Skip to main content

Locating & Borrowing Shares

Updated Jun 24, 2026 at 2:35pm

Research Draft High 1,234 words

To sell a stock short, a trader must first deliver shares they do not own to the buyer at settlement — which means those shares have to come from somewhere. "Locating & borrowing" is the plumbing that makes this possible: before the short sale executes, the broker must find a lender willing to part with the shares temporarily, and the trader pays a fee to rent them. The central tension is that this borrowed-share supply is finite, repayable on demand, and re-priced continuously, so the availability and cost of the borrow is itself a market — one that can make an otherwise-correct short thesis unworkable or unprofitable.

How it works (the locate, then the borrow)

The process has two regulatory-and-mechanical stages:

1. The locate (pre-trade). Under SEC Regulation SHO, Rule 203(b)(1), a broker-dealer may not execute a short sale unless it has, before the sale: (a) borrowed the security, (b) entered into a bona-fide arrangement to borrow it, or (c) has reasonable grounds to believe the security can be borrowed for delivery by settlement date (SEC; Databento). In practice brokers maintain an "easy-to-borrow" (ETB) list — highly liquid names where availability is presumed — and require an explicit locate for everything else. A locate is a reservation, not yet a loan.

2. The borrow (settlement). The actual stock loan is sourced through a supply chain: the ultimate lenders are large long-term holders — index/mutual funds, pension plans, insurers, ETFs — who lend through custodian banks acting as agent lenders; prime brokers intermediate to hedge funds and retail brokers (Wikipedia: Securities lending). The borrower posts collateral exceeding the loan value — roughly 102% in the US (105% common in Europe), marked-to-market daily (Wikipedia).

Pricing. When collateral is cash, the lender invests it and pays the borrower a rebate rate. For general collateral (GC) — abundant, easy-to-borrow names — the rebate has historically run near a benchmark short-term rate (overnight Fed funds / OBFR) minus a modest fee, commonly cited at roughly 25–75 bps (Lamont survey; Marquette Associates). For hard-to-borrow ("on special") names, the rebate falls and can go negative — the borrower effectively pays a borrow fee on top of forgoing the cash interest (SEC comment letter). Hard-to-borrow annualized fees commonly cited range from ~1% to over 300% in extreme cases (Public FAQ). Most loans are "open" — terminable/recallable by either side daily — rather than term loans.

How it's used in practice

  • Borrow cost is part of expected return. A short carries a daily holding cost equal to the borrow fee; at a 50% annualized fee, a position bleeds roughly 4%+/month regardless of whether the thesis plays out. Traders check the indicative borrow rate and availability before entering, not after.
  • Utilization and short interest as signals. Utilization = shares on loan ÷ shares available to lend. Near 100% means the borrow is nearly tapped out — fees spike and recall risk rises. Vendors (S3 Partners, DataLend, Ortex) sell this data; it is widely used as both a crowding gauge and a squeeze-risk gauge.
  • Recall is a live risk. Because most loans are open, a lender can recall shares, forcing the short to either re-borrow elsewhere (possibly at a worse rate) or buy in to cover involuntarily — at the worst possible time if the stock is rising.

Adoption, debate & evidence

Securities lending is a mature, multi-trillion-dollar institutional business, and the locate requirement has been mandatory under Reg SHO since 2005; this is settled infrastructure, not a contested technique. The substantive debates are about consequences:

  • Short-sale constraints cause overpricing. This is well-supported. Ofek, Richardson & Whitelaw documented that put–call parity violations are asymmetric toward short-sale-constrained stocks and scale with borrow cost — evidence that hard-to-borrow stocks are systematically overpriced and correct only slowly over months (Ofek-Richardson-Whitelaw, NBER). The high borrow fee itself is a return predictor: expensive-to-short stocks tend to underperform.
  • Recall/utilization risk is quantifiable. One analysis cited that for easily-borrowed stocks forced recalls occur roughly once in eight years, but for stocks with utilization above 75% a forced recall can occur about once every ~26 days (Schultz, "What Makes Short Selling Risky") — figures specific to that study's sample, not universal constants.
  • Naked shorting & fails. Reg SHO's close-out rule (Rule 204) requires a participant to close out a fail-to-deliver, generally by the open of trading on the day after settlement, and "threshold securities" with persistent fails face the well-known 13-consecutive-settlement-day mandatory buy-in (NYSE Reg SHO Guide). Note: US settlement is now T+1 (since May 2024), tightening these timelines versus older T+2/T+3 references.

A note on what borrow data does not tell you: high short interest is frequently mis-read as a directional bear signal. It is at least as much a fuel-for-squeeze signal, and the academic edge belongs to the borrow-fee/constraint literature, not to naive "shorts are smart money" folklore.

Strengths & limitations

When the framework helps: Borrow fee and utilization are among the few genuinely predictive, hard-to-game datasets in short-selling — they reflect real supply/demand for the loan, and extreme readings reliably flag both overpricing and squeeze danger.

When it fails / #1 misuse: The classic error is treating a short as "free" once located. The locate guarantees you can sell short today; it guarantees nothing about (a) tomorrow's borrow fee, which can ratchet up without notice, or (b) whether your shares get recalled. A trader who is right on direction can still lose because the borrow became uneconomic or the position was bought-in. Indicative locate rates can also differ materially from the actual fee charged after the fact.

Sources

Flags: Rule 204 close-out deadline and the 13-day threshold-securities rule are summarized from secondary guides (the primary PDF would not parse cleanly); the "once every 26 days at >75% utilization" figure is sample-specific to one paper, not a universal constant.