美联储主席凯文·沃什（Kevin Warsh）正重拾半个世纪前阿瑟·伯恩斯（Arthur Burns）面临的挑战，试图精准测算“潜在通胀”（underlying inflation）。沃什将剔除食品和能源的传统核心通胀指标斥为“粗略估计”，转而青睐截尾平均数（trimmed averages）与通胀广度指标；由于8月份移动电话服务价格单月异常跃升近6%，且9月16日通胀数据显示“过多”商品品类价格涨幅超过3%，美联储实施了三年来首次加息。然而，历史警示历历在目：伯恩斯在1970年代石油禁运与粮食短缺期间随意剔除各类所谓“特殊因素”，反而掩盖了全面通胀的真实蔓延。

为了更准确地过滤杂音，经济学界自1994年布赖恩与切凯蒂的研究以来，普遍推崇截尾均值与中位数通胀指标。例如达拉斯联储编制的截尾平均PCE指标，长期采用非对称截断（剔除底部24%与顶部31%的价格变动），但该方法在2021年通胀飙升期间因剪裁掉过多真实的价格上涨而失效；达拉斯联储研究人员随后开发了对称截尾替代模型，该模型将截至今年6月的一年潜在通胀率估算为2.6%，明显低于核心PCE的3.3%。这表明，任何依赖历史价格分布设定的统计修剪规则，在价格变动模式发生剧烈转变时都会产生严重偏差。

尽管沃什倡导引入类似“十亿价格项目”（Billion Prices Project）的实时海量高频微观数据，并坚持聚焦于这十亿价格中排序居中的“第5亿零1个价格”（即中位数），但中位数指标本身存在显著盲区。当相当一部分少数商品价格集中暴涨而尚未触及分布中点时，中位数同样会严重低估潜在通胀风险，正如2021年初期发生的失真那样。没有任何单一统计法则能够完美捕捉通胀全貌，若中央银行家试图通过主观干预来修正模型缺陷，极易重蹈伯恩斯当年以主观意志粉饰通胀的覆辙。

Federal Reserve Chairman Kevin Warsh is confronting the perennial challenge of quantifying “underlying inflation”, echoing Arthur Burns’s fraught tenure in the 1970s. Dismissing conventional core indices that exclude food and energy as mere “rough swags”, Warsh champions trimmed-mean metrics and inflation breadth measures. Spurred by idiosyncratic distortions—such as mobile phone service costs jumping nearly 6% in August alone—and noting that prices rose by more than 3% across “too many” categories, the Fed enacted its first interest rate hike in three years on September 16th. Yet historical precedent urges caution, as Burns famously obscured structural price spirals by arbitrarily categorising oil, food, used cars, and jewellery as transient anomalies.

To construct cleaner inflation signals, central bankers have long leveraged trimmed and median measures, rooted in foundational 1994 research by Michael Bryan and Stephen Cecchetti showing the median CPI outperformed headline and core measures in forecasting inflation. However, static pruning models struggle when structural distributions shift. The Dallas Fed’s trimmed-mean personal consumption expenditures (PCE) gauge historically pruned the bottom 24% and top 31% of price changes, inadvertently filtering out genuine inflationary spikes during the 2021 surge. Researchers recently developed a symmetric trimmed alternative, which recorded underlying inflation at 2.6% for the year to June, compared with 3.3% for standard core PCE, illustrating how rigid pruning rules risk discarding vital price signals.

Seeking enhanced granularity, Warsh advocates harnessing massive real-time datasets akin to the Billion Prices Project, asserting that tracking the “500 million and one” median price will unveil true economic equilibrium. Nevertheless, median metrics harbour critical blind spots: when a sizeable minority of product categories experience violent price surges without immediately shifting the 50th percentile, median gauges substantially understate inflationary momentum, as occurred in 2021. Because no quantitative shortcut offers flawless foresight, policymakers risk repeating Burns’s cardinal error if they substitute rigorous statistical discipline with subjective discretion over which prices to heed or ignore.

Source: How the Fed should measure inflatio

Subtitle: Pinning down the underlying rate may be harder than Kevin Warsh thinks

Dateline: Sep 24th 2026\n