Charles University Faculty of Social Sciences Institute of Economic Studies

Charles University Faculty of Social Sciences Institute of Economic Studies MASTER'S THESIS The Impact of Oil Prices in Norway on Macroeconomic Indicators Author: Bc. Peter Bogren Supervisor: prof. Roman Horváth Ph.D. Academic Year: 2016/2017 1

Declaration of Authorship The author hereby declares that he compiled this thesis independently; using only the listed resources and literature, and the thesis has not been used to obtain a different or the same degree. The author grants to Charles University permission to reproduce and to distribute copies of this thesis document in whole or in part. Prague, June 14, 2017 Signature 2

Acknowledgments The author is thankful to Prof. Roman Horváth Ph.D for his professional and kind guidance during the whole process of finalizing this Master Thesis, it would not have been possible without his support, advice and assistance. The author finds all chosen sources useful and interesting. All visited libraries, such as library in IES FSV UK, CERGE-EI, Czech Academy of Sciences and National Library of Technology have a lot of professional sources and databases available. 3

TABLE OF CONTENTS ABSTRACT... 5 INTRODUCTION... 6 BACKGROUND INFORMATION INCLUDING LITERATURE REVIEW... 8 HOW DOES THE NORWEGIAN STATE RECEIVE REVENUE... 8 DO OIL PRICES DIRECTLY AFFECT THE ECONOMIC ACTIVITY IN NORWAY?... 9 LITERATURE REVIEW... 10 OIL PRICE AND EXCHANGE RATE... 11 OIL PRICE AND GDP... 11 OIL PRICES, UNEMPLOYMENT RATE AND GDP... 12 OIL PRICES AND EXPORTS... 13 OIL PRICES AND INTEREST RATES... 13 OIL PRICES, INTEREST RATE AND UNEMPLOYMENT RATE... 14 OIL PRICES AND INDUSTRIAL PRODUCTION... 14 OIL PRICES, EXCHANGE RATE AND GDP... 15 THE MAIN POINTS OF AUTHORS... 15 CRITICISM OF STUDIES... 16 METHODOLOGY... 17 INTRODUCTION TO SVAR MODEL... 17 ADVANTAGES OF SVAR... 17 THE MODEL-IDENTIFICATION IN SIMULTANEOUS EQUATIONS... 18 IMPULSE RESPONSE FUNCTIONS... 20 THE MODEL-SVAR ORTHOGONAL IDENTIFICATION... 21 NORMALIZATION OF SVAR MODEL... 21 RESTRICTION IN MATRIX Γ... 22 MACROECONOMIC ASPECT OF THE THESIS... 23 RELATIONSHIP BETWEEN OIL PRICE AND INTEREST RATES... 23 RELATIONSHIP BETWEEN OIL PRICE AND MACROECONOMIC VARIABLES... 24 RELATIONSHIP BETWEEN OIL PRICE AND UNEMPLOYMENT RATE... 24 RELATIONSHIP BETWEEN OIL PRICE AND INDUSTRIAL PRODUCTION INDEX... 25 MACROECONOMIC BEHAVIORS BETWEEN BETWEEN THE VARIABLES... 25 DATA... 26 THE DRAWBACK... 26 DESCRIPTIVE DATA... 26 HISTOGRAM AND Q-Q PLOT OF EACH VARIABLE... 27 CHECKING THE STABILITY CONDITION OF LOG VARIABLES... 34 DATA STATIONARITY ANALYSIS OF EACH VARIABLE... 35 DIFFERENCING THE DATA AND AUTOCORRELATIONS... 42 THE STRUCTURAL BREAK TEST... 48 ESTIMATING SVAR... 49 ESTIMATION RESULTS FROM THE ESTIMATION TABLE... 51 ESTIMATION RESULTS FROM MATRICES A AND B... 51 STABILITY OF SVAR ESTIMATES... 52 IMPULSE RESPONSE FUNCTIONS... 52 CONCLUSION... 56 APPENDICES... 59 REFERENCES... 75 4

ABSTRACT This thesis analyzes the impact of oil price shocks on the selected macroeconomic variables in Norway for the period of 1990 till 2016. Lag-length test and structural vector autoregressive models are also applied to determine the oil price shocks effect on macroeconomic indicators. We are incorporating 1990-2016-time horizon to show the before-crisis, during and after the crisis oil price behavior. We will show that Norwegian oil prices have a strong and significant impact on exchange rate and export per capita. In the context of significant impact, we mean the oil prices will be dependent on these macroeconomic variables. We assume to have the correlation strong. We will show that oil price increase has opposite effect on export per capita and negative impact (decrease) on industrial production index. Last hypothesis states that oil price increase cause interest rate to rise and exchange rate to appreciate. 5

