EconPlot › Explainers › What "inflation cooled" actually means

Inflation cooled — so why did my bill still go up?

Updated 2026-09-06 · how we source every figure

Cooler inflation is a change in speed, not a change in direction

Prices did not come down; they went up by less than they had been going up. The Consumer Price Index, published by the Bureau of Labor Statistics and known as CPI, stood at 332.813 in July 2026 against 322.169 in July 2025 — the basket cost 3.3 percent more than a year earlier. Nothing in that comparison describes a price falling.

The month-to-month reading points the same way. The index was 332.568 in June 2026 and 332.813 in July 2026, a rise of 0.245 index points. So over the year the basket got more expensive, and over that single month the basket got more expensive too.

The word "cooled" refers only to the rate. Over the year to June 2026 the index rose 3.5 percent; over the year to July 2026 it rose 3.3 percent. Both numbers are positive, and a positive number added to last year's already-higher prices is a bigger bill.

For a typical bill to actually shrink, the index level itself would have to fall — a general decline in the price level, which economists call deflation. That is not what the readings show. A cooling inflation rate is a smaller addition on top of the existing price level, never a subtraction from that level. The highest reading in this series is 333.979 in May 2026, and July 2026's 332.813 sits just below it, which is nothing like a return to the levels of earlier years.

What the 3.3 percent actually measures: one basket, two dates, twelve months apart

The 3.3 percent is the ratio between the cost of one basket of goods and services in July 2026 and the cost of the same basket in July 2025. The Bureau of Labor Statistics builds that basket from what urban consumers actually buy, and the basket includes shelter (which covers rent), food, and motor fuel such as gasoline. The index is a single number standing in for thousands of prices.

The index level itself is not a dollar amount. A reading of 332.813 does not mean anything costs $332.81. The level is only meaningful against another level from the same series — which is why every honest CPI sentence names two dates.

The twelve-month window is wider than the headline suggests. The year to July 2026 runs from August 2025 through July 2026, and the year to June 2026 runs from July 2025 through June 2026. Those two windows share eleven of their twelve months. Because those eleven months are common to both readings, they cancel out of the comparison: the whole move from 3.5 percent to 3.3 percent comes from the two months that differ — the month leaving the back of the window rose by more than the month entering at the front. The move is a fact about those two months, not a description of the year.

The level also moves in both directions inside a rising year. The index reached 333.979 in May 2026, slipped to 332.568 in June 2026 — a fall of 1.411 index points — then rose again to 332.813 in July 2026. A single month of the index going down is normal inside a year of the index going up, which is why a one-month move settles very little on its own.

Your bill is not the average bill, and one price is not the basket

A household's own bill can outrun 3.3 percent because the index is an average across many urban households and across the whole basket, and nobody buys the average basket. A household that spends an unusually large share of its money on the categories that rose fastest experiences more than the headline rate; a household weighted toward categories that rose slowly experiences less. The published number describes the average, and the average is not a promise to anyone.

The reverse trap is just as easy. Eggs — grade A, large, US city average — cost $2.500 per dozen in February 2026 and $2.189 per dozen in July 2026, a fall of $0.311 a dozen. That was not one lucky month: every month from March through July 2026 came in below the February price. A shopper watching only the egg carton in that window would have concluded that prices were falling.

Over exactly the same months, the whole index went the other way: 327.460 in February 2026 to 332.813 in July 2026. One item fell while the basket rose. That is not a contradiction — the headline rate is a summary in which a falling egg price is one small pull against many larger pushes.

So two people can both be right and disagree completely: one points at a cheaper carton of eggs, the other at a bill that grew, and the index agrees that the bill grew while still containing the cheaper eggs.

The Federal Reserve wants 2 percent inflation; consumer prices are running at 3.3 percent

Inflation measured by the CPI is running above the Federal Reserve's stated goal — 3.3 percent against a goal of 2 percent a year, a difference of 1.3 percentage points on those two figures, so prices are rising faster than the central bank wants. The Federal Reserve is the central bank of the United States, and its Federal Open Market Committee sets the target range for the interest rate banks charge each other for overnight loans. In its Statement on Longer-Run Goals and Monetary Policy Strategy, that committee judges inflation at 2 percent to be most consistent with its mandate.

