Hackathon — NGO Campaigns & Responsible Investment
Two weeks, real research data, a menu of tasks: do stock markets react when NGOs campaign against banks? Event studies, sentiment analysis, and sustainable finance in one project.
What is a Hackathon?
Not a normal lab. A menu of tasks, a 2-week deadline, and real research data — the kind used in actual academic work. It counts as a mandatory PC Lab, and replaces the Week 6 lecture.
The research question
Do markets react to NGO press campaigns?
When Greenpeace or another NGO publicly targets a bank — for financing fossil fuels, for laundering, for anything — does the bank’s stock price care? You have the Sigwatch database of NGO campaigns and Datastream stock data for US, UK, and EU banks. The method is the event study of Krueger (2015, JFE): estimate each bank’s CAPM parameters on clean data, then measure cumulative abnormal returns (CARs) around campaign dates. Krueger’s headline number: a negative CSR event costs shareholders on average ~$76 million.
The assignment slides also give you a compact lecture on sustainable finance — SRI, ESG ratings, greenwashing, the ESG-efficient frontier — the economics your results will speak to.
The data
- NGO campaigns:
ngo_data_bocconi.zip(Sigwatch, 2010–2018, Stata files) — distributed on BBoard (too large to host here). The Sigwatch data note explains the variables. - Banks & market data (download above): bank prices/returns from Datastream (
banks_ri,banks_pi), bank info (ISIN, country), Fama–French factors for US and EU, risk-free rates. - A worked name-cleaning example (matching NGO-campaign firm names to Datastream names is half the battle).
Your assignment
Day 1 — Explore the NGO campaign data
Import and append all Sigwatch years into one dataset; keep only campaigns targeting US, UK, EU banks. Describe what you have: how many campaigns (positive vs. negative)? How many NGOs? Which firms are targeted most? What do prominence, ngo_power, and sentiment measure? Any striking pattern?
The menu — pick one or many (difficulty in “+“s)
- Task 1 (++) — Media coverage: did major news outlets echo each campaign? (Needs a newspaper list and translation.)
- Task 2 (++) — Social media coverage: did Twitter/Reddit/Facebook talk about it? (Translation + sentiment analysis.)
- Task 3 (++) — Better sentiment: use each campaign’s link to build your own sentiment measure.
Day 2 — The event study
- Task 4 (+) — Estimate CAPM parameters: rolling-window CAPM (or better, FF-5) per bank. Careful: the risk-free rate differs between US and EU.
- Task 5 (++) — Compute CARs: merge Sigwatch with the stock panel to define event dates; compute CARs in a (−10, +10) window around campaigns (try other windows too).
- Task 6 (+++) — Do NGO campaigns matter?: statistical tests of the impact; split by period (before/after COP21), campaign type, prominence, sentiment. (Tasks 4–5 are prerequisites.)
- Task 7 (+) — “the dessert”: bring in political preferences — green voting data across the EU. Do markets react differently where voters care about climate?
Deliverables
- A Jupyter notebook (same format and guidelines as the PC Labs — comments and interpretation are key), plus a 4-slide summary as if presenting to your company’s board: which tasks you chose, what you did, what you found, what you’d do with more time. One plot may summarize your work.
- Deadline: two weeks — Friday, midnight of Week 7.
“You’re working for yourself, not for a grade: please don’t spend too much time, and do what you like.”