Impact of the Main Currencies Exchange Rates on the Romanian Economic Policy Transformation

Received March 24, 2018 Revised from March 29, 2018 Accepted May 23, 2018 Available online June 15, 2018 Evolutions of the exchange rates in a contemporary competitive economy represent an influential marker in evaluating the effects of the governmental policies in field of monetary policy and marks future directions in developing the specific policies in the field. This study investigates how the exchange rates of EUR, USD, GBP and CHF were influenced by disturbing factors and the existence of potential quantitative correlations and dependencies among the four exchange rates in terms of uncertainty. In this context the starting premise of the research took into account the exchange rate developments for these four currencies for the period January 2008 – October 2015, period marked by two major disruptive events, which highly influenced the exchange rates of the main currencies on the Romanian market, such as the financial crisis started in late 2008 and CHF exchange rate’s significant variation at the beginning of 2015 that triggered a deep debt crisis in the case of mortgage loans in Romania. The research results reveal particular evolution patterns in the case of the exchange rates of the four currencies taken into analysis, confirming the existence of disturbing factors and the correlations and quantitative dependencies between them. JEL classification: C54, E52, G18. DOI: 10.14254/1800-5845/2018.14-2.1


INTRODUCTION
The analysis of the relationship between exchange rate volatility of various currencies and the national one constitutes a defining element in understanding the behavior in terms of monetary policy and to outline intervention measures and instruments in order to counteract its effects' propagation within the national economy.
The available literature points out various opinions regarding the characteristics of exchange rates and the determinants of their evolution. Such studies as Chinn and Meese (1995) and Mark (1995) argue that the evolution of the exchange rate can be predicted and somehow managed for long periods of time, while other views (Della Corte and Tsiakas, 2012;Verdelhan, 2013) highlight the randomness of the exchange rate evolution and the impossibility of predicting it. Ahmed, Liu and Valente (2016) analyzing the predictability of the exchange rates using linear factor models in the case of three currency-based risk factors reach the same conclusions expressed in previous studies (Meese and Rogoff, 1983), according to largely fail to outperform the benchmark random walk. On the other hand, Ahmed and Straetmans (2015) try to predict the cyclical behavior of exchange rates by using five risk factors, namely violations of uncovered interest rate parity (UIP), relative purchasing power parity (RPPP), pseudo-parity for equity returns, relative (cross-country) TED spreads and relative term spreads. Their conclusion highlights the usefulness of using financial variables and exchange rates as warning financial indicators for investors and especially for policy makers.
On one hand, the exchange rate evolution reflects to a high extent both the effects of monetary policy promoted by the National Bank and the result of government measures taken in order to strengthen fiscal strategy applied in the case of excised products, which depend on the exchange rate, particularly against the EUR, and on the other hand, major implications are also received from main retailers which are forced to make transnational payments in order to pay off capitalized goods within the retail chain.
The existence of substantial economic imbalances in the national economy correlated with enhanced effects of the CHF crisis on an important part of bank loans has led to persistent disparities regarding the evolution of the main currencies in the forex market. This situation derives from a multitude of factors that influence aggregate currency demand and supply tending to propagate negative effects on the entire economy. Between January 2008 and October 2015, the exchange rates of EUR, USD, GBP and CHF against RON pointed out developments with alternating increases and decreases spread around upward trends which characterized the period under review taken as a whole.
Most empirical studies conducted in this field reveal the existence of diffusion (at least indirect and incomplete) of exchange rate fluctuations within the price mechanism in the national economy. In this context, understanding the trends of the exchange rates of the four currencies taken into analysis -EUR, USD, GBP and CHF -and the factors that influenced their evolution, such as identifying and highlighting possible correlations and quantitative dependencies between the four exchange rates constitute the central objective of this article.
The models developed in this paper complement previous research conducted in this field (Berg and Miao, 2010;Ghiba, 2010;Fidrmuc and Horváth, 2008;Nuti 1996or Babetskii et al., 2004, thus contributing to outspreading this knowledge area by highlighting both the characteristics of the evolution of the four main currencies considered for this analysis and the implication of any factors and quantitative dependencies in determining exchange rates. We contribute to the empirical literature by identifying, building and testing various models type AR (p), MA (q) and ARI-MA (p, i, q).
We have organized the paper in four sections. The following section summarizes the methodology employed in the study. Third section called Results and discussions contains the analysis of the characteristics of the trend of exchange rates and the correlations between exchange rate developments as well as a time series analysis. The empirical results are also reported in this section.
The last section of the paper is represented by the conclusions.