1. Introduction Oil is one of the world s most important commodities with its importance to the functioning of the global economy. Countries with large oil reserves shape their economies and strategic interactions with the world through maximizing oil extraction profits. Since oil is the most traded commodity among the economies (What are the most commonly traded commodities, 2017), oil price shocks can have important effects on economic growth and other macroeconomic variables. The main objective of this thesis to analyze how the oil prices in Norway impact the macroeconomic indicators, such as interest rate, total industrial production index, unemployment and so on. We can see that the country is very special because it is way different from the other Nordic countries because of its history of renewed resource wealth. Oil is the main resource that drove the economy upwards when it was discovered. Norway is the 6 th largest supplier of crude oil. The oil and gas sector constitutes around 22% of Norwegian GDP and 67% of Norwegian exports (Norway, [no date]). In our thesis, we will study the correlations between the economy and oil prices using structural vector autoregressive model, as well as whether Norway benefits from oil price increases. It will allow us to gain an insight into the mechanism of oil effects on the economy. The time horizon will be between 1990 and 2016 since it will explore the pre-crisis, during crisis and after crisis behavior of oil prices and the macroeconomic indicators. The period is very vital because there were several oil price fluctuations and more importantly, sharp oil price decline in the year of 2015. Moreover, Dutch Disease is an important issue that will be mentioned in the thesis as well. The Dutch disease explains the casual relationship between economic development and oil sector and decline in other sectors such as manufacturing sector or agriculture. The thesis will examine the Dutch Disease in terms of unemployment rate. Labor demand will increase in the booming sector (oil); however, will decrease in other sectors (manufacturing). The Dutch disease will not be analyzed through estimation, since the thesis will entirely focus on unemployment in a general level (across sectors). Before we conduct structural vector autoregressive model (SVAR) estimation, we will use VAR stability test to see whether our estimates are stable. After the stability test, we will test for stationarity for each of the variable. If the variables turn out to be non-stationary, we will use the first differencing to make them stationary. The Dickey-Fuller test will show us whether the variables are stationary or not. Moreover, the autocorrelations will also give us greater insight about the characteristics of the variables, such as what are their significant 6

lagged values and whether they are stationary or not. The thesis will provide SVAR models for time series and forecasting. After choosing variables, selection of lag length and checking stability will be conducted, and then SVAR models will be estimated. Impulse response functions will also be used to examine the dynamic effects of structural shocks (shock such as oil price increase on economic indicators). Since SVAR model includes several variables, relationship between the independent variables will be examined. OECD and other source data will be used for the observations. These data will be applied to all relevant models. Oil price will be treated as exogenous, which means that the other macroeconomic variables will have no impact on oil prices. My thesis only focuses on the effects of oil prices. Since Norway has effective economic policies (institutional quality is high in Norway) and their oil and gas sector accounts only 22% of their GDP, we might expect that the oil prices do not directly affect the macroeconomic indicators. The hypotheses that we are testing are as follows: Firstly, movement of oil prices have significant impact on economic indicators, such as exchange rate and export per capita. When oil prices increase, we expect the inflation to increase; thus, affecting the exchange rate. When the exchange rate is affected, then we expect the exports to be greatly influenced. The fluctuations of oil prices will be analyzed in order to see the overall pattern and be able to forecast. Second hypothesis is that higher oil prices have opposite effects on export per capita and negative effect on total industrial production index. Lastly, we expect that the increase of oil prices will cause interest rate to increase and exchange rate to appreciate. All the hypotheses are based on macroeconomic theories. These findings will provide important implications for petroleum investors by contribution of better understanding of oil price effects in Norway. The thesis will also look at how the Norwegian state receives revenue from the oil industries to examine how the oil price affects the government revenue. We will also have a greater insight about Norwegian economic policies, which are effective in combating oil price fluctuations. Moreover, we will also focus on how the decrease of oil prices in 2015 affected the Norwegian economy. The thesis will proceed as follows: Section 2 will deal with background information (consisting of literature review) on how Norwegian state gets oil revenue, but also whether oil prices directly affect economic activity in the context of Norway. Section 3 will deal with the literature review. We will compare our thesis with other influential studies. Section 4 will consist of methodology, specifically the description of the SVAR model and impulse response functions. Section 5 will focus on the macroeconomic aspects of the independent 7

variables. It will provide how these variables behave and how they affect each other according to various macroeconomic theories. The section also provides the reasons for our chosen hypotheses. The next section will include the data drawback and description of the data given the the number of observations and volatility. The ordering of the variables will be explained. The Q-Q plot and histogram will be part of the data description. The stability of the model will be conducted to see if our SVAR model is stable. Stationarity of each of the variables will be analyzed as well. If some variables are non-stationary, differencing the data will be done. The autocorrelations of differenced variables will be shown. The structural break tests will be conducted to see whether unexpected changes of variables occurred. The seventh section will include SVAR estimation and its results. The impulse response functions will be part of the section. The last section, which is the conclusion, will enable us to see whether our hypotheses were valid or not. It will also provide the summary of the results and what were the overall challenges. 2. Background information including literature review a. How does the Norwegian state receive oil revenue? Since it is vital to be inspired from studies which analyze the effect of oil prices on Norwegian macroeconomic indicators using econometric models, it is mostly important to know how the state gets revenue from the oil industries and from that point we can find out more easily how the oil price impacts the macroeconomic indicators. The article called The Government s Revenues describes facts and explain how the Norwegian state gets revenue and what impacts has it on society. The article emphasizes that principles of Norway s management of petroleum resources and government revenue must benefit the society. The government is willing to spend money by investing in future. The Norwegian government receives money from petroleum activities through taxes, environmental taxes and area fees, dividend from state owned companies and through system known as System s Direct Financial Interest (Norwegian Ministry of Petroleum and Energy 2016). The SDFI is a system under which the state owns holdings in oil and gas fields, pipelines and and other facilities. For oil and gas fields, the system awards them with production licenses and from that the government receives share of income. The government s total net cash flow in 2015, including the dividend from Statoil and different types of fees was 218.3 billion kroner; however, in 2014 it was higher, totaling 312 billion kroner. The decrease of the net cash flow was particularly due to decrease of oil price. The article is important since it shows how the 8

