KASPR Datahaus · Research Brief

When the Internet Slows,
Prices Follow

Day-to-day internet congestion measurably delays how quickly stock prices absorb public news — creating systematic, predictable patterns that sophisticated investors can exploit.

23% reduction in same-day positive-news price impact on high-latency days
10× larger effect for small-cap vs. large-cap stocks
1–2 trading days for delayed price discovery to complete
Explore the research
01 — The Signal · 2020

The pandemic exposed a hidden vulnerability in market infrastructure

KASPR Datahaus measures round-trip time (RTT) continuously across millions of U.S. IP addresses. The cross-sectional variance — LatencyVar — captures how unevenly internet quality is distributed on any given day.

62

Peak LatencyVar index reading, March 2020. The COVID-19 lockdown triggered the largest latency spike in the sample — 70% above the high-latency threshold of 36.3.

01 — The Signal · 2021

High-latency days persisted well beyond the initial shock

Through 2021 multiple distinct spikes recurred — driven by infrastructure stress from sustained remote-work demand, winter weather events, and shifting network load patterns. This was not a COVID artefact.

4

months in 2021 where mean LatencyVar exceeded the P75 threshold — including a sharp November spike reaching 41, nearly as elevated as the 2020 COVID period.

01 — The Signal · 2022

2022: the most persistently congested year in the sample

Six separate months in 2022 exceeded the P75 threshold, with Q4 seeing the most sustained elevation — coinciding with peak Fed rate-hiking intensity and heightened financial news volume.

47

Peak LatencyVar reading, December 2022 — the highest reading outside the COVID period and evidence that market structure frictions intensify during macroeconomic stress.

01 — The Signal · Full Sample

Across four years: high-latency days are a permanent feature

Orange shading marks days above the P75 threshold across all four years of the sample. High-latency episodes are distributed across every market regime — bull, bear, and volatile — confirming this is a structural property of U.S. internet infrastructure, not an outlier phenomenon.

25%

of all trading days in the 2020–2023 sample qualify as "high-latency days" by construction of the P75 threshold — roughly one in every four sessions.

U.S. Internet Latency Dispersion · 2020

LatencyVar: cross-sectional variance of RTT across U.S. Autonomous Systems, normalised to historical mean. Orange shading = days above P75 threshold (36.3). Source: KASPR Datahaus.


02 — The Finding

On a normal day, news moves prices exactly as expected

Using a firm-day panel of U.S. equities, RavenPack news sentiment, and KASPR latency data (2020–2023), the baseline is clean:

+1.95%

same-day return on positive-news days under low latency. The negative-news mirror effect is −1.79pp.

The Attenuation — Positive News

High latency silences 23% of positive news

When LatencyVar exceeds the P75 threshold, the same-day positive-news return falls from +1.95% to just +1.51% — a 0.44pp reduction.

−23%

attenuation in same-day impact. Significant at p<0.01, robust to excluding COVID-19.

The Attenuation — Negative News

Negative news is muted too — but less so

High latency attenuates the negative-news return by 0.25pp — a 14% reduction. The 23% vs 14% asymmetry is consistent with institutional sell-side desks partially absorbing the delay.

−14%

attenuation in same-day negative-news impact.

The Volume Confirmation

Retail trading volume confirms the channel

On high-latency days, retail buy volume on positive-news events is significantly dampened — direct evidence that congestion physically delays retail investors from acting on good news.

Retail sell volume on negative-news days is amplified — consistent with over-weighting of negative signals under noisy transmission.

Baseline: Same-Day News Return Sensitivity

High-dimensional FE regressions, 2020–2023 U.S. equity panel. SE clustered two-way at stock-year and date.


Delayed news doesn't disappear — it arrives late

When latency mutes day-zero price adjustment, the information isn't lost — it propagates over the next one to two sessions. The high-latency group shows a compressed day-zero return followed by persistent drift in the direction of the news.

Positive News Events — Cumulative Abnormal Return

Low-latency (blue/solid) shows strong day-zero jump then partial reversal. High-latency (light blue/dashed) shows muted day-zero reaction and upward drift over t+1 to t+3. CARs anchored at day −1 = 0. Source: author calculations, 2020–2023.

Negative News Events — Cumulative Abnormal Return

Low-latency (red/solid) shows sharp day-zero decline then partial reversal. High-latency (salmon/dashed) shows attenuated day-zero drop and continued downward drift. CARs anchored at day −1 = 0. Source: author calculations, 2020–2023.

68% recovery in two trading days for positive news. Delayed, not permanently lost price discovery.

Negative news continues to drift after day zero in the high-latency group — investors act on delayed bad news in the next session.

Investor implication: high-latency days create a predictable short-term continuation signal in both directions.


Latency Attenuation by Firm Segment · Positive News (pp)

Reduction in same-day positive-news return (pp) on high-latency days, by firm segment. Subgroup regressions within the main specification.

Small-cap and retail-heavy stocks bear the brunt

The latency effect concentrates precisely where retail investors dominate and institutional co-location provides the least offset.

For small-cap stocks (bottom market-cap quintile), attenuation reaches ~1.5pp — roughly 10× larger than large-cap firms.

Stocks with above-median retail participation show 0.32pp attenuation. For low-retail stocks, the effect is near zero and statistically insignificant.

Bid-ask spreads: the normal spread-tightening after positive news is dampened under high latency — market microstructure is distorted beyond return means.


Three data sources. One novel finding.

The research merges three firm-day datasets across the U.S. equity market 2020–2023. Identification relies on LatencyVar variation orthogonal to firm characteristics and news incidence.

📡

KASPR Latency Data

Continuous RTT measurements across millions of U.S. IP addresses, aggregated daily to LatencyVar — cross-sectional variance across U.S. Autonomous Systems.

📰

RavenPack News Analytics

Machine-readable firm-day sentiment from 19,000+ sources in real time. Binary and continuous measures of positive and negative firm-specific news shocks.

📈

CRSP Equity Returns

Daily U.S. stock returns, trading volume, retail buy/sell fractions, and bid-ask spreads. Two-way clustered SE at stock-year and date level.

Placebo validation: a non-news-day placebo assigns random shocks on days with no RNA coverage. The latency interaction is precisely zero — confirming the effect is specific to actual news events. Results hold excluding COVID-19, with continuous latency measures, and under alternative clustering structures.


Get the Data

Turn latency into alpha

KASPR's daily LatencyVar data is the input signal behind this research. Contact us to discuss integration into your systematic or discretionary investment process.

Request Data Access Learn More →

Ackermann, K., Angus, S., Cui, B., & Raschky, P.A. (2026). News, Latency, and Stock Prices. Monash University / SoDa Labs.

Based on U.S. equity panel 2020–2023. Past patterns do not guarantee future results. For institutional and professional investor use only.