The Shocking “Discovery”
The Data
- Correlation: 99.26% correlation between Maine divorce rate and US margarine consumption (2000-2009)
- Source: Tyler Vigen’s “Spurious Correlations” website
- Pattern: Both declined steadily over the decade
Year-by-Year Breakdown
| Year | Maine Divorce Rate | Margarine Consumption |
|---|---|---|
| 2000 | 5.0 per 1,000 | 8.2 lbs per person |
| 2001 | 4.7 per 1,000 | 7.0 lbs per person |
| 2002 | 4.6 per 1,000 | 6.5 lbs per person |
| 2003 | 4.4 per 1,000 | 5.3 lbs per person |
| 2004 | 4.3 per 1,000 | 5.2 lbs per person |
| 2005 | 4.1 per 1,000 | 4.0 lbs per person |
| 2006 | 4.2 per 1,000 | 4.6 lbs per person |
| 2007 | 4.2 per 1,000 | 4.4 lbs per person |
| 2008 | 4.2 per 1,000 | 4.5 lbs per person |
| 2009 | 4.1 per 1,000 | 4.2 lbs per person |
The Absurd “Theories”
If We Believed the Correlation:
- 🧈 “Margarine makes marriages unstable!”
- 💔 “Divorce stress drives people to butter!”
- 🏠 “Maine residents are particularly sensitive to margarine!”
- 📊 “We should monitor margarine sales to predict divorce rates!”
What’s Really Happening
Actual Explanations:
Maine Divorce Rate Decline:
- Changing social attitudes
- Economic factors
- Demographic shifts
- Legal/procedural changes
- Cohabitation trends
Margarine Consumption Decline:
- Health consciousness rising
- Trans fat awareness
- Return to “natural” foods
- Butter rehabilitation
- Cooking trend changes
Why This Correlation Exists
Pure Mathematical Coincidence:
- Both variables happened to decline
- Similar time periods
- No causal mechanism
- Random statistical alignment
- Cherry-picked comparison
The Broader Lesson
This Example Teaches Us:
1. Data Mining Dangers
- With enough variables, you’ll find correlations
- Computers can find millions of meaningless patterns
- Statistical software doesn’t understand causation
2. The Correlation Trap
- High correlation ≠ meaningful relationship
- 99.26% correlation can be 100% meaningless
- Numbers alone don’t tell the story
3. Common Sense Check
- Ask: “Does this make logical sense?”
- Consider: “What’s the proposed mechanism?”
- Question: “Could this be coincidence?”
Other Ridiculous Correlations from the Same Dataset
| Variables | Correlation | Obviously Spurious Because |
|---|---|---|
| Pool drownings vs Nicolas Cage films | 66.6% | Actors don’t control water safety |
| Cheese consumption vs bed sheet deaths | 94.7% | Dairy products don’t affect bedding |
| Honey production vs juvenile arrests | 70.3% | Bees don’t influence crime rates |
| Math doctorates vs uranium stored | 99.5% | Academic degrees don’t affect nuclear storage |
Red Flags for Spurious Correlations
Warning Signs:
- ✋ No logical connection between variables
- ✋ Cherry-picked time periods
- ✋ Unusual geographic specificity (why Maine?)
- ✋ No proposed causal mechanism
- ✋ Correlation hunting through massive datasets
Questions to Ask:
- 🤔 “Why would these be related?”
- 🤔 “What’s the proposed cause-and-effect?”
- 🤔 “Could this be random chance?”
- 🤔 “Are there confounding variables?”
- 🤔 “Does this replicate in other contexts?”
The Statistical Reality
Why Spurious Correlations Happen:
- Multiple Comparisons: Test enough pairs, some will correlate by chance
- Data Dredging: Mining datasets for interesting patterns
- Confirmation Bias: Remembering hits, forgetting misses
- Publication Bias: Weird correlations get attention
Practical Applications
In Business:
- Don’t base decisions on correlation alone
- Demand causal explanations
- Test theories with experiments
- Consider alternative explanations
In Research:
- Replicate findings
- Control for confounding variables
- Use theory to guide analysis
- Be skeptical of surprising correlations
In Media Consumption:
- Question sensational statistical claims
- Look for logical explanations
- Consider the source’s methodology
- Remember: extraordinary claims need extraordinary evidence
The Bottom Line
The margarine-divorce correlation is a perfect example of why:
- 📊 Statistics require interpretation
- 🧠 Common sense matters
- 🔍 Correlation hunting is dangerous
- ⚠️ High correlation can be meaningless
Remember: Just because two things move together doesn’t mean one causes the other – sometimes they’re just dancing to completely different tunes that happen to have the same rhythm! 🎵