government taxes the petroleum. The petroleum taxation is based on the rules of ordinary company taxation (Norwegian Ministry of Petroleum and Energy 2016); however, the oil companies are subject to an additional special tax (Norwegian Ministry of Petroleum and Energy 2016). In Norway, the current company tax rate is 25% and the special tax rate is 53%. Furthermore, the Norwegian state owns 67% of shares in Statoil (state oil company) and thus receives dividends. The article shows that area fee is a huge contributor to the government revenue. The area fees ensure that the explorations are efficient. The environmental tax is also imposed under which companies must pay purchase emission allowances if their greenhouse gas emissions are more than their allocated amount. The government revenue is vital for the Norwegian economy since it impacts its expenditure. The article does not show econometric analysis; however, it shows very useful information; especially how the state receives revenue. From that point, we can see the relation between the oil price and government revenue. b. Do oil prices directly affect the economic activity in Norway? According to Gjedrem (1999), the economic activity is not only affected by oil prices. He states that exchange rate is not the only determinant of oil prices. The impact of oil price fluctuations on the exchange rate and on economic stability in general will depend on the speed at which petroleum revenues are used (Gjedrem, 1999). Moreover, he states that Norwegian economic policy is organized against instability in the Norwegian economy and large fluctuations in the exchange rate over time (Gjedrem, 1999). Fluctuations of the oil revenues would lead to economic instability; thus, the Norwegian policy can counteract that. In order to reduce dependence of government spending on oil revenues, petroleum fund was created. Interestingly, government can use 3% of the petroleum fund for spending. Norway has created a Petroleum Fund, which receives revenue from the petroleum sector, transfers the amount necessary to produce a balanced government budget and invests the surplus abroad. As long as the increasing petroleum income is kept outside the domestic economy, there will be less need for structural change and thus less need for exchange rates to change (Gjedrem, 1999). The fund has many objectives. One of the objectives is to store wealth and buffer against changes in oil prices. In a way, the fiscal policy in Norway is aiming at maintaining the desired size of the exposed sector and stabilizing the economy (Gjedrem, 1999). Since high share of state revenue consists of petroleum revenues, any fluctuations in oil prices will result in changes in allocations to the Fund; however since all of the Fund s capital is invested abroad, such changes will in principle not influence 9

economic activity (Gjedrem, 1999). Norwegian economy is therefore more robust to oil price fluctuations (Gjedrem, 1999). To make a point, it might be a case that oil price do not affect macroeconomic indicators directly. It is the economic policy decisions that influence the economic activity of Norway. In one way or another, the Norwegian economy can be regarded as insured against oil price fluctuations. Moshiri (2015) included in his study a hypothesis that states, oil price shocks do not have significant effects on developed oil-exporting countries. The developed oil-exporting countries have oil revenue comparable to those of developing oil-exporting countries, but their oil shares of GDP are small and their economies are more diversified and making them less vulnerable to oil price shocks (Moshiri, 2015). He also states in his third hypothesis that the differences in responses of the developed and developing oil-exporting countries to oil price shocks can be explained by differences in their institutional quality (Moshiri, 2015). The institutional quality is high in Norway; thus, we can expect that Norway benefits from strong and established institutions. Moshiri (2015) also explains that developed oil-exporting countries (including Norway) are not as dependent on oil as developing oil-exporting countries, since oil export shares of total exports in developed countries are 6 to 29 percent (Moshiri, 2015). Norway s oil export share of total exports is currently 22% (Norway, [no date]). Moshiri (2015) found out that the average oil dependency for OPEC members was 28% and for non-opec members in the 1990 s, but increased 2 percentage points for non- OPEC countries in 2000 s. He also discovered using Granger causality test that oil price shocks in Canada, Norway and UK do not have significant effects on their economic performance. The reason is that the structure of these countries is well diversified (Moshiri, 2015). Regarding institutional quality, Moshiri (2015) found out by using panel data growth model that when the institutional quality is better, the more it alleviates the negative effect of negative oil price effects. Norway and Canada have managed to avoid economic fluctuations associated with oil price volatility through establishing oil-specific funds and proper and fiscal policies (Moshiri, 2015). Due to the study of Moshiri (2015) we expect that the oil price in Norway will not significantly impact most of the macroeconomic variables. 3. Literature review Given the importance of oil price shocks on macroeconomic shocks, some studies have observed the how the oil price effects the exchange rate, total industrial production index, export per capita, interest rate, unemployment rate and exchange rate. These studies used 10

