Factors affecting the exchange rate of Russia

(1)

Factors affecting the exchange rate of Russia

By

Gulshat Ustabayeva

10679464

Supervised by Razvan Vlahu

August 2015

University of Amsterdam

Business School

(2)

1

Abstract

In this thesis we examine how factors such as government expenditure, oil prices, interest rate, import and export may affect the real exchange rate of Russian ruble. We use International Monetary Fund, OECD and Rosstat quarterly data and cover the period starting from year 1995 to 2014. The main finding is that the real exchange rate at preceding period, the oil price and the government expenditure have the most significant effect on real exchange rate, while export, import and interest rate have insignificant effect.

Keywords: oil prices, exchange rate, Russian ruble

I would like to thank Razvan Vlahu for guiding me through the process of writing this thesis, my fiancé Michael for supporting me for these two years and my family for their endless love.

(3)

2

1. Introduction

There are three main basic perspectives on exchange rates such as a multiple equilibria, monetary regulations and fiscal regulations. Multiple equilibria – the exchange rate is determined by currency traders, large institutions recommendations; monetary regulations- central bank intervention; and fiscal regulations– future fiscal balance of the government, expectations if the government will be able to repay the debts.

In this thesis we examine the impact of factors like government expenditure, oil prices, interest rate, import and export on real effective exchange rate of Russian ruble using International Monetary Fund, OECD and Rosstat quarterly data starting from year 1995 to 2014.

Russia is the largest crude oil, petroleum products and natural gas exporter in the world which accounted to more than 60% of total exports revenues of the country according to Rosstat.

There are numerous studies about the oil price fluctuations effect on exchange rate of oil exporting countries, especially over the post-Bretton Wood era. Sharpe increase in oil prices starting from 2003 contributed to large current account surplus of oil exporting countries with Russia being one of them. The oil price is a very important factor in determining the economic activity of the country. The rise of oil prices increases a transfer of income from importing countries to exporting countries. Increase in income positively affects the internal investment in exporting country which leads to an increase in employment and this in turn increases the money supply and appreciation of the currency. Appreciation of an exchange rate discourages the export and encourages the import, while depreciation of the exchange rate has the opposite effect. The rise in oil prices will also decrease the demand of oil

(4)

3

importing countries. This will force importers to search and shift to alternative resources to oil which negatively effects the balance of trade of exporting countries. It was projected that oil demand will increase to 118 million barrels per day by 2030 where average demand in 2015 is 93 million barrels a day according to US Energy Information Administration. The most significant demand continues to come from China and India, with almost half of the oil demand growth being expected to come from China alone.

After the collapse of Soviet Union in December 1991, the monetary policy of the Bank of Russia was exchange rate oriented because of financial instability and hyper and high inflation between 1992 and 1998. Russian ruble was tightly managed especially after the government debt crisis in 1998 which was triggered by sudden fall of energy prices and Southeast Asia crisis in 1997. During the crisis of 1998 Russian ruble dropped sharply from 5.96 to 20.65 to US dollar over 6 months; the interest rates were increased sharply from 30% to 150% in May 1998, the government had very high fiscal deficits in 1998, the Bank of Russia had limited foreign currency reserves and Russian ruble was pegged to US Dollar. The tight monitoring continued until 2005 when the Bank of Russia introduced dual-currency basket as an operational indicator for exchange rate policy with a purpose of decreasing the volatility of the ruble against major currencies. Currently the value of the dual-currency basket is calculated as the sum of ruble values of 55% US dollar and 45% Euro. Still the exchange rate of Russia ruble is monitored and controlled although intervention volumes have decreased until 2014 except the period of financials crisis of 2008. The recent Ukraine crisis of 2014 brought tight interventions back. This crisis share similarities with the crisis of 1998, such as high dependency on oil exports, plunge of the oil prices, depreciation of Russian ruble and increase of interest rate by Bank of Russia. The

(5)

4

main difference is that this time the government had enough financial sources to support a declining ruble. As of December 2014 the reserves amount to 420 billion US dollars. These reserves were built during the time of high oil prices over the period of 2003-2014.

