Correlation vs. Causation — I can distinguish correlation from causation using study descriptions. - Grade 9 Math

SECTION 1

Historical Context & Motivation

Humans are natural pattern-seekers. For thousands of years, people noticed that certain events seemed to go together—roosters crow before sunrise, so early civilizations sometimes believed the rooster caused the sun to rise. This kind of faulty reasoning is at the heart of one of the most important distinctions in statistics: the difference between correlation (two things happening together) and causation (one thing actually making the other happen). Understanding this distinction helps us avoid jumping to wrong conclusions in science, medicine, and everyday life.

1747 • First Controlled Experiment

Scottish physician James Lind conducted one of the first controlled clinical trials, testing six treatments for scurvy aboard a ship. By isolating variables, he showed that citrus fruit caused recovery—not just correlated with it.

1880s • Galton & Correlation

Sir Francis Galton studied how parents' heights related to their children's heights. His work laid the groundwork for the statistical concept of correlation, a way to measure the strength of a relationship between two variables.

1900 • Pearson's Correlation Coefficient

Karl Pearson formalized the correlation coefficient (r), giving researchers a precise number between −1 and 1 to describe the direction and strength of a linear relationship.

1950s • Smoking & Lung Cancer Debate

Researchers found a strong correlation between cigarette smoking and lung cancer rates. Tobacco companies argued it was "just correlation." Decades of controlled studies eventually proved the causal link, highlighting why distinguishing correlation from causation matters so much.

Today • Big Data & Spurious Correlations

With massive datasets, we can find thousands of random correlations—like the correlation between U.S. cheese consumption and deaths from bedsheet tangling. Statisticians must constantly guard against mistaking coincidence for causation.

These historical moments reveal a recurring challenge: when two variables move together, how do we know if one is truly causing the other? This is the central question that the study of correlation vs. causation helps us answer.

SECTION 2

Core Principles & Definitions

Before diving into examples, you need a solid grasp of the key terms. The ideas below form the foundation for telling the difference between correlation and causation in any study or data set you encounter.

Correlation

A statistical relationship in which two variables tend to change together. When one increases, the other may increase (positive correlation) or decrease (negative correlation). Correlation does NOT tell us why they move together.

Causation

A relationship in which changes in one variable directly produce changes in another. Causation requires that the cause comes before the effect, and other explanations have been ruled out.

Lurking (Confounding) Variable

A hidden third variable that influences both of the variables being studied. Lurking variables can create the appearance of a causal link where none exists.

Observational Study

A study where researchers observe and collect data without changing anything. Observational studies can show correlation but generally cannot prove causation because variables are not controlled.

Controlled Experiment

A study in which researchers manipulate one variable (the independent variable) while keeping everything else the same, then measure the effect. This design can establish causation.

✦

KEY TAKEAWAY

Think of correlation like noticing that your neighbor's car is always wet when your lawn is wet. You might think your sprinkler is hitting their car, but the real cause could be rain—a lurking variable that affects both. Correlation is the observation; causation is the explanation. You need more evidence—like a controlled experiment—to move from one to the other.

SECTION 3

Visual Explanation — How Correlation Can Mislead

The diagram below illustrates three different ways two variables (A and B) can appear related. Understanding these three scenarios is the key to avoiding the common trap of assuming that because two things are correlated, one must cause the other.

Three scenarios that explain how two variables A and B might appear correlated. Only Scenario 1 represents true causation, where A directly changes B. In Scenario 2, a lurking variable C causes both A and B to change, creating a misleading correlation. In Scenario 3, the relationship is pure coincidence—no connection at all.

When you encounter a claim like "students who eat breakfast get higher grades," ask yourself which scenario fits best. Is eating breakfast directly causing the higher grades? Or could a lurking variable—like having a stable home environment—explain both habits? This kind of critical thinking is at the core of statistical literacy.

SECTION 4

Mathematical Framework — The Correlation Coefficient

While the focus of this lesson is on the logic behind correlation and causation, it helps to understand the number that statisticians use to measure correlation. The correlation coefficient, written as r, is a value between −1 and 1 that tells you the direction and strength of a linear relationship between two quantitative variables.

