Event response
What happens after the headline?
Upgrades, earnings surprises and news bursts do not behave alike. Compare the average response with the interval and sample behind it.
Jun 2, 2025–Oct 15, 2026
Look at the interval, not just the dot
Dots are average next-day returns against QQQ, signed in the event’s direction. A negative value is a reversal. Lines show 95% bootstrap intervals for the mean.
Analyst upgrades have 45 observations; earnings beats have 39. Earnings misses have only four, so their interval is omitted. Positive news bursts have a much larger sample, but their next-day interval crosses zero.
The response can change with the horizon
| Event | n | 1 hour bp | To close bp | Next day bp | 3 days bp |
|---|---|---|---|---|---|
| Analyst downgrade | 24 | +18.1 | −13.7 | −130.0 | −192.0 |
| Analyst upgrade | 45 | +62.1 | +86.8 | +104.0 | +166.5 |
| Earnings miss | 4 | +138.2 | −46.1 | +104.7 | +44.1 |
| Earnings beat | 39 | −76.2 | −32.0 | −125.0 | −107.1 |
| Negative news burst | 102 | −9.1 | −20.0 | −27.0 | −51.9 |
| Positive news burst | 595 | −8.0 | −16.2 | −17.9 | −39.3 |
These are event-direction excess returns, not raw stock returns. The same event can enter multiple overlapping horizons. No trading costs are included.
Next-day uncertainty and sample size
| Event | n | Mean bp | 95% low | 95% high | In direction |
|---|---|---|---|---|---|
| Analyst downgrade | 24 | −130.0 | −334.8 | +39.5 | 45.8% |
| Analyst upgrade | 45 | +104.0 | +11.2 | +199.4 | 60.0% |
| Earnings miss | 4 | +104.7 | Not estimated | Not estimated | 50.0% |
| Earnings beat | 39 | −125.0 | −227.4 | −25.0 | 33.3% |
| Negative news burst | 102 | −27.0 | −86.3 | +32.8 | 51.0% |
| Positive news burst | 595 | −17.9 | −37.3 | +3.0 | 46.0% |
Intervals use 2,000 bootstrap draws. A confidence interval for a mean is not a prediction interval for one future event. Samples and tests are exploratory.
What this view establishes
- Each response starts at the first 15-minute bar after the event was public, excluding the overnight gap for after-hours events.
- Returns are measured against QQQ and signed in the event’s direction. Negative values mean a reversal, including after negative news.
- Intervals use a 2,000-draw bootstrap. They describe uncertainty in the estimated mean, not the range of individual future outcomes.
- Samples range from 4 earnings misses to 595 positive news bursts. Several tests and overlapping outcomes limit broad conclusions.