Counterfactual Framework and Potential Outcomes

Research Depth 187 in the knowledge graph I know this Set as goal
Unlocks 21 downstream topics
causal-inference potential-outcomes rubin-causal-model

Core Idea

The counterfactual framework defines causal effects as contrasts between potential outcomes under different exposure levels, observed for the same individual in hypothetical scenarios. The fundamental problem of causal inference is that only one potential outcome is observed per person; valid causal inference requires assumptions about missing counterfactuals (e.g., consistency, positivity, exchangeability).

How It's Best Learned

Work through simple numerical examples: two identical people who differ only in treatment received, calculate what the ATE would be if you could observe both potential outcomes, then see how the naive observational comparison can diverge. This makes the fundamental problem concrete and the need for assumptions intuitive.

Explainer

From your study of directed acyclic graphs (DAGs), you already understand that causation has a direction and that confounding arises when common causes of both exposure and outcome distort an observed association. The counterfactual framework takes this further: it defines what a causal effect actually *means* for a single individual. The claim "smoking caused her lung cancer" is a counterfactual claim — it asserts that if, contrary to fact, she had not smoked, she would not have developed cancer. Causation is always a comparison between what happened and what *would have happened* under a different world.

The formal notation makes this precise. Write Y(1) for the outcome a person would experience if exposed (treatment = 1) and Y(0) for the outcome they would experience if unexposed (treatment = 0). These are called potential outcomes — not observed outcomes, but outcomes that would be realized under each possible treatment state. The individual causal effect is Y(1) − Y(0): did the treatment change this person's outcome? This is the exact quantity we care about. But here is the inescapable problem: every person receives one treatment. If a patient takes the drug, we observe Y(1) and never learn Y(0). If they don't take it, we observe Y(0) and never learn Y(1). One of the two potential outcomes is always a counterfactual — literally counter to the observed fact. This is the fundamental problem of causal inference.

Because we cannot observe both potential outcomes for the same person, we cannot directly measure individual causal effects. The solution is to shift the estimand: instead of the individual effect, we target the average treatment effect (ATE) — E[Y(1) − Y(0)] — averaged across a population. This is estimable if the treated and untreated groups are *exchangeable*: statistically comparable in their potential outcomes, so that the untreated group's observed Y(0) can stand in for the treated group's counterfactual Y(0). Randomization achieves this mechanically; observational studies must achieve it through design and modeling.

Three key assumptions underpin valid counterfactual inference. Consistency requires that the potential outcome Y(a) for treatment a is precisely what you observe when treatment a is received — there is one well-defined version of each treatment level, not ambiguous variations. Positivity (also called the overlap assumption) requires that every subgroup defined by measured covariates has some positive probability of receiving each treatment level; if certain people *never* receive the treatment, we cannot estimate the effect for them. Exchangeability (no unmeasured confounding) is the most demanding: it requires that, conditional on measured covariates, treatment assignment is independent of the potential outcomes — no hidden common causes remain.

The connection to your DAG prerequisite is direct. A DAG encodes the assumed causal structure of the data-generating process. Exchangeability conditional on covariates Z translates to the DAG condition that blocking all backdoor paths from exposure to outcome is achievable by conditioning on Z. The three assumptions are not mere statistical niceties; they are substantive claims about the world that must be justified on scientific grounds. Sensitivity analysis — the topic this builds toward — is specifically the practice of asking how badly your conclusions break down if exchangeability is violated.

