Nomura: Trimmed-Mean PCE Inflation Understates True Level by ~48 bps
AI summary card
Nomura: Trimmed-Mean PCE Inflation Understates True Level by ~48 bps
The report finds that the Federal Reserve’s preferred trimmed-mean PCE inflation measure suffers from a downward bias due to shifts in the skewness of goods inflation distributions, potentially misleading policymakers about inflation stickiness.
- Trimmed-mean PCE inflation lags turning points in inflation trends—particularly during the post-pandemic inflation surge.
- Since the methodology was last updated in 2009, price-change distributions have become more frequently positively skewed, causing the metric to systematically underestimate underlying inflation.
- After adjusting for this bias, trimmed-mean inflation stands at approximately 2.8%, roughly 48 basis points above the official reading.
- The AI investment boom and semiconductor shortages may keep core goods inflation positive for several years, further exacerbating distributional skew.
- Overreliance on this flawed inflation metric increases the risk that the Fed falls 'behind the curve,' necessitating more aggressive rate hikes later.
Report interpretation
Overview
This report provides an in-depth assessment of the alternative inflation metric—trimmed-mean PCE inflation—championed by former Fed official Kevin Warsh. Nomura’s macro team argues that although the metric is designed to filter out noise, its current trimming thresholds are no longer optimal given evolving post-pandemic goods price dynamics, leading to a systematic underestimation of underlying inflation trends. Through historical backtesting and skewness analysis, the report identifies a downward bias of approximately 48 basis points in the trimmed-mean PCE inflation measure. Overreliance on this indicator could mislead policymakers about the true inflation outlook and heighten the risk that monetary policy lags economic reality.
Core views
Lag and limitations of trimmed-mean PCE inflation. Trimmed-mean PCE inflation filters noise by excluding components at both ends of the price-change distribution (currently the bottom 24% and top 31%). However, this mechanism renders it sluggish in adapting to new inflation regimes. During the post-pandemic inflation surge, trimmed-mean inflation clearly lagged behind other 'true' trend inflation measures (e.g., band-pass filtered headline PCE), while core PCE inflation captured accelerating trends more promptly. Similarly, during the disinflationary period following the Global Financial Crisis, it failed to timely reflect downward momentum. Systematic underestimation driven by shifting distributional skew. Since the Dallas Fed last updated its trimming thresholds in 2009, the cross-sectional distribution of monthly price changes has undergone a structural shift—increasingly exhibiting positive (right-tailed) skew. This implies that the upper tail—which is trimmed—contains disproportionately more inflationary drivers, and the trimmed-mean metric excessively discounts these high-inflation components. Nomura’s analysis using Bowley and Kelly skewness coefficients reveals that this skew shift is primarily driven by goods inflation dynamics. Results indicate that the current trimmed-mean inflation reading is approximately 48 basis points lower (annualized year-on-year) than the skew-adjusted value. Structural shift in goods inflation and forward outlook. Core goods inflation—a persistent deflationary force historically (largely due to quality adjustments in electronics)—has shifted to an inflationary driver in the post-pandemic era. With the AI investment boom continuing, critical electronic components such as semiconductors face supply shortages; combined with supply-chain disruptions stemming from the Iran conflict, core goods inflation may remain positive for several years. Moreover, corporate pricing behavior is evolving: manufacturers adjust prices more frequently, and the inventory-sales ratio has strengthened its correlation with goods inflation—indicating that goods inflation is becoming more cyclical. Policy implications and risks. Although dovish officials—including former Fed Chair Jerome Powell—have tended to view goods price increases as transitory, the persistence of positive core goods inflation—and the resulting distributional skew—undermines the relevance of alternative metrics like trimmed-mean inflation. If monetary policy is guided by such suboptimal indicators, the Fed may underestimate inflation stickiness, fall 'behind the curve,' and ultimately be forced to deploy more aggressive rate hikes to rein in inflation.
Analysis framework
The report employs a hybrid methodology combining statistical analysis and historical backtesting. First, it constructs four 'true' inflation trend series based on ex-post data (e.g., 36-month centered moving averages, band-pass filtering) to serve as benchmarks for evaluating the historical tracking performance of trimmed-mean inflation. Second, it quantifies shifts in price-distribution symmetry using Bowley and Kelly skewness coefficients and applies Bai-Perron tests to identify structural breakpoints, linking skew shifts to breakpoints in goods inflation. Finally, following the Cleveland Fed’s methodology, it computes the gap between trimmed-mean and headline PCE inflation, removes outliers, and estimates the magnitude of the systematic bias attributable to skew—thereby quantifying the current degree of underestimation.
Methodology notes
Statistical Bias in Inflation Metrics
By analyzing changes in price-distribution skewness, this approach assesses whether a specific statistical indicator (e.g., trimmed mean) systematically over- or underestimates true values due to shifts in distributional shape. In this report, it reveals how trimmed-mean PCE inflation underestimates true inflation due to thickening of the right tail.
Volume-Price Decomposition
Decomposes overall inflation into components—such as core goods, supercore, and shelter services—to identify specific drivers behind shifts in inflation distribution. The report finds that the shift of core goods inflation from deflationary to inflationary is the primary cause of increased distributional skew.
Key data
- April Trimmed-Mean PCE Inflation2.35%94 basis points below core PCE inflation
- April Core PCE Inflation3.29%The Fed’s long-favored inflation gauge
- Downward Bias in Trimmed-Mean InflationApproximately 48 basis pointsSkew-adjusted trimmed-mean inflation is ~2.8%
- Trimming ThresholdsBottom 24%, Top 31%Current Dallas Fed trimming standards, calibrated on 1977–2009 data
Impact & implications
The report cautions that market interpretations of declining trimmed-mean inflation may be overly optimistic. Given its significant downward bias, underlying inflation pressure is likely stronger than surface-level data suggest. Investors should be wary that the Fed—relying on distorted inflation signals—may maintain an accommodative stance too long, raising the cost of subsequent policy correction (i.e., sharper rate hikes). The stickiness of core goods inflation will be a critical variable in assessing whether inflation is truly coming under control.
Risks
- If the AI investment boom falls short of expectations, core goods inflation could revert downward, weakening positive skew.
- Supply-chain disruptions (e.g., those linked to the Iran conflict) may ease rapidly, reducing goods-price pressures.
- The Fed may adopt alternative inflation metrics better aligned with the current economic structure, thereby recalibrating its policy path.
What to watch
- The persistence of core goods inflation and its impact on overall PCE distributional skew.
- Supply-demand conditions and price trajectories for critical electronic components such as semiconductors.
- The Fed’s evolving weighting of alternative inflation metrics (e.g., trimmed-mean, median CPI).
- Survey data on firms’ pricing frequency and changes in the inventory-sales ratio.