Evidence And Effects Of Social Referencing Investor Behavior During Market BubblesVW Staff
Evidence And Effects Of Social Referencing Investor Behavior During Market Bubbles
November 23, 2010
“Evidence of Social Referencing Behaviour during Market Bubbles” (Stephen Chen, Brenda Spotton Visano, Michael Lui, Chaohui Lu) IAENG International Journal of Computer Science, Volume 37 Issue 4, Pages 359-366 London, U.K , 2010
Market bubbles often occur around the same time that new means of investing become available to enable increased market participation. An important aspect of increased market participation is the possible introduction of new investors who behave differently from existing traditional investors. Preliminary evidence from a new data set constructed from publicly available information suggests that these new investors display social referencing behaviour – their investment decisions are based more on social information (e.g., members of their peer group have purchased a stock) and less on typical financial information (e.g., the price of a stock). During the internet bubble of the late 1990s, our collected data show how investors using newly introduced on-line brokerages may have invested differently than investors using traditional and established brokerages. Using this model, we simulate an influx of these new social referencing investor agents in a proportion that is similar to the market weight that new on-line investors had during the internet bubble. The ability of our model to cause a quantitatively accurate, multi-agent simulation of traditional investors to similarly produce a price bubble demonstrates the potential that multi-agent models can have to produce quantitative results for qualitative investor models.
Evidence And Effects Of Social Referencing Investor Behavior During Market Bubbles – Introduction
In conventional models of financial asset prices, the price of equities reflects fully and accurately the existing information on the income earning potential of an asset. This “efficient market” outcome as explored by Fama – suggests that the present discounted value of the expected future income over the life of the asset – its “fundamental value” – will ultimately govern the asset’s market price. In addition to fundamentals-based investors, some models of market behavior also include “noise” traders or “chartists” to help explain the excess volatility observed in stock markets. Chartists attempt to exploit short-term momentum in the movement of stock prices, and their actions (e.g., buying when prices are rising, and selling when prices are falling) can exaggerate any movement in prices.
The presence of noise traders alters, however, neither the ultimate equilibrium market price for stocks (as fixed by the fundamental value of the underlying assets) nor the fact that the market will eventually reach it. In the extant literature, the formal introduction of “noise” traders creates a mean-reverting market dynamic to explain temporary deviations from fundamentals ,. The presence of noise traders can confound market dynamics to such an extent that under some conditions or for some time, it is profitable for the more sophisticated traders to disregard the intrinsic value of the asset, follow the herd, and thus contribute to the resulting asset bubble . Alternatively, studies by Lux and Marchesi – which employ an agent-based model suggest that herding may explain the excess kurtosis observable in high-frequency market data.
Both fundamental and “noise” traders base their decisions solely on objective market information. Traditional financial models commonly exclude by assumption the possibility that investment activity may also be a social activity. In certain situations, and where individuals are motivated to belong to a group, the possibility of fads, fashions, and other forms of collective behavior can exist. Spotton Visano  suggests that investing in equity markets is not immune from social influences, especially when investors face true uncertainty. Consistent with the early views of financial markets as “voting machines” when the future is uncertain ,, Spotton Visano’s result explains the fad and contagion dimensions of investing which relate to Lynch’s  explanations of the recent internet bubble.
Investors in internet stocks in the 1990s faced considerable uncertainty about the future commercial prospects of the internet companies in which they were investing. As such their investment decisions would have been motivated by reasons other than the typical analysis of financial information which was contemporaneously unavailable. As “new” investors, facing an absence of financial information for this revolutionary industry, they would have been motivated by other sources of information. Emulating the behavior of others in their social reference group is one such known motivation.
During the internet bubble, the internet itself enabled the increase in investor participation through on-line trading. Since many of these on-line brokerages were in fact central to the internet bubble, their quarterly reports offer valuable information about the manner in which the internet bubble unfolded. We extracted from these quarterly reports the data required to examine our hypotheses that both social referencing investor behaviour occurred and that this behavior could have contributed to a bubble in the stock market.
Within the limits of such sparse and coarse data, we find preliminary evidence suggestive of two investor classes. There appears to have been some degree of investor differentiation and market segmentation. Inferences drawn from correlations in portfolio returns suggest that traditional investors using traditional brokerages such as Merrill Lynch invested in traditional stocks as represented by the S&P 500. New investors using new internet brokerages such as E*Trade appear to have invested primarily in internet-related stock portfolios as approximated by the NASDAQ. By the peak of the bubble in the spring of 2000, total assets invested through these new internet brokerages were equal to approximately 10 percent of the market capitalization of the NASDAQ .
Previous research describes in a qualitative manner how social referencing investor behavior can affect market dynamics. Yet, quantifying the possible effects of these social influences eludes these models. Fundamentally, it is difficult to acquire quantitative results from a qualitative model. Even multi-agent models of the type used by Lux and Marchesi –, for example, rely on mathematics-based aggregating equations to describe and constrain the overall behavior of a set of agents.
In this paper, we employ a hybrid multi-agent model with both software-based agents and “slave” agents controlled by aggregating mathematical equations. This hybrid model allows the effects of qualitative investor behavior (programmed into the software agents) to be observed within the context of a quantitatively accurate financial model. The overall goal is to explore the conditions necessary to reproduce market dynamics – such as the spike in stock prices observed during the internet bubble – by using a model that simulates the perceived market conditions as accurately as possible.
See full PDF below.