Shrinkflation touched 29.6% of Canadian grocery items in three years yet added just 0.06 points to US inflation. The gap between the two is the story.
Two Numbers, Both True
and the argument that never resolves between them
In the three years to 2023, 29.6% of the grocery items Statistics Canada tracks in its consumer price basket were sold in a smaller quantity than before for the same money ✓ Established Fact [3]. Over almost exactly the same window, package downsizing added 0.06 of a percentage point to the 34.5% rise in United States consumer prices ✓ Established Fact [1]. Neither figure is wrong. The distance between them is the subject of this report, because that distance is where a measurement problem turns into a legitimacy problem.
Shrinkflation is not hidden from the consumer price index, and it is not a large driver of measured inflation. Those two statements are the settled part of the evidence, and almost every public argument about shrinkflation proceeds as though neither were true. When a box shrinks and the price holds, national statistical agencies convert the change into a per-unit price increase and record it as inflation ✓ Established Fact [2]. The United States Government Accountability Office, reviewing Bureau of Labor Statistics data from December 2019 to December 2024, found that downsizing contributed 0.06 of a percentage point to a 34.5% cumulative price rise — an average of roughly 0.01 percentage point a year ✓ Established Fact [1]. That is the aggregate. The rest of this report is about why the aggregate is the least informative number available, and why treating it as the answer has been so expensive.
Begin with the Canadian measurement, because it is the most direct count of the practice rather than of its price effect. Statistics Canada examined every eligible grocery item in the consumer price index between 2021 and 2023 and found that 29.6% of them underwent a quantity adjustment — the technical term for a package that shrank ✓ Established Fact [3]. Nearly half of those instances, 49.5%, fell in a single year, 2022, when Canadian grocery inflation reached its highest annual average in 41 years at 9.8% ✓ Established Fact [3]. The remaining instances split fairly evenly, 24.4% in 2021 and 26.2% in 2023 [3]. Downsizing, on this evidence, is not a permanent background hum. It is a device that firms reach for hardest when input costs are moving fastest and a headline price rise would be most visible.
Now set that against the American measurement, which counts something different. The GAO was not asking how many products shrank; it was asking how much of the observed inflation those shrinkages explain. Because the consumer price index is a weighted average across every category a household buys, and because downsizing is overwhelmingly a packaged-goods phenomenon, the arithmetic dilutes it to near-invisibility. Shelter, medical care, motor vehicles, insurance and energy carry the weight of the index, and none of them can be downsized in a box. The 0.06 figure is therefore not a rebuttal of the 29.6% figure. It is an answer to a question almost nobody was asking.
The third measurement is the one that decides the politics. In the European Commission 2025 Consumer Conditions Scoreboard, 74% of consumers reported noticing that packaged goods they buy regularly had shrunk without a corresponding fall in price ✓ Established Fact [7]. Roughly three-quarters of a continent has registered the practice directly, at the shelf, without any statistical agency telling them about it. In the United Kingdom, YouGov tracking through July 2025 found reported shrinkflation rising across categories over two years: frozen foods from 26% to 32%, beverages from 24% to 31%, cleaning products from 26% to 29%, and beauty and personal care from 20% to 28% [24]. Awareness is climbing even as headline food inflation falls.
Perception of inflation has decoupled from the measurement of it, and the decoupling is durable rather than transient. European Central Bank consumer survey data put the median perceived rate of inflation over the preceding twelve months at 3.5% in July 2026 ✓ Established Fact [18]. Research published through the Centre for Economic Policy Research frames the difference as a perceived inflation wedge that widens sharply during periods of economic stress — the global financial crisis, the pandemic recession, the 2021 and 2022 surge — and then closes only slowly, because once households begin paying close attention to prices, that attention persists well after the rate itself falls back ◈ Strong Evidence [19]. The wedge is largest among lower-wealth households, whose decisions are most sensitive to perceived purchasing power [19].
This report treats shrinkflation as a measurement question with a political answer. The measurement question — does the index capture it — has been settled for years and settled in the affirmative. The political question is why a practice that moves the aggregate index by six hundredths of a point has produced labelling statutes in France, South Korea and Brazil, a failed federal bill in the United States, an enforcement regime in the United Kingdom and a durable collapse in the credibility of official price statistics. The answer is that shrinkflation is not primarily an inflation event. It is a disclosure event, and disclosure failures compound in a way that price rises do not.
