Nonparametric Modeling of the Relationship between the Series: Qualitative Co-integration

By Christophe Boya, Jean-Louis Monino


In this paper, we examine the behavior of two time series with an approach based on qualitative data. We put into parentheses the problem of integration or stationarity, sources of growing difficulties, and then focus on the quantitative aspect in the short term of a time series. These values are modulated by the qualitative data, not by the quantitative. In other words, we are interested in two complementary events that are “the rise” and “the decrease.” This binary choice led us to consider the time series as a binary response model. We use Markov chains to get a decisional coefficient. The problem caused by the fact that we do not know the underlying laws of a phenomenon forces us to propose more or less constraining hypotheses to make up statistical tests. Resampling gives us percentiles of a time series and can define an interval confidence of our coefficient. This permits us to make conclusions about the existence of a relation between the two series, as does the “co-integration.”
JEL Codes: C01, C53, C14


  • Bootstrap
  • Markov Chains
  • cointegration
  • qualitative data
  • binary methodology
  • returns series
Go to the article on