METHODOLOGY
In order to determine the factors that influenced the evolution of exchange rates for the four currencies under analysis and to build econometric models the ANOVA methodology was used (Montgomery, 2001, Moore andMcCabe 2003). We have considered it more suitable for this case, although specialized studies on models for determining exchange rates tend to use more frequently the so-called Taylor rule models deriving from West (2005 and2006) previous works, which are very popular in this field.
The period January 2008 -October 2015 was marked by at least two events that influenced the exchange rates of the main currencies on the Romanian market: the financial and economic crisis that started in late 2008 and CHF exchange rate significant alteration in early 2015. Based on these findings, the empirical model employed in this study aimed at proposing and testing three hypotheses, as follows:  The first hypothesis is based on the argument that rates have been significantly influenced by disturbing factors;  The second hypothesis envisages a development that is not significantly different in the case of the four currencies EUR, USD, GBP and CHF;  The third and final tested hypothesis assumes the existence of significant correlations between EUR, USD, GBP and CHF respectively.
In order to determine and analyze the statistical significance of the trend for the case of all 4 currencies (ER) addressed in this paper, several models were developed under the (1): (1) while for emphasizing the quantitative relation between EUR, USD, GBP and CHF, a model has been tested using (2): Determining the parameters of models (1) and (2) has been conducted using the least squares method, ANOVA methodology (Montgomery 2001, Moore andMcCabe 2003) was used for testing the validity of the acquired models, while the validation of their parameters was achieved using t-Test (2-tailed). For primary regression testing, we employed the Durbin-Watson test (Durbin and Watson, 1951), using the following equation: and to eliminate autocorrelation of order p, values p  were identified by applying the least squares method on the following model: where t e are not auto correlated (Andrei et al. 2008, Săvoiu andNecşulescu 2009).
The series EUR, USD, GBP and CHF were analyzed as time series. To test their stationarity Augmented Dickey-Fuller was used with Null Hypothesis: The series has a unit root. For obtaining stationary series the first difference was employed (5): In (5), L is backward shift operator in order to achieve For all 4 time series we have tested AR(p), MA(q) and ARIMA(p,i,q). A stationary series   Z t t y  follows a process ) ( p AR if the following condition is met (Zaharia and Gogonea, 2009): Using lag operator L and denoting by , whose characteristic polynomial is: The stationary series condition is In the case of moving average processes MA(q), the stationary series   Z t t y  can be represented by (8): In order to process the evolution of the 4 analyzed currencies, after developing stationary data series for EUR, USD, GBP and CHF, ARMA (p, q) models were tested whose general form is (Zaharia and Gogonea, 2009): In carrying out the analysis contained herein, we have used initial EUR, USD, GBP and CHF data series taken from the database of The National Bank of Romania (The National Bank of Romania, 2015).

