The Adaptation Tax: How Trailing Averages in Policy Tools Systematically Misprice Downturns

Abstract

Trailing averages are embedded in the institutional infrastructure that governs monetary policy, banking regulation, and retirement security. Because a trailing average mathematically converges toward the series it tracks, it absorbs economic shocks into its own definition of "normal" — understating downturns while they happen and overstating them afterward. This paper measures that bias in five institutional systems using only public data, under a pre-registered design with committed decision rules.In our sample, the real-time Hodrick-Prescott inflation trend is revised asymmetrically by 0.60 percentage points between rising- and falling-inflation regimes (block-bootstrap p = 3.0e-03), and the trend itself performs 85.4% of the gap-closing work when inflation deviates from it. The Basel III credit-to-GDP gap absorbed 81.4% of the 2003–2007 credit expansion into trend, an understatement worth an estimated USD 154.73 billion in uncalled countercyclical capital on the primary Y-9C risk-weighted-asset series (USD 163.02 billion on the FDIC fallback). The Social Security COLA lagged actual inflation in 96% of rising-inflation years — a lag the statutory chain makes near-definitional in rising years, measured here to confirm it holds outside rounding, floor, and freeze artifacts. A quarterly application of the same statutory formula reduces tracking error by 38% at approximately zero cumulative cost; that improvement falls short of the pre-registered 50% assertion bar, so the design's own decision rule records the reform as neutral-but-not-better — a measured candidate, not a demonstrated fix. Public pension actuarial smoothing produced a median gap of 20.6% between reported and market asset values in fiscal 2009, with actuarial values exceeding market values in 8 of 8 downturn years. The GDP and CPI asymmetries replicate with the same sign in 5 of 5 and 5 of 5 economies tested (United States, United Kingdom, Euro Area, Japan, South Korea), with cross-country coefficients of variation of 14.0% and 46.8% — dispersion we read as consistent with a property of the filter rather than of any one economy. A conditional Monte Carlo stress exercise, run because the pre-registered falsifier board supported it, estimates the compound mispricing exposure at USD 116.36 per US household per year averaged across all households during moderate stress, concentrated at roughly USD 3,642 per affected homebuyer per year — an absolute-mispricing metric that sums transfers and losses alike (a rising-regime error favors the borrower at the other side's expense; incidence and conditionality are stated in Sections 3.4 and 8). For each in-scope system, the diagnostic that measures the bias is itself the proposed replacement, and each replacement is computable from public data in a spreadsheet.The paper registers three dated, falsifiable forward predictions targeting the next downturn, with explicit falsification criteria and full replication materials.Verified rebuild under the Research-to-Publication Standard v1.9.1: every load-bearing number is registered in a machine-checked ledger (claims.lock) and regenerated on demand by verify.py on hash-pinned inputs, then independently reproduced from a clean checkout; a capped two-round adversarial review under a fix-or-rebut protocol and a public corrections log are committed in the repository. The three registered predictions stand as prospective out-of-sample falsifiers, resolvable by any reader from named public data series on a July 2031 horizon. The paper is licensed CC BY-NC-ND 4.0; the accompanying analysis and verification code is MIT-licensed.