Causal Reasoning & Correlation vs. Causation

1. What You Need to Know

Causal reasoning questions test whether you can tell the difference between:

• A correlation: two things occur together.
• A causal claim: one thing produces (or prevents) the other.

On the LSAT, causal arguments are everywhere (Strengthen/Weaken, Flaw, Necessary Assumption, Sufficient Assumption, Evaluate, Resolve the Paradox). Your job is usually to spot that the author has moved from a correlation (or a before/after change) to a cause without ruling out other explanations.

Core rule (the entire game)

Correlation does not by itself establish causation.

To justify “A causes B,” you generally need to show (at least in reasoning terms):

• Temporal order: A happens before B.
• No better explanation: no alternate cause or common cause explains B.
• Not reversed: B isn’t causing A.
• A credible link: some mechanism or support that A could plausibly produce B.

When you use this

Use causal reasoning tools whenever you see language like:

• causes, leads to, results in, produces, increases, reduces, prevents, responsible for
• or the author infers a cause from:
  • a correlation (“people who do A have more B”)
  • a trend (“after policy A, outcome B fell”)
  • a group difference (“Group 1 has more B than Group 2, so difference is due to A”)

2. Step-by-Step Breakdown

A. How to attack any causal argument (fast and systematic)

1. Identify the conclusion: is it causal? (Look for causal verbs.)
2. Identify the evidence type:
   • Correlation (co-occurrence)
   • Before/after change
   • Comparison of groups
   • Anecdote / small sample
3. Translate into a causal diagram in your head:
   • Claimed: A \rightarrow B (or A \rightarrow \neg B)
4. Run the “3 classic gaps” test (this is most LSAT causal damage):
   • Reverse causation: B \rightarrow A
   • Common cause (confounder): C \rightarrow A and C \rightarrow B
   • Alternative cause: D \rightarrow B (something else causes B)
5. Check direction + timing:
   • Did A happen **before** B?
   • Is the measured relationship possibly backwards?
6. Check the “apples-to-apples” issue (group differences):
   • Are groups comparable, or is there selection bias?
7. Predict what would strengthen vs. weaken:
   • Strengthen = rules out alternatives / supports mechanism / confirms timing.
   • Weaken = introduces alternative explanation / shows reversal / attacks measurement.

B. Quick worked mini-example (annotated)

Stimulus: “Cities with more police officers have more crime. Therefore, hiring more police increases crime.”

• Conclusion: causal, \text{More police} \rightarrow \text{More crime}
• Evidence: correlation
• Gaps:
  • Reverse causation: \text{More crime} \rightarrow \text{More police} (very plausible)
  • Common cause: larger cities C create both more crime and more police
• Weaken: “Crime levels rose before police levels rose” (supports reversal) or “Large cities have both more crime and more police.”
• Strengthen: “When cities randomly increased police staffing, crime decreased” (hurts the original conclusion; strengthens the opposite). If strengthening the original conclusion specifically: “After controlling for city size and prior crime rate, more police predicts subsequent increases in crime.”

C. How to handle common causal question types

Strengthen

Pick answers that do one (or more) of:

• Rule out alternative causes of B
• Rule out common causes of A and B
• Rule out reverse causation
• Show timing (cause precedes effect)
• Provide mechanism, dose-response, or stronger data (controlled study)

Weaken

Pick answers that do one (or more) of:

• Provide a plausible alternative cause for B
• Provide a confounder C that explains both
• Show B happens before A (reverse causation)
• Attack the data: biased sample, measurement error, changing definitions

Necessary Assumption (causal)

The necessary assumption often asserts:

• No confounder / no alternative cause doing the work
• Causal direction (not reversed)
• The effect isn’t already happening for another reason

Use the negation test: if negating the answer destroys the causal link, it’s necessary.

Sufficient Assumption (causal)

Often gives you a strong “plug the gap” statement like:

• “No other relevant differences exist between the groups.”
• “The only change during the period was A.”
• “A precedes B and nothing else could explain B.”

Evaluate / Method of Reasoning

Good evaluation questions ask what would help decide between:

• A \rightarrow B vs. B \rightarrow A
• A \rightarrow B vs. C causes both

3. Key Formulas, Rules & Facts

A. Correlation vs. causation: what each allows

B. The “causal gap” suspects (memorize these)

C. Stronger vs. weaker evidence (in LSAT terms)

D. Causal language that hides bad reasoning

• “Since A, B.” (mere sequence)
• “After A, B decreased, so A caused the decrease.” (ignores other changes)
• “People who do A are healthier, so A makes you healthy.” (selection + confounding)

Warning: The LSAT loves “post hoc” reasoning: after this, therefore because of this.

4. Examples & Applications

Example 1: Trend / before-after (classic post hoc)

Stimulus: “After the city installed LED streetlights, nighttime crime fell. Therefore, LED streetlights reduce crime.”

