The Relationship Between Unemployment and
Oil Price, Oil Price Uncertainty, and Interest
Rates in Small Open Economies
A study on Sweden, Norway, Denmark, and Finland
MASTER THESIS WITHIN: Economics NUMBER OF CREDITS: 30 ECTS PROGRAM OF STUDY: Civilekonom AUTHOR: Emil Sköld
TUTOR: Kristofer Månsson JÖNKÖPING May 2020
Master Thesis in Economics
Title: The Relationship Between Unemployment and Oil Price, Oil Price Uncertainty, and Interest Rates in Small Open Economies: A study on Sweden, Norway, Denmark, and Finland
Author: Emil Sköld
Tutor: Kristofer Månsson Date: 2020-05-18
Key terms: Unemployment, oil price, oil price uncertainty, interest rate, cointegration, causality, small open economies
Abstract
This study examines the relationship between unemployment rates and oil price, oil price uncertainty, and interest rates. This relation is examined by testing for both cointegration and causality between the variables. By employing the Autoregressive Distributed Lag (ARDL) method this study managed to examine the long-run cointegration between unemployment rates oil price, oil price uncertainty, and interest rates. A modification of the ARDL method is the error correction method which was used to find the short-run dynamics and the speed of convergence back to equilibrium after a shock. Fully modified ordinary least squares (FMOLS) regression was then applied to find the optimal estimates of the long-run coefficients for the regressions. The Toda-Yamamoto Granger causality test is used to find the direction of causality between the variables. These tests were conducted on Sweden, Norway, Denmark, and Finland on monthly data from January 2008 to February 2020. A cointegration relationship was found for Sweden, Norway, and Denmark. The long-run coefficients from the FMOLS regression showed that increased oil prices lead to increased unemployment rates for Sweden and Denmark. All countries except Denmark show evidence of causality from oil prices on unemployment indicating a strong relationship between these two variables. Some countries show causality from oil price uncertainty and interest rates on unemployment rates. These results provide important guidance for policymakers on how to design good economic policies.
Acknowledgment
First, I want to thank my tutor Kristofer Månsson for inspiration helpful guidance and inspiration throughout this thesis. I would also like to thank my fellow student, and especially the members of my seminar group for providing me with a lot of ideas and guidance for improvement of this thesis.
Table of Contents
1.Introduction
... 1 1.1 Purpose ... 3 1.2 Outline ... 3 2.Background
... 4 2.1 Previous research... 4 2.2 Theoretical background ... 83.
Data and Methodology
... 103.1 Descriptive statistics ... 10
3.2 Unit root test ... 11
3.3 Autoregressive Distributed Lag bounds test ... 12
3.4 Error Correction Method ... 13
3.5 Fully Modified Ordinary Least Square test ... 14
3.6 Stability test ... 14
3.7 Toda-Yamamoto Granger causality test ... 14
4.
Empirical results
... 154.1 Unit root test results ... 15
4.2 Autoregressive Distributed Lag bounds test results ... 17
4.3 Error Correction Method results ... 18
4.4 Fully Modified Ordinary Least Square test results ... 19
4.5 Stability test results ... 20
4.6 Toda-Yamamoto Granger causality test results ... 21
5.
Discussion
... 226.
Conclusion
... 257.
Limitations and ideas for further research
... 26References
... 27Appendix 1
– ARDL bounds test results ... 31Figures
Figure 1. CUSUM results for Sweden……… 20
Figure 2. CUSUM results for Norway……….... 20
Figure 3. CUSUM results for Denmark……….. 20
Figure 4. CUSUM results for Finland………...………... 20
Tables
Table 1. Summary of previous research. ... 6Table 2. Descriptive statistics ... 11
Table 3. ADF Unit root test results ... 16
Table 4. ARDL bounds test results ... 17
Table 5. Error Correction Method results ... 18
Table 6. FMOLS regression results ... 19
Table 7. Toda-Yamamoto Granger causality test results ... 21
Table 8. ARDL bound test results... 31
1
1. Introduction
This study will examine if there is a relationship between unemployment rates and oil price, oil price uncertainty, and interest rates in Sweden, Norway, Denmark, and Finland. The relationship will be examined by checking for cointegration and causality between the variables. Cointegration indicates that the economic variables share a stochastic trend meaning that they have a relationship in the long run and even if they move away from each other in the short run, the variables tend to move back to the trend in the long-run (Engle & Granger, 1987). An Autoregressive Distributed Lag (ARDL) model will be used to test for cointegration. If a variable has causality on another variable, it means that changes in this variable causes the other variable to change (Granger, 1969). A Toda-Yamamoto Granger causality test is used to examine if one time series helps forecast another time series. The countries that these tests will be conducted on have quite different roles in the oil market. Sweden and Finland are oil importers, Denmark has been a small oil-exporter but has lately turned into a small importer of oil and is therefore considered to have a neutral position in the oil market in this paper and Norway is a huge exporter of oil ("The World Factbook — Central Intelligence Agency", 2020). The reason for using these countries is that they are all small and open economies and represent different roles in the oil market. This study uses Brent crude oil price as the measure of oil price and the crude oil volatility index (OVX) will be used as a measure of the oil price uncertainty. Unemployment rates and interest rates for each country are collected and monthly data are used for all variables.
These relationships are of huge interest since unemployment is one of the major objects of study in macroeconomics and the reasons causing it needs to be assessed carefully. In addition to the negative economic effect of increasing unemployment there is also a negative social effect from increasing unemployment rates. When fewer people have an income, fewer people can spend money in the same way as they used to do which will harm companies which in the end causes even more unemployment. Historically the oil price has been used as an indicator of how the economy is doing and another good measure of how the economy is doing is the unemployment rate. It has been shown that large economies have gone into great recessions due to oil price shocks (Bade & Parker, 2003). However, this paper wants to investigate if it is possible to provide evidence that unemployment rates in smaller economies are also affected by the oil price, oil price uncertainty and interest rates, in both oil-importing and oil-exporting countries. Oil is used in the production of many products but also in the transportation of goods which means that an increase in the oil price will make production and transportation more expensive for net oil-importing countries. According to theories, changes in oil prices, oil price uncertainty, and interest rates will lead to changes in the unemployment rate through some different mechanisms. Pederson and Mork (1982) describes the wealth transfer effect from the
2 change in purchasing power. According to this theory, a rise in oil price will lower the Gross Domestic Product (GDP) growth for oil-importing countries which leads to higher unemployment rates. Efficiency wages theory (Carruth et al., 1998) states that decreased profit margin due to higher production cost, caused by increases in oil prices and interest rates, leads to lower wages which leads to higher unemployment. The real balance effect explained by Pierce et al. (1974), which occurs when oil prices increase, is that money demand increases which leads to higher interest rates if monetary authorities do not manage to meet this increased money demand. Higher interest rates increase the real price of the products and people will buy less which decreases companies profit and causes higher unemployment rates. Real options theory developed by Bernanke (1983) explains that big changes in the oil price leads to uncertainty about the future oil price, which can lead to postponing of investments by the firms. Postponing of investment has a decreasing effect on output and employment which leads to higher unemployment.
