Organizations: Tailstate Intelligence Ltd London, United Kingdom · Independent Researcher Antibes, France · Zanista AI Ltd London, United Kingdom · Increase Alpha, LLC Miami, United States
Two old market sayings hold that news is already priced in by the time it is published, and that the rumor is bought while the news is sold. Both are empirical claims about event time: they place the price move associated with a piece of news before and at publication rather than after it. Whether the claims hold, for which kinds of news, and by how much are basic questions about how fast markets absorb public information. We test them on 4.57 million financial news articles covering roughly 3,000 US stocks (2023-2026). A large language model teacher, distilled into a compact classifier through active learning, assigns each article one of 17 event tags and five attributes; articles are clustered into stories to separate first reports from follow-up coverage; and beta-adjusted abnormal returns are measured around the resulting 1.68 million stock-day events, with 364{,}405 neutral-sentiment events as a placebo group. Three results follow. First, the price move associated with news concentrates before and at publication: pooled across all signed events, the cumulative move in the news direction by the close of publication day is 2.8 times its value twenty days later, and for rumor-flagged events the directional return is concentrated around the rumor event, with little incremental return at the subsequent linked non-rumor event. Second, measured against the placebo of comparable stocks, markets underreact to numbers and overreact to stories: quantified fundamental news (earnings, dividends, guidance, analyst actions) keeps drifting in the direction of the news for weeks, while soft story-driven news (launches, macro commentary, leadership) gives back its move. Third, news carries width as well as direction: neutral news raises publication-day volatility. The study also produces a table of measured drift for each event tag, usable as a prior in news-conditioned forecasting models.
Figures & tables
Tag
Stories
Days
Tag
Stories
Days
M&A deal
22,596
12.9
Macro through stock
322,666
1.1
Legal, regulatory
46,846
10.6
Financing, dilution
11,497
0.9
Insider, ownership
119,394
4.0
Product launch
159,672
0.9
Analyst action
99,847
2.5
Competition
32,131
0.8
Leadership
53,794
2.4
Partnership
103,199
0.8
Operations, supply
89,071
2.3
Price commentary
220,451
0.6
Table 1. Story persistence by tag, measured from the clustering: number of stories and mean lifetime in trading days.
Event tag
n
%pos
−5..0
day 0
+1..+5
+6..+20
adj. +6..+20
padj
Capital returns
26,273
90
+0.22
+0.06
-0.01
-0.14
+0.35
< 0.001
Earnings results
107,174
72
+1.41
+0.42
-0.04
-0.05
+0.22
< 0.001
Guidance and outlook
24,828
69
+1.43
+0.56
-0.02
-0.09
+0.13
0.046
Analyst action
134,464
70
+1.12
+0.22
-0.05
-0.13
+0.10
0.012
Price commentary
174,087
61
+1.55
+0.43
-0.02
-0.03
+0.09
0.053
Promotional content
55,230
96
-0.01
0.00
-0.11
-0.43
+0.08
0.274
Table 2. Signed abnormal drift by event tag, sorted by adjusted post-event drift. All columns are percent, in the direction of the news. “%pos” is the share of positive-sentiment events. “adj.” subtracts the baseline of the event’s size bucket before signing (Section 4.3 ); padj is the date-bootstrap p-value of the adjusted days +6..+20 drift.
Figure 1. Cumulative abnormal return in the news direction around publication (day 0), five event tags plus the neutral placebo. Left: raw. Right: net of the background drift. Bands are ±2 date-clustered standard errors. Raw curves show the priced-in signature (the move precedes publication) and near-universal post-news decay tracking the placebo. Adjusted curves separate genuine continuation (earnings, analyst) from genuine reversal (launches, macro), and flatten legal and regulatory news to the baseline.
Sample
n
Rumor day
Rumor → news
News day
+1..+5
+6..+20
All rumor events
17,510
+0.36
-0.09
+0.01
-0.14
-0.06
M&A deal
3,335
+0.24
-0.32
-0.10
-0.16
-0.37
Legal and regulatory
1,452
+0.55
-0.01
-0.01
+0.17
-0.17
Product launch
2,316
+0.13
+0.12
-0.08
-0.35
+0.02
Table 3. Buy the rumor, sell the news. Mean abnormal return in the rumor’s direction (percent) at each stage, for rumor-flagged events with a same-tag non-rumor event within 60 trading days.
Figure 2. The background drift: abnormal returns of stocks receiving neutral news, by window and size group. Quiet stock-days with no recent news drift similarly (Section 5.2 ): the drift is the residual of the single-beta benchmark over this sample, not an effect of publicity, and the placebo adjustment removes it either way.
Figure 3. The asymmetry illusion. Raw post-news drift suggests good news reverses and bad news continues. Once the baseline is subtracted, both collapse to zero: the apparent asymmetry is the baseline interacting with the 73% positive share of news sentiment.
Figure 4. The tag map: adjusted drift over days +6 to +20 in the news direction (bars; pale bars not significant at 5%) against raw drift (circles). Adjustment moves every tag toward zero and reorders them: quantified fundamentals drift, soft news reverses.
