Three Essays on Applied Time Series Econometrics

from effects of the financial crisis. Precise data sources are given in the appendix. Time series plots for the accession country data series are given in Figures 2.1-2.4. It is obvious from Figures 2.1-2.4 that at least some variables show trending behaviour. A unit root analysis (results not shown) indicates that the unit root hypothesis cannot be rejected for the all the considered time series and that most of them can be characterized as integrated of order 1, I(1). Against the background that the variables may be cointegrated, we follow the standard practice in this line of the literature and specify VAR models for the levels of the variables. This avoids the false cancelation of long-run relationship between the variables. Also note that the beginning of the sample is characterized by more volatile movements e.g. in interest rates, most likely due to effects related to transitional processes. For instance, Czech Republic and Slovakia abandoned the fixed exchange rates to the Deutsche Mark an inflation-targeting framework was adopted by the Czech national bank in December 1997 and in October 1998 a similar policy was adopted by the National Bank of Slovakia. To account for external influences we extract factor time series from (a) a large set of macroeconomic time series from EMU countries and (b) from sets of accession countries. To extract the EMU factors we use data from 11 EMU member states4 on industrial output, CPI and producer price index, shortand long-term interest rates, exchange rates, share prices, unemployment, imports and exports. Again, the data sources are listed in the appendix. The EMU data set is composed of 107 time series.5. Similarly, for a similar set of variables we collect data for the four accession countries Czech Republic, Hungary, Poland and Slovakia. The list of variables coincides with the one given for the EMU countries except that we have no observation on the long-term interest rates. Thus for each country we have only 9 variables available. 2.3.2 EMU and Accession Country Factors Time Series For the factor extraction step, the time series have to be transformed to stationarity. We follow the standard practice in the factor literature (see e.g. Stock and Watson, 2002b) and transform the variables to stationarity by either taking first differences or first differences The countries used are the original 11 EMU members Austria, Belgium, Finland, France, Germany, Italy, Ireland, Luxembourg, Netherlands, Spain and Portugal. There are no series for the short term interest rate and share prices for Luxembourg, and no unemployment series for Austria is available. Consequently, the EMU data consists of 10 time series for each 11 countries minus three (10 × 11 − 3 = 107).