What the Index Does With a Smaller Box
the quantity adjustment, and the three things it cannot see
When a data collector records that a carton of ice cream has gone from 64 ounces to 60 ounces while the shelf price stays at 5.99 dollars, the Bureau of Labor Statistics does not record a flat price. It records a 6.7% increase in the effective price per ounce ✓ Established Fact [2]. This procedure — quantity adjustment — is decades old, is documented publicly, and is applied by every major statistical agency. Understanding exactly what it does, and where it stops, is the precondition for any honest argument about hidden inflation.
The mechanics are unglamorous and worth stating precisely. Price collectors in the field record not only the marked price of an item but its size, count and net weight. When a size changes between collection periods, the collector flags it, updates the item description and refers it to a commodity analyst, who converts both the old and the new observation into a price per standard unit — usually per ounce, per sheet or per millilitre ✓ Established Fact [2]. If the number of sheets on a toilet roll falls from 220 to 200 while the price holds, the index registers a 10% rise in the price per sheet [2]. The consumer experiences no change at the till. The index registers inflation. That is the intended behaviour, not a loophole.
The Bureau of Labor Statistics has published the size of the effect twice. Between 2015 and 2019, the all-items index differed from a research index holding sizes constant by 0.05% over the full five years, roughly 0.01% a year ✓ Established Fact [2]. Within food and beverages the difference was 0.21%, with baby food at 2.81%, snacks at 2.64%, fresh biscuits and rolls at 1.59% and household paper products at 1.30% [2]. A second window, January 2019 to October 2023, put household paper products at a 3.6% difference and snacks at 2.6% [2]. The aggregate effect is trivial and the category effects are not, which is the same finding the GAO reached six years later with newer data [1].
The Bureau of Labor Statistics converts every observed size change into an effective price per standard unit, so a 64-ounce carton falling to 60 ounces at an unchanged 5.99 dollars enters the index as a 6.7% price rise ✓ Established Fact [2]. Statistics Canada applies the same procedure and calls it quantity adjustment, identifying 29.6% of eligible grocery items as having undergone one between 2021 and 2023 [3]. The claim that shrinkflation constitutes inflation the government does not count is false as a matter of documented methodology, and has been for decades [22].
Canada documents the same procedure and reports it more granularly. The February 2025 study from Statistics Canada is explicit that when product sizing changes are captured, adjustments are made in the index to account for the impact on price ✓ Established Fact [3]. The categories where quantity adjustments clustered are exactly the ones a shopper would name from memory: margarine, pasta mixes, cookies and crackers, mozzarella cheese, breakfast cereal and cheddar cheese [3]. The United Kingdom is moving in the same direction from a different angle. The Office for National Statistics has published an indicative impact analysis for grocery scanner data ahead of its introduction into consumer price statistics in March 2026, which replaces hand-collected shelf observations with the retailer transaction record itself [25].
Scanner data matters here more than any labelling statute, because it changes what the index can see. A field collector visits a store, samples a defined item and records what is on the shelf that day. Scanner feeds capture every transaction, every size variant and every substitution across the whole assortment. The GAO identified retail scanner integration as one of five policy levers available to statistical agencies and noted that eight countries already use it ✓ Established Fact [1]. Where it is in place, downsizing is not something an analyst has to catch. It arrives in the data as a change in the size distribution of what households actually bought.
There are three things quantity adjustment cannot do, and they matter more than the thing it does. It cannot see quality. A recipe that substitutes a cheaper fat, a garment that drops a stitch count, a hotel that stops cleaning rooms daily and a support line that adds forty minutes to its queue all deliver less for the same money, and none of them changes a declared net weight ◈ Strong Evidence [23]. It cannot see the fee that appears at the final checkout step rather than in the advertised price. And it cannot see the composition of the basket a household is forced into when the cheap variants of a category inflate faster than the expensive ones.
That last effect has a name and a measurement. Analysis of micro price data covering 91 large multi-channel retailers across ten countries between 2018 and 2024 found that temporary discounts grew more slowly than regular prices during the post-pandemic surge, which mitigated the household burden, while cheapflation — the faster rise in the prices of cheaper varieties relative to expensive varieties of the same good — worked in the opposite direction and made it worse ◈ Strong Evidence [6]. A household that trades down to protect its budget is trading into the segment where prices are rising fastest. The index records the price of each variety correctly. It does not record that the household had to move.
