Financial social trading networks: The case of copy trading platforms

Jose Apesteguia, Joerg Oechssler, Simon Weidenholzer 29 September 2018

a

A

The rise of network platforms such as Uber, Twitter, and TripAdvisor has profoundly shaped social interactions and fundamentally changed entire industries (transport, news media, and tourism, respectively, in the cases above). Integrating the features of these online social networks into financial brokerage platforms has given rise to social trading.  While still in its infancy, social trading might have transformative impact on the finance industry.

Many social trading platforms not only facilitate the exchange of information among traders, but also allow investors to automatically copy investment strategies of others.  That is, copy trading platforms offer the possibility of allocating a monetary endowment to reproduce the financial strategies of the user one wants to copy. At the time of writing there are at least a dozen active copy trading platforms, with millions of users spread all over the world. One of the larger of these, eToro, currently has 9 million subscribers and according to its CEO has had an annual trading volume in excess of $300 billion in 2016.1

While copy trading in principle offers copiers the opportunity to benefit from superior information of other traders as well as providing incentives for those copied through discounts and commissions, there are also potentially negative implications for the welfare of traders. These are conveyed through the adverse role imitation may play in stochastic environments (for an earlier exploration in a different context, see Offerman and Schotter 2009).  To see this, note that previously successful traders may have just been lucky.  Under copy trading other traders are inclined to imitate those lucky traders. To make matters worse, high returns might be associated with high risk taking by the copied agents, i.e. successful traders might not only have been lucky, but may have also taken more risk. Copiers may therefore be more likely to adopt risky investment strategies. Thus, copy trading as a business model may well result in excessive risk taking and individually and socially suboptimal outcomes.

In a recent paper, we examine copy trading and its implications for risk taking in a series of laboratory experiments (Apesteguia et al. 2018). The experimental laboratory allows us to control for a number of key variables that would be very difficult, if not impossible, to do control for in the field. For example, in our studies we measure risk preferences outside the financial market.  This allows us to determine optimal behaviour in the choice of assets at the individual level. Moreover, the experimental approach enables us to directly test the influence of the main characteristics of copy trading platforms, namely, the provision of information on the financial decision and success of others, and the possibility of copying others. That is, it enables us to compare outcomes under copy trading to the counterfactual of not being able to copy trade, and to test whether this induces more risk taking behaviour.

Our experiment

Our experiment consisted of three parts. In the first part we elicited the subjects’ risk preferences. The second part was composed of two blocks of investment decisions, whereby subjects had to choose one of four assets whose prices evolved according to a Brownian motion (approximated by a Binomial tree model in discrete time, see Cox et al. 1979). The assets were characterised by different state-dependent rates of return.  Further, some assets featured tail risk, which we modelled as the probability of a crash to a relatively low price. The subjects were made aware of all attributes of the available assets. After choosing their assets, the subjects had to decide for a number of periods whether to sell the asset at the current price or keep it. In the second block the subjects were confronted with the same investment problem. 

Depending on the treatment, there were additional components. In the ‘Baseline’ treatment, the second block consisted of the exact repetition of the investment situation the subjects confronted in the first block. In our main treatment, ‘Copy’, agents received a list containing the decisions and realised profits in the first block of all the agents in ‘Baseline’, ordered from highest to lowest realised payoffs. Then, the subjects could either make their own investment choice, or could choose to copy the unknown investment decisions in the second block of a subject of their choice from the list. In the latter case, copiers then simply received the (as yet) unknown payoffs the copied subjects had earned in the second block.  In addition, we ran an ‘Info’ treatment in which the subjects received the same information on the first block decisions and outcomes of the ‘Baseline’ participants than in ‘Copy’, but in this case did not have the option of copying any investment strategy.

Before any investment decisions were made, the subjects were provided with a tool that allowed them to simulate price path realisations for each of the assets. The purpose of this simulator was to familiarise subjects in a user-friendly way with the possible outcomes of the various assets and to mitigate the role of the additional information subjects received from peers in the ‘Copy’ and ‘Info’ treatments.  Analogous tools are being offered in practice by financial institutions to private investors at the time of buying financial products such as mortgages or pension plans.

Our findings

The main results are as follows. Traders take significantly more risks when they see the ranking and the results of other traders. Figure 1 shows a substantial shift of traders towards riskier asset choices in the INFO treatment in Block 2 as compared to the choices of traders in the ‘Baseline’ treatment.  A similar shift towards more risk taking occurs in the ‘Copy’ treatment among those traders who chose not to copy another trader. 

Moreover, we observe that when participants are given the possibility of copying others, a sizable fraction does so, and that the distribution of asset choices shifts markedly towards riskier ones. Concretely, 35% of participants in ‘Copy’ chose to copy someone in the list and, of these, the vast majority copied somebody who had chosen the riskiest possible asset in Block 1. Figure 2 plots the distribution of asset choices in this treatment alongside the implied choices of those who decided to copy somebody. The opportunity of copying others thus leads to an additional increase in risk taking in comparison to the pure effect of providing information on other traders.

Figure 1 Distribution of asset choices across treatments (excluding copiers in ‘Copy’)

Note: Asset A is the least risky asset, and D the most risky asset.

Figure 2 Distribution of asset choices and copiers in the ‘Copy’ treatment

 

We therefore observe that the type of information provided and the possibility of copying others present in copy trading platforms is ex ante welfare reducing, as investors choose suboptimal assets when compared to either i) the level of risk aversion revealed in the asset choices of Block 1, ii) the risk aversion elicited in the lottery choices in Part 1, or iii) the counterfactual scenario when agents neither can copy trade nor receive information on others.

We further address the question of who decides to become a copier. Here we find that risk aversion plays a determinant role. The more risk averse subjects are, the more likely they are to copy others. Ironically, it is thus those with a revealed low tolerance for risk taking who are enticed through copy trading to take on more risk.

Policy implications

We believe that the implications of copy trading on risk taking may be even stronger on real world copy trading platforms. While we recruited our participants from a student subject pool, investors are likely to join copy trading platforms with the explicit intent to engage in copy trading.  Further, the design of our experiment made the role of luck very salient.  In the real world, however, investors’ beliefs on the skills and information of leaders might be more optimistic.  In addition, whereas our experimental setup allowed subjects to easily asses how risky previous investments of other investors were, such an assessment is much more difficult in the real world.  From a social perspective, imitation encourages traders to follow similar investment strategies, and thus could lead to financial risk through herding and contribute to the formation of financial bubbles. We believe our results are a serious call for attention to copy trading platforms that are currently proliferating, and hope it will trigger more research in the near future.

References

Apesteguia, J, J Oechssler and S Weidenholzer (2018), “Copy trading,” mimeo.

Cox, J C, S A Ross and M Rubinstein (1979), “Option pricing:  A simplified approach,” Journal of Financial Economics 7(3): 229–263.

Offerman, T  and A Schotter (2009), “Imitation and Luck:  An Experimental Study on Social Sampling", Games and Economic Behavior 65(2): 461–502.

Endnotes

[1] See https://uk.reuters.com/article/us-tech-etoro-fundraising/israeli-social-trading-firm-etoro-raises-100-million-in-private-funding-idUKKBN1GZ15S and 09:16 in an interview with  eToro CEO Yoni Assia at https://www.youtube.com/watch?v=P2yRjHAAPeU&vl=en

a

A

Topics:  Financial markets Frontiers of economic research

Tags:  of copy trading platforms, investors, risk taking, social networks

Professor of Economics, Universitat Pompeu Fabra

Professor of Economics, University of Heidelberg

Professor of Economics, University of Essex

Events

CEPR Policy Research