# Contract — F6-conf-03 (Okun's law, US 1960Q1–2019Q4)

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

Estimands. β (and α) in Δu_t = α + β g_t + ε_t, OLS, 1960Q1–2019Q4 (n = 240). u_t = mean of the 3 monthly UNRATE values in quarter t; Δu_t = u_t − u_{t−1}; g_t = 100·Δln GDPC1_t (not annualised). β is a reduced-form conditional association (an "Okun coefficient"), meaningful as a descriptive/predictive slope under linearity and approximate stationarity of the relationship within each sample; it is not a structural causal effect. Inference: Newey–West (Bartlett) HAC SE for full-sample β, lag reported; 95% CI from that SE. Stability: subperiod OLS slopes (1960Q1–1983Q4, n=96; 1984Q1–2019Q4, n=144) and classical Chow SSR F-test, df (2, n−4), 5% level.

Outputs. out/results.json (keys: 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 (scatter + two subperiod lines); out/table1.csv (period,n,intercept,slope,slope_se; 3 rows); out/report.md (6 sections in order); out/analysis.py; out/claims.json.

Information rules. Only data/GDPC1.csv and data/UNRATE.csv. No quarter after 2019Q4 may enter any estimate (1959Q4 levels used only to form the 1960Q1 differences). Data vintage is the 2026 FRED vintage (as given).

Invalidating conditions. Missing monthly UNRATE or GDPC1 values inside 1959Q4–2019Q4; incomplete quarters; using post-2019 data; non-HAC SE for full-sample β; wrong Chow df.