The price elasticity of demand formula, worked end to end

Price elasticity of demand is the percentage change in units sold divided by the percentage change in price. It is almost always negative, because raising a price usually reduces volume, and what matters is its size. Below -1 and demand is elastic, meaning units respond more than proportionally. Between 0 and -1 it is inelastic. The number is easy to compute and hard to compute honestly, because almost every price change in a real business happened for a reason that also moved demand.

The price elasticity of demand formula

                     percentage change in quantity demanded
  Price elasticity =  ----------------------------------------
                          percentage change in price


  Simple (point) method, using the starting values as the base:

      % change in Q = (Q1 - Q0) / Q0
      % change in P = (P1 - P0) / P0


  Midpoint (arc) method, using the average of start and end as the base:

      % change in Q = (Q1 - Q0) / ((Q1 + Q0) / 2)
      % change in P = (P1 - P0) / ((P1 + P0) / 2)

Use the midpoint method. It matters more than it looks. The simple method gives a different answer depending on which end you start from, so a cut from $199 to $179 and a rise from $179 back to $199 produce two different elasticities for the same pair of observations. The midpoint method divides by the average, so the direction of travel cancels out and one price pair gives one number.

A full worked example, with the arithmetic

A retailer sells a cordless drill kit at $199.00 and moves 420 units in a four-week window. The price drops to $179.00 and units reach 520 in the next comparable four weeks. These are illustrative numbers chosen to make the arithmetic clear.

  Observed
    P0 = $199.00    Q0 = 420 units (four weeks)
    P1 = $179.00    Q1 = 520 units (next four weeks)

  Midpoint percentages
    average quantity = (420 + 520) / 2 = 470
    % change in Q    = (520 - 420) / 470  =  100 / 470  = +0.21277  = +21.28%

    average price    = (199 + 179) / 2 = 189
    % change in P    = (179 - 199) / 189  =  -20 / 189  = -0.10582  = -10.58%

  Elasticity
    E = +21.28% / -10.58% = -2.011   ->   -2.01


  The same data with the simple method, from the starting values
    % change in Q = 100 / 420 = +23.81%
    % change in P = -20 / 199 = -10.05%
    E = +23.81% / -10.05% = -2.37

  Two methods, one dataset, a difference of 0.36. Pick one and never mix them.

An elasticity of -2.01 says demand is elastic: a one percent price reduction bought roughly two percent more units. In most presentations that is where the analysis stops, and it is exactly where it should not, because the revenue and margin outcomes point in opposite directions.

That table is the single most useful thing on this page. Demand was elastic. Volume rose sharply. Revenue rose by $9,500. And the business made $3,500 less than it did before, because each of those extra units carried $20 less contribution. Any pricing conversation that stops at revenue will approve this decision.

Elastic, inelastic, and how to read the number

The dividing line is -1.0. Below it, meaning larger in magnitude, quantity responds proportionally more than price and demand is elastic. Between -1.0 and 0, quantity barely responds and demand is inelastic. At exactly -1.0 the two changes offset, which is the textbook definition of unit elasticity.

Two limits are worth stating plainly. Elasticity is local, not global: a value measured between $179 and $199 tells you nothing reliable about $129, because demand curves bend and price thresholds are real. And elasticity says nothing about profit on its own. That requires the contribution margin, which is the next section.

There is a second elasticity worth naming, because retailers meet it constantly and rarely separate it from the first. Own-price elasticity is how your units respond to your own price. Cross-price elasticity is how your units respond to somebody else's price, and for substitutable products it is normally positive: a competitor cutting their price reduces your volume. Anything you compute from a period when a competitor also moved is a blend of the two. That is not a technicality. It is the most common reason an elasticity estimate produced during a competitive promotion looks nothing like the one produced a month later.

Break-even volume: the calculation that should decide a price change

Before asking what elasticity is, ask what it would have to be. The break-even volume calculation answers that, it needs no historical data at all, and it takes thirty seconds. It is the single highest-value piece of pricing arithmetic most teams do not run.