Where the Shrinking Actually Lands
concentration is the finding, aggregation is the distraction
In the five product categories most exposed to downsizing between 2019 and 2024, size changes contributed between 1.6 and 3.0 percentage points to category inflation ✓ Established Fact [1]. Household paper products sat at the top at 3.0 points, snacks at 2.6, candy and chewing gum at 2.3, ice cream at 1.8 and breakfast cereal at 1.6 [1]. These are between 27 and 50 times the aggregate effect, and they fall on the part of the basket a household restocks every week.
The distributional structure of downsizing is the whole story, and it is systematically lost when the practice is discussed at the level of the all-items index. Downsizing does not occur at random across the consumption basket. It concentrates in packaged goods with a discrete declared quantity, a habitual purchase cycle and a shelf price that shoppers have memorised — precisely the conditions under which a size change is least likely to be noticed and a price change is most likely to be. A household does not have a memorised reference price for a hospital procedure or a car insurance renewal. It has one for a bag of coffee.
The category concentration is visible in both the American and the Canadian evidence, using different methods. The GAO figures come from a comparison of the published index against a research index that holds sizes constant, so they isolate the pure size contribution to category inflation ✓ Established Fact [1]. The Bureau of Labor Statistics results from the earlier window point at the same categories: baby food, snacks, fresh biscuits and rolls and household paper products [2]. Canada names margarine, pasta mixes, cookies and crackers, mozzarella, breakfast cereal and cheddar [3]. Three agencies, three methods, one list.
Counting items understates the exposure, because downsizing is applied to the products people actually buy. In breakfast cereal, 1.1% of items were downsized between 2021 and 2023, but those items accounted for 8.6% of dollar sales in the category ✓ Established Fact [1]. In paper towels, 2.57% of items covered 38.6% of sales [1]. In toilet paper, 2.40% of items covered 10.7% of sales [1]. A shopper reading a headline that says under 3% of products shrank is reading a number that bears almost no relation to the probability that the specific product in their trolley shrank. Firms downsize their volume lines, not their long tail.
The per-unit price consequences on the affected items are large in absolute terms. Among downsized products, the average increase in price per unit ran from 11.6% for paper towels to 32.4% for coffee ✓ Established Fact [1]. A decade of NielsenIQ scanner data covering United States grocery retail from 2010 to 2020 puts the average per-unit-volume increase across all downsizing events at about 12%, with roughly 1.92% of products downsized over the period at more than five times the rate of upsizing ◈ Strong Evidence [4]. In absolute terms that is close to 77,000 individual products, with almost no accompanying reduction in shelf price [32].
Brand structure adds a further asymmetry that has gone almost unremarked. Of the shrinkflation instances Statistics Canada identified between 2021 and 2023, 77.6% occurred in name-brand products against 22.4% in house brands ✓ Established Fact [3]. The gap is not primarily a story about corporate character. Retailer own-label lines compete on an explicitly posted price comparison and are frequently priced against a named competitor product, which makes a silent size cut strategically useless. Manufacturer brands compete on brand equity and shelf familiarity, which is exactly the asset that lets a size change pass unremarked.
A household does not consume the all-items index. It consumes a basket weighted towards the categories in which downsizing is concentrated and away from the categories that dilute it. For a household whose discretionary spending is dominated by groceries, household paper and personal care, the relevant contribution of downsizing is not 0.06 of a point but something between 1.6 and 3.0 points ✓ Established Fact [1]. Reporting the aggregate as the answer is technically correct and practically misleading, and shoppers correctly perceive it as an evasion even when they cannot articulate why.
This is where the credibility damage originates. When a statistical agency reports the all-items effect and a consumer group reports the category effect, both are telling the truth and the public hears a contradiction. The agency has no incentive to lead with the category number because its mandate is the aggregate. The consumer group has no incentive to lead with the aggregate because its mandate is the shelf. Neither is lying, and the resulting public impression — that the official numbers are constructed to understate what people are experiencing — is close to unshakeable once formed [15].
The Inattention Premium
why a smaller box outperforms a higher price
Consumers punish a price increase far harder than they punish an equivalent size reduction, and the gap is large enough to constitute a standing commercial incentive. Across the studies the GAO reviewed, shoppers were about two to five times more responsive to total price increases than to product downsizing ◈ Strong Evidence [1]. That ratio is the entire economic explanation for why the practice exists, and it is not a story about deception so much as about the limits of attention.