different time periods and methodologies to understand the relationship between the oil prices and the macroeconomic indicators mentioned above. Even though Moshiri (2015) showed that economic activity is not greatly affected by oil price fluctuations in developed oil-exporting countries, we still want to investigate other studies that might show otherwise. a. Literature review: Oil Price and Exchange rate One of the studies (Gocekli and Peker, 2015) highlight the effect of oil prices on exchange rate in case of Turkey. Peker (2015) have shown that the increase of oil leads to exchange rate to increase in oil-exporting countries, but in the long term it will depreciate. The time period used in the study is between 2003 and 2014 via monthly data. In order to observe the relationship between the oil price and exchange rate, Gocekli and Peker (2016) used Johansen co-integration method. Before applying the co-integration, he tested whether the variables are stationary through using Dickey-Fuller test. The oil prices and the exchange rate (USD against Turkish lira) were differenced; thus, becoming stationary. After conducting differencing and finding out that the results were stationary, Gocekli and Peker (2015) found out the optimal lag length before conducting Johansen integration test. The test showed that at least one co-integration relationship exists. Gocekli and Peker (2015) found out that 1% increase in exchange rate of dollar leads a 1.32% decrease in the crude of oil prices. He also concluded (using error correcting model) that the exchange rate is the causal of the oil price and its effect is negative. An article made by Holter (2016) takes aim at the role of exchange rate when the oil price decreases. The director general of the Norwegian Petroleum Directorate told to Bloomberg that Norway s oil industry is in the crisis since krone lost 0.8 percent against euro with crude oil down about 2.8% (Holter 2016). Holter (2016) also describes the structure of Norwegian economy. It shows that Norway depends on oil and gas about one fifth of its economic output (Holter 2016). The Norwegian finance minister said that the krone is often correlated with oil price; however, the devaluation of krone also represents improved competitiveness for the rest of the Norwegian economy (Holter 2016). b. Literature review: Oil prices and GDP Another influence comes from the studies called Macroeconomic Responses to Oil Price Increases and Decreases in seven OECD countries (Mysen 1994), and Oil Price shocks and real GDP growth: Empirical evidence for some OECD countries (Sanchez 2005). Mysen (1994) investigates the correlation between oil price movements and GDP fluctuations. His 11

study focuses on seven countries, not only Norway. He utilizes bivariate correlations in which he found out that oil price correlations are negative and significant for most countries, but positive for Norway, whose oil producing sector is large relative to the economy as a whole. Sanchez (2005) uses on his study VAR model to show the effects of oil price shocks on the real economic activity of the main industrialized countries. He investigated that the effect of oil shocks on GDP growth differs between UK and Norway. UK is negatively affected by an oil price increase and Norway benefiting from it. Both the authors use VAR models to examine the relationship between oil price shocks and macroeconomic indicators. Makau (2017) studied the effects of oil prices effects on Kenya s GDP growth. He used exchange rate and inflation rate as intervening variables. Moreover, he included quarterly data, from year 2004 to 2013. He uses an OLS regression to get the parameter estimates (Makau,2017). One of his results show that when crude oil price goes up KSh 1,000, GDP growth rate dips by 0.165 percentage points. Makau (2017) also found out that when oil price increases, inflation also increases (Makau, 2017). From using the multivariate regression model, when the oil price increase by 1,000 KSh per barrel, the Kenya shilling weakens by a single shilling for every US dollar and the inflation rate goes up by 1 per cent, the GDP growth decreases by 0.132 percentage points (p=0.000). The decrease in GDP growth rate is 86.9% of the actual decline as the model assumes in the absence of other indicators such as interest rates (Makau, 2017). Furthermore, Makau (2017) also explained the effects of intervening variables to the multivariate regression model. He concluded that exchange rate was the most significant intervening variable compared to inflation rate. He shows that fluctuations in both the crude oil price and exchange rate explains 83.2% of the behavior of the GDP growth rate (Makau 2017). Regarding inflation rate, Makau (2017) shows that including inflation rate to the model is insignificant in explaining the trend in GDP growth rate. c. Literature review: Oil Prices, Unemployment rate and GDP Mignon (2008) investigates the links between oil prices and various macroeconomic variables for three groups of groups of countries: OPEC, oil exporting countries and oilimporting countries. The macroeconomic variables are GDP, consumer price index, household consumption, unemployment and share prices. He focuses on both the short-run and long-run interactions between oil prices and the various macroeconomic variables. For the short-run, Mignon (2008) uses causality tests and for the long-run, he uses co-integration 12

analysis, both in time series and in a panel framework. Mignon (2008) investigated that long term analysis concerns GDP, unemployment rate and share price (financial variable). GDP and oil prices evolve together in the long run for most of the countries studied (Mignon 2008). The relationship between oil prices and CPI and the relationship between oil price and unemployment rate is particularly strong in Norway. Moreover, the study showed that conducting short term analysis results in causality running from oil prices to the other considered variables. There is a strong causality running from oil to share prices, especially for the oil exporting countries. In the long run analysis, Mignon (2008) found out that GDP and oil prices have a strong relationship. The relationships between oil prices and unemployment rates or share prices only concern non-opec members (Mignon 2008). d. Literature review: Oil Prices and Export Wengi Zhou (2017) examines the correlation between the export trade and oil prices. She mentions that during the 2011-20015 global crisis, the global crude oil prices fell, China, Europe, The United States and other regions showed negative correlation between export trade and oil prices (Zhou, 2017). The author, Wangi Zhou, uses dynamic factor model to analyze the correlations. She found out that strong correlation exists between these variables. Moreover, she states that for the oil exporting countries, like Middle East, Mexico and Canada, the rising crude prices will lead to a decline in exports, the falling crude oil prices will rise the exports. Since Norway is an oil exporting country, we can assume that oil price increase causes export to decrease in Norway. The author uses a different model to analyze the correlation between the oil price and exports. SVAR will be the main target. Moreover, she also used Granger causality test to show the relationship between oil price and the export of major economies. Her results show that the volume of export trade an all countries is the non-linear Granger cause of oil prices. Japan is the only country, where their export trade is not affected by oil prices. e. Literature review: Oil Prices and Interest rates Study written by Krishna and Malhotra (2015) tries to show the correlation by oil prices and interest rates (case of India) by using GARCH model. The increase of oil prices produces some positive impact on economy of oil exporting countries, but not for the oil importing countries (Krishna and Malhotra, 2015). The study is focusing on the period between April of 2004 through September 2014. Monthly data are used for the model. Interestingly, the whole price index is utilized in the study as a measure of inflation and interest rate as repo rate. 13