2. Literature Review

Many empirical relationships were established in the existing literature between oil price and Russian ruble. There is no “fit for all” model for all oil exporting countries, due to their different exchange rate regimes (Habib and Kalamova, 2007). In their study they considered Norway, Saudi Arabia and Russia. The study of Habib and Kalamova has shown that the long- term relationship between exchange rate and oil price was identified in case of Russia only.

On the other hand, Shafi and Hua (2014) argue that oil price and exchange rate have positive relation with economic growth of the Russia. They find that imports, exports, interest rates, inflation, government expenditure and foreign direct investment also have significant effect on exchange rate.

Cashin et al. (2004) investigate the relationship between real exchange rate and real commodity price of 58 developing countries and show that in one third of commodity exporting countries the long-run relationship between these two exists.

Sosunov and Zamulin (2006) demonstrate that the appreciation of Russian ruble in 1999-2005 is explained by growth of oil price which were due increase in both to oil prices the export.

Obstfeld (2004) shows that the macroeconomics has a heterogeneous effect across different economies. Financial integration for emerging economies was less beneficial than for developed countries. One of the reasons is that emerging markets should deal with exchange

(6)

5

rate risk. Fixing the exchange rate is required which makes it prone to crisis, alternatively the flexibility will lead to high volatility during the market stress.

Analysis of Saif Ali (2013) focuses on the ability of nominal oil prices to predict nominal exchange rates of several countries such as Canada, Mexico, Norway, UK, India, Singapore, Thailand, South Korea, Japan and Russia. He used lagged oil price and Vector Autoregression models to forecast exchange rates and compare them to random walk benchmarks. Besides this he tried to investigate the exchange rates based on oil price is different for oil exporting and oil importing countries taking into account the size of net export figures of the economies. The results demonstrate that oil price model can outperform random walk but this was valid in case of Russia only. He also find that the oil price significantly affect the exchange rate of net exporters and countries who has high share export or import of the oil.

Chen and Chen (2007) made test on Canada, France, Germany, Italy, Japan, United Kingdom and Japan countries using a cointegration test using data from 1972 to 2005. They found out that the real oil price may have been a dominant source for real exchange rate movements in these countries during given period.

The study of Lizardo and Mollick (2010) investigated the relationship between the real oil price and US dollar exchange rate against other oil-importing and oil-exporting such as Brazil, Canada, Euro Area, Iran, Japan, Mexico, Norway, Russia, south Africa, Saudi Arabia, Sweden, United Kingdom, United States of America, by including output inflation, and the money supply. The result is that an increase in oil prices leads to a depreciation of US dollar against net oil exporter countries such as Mexico, Canada and Russia but appreciation against oil-importer countries such as Japan.

(7)

6

US dollar is the most frequently used currency in international trade and unit of account to oil price worldwide. Thus there is the possibility that oil price fluctuations are influenced by US dollar. So in case of depreciation of US dollar the oil price is relatively cheaper for oil-importing countries. So what is the relationship between oil prices and US dollar? There are numbers of studies done to measure the relationship between oil price and US dollar. Ferraro et al (2011) and Chen et al (2013) and both find no long-run relationship. The effect of US dollar on oil price is found in studies of Alhajji (2004) and Haughton (1989) where the depreciation of US dollar makes it cheaper for oil-importing countries so increasing the demand in oil which eventually leads to increase in oil price.

Initially US dollar was pegged to gold and everything was pegged or floated to the dollar, so did the oil. The above literature considers the effect of oil price on exchange rates but there are also studies done on effect of US dollar on other currencies. The US dollar became the widest used currency globally and a measure of value. The trade can be concluded in any agreed currency which is at that currency’s current US dollar exchange rate. So below studies are done on exchange rates of other countries where US dollar is used as a main exchange currency.

