Endogenous vs Exogenous Variables

Why the Distinction Exists

A model is a system of equations. Equations determine values—they take inputs and produce outputs.

The causal chain:

You have equations (structure)
THEREFORE some variables must be solved for (necessity)
THEREFORE those variables are outputs of the system (role)
THEREFORE we call them endogenous (label)

The label “endogenous” is the last step, not the first. It describes what the math already requires.

The same logic in reverse:

Some variables have no equation determining them
THEREFORE you must provide their values before solving (necessity)
THEREFORE those variables are inputs to the system (role)
THEREFORE we call them exogenous (label)

The Core Distinction

Classification	Meaning	What you do with it
Endogenous	Determined inside the model	Solve for it
Exogenous	Given from outside the model	Plug it in

This table is a summary. The mechanism above is the understanding.

The Causation Principle

Causation flows FROM whatever is fixed (exogenous) TO whatever must adjust (endogenous).

This is the same distinction restated as a directional claim. If you know what’s exogenous, you know where causation originates. If you know what’s endogenous, you know what responds.

The “Influenced By” Test

Equations influence variables by determining their values. THEREFORE:

If an equation in the model determines this variable → it’s influenced → endogenous
If no equation in the model determines this variable → it’s not influenced → exogenous

The classification follows from equation structure, not from labels applied first.

Limitation: This assumes binary classification. In reality, influence can be partial—a large firm influences market price, a small firm doesn’t. The binary is a modeling simplification.

Why the Distinction Matters: Comparative Statics

The classification exists to enable prediction:

You have a question: “What happens if X changes?”
THEREFORE X must be something you can hypothetically vary → exogenous (the dial)
The equations take the new X value and produce new outputs → endogenous (what you observe)
THEREFORE you can compare old equilibrium to new equilibrium
WHICH IS comparative statics—the core predictive method in economics

The workflow:

Step	Action
1	Set exogenous values (turn the dial)
2	Run equations (model processes)
3	Observe endogenous outputs (read the results)
4	Compare before/after (comparative statics)

Without this classification, you wouldn’t know what you can vary vs. what the model determines.

The Framing Principle

Classification flows from the question you’re asking—just like relevance flows from framing a decision.

“What is this model trying to explain?”

The answer to that question determines everything:

Variables the model must solve for to answer that question → Endogenous
Variables the model takes as given → Exogenous

This mirrors The Decision Lifecycle Stage 1 (ROUGH FRAME): you can’t classify until you’ve identified what you’re trying to figure out.

Systems of Equations: The Foundation

Economic models are systems of equations. The economic vocabulary is just a layer on top of the math.

What is a system of equations?

Two or more equations that share variables and must all be true at the same time.

Simplest example:

$x + y = 10$
$x - y = 4$

Two equations. Two unknowns (x and y). Both equations must hold simultaneously.

Solving means: Finding the values of x and y that make both equations true at once.

Method (substitution):

From equation 2: $x = y + 4$

Plug into equation 1: $(y + 4) + y = 10$

Solve: $2 y = 6$ → $y = 3$

Then: $x = 3 + 4 = 7$

Check: Does $7 + 3 = 10$ ? Yes. Does $7 - 3 = 4$ ? Yes. Both hold.

The core rules:

Rule	What it means
n equations can solve for n unknowns	2 equations → 2 unknowns. 3 equations → 3 unknowns.
Equations must be connected	They must share at least one variable, or they’re separate problems
Everything else is given	If you have 3 variables but only 2 equations, one variable must be handed to you as a number

Example with a “given”:

$y = 2 x + b$
$y = 5 x - 1$

Three symbols (x, y, b). Two equations. You can only solve for two unknowns.