The commercial logic is simple enough to state in one sentence: if a firm needs to raise revenue per unit by 10%, it can raise the shelf price by 10% and lose measurable volume, or reduce the contents by 10% and lose considerably less. Everything else in the shrinkflation debate follows from the size of that difference. The empirical work establishing it is now substantial, uses transaction data rather than surveys, and converges on the same asymmetry from several directions.
The most direct estimate comes from a decade of NielsenIQ retail scanner data covering United States grocery stores from 2010 to 2020, published in Marketing Science. A 1% increase in price reduced sales by about 1.19%, while a 1% reduction in size reduced sales by only 0.56% ◈ Strong Evidence [4]. The elasticity to price is more than double the elasticity to quantity for what is, in purchasing-power terms, an identical change. The authors attribute the gap to limited consumer attention to size information and conclude that downsizing is an effective margin strategy precisely because of it [4].
Consumers were about two to five times more responsive to total price increases than to product downsizing.
— United States Government Accountability Office, GAO-25-107451, July 2025The GAO reached a wider version of the same conclusion from a review of the available literature, reporting the two-to-five-times range and citing one study in which a 1% increase in the total price of an item led to a 1.95% decline in sales while a 1% decrease in size was associated with a slight increase of 0.76% in the number of packages sold ◈ Strong Evidence [1]. The second half of that finding is the uncomfortable one. Under some conditions, shrinking a product does not merely avoid a volume penalty; it moves more units, because households buying to a fixed consumption need simply purchase more packages.
Post-downsizing sales behaviour supports that reading. Analysis of the decade-long United States scanner series found that sales typically rise by around 6% after a product is downsized, which is consistent with shoppers replacing the missing volume without registering why they are running out sooner ◈ Strong Evidence [32]. Nearly 2% of products, roughly 77,000 individual items, were downsized over the period, more than five times as often as products were made larger, and with almost no reduction in price [32]. Firms are not experimenting. They are acting on a well-characterised regularity.
Scanner-data estimates published in Marketing Science put the sales response to a 1% price rise at 1.19% against 0.56% for a 1% size cut across United States grocery retail from 2010 to 2020 ◈ Strong Evidence [4]. The GAO literature review reports a wider range of two to five times, including one study in which downsizing was associated with a 0.76% increase in packages sold [1]. The asymmetry is stable across datasets, periods and methods, and it is the proximate cause of the practice.
It is worth being precise about what this evidence does and does not show. It does not show that shoppers are incapable of detecting a size change, and it does not show that firms conceal net weights, which are printed on the pack and legally required. It shows that under real shopping conditions — a few seconds per item, dozens of items, a memorised price anchor and no memorised quantity anchor — the price is the salient variable and the quantity is not. That is rational inattention, not gullibility, and it is exactly the condition that unit pricing was invented to correct.
The asymmetry also explains the shape of the regulatory response. Every jurisdiction that has acted has acted on salience rather than on the practice itself. France, South Korea and Brazil all require disclosure of the change; none of them prohibits it ✓ Established Fact [11]. The theory of change is that if the size cut is made as visible as a price rise, the demand penalty equalises and the incentive disappears on its own. Whether that theory survives contact with retail compliance is the subject of section seven.
Quality Fade and the Fee at Checkout
the two channels no quantity adjustment can reach
Shrinkflation is the tractable case, because a net weight is a number and a number can be adjusted for. The two adjacent practices are not tractable. Skimpflation — the substitution of cheaper inputs or reduced service at an unchanged price and quantity — leaves no numeric trace in any index ◈ Strong Evidence [23]. Drip pricing moves mandatory charges out of the advertised price and into the final checkout step, where no shelf-price collection method can see them ✓ Established Fact [13].
The Federal Reserve Bank of St. Louis set out the distinction in December 2022 and it has held up. Shrinkflation reduces quantity at a constant price; skimpflation reduces quality at a constant price and quantity ◈ Strong Evidence [23]. The second is harder to measure because quality is not declared on the pack, is frequently not observable before purchase, and in services is often not observable at all until the moment the service fails. A thinner tissue, a diluted concentrate, a hotel that has stopped daily housekeeping and a helpline that now answers in forty minutes rather than four are all real reductions in what a household receives for its money, and none registers as a price change anywhere.
There is now direct experimental evidence on how consumers rank the three strategies, and it runs against the firms. A series of ten preregistered experiments published in the Journal of Consumer Research finds that reductions in product quality trigger substantially stronger negative reactions than either size reductions or straightforward price increases, an effect the authors label the skimpflation penalty ◈ Strong Evidence [5]. The mechanism they identify is perceived transparency: quality cuts are judged less transparent and to affect more central aspects of the product, and communicating the change openly reduces the penalty substantially [5].
