Stock lending and short-selling: evidence from national pension service in Korea

Abstract

The National Pension Service (NPS) of Korea suddenly announced that they would suspend their stock lending business from October 22, 2018. Using this ideal setting, the authors investigate the effects of this suspension on market quality and short-selling activities. The authors find that stock return does not increase after the suspension of stock lending for both the KOSPI and KOSDAQ markets. However, the returns of stocks with NPS ownership decline less than those without NPS ownership. The authors also find that the institutional and foreign investors' short sales did not increase in both markets after the lending business suspension by the NPS. In addition, the effect of suspension of stock lending on market quality is mixed, so the authors cannot conclude that market quality has improved. Overall, the authors’ results indicate that the stock market, especially for short-sales activity, has not been affected by the suspension of the stock lending service by the NPS.

Citation

Lee, K.-H. and Wang, S.-F. (2022), "Stock lending and short-selling: evidence from national pension service in Korea", Journal of Derivatives and Quantitative Studies: 선물연구, Vol. 30 No. 3, pp. 146-171. https://doi.org/10.1108/JDQS-02-2022-0005

1. Introduction

“We have stopped lending stocks from October 22 after an internal discussion,” said Kim Sung-Joo, NPS CEO in the Congressional audit hearing on October 23, 2018. “Existing stock lending transactions will be resolved by the end of the year.”

JoongAng daily

The National Pension Service (NPS), Korea's largest public retirement pension fund, suddenly suspended lending on the stocks held from October 22, 2018. The NPS has been constantly criticized by investors who are concerned that stocks lent by NPS could be used by short-sellers, which may lead to a decline in stock price. Using this exogenous event, this study investigates the impact of short-selling constraints on stock market quality and short-selling activity when an exogenous shock arrives in the market. The effect of short-sale constraints on market quality and stock price has been studied extensively. Most studies conclude that short-sales constraints reduce market quality. For example, Miller (1977) argues that short-sale constraints impede negative information in the stock market, leading to overpricing. D'avolio (2002) and Nagel (2005) conclude that restrictions on short sales exclude negative information from the market, leading to overpricing of stocks. Saffi and Sigurdsson (2011) investigate short-sales constraints on stock price efficiency using data from 25 countries. They find that stocks with higher short-sales constraints lower price efficiency. However, several studies have shown mixed results regarding short-sale constraints on price efficiency. For example, Bai et al. (2006) show that restricting short-sellers may result in both over- or under-valuations, depending on the degree of risk-sharing traders and traders who are speculating on private information. Diamond and Verrecchia (1987) show that short-sale constraints impede incorporation of negative information into stock prices than positive information. However, they argue that investors will recognize short-sale constraints and adjust their beliefs so that stock prices do not tend to be upward biased. Moreover, Kaplan et al. (2013) argue that empirically the results of short-sales constraints on stock prices are mixed because it is difficult to separately identify the demand and supply effects. Kim and Cho (2021) investigate the information effect on stocks designated as overheated short-selling stocks. They find that stocks are stabilizing as they have been designated as overheated short-selling stocks.

In this paper, we contribute to the literature by providing direct evidence using an ideal setting in which the NPS suddenly suspends stock lending, and this event provides us with an exogenous shock to the supply of lendable stocks. Given that short-sale in Korea is usually performed by foreign investors [1], who are informed traders (Froot et al., 2001; Froot and Ramadorai, 2008; Wang and Lee, 2015), it is possible that suspension of stock lending to short-sellers may alter the trading behavior of foreign investors and, in turn, affect stock market quality. Moreover, NPS is one of the largest institutional investors in Korea and its investment or management decision has received a lot of attention. Therefore, short-sales constraints arising from the NPS might affect the investment decisions of other institutional or individual investors, which further affect the stock market quality and short-selling trades.

We conduct two sets of empirical tests. The first examines the role of short-sales constraints in market quality, and the second examines the effect of short-sales constraints in the short-selling activity by different investor types and stock return. We find that stocks with NPS ownership tend to have high short-selling activity, low illiquidity (measured by Amihud's illiquidity and the bid-ask spread), large size and low stock volatility (measured by intraday price volatility and standard return of daily return) than stocks without NPS ownership. The relative short-selling activity of stocks with NPS ownership is over 5%; it is 1.0% for stocks without NPS ownership. We also examine whether the NPS suspends stock lending to improve market quality and reduce short-selling activity, especially for institutional and foreign investors. We perform the difference-in-differences (DID) test and regression analysis and find no significant change in the short-selling activity for institutional and foreign investors in the KOSPI market. In the KOSDAQ market, short-selling decreases for institutional investors after the NPS stops stock lending. Interestingly, we find that individual short-sellers increase their shorting after NPS stops stock lending for both markets. We also find that stock return does not rise after the event date, suggesting that the NPS share lending service does not cause a stock price drop. We find mixed results for stock liquidity and volatility. In the DID analysis, we show that stocks tend to be more illiquid and have high volatility after NPS suspends stock lending for both NPS and NonNPS stocks, but the increasing magnitude is smaller for NPS stocks than for NonNPS stocks. However, we find contradictory results in the regression analysis: the bid-ask spread and standard deviation of daily return are reduced for stocks with NPS ownership after NPS suspends stock lending in the KOSPI market. For the KOSADQ market, we find a significant effect on the bid-ask spread, turnover, and price volatility. After the NPS suspension of stock lending, stocks with high NPS ownership tend to be more liquid and have low volatility. We cannot conclude that stock market quality is improved after the NPS suspends stock lending.

In the DID analysis, we find that short-selling relative to trading volume (relss) and abnormal short-selling relative to trading volume (abrelss) by individual investors are increased for stocks with NPS ownership after NPS suspends stock lending for both markets. Interestingly, we find that institutional short-sellers reduce short-selling after NPS stops stock lending in the KOSDAQ market, suggesting that institutional investors mainly borrow shares from the NPS. In the regression framework, we find the same results as the DID analysis, suggesting that short-selling behavior of institutional and foreign investors does not change. We also examine the presence of a significant change in stocks with NPS ownership relative to stocks without NPS ownership in stock return. Jones (2003) shows that shorting restriction is associated with the positive abnormal return, consistent with Miller (1977), who shows that stocks tend to be upward biased in the presence of short-sale restrictions. In contrast to Jones (2003), Miller (1977), and Diamond and Verrecchia (1987) show that investors will recognize short-sale constraints and adjust their beliefs so that stock prices do not tend to be upward biased. We find evidence that stock returns decrease for both samples, that is, stocks with and without NPS ownership for the KOSPI and KOSDAQ markets. We find that after NPS suspends stock lending, stock returns continuously drop, consistent with Diamond and Verrecchia (1987). However, the level of decrease is larger for NonNPS stocks than NPS stocks.

To the best of our knowledge, this is the first study that examines the impact of stock lending of the public pension fund on stock market quality and short-selling activity. Our study makes several contributions to the literature. First, we investigate the largest public pension fund stock lending effect on short-selling activity, which has received relatively little coverage in the literature. Second, our event is an exogenous shock; lendable share is suddenly binding without any advance notice, which provides an ideal setting to investigate short-sale constraints on the stock market. Third, Wang and Lee (2015) show that foreign investors are the major players in the short-sale market. Similar to Wang and Lee (2015), we classify the short-sellers into the individual, institutional and foreign investors to examine which investor group's short-selling behavior significantly changes.

The remainder of this paper is organized as follows. Section 2 briefly describes the NPS of Korea. Section 3 describes our dataset and variables. Section 4 provides an empirical analysis and Section 5 presents the conclusions.

2. The National Pension Service of Korea

The National Pension Service (NPS) of Korea was founded in 1988 and launched the National Pension Fund Investment Management in 1999. The NPS is a public pension fund for retirement, and its management and investment is governed by the National Pension Act. The Minister of Health and Welfare oversees the management and operation of the National Pension Fund, and the National Pension Fund Investment Management is in charge of fund operation.

The NPS is the third-largest pension fund in the world, with 930 trillion Korean won in the total fund as of the end of August 2021. The investment portfolio consists of domestic fixed income (37.0%), global equity (26.7%), domestic equity (19.0%), alternative investment (10.6%), global fixed income (6.3%) and others (0.2%). The NPS is also among the largest institutional investors in the Korean stock market – the proportion of NPS investment in the domestic equity market accounts for 6.7% of total stock market capitalization [2].

The NPS commenced the stock lending service from April 2000 for local stocks and from June 2005 for foreign equity. On October 23, 2018, the CEO of the NPS announces the suspension of stock lending service in the Congressional audit hearing; however, stock lending on their foreign equity portfolio has not been suspended. The NPS stock lending service has been criticized over the past few years, as lending stocks may lead to a decline in stock price because stocks are lent out to short-sellers. From 2014 to 2017, the total profit of the NPS generated from stock lending was 62.1 billion KRW [3], which is approximately 16 billion KRW per year. Compared to the total performance of the NPS, which was 41.2 trillion KRW in 2017, the income from the stock lending service is minuscule.

