Empirical Investor Networks
With the surge algorithmic trading, and the evolving structure of the stock market, it has become nearly impossible to achieve short term gains for individuals. While long term, value based, investment strategies proposed by Benjamin Graham and popularized by Warren Buffet, remain sound sources of gain, it is nearly impossible for small investors to make short term profits. The vast disparity in accessible information and market data between professional traders and retail traders causes the professional trader to nearly always be on the “winning” side of a trade. Any individuals’ possible short term gains are even further mitigated by high frequency and algorithmic trading’s ability to capture profits from just one cent fluctuations in stock prices.
Such a disparity exists not only between professional and retail investors, but among professional traders themselves. Analyzing information networks among professional traders, one will typically find, as consistent with information theory, that central investors have an increased likelihood for higher profits. While there exist many reasons for the disparities between traders, I will focus primarily on such disparities due to differing information.
Network theory allows us to construct a network to help model the underlying behavior, often referred to in the field as an Empirical Investor Network (EIN). The network differs slightly from the one discussed in class as well as the typical information theory network as it allows for an addition, semi-public channel, through which information may affect asset prices. This differs from the traditional approach which applies the constraint that assets are priced with heterogeneous information (assuming completely private signals). Thus, such the EIN model permits us to model slow information diffusion though the network of information that is neither completely public, nor private, such as online investment boards and word-of-mouth.
Constructing an accurate EIN is a significant challenge, but those details shall be omitted from this post. The rest of the post shall focus on the resulting analysis of a constructed EIN. First, the EIN models allows the identification of idiosyncratic information events that are associated with large fluctuations. Research finds that central agents (investors) tend to trade in the correct direction and earlier the peripheral investors. Often large price fluctuations are not accompanied by public information, but are often accompanied by spread of information in a local arena of investors. The result suggests that the constructed EIN model is consistent with a decentralized diffusion mechanism, which is not possible to model using typical information networks which only account for diffusion through public channels.
Furthermore, it is interesting to contrast the EIN with social networks and job finding. We note the EIN have two primary channels incorporated, public news channels and a decentralized diffusion channel similar to social networks. As discussed in class, it was likely that jobs opportunities came over network bridges, rather than strong ties, while investors can more clearly benefit from a large number of strong ties. Two primary reasons exist for this observation. First, the quality of information passed along strong ties is typically larger. It is less likely that a great insight is passed along weak edge, while a job opportunity are typically considered less valuable, as not everyone can benefit from every job opportunity (and is thus more like to be passed along a weak edge than a piece of investment information). Secondly, the model in class does not account for the diffusion time of such information. There is typically a much quicker spread across strong ties, so a piece of information will typically more rapidly diffuse through an arena of strong ties, rather than across weak ties. As the value of investment information declines rapidly with time, receiving the information as early as possible is of upmost importance for maximizing gains from stock price fluctuations.
In conclusion, centralized investors in EINs typically outperform their peripheral neighbors. EIN’s allow us to model to what degree a piece of information is public based on the diffusion time, the point at which the information entered the network, and by the topological properties of the network. Such a network permits asset price movements occurring independently of public information events to be modeled. Yet, there is still a significant amount of research to be done in EINs regarding the exact channels through which information diffuses and what other factors determine the market’s information network.
-esh
Relevant Links:
http://money.cnn.com/2012/10/01/investing/individual-investors-stocks/index.html
http://www.ccfr.org.cn/cicf2012/papers/20120127024534.pdf
http://www.it.uu.se/edu/course/homepage/projektTDB/ht11/projects/TDBproject2.txt
http://faculty.haas.berkeley.edu/walden/HaasWebpage/empiricalnetworks.pdf
http://phys.org/news/2012-01-stock-network-reveals-investor-clustering.html