Characteristics of the exchange rates' trend
Within the analyzed timeframe the largest fluctuations were recorded by the GBP (Figure 1). On the opposite side in terms of fluctuations' amplitude was EUR. Regarding the relationship between the exchange rates of the four currencies, from the start of the timeframe up to September 2010, GBP ranks first, with quotations from 4.0777 RON/GBP (30 th December 2008, the absolute minimum for GBP) to 5.3722 RON/GBP (29 th June 2010), followed by EUR with quotations from 3.4719 RON/EUR (6 th August 2008, the absolute minimum for the EUR)    Economics, Vol. 14, No. 2 (2018), 7-19 A first examination aimed at emphasizing the characteristics of exchange rates' developments, expressed in RON for the entire period of time. The null hypothesis was tested regarding the randomness of these developments. In order to test the null hypothesis several type (1) models were settled, whose statistical significance was tested by F test ( , the null hypothesis is rejected and the alternative hypothesis (H1) is accepted: the exchange rate was not significantly influenced by random factors. Particularly, trends in exchange rates of EUR, CHF, GBP and USD during the time frame January 2008 -October 2015 turned out to be linear.
Regarding the coefficients a0 and a1 for the 4 models, they are statistically significant (for all P-value=0.00<=0.05). By analyzing the values of the regressors (a1), the CHF recorded the most pronounced evolution (an increase of 0.00089 RON per trading day, in a confidence interval between 0.00088 and 0.00092 RON per day), followed by GBP with a daily increase of 0.00071 RON (the confidence interval being comprised between 0.00069 and 0.00073 RON per day). At the opposite side we find EUR whose average daily increase was 0.00035 RON (2.54 times lower than CHF and 2.03 times lower than GBP), the confidence interval being between 0.00034 and 0.00036 RON. During the same period, USD recorded an average daily increase of 0.00056 RON (1.69 times higher than the EUR, but 1.2 times lower than the GBP and 1.53 times lower than CHF), the confidence interval being situated between 0.00056 and 0.00059 RON .
For testing the hypothesis of no significant differences between the EUR, USD, GBP and CHF exchange rates' evolution during the time frame under review, the dispersion analysis of the 4 data series was conducted, determining variances between groups and with groups, and thus applying the F test. The resulting information is comprised in Table 2, where SS -Sum of Squares, df -Degrees of Freedom, MS -Mean Square, P-value -probability corresponding to the value of F and F crit -critical value for , the null hypothesis (H0) is rejected and the alternative hypothesis (H1) is accepted: there are significant differences between the data series corresponding to analyzed exchange rates. In conclusion, although the exchange rates for the 4 currencies recorded an upward trend, the influence of random factors proving to be insignificant, during January 2008 -October 2015 the evolution of the exchange rates for EUR, USD, GBP and CHF differ significantly.

Correlations between exchange rates' fluctuations
Based on the series of data containing the exchange rates of EUR, GBP, CHF and USD against RON, over the time interval January 2008 -October 2015, table 3 shows the Pearson correlation coefficients corresponding to exchange rates' fluctuations.
A first observation relates to the fact that by testing their statistical significance (t-Test 2-tailed) for the significance threshold =0.005, it results that all are statistically significant (therefore null hypothesis regarding their insignificance was rejected).
A second observation relates to their values (r). Except for the relationship between the EUR and GBP for which r=0.702 shows a direct medium correlation, in all other cases the values of correlation coefficients highlight direct correlation of strong intensity. The strongest direct correlation was determined between USD and CHF (r=0.879), which emphasizes that the modification of one's exchange rate will determine, with 95% probability, the modification in the same direction of the other. Moreover, for the time frame subject to analysis, strong correlations have been determined between CHF and EUR (r=0.870), USD and GBP (r=0.867), GBP and CHF (r=0.857) and also between USD and EUR (r=0.826). To conclude, both the values of correlation coefficients between EUR, CHF, GBP and USD exchange rates (Table 3) and the results of the statistical significance testing confirmed the hypothesis concerning their interactions for the period January 2008 -October 2015. In order to highlight the connection between EUR the other 3 currencies, a linear model was tested using (10): The results obtained by applying the least squares method for determining the parameters (c,a1,a2,a3) and the ANOVA method for validating model (10) are shown in table 4.  (10) and coefficients thereof are statistically valid (Prob(Fstatistic)=0.000<=0.05, respectively Prob.=0.000 for all coefficients), the value d=0.015848 (Durbin-Watson test) indicates a strong positive autocorrelation of residual values and therefore the model (10) cannot be used.
Under these circumstances, with 95% probability, the linear regression model which describes the quantitative dependencies between all 4 currencies taken into analysis (Table 6) is: Bearing in mind the value of Prob(F-statistic)=0.000<=0.05 it turns out that model (12) is statistically valid. Likewise, the value of Prob=0.000 indicates that all model's parameters are statistically significant. Parameter values a1=0.195492, b=-0.02758 and c=0,403236 point out which was, over the observed time frame, the effect of modification by 1 RON of the exchange rate of CHF, GBP and USD respectively upon the exchange rate of EUR.

