eThesis - School of Business electronic theses

» hse_ethesis_13344.pdf   size:693 KB (709325)

In this paper I use an inter-temporal model to analyze the variability of the current accounts (CA) of the 12 countries which have formed the initial European Monetary Union and test whether this approach is successful in accounting for the evolution of the CA imbalances over the last decades. The use of current account balances is assumed to be comparable to the use of savings, so that the main hypothesis is that current account could be used to smooth (national) consumption and its variation would be driven by expectations about future income (net output) and relative prices. The model that I choose to adopt encompasses sources of external shocks for small economies such as changes of the real interest rate and of the real exchange rate. This paper finds that, on the one hand, the country specific VAR model passes informal tests such as graphical representations and standard deviation comparisons. On the other hand, although the main formal statistical test (k-test) fails for all the 12 countries, the secondary formal test R-test can be considered successful in the case of Austria, France, Germany, Ireland and Portugal while the Granger-causality is passed by Austria, Belgium, Netherlands and Spain. This paper also warns on the validity of the Vector AutoRegressive (VAR) model results for those cases which have proven statistically significant. In fact, analyzing the same set of equations through a Seemingly Unrelated Regression Equations (SURE) system shows important differences in the magnitude, standard deviations and t-statistics of the estimated coefficients. The paper concludes that an intertemporal approach to current account, executed through VAR modelling, cannot statistically assess the reasonability of recent CA imbalances in the studied countries for the sampled period but informally suggests that the current account of the studied countries behaved as the theoretical model would predict. In order to improve the performance of the model, this study suggests taking into consideration the correlations found among the error terms of the equations estimated from the country specific VAR models.