ATTACK A: is the -38% realistic, or an artifact of permanent real-wage erosion?
  our model (permanent phi=0.7), hetero -1/+3:    -37.6%  (compounds over 40y)
  realistic 1-yr indexation lag, hetero -1/+3:     -3.9%  (CONSTANT over 40y)
  realistic 2-yr indexation lag, hetero -1/+3:     -7.6%  (CONSTANT over 40y)
  realistic 3-yr indexation lag, hetero -1/+3:    -11.2%  (CONSTANT over 40y)
  realistic 5-yr indexation lag, hetero -1/+3:    -18.0%  (CONSTANT over 40y)
  -> the -38% assumes wages NEVER catch up. Realistic lags give a small, constant offset.

ATTACK F: does a COMMON level really cancel at any wage stickiness? Test HETEROGENEOUS stickiness.
  common +3%, but H indexes fast (lag 1) and L indexes slow (lag 4):
     distortion =   -8.5%   <- NOT zero! a common POSITIVE level leaves a residual
  common  0%, same heterogeneous stickiness (lag 1 vs 4):   -0.0%   <- zero
  (phi-form) common +3%, phiH=0.9 vs phiL=0.5:
     distortion =  -37.5%   common 0%, same:   +0.0%
  -> 'any common level' was wrong under heterogeneous stickiness. Only common ZERO is fully robust.

ATTACK C: is '8x cheaper' robust to the baseline? excess price level over 40y vs different baselines:
  vs natural -1%       : corridor +   388%   zero +   49%   ratio 7.8x
  vs price stability 0%: corridor +   226%   zero +    0%   ratio inf (zero has no excess)
  vs Friedman -2%      : corridor +   632%   zero +  124%   ratio 5.1x
  -> the '8x' is specific to the -1% baseline. Robust claim: corridor adds 200-400% cumulative; zero adds ~0.

ATTACK E: does variance still beat level if shocks are CORRELATED across countries?
  high variance, independent (rho=0):   2.78 %/yr
  high variance, correlated  (rho=0.8): 0.56 %/yr  <- common shocks cancel in the differential
  -> it's the variance of the DIFFERENTIAL that matters; correlation reduces it. Level still irrelevant.
