Updated: 4 days ago
U.S. Macroeconomic Analysis and August 2026 Consumer Price Index Forecast: A Comprehensive Research Report
Editor´s Note
This is a living CPI research article, updated through the month as new inflation-relevant developments emerge. It combines weekly macro and market monitoring with XTech’s internal CPI forecast, added ahead of the official BLS release (roughly 10 days before).
This version was last updated September 11, 2026, after the official BLS release. It covers August 1–31 and scores XTech's forecast against the actual print
What changed: a post-release review was added — XTech was correct on all four measures and final consensus missed core month-over-month, though the core call landed through the named risk cases rather than the base case.
XTech's CPI Advanced Forecasting Methodology
XTech's CPI forecasts are produced by a machine learning model trained on historical data, consumer survey data, and bespoke alternative datasets, using a one-step-ahead ("teacher forcing") technique that continuously recalibrates against the most recent actual release.
Two forecasts are published each month: the First Forecast (made available to clients ~20 days before the BLS release, before sell-side consensus forms; headline MoM correlation 87%, MAE 0.001) and the Second and Final Forecast (made available to clients ~5 days before, with more data inputs; correlation 88%, MAE 0.0009).
Both cover headline and core CPI (MoM/YoY); category-level forecasts are also available for Gasoline, Shelter, Food, Transportation, Medical, Used Cars, and others.
Approximate basket weights: Shelter 35% · Commodities 19% · Food 14% · Medical 7% · Transportation 6%. Gasoline is the model's highest-conviction category (correlation 96%); Used Cars & Trucks the most uncertain (correlation 73%).
Why XTech forecasts outperform consensus
Our empirical study benchmarking XTech CPI forecast against the world's leading economists and consensus polls (Nov 2017 – Sep 2025) found:
Metric | XTech Final | Consensus |
Headline CPI MoM directional accuracy | 81.9% | 75.5% |
Headline CPI MoM hit rate (exact bp) | 48.4% | 32.6% |
Headline CPI MAE | 0.0009 | 0.0010 |
Core CPI MoM directional accuracy | 55.3% | 20.2% |
Access the White Paper:
XTech CPI Forecast — August 2026
XTech 4 of 4. Final consensus 3 of 4.
At publication on September 7, XTech differed from consensus on both core measures and took the right side of each. Consensus moved on core year-over-year from 2.5% to 2.4% before the release, arriving at the figure XTech had published four days earlier. It never moved on core month-over-month, which is where it finished wrong.
Post-Release review
The scoreline.
XTech was correct on all four measures. Final consensus was correct on three, missing core month-over-month, where the street said +0.2% and the print was +0.3%.
The divergences were entirely in core. When this forecast was published on September 7, XTech's numbers differed from consensus on both core measures — +0.3% against +0.2% on the month, and 2.4% against 2.5% on the year — and both calls were right.
Consensus closed the year-over-year gap before the release, arriving at the 2.4% published here four days earlier. It never closed the monthly gap.
The energy mechanism held, and BLS said so.
The central claim here was that August's headline would be carried by an arithmetic flip in the gasoline line rather than by anything happening at the pump. Gasoline rose 3.9% seasonally adjusted, and BLS attributed "over one third of the monthly all items increase" to that single index.
This article published that the swing would account for roughly 70% of the acceleration from July's +0.1%. On the actual prints it was about 70%.
The magnitude was light. The monthly-average method used here produced +3.2% against a seasonally adjusted +3.9% — a 0.7 point understatement. Last month the same method landed on the seasonally adjusted figure exactly. It reads direction reliably and magnitude only approximately, and that is worth saying plainly rather than quietly.
The core call was right. The reasoning derived from our surveillance was not.
This surveillance article argued core could not stay at 0.2% because owners' equivalent rent and rent of primary residence would hold the +0.3% pace they set in July. They did not. Both decelerated to +0.2%.
Core printed +0.3% anyway, and it got there through the risk cases rather than the base case. Lodging away from home rose 2.4% — named here as the print's cleanest upside risk — and airline fares rose 2.7%, extending July's +2.2% exactly as the second named risk described. Shelter still printed +0.3% in aggregate, but on lodging rather than on rents. The number was right; the composition was not the one this article described.
What else the forecast called correctly.
Natural gas fell 1.1% and electricity 0.2%, the disinflationary energy offset set out in Section 1. Motor vehicle insurance fell 0.8%, a third consecutive decline and the named downside risk. Core goods rose just 0.1% — tariff pressure is still being absorbed in margins rather than cleared, which is now the second consecutive month this article has made that call.
What nobody had.
Medical care services fell 0.2% after rising 0.6% in July. That reversal appears nowhere in the evidence gathered for this cycle, and it offset part of what lodging and airfares added to core.
A note on what this does and does not show.
