Why Price Elasticities From the Cost-of-Living Squeeze Need Retesting

What has happened to real pay?

In the year to June 2022 real average hourly earnings in America fell 3.6 percent, and British real regular pay fell 3.0 percent in April to June 2022, a record fall at the time. Pay then grew faster than prices again, as the euro area chart below shows.¹

In the year to September 2026, by contrast, real average hourly earnings in America fell 0.7 percent, and British real regular pay grew 0.6 percent in the three months to July 2026. In the euro area hourly wages rose 3.0 percent in the second quarter of 2026, the same as consumer prices. Energy prices are rising fastest. In the year to September 2026 they rose 20.4 percent in America and, on Eurostat's first estimate, 18.8 percent in the euro area. Food inflation is low, at 1.8 percent for American groceries in September 2026 and 0.7 percent in the euro area in August 2026.¹

Low food inflation in 2026 has not undone the squeeze. Euro area food prices rose 33.1 percent between the first quarter of 2021 and the second quarter of 2026, and hourly wages 21.4 percent. An hour's pay therefore buys about 9 percent less food than before the squeeze (1.214 divided by 1.331 is about 0.91).

Real pay growth has stalled in 2026, and an hour's pay still buys less food than it did before the squeeze.

Why were elasticities from the squeeze years so low?

Price elasticity is the percentage change in volume divided by the percentage change in price, so at -1.2 a 5 percent price rise cuts volume by about 6 percent (5 times 1.2). An estimate fitted on sales from 2022 to 2024 learned most of what it knows about price from 2022 and 2023, when prices moved most. In those two years nearly every brand raised its prices at the same time, shoppers expected rises, and until mid-2023 real pay was falling.

Companies said at the time that volumes were falling less than usual. In March 2023 General Mills reported that "volume elasticities remained below historical levels". In January 2024 Mondelez's chief executive put elasticity "at or below historical norms", and its finance chief gave one reason for confidence: "competitors ... will have to price too".²

In those years a brand's own price and its category's price moved together, so a model found it hard to tell them apart. The number it produced is called a brand elasticity, but much of what it measured was the category's response, unless the model also included competitors' prices and those prices moved at different times from the brand's. No published study has directly measured that gap between what the number was called and what it measured; the company remarks above point to it. Falling real pay in the same months blurs the estimate further. The model has to separate the effect of falling pay from the effect of higher prices, and an error in that split could push the estimate either way.

Company results since then point to stronger reactions. PepsiCo raised prices by double-digit percentages for eight straight quarters in 2022 and 2023, and sales of its drinks and Frito-Lay snacks then fell. Early in 2026 it cut the suggested retail prices of its main chip brands in America by up to nearly 15 percent; the shelf prices were the retailers' decision.³ On 30 September 2026 Conagra's finance chief said the company was "factoring in greater-than-historical volume elasticities" in parts of its business.²

An elasticity fitted on the squeeze years can understate what a brand loses when it raises its price alone, and PepsiCo's and Conagra's experience since then points the same way.

Does price sensitivity rise when times are hard?

Past downturns suggest that it does on average, with large differences between categories. A study of 19 American grocery categories from 2001 to 2006 found that price sensitivity rose in weak economic times in seven categories, fell somewhat in six and did not change in six. Categories that take a large share of the household budget tended to become more sensitive.⁴

Shoppers who traded down during the squeeze have not all traded back, so private label is now a familiar alternative when a brand raises its price alone. Part of the share that private label gains in a downturn is permanent, a study of four countries found.⁵ In a 2026 survey by Simon-Kucher, a consultancy, only a quarter of 14,000 shoppers in 14 countries said they would return to brands if prices fell.⁶ In Britain, sales of branded groceries grew faster than private label's in the four weeks to 6 September 2026, 3.7 percent against 2.9 percent. That reverses the pattern of much of 2026, but is one month of data.⁷

The large everyday categories are the likeliest to have become more price sensitive since the last estimate, because food takes more of an hour's pay than in 2021 and energy is squeezing the rest of the budget.

What does a wrong elasticity cost?

Take an invented snack brand that sells 1,000,000 cases a year at a net price of 20.00 euros a case after its trade terms with retailers. Each case costs 12.00 euros to make and deliver, so the brand keeps 8.00 euros a case (20.00 less 12.00). That is its contribution, the money left to cover fixed costs and profit: 8 million euros a year (1,000,000 cases at 8.00 euros). It plans a 5 percent list price rise. If its trade terms stay the same share of list price, its net price also rises 5 percent, to 21.00 euros a case (20.00 plus 5 percent). Its contribution rises to 9.00 euros a case (21.00 less 12.00). Its finance team's elasticity, fitted on 2022 to 2024 sales, is -1.2.

