Research projects

Project CEMAPRE internal

TitleMTD-Probit Model
ParticipantsJoão Nicolau (Principal Investigator), Flávio Ivo Riedlinger
SummaryOne of the most complex tasks in finance is to forecast stock market. The main challenge is how to
choose a methodology that is theoretically consistent and feasible when applied in the real world
in the presence of market inefficiencies. Usually, the best choice should be suggested by the
underlined theoretical background; however, the model is often selected by its capacity to reproduce
the key data series characteristics. The main goal of this project is to study financial market
behavior using the multivariate Markov chain (MMC, hereafter) modelling framework. To the best of
our knowledge, this is the first time that the MMC methodology is applied in finance to forecast
stock markets. Indeed, the Markov chain research in this area, is based on the standard
discrete-time univariate state-space case, adopting the restrictive hypothesis that a security price
behavior depends only on its own past price information. In this context, we proposed to use a new
MMC forecast methodology: the MTD-Probit model. This model has very attractive features since it is
not only completely free of constraints, but also it is more precise in estimating the transition
probabilities than the standard MMC models. We believe that our project must leverage new lines of
research in a variety of fields.