Is the incremental transparency necessary?

ABSTRACT

This paper examines how liquidity, price discovery and the causality between them are affected by incremental transparency on the Taiwan Stock Exchange (TSEC). We find that it has small effects on market quality, with the impact being focused on small- or medium-sized firms. We also find that the causality between liquidity and price discovery is weakened after the event. This study is a supplement to the existing literature and our results imply that if the market is already considerably transparent, then the effects of further opening up the limit-order book may be limited.

KEYWORDS:

• Transparency
• liquidity
• price discovery
• limit-order book
• causality

Notes

1 The best one bid/ask respectively mean the highest unexecuted bid price and the lowest unexecuted ask price from the limit-order books.

2 The daily price variation limit for stocks on the TSEC has been loosened to 10% since 1 June 2015.

3 In fact, informed traders are not obviously identified since they tend to disguise themselves and their trading motives. Therefore, empirical researchers try to propose reasonable proxies for what comprise informed (uninformed) traders. Determining proxies of what comprise informed (uninformed) traders is beyond the scope of this study. Institutions, for example, are assumed to be informed traders, while individuals are viewed as uninformed traders when they are simultaneously involved in a trading market (Anand, Chakravarty, and Martell, 2005). The other case for informed traders versus uninformed traders takes places when looking at market order traders versus limit order traders. See Madhavan, Porter, and Weaver (2005) for a more detail discussion.

4 We follow Lee and Ready’s (1991) method to decide the trading direction. If the traded price is larger than the mean of the bid and ask quotes, then we assume that traders will buy the stock and q_it is +1. If the traded price is smaller than the average of the bid and ask quotes, then it is thought that traders will sell the stock and q_it is −1.

5 We follow the setting of Hasbrouck (1991b). In the following test, we also try lagging five periods, but the results are similar.

6 The stock subscript j is omitted from now on.