Event studies measure how markets react to specific events by isolating abnormal returns — the difference between what a stock actually returned and what it was expected to return. The methodology applies across many domains, each with different data requirements and model choices.

At a Glance

Application	Typical Events	Recommended Approach	Time Scale
M&A / Corporate Finance	Merger announcements, spinoffs	Market Model, Fama-French	Daily
Earnings & Accounting	Earnings surprises, restatements	Market Model, GARCH	Daily / Intraday
Regulation & Policy	New laws, sanctions, tax changes	Panel DiD (TWFE, Sun-Abraham)	Daily
ESG & Reputation	Scandals, ESG ratings, product recalls	Market Model, Volume Model	Daily
Macro & Central Bank	Rate decisions, CPI releases	Market Model	Intraday

Mergers & Acquisitions

The classic event study application: does a merger announcement create or destroy shareholder value?

plot_event_study(task_ma, type = "car")

Target firms typically show large positive abnormal returns at announcement; acquirers often show small negative or zero abnormal returns.

# Abnormal returns around the M&A announcement
task_ma$get_ar(event_id = 1)

# A tibble: 16 × 2
   relative_index abnormal_returns
            <int>            <dbl>
 1             -5          0.0183 
 2             -4         -0.00253
 3             -3         -0.0295 
 4             -2          0.0182 
 5             -1          0.0209 
 6              0          0.0401 
 7              1          0.0185 
 8              2         -0.00878
 9              3          0.00609
10              4          0.00457
11              5         -0.00328
12              6         -0.0267 
13              7          0.0154 
14              8          0.00953
15              9         -0.0167 
16             10         -0.0309

Model choice: The Market Model is standard. For large-cap stocks, Fama-French 3-Factor controls for size and value effects.

Earnings Announcements

Earnings surprises are among the most studied events in finance. Positive surprises drive prices up; negative surprises drive them down.

plot_event_study(task_earn, type = "car")

Intraday timing matters. If earnings are released at 4:05 PM, a daily study misses the immediate reaction. Use Intraday Event Studies for precise measurement.

Model choice: Market Model for most cases. GARCH if volatility clustering around announcements is a concern.

Regulation & Policy Changes

When a regulation affects many firms at different times (e.g., staggered adoption of a new law across states), a standard cross-sectional event study can be biased. Panel event studies solve this.

plot_panel_event_study(result)

The flat pre-treatment coefficients confirm parallel trends; the jump at period 0 shows the policy effect.

Staggered treatment? Standard TWFE can be biased. Use the Sun & Abraham estimator for unbiased estimates with heterogeneous treatment timing.

Model choice: Panel DiD — static TWFE for a single treatment effect, dynamic TWFE or Sun-Abraham for event-time coefficients.

ESG, Reputation & Operational Risk

Event studies quantify the cost of scandals, data breaches, product recalls, or ESG rating changes.

Typical findings:

Data breaches: -1% to -5% CAR over 3 days
Product recalls: -2% to -8% CAR depending on severity
ESG downgrades: -0.5% to -2% CAR

Model choice: Market Model for price impact. The Volume Model can capture abnormal trading activity around reputation events. See Customization Examples for extending the framework to operational risk.

Central Bank & Macro Announcements

Fed rate decisions, CPI releases, and employment reports move markets within minutes. Daily data is too coarse — you need intraday precision.

Sub-minute reactions. Central bank announcements are priced in within 5-15 minutes. Use Intraday Event Studies with 1-minute bars and the non-parametric Rinaudo & Saha test.

Model choice: Market Model on intraday data with 60-120 minute estimation windows.

Which Approach Should I Use?

flowchart TD
    A[Start] --> B{Multiple firms treated<br>at different times?}
    B -->|Yes| C[Panel DiD<br>Sun-Abraham / TWFE]
    B -->|No| D{Need sub-daily<br>precision?}
    D -->|Yes| E[Intraday Event Study]
    D -->|No| F{Long horizon<br>> 12 months?}
    F -->|Yes| G[BHAR Model]
    F -->|No| H[Cross-Sectional<br>Event Study]
    H --> I{Volatility clustering<br>around event?}
    I -->|Yes| J[GARCH Model]
    I -->|No| K[Market Model or<br>Fama-French]

Next Steps

Getting Started — quick start guide with R code
Expected Return Models — all 13 models explained
Test Statistics — choosing the right significance test
Diagnostics & Export — validate and publish your results