| Efficient market theory posits that market prices reflect at any instant of time the fundamental value of assets, and can only change because of unpredictable news or other information items that affect this fundamental value. If true, well, systematic quantitative strategies should not work. But of course this picture cannot strictly hold – for one thing someone should process information and push the price towards its putative true value. There should be at least some kind of tatonnement and arbitrage opportunities at high frequencies. In order to dissect these possible mispricing mechanisms and to devise profitable high frequency trading and execution models, the detailed study of order flow and order books has become mandatory. Much as in physics, where the detailed understanding of the microscopic world provides invaluable insight on macroscopic phenomena, we believe that a consistent picture of the microstructure mechanisms will help put in perspective some of the traditional questions about markets and prices, such as: “Are prices in equilibrium”, “What is the information content of these prices”, or “Why is the volatility so high”. It will also allow one to optimize execution costs, which for large AUMs is mostly due to impact. Empirical data reveals an unexpectedly subtle price formation mechanism. Order flow turns out to be a highly persistent, long memory process, both in sign and volume. This reflects the fact that even on very liquid markets, the revealed liquidity is in fact extremely small (typically 0.001% of the market cap of a stock). Large orders to buy or sell can only be traded incrementally, over periods of time as long as months. Hence prices cannot be instantaneously in equilibrium, and cannot instantaneously reflect all available information. There is nearly always a substantial offset between latent offer and latent demand that only slowly gets incorporated in prices. Maintaining compatibility with market efficiency has profound consequences on price formation, on the dynamics of liquidity, and on the nature of impact. On anonymous, electronic markets, there cannot be any distinction between “informed” trades and “uninformed” trades. The average impact of all trades must be the same, which means that impact must have a mechanical origin: if everything is otherwise held constant, the appearance of an extra buyer (seller) must on average move the price up (down). A body of theory that makes detailed quantitative predictions about the volume and lag dependence of market impact, the bid-ask spread, order book dynamics, and volatility has been recently put forth [1,2,3,4]. This framework allows one to make quantitative models of execution costs in terms of a time dependent (non-local) friction kernel. It also suggests a novel interpretation of financial information. The theory of rational expectations and efficient markets has increasingly emphasized information and belittled the role of supply and demand, in contradiction with the intuition of traders and of the layman. Our recent work on the role of news on price jumps [5] also shows that information in the traditional sense is not the main driver of market volatility. Rather, we highlight the role of fluctuations in supply and demand, which may or may not be exogenous, and may or may not be informed – it does not really matter. Attempts to estimate ex-post the fraction of truly informed trades actually leads to very small numbers, at least judged on a short time basis, meaning that the concept of informed trades is not very useful to understand what is going on in markets at high frequencies. But still, prices manage to be almost perfectly unpredictable, even on very short time scales. The conclusion is that any useful notion of information must be internal to the market: trades, order flow, cancellations are information, whatever the final cause of these events may be. |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