• Key insight: Could be other simultaneous changes (more patrols, economic improvement, seasonal effect), or crime was already trending down.
• Strong strengthen: “Comparable neighborhoods without LED streetlights did not experience a similar decline during the same period.” (controls for time trend)
• Strong weaken: “During the same period, the city also increased police patrols at night.” (alternative cause)

Example 2: Group difference (selection bias)

Stimulus: “Employees who use the company gym take fewer sick days. So using the gym reduces illness.”

• Key insight: Healthier/more disciplined people might be the ones choosing the gym (common cause/selection).
• Strengthen: “After employees were randomly given free gym memberships, those who used the gym showed a subsequent decrease in sick days compared to those who did not.”
• Weaken: “Employees with chronic health conditions are less likely to use the gym.” (selection explains correlation)

Example 3: Reverse causation trap

Stimulus: “People who drink herbal tea report more stress. Thus, herbal tea increases stress.”

• Key insight: Stressed people may be more likely to drink herbal tea.
• Weaken: “Many people start drinking herbal tea specifically to cope with stress.” (reverse causation)
• Strengthen (for the conclusion): “In a study, participants who began drinking herbal tea showed increased stress levels compared to matched controls, even though they were not stressed beforehand.”

Example 4: “Explaining away” a surprising correlation (Resolve the Paradox)

Stimulus: “A safety course reduces accidents in simulations, but factories whose workers took the course have higher accident rates.”

• Key insight: The real-world factories might be different (riskier jobs) or the course is assigned when accidents are already high.
• Good resolve: “Factories with the most dangerous machinery were the ones that required the course.” (common cause/selection)

5. Common Mistakes & Traps

1. Mistake: Treating correlation as proof of causation

   • What goes wrong: You accept “A and B occur together” as “A causes B.”
   • Why wrong: Confounders, reversal, coincidence.
   • Fix: Automatically ask: reverse? common cause? alternative cause?

2. Mistake: Ignoring time order

   • What goes wrong: You don’t check whether the proposed cause precedes the effect.
   • Why wrong: Causes must come before effects (in LSAT-world reasoning).
   • Fix: Look for “before/after,” “prior to,” “subsequent,” “already trending.”

3. Mistake: Missing the “only cause” leap

   • What goes wrong: The argument subtly assumes A is the explanation.
   • Why wrong: Most real outcomes have multiple causes.
   • Fix: In necessary assumption/weakening, consider other contributors to B.

4. Mistake: Overvaluing a single study without checking design

   • What goes wrong: You treat any “study found” as causal.
   • Why wrong: Observational studies can’t easily rule out confounders.
   • Fix: Ask: random assignment? controlled variables? self-selection?

5. Mistake: Confusing “mechanism” with “proof”

   • What goes wrong: If an answer gives a plausible story, you think it confirms causation.
   • Why wrong: A mechanism helps, but doesn’t rule out other causes.
   • Fix: Prefer answers that eliminate alternatives or improve comparison.

6. Mistake: Falling for “numbers went up” without considering base rates/definitions

   • What goes wrong: You accept an increase/decrease as real.
   • Why wrong: Measurement methods, reporting incentives, or definitions may have changed.
   • Fix: Check whether the way B is measured stayed constant.

7. Mistake: Not matching the answer choice to the task (strengthen vs. weaken)

   • What goes wrong: You pick a true-sounding critique on a strengthen question.
   • Why wrong: LSAT rewards task alignment, not general skepticism.
   • Fix: Predict the gap first; then pick the choice that addresses it in the correct direction.

8. Mistake: Forgetting that “lack of evidence” can weaken

   • What goes wrong: You ignore answers showing the effect occurs without the cause (or vice versa).
   • Why wrong: If B happens when A doesn’t, A looks less causal.
   • Fix: Use the “cause without effect” / “effect without cause” checks.

6. Memory Aids & Quick Tricks

Quick trick: If an answer choice says “some other factor changed,” it often weakens a before/after causal claim. If it says “nothing else relevant changed” or “a control group didn’t show the change,” it often strengthens.

7. Quick Review Checklist

• Identify whether the conclusion is causal.
• If the evidence is correlation/trend/group difference, assume a gap exists.
• Run ARC:
  • Alternative cause for B?
  • Reverse causation B \rightarrow A?
  • Confounder C causing both?
• Check timing: did A happen before B?
• Check selection bias: are the groups comparable, or self-selected?
• For strengthen: look for controls, randomization, mechanism, dose-response, or ruling out alternatives.
• For weaken: look for other changes, confounders, reverse direction, effect without cause, cause without effect, or measurement problems.
• For necessary assumptions: negate-test statements that prevent confounding/reversal.

You’ve got this: treat every causal claim like a suspect and make it earn the conclusion.