3. Data

3.1 Sample and sample construction

We obtained the ownership data of the NPS shareholding information from the website of the National Pension Service Investment Management [4]. The NPS discloses the fiscal year-end information of investment portfolio composition in the third quarter of the following year, including the dollar amount of shareholding and the percentage of ownership for each stock [5]. Stock returns, short-selling shares and other stock characteristics are from DataGuide Pro, which is among the largest data vendors in Korea. Given that the NPS suspended stock lending service from October 22, 2018, our sample period is from days t – 50 to t + 50 (August 3, 2018 to January 3, 2019). Our sample stocks consist of all stocks listed on the KOSPI and KOSDAQ stock markets. We exclude stock-day observations with a missing daily return, the bid-ask spread, order imbalance [6] and the number of shorted shares. To construct abnormal short-selling activity measures, we require stocks to have trading activity during the period of days t – 370 to t – 70 (from April 11, 2019 to July 5, 2018). Specifically, we require stocks to have at least 200 trading days during the period of days t – 370 to t – 70. We further restrict the sample stocks to have at least 101 trading days during the sample period. The final sample consist of 1,868 stocks: 734 from the KOSPI market and 1,134 from the KOSDAQ market.

The main variables used in this paper are market quality and short-selling activity measures. We use Amihud's (2002) illiquidity, the bid-ask spread, turnover, price range and standard deviation of daily return (stdev) as a proxy for market quality. Amihud's illiquidity is calculated as an absolute value of daily return divided by the daily trading dollar value. The bid–ask spread is a daily relative quoted spread defined as the difference between the ask price and bid price and scaled by the midpoint of the bid and ask prices. Specifically, we calculate all transactions relative quoted spread in a given day and then average those relatives quoted spread within a day for a stock. Turnover is defined as the daily trading volume divided by the number of shares outstanding. The price range is the intra-day highest price minus the intra-day lowest price scaled by the highest price. The standard deviation of daily return in day t is defined as the standard deviation of the daily return from days t – 20 to t – 1.

The short-selling activity measure we used in this study is relative short-selling activity (relss), which was first introduced by Diether et al. (2009b). The daily relative short-selling activity is as follows:

(1)relssi,t=SSvoli,tvoli,t

SSvol is the number of shorted shares for stock i in day t. Vol is the number of traded shares for stock i in day t. We also construct the daily relative short-selling activity for individual investors, institutional investors, and foreign investors. The daily relative short-selling activity for each investor type is as follows:

(2)relssi,t,j=SSvoli,t,gvoli,t

where j is individual, institution and foreign investors.

Figure 1 shows the time-series trend of the market relss and stock market index for the days t – 50 to t + 50 (August 3, 2018 to January 3, 2019). Panel A shows the figure of KOSPI stocks, and Panel B reports the KOSDAQ stocks. For each day, we average relss across stocks as market relss. On average, the relss is inversely related to the stock market index, and it appears that short-selling activity increases before the stock market index drops. This result is consistent with Wang and Lee (2015) and Diether et al. (2009b), who show that short-sellers are contrarian in the Korean and US stock markets, respectively. After the event date (October 22, 2018), both the KOSPI market and the KOSDAQ market indexes declined, which suggests that the stock market does not positively react to the suspension of stock lending by the NPS. Another possible explanation is that around our event date, the stock market has a shock that causes a decline in the stock market.

3.2 Summary statistics of stock characteristics

In this section, we present the summary statistics of market quality and short-selling activity measures for sample stocks. Table 1 presents the summary statistics of the market quality measures for our sample stocks. We construct several market quality characteristics, such as daily return (Daily ret), Amihud's illiquidity (Illiq.), the bid-ask spread, price range, turnover (tv), order imbalance (OIB) and size. Panels A and B are the KOSPI and KOSDAQ markets, respectively.

For each panel, we divide the sample into three subsamples and two subperiods. The three subsamples are stock, stocks with NPS ownership and stocks without NPS ownership (hereafter NPS and NonNPS). The two subperiods are pre- and post-event periods: 50 days before the NPS suspends stock lending service (days t – 50 to t – 1) and 51 days after the NPS suspends stock lending service (days t to t + 50). The daily return for all sample stocks is negative for both the pre- and post-periods. Moreover, both NPS and NonNPS are negative for the pre-and post-periods. The daily stock return increases for both NPS and NonNPS after NPS suspends stock lending services for the KOSPI market. The average daily return is −0.124% in the pre-period and increases to −0.096% in the post-period for the NPS sample. Meanwhile, the NonNPS sample average daily return is −0.080% in the pre-period and increases to −0.049% in the post-period. In the KOSDAQ market, the post-period returns decrease for both NPS and NonNPS stocks. These results are inconsistent with Miller's (1977) overvaluation story, showing that stocks tend to have an upward bias in the presence of short-sale constraints. Our results are more close to Diamond and Verrecchia (1987). This issue is examined in more detail in the empirical analysis section.

Amihud's illiquidity, the bid–ask spread, price range and stdret increase in the post-period for the two markets, suggesting that market quality is not improved after NPS suspends stock lending. In the next section, we formally test the difference between the NPS and NonNPS stocks in terms of market quality. In Table 1, we can also find that NPS tends to be large, liquid, and have low volatility low turnover compared to NonNPS in both markets.

Table 2 shows summary statistics of short-selling activity measures. The daily relss for all KOSPI stocks are 3.539 and 3.498% in the pre- and post-periods, respectively. For all stocks listed in the KOSDAQ market, the daily relss is 1.587 and 1.505% in the pre- and post-periods, respectively. Diether et al. (2009b) show that in the US, the relss is 24 and 31% in NYSE and Nasdaq, respectively. Compare to the US, the daily trading volume generated from short-selling trading in Korea is tiny. We further classify the relss into the individual, institution, and foreign investors and find that foreign investors have high relss, which suggests that they are major players in the Korean stock market in terms of short-selling trading.

In Table 2, we can also find that relss is higher for NPS stocks than NonNPS stocks in both markets, consistent with Wang and Lee (2015), who show that large and liquid stocks have high short shares in the Korean stock market. In Panel A, the relss decreases from 5.135% to 5.084% and from 1.053% to 1.029% for NPS and NonNPS, respectively. These results are inconsistent with our expectations that as stock lending is binding, short-selling is reduced for NPS stocks rather than NonNPS stocks. We also compare the pre- and post-periods by investor type and find that individual and institutional investors short more, while foreign investors short less for NPS stocks after NPS suspends stock lending. However, the magnitude of the difference between pre and post is not enormous. This result is similar to that of the KOSDAQ market (Panel B), except for the case of institutional investors. The institutional relss decreases from 1.079% to 0.856% and from 0.120% to 0.100% for NPS and NonNPS, respectively.

Table 3 shows the ownership of the NPS for domestic stock at the end of 2017. Panel A shows the results for the KOSPI market, while Panel B shows the results for the KOSDAQ market. We divided the sample stocks into five groups based on NPS ownership at the end of 2017. On average, NPS ownership is 5.676% in the KOSPI market and 2.434% in the KOSDAQ market. The high group has an average ownership of 11.895%, suggesting that the NPS is the largest shareholder for some firms in the KOSPI market. Table 3 also shows that the NPS prefers large firms to small firms in the KOSPI market, but in the KOSDAQ market, firm size is inversely related to NPS ownership.

4. Empirical analysis

In this section, we provide evidence on the impact of NPS suspension of stock lending on stock market quality and short-selling activity. We first analyze whether the market quality and shorting activity change after the event date using the difference-in-differences (DID) analysis and regression analysis. We then examine the market reaction to the announcement of NPS suspending stock lending.

4.1 Does market quality improve after suspending stock lending?

In the previous section, we see that market quality measures are exacerbated after the event date. As the short-sales constraint is only for NPS, not for NonNPS, we need to test whether market quality or short-selling activity changes significantly for NPS stocks relative to NonNPS stocks. Table 4 presents the results of the DID analysis for the NPS and NonNPS stocks. Panels A and B report the results for the KOSPI and KOSDAQ markets, respectively. Our DID analysis follows Diether et al. (2009a). Specifically, for each variable each day, we compute the daily cross-sectional mean and then run a time-series regression with a dummy variable that equals one for the post-period and zero for the pre-period. To deal with the problem of possible cross-sectional correlation as well as time-series correlation in the sample stocks, we adjust the standard error following the Newey–West method based on lags of 20 days. The coefficient of the post-period dummy is “Diff” in the table. To test whether the market quality and short-selling activity change significantly for NPS relative to NonNPS stocks, we also use a time-series regression. For each day, we compute the difference between NPS stocks and NonNPS stocks and then run a time-series regression with a dummy variable that equals 1 for the post-period and zero for the pre-period. The coefficient of a dummy variable is “Diff-in-Diff” in the tables, which represents the changes for NPS stocks relative to NonNPS stocks.

