This thesis addresses the problem of how to identify and model sources of common fluctuations of economic variables. It is an interesting question not only for researchers but also for policy makers and other authorities. The literature presents two approaches. The first one is based on an assumption that the important structural shocks can be captured by a small set of macroeconomic variables. The most popular models used in this context are structural vector autoregression models (SVAR). The second approach follows from a belief that there exists a small number of factors that affect many economic processes. Therefore, it involves analysis of large data sets, with both time and cross- sectional dimensions large enough to describe the factor structure. We dedicate the first part of the thesis to the problem of identification and estimation of structural shocks in small SVAR models. We follow the ideas of Rigobon (2003) and Lanne and Lütkepohl (2008), which show that the statistical property of the data may provide enough information to identify the structure of the model. The papers argue that a shift in the error covariance matrix allows for the estimation of the structural parameters of interest. The literature concentrates on models in which the shift is a result of a structural brake or a mixed distribution of errors.