CORRELATION COEFFICIENT

r = (1 / (n − 1)) × Σ [((xᵢ − x̄) / sₓ) × ((yᵢ − ȳ) / sᵧ)]

Where n = number of data pairs, x̄ and ȳ are the means of x and y, and sₓ and sᵧ are the standard deviations. You won't need to calculate this by hand in Grade 9, but knowing what it measures is essential.

What r Tells You

Value of r	Direction	Strength	Example
r = 1	Positive	Perfect	Every point falls on an upward line
0.7 ≤ r < 1	Positive	Strong	Height and shoe size
0.3 ≤ r < 0.7	Positive	Moderate	Study hours and GPA
0 < r < 0.3	Positive	Weak	Age and happiness level
r = 0	None	None	Shoe size and math test score
−1 ≤ r < 0	Negative	Varies	TV hours and test scores

Interpreting the correlation coefficient r

⚠

Critical Rule

No matter how strong the correlation (even if r = 0.99), correlation alone never proves causation. A high r value tells you the relationship is strong and linear, but it says nothing about why the relationship exists.

SECTION 5

Study Types — Observational vs. Experimental

The type of study determines whether you can claim causation. Learning to identify the study design is your most powerful tool for distinguishing correlation from causation. There are two main categories: observational studies and controlled experiments. Each has specific features that you can learn to recognize.

Side-by-side comparison of the two main study types. The key difference: in a controlled experiment, researchers actively manipulate a variable and randomly assign participants to groups, which allows them to rule out lurking variables and claim causation.

How to Identify the Study Type

✓ Look for random assignment: If participants were randomly placed into treatment and control groups, it's an experiment.
✓ Check for a manipulated variable: If researchers changed something on purpose (gave a drug, changed a teaching method), it's an experiment.
✓ Look for surveys or records: If researchers just collected existing data or surveyed people about their habits, it's observational.
✓ Check the conclusion wording: Words like "is associated with" or "is linked to" suggest correlation (observational). Words like "causes" or "leads to" suggest the study claims causation (experiment).

SECTION 6

Worked Example — Analyzing a Study Description

Let's walk through a realistic study description step by step, applying the principles you've learned to determine whether the study shows correlation, causation, or neither.

ℹ

Study Description

A school district surveyed 2,000 high school students about their daily screen time (hours spent on phones, tablets, and computers for non-school purposes) and their grade point averages (GPAs). The results showed that students who reported more than 4 hours of daily screen time had an average GPA of 2.5, while students who reported less than 1 hour had an average GPA of 3.6. The correlation coefficient was r = −0.72.

Analyzing the Screen Time Study

1
Step 1 — Identify the Variables

The two variables in this study are daily screen time (measured in hours) and GPA (measured on a 4.0 scale). Both are quantitative variables, so calculating a correlation coefficient makes sense.

Variable 1: Screen time (hours) | Variable 2: GPA
2
Step 2 — Identify the Study Type

The study description says the school district "surveyed" students. There is no mention of random assignment, no treatment group, and no control group. Researchers did not assign students to use their phones for a certain number of hours—they simply asked students about their existing habits. This is an observational study.

Study type: Observational
3
Step 3 — Interpret the Correlation

The correlation coefficient r = −0.72 tells us two things. First, the negative sign means the relationship is negative: as screen time increases, GPA tends to decrease. Second, the magnitude (0.72) indicates a strong relationship. However, strong does not mean causal.

Strong, negative correlation between screen time and GPA
4
Step 4 — Consider Lurking Variables

Since this is observational, we must consider lurking variables. Students with more screen time might also have less parental supervision, fewer extracurricular commitments, different socioeconomic backgrounds, or less interest in academics. Any of these factors could independently affect both screen time and GPA, creating a correlation without direct causation.

Possible lurking variables: parental involvement, motivation, socioeconomic status
5
Step 5 — State the Correct Conclusion

Because this is an observational study, we can conclude that screen time and GPA are negatively correlated. We cannot conclude that excessive screen time causes lower GPAs. To test causation, researchers would need a controlled experiment—for example, randomly assigning students to limited vs. unlimited screen time groups and comparing their GPAs after a semester.