Numbers of studies were done on efficiency of foreign exchange market on major currencies such as US dollar, Euro, Japanese Yen and other advanced economies currencies, but limited number of studies were focused on emerging markets currencies. Chong and Ip (2009) study looks at currencies of Mexico, Philippines, South Korea, Sri Lanka, Thailand and South Africa and indicates considerable excess return in currency markets between January 1985 and December 2004. The study of Tabak and Lima (2009) indicates that Brazilian real is predictable in long run rather than in short run. There are many factors suggested to

(8)

7

influence exchange rate but in case of Russian ruble central bank’s interventions are prevailing. The main reason of their intervention is to stabilize foreign exchange volatility and bring the currency to the politically acceptable level as it is suggested in study of Okunev and White (2003). The purpose of interventions is to push the exchange rate opposite to market direction. Otherwise there is no incentive for central bank’s intervention if market direction goes in line with their policy.

Mees and Rogoff (1983 and 1988) found that economic models explain and forecast exchange rates worse than random walk. The study of Rossi (2005) shows that investigating the economic model parameters and their nature carefully, gives the possibility to find some economic models that can be better than random walk in explaining the exchange rate and its future behavior.

3. Data Analysis

3.1. Data Description

The collected data are represented in the quarterly basis in range from 1999 Q1 to 2014 Q4 which makes 60 observations in total. All data including real effective exchange rate (REER), Money market rates of Russian ruble (INT), Import of Russia (IMP) in US dollars, and Export of Russia (EXP) in US dollars, Brent UK Oil price (OIL) in US dollar per barrel are collected from IFS, Rosstat and OECD.

The Bank of Russia’s policy is to maintain money market rates within the borders of interest rate corridor and keep the market rates to the key rate which is in line with monetary policy. So the instruments of monetary policy are used to control overnight borrowing between commercials banks. By moving this rate the Bank of Russia can influence the inflation level and currency exchange rate.

(9)

8

The oil price is a very important factor in determining the economic activity of the country. The rise in oil prices increases the transfer of income from importing countries to exporting countries. Increase in income positively effects the internal investment in exporting country which leads to an increase in employment and an increase in the money supply and appreciation of the currency. Appreciation of exchange rate discourages the export and encourages the import, while depreciation of the exchange rate has opposite effect. The rise in oil prices will decrease the demand of oil importing countries because those countries try to search and shift to alternative resources to oil which negatively effects the balance of trade of exporting country. Increase in oil prices positively effects economic growth and appreciation of national currency.

To increase the expenditures the government should have available funds which can be raised by issuing bonds which may increase the interest rate and attract investors which will lead to appreciation of the currency. Alternative source of funds in case of Russia is export revenues In case of Russia availability of the funds depends on oil revenues. Russia is producing 10.7 million oil barrel per day which makes 13.28% of world’s oil product in August 2015 (Energy Ministry of Russia) and exporting 7.5 million barrel oil per day. The main objective of the study is to investigate whether macroeconomic variables such as interest rate, government expenditure, oil price, import and export do affect currency value of Russian ruble if so to what extent. The objective is examined through regression analysis.

3.2 Regression Analysis

Time series modelling is widely used for the last few decades. Study of the past observations helps to describe inherent structure of the series which is used in constructing a model for forecasting the future by understanding the past. The impact of interest rate, oil prices,

(10)

9

government expenditures, export and import are is below on real exchange rate is expressed as following:

REER = β0+ β1EXP + β2 IMP + β3IR + β4OIL+β5GE

Multiple regression analysis was conducted which initially requires testing for normal distribution of the data which is a prerequisite in order to get a valid model.

3.3. Normality Test

Jarque-Bera test was conducted to test the normal distribution, the results are summarized in Table 1. The test determines the skewness and kurtosis, checking for asymmetry of the data and “peakedness” of the shape. Jargue-Bera test null hypothesis is that skewness and excess kurtosis are being zero.

