Can earnings surprises truly be predicted in advance? Two ACL 2026 approaches tackle the problem from both the dataset and methodology angles.

Predicting whether a company’s next-quarter earnings report will result in a 'surprise' or a 'disappointment' is among the most valuable—and most difficult—tasks in quantitative investing. Yet there’s an awkward reality: the data historically used for this purpose is either proprietary text available only at great expense, or stripped-down textual transcripts that omit critical contextual cues—such as the CEO’s tone during the earnings call or the key points highlighted in presentation slides. Both papers presented at ACL 2026 share a common insight: rather than continuing to circle in the dead ends of data monopolies and plain-text-only modeling, the field should instead (1) fully open-source and multi-modalize its data, and (2) surgically strip away redundant signals inside models themselves. These two works push the frontier forward—one by *building better data*, the other by *rethinking the method*. __________________________________________________ I. Why Earnings Surprise Prediction Matters So Much An 'earnings surprise' refers to the gap between a company’s actual reported earnings per share (EPS) and the consensus forecast by Wall Street analysts. Once interpreted by the market as 'better-than-expected' or 'worse-than-expected', this gap often triggers sharp stock price movements within days of the announcement—making it one of the most actionable signals for institutional event-driven trading strategies. Accurately anticipating the direction of the next surprise thus yields tangible, real-world returns. But the challenge lies in data access. Primary materials—earnings call audio recordings, verbatim transcripts, and presentation decks—have long been monopolized by a handful of commercial data vendors, sold at premium prices. Most academic and even many industry studies rely solely on plain-text transcripts, leaving audio and slide-based multimodal signals entirely untapped. This severely constrains the potential of state-of-the-art large models on this task. Compounding the issue, earnings call transcripts routinely exceed 5,000 words and are saturated with managerial boilerplate and industry jargon—presenting models with long, high-noise, high-redundancy text. Extracting truly discriminative signals from such environments remains the field’s core bottleneck. __________________________________________________ II. Notable Works Worth Reviewing FinCall-Surprise: Open-sourcing audio-video-text tri-modal data to establish a benchmark for earnings surprise prediction First up is FinCall-Surprise—a large-scale, open-source, multimodal benchmark dataset for earnings surprise prediction, introduced in the paper titled 'FinCall-Surprise: A Large-Scale Multi-modal Benchmark for Earning Surprise Prediction'. What problem does it address? Prior research on earnings surprise prediction has largely relied on expensive, proprietary, text-only data—leaving advanced models unable to leverage audio or presentation slides. The authors directly target this gap: they construct a publicly available dataset that simultaneously includes verbatim transcripts, full-length audio recordings, and corresponding presentation decks—truly embedding the 'multimodal' dimension and providing future models with a fair, standardized evaluation arena.

Macroeconomic indicators for H1 2026 + summary of listed companies’ earnings previews.

The monthly average of new RMB credit fell from approximately ¥2.1227 trillion in H1 2025 to about ¥1.7934 trillion in H1 2026, with April registering a contraction of ¥400.6 billion. Credit deployment remains volatile, and improvements in financing conditions have yet to stably translate into corporate investment or end-user demand. III. Earnings Previews Through the Lens of Macroeconomic Indicators 3.1 Overview Rising PPI coupled with a weak recovery in PMI suggests improvement in industrial commodity prices, certain margin spreads, and supply-constrained segments—consistent with profit improvements observed in resources, petrochemicals, and memory chips. However, sluggish retail sales (social consumption) and markedly negative fixed-asset investment indicate pressure across real estate, home furnishings, automobiles, and traditional manufacturing end markets—resulting in a profit structure characterized by 'upstream/supply-side improvement, but downstream demand-side divergence'. 3.2 Coal Among 14 coal-sector companies sampled: 7 reported losses, 4 forecast growth, 2 forecast declines, and 1 expects to turn profitable. Most remain in weak operating conditions. The PPI turning positive and the manufacturing PMI approaching 50 create a more favorable environment for resource commodity prices and industrial spreads—but weak retail sales and fixed-asset investment signal that downstream demand is not broadly robust. At the firm level: Shanxi Coal International, Gansu Energy Chemical, Shaanxi Coal Industry, and China Shenhua benefit from coal prices, sales volume, or resource advantages; whereas Shuanghuan Technology and Benxi Iron and Steel face dual pressure from rising coal/iron ore input costs and weak downstream pricing. 3.3 Petrochemicals Among 44 petrochemical-sector companies sampled: 14 forecast concurrent growth, 13 report losses, 8 forecast concurrent declines, 6 expect to turn profitable, and 3 forecast growth ≥50%. The sharp turnaround in PPI aligns with statements from Hengli Petrochemical, Rongsheng Petrochemical, and Orient Prosperity Petrochemical citing 'improved industry supply-demand balance and recovered product spread'. However, cases like Jinneng Technology and Shuanghuan Technology illustrate how upstream raw material price hikes—without commensurate downstream price pass-through—can transform PPI gains into cost pressures for downstream players. For petrochemicals, the central theme remains *product–raw-material spread*; sustained monitoring of spread trends is essential. 3.4 Real Estate Among 35 real estate–related companies sampled: 19 report losses, 6 forecast concurrent declines, 5 forecast concurrent growth, 3 expect to turn profitable, and 1 each forecasts growth and decline. H1 2026 fixed-asset investment posted a clear year-on-year decline, and retail sales growth lagged behind the same period in 2025—consistent with announcements across real estate sales, home furnishing, renovation, and construction chains citing 'lower revenue recognition, declining gross margins, weak demand, impairment charges, and debt resolution efforts'. For real estate firms reporting profitability turns or growth, improved profits may stem from debt-resolution gains—not necessarily signaling a recovery in core operations. 3.5 Securities Among 16 securities firms sampled: 8 forecast growth ≥50%, 5 forecast concurrent growth, 1 expects to turn profitable, 1 forecasts a concurrent decline, and 1 reports a loss. The securities sector is primarily driven by trading volumes, margin financing, proprietary trading, wealth management, investment banking, and overall capital-market risk appetite. Firms including Huachuang Cloud Information, GF Securities, Huaxi Securities, Northeast Securities, Yangtze River Securities, and Guotai Haotong cited improvements in proprietary trading, brokerage, wealth management, investment banking, or overall capital-market activity—linking their performance to broader 'bull market' sentiment and heightened capital-market vitality. 3.6 Photovoltaics Among 30 photovoltaic-sector companies sampled: 19 report losses, 7 report positive or turnaround results, and 4 forecast declines. The sector remains dominated by losses, with recurring themes in earnings previews including supply-demand imbalances, persistently low product prices, intense competition, trade frictions, export tax rebate adjustments, production halts, and asset impairments.

Earnings call transcript: Rayonier Advanced Materials Q1 2026 misses EPS expectations

May 6, 2026

Rayonier Advanced Materials reported a Q1 2026 EPS miss of -$1.22, though revenues hit $319 million. Stock rose 2.48% as the company prioritizes its High Purity Cellulose segment, which saw a 17% price increase.

Investing.com

2026-05-05

Rayonier Advanced Materials Reports Q1 Loss of $81 Million

2026-05-06

Rayonier Advanced Materials Q1 2026 Earnings Report

2026-05-06

Rayonier Advanced Materials (RYAM) Reports Q1 Loss, Tops Revenue Estimates

2026-05-06

Earnings call transcript: Rayonier Advanced Materials Q1 2026 misses EPS expectations

2026-05-07

Rayonier Advanced Materials Inc. (NYSE:RYAM) Q1 2026 Earnings Call Transcript