# Contract — F6-conf-03 (Okun's law, US 1960–2019)

Question. How strongly does the quarterly change in the US unemployment rate respond to quarterly real GDP growth, and is that response stable across 1960Q1–1983Q4 vs 1984Q1–2019Q4?

Estimands.

• β (and α) in Δu_t = α + β g_t + ε_t, OLS, 240 quarters 1960Q1–2019Q4. u_t = mean of 3 monthly UNRATE values; Δu_t = u_t − u_{t−1} (pp); g_t = 100·Δln GDPC1 (percent, not annualised).
• HAC (Newey–West, Bartlett) SE for β with stated lag; 95% CI from it.
• β in subperiods (n=96, n=144); classical Chow SSR F-test of joint (α,β) equality, df (2, n−4)=(2,236).
• Baseline: RMSE of mean-only prediction of Δu; model RMSE of OLS residuals.
• Meaningful under: linear, time-invariant relation within each (sub)sample; g_t exogenous enough for β to be a descriptive (not causal) projection coefficient; Chow test assumes homoskedastic, serially uncorrelated errors (classical) — a known limitation.

Outputs. out/results.json (keys per TASK.md table: sample_start, sample_end, n_sub1, n_sub2, baseline_metric, model_metric, effect_estimate, intercept, effect_se, hac_lags, effect_ci_low, effect_ci_high, beta_sub1, beta_sub2, chow_f_stat, chow_p_value, stability_conclusion, conclusion, task_id, feasibility, feasibility_reason, data_issues, n_input_rows, n_analysed, random_seed); out/figure1.png; out/table1.csv (period,n,intercept,slope,slope_se; 3 rows); out/report.md (6 sections); out/analysis.py; out/claims.json; plus protocol state files and out/exclusions.csv.

Information rules. No quarter after 2019Q4 enters any estimate; 1959Q4 levels used only to form 1960Q1 differences. Only data/GDPC1.csv and data/UNRATE.csv.

Invalidity / impossibility. Missing months or quarters inside 1959Q4–2019Q4; non-numeric values; wrong frequency; sample ≠ 240. Use of post-2019 data. Note: GDPC1 is the current (2026) vintage, so revisions are embedded — acceptable since the task is an ex-post descriptive estimate, no real-time rule.