A lot of the earlier studies are made on data from the US, which is a huge economy and importer of oil, therefore it will be interesting to see if the results of this paper are in line with these results or if the results are different for smaller economics and also if the countries role in the oil market will affect the results.
Previous research emphasizes that some of the macroeconomic factors that have an impact on unemployment levels are oil prices and interest risk because of their huge impacts on economic activities. (Carruth et al., 1998; Lardic & Mignon, 2008; Doğrul & Soytas, 2010). These three studies found evidence that oil price shocks affect the unemployment rates. Doğrul and Soytas (2010) and Carruth et al. (1998) also found a significant relationship between interest rate and unemployment rate. In addition to this, some studies used a GARCH in mean VAR model to stress that oil price uncertainty has a central role in influencing aggregate output, investment, and unemployment rates in the US. (Elder & Serletis, 2009, Elder & Serletis, 2010; Kocaaslan, 2019). These studies found that uncertainties about the oil price lead to postponing of investment opportunities which lead to increased unemployment rates. Kisswani and Kisswani (2019) used a non-linear ARDL method to investigated interactions between oil prices and unemployment on data from the US and found that increasing oil prices lead to higher unemployment. Karlsson et al. (2018) investigated the relationship between real oil prices, real interest rates, and unemployment rates in Norway. The most interesting result in this study show that an increase in oil prices caused lower unemployment rates. This result is different than the results found in studies made in oil-importing countries, indicating that the countries role in the oil market can affect the way oil price changes affect unemployment.
3
1.1 Purpose
The purpose of this thesis is to examine the relationship between unemployment rates and oil price, oil price uncertainty, and interest rates, by using cointegration techniques. This paper uses an ARDL cointegration model on data for small open economies. Besides, a Toda-Yamamoto causality test will be conducted to see if there exists a causal relationship between the variables and also the direction of the causality to add robustness to the study. This will hopefully help to get interesting results and give information if the countries role as an importer or exporter of oil, influence the effect of oil price, oil price uncertainty, and interest rates on unemployment rates. Lots of studies have been conducted on this interesting and important relationship using a range of different methods, however, almost all of them only investigate the impact of one independent variable, or uses all three of them but then look at the relationship in one country only. Up to my knowledge no previous research examines this relationship using multiple variables on countries having different roles in the oil market. This study will fill this gap by checking for the relationship between oil price, oil price uncertainty, interest rate and unemployment using an ARDL model and a Toda-Yamamoto Granger Causality test on data for both oil-importing and oil-exporting countries to examine if the effects of these variables on unemployment differs due to the country’s role in the oil market.
1.2 Outline
This paper will proceed in the following order. Sector 2 will present related research together with theoretical background, discussing theories and the connection between oil price, oil price uncertainty, interest rates, and unemployment. In section 3 the data that is used in this thesis will be discussed and all methods used will be presented and explained in detail. Section 4 presents the empirical results found in for each country. This is followed by a discussion about the results in section 5. Section 6 will be a conclusion where the most important findings will be summarized. Lastly there will be a discussion about the limitations of this study and ideas for future research in section 7.
4
2. Background
2.1 Previous research
Previous research emphasizes that some of the macroeconomic factors that have an impact on unemployment levels are oil prices and interest rate risk because of their huge impacts on economic activities. (Carruth et al, 1998; Lardic & Mignon, 2008; Doğrul & Soytas, 2010). By using the efficiency-wage theory as a theoretical framework, the effect of real oil prices and interest rates on unemployment rates for the US was explored empirically by Carruth et al., (1998). In addition to this, some studies stress that uncertainties about the oil price fluctuation has a central role in influencing aggregate output, investment, and unemployment. (Elder & Serletis, 2009, Elder & Serletis, 2010). Kocaaslan (2019) found that a positive oil price shock increased unemployment in the US while a negative oil price shock decreased the unemployment.
The relationship between oil price and unemployment rates has been widely discussed and examined in the economic literature. Loungani (1996) suggests that the main reason for increased unemployment rates in the case of an oil price increase is the reallocation of labor between different sectors in US labor markets. If oil price increases for a long period it can force firms to change their production structure which would harm employment. A rise in oil prices decreases the profits of sectors that are oil-intensive which gives these firms incentives to construct new methods to produce which are less dependent on oil. A strong relationship between oil prices and unemployment rates in the US was provided by Hamilton (1983). This study found a significant correlation between oil price shocks and lots of macroeconomic indicators, including unemployment. The results of the study by Hamilton (1983) are supported by Gisser and Goodwin (1986) who also found that oil price in the US economy has a big influence on different macroeconomic indicators, and unemployment was one of them. In a study by Uri (1996) he investigated the effects of crude oil price changes on employment in the United States between 1947 and 1995. He used Granger causality test and found an empirical link between crude oil price changes and employment. By using the efficiency-wage theory as a theoretical framework, the effect of real oil prices and interest rates on unemployment rates for the US was explored empirically by Carruth et al. (1998). In this paper, the authors developed an efficiency-wage model and showed that by only using real oil prices and real interest rates, the simple framework could explain postwar movements in the US unemployment rates. Ewing and Thompson (2007) investigated cyclical movements of crude oil prices and unemployment rates in the US business cycles. The findings suggest that there were cyclical co-movements between crude oil price and unemployment rates, meaning that when the crude oil price increases, the unemployment rates also increased. Lescaroux and
5 Mignon (2008) investigate the relationship between oil prices and many macroeconomic variables including unemployment rates in several countries. A Granger causality test found that the oil price has a significant causing effect on unemployment rates in both oil-exporting and importing countries. In a study made by Karlsson et al. (2018), they investigated the causal link between real oil price, real interest rate, and unemployment rates in Norway. The results show no short-run causal relations but a long-run causal relation was found from interest rates and oil prices on unemployment rates. In addition to this, the impulse response function found that increases in oil prices decreased the unemployment rate after approximately two years. These results are different from the results earlier discussed, in the way oil price increases affect unemployment rates, but since Norway is a huge oil-exporting country, while most of the other countries investigated has been oil importers, the results are in line with the theories.
Research on the relationship between oil price and employment is very limited when it comes to developing countries but some examples are found. Papapetrou (2001) used the multivariate vector-autoregression method to examine the relation between oil price and employment for Greece and found that oil price shocks harmed the employment for Greece. Doğrul and Soytas (2010) testes for causality between real oil prices and unemployment rates using the Toda-Yamamoto Granger causality procedure on data for Turkey and found significant causality from oil price to unemployment rates, meaning that changes in oil price had a significant effect on the unemployment in Turkey. Since both Greece and Turkey were net importers of oil during the time frame of the study ("The World Factbook — Central Intelligence Agency", 2020), these results indicate that increases in oil prices harm employment for net oil-importing countries. Cuestas and Gil-Alana (2018) conducted a study on countries in Central and Eastern Europe using a non- linear ARDL procedure that showed that rising oil prices harmed employment, while a decreasing oil price had a positive effect on employment.
