Wikipedia Pageviews
Daily article-view counts for each company's primary Wikipedia page. Abnormal attention is measured as a trailing 60-day z-score, then cross-sectionally ranked within date and sector.
Signal construction
For each trading day t, we compute z = (views_t − μ_{t-60,t-1}) / σ_{t-60,t-1}. Securities with fewer than 20 valid observations in the window are excluded. The z-score is then ranked cross-sectionally within sector on that date. Empirical Bayes shrinkage toward the sector mean is applied for companies with thin coverage.
Limitations & caveats
Company→article mapping requires manual curation. Ambiguous names (e.g., 'Block', 'Target') are handled with documented rules; precision is audited quarterly. Wikipedia pageviews can spike for reasons unrelated to financial news (celebrity death, product launch, sports).
Evaluation results
Rank IC (Spearman) with Newey-West standard errors, walk-forward with 21-day embargo. Multiple-testing correction: Benjamini-Hochberg at q=0.05 across all signal × horizon × cohort tests.