This article is devoted to the study of the problem of retrospective analysis of structural changes in SEM (simultaneous equation) models with varying structure. We consider main assumptions about statistical dependence of observations: strong mixing and ψ-weak dependence conditions, as well as the main criteris for effectiveness of a method of retrospective analysis. A new nonparametric method of retrospective detection of structural changes is proposed which does nor require knowledge about distributuinal laws of data and its statistical properties are studied. We formulate theorems about convergence to zero of type 1 and type 3 error probabuilities for the proposed method with an increasing sample size. Results of the simulation sudy of the proposed method are given the second part of the paper. Unlike earlier published papers, this article considers the problem of macroeconometric modeling with account of structural changes in data. Here we consider the method for the retrospective detection of multiple structural changes in data, simulation study of this method, as well as applications to the problems of macroeconometric modeling with account of structural changes in data. In particular, we consider the macromodel of the USA economy proposed by L. Klein (the structural change in the year 1929 is detected) and the disaggregated model of the Russian economy (quarterly data in 1995–2016s). Here we detect two instants of structural changes in 2002 and 2010. Results of the simulation study wittness about the fact that the proposed method can effectively detect structural chsnges in SEM models.