Template-Type: ReDIF-Paper 1.0
Series: Tinbergen Institute Discussion Papers
Creation-Date: 2003-12-19
Number: 03-103/1
Author-Name: Cees Diks
Author-Email: C.G.H.Diks@uva.nl
Author-Workplace-Name: Faculty of Economics and Econometrics, Universiteit van Amsterdam
Author-Name: Roy van der Weide
Author-Email: rvanderweide@worldbank.org
Author-Workplace-Name: Faculty of Economics and Econometrics, Universiteit van Amsterdam
Title: Herding, A-synchronous Updating and Heterogeneity in Memory in a CBS
Abstract: This paper considers a simple Continuous Beliefs System (CBS) toinvestigate the effects on price dynamics of several behavioralassumptions: (i) herd behaviour; (ii) a-synchronous updating ofbeliefs; and (iii) heterogeneity in time horizons (memory) amongagents.  The recently introduced concept of a CBS allows one to model the co-evolution of prices and the beliefs distribution explicitly, while keeping track of the unpredictable nature of individual preferences (Diks and Van der Weide, 2003).  As a benchmark model we take a simple CBS, which in a market withmany traders exhibits a random walk driven by news.Using the explicit nature of the dynamics of the CBS we show that the introduction of herding modifies the random walk to an ARIMA($0,1,1$) process, which is observationally equivalent to areduction of the number of market participants. In terms of returns the model predicts MA(1) structure with a negative coeffient.  Asynchronous updating leads to an MA(1) model for returns with GARCH($1,1$) innovations, and predicts a relation between the ARCH and GARCH coefficients. Heterogeneity in memory leads to long-range dependence in returns. In the empirical section we perform a modest `reality check' concerning the predicted sign of the MA coefficient and the relation between the ARCH and GARCH coefficients for exchange rate data.
Classification-JEL: C5; G1; D8; F31
Keywords: Continuous beliefs systems; Random dynamical systems; Heterogeneity; Herding; ARIMA; GARCH; Long-range dependence
File-Url: https://papers.tinbergen.nl/03103.pdf
File-Format: application/pdf
File-Size: 337205 bytes
Handle: RePEc:tin:wpaper:20030103