One month is one month. What is checkable is narrower and more useful than a scoreline: the disagreement with consensus was stated in advance, in writing, with the mechanism attached — and the mechanism is where the review above finds the fault.
The Fed now has its number.
Core at 0.3% with headline holding at 3.4% is the outcome that strengthens the hawkish case, four days before the Committee meets on September 15–16. The claim that this print would be the decisive input stands.
XTech CPI Forecast — Surveillance Elements August 2026
The energy drag that held July to +0.1% flipped to an energy contribution of similar size: the monthly-average gasoline series swung from a 2.9% drop in July to a 3.2% increase in August—a shift worth roughly two-tenths on headline inflation despite pump prices remaining flat all month.
Meanwhile, the shelter anchor has dissolved; July’s soft print was driven by lodging rather than rent, and both independent market-rent series inflected upward together in August.
The labor market also reversed, pivoting from an inflationary argument into the strongest disinflationary signal in the data after shedding 23,000 jobs and revising away 103,000 more—an argument the Fed declined to make.
Only the tariff channel remains unchanged: accelerating upstream, absorbed at the consumer stage, and still waiting to clear.
Ultimately, the August print is likely to read hot for reasons that are arithmetic rather than structural, arriving just four days before a Federal Reserve Committee that has told markets it is willing to act on exactly this kind of number.
Energy and Global Commodities
August’s energy line flips sign, and that is the arithmetic of this print. Retail gasoline averaged $4.058 a gallon across the month’s five EIA weekly readings against $3.932 across July’s four — a 3.2% increase on the monthly-average basis CPI uses. The same method produced −2.9% for July and matched the seasonally adjusted figure to the decimal. On a 3% basket weight that is roughly a 0.2 percentage point swing on the headline, and nothing else this month is close to it.
The detail that matters is that nothing happened in August. Prices traded between $4.006 and $4.085 all month and finished lower than they started. The increase is a base effect from July’s $3.777 trough in the week to July 6, not a fresh shock.
Crude stayed constrained rather than spiking. EIA’s August 11 outlook raised its estimate of shut-in Middle East production on continuing Strait of Hormuz transit constraints, putting Brent at $85 a barrel for the third quarter and $69 in 2027. Natural gas runs the other way: the Henry Hub third-quarter forecast was cut 50 cents to $2.87, with October inventories heading for their highest pre-winter level since 2016.
Trade Policy and Tariffs
Tariff pressure is accelerating upstream and stalling at the consumer. Nonfuel import prices rose 0.4% in July and 4.5% over twelve months — the largest over-the-year advance since the year ended June 2022. The category split is the mechanism: capital goods rose 0.9% and foods, feeds and beverages 0.9%, while consumer goods excluding autos were unchanged.
The producer data says the same. Final demand PPI was flat in July, but that was energy — final demand less foods, energy and trade services rose 0.4% on the month and 4.7% over the year. The pipeline is filling faster than the shelf.
Core goods CPI at +0.2% in July was the first crack, ending three months of flat or negative readings. Margins are still absorbing most of it, which makes pass-through a question of timing, not of whether.
New actions queue behind that. The 50% Section 338 duties on roughly $20bn of Canadian goods took effect around August 22 after trade talks collapsed, with Canadian retaliation on some 700 U.S. products due September 8. That lands in the final ten days of August’s collection window: it is a September and October story, not this print.
Labor Market and Wages
The labour market is now the strongest disinflationary force in the data, and it is deteriorating faster than the unemployment rate admits. Nonfarm payrolls fell 23,000 in July against an average monthly gain of 34,000 over the prior year, and revisions removed a further 103,000 from May and June — May cut from +129,000 to +63,000, June from +57,000 to +20,000.
The jobless rate held at 4.1%, which understates the change. Labour force participation is 61.4%, down 0.7 percentage points since January. People are leaving the count rather than appearing in it.
Wages followed. Average hourly earnings rose two cents to $37.62, up 3.2% over twelve months, down from the 3.4% employment cost index pace recorded in the second quarter. Wage growth at 3.2% is broadly consistent with 2% inflation once productivity is allowed for.
Labour costs reach CPI through the most labour-intensive services — food away from home, medical care, recreation, personal care — with a two-to-four-quarter lag. So this is a statement about core services in 2027. It is also the argument against a September hike that nobody at Jackson Hole made.
Manufacturing, Logistics, and Supply Chain
The input-cost deceleration stopped in August. The ISM manufacturing prices index held at 71.1, unchanged from July after falling from 82.1 in May and 73.0 in June. Any reading above 50 signals rising input prices, and raw materials have now risen for 23 consecutive months.
The rest of the report softened: headline PMI 54.6 from 55.6, new orders 53.7 from 56.7, employment 51.2 from 52.8. Slowing growth alongside sticky input costs is the least comfortable combination for core goods.