The causal chain:

Two equations exist, THEREFORE two variables can be solved for
Three variables exist, THEREFORE one must be given
b has no equation determining it, THEREFORE b is exogenous
x and y are solved by the system, THEREFORE x and y are endogenous

If I tell you $b = 5$ :

You plug it in: $y = 2 x + 5$
Now solve the system: $2 x + 5 = 5 x - 1$ → $6 = 3 x$ → $x = 2$
Then: $y = 2 (2) + 5 = 9$

Without being given b, you’re stuck. The system cannot run without its inputs.

The connection to economics:

Math	Economics	Role
Variables you solve for	Endogenous	Unknowns the system determines
Parameters given to you	Exogenous	Inputs you plug in before solving
System of equations	Model	The relationships that define the problem
Solution	Equilibrium	The values that satisfy all equations simultaneously

Context-Dependence: Classification Is a Modeling Choice

The same variable can be endogenous or exogenous because:

Classification depends on equation structure
Different models have different equations
THEREFORE the same variable gets different classifications

Model	Equations for Price?	Price Classification
Demand only	No	Exogenous
Supply-demand equilibrium	Yes (equilibrium condition)	Endogenous
Price-setting firm	Yes (profit maximization)	Endogenous

The variable didn’t change. The model did.

Trade-off (why not always use the biggest model?):

Making more variables endogenous requires more equations, WHICH requires more assumptions, WHICH increases complexity, WHICH may exceed available data or tractability. Simpler models sacrifice realism for clarity.

Scope check:

There is no “true” classification outside a model. Reality doesn’t label variables as endogenous or exogenous—models do. Asking “is price really endogenous?” without specifying a model is a malformed question.

Direction of Causation: The Textbook’s Claim and Its Limits

The textbook states that the direction of causation assumption is:

Price → Quantity demanded (not the reverse)

This is true when analyzing individual responses to observed prices. A consumer sees a price tag and decides how much to buy. They don’t set the price—they respond to it.

But this framing quietly assumes the actor treats price as given—that is, price is exogenous to the individual’s decision. In wholesale and auction markets with fixed supply, price isn’t given—it’s discovered by the quantity hitting demand.

The more complete frame:

Context	What’s Fixed (Exogenous)	What Adjusts (Endogenous)	Direction of Causation
Individual response	Price (taken as given)	Quantity demanded	P → Qd
Market equilibrium (elastic supply)	Demand & supply curves	Price and quantity simultaneously	Neither → Both determined together
Market clearing (inelastic supply)	Quantity supplied	Price	Qs → P

The principle holds: causation runs FROM whatever is fixed TO whatever must adjust.

The Fish Market Example (Inelastic Supply)

The boats arrive at 4am with 500 kilos of fish. That’s it. The fish are perishable—they will sell today. The fishermen can’t hold out for a better price tomorrow.

This is perfectly inelastic supply: the supply curve is vertical.

Price
  │
  │     S (vertical—quantity fixed)
  │     │
  │     │╲ D
  │     │ ╲
  │     │  ╲
  │─────┼───╲────
  │     │    ╲
  └─────┴─────────
       500    Quantity

No matter what price emerges, 500 kilos will sell. The only question is at what price.

What Determines Price Here?

The fixed quantity (500 kilos) intersects the demand curve at exactly one point. That intersection is the price.

Causal chain:

Catch size (exogenous) → Quantity supplied (fixed at 500)
500 kilos hits the demand curve → Price discovered at intersection

If only 300 kilos arrived, the vertical supply line shifts left, hits the demand curve higher, and price rises.

Q genuinely causes P here—not as a response function, but as market-clearing mechanics with fixed supply. The exogenous variable (quantity) determines the endogenous variable (price).

Temporal Sequence vs. Causal Priority

These are different concepts that often get conflated.

Temporal Sequence = What Happens First in Time

At the fish market:

Boats arrive with their catch (quantity exists)
Buyers gather and observe what’s available
Bidding/negotiation happens
Price emerges from the process

Temporally: Q arrives → P discovered

Causal Priority = Which Variable Is the “Input” in the Response Function

During bidding, each buyer asks: “At this price, how many kilos do I want?”

