Digital Finance
Over the last ten years, we’ve witnessed a large technology breach, which has led to several impacts and potentially disruptive effects on many economic sector, uber and transportation per example. Today, Finance seems to be one ot the next target that would be hit by this digital revolution. This series of article aims to present some of the last evolutions and their effects on the financial industry.
- Big Data algorithms
Every day we create 2.5 quintillion bytes of data — so much that 90% of the data in the world today has been created in the last two years alone[1]. Most of those data are issued from our online activity either it’s from our computer, tablet or phone. Those data represent a part of what we are, our hobbies, our interests etc. By analyzing this data it’s possible to identify patterns that we unconsciously follow. Tech. companies have been exploiting those patterns, to sell ad spot such as Google or product such Amazon. To identify your patterns and the products you like they use algorithm that acknowledge correlation within your data. In the end, the machine knows you better than you do.
Those technics are used by the financial industry for three main purposes:
· Trying to determine the future, where and when to invest? By using perception technics
· Credit scoring (also address in another article).
· Offer a better service to their customers by knowing them better (we will address this topic in another article).
These evolutions can bring major turbulences in the financial environment and we will try to highlight some of those.
The progress of IT has led to new sets of data and its exploitation are used to help the decision making process in terms of investments. One illustration of an application was provided with the Brexit referendum. Before the brexit elections macro hedge funds used big data algos to try to predict the outcome of the vote. Actors such as, Breven Howard, Discovery and Odey Absolute Management gained significantly from the result they have been able to predict.
Big data algos uses statistical formula to determine a trend. The main issue raised here, which data feed your algo? For some it’s “classic” market data such as gold price, pound vs dollar etc. Others such has BH have chosen to include social media data to take the pulse of the electorate.
Those funds are not the only one to invest in prediction; the historical expert is Renaissance Technologies with a tremendous track record but other major HF such as Brigewater Associates or Highbride Capital Management are also investing in the domain to gain a competitive advantage over the market[2].
This trend is totally logic, prediction has always been a big concern in science and big data has foster the interest. In Finance, hedge funds are known to always be looking at new ways to enhance their performance.
In the near future it will be interesting to see where this goes. In term of infrastructure, big data is an area where a lot of investment is needed to build and maintain your algos. In term of staff, an economical war between HF, banks, tech firms and governments is occurring to attract the best talents and obtain the best technology.
If staff and infrastructure are important factor of success, in big data, as mention earlier, the main topic is data. In this very competitive world where states and several industries are involved, we can imagine that the price for data will increase following the demand (even though the offer is also growing cf introduction).
Financial actors also use big data algo for the fast decision-making and objectivity for long-term investment decision. For instance in 2014 a Hong Kong Venture capitalist promote to its board a program. The software votes on investment decision with the same vote power than other member of the board. He makes his choice by pulling large number of data on the sector of the startup and analyzing it[3].
Questions will remain on data:
Will a company be able to obtain the exclusivity of data from certain provider or does the provider will sale to all of the actors? In the first case scenario the actor will have a significant competition advantage, which could disrupt the market. Do state actors need to secure access to data?
It’s interesting to note that market regulators are also asking for more and more data from financial actors. The MIFID series, SFTR etc. are regulations which are asking financial actors to public a lot of data from there trades, these demands are understandable, regulators want more transparency for several reasons, we see two main one:
· Market understanding to avoid systemic crisis such as the subprime crisis
· Economical information to have a better understanding of the actual economy and drive more efficiently the monetary policy
We think the authorities are fully in their mandate to secure markets and the economy but we think that if this data is accessible to everyone pure speculative actors such as the HF listed before would be able to gain significant advantageous from it.
Possible limits:
Moreover, regarding market impact if all actors have the same data it’s possible — even if each actor build prediction algos individually — that the same trends will appear. The HFs will be betting on the same trend they will probably have a significant market impact. How long only actors such as Pension funds, Asset Managers will react to this new paradigm in financial markets? Their performance will decrease, AM will lose customer but how to explain a worker that his pension will reduce?
We have notice the importance taken by big data algos in the investment process of financial actors. If what is done already and the future is full of promises some limits have come to our knowledge. Hacking is probably the first limit that we think of, hacking an algo to alter its findings could largely benefit the hacker.
There also other limits to big data algos. For instance, financial actors could be tempted to rely a lot on algos for their investments, they do procure a fast and objective insight but they are “only” correlation creators they don’t “think” properly speaking. Algos are not going to ask themselves on the quality of the data they have, they will just use it. Therefore we can imagine this alteration:
An investment manager knows that some of his competitors are using social media to have sentiment of a stock or a market segment. To alter his sentiment this investment manager could create bots on Tweeter[4] that will systematically retweet wrong information (the same stratagem could be used on other social media). This would create a large amount of data that other investment managers will wrongly use in their investment process. This illegal stratagem is not new; we often see an actor trying to alter the market via disinformation but it highlights the importance of “critical thinking” that we only see in human’s capability today[5].
Big data algos are a very promising tool for financial actors to have an insight of the future. It create new indicators to better manage their investment but the competition is hard and a disruption is possible we think specially in asset management where Hedge funds seems to have taken a big advantage on long only actors.
We think regulators should have a good look of what’s happening and especially on two aspects:
· Market impact: how can the regulator mitigate risk of a market crash if most of the actors using big data algos trade in the same direction?
· Transparency and availability of data:
o Regulators must be aware that big old-fashioned actors might wait for regulators to understand what they can do or not, whereas quickly adaptable actors such as Hedge funds might take the risk to be very disruptive as it has been seen in the transport industry where Uber forced regulators to react.
o Access to data, do we need to make data available free to everyone to avoid that some actors capture some large amount of data feed and gain large competitive advantage that can be harmful? Or is there enough sources of data to avoid this risk?
[1] IBM : https://www-01.ibm.com/software/data/bigdata/what-is-big-data.html
[2] Banks must also have skin in the game but it’s more an in house business. At this time developments by these actors haven’t been advertised.
[3] Source : http://www.businessinsider.com/vital-named-to-board-2014-5?IR=T
[4] This method is used for political reason and has been highlighted through the « Maidan revolution » source : Hegelich and Janetzko, 2016, Are Social Bots on Twitter Political Actors ? Emperical evidence from a Ukrainian Social Botnet. AAAI Conference on Web and Social Media.
[5] Research are ongoing to allowed program to have this capability Google for instance : http://research.google.com/pubs/MachineIntelligence.html