Table 1. Jargue-Bera Test

Skewness Kurtosis Jarque-Bera Probability

GE 0.42 1.85 5.45 0.065 OIL 0.25 1.69 5.25 0.071 REER -0.37 1.88 4.8 0.091 IMP 0.18 1.51 6.22 0.044 EXP 0.091 1.5 6.02 0.049 IR 1.77 8.01 100 0.000

We found out that interest rate and oil price are not normally distributed, so transformed variables were created. The common transformations are the logarithmic transformation, the square root transformation and the inverse transformation. Logarithmic transformation was used and the variables were included in regression. Including the variables that were not

Notes: Kurtosis is quantified by Gaussian distribution where it has zero kutrosis, flatter distribution has negative kurtosis and more peaked distribution has positive kurtosis. In case the data is normally distributed the skewness coefficient is closer to zero and kurtosis is closer to 3. So the null hypothesis of skewness and kurtosis being zero is rejected in case the p value of the test is less than 0.05. For visual cues please see Appendix 1, figure 1.

(11)

10

transformed which reduces the effectiveness and estimated values of the dependent variable, so the analysis may be biased or systematically incorrect.

3.4 Correlation Test

Correlation together with linear regression analysis is the most used techniques to investigate the relationship between two variables. Pearson correlation coefficient test is used to test the strength of the linear relationship between normally distributed variables and the results are shown in Table 2. The test for linear dependency of variables was conducted to determine whether two variables co-vary and to quantify the direction and strength of the relationship of the variables. The magnitude of the correlation coefficient indicates the strength of dependency and varies between 1 and +1; where +1 is positive correlation and -1 is negative correlation.

Table 2. Pearson Correl. Test

Correlation REER REER 1 OIL 0,916236 GE 0,880294 IR -0,295402 EXP 0,952277 IMP 0,918482

From the table above we can see that REER has the highest correlation with oil price, export, import and government expenditure and a negative correlation with interest rate.

Negative correlation between interest rate and exchange rate can be explain by the fact that an increase in RUB interest rate may attract more foreign investors to invest in ruble which will lead to appreciation of the local currency so the exchange rate USD/RUB or EUR/RUB will decrease. Rise in the real interest rates attracts capital which leads to demand in

(12)

11 currency consequently to appreciation of the currency. It does not affect the currency if higher interest rate is simply due to inflation differential.

Positive correlation between oil price and exchange rate is in line with international trade theory which indicates that when the price of exporting commodity rises, if it is inelastic, this creates a demand for the exporting country’s currency so driving up the value of the currency. The oil price rise can be driven by increase in demand or reduction of the supply. Increase in value of the currency caused by rise in oil price is referred to as terms of trade effect, when export revenues increased with no large impact on import costs. Improvement in terms of trade puts upward pressure on the exporting country’s currency.

After all transformations the model is constructed based on stepwise multiple regressions, where the dependent variable is REER and independent variables are OIL, EXP, IMP, IR, and GE. Regression equation allows forecasting the value of dependent variable on the basis of the values of the independent variables, to compare the actual values with the expected values. First regression results in Table 3 indicate that the import, oil price and government expenditure have insignificant effect since the probabilities are 0.4939, 0.4180 and 0.1306 respectively.

Table 3. First Regression

First Regression t-stat Prob C 3.424372 0.0011 GE 1.533550 0.1306 IMP -0.688411 0.4939 IR_LOG -1.962611 0.0545 OIL 0.815629 0.4180 EXP_ROOT 3.182910 0.0023

(13)

12

3.4 Autocorrelation test

While working with time series data the most common problem people face is the autocorrelation. Autocorrelation occurs when observations are related with each other. If assumption that residuals are not correlated is violated, the model will lead to biased and inefficient estimates. Autocorrelation might lead to false results like standard error of regression may be much smaller than the real, standard errors of coefficients can be less than actual, so conclusions drawn on the basis of t and F cannot be used since the model will lead to biased and inefficient estimates.

Correlogram can be a good tool for visual diagnosis of autocorrelation, but is more intuitive not definitive way. To be sure to detect the presence of serial correlation we used Durbin-Watson statistic which is usually denoted by d and can be interpreted as follows:

 If d is close to zero (0), then positive autocorrelation is probably present;

 If d is close to two (2), then the model is likely free of autocorrelation;

 If d is close to four (4), then negative autocorrelation is probably present.

