RESEARCH.md: placement premise (IRA distributions taxed ordinary at withdrawal, UBTI excepted) + constrained-taxable volatility rule

fundlab/RESEARCH.md

@@ -682,3 +682,53 @@ when relevant. All of 3-4 above are taxable-only by nature.
 Post-OBBBA (Pub. L. 119-21) items I could not re-verify live (SearXNG
 index down): 199A specified-investment-asset scope/sunset, SALT cap,
 QCD dollar limit. Verify with tax pro before relying.
+
+### Placement premise + constrained-taxable rule (2026-08-27)
+Premise (user-stated, verified): distributions on assets held in a
+TRADITIONAL IRA are also taxed - deferred, but at the ORDINARY rate
+at withdrawal.  Price appreciation and every distribution (div,
+interest, cap-gain dist) come out as ordinary income; the character
+is destroyed.  Exception: K-1 UBTI is taxed CURRENTLY, not deferred.
+So the IRA is an ordinary-rate bucket: tax-favorable character
+(LTCG / QD / tax-exempt / ROC-deferral) is worthless inside it;
+ordinary character only gets deferral there.
+
+Constrained-taxable rule (user's proposal, agreed with refinements):
+the tax-exempt accounts are larger in aggregate, so some funds that
+prefer the taxable account will end up in the IRA.  User's rule: for
+the scarce taxable space, select for VOLATILITY (big potential
+returns AND losses): in down years harvest and swap into a similar
+non-substantially-identical fund (no wash sale), in up years hold
+and register the LTCG at the LTCG rate that the IRA would tax as
+ordinary.
+
+Agreed.  The counterfactual makes it exact: the same swing inside
+the IRA generates ZERO tax events; inside the taxable account, up ->
+(15-20%) LTCG, down -> a CURRENT deduction (offsets gains 1:1 +
+$3k ordinary + carryforward).  Both terms scale with the size of the
+swing, so volatility is the right selection variable.  Refinements:
+- The exact objective is tax-arb per dollar = (future tau_ord -
+  tau_char) x expected favorable return + harvest value.  Volatility
+  proxies both terms, but it must be volatility in POSITIVE
+  expected-return strategies (risk premium), not variance for its
+  own sake.
+- Ordering for scarce space: (1) MUNIS first - the IRA destroys
+  100% of the exemption, the saving is certain and immediate, and
+  munis are wash-sale EXEMPT (1091(c)(3)(B)): harvest and rebuy the
+  SAME fund next day.  (2) high-return/high-vol accumulators - the
+  deferral term is largest, and the user's premise (future ordinary
+  rate rises) widens the gap every year.  (3) steady low-yield
+  favorable-character funds last - they lose the least in the IRA.
+- The harvest term is a BONUS, the deferral term the MAIN EVENT: a
+  loss in a no-gain year is worth only $3k x tau_ord (~$1k) unless
+  there are concurrent gains to offset.  The deferral term is
+  (tau_ord - 15%) on every dollar of appreciation - uncapped.
+- Basis management: harvest size = NAV vs basis.  Buying in up years
+  builds high-basis lots = fatter future harvests.
+- Use the rule to PRIORITIZE among funds you would own anyway; the
+  tax benefit (~1.5-2.5%/yr on balance) does not justify ADDING
+  risk by itself.
+- Swap mechanics: different mutual funds are not substantially
+  identical (different holdings/NAV) - the standard harvest-and-swap
+  is safe; the 250-candidate universe makes a like-for-like
+  substitute usually available.