One precision the headline usually skips: the Fed's 2 percent goal is written in terms of the annual change in the price index for personal consumption expenditures, which is a different index from the CPI quoted here. Setting 3.3 against 2 is a useful comparison of scale, not an apples-to-apples subtraction on the Fed's own scorecard.

Why 2 rather than 0? The Fed's own framing is that price stability means inflation low and stable enough that households and businesses can plan without having to think about it. That is a definition about predictability, not about prices holding still.

The tool is interest rates, and the Fed describes the channel in three steps. A higher federal funds rate makes it dearer for banks to borrow from each other, and banks pass that cost on, so mortgages and business loans get more expensive. Households and firms then borrow and buy less. And when fewer buyers are competing for the same goods, sellers cannot raise prices as fast without losing the sale — that is the downward pressure on prices the Fed is aiming for.

Two current readings sit alongside each other, without this page claiming that either one produced the other. The effective federal funds rate was 3.63 percent in August 2026. The average rate on a thirty-year fixed mortgage in Freddie Mac's weekly survey was 6.7 percent in the week of August 27, 2026.

The crucial point is that 2 percent is a goal for the rate, not for the level. If inflation reached exactly 2 percent tomorrow, prices would still be higher than last year — just higher by less.

The index that measures the bill also sets some raises and rents, and a sibling index sets the Social Security adjustment

Cooler inflation tends to produce smaller increases in incomes and payments that are tied to the Consumer Price Index, which is why a cooling number can feel like it takes with one hand what it gives with the other. The Bureau of Labor Statistics documents this use of its own index: cost-of-living adjustments in many union contracts and employer pay reviews reference the CPI, and some leases carry escalation clauses that adjust rent by it.

Social Security works through a close relative of the headline number. The Social Security Administration bases its annual cost-of-living adjustment on the third-quarter average of the CPI for Urban Wage Earners and Clerical Workers, known as CPI-W — the same kind of price measure, but built for wage-earning households rather than all urban ones.

The mechanism is arithmetic, not policy: when the referenced inflation rate is lower, anything that scales with that rate produces a smaller increase. A contract that adds the CPI rate to a wage adds less at 3.3 percent than it did at a higher rate.

Two qualifications keep this from being a rule:

• These arrangements often follow the CPI; they do not always. Many wages are not indexed to anything, and are set by negotiation, budget or market conditions rather than by a published series.
• Different indexed contracts read the index at different moments and over different windows, so two adjustments announced in the same year need not use the same figure.

The result is that one monthly release does double duty. The release is the best available description of how much a household's costs rose, and it is also an input to how much some households' income rises in response.

The series behind this page

The same data the video plots, from the same file. Point at the line to read a value off it.

What this page does not establish

• The 3.3 percent is an average across urban consumers and across the entire basket, so it is nobody's personal inflation rate. A reader who treats it as a prediction of their own bill will be wrong whenever their spending is weighted differently from the average household's — and everyone's is, to some degree.
• One price is not the index, in either direction. Eggs fell from $2.500 a dozen in February 2026 to $2.189 in July 2026 while the whole index rose from 327.460 to 332.813 over the same months, so a reader who generalises from the item they buy most often can conclude that inflation is falling when the measured basket says it is rising.
• The year-to-June and year-to-July readings share eleven of their twelve months. Because those eleven months are common to both, the move from 3.5 percent to 3.3 percent reflects only the two months at the ends of the window — it says nothing about the other eleven, and nothing about the level of prices, which rose in both readings.
• The comparison of 3.3 percent CPI inflation with the Federal Reserve's 2 percent goal is indicative, not exact: the Fed states its goal in terms of the annual change in the price index for personal consumption expenditures, a different index from the CPI used throughout this page. A reader who treats 1.3 points as the Fed's own measured shortfall is using the wrong yardstick.
• The Federal Reserve's description of how interest rates cool prices — higher borrowing costs slow spending, which presses down on prices — is a mechanism, not proof about this particular month. The federal funds rate at 3.63 percent in August 2026 and the mortgage rate at 6.7 percent in the week of August 27, 2026 sit alongside 3.3 percent inflation; nothing here establishes that those rates caused this inflation reading, or that the funds rate set the mortgage rate, rather than something else in the economy doing so.
• The figures come from a local copy of series republished by FRED — the free economic-data service run by the Federal Reserve Bank of St. Louis, which redistributes the Bureau of Labor Statistics data used here — read on 2026-09-03. In that copy July 2026 was the most recent month available and October 2025 has no value at all. A later monthly release, or a revision to a month already published, can move the headline rate after this page was written.