DW statistic is 0.674836 Appendix 2, Regression 1 which is closer to zero, so the presence of autocorrelation was detected and some interventions for its elimination are required. The most common method is to create a new variable which is equal to dependent variable at preceding period (t-1) and include it to the right hand side variable. It should be noted that this correction is based on theoretic belief, which may cause more problems than it solves, additionally it costs degree of freedom.

After adding another variable lag REER, DW statistics changed to 1.860101 which is closer to 2 indicating that no autocorrelation is present and can also be observed graphically. Please see the test result in Appendix 2, Regression 2.

(14)

13

3.5 Residuals Testing

Last after obtaining the regression model we need to check the residuals in order to validate our model. Randomness and unpredictability are crucial components of any regression model, if these two components are missing the model is not valid. Residuals – error is the difference between expected value and observed value, so the difference between observed and expected value should be unpredictable meaning none of the explanatory or predictive variables should be in the error. If error contains explanatory or predictive power it means the model is missing some of the predictive information. The main idea is that the deterministic part of the model is good at predicting that only resident randomness remains in the error. Residuals should be centralized around zero, should be neither systematically high nor low. So the model is considered correct if errors fall in symmetrical pattern and have a constant spread along the range. Dickey - Fuller and Kwiatkowski-Phillips-Schmidt-Shin tests were used to check the stationarity of the residuals.

The main purpose of Dickey and Fuller test is in checking the hypothesis of presence of stationarity. Null hypothesis is the presence of non-stationarity and alternative hypothesis is stationarity.

The null hypothesis is rejected if DF t-statistic is higher than or equal to critical values at specified significance level, often 5%, or 1% and even 10%.

Kwiatkowski-Phillips-Schmidt-Shin test’s interpretation differs from Dickey and Fuller test. Kwiatkowski-Phillips-Schmidt-Shin test is developed to complement unit root tests because unit root test has low power with respect to near unit-root and long-run trend processes. The null hypothesis is trend stationarity against the alternative of a unit root. The hypothesis is

(15)

14

rejected in case statistical value is less than critical level. The results of both tests are summarized in Table 4 and Eviews results are in Appendix 3, Tables 1&2.

In DF test the null hypothesis is that residuals are non-stationary and alternative hypothesis is that residuals are stationary. We reject the null hypothesis and accept alternative hypothesis, because DF test statistics is 10.2772 which is significantly higher than critical values 2.60, 1.94 and 1.61 at significant levels 1%, 5% and 10% respectively.

KPSS null hypothesis is that residuals are stationary and alternative hypothesis is that residuals are non-stationary. We accept the null hypothesis that residuals are stationary because KPSS t-statistic is significantly higher than critical value 0.21, 0.14 and 0.11 at significant levels 1%, 5% and 10% respectively.

Both tests produce the same results for residuals which appear to be stationary.

Table 4. ADF and KPSS Tests

ADF KPSS

H0: x ̴ I(1) H0: x ̴ I(0)

DF t-statistics 10.27725 KPSS t-statistics 0.82490 Significance

Level 1% 2.60 Significance Level 0.21

Significance

4. Discussion of the results

The regression results are summarized in Table 5, Eviews results are in Appendix 2. After conducting the first regression, autocorrelation of residuals was determined. Autocorrelation was eliminated by adding the real exchange rate at preceding period (REER_LAG) to the left hand side of the equation as it is discussed in Section 3.4. All test are conducted

Absolute values of statistics for Kwiatkowski-Phillips-Schmidt- Shin (KPSS) tests for null of stationarity in the series; the Augmented Dickey Fuller (ADF) test for the null of non-stationarity

(16)

15

according to 5% (0.05) significance level. The second regression results suggest that export, import and interest rate have less significance, because their probabilities are 0.6988, 0.6756 and 0.2210 respectively and significantly higher than 0.05. Insignificant variables removed from equation one by one. First the import is removed, the results of third regression suggest that export and interest rate are still insignificant; their probabilities are 0.34 and 0.13 respectively, still significantly higher than 0.05. Consequently, export is excluded from the fourthregression; the result suggests that import is still insignificant with probability value 0.15. Final fifth regression suggests that government expenditure, oil price and real exchange rate at preceding period are significant because their probability values are 0.003, 0.000 and 0.00 respectively and all of them are less than 0.05 which is the significance level of our test. There are three variables will such as government expenditure; the oil price and real exchange rate at preceding period are included into the final regression.

