A latticework reading of Ashley Hodgson on elasticity — the economic tool for measuring how strongly one variable responds to another, and what it reveals about cause, magnitude, and the limits of intervention.
Ashley Hodgson on elasticity in economics
Most mental models describe mechanisms: how leverage works, how feedback loops compound, what inversion produces. Elasticity is different. It is a measurement framework — a way of attaching a number to the strength of a causal relationship. Defined as the percentage change in one variable divided by the percentage change in another, it forces two questions that most informal reasoning skips: which direction does the relationship run, and how strong is it? These questions seem elementary. Getting them wrong is surprisingly common and surprisingly costly.
Ashley Hodgson's lecture works through elasticity with everyday examples before arriving at its economic applications — price and quantity, income and demand, supply responsiveness. The non-economics opening is the latticework gift: it reveals that elasticity is not really about economics at all. It is about how to think about cause and effect in any domain where the relationship is real but the magnitude is unknown. Farnam Street's canon covers many specific mechanisms. Elasticity covers the gap between knowing that a mechanism exists and knowing how hard it pulls.
The reinforcements are in the models that get quantified. The contradictions are in the models that quietly assume magnitude where they should be agnostic. The new additions are frameworks for classifying what kind of relationship you're actually dealing with before you design any intervention.
The model most directly clarified is leverage. The standard account of leverage is qualitative: a small input produces a large output. Elasticity makes this precise. High elasticity is high leverage — the dependent variable moves a lot for each unit of independent variable. Low elasticity is low leverage. When Farnam Street's canon says "find the high-leverage point," elasticity is the tool that makes that search quantitative rather than intuitive. A ten-percent price increase that cuts demand by fifty percent (elasticity of −5) is a very different lever than a ten-percent price increase that cuts demand by two percent (elasticity of −0.2). Both are "price affects demand." Only one is a lever worth pulling.
Feedback loops get sharper from elasticity too. A feedback loop with high elasticity — where a small perturbation causes a large response which feeds back into the next perturbation — is genuinely dangerous in ways that a low-elasticity loop is not. The model of second-order effects also benefits: a policy intervention has strong second-order effects when the system it touches has high elasticity. A $5 tax on a perfectly inelastic good produces no behavioural change and pure revenue. The same $5 tax on a highly elastic good produces near-total substitution and near-zero revenue. The second-order effect is embedded in the elasticity.
Finally, map and territory gets a sharper diagnostic: when your theoretical model predicts a direction but the observed elasticity is near zero, the map is wrong. Hodgson's example is perfectly inelastic demand for marriage licenses — raising the price from twenty to forty dollars produces no detectable change in the number of marriages. The map (demand curves slope down) is directionally correct; the magnitude tells you that, for this product, at this price range, the model contributes nothing actionable.
The standard account of incentives in the Farnam Street canon — summarised as "show me the incentives and I'll show you the outcome" — is directionally correct but mute on magnitude. It predicts that higher pay will attract more workers. It does not predict whether a ten-percent wage increase will attract two percent more workers or two hundred percent more. The difference matters enormously for policy design. Low wage elasticity of supply means you need very large wage increases to fill positions; high elasticity means you barely need to move. The "incentives shape behaviour" model is almost always right on direction; elasticity is the tool that tells you whether the direction is actionable at realistic magnitudes.
Cause and effect as a mental model — the idea that identifying a causal relationship is the hard part, and the application follows — gets the same critique. In many real-world interventions, the causal direction is established and uncontested; it's the magnitude that is unknown and it's the magnitude that determines whether the intervention is worth running. Alcohol education programs on campuses presumably reduce binge drinking in some direction. Whether they reduce it by half a percent or thirty percent determines whether the program is worth its cost. The casual model says "cause found, effect follows." Elasticity says "effect size unknown until measured."
Perhaps the most counterintuitive bend involves inversion. Inversion says: to achieve a goal, ask what would prevent it. But the useful inversion for price-setting isn't just "what would reduce demand?" — it's "at what elasticity does a price increase hurt revenue?" When elasticity is greater than one (elastic demand), a price increase reduces quantity so much that total revenue falls. When elasticity is less than one (inelastic demand), a price increase raises revenue even as quantity falls. Inversion, without the elasticity quantification, misses this sign flip.
The most generalisable new model is Direction-Magnitude Decomposition. Any causal claim can be decomposed into two independent questions: (1) which direction does the relationship run (positive, negative, or null)? and (2) how strong is it (elastic or inelastic)? These questions are logically independent — a relationship can be positive and weak, positive and strong, negative and weak, or negative and strong. Most informal reasoning conflates them: "raising tuition will reduce enrollment" answers only the direction question. The magnitude determines whether the effect is large enough to matter for any real decision. Making the decomposition explicit before acting is the discipline the model teaches.
The second new model is the Substitution Proximity Test. Elasticity in economic markets is driven primarily by how easily buyers can switch to a substitute. Close substitutes → high elasticity (small price increase, large substitution). No substitutes → low elasticity (insulin, salt, marriage licenses at low absolute prices). Generalised beyond economics: any dependent variable will be highly elastic with respect to an independent variable if there is a cheap, accessible alternative behaviour. Alcohol education may have low elasticity partly because the substitute (not attending training) is blocked by mandate. The test — "how available is the alternative?" — predicts magnitude before measurement.
The third new model is the Inferior-Good Reversal. Most mental models assume that more income produces more of what you value. For inferior goods — ramen, generic brands, certain transit options — more income produces less demand, because the product exists as a substitute for something better. The inferior-good frame generalises: in any domain where a behaviour is a second-best response to a constraint, removing the constraint doesn't increase the behaviour — it eliminates it. This is a reliable way that first-order analysis generates the wrong sign. Ask: is this behaviour a choice, or a workaround?
The Farnam Street canon is rich in mechanisms and weak on measurement. Elasticity fills one part of that gap. It doesn't add a new causal story; it adds a way of asking how strong the existing stories are. The mental model most changed by this concept is not any specific one — it's the implicit assumption that identifying a direction is enough to act on. It is not, quite, enough.
Elasticity is how we hold our theoretical models in check with reality. — Ashley Hodgson
The practical residue is a habit: before designing any intervention — a price, a policy, a campaign — pause to estimate the elasticity of the target variable. Direction first, then magnitude. The estimate doesn't need to be exact. It needs to tell you whether the relationship is strong enough to matter at the scale you can realistically implement. Most interventions fail not because the causal direction was wrong, but because the magnitude was never quantified and turned out to be too small.