Considering earlier studies investigating oil price uncertainties and its impact on unemployment rates, Kocaaslan (2019) used a GARCH-in-mean VAR model to investigate if oil price uncertainty had any effects on unemployment rates in the US. The findings show that an increase in oil price uncertainty negatively affected employment, leading to higher unemployment rates. Jo (2014) used oil price volatility to investigate if oil price uncertainty affects global economic activity. The finding suggests that an increase in oil price uncertainty led to a decrease in world industrial production, leading to an increase in unemployment rates. The study by Kocaarslan et al. (2020) used a non-linear ARDL method to investigate the relationship between oil price, oil price uncertainty, interest rate, and unemployment rates in the United States. The results from this study show that in the long-run, increases in oil prices lead to higher unemployment rates, increased oil price uncertainty caused the unemployment rates to increase and a reduction in interest rates leads to higher unemployment rates.
6
Table 1. Summary of previous research.
Authors Country Method Findings
Carruth et al. (1998)
United States Granger causality test
Equilibrium unemployment rate depends on the price of energy and the cost of borrowing.
Lardic and Mignon, (2008)
US, Europe and the counties in G7
Ordinary least square and Johansen test
Economic activities have an asymmetric reaction to oil price shocks. Doğrul and Soytas, (2010) Turkey Toda-Yamamoto Granger causality test
The real price of oil and real interest rate affect unemployment.
Elder and Serletis, (2009)
Canada Var and
Multivariate GARCH-in-mean VAR
Uncertainty about oil prices makes firms postpone investment leading to a decline in aggregate output.
Elder and Serletis, (2010)
United States Vector
autoregression and Multivariate GARCH-in-mean VAR
Uncertainty about oil prices makes firms postpone investment leading to a decline in aggregate output.
Kocaaslan (2019)
United States GARCH-in-mean VAR
Uncertainty in oil prices increases the unemployment rate and a positive price shock for oil increases the unemployment rate.
Papapetrou (2001)
Greece Multivariate VAR An increase in oil prices harmed employment.
Cuestas and Gil-Alana, (2018) Central and Eastern Europe Non- linear Autoregressive Distributed Lag
Change in oil prices and the
unemployment rate goes is the same direction.
Loungani (1986)
United States Dispersion index Increasing oil prices have a large long-run impact on unemployment.
Hamilton (1983) United States Granger causality test
Found out that there is a strong relationship between oil price and unemployment.
7 Gisser and
Goodwin (1986)
United States ST. Louis-type equations and multivariate Granger – causality tests
Crude oil prices had a clear impact on several macroeconomic indicators, including unemployment.
Ewing and Thompson, (2007)
United States Hodrick-Prescott filter
Negative relationship between oil prices and unemployment cycles.
Lescaroux and Mignon (2008)
Large set of oil-importing and exporting countries Granger causality tests and Multivariate model
It is found that in the long run, oil prices cause unemployment.
Kim Karlsson et al, (2018)
Norway Wavelet
multi-resolution analysis and Granger causality
An increase in the price of oil leads to a decrease in unemployment rates for Norway.
Uri (1996) United States Granger causality test
Empirical relationship between crude oil price changes and employment. Jo (2014) United States Vector
autoregressive model
An increase in oil price uncertainty leads to higher unemployment rates.
Kocaarslan et al, (2020)
United States Non- linear Autoregressive Distributed Lag
Increased oil price and increased oil price uncertainty and decreased interest rates all lead to increased unemployment rates.
Kisswani and Kisswani (2019)
United States Non- linear Autoregressive Distributed Lag
Increasing oil prices lead to higher unemployment.
8
2.2 Theoretical background
The economy can be affected by the fluctuation of oil prices and interest rates through multiple different mechanisms (Lardic & Mignon, 2008). One of the mechanisms is the wealth transfer effect implying the change in purchasing power, negative or positive depending on the country’s role as oil-importing or oil-exporting, as an effect of a change in the oil price (Pederson & Mork, 1982). Purchasing power is a measure of how many goods and services that can be bought with a unit of the country’s currency (Gärtner, 2016). This mechanism means that a rise in oil prices will reduce consumers demand for products and therefore reduce GDP growth in countries that are oil importers, while it will increase GDP growth in oil-exporting countries. In terms of unemployment rates, this means that an oil-importing country will experience an increase in unemployment rates when there is an increase in oil price, while oil-exporting countries will experience a decrease in unemployment rates if the oil price increases.
Another mechanism is the effect on the supply-side that an increase in oil price and interest rates causes. Increased oil prices indicate a decreased availability of oil, which is a basic input in production and leads to an increase in production costs and decreasing productivity. This causes the growth of production and output to slow down. Declining production growth decreases the real wage growth and leads to increased unemployment. Since wages are nominally sticky downwards, the decreased GDP growth will cause unemployment rates to increase which will in turn lead to a further decrease in GDP growth. The initial decrease in the growth of GDP is followed by a decrease in the productivity of labor. If the fall in labor productivity is bigger than the fall in real wages, firms will lose profit and be forced to let go of more workers, increasing unemployment rates which leads to even more loss in GDP. This can be related to the efficiency wage theory where lower wages lead to lower productivity since the workers are less motivated to work. (Brown & Yücel, 2002). Efficiency wage theory model by Carruth et al., (1998) is a classical economic model that is built on the fact that equilibrium unemployment rate is depending on demand for labor which is changing dependently on the changes of real input prices. An increase in oil prices is leading to higher production expenses and hence decreased profit margins. To adjust for this decrease in profit margins and keep the economy in equilibrium, the wages for labor decrease and hence the price of labor decreases. Since there is an inverse relationship between unemployment and wages (Romer, 2019), the decrease in wages increases unemployment rates. A similar chain of outcomes happens when
9 there is an increase in the interest rate. When there is a change in the interest rates, also called the price of credit, the real input prices fluctuate, which changes the equilibrium unemployment rates. In the same way as increased oil prices, this will lead to higher production costs and consequently lead to lower profit margins. To adjust to equilibrium in the economy, wages decline and hence unemployment rates increases. This theory is used to theoretically relate oil price changes and changes in interest rates with the fluctuations of unemployment rates. The third mechanisms are the real-balance effect stating that a rise in oil prices increases the money demand. When monetary authorities fail to meet this increased money demand, it causes the interest rates to increase, which harms economic activities and causes the unemployment rates to increase. When the interest rate increases, the real price to buy products increases which makes people buy less, harming companies profit and cause higher unemployment rates. (Pierce et al., 1974; Mork, 1994)
The fourth mechanisms for this study are the effect oil price shocks have on the labor market. An increase in oil prices leads to changes in the production structures that firms have to adjust for, which possibly lead to reallocations of labor and capital in different sectors within the firms, having a big negative influence on the long-run unemployment rates in oil-importing countries (Loungani, 1986).