Conclusion: Correlation, not causation. More screen time is associated with lower GPA, but we cannot say one causes the other.

SECTION 7

Common Mistakes & How to Avoid Them

Even after learning the difference between correlation and causation, students (and even professional journalists!) frequently make certain errors. The table below lays out the most common mistakes alongside the correct reasoning.

Common Mistake	Why It's Wrong	Correct Approach
"Strong correlation = causation"	Even r = 0.95 only means the variables are strongly associated. It says nothing about why.	State that the variables are strongly correlated and explain that a controlled experiment is needed to establish causation.
Ignoring lurking variables	A hidden third factor may be driving both variables. If you don't consider it, your conclusion is incomplete.	Always brainstorm at least one possible lurking variable that could explain the correlation.
Reversing cause and effect	Even if A causes B, the correlation looks the same as B causing A. Direction cannot be read from r alone.	Ask whether a logical mechanism exists for A → B, and whether time order supports it.
Claiming causation from an observational study	Without random assignment, lurking variables remain uncontrolled, so causal claims are not supported.	Use language like "is associated with" or "is correlated with" instead of "causes."
Assuming no correlation means no relationship	r measures linear relationships only. A curved or complex pattern could exist with r ≈ 0.	Always look at a scatter plot before interpreting r. The shape of the data matters.

Five common reasoning errors about correlation and causation

✦

KEY TAKEAWAY

Think of it this way: wearing a winter coat is strongly correlated with it being cold outside, but wearing the coat doesn't cause the cold weather. The cold weather (a third factor) causes both the low temperature and your decision to wear the coat. Before claiming causation, always ask: "Is there something else that could explain both of these things?"

SECTION 8

Connection to Advanced Statistical Thinking

The distinction between correlation and causation that you're learning now is the foundation for more advanced statistical methods. As you progress in math and science, you'll encounter more sophisticated tools for analyzing relationships between variables.

What You Learn Now (Grade 9)	What Comes Later (AP & College)
Identifying correlations using scatter plots and r	Linear regression models that predict one variable from another
Recognizing lurking variables	Multivariate analysis that statistically controls for confounding variables
Distinguishing observational from experimental studies	Randomized controlled trials (RCTs), double-blind designs, and meta-analyses
Using careful language ("associated with" vs. "causes")	Hypothesis testing with p-values and confidence intervals to quantify evidence
Identifying spurious correlations	Causal inference frameworks (Bradford Hill criteria, counterfactual reasoning)

How today's concepts connect to future coursework

The good news is that the fundamental question never changes, no matter how advanced the statistics become: "Does the study design allow us to rule out other explanations?" If you master this question now, you'll have a significant head start when you encounter these advanced topics in AP Statistics, college courses, or even when reading scientific studies in the news.

SECTION 9

Practice Problems

Question 1: CONCEPTUAL

A news headline reads: "People who eat dark chocolate have lower rates of heart disease." Does this headline suggest correlation or causation? Explain your reasoning and identify what information you would need to decide whether the claim is causal.

Show answer

The headline suggests correlation, not causation. The phrase "people who eat" implies this is an observational finding—researchers observed existing eating habits rather than assigning people to eat chocolate. People who eat dark chocolate may also have other healthy habits (exercise, balanced diets, higher income allowing healthier food choices), which are lurking variables. To determine causation, you would need a controlled experiment in which participants are randomly assigned to eat dark chocolate or not, with all other factors held constant, and heart disease rates are compared after a significant period.

Question 2: BASIC CALCULATION

A study finds that the correlation between hours spent exercising per week and resting heart rate is r = −0.65. (a) Is this a positive or negative correlation? (b) Is the correlation weak, moderate, or strong? (c) What does this correlation tell you in plain language?

