Measuring Inflation: CPI vs PCE

Every month, a number comes out that decides whether the Federal Reserve raises rates, whether your landlord's lease escalator kicks in, and whether $2 trillion of Treasury bonds pay out more cash. That number is supposed to answer a simple question: how much more does the same life cost than it did a year ago? Two US government agencies answer it differently, and the gap between their answers — usually a few tenths of a percent, sometimes a full point — is one of the more consequential disagreements in economics.

Think of two people pricing the same grocery basket. One prices exactly what a typical urban shopper bought last month and re-prices those same items this month. The other watches what shoppers actually bought this month, including the cheaper store-brand pasta they switched to when the name brand got expensive. Both are honest measurements of the same economy, but they will not agree, because people change what they buy when prices change, and only one measure lets the basket change with them. The first is roughly the Consumer Price Index (CPI); the second is roughly the Personal Consumption Expenditures price index (PCE).

The mechanics, one piece at a time

Both indices are built the same general way: pick a basket of goods and services, weight each item by how much of spending it represents, price the basket every month, and express the change as a percentage.

$$\text{Index}_t = \sum_i w_i \times \frac{P_{i,t}}{P_{i,0}}$$

In words: for every item $i$ in the basket, take how much its price today, $P_{i,t}$, has moved relative to a base period price $P_{i,0}$, multiply by that item's weight $w_i$ in total spending, and sum across all items. The weights $w_i$ are what differ between CPI and PCE, and they are the whole story.

Scope of spending. CPI is built from a household survey — the Bureau of Labor Statistics asks urban consumers directly what they bought. PCE is built from business-side data — what businesses report selling to consumers, which also captures spending done on behalf of consumers, like employer-paid health insurance premiums and Medicare payments. PCE's basket is therefore broader; healthcare alone is roughly twice the weight in PCE that it carries in CPI.

How weights update. CPI updates its basket weights only every two years, using a fixed basket in between (with partial substitution allowed since 1999). PCE reweights every single month, using whatever businesses actually sold. That is the pasta-shopper story again: PCE lets the basket bend toward cheaper substitutes in real time, which mechanically produces a lower measured inflation rate than a basket that updates slowly, because substitution always looks like "saving money" in the index math even when the shopper is worse off eating store-brand pasta.

Housing weight. Shelter is about a third of CPI, priced through owners' equivalent rent — what a homeowner would pay to rent their own home, estimated by survey. In PCE, housing carries roughly half that weight. Because shelter costs move slowly and with a lag (rent leases reset annually, not monthly), CPI inherits more of that lag and more of that weight, which is a major reason CPI often prints hotter or stickier than PCE during turning points in the housing cycle.

Worked example: the same year, two numbers

Suppose over one year: food +3.0 percent, energy +8.0 percent, shelter +5.5 percent, healthcare +2.0 percent, everything else +2.5 percent.

CPI weights (roughly): food 13 percent, energy 7 percent, shelter 34 percent, healthcare 8 percent, other 38 percent.

$$\text{CPI} = 0.13(3.0) + 0.07(8.0) + 0.34(5.5) + 0.08(2.0) + 0.38(2.5)$$

Step by step: $0.13 \times 3.0 = 0.39$; $0.07 \times 8.0 = 0.56$; $0.34 \times 5.5 = 1.87$; $0.08 \times 2.0 = 0.16$; $0.38 \times 2.5 = 0.95$. Summing: $0.39 + 0.56 + 1.87 + 0.16 + 0.95 = 3.93$, so CPI reads about 3.9 percent.

PCE weights (roughly): food 8 percent, energy 4 percent, shelter 17 percent, healthcare 17 percent, other 54 percent.

$$\text{PCE} = 0.08(3.0) + 0.04(8.0) + 0.17(5.5) + 0.17(2.0) + 0.54(2.5)$$

Step by step: $0.08 \times 3.0 = 0.24$; $0.04 \times 8.0 = 0.32$; $0.17 \times 5.5 = 0.935$; $0.17 \times 2.0 = 0.34$; $0.54 \times 2.5 = 1.35$. Summing: $0.24 + 0.32 + 0.935 + 0.34 + 1.35 = 3.19$, so PCE reads about 3.2 percent.

Same underlying price moves, a 0.7 percentage point gap, purely from how much weight each index puts on the housing-heavy, sticky part of the economy versus the broader, faster-adjusting healthcare and services categories.

CPI PCE

What this means in practice

The Fed's official 2 percent target is a PCE target, specifically core PCE (PCE excluding food and energy, which are volatile and mostly driven by supply shocks the Fed cannot control with interest rates). Traders who watch CPI headlines to predict Fed decisions without translating to PCE terms are reading the wrong gauge — a hot CPI print does not automatically mean the Fed's actual target measure is hot too, especially when the gap is being driven by shelter.

The gap also has direct payout consequences. Treasury Inflation-Protected Securities (TIPS) have their principal adjusted using CPI, not PCE, so a persistent CPI-PCE gap is effectively a persistent difference between what bondholders get paid and what the Fed thinks "true" inflation is running at. Wage contracts, Social Security cost-of-living adjustments, and many commercial leases are also CPI-indexed, which is why CPI, despite not being the Fed's target, remains the number that moves headlines and household budgets.

Key terms

• CPI — Consumer Price Index; household-survey basket, updated every two years, used for TIPS and COLA adjustments.
• PCE — Personal Consumption Expenditures price index; business-side data, reweighted monthly, the Fed's actual policy target.
• Core — either index excluding food and energy, used to strip out volatile supply-driven swings.
• Owners' equivalent rent — CPI's method for pricing owner-occupied housing as if it were rented.
• Substitution bias — the tendency of a fixed, slow-updating basket to overstate inflation because it can't reflect shoppers switching to cheaper alternatives.