Questions people also ask

Sources

The evidence behind every claim above, as recorded when the piece was made.

inflation 3 3 july 2026

BLS Consumer Price Index for All Urban Consumers via FRED, series CPIAUCSL, local file data/CPIAUCSL.csv: July 2026 = 332.813, July 2025 = 322.169; 332.813 / 322.169 = 1.0330, a 3.3 percent rise over the year, matching the brief. July 2026 is the last row in the file, so 'cooled' describes the current print

inflation 3 5 june 2026

Same series: June 2026 = 332.568, June 2025 = 321.435; 332.568 / 321.435 = 1.0346, rounding to 3.5 percent over the year; 3.5 in the year to June to 3.3 in the year to July is the 'cooled' in the title and stakes scene. Narration and stamps say 'over the year to' so the June figure cannot be heard as a one-month change

cpi definition

BLS CPI handbook: the Consumer Price Index measures the average change in prices paid by urban consumers for a market basket of consumer goods and services, including shelter (rent), food and motor fuel (gasoline); the narration glosses it in plain words and the on-screen line names rent, food and gasoline

cooler is not cheaper

Derived from the same series: the CPI level rose from 322.169 in July 2025 to 332.813 in July 2026 and from 332.568 in June 2026 to 332.813 in July 2026, so prices rose over the year and over the month; the slower yearly rate (3.5 to 3.3) is a change in speed, not a fall in the level. 'Your bill still went up' in the title is this aggregate-index claim

fed target 2 percent

Federal Reserve, Statement on Longer-Run Goals and Monetary Policy Strategy: the FOMC judges inflation at the rate of 2 percent, as measured by the annual change in the price index for personal consumption expenditures, to be most consistent with its mandate. The chart draws the 2 percent line against headline CPI as the brief instructs; the narration says the Fed 'wants 2 percent inflation' and does not claim CPI is the Fed's preferred gauge. Narration and cover stamp both say 'inflation' so the 2 percent cannot be heard as an interest rate

why 2 percent

Same statement: the Fed explains price stability as inflation low and stable enough that households and businesses can make decisions without worrying about it; glossed as 'slow enough that few notice'. No claim is made about why 2 rather than another low number beyond the Fed's own framing

gap meaning

Derived: 3.3 percent CPI inflation against a 2 percent target is 1.3 points above target, so prices are still rising faster than the Fed wants; the narration states the meaning of the gap rather than only the numbers. Every monthly year-over-year reading from January 2024 to July 2026 in data/CPIAUCSL.csv is above 2 percent, which supports 'still'; October 2025 has no value in the file and is left as a gap

fed definition

Federal Reserve Act and federalreserve.gov: the Federal Reserve is the central bank of the United States and its Federal Open Market Committee sets the target range for the federal funds rate; glossed as 'the US central bank that sets interest rates'

rates mechanism

Federal Reserve, 'How does monetary policy influence inflation and employment?' (federalreserve.gov FAQ): a higher federal funds rate raises borrowing costs, which slows spending and puts downward pressure on prices; glossed as 'High rates make loans costlier and slow price rises'