In Panel A of Table 4, the NPS stocks become illiquid after suspension of stock lending by NPS, and Amihud's illiquidity is 0.004 and 0.006 in the post- and pre-periods, respectively. The bid-ask spread is wider, the pre-period is 0.352, and the post-period is 0.388. NonNPS stocks are also illiquid after suspension of stock lending. Amihud's illiquidity is 0.030 and 0.024 in the post- and pre-periods, respectively. The bid–ask spread is also wider: from 0.594 in the pre-period to 0.711 in the post-period. However, we find no significant improvement in market quality for NPS stocks relative to NonNPS stocks in the KOSPI market. The Diff-Diff for Illiq. is negative, but not significant. The bid-ask spread result shows that NonNPS stocks are more illiquid than NPS stocks, the Diff-Diff column shows −0.081, and is significant at the 5% level. Alternatively, the volatility measures measured by price range and standard deviation of daily return show that volatility is higher for both NPS and NonNPS stocks in post-period. The NonNPS stocks are more volatile than the NPS stocks; the Diff-Diff column shows −0.217 and −0.261 for the price range and stdret, respectively. However, only stdret is significant, suggesting that NonNPS stocks are more volatile than NPS stocks after the suspension of stock lending. Other variables, such as daily return, tv, and OIB, are all insignificant. These results are more prominent in the KOSDAQ market (Panel B), in that the market quality worsens for NonNPS stocks than for NPS stocks. The Diff-Diff column of Illiq, the bid-ask spread, price range and stdret are all negative and significant, suggesting that after the suspension of stock lending by the NPS, NonNPS stocks are more illiquid and volatile than NPS stocks. This is inconsistent with our hypothesis in that restrictions on short-selling may deteriorate market quality.

Next, we investigate the change in short-selling activity for NPS and NonNPS stocks. The results are presented in Table 5. We calculate the relss and abnormal relss from days t – 50 to t + 50 and investigate which investor group has a significant change in short-selling activity. We also calculate the relss and abnormal relss for individual, institutional, and foreign investors. Specifically, abnormal relss is defined as the difference between the daily relss and normal relss. We run the following regression to estimate the normal relss of individual stocks. The period we estimate is from days t – 370 to t – 70, and we skip one month:

(3)relssi,t=αi+β1i×ri,t−1+β2i×tvi,t+β3i×Bid−Ask spreadi,t+β4i×OIBi,t++β5i×relssi,t−1+β6i×mtreti,t

The abnormal relss is as follow

(4)abrelssi,t=relssi,t−(αi+β1i×ri,t−1+β2i×tvi,t+β3i×Bid−Ask spreadi,t+β4i×OIBi,t++β5i×relssi,t−1+β6i×mreti,t)

Panels A and B show the results of the KOSPI and KOSDAQ markets, respectively. In Panel A of Table 5, the overall and foreign relss are decreased for NPS stocks after suspension of stock lending – the diff column is −0.051 and −0.208, respectively. While individual and institutional relss are increased, the diff column is 0.010 and 0.147, respectively. For NonNPS stocks, only institutional relss increases, and overall, individual and foreign relss decrease. The Diff-Diff column shows that only individual relss is significant at the 1% level, and the NPS stock's individual relss increases more relative to NonNPS stocks. It is possible that individual investors may expect the stock will overvalue after NPS suspends stock lending, therefore they arbitrage against overpricing (Chang et al., 2014). The decrease in relss for institutional and foreign investors is because as NPS usually lends stock to institutional or foreign investors, these results imply that the suspension of stock lending does not affect the short-selling activity of institutional and foreign investors. The results are the same when we use abnormal relss. Panel B of Table 5 presents the results of the KOSDAQ market. Except for individual investors, overall, institutional, and foreign relss decrease after suspension of stock lending by the NPS. The diff column shows −0.307, −0.223 and −0.095 for the overall, institutional and foreign relss, respectively. This is also similar for NonNPS stocks, except that foreign relss is slightly increased. The Diff-Diff column shows that the overall relss is reduced for NPS stocks relative to NonNPS stocks. We classify the overall relss into three investors types; individual relss increases and institutional relss decreases. This result implies that institutional investors who want to short the stocks in the KOSDAQ market should borrow stocks first, and they usually borrow from the NPS. Therefore, the restriction of NPS stock lending will reduce institutional investors' short-selling activity rather than foreign investors' short-selling activity. The results are the same for abrelss.

To summarize, the DID analysis results are partly inconsistent with our expectation in that relss and abrelss are not decreased more for NPS stocks relative to NonNPS stocks after the NPS suspension of stock lending in the KOSPI market. However, there is a significant difference between NPS and NonNPS stocks for individual and institutional investors' short-selling activity after NPS suspends stock lending in the KOSDAQ market.

4.2 Regression analysis

4.2.1 Market quality

So far, we confirm that market quality and short-selling activity are not different between NPS and NonNPS stocks after the NPS suspension of stock lending services in the KOSPI market. In this section, we use regression analysis to examine the previous findings. Table 6 shows the results of regress market quality on the suspended lending dummy (Post), the NPS shareholding, and other control variables for the period from days t – 50 to t + 50. Panels A and B are the KOSPI and KOSDAQ markets, respectively. The specific regression equation is as follows:

(5)MQi,t=α+β1×Postt+β2×NPSi+β3×(Postt×NPSi)+other controls+ϵ

where MQ_i,t is the market quality measures for stock i at day t. The market quality measures are Amihud's illiquidity, the bid-ask spread, tv, stdret, price range and daily return.

Post is a dummy variable that equals 1 if the date is after suspension of stock lending, which is October 22, 2018. NPS is a dummy variable that equals 1 if stocks are held by the NPS, and zero otherwise. Other controls include firm size, the value-weighted average price (VWAP), market return and stock return. Overall, we find that the stop lending event deteriorates market quality, and the coefficients are positive and significantly related to the bid-ask spread, stdret and price range for the KOSPI market. NPS is negative and significant for all variables except Illiq, suggesting that stocks invested by the NPS tend to be liquid and less volatile. We interact Post with NPS to investigate whether NPS stocks are more likely to have high liquidity and low price volatility after the suspension of stock lending. The coefficients of Post × NPS are insignificant for Illiq, tv, price range, and daily return. The insignificant of Post × NPS for daily return suggesting that stocks are not overvalued after NPS suspends stock lending, consistent with Diamond and Verrecchia (1987). The coefficients of Post × NPS are negative and highly significant for bid-ask spread and stdret, suggesting that the restriction on short-selling activity leads to a narrower bid-ask spread and lower stock volatility for stocks with high NPS stocks. The market quality results are different for the KOSDAQ market. In Panel B, we can see that Post × NPS is significant for all variables except illiquidity and daily stock return measures. The NPS suspends stock lending, leading to a narrower bid-ask spread, reduced price volatility and increased turnover. The suspension of NPS stock lending may convey a positive signal to the stock market partly, because some of the market quality measures are not significant. Overall, in this section, we cannot conclude that suspension of NPS for stock lending improved market quality.

4.2.2 Short-selling and abnormal short-selling

Next, we conduct a regression analysis by regressing short-selling and abnormal short-selling on the suspended lending dummy (Post), NPS shareholding, and other control variables. The regression equation is as follows:

(6)relssi,t,j=α+β1×Postt+β2×NPSi+β3×(Postt×NPSi)+other controls+ε

relss is a relative short-selling activity, defined as the daily number of shorted shares divided by the daily trading volume. i, t, and j represent stock, day, and investor type, respectively. Investors are individual investors, institutional investors, and foreign investors. Post is a dummy variable that equals 1 if the date is after suspension of stock lending, which is October 22, 2018. NPS is the dummy variable that equals 1 if stocks are held by the NPS, and zero otherwise. Other controls include previous day relss, daily return, market return, firm size, the bid-ask spread, price range and order imbalance.

Table 6 shows the results of the relss. Panel A reports the results for the KOPSI market, while Panel B provides the results for the KOSDAQ market. In Panel A, the coefficients of Post × NPS are insignificant for all variables except Indi. relss. The coefficient of Post × NPS for Indi. relss is 0.0090 with a t-value of 3.33, implying that after NPS suspends stock lending, the Indi. relss increases if stocks are held by the NPS. This result is consistent with the DID analysis. The previous day's return (Ret_t−1) is positive and significant for all relss, Inst. relss, and Fore. relss, which means that short-sellers are contrarian. This is consistent with Diether et al. (2009b) and Wang and Lee (2015), who show that short-sellers are contrarian in the US and Korean stock markets, respectively. We also provide evidence that individual short-sellers are momentum traders, a finding consistent with Wang et al. (2017). Interestingly, we find that the bid-ask spread is negative and the price range is positive; the results are inconsistent with Diether et al. (2009b) and Wang and Lee (2015). Diether et al. (2009b) show both positive coefficients for the bid-ask spread and price range, and based on this result, they argue that short-sellers are opportunistic risk bearers in periods of high uncertainty caused by information asymmetry. Using Korean data, Wang and Lee (2015) find that both variables are negative and significant. However, in this study, we have a mixed result. The positive coefficient of the price range implies that short-sellers are opportunistic risk bearers, but the negative coefficient of the bid-ask spread shows that short-sellers increase short-selling when liquidity is high. We also find that OIB⁺ is positive and significant for all relss, Inst. relss, and Fore. relss, which suggests liquidity provision by short-sellers, in that they increase short-selling when the buying pressure is high. This result is also consistent with Diether et al. (2009b) and Wang and Lee (2015).

