1\section{Research design}2\label{sec:method}34\subsection{Clocks, cutoffs and the information set}56Figure~\ref{fig:timeline} summarises the event-time design. For firm $i$ and fiscal quarter $q$ we observe the period end7$E_{i,q}$, the earnings-release date $A_{i,q}$ (the reveal date, with its before-open/after-close timing when known), the810-Q/10-K filing date $F_{i,q}$ and the vendor availability instant $V_{i,q}=F_{i,q}+1$ day. The prediction cutoff is9$c=A_{i,q}-h$ trading days for $h\in\{40,20,10,5,1\}$, with $h=20$ as the preregistered primary horizon. A cutoff is10admissible only if $c<\min(A_{i,q},F_{i,q})$ and $c>A_{i,q-1}$, so the feature window never contains the release of11quarter $q$ and never straddles two reveals. The information set $\mathcal{I}(c)$ contains regular-session bars with12session date $\le c$ and accounting facts with $V_{i,j}\le c$ only; the latest such quarter is $j^*(c)=q-1$ in the13retained sample (we require that the forecaster predicts the quarter immediately following the latest available one).14Twelve leakage guards (future facts, fiscal-end joins, post-cutoff bars, test-period transforms, daylight-saving errors,15identifier drift, split artefacts, release-day bars, overlapping windows across splits, survivorship, ingestion dates,16randomised placebos) are implemented as unit tests in the replication package; the local as-of join was verified live17against the FinData API's \texttt{as\_of} semantics on random entities and instants.1819\begin{figure}[H]20\centering21\includegraphics[width=\textwidth]{figures/fig01_timeline.pdf}22\caption{Event-time design and the no-leakage rule. Cutoffs $c=A_q-h$ precede the earnings release $A_q$; the 20-session23feature window ends at $c$; structured facts of quarter $q$ become available at $V_q=F_q+1$ day. Schematic of the24protocol, not data.}25\label{fig:timeline}26\end{figure}2728A fixed-calendar design (month-end cutoffs, target = the next release within 65 trading days) is used for the pricing29tests and as a robustness check because it does not condition on knowledge of the release date.3031\subsection{Targets}3233The primary outcome is the scaled operating-income surprise $S_{i,q}$ of equation~\eqref{eq:surprise}, with34$\mathbb{E}[X_q]=X_{q-4}+\tfrac14\sum_{k=0}^{3}(X_{q-1-k}-X_{q-5-k})$ computed from quarters available at $c$ and scaled by35the latest available total assets. The binary adverse outcome is $D_{i,q}=\mathbb{1}[S_{i,q}\le-0.5\%]$, i.e., a36shortfall of at least half a percent of assets (roughly a fifth of a typical quarterly return on assets). Secondary37outcomes are the net-income analogue, a year-on-year ROA decline of at least one point, a negative operating-cash-flow38innovation of at least one percent of assets, a liquidity/coverage deterioration (current ratio down 10\% or interest39coverage falling below one), a symmetric placebo $|S_{i,q}|\ge0.5\%$ that a pure volatility channel would predict, and an40exploratory EPS miss against the vendor's estimate (available for \epsEstimateShare\% of releases; vintage unknown).41The continuous surprise, winsorised at the training-period 1st/99th percentiles, is modelled by ridge regression. No42target uses prices.4344\subsection{Features by sampling frequency}4546All market features are derived from the same raw 1-minute bars of the regular session (09:30 to the closing print),47aggregated to 5-minute and 30-minute buckets and to daily bars. Each intraday \emph{family} is computed per session48on bars of the given length and averaged (or summed) over the 20-session window ending at $c$: realized variance,49bipower variation, relative jump, negative-semivariance share, signed-jump variation, realized skewness and kurtosis,50concentration of variance in the largest bar, first-order autocorrelation, intraday drift, opening and closing volume51shares, volume dispersion (Gini), down-bar volume share, share of expected bars present, share of variance on52market-down bars, realized beta to SPY (ratio of summed cross-products), variance ratio against the daily return variance,53and volatility-of-volatility. The 30-minute, 5-minute and 1-minute sets therefore have the \emph{same} \kIntra{} columns.54Daily features (\kDaily{} columns) are 20- and 60-day returns, volatilities, Parkinson range, skewness, kurtosis,55overnight and intraday return components, volume ratios, Amihud illiquidity, zero-return share, dollar volume, maximum56drawdown and bar activity. Point-in-time accounting features (\kFund{} columns) are margins, ROA, accruals, cash,57leverage, current ratio, coverage, capex, R\&D, issuance, revenue growth, size, the last available surprise, the trailing58mean surprise, the trailing adverse rate and the age of the information. Bar statistics are descriptive; nothing is59interpreted as order flow, spread or informed trading.6061\subsection{Baseline ladder and equal budgets}6263\begin{table}[H]64\centering\footnotesize65\caption{Nested information sets}66\label{tab:infosets}67\input{tables/tab04_info_sets}68\end{table}6970Table~\ref{tab:infosets} nests the information sets. B0 is the training base rate (and a shrunk firm-history rate);71B1--B5 are $\ell_2$-penalised logistic regressions whose penalty is tuned on the validation period only; LB1--LB5 are72gradient-boosted trees \citep{ke2017lightgbm} with an identical random-search budget (24 trials, early stopping on the73validation period) for every information set, so that the frequency comparison is a comparison of information sets,74not of capacity. Winsorisation bounds, imputation medians and standardisation constants are estimated on training rows75only. A supplementary set (B5all) concatenates all three intraday resolutions. Neural and text models are not used.7677\subsection{Validation and inference}7879The chronological split trains on cutoffs in 2012--2018, tunes on 2019--2020 and evaluates on 2021--2026, with purging80of training outcomes revealed after the next split's start minus a one-quarter embargo \citep{lopezdeprado2018}.81Expanding-window annual refits from 2016 provide out-of-fold predictions for the panel regressions and the timing and82pricing exercises; a 20\% firm hold-out evaluates transfer to unseen firms. The primary metric is log loss; Brier score,83AUC, PR-AUC and calibration are reported for every model. Paired differences (B5$-$B2, B5$-$B4, B5$-$B3) are assessed84with a block bootstrap over calendar quarters of the reveal date \citep{kunsch1989, politis1994} and with the85Giacomini--White test on quarterly mean loss differentials with Newey--West variance \citep{diebold1995,86giacomini2006}; Holm's correction is applied within each hypothesis family \citep{holm1979}. Panel regressions87\begin{equation}88 S_{i,q}=\alpha_i+\delta_{s(i)\times t(q)}+\beta\,\mathrm{LFDS}_{i,c}+\gamma'X_{i,c}+\varepsilon_{i,q}89 \label{eq:panel}90\end{equation}91use the out-of-fold B5 probability as the Latent Fundamental Deterioration Signal (LFDS), firm and sector-by-quarter92fixed effects, and two-way clustering by firm and calendar quarter \citep{petersen2009, cameron2011}. Mechanism tests93(H3) interact the \emph{incremental} signal $\mathrm{LFDS}^{\mathrm{B5}}-\mathrm{LFDS}^{\mathrm{B2}}$ with ex-ante94terciles of size, illiquidity and leverage formed within each calendar quarter.95