Their decision rule is: Given P → decide Qd

Even though the fish existed before the price, the buyer’s response treats price as the input and quantity-demanded as the output. The demand curve is that response function.

Why They Diverge

Frame	What’s “First”	Why
Temporal	Quantity (fish arrive)	Physical reality: stuff exists before transactions
Causal (individual)	Price	Decision rule: buyer responds to price
Causal (market)	Neither—simultaneous	Equilibrium: P* and Q* determined together
Causal (inelastic supply)	Quantity	Fixed supply: price adjusts to clear

The theory assigns causal priority to price in the demand function because it describes how actors make decisions, not when things physically appear.

The Demand Curve’s Shape vs. Market Determination

Two separate things are true simultaneously:

The demand curve slopes downward — At lower prices, buyers want more. This is always true.
What determines equilibrium price varies by market structure — Sometimes price is given (retail), sometimes quantity is given (wholesale/auction with fixed supply).

The demand curve’s internal logic (P → Qd) is constant. What changes is which variable the market treats as the input for determining equilibrium.

This is another application of context-dependence: the demand curve doesn’t change, but what’s exogenous does.

When Does Classification Change? (Conditions Lens)

The classification isn’t fixed. It responds to modeling choices.

Change	Effect
Add an equation for a variable	Moves from exogenous → endogenous
Remove an equation	Moves from endogenous → exogenous
Change the question you’re asking	Entire classification may shift

Example:

Simple demand model: $Q_{d} = 100 - 2 P + 0.5 Y$

One equation → solves for one variable (Q)
P and Y are exogenous (given)

Add supply and equilibrium:

$Q_{d} = 100 - 2 P + 0.5 Y$
$Q_{s} = 20 + 3 P$
$Q_{d} = Q_{s}$

Three equations → solves for three variables (Qd, Qs, P)
P moved from exogenous to endogenous
Y is still exogenous (no equation for it)

The variable didn’t change. The model did.

What’s Sacrificed by Calling Something Exogenous? (Trade-offs Lens)

When you treat a variable as exogenous, you ignore any feedback loops.

Example: Income in a consumption model

Reality:

Income → Consumption (people spend based on what they earn)
Consumption → Employment (spending creates jobs)
Employment → Income (jobs generate earnings)

This is a loop. But if you treat income as exogenous, you cut the loop:

Income → Consumption ✓
Consumption → Employment ✗ (ignored)
Employment → Income ✗ (ignored)

The trade-off: Simplicity vs. realism.

Every exogenous assumption is a claim: “For my purposes, I can ignore how this variable gets determined.”

This is legitimate when:

The feedback is weak (my coffee consumption doesn’t affect Starbucks’ pricing)
The feedback is slow (income might respond, but not within the time frame I’m studying)
The complexity cost exceeds the accuracy gain

This is dangerous when:

Strong feedback exists and you ignore it
Policy analysis depends on the loop you cut

Procedure for Word Problems

Step	Ask yourself	Output
1. Identify the model	”What is this trying to explain?”	Name the model type
2. List the equations	”What relationships define this model?”	Write them out
3. Count unknowns solved for	”Which variables do these equations determine?”	→ Endogenous
4. Everything else	”What’s taken as given input?”	→ Exogenous

The rule: One equation solves one unknown. This scales linearly.

Equations	Endogenous variables solvable
1	1
2	2
n	n

For equations to form a system: They must share variables. Unconnected equations are separate problems, not a system.

Quick Classification for Standard Models

If you recognize the model type, you already know what’s endogenous:

Model type	Endogenous (solved for)	Exogenous (given)
Supply-demand equilibrium	P, Q	Income, tastes, technology, taxes
Demand only (firm sets price)	Q	P, income, preferences
Consumer choice	Quantities consumed	Prices, income, preferences
Production	Output level, input mix	Input prices, technology
Inelastic supply (fish market)	P	Q (fixed by harvest/catch)

Graph shortcut: Whatever is on the axes and jointly solved → endogenous. Whatever shifts the curves → exogenous.

