Does an increase in stock lending balance mean short selling increased?

Not necessarily. The stock lending balance includes all shares borrowed but not yet returned. These data cannot identify which portion was used for short selling. This experiment compared only the increase in the balance with share prices over the following week.

Does holding for 4 weeks change the result?

The direction is the same, but the difference becomes less clear. With a 4-week holding period, the portfolio underperformed the market by 0.62%p when short selling was allowed (t -1.46) and outperformed by 0.34%p during the ban (t 0.67). Of the two holding periods, the 1-week period showed a clearer difference.

When short selling was banned, did piles of borrowed shares still matter?

Mr. Latte

Published 2026-09-30 00:40 KST

Short selling was banned for all stocks in the Korean market from November 6, 2023, to March 30, 2025. Buying the 30 stocks with the largest weekly increases in stock lending balance and selling them 1 week later for 73 rounds would have turned 10M won into 9.98M won (-0.2%). A portfolio holding all stocks that met the same threshold would have ended at 9.58M won (-4.2%). The average weekly gap was +0.06%p (t 0.39). Across the 206 weeks in the two periods before and after the ban when short selling was allowed, the same rule underperformed the market by an average of 0.24%p per week (t -2.23).

Hypothesis

"Stocks with a sharp weekly increase in stock lending balance subsequently underperform the market. This holds only when short selling is allowed."

Experiment setup

During the full short selling ban, we compared each Wednesday's stock lending balance with the previous week's and selected the 30 stocks whose borrowed share counts had increased the most. Only KOSPI and KOSDAQ common shares with a stock lending balance value of at least 10B won on both dates were included. We bought the stocks in equal weights at the next session's close, sold them at the close 1 week later, and switched to a new list. This was repeated 73 times. The first purchase was on November 9, 2023, in the first week of the ban. The final sale was on April 3, 2025, when the last list selected during the ban was sold.

First purchase: 2023-11-09
Valuation date: 2025-04-03
Total invested: ₩10M
Stocks: 30 stocks with the largest stock lending balance increases (replaced weekly)

Result

Final value

₩9.98M-0.2%

Hypothesis held

30 stocks with the largest stock lending balance increases₩9.98M

All stocks meeting the threshold₩9.58M, -4.2%

Trend chart

Unit: KRW

30 stocks with the largest stock lending balance increasesAll stocks meeting the threshold

Each point shows the portfolio value on the day the previous week's holdings were sold and new holdings were bought. The chart covers only the 73 weeks of the short selling ban. Even at its widest, the gap between the two lines was 1.23M won.

Verdict

Hypothesis held

During the 73-week short selling ban, the weekly gap between the 30 stocks with the largest increases in borrowed shares and the market was +0.06%p, effectively no difference. In the 206 weeks when short selling was allowed, stocks selected by the same rule lagged by 0.24%p per week.

Short selling rules changed twice

The stock lending balance counts shares that institutions and foreign investors have borrowed but not yet returned. Borrowed shares can be used for short selling, so a sharp weekly increase is often read as a buildup of bets on falling prices. If this interpretation is correct, one condition follows. When short selling is banned, the route for using borrowed shares to bet on a decline is blocked. The same signal should therefore say nothing about share prices. This condition occurred during the sample period. Until November 3, 2023, short selling was allowed only for KOSPI 200 and KOSDAQ 150 stocks. From November 6, 2023, to March 30, 2025, it was banned for all stocks. Only covered short selling by market makers and liquidity providers was exempt. Short selling reopened for all stocks on March 31, 2025. Running the same rule separately for the three periods provides a comparison in which only the short selling rules differ.

The 73 weeks under the ban: No difference

The first list during the ban was selected using balances from November 8, 2023, and bought at the close on November 9, 2023. The 30 stocks with the largest increases beat the market average in 36 of the 73 weeks, close to half. The average weekly gap was +0.06%p, with a t value of 0.39. In portfolio terms, 10M won became 9.98M won, while a portfolio holding all stocks that met the same threshold ended at 9.58M won. The 30-stock portfolio fell to 9.23M won on January 25, 2024, before rising to 11.71M won on June 13, 2024. Its gap against the market portfolio ranged from trailing by 0.8M won on February 8, 2024, to leading by 1.23M won on January 9, 2025. The gap did not accumulate in either direction.

The 206 weeks when short selling was allowed: Lagging by 0.24%p per week

Across the combined 206 weeks in the two periods when short selling was allowed, the same 30-stock selection underperformed the market average by 0.24%p per week, with a t value of -2.23. It beat the market in 43% of weeks. For the 130 weeks from May 2021 to November 2023 alone, the gap was -0.14%p (t -1.31), and 10M won became 6.45M won. The market portfolio ended at 7.89M won over the same period. In the 76 weeks after short selling resumed, from April 2025 to September 2026, the gap was wider at -0.42%p (t -1.83). The market portfolio grew from 10M won to 14.98M won during this period, while the 30-stock portfolio reached only 10.61M won. The 30-stock portfolio rose to 16.52M won on February 26, 2026, before falling to 8.08M won on July 30, 2026.