Oil price uncertainty occurs when there are large oil price changes, both large increase and decrease leads to uncertainty about the future oil price which can delay business investment or lead to resource allocation which affects employment. Crude oil price increases work in the same way as a tax on consumption, harming employment and output. Uncertainty about the future oil price leads to postponing investment in capital good which slows down the output growth leading to decreased employment growth since employment growth seems to depend on output growth, which in the end leads to increased unemployment rates. This is connected to a theory concerning the oil price uncertainty effect on unemployment rates which is the real options theory developed by Bernanke (1983). This theory is an extension of the financial options theory and suggests that uncertainties about the oil price, meaning that there is also un uncertainty about the price of goods in the future, will make firms postpone or even abandon their investments since they cannot be certain that the investment will lead to a profit. A firm has three options when it comes to investment projects, expanding, postpone or abandon. High uncertainty concerning the future oil prices causes postponing or abandoning of investments
10 and has a depressing effect on both output and employment, leading to higher unemployment rates (Rernanke, 1983).
3. Data and Methodology
In this study, to examine the relationship between oil price, oil price uncertainty, interest rate, and unemployment rates, monthly data on unemployment rates and oil price uncertainty has been collected from Thomson Reuters DataStream. Brent oil price index will be used as the oil price in this paper and is a commonly used oil price index used in Europe. Monthly data on the for Brent oil spot price is collected from U.S Energy Information Administration ("Europe Brent Spot Price FOB (Dollars per Barrel)", 2020). The crude oil price volatility index (OVX) will be used as a measure for oil price uncertainty, as is done in the study by Kocaarslan et al., (2020). This index measures the monthly volatility expectations in the oil option prices (Kocaarslan et al., 2020). The interest rates have been collected from Thomson Reuters DataStream for Sweden and Finland, from the European central bank for Denmark and Norwegian interest rates are collected from the Norwegian central bank. The three-month yield on a treasury bill have been used for monthly interest rates, this is a common measure of interest rates and used a lot of studies, including the study by Kocaarslan et al. (2020). Since Finland uses the Euro as its main currency, the interest rate for the Euro has been used for Finland. Some of the countries have experienced negative interest during the chosen period, and since it is not possible to take the natural logarithm of a negative number, the interest rate has been modified by adding 1 to the percentual interest rate to get the interest in a convenient form to work with. Presenting the interest in this way shows the final amount you will have after the period of the investment, instead of showing the change in money (Churchill et al., 2015). The data spans from January 2008 to February 2020 and is chosen due to the availability of the data.
3.1 Descriptive statistics
The first step in this study is to check the descriptive statistics to test if the time series are normally distributed. This Jarque-Bera test is a goodness of fit test and compare if the sample data have the kurtosis and skewness matching a normal distribution. The null hypothesis of this test is a mutual hypothesis saying that the skewness and kurtosis are equal to zero, meaning that the series is normally distributed and the alternative hypothesis is that the skewness and kurtosis is not equal to zero meaning that the time series are not normally distributed (Jarque
11 & Bera, 1980). These descriptive statistics show that all times series except unemployment rates for Sweden are not normally distributed. Since the time series of Sweden’s unemployment rates were the only time series where it was not possible to reject the null hypothesis, this could be due to a type 2 error, which means that a false null hypothesis was not rejected. It could also be because unemployment rates for Sweden are indeed normally distributed. The results are shown in Table 2. To decrease this non-normality in the times series, the natural logarithm is taken on every time series.
Table 2. Descriptive statistics
Variable UNE SWE UNE NOR UNE FIN UNE DEN OIL P UNC
Mean 7.405 2.591 8.014 3.586 36.408 Median 7.500 2.600 8.250 3.700 33.575 Maximum 9.300 3.400 9.500 4.600 88.930 Minimum 5.600 1.500 6.100 1.600 15.610 Std. Dev. 0.830 0.382 0.904 0.732 13.460 Skewness -0.062 -0.748 -0.491 -0.869 1.373 Kurtosis 2.229 3.600 2.156 3.400 5.507 Jarque-Bera 3.713 15.814 10.189 19.347 84.101 Probability 0.156 0.000 0.006 0.000 0.000
Variable OIL PRICE INT SWE INT NOR INT FIN INT DEN
Mean 79.228 0.891 1.594 0.598 0.849 Median 74.330 0.520 1.390 0.198 0.275 Maximum 132.720 5.235 5.970 5.250 6.076 Minimum 30.700 -0.630 0.260 -0.485 -0.448 Std. Dev. 26.391 1.458 1.260 1.316 1.595 Skewness 0.179 1.342 2.040 2.189 1.853 Kurtosis 1.721 4.419 7.100 7.360 5.626 Jarque-Bera 10.727 56.050 203.510 232.286 125.533 Probability 0.005 0.000 0.000 0.000 0.000
3.2 Unit root test
Before moving on with the ARDL model, the order of integration has to be investigated, since the ARDL model only works when the time series are integrated of order I(0) or I(1), but not I(2) (Pesaran et al., 2001). To test the order of integration in the time series, the Augmented Dickey-Fuller unit root test was conducted. The test was conducted on the logged variables with an intercept but no trend, using the Akaike Information Criteria (AIC), and the results suggest that it is appropriate to move on to the ARDL long run form and bound test.
The formula for the ADF test is shown in Equation 1.
12
3.3 Autoregressive Distributed Lag bounds test
Moving on to employ the Autoregressive Distributed Lag (ARDL) bounds testing method to test for cointegration suggested by Pesaran et al. (2001) which lately has been a widely spread method to use and it is based on the Unrestricted Error Correction Model. The ARDL model has many advantages compared to other cointegration methods, for example when the sample is relatively small and includes multiple endogenous regressors, this method is still an efficient estimator. Another advantage is that the ARDL method removes problems connected with autocorrelation and excluded variables by estimating short and long-run components simultaneously for the model. The variables do not need to have the same optimal lags. The standard F-statistics used when conducting the bounds test do not have a standard distribution under the null hypothesis, stating no- cointegration relationship between the variables tested, no matter if the variables are integrated of I(0) or I(1) (Pesaran et al., 2001). The basic framework for investigating the long-run effect off oil price, oil price uncertainty and interest rate on unemployment rates is expressed in a linear equation shown in equation 2. (Elian & Kisswani, 2017)
𝑌 = 𝛽0+ 𝛽1𝑋𝑡+ 𝛽2𝑍𝑡 + 𝛽3𝑊𝑡 + ε𝑡 (2)
where Y is the logged unemployment rate, X is the logged oil price, Z is the logged oil price uncertainty, W is the logged interest rate. These letters will be used for the corresponding variable in all equations throughout this paper. εt is the error term.
Ghosh and Kanjilal (2014) state that the ARDL model uses the lags of the endogenous variable and the lagged and contemporaneous values of the explanatory variables, which makes it possible to estimate the long-run equilibrium relationship. The ARDL model uses a modification of equation (2) to test the long-run cointegration relationship between oil price, oil price uncertainty, interest rates, and unemployment rates. This modified equation is shown in equation 3.
