# Data audit

• File: data/firm_panel.csv, SHA-256 ae54378f63f4c9527cb2a1061d96d42eb1e45173bb8762a5998966cbfc9672e0 (matches bundle.json)
• 1200 rows × 4 columns: firm_id (int64), year (int64), x (float64), y (float64); dimensionless; synthetic data.
• Firms: 100 (ids 1001–1100), each with 12 rows; years 2008–2019, 100 rows each → balanced.
• Missing values: none. Exact duplicate rows: 0. Duplicate (firm_id, year) keys: 0.
• x: mean −0.162, sd 1.258, range [−4.52, 3.30]; y: mean −0.424, sd 2.003, range [−6.29, 5.33]. No implausible values (no units to check against).
• Rows in random order (as documented) — sorted in analysis; irrelevant to estimates.
• Information-availability: only input is the permitted file; no violation.