  Contribution margin per unit, before   CM0 = P0 - variable cost
  Contribution margin per unit, after    CM1 = CM0 + (P1 - P0)

  Units needed to hold total contribution flat:

      Q_breakeven = Q0 x (CM0 / CM1)

  Required percentage change in volume:

      = (CM0 / CM1) - 1        which is the same as       -(P1 - P0) / CM1


  Worked on the example above, variable cost $130.00 per unit:

      CM0 = 199 - 130 = $69.00
      CM1 = 179 - 130 = $49.00

      Q_breakeven       = 420 x (69 / 49) = 420 x 1.40816 = 591.4 -> 592 units
      Required increase = 1.40816 - 1 = +40.8%

      Actually achieved = 520 units, an increase of +23.8%
      Shortfall         = 72 units

  Break-even elasticity for this cut:
      +40.8% / -10.05% = -4.06

  The cut needed demand around -4.06 and demand was -2.37. It was never
  going to pay, and the arithmetic said so before a single unit shipped.

Run that calculation before the price change, not after. It converts an argument about whether customers are price sensitive into a specific, falsifiable claim: we need 40.8 percent more units. Nobody in the room has to guess an elasticity. They only have to decide whether a forty percent volume lift is plausible, and that is a question merchandising can answer.

The asymmetry in that table is the practical lesson. On thin margins, price increases are far more forgiving than price cuts, and the intuition most teams carry is the opposite. If you want to run these numbers against your own costs, the price elasticity calculator does the elasticity and break-even arithmetic together, and the margin calculator handles the landed-cost side.

Estimating elasticity from observed data, without an experiment

Most teams do not run randomised price tests, so they try to read elasticity from history they already have. That is legitimate, and it produces a number worth having as long as everyone treats it as an order of magnitude rather than a measurement.

The standard approach is a constant-elasticity regression. Take the natural log of units and the natural log of price for each period, fit a straight line, and the slope is the elasticity directly. That is convenient enough to be worth knowing on its own.

  ln(Q) = a + b x ln(P)        b is the elasticity

  Minimum viable data before the number means anything:

    - at least 3 to 4 genuinely distinct price levels, not two
    - several periods held at each level, so a level is not one week
    - enough units per period that weekly variance is small next to the effect
    - a record of what else changed: promotions, placement, competitor prices,
      stock outages, marketing spend, season

  Pool across similar SKUs when a single product is too thin. A category-level
  elasticity from 40 comparable products is usually more trustworthy than a
  SKU-level one from 12 noisy weeks.

Three cheaper approaches are worth knowing because they need less data and lie less often. Historical promotions are natural experiments if the promotion was price-only, with no email, no placement change and no paid support, which is rarer than anyone expects. Staggered regional or channel pricing gives you a comparison group at the same moment in time, which removes seasonality entirely. And a deliberate small test on twenty SKUs with a matched holdout costs one quarter and produces something you can actually defend, which is the same rollout discipline described in the dynamic pricing guide.

The confounders that ruin naive estimates

Here is the core problem with observational elasticity, stated once. Your historical price changes were not random. You cut prices when demand was weak, when stock was heavy, when a competitor moved, or when marketing had a campaign to support. Every one of those reasons also moved demand. The regression cannot tell your price change apart from the reason for it, so it attributes both to price.

None of this makes elasticity useless. It makes it a prior rather than a measurement. Use it to rank which categories are worth testing, use the break-even calculation to decide whether a specific change could possibly pay, and use a holdout test to find out whether it did. That ordering is much more reliable than trying to extract a precise elasticity from data that was never generated to answer the question.

The competitor half of the picture matters too. A price change made while three competitors were also moving is not a clean observation of your own demand curve, which is why dated competitor series are a prerequisite for any of this, and why the rules described in repricing strategies mostly avoid depending on an elasticity estimate at all.