Moreover, the author used Granger causality test to observe the relationship between oil prices and interest rates. It indicates that oil prices will be able to affect interest rates (Krishna and Malhotra, 2015). His empirical results suggest that oil prices have significant effect on inflation, but no direct and consistent effect on interest rates (Krishna and Malhotra, 2015). The reason is that it could be attributed to lagged response in changing interest rates by monetary authorities to check inflation fueled by high energy prices (Krishna and Malhotra, 2015). Krishna and Malhotra (2015) confirmed by using Granger causality test that the crude oil prices statistically significantly affect Indian interest rates at 3 lags included. f. Literature review: Oil Prices, Interest Rate and Unemployment Rate Soytas (2010) perfectly describes the relationship between oil prices, interest rate and unemployment in Turkey. The author uses a new technique called Toda-Yamamoto to test for a long run Granger causality between the variables, which finds that the real price of oil and interest rate improve the forecasts of unemployment in the long run (Soytas, 2010). Monthly data are used, from 2005 till 2009. Soytas describes that unemployment rate will also follow closely the local factors such as the state of the economy, business cycles, the technology level, and population demographics, as well as global factor like energy prices. He found out that both oil prices and interest rates affect unemployment in Turkey using the new technique. The author states the relationship between unemployment and oil prices in a macroeconomic aspect. The study shows that another way how energy prices may influence unemployment is through the relative prices of factors of production. Moreover, regarding elasticity of supply, the change in the equilibrium in labor market can be attributed to demand changes caused by the oil prices and interest rates. On the other hand, impulse response function indicates that shocks to oil price and interest rate have respectively negative and insignificant impacts on unemployment. Another finding through the impulse response functions is that the oil price shocks have a positive impact on interest rates. g. Literature review: Oil Prices and Industrial Production (GDP) Cobo-Reyes (2005) focuses his study on analyzing the relationship between the oil prices and industrial production and between oil prices and stock returns. He developed a Markov switching model, which determines the mean of the industrial production. Monthly data is used, from January 1963 to May 2004. Reyes (2005) found out that oil price increases negatively and significantly affects the industrial production. The reaction on stock returns is 14

higher in the immediate period, but after four periods, the response of industrial production is higher. h. Literature review: Oil Prices, Exchange Rate and GDP Al-mulali (2010) aimed at studying the impact of oil shocks on the real exchange rate and the gross domestic product in Norway from 1975 till 2008. He used co-integration test and Granger causality test. His results indicate that increase in oil price is behind Norway s GDP increase and increase of its competitiveness to trade by its real exchange rate depreciation (Al-mulali, 2010). The study states that the OPEC members were badly influenced from high level of oil prices during the period of 2003 till 2008, while Norway as an oil exporting country its inflation rate remained low and stable (Al-mulali, 2010). Most of his variables resulted to be non-stationary because they have a time trend. His variables are GDP, real effective exchange rate, net foreign direct investment, trade balance of goods and services, total trade of goods and services, consumer price index, inflation rate, and employment rate (Al-mulai, 2010). His ADF results show that all the variables resulted in stationarity after the first differencing. From using the Johansen test, he found out that oil price has a long run positive relationship with the real exchange rate (Al-mulali, 2010). One percent increase in oil prices cause the real exchange rate to depreciate by 0.22%. According to the study, increase in oil prices will therefore cause higher exports due to depreciation of the currency. Furthermore, the co-integration equations showed that 1% increase in oil prices will increase Norway s GDP by 0.0957%, indicating the higher oil price the better economic growth in Norway (Al-mulali, 2010). Granger causality test showed that oil price influences GDP in the long run. Usama (2010) concludes that increase in oil price is a blessing for Norway due to positive effect on exchange rate and GDP. The reason is that Norway uses the floating exchange rate regime which always act as a shock absorber (Al-mulali, 2010). The monetary policy is independent; thus, it can choose the inflation rate independently (Almulali, 2010). i. The main points of authors We can see that Gocekli and Peker (2015) showed that increase in oil prices affect exchange rate. The exchange rate appreciates. In the long run, Gocekli and Peker (2015) found out that the exchange rate will depreciate (the case of oil-exporting countries). On the other hand, Holter (2016) witnessed that when oil price decreased, the krone devaluation took place. 15

