International Statistical Institute, 55th Session 2005

Forecasting Foreign Currency Deposits with Banks in Croatia

Ante Rozga, Ph.D.

Faculty of Economics, Department of Quantitative Methods
Matice Hrvatske 31
21000 Split, Croatia

1.  Introduction

In this paper we have examined seasonal variations using X-12-ARIMA and TRAMO/SEATS with the comparison of results obtained by both methods. Also, we tested if other components were present, such as intervention variables, Easter effect, outliers and trading day effect. Also, forecasts of foreign currency deposits were done.

2.  The Analysis of Historical Data

Figure 1. Foreign currency deposits with trend using TRAMO/SEATS.

Analyzing historical data there were three level shifts detected by X-12-ARIMA and four detected by TRAMO/SEATS. First level shift (downwards) in May 1999 was caused by crisis in banking system in Croatia. Level shift in August 2001 could be the result of very successful tourist season after years of troubles caused by the war and reintegration of Croatian territories. The other level shifts (upwards) were in November 2001 and December 2001 anticipating the Euro as an official currency of EU. This was due to fact that "official" currency in Croatia was DEM as 80% of all savings were in that currency. Method TRAMO/SEATS proved to be more sensitive in detecting level shifts since it detected November 2001, not recognized by X-12-ARIMA.There was one transitory change in March 1995 recognized by both methods, the reason unknown for us. Overall seasonal quality index was also slightly better for TRAMO/SEATS as well as statistics for residuals. Also, TRAMO/SEATS found trading days effects to be statistically significant which is very reasonable looking at the original data.

Generally speaking, it could be said that foreign currency deposits have had strong trend upwards throughout the whole time span used. The main reason for this is establishment of peace process thru the whole region and regaining confidence in Croatian banks. Nowadays 95% of all savings in Croatia is in foreign banks as a result of process of privatization and penetration of foreign banks (all of them from EU) into Croatian financial system. The situation was rather strange because interest rates for deposits in Croatian kuna were always significantly higher that for deposits in foreign currency. It could be explained by history of very high price inflation in former Yugoslavia and uncertainty about the future developments of financial and other politics. If Croatia is to be part of EU, very soon official currency would be Euro.

3.  Forecasting Foreign Currency Deposits

Figure 2. Forecasting foreign currency deposits using both methods.

As we can see from the above forecasts, the rise in foreign currency deposits is slowing the pace, maintaining statistically significant seasonal pattern. This could be explained by the fact that effect of switching from DEM to Euro has been exhausted. On the other side Croatian population is becoming more and more indebted since the purchasing fever is catching Croatian population as well as other East European nations. There should be limit in debt burden which is to be achieved rather soon according to some financial analysts. Also, people trust more and more in Croatian currency (kuna) since it doesn't fluctuate very much from its introduction. Since the implementation of stabilization measures in Croatian financial system in 1993 there were no significant changes in exchange rate to Euro, contrary to US$, for example, who lost more than 30% against Croatian currency.

As far as the forecasting is concerned, method TRAMO/SEATS gives higher forecasts then X-12-ARIMA.

ReferenceS

1. Findley, D.F., Wills, K.C., Aston, J.D.A., Feldpausch, R., and Hood, C.C. (2003), "Diagnostics for ARIMA-Model-Based Seasonal Adjustment", ASA Proceedings , November 2003.

2. Monsell, B.C., Aston, J.A.D., and Koopman S.J. (2003), "Toward X-13?". ASA Proceedings , November 2003.

3. Findlay, D.F.,and Martin, D.E.K. (2003), "Frequency Domain Analyses of SEATS and X-11/12-ARIMA Seasonal Adjustment Filters for Short and Moderate-Length Time Series", SRD Research Report.

4. Hood, C.C. (2002), "Comparison of Time Series Characteristics for Seasonal Adjustments from SEATS and X-12-ARIMA. ASA Proceedings, October 2002.