Not all borrowed shares are used for short selling

The stock lending balance includes shares borrowed for purposes other than short selling, such as arbitrage and hedging. The results therefore mean that a concentration of borrowed shares coincided with weaker share prices only when short selling was allowed. They do not mean that the entire balance represented bets on falling prices. The direction was unchanged when the holding period was extended to 4 weeks. The gap was -0.62%p (t -1.46) when short selling was allowed and +0.34%p (t 0.67) during the ban. However, the t values are less negative than -2 when each period is examined separately, and the sample after short selling resumed covers only 1.5 years. The calculation assumes that all 30 stocks are replaced every week. Including fees and taxes would lower every portfolio value.

Key numbers

30-stock portfolio during the ban: 9.98M won (-0.2%)
Market portfolio during the ban: 9.58M won (-4.2%)
Weekly gap during the ban: +0.06%p (t 0.39, 73 weeks)
Weekly gap when short selling was allowed: -0.24%p (t -2.23, 206 weeks)
30-stock portfolio, 2021 to 2023: 6.45M won (market 7.89M won)
30-stock portfolio after short selling resumed: 10.61M won (market 14.98M won)

FAQ

Does an increase in stock lending balance mean short selling increased?: Not necessarily. The stock lending balance includes all shares borrowed but not yet returned. These data cannot identify which portion was used for short selling. This experiment compared only the increase in the balance with share prices over the following week.
Could investors still borrow shares during the short selling ban?: Yes. The ban covered short selling, while stock lending itself remained available. The Financial Services Commission allowed covered short selling by market makers and liquidity providers during the ban. Even during this period, 317 to 365 stocks each week had stock lending balances worth at least 10B won.
Does holding for 4 weeks change the result?: The direction is the same, but the difference becomes less clear. With a 4-week holding period, the portfolio underperformed the market by 0.62%p when short selling was allowed (t -1.46) and outperformed by 0.34%p during the ban (t 0.67). Of the two holding periods, the 1-week period showed a clearer difference.

How this was calculated

We obtained the Financial Services Commission's Stock Lending Transaction Information (balance in shares and balance value) and Stock Price Information (closing prices and listed shares) from the Public Data Portal once a week from April 28, 2021, to September 17, 2026. The lending signal date was each Wednesday, or the previous trading session if the market was closed. During Lunar New Year and Chuseok weeks when the market was closed from Monday through Wednesday, the previous week's Thursday or Friday was used. The portal returned only 4 rows of lending data for February 21, 2024, so we substituted data from the previous session, February 20, 2024. The rate of change was the share balance on the signal date divided by the share balance on the previous week's signal date, minus 1. Eligible stocks were KOSPI and KOSDAQ common shares with balance values of at least 10B won on both dates, excluding preferred shares and SPACs. The top 30 by rate of change formed the surge group. Purchases used the closing price of the session after the signal date, normally Thursday that week, or the next session if the market was closed. The same stocks were sold at the following week's closing prices. The comparison portfolio held all stocks meeting the same threshold in equal weights. Stocks with price discontinuities caused by stock splits, reverse splits, or bonus issues during the holding period were excluded from that week's average. Periods were assigned by purchase date: short selling allowed through November 3, 2023, banned through March 28, 2025, and resumed thereafter. Each period started afresh with 10M won. Fees, taxes, tick sizes, and dividends were excluded. The t value is the mean weekly gap divided by its standard error. For the 4-week holding period, t was computed on non-overlapping subsamples and averaged. The calculation can be rerun using tools/lib/lending-backtest.mjs and tools/scripts/backtest-whatif.mjs in the inverseone repository.

Data sources

[1] Financial Services Commission stock lending data (balance shares, balance value), Public Data Portal (data.go.kr), 2026-09-16
[2] Financial Services Commission stock price data (close, listed shares, market cap), Public Data Portal (data.go.kr), 2026-09-17
[3] Press release: short selling banned on all stocks through the first half of 2024, Financial Services Commission, 2023-11-05
[4] Press note: short-selling ban extended to March 30, 2025, Financial Services Commission, 2024-06-13
[5] Press note: short selling fully resumes on March 31, 2025, Financial Services Commission, 2025-03-21

Stocks in this report

30 stocks with the largest stock lending balance increases (replaced weekly) KRX
All contrarian reports →

These are hypothetical results computed from past prices. A real trade would differ with fees, taxes, lot sizes and execution timing, and past returns say nothing about future returns. This is not a solicitation to buy or sell any security. The money flow and stock signals this service shows are estimated, for-reference information based on price and volume. It is not a solicitation to buy or sell any security, nor investment advice. It is a prior-close daily batch, not real-time quotes, and all investment decisions and responsibility are your own. The Trade button is a convenience link to your chosen brokerage's app/website; orders are placed directly at the brokerage and InverseOne is not involved in any order.