direction/strength, draw scatter plots to visualize linearity and outliers, and use r² to discuss how much variation is explained by a linear trend.

What to interpret from a correlation coefficient

The correlation coefficient (r) measures the degree and direction of linear association between two variables X and Y.

• Sign (positive/negative): If r > 0 the variables move in the same direction (positive correlation). If r < 0 they move in opposite directions (negative correlation). If r = 0 there is no linear relationship.
• Magnitude (strength): The absolute value |r| (0 to 1) indicates how strong the linear relationship is. Larger |r| means a stronger linear association.
• Direction vs causation: Correlation shows association only. It does not prove that changes in X cause changes in Y.
• Linear limitation: Pearson's r measures linear association. Non-linear but strong relationships can give low |r|.
• Effect of outliers: Outliers can substantially change r, so inspect scatter plots before concluding.

Common strength classification (approximate and widely used)

• |r| = 1 : Perfect linear relationship
• 0.8 ≤ |r| < 1 : Very strong (high) linear relationship
• 0.5 ≤ |r| < 0.8 : Moderate linear relationship
• 0.3 ≤ |r| < 0.5 : Weak linear relationship
• |r| < 0.3 : Negligible or very weak linear relationship

Note: these cut-offs are conventional — different textbooks use slightly different class boundaries. Always report the actual r and show a scatter plot.

Interpretation using coefficient of determination

The square of the correlation coefficient, r² (called the coefficient of determination), gives the proportion (or percentage) of variance in Y explained by X in a linear model. For example, r = 0.7 gives r² = 0.49, so about 49% of the variation in Y can be explained by X (linearly).

Practical guidance

• Always inspect the scatter plot and, if appropriate, fit a regression line to visualise the linear trend.
• Check for outliers and non-linearity; a low r does not mean no relationship if the pattern is curved.
• Remember correlation ≠ causation; use controlled studies or additional knowledge to support causal claims.

What it is
Spearman's Rank Correlation Coefficient (denoted r_s) is a non‑parametric measure of the strength and direction of the monotonic relationship between two variables. It is used when data are ordinal (ranks) or when assumptions for Pearson correlation (linearity, interval scale, normality) are not satisfied.

When to use

• Comparing ranks (e.g., positions, ratings).
• Detecting monotonic relationships (not necessarily linear).
• Small samples or when data contain outliers or are not normally distributed.

Procedure (step by step)

1. Rank the values of each variable separately. If ties occur, assign average ranks to tied values.
2. For each observation i, compute the difference d_i = R_xi - R_yi between the two ranks.
3. Compute d_i² and find Σd_i².
4. If there are no tied ranks, apply the formula r_s = 1 - (6 Σd_i²)/(n(n² - 1)). If ties are present, compute Pearson correlation on the ranks (see note below).

Interpretation

• r_s ranges from -1 to +1.
• r_s = +1: perfect positive monotonic agreement of ranks.
• r_s = -1: perfect negative monotonic (one rank increases while the other decreases).
• r_s ≈ 0: little or no monotonic association.

Tie handling (brief)
When ties exist, use average ranks for tied observations and either compute Pearson's correlation coefficient on the ranks or apply a tie-correction version. The Pearson-on-ranks method is commonly used and gives a correct r_s in presence of ties.

Worked numeric example (n = 6)

Apply formula: r_s = 1 - [6 × 6] / [6(6² - 1)] = 1 - 36 / 210 = 0.8286 (approx). This indicates a strong positive monotonic association.

Notes on statistical testing
For inference, critical values or t-approximation (for large n) may be used to test if r_s differs significantly from zero.

Key advantages

• Robust to outliers and non-normal data.
• Works with ordinal data.

Limitations

• Detects monotonic relationships only; a non-monotonic but strong relationship may give low r_s.
• Ties complicate computation and interpretation slightly.

What it means
Correlation for grouped data measures the strength and direction of linear association between two quantitative variables when observations are given in class intervals (grouped) rather than as individual pairs. In practice we replace each class by its class-midpoint and use the joint frequencies to compute the Pearson correlation coefficient approximately.

When to use
Use this method when raw paired data are summarized into a bivariate grouped (contingency) table: classes of X on one axis, classes of Y on the other, with joint frequencies f_ij in each cell.

Basic idea (midpoint method)

1. For each X-class i take its midpoint x_i. For each Y-class j take its midpoint y_j.
2. Use the joint frequencies f_ij as weights to compute sums involving x and y: Σf_ijx_i, Σf_ijy_j, Σf_ijx_iy_j, Σf_ijx_i², Σf_ijy_j². Let N = Σf_ij (total frequency).
3. Compute the Pearson correlation coefficient r using these weighted sums (formula given below).

Step-by-step procedure

1. Create a table where each cell contains f_ij and the corresponding midpoints (x_i, y_j).
2. Compute products f_ijx_i, f_ijy_j, f_ijx_iy_j, f_ijx_i², f_ijy_j², and sum each column.
3. Plug sums into the correlation formula to get r.
4. Interpret r: r close to +1 strong positive linear relation, r close to −1 strong negative linear relation, r near 0 little or no linear relation.

