# Contract (F2-conf-02)

• Question: does the memo's claim (OLS slope of y on x significant at 5%, based on iid SEs) survive once the SE accounts for serially correlated errors?
• Estimand: β in the linear projection y_t = α + β x_t + u_t (full sample OLS, 400 months). Meaningful under stationarity/weak dependence of (x_t, u_t) with E[x_t u_t]=0 by construction of the projection; HAC SE consistent under mixing + finite moments with bandwidth growing slower than T.
• Outputs: out/results.json (task_id, beta_hat, alpha_hat, se_iid, rho_resid, se_hac, hac_lags, t_hac, p_hac, beta_ci_low, beta_ci_high, memo_claim_supported, n_analysed, feasibility, feasibility_reason, data_issues, n_input_rows, random_seed), out/report.md (6 sections), out/analysis.py, out/claims.json (incl. claim restating memo with verdict), plus protocol files.
• Information rules: only data/series.csv and data/memo.md; memo numbers may be reproduced, not assumed.
• Invalidating conditions: missing/corrupt data, gaps in months, inability to compute HAC; result invalid if memo numbers copied rather than computed.