Moreira-Muir Volatility-Managed International Equity Sleeves: Monthly c/RV^2 Exposure on EFA, EEM and EWJ, c Calibrated Causally to Match Buy-and-Hold Volatility (USEQ 1-DAY, long/scaled, monthly rebalance)

Derived by the Research Lead from the quant.stackexchange question 'implementing Volatility Managed Portfolios' (axl, 2018) and the answer by Alex C, which cite Moreira & Muir (2017) Eq. 1: managed return_{t+1} = (c / RV_t^2) * f_{t+1}, with c chosen so the managed series has the same standard deviation as buy-and-hold. The source gives the weighting rule and the calibration of c; it gives no market, sizing cap, execution or cost model, so those are the Research Lead's choices. NOTE: both linked pages (S1, S2) answered HTTP 403 and were NOT read; only the question and answer text quoted in the seed was used. MARKET CHOICE: US single-index versions of this idea (QQQ, IWM) and crypto long-only versions are already in the corpus, so this tests the textbook 'market factor' reading on three markets the corpus has not used for it: developed ex-US (EFA.USEQ, from 2001), emerging markets (EEM.USEQ, from 2003) and Japan (EWJ.USEQ, from 1996), each run as an independent sleeve. Each sleeve is one 'factor' f = the ETF's monthly return. RULES (per sleeve i, evaluated on daily bars): (1) At the close of the LAST session of each calendar month t, compute RV_t^2 = sum of squared daily log returns over the last rv_lookback_sessions sessions (baseline 21 = one month, the source's 'realized variance over the past month'). (2) Raw scale s_t = 1 / RV_t^scaling_power (scaling_power 2 = the source's variance rule; 1 = the common volatility variant, registered as a categorical alternative). (3) Calibration of c, made causal: the source runs a trial with c0 and rescales c by sigma_BH / sigma_0 over the SAME historical period, which uses future data if applied in-sample. Here c_t is recomputed every month from history only: over all completed months up to t (at least calib_min_months), sigma_BH = std of the sleeve's unmanaged monthly returns and sigma_0 = std of the trial series s_{k-1} * f_k (c0 = 1); c_t = sigma_BH / sigma_0. Before calib_min_months of history exist, the sleeve is flat. (4) Target weight w_t = c_t * s_t, as a multiple of the sleeve's base allocation (1/3 of equity), capped at max_sleeve_mult = 2.0 (so a sleeve's notional never exceeds 0.667 x equity and the book's gross never exceeds the Reg-T 2.0x). (5) Execution: the decision on the month's last close fills at the NEXT session's open (platform next_open_fill on USEQ, supported). Each rebalance closes the prior month's sleeve position and opens the new target size, so every sleeve-month is one closed trade (clean per-trade accounting; ~0.03-0.05% RT cost per sleeve-month on these ETFs). PRE-REGISTERED COMPARATOR (the source's own test): the equal-weight, monthly-rebalanced, unmanaged buy-and-hold of the same three ETFs at the same base allocation over the same span; the premise is that the managed book has higher Sharpe and positive alpha against it at matched volatility. Dividends are not in PnL on USEQ, which lowers both the strategy and the comparator equally (about 2-3%/yr on these ETFs); the comparison is unaffected in sign. No stop-loss: risk control route (b) - a time/rebalance exit every month plus capped sizing (no position above 0.667x equity; gross <= 2x). Intended book volatility ~15-18%/yr (matched to buy-and-hold of the basket by construction), leverage <= 2x.