𝛥𝑌𝑡 = 𝛽0+ ∑𝑝𝑖=1𝛿1𝑖𝛥𝑌𝑡−𝑖+ ∑𝑞𝑖=1𝛿2𝑖𝛥𝑋𝑡−𝑖+ ∑𝑟𝑖=1𝛿3𝑖𝛥𝑍𝑡−𝑖 + ∑𝑠𝑖=1𝛿4𝑖𝛥𝑊𝑡−𝑖+
𝜃1𝑌𝑡−1+ 𝜃2𝑋𝑡−1+ 𝜃3𝑍𝑡−1+ 𝜃4𝑊𝑡−1+ µ𝑡 (3)
13 where Δ represents the difference operator, p, q, r, and s are the selected lag lengths for the model and µt represents the error term, which is assumed to be serially uncorrelated.
According to Pesaran et al. (2001) ARDL bounds testing begins with estimating equation (3) using ordinary least square to test if there exists a long-run relationship between the variables through conducting an F-test that uses the mutual significance of the coefficients for the lagged variables. The null hypothesis states that there is no cointegration (H0: θ1 = θ2 = θ3 = θ4 = 0)
whereas the alternative hypothesis states that there is cointegration (H1: θ1 ≠ θ2 ≠ θ3 ≠ θ4 ≠ 0).
The F-statistics generates two critical values for each significance level, one lower bound and one upper bound, constructed by Pesaran et al. (2001). The lower bound is used if all variables are I(0) while the upper bound is used if all variables are I(1) or a combination of both. The null hypothesis cannot be rejected if the collected F-statistic is lower than the lower band, but if the collected F-statistic is above the upper bound the null hypothesis is rejected, indicating that there is a long-run cointegration between the variables. If the observed F-statistic lies between the lower and upper bound the test is inconclusive (Pesaran et al., 2001).
3.4 Error Correction Method
If cointegration is found, an error correction model including the long-run estimates are used to estimate the short-run dynamic parameters, the speed of convergence to equilibrium. The error correction model equation is shown in equation 4.
𝛥𝑌𝑡= 𝛽0+ ∑𝑝𝑖=1𝛿1𝑖𝛥𝑌𝑡−𝑖+ ∑𝑞𝑖=1𝛿2𝑖𝛥𝑋𝑡−𝑖+ ∑𝑟𝑖=1𝛿3𝑖𝛥𝑍𝑡−𝑖 + ∑𝑠𝑖=1𝛿4𝑖𝛥𝑊𝑡−𝑖+
⍴𝐸𝐶𝑀𝑡−1 + µ𝑡 (4) where δ1, δ2, δ3, and δ4 represents short-run dynamic coefficients of the model’s convergence
against the equilibrium, parameter ⍴ measures how fast the model converges to the equilibrium level after experiencing a shock, this value should have a coefficient between negative one and zero and should be statistically significant. Δ represents the first difference while ECMt-1
represents the error correction term derived from the calculated equilibrium relationship in equation 3. (Shahbaz et al., 2013)
14
3.5 Fully Modified Ordinary Least Square test
Ordinary Least Square estimations for the coefficients cannot be used when the series are cointegrated since this generates estimates that contain non-Gaussian asymptotic distribution and therefore they present asymptotic biasedness (Elian & Kisswani, 2017). To overcome this problem with biasedness, Fully Modified Ordinary Least Square (FMOLS) is applied to empirically estimate the long-run relationship between unemployment rates and oil price, oil price uncertainty, and interest rates. FMOLS regression was designed to find optimal estimations of regressions with cointegration and was originally designed by Phillips and Hansen (1990). FMOLS can be used by modifying the least squares to take the serial correlation effects into account and account for endogeneity in the regressors that occurs when a cointegrating relationship exists. This model can study the behavior of FMOLS in models with I(0), I(1) or a mix of them. This method is applied to equation (2) to estimate the long-run coefficients and see how the independent variables affect unemployment rates. (Phillips & Hansen, 1990)
3.6 Stability test
To test if the model has stability, the cumulative sum of recursive residuals (CUSUM) stability test, originally propounded by Brown et al. (1975) are applied. Recursive CUSUM is applied to check if any autoregressive structure is hidden in the model. The parameters of the model are considered stable and consistent if the line is constantly inside the 5% critical bound (Brown et al., 1975).
3.7 Toda-Yamamoto Granger causality test
The null hypothesis on no cointegration in the ARDL bounds test could not be rejected for all countries, therefor this paper continues by checking for causality between unemployment and all independent variables for all countries. Causality and the direction of the causality will be tested between the dependent variable unemployment rates and the independent variables oil price, oil price uncertainty and interest rates by using the Toda-Yamamoto (1995) Granger causality test. This test examines causality by ignoring all possible non-stationary and cointegration between the time series. The Toda-Yamamoto technique works in three steps which are clarified in an article by Kisswani (2015). Step one is to construct the maximum order of integration for each series (dmax). The next step is to estimate a VAR model in levels
15 and determine the optimal lag order (m). This is done by open the time series as a VAR model and allow a maximum of 12 lags in the lag structure. A table shows the optimal number of lags to include for different criteria, this paper used AIC to choose the optimal lag. The last step is re-estimating the VAR model, this time including dmax lags so that the VAR order becomes m
+ dmax = k. When applying the Toda-Yamamoto Granger causality test between unemployment
rates and the oil price, the following two VAR specification equations were used:
𝑌𝑡 = Υ + ∑𝑚𝑖=1𝛿1𝑖𝑌𝑡−𝑖+ ∑𝑘𝑖=𝑚+1𝛿2𝑖𝑌𝑡−𝑖+ ∑𝑚𝑖=1𝜃1𝑖𝑋𝑡−𝑖 + ∑𝑘𝑖=𝑚+1𝜃2𝑖𝑋𝑡−𝑖+ 𝜀1𝑡 (5)
𝑋𝑡 = τ + ∑𝑚𝑖=1𝜐1𝑖𝑋𝑡−𝑖+ ∑𝑘𝑖=𝑚+1𝜐2𝑖𝑋𝑡−𝑖 + ∑𝑚𝑖=1𝜂1𝑖𝑌𝑡−𝑖 + ∑𝑘𝑖=𝑚+1𝜂2𝑖𝑌𝑡−𝑖+ 𝜀2𝑡 (6) The same equation is used to test for causality between unemployment rates and oil price
uncertainty and between unemployment rates and interest rates, changing X to Z and W respectively. Equation (5) has a null hypothesis of non-causality from the independent variable to unemployment rates while the alternative hypothesis states that there is causality from the independent variable to unemployment rates. Equation (6) has the null hypothesis that there is no causality from unemployment rates to the independent variable investigated and the alternative hypothesis states that there is causality from unemployment rates to the independent variable.