Regarding oil prices and GDP, Mysen (1994) illustrated to us that Norway (its GDP) is positively and significantly affected by the oil price increase. In case of Kenya, Makau (2017) indicates that rise in oil prices leads to decline in GDP growth. Mignon (2008) has shown that oil prices significantly GDP, especially in the long run. Cobo- Reyes illustrated that oil price increase significantly and negatively impacts the industrial production (GDP). Usama (2010) showed that oil prices affect positively GDP, but caused krone to depreciate. Mignon (2008) also showed that relationship between oil price and unemployment is significant, especially in the oil-exporting countries. The study of Soytas (2010) has indicated that both oil prices and interest rates affect unemployment, but the effect is insignificant through using impulse response functions. Moreover, Soytas (2010) has shown that oil price shock has positive impact on interest rates in Turkey. On the other hand, Malhotra (2015) showed that oil prices significantly affect inflation, but has no direct effect on interest rates in the immediate period (in India). Zhou (2017) examined the effect of oil prices on exports. She found out that oil prices affect exports significantly. In Canada and other oil-exporting countries, the rise in oil prices lead to decline in exports. Moreover, Al-mulali (2010) found out that in Norway, the oil price increase causes GDP to increase, but exchange rate to decrease, which in turn will cause exports to increase. j. Criticism of the studies In general, many authors focus on few or many countries rather than analyzing a specific country. For the most part, we will primarily focus on Norway and analyze how its economy is related to oil prices. We will compare Norway with other countries as well; however, that will not be the primary aim of this thesis. Moreover, many studies have not shown the impact of oil prices on export and imports specifically. Very few studies have shown that relationship. Moreover, many studies do not use structural break tests to see whether unexpected changes occurred in the time series. Many authors focused on analyzing effects of oil price on few macroeconomic variables. Some author focused entirely on impacts on exchange rate, but this thesis analyzes the effects of oil prices on six macroeconomic variables. We want to see the overall reaction of common macroeconomic variables. 16

4. Methodology a. Introduction to SVAR model SVAR analysis is very useful tool in observing unexpected shocks. The model has evolved into one of the most used models in empirical research using time series data. The SVAR has been used in macroeconomics ; thus, the model is very useful in analyzing effects of oil prices on macroeconomic indicators (Kilian, 2016). More importantly, SVAR s can address following question: 1) How does the economy respond to different type of shocks? (Bjornland [no date]) 2) What is the contribution of different shocks to the business cycle? (Bjornland [no date]) Identifying SVAR is based on non-data information. We need to set up an economic theory or beliefs about the speed at which shocks in one sector are transmitted to the rest of the economy (Gordon, 2001). We need to identify which variables react fastest to various types of shocks. Moreover, traditional approach has been to include more variables, so that the equations can be separately expressed and identified (Bjornland) For the SVAR model, simultaneous equations will be the tool to analyze the effects of oil price on macroeconomic indicators. The simultaneous equations will allow us to estimate the events of oil price shocks and the economy. We will carefully observe the effects of the oil price increase as well as oil price decrease on the Norwegian economy. We expect that higher oil prices have significant effects on economic performance of Norway. The study uses six macroeconomic variables including oil Brent price (dependent variable), unemployment rate, total industrial production index, export per capita (in US dollars), import per capita (in US dollars), average exchange rate (at ask rate) and key policy interest rate (set by Norges Bank). Data from 1990 until 2016 will be used for Norway. The methodology section will only focus on two variables only, oil price and industrial production index. b. Advantages of SVAR (Structural Vector Autoregression) model SVAR is an extension on the traditional VAR because it combines economic theory with time-series analysis to analyze and determine response of economic variables to various disturbances. One of the main advantages of using SVAR model is that the necessary restrictions on the estimated reduced model, required for identification of the underlying structural model, can be provided by economic theory (Mccoy, 1997). When estimating SVAR, we can put some 17

restrictions on some of the effects so that it is in line with our economic theory. We create a theory by putting the restrictions, so to speak. Once we achieve a certain identification, we can recover the structural shocks. The shocks can be then used to generate impulse response functions to analyze specific impacts on different economic variables. Another advantage is that SVAR analysis can be used for variety of research topics, for example monetary policy effects or impacts of exchange rate movements. Third advantage is that structural VAR models allow the construction of forecast scenarios of forecast scenarios conditional on hypothetical sequences of future structural shocks (Kilian, 2016). SVAR s ability is to also look at forecasting. For the SVAR, the first step would be to estimate the data. By estimating the data we will see what the results are. The restrictions will be imposed on the relations among the variables. The aim is to see how the factors (brent oil price) affect the Norwegian economy. We will show the structure of the SVAR model by explaining its identification and how to interpret it. c. The model identification in simultaneous equations When characterizing the structural model, we need to show an identification problem, which represents the structure of an economy. (1) Γ" # = %& # + ( #, where " # is a (n x 1) vector of endogenous variables, & # represents the exogenous and lagged endogenous variables and + =,(((. ) gives the variance-covariance matrix of structural innovations (Gottschalk, 2001). Moreover, the coefficients in Γ and % are the parameters of interest. The main problem with the (1) equation is that we cannot directly estimate it. We cannot obtain the true values of Γ and %. The (1) equation is not enough, since we need to further identify the restrictions. The infinite number of different values for Γ and % makes it difficult for us to identify the true values for Γ and %. As a result, the parameters are called unidentified (Gottschalk, 2001). In order to solve the identification problem, we will show how a dynamic simultaneous equation model is identified. The simple bivariate model consists of oil prices (0 # ) and a monetary policy instrument, total industrial production index (1 # ). The first variable represents the oil market and the second variable represents the economy. We expect oil 18

