Multi-Market Direction-of-Change Modeling Using Dependence Ratios

Abstract

We consider a multivariate dynamic model for the joint distribution of binary outcomes associated with directions-of-change for several markets or assets. The marginal distribution of each binary outcome follows a dynamic binary choice model, while the association structure is parameterized via possibly time varying dependence ratios. We illustrate the technique using daily stock index returns from three European markets, from three Baltic markets, and from two Chinese exchanges.

International sign predictability of stock returns: The role of the United States

“…Typically very little out-of-sample predictive power is found, if any (see Goyal and Welch (2008) and Campbell and Thompson (2008)). In contrast to the usual predictive models, the previous research on (out-of-sample) sign predictability is rather scant and, to the best our knowledge, so far only Leung et al (2000) and Anatolyev (2009) have examined international datasets (containing only a few countries).…”

“…Leung et al (2000) consider the U.S., U.K. and Japanese markets, but unlike us, they do not explore international linkages between the markets but concentrate purely on country-specific models. Furthermore, Anatolyev (2009) considers directional cross-predictability of daily returns from three European markets, three Baltic markets, and from two Chinese exchanges in a different multivariate model compared to ours.…”

A Bivariate Autoregressive Probit Model: Business Cycle Linkages and Transmission of Recession Probabilities

“…The proposed bivariate autoregressive model is related to the multivariate dynamic binary model of Anatolyev (2009). The main difference is that Anatolyev (2009) suggests using the so-called "dependence ratios" [see also Ekholm et al (1995)] between the dependent variables to construct the conditional joint probabilities of the different outcomes of (y 1t , y 2t ). In our model, the dependence structure between y 1t and y 2t is modeled using the bivariate cumulative normal distribution function (10) with the correlation coefficient ρ instead of the dependence ratios.…”

“…This is not typically the case in dynamic models based on latent variables, such as the dynamic univariate model of Chauvet and Potter (2005) and the bivariate model of Mosconi and Seri (2006) [see also Dueker (2005)]. Instead of the latent variable approach, the bivariate autoregressive model has some similarities to the model proposed by Anatolyev (2009), even though the dependence between the two binary time series is modeled in a different way.…”

A non-linear forecast combination procedure for binary outcomes

“…The stream of literature on a general introduction to the modeling strategies based on copulas includes Trivedi and Zimmer (2005), Nelsen (2006), andPatton (2012). Anatolyev(2009), Patton(2006 and Scotti(2011) applied this methodology to predict multiple economic events. Patton (2013) provided a recent survey on copula methods to forecasting multivariate time series.…”