Panel B shows the results of the KOSDAQ market. The results are identical to those of the KOSPI market, except that OIB⁺ is not significant, and a few coefficients are negative. This result implies that unlike the KOSPI market, the short-sellers are not acting as liquidity providers in the KOSDAQ market.

Table 7 shows the results of the regression relss on interesting variables. We now consider the abnormal relss. As we want to investigate the effect on change in lending supply, we should control for normal relss that is not related to the event. We construct an abnormal relss: abrelss is an abnormal relss defined as the difference between the daily relss and the normal relss. We run Equation (3) by using the period from days t – 370 to t – 70 to estimate the normal relss. We also calculate abnormal individual relss, abnormal institution relss, and abnormal foreign relss in the same way.

(7)relssi,t=αi+β1i×ri,t−1+β2i×tvi,t+β3i×Bid−Ask spreadi,t+β4i×OIBi,t++β5i×relssi,t−1+β6i×mtreti,t

We calculate abnormal relss as follows:

In Table 8, we run the following regression.

(8)abrelssi,t,j=α+β1×Postt+β2×NPSi+β3×Postt×NPSi+other controls+ε

abrelss_i,t,j is an abnormal relss defined as in equations (3) and (4). i, t and j represent stock, day and investor type, respectively. Investors are individual investors, institutional investors, and foreign investors. Panels A and B represent the KOSPI and KOSDAQ markets, respectively. The results in Table 8 are identical to those listed in Table 7. After a suspension of stock lending, the abnormal relss of NPS stocks is not changing for all, Inst., and Fore. relss. However, individual investors short more for NPS stocks after the NPS suspends stock lending. It is also held for the KOSDAQ market, except that institutional investors short less after the NPS suspends stock lending, suggesting that the source of stocks supply by the NPS is important for institutional short-sellers but not for foreign short-sellers.

4.3 Does the stock market react to the announcement of suspension of stock lending?

In this section, we investigate the stock market's reaction to the suspension of stock lending by computing cumulative abnormal returns (CAR) after the NPS suspends stock lending. Miller (1977) argues that restrictions on short-sale leads to overvalued stocks. Meanwhile, Diamond and Verrecchia (1987) argue that investors will recognize short-sale constraints and adjust their beliefs so that stock prices do not tend to be upward biased, as suggested by Miller (1977). Using this exogenous shock, we test these two contradicting hypotheses. Following Brown and Warner (1980), we use the market model to generate ex ante expected returns. The estimated period is from days t – 300 to t – 51; we estimate Equation (10) for each stock to obtain alpha and beta, and use these coefficients to calculate event period abnormal return (AR).

(9)Ri,t=αi+βi×Rm,t+ϵi,t

(10)ARi,t=ri,t−αi+βi×rm,t

where R_i,t is the return of stock i at day t in the estimation period; R_m,t is the market return of day t in the estimation period. r_i,t is the return of stock i at day t in the event period; r_m,t is the market return of day t in the event period; and AR_i,t is the abnormal return of stock i in day t. To examine the stock return reaction to the announcement, we build two samples. First, we divide all sample stocks into two groups: stocks with NPS ownership and stocks without NPS ownership, and then calculate a variety of periods of CAR. Second, we divide stocks into four groups based on sample stocks' NPS ownership at the end of 2017 and examine a variety of periods of CAR. Tables 8 and 9 show the CARs around the announcement date.

We average stocks' CAR and report the average number and t-value in Table 8. CAR (0,1) is negative and significant for both NPS and NonNPS stocks, but the magnitude is larger for NonNPS stocks and the difference is not significant. The insignificant announcement return suggests that the stock market does not react to the event of NPS suspending stock lending. To test Miller's overvaluation hypotheses, we examine the CAR after the announcement date. As we can see from the table, CAR(0,1) to CAR(0,5) are all negative and significant for both NPS and NonNPS stocks, and the difference between these two samples is significant. However, based on these results we cannot argue that our results are consistent with Miller's overvaluation story. Our results are more close to Diamond and Verrecchia (1987). However, stock returns drop less for NPS stocks than NonNPS stocks which may provide little evidence that the market reaction is different for these two groups.

We also hypothesize that stocks with high NPS ownership are more likely to be affected by the suspension of stock lending events than stocks with low NPS ownership because NPS owns more shares for high NPS stocks so that the lendable shares are sharply reduced after the suspension of stock lending. We test this hypothesis in Table 10. We divided the stocks with NPS and low ownership into five groups based on the NPS ownership, and then we investigate the stock returns for these five groups. First, we present the KOSPI market results in Panel A. Our results show that CAR(0,1) to CAR(0,5) are all negative and significant for high NPS ownership stocks and low NPS ownership stocks. The difference between low and high NPS ownership groups is insignificant for CAR(0,1) but significant for CAR(0,2) to CAR(0,5), suggesting that after the event day, stock returns decline, but high NPS ownership stock returns drop less than low NPS ownership stock returns. In Panel B, the CAR(0,1) to CAR(0,5) are insignificant for both high and low NPS ownership stocks, and the difference between low and high ownership groups is not statistically significant for CAR(0,1) to CAR(0,5). These results are inconsistent with our hypotheses and suggest that in the KOSDAQ market, the suspension of stock supply event does not affect stock returns with NPS ownership. In conclusion, the overall results are inconsistent with Miller's (1977) overvaluation story and our results are more close to Diamond and Verrecchia (1987). We also find that there is no difference between low and high NPS ownerships in the KOSDAQ market.

5. Conclusion

The sudden suspension of the stock lending service by the NPS is a major shock to the stock market because the NPS is among the largest pension funds for retirement in Korea, as well as worldwide. Our findings are as follows. First, stocks with NPS ownership tend to have high relative short-selling, high liquidity, large size, and low stock volatility than stocks without NPS ownership. Second, we examine whether the NPS suspension of stock lending improved market quality. We find that stock return remained unchanged, suggesting that unlike investors' criticism, stock return is not affected by NPS lending the shares to short-sellers, especially to foreign short-sellers. We also find a mixed result on stock liquidity and volatility. In the DID analysis, we find that stocks are more illiquid and volatility is higher after the suspension of stock lending for both NPS and NonNPS stocks, but the increasing magnitude is smaller for NPS stocks than NonNPS stocks. However, we find contradictory results in the regression analysis: the bid-ask spread and standard deviation of daily return are reduced for stocks with NPS ownership after NPS suspends stock lending in the KOSPI market. For the KOSADQ market, we find a significant effect on the bid-ask spread, turnover, and price volatility. After the NPS suspension of stock lending, stocks with high NPS ownership tend to be more liquid and have low volatility. We cannot conclude that stock market quality is improved after the suspension of stock lending.

Third, we also examine whether the presence of restrictions on short-selling affected the short-selling activity. In the DID analysis, we find that individual relss and abrelss are increased for stocks with NPS ownership after NPS suspended stock lending for both markets. Interestingly, we find that institutional short-sellers reduce short-selling after NPS stops stock lending in the KOSDAQ market, suggesting that institutional investors mainly borrow shares from the NPS. In the regression framework, we find the same results as the DID analysis, suggesting that short-selling behavior by institutional and foreign investors remains unchanged. Lastly, we find evidence that stock returns are decreased for both samples, that is, stocks with and without NPS ownership for the KOSPI and KOSDAQ markets. We find that after NPS suspends stock lending, stock returns continuously drop, consistent with Diamond and Verrecchia (1987). However, the level of decrease is larger for NonNPS stocks than for NPS stocks, suggesting that market reaction is different for NPS and NonNPS stocks.

