Consuming Values

Battle Count: 72

The paper has limited direct relevance to quantitative trading. It studies consumer responses to corporate social stances and firm revenue impacts, which could inform fundamental analysis of consumer-facing companies or ESG-related investment strategies. The stock price analysis finds no clear average impact on returns. The methodology (synthetic DiD, ML-based prediction) is econometric rather than trading-oriented. Potential indirect relevance includes understanding demand shocks to consumer brands and political risk in equity valuations.

Extremely high complexity due to: (1) proprietary transaction data requiring special access agreements; (2) multi-step event identification pipeline combining Google Trends, news scraping, BrandIndex surveys, and LLM queries with manual filtering; (3) XGBoost model trained on >30 million donor records with 1,000 merchant predictors; (4) synthetic DiD with thousands of control units, hyperparameter tuning via rolling out-of-sample validation; (5) wild cluster bootstrap for inference; (6) multiple complementary datasets (YouGov, Nielsen, Numerator, Revelio Labs, CRSP, D&B Hoovers, OpenSecrets, Bonica). The full pipeline requires significant computational resources and domain expertise in both econometrics and ML.

The core transaction data from a major payment card company is proprietary and not publicly available. The analysis requires access to this data under a data agreement. YouGov BrandIndex, Google Trends, ProQuest, Nielsen, Numerator, Revelio Labs, CRSP, and other datasets are commercially available but require subscriptions. The event selection procedure involves manual filtering. The XGBoost model parameters and synthetic DiD hyperparameter tuning are described in detail, but the underlying data cannot be replicated.