Regression Analysis: Effect of AlphaFold 2 Release

Treatment: AlphaFold 2 public release (July 2021)

Panel: Gene-month level, Jan 2020 - Dec 2023

Model Specification

Yit = β × Postt + αi + γt + εit

where:
• Yit = outcome for gene i in month t
• Postt = 1 if t ≥ July 2021
• αi = gene fixed effects
• γt = year-month fixed effects
• Standard errors clustered at gene level

1. Fixed Effects DiD Results

Outcome β (Post) SE 95% CI N R² (within)
n_papers 2.0620 (1172948480069.1499) [-2298979020933.472, 2298979020937.596] 949,824 0.0016
n_papers_newcomer 1.3965 (989998539874.2324) [-1940397138152.099, 1940397138154.892] 949,824 0.0007
n_papers_veteran 0.7021 (468392797057.0117) [-918049882231.041, 918049882232.445] 949,824 0.0019
n_papers_healthcare 0.9580 (624428699138.7983) [-1223880250311.087, 1223880250313.003] 949,824 0.0038
n_papers_education 1.0790 (850644814431.4664) [-1667263836284.595, 1667263836286.753] 949,824 0.0003
n_papers_company 0.0326 (90049604062.4330) [-176497223962.336, 176497223962.401] 949,824 -0.0009
n_top10_q 0.4925 (396983730191.3441) [-778088111174.542, 778088111175.527] 949,824 0.0023
n_unique_diseases 12.3252 (8355436477366.5693) [-16376655495626.152, 16376655495650.801] 949,824 0.0062
n_papers_openalex 3.6255 (1076344105056.0890) [-2109634445906.309, 2109634445913.560] 949,824 -0.0039
n_papers_openalex_newcomer 2.5010 (892808087378.8148) [-1749903851259.976, 1749903851264.978] 949,824 -0.0055

*** p<0.01, ** p<0.05, * p<0.1

FE Coefficients

2. Event Study Results

Event study specification with leads and lags around the treatment date (t = -1 as reference):

Event Studies Combined
Interpretation: The event study plots show the dynamic treatment effects. Coefficients before July 2021 (t < 0) test the parallel trends assumption - they should be close to zero if pre-treatment trends were similar. Coefficients after (t ≥ 0) show the treatment effect evolution.

Individual Event Studies

n_papers

Event Study: n_papers

n_papers_newcomer

Event Study: n_papers_newcomer

n_papers_veteran

Event Study: n_papers_veteran

n_top10_q

Event Study: n_top10_q

n_unique_diseases

Event Study: n_unique_diseases

n_papers_openalex

Event Study: n_papers_openalex

3. Log Specifications (Intensive Margin)

Outcome β % Effect p-value
log(n_papers+1) 0.0474 4.9% 1.0000
log(n_papers_newcomer+1) 0.0426 4.4% 1.0000
log(n_papers_openalex+1) 0.1545 16.7% 1.0000