Figures

Figure 1

Short-selling activity

Table 1

Summary statistics of market quality variables

		Daily ret (%)	Illiq.	Bid-Ask spread	Price range (%)	Stdret (%)	tv (%)	OIB (%)	Size	N firms
Panel A: KOSPI market
All	Pre	−0.107	0.012	0.447	3.373	2.129	0.851	−12.107	1,903	729
	Post	−0.077	0.015	0.515	4.062	2.735	0.897	−10.560	1,739	729
NPS	Pre	−0.124	0.004	0.352	3.020	1.923	0.430	−9.314	3,026	444
	Post	−0.096	0.006	0.388	3.624	2.427	0.423	−7.615	2,770	444
NonNPS	Pre	−0.080	0.024	0.594	3.923	2.449	1.506	−16.458	153	285
	Post	−0.049	0.030	0.711	4.745	3.215	1.636	−15.150	133	285
Panel B: KOSDAQ market
All	Pre	−0.127	0.008	0.551	4.184	2.645	1.548	−16.768	208	1,115
	Post	−0.153	0.040	0.669	5.009	3.375	1.345	−14.872	178	1,116
NPS	Pre	−0.158	0.003	0.378	3.801	2.504	0.897	−12.831	486	253
	Post	−0.186	0.004	0.431	4.371	3.030	0.806	−11.758	425	253
NonNPS	Pre	−0.117	0.009	0.602	4.297	2.687	1.740	−17.923	126	862
	Post	−0.144	0.050	0.739	5.197	3.477	1.503	−15.785	105	863

Note(s): The table shows summary statistics of market quality variables for stocks with NPS ownership (NPS) and stocks without NPS ownership (NonNPS). Pre is the pre-event period from days t – 1 to t – 50. Post is post-event period from days t to t + 50. The event day is October 22, 2018, which is the date that NPS announces suspending stock lending. Daily Ret is the daily return, expressed as a percentage. Illiq. is Amihud illiquidity measure, defined as an absolute value of daily return divided by daily trading dollar value. Bid-Ask spread is a daily relative quoted spread defined as the difference between ask price and bid price and scaled by the midpoint of the bid and ask price. Price range is the intra-day highest price minus the intra-day lowest price scaled by the highest price. stdret is the standard deviation of daily return from t – 20 to t – 1. tv is turnover, which is daily trading volume divided by the number of shares outstanding. OIB is the order imbalance, defined as the number of daily buy-initiated shares minus daily sell-initiated shares divided by daily trading volume. Size is the market capitalization in billion KRW. All the numbers in the table are first average across stocks in a given day and then calculate the time-series average. Panels A and B are the KOSPI market and KOSDAQ market, respectively

Table 2

Summary statistics of short selling and stock lending

		SSvol	Indi. SSvol	Inst. SSvol	Fore. SSvol	SSval	Indi. SSval	Inst. SSval	Fore. SSval	relss	Indi. relss	Inst. relss	Fore. relss
Panel A: KOSPI market
All	Pre	13.314	0.215	3.673	9.426	516.270	3.890	173.495	338.884	3.539	0.016	1.007	2.516
	Post	13.509	0.243	4.029	9.237	486.460	4.619	179.532	302.309	3.498	0.022	1.110	2.367
NPS	Pre	16.866	0.127	5.472	11.267	826.987	5.344	281.351	540.292	5.135	0.016	1.561	3.557
	Post	17.183	0.187	5.942	11.054	782.276	7.016	292.499	482.761	5.084	0.026	1.708	3.349
NonNPS	Pre	7.779	0.352	0.869	6.558	32.205	1.625	5.468	25.112	1.053	0.016	0.142	0.895
	Post	7.786	0.330	1.049	6.407	25.563	0.886	3.526	21.151	1.029	0.015	0.177	0.837
Panel B: KOSDAQ market
All	Pre	5.390	0.150	0.749	4.492	84.708	2.381	25.572	56.755	1.587	0.013	0.337	1.236
	Post	5.581	0.179	0.749	4.653	82.810	1.623	25.303	55.884	1.505	0.018	0.271	1.216
NPS	Pre	7.515	0.125	1.982	5.408	248.020	4.263	88.525	155.232	3.536	0.022	1.079	2.435
	Post	7.542	0.165	2.109	5.267	253.535	3.617	93.304	156.614	3.229	0.034	0.856	2.340
NonNPS	Pre	4.767	0.157	0.387	4.223	36.802	1.829	7.104	27.870	1.015	0.011	0.120	0.885
	Post	5.007	0.183	0.350	4.473	32.757	1.038	5.367	26.352	1.000	0.013	0.100	0.886

Note(s): This table reports summary statistics of stock lending and short-selling variables for stocks with NPS ownership (NPS) and stocks without NPS ownership (NonNPS). Pre is the pre-event period from days t – 1 to t – 50. Post is post-event period from days t to t + 50. The event day is October 22, 2018, which is the date that NPS announces suspending stock lending. SSvol (Indi., Inst., and Fore) is the daily number of shorted shares (by the individual, institution and foreign investors) in thousand shares. SSval (Indi., Inst., and Fore) is the daily trading value (by the individual, institution, and foreign investors) in million KRW. relss (Indi., Inst., and Fore.) is a relative short-selling activity defined as the daily number of shorted shares (by the individual, institution and foreign investors) divided by the daily number of traded shares. All the numbers in the table are first average across stocks in a given day and then calculate the time-series average. Panels A and B are the KOSPI market and KOSDAQ market, respectively

Table 3

Summary statistics of NPS ownership

	Mean			Median			N firms
	Ownership value	Ownership (%)	Size	Ownership value	Ownership (%)	Size
Panel A: KOSPI market
All	264.97	5.676	3,315	28.25	5.366	599	458
Low	5.27	0.566	1,123	0.95	0.556	263	92
2	14.78	2.507	571	6.92	2.306	282	92
3	155.24	5.346	2,812	39.68	5.507	731	91
4	782.20	8.128	8,708	139.77	8.147	1,771	92
High	367.29	11.895	3,355	179.68	12.091	1,461	91
Panel B: KOSDAQ market
All	11.26	2.434	551	4.35	1.513	285	269
Low	0.81	0.219	719	0.60	0.227	316	54
2	4.08	0.668	703	1.78	0.683	266	54
3	6.43	1.600	373	3.73	1.537	250	54
4	14.71	3.054	475	9.43	2.881	302	54
High	30.64	6.708	482	22.68	5.261	352	53

Note(s): This table reports NPS ownership as of 2017 year-end. We divided the stocks into five groups based on NPS Ownership. NPS Ownership (%) is ownership of NPS at the end of 2017 defined as the number of shares held by NPS for a stock divided by the number of shares outstanding. Ownership value is the dollar value of NPS ownership defined as the number of shares held by NPS times the closing price at the end of 2017 expressed in billion KRW. Panels A and B are the KOSPI market and KOSDAQ market, respectively

Table 4

Changes in market quality around NPS suspension of stock lending

	NPS			NonNPS			Diff-Diff (NPS – NonNPS)
	Post	Pre	Diff (Post-Pre)	Post	Pre	Diff (Post-Pre)
Panel A: KOSPI market
Daily ret	−0.096	−0.124	0.028	−0.049	−0.080	0.031	−0.002
			0.12			0.10	−0.02
Illiq	0.006	0.004	0.002***	0.030	0.024	0.005	−0.003
			4.29			0.56	−0.36
Bid-Ask spread	0.388	0.352	0.036**	0.711	0.594	0.117**	−0.081**
			2.05			2.06	−2.04
Price range	3.624	3.020	0.605**	4.745	3.923	0.822*	−0.217
			2.00			1.89	−1.31
Stdret	2.427	1.923	0.504***	3.215	2.449	0.765***	−0.261
			3.03			3.28	−2.57**
Tv	0.423	0.430	−0.008	1.636	1.506	0.130	−0.137
			−0.23			0.83	−0.85
OIB	−7.615	−9.314	1.700	−15.150	−16.458	1.308	0.391
			1.59			0.76	0.45
Panel B: KOSDAQ market
Daily ret	−0.186	−0.158	−0.028	−0.144	−0.117	−0.026	−0.001
			−0.09			−0.08	−0.01
Illiq	0.004	0.003	0.002***	0.050	0.009	0.041*	−0.039*
			5.07			1.84	−1.76
Bid-Ask spread	0.431	0.378	0.053**	0.739	0.602	0.137**	−0.084**
			1.98			2.22	−2.39
Price range	4.371	3.801	0.569	5.197	4.297	0.900*	−0.331**
			1.47			1.85	−2.39
Stdret	3.030	2.504	0.525***	3.477	2.687	0.790***	−0.264**
			2.66			2.82	−2.29
Tv	0.806	0.897	−0.091	1.503	1.740	−0.236***	0.146
			−1.52			−2.64	1.38
OIB	−11.758	−12.831	1.073	−15.785	−17.923	2.138	−1.066
			0.69			1.07	−1.18

