Trading Network Traffic and Coronavirus Growth
The turmoil in world financial markets since the outbreak of the Coronavirus has been well documented in the press and online. As of the writing of this blog, the Dow Jones industrial index is down -22% from the beginning of the year. Bad, but way better than the -35% on March 23rd. Similar falls have been recorded on exchanges around the world. Emphasis in the media is on the absolute dollar value of shares or a bucket of shares such as the Dow Jones Industrial.
However, for capacity planners or network operations staff who run trading networks, it's not the value of shares that are traded that drives the amount of traffic on the network, it's the number of trades. For example a market "correction" that occurs over a period of many weeks or months is unlikely to have a major impact on network traffic, but the same "correction" occurring over a single day is likely to be associated with a rapid increase in trades, which then leads to a significant increase in trading network traffic. Keysight visibility equipment (such as optical taps, packet brokers and TradeVision) are at the heart of many trading networks and so are critical to network architects and operations staff responsible for the correct sizing of trading network capacity.
So two questions...
1. What has been the increase in the number of trades?
2. Can network operations staff use data on the number of coronavirus infections announced by public health authorities to help predict the likely increase in traffic?
INCREASE IN THE NUMBER OF EQUITY TRADES
I've put together a graph showing the total trades reported by each of the major equity Exchanges in North America over the last month - this includes NASDAQ, ARCA, BATS Y/X, EDGEA/X, NYSE, IEX, AMEX, BX and PSX venues (though it excludes off Exchange trades reported to FINRA via the TRF) [Source NASDAQ Trader]. Historically the volume of trades has been around 25 million trades per day, but this roughly trebled in the last month. From speaking to customers I have heard of up to four fold increases, so it's likely the analysis is roughly correct. See the graph below with a best fit curve applied to it.
I don't have actual traffic statistics, but I assume that actual traffic on a trading network will approximately follow the same trajectory - [If anyone has real traffic data, please reach out via LinkedIn or email. I'll be interested to know]. So hopefully network planners had planned plenty of network capacity to handle at least a 3:1 increase. If not, it's likely there would have been congestion, queuing delays and even dropped packets - just in an environment where dramatic rises and falls in share prices means that low latency access to venues were critical.
CORONAVIRUS AS A DRIVER OF TRADES IN US EQUITY MARKETS
So we can see from the above that the volume of trades on the US Equity markets increased dramatically in the last month. How correlated was this with the growth in reported infections in the USA? Everyday the Center for Disease Control (CDC) publishes a consolidated view of the number of positive Covid-19 tests in the USA:
significant growth. However, comparing this graph with the graph of the number of shares traded, its obvious that there is no direct correlation between the number of cases and the volume of trades. Maybe the volume is related to the rate of growth, rather than the absolute increase in cases.
In the following graph I have plotted the rate of increase vs the volume of trades on a scatter diagram. Now things look more related.
Clearly when the growth rate was low, the average trade volumes were low. The higher the growth rate the higher the volume of trades, but it's not a very "tight" relationship. I calculated the correlation coefficient (one of the benefits of Excel!) and it's 0.53. A correlation, but a "weak" one.
I also took the same data and looked at the relationship between the absolute rate of change of the growth rate (the flattening of the curve mentioned so much on the news) and could find very little correlation. In fact the correlation function was only 0.3.
So what does this all mean?
I think the overall conclusion is really summarized in the Graph 1. Market volumes, and trading network traffic, tripled or even quadrupled against historical averages. These were driven as much by a realization of the impact on the global economies and the reaction of governments and politicians. US markets are responding somewhat to day by day growth rates in infections in the US, but the linkage is weak. Once the markets got used to the impact, then market volumes and volatility have started to decrease. I suspect that trade volumes will continue to decrease as the medical responses to the virus and therefore its impact on the economy become clearer.
The above analysis and discussion was focused around the impact of the coronavirus on trading networks. The visibility into trading networks is very much a key focus for Keysight and we have special products such as TradeVision to address it. However, we are also heavily involved in the visibility into the performance or standard Enterprise networks where the impact of the virus has also been significant. A key challenge in these environments has been the performance of VPN gateways and the testing of them under high loads - e.g. how many VPN sessions can be supported at once? If I get chance I'll look at a blog on this subject. In the meantime, if you are interested in this area...VPN Testing
As ever, please reach out if you have any thoughts on the above..good or bad!