4. Empirical results
4.1 Unit root test results
Before conducting the ARDL bounds test, it is vital to examine the stationary conditions for the variables to find their order of integration. Since the computed F-statistics in ARDL only are valid when the variables are integrated of I(0) or I(1) it is important to assure that no variables are integrated of I(2). Therefore, the first thing to do is the unit root test to make sure that all the variables are integrated of I(0) or I(1) and hence satisfy the basic assumptions for the ARDL bounds test of cointegration. Augmented Dickey-Fuller (ADF) unit root test is used, applying the Akaike Information Criterion (AIC).
The null hypothesis for this Augmented Dickey-Fuller unit root test is that there is a unit root, meaning that the variable is non-stationary whereas the alternative hypothesis is that the variable has no unit root and hence are stationary. The results for the unit root test, according
16 to AIC optimal lag length, are shown in table 3. The results show that for the dependent variables, only Sweden´s logged unemployment rate is stationary at the most used 5% significance level. Logged interest rates for all countries and logged oil price uncertainty are integrated of order I(0) at a 5% significance level, indicating that they are stationary and deviations are mean-reverting. For the rest of the variable, the results indicate that they are not stationary at levels, but get stationary at their first difference at the 1% significance level, indicating that deviations are not mean-reverting. The results show that no variable is integrated of order two which indicates that it is possible to use the ARDL bounds test to examine the cointegration between unemployment rates, oil price, oil price uncertainty, and interest rates.
Table 3. ADF Unit root test results
Variables ADF Statistics p-values
LNUNEMP SWE 0.0449** LNUNEMP DEN 0.2197 LNUNEMP FIN 0.0995* LNUNEMP NOR 0.5194 LNINT FIN 0.0001*** LNINT SWE 0.0090*** LNINT DEN 0.0341** LNINT NOR 0.0000*** LNOILPRICE 0.1798 LNOILPRICE UNC 0.0476** DLNUNEMP SWE 0.1716 DLNUNEMP DEN 0.0000*** DLNUNEMP FIN 0.0020*** DLNUNEMP NOR 0.0068*** DLNINT FIN 0.0000*** DLNINT SWE 0.0012*** DLNINT DEN 0.0000*** DLNINT NOR 0.0011*** DLNOILPRICE 0.0000*** DLNOILPRICE UNC 0.0000***
The table shows the results from the ADF unit root test. The null hypothesis is that the variable has a unit root and hence are non-stationary where the alternative hypothesis is that the variable does not has a unit root and hence is stationary. The numbers shown represent the p-value. * is significant at the 10% level, ** means significant at 5% level and *** is significant at the 1% level. LN is in the natural logarithm and D stands for the first difference.
17
4.2 Autoregressive Distributed Lag bounds test results
None of the variables are integrated of order 2, therefor the paper continues with the ARDL bounds test to check for a cointegration relationship between the variables based on Equation 2. The orders for selecting ARDL (p, q, r, s) model was done by choosing a maximum lag of 12 since the time series consists of monthly data, for a value whereby EViews then choose the best fitting model for each country, where p, q, r and s represent the optimal lag length for unemployment rates, oil price, oil price uncertainty and interest rates respectively.
The F-statistics for cointegration analysis for every country, based on the selected ARDL models are shown in Table 4. The null hypothesis for this ARDL long run form and bounds test is that there is no cointegration between the variables, while the alternative hypothesis states that there is cointegration. The F-statistics for Norway and Denmark lie above the upper bond critical F-value for 1% significance and the F-statistics lies above the critical upper bound F-value for the 5% significance level for Sweden, therefor we can with a 5% significance level say that cointegration is established in these countries. The F-statistics for Finland however lies between the lower and upper bound so the test results for Finland are inconclusive and the null hypothesis of no cointegration cannot be rejected. The complete results for the ARDL bound test, including all coefficients and p-values for each country can be found in Appendix 1 and the critical F-values for the same test can be found in Appendix 2.
Table 4. ARDL bounds test results
Country Sweden Norway Denmark Finland
Selected ARDL model
ARDL(12,2,11,10) ARDL (12,8,1,12) ARDL(2,2,0,11) ARDL(12,6,9,12) Computed
F-statistic
5.783*** 6.731*** 9.899*** 3.331
The null hypothesis is that there is no level relationship and the alternative hypothesis states that there is cointegration. *** represents a 1% significance level. The numbers in parentheses stand for the optimal lag chosen for the unemployment rates, oil price, oil price uncertainty, and interest rates.
18
4.3 Error Correction Method results
The error correction method is used to estimate the short-run dynamic parameters. The error correction coefficient is negative and highly significant for all countries except for Finland, where the coefficient is positive. This is in line with the previous results since cointegration could not be established in Finland either. A positive coefficient implies that the model does not converge in the long run, instead it is moving away from the equilibrium every period. The negative coefficient indicates how fast the model returns to its equilibrium level after experienced a shock causing it to move away from the equilibrium. The results from the error correction model are listed in Table 5.
Table 5. Error Correction Method results
Country Variable Coefficient SE t-statistic Prob
Sweden Error correction -0.193 0.040 -4.885 0.000***
Norway Error correction -0.055 0.010 -5.268 0.000***
Denmark Error correction -0.076 0.012 -6.373 0.000*** Finland Error correction 0.006 0.002 3.710 0.000***
It can be seen from the results that there is a varied speed of adjustment to equilibrium for the different countries. Any deviation from the long-run equilibrium in Sweden is corrected and adjusted approximately 19% percent each period takes around 5 periods to get back to the long-run equilibrium level. For Norway, which has the slowest adjustment to equilibrium of the cointegrated countries, it takes about 18 periods to move back to the equilibrium level after a deviation from equilibrium. The speed of convergence back to equilibrium for Denmark is 7.6 percent each period while the values for Finland move away from the long-run equilibrium with 0.6 percent each period.
19
4.4 Fully Modified Ordinary Least Square test results
The FMOLS regression was conducted and the results are shown in Table 6 show the elasticity of the independent variables with respect to unemployment rates. This is the long-run coefficients and shows how unemployment rates are affected by changes in one of the independent variables, holding everything else equal.
Table 6. FMOLS regression results
Variable Sweden Norway Denmark Finland
LNOILPRICE 0.393 (0.000***) 0.0573 (0.536) 0.458 (0.000***) -0.091 (0.270) LNOILUNC 0.205 (0.013**) 0.069 (0.437) 0.174 (0.164) -0.072 (0.383) LNINTR -6.567 (0.005***) -9.855 (0.000***) -13.588 (0.000***) -2.226 (0.304) C -0.365 (0.596) 0.604 (0.343) -1.241 (0.137) 2.740 (0.000***)
Observations 145 145 145 145
The number before the parentheses is the coefficient while the number in parenthesis is the p-value. ** is significant at a 5% level and *** is significant at a 1% level.
