"""Why does slower capital lower cognitive unemployment? Decompose the January 2030 cognitive wage
into the common (potential) wage and the sticky cognitive discount (Eq. 30, p. 19), and report the
cumulative layoffs, for the extreme scenario at the paper's setting and two adjusted settings.

Usage: python -I mechanism_check.py PKG_DIR
"""
import math
import sys
from dataclasses import replace

sys.path.insert(0, sys.argv[1])
from econ_scenarios import EXTREME, Calibration, simulate  # noqa: E402

print(f"{'eps':>4} {'half-life':>9} {'common wage':>11} {'discount':>9} {'paid w_C':>9} "
      f"{'layoffs to 2030':>15} {'u_C':>6}")
for eps, hl in ((3.0, 0.0), (3.0, 12.0), (6.0, 0.0), (6.0, 24.0)):
    sim = simulate(EXTREME, replace(Calibration(), eps=eps), capital_half_life=hl)
    s = sim.series
    k = sim.index(2030.0)
    layoffs = float(sum(s["D_C"][: k + 1]))
    u_c = 100 * s["U_C"][k] / (s["U_C"][k] + s["ell_C"][k])
    print(f"{eps:4.0f} {hl:9.0f} {100 * math.expm1(s['pot_lnW'][k]):11.2f} {100 * s['wage_gap'][k]:9.2f} "
          f"{100 * math.expm1(s['lnW_C'][k]):9.2f} {100 * layoffs:15.2f} {u_c:6.2f}")
print("common wage and paid w_C: percent above the no-AI path; discount: 100 x ln(w_C / w);")
print("layoffs: cumulative D_C, percent of the labor force.")