Research Methodology

1. Event Identification

Three types of GitHub events are analyzed:

• Releases: Tagged version releases (excluding pre-releases)
• Commit Spikes: Unusual increases in commit activity (2+ standard deviations)
• Milestones: Major version releases, first commits, repository creation

2. Expected Returns Calculation

We use the market model to estimate expected returns:

R_it = α_i + β_i × R_mt + ε_it

Where:

• R_it = Stock return for company i at time t
• R_mt = Market return (S&P 500)
• α_i, β_i = OLS regression parameters

Parameters are estimated using 100 trading days (-130 to -31 days before event).

3. Abnormal Returns

AR_it = R_it - E(R_it)

Where E(R_it) is the expected return from the market model.

4. Cumulative Abnormal Returns

CAR_i(t1, t2) = Σ AR_it for t = t1 to t2

We calculate CAR for multiple windows:

• Day 0: Event day only
• Days 0 to +1: Event day plus next day
• Days -1 to +1: Three-day window
• Days -5 to +5: Full event window

Parametric Tests

• t-test: Tests if mean abnormal return differs from zero
• ANOVA: Compares abnormal returns across event types

Non-Parametric Tests

• Sign Test: Tests if median AR equals zero (no distributional assumptions)
• Wilcoxon Signed-Rank: Robust alternative to t-test