Re-thinking Speed in Financial Markets: Part 2
The IEX decision has re-ignited a polarizing debate about speed being good or bad for markets. What is the alternative to those two camps? If we think of speed as a proxy for price certainty, how would our perception of it’s role in modern market structure change?
If we assume that we are unlikely to make trading via computers illegal any time soon, I have come to the belief that all attempts to regulate and police speed to artificial limits will ultimately prove ineffective.
Faster markets enable narrower spreads and greater price certainty. Slower markets tend to have wider spreads and less price certainty. No market is uniformly slow or fast. Proximity to market is not the only determining factor. Varying traffic patterns on trading networks can cause corresponding variation in delay due to congestion. These delays are sometimes larger and more difficult to manage compared to deterministic transit delays. Unless there is zero traffic, this phenomenon will always be in play. This is why it is possible for someone in a co-location facility to lose a fill opportunity to a participant not in the co-location facility. We have seen this happen and it is straightforward to understand why it happened when you have the right framework and right data to analyze.
Adding speedbumps is not a good general answer but does provide differentiation local to the venue e.g. IEX. Speedbumps introduce speed asymmetry between interacting participants and venues. This adds more complexity to an already complex system because it is harder to figure out the likely execution outcome when there is a mix of fast venues and slow venues. It is speed asymmetry that creates the opportunity for speed arbitrage. Even if you block off a speed play at one venue, you typically end up creating (not intentionally) multiple new speed plays across the population of interacting venues, resulting in proliferation of new order types, rules and infrastructure enhancements that either encourage or discourage exploitation of the effect.
It is impractical to build a marketplace where speed is guaranteed to have no impact. To do this, we need an environment where everything happens so fast relative to the time required to execute a transaction that we can consider all actions to effectively happen simultaneously. The SEC recently declared that events in US Equities Markets within a millisecond to be de minimis i.e. they can be treated as if they happened simultaneously. In today’s markets, high speed algorithms respond to price updates and make trade decisions in less than 10 microseconds. Participants can execute orders within 25 microseconds. Hundreds of market orders can be executed across multiple venues within one millisecond of time. Unfortunately, the SEC declaration is out by about a factor of 1000. It would be more accurate to say events within a microsecond are de minimis. To make this a reality, we would need all venues and all participants to be no more than a microsecond apart in time, i.e. everyone co-located within the same building. Even then, we would have to deal with the random delay variations that occur over short-timescales due to varying traffic patterns and dynamic capacity limits causing random congestion and uncertainty in outcome. In fact, as transaction times get faster we would need to co-locate all participants and venues on the same computer chip. This is theoretically possible in the future but not really a practical solution. Therefore, we have to live with the realities of speed and be able to deal with the impact of co-location, speedbumps, direct data feeds, consolidated data feeds, random delay gateways, random traffic congestion, and venues at distant locations e.g. west coast.
A good market structure design must be able to handle diversity and unforeseen actions. No one rule or one size will work for all people all of the time. Just as democracy allows for people with opposing views, beliefs and actions to coexist within its system, our market structure should allow for any expression of speed (fast or slow) or speed variability that may arise naturally or may be employed purposefully.