Hypotheses

Given that institutional traders are informed but that they on average sell stocks for non-information purposes, and also given that the institutional trading data are disclosed

to the public at the end of each trading day, we propose three hypotheses to explore the contemporary relationship between volatility and institutional trading: the dispersion of beliefs hypothesis, which is applicable to institutional purchases; the discretionary liquidity hypothesis, which focuses on the relationships for expected institutional purchases and sales; and the liquidity demand surprise hypothesis, which provides predictions for unexpected institutional purchases and sales.

The dispersion of beliefs hypothesis gives rise to a negative relationship between volatility and institutional purchases. Daigler and Wiley (1999) suggest that informed traders have relatively homogeneous beliefs because these traders possess more resources that they can use to obtain and analyze private information; as a result, informed trades are executed at prices relatively close to the fair value of the asset, which helps stabilize prices. Accordingly, institutional purchases and volatility should be negatively related since institutional traders utilize their private information through purchases. The hypothesis, however, gives no prediction for institutional sales because selling overall does not convey much private information.

The discretionary liquidity hypothesis, enlightened by Admati and Pfleiderer (1988), interprets the relationship between volatility and expected institutional trading as a result of the optimizing behavior of discretionary liquidity traders. Discretionary liquidity traders are uninformed liquidity demanders who can choose when to trade within a given period of time. As Admati and Pfleiderer (1988) have indicated, these traders prefer to trade together in order to reduce their trading impact on prices. Since institutional trading data are disclosed to the public at the end of each trading day, discretionary liquidity traders utilize the data to decide whether to trade on the next day.

When an increase in institutional sales has been expected, this expected heavier uninformed trading will attract more discretionary liquidity traders to join the market,

and these new entries of discretionary liquidity traders will help stabilize prices.

Similarly, an expected decrease in institutional sales will reduce the discretionary liquidity traders’ incentive to participate in the market; accordingly, there will be fewer discretionary liquidity traders in the market, and prices will be more volatile. To sum up, the discretionary liquidity hypothesis predicts a negative relationship between volatility and expected institutional sales.

In contrast to the decisive relationship for expected institutional sales, the hypothesized relationship between volatility and expected institutional purchases depends on whether private information is diversified among institutional traders. As Admati and Pfleiderer (1988) have shown in their model, because informed traders who possess the same information will compete with each other, more entries by these informed traders will improve the welfare of discretionary liquidity traders; on the other hand, however, if private information is diversified among informed traders, more entries by informed traders will increase the total amount of private information, which worsens the terms of trade for discretionary liquidity traders. In other words, when private information is identical (complementary), discretionary liquidity traders will have greater (less) incentive to join the market once they have expected an increase in institutional purchases, which translates into a negative (positive) relationship between volatility and expected institutional purchases.

The liquidity demand surprise hypothesis suggests that a positive relationship exists between volatility and unexpected institutional trading. This hypothesis postulates that liquidity providers predetermine their daily liquidity supply before each trading day.

When institutional traders’ actual liquidity demand surprises liquidity providers and is greater than the expected level, there will be a shortage in liquidity supply. As a result, price will be pushed to an unusual level, and a greater degree of volatility will be

Figure 1. The Hypothesized Relationships between Volatility and Institutional Trading

Dispersion of Beliefs

Volatility ↓

Expected Purchases↑ Identical

Information

Discretionary Liquidity Volatility ↓

Volatility ↑

Complementary Information

Expected Sales↑ ^Volatility ↓

Volatility ↓

Unexpected Purchases↑

Liquidity Demand Surprise

Volatility ↑

Unexpected Sales↑ ^Volatility ↑

Discretionary Liquidity

Dispersion of Beliefs

Liquidity Demand Surprise

observed. Given that institutional traders are net liquidity demanders both in buying and selling, unexpected institutional purchases and sales respectively represent institutional traders’ excess liquidity demand in buying and selling and hence are both positively related to volatility. This hypothesis is parallel to Fagan and Gencay (2008)’s finding that the scarcity of counterparties occasionally occurs even in a large and active market;

the counterparty scarcity can exhaust liquidity, thereby causing an increase in volatility.

Figure 1 summarizes the hypothesized contemporary relationship between volatility and institutional trading. The hypotheses provide decisive relationships for institutional sales. Specifically, the discretionary liquidity hypothesis predicts a negative relationship between volatility and expected institutional sales, while the liquidity demand surprise hypothesis predicts a positive relationship between volatility and unexpected institutional sales.

Different information content for institutional purchases and sales translates into asymmetric effects on the volatility-volume relationship. The liquidity demand surprise hypothesis predicts a positive relationship for both unexpected purchases and sales, while the dispersion of beliefs hypothesis predicts a negative relationship only for unexpected purchases. As a result, unexpected institutional purchases tend to have a less positive effect on volatility than unexpected institutional sales.