egg prices

BLS Average Price Data via FRED, series APU0000708111 (eggs, grade A, large, per dozen, US city average), local file data/APU0000708111.csv: February 2026 = 2.500, March 2.348, April 2.250, May 2.191, June 2.141, July 2026 = 2.189; every month from March to July 2026 is below February, so the endpoints are not cherry-picked; narration and stamps round to $2.50 and $2.19

basket still rose

Derived from CPIAUCSL over the same window as the egg comparison: February 2026 = 327.460, July 2026 = 332.813, so the whole index rose while the egg series fell, which is the contrast the scene draws between one item and the whole basket

raises tied to cpi

Explanation of common practice, not a measured share: cost-of-living adjustments in many union contracts and employer pay reviews reference the CPI (BLS, 'How the CPI is used'); the narration says 'often follow CPI', not always

rent increases tied to cpi

BLS 'How the CPI is used': the CPI is used to adjust rent in some leases and other contracts with escalation clauses; the narration says 'often follow CPI', not always

social security tied to cpi

Social Security Administration, cost-of-living adjustment (COLA) page: annual COLAs are based on the third-quarter average of the CPI for Urban Wage Earners and Clerical Workers (CPI-W), a sibling BLS index built from the same price data as headline CPI; the narration says the Social Security bump 'often follows CPI' and does not name a COLA figure or claim a specific 2027 adjustment

cooler means smaller increases

Derived from the indexing mechanism above: when the referenced CPI rate is lower, contracts and COLAs that scale with it produce smaller increases; stated as a mechanism with the three items named in the same sentence, not as a forecast of any specific raise

fed rate 3 63

Effective federal funds rate via FRED, series FEDFUNDS, local file data/FEDFUNDS.csv: August 2026 = 3.6 percent, the last row in the file; used only in the ending stamp, labelled as the Fed's base rate so the pair with the home-loan stamp reads as cause and effect. 3.63 is the low of the 2024 to 2026 range in the file, so the narration says the Fed 'holds' rather than that rates 'stay high'

mortgage 6 66

Freddie Mac Primary Mortgage Market Survey via FRED, series MORTGAGE30US, local file data/MORTGAGE30US.csv and the data digest: week of August 27, 2026 = 6.7 percent, the latest week on file, with the prior weeks at 6.67 (August 13) and 6.65 (August 20), so 'stays near 6.7 percent' describes a level that has held; stated as the current level, not a forecast. Re-check for a newer weekly print before render

two sided ending

Scenario framing, not a prediction: 'the Fed can cut' if inflation keeps cooling and 'the Fed holds' if inflation stops cooling are the two directions the Fed's own framework allows; no outcome is stated as fact and no action is recommended to the viewer. The on-screen headline names inflation as the subject of 'keeps cooling' and 'stops cooling'

chart

BLS via FRED, series CPIAUCSL, monthly, year-over-year percent change computed by the builder from January 2024 with the 2 percent target line drawn from the Fed's Statement on Longer-Run Goals; the series ends at 3.3 in July 2026 and shows 3.5 in June 2026

editorial

Explanation only: no advice on spending, borrowing, saving or timing; no prediction stated as fact; no ironic aside, per the humor arm 'none'

punchline and length

Owner rule 2026-09-03 ('we aim for memorable and not for completeness'): the video ends on 'cooler is not cheaper' and every scene after it was cut. Narration 1004 -> 717 characters (78s -> 56s at the measured 12.9 chars/second). No claim was added; cut scenes only removed material.

number discipline

Owner rules 2026-09-03 (at most three numbers, none repeated, at most two significant digits, first slide provokes, last slide is the punch line). Numbers kept: 3.3% and 3.5% (scene 2), the Fed's 2% goal (the chart). The hook no longer states 3.3%; it lands in scene 2 and the punch line repeats none of them.

cover headline length

Owner catches 2026-09-03: cover headlines were being cut off where they ran under the artwork. The Cover component now sizes the headline to the band above the image, and the headlines themselves were cut to one provoking line.