Worked Example: Rental Market

“A city government is studying the market for rental apartments using a supply-demand model.”

Step 1 — Identify the model: Supply-demand equilibrium for rentals.

Step 2 — List the equations:

$Q_{d} = a - b P + c Y$ (demand)
$Q_{s} = d + e P$ (supply)
$Q_{d} = Q_{s}$ (equilibrium)

Step 3 — What do these equations solve for? P (rent) and Q (apartments rented).

Step 4 — What’s taken as given? Population growth, rent control law, income.

Variable	Classification	Why (causal chain)
Average rent	Endogenous	Three equations exist → system solves for P → rent is determined inside
Apartments rented	Endogenous	Three equations exist → system solves for Q → quantity is determined inside
Population growth	Exogenous	No equation determines it → must be given → affects demand but from outside
Rent control law	Exogenous	No equation determines it → must be given → constrains market but from outside

Worked Example: Computer Demand

“An economist hypothesizes that the annual quantity demanded of a specific computer brand ( $Q^{D}$ ) is determined by the price (P) and average income (Y) according to $Q^{D} = Y - 3 P$ .”

Step 1 — Identify the model: Demand-only model (one equation).

Step 2 — List the equations:

$Q^{D} = Y - 3 P$

One equation.

Step 3 — Apply the causal chain:

One equation exists
THEREFORE one variable can be solved for
THEREFORE two variables must be given as inputs
The equation is structured to solve for $Q^{D}$ (isolated on left side)
THEREFORE $Q^{D}$ is endogenous
P and Y have no equations determining them
THEREFORE P and Y must be provided before the equation can run
THEREFORE P and Y are exogenous

Step 4 — Trade-offs check:

By treating P as exogenous, we’re using a demand-only model. Both buyers and sellers are price-takers in a competitive market—neither sets the price. But this model has no supply equation, so it cannot explain HOW price is determined. It can only answer: “Given a price, how much is demanded?”

By treating Y as exogenous, we assume income is determined outside this model—by labor markets, macroeconomic conditions, etc. We ignore any feedback from computer purchases to income.

Variable	Classification	Causal Chain
$Q^{D}$	Endogenous	One equation exists → it solves for $Q^{D}$ → $Q^{D}$ is the output
P	Exogenous	No equation determines P → must be given → P is an input
Y	Exogenous	No equation determines Y → must be given → Y is an input

When Word Problems Are Ambiguous

If a problem doesn’t specify the model:

State your assumption: “Assuming this is a supply-demand equilibrium model…”
Classify based on that assumption
Your reasoning is defensible because you made your model explicit

Ambiguity in word problems transfers the cost of model selection to you. Professionals resolve this by being explicit about their framing—the same move as ROUGH FRAME in The Decision Lifecycle.

Common Traps

Trap	Fix
Memorizing “weather = exogenous, price = endogenous” as fixed facts	Always ask “in what model?” first
Classifying before identifying the model	Frame first, then classify
Assuming endogenous = dependent, exogenous = independent	They overlap but aren’t identical—in supply-demand, P and Q are both endogenous but jointly determined
Thinking tests can replace theory	Statistical tests for endogeneity require valid assumptions you can’t test. Theory comes first.
Stating the label without the mechanism	Use “therefore” chains: “No equation determines Y, THEREFORE Y must be given, THEREFORE Y is exogenous”
Asking “is X really endogenous?” without specifying a model	Malformed question—classification only exists relative to a model
Assuming “Price → Quantity demanded” is universal	True for individual response functions, but at market level with inelastic supply, Q → P
Confusing temporal sequence with causal priority	Fish exist before price is discovered (temporal), but that doesn’t mean fish “cause” price in the response function sense