Show answer

(a) The correlation is negative because r is less than 0. (b) The magnitude |−0.65| = 0.65 is close to 0.7, so this is a moderate-to-strong correlation. (c) In plain language, as the number of hours of exercise per week increases, resting heart rate tends to decrease. People who exercise more generally have lower resting heart rates. However, this correlation alone does not prove that exercise causes the lower heart rate—though in this case, controlled experiments have indeed established that causal link.

Question 3: INTERMEDIATE

A researcher wants to test whether listening to classical music while studying improves test scores. She designs the following study: 100 students are randomly divided into two groups. Group A studies for one hour while listening to classical music. Group B studies for one hour in silence. Both groups take the same test the next day. Group A scores an average of 82, while Group B scores an average of 76. (a) Is this an observational study or a controlled experiment? (b) Can the researcher conclude that classical music caused the higher scores? Explain why or why not.

Show answer

(a) This is a controlled experiment because the researcher manipulated one variable (music vs. silence), randomly assigned students to the two groups, and kept other conditions (study time, test content) the same. (b) Yes, the researcher can reasonably conclude that classical music caused the higher scores. Random assignment means that lurking variables (prior knowledge, motivation, study habits) should be roughly evenly distributed between the two groups, so the only systematic difference is the music. The 6-point difference can therefore be attributed to the independent variable. However, it's worth noting that one experiment alone is not final proof—replication strengthens causal claims.

Question 4: APPLIED

A city government notices that neighborhoods with more fire stations tend to have more fires. A city council member proposes closing fire stations to reduce fires. Using your understanding of correlation vs. causation, explain the flaw in this reasoning and identify a likely lurking variable.

Show answer

The council member is making a classic correlation-implies-causation error. Fire stations do not cause fires. The lurking variable is population density (or neighborhood size). Larger, more densely populated neighborhoods naturally have more buildings and more people, which leads to more fires. The city responds by building more fire stations in those areas. So population density causes both the higher number of fires and the higher number of fire stations, creating a positive correlation between fire stations and fires. Closing fire stations would actually make the fire problem worse, not better. This example shows how dangerous it can be to act on correlation alone without understanding the underlying causal mechanism.

Question 5: CRITICAL THINKING

Consider these two studies about vitamin D and depression: Study 1: Researchers surveyed 5,000 adults and found that people with higher vitamin D levels reported fewer symptoms of depression (r = −0.45). Study 2: Researchers randomly assigned 200 adults with depression to two groups. One group received daily vitamin D supplements for 3 months; the other received a placebo. The supplement group showed a 20% greater reduction in depression symptoms. Compare the two studies. Which provides stronger evidence for causation? Could both studies be correct yet tell you different things? Explain.

Show answer

Study 2 provides stronger evidence for causation because it is a controlled experiment with random assignment and a placebo group. Study 1 is observational and can only show correlation. Both studies can absolutely be correct and complementary. Study 1 tells us that there is a moderate negative correlation between vitamin D levels and depression in the general population—this is a useful observation that motivates further research. Study 2 goes further by testing whether actively changing vitamin D levels (through supplements) causes a reduction in depression symptoms. Notice that Study 1 has a much larger sample (5,000 vs. 200), which gives it better generalizability, while Study 2 has a stronger design for causal inference. In real science, researchers often start with observational studies to identify promising correlations, then design experiments to test causation. Also, even Study 2's result could have limitations—maybe vitamin D only helps certain populations, or the effect diminishes over longer periods. Science relies on multiple studies of both types to build a complete picture.

SUMMARY

Lesson Summary

Summary

Correlation means that two variables tend to change together, while causation means that one variable directly produces a change in another. The correlation coefficient (r) measures the direction and strength of a linear relationship on a scale from −1 to 1, but even a very strong r value does not prove causation. Lurking variables—hidden factors that affect both variables—are the most common reason a correlation can mislead us into thinking one thing causes another.

To determine whether a causal claim is justified, identify the study type. Observational studies (surveys, data reviews) can reveal correlations but cannot prove causation because researchers do not control variables. Controlled experiments with random assignment can establish causation because they isolate the effect of one variable while keeping others constant. Always use careful language: say "is associated with" for correlations and reserve "causes" for results supported by experimental evidence.