Tax-loss harvesting across multiple brokerages: a DIY guide

Why multiple brokerages break naive harvesting

IRS §1091 disallows a loss if you buy a substantially identical security within 30 days before or after the sale — in any of your accounts. Brokers, though, only detect wash sales within a single account at a single firm (same CUSIP, same account); that's all the 1099-B reflects. Selling at a loss at Schwab and rebuying at Fidelity is invisible to both firms and fully visible to the IRS — compliance across accounts is entirely on you.

It gets sharper: under Rev. Rul. 2008-5, a purchase in your IRA can trigger the wash sale — and because basis can't be adjusted inside an IRA, that loss isn't deferred, it's gone permanently. Automated dividend reinvestment is the classic silent trigger: a DRIP purchase inside the 61-day window is a replacement purchase you never consciously made.

The rules that actually bind

Five facts do most of the work:

• The window is 61 days: 30 days before the sale, the sale date, and 30 days after.
• "Substantially identical" certainly covers the same security (same CUSIP). Two funds tracking the same index are a gray zone the IRS has never precisely defined — conservative practice treats them as risky.
• Your spouse's accounts count. Selling in your account and buying in theirs is still a wash sale.
• Controlled IRAs count (Rev. Rul. 2008-5) — and a wash sale absorbed by an IRA disallows the loss permanently, the worst outcome on this page.
• Outside the IRA case, a wash sale defers rather than destroys the loss: the disallowed amount is added to the replacement shares' basis. Annoying, not fatal — but it wrecks the current-year plan.

The manual process, end to end

Without software, safe cross-custodian harvesting looks like this:

• Inventory every account that counts: yours, your spouse's, and all IRAs — across every custodian.
• Export tax lots from each brokerage (not just positions — you need dates and basis per lot).
• Unify by security identifier. Match on CUSIP/ISIN, not ticker alone — share classes and near-duplicates hide behind similar tickers.
• Identify harvestable lots: positions below basis, worth the trade after costs.
• Scan the full ±30-day window across all accounts for purchases of the same security — including dividend reinvestments. Turn off DRIP on any security you plan to harvest, everywhere you hold it.
• Choose a replacement that keeps market exposure without being substantially identical. When no clearly-distinct equivalent exists, the conservative answer is cash or your CPA's judgment — not hope.
• Log the trade, and calendar day 31 before re-entering the original position.

What good software should check for you

The manual process is a data problem — five custodians' lot files and a 61-day rolling window — which is why software should do the watching: a unified cross-custodian feed, continuous ±30-day scanning that includes controlled IRAs, DRIP awareness, and a conservative line on replacements.

Formation's implementation is deliberately strict: every candidate is scored clear, risk, or blocked; blocked positions disable the action entirely and are excluded from any projected-savings figure; and where no vetted non-identical replacement exists, the answer shown is "consult your CPA," not a workaround. Formation does not execute — it produces the cited picture and a plain-text trade list you take to your own broker, on your own judgment. Education, not a recommendation.

What the savings are — and aren't

Honest math: harvested losses first offset realized capital gains; up to $3,000 of any excess offsets ordinary income each year; the remainder carries forward indefinitely. The value depends on your bracket, your future rates, and the basis reset you accept on the replacement — harvesting is a deferral engine with option value, not free money. A poorly-chosen replacement that lags the original can cost more than the tax it saved. Treat every projected-savings number, from any tool, as an estimate against assumptions you should be able to see.