10.11.2014 Views

CES VŠEM o aktuálních ekonomických problémech

CES VŠEM o aktuálních ekonomických problémech

CES VŠEM o aktuálních ekonomických problémech

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

for panels to eliminate cross-sectional dependence. 46 This test<br />

enables us to test for different long run trajectories and<br />

regression may also include different intercepts and trends<br />

within a set of countries. Other pURTs such as the Harris and<br />

Tzavalis's (1999) test, the Bai and Ng (2004) test or the Moon and<br />

Perron (2004) test, are not covered in this paper as the main<br />

purpose is to apply pURTs to data and not to discuss properties of<br />

individual pURTs. For an overview see e.g. Gegenbach al., 2010.<br />

In addition, if time series in a panel are assumed to have<br />

a common numeraire country’s currency, 47 cross-serial<br />

correlation is to be expected. Therefore, this setting requires an<br />

application of pURTs that allows for serial correlation such as the<br />

MADF test, see Harwey and Bates, 2003 or the panel corrected<br />

standard error test (PCSE) see Jonsson, 2005. 48 But this does not<br />

seem to be true in this case given the fact that our time series<br />

are calculated against the euro. 49 As time series of exchange<br />

rates in our sample cover a set of different countries, 50 it is not<br />

certain whether all will follow a common unit root process or<br />

not. Moreover, it seems to be reasonable to assume crosssectional<br />

correlation between them. 51 Since our panel is large<br />

with a medium dimension of cross sections, we will apply and<br />

46 If is a demeaned time series (an exchange rate in our case), it does hold:<br />

, where , i.e. is the exchange rate in relation<br />

to the cross-section average.<br />

47 This problem is described e.g. in [Banerjee al. (2005)].<br />

48 While the LLC test may reject the null hypothesis even when one of the time<br />

series is stationary, the MADF test is robust against that and the null is rejected<br />

only if all time series are I(0). This test can also be applied in the case of unbalanced<br />

panels.<br />

49 Since the euro did not exist before January 1999 and they are based on artificial<br />

calculations using the conversion rates for invidual countries and the euro<br />

for the period (1995–1998) before euro adoption.<br />

50 One may think of possible structural differences among them or the necessity<br />

to overcome the ‘burden’ of past decades of communistic regimes.<br />

51 Since correlation coefficients show relatively high correlation between time<br />

series in our panel.<br />

85

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!