That finding inverts the usual assumption about which cost-management strategy is safest. A price rise is the most visible and the most immediately punished, but it is also the most quickly forgiven, because nothing has been concealed. A quality cut is the least visible at the point of sale and the most damaging once detected, because detection carries an inference about intent. Firms optimising for the next quarter choose the least visible option. Firms optimising for the next decade would choose differently, and the aggregate result is a market in which the least trust-destructive strategy is the one least often selected.
Ten preregistered experiments reported in the Journal of Consumer Research find that consumers judge quality reductions substantially more unfair than size reductions or price increases, and are correspondingly less willing to buy ◈ Strong Evidence [5]. The authors attribute the gap to perceived transparency and to the centrality of the attribute being cut, and show that explicit communication of the change materially reduces the penalty [5]. No consumer price index in any country adjusts for the practice, because there is no declared quantity to adjust [23].
The third channel is the one regulators have actually moved on, because it is the one with a clean legal test. Drip pricing advertises a headline figure and then reveals mandatory charges at successive steps of a transaction. The United States Federal Trade Commission Rule on Unfair or Deceptive Fees took effect on 12 May 2025, requiring that all mandatory charges other than shipping and taxes appear in the advertised total price for short-term lodging and live event ticketing ✓ Established Fact [13]. Crucially, the rule does not cap any charge. Sellers may charge what they like; they must disclose it at the start [13].
The Federal Trade Commission fee rule, the United Kingdom total-price requirement and the French labelling order all share one structural property: they change when a consumer learns the price, not what the price is ✓ Established Fact [13]. That is the correct instrument if the harm is informational and the wrong one if the harm is distributional. Every jurisdiction reviewed here has chosen the informational instrument, and every jurisdiction is therefore betting that salience alone will discipline the practice [11].
The United Kingdom went further and attached serious penalties. Drip pricing became unlawful under the Digital Markets, Competition and Consumers Act 2024 when the relevant provisions took effect in April 2025, requiring the total price including all mandatory fees to be shown from the beginning of the transaction, with exposure of up to 10% of global annual turnover ✓ Established Fact [14]. The Competition and Markets Authority opened its first public enforcement actions under the Act on 18 November 2025, alongside more than 100 advisory letters and final price transparency guidance [14]. In one case the authority found that over 80,000 consumers had been shown prices excluding a mandatory booking fee between April and December 2025 [14].
The American enforcement track is running in parallel and widening. In September 2025 the Federal Trade Commission and seven states sued a major live entertainment group over historic drip pricing practices, and state attorneys general have continued independent enforcement into 2026 ✓ Established Fact [29]. In April 2026 the Commission issued an advance notice of proposed rulemaking considering extension of the fee rules to online food delivery, describing delivery platforms as among the most significant sources of drip pricing in the modern economy [29]. Three channels, three regulatory regimes, and one index that adjusts for exactly one of them.
Who Absorbs It
the categories that shrink are the categories the poorest buy
Downsizing concentrates in food, household paper, cleaning products and personal care. Those categories occupy a far larger share of a low-income household budget than a high-income one, which means the category-level contribution of 1.6 to 3.0 percentage points lands unevenly by construction ✓ Established Fact [1]. Experimental United States federal data show low-income households facing roughly 10% higher inflation over time than the highest-income group ◈ Strong Evidence [21].
The incidence question has a clean structural answer and a contested empirical one. The structural answer is arithmetic. If downsizing contributes 3.0 percentage points to household paper inflation and 0.06 points to the aggregate, then the burden falls in proportion to how much of a household budget sits in the affected categories ✓ Established Fact [1]. Food at home, cleaning products and paper goods are close to non-compressible for a household near the bottom of the distribution and are a rounding error for a household near the top. The same practice therefore produces very different effective inflation rates without anyone designing it that way.
The empirical work on inflation inequality supports the structural reading. Experimental distributional price data reviewed by the Federal Reserve Bank of Minneapolis show that low-income households experience roughly 10% higher inflation over time than the highest-income households, driven by the concentration of their spending in necessities whose prices rose fastest ◈ Strong Evidence [21]. Wealthier consumers can also stock up when prices are low, which converts a storage cupboard into a hedging instrument that lower-income households do not have [21]. Downsizing degrades that hedge directly, because the unit being stockpiled is smaller.