Table 5. Regression

2n St Regression 3rd Regression 4th Regression 5th Regression

t-stat Prob t-stat Prob t-stat Prob t-stat Prob Coefficient

C 1.33 0.1882 2.16 0.03 2.04 0,05 1.43 0.1568 5,857541 GE -1.66 0.1031 -1.81 0.07 -1.99 0,05 -3.10 0.0030 -0,00000000000391 IMP -0.42 0.6756 IR_LOG -1.24 0.2210 -1.53 0.13 -1.45 0,15 OIL 3.91 0.0003 -9.70 0.00 4.75 0,00 5.18 0.0000 0,170484 EXP_ROOT -0.39 0.6988 0.96 0.34 REER_LAG 8.45 0.0000 8.55 0.00 10.57 0,00 1.19 0.0000 0,875228

The final regression model is as following:

REER = 5.85 – 0.00000000000319GE + 0.17OIL + 0.87REER_LAG

The equation indicates that 1 unit increase in government expenditure will cause 0.00000000000319 unit decrease in real exchange rate. Also 1 unit increase in oil price will cause 0.17 units of increase in real exchange rate and 1 unit change in real exchange rate in preceding period will cause 0.87 unit increase in real exchange rate.

(17)

16

By changing the size and structure of the government spending the state can sufficiently maneuver in choosing objectives to pursue social and economic policy. However, not only the state can influence the size and structure of the government spending also external and internal policy of the state, regime, the welfare of the population and many other factors. The importance of government expenditure in developing countries including Russia is crucial since it stimulates development of the economy, acceleration of renovation, encourages development and implementation of advanced scientific achievements. Government spending has significant impact on social and cultural welfare of the country. Emerging markets especially exporting countries experience appreciation of their currencies during the commodities price increase. One of the important instruments to reduce the appreciation is government expenditure structure with focus and increase in public investment.

In open economies increase in government expenditure creates excess demand and consumption of local currency which leads to appreciation of the local currency.

In case of Russia it is worth to mention that the Central Bank of Russia prevents appreciation and monitors ruble within determined corridor which fits the government policy. Also necessary to mention that coefficient of government expenditure in regression is not significant.

Russia is commodities exporting country and its economy heavily depends especially on oil export revenues. The main factors like the government expenditure, real exchange rate, and competitiveness of the economy are based on management of oil revenues. Last year’s surge in oil prices lead improvement of terms of trade, consequently to appreciation of local currencies of oil exporting countries where Russia is one them.

(18)

17

Despite official announcement of floating regime of Russia, the Bank of Russia continues to monitor and influence the exchange rate. As it states in website of Central Bank of Russia “The official exchange rates of foreign currencies against the ruble are set by the Bank of the Russia without assuming any liability to buy or sell foreign currency at the above rate. The official exchange rates of foreign currencies against the Russian ruble are set by an order of the Bank of Russia and remain valid until the next order becomes effective” which explains the influence of real exchange effective rate of previous period on today’s rate. The final step is checking regression model’s predictive power. Below graph shows that the residuals are centralized around zero, neither systematically high nor low. The model is considered correct if errors fall in symmetrical pattern and have a constant spread along the rang. The red line is actual and green is fitted models, blue line is the difference between the green and red lines. The blue line is also called residuals (errors). The residual fall around zero and no large fluctuations are observed. We may conclude that the regression model’s predictive power is acceptable.

(19)

18

5. Conclusion

The results suggest that the real effective exchange rate can be explained by real effective exchange rate at preceding period, oil prices and by government expenditures. This can be explained by fact that Russia’s export revenues largely depend on energy exports, on management of these revenues and market interventions of the government.