prices to influence industrial production index, but not the other way around; thus, we identify a certain restriction. The structural model has the form (2) 1 # = 2 3 0 # + % 44 5 1 # + % 46 5 0 # + ( 78 (3) 0 # = 2 9 1 # + % 64 5 1 # + % 66 5 0 # + ( :;8, where %(5) denotes polynomials in the lag operator 5 and Σ + is the variance-covariance matrix of the structural disturbances (Gottschalk, 2001). The (2) equation shows the impact of oil prices on industrial production index. The (3) equation can be interpreted as how industrial production index affects the oil prices. We need to identify the restriction, in which we will obtain estimates of the structural parameters of interest. The disturbance ( :;8 represents the oil price shock. Again, we don t want the industrial production index to affect the oil prices. The reduced form of (2) and (3) is shown below (4) 1 # = % 44 5 1 # + % 46 5 0 # + > 78 (5) 0 # = % 64 5 1 # + % 66 5 1 # + > :;8, where % = Γ?3 % and > = Γ?3 (. Assuming a uniform lag length of @ it is noticeable that the reduced form shown by (4) and (5) has 4@ coefficients while the structural model shown by (2) and (3) has (4@ + 2) coefficients; thus, one identifying restriction for each equation is needed to obtain estimates of the structural parameters from the data (Gottschalk, 2001). Regarding restriction on Γ, we don t want the industrial production index to affect the oil prices. The oil prices are treated as exogenous. The aim of our thesis is to see the effects of oil prices. By solving the problem, we impose restriction on 2 9 to zero. Moreover, we might argue that the response of industrial production index to oil prices might not be immediate (lags). In the thesis, we assume that the industrial production index does not react as fast to oil prices. According to this argument, the parameter 2 3 could be set to zero. With these two restrictions, the matrix Γ becomes the identity matrix and the reduced model of (4) and (5) represent a structural model of the economy. We will not put restrictions on the simultaneous relationships between the variables further, but instead use this issue in the SVAR analysis. All the equations above showed the traditional approach to identification. They serve as useful background for the SVAR methodology. 19

d. Impulse response functions In order to set up the impulse response functions, we need to have a reduced form of structural model (Γ" # = % 5 " # + ( # ). The matrix notation is (6) " # = % 5 " # + > #, where, % = Γ?3 %, and > # = Γ?3 ( #. The variance-covariance matrix can be transformed into reduced form as Σ C = Γ?3 Σ + Γ?3. The next step is to compute the moving average representation of (6). By doing that, we need to reparametrize the system to show the endogenous variables in " #, as function of current and past reduced form shocks, > # (Gottschalk, 2001). We obtain the moving average by reshuffling (7), which leads to (7) " # = E 5 > #, with E 5 = (F % 5 )?3. The MA representation (7) shows the industrial production function index expressed as a function of current and past innovations in > 7 and > :; (Gottschalk, 2001). On the other hand, the AR (autocorrelation) representation in (6) shows the industrial production index as a function of past values of industrial production index and oil price. The clear representation of MA representation is (8) 1 # 0 # = > 7,# > :;,# + E 77,3 E 7:;,3 E :;7,3 E :;:;,3 > 7,#?3 > :;,#?3 + E 77,9 E 7:;,9 E :;7,9 E :;:;,9 > 7,#?9 > :;,#?9 +.. E 77,9 represents the response of industrial production index in period t+2 to a unit innovation in the disturbance term > 7 happening in period t. The plot of E 77,J as a function of s gives the response of industrial production index in time to a unit innovation in > 7,#. As a result, 20

impulse response function is formed as function of industrial production index to a unit innovation > 7,#. e. The model SVAR orthogonal identification The main assumption in SVAR is that these models is that their structural innovations are orthogonal. The error terms ( 7 and ( :; are uncorrelated. We put restrictions in the variancecovariance matrix by restricting the covariance to zero. The matrix can be shown as (9) Σ + = K 7 9 0 9. 0 K :; From the (9) equation, we have one non-linear restriction on Γ since Γu = e. The reduced and the structural variance-covariance matrix are connected to each other by ΓΣ C Γ. = Σ +. f. Normalization of SVAR model In order to normalize the SVAR model, we need to set the diagonal elements of Γ to one. The structure of matrix Γ is (10) Γ = 1 2 3 2 9 1. We need to normalize this matrix to show the impulse response functions because the impulse response functions show the response of model to standard deviation shock to the structural innovations (Gottschalk, 2001). The unit innovations are ( 7 and ( :;. As a result, we have a variance-covariance matrix of the structural innovations. The form of the matrix is as follows (11) Σ + = 1 0 0 1. The matrix can also be noted as Σ + = F. The matrix Γ is therefore normalized (ΓΣ C Γ. = Σ + = F) because the structural innovations are related to the reduced form disturbances by Γu = e. 21