cover is one line

Owner, 2026-09-03: 'we just need catchy top phrases... nobody can read [the rest] in a fast scrolling environment, so just remove them'. The cover is the headline plus the artwork; the stamps and the bottom kicker bar are no longer rendered at all.

two model review

Fixed 2026-09-03 from findings that BOTH reviewers (Claude and DeepSeek, run separately by scripts/validate_script.py) raised on the same line. 'ride on' was a metaphor carrying the only statement of the mechanism.

doccard retired

Owner, 2026-09-03: 'why we need doc card? when we already have another card like this'. The doccard scene type rendered a headline on a bare paper card with no artwork at all, which is newscard minus the picture; it was the only text-only type in the renderer and every project had unused card art sitting beside it. DocCard was deleted from remotion/src/MarketShort.tsx and these scenes are newscard now. NewsCard took over its two extras: the kicker renders above the headline, and a highlight that is a whole sentence not lifted from the headline renders as a caption bar over the artwork instead of painting headline words yellow. Not one word of narration or on-screen text changed, so the script fingerprint and its validation still stand.

sentence caption removed

Owner, 2026-09-05, on the staged drafts: 'the label ... should be removed. Way too many words and too little time to read ... they look like a caption, and is redundant with the caption almost. can we just please remove this type of text.' The caption bar was added on 2026-09-03 so the retired doccard's second sentence had somewhere to live; that was the wrong instinct. A short already burns in word-by-word captions, so a second block of prose over the artwork competes with the text the viewer is reading, in a scene lasting a few seconds. Removed from the renderer and from these scenes: scene 3: Rent, food, gasoline, all in one basket.; scene 5: All three often follow CPI.

no bare comparatives

Owner, 2026-09-05, on the staged draft: 'so remember cooler is not cheaper -> so remember cooler inflation is not cheaper prices. Can we not skip nouns for entire phrase.' The closing line used two bare comparatives as if they were nouns - 'cooler' standing in for cooler inflation and 'cheaper' for cheaper prices - which asks the listener to supply both nouns from memory in the one sentence the video wants them to remember. The on-screen headline already said it in full ('Cooler inflation is not cheaper prices'), so the narration was the shorter of the two and the harder to follow. Both nouns are now spoken.

on screen headlines cut

validations/on-screen-brevity.md, first application, 2026-09-05. Owner: 'Way too many words and too little time to read.' Five headlines were paragraphs in a big font, each a near-verbatim restatement of the same scene's narration, so the frame was reading aloud what the burned-in captions were already spelling out. Cut: scene 3 from 90 characters and two sentences to 35 (the two stamps already carry both figures); scene 4 from 76 to 31, a label rather than the definition; scene 6 from 104 to 44, the narration keeps the rent and Social Security detail; scene 7 from 85 characters and three sentences to 39, which also makes the 'not cheaper' highlight a real substring again. Scene 7's two stamps ('PRICES STILL UP', 'JUST SLOWER') were dropped because the shortened headline now says both. No narration changed and no claim moved - this is entirely a reduction of on-screen text.

scene art replaced

validations/scene-artwork.md, owner review 2026-09-05. The picture must say what the scene says: 'if I look at the image alone I can't tell what it wants to say.' scene 3 reused the cover's picture; scene 4 said 'a basket of everyday goods' and showed one egg carton; scene 6 was a balance scale with a house and coins. A price label was tried first for scene 3 and abandoned: the house style says 'no text', so an image whose meaning lives in a printed number renders as an unreadable smudge. Two identical bags with unequal coin stacks says 'same goods, more money' with no number at all. Owner picked it over a single loaf beside a pile of cash: 'price-a is much better.'

chart axes labelled

validations/chart-axes.md, owner catch 2026-09-05: 'why is there no x-axis and no y-axis labels? add, and add a validation rule. x and y must be clearly labeled.' The renderer drew gridlines with no values on them at all, so every chart in the library showed a shape rather than a measurement - a viewer could not tell a two-point move from a two-hundred-point one. ChartLine now writes the value at the top, quarters and bottom of the plot from yMin/yMax, with a unit from the new yUnit field, and takes an optional xLabel for the horizontal axis. A second defect surfaced with it: xTicks are generated on year changes, so a series that never crosses a year boundary produced NO horizontal labels - which is exactly what happened to the 2026-only jobs chart. build_video.py now falls back to month names in that case.