A time series analysis
For obtaining more information regarding the evolution trends of the EUR, CHF, GBP and USD exchange rates in Romania, for the period January 2008 -October 2015, the quotations of the 4 currencies were analyzed as time series. Testing of the stationarity hypothesis has been conducted using Augmented Dickey-Fuller test. For the chosen significance threshold   05 . 0   , both Prob. values (table) which are strictly higher than 0.05 and the fact that all ADF values are higher than the critical value (-2.862803) lead to accepting the null hypothesis for all 4 data series. Subsequently, all series are not stationary. Still, we need to make the observation that for the significance threshold (Confidence Level 90%), for the EUR data series the stationarity hypothesis was not accepted. For obtaining stationary series, the transformation provided by equation (5) was used, based on which the series DEUR, DCHF, DGBP and DUSD were generated. Testing their stationarity led to rejecting the null hypothesis (The series has a unit root) and to accepting alternate hypothesis. Therefore, DEUR, DCHF, DGBP and DUSD series are stationary and can be employed in identifying autoregressive and moving average processes which describe the fluctuations of the 4 currencies.
Following the testing of several types of models and the analysis of their quality, the AR-MA(3,4) models have been selected for DEUR series. For this particular series we identified an autoregressive and moving average type ARMA(3,4) model. Except for constant C (Table 8)  and thus the null hypothesis is accepted: constant C is not significantly different than zero, for all other variables the coefficients' values are statistically significant for the confidence level of 95% (coefficients of AR(3) and MA(1) are statistically significant for a confidence level of 99%).
The moment in time t=5 corresponds to the date of 9 th of January 2008.
Based on (13) and on the fact that On the other hand, for the DCHF series the autoregressive and moving average model is type ARMA(1,3). In this case too, the constant C (Table 9) is not statistically significant for a confidence level of 95% ( ). The coefficients of AR(1) and MA(3) variables are statistically significant for a confidence level higher than 99%).
The autoregressive and moving average model assigned to the evolution of DGBP is type AR-MA(3,1). Again, except for constant C (Table 10) which is not statistically significant for a confidence level of 95% ( , the coefficients of variables AR(1) and MA(1) are statistically significant for confidence level above 99%, while the coefficient of variable AR(3) is statistically significant for a confidence level of 95%.
The moment t=4 corresponds to the date of 9 th of January 2008. Since the model best describing the evolution of the RON/GBP exchange rate is type ARIMA(3,1,1), its explicit form turns out: Unlike the autoregressive and moving average processes corresponding to the other 3 currencies, in which cases the impact of previous trends became significant after 1 trading day for CHF and after 3 trading days for EUR and GBP, for the RON/USD exchange rate the time frame is wider (10 trading days). Also, while the time interval taken into calculation for the moving averages was comprised between 1 day (in the case of GBP) and 4 days (for EUR), the case of USD showed a 10 day-interval.

CONCLUSIONS
During 2008-2015, economic developments in most EU countries and Romania as well were significantly influenced by the economic crisis which started in the beginning of the proposed period. In Romania, even though at the beginning of 2008 the RON has appreciated against the other currencies subject to analysis, both during the crisis and afterwards the exchange rates presented an increasing trend, thus determining (between July 2008 -May 2015) an almost doubled RON/CHF exchange rate after just an average increase of 0.089 RON per trading day. During the same period, the price of USD had increased by 79%, and GBP and EUR were more expensive by 38% and 25% respectively.
Despite all economic oscillations, in Romania, their influence upon the exchange rates of main currencies did not create severe turbulences, the fluctuations associated with ascending trends of the main currencies being insignificant. Even though the RON/CHF exchange rate had 2 sudden increases in August 2011 (when the exchange rate exceeded 4 RON/1 CHF) and in January 2015 (when CHF exceeded 4.5 RON), these did not influence significantly the linear increasing trend observed for the whole period of time.
Another characteristic of the evolution of exchange rates in Romania is that although exchange rates for the four currencies recorded an increasing trend, and the correlations between them are significant, the correlation coefficients' values range from 0.702 (between EUR and GBP) and 0.879 (between USD and CHF), the dispersion analysis pointing out significant differences between their evolutions. This aspect is also underlined by the Durbin-Watson statistical value, corresponding to the first regression model showing the dependence between EUR and the other three currencies (Table 11), highlighting a strong autocorrelation of residual values and the influence of a factor not considered in the analysis carried out on EUR. Eliminating its influence has yielded a valid regression model which emphasizes that for a significance level of 95% the highest pressure on EUR was in fact exercised by CHF (a modification of one monetary unit leading to a change in the same direction of the EUR exchange rate of approximately 0.19 monetary units), followed by GBP and USD.