Note(s): This table reports the difference in difference analysis results of market quality measures for stocks with NPS ownership (NPS) and stocks without NPS ownership (NonNPS). The numbers in Pre and Post columns are time-series average of cross-stocks mean for the period of days t – 50 to t – 1 (August 3, 2018 to October 21, 2018) and days t to t + 50 (October 22, 2018 to January 3, 2019), respectively. The diff is the coefficient of the post dummy variable which regresses each variable listed below on an intercept and the post dummy variable. The post period is defined as after NPS suspends lending stocks from October 22, 2018 to January 3, 2019. The diff-in-Diff column shows the coefficient of the post dummy variable which regresses the difference between NPS stocks and NonNPS stocks for each variable on an intercept and post dummy variable. The t-value is for Diff-Diff, and we compute the standard error of t-value by using Newey–West adjusted with 20 lags. Daily Ret is the daily return, expressed as a percentage. Illiq. is Amihud illiquidity measure, defined as an absolute value of daily return divided by daily trading dollar value. Bid-Ask spread is a daily relative quoted spread defined as the difference between ask price and bid price and scaled by the midpoint of the bid and ask price. Price range is the intra-day highest price minus the intra-day lowest price scaled by the highest price. stdret is the standard deviation of daily return from t – 20 to t – 1. tv is turnover, which is daily trading volume divided by the number of shares outstanding. OIB is the order imbalance, defined as the number of daily buy-initiated shares minus daily sell-initiated shares divided by daily trading volume. Panels A and B are the KOSPI market and KOSDAQ market, respectively. The t-values are in italic and the ^***, ^**, and ^* are significance levels at 1%, 5%, and 10%, respectively

Table 5

Changes in shorting flow around NPS suspension of stock lending

	NPS			NonNPS			Diff-Diff (NPS – NonNPS)
	Post	Pre	Diff (Post-Pre)	Post	Pre	Diff (Post-Pre)
Panel A: KOSPI market
Relss	5.084	5.135	−0.051	1.029	1.053	−0.024	−0.027
			−0.17			−0.26	−0.12
Indi. relss	0.026	0.016	0.010***	0.015	0.016	−0.001	0.011***
			4.07			−0.20	3.88
Inst. relss	1.708	1.561	0.147	0.177	0.142	0.035	0.112
			1.16			1.45	0.91
Fore. relss	3.349	3.557	−0.208	0.837	0.895	−0.058	−0.151
			−0.94			−0.63	−1.05
abrelss	3.067	3.104	−0.037	0.683	0.690	−0.007	−0.030
			−0.19			−0.10	−0.21
Indi. abrelss	0.010	0.001	0.009***	0.004	0.004	0.000	0.009***
			3.69			−0.11	3.36
Inst. abrelss	1.224	1.108	0.116	0.143	0.102	0.041*	0.075
			1.22			1.84	0.81
Fore. abrelss	1.969	2.108	−0.139	0.542	0.579	−0.037	−0.102
			−0.97			−0.58	−1.17
Panel B: KOSDAQ market
Relss	3.229	3.536	−0.307	1.000	1.015	−0.016	−0.291*
			−1.42			−0.18	−1.78
Indi. relss	0.034	0.022	0.012***	0.013	0.011	0.003	0.009***
			2.82			1.59	2.83
Inst. relss	0.856	1.079	−0.223**	0.100	0.120	−0.020**	−0.204**
			−2.17			−2.09	−2.09
Fore. relss	2.340	2.435	−0.095	0.886	0.885	0.002	−0.097
			−0.54			0.02	−0.86
abrelss	2.156	2.337	−0.181	0.669	0.679	−0.011	−0.170
			−1.24			−0.17	−1.53
Indi. abrelss	0.011	−0.001	0.012***	0.002	−0.002	0.004	0.008***
			3.36			1.66	3.41
Inst. abrelss	0.617	0.785	−0.168**	0.066	0.081	−0.015*	−0.153**
			−2.16			−1.79	−2.07
Fore. abrelss	1.537	1.595	−0.058	0.594	0.595	−0.001	−0.057
			−0.49			−0.02	−0.77

Note(s): This table reports the difference in difference results analysis of short-selling measures for stocks with NPS ownership (NPS) and stocks without NPS ownership (NonNPS). Panels A and B report the KOSPI and KOSDAQ market, respectively. The numbers in Pre and Post columns are time-series average of cross-stocks mean for the period of days t – 50 to t – 1 (August 3, 2018 to October 21, 2018) and days t to t + 50 (October 22, 2018 to January 3, 2019), respectively. The diff is the coefficient of the post dummy variable which regresses each variable listed below on an intercept and the post dummy variable. The post period is defined as after NPS suspends lending stocks from October 22, 2018 to January 3, 2019. The Diff-Diff column shows the coefficient of the post dummy variable which regresses the difference between NPS stocks and NonNPS stocks for each variable on an intercept and post dummy variable. The t-value is for Diff-Diff, and we compute the standard error of t-value by using Newey–West adjusted with 20 lags. relss (Indi., Inst., and Fore.) is a relative short-selling activity defined as the daily number of shorted shares (by the individual, institution and foreign investors) divided by the daily number of traded shares. abrelss is abnormal relss defined as the difference between daily relss and the normal relss. We run eq. (3) by using the period from days t – 370 to t – 70 to estimate normal relss. We also calculate abnormal individual relss, abnormal institution relss and abnormal foreign relss in the same way

relssi,t=αi+β1i×ri,t−1+β2i×tvi,t+β3i×Bid−Ask spreadi,t+β4i×OIBi,t++β5i×relssi,t−1+β6i×mtreti,t (3)

We calculate abnormal relss as follows:

abrelssi,t=relssi,t−(αi+β1i×ri,t−1+β2i×tvi,t+β3i×Bid−Ask spreadi,t+β4i×OIBi,t++β5i×relssi,t−1+β6i×mreti,t) (4)

We also calculate abnormal individual relss, abnormal institution relss, and abnormal foreign relss in the same way. The t-values are in italic and the ^***, ^**, and ^* are significance levels at 1%, 5%, and 10%, respectively

Table 6

Regression of market quality around NPS suspension of stock lending

	Illiq	Bid-Ask spread	Tv	stdret	Price range	daily ret
Panel A: KOSPI market
Post	0.0064	0.1037***	−0.0040**	0.0077***	0.0086***	−0.0012
	0.72	3.84	−2.22	7.36	3.61	−0.50
NPS	−0.0093	−0.0605***	−0.0063***	−0.0062***	−0.0078***	0.0011
	−1.42	−2.96	−5.56	−6.38	−6.44	1.30
Post × NPS	−0.0036	−0.0783***	0.0000	−0.0026***	−0.0022*	0.0003
	−0.42	−3.85	0.03	−3.22	−1.68	0.18
ln(Size)	−0.0078***	−0.1011***	0.0009***	0.0007***	−0.0004	0.0003*
	−2.94	−14.23	3.24	2.75	−1.30	1.84
VWAP	0.0272*	0.2062**	−0.0006	−0.0055**	−0.0068**	0.0007
	1.89	2.56	−0.35	−2.52	−2.48	0.89
Price range	−0.2555***	0.1829	0.6601***			0.0164
	−3.16	0.38	10.45			0.20
Mret	−0.0260	−0.2286	−0.0572	0.0055	−0.2080**	0.5217**
	−0.40	−0.32	−1.43	0.25	−2.27	2.49
Ret	0.1418	−0.4813**	0.2417***	0.0089	0.1147*
	1.26	−2.36	6.97	0.94	1.92
Illiq						0.0005***
						4.53
Tv						0.1798***
						6.50
R2	0.001	0.250	0.280	0.109	0.080	0.1194
N	73,628	73,628	73,628	73,628	73,628	73,628
Panel B: KOSDAQ market
Post	0.0384	0.0980***	−0.0089***	0.0083***	0.0085***	0.0006
	1.43	3.68	−5.59	8.29	3.58	0.43
NPS	0.0568	0.0047	−0.0061***	−0.0042***	−0.0050***	0.0005
	1.27	0.29	−4.80	−4.77	−4.70	0.68
Post × NPS	−0.0397	−0.0779***	0.0039***	−0.0027***	−0.0033***	0.0000
	−1.38	−4.44	3.98	−3.96	−2.91	0.03
ln(Size)	−0.0709	−0.2052***	0.0016**	0.0022***	0.0003	0.0007
	−1.40	−16.84	2.28	5.07	0.46	1.63
VWAP	0.8027	0.8456***	−0.0199**	−0.0113	−0.0095	−0.0002
	1.27	2.85	−1.97	−1.49	−0.94	−0.06
Price range	−0.8397	0.7552	0.7497***			0.0846
	−1.29	1.17	12.94			1.30
Mret	0.0663	0.2466	−0.0813	0.0524**	−0.4387***	0.9425***
	0.41	0.16	−1.10	2.15	−3.98	15.30
Ret	0.1412	−0.7988***	0.2744***	−0.0008	0.1940***
	1.29	−3.63	9.83	−0.13	3.96
Illiq						0.0000
						1.33
Tv						0.1554***
						7.71
R2	0.001	0.223	0.298	0.074	0.086	0.2582
N	112,686	112,686	112,686	112,669	112,686	112,686

