Skúmanie exogenity v nestacionárnych časových radoch: aplikácia na spotrebiteľské ceny a nominálne mzdy na Slovensku v rokoch 1990–1996
INVESTIGATING EXOGENEITY IN NONSTATIONARY TIME SERIES: AN APPLICATION TO NOMINAL WAGES AND CONSUMER PRICE INDEX IN SLOVAKIA DURING...

Summary/Abstract: This article is devoted to problems of estimation and analysis of relationships in the case of nonstationary time series. It can be divided into three parts. The first part dis-cusses the limitation of static regressions using ordinary least squares (OLS), when both regressee and regressor(s) are nonstationary and the author suggested solution of these problems. The second part redefines exogeneity in context of Lucas’ critique and shows how to test it. The third part presents an application of defined approaches to nominal wages and consumer price index in Slovakia in the period 1991–1996. If the analyzed time series have cumulated from relatively independent increases, they are called „integrated“. Integrated series are nonstationary and usually their linear combinations are also nonstacionary. However, there are also stationary linear combina-tions, if a long run equilibrium has existed between the time series. In the first case, we often encounter meaningless regressions, i. e. regressions with a „good” match and t-tests values amongst evidently unrelated variables. Similar results are obtained, if the series are nearly integrated, i. e. highly positively autocorrelated. By differentiating the data we can rid of the problem of incorrect regressions, but in case of data which are only in a status nearly integrated, (which is usually the case) we damage also the long term information. The preferred solution is to use differenced time series and add an agent with shifted left- and right-hand variables in particular levels (an error correcting agent). The error correcting agent is interpreted as a deviation from the equilibrium and the corresponding parameter should be negative, but greater then -1. Three estimation procedures are presented in the article. The parameters of error correcting agent can be estimated in a separate degree, together with other parameters or they can be set up according to a priori information....

• Details
• Contents