The Information Environment, Informed Trading, and Volatility

ABSTRACT

The relation between informed trading and volatility is analyzed using the change in the proportion of informed transactions calculated through the probability of informed trading variable. The analysis relates to the Spanish market during 1997–2010, given that the Spanish market covers a very diverse range of listed companies. Some companies are comparable to companies listed on U.S. markets while others are smaller in size and have a lower trading volume and inferior quality of information. The methodology is based on a modification of the model proposed by Avramov, Chordia, and Goyal [2006]. The authors’ proposal incorporates the change in the proportion of informed transactions, calculated with intraday data, into the volatility model. The results are also presented using a conditional volatility model in which the change in the proportion of informed transactions is incorporated. These results attest to the influence of informed trading as a price-stabilizing factor in heavily traded and highly capitalized stocks (familiar stocks). Informed trading leads to a marked decrease in volatility for these particular stocks both in periods of calm and crisis.

KEYWORDS:

• Behavioral finance
• Time series
• Microstructure
• Informed trading
• Volatility

Acknowledgments

The authors would like sincerely to thank the reviewer for the helpful comments and suggestions made with respect to the first version of the article.

Funding

This article has received financial support from the Spanish Ministry of Economy and Competitiveness (ECO2012-35946-C02-01, ECO2016-77631-R AEI/FEDER, UE, and ECO2013-45568-R), and from the Government of Aragón/European Social Fund (S14/2).

Notes

1. The reason for excluding trades outside normal hours is that these operate under a different trading mechanism than that used during the rest of the day.

2. There are alternative ways of classifying a transaction as being initiated by the buyer or by the seller. Finucane [2000] showed that the tick-test method produces similar results to those of other classification methods. Given this finding and the fact that there is no database available, which includes the bid-ask differential, we have decided to use the tick-test to classify operations. Specifically, a transaction is classified as being initiated by the buyer if the price of the transaction is higher than that of the previous transaction (up-tick), and as initiated by the seller if the transaction price is lower than that of the previous transaction (down-tick). If there is no price difference between a transaction and the previous transaction, it is classified as zero tick.

3. The estimation of the PIN variable with intraday data involves a very high number of iterations and a high computational cost. It was therefore decided to take a representative majority of stocks traded on the Spanish stock exchange. This takes into account criteria including the whole range of traded stocks in terms of size, liquidity, and volatility.

4. The figure of the market maker does not specifically exist in the Spanish market, but the trades placed through the order book enable solicited transactions to be observed. The experimental study of Bloomfield et al. [2005] suggests that a market-making role arises endogenously in the electronic markets.

5. The Newton-Raphson method has been used for maximizing the likelihood function in Equation 3(3) . This method was used by authors such as Brockman and Chung [2003], Brown and Cliff [2004], Pang, Hou, Troutt, Yu, and Li [2007], or H. W. Lin and Ke [2011], among others.

6. We have also used the H. W. Lin and Ke [2011) factorization. Nevertheless, we do not find higher estimates than the estimates based on the Easley et al. [2010] factorization, and boundary solutions appear with a very high frequency.

7. This number is not the same for all stocks and depends on the autocorrelation detected. The range varies from 1 to 5 lags.

8. This variable is included because there are numerous papers in the empirical financial literature that show a positive and significant relation between volume and volatility (Chan and Fong [2000, 2006], Epps and Epps [1997], Gallant, Rossi, and Tauchen [1992], Jones et al. [1994], Karpoff [1987], and O'Hara [1995], among others). The 2 paradigms that attempt to explain this relationship are the mixture of distributions (Epps and Epps, [1997]) and the microstructure paradigm (O'Hara, [1995]). From a number of empirical studies that use different measures of volume to test these paradigms, we have taken Jones et al. [1994], Chan and Fong [2000, 2006], and ACG06. Following these papers, we use 2 different measures of volume: the total number of transactions and the total number of stocks traded.

9. The number of lags included varies for stocks depending on the significance of the correlation.

10. For reasons of clarity only the results when the volume is approximated by the number of transactions are shown. The conclusions obtained from the results when the volume is approximated by the number of traded stocks coincide with those shown for the number of transactions.