Baca buku ini gara-gara penasaran aja. Judulnya unik, cover-nya menarik. Ga expect apa-apa.
Ternyata changed how I see things.
Everything is about incentives
Premis utama super simple: "Incentives are the cornerstone of modern life."
Manusia bereaksi terhadap insentif. Always.
Tiga tipe:
- Economic - duit, harta
- Social - reputasi, status
- Moral - nilai, prinsip
Yang menarik: kadang ketiganya interact dalam cara yang counterintuitive.
Daycare late fee story
Ada daycare di Israel. Masalah: parents sering telat jemput.
Solution: kasih fine.
Result: lebih banyak yang telat.
Wait, what?
Ternyata sebelum ada fine, parents feel moral obligation - merasa ga enak bikin teacher nunggu.
Pas ada fine, moral obligation hilang. Jadi transaksional: "Gue bayar, berarti boleh telat."
Economic incentive accidentally eliminated moral incentive.
Ini yang bikin incentive design tricky.
Dan gue liat ini juga di tech companies. Ada yang kasih bonus based on lines of code. Result: codebase penuh unnecessary code.
Wrong incentive = wrong behavior.
Teachers and sumo wrestlers both cheat
Chapter favorit. Pertanyaannya udah aneh: apa kesamaan guru dan pegulat sumo?
Jawabannya: keduanya nyontek.
Teachers:
Di Chicago, gaji guru tied ke nilai ujian siswa. Sounds like good incentive.
Tapi Levitt analyze pattern dan nemu something fishy:
- Siswa yang biasa jelek suddenly benar di soal sulit
- Pattern terlalu "clean"
- Consistent across multiple students in same class
Conclusion: teachers ngubah jawaban setelah ujian.
Ditest dengan monitoring ketat - cheating rate drop.
Sumo wrestlers:
Wrestling record 7-7 (butuh 1 win lagi untuk maintain rank) punya suspiciously high win rate di match terakhir.
Way higher dari statistical probability.
Match fixing. Wrestler yang udah aman "ngalah" ke yang desperate, dengan implicit deal: "Next tournament lo kasih balik."
The point:
When stakes high and opportunity exists, people respond to incentives - even if means bending rules.
Relate sama ini. Pernah liat dev:
- Inflate story points biar velocity bagus
- Close tickets tanpa actually fixing
- Write tests yang pass tapi ga test anything meaningful
We're human. We respond to incentives.
Question: are we designing right incentives?
Why drug dealers live with their moms
This one shattered stereotypes.
Media portrayal: fancy cars, bling, mansions.
Reality: average street-level dealer makes less than minimum wage.
Levitt got actual financial records dari Chicago drug gang.
Data:
- Foot soldiers: $3.30/hour average
- Risk of being killed: 1 in 4 per year
- Most live with moms, can't afford rent
Why do it?
Tournament economics.
Like pro sports. Ribuan di minor leagues with small pay, hoping to make big leagues.
Drug gangs sama. Top leaders make huge money. Tapi very few spots.
Foot soldiers bertahan karena:
- Hope of climbing (tournament incentive)
- Status (social incentive)
- Limited alternatives (economic constraint)
Same structure as corporate pyramid, startup ecosystems.
Startup parallel
Surprisingly similar.
Berapa banyak startup founders:
- Work 80+ hours/week
- Below market rate gaji (atau no gaji)
- High stress, 90% fail rate
Why? Tournament economics. Mayoritas fail, tapi yang succeed jadi decacorns.
Gojek, Tokopedia success stories create incentive untuk thousands take the shot, even though odds against them.
Understanding tournament economics explains so much about competitive environments.
As tech worker, I see this - kenapa orang mau kerja di early-stage startups dengan high uncertainty? Equity upside. Tournament economics.
The abortion-crime link
Chapter paling controversial. Made me uncomfortable. But that's why it's important.
Question: Kenapa crime rate di US drop dramatically tahun 1990-an?
Standard answers: better policing, stronger economy, tougher sentencing.
Levitt's data: none of these explain magnitude.
His hypothesis: Roe vs Wade (1973) legalized abortion → fewer unwanted children → fewer potential criminals 18 years later.
Logic: unwanted children more likely grow up in unstable environments, poverty, neglect. More likely engage in crime as teens.
Legal abortion gave women choice. Women least ready could wait. Fewer children born into high-risk situations.
18 years later, crime drops.
Why controversial:
Ethical implications massive. Correlation vs causation. Not making moral judgment, just analyzing data patterns.
But can't separate from moral and political implications.
Why this matters to me:
Because uncomfortable truths matter.
Data might reveal patterns that challenge assumptions. Doesn't mean ignore data - means think more carefully.
As engineer, I encounter uncomfortable data constantly:
- Feature we spent months on? Users hate it
- "Brilliant" redesign? Decreased engagement
- Assumption about user behavior? Completely wrong
Don't ignore because uncomfortable. Adapt.
Note: Levitt's methodology heavily criticized and debated. Not settled science. But conversation valuable.
Perfect parenting
Question: what actually matters for child development?