Note(s): The table reports regression results of market quality measures on suspension of stock lending date dummy, NPS ownership, and other characteristics. Illiq. is Amihud illiquidity measure, defined as the absolute value of daily return divided by daily trading dollar value. Bid-Ask spread is the daily relative quoted spread, defined as the difference between ask price and bid price and scaled by the midpoint of bid and ask price. tv is turnover, which is daily trading volume divided by the number of shares outstanding. Price range is the intra-day highest price minus the intra-day lowest price scaled by the highest price. Post is a dummy variable that takes a value of 1 for the period after NPS suspension of stock lending, and zero otherwise. NPS is a dummy variable that takes a value of 1 if stocks are owned by NPS and zero otherwise. ln(Size) is log market capitalization. VWAP is value-weighted average price, defined as the total trading dollar amount divided by the total number of shares traded on a given day. Mret is daily market return, and Ret is the daily stock return. Panels A and B report the KOSPI and KOSDAQ market, respectively. The intercepts are estimated, but are not report. The t-values are in italic and standard errors are clustered by stock and day. The ^***, ^**, and ^* are significance levels at 1%, 5%, and 10%, respectively

Table 7

Regression of shorting flow around NPS suspension of stock lending

	relss	Indi. relss	Inst. relss	Fore. relss
Panel A: KOSPI market
Post	0.0936	−0.0023	0.0678**	0.0414
	1.59	−1.08	2.06	0.88
NPS	0.4438***	−0.0076**	0.0832	0.3749***
	3.92	−2.47	1.30	4.59
Post × NPS	−0.0639	0.0090***	0.0548	−0.1127
	−0.53	3.33	0.71	−1.28
Dept-1	0.5389***	0.2177***	0.4323***	0.5447***
	35.88	6.63	17.14	35.41
Ret	−0.1073***	0.0020***	−0.0451***	−0.0656***
	−7.67	5.30	−6.85	−6.80
Rett-1	0.0773***	−0.0007**	0.0152***	0.0590***
	5.69	−2.11	2.62	6.26
Mret	−0.0482	−0.0012**	−0.0164	−0.0310
	−1.33	−2.56	−1.04	−1.38
ln(Size)	0.7255***	0.0051***	0.3804***	0.4070***
	11.44	5.41	10.33	10.12
Bid-Ask spread	−0.3454***	−0.0084***	−0.0270	−0.3222***
	−3.75	−4.56	−0.46	−5.04
Price range	0.0316**	0.0042***	0.0082	0.0206**
	2.43	9.47	1.14	2.33
OIB+	1.5309***	0.0045	0.3901***	1.1650***
	6.18	1.12	3.46	6.02
R2	0.470	0.064	0.301	0.413
N	73,627	73,627	73,627	73,627
Panel B: KOSDAQ market
Post	0.0987**	0.0023*	0.0324***	0.0687*
	2.31	1.85	2.81	1.74
NPS	0.4944***	−0.0042	0.2343***	0.2899***
	4.90	−1.20	3.82	4.04
Post × NPS	−0.1680	0.0069***	−0.1160	−0.0717
	−1.57	2.64	−1.53	−0.89
Dept-1	0.4495***	0.2595***	0.3840***	0.4484***
	40.50	6.91	17.01	37.70
Ret	−0.0421***	0.0016***	−0.0154***	−0.0291***
	−6.24	8.78	−7.48	−4.91
Rett-1	0.0351***	−0.0009***	0.0051***	0.0300***
	6.39	−5.73	2.82	6.75
Mret	−0.0399**	−0.0009**	0.0040	−0.0425***
	−2.21	−2.57	0.72	−2.72
ln(Size)	0.7479***	0.0105***	0.3017***	0.4713***
	15.89	5.59	10.13	15.28
Bid-Ask spread	−0.0414	−0.0068***	0.0420***	−0.0748**
	−0.97	−5.27	2.85	−2.02
Price range	0.0053	0.0023***	−0.0023	0.0054
	0.72	10.33	−1.06	0.83
OIB+	−0.0307	0.0023	0.0258	−0.0436
	−0.25	0.67	0.61	−0.43
R2	0.373	0.087	0.256	0.307
N	112,679	112,679	112,679	112,679

Note(s): The table reports regression results of relative short-selling on suspension of stock lending date dummy, NPS ownership, and other characteristics. relss (Indi., Inst., and Fore.) is a relative short-selling activity defined as the daily number of shorted shares (by the individual, institution and foreign investors) divided by the daily number of traded shares. abrelss is abnormal relss defined as the difference between daily relss and the normal relss. We run eq. (3) by using the period from days t – 370 to t – 70 to estimate normal relss. We also calculate abnormal individual relss, abnormal institution relss, and abnormal foreign relss in the same way

relssi,t=αi+β1i×ri,t−1+β2i×tvi,t+β3i×Bid−Ask spreadi,t+β4i×OIBi,t++β5i×relssi,t−1+β6i×mtreti,t (3)

We calculate abnormal relss as follows:

abrelssi,t=relssi,t−(αi+β1i×ri,t−1+β2i×tvi,t+β3i×Bid−Ask spreadi,t+β4i×OIBi,t++β5i×relssi,t−1+β6i×mreti,t) (4)

We also calculate abnormal individual relss, abnormal institution relss, and abnormal foreign relss in the same way. Post is a dummy variable that takes a value of 1 for the period after NPS suspension of stock lending, and zero otherwise. NPS is a dummy variable that takes a value of 1 if stocks are owned by NPS and zero otherwise. Dep_t-1 is day t – 1 dependent variable. Ret is the daily return. Ret_t–1 is the daily return on t – 1. Mret is daily market return. Bid-Ask spread is the daily relative quoted spread, defined as the difference between ask price and bid price and scaled by the midpoint of bid and ask price. Price range is the intra-day highest price minus the intra-day lowest price scaled by the highest price. OIB is the order imbalance defined as number of daily buy-initiated shares minus daily sell-initiated shares divided by daily trading volume. OIB⁺ equals OIB if OIB greater than 0, and zero otherwise. Panels A and B report the KOSPI and KOSDAQ market, respectively. The intercepts are estimated, but are not report. The t-values are in italic, and standard errors are clustered by stock and day. The ^***, ^**, and ^* are significance levels at 1%, 5%, and 10%, respectively

Table 8

Regression of abnormal shorting flow around NPS suspension of stock lending

	abrelss	Indi. abrelss	Inst. abrelss	Fore. abrelss
Panel A: KOSPI market
Post	0.1309*	−0.0023	0.0978***	0.0532
	1.90	−0.99	2.69	0.94
NPS	0.6245***	−0.0010	0.1907**	0.4306***
	4.35	−0.25	2.48	4.14
Post × NPS	−0.0703	0.0098***	0.0666	−0.1349
	−0.53	3.20	0.78	−1.35
Ret	−0.1221***	0.0020***	−0.0471***	−0.0783***
	−8.18	5.09	−6.92	−7.60
Rett-1	0.0581***	−0.0026***	0.0067	0.0450***
	4.38	−7.22	1.18	4.72
Mret	−0.0477	−0.0012**	−0.0182	−0.0272
	−1.38	−2.32	−1.39	−1.17
ln(Size)	0.8920***	−0.0005	0.4229***	0.5327***
	12.78	−0.28	10.37	10.79
Bid-Ask spread	−0.4393***	−0.0026	−0.0475	−0.3919***
	−3.92	−1.00	−0.63	−5.12
Price range	0.0403***	0.0029***	0.0073	0.0309***
	2.74	5.25	0.88	3.05
OIB+	1.8661***	0.0243***	0.5582***	1.3734***
	6.44	4.97	4.24	5.94
R2	0.137	0.134	0.078	0.087
N	73,627	73,627	73,627	73,627
Panel B: KOSDAQ market
Post	0.1225***	0.0032**	0.0357***	0.0873**
	2.67	2.16	2.74	2.05
NPS	0.6072***	−0.0053	0.3145***	0.3092***
	5.38	−0.96	4.66	3.64
Post × NPS	−0.1762	0.0093***	−0.1513*	−0.0645
	−1.48	3.05	−1.90	−0.69
Ret	−0.0471***	0.0016***	−0.0165***	−0.0329***
	−6.59	6.97	−7.76	−5.31
Rett-1	0.0260***	−0.0027***	0.0012	0.0224***
	4.94	−12.82	0.60	5.29
Mret	−0.0393**	−0.0008**	0.0045	−0.0433***
	−2.22	−2.34	0.79	−2.81
ln(Size)	0.8797***	0.0049*	0.3334***	0.5705***
	15.44	1.68	9.37	15.40
Bid-Ask spread	−0.0425	−0.0013	0.0439***	−0.0760**
	−0.93	−0.64	2.59	−1.96
Price range	0.0077	0.0014***	−0.0040*	0.0078
	1.00	4.60	−1.71	1.17
OIB+	0.0642	0.0417***	0.0861*	0.0297
	0.49	8.41	1.95	0.27
R2	0.125	0.012	0.076	0.076
N	112,603	112,603	112,603	112,603