The counter-evidence deserves equal weight, because it is strong and it is frequently omitted. The Federal Reserve Bank of Cleveland examined post-pandemic inflation and wage growth together and found that households in the bottom 40% of the income distribution gained roughly 4.5 percentage points of cumulative purchasing power between January 2019 and the end of 2024 — the same gain as the middle 40% and more than the roughly 3.5 points gained by the top 20%, despite facing the highest inflation of the three groups ✓ Established Fact [20]. Higher inflation was more than offset by higher wage growth at the bottom [20].
The Cleveland Federal Reserve finding and the lived experience of a shrinking package are not in conflict, because they measure different objects ✓ Established Fact [20]. Wages caught up with the rate of inflation; they did not restore the pre-2021 price level, and households experience levels rather than rates. A package that is 12% smaller is 12% smaller permanently, and no wage settlement reverses it ◈ Strong Evidence [4]. The recovery in purchasing power is real and it is invisible at the shelf.
European survey evidence shows the perception gradient the structural argument predicts. European Central Bank consumer expectations data consistently record respondents in lower income quintiles reporting higher inflation perceptions and higher expectations than those in upper quintiles ✓ Established Fact [18]. The pattern persists month after month across 2026, alongside a median perceived rate of 3.5% for the twelve months to July 2026 [18]. Research on the perceived inflation wedge finds the same gradient by wealth, with effects most pronounced among lower-wealth households whose financial decisions depend most directly on perceived purchasing power ◈ Strong Evidence [19].
The brand asymmetry compounds the incidence. Because 77.6% of Canadian shrinkflation instances occurred in name brands rather than house brands, the household that can and does trade down to own-label is partially insulated ✓ Established Fact [3]. The household that is brand-loyal, or that shops in a format without a developed own-label range, is not. Meanwhile cheapflation pushes in the opposite direction: the evidence from ten countries shows the prices of cheaper varieties rising faster than expensive ones during the surge, so the trade-down itself carries a penalty ◈ Strong Evidence [6]. Both escape routes are partially blocked.
Japan illustrates the same mechanics under different macro conditions. Teikoku Databank recorded 20,609 food and drink items marked up across the major Japanese food manufacturers in 2025, 64.6% more than the 12,520 items of 2024 ✓ Established Fact [26]. By December 2025 the announced total for 2026 stood at 1,044 items against 6,121 at the same point a year earlier, roughly 40% fewer, with the surveying firm expecting price rises to settle at around a thousand items a month rather than the four-thousand-item waves of the preceding years [26]. Normalisation of the practice, not its reversal, is the pattern in every market examined here.
Twelve Years of Instruments
what the labelling regimes have and have not achieved
Five jurisdictions now regulate package downsizing or hidden fees directly, and the earliest hard evidence on whether the instruments work is not encouraging. A survey of 423 French stores in the first week of the country's labelling obligation found no labelling at all in 95% of them ✓ Established Fact [9]. The instrument was correctly designed, lawfully in force, and almost entirely unobserved.
The European baseline predates the shrinkflation debate by a quarter of a century. Directive 98/6/EC requires traders to indicate both the selling price and the price per unit of measure for products offered to consumers, specifically to enable price comparison, and it was amended and strengthened by Directive 2019/2161 ✓ Established Fact [30]. What European Union law does not contain is any provision addressing downsizing as such; the practice is assessed case by case under the general prohibition on unfair and misleading commercial practices [30]. Unit pricing was the answer Europe adopted long before anyone used the word shrinkflation, and it is a passive answer: the information is on the shelf, and the shopper must use it.
Brazil moved first on active disclosure. Ordinance 392 of the National Consumer Secretariat, issued on 29 September 2021 and in force from March 2022, requires that any reduction in the quantity of a product already on the market be communicated clearly and prominently on the principal label, stating the change in both absolute and percentage terms, for a minimum of 180 days ✓ Established Fact [12]. The scope is broad, the duration is long by international standards, and the obligation sits on the manufacturer rather than the retailer — a design choice that later regimes did not copy and probably should have.
France produced the natural experiment, and the result was a compliance failure rather than a policy failure. The order of 16 April 2024 obliged retail outlets above 400 square metres to post a notice for two months on any prepacked food, drink or household cleaning product whose quantity had fallen while its price per kilogram or per litre had risen ✓ Established Fact [8]. Between 1 and 6 July 2024, the consumer organisation UFC-Que Choisir surveyed 423 stores across France and found no labelling whatsoever in 95% of them, with the information incomplete where it did appear [9]. The exemptions were also wide: stores between 120 and 400 square metres, small food shops, online and drive-through channels, and manufacturers were all outside the scope [8].