Interpreting these long-run estimates effects on the dependent variable unemployment rates, having the values in the natural logarithm form, saying that a 1% change in one of the independent variables, will cause a percentual change on unemployment rates equal to the coefficient of this independent variable, holding everything else equal. For example, looking at the results for Sweden, a 1% change in the oil price, would cause the unemployment rate to increase with 0.393%. For Norway there is a positive coefficient for oil price indicating that there would be a positive relationship between oil price and unemployment rates in Norway but it is highly insignificant. Increases in interest rates will lead to lower unemployment rates for all countries and is highly significant for all countries except Finland. The results will be analyzed and discussed in depth in the discussion section.
20
4.5 Stability test results
To test if the model has stability, the cumulative sum of recursive residuals (CUSUM) stability test, originally propounded by Brown et al. (1975) are applied. The results for each country are shown graphically in Figures 1 to 4. The parameters of the model are considered stable and consistent if the line is constantly inside the 5% critical bound. As can be seen in the results presented, the CUSUM for Sweden, Denmark, and Finland are inside the boundaries, meaning that the model is stable and consistent for these countries. The CUSUM results for Norway shows that it fluctuates outside the boundaries for a short time and cannot be considered stable.
Figure 1. CUSUM results for Sweden Figure 2. CUSUM results for Norway
Figure 3. CUSUM results for Denmark Figure 4. CUSUM results for Finland
21
4.6 Toda-Yamamoto Granger causality test results
The next step is to conduct a Toda-Yamamoto Granger causality test to examine if there is any causality between the variables and also the direction of causality will be examined. For all countries except Denmark there is a causal effect from oil price on unemployment rates. As can be read from the results, oil price uncertainty cause changes in the unemployment rate in Sweden and Finland, but has no significant effect in Norway and Denmark. The casual effect of interest rates on unemployment are highly significant in most of the countries and a causing effect from unemployment on interest rates is highly significant for Norway and Denmark. There is no causal effect from unemployment on oil price and oil price uncertainty for any country. All these results are shown in Table 7.
Table 7. Toda-Yamamoto Granger causality test results
Countries Sweden Norway Denmark Finland
Cause → Effect P-value P-value P-value P-value
LNOILPRICE→ LNUNE 0.082* 0.024** 0.277 0.049** LNUNE→LNOILPRICE 0.485 0.494 0.654 0.509 LNOILUNC→ LNUNE 0.012** 0.402 0.221 0.076* LNUNE →LNOILUNC 0.970 0.705 0.871 0.860 LNINTR→ LNUNE 0.069* 0.005*** 0.187 0.000*** LNUNE→ LNINTR 0.492 0.000*** 0.000*** 0.134
22
5. Discussion
The first part of this discussion will discuss the result from each test for the countries individually. Thereafter a discussion about some of the different results for the countries will be discussed simultaneously. Starting with Sweden, the ARDL bounds test of cointegration show that there is cointegration between oil price, oil price uncertainty, interest rates, and unemployment rates. The result from the Error Correction model gives robustness to this result since the error correction coefficient was negative and significant, indicating that the values converge back to equilibrium after a shock. Looking at the long-run coefficient from the FMOLS regression shows that significant results are indicating that an increase in oil prices has an increasing effect on unemployment rates, which is in line with the results found by Uri (1996). Interest rates have a negative and significant coefficient, meaning that increases in interest rates have a decreasing effect on unemployment rates. The results for Sweden are the same as the results in studies made in the US by Ewing and Thompson (2007) and by Koocarslan et al. (2020), which is also an oil-importing country. The results from the Toda-Yamamoto Granger causality test indicates that all independent variable has a causing effect on unemployment in Sweden which strengthen the results on cointegration. These results are in line with stated theories that oil price, oil price uncertainty, and interest rates have a big impact on unemployment.
Cointegration was also found between oil price, oil price uncertainty, interest rates, and unemployment rates in Norway. The only statistically significant coefficient found from the FMOLS regression was for interest rates, which was negative and very high, indicating that increases in interest rates will have a large decreasing effect on unemployment rates, everything else equal. Norway is the only big oil-exporting country in this study and according to the theory by Pederson and Mork, (1982) about oil prices and unemployment rates, they state that increased oil prices should lead to lower unemployment rates in oil-exporting countries. The FMOLS coefficient for oil prices is positive indicating that increases in oil prices will lead to higher unemployment rates, which contradicts this theory, but since it is not statistically significant no conclusion can be drawn from this coefficient. The results from the study by Karlsson et al, (2018) on Norway, considering oil price and unemployment, found that increases in oil price lead to a decrease in unemployment rates only in the long-run and after as much as two years. Looking at the results from the Toda-Yamamoto Granger causality test, there is a very significant causality from oil price to unemployment rates, strengthening the
23 original thought of this paper that oil prices affect unemployment rates. Another interesting result is that interest rates and unemployment rates have a causing effect on each other. This is another reason why it is interesting and important to investigate the relation between these variables because it indicates that when Norway experience changes in its unemployment rates, they can change their interest rate to get their unemployment rates to move back to its equilibrium level. To do this it is important to know if they should increase or decrease the interest rate to push the unemployment rate in the right direction.
The results for Denmark indicate that there is Cointegration between oil price, oil price uncertainty, interest rates, and unemployment rates. The results of the FMOLS regression show that increases in oil prices have an increasing effect on unemployment rates. Oil price uncertainty had no significant effect on unemployment rates while interest rates have a significant negative relationship with unemployment rates. The Toda-Yamamoto Granger causality test shows that the only causal effect that could be found for Denmark was from unemployment rates on interest rates. These results contradict with the significant positive oil price coefficient found in FMOLS regression. Since Denmark is considered neutral in the oil market, it is logical to think that there should not be a causal relationship between oil price and unemployment rates. Since the results from FMOLS and Toda-Yamamoto Granger causality test contradict with each other, no clear conclusion can be drawn considering the effect of oil price changes on unemployment. The test further shows that there is a causing effect from unemployment rates on interest rates, meaning that changes in unemployment rates causes changes in the interest rates.
Finland is the only country where cointegration could not be found since the F-statistics for the ARDL bound test lie between and critical bounds and therefore the test was inconsistent. This result was confirmed by the error correction coefficient which indicates that the variables move away from each other after a shock that forced them away from the equilibrium level. The results from the Toda-Yamamoto Granger causality test indicates that oil price, oil price uncertainty, and interest rates have a causing effect on unemployment rates. Even if a cointegration relationship could be established for Finland, a causal relationship from the oil price, oil price uncertainty, and interest rates on unemployment rates was found, which indicates that these variables do have an impact on unemployment indicating that there is some kind of relationship between these variables.