one reading not two

Owner, 2026-09-05: 'no need to put both June and July numbers together on one slide - both are one year change, and the measurement intervals overlap a lot. just put July numbers.' Right, and it is a measurement point, not only a clutter one: the year to June 2026 and the year to July 2026 share eleven of their twelve months, so setting 3.5 against 3.3 invites the viewer to read a 0.2-point fall as new information about one month when the two windows are almost the same window. The scene now carries one reading, in the narration and in a single stamp. The chart in scene 5 still shows the whole series, which is where a trend belongs.

on screen text swept

validations/on-screen-brevity.md sweep, 2026-09-05. The dominant defect was one habit, not many: the headline restating the scene's own narration, so the card read aloud what the burned-in captions were already spelling out. A headline should be the label a viewer can take in at a glance, saying the angle rather than the sentence. Also fixed: highlights appearing nowhere in their headline (they paint nothing), stamps repeating a figure the headline already carries, and headlines of three sentences. Scene 2's highlight painted nothing; scene 3 and scene 7 were verbatim narration.

yellow belongs to type

Owner, 2026-09-06: 'Pull the yellow prompt.' Every image prompt asked for 'one marker-yellow highlight strip'. In black and white that strip WAS the accent and it worked; in colour the generator started using yellow as a compositional block, so the artwork's yellow competed with the yellow our typography uses to point at a word — two things claiming the same signal on one frame. Yellow now belongs to the type alone. The collage keeps its torn-paper colour blocks, which no longer include a deliberate yellow marker.

no marks in artwork

Owner, 2026-09-06: 'costco card usually have costco on it. is it a fair use?' Naming a company in the narration is ordinary editorial use and this channel depends on it; rendering its mark is a different act, and a generated mark is always a wrong mark - an approximation that misrepresents the logo rather than reporting on it, on top of platform brand complaints that are decided without adjudicating anything. The Costco card was already specified as 'a plain white plastic membership card'. The real exposure was incidental branding nobody had thought of as a brand shot: supermarket shelves, storefronts, packaged goods, and a container port whose boxes carry shipping-line marks. 42 of 63 prompts in the library lacked the clause; every image and video prompt now ends 'no text, no logos, no brand names'.

house style 2026 09 06

Applied the house changes settled on the egg-prices and costco pilots: image prompts moved from newspaper halftone black and white to full-colour magazine collage (saturation 41 -> 84, colourfulness 48 -> 74, against Economics Explained's 91 / 54); no marker-yellow in any prompt, so yellow belongs to the typography alone; and scenes after the stakes card carry the narration captions plus at most one stamp holding a figure, which is what the field's inner slides actually do. The colour-era prompt constraints - pin real colours, state how people are framed, state relative scale - are enforced by validations/scene-artwork.md and repaired by the heal loop below.

silent outro card

Owner, 2026-09-06: 'Can we add a call to action slide at the end ... for users to follow my econplot videos to understand what moves your wallet.' Implemented as the one shape this channel's own rules allow. validations/punchline.md: 'A silent end card (no words) after the punch line is allowed - it is a held frame, not more video', so the punch line still lands last in the NARRATION and nothing follows it aloud. validations/editorial-register.md bans comment-bait specifically ('comment X and I'll send you', 'link in bio for') and says a plain editorial ending is fine - a channel name and what the channel is about is neither bait nor financial advice, which is what STRATEGY.md's 'no call to action' was written to exclude. validations/scene-artwork.md permits shared artwork for 'a channel-level element (a logo, an outro card)', so one card serves every video rather than each generating its own. The line is topic-tied where the topic earns it - 'what moves your mortgage' on the bond video - and the house line otherwise. 1.8 seconds, no narration, no audio.

Related

• Why did egg prices fall so far?