(20)

19

References

Algozhina, A. (2006): “Inflation Consequences of “Dutch Disease in Kazakhstan: The Case of Prudent Fiscal Policy”. Research paper CERGE-EI.

Beckmann J. and Czudaj R. (2012): “Oil prices and U.S dollar exchange rate dynamics”.

Research Paper, University of Duisburg –Essen.

Cahtterjee S. and Mursagulov A. (2012): “Fiscal policy and the real exchange rate”. IMF

Working Paper, No.12/52.

Cakrani E., Resulaj P and Koprenska L. (2013): “Government spending and real exchange rate case of Albania”. European Journal of Sustainable Development, 2, 4, 303-310.

Cashin,P.: L.F. Cespedes and R. Sahary (2004): “ Commodity currencies and the real exchange rate”. Journal of Development Economics 106, 315-337.

De Castro F. and Garrote D. (2012): “The effects of fiscal shocks on the exchange rate in the EMU and differences with the US”. Banco de Espana, Working Papaers, No.1224.

De Gregorio, J and Wolf H. (1994): “Term of trade, productivity and the real exchange rate”. NBER Working Paper Series, No.4807.

Galkovskaya V. (2011): “Does nominal exchange rate have a predictive power for oil and

gas prices? The case of Russia”. Southern Illinois University Carbondale, Research Paper, Master of Science Degree.

Golub S:”Oil prices and exchange rates”. The Economic Journal 93, 576-593.

Gregory A. and Hansen B. (1996). “Residual-based tests for cointegration in models with regime shifts”. Journal of Econometrics 70, 99-126.

Habib and Manolova Kalamova (2007): “Are there oil currencies the real exchange rate of oil exporting countries”? Working Paper Series, No. 839

Hasanov F. (2010): “The impact of real oil price on real effective exchange rate: the case of Azerbaijan”. DIW Discussion Papers, No. 1041.

IMF(2006): “World Economic Outlook: Oil and Global Imbalances”

Johansen S. (1991): “Estimating and hypothesis testing of cointegration vectors”. Econometrica 59, 1551-1580.

Krugman P. (1983): “Oil and the Dollar”. NBER Working Papers 0554.

Obstfeld M. and Rogoff K. (1996): “Foundations of International Macroeconomics”. MIT Press, Cambridge, MA

(21)

20

Shafi K and Hua L. (2014): “Oil prices fluctuations and its impact on Russian’s economy: an exchange rate exposure”.

Spatafore N. and Stavrev E. (2003): “The equilibrium real exchange rate in a commodity exporting country: The Case of Russia”. IMF Working Paper Series, WP0/03/93.

Zanello A. and Desruelle A. (1997):””A Prime on the IMF’s Information Notice Sysytem”. IMF Working Paper Series, WP/97/71.

Zhuk O. and Gharleghi B. (2015): “The Augmented Monetary Model and Exchange Rate Fluctuations”

(22)

21

Appendix 1

Figure 1

Normality Test Histogram

Gov.Exp

Oil Price REER

IMP EXP

Int.Rate

A histogram is used to test whether the data is normally distributed. It helps to visualize if whether it approximates the bell curve for normal distribution.

(23)

22

Appendix 2

Regression 1

Dependent Variable: REER Method: Least Squares Sample: 1999Q1 2014Q4 Included observations: 64

Variable Coefficient Std. Error t-Statistic Prob.