24 The overall results of this study confirm a lot of the results in the previous research and is in line with stated theories. A strong relationship between oil price and unemployment was found, where changes in oil price caused changes in unemployment rates in most of the countries, same results as was found by Hamilton (1983) and Loungani (1986). The results from the FMOLS regression found negative coefficient for interest rates, indicating that increases in interest rates should lead to lower unemployment rates for all countries, which are contradicting with the theory discussed by Pierce et al. (1974), saying that increasing interest rates should lead to higher unemployment. These results are however in line with results from some previous research, for example, the study by Kocaarslan et al, (2020), which also found that increases in interest rates lead to lower unemployment rates. Therefore, these results need to be analyzed carefully. One reason for these contradicting results could be to the fact that there have been extremely low interest rates, even negative during some periods, which might have influenced the effect of changes in interest rates. The results from the Toda-Yamamoto Granger causality test show that for all countries, the unemployment rates had no causal effect on neither oil prices or oil price uncertainty. These results are very logical since it would be strange if the unemployment rates in small open economies would affect international oil prices. The same test shows a causing effect from oil price uncertainty on unemployment rates for Sweden and Finland indicating that for oil-importing countries, uncertainties about the future oil price have an impact on the unemployment rates. This can be connected to the real option theory by Bernanke (1983) stating that uncertainties concerning future oil prices will lead to postponing or abandoning investments, leading to higher unemployment. This is strengthened by results from the FMOLS regression where the positive coefficient for oil price uncertainty in Sweden is significant and indicates that increased oil price uncertainty will lead to higher unemployment rates. This relationship between oil price uncertainty and unemployment rates only found with statistical significance for one country in this study but the same relationship was found in the US by Jo (2014).
25
6. Conclusion
This study focused on analyzing the relationship between unemployment rates and oil price, oil price uncertainties, and interest rates for Sweden, Norway, Denmark, and Finland from January 2008 to February 2020. After conducting the Augmented Dickey-Fuller unit root test, results show that the variables had a mix of integration I(0) and I(1) which suggest that it was appropriate to move on with the Autoregressive Distributed Lag (ARDL) bounds test of cointegration. By conducting the ARDL bounds test of cointegration, this paper tests if there is a long-run cointegration between unemployment rates, oil price, oil price uncertainty, and interest rates. Error correction model was applied to estimate the short-run dynamics to find the speed of adjustment back to the equilibrium level after experiencing a shock. Fully Modified Ordinary Least Squares (FMOLS) regression was applied to find estimations of the long-run coefficients. The stability of the ARDL bound test model was tested by conducting a cumulative sum of the recursive residuals (CUSUM) test. Additionally, a Toda-Yamamoto Granger causality test was conducted to increase robustness and examine the direction of causality between the variables. This paper manages to find an indication of cointegration between unemployment rates, oil price, oil price uncertainty and interest rates for three of the four countries examined. The results indicate that oil price increases are associated with increases in unemployment rates for oil-importing countries. However, no differences could be found between oil-importing and oil-exporting countries since the long-run coefficient of oil price on unemployment rates was insignificant for Norway. The error correction coefficient shows that Sweden has the fastest adjustment back to equilibrium only taking around 5 periods to move back to equilibrium. In addition to this, causality was found from oil prices on unemployment rates for most countries, which is in line with what earlier studies have found and also what theories state. Oil price uncertainty and interest rates also show evidence of causal effects on unemployment for some countries. No causality was found from unemployment on oil price and oil price uncertainty, which is in line with what can be expected since this paper is looking at small open economies and unemployment rates of these countries should not affect the oil price. These results show that oil prices, oil price uncertainties, and interest rates are interesting when investigating factors that influence unemployment rates and hence the economy of countries, which makes it an important subject to investigate further.
26
7. Limitations and ideas for further research
When writing this paper, some limitations brought my attention. Starting with the sample period from 2008 can have affected the results of this paper because of the financial crises starting in the same year. The extreme decreases in oil prices and increases in unemployment rates during this period and the time it took to adjust for these movements and move back to equilibrium level might have affected the results of this paper and the relationship found might be incorrect. Another limitation of this paper is the limited use of independent variables, as there are a lot more things that can affect unemployment and would be interesting to include. As discussed earlier, the fact that Denmark has changed its role in the oil-market during this period made this paper to treat it as neutral in the oil market. This might cause some interesting interpretations of oil-exporting countries to be left out.
When it comes to future studies, it would be interesting to include more independent variables to see how they affect unemployment rates, for example GDP. An interesting extension of the study would be to use a non-linear ARDL model to estimate the asymmetric impact on unemployment. This would make it possible to see if positive and negative changes in the independent variables would have different magnitude of impacts on unemployment rates. Further research could also include more countries and a longer period to see the long-run effects and the differences between oil-importing and oil-exporting countries.
.
27
References
Bade, R, & Parkin, M. (2003). Foundations of Macroeconomics. Boston: Pearson.
Bernanke, B., 1983. Irreversibility, Uncertainty, and Cyclical Investment. The Quarterly Journal of Economics, 98(1), p.85.
Brown, R., Durbin, J., & Evans, J. (1975). Techniques for Testing the Constancy of Regression Relationships Over Time. Journal Of The Royal Statistical Society: Series B
(Methodological), 37(2), 149-163. doi: 10.1111/j.2517-6161.1975.tb01532.x
Brown, S., & Yücel, M. (2002). Energy prices and aggregate economic activity: an interpretative survey. The Quarterly Review Of Economics And Finance, 42(2), 193-208. doi: 10.1016/s1062-9769(02)00138-2
Carruth, A., Hooker, M., & Oswald, A. (1998). Unemployment Equilibria and Input Prices: Theory and Evidence from the United States. Review Of Economics And Statistics, 80(4), 621-628. doi: 10.1162/003465398557708
Churchill, G., Brown, R., Chandler, G., & Davis, W. (2015). Compound Interest Simplified. Kent: Elsevier Science.
Cuestas, J., & Gil-Alana, L. (2018). Oil price shocks and unemployment in Central and Eastern Europe. Economic Systems, 42(1), 164-173. doi: 10.1016/j.ecosys.2017.05.005 Doğrul, H., & Soytas, U. (2010). Relationship between oil prices, interest rate, and
unemployment: Evidence from an emerging market. Energy Economics, 32(6), 1523-1528. doi: 10.1016/j.eneco.2010.09.005
ECB Statistical Data Warehouse. (2020). Retrieved 4 Mars 2020, from
http://sdw.ecb.europa.eu/browseTable.do?node=SEARCHRESULTS&q=FM.M .DK.DKK.DS.MM.CIBOR3M.ASKA
Elder, J., & Serletis, A. (2009). Oil price uncertainty in Canada. Energy Economics, 31(6), 852-856. doi: 10.1016/j.eneco.2009.05.014
Elder, J., & Serletis, A. (2010). Oil Price Uncertainty. Journal Of Money, Credit And Banking, 42(6), 1137-1159. doi: 10.1111/j.1538-4616.2010.00323.x
Elian, M., & Kisswani, K. (2017). Oil price changes and stock market returns: cointegration evidence from emerging market. Economic Change And Restructuring, 51(4), 317-337. doi: 10.1007/s10644-016-9199-5
Engle, R., & Granger, C. (1987). Co-Integration and Error Correction: Representation, Estimation, and Testing. