US inflation measurement for Federal Reserve policy Report Interpretation

Overview

Goldman Sachs examines how the Fed should measure inflation when commonly used gauges give sharply different signals. It concludes that a combination of inflation measures, adjusted for the broader macroeconomic context, is preferable to relying on PCE or any single statistical measure.

Core views

The report starts from the unusually wide divergence among current inflation gauges: trimmed-mean PCE rose 2.2% over the past year while core PCE rose 3.3%. Goldman Sachs argues that the interpretation also depends on whether analysts take the data at face value or adjust for tariffs, the Iran-war-related oil spike, and statistical mismeasurement. This matters as Chairman Warsh has said he is examining a broader set of inflation data than PCE and a new Fed task force will reassess the inflation framework. Goldman Sachs argues that the appropriate inflation index depends on its purpose. CPI is intended to track prices paid by consumers, while PCE is designed as a deflator for all consumption in GDP statistics. PCE has advantages including monthly expenditure-weight updates to reflect substitution, revisability, and broad household-expenditure coverage. But those design features can conflict with a central bank's need to measure persistent inflation and the prices consumers perceive. Expenditure weights can give too much importance to idiosyncratic components, and PCE includes items consumers do not pay directly or may not observe, including employer-paid healthcare, nonprofit expenditures, and some financial-services fees. The report also notes that quality adjustment can make products such as phones and televisions appear deflationary despite visible sticker-price increases, while insurance-cost conventions can diverge from consumers' inflation experience. To better match central-bank objectives, the authors build a Central Bank PCE measure across the 18 top-level PCE categories. They use LASSO estimation beginning with official PCE weights and alter weights only where robust evidence indicates that a category is especially relevant or irrelevant to two goals: identifying the underlying trend and explaining inflation expectations. Trend relevance is assessed through categories' ability to predict headline inflation one year ahead, correlation with cyclical cost pressure as captured by unemployment, and co-movement with other categories using principal-component analysis. Expectations relevance is estimated against one- and five-year Michigan consumer expectations and the authors' business-expectations tracker; non-market PCE categories with imputed prices are excluded from that exercise. The resulting measure overweights food and rent and underweights energy and cars. It retains some food and energy exposure, making it more volatile than core PCE, but underweights energy and idiosyncratic categories, making it less volatile than headline inflation. During the pandemic-era inflation surge it ran above core PCE in 2022, then somewhat below it by mid-2023. It currently stands at 3.4% year over year, close to core PCE at 3.3%. International practice and recent research reinforce the case for using several gauges. Most foreign central banks target CPI because a monthly PCE equivalent is unavailable; countries also differ materially in their treatment of owners' equivalent rent. The Bank of Canada monitors preferred trimmed and median measures and previously followed a factor-model common-trend measure, though it stopped because real-time performance was unreliable. The report cites research finding that multivariate models using category-level prices can distinguish persistent trends from transitory shocks, and that averaging many trend-oriented measures can forecast inflation better than any one measure. Goldman Sachs therefore proposes that the Fed begin with an average of its Central Bank PCE measure, core PCE, trimmed-mean PCE, median PCE, and multivariate core trend. That five-measure average rose 2.9% over the last year and has run 0.1-0.2 percentage point above core PCE on average since 2000. The authors stress that statistics cannot eliminate the need for judgment: trimming should depend on whether outliers reflect sampling noise or a major development and, if so, how temporary it is. They view trimmed mean as handling current tariff effects, war-driven energy spikes, and AI-related measurement issues relatively well, while cautioning that it did not handle all earlier episodes, including the pandemic, equally well. Their more benign inflation view rests on treating current one-off effects as fading and either estimating and subtracting them or trimming outlier categories; on that basis, the 2% target appears closer than core PCE suggests.

Analysis framework

The report compares the design and latest readings of established inflation measures, explains why their objectives and category weights differ, constructs a central-bank-focused PCE variant, then tests that framework against foreign central-bank practices and academic evidence. It ultimately combines several gauges and applies judgment to current temporary distortions.

Methodology notes

• Other

  LASSO-based reweighting of PCE categories

  The authors start with official PCE expenditure weights and adjust them only where statistical evidence supports a different weight for identifying persistent inflation or explaining inflation expectations.

• Other

  Principal-component and multivariate common-trend analysis

  The report uses category co-movement to assess underlying inflation and discusses dynamic factor-style common-trend measures that separate persistent inflation from transitory category shocks.

Key data

• Trimmed-mean PCE inflation2.2%Year-over-year increase over the last year
• Core PCE inflation3.3%Year-over-year increase over the last year
• Central Bank PCE measure3.4%Current year-over-year reading
• Average of five trend measures2.9%Year-over-year rate over the last year
• CPI inflation3.5%Latest year-over-year reading in the report's comparison table
• Market-based core PCE3.0%Latest year-over-year reading
• Median PCE2.7%Latest year-over-year reading

Impact & implications

The report says a broader inflation dashboard would give the Fed a more balanced view of persistent price pressure and public inflation perceptions. In the current environment, adjusting for fading one-off shocks or trimming outlier categories supports Goldman Sachs' more benign assessment of progress toward the 2% target.

Risks

• Statistical measures can misclassify major economic developments as temporary outliers or fail to handle unusual episodes reliably.
• Judgmental adjustments can be affected by confirmation bias.
• The report notes that trimmed-mean measures did not handle all past challenges, including the pandemic, equally well.

What to watch

• The Fed's inflation task force and any shift toward a broader set of inflation indicators.
• The persistence of tariff effects, war-driven energy-cost spikes, and AI-related statistical mismeasurement.
• Whether the divergence between core PCE, trimmed mean, and other trend measures narrows.