A business report states, cleanly: "demand elasticity for this product is -1.4." It's presented as a fact about the product, the way weight or dimensions would be. It isn't. That number is a fact about one specific price range, one specific time period, and one specific set of market conditions — and every one of those conditions, left unstated, quietly limits how far the number can actually travel.
Elasticity is range-dependent, not constant — and a single number hides that
Demand curves are very rarely straight lines. The elasticity measured between $40 and $45 is frequently different from the elasticity measured between $60 and $65 for the same product, because different price ranges put the product in front of different buyer segments with different sensitivity. Reporting one elasticity figure without the price range it was estimated over silently assumes the curve is a constant-slope line across all prices — an assumption that's convenient for a slide but rarely true, and one that specifically breaks down exactly where it matters most: extrapolating to a price point outside the range that was actually tested.
Historical sales data usually has an endogeneity problem, and it biases the estimate in a specific direction
The naive approach: pull historical sales at different price points, regress quantity on price, read off the elasticity. This has a well-known and serious flaw — prices in the real world are usually not set independently of demand. A retailer raises prices when demand is strong and discounts when it's weak, which means the historical data reflects the pricing decisions made in response to demand, not demand's independent reaction to price. Naively regressing on this data tends to understate true elasticity, because the price-setting process itself is correlated with the very demand shocks you're trying to isolate. Correcting for this requires an instrument or a genuine natural experiment (a price change driven by something unrelated to demand, like a tax change or a supply cost shift) — without one, the estimate is a description of historical pricing behavior, not a reliable input for a forward-looking pricing decision.
An elasticity from last year's data doesn't necessarily hold in this year's environment
Elasticity reflects the competitive and macroeconomic conditions the data was generated under. A product's demand became meaningfully more price-sensitive during a period of tightened consumer spending than it was the year before, or meaningfully less sensitive after a competitor exited the category — same product, different number, because the conditions changed. An elasticity estimate carried forward without checking whether the underlying conditions still hold is being applied outside the range it was validated for, in time instead of in price.
What a more honest report actually states
Instead of a single point elasticity, a report that's actually useful for a pricing decision states: the price range the estimate covers, the time period and competitive conditions the data reflects, and whether the estimation method accounted for endogeneity (a genuine experiment or instrument) or is a naive historical regression (in which case, say so, and treat the number as directional at best). None of this requires a more complex model — a naive regression reported honestly, with its limitations stated, is more useful to a decision-maker than a sophisticated model's output presented as an unconditional fact.
The practical check before using an elasticity number in a pricing decision
- What price range was this estimated over, and does the decision in front of you fall inside or outside that range?
- Was the underlying price variation driven by something independent of demand, or could prices have been set in response to demand?
- How old is the underlying data, and have competitive or macro conditions changed since then?
- Is this being presented as a single point estimate, or with the uncertainty and conditions that actually surround it?
An elasticity number without its conditions isn't wrong because the math is wrong — it's wrong because it's being asked to answer a question (what happens at a price we haven't tried, in a market that may have changed) that the underlying data was never able to answer in the first place.