Abstract. Time series data from the capital market exhibits certain qualities which invalidate the hypotheses necessary for obtaining meaningful results from statistical modeling. This paper presents some of these qualities by looking at the time series for prices and returns on the Romanian Stock Exchange. The examples are based on the price time series and return time series of the Antibiotice securities and the BET-C index. The choice of a certain security and of the stock exchange index has been made with the intention of analyzing, in the future, the correlation between these two variables, and drawing significant conclusions which can be used for forecasts.

Firstly, we will identify the empirical proprieties of the capital market, as they are described in the field research. Secondly, we will investigate the prerequisites for modeling chronological data series; these are stationary mean and variance. In the paper, the three methods are used: graphical representation, autocorrelation and the ADF test (Augmented Dickey-Fuller). For the frequent cases where the mean is not stationary, we will present the time series differentiation method, which can be used to obtain stationary values.

Lastly, we will investigate the normality of the time series through the skewness and kurtosis methods, and through the Jarque-Bera statistic. We find out a characteristic for the capital market, in that the majority of the time series for securities have non-normal distributions.

Keywords: statistical proprieties; stationarity; autocorrelation; ADF test; differentiation; skewness; kurtosis; Jaques-Bera statistic.