g. Restriction to matrix Γ To identify restrictions on Γ, the SVAR model focuses on the relationship Γ> # = ( #. The relationship between the shocks are portrayed as various relationships in the economy in the SVAR models. At first, we impose orthogonality restriction on the basic structural model (12) Γ" # = % 5 " # + ( #, where the variance-covariance matrix is represented as Σ + = F and the vector ( represents the structural shocks. From each side of the equation we subtract the expected value of " # based on the condition that time is P 1,, #?3 " #. The reason the " # becomes > # on the left side of the equation is that the information on " # (in time P 1) is summarized in the term % 5 " # "(Gottschalk, 2001), shown in equation (6). According to that information, we can say that the forecast error " #, #?3 " # is equal to the reduced form error > #. The term % 5 " # vanishes on the right side of the equation since the term represents only variables at time P 1. As a result, the structural innovations ( # remains in the equation. We cannot forecast the term ( #. The new equation is formed, which is (13) Γ> # = ( #. According to (13) equation, we focus only on unexpected changes in " #. We can restrict the relationship between the oil price and industrial production index in matrix Γ. We do not want the industrial production index to affect the oil prices, so we restrict that relationship. We put this restriction to (13) equation, which results in (14) 2 33 2 39 0 2 99 > 4,# > :;,# = ( 7,# ( :;,#. The equation (14) shows us that causality is running from oil prices to industrial production index, but not the other way around. It shows that the industrial production (aggregate demand) shock, which is represented by ( 7,#, leads to a forecast error within periods in the 22

industrial production variable, but not in the oil price variable. The oil industry does not realize that the industrial production shock occurs. When we set up the matrix Γ, we need to impose restrictions on it. We need to restrict some relationships, which either do not make economically sense or is not in line with economic theory. By using SVAR, we can choose our own economic model, which corresponds to our economic intuition. We want the oil prices to be treated as exogenous; thus, we impose any restrictions on variables that would cause the oil price to move. 5.The macroeconomic aspect of the thesis The thesis mainly deals with macroeconomic variables. Behind the behavior of these variables are macroeconomic theories. The main question we must ask is what is the common behavior of these variables, meaning how do they macro economically response various effects. The hypotheses were chosen due to the common macroeconomic behaviors. a. Relationship between oil prices and interest rate In many studies and articles, the increase in oil price causes increase in inflation. For example, the increased price of gasoline in Mexico is causing inflation to increase (Calcuttawala, 2017). Another study indicates that a marked rise in oil prices will contribute to higher inflation level (Pettinger, 2009). Since many studies explain this trend, we will assume in this thesis that higher oil prices lead to higher inflation. This thesis will therefore portray movements of oil price as movements in inflation. Central bank of Norway (Norges Bank) uses monetary policy for mainly targeting inflation levels. The Norges Bank uses key policy interest rate to set the level of interest rate, which affects the economic activity. In 2002, Norges Bank shall set the key interest rate with a view to maintaining low and stable inflation (Gjedrem, 2002). The inflation target of Norges Bank is at 2.5%. Norges Bank sets the interest rate based on relationships between interest rate, the krone exchange rate, output, employment and inflation (Giedrem, 2002). Furthermore, the central bank in Norway was always aiming at setting low interest rate. During 2014, the monetary policy has to a great extent been influenced by weaker growth, resulting from fall in oil prices and lower activity in petroleum activity (Olsen, 2016). To solve the problem of declining of oil prices, the key policy rate has been reduced by a total of one percentage point since 23

December 2014 (Olsen, 2016). As a result, we can assume that when oil prices decrease, the interest rate decreases as well. b. Relationship between oil price and macroeconomic variables One of the basic macroeconomic theories is that when higher interest rate is set, it will cause aggregate demand to shift down (decrease), leading to lower economic growth and higher unemployment. Moreover, higher interest rates tend to increase value of country s currency (How do changes in national interest rates affect currency s value and exchange rate?, 2017). Higher interest rates will cause consumers to deposit money and get a better rate of return (Economics help, [no date]). As a result, the demand for the currency will rise. We therefore assume that a rise in oil price tends to translate into a stronger krone exchange rate (Bernhardsen, 2000). Since stronger exchange rate is the result of higher interest rate, we assume that stronger exchange rate will lead to lower exports and higher imports. c. Relationship between oil prices and unemployment rate Regarding unemployment, higher oil prices tend to decrease unemployment. Since oil prices affect unemployment mainly in the petroleum industry, we will assume that correlation between the oil prices and unemployment in Norway (across all sectors) is insignificant. The thesis does not analyze effect of oil prices on different sectors. The oil price increase affect increases employment in the oil sector, since oil companies would earn higher revenues; thus, increasing wages for labor. In this thesis though, we take account the unemployment in whole Norway, meaning all the other sectors. Service industry is the second largest sector in Norway. In 2014, many jobs in the Norwegian economy were likely linked to the petroleum sector (Stensland, 2015). Norway is a small open economy. The theoretical framework is that the small open economy consists of three sectors. Each produces its own goods, but two of the goods are tradeable and one is non-tradeable (cannot be imported). The firms in the various sectors compete for a given supply of labor, which is mobile between the sectors (Stensland, 2015). Wage level must therefore be same across the sectors. One tradeable sector is an oil sector and another tradeable sector is manufacturing. Increase in the oil price will cause return factors of production used in the oil related sector to be higher (companies extracting oil, but also delivering goods and services to oil companies). For a given wage level, the demand for the labor will rise. In the non-tradeable sector, share of employed will decline. In the manufacturing sector, the employment will fall as well. The oil sector will expand at the expense of the other two (Stensland, 2015). 24