Causal Reasoning & Correlation vs. Causation

1. What You Need to Know

Causal reasoning is everywhere on the LSAT: arguments conclude that one thing causes another (or prevents it), often based on correlations, trends, or before/after comparisons. Your job is to spot when the evidence only supports association and what would be needed to justify causation.

Core idea

Correlation/association: two things vary together (they co-occur, rise/fall together, etc.).
Causation: one thing produces or brings about the other.

A correlation can be consistent with causation, but it does not by itself prove causation.

The LSAT’s “causal gap”

Most causal arguments jump from:

“ $A$ and $B$ are correlated”
or “After $A$ happened, $B$ happened”

to:

“Therefore, $A$ caused $B$ .”

The test loves the alternative explanations that keep correlation true while making causation false.

When you’ll use this

Causal reasoning shows up in:

Strengthen/Weaken (most common)
Assumption (necessary/sufficient)
Flaw (correlation vs causation, reverse causation, confounding)
Resolve/Explain (two facts that seem inconsistent)
Method of Reasoning (identifying causal structure)

2. Step-by-Step Breakdown

Use this procedure any time you see causal language: cause, leads to, results in, due to, because of, responsible for, makes, produces, prevents, reduces, increases, contributes to.

Step 1: Translate into a simple causal claim

Write a clean version:

Conclusion: “ $A$ causes $B$ ” or “ $A$ prevents $B$ .”
Evidence: usually a correlation, trend, or comparison.

Example translation:

Evidence: “Cities with more police have more crime.”
Conclusion: “More police cause more crime.”

Step 2: Identify what type of evidence it is

Most common evidence types:

Correlation: “When $A$ , $B$ .”
Before/after: “After $A$ started, $B$ changed.”
Group comparison: “Group with $A$ has more $B$ than group without $A$ .”

Each type has standard vulnerabilities (below).

Step 3: List the 4 classic alternatives (the “causal suspects”)

Given correlation between $A$ and $B$ , any of these could explain it:

$A \rightarrow B$ (the claimed direction)
$B \rightarrow A$ (reverse causation)
$C \rightarrow A$ and $C \rightarrow B$ (common cause / confounder)
Coincidence / bias / bad data (selection bias, measurement issues, cherry-picking)

Step 4: For Strengthen vs Weaken, know what “wins”

Strengthen: make $A \rightarrow B$ more likely

Good strengthen moves:

Rule out alternative causes ( $C$ )
Show temporal order: $A$ happens before $B$
Show dose-response: more $A$ , more $B$
Show mechanism (how $A$ produces $B$ )
Show cause without other changes (everything else held constant)
Show effect disappears when cause removed

Weaken: make $A \rightarrow B$ less likely

Good weaken moves:

Provide an alternative cause $C$
Show reverse causation: $B$ starts first
Show no co-variation when controlled
Show cause occurs without effect (counterexample)
Show effect occurs without cause
Attack the data (biased sample, bad measurement)

Step 5: For Assumption questions, find what must be true

Causal conclusions typically require an assumption like:

“No other factor caused $B$ .” (or at least: no other factor explains the change)
“ $B$ didn’t cause $A$ .”
“The correlation isn’t driven by selection/measurement bias.”

Necessary assumption test: negate the choice; if the argument collapses, it was required.

Step 6: For Resolve/Explain, don’t “prove causation,” reconcile facts

You’re usually given:

A causal generalization (“ $A$ reduces $B$ ”)
But a surprising outcome (“Yet in this case $B$ didn’t drop”)

Resolution often introduces:

A new factor blocking the causal link
A different definition/measurement
A different population (external validity)

