Way back in 2008 in our Toxic Equity Trading on Wall Street , we were one of the first to recommend a minimum order life.  At the time, we recommended a one second minimum on orders.  Since then  many others have called for a minimum order life to try to limit the amount of phantom orders and pinging that goes on in the market.  The SEC has even said they are looking at a minimum order life  but they have been looking at this since the Flash Crash and still have not proposed anything.  The objection that we hear often to a minimum order life requirement is that liquidity would contract and spreads would widen.  We also hear that there is no data that proves orders that are cancelled in less than 50 milliseconds are harmful to the market.  Well, a new academic paper proves that minimum order life requirement would not harm liquidity or price efficiency.

The paper titled, “The Externality of High Frequency Trading” , was written by three researchers from the University of Illinois.  It is a must read for all market structure critics and especially a must read for the SEC.  It is filled with data pulled directly from a NASDAQ Total View-ITCH feed and processed by two supercomputers.  The authors believe that their paper is the largest computing effort ever in academic finance.  The authors propose:

“We argue that minimum quote life would be ideal for this purpose. An increase in trading speeds increase the cancellation ratio but not market quality. Therefore, a decrease in speed has the potential to decrease the cancellation ratio, but has no negative impact on market quality.”

And they have the data to back this statement up.

“Some of the most liquid stocks in our sample have very high cancelation ratios and a high percentage of fleeting orders. For example, 96.09 percent of orders of Apple (AAPL) are canceled, with almost 40 percent canceled within 50 milliseconds. For orders of Google (GOOG), 95.92 percent of orders are canceled, with 30 percent canceled within 50 milliseconds.”

The authors sampled a cross section of 120 NASDAQ stocks representing various capitalizations and found:

“The average quoted spread for the whole book is about 5.971 cents and the median is about 2.805 cents.”

They then extracted the fleeting orders (orders that are cancelled with 50 milliseconds and found that:

“We find that a limit order book without fleeting orders has a quoted spread of 5.996 cents, reflecting a 0.0251 cent increase in quoted spread on average.”

They also found that eliminating fleeting orders would not harm liquidity:

“Fleeting orders contribute 3.96 shares of liquidity in terms of the depth at the best ask, and 3.59 shares in terms of the depth at the best bid.”

Not only does this academic paper prove that orders that are cancelled in less than 50 milliseconds add no value to the market, but this paper also proves that the concept of “quote stuffing” is a real phenomenon that is hurting the market.  Almost daily, our good friends at Nanex point out the many quote stuffing issues that exist in the market.  Many industry professionals often try to discount their findings and say that there is no proof of quote stuffing.  Well, this academic paper proves that there is such a thing as quote stuffing and we are all paying the price for it.

“Our identification strategy for stuffing is based on the channel assignment of NASDAQ stocks. The trading data of NASDAQ stocks are handled by six independent channels based on the alphabetic orders of ticker symbols…we document clear evidence of co-movement of message flow for stocks in the same channel through factor regression. This result is consistent with quote stuffing, because message flow of a stock slows down the trading of the stock in the same channel, but does not have the same effect on stocks in a different channel.”

“Traders who cause stuffing can slow down other traders either because 1) there are more messages in the queue for the exchange to process; 2) the dissemination of trading data from the exchange is delayed so that their competitors cannot react promptly to the market condition; 3) their competitor needs to analyze the data generated by quote stuffing, but the trader who generates the quote stuffing does not. Quote stuffing is certainly an externality-generating activity. It is the noise or pollution in the financial market.”

 “We find that tickers randomly grouped into the same channel have an abnormal correlation in message flow, which is consistent with the quote stuffing hypothesis.”

The authors conclude:

“An increase in speed does not improve liquidity and price efficiency. However, speed may create several externalities. Quote stuffing is certainly one type of externality generating activity. Even without quote stuffing, we argue that investment in speed in sub-millisecond may provide benefit to the investor without consummate social benefit, therefore, there will be an overinvestment in speed.”

 “Finally, the exchange continuously make costly enhancement to its systems to accommodate more message flow, but this enhancement facilitates more cancellation instead of trading. However, the current exchange fee structure only charges trades but not cancellation. Therefore, people who want to trade subsidize people who cancel, reflecting a wealth transfer from long term investor to high frequency traders.”

And to drive home their point:

“An increase in speed increases cancellation, generate more noise to the message flow, and non–high frequency traders need to subsidize the high frequency traders because only trade but not cancellations are charged. Also, the trading games in stressed time would become so complex in stressed times that all the traders stop to play the game, resulting in such event as the flash crash.”

Add another paper to the growing list of academic literature that states high frequency trading is bad for the market.