After short selling resumed, what if you bought the stocks being covered the most?

The signal from a large volume of returned shares

A sharp weekly decline in stock lending balance means that far more borrowed shares were returned than newly borrowed. Short sellers have to buy shares back in the market to return them. This leads to the view that stocks with falling balances receive buying demand. To test this hypothesis, we compared balances each Wednesday with the previous week's and selected the 30 common stocks with the largest declines. Only stocks with a balance value of at least 10B won at both dates were eligible. For stocks with balances worth several billion won, a single lending transaction can cause a large percentage change. We bought the selected stocks at the next trading session's closing prices and sold them at the closing prices one week later. The benchmark portfolio bought all stocks meeting the same threshold at equal weights.

76 weeks after resumption: 10M won became 20.09M won

The first list was based on balances as of April 2, 2025. We bought those stocks at the closing prices on April 3. The 30-stock portfolio reached its low of 9.95M won on April 10, then grew faster than the benchmark portfolio. It peaked at 28.52M won on May 7, 2026, when the benchmark portfolio was worth 19.91M won. Both portfolios then fell sharply. By July 30, the 30-stock portfolio was worth 17.43M won and the benchmark 12.19M won. At the final valuation on September 17, the 30-stock portfolio was worth 20.09M won and the benchmark 14.98M won. The 30 stocks with the largest declines outperformed the benchmark in 46 of 76 weeks. The average weekly gap was +0.41%p, with a t value of 2.14.

It underperformed the benchmark in 2021-2023

Applying the same rule to the two earlier periods produced different results. During the 130 weeks from May 2021 to November 2023, short selling was allowed only in KOSPI 200 and KOSDAQ 150 stocks. The 30 stocks with the largest balance declines underperformed the benchmark by 0.08%p per week (t -0.75). In this period, 10M won became 7.05M won, compared with 7.89M won for the benchmark portfolio. During the 73 weeks from November 2023 to March 2025, short selling was banned for all stocks. The gap was +0.12%p (t 0.84), which was not clear-cut. The 10M won became 10.42M won, while the benchmark portfolio ended at 9.58M won. The same rule doubled the portfolio in one period but trailed the benchmark in another. The hypothesis held only after short selling resumed, so it is hard to treat it as a rule that works consistently.

The effect was brief and coincided with a rising market

Extending the holding period to 4 weeks made the gap after resumption less clear, at +0.54%p (t 0.67). If the 1-week gap had accumulated at the same rate over 4 weeks, it would have been around 1.6%p. This suggests that most of the price move after a balance decline ended within a week. The 76 weeks after resumption also coincided with a rising market, with the benchmark portfolio gaining nearly 50%. This calculation cannot distinguish between stocks with sharp balance declines rising more with the market and rising because of purchases to return borrowed shares. The sample covers only a year and a half. The calculation assumes that all 30 holdings are replaced every week. Fees and taxes would reduce the reported twofold portfolio value.

Key numbers

30-stock portfolio after resumption

20.09M won (+100.9%)

Benchmark portfolio after resumption

14.98M won (+49.8%)

Weekly gap after resumption

+0.41%p (t 2.14)

Weeks outperforming the benchmark

46 of 76 weeks

30-stock portfolio in 2021-2023

7.05M won (benchmark 7.89M won)

With a 4-week holding period

+0.54%p (t 0.67)

FAQ

Does a decline in stock lending balance mean short sellers bought shares back?

Not necessarily. A decline only means that borrowed shares were returned. These data do not show whether the shares were bought back in the market or came from existing holdings. The figures also include returns of shares borrowed for purposes other than short selling.

Could the result after resumption be due to a rising market?

Partly, yes. Over the same 76 weeks, the benchmark portfolio also grew from 10M won to 14.98M won. The 30-stock portfolio gained an average of 0.41%p more per week. But this calculation cannot determine whether the list contained many stocks that tend to rise more in a rising market.

How did stocks with the largest balance increases perform over the same period?

They moved in the opposite direction. Over the same 76 weeks, a portfolio buying the 30 stocks with the largest increases in stock lending balance ended at only 10.61M won. It underperformed the benchmark by an average of 0.42%p per week. Results for stocks with rising balances are covered separately in the Contrarian Report on the short selling ban period.