Note(s): The table reports regression results of abnormal relative short-selling measures on suspension of stock lending date dummy, NPS ownership, and other characteristics. relss (Indi., Inst., and Fore.) is a relative short-selling activity defined as the daily number of shorted shares (by the individual, institution, and foreign investors) divided by the daily number of traded shares. abrelss is abnormal relss defined as the difference between daily relss and the normal relss. We run eq. (3) by using the period from days t – 370 to t – 70 to estimate normal relss

relssi,t=αi+β1i×ri,t−1+β2i×tvi,t+β3i×Bid−Ask spreadi,t+β4i×OIBi,t++β5i×relssi,t−1+β6i×mtreti,t (3)

We calculate abnormal relss as follows:

abrelssi,t=relssi,t−(αi+β1i×ri,t−1+β2i×tvi,t+β3i×Bid−Ask spreadi,t+β4i×OIBi,t++β5i×relssi,t−1+β6i×mreti,t) (4)

We also calculate abnormal individual relss, abnormal institution relss and abnormal foreign relss in the same way. Post is a dummy variable that takes a value of 1 for the period after the NPS suspension of stock lending, and zero otherwise. NPS is a dummy variable that takes a value of 1 if stocks are owned by NPS and zero otherwise. Ret is the daily return. Ret_t-1 is the daily return on t – 1. Mret is daily market return. Bid-Ask spread is the daily relative quoted spread, defined as the difference between ask price and bid price and scaled by the midpoint of bid and ask price. Price range is the intra-day highest price minus the intra-day lowest price scaled by the highest price. OIB is the order imbalance, defined as number of daily buy-initiated shares minus daily sell-initiated shares divided by daily trading volume. OIB⁺ equals OIB if OIB greater than 0, and zero otherwise. Panels A and B report the KOSPI and KOSDAQ market, respectively. The intercepts are estimated, but are not report. The t-values are in the parentheses and standard errors are clustered by stock and day. The ^***, ^**, and ^* significance levels at 1%, 5%, and 10%, respectively

Table 9

Cumulative abnormal return around NPS suspension of stock lending

	Panel A: KOSPI			Panel B: KOSDAQ
	NPS	NonNPS	Diff (NPS-NonNPS)	NPS	NonNPS	Diff (NPS-NonNPS)
CAR(-1,1)	−1.952***	−2.383***	0.431	−0.288	−1.418***	1.129***
	−13.12	−7.04	1.17	−1.09	−8.12	3.57
CAR(0,1)	−1.920***	−2.333***	0.413	−0.280	−0.867***	0.587***
	−14.66	−8.00	1.29	−1.33	−6.23	2.32
CAR(0,2)	−2.188***	−4.283***	2.095***	−0.103	−1.728***	1.625***
	−10.10	−10.77	4.63	−0.35	−9.11	4.64
CAR(0,3)	−3.503***	−6.399***	2.896***	−0.590*	−3.207***	2.618***
	−13.63	−13.80	5.46	−1.70	−14.23	6.32
CAR(0,4)	−5.305***	−8.861***	3.556***	−1.178***	−4.603***	3.425***
	−18.03	−16.60	5.84	−2.61	−16.84	6.49
CAR(0,5)	−8.690***	−15.868***	7.179***	−1.733***	−9.861***	8.128***
	−23.19	−24.79	9.68	−2.79	−28.76	11.45
CAR(0,10)	−4.812***	−6.116***	1.304*	−0.750	−3.256***	2.506***
	−13.92	−10.30	1.90	−1.47	−10.81	4.03
CAR(0,30)	−0.063	−3.155***	3.092***	1.571**	−0.931*	2.502***
	−0.12	−3.36	2.87	2.04	−1.88	2.73
CAR(0,50)	−5.400***	−5.819***	0.419	−0.063	−0.198	0.135
	−7.07	−3.97	0.25	−0.06	−0.26	0.10

Note(s): This table reports the mean of cumulative abnormal return (CAR) for stocks with NPS ownership (NPS) and stocks without NPS ownership (NonNPS). Abnormal return is defined as the difference between a daily return and a daily normal return. We estimate normal return using a market model for the past 250 days (days t – 300 to t – 51). Panels A and B report the KOSPI and KOSDAQ market, respectively. The t-values are in the parentheses and ^***, ^**, and ^* represent the significant levels at 1%, 5%, and 10%, respectively

Table 10

Cumulative Abnormal Return (CAR) by NPS ownership

	Low	2	3	4	High	Diff (H-L)
Panel A: KOSPI
CAR(−1,1)	−2.187***	−2.125***	−1.654***	−1.849***	−1.933***	0.254
	−7.19	−6.43	−4.75	−4.96	−6.32	0.59
CAR(0,1)	−2.327***	−1.923***	−1.945***	−1.735***	−1.672***	0.654
	−8.27	−6.71	−6.02	−5.84	−6.01	1.65
CAR(0,2)	−3.485***	−2.988***	−1.939***	−1.636***	−0.861*	2.625***
	−8.55	−5.36	−4.32	−3.66	−1.72	4.07
CAR(0,3)	−5.018***	−4.776***	−3.048***	−2.339***	−2.298***	2.720***
	−10.59	−7.61	−5.55	−4.27	−3.82	3.55
CAR(0,4)	−6.733***	−6.737***	−4.994***	−3.825***	−4.209***	2.524***
	−10.86	−9.09	−8.71	−6.25	−6.29	2.77
CAR(0,5)	−10.925***	−10.870***	−8.256***	−6.190***	−7.177***	3.748***
	−12.62	−12.20	−11.40	−8.04	−8.90	3.17
CAR(0,10)	−5.428***	−5.806***	−4.691***	−3.781***	−4.341***	1.087
	−7.16	−7.55	−7.56	−4.81	−4.87	0.93
CAR(0,30)	−0.251	−2.004	−0.212	1.037	1.140	1.391
	−0.21	−1.50	−0.19	0.87	1.07	0.88
CAR(0,50)	−5.812***	−6.812***	−5.930***	−3.267**	−5.216***	0.596
	−3.55	−3.92	−3.22	−1.96	−3.11	0.25
Panel B: KOSDAQ
CAR(−1,1)	−1.230**	−0.612	0.311	−0.021	0.124	1.354*
	−2.29	−1.15	0.43	−0.04	0.21	1.70
CAR(0,1)	−0.610	−0.427	−0.036	−0.445	0.125	0.734
	−1.58	−0.90	−0.07	−0.96	0.25	1.18
CAR(0,2)	−0.096	−0.626	−0.084	−0.235	0.518	0.613
	−0.16	−0.94	−0.11	−0.43	0.73	0.67
CAR(0,3)	−0.667	−1.132	−0.860	−0.502	0.207	0.875
	−0.99	−1.40	−1.09	−0.63	0.25	0.82
CAR(0,4)	−0.794	−2.418**	−1.690	−0.814	−0.213	0.581
	−0.92	−2.23	−1.46	−0.84	−0.22	0.45
CAR(0,5)	−1.098	−3.024**	−3.768***	−0.399	−0.412	0.686
	−0.91	−2.16	−2.67	−0.27	−0.29	0.37
CAR(0,10)	0.918	−2.213**	−0.998	0.023	−1.583	−2.501
	0.83	−2.37	−0.88	0.02	−1.27	−1.50
CAR(0,30)	3.047*	−0.731	3.794**	1.448	0.148	−2.899
	1.90	−0.41	2.29	0.68	0.11	−1.39
CAR(0,50)	3.022	−0.990	2.162	−1.334	−3.337	−6.359**
	1.34	−0.45	0.89	−0.42	−1.60	−2.07

Note(s): This table reports the mean of cumulative abnormal return (CAR) for stocks with NPS ownership (NPS). We sort the stocks into four groups based on NPS ownership. Abnormal return is defined as the difference between a daily return and daily normal return. We estimate normal return by using a market model for the past 250 days (days t – 300 to t – 51). Panels A and B report the KOSPI and KOSDAQ market, respectively. The t-values are in the parentheses and ^***, ^**, and ^* represent the significant levels at 1%, 5%, and 10%, respectively

Notes

1. Wang and Lee (2015) find that the average daily short-selling shares to daily traded shares is 3.16% from January 1, 2006 to May 31, 2010; it is 24 and 31% for NYSE and Nasdaq, respectively (Diether et al., 2009b).

2. As of the end of June 2021.

3. The source of data is from Kim, Sang-hee, a member of the Democratic Party of Korea (http://www.businesskorea.co.kr/news/articleView.html?idxno=25932).

4. https://fund.nps.or.kr/jsppage/fund/fund_main_e.jsp

5. Under the fund management guidelines and regulations, the details of each asset class at the end of the fiscal year is disclosed in the third quarter of the following year (Article 25 of the Fund Management Guidelines and Article 39 of the Fund Management Regulation).

6. We can clearly distinguish which trade is buyer-initiated trade or seller-initiated trade without relying on Lee and Ready's (1991) trade classification algorithm.