The Bank of Canada has released a working paper by George J. Jiang and Ingrid Lo titled Private Information Flow and Price Discovery in the U.S. Treasury Market:
Existing studies show that U.S. Treasury bond price changes are mainly driven by public information shocks, as manifested in macroeconomic news announcements and events. The literature also shows that heterogeneous private information contributes significantly to price discovery for U.S. Treasury securities. In this paper, we use high frequency transaction data for 2-, 5-, and 10-year Treasury notes and employ a Markov switching model to identify intraday private information flow in the U.S. Treasury market. We show that the probability of private information flow (PPIF) identified in our model effectively captures permanent price effects in U.S. Treasury securities. In addition, our results show that public information shocks and heterogeneous private information are the main factors of bond price discovery on announcement days, whereas private information and liquidity shocks play more important roles in bond price variation on non-announcement days. Most interestingly, our results show that the role of heterogeneous private information is more prominent when public information shocks are either high or low. Furthermore, we show that heterogeneous private information flow is followed by low trading volume, low total market depth and hidden depth. The pattern is more pronounced on non-announcement days.
I have often made the point that the reverence shown by the media for traders is misplaced. They don’t make huge profits by keen analysis; they only need to be smart enough to sell at higher prices than they buy. The paper describes how this works:
One challenge of our study is that compared to public information flow in the Treasury market, which generally coincides with news announcements, private information flow is not directly observed. In this paper, we use the impact of order flow on bond prices to infer private information flow. For example, Brandt and Kajeck (2004) argue that order flow impact effectively captures heterogeneous information flow in the U.S. Treasury market. Empirically, Green (2004), Pasquariello and Vega (2007) use order flow impact to proxy for the level of information asymmetry on announcement versus nonannouncement days in the Treasury market. Loke and Onayev (2007) also find state-varying level of order flow impact in the S&P futures market. Using information from order flow impact, we specify a Markov switching model to identify private information flow. Using high frequency transaction data for the 2-, 5-, and 10-year Treasury notes, we obtain 5-minute estimates of the probability of private information flow (PPIF hereafter). In this aspect, our model can be viewed as an extension of the existing PIN model by Easley et al. (2002) and Li et al. (2009).
The data used in our study is obtained from the BrokerTec electronic limit order book platform on which secondary interdealer trading occurs. It contains not only tick-by-tick information on transaction and market quotes but also information of the entire limit order book for the on-the-run 2-year, 5-year, and 10-year notes. This allows us to examine the effect of heterogeneous private information in high frequency. The detailed information on the limit order book also allows us to examine how liquidity dynamics interact with private information. A novel aspect of our paper is that we examine how liquidity reacts to information uncertainty. Given that the timing and the context of information arrival is unknown on non-announcement days, we look at how trading activities and placement of limit orders differs from that on announcement days. Data on announcements comes from Bloomberg and includes date, time and values for expected and actual announcements. Since surveys of market participants provide ex ante expectations of major economic announcements, measures of announcement surprises or unexpected information shocks can be constructed.
Our results show that PPIF is higher for longer maturity bonds, and higher on announcement days than on non-announcement days. The finding is consistent with Brandt and Kavajecz (2004) that price discovery manifests in less liquid markets. In addition, on announcement days, PPIF coincides with public information shocks as measured by announcement times.This is consistent with Green(2004) finding that the role of private information is hihger [sic] at and after announcements.
If you’re a trader and do this, you’re smart. But if you create a computerized expert system to do this, you’re an evil High-Frequency Trader.