Practice Questions 5 questions

Prerequisite Chain

Counting to 10Counting to 20Understanding ZeroThe Number ZeroCounting to FiveOne-to-One CorrespondenceCombining Small Groups Within 5Addition Within 10Addition Within 20Two-Digit Addition Without RegroupingTwo-Digit Addition with RegroupingAddition Within 100Repeated Addition as MultiplicationMultiplication Facts Within 100Division as Equal SharingDivision as Grouping (Measurement Division)Division: Grouping (Repeated Subtraction) ModelDivision: Fair Sharing ModelDivision as Equal SharingDivision as GroupingBasic Division FactsDivision Facts Within 100Two-Digit by One-Digit DivisionDivision with RemaindersRemainders and Quotients in DivisionDivision Word ProblemsIntroduction to Long DivisionFactors and MultiplesPrime and Composite NumbersEquivalent FractionsRelating Fractions and DecimalsDecimal Place ValueReading and Writing DecimalsComparing and Ordering DecimalsAdding and Subtracting DecimalsMultiplying DecimalsDividing DecimalsDividing FractionsMixed Number ArithmeticOrder of OperationsInteger Order of OperationsVariable ExpressionsCombining Like TermsOne-Step EquationsTwo-Step EquationsSolving Multi-Step EquationsEquations with Variables on Both SidesAngle Pairs: Complementary, Supplementary, and VerticalParallel Lines and TransversalsCorresponding AnglesAlternate Interior AnglesTriangle Angle Sum TheoremExterior Angle TheoremTriangle Inequality TheoremSimilar Triangles: AA SimilaritySimilar Triangles: SSS and SAS SimilarityProportions in Similar TrianglesRight Triangle Trigonometry IntroductionTrigonometric Ratios ReviewRadian MeasureConverting Between Degrees and RadiansThe Unit CircleGraphing Sine and CosineGraphing Tangent and Reciprocal Trigonometric FunctionsDerivatives of Trigonometric FunctionsAntiderivativesIterated Integrals and Fubini's TheoremDouble Integrals in Cartesian CoordinatesDouble Integrals over Rectangular RegionsDouble Integrals in Polar CoordinatesDouble Integrals: Definition and SetupIterated Integrals and Fubini's TheoremDouble Integrals over Rectangular RegionsDouble Integrals over General RegionsApplications of Double Integrals: Area, Mass, and MomentsTriple Integrals in Cartesian CoordinatesTriple Integrals in Cylindrical and Spherical CoordinatesChange of Variables and the Jacobian DeterminantApplications of Triple Integrals: Volume and MassVector Fields and Their RepresentationsLine Integrals of Vector FieldsGreen's TheoremSurface Integrals and Flux of Vector FieldsSurface Integrals and Flux of Vector FieldsDivergence Theorem: Flux and OutflowDivergence TheoremElectric FluxGauss's LawConductors in Electrostatic EquilibriumCapacitance and CapacitorsDielectricsDielectric Constant and Relative PermittivityElectric Field Inside Dielectric MaterialsDielectric Materials and PolarizationDielectric Susceptibility and PermittivityEnergy Density in Electric FieldsElectric Current and Current DensityElectrical Resistance and ResistivityOhm's Law and Circuit ElementsElectromotive Force (EMF) and BatteriesKirchhoff's Circuit Laws: Voltage and CurrentDC Circuit Network Analysis MethodsTransient Response in RC CircuitsRC CircuitsLC and RLC CircuitsAC Circuits: FundamentalsImpedance and ReactanceAC Power and ResonanceElectromagnetic WavesThe Electromagnetic SpectrumBlackbody Radiation and Planck's LawPhotoelectric EffectThe Photon: Light as QuantaCompton ScatteringWave-Particle Dualityde Broglie WavelengthHeisenberg Uncertainty PrincipleWavefunction and the Born RuleThe Schrödinger EquationState Vectors and WavefunctionsQuantum SuperpositionQuantum EntanglementBell Theorem and Bell InequalitiesPostulates of Quantum MechanicsScattering TheoryIntroduction to Scattering TheoryPartial Wave Analysis in ScatteringSpin Angular MomentumElectron Spin and Intrinsic Magnetic MomentStern-Gerlach Experiment: Spin Quantization and MeasurementElectron Diffraction and Matter Wave PropertiesDavisson-Germer Experiment: Crystal Diffraction of ElectronsElectron Diffraction and Matter Wave InterferenceWavefunctions and Probability Density InterpretationQuantum Superposition and Linear Combinations of StatesQuantum Operators and ObservablesCanonical Commutation Relations and UncertaintyHeisenberg Uncertainty Principle and Measurement LimitsTime-Independent Schrödinger Equation and EigenvaluesHydrogen Atom in Quantum MechanicsSpectral Lines and Energy TransitionsSelection Rules for Atomic TransitionsLS and jj Coupling Schemes in Multi-Electron AtomsPauli Exclusion Principle and Antisymmetric WavefunctionsElectron Configuration and the Aufbau PrincipleThe Periodic Table and Atomic Electronic StructureThe Periodic TableElectron ConfigurationPeriodic TrendsIonization EnergyIonic BondingLewis StructuresResonance Structures and Delocalized ElectronsResonance and Formal ChargeMolecular Polarity and Dipole MomentsIntermolecular ForcesStates of Matter and Phase Changes: Melting, Boiling, and SublimationGas Laws and the Ideal Gas EquationGas Stoichiometry and Volume-Volume CalculationsThermochemistry and EnthalpyHeat Capacity and CalorimetryEntropy and Molecular DisorderSpontaneity and ΔGEntropy and Gibbs Free EnergyChemical EquilibriumAcid-Base ChemistryOrganic Reaction Mechanisms and Arrow PushingElectrophilic Addition to AlkenesAromaticity and BenzeneDNA StructureCentral Dogma of Molecular BiologyThe Genetic CodeDNA MutationsDNA Repair MechanismsCell Cycle Checkpoints and Cancer PreventionMitotic Spindle Checkpoint and Chromosome SegregationKinetochore Structure and FunctionMitochondria: Structure and FunctionCellular Respiration OverviewBacterial Metabolism OverviewAntibiotic Resistance MechanismsInfectious Disease EpidemiologyFoundations of EpidemiologyMeasuring Disease Frequency: Incidence and PrevalenceEpidemiologic Study DesignsConfounding: Definition, Identification, and Causal CriteriaDirected Acyclic Graphs for Causal ModelingCounterfactual Framework and Potential Outcomes

Longest path: 188 steps · 929 total prerequisite topics

Prerequisites (2)

Leads To (6)