Analysis of sectorial indexes at Prague stock exchange

Statistical analysis of sectorial and global indexes at Prague Stock Exchange in 1997 was performed. The relative variability of sectorial indexes ranged from 5 % in Mining to 29 % in Agriculture. About one half of index returns can be approximated by normal distribution. The correlation between different sectorial daily and weekly returns was statistically significant particularly between Finance and Banking, Investment funds and heavy industry sectors. As a rule, weekly correlations were roughly twice greater than daily ones. Quasi-periodic behaviour of indexes enables the trend modelling through harmonic components superposition. The first four harmonic terms were shown to be sufficient to explain 80 - 90 % of values observed. Box-Jenkins method was employed for index returns modelling. Except for two cases, simple models up to ARMA(2,2) proved to be sufficient. Further, two thirds of index returns were explained first-order models, particularly MA(1). Prevailing part of sectorial index returns exhibited random behaviour with respect to number of positive and negative events, but the opposite is true from the point of view of runs number expected. Statistically significant difference between Monday/Tuesday and Monday/Wednesday returns was revealed. Monday returns are systematically lower (weekend effect). It was found, that PX GLOB returns variance depends on the PX GLOB values. Thus, we can not exclude, an underlying generating mechanism has different structure.

Keywords: stock analysis, return variability, financial statistics, Box-Jenkins modelling, capital market