CPI vs Core CPI: Why the US Prints 3.4 and 2.5 While Korea Prints 3.1 and 3.4
Korea's August CPI came in at 3.1% while core ran at 3.4%. The US July print was the mirror image: headline 3.4%, core 2.5%. Same word, opposite ordering, and the two central banks are reading those orderings very differently. This is a field guide to what CPI and core CPI actually measure, why the numbers diverge, and what the ordering tells you.
KEY TAKEAWAYS
1. CPI is a spending-weighted basket. Shelter alone is 35.3% of the US index; Korea's index tracks 458 representative items on a 2020=100 base.
2. Core strips food and energy for volatility, not for irrelevance. That leaves 79.0% of the US basket and 309 of Korea's 458 items.
3. In summer 2026, US energy (+14.7% y/y) pushed headline above core; Korea's mobile-fee base effect (+26.7% y/y) pushed core above headline. The central bank whose core sits above headline is the more nervous one.
What the latest prints say
US, July 2026 (BLS, released August 12). CPI-U rose 0.1% m/m and 3.4% y/y. Core (all items less food and energy) rose 0.2% m/m and 2.5% y/y. Energy is up 14.7% over twelve months, gasoline 24.6%, food 3.0%, shelter 3.2%. June's −0.4% headline was the biggest monthly drop since April 2020, and it was an energy number (−5.7%). August CPI lands Friday, September 11.
Korea, August 2026 (Statistics Korea, released September 2). The index reached 120.05 (2020=100), up 3.1% y/y and 0.2% m/m. The core measure excluding food and energy jumped to 3.4% from 2.6% in July, the highest since May 2023. The older Korean core measure excluding farm products and petroleum printed 3.1%. The living-cost index rose 3.2%.
Korean headlines ran with "core at a 3-year-3-month high." True, but look at the composition. A one-month 50% mobile-fee discount in August 2025 dropped out of the base, so mobile phone fees rose 26.7% y/y. Statistics Korea noted that this item carries a larger weight inside the core index than in the headline index, so the base effect hit core harder than headline. The Ministry of Economy and Finance estimates headline would have been about 2.5% without it. Processed food (+1.5%) and services are the parts that will not roll off.
How a CPI basket is built
A consumer price index measures the price change of a fixed basket of goods and services bought by urban households. The mechanics that matter are the weights. A 10% rise in rice and a 10% rise in electricity do not hit a household equally, so each item enters in proportion to its share of spending.
BLS collects prices monthly across 75 urban areas from about 6,000 housing units and roughly 22,000 retail establishments. Relative importance as of June 2026: food 13.5%, energy 7.4%, everything else (core) 79.0%. Inside core, shelter is 35.3% of the whole index, and 25.8 points of that is owners' equivalent rent, a survey-based estimate of what homeowners would charge to rent their own home.
Statistics Korea prices 458 representative items and weights them with household expenditure survey shares. The base is 2020=100, so August's 120.05 means prices are 20.05% above the 2020 average. One structural difference: Korea's CPI includes rent but not owner-occupied housing costs, which is one reason its shelter weight is far smaller than the US figure.
The investor takeaway from the weights is simple. A third of US CPI is housing, and housing moves slowly, so US core moves slowly. In July, a 0.1% shelter increase accounted for roughly two-thirds of the headline gain. That is what "sticky core" means in practice.
Why core drops food and energy
Core exists to separate trend from noise. Food follows weather and energy follows crude, and both can swing 10% in a month. Neither is unimportant; they are just poor signals of where underlying inflation is heading.
March 2026 is the textbook case: headline +0.9% m/m on a 10.9% energy spike, core +0.2%. Multiplying July's y/y rates by relative importance gives a rough decomposition of the 3.4% headline: energy about 1.1 points (14.7% × 7.4%), food about 0.4, core about 2.0. Illustrative arithmetic, not an official contribution table.
Korea publishes two core series. The ex-food-and-energy index follows the OECD convention, removing 149 items and keeping 309 (since 1990). The ex-farm-products-and-petroleum index is a Korea-specific measure that removes only 57 items (non-grain farm products, petroleum, city gas) and keeps 401 (since 1975). The Bank of Korea forecasts and communicates on the first one. The 3.4% vs 3.1% split in August comes entirely from the different exclusion sets.
