Thanks for the question! We did not consider this tradeoff in our model, as the agents’ trading strategies are relatively simple so the time spent computing what order(s) to submit is negligible. To my knowledge, other related analyses have also not examined this tradeoff, largely because computation time is currently several orders of magnitude faster than the latency in accessing and responding to market information. (For example, current general purpose microprocessors can perform billions of operations per second, while the latencies exploited by HFTs are on the order of milliseconds.)

Thanks for the question. We hope that our work catalyzes a shift from the continuous trading employed by current stock markets to a centralized call market, which eliminates the exploitation of latency advantages by matching orders at fixed, regular intervals. As demonstrated by our results, latency arbitrage is significantly detrimental to overall market efficiency (measured by surplus) and thus harmful to regular investors.

Our two-market model is novel in that it is the first to capture the interplay between market fragmentation, latency arbitrage, and current U.S. securities regulations. In addition, by combining agent-based modeling and discrete-event simulation, our system can be used to answer counterfactual questions about the effects of HFT and fragmentation, unlike prior models which have relied primarily on the analysis of historical order data. As such, our work provides a tool that we hope policymakers and other researchers will use in the future to more rigorously evaluate the various proposed regulatory responses to high-frequency trading.

In addition, penny by penny, share by share, latency arbitrage takes an estimated $3 billion a year from regular investors (Cohan, 2010). Switching to a centralized call market would improve efficiency, meaning that resources would be better distributed among market participants, and therefore regular investors would see improvements (in aggregate) in the gains from their investments—which means more money to go towards financial goals such as buying a house, saving for retirement, or raising capital for a business.

Thanks for the questions. I used Java to develop our simulation system and for the implementation of our two-market model. In our experiments, we simulated two trading venues (i.e., markets) in each run, and we collected data from a total of 200 runs. All simulations were performed on a dual CPU quad core (Intel Xeon E5-1620) machine with 16 GB RAM. As an extension to this work, we are planning on running further simulations on a high-performance computing cluster in order to evaluate the agent trading strategies from a game-theoretic perspective.

Thanks for the question! In order to switch from continuous-time trading to a centralized call market, there are certainly several hurdles that would have to be overcome.

A fully centralized stock market, i.e. one without fragmentation, is most likely infeasible, given current market structure; it is simply not possible to ensure that information is communicated to all participants simultaneously. However, it is more specifically the lack of synchronization between trading venues—a direct result of fragmentation—that permits HFTs to gain an informational advantage via speed.

Therefore, one of the first hurdles would be the development of a robust method for synchronizing multiple trading venues. Such a method would require precise coordination between exchanges and would need to ensure that trade executions and quote updates all occur simultaneously, across all markets, in order to eliminate the potential for latency arbitrage. This is a challenging task, as there are several issues that will necessitate careful consideration, such as how to determine the unique transaction price across all venues for a given matching interval, how to ensure that the global price quote is accurate, and how to handle orders that match across multiple venues (Sparrow, 2012).

Our work is a preliminary exploration of the benefits of synchronization and periodic discrete-time trading via simulations of our two-market model compared to a centralized call market. Further study will be necessary to investigate possible ramifications of any proposed synchronization method. In addition, given the size of the stock market, implementing a new method for synchronizing trading venues will likely be a costly and lengthy process, as it will require significant changes to the technological infrastructure underlying order matching and the communication of public price quotes.

On the regulatory side, there is resistance from HFT firms and exchanges to this and other proposed measures for curbing the exploitation of latency advantages, as both benefit substantially from these practices within the current market structure. High-frequency trading firms make billions of dollars every year in nearly risk-free profit by exploiting latency advantages, and stock exchanges make an estimated $1.8 billion a year from practices such as co-location (Cohan, 2010). Overcoming this and the aforementioned issues will be necessary before latency arbitrage can be fully eliminated from current financial markets.

Thanks, Kathy! Switching from continuous trading to a centralized call market would be a fairly complex process. A fully centralized stock market, i.e. one without fragmentation, is most likely infeasible. However, it is more specifically the lack of synchronization between trading venues—a direct result of fragmentation—that permits HFTs to gain an informational advantage via speed.

Therefore, it will be necessary to develop a robust method for synchronizing multiple trading venues. Such a method would require precise coordination between exchanges and would need to ensure that trade executions and quote updates all occur simultaneously, across all markets, in order to eliminate the potential for latency arbitrage. This is a challenging task, as there are several issues that will necessitate careful consideration, such as how to determine the unique transaction price across all venues for a given matching interval, how to ensure that the global price quote is accurate, and how to handle orders that match across multiple venues (Sparrow, 2012).

Our work is a preliminary exploration of the benefits of synchronization and periodic discrete-time trading via simulations of our two-market model compared to a centralized call market. Further study will definitely be necessary to investigate the possible ramifications of any proposed synchronization method.

As for your second question, I can’t say for certain. Public outcry over HFT activity has been getting stronger in recent years, however, so perhaps we’ll see such a regulatory response not too far in the future.

Thanks, Samson. I hope so too!