Phase 8: Annual Earnings Signal and Post-Final Date Drift in Japanese Equity Markets, Prospective Time-Series Standardised Unexpected Earnings (TS-SUE) Cross-Sectional IC Test on Tokyo Stock Exchange (2003-2024)

Read full text Summarise with VIRAT
Authors: Sungmin Hong
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This study tests whether an annual Time-Series Standardised Unexpected Earnings (TS-SUE) signal predicts cross-sectional stock returns among Tokyo Stock Exchange (TSE) listed equities, using Compustat's fdateq (Final Date) as the point-in-time anchor and restricting the IC cross-section to the 90-day post-fdateq window. Background: Jinushi (2023, The Japanese Accounting Review, Vol. 13) documents post-earnings drift on the modern TSE and shows the effect attenuates with ownership structure and institutional sophistication. Phase 8 tests whether a TS-SUE signal computed from annual earnings data (iby/cshoc per-share) produces a detectable cross-sectional IC in the post-fdateq drift window, and whether that IC shows directional attenuation between the early (2003–2014) and late (2015–2024) sub-periods consistent with Jinushi's ownership-structure mechanism. Event-anchor Caveat: Compustat Global does not provide an explicit earnings announcement date (rdq equivalent) for Japanese firms. This study uses fdateq (Compustat Final Date) as the closest available PIT anchor, available for 74% of observations. The study therefore measures post-fdateq drift rather than clean post-announcement PEAD. If fdateq lags the actual exchange announcement date, the 90-day IC window may partially capture stale-signal momentum rather than pure drift. This limitation is disclosed prominently in §3 and §9 of the pre-registration document. Signal Construction: Annual TS-SUE is computed as the year-over-year per-share earnings change scaled by the historical standard deviation of such changes, using annual income before extraordinary items (iby) normalised by shares outstanding (cshoc) from Compustat Global g_secd merge. One signal update per fiscal year per firm (fqtr=4 year-end records only). No freshness weight is applied — annual reporting frequency with 70% March fiscal year concentration creates a seasonal artefact under continuous decay weighting. Staleness cutoff: 14 months. PIT anchor: fdateq + 1 business day; fallback datadate + 90 calendar days + 1 business day (universe construction only; fallback firms excluded from IC cross-section). IC restricted to calendar months t where 0 < EOM(t−1) − fdateq ≤ 90 days. Confirmatory window: 2003–2024. Hypotheses. One confirmatory hypothesis (H1) and one exploratory directional hypothesis (H2): H1 (confirmatory): Annual TS-SUE predicts cross-sectional TSE returns in the post-fdateq window (IC > 0). Single pre-specified test; block-bootstrap p(mean IC > 0) ≤ 0.100. Block = 12 months (structural 12-month seasonality from 70% March fiscal year concentration). H2 (exploratory, directional only): IC is weaker in 2015–2024 than in 2003–2014, consistent with Jinushi's institutional-sophistication attenuation mechanism. No significance threshold; reported as directional evidence only. Test Statistic: Cross-sectional Spearman IC between annual TS-SUE rank and next-month adjusted price return in JPY (split/rights-adjusted via ajexdi; dividends omitted). Only firms with observed fdateq (not fallback-estimated) enter the IC cross-section. Subtracting any common benchmark does not change rank ordering under Spearman; there is only one IC. Market-adjusted framing applies to the prospective portfolio section (Phase 8b, future extension) only. IC Structure: The post-fdateq restriction produces a structurally sparse IC time series: dense months (expected June–October, approximately 5 per year) with large cross-sections from March fiscal year firms, interspersed with sparse or NaN months from non-March fiscal year firms only. Block = 12 months in the bootstrap operates on the compact non-NaN series; this is a conservative choice (longer blocks → more conservative inference) and is pre-committed. A block = 5 sensitivity is reported alongside the primary result. Programme Context: Phase 8 is one of two international mechanism-scope studies (Phase 7 Korea, osf.io/semk3; Phase 8 Japan) testing whether the PEAD earnings-underreaction mechanism documented in Phase 4 (US, OSF: ngxud) extends across market structures. Both pre-registrations were drafted in parallel on 2026-03-21 and submitted to OSF in the same session (Phase 7 first, Phase 8 immediately after), establishing that Japan was not added post-hoc after observing Korean results. The programme-level primary criterion is "at least one of the two studies produces a statistically significant result under its pre-specified procedure" (joint false positive rate: 1 − (1−0.05)(1−0.10) = 14.5%, assuming independence).

Fields of Study & Subjects

Financial Markets and Investment Strategies Financial Analysis and Corporate Governance Auditing Earnings Management Governance

Publication details

Publication date
1 January 2026
Language
EN
Document type
Preprint