Estimation of an Adaptive Stock Market Model with Heterogeneous Agents

Abstract

Standard-Nutzungsbedingungen:Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden.Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen.Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. The Working Paper series presents reports on matters in the sphere of activities of the Riksbank that are considered to be of interest to a wider public. The papers are to be regarded as reports on ongoing studies and the authors will be pleased to receive comments. Terms of use: Documents inThe views expressed in Working Papers are solely the responsibility of the authors and should not to be interpreted as reflecting the views of the Executive Board of Sveriges Riksbank. Estimation of an Adaptive Stock Market Model with Heterogeneous AgentsHenrik Amilon * Sveriges Riksbank Working Paper Series No. 177 January 2005Abstract Standard economic models based on rational expectations and homogeneity have problems explaining the complex and volatile nature of Þnancial markets. Recently, boundedly rational and heterogeneous agent models have been developed and simulated returns are found to exhibit various stylized facts, such as volatility clustering and fat tails. Here, we are interested in how well the proposed models can explain all the properties seen in real data, not just one or a few at a time. Hence, we do a proper estimation of some simple versions of such a model by the use of efficient method of moments and maximum likelihood and compare the results to real data and more traditional econometric models. We discover two main Þndings. First, the similarities with observed data found in earlier simulations rely crucially on a somewhat unrealistic modeling of the noise term. Second, when the stochastic is more properly introduced we Þnd that the models are able to generate some stylized facts, but that the Þt generally is quite poor.

Structural stochastic volatility in asset pricing dynamics: Estimation and model contest

“…Defining a (moment-specific) bootstrapped p-value to quantify a model's goodness-of-fit, it turns out that roughly one-third of all simulation runs cannot be directly rejected by the data. On the other hand, the joint MCR amounts to more than 25 per cent, which we think is a fairly respectable order of 3 See Gilli and Winker (2003), Alfarano et al (2005), Manzan and Westerhoff (2007), Winker et al (2007), Boswijk et al (2007), Amilon (2008), Franke (2009), Li et al (2010), Chiarella et al (2011), Franke and Westerhoff (2011). 4 The choice of MSM does not rule out that also other estimation approaches may be tried.…”

Estimation of financial agent-based models with simulated maximum likelihood

“…All these studies are based on daily stock market or foreign exchange data. As a matter of interest, we can compare the performance of models and selected moments based on the S&P500 dataset that is shared by all six stock market models using the method of moments and that covers with a small exception of the older work of Amilon (2008) almost the similar span of data. We can simply follow Franke andWesterho↵ (2012, pg.…”

“…Strategy-specific a h or time-varying 2 h,t are appealing concepts mainly for simulation analyses of the model (see e.g. Gaunersdorfer, 2000;Chiarella and He, 2002;Amilon, 2008). Moreover, we intentionally use relatively small number of possible trading strategies following Kukacka and Barunik (2013), H = 5, for the general model setting or H 2 {2, 3} for so called 2-type and 3-type model, respectively (Chen et al, 2012).…”

“…Next, as discussed by Amilon (2008), the magnitude of noise term has to be considered carefully. Noise is an inevitable part of the model as it represents the market uncertainty and unpredictable market events, but it must not overshadow the e↵ect of variables under scrutiny.…”

Testing of a market fraction model and power-law behaviour in the DAX 30

“…Alfarano et al (2005) estimate a simplified herding model by maximum likelihood method. Amilon (2008) estimates two specifications of the extended Brock and Hommes switching models described in De Grimaldi (2003, 2006) by using the efficient method of moments and maximum likelihood method. He concludes that the simple prototype models he estimated seem to have potential to explain empirical facts although the fit is generally not quite satisfactory.…”