Self-Check Before Submitting Answers

From The Causality Meta-Framework:

Check	Question	Fail condition
Therefore/But	Does my answer use “therefore” or just “and”?	”X is endogenous AND Y is exogenous” (no causal chain)
Five Whys	Can I go 3+ levels deep on my conclusion?	Answer dies at “because that’s the definition”
Process Verb	Did I name what’s happening (determines, solves, constrains)?	Only labels, no verbs
Mechanism	Did I state HOW, not just WHAT?	”Q is endogenous” without “because the equations solve for it”

If any fail → rewrite with explicit causal chain.

Connection to Decision-Making

Decision Lifecycle	Economic Modeling
ROUGH FRAME: “What am I deciding?"	"What is this model trying to explain?”
Inside the frame = relevant	Inside the model = endogenous
Outside the frame = context accepted as given	Outside the model = exogenous
Reframe if the scope is wrong	Choose different model if boundaries are wrong

Both are acts of scoping. The frame determines what counts as inside vs outside. Without the frame, classification is impossible.

How This Understanding Was Built: Framework Application Log

This note was developed by systematically applying frameworks to stress-test and improve initial explanations. The process is documented here for future reference.

The Problem Being Solved

Initial textbook definition: “Endogenous variables are determined within the model; exogenous variables are determined outside.”

This created confusion because:

“Within” and “outside” were unclear without knowing what a model IS
Word problems didn’t always specify which model to use
Heuristics like “push-back test” didn’t work consistently
The textbook’s “Price → Quantity demanded” causation claim seemed to conflict with wholesale market dynamics

Frameworks Applied

1. The Causality Meta-Framework (Therefore/But)

Before	After
”Endogenous = determined inside AND you solve for it"	"The model contains equations, THEREFORE some variables are determined by those equations, THEREFORE they’re endogenous”

Key contribution: Forced causal chains instead of lists. Revealed that the label “endogenous” is the LAST step, not the first.

2. Seven Lenses for Decomposing Claims

Applied to three core claims:

Claim: “Exogenous variables are NOT influenced by the relationships in the model”

Lens	Question Asked	What It Revealed
Mechanism	HOW does “influence” work?	Influence = an equation takes this variable as output. No equation = no influence.
Actors	WHO is doing the influencing?	The equations themselves—not people. Economists choose the model.
Conditions	What if the equations changed?	Add an equation → variable becomes influenced → moves to endogenous
Scale	Does this hold at micro and macro?	Yes—the logic is universal across model sizes
Sequence	What comes first?	Equation first. Classification is derived, not assigned.
Trade-offs	What’s lost by calling something “not influenced”?	You ignore feedback loops. Simplicity vs. realism trade-off.
Scope	Is this ALWAYS true?	Partially endogenous variables exist. The binary is a simplification.

Claim: “Exogenous = dial you turn; Endogenous = output you observe”

Lens	Question Asked	What It Revealed
Mechanism	HOW does turning the dial produce output?	Change exogenous → feed into equations → equations produce new endogenous values
Actors	WHO turns the dial?	The economist—hypothetically. This is “what if” analysis.
Sequence	What’s the order?	Set exogenous → Run equations → Observe endogenous
Trade-offs	What’s sacrificed?	”Dial” is a metaphor—you’re not literally controlling anything

Claim: “The same variable can be endogenous or exogenous depending on the model”

Lens	Question Asked	What It Revealed
Mechanism	HOW does adding an equation change classification?	More equations → more variables determined → more endogenous
Actors	WHO decides the model structure?	The economist. Model choice is a human decision.
Conditions	What if models disagree?	Neither is “right”—they answer different questions
Trade-offs	What’s the cost of making something endogenous?	Complexity. More equations = harder to solve, more assumptions.
Scope	Is there a “true” classification?	No. Classification is always model-relative. Reality doesn’t label variables.

