All judgments are relative. “Good,” “bad,” “better,” “improved,” “created value” — none of these exist in isolation. They only exist compared to something else.

That something else is the baseline. And the baseline is usually invisible.

There is no absolute "good" — only "better than the alternative"

Whoever picks the alternative controls the conclusion


How This Idea Emerged

Starting point: Studying GDP and asking “why do we care if GDP goes up?”

The chain:

  1. GDP measures production → but more production isn’t automatically good
  2. Building houses raises GDP → but what if nobody needs them?
  3. Oil spill cleanup raises GDP → but we’re just fixing damage
  4. “At least we created jobs” → compared to what?

The unlock: The Broken Window Fallacy assumes workers would have had other jobs. But what if they wouldn’t? Then the baseline shifts — and so does the conclusion.

The generalization: This pattern appears everywhere. CPI base years, “improvement” metrics, policy evaluations. Every claim contains a hidden comparison.


The Core Principle

ComponentDescription
Claim”This is good” / “This worked” / “This created value”
Hidden structure”This is better than [baseline]“
The baselineThe alternative being compared against — often unstated
The leverChange the baseline → change whether the claim is true

Every evaluation is a comparison

The comparison has a baseline The baseline is often invisible Change the baseline, change the conclusion


The Broken Window Fallacy — A Case Study

Scenario: A kid breaks a shopkeeper’s window.

The claim: “At least it creates work for the glazier!”

FramingBaselineConclusion
”Glazier gets paid”vs. nothingGood — jobs created!
”Glazier gets paid”vs. shoemaker getting paidNeutral — just redirected spending
”Glazier gets paid”vs. new shoes + intact windowBad — net loss of value

What’s actually true?

Economic ConditionRelevant BaselineConclusion
Mass unemploymentWorkers would have been idleBreaking window = some stimulus
Full employmentWorkers would have been productive elsewhereBreaking window = waste

The fallacy isn't always a fallacy

It depends on which baseline matches reality

Full employment → fallacy holds (opportunity cost is real)

Mass unemployment → your objection holds (opportunity cost is zero)


The Pattern Across Domains

DomainWhat Has a BaselineHow Baseline Changes Conclusion
Broken Window”Created jobs” compared to what?vs. nothing → good; vs. other jobs → neutral
CPIPrice change since when?Pick 2019 vs. 2021 → different inflation narrative
GDP Deflator”Real” growth relative to what base year?Different base → different growth rate
Output GapY* (potential GDP)Different model → different gap → different policy
Performance Review”Improved” since when?Pick a bad quarter → everything looks good
Policy Evaluation”This policy worked”vs. doing nothing? vs. a better policy?
Investment Returns”Beat the market”Which market? Which time period?

GDP Is a Gauge, Not a Goal

This baseline problem is why GDP growth isn’t automatically good.

What GDP MeasuresWhat People Think It Means
Total market production”How well we’re doing”
Economic activityProgress
Stuff made and soldQuality of life

The hidden baseline: “More GDP is good” assumes the alternative was less GDP with everything else equal.

But what if:

ScenarioGDPActual Outcome
Car accident → hospital bills, repairsNobody better off
Oil spill → cleanup crewsJust fixing damage
Work 80 hrs instead of 40Burnout, lost family time
Cut down forest → sell lumberLost ecosystem, recreation

GDP counts activity, not whether activity made life better

“GDP went up” is meaningless without asking: compared to what alternative?

Goodhart’s Law connection:

“When a measure becomes a target, it ceases to be a good measure.”

StageWhat Happens
1GDP created as a gauge (useful)
2GDP becomes a goal (“grow GDP!“)
3Optimization begins (maximize the number)
4Gauge breaks (GDP rises, life doesn’t improve)

Detecting Hidden Baselines with the Seven Lenses

The Seven Lenses for Decomposing Claims are partly tools for surfacing hidden baselines:

LensThe QuestionHow It Exposes Baselines
Conditions”What if [condition] changed?”Tests whether conclusion holds under a different baseline
Trade-offs”What’s the cost? What’s sacrificed?”Names the foregone alternative — the baseline you’re NOT choosing
Scope”Why ‘all’? Why ‘always’?”Challenges whether one baseline works universally
Actors”Who decides? Who benefits?”Asks who CHOSE this baseline — and who benefits from that choice

Example — Broken Window through the lenses:

Claim: “Breaking windows creates jobs.”

LensQuestionWhat It Reveals
ConditionsWhat if full employment vs. recession?Baseline shifts → conclusion shifts
Trade-offsWhat job is NOT created?Hidden baseline: productive alternative
ScopeIs this always true?Baseline isn’t universal
ActorsWho benefits from this framing?Someone chose this baseline

The Defense

When evaluating any claim:

QuestionWhat It Does
”Compared to what?”Makes the baseline explicit
”Who chose this baseline?”Reveals potential agenda
”What if we used a different baseline?”Tests robustness
”Is this baseline disclosed?”Checks intellectual honesty

Common Traps

TrapExampleThe Problem
Floor thinking only”At least we got something”Ignores better alternatives
Ceiling thinking only”We could have done better”Ignores that worse was possible
Shifting baselinesDifferent baseline for praise vs. criticismInconsistent reasoning
Hidden baselines”This policy worked”Worked compared to what?
Baseline as given”CPI says inflation is 3%“Forgetting the base year is a choice

The Practical Takeaway

When someone presents a conclusion, find the baseline

  • If the baseline is hidden → ask for it
  • If the baseline seems wrong → propose a different one
  • If the baseline is cherry-picked → call it out
  • If you’re making a claim → disclose your baseline

North: Where this comes from

East: What opposes this?

South: Where this leads

West: What’s similar?