C 33.56210 9.800951 3.424372 0.0011 GE 3.10E-12 2.02E-12 1.533550 0.1306 IMP -9.32E-05 0.000135 -0.688411 0.4939 LOG_IR -3.068799 1.563631 -1.962611 0.0545 OIL 0.055999 0.068657 0.815629 0.4180 ROOT_EXP 0.194933 0.061244 3.182910 0.0023

R-squared 0.917733 Mean dependent var 83.38688

Adjusted R-squared 0.910641 S.D. dependent var 19.43605

S.E. of regression 5.810018 Akaike info criterion 6.446104

Sum squared resid 1957.866 Schwarz criterion 6.648499

Log likelihood -200.2753 Hannan-Quinn criter. 6.525838

F-statistic 129.4040 Durbin-Watson stat 0.674836

Prob(F-statistic) 0.000000

Regression 2

Dependent Variable: REER Method: Least Squares

Sample (adjusted): 1999Q2 2014Q4 Included observations: 63 after adjustments

C 9.613669 7.216875 1.332110 0.1882

GE -2.51E-12 1.51E-12 -1.656954 0.1031

IMP -3.87E-05 9.20E-05 -0.420636 0.6756

LOG_IR -1.350401 1.090968 -1.237800 0.2210

OIL 0.191402 0.049011 3.905262 0.0003

ROOT_EXP -0.018875 0.048525 -0.388976 0.6988

REER_LAG 0.897719 0.106267 8.447762 0.0000

(24)

23

Regression 3

C 11.62062 5.375426 2.161805 0.0348 GE -2.66E-12 1.46E-12 -1.818142 0.0743 LOG_IR -1.528865 0.997807 -1.532225 0.1310 OIL 0.193126 0.048486 3.983167 0.0002 ROOT_EXP -0.033224 0.034260 -0.969767 0.3363 REER_LAG 0.900543 0.105286 8.553258 0.0000

Adjusted R-squared 0.958171 S.D. dependent var 19.01024 S.E. of regression 3.888019 Akaike info criterion 5.644069 Sum squared resid 861.6515 Schwarz criterion 5.848177 Log likelihood -171.7882 Hannan-Quinn criter. 5.724346

Regression 4

C 10.85925 5.315051 2.043113 0.0456

GE -2.87E-12 1.44E-12 -1.988967 0.0514

LOG_IR -1.441492 0.993221 -1.451330 0.1521

OIL 0.159160 0.033510 4.749564 0.0000

REER_LAG 0.832846 0.078777 10.57220 0.0000

Adjusted R-squared 0.958214 S.D. dependent var 19.01024 S.E. of regression 3.886023 Akaike info criterion 5.628688 Sum squared resid 875.8680 Schwarz criterion 5.798778 Log likelihood -172.3037 Hannan-Quinn criter. 5.695585

(25)

24

Appendix 2 cont.

Regression 5

C 5.857541 4.084098 1.434231 0.1568

GE -3.91E-12 1.26E-12 -3.096025 0.0030

OIL 0.170484 0.032894 5.182891 0.0000

REER_LAG 0.875228 0.073847 11.85188 0.0000

(26)

25

Appendix 3

Table 1

Null Hypothesis: D(RESID_REER) has a unit root Exogenous: None

Lag Length: 1 (Automatic - based on SIC, maxlag=10)

t-Statistic Prob.* Augmented Dickey-Fuller test statistic -10.27725 0.0000

Test critical values: 1% level -2.604073

5% level -1.946348

10% level -1.613293

Augmented Dickey-Fuller Test Equation Dependent Variable: D(RESID_REER,2) Method: Least Squares

Variable Coefficient Std. Error t-Statistic Prob. D(RESID_REER(-1)) -2.342589 0.227939 -10.27725 0.0000 D(RESID_REER(-1),2) 0.636230 0.133128 4.779068 0.0000 R-squared 0.753482 Mean dependent var -0.225674 Adjusted R-squared 0.749232 S.D. dependent var 8.154489 S.E. of regression 4.083505 Akaike info criterion 5.684553 Sum squared resid 967.1506 Schwarz criterion 5.754365 Log likelihood -168.5366 Hannan-Quinn criter. 5.711860 Durbin-Watson stat 1.995891

Table 2

Null Hypothesis: D(RESID_REER) is stationary Exogenous: Constant, Linear Trend

Bandwidth: 4 (Newey-West automatic) using Bartlett kernel

LM-Stat. Kwiatkowski-Phillips-Schmidt-Shin test statistic 0.082490

Asymptotic critical values*: 1% level 0.216000

5% level 0.146000