South Korea attached a penalty, which is the single design difference most likely to matter. From August 2024 manufacturers of daily necessities — the enumerated list includes milk, coffee, instant noodles, toilet paper and shampoo — must display a downsizing notice on the package, in store or on their website for three months from the day the quantity changes, with a fine of 5 million won for a first violation and 10 million won for a second ✓ Established Fact [10]. The amounts are modest in absolute terms. The mechanism is not the fine but the compliance reporting it forces, and the obligation sits on the manufacturer, who controls the pack, rather than on the retailer, who does not.
| Risk | Severity | Assessment |
|---|---|---|
| Enforcement failure at the shelf | A survey of 423 French stores in the first week of the labelling obligation found no notices in 95% of them [9]. A rule that is lawful, correctly drafted and unobserved changes nothing except public expectations. | |
| Quality reduction remains unmeasured | No index adjusts for skimpflation because no quantity is declared, yet experimental evidence finds it provokes the strongest consumer reaction of the three strategies [5]. | |
| Category effects hidden by aggregation | The headline 0.06 point contribution conceals 3.0 points in household paper products and 2.6 in snacks, and the aggregate is what gets reported [1]. | |
| Statistical agency capacity erosion | Eight of thirteen United States federal statistical agencies have lost at least 16% of purchasing power since 2009, with staff reductions of 20% to 30% [16]. | |
| Legitimacy spillover to all official data | Trust in federal statistics fell from 57% to 50% between June 2025 and February 2026, a decline that does not distinguish between price data and any other series [16]. |
The United States has legislated at state level and failed at federal level. Sixteen states maintain unit pricing laws, and the GAO listed uniform unit pricing as one of five policy options alongside labelling requirements, retail scanner integration, consumer education and the prohibition of deceptive practices ✓ Established Fact [1]. At federal level, S.3819 was introduced on 26 March 2024 and would have directed the Federal Trade Commission, within eighteen months, to treat downsizing without a commensurate price reduction as an unfair or deceptive act, enforceable by the Commission and by state attorneys general [27]. It never received a vote [27].
The strongest evidence for any instrument concerns unit pricing rather than event labelling, and it is modest. Research on mandatory unit price disclosure in the canned tuna market puts the annual consumer benefit at 8.33 million dollars once firms' longer-run strategic price adjustments are accounted for, with part of the gain arising because firms lower prices in response to the disclosure regime ◈ Strong Evidence [28]. Field work on shelf labels finds that more prominent unit price information leads shoppers to select lower unit prices and avoid quantity surcharges, and that consistency of presentation is what determines whether the information is used at all [28]. The instrument works when it is legible. It is very often not legible.
The Trust Recession
the cost that appears in no price index
Between June 2025 and February 2026, the share of Americans expressing trust in federal statistics fell from 57% to 50% ✓ Established Fact [16]. Over a shorter window, June to September 2025, agreement that federal agency statistics are generally accurate fell from 40% to 32% [16]. Among people who actually use federal statistics, trust held near 80% and barely moved [16]. The collapse is in the population that experiences the numbers only as a claim about their own lives.
The argument this report has traced is not really about six hundredths of a percentage point. It is about what happens when a technically correct aggregate is offered as an answer to a question about lived experience, repeatedly, over four years, by institutions whose authority depends on being believed. Shrinkflation is a small inflation event and a large disclosure event, and the institutions best equipped to explain the first have almost no mandate to address the second. The gap has been filled by the assumption that the numbers are constructed to reassure.
Two coherent positions exist and both are held by serious people ⚖ Contested. The measurement position holds that the index does its job, that the aggregate effect is genuinely tiny, and that political attention to shrinkflation, greedflation and junk fees is a displacement of responsibility for inflation that was caused by macroeconomic policy ⚖ Contested [22]. The legitimacy position holds that an index designed to measure the aggregate cost of living cannot, by construction, answer the question households are actually asking, and that dismissing the question as innumeracy is how an information institution loses its constituency [15].
The Measurement Case
Downsizing contributed 0.06 of a percentage point to a 34.5% cumulative price rise over five years, an average of about 0.01 point a year [1].