Parents stress about:
- Enrichment classes?
- Screen time?
- Name effects?
- "Good schools"?
Levitt analyzed data: most things parents worry about don't matter that much.
What doesn't matter (as much as you think):
- Museum trips
- Daily reading (though reading environment matters)
- Strict TV rules
- Spanking vs not
What matters:
- Parents' education level
- Socioeconomic status
- Parents' involvement quality
- Books in home (indicator of education culture)
Key: correlation vs causation.
Museum trips don't make kids smarter. But parents who value education both take kids to museums AND raise them in stimulating environments. The latter matters.
Names and future
Famous case: two brothers named Winner and Loser.
Winner became criminal. Loser became detective.
Names are signals of socioeconomic background. Not causa success/failure. Context and upbringing do.
Lesson:
Stop sweating small stuff. Focus on what matters: environment, culture, genuine engagement.
In tech world, sama - people obsess over tools (Vim vs VSCode, tabs vs spaces) when what matters: solving real problems and learning effectively.
The Freakonomics mindset
Not just random facts. About developing specific way of thinking.
Question conventional wisdom
"Everyone knows that..." = red flag.
Examples:
- "Everyone knows crack caused crime spike" (data says otherwise)
- "Everyone knows good parenting requires X, Y, Z" (most don't matter)
In tech: "Everyone knows users want more features" (often, they want simplicity)
Correlation ≠ causation (but can point to it)
Stats 101, but applying in real life hard.
Ice cream sales correlate with drowning deaths. Ice cream cause drowning? No. Both correlate with summer (confounding variable).
But sometimes correlation does suggest causation - need to dig deeper, control variables, test hypotheses.
Think incentives, not just actions
Don't just observe what people do. Ask: what are incentives?
Why car salesman pushing this model? (Higher commission)
Why doctor recommending this procedure? (Necessary or profitable?)
Why team advocating this feature? (Valuable or easy?)
Data over anecdotes
Personal experience valuable. But data reveals broader patterns.
"I know someone who smoked and lived to 90" ≠ "smoking is safe"
One success story doesn't validate strategy.
Second-order effects matter
Every action has immediate effect AND ripple effects.
Daycare fine: first-order = economic penalty. Second-order = eliminates moral obligation.
Need to think about second-order effects in product decisions.
Example: gamification might increase engagement (first-order) but decrease quality (second-order).
Application to engineering work
Product decisions: understand user incentives
Before building feature, ask: what are users' actual incentives?
Personal experience:
Built social sharing feature assuming users want to share achievements. Low adoption.
Why? Wrong incentive assumption.
Actual: users don't want to look like bragging. Social disincentive > desire to share.
Solution: reframe as "help friends discover" vs "show off". Adoption increased.
Lesson: understand psychological and social incentives, not just functional needs.
Metrics: vanity vs actionable
Look beyond surface-level data.
Bad: total signups
Better: active users, retention, engagement depth
Bad: feature usage count
Better: usage by cohort, correlation with retention
Ask: does this metric actually inform decisions?
Incentive design in teams
How incentivize good engineering?
Wrong: lines of code → bloat
Wrong: tickets closed → quick fixes without quality
Wrong: features shipped → feature bloat
Better: measure outcomes (user satisfaction, reliability, maintainability)
But remember daycare - explicit incentives might crowd out intrinsic motivation.
Best engineers motivated by:
- Craft pride (moral)
- Peer respect (social)
- Growth and learning (personal)
Don't destroy these with wrong explicit incentives.
Limitations
Love this book, but limitations exist.
Oversimplification
Humans aren't purely rational. We're emotional, irrational, influenced by biases, social creatures with complex motivations.
Economic models can't capture everything.
Correlation-causation issues
Despite Levitt's skill, some claims controversial. Causation hard to prove definitively.
Missing the "why"
Great at identifying incentives, but sometimes doesn't dig deep into why they exist or how shaped by culture, history, power structures.
Ethical blind spots
Not everything should be reduced to incentives and rational analysis. Some things have moral dimensions that can't be captured by data.
What changed
After reading, I started seeing incentives everywhere:
- Why gym has annual contracts? (Lock-in)
- Why airlines complicate boarding? (Extract revenue from priority)
- Why companies offer free snacks? (Keep people in office longer)
- Why apps make it hard to unsubscribe? (Maximize passive subscribers)
Once you see incentives, can't unsee.
In tech, incredibly valuable for:
- Designing products (what incentives creating?)
- Building teams (what incentives for engineers?)
- Making decisions (what incentives responding to?)
Understanding incentives = understanding behavior = better decisions.
Bottom line
This book didn't teach me economics. It taught me how to think.
Every product decision, A/B test, feature prioritization - all about understanding incentives, analyzing data, predicting behavior.
The core lesson: don't assume. Test. Look at data. Question conventional wisdom.
Simple framework. Powerful results.
I apply this beyond work:
- Learning new tech - build small project first to validate interest before commit to course
- Career moves - experiment with side gig before full jump
- New habits - test with smallest commitment before scale
And now can't stop seeing patterns everywhere.
That's probably a good thing.