3. Key Formulas, Rules & Facts

Causal structures you should recognize

Pattern in stimulus	What it suggests	LSAT-friendly attack/defense
“ $A$ and $B$ are correlated”	Association only	Alternative cause, reverse causation, bias
“After $A$ , $B$ changed”	Temporal sequence, not proof	Another change occurred at same time; trend already ongoing
“Where $A$ is higher, $B$ is higher”	Cross-sectional correlation	Confounders differ between groups
“Implement $A$ , $B$ decreases”	Possible causation	Regression to mean, other interventions, measurement change
“Only factor that changed was $A$ ”	Stronger causal case	Check if true; hidden changes, selection effects

Correlation vs causation: what’s valid

Claim	Valid?	Notes
“ $A$ correlates with $B$ , so $A$ causes $B$ ”	No	Classic flaw
“ $A$ causes $B$ , so $A$ correlates with $B$ ”	Not guaranteed	Causal effects can be masked by other forces or poor measurement
“ $A$ causes $B$ , so changing $A$ will change $B$ ”	Usually intended, but still testable	Requires no blockers and correct population/conditions

Classic causal flaws (Flaw questions love these)

Flaw	What it is	What it looks like
Confounding / common cause	$C$ causes both $A$ and $B$	“Coffee drinkers have more heart disease; coffee causes it” (maybe smoking)
Reverse causation	$B$ causes $A$	“People who exercise are happier; exercise causes happiness” (maybe happy people exercise)
Post hoc	“After this, therefore because of this”	“After the mayor took office, crime rose; mayor caused crime”
Selection bias	Sample differs in relevant way	“Treatment group improved” but they volunteered / were more motivated
Measurement / definition shift	Outcome not measured consistently	New test makes scores ‘rise’
Overgeneralization	Causal claim too broad	Works in one city, claimed for all cities
Causal oversimplification	Treats multi-cause effect as single-cause	“Traffic is caused by taxis”

What strengthens a causal claim (high yield)

Strengthener	When to use	Why it helps
Rule out alternative causes	Anytime there could be $C$	Narrows explanation to $A$
Show $A$ precedes $B$	Reverse causation possible	Supports direction $A \rightarrow B$
Dose-response (more $A$ , more $B$ )	Correlation/trend arguments	Harder to explain by coincidence
Mechanism	Evidence is purely statistical	Makes causal story coherent
Controlled comparison (hold others constant)	Group comparisons	Reduces confounding
Remove $A$ and $B$ disappears	Policy/treatment scenarios	Suggests dependency

What weakens a causal claim (high yield)

Weakener	When to use	Why it helps
Alternative cause $C$	Correlation-based arguments	Explains $B$ without $A$
Reverse causation	Direction unclear	Flips arrow to $B \rightarrow A$
Cause without effect	Claimed cause present but effect absent	Undermines “ $A$ is sufficient” vibe
Effect without cause	Effect occurs without $A$	Undermines “ $A$ is necessary” vibe
Data problems (biased sample, mismeasurement)	Surveys, studies, polls	Makes evidence unreliable

Causal language to treat as strong conclusions

If the conclusion says “the reason,” “responsible for,” “results from,” “will lead to,” “will reduce,” assume it’s making a causal commitment and look for the gap.

4. Examples & Applications

Example 1 (Flaw): Correlation $\neq$ causation

Stimulus: “Neighborhoods with more streetlights have higher crime rates. Therefore, streetlights increase crime.”

Key insight: Correlation doesn’t establish direction or rule out confounders.

Reverse causation: high crime areas install more lights.
Confounder: dense commercial areas have more lights and more crime.

Typical correct flaw description: treats correlation as proof of causation.

Example 2 (Strengthen): Ruling out alternative causes

Stimulus: “After the company introduced remote work, productivity increased. So remote work caused the increase.”

Best strengthen: something like “No other major policy changes occurred during that period, and teams that did not adopt remote work did not improve.”

Why it works:

Controls for other changes
Adds a comparison group
Makes $A$ (remote work) the key differentiator

Example 3 (Weaken): Alternative cause + timing

Stimulus: “People who take supplement X get fewer colds. So supplement X prevents colds.”

Strong weaken: “People start taking supplement X when they begin exercising regularly and improving sleep habits.”