Quantity adjustment converts every recorded size change into an effective per-unit price change, and has done so for decades in the United States, Canada and elsewhere [2].
Package quantities are printed on the pack and unit prices are posted on the shelf across the European Union and in sixteen United States states [30].
Critics argue the post-pandemic inflation was driven by macroeconomic stimulus, and that shrinkflation and junk fees serve as scapegoats for policy failure [22].
Households in the bottom 40% of the income distribution gained about 4.5 percentage points of cumulative purchasing power from January 2019 to the end of 2024 [20].
The Legitimacy Case
Category contributions of 1.6 to 3.0 percentage points fall on the weekly basket, and the household basket is not the all-items index [1].
74% of European Union consumers report seeing packaged goods shrink without a price cut, a direct observation that no methodological explanation dislodges [7].
Quality reduction and drip pricing leave no declared quantity to adjust, and experimental evidence shows quality cuts provoke the strongest reaction of the three [5].
The perceived inflation wedge widens during stress episodes and closes slowly, because attention to prices persists after the rate itself falls [19].
Trust in federal statistics fell seven points in eight months, while trust among actual users of those statistics held near 80% [16].
The institutional dimension turned acute in 2025. The commissioner of the Bureau of Labor Statistics was dismissed on 1 August 2025 amid allegations that agency data had been manipulated ✓ Established Fact [15]. Analysis published through the Centre for Economic Policy Research found that economic policy uncertainty rose by roughly 127 index points, more than 50%, comparing the week before with the week after, and that on a conservative reading the resulting loss of confidence may have reduced United States gross domestic product by around 20 billion dollars [15]. A 2019 survey had found that 95% of business economists regarded government statistics as important to their work [15].
Once credibility is questioned publicly, rebuilding confidence can take considerable time.
— Centre for Economic Policy Research, When Trust in Official Statistics Declines, 2026The capacity problem underneath is older and less dramatic. The American Statistical Association finds that eight of thirteen federal statistical agencies have lost at least 16% of purchasing power since 2009, with most experiencing staff reductions of 20% to 30% ✓ Established Fact [16]. Survey response rates have fallen across the board, which raises the cost and lowers the precision of every estimate produced [16]. An agency that is asked to measure a faster-moving economy with fewer people and a smaller real budget will produce estimates that are more often revised, and each revision is read by a suspicious public as confirmation.
Net quantities are printed on packs and unit prices are posted on shelves, which supports the view that no material fact is concealed ⚖ Contested [30]. Against that, transaction-data estimates show demand responding roughly twice as strongly to a price change as to an identical size change, which implies the disclosure is not functionally reaching the shopper [4]. A United States federal bill would have resolved the question by statute in 2024 and was never voted on [27]. Five jurisdictions have resolved it in favour of mandatory disclosure without prohibiting the practice [11].
The wider trust environment gives the price argument nowhere soft to land. The 2026 Edelman Trust Barometer places developed markets at the bottom of its trust index for the second consecutive year — Japan at 38, France at 42, Germany and the United Kingdom at 44, Spain at 45, South Korea at 46 and the United States at 47 — against China and the United Arab Emirates at 80, India at 74 and Indonesia at 73 ✓ Established Fact [17]. The barometer attributes the deterioration to record economic anxiety, collapsing optimism, eroding institutional trust and an information crisis, and finds seven in ten respondents unwilling or hesitant to trust people whose values or information sources differ from their own [17].
Package downsizing is measured, it is small in aggregate, it is large in the categories where it occurs, and it is concentrated in the products households buy most often ✓ Established Fact [1]. Its companion practices, quality reduction and drip pricing, are not measured at all, and one of them provokes the strongest consumer reaction of the three ◈ Strong Evidence [5]. Disclosure regimes address salience rather than price, and the first serious compliance evidence shows them failing at the shelf ✓ Established Fact [9]. The durable cost is not the six hundredths of a point. It is that the aggregate was offered as a complete answer, and enough people concluded the aggregate was the point [15].
A practical conclusion follows for statistical agencies rather than for legislators. The instrument that most reliably closes the gap between the measured number and the experienced one is not a labelling mandate — France has demonstrated how those fail — but routine, prominent publication of the category-level downsizing contribution alongside the headline index, in the same release, with the same standing. The GAO already produces the number [1]. The Bureau of Labor Statistics already produces the research index it comes from [2]. Nothing prevents publishing it every month except the institutional habit of leading with the aggregate, and that habit has now cost considerably more than it has saved [16].