Supplier deliveries lengthened for a ninth straight month at 59.3. Respondents named tariffs and the Strait of Hormuz conflict in the same breath; steel has risen for ten months, with aluminium, copper, semiconductors and resins behind it. Pricing volatility appeared in 57% of negative comments.
Diesel extended rather than reversed. On-highway diesel reached $5.599 a gallon on August 31, up from $5.313 in late July and $1.865 above a year earlier — a 50% annual increase. Diesel barely enters CPI directly; it surfaces in goods and food distribution costs with a one-to-three-month lag. The August level is a fourth-quarter story.
Shelter and Housing
Shelter was the reason core was supposed to keep decelerating. That case weakened on both the official and the market side this month.
July’s shelter index rose 0.1% and BLS attributed roughly two-thirds of the entire monthly increase to it. But the two lines carrying the weight accelerated: owners’ equivalent rent and rent of primary residence both rose 0.3%. The soft aggregate came from lodging away from home at −2.8% — a category near 1% of the basket that mean-reverts.
The market data turned too. Zillow’s typical asking rent reached $1,962, up 2.3% year over year and the fastest pace in more than a year, with single-family rents up 3.0% against 1.7% for multifamily. Apartment List recorded a seventh consecutive monthly increase and a vacancy index of 7.1%, down from February’s 7.3% peak and the first decline since late 2021. August is normally a seasonally soft month.
Concessions still cushion this — 39.8% of listings offered one, against 35.9% a year ago — and CPI shelter lags market rents by nine to twelve months, so the inflection is a 2027 problem. What has gone is the assumption that shelter keeps pulling core down.
Food and Agriculture
Food is the quiet offset this month. USDA’s August 25 update cut the 2026 food-at-home forecast to 2.5% from 2.8% and trimmed beef and veal to 9.8% from 10.7%. All food is projected at 3.0% and food away from home at 3.6%.
Eggs do most of the work at −30.8% for the year as avian-influenza flocks recover — a large disinflationary swing in the category that drove headlines the other way eighteen months ago. Pork is forecast at 0.8% and poultry at 0.5%.
Beef remains the structural exception. Federally inspected beef production fell almost 5% in July, holding wholesale prices up against a cattle herd near a 75-year low. Herd rebuilding takes two to three years, which puts this beyond the reach of monetary policy, and diesel 50% above a year earlier layers distribution cost on top.
The aggregate is not showing it yet. CPI food rose 0.1% in July, food at home actually fell 0.1%, and food away from home rose 0.3% — the labour-intensive half of the basket is the half that is still rising.
Services Inflation and Fed Signaling
Warsh used Jackson Hole to open a door rather than to give guidance. “Inflation is running above our 2 percent target,” he said on August 28, “so the Fed’s predominant focus right now should be on prices.” He cited PCE at 3.7% over twelve months and 4.1% annualised over six, and noted that 49% of PCE components were running above 3%. The operative sentence: “We must be confident that underlying inflation is moving to our objective, clearly and at sufficient speed. Otherwise, we have work to do.”
He rejected forward guidance, preferring “clear market signals, as unfiltered as possible.” Markets duly supplied one: September hike pricing moved to roughly a coin flip. There was no FOMC meeting in August, and the next is September 15–16 — four days after this release. That makes this print the decisive input.
The consumer data splits. University of Michigan sentiment fell 6.3% to 51.7, more than eleven points below a year earlier. Yet year-ahead inflation expectations eased to 4.0% from 4.2%, a fourth consecutive decline, with the five-to-ten-year measure unchanged at 3.3%. Households are reacting to the level of prices, not the rate of change.
High-Volatility CPI Components
The volatile components point past this print, in both directions. Wholesale used-vehicle values are rolling over: the Manheim index fell 1.2% to 207.4 in mid-August, flat against a year earlier, after 211.5 in mid-July and 212.9 in June. Wholesale reaches CPI with roughly a two-month lag, so this lands in October and November.
Caution is warranted on that channel. CPI used cars rose 0.4% in July against soft wholesale data, and this is XTech’s weakest forecast category at 73% correlation. Cox also noted buyers rotating into compacts, up 2.2% year over year, and away from pickups and SUVs — a fuel-cost effect.
Airfares are the opposite risk. The category rose 2.2% in July and sits 25.5% above a year ago, and jet fuel rose again through August. Against that, the seasonal factor already anticipates the usual late-August fare decline.
Lodging away from home is the cleanest upside risk in this print. A −2.8% month in a roughly 1% weight is an outlier, and simple reversion toward zero adds a few basis points to the headline while visibly lifting shelter. Motor vehicle insurance, down 0.3% after −2.0%, cuts the other way.
Our CPI Forecast is based on research
Our empirical study is benchmarking XTech CPI forecast against the world's leading economists and consensus polls.
Comments