Coding (short) method
To reduce arithmetic, you may code midpoints. For X use u_i = (x_i − a)/k_x, for Y use v_j = (y_j − b)/k_y where a,b are convenient assumed means and k's are scaling factors (often class widths). Then compute sums with u and v and use the same algebraic form for r. After computing r with coded values, r is invariant to linear scaling so the result is the same as with original midpoints.

Assumptions & limitations

• Using class midpoints is an approximation: within-class variation is ignored.
• The Pearson r measures only linear association. Nonlinear associations may not be detected.
• Sensitive to outliers (but grouping tends to mask individual outliers).

Interpretation tips
Always report N (total frequency), the value of r, and a brief comment on strength and direction (e.g. "r = 0.78, strong positive linear association"). Consider plotting the grouped scatter (see graph suggestions) to visually confirm linearity.

Overview: Correlation measures the degree and direction of association between two variables (e.g., income and consumption). In Class 11 Economics we mainly use Karl Pearson's coefficient (for linear relationships) and Spearman's rank correlation (for ordinal/ranked data). Understanding the assumptions, limitations and precautions is essential to use correlation correctly.

Assumptions:

• Linearity (for Pearson): The relationship between the two variables should be approximately linear. Pearson's r measures only linear association.
• Type of data: Pearson requires interval or ratio scale (continuous) data; Spearman requires ordinal or rankable data (or continuous data converted to ranks).
• Independence: Observations should be independent of each other (no repeated measures on same unit unless handled appropriately).
• Homoscedasticity: The variance of one variable should be roughly constant across values of the other (important for inference).
• Absence of influential outliers: Outliers can distort the value of Pearson's r; Spearman is less sensitive to outliers.
• Monotonicity (for Spearman): Spearman assumes a monotonic relationship (values consistently increase or decrease) but not necessarily linear.
• Sample size: A reasonable sample size is needed for stable estimates and reliable significance testing.

Limitations:

• Correlation is not causation: A high correlation does not prove that one variable causes the other. There may be reverse causation or a common causal factor.
• Only measures linear association (Pearson): Non-linear relationships (e.g., quadratic) can produce low Pearson r even when variables are strongly related.
• Sensitive to outliers: A single extreme value can substantially change Pearson's r and give misleading impression.
• Range restriction: If data cover a narrow range of values, correlation may be underestimated (attenuation).
• Spurious correlations: Two variables may correlate because of coincidence or because they both depend on a third variable (confounder) or seasonality.
• No multivariate insight: Simple correlation considers only two variables; it ignores other influencing variables.
• Measurement error: Errors in measuring variables reduce correlation.

Precautions when using correlation:

• Draw a scatter plot first to visually inspect linearity, outliers, clusters, and heteroscedasticity.
• Choose appropriate coefficient: Use Pearson for linear continuous data; use Spearman for ranks or monotonic but non-linear relationships.
• Check for outliers and investigate them (data entry errors, special cases); compute correlations with and without them to assess influence.
• Consider confounding variables: Think whether a third factor might explain the observed correlation.
• Test statistical significance (e.g., t-test for Pearson) before making inferences.
• Be careful interpreting small samples—correlations can be unstable.
• Report direction and strength and, where useful, r^2 (percentage of variance explained) rather than just saying 'related'.

Short practical checklist for students:

• Plot the data (scatter diagram)
• Look for linearity, outliers and range issues
• Decide Pearson vs Spearman
• Compute r and r^2; test significance if needed
• Interpret carefully: state association, not causation

Correlation measures the strength and direction of the relationship between two quantitative variables. The topic 'Applications and Examples' shows where correlation analysis is useful in economics and everyday decisions, how to compute and interpret correlation coefficients, and what cautions to keep in mind (especially that correlation does not imply causation).

Why it is useful

• Helps identify relationships for forecasting (e.g., sales vs advertising).
• Supports policy and business decisions (e.g., income and consumption patterns).
• Summarises degree of linear association and guides further analysis (regression, causality tests).

How to apply (practical steps)

1. Collect paired observations (x, y) for the variables of interest.
2. Draw a scatter plot to look for pattern: positive, negative, none, or curvilinear.
3. Compute the correlation coefficient (e.g., Pearson's r for linear association or Spearman's r_s for ranks).
4. Interpret sign (positive/negative) and magnitude (close to 1 or 0) and check significance if required.
5. Remember: a strong correlation is not proof of causation — look for confounding variables and use controlled analysis if causality is claimed.

Key points on interpretation

• r > 0: positive association (as x increases, y tends to increase). r < 0: negative association.
• |r| near 1: strong linear relationship; |r| near 0: weak or no linear relationship.
• Outliers can distort r. Curvilinear relationships may show low r even when variables are related.

Common economic applications

• Price and demand analysis (to examine direction and strength of association before regression).
• Income and consumption (to identify propensity to consume patterns).
• Advertising expenditure and sales forecasting.
• Education: study time and exam performance.
• Macroeconomics: investment and GDP growth, unemployment and inflation trade-offs (in exploratory analysis).

Limitations and cautions

• Correlation does not establish causation — other factors or reverse causality may exist.
• Correlation measures linear association; non-linear relationships require other tools.
• Ties, grouped data, and small samples need appropriate formulas or careful interpretation.