Correlation
Correlation is the statistical study of the direction and intensity of the relationship between two variables — measuring how they move together without claiming one causes the other.
It builds the core analytical reflex UPSC loves to test: 'correlation is not causation', plus the ability to read statistical relationships in data — directly useful for CSAT data interpretation and reasoning. For GS-III (Economy), it underpins how aspirants interpret Economic Survey figures and relationships like money supply and price index, agricultural rainfall-yield links, and supply-price movements. Prelims may probe the −1 to +1 range, properties of r, and who developed Spearman's rank correlation.
Understand the chapter
What Correlation Measures
Correlation studies and measures the direction and intensity of the relationship among variables. Crucially, it measures covariation, not causation — the presence of correlation only means that when one variable changes, the other changes in a definite way (same or opposite direction). It should never be read as implying a cause-and-effect relation.
- Direction = whether variables move together (positive) or oppositely (negative)
- Intensity = how strong the association is
- For simplicity, the chapter assumes correlation, if present, is linear (representable by a straight line)
Types of Relationship: Genuine, Coincidence, Spurious
Some relationships permit a cause-effect reading (low rainfall and low farm productivity), but others are pure coincidence (migratory birds' arrival and local birth rates; shoe size and money in pocket). A third type is spurious — a hidden third variable drives two others. Brisk ice-cream sales correlating with drowning deaths is explained by a third factor: rising temperature increases both ice-cream sales and swimming, hence drownings.
- Cause-effect: rainfall → agricultural productivity
- Coincidence: migratory birds & birth rates; shoe size & pocket money
- Spurious (third variable): ice-cream sales & drownings, both driven by temperature
- Even real relationships may be hard to explain
Positive vs Negative Correlation
Correlation is commonly classified as positive or negative. It is positive when variables move together in the same direction — income rising with consumption, or temperature with ice-cream sales. It is negative when they move in opposite directions — apple price falling while demand rises, or more study hours reducing chances of failure.
- Positive: income↑ → consumption↑ (same direction)
- Negative: price↑ → demand↓ (opposite direction)
- Assumed linear: relative movement shown by a straight line on graph
Three Techniques for Measuring Correlation
The chapter gives three tools: scatter diagram, Karl Pearson's coefficient of correlation, and Spearman's rank correlation. A scatter diagram visually shows the form and closeness of association without any numerical value. Karl Pearson's coefficient gives a precise number for linear relationships, while Spearman's measures the linear association between ranks — useful for attributes that cannot be numerically measured.
- Scatter diagram: visual only, no numerical value; closeness + direction of points reveal the relationship
- Karl Pearson's coefficient: precise numerical degree of linear relationship
- Spearman's rank correlation: for ranks/attributes like intelligence, honesty, appearance
Karl Pearson's Coefficient and Its Properties
Also called the product moment correlation coefficient or simple correlation coefficient, it must be used only for linear relations — applying it to non-linear data is misleading, so examining the scatter diagram first is advisable. The sign of the covariance determines the sign of r, since standard deviations are always positive. Its key properties are central to exams.
- r has no unit — a pure number; lies within −1 ≤ r ≤ 1 (outside = calculation error)
- Magnitude of r is unaffected by change of origin and scale (basis of step deviation method)
- r = 0 → uncorrelated (no LINEAR relation; non-linear may still exist); r = ±1 → perfect, exact linear relation
- Worked examples: schooling vs yield r = 0.644; price index vs money supply r = 0.98
Spearman's Rank Correlation and Step Deviation Method
Spearman's rank correlation, developed by British psychologist C.E. Spearman, is used when variables cannot be measured numerically but can be ranked — like ranking students by height/weight without instruments, or judging fairness, honesty and beauty. Separately, the step deviation method exploits r's independence of origin and scale to drastically cut calculation burden when values are large.
- Attributes = variables that cannot be numerically measured (intelligence, honesty, beauty)
- Spearman works on ranks assigned to items by their attributes
- Step deviation method: transform U=(X−A)/B, V=(Y−C)/D; r_uv = r_xy
- Simplifies computation for large data values
Key terms
- Correlation
- A statistical measure of the direction and intensity of the relationship between two variables.
- Covariation
- Joint variation of two variables; what correlation measures — distinct from causation.
- Positive correlation
- Variables move in the same direction (both rise or both fall together).