Claim: “Price → Quantity demanded (direction of causation)“

Lens	Question Asked	What It Revealed
Conditions	What if supply is inelastic?	With fixed supply, quantity is exogenous and price adjusts—Q → P
Scale	Individual vs market level?	Individual response: P → Qd. Market clearing with fixed supply: Qs → P
Sequence	Temporal vs causal?	Fish arrive before price (temporal), but response function treats P as input (causal)
Scope	Is this ALWAYS true?	True for response functions, not for all market structures

3. Five Whys (Depth Test)

Applied to initial understanding:

"Why is Q endogenous?"
→ "Because you solve for it"
→ Why do you solve for it?
→ "Because there's an equation"
→ Why does having an equation make it endogenous?
→ [Initial understanding stopped here]

Key contribution: Revealed the gap. Full understanding requires: “Because equations DETERMINE values—they take inputs and produce outputs. A variable with an equation acting on it is an output of the system.”

4. Connection to Existing Frameworks

Existing Framework	How It Connected
The Decision Lifecycle	ROUGH FRAME = identifying the model. Both are scoping decisions that must precede classification.
The Cost Transfer Principle	Ambiguous word problems transfer the cost of model selection to the student. Professionals avoid this by being explicit.
Methodology as Power	Model choice determines classification, which affects conclusions. Choosing what to treat as exogenous is a power move.

Summary: What Each Framework Contributed

Framework	Contribution
Causality Meta-Framework	”Therefore” chains; revealed classification is derived, not assigned
Mechanism Lens	Required explaining HOW determination/influence works
Actors Lens	Clarified equations (not people) determine; economists choose models
Conditions Lens	Showed adding/removing equations changes classification; revealed inelastic supply case
Scale Lens	Confirmed logic holds across micro and macro; distinguished individual vs market causation
Sequence Lens	Established order: model → equations → classification; separated temporal from causal
Trade-offs Lens	Exposed costs of exogeneity (ignoring feedback) and endogeneity (complexity)
Scope Lens	Identified limits: binary is simplification; no “true” classification exists; P→Qd not universal
Five Whys	Depth-tested understanding; revealed where explanation stopped too early
Decision Lifecycle	Connected to framing principle (scope before classify)
Cost Transfer Principle	Explained why ambiguous problems are pedagogically weak

The Protocol for Future Concepts

When building understanding of a new economic concept:

State the textbook definition — What claim is being made?
Apply Causality Meta-Framework — Rewrite as “therefore” chain. Where does it break?
Apply Seven Lenses — Run through each lens, capture questions with traction
Apply Five Whys — Depth-test. Where does understanding stop?
Connect to existing frameworks — What does this relate to in your vault?
Document the process — Future you will forget how you got here

North: Where this comes from

Economic Modeling (this is fundamentally about how models work)
The Decision Lifecycle (framing determines classification)
Ceteris Paribus (exogenous variables are what we hold constant)
Systems of Equations (the mathematical foundation)
The Causality Meta-Framework (therefore/but structure)
ECON-1221 Chapter 2 - Notes from the Textbook (direction of causation assumption)

East: What’s the opposite?

Endogenous ↔ Exogenous (they define each other by the boundary)
Methodology as Power (model choice distributes consequences)
Correlation vs Causation (different problem—this is about which direction, not whether causation exists)

South: Where this leads

Comparative Statics (change exogenous input → see endogenous output change)
Simultaneous Equations (multiple endogenous variables solved together)
Price Elasticity of Supply (determines which causation context you’re in)
Agricultural Economics (classic case of inelastic short-run supply)

West: What’s similar?

The Decision Lifecycle Stage 1: ROUGH FRAME (scoping before classifying)
The Cost Transfer Principle (ambiguous models transfer cost to the reader)
Seven Lenses for Decomposing Claims (Conditions and Trade-offs lenses applied here)

Kai Mallari's Notes'

Explorer