CPI, PCE and the 0.8-point gap
One more layer for US readers: the Fed's 2% target is defined on the BEA's PCE price index, not CPI. July PCE printed 3.7% headline and 3.3% core, released August 26. That is 0.8 points above core CPI for the same month.
| US CPI | US PCE | Korea CPI | |
|---|---|---|---|
| Compiler | Bureau of Labor Statistics | Bureau of Economic Analysis | Statistics Korea |
| Weight source | Consumer expenditure survey | National accounts, incl. business-side data | Household expenditure survey |
| Shelter | 35.3%, incl. owners' equivalent rent | Lower than CPI | Rent only, no owner-occupied cost |
| Core definition | Ex food and energy (79.0%) | Ex food and energy | Ex food and energy (309 items) / ex farm and petroleum (401 items) |
| Latest core | 2.5% (Jul) | 3.3% (Jul) | 3.4% / 3.1% (Aug) |
| Policy anchor | — | Fed 2% target | BoK 2% target, forecast on ex-food-and-energy |
Historically PCE runs 0.2 to 0.4 points below CPI thanks to a lower housing weight and substitution effects. In 2026 the relationship flipped. Employ America's decomposition attributes much of the core PCE overshoot to portfolio management fees (a lagged proxy for equity prices), tech equipment, healthcare services and airfares. Portfolio fees do not exist in CPI at all; they enter PCE from producer price data. The practical consequence is that a rising stock market lifts core PCE. The same analysts flagged that this category may be revised down in the September 30 annual update.
For traders the sequencing matters. A soft core CPI on CPI day can be reversed two weeks later by a firm core PCE. When a Fed official says "core," assume core PCE.
Reading the ordering
Headline above core (US) points to a supply shock in food or energy. The question for the Fed is second-round effects: does the energy impulse leak into core services? The target range has sat at 3.50–3.75% since December 2025. The July minutes recorded that markets were pricing a hike by September. The September 15–16 meeting includes a new dot plot.
Core above headline (Korea) says the pressure is domestic, in services and manufactured goods, which is the pattern that worries a central bank most. The Bank of Korea hiked in July (to 2.75%) and again on August 27 (to 3.00%), and raised its core inflation forecasts for both 2026 and 2027 to 2.5%. But with the phone-fee base effect doing much of the work in August's 3.4%, the real test is September CPI, released in early October, once the base effect is gone. Whether core settles back into the high 2s will frame the October 22 policy meeting.
What I actually watch
| Date | Release | What matters |
|---|---|---|
| Sep 11 | US August CPI | Core m/m at or below 0.2%; owners' equivalent rent cooling from July's +0.3% |
| Sep 15–16 | FOMC + dot plot | The end-2026 median more than the decision itself |
| Sep 30 | US August PCE + annual revision | Portfolio-fee revision; whether the core PCE–core CPI gap narrows |
| Early Oct | Korea September CPI | Ex-food-and-energy level after the phone-fee base effect rolls off |
| Oct 22 | Bank of Korea | How the board reads September core and the won |
Value chain read-through
| Channel | Mechanism | Indicator |
|---|---|---|
| Discount rate | Core PCE above target delays cuts and compresses growth-stock multiples | Core PCE, 10-year Treasury yield |
| FX | US–Korea rate differential and hiking pace move USD/KRW, which flows into won-denominated exporter earnings | Policy rate gap, USD/KRW |
| End demand | Device price increases show up in core goods CPI and weigh on PC and smartphone replacement | US core goods CPI (+0.8% y/y in July), Korea manufactured goods |
| Base-effect noise | One-off items distort core and invite overreaction | Statistics Korea special-factor notes, MOEF estimates |
Risks to this view
• The contribution arithmetic (energy ≈ 1.1 points, etc.) multiplies relative importance by y/y rates. It is an approximation, not the official contribution table.
• US seasonally adjusted series are revised for up to five years, and PCE faces an annual revision on September 30. Figures in this post can change.
• Korea's "2.5% ex phone-fee" is a Ministry estimate, not an index. The government's fuel price cap is also estimated to have shaved 0.5 points off August headline; when it ends, headline can jump.
• A renewed oil spike tied to the Middle East can flip the headline–core ordering in either country within a month.
Once the August US CPI is out on September 11, I will post a short follow-up reading the print through this framework: weights, then core, then the PCE gap.
Sources: BLS, Consumer Price Index July 2026 (Aug 12, 2026), Tables 1 and A · BEA, Personal Income and Outlays July 2026 (Aug 26, 2026) · Statistics Korea (Korea Data Agency), August 2026 Consumer Price Index (Sep 2, 2026) and core index methodology notes · Ministry of Economy and Finance press release (Sep 2, 2026) · Bank of Korea, Monetary Policy Decision (Aug 27, 2026) and Economic Outlook (Aug 2026) · FOMC Minutes, July 28–29, 2026 · Employ America Core-Cast (Aug 2026)
Comments
Post a Comment