- Negative correlation
- Variables move in opposite directions (one rises as the other falls).
- Linear correlation
- A relationship that can be represented by a straight line on a graph.
- Scatter diagram
- A graph plotting paired values as points to visually show the relationship, without any numerical value.
- Karl Pearson's coefficient (r)
- Product moment / simple correlation coefficient giving a precise numerical degree of LINEAR relationship.
- Spearman's rank correlation
- Correlation between ranks, used for attributes that cannot be numerically measured; by C.E. Spearman.
- Attributes
- Qualitative variables that cannot be numerically measured, such as honesty, beauty or intelligence.
- Step deviation method
- A simplification using r's invariance to change of origin and scale to reduce calculation for large values.
Must-know facts exam-ready
- Correlation measures direction and intensity, and is covariation — NOT causation.
- The correlation coefficient lies strictly between −1 and +1; any value outside indicates a calculation error.
- r = 0 means uncorrelated (no linear relation), but a non-linear relation may still exist.
- r = +1 or −1 means perfect correlation with an exact linear relation.
- r has no unit — it is a pure number, unaffected by change of origin and scale.
- Karl Pearson's coefficient is also called the product moment correlation coefficient or simple correlation coefficient.
- Karl Pearson's r should be used only for linear relations; it misleads for non-linear data.
- The three techniques are: scatter diagram, Karl Pearson's coefficient, and Spearman's rank correlation.
- Spearman's rank correlation was developed by British psychologist C.E. Spearman.
- Scatter diagram is purely visual (no numerical value); Karl Pearson gives the numerical value.
- The sign of covariance determines the sign of r; standard deviations are always positive.
- Chapter examples: schooling vs annual yield r = 0.644; price index vs money supply r = 0.98.
Memory tricks remember it for good
Traps to avoid
- Treating correlation as causation — the chapter stresses it measures covariation only (deaths vs doctors during an epidemic looked positive but doctors don't cause deaths).
- Reading r = 0 as 'no relationship' — it only rules out a LINEAR relation; a non-linear relation may still exist.
- Assuming Karl Pearson's r works for any data — it is valid only for linear relations and misleads on non-linear scatter (Fig 6.6, 6.7).
- Confusing scatter diagram (visual, no number) with Karl Pearson's coefficient (precise numerical value).
- Mixing up Karl Pearson (numerical, measurable variables) with Spearman (ranks, attributes like honesty/beauty).
- Forgetting r is bounded: a value outside −1 to +1 signals a calculation error, not a strong relationship.
Exam focus
🧠 Prelims angles
- The fixed range of the correlation coefficient: −1 ≤ r ≤ 1, and meaning of r = 0, +1, −1.
- Correlation vs causation distinction and spurious correlation via a third variable (CSAT reasoning/data interpretation).
- Properties of r: unit-free, pure number, unaffected by change of origin and scale.
- Who developed Spearman's rank correlation (British psychologist C.E. Spearman).
- Identifying positive vs negative correlation from given variable movements or scatter diagrams.
- Karl Pearson's alternate names: product moment / simple correlation coefficient, and its linear-only validity.
✍️ Mains angles GS-III
- 'Correlation does not imply causation' — examine with reference to interpreting economic and policy data.Use chapter cases (ice-cream–drowning, doctors–deaths) to argue for checking hidden third variables before drawing policy conclusions.
- Role of statistical correlation in reading macroeconomic relationships such as money supply and price index.Cite the strong positive r (0.98) as a premise of monetary policy, while cautioning against over-reading correlation.
Last-minute revision tick as you recall
- Correlation = direction + intensity; it is covariation, NOT causation.
- Positive = same direction; Negative = opposite direction; assumed linear.
- Three tools: Scatter diagram, Karl Pearson's coefficient, Spearman's rank correlation.
- −1 ≤ r ≤ 1; r = 0 uncorrelated (linearly), r = ±1 perfect linear relation.
- r is unit-free and unaffected by change of origin and scale.
- Karl Pearson = product moment coefficient; valid only for linear relations.
- Spearman (by C.E. Spearman) handles ranks/attributes like honesty and beauty.
- Step deviation method simplifies calculation for large values using r's invariance.
- Watch spurious correlation: a hidden third variable (e.g., temperature) can fake links.
Distilled from NCERT Class 11 · Statistics for Economics for UPSC. Always cross-check facts with the original NCERT.