In addition, as a result of different information content, the relationships for expected institutional purchases and sales are grounded in different scenarios and hence are unlikely to be symmetric. The relationship between volatility and expected institutional purchases is negative if the dispersion of beliefs hypothesis dominates or institutional traders possess identical private information.

3. Data

The Taiwan Stock Exchange is an order driven market where orders are automatically matched through a fully computerized order book system. Detailed order book information, however, is not publicly available. Despite the fact that the system only accepts limit orders, traders can submit aggressive price-limit orders to obtain matching priority. Since 2001, trading has taken place between 9:00 a.m. and 1:30 p.m.

Monday to Friday.

To test our hypotheses, we examine the trading from foreigners and mutual funds.

Previous research suggests that foreigners and mutual funds trade on information because they are aggressive traders and consistently make profits from trade (Barber, Lee, Liu, and Odean, forthcoming). Both groups of investors have also been prohibited

by regulations from selling short until June 25, 2005.¹ Starting from June 25, the Taiwan Stock Exchange has allowed institutional investors to borrow shares from a centralized system and sell them on the exchange.

Our sample period, extending from December 12, 2000 to June 24, 2005, spans the day on which the market began to disclose the daily number of shares bought and sold by institutional traders, and the day before on which short selling by institutional traders was prohibited. Since the objective is to investigate the effect of institutional trading on volatility, we restrict our sample to stocks that are heavily traded by institutional traders

— the Taiwan 50 index constituent stocks. In 2007, institutional traders traded a total of US$257 billion for all 698 stocks listed on the exchange, while 57% of the money was concentrated in these 50 stocks. We require each stock to have a minimum of 150 daily observations, with the result that 46 stocks remain in the sample. The data source is the Taiwan Economic Journal.

We measure trading activities by share volumes. Specifically, total volume is defined as the total number of shares of a stock traded on a particular day. Institutional purchases (sales) are defined as the total number of shares of a stock bought (sold) by institutional traders on a particular day. To check robustness, we also measure trading activities based on turnover instead of share volume, where turnover is defined as the ratio of share volume to the number of shares outstanding. Equivalent results are obtained.

Panel A of Table 1 provides time-series statistics for the return, absolute return, and trading activity variables. Trading is very frequent, with a daily average of 3,724 trades

1 Article 10 of the Regulations Governing Securities Investment Trust Funds forbids mutual funds and Article 21 of the Regulations Governing Investment in Securities by Overseas Chinese and Foreign Nationals forbids foreigners from selling short.

Table 1: Summary Statistics

Panel A: Time-series statistics

The sample comprises daily data of the Taiwan 50 index constituent stocks from 2000/12/12 to 2005/6/24; conforming to the requirement of a minimum of 150 daily observations, 46 stocks remain in the sample. For each variable of each stock, we calculate several time-series statistics, including the mean, standard deviation, and the partial autocorrelations at the first five lags. The cross-sectional means of the time-series statistics are reported.

Partial Autocorrelation at Lag Mean

Standard

Deviation 1 2 3 4 5

Return (%) 0.007 2.555 0.033 -0.015 0.008 -0.011 -0.008

Absolute Return (%) 1.840 1.768 0.144 0.125 0.094 0.085 0.072

Total Volume (10⁷ shares) 2.036 1.711 0.637 0.190 0.121 0.088 0.077 Number of Trades (10⁴ trades) 0.372 0.269 0.676 0.193 0.129 0.097 0.083 Institutional Purchases (10⁷ shares) 0.535 0.518 0.505 0.155 0.090 0.074 0.071 Institutional Sales (10⁷ shares) 0.486 0.458 0.467 0.165 0.114 0.087 0.093

Panel B: Pearson correlation coefficients across trading activity variables

The sample comprises daily data of the Taiwan 50 index constituent stocks from 2000/12/12 to 2005/6/24; conforming to the requirement of a minimum of 150 daily observations, 46 stocks remain in the sample. For each stock, we calculate the Pearson correlation coefficients across five trading activity variables, including expected institutional purchases, expected institutional sales, unexpected institutional purchases, unexpected institutional sales and the number of trades. The cross-sectional means of the correlation coefficients for all pairs of trading activity variables are reported.

Expected Unexpected

Institutional Sales -0.03 -0.03

Unexpected Institutional Purchases 0.25

Number of Trades 0.33 0.24 0.35 0.27

per stock. Institutional trading constitutes one quarter of total volume. Trading activity variables are highly autocorrelated; each trading activity variable has a first-order partial autocorrelation that is three or four times greater than the first-order partial autocorrelation of the absolute return series.