Event tag
All events
NEW only
Single-tag days
Teacher labels
Capital returns
+0.35
+0.33
+0.47
+0.34
Earnings results
+0.22
+0.11
+0.17
+0.11
Guidance and outlook
+0.13
+0.06
+0.34
+0.08
Analyst action
+0.10
+0.07
-0.05
+0.08
Legal and regulatory
-0.08
-0.07
-0.25
-0.01
Product launch
-0.18
-0.18
-0.48
-0.14
Table 4. Robustness of the adjusted days +6..+20 drift (percent). Columns: all events; NEW articles only; stock-days with one event tag; events labeled directly by the GPT teacher.
Source group
n
NEW share
−5..0
day 0
day-0 ∣AR∣ ratio
adj. +6..+20
Retail analysis
224,122
50%
+1.07
+0.41
1.28
+0.09
Portals
294,359
55%
+0.92
+0.28
1.22
+0.11
Mainstream media
70,616
49%
+0.86
+0.38
1.26
+0.01
Aggregators
544,809
38%
+0.80
+0.21
1.12
+0.06
Other
475,705
63%
+0.46
+0.17
1.07
-0.12
Press wires
117,517
60%
-0.06
+0.05
1.17
-0.19
Table 5. News source groups. NEW share is the fraction of articles that are the first report of their story; day-0 ∣AR∣ ratio is the publication-day absolute move as a multiple of the stock’s normal day. Events here are (stock, day, tag, source group) aggregates, so a story covered by several source groups appears once per group; counts therefore exceed the number of signed events.
Strategy
Events
Ann. ret.
Sharpe
Sharpe 5 bps
10 bps
20 bps
Fade small-cap launch, partnership
52,149
15.8%
1.35
1.20
1.06
0.77
Short any covered small cap
260,472
15.9%
1.35
1.21
1.07
0.78
Follow legal and regulatory
46,057
4.1%
0.66
0.39
0.12
-0.43
Table 6. Calendar-time portfolios, days +6 to +20, abnormal returns, 2023 to 2026. Net Sharpe ratios charge the stated cost per side.
Figure 5. Cumulative returns of the calendar-time portfolios, each hedged with a beta-sized index position, next to an S&P 500 buy and hold. The sentiment-fading strategy (orange) is indistinguishable from shorting every covered small cap (dashed): the news direction adds nothing.
We examine whether financial news can improve realised volatility forecasting using a parsimonious NLP-based framework that incorporates specialised financial word embeddings alongside general-purpose alternatives. News-only forecasts contain useful predictive information but generally do not outperform strong volatility-history benchmarks. Crucially, combining stock-related news forecasts with a strong volatility-history benchmark lowers forecast losses for several specifications and increases realised utility, providing evidence consistent with forecast complementarity. Performance varies across news types, embedding representations, and volatility regimes. SHAP attributions associate forecast variation with economically interpretable firm-specific and macroeconomic news themes.
Eghbal Rahimikia, Stefan Zohren, Ser-Huang Poon
Alliance Manchester Business School at the University of Manchester · Department of Engineering Science, University of Oxford · Alliance Manchester Business School, University of Manchester
Can financial news reliably predict short-term stock movements? Despite advances in large language models, this question remains unresolved. We revisit this problem using a zero-shot natural language processing framework, investigating whether models can extract actionable signals from financial news without domain-specific training. We design a structured pipeline that combines zero-shot natural language inference with temporal aggregation, explicitly modelling recency and event-dependent impact horizons when integrating information across articles. To address the need for transparency in high-stakes settings, we introduce a multi-layered explainability framework that links predictions to token-level, article-level, and aggregate evidence, and produces grounded natural language rationales. Across multiple models and prediction horizons, we find that zero-shot approaches consistently fail to outperform simple baselines, with particularly weak performance on negative movements, suggesting deeper structural limitations in mapping news sentiment to short-term price dynamics. However, explainability signals reliably distinguish between trustworthy and unreliable predictions, offering practical value even when accuracy is limited. These findings highlight the limits of zero-shot financial NLP and motivate a shift toward decision-support systems that prioritise transparency and uncertainty awareness. Code: https://github.com/alimert05/zero-shot-stock-xai
Ali M Karaoglu, Shreyank N Gowda
School of Computer Science, University of Nottingham, Nottingham, United Kingdom
Machine-learning signals built from financial text treat what institutions say, and what the media repeat, as evidence about value. But whoever shapes a narrative may be trading against it. We study markets with three observable voices: institutional statements (Say), media repetition (Echo) and revealed positioning (Do). We ask when words should be followed and when they should be faded. In a linear-quadratic model of an informed institution that speaks and trades before a partly credulous crowd, talking an asset down while buying it is optimal exactly when φ2<2λk<φ. A distribution-free identity then shows that when the observable Say-Do covariance is negative, words carry negative predictive content and should be faded. For measurement, we derive (i) an exact factorised posterior over which articles are echoes, combining arrival times with embedding similarity; (ii) a return-aligned contrastive objective that attains its bound exactly when squared embedding distances are an increasing affine function of squared outcome distances, with the tightest loss-based certificate of which neighbour rankings survive imperfect training; and (iii) a path-signature statistic for who moved first. In a controlled market with known ground truth, echo sentiment predicts returns with a significantly negative sign in all 29 simulated markets, the rolling Say-Do correlation flags false-alarm events with an AUC of 0.90, and return-aligned embeddings organise headlines by consequence rather than topic. We also report where the tools fail.