Marginal Substitution Estimator¶
Estimation model — instantiates Revealed Preference Validation Against Indifference Curves
Estimates the local rate at which a chooser traded one attribute for another, reading marginal substitution rates off choices made near the margin.
Once options live on comparable axes, the question becomes: at what exchange rate did this person swap one attribute for another? Marginal Substitution Estimator answers it numerically. It finds choices made near a boundary — where the chooser was nearly indifferent and a small change would have flipped the decision — and reads the local trade-off rate directly off them: how many dollars of price the chooser accepted per minute of time saved, per unit of quality gained. Its defining product is a number with a slope — the marginal rate of substitution at a point — not a picture and not a verdict. It quantifies the trade-off; it leaves the drawing, the consistency-checking, and the judging to other mechanisms.
Example¶
A transit authority wants to know how commuters trade travel time against fare. From normalized trip records it isolates the near-margin cases: riders who, week to week, switched between a slower cheap bus and a faster costlier express as the express fare and the bus's reliability drifted. Around each switch, the estimator fits the local slope — the fare increment that just tipped a rider off the express and onto the bus — and recovers a value of travel time of, say, roughly $18 per hour for the commuter segment (illustrative), rising steeply for trips with tight transfer windows.
That single estimated rate is the payload. It tells the authority not merely "people like fast trips" but how much speed is worth at the margin, in the same units as the fare — enough to price a new express tier or to judge whether a time-saving investment clears its cost. Crucially, the estimator reports the rate as a local quantity: it is trustworthy near the observed switches and says nothing, on its own, about trade-offs far from them.
How it works¶
- Locate the margins. Find choices where the chooser was close to indifferent — small perturbations in price or an attribute flipped the selection — since only there is the local trade-off identified.
- Fit the local slope. Estimate the exchange rate between the two attributes that rationalizes the switch, using the normalized bundle coordinates.
- Attach the rate to the bundle space. Record the estimated marginal valuation as a slope at that point in attribute space — a fragment of the underlying trade-off surface.
- Report the trace, not the whole map. Emit a set of local rates with the region each was estimated in, explicitly local rather than a global utility function.
Tuning parameters¶
- Margin window — how close to indifferent a choice must be to count. Tight windows give cleaner local identification but fewer data points; wide windows borrow strength at the cost of conflating distinct trade-offs.
- Functional form — the local model fitted to the switches (linear, log, flexible). More flexible forms bend to the data but overfit noise into spurious curvature.
- Pooling level — estimate one rate per individual or pool across similar choosers. Pooling stabilizes noisy individual data but blurs heterogeneity.
- Attribute pair scope — which trade-off is being estimated; the rate between time and money differs from the rate between quality and money.
- Noise model — how much of a switch is attributed to taste versus random error, which sets how sharply the slope is pinned down.
When it helps, and when it misleads¶
Its strength is precision where the data is richest: near observed switches it turns behavior into a defensible exchange rate expressed in the decision's own units, which is exactly what pricing, compensation design, and cost-benefit analysis need. It is the quantitative heart of the archetype.
Its failure mode is over-extension. A marginal rate of substitution is local by construction — valid in the neighborhood of the choices it was fit to — and treating it as a global constant is the recurring error.[n1] Extrapolate the transit rate to a fare ten times higher and it will mislead, because trade-off rates bend as bundles move. The estimate is also only as clean as the margins: if few choices sit near indifference, the slope is barely identified and noise masquerades as signal. The guarding discipline is to report the rate with its region of validity and to hand the drawing of a full contour — with its widening uncertainty away from the data — to a visualization step rather than pretending one local slope describes the whole surface.
How it implements the components¶
Marginal Substitution Estimator realizes the archetype's quantitative-inference core:
revealed_substitution_trace— its primary output: the estimated local marginal rates of substitution read off near-margin choices.attribute_bundle_model— it enriches the normalized bundle space with marginal valuations, attaching an estimated slope to points within it.
It does not draw the contours or band their uncertainty (inference_confidence_band, segment_specific_preference_map) — that is Indifference Region Visualization, which consumes this trace — and it runs no consistency tests (consistency_test_suite) on the resulting relations, which belongs to Revealed Preference Consistency Matrix.
Related¶
- Instantiates: Revealed Preference Validation Against Indifference Curves — it supplies the numeric trade-off rates the validation is built to test and use.
- Consumes: Choice Bundle Normalization supplies the comparable attribute coordinates the slopes are fit in.
- Sibling mechanisms: Budget Set Reconstruction · Choice Bundle Normalization · Indifference Region Visualization · Dominance Violation Scan · Revealed Preference Consistency Matrix · Preference Reversal Probe · Choice Architecture Confound Audit · Stated vs Revealed Gap Report · Ethical Preference Inference Review
Editorial Notes¶
Form Classification¶
Form family: Analysis, Modeling & Optimization
Rationale: Marginal Substitution Estimator operates as a computation, comparison, model, or analytic representation used to infer, estimate, or choose because it estimates the local rate at which a chooser traded one attribute for another, reading marginal substitution rates off choices made near the margin.
Independent corroboration: The frozen evidence defines Marginal Substitution Estimator as 'Estimates the local rate at which a chooser traded one attribute for another, reading marginal substitution rates off choices made near the margin', so its operative form is Analysis, Modeling & Optimization.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Economics & Finance
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Specialized
Rationale: Marginal rates of substitution and indifference curves are canonical microeconomic concepts.
Related originating lineages:
- Psychology — Psychophysics and preference research independently study attribute tradeoffs in human choice.
- Statistics & Experimental Design — Estimating local tradeoff rates from revealed choices materially supplies the empirical model.
Review resolution: Both independent reviews assign primary provenance to economics_finance. The queued secondary differences (alternate_origin_disagreement) are reconciled by retaining psychology, statistics_experimental_design only as formative or independently established lineage(s), not merely as application domains. origin_mode=cross_disciplinary_synthesis records the provenance relationship, while domain_reach=specialized separately records applicability breadth. confidence=high preserves the more cautious assessment, and encyclopedia_synthesis=false records whether either reviewer identified a corpus-specific synthesis.
Review outcome: Reconciled after independent review; high confidence.
Notes¶
[n1] The marginal rate of substitution is the slope of an indifference curve at a point — the rate at which a chooser will swap a little of one good for a little of another while staying equally satisfied. It is defined locally; its variation along the curve (a diminishing MRS) is precisely why a single estimated slope cannot stand in for the whole trade-off surface. ↩