1 / 26

Investor Contagion: A Network Model for Asset Markets

Investor Contagion: A Network Model for Asset Markets. James Luo ELE 381 Mini-project Presentation. Introduction. Traditional finance models make many assumptions when formulating asset pricing theories In reality, many of these assumptions are unrealistic

elkan
Télécharger la présentation

Investor Contagion: A Network Model for Asset Markets

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Investor Contagion: A Network Model for Asset Markets James Luo ELE 381 Mini-project Presentation

  2. Introduction • Traditional finance models make many assumptions when formulating asset pricing theories • In reality, many of these assumptions are unrealistic • New field of Behavioral Finance attempts to relax or redefine assumptions

  3. Information • Rational investors use some updating rule to weigh information equally • E.g. Bayes’ rule • Traditional finance largely assumes that the path of information transfer does not matter • In reality, the source of information and how that information travels matter because different sources have different influence • Concerned about potential herding and contagion

  4. Relevant Literature • Bikhchandani and Sharia (2000) argue that herding can be caused by peer influence • Kodres and Pritsker (2002) show that cross-market contagion can occur when investors rebalance portfolios

  5. A Novel Application of Networks • Assume that the investors in an asset market can be modeled by a network of N nodes • Three topologies: Watts-Strogatz, Preferential Attachment, and an Artificial Network • Artificial Network has clusters of roughly size qN, and there exists a central cluster such that exactly one link connects the central cluster to each other cluster • We test both un-weighted and weighted networks, where a weighted network implies varying influence

  6. Model Setup • An asset pays a number on [0, 1] at maturity T • At time 0, each investor receives private information about that payoff, with information distributed i.i.d and normally with mean equal to the true payoff and some variance • Market maker first sets a price, equal to the expected payoff of the asset (initially 0.5) [Heuristic 1]

  7. Investor Behavior • First period, investors submit “buy” or “sell” based on their signal, and we assume the market clears due to some large enough mass of “noise” traders • Each period after, the investors can see the beliefs of their neighbors from the period prior • They update their beliefs by taking a weighted average of neighbors’ prior beliefs and their own [Heuristic 2] • Market maker increments/decrements price by 0.01 if there are more buyers/sellers

  8. Assumptions • Risk-neutral investors • No discounting • Long-run equilibrium, if it exists, should be the same • No transaction costs • Should just lower valuations and price • Endowments of the asset do not matter • Can short sell and liquidity can be provided by noise traders

  9. Key Results • Assumed N = 1000, T = 100, and variance = 0.05 • Four sections • Un-weighted networks and random signals • Un-weighted networks and seeding of negative outlook • Weighted networks and random signals • Weighted networks and seeding of negative outlook

  10. Un-weighted Networks and Random Signals

  11. Un-weighted Networks and Random Signals

  12. Un-weighted Networks and Random Signals

  13. Un-weighted Networks and 10% seeding

  14. Un-weighted Networks and 10% seeding

  15. Un-weighted Networks and 10% seeding

  16. Weighted Networks with Random Signals

  17. Weighted Networks with Random Signals

  18. Weighted Networks with Random Signals

  19. Weighted Networks with 10% Seeding

  20. Weighted Networks with 10% Seeding

  21. Weighted Networks with 10% Seeding

  22. Weighted Networks with 1% Seeding

  23. Weighted Networks with 1% Seeding

  24. Weighted Networks with 1% Seeding

  25. Conclusion • Topologies that follow power law are most susceptible to outliers, and those that are small world tend to be the most fragile when exposed • Small group of investors can have disproportionate effects on the price if they are influential • Under incomplete and asymmetric information, price can deviate from fundamental value and large mispricings can occur

  26. Future Work • Alternative model that allows for period-by-period Bayesian updating • Addition of a bid-ask spread and limit orders • Time-dependent preferences (discounting) • Empirical test of some well-defined investor network

More Related