Elasticities are usually measured against the shelf price that shoppers pay, and each retailer decides its own shelf price; a recommended retail price from the brand is only a direction. The example assumes full pass-through, with each retailer raising its shelf price by 5 percent. A retailer that keeps the same percentage margin raises its shelf price by that same 5 percent. One that keeps the same margin per case in euros raises its shelf price by less than 5 percent, because its own margin is part of the shelf price and does not rise. Volume then falls by less than the figures below show.

At -1.2, volume falls 6 percent (1.2 times 5 percent) to 940,000 cases. At 9.00 euros of contribution a case, that is 8.46 million euros, so the brand's contribution rises by 460,000 euros a year (8.46 million less 8 million). If the brand's elasticity has moved to -2.0, volume falls 10 percent to 900,000 cases, which give 8.10 million euros, and the gain shrinks to 100,000 euros. At -2.6, volume falls 13 percent to 870,000 cases, which give 7.83 million euros, and the rise loses 170,000 euros. That elasticity is a stress case, close to the average of -2.62 across 81 published studies.⁸

The break-even lesson also calls a move a coin flip when the forecast loss lands within 1.5 points of the BESC, because forecasts carry error. At -2.0 the forecast loss of 10 percent sits 1.1 points inside the 11.1 (11.1 less 10), so even the case that still earns 100,000 euros falls in that zone.

Shoppers usually respond less to a price cut than to an equal rise. In the Academy's elasticity lesson a 5 percent rise costs about 9 percent of volume and an equal cut gains about 6 percent, a ratio of 1.5. An elasticity estimated on rises therefore overstates what a later cut wins back.

Suppose the rise loses 100,000 cases at -2.0, 10 percent of the brand's volume, and the brand then takes its net price back to 20.00 euros a case, with each retailer assumed to take its shelf price back too. If shoppers responded to the cut as strongly as to the rise, it would win back all 100,000 cases. At the lesson's ratio the cut wins back two-thirds of them (1 divided by 1.5), about 66,700. The brand ends at about 966,700 cases (900,000 plus 66,700), 3.3 percent below where it started at the same price. Its contribution ends about 266,700 euros a year lower: the 33,333 cases it did not win back (a third of the 100,000 lost), at 8.00 euros a case.

A price rise planned on too low an elasticity can lose money, and cutting the price back later does not win back all the volume it lost.

How can a team tell whether its elasticity still holds?

The Academy's elasticity lesson names the use of stale estimates as a pitfall and advises re-estimating at least once a year. The first test is to fit the model on data to the end of 2024, give it the prices and promotions that actually ran from January 2025 to June 2026, and compare the sales it predicts with what the brand sold. Errors that lean the same way period after period mean the model has gone stale. If the misses are largest when the brand's own price moved, the elasticity is the likely cause.

A second test is to re-estimate the elasticity on the latest two years of data, then move that two-year window forward a quarter at a time, to see whether the estimate has left its old range or is only varying within it. A structural-break test, a statistical check for the point at which a relationship changed, then shows when the change happened.

A model that includes real pay or energy prices as separate variables, and measures the brand's price against both its competitors' prices and the category's price level, is less likely to need rebuilding each time the economy turns. National pay data move slowly, though, and give it little to learn from. Whichever model the team uses, each price move can be shown to the board at three elasticities (the old estimate, the new one and a stress case near the published average), and the moves that make money at all three can go ahead first.

A project that estimated more than 7.5 million elasticities from retail scanner data found that their distribution did not shift ("stationary over time") in the years it covers.⁹ In a 2024 study of a 35-week price experiment across 82 test stores, the average elasticity was -0.34 in the experiment against -1.97 from standard estimates on observational data, the sales recorded in normal trading. A standard estimate can therefore overstate price sensitivity too.¹⁰ In a category that takes a small share of the household budget, where competitors and private label move with the brand, a squeeze-era estimate may still be close enough. But an estimate that has not been tested against the conditions of 2026 should not be assumed to hold.

An elasticity fitted on the squeeze years should be tested against sales from 2025 and 2026 before it is used to set list prices for 2027.

References