Why it works: introduces $C$ (health behaviors) causing fewer colds and correlated with taking X.

Example 4 (Resolve): Blocking factor / different conditions

Stimulus:

General: “Using salt on icy roads prevents accidents.”
Surprise: “This winter, the city used more salt than ever, yet accidents increased.”

Good resolution: “This winter had far more days of freezing rain, creating ice conditions in which salt is less effective, and overall driving volume increased.”

Why it works: salt can still help per unit of ice, but stronger opposing factors overwhelmed the effect.

5. Common Mistakes & Traps

Treating correlation as enough: You see “associated with” and accept “causes.”
- Why wrong: correlation is consistent with multiple causal stories.
- Fix: force yourself to ask: reverse causation? confounder? bias?
Forgetting reverse causation: You only look for third factors.
- Why wrong: many LSAT correlations are best attacked by flipping direction.
- Fix: always ask “Could $B$ cause $A$ ?” especially with behavior/choices (exercise, buying, studying, calling the doctor).
Ignoring selection bias: You assume groups are comparable.
- Why wrong: “People who chose X” often differ from “people who didn’t.”
- Fix: check for self-selection, non-random assignment, attrition.
Confusing “cause” with “sufficient condition”: You treat “ $A$ causes $B$ ” as “If $A$ , then always $B$ .”
- Why wrong: causes can be probabilistic, partial, or blocked.
- Fix: weakeners like “ $A$ happened but $B$ didn’t” help, but only if the argument implied $A$ was enough under the relevant conditions.
Overvaluing temporal order: You think “ $A$ happened first” proves causation.
- Why wrong: many other things can happen between $A$ and $B$ .
- Fix: time order helps, but you still need to rule out other changes.
Missing “third variable” phrasing: The stimulus may hide $C$ as “economic conditions,” “demographics,” “infrastructure,” “health awareness.”
- Why wrong: those are classic confounders.
- Fix: in group comparisons, ask what else differs systematically between groups.
Picking weakeners/strengtheners that are irrelevant: You choose an answer that sounds causal but doesn’t touch the gap.
- Why wrong: LSAT wrong answers often discuss a different variable, different time frame, or different population.
- Fix: keep the causal claim in one line (“ $A \rightarrow B$ ”) and ask whether the choice changes its likelihood.
In Resolve questions, trying to ‘disprove’ one fact: You attack the data instead of reconciling.
- Why wrong: correct resolution usually lets both statements remain true.
- Fix: look for a new factor, different conditions, or a threshold effect.

6. Memory Aids & Quick Tricks

Trick / mnemonic	Helps you remember	When to use
ARC = Alternative cause, Reverse causation, Coincidence/bias	The 3 biggest correlation-to-causation threats	Any correlation-based argument
ACE = Alternative cause, Cause without effect, Effect without cause	3 top weaken moves	Weaken causal conclusions
TIME = Temporal order, Isolate variables, Mechanism, Eliminate alternatives	4 top strengthen moves	Strengthen causal conclusions
“Compare like with like”	Group comparisons need comparable groups	Studies, surveys, policy comparisons
“What changed besides $A$ ?”	Before/after arguments often ignore other changes	Trend and intervention stimuli

Quick rule: If the stimulus is a study, your first thought should be selection bias and confounders.

7. Quick Review Checklist

You can restate the argument as “ $A \rightarrow B$ ” (or prevents).
You automatically test reverse causation: “Could $B \rightarrow A$ ?”
You look for a common cause $C$ explaining both.
You check for selection bias (who got into which group) and measurement issues.
Strengthen = rule out alternatives, show time order, add mechanism, add controlled comparisons, show dose-response.
Weaken = alternative cause, reverse causation, cause without effect, effect without cause, or bad data.
Resolve = keep both facts true; add a blocking factor, different conditions, or changed definition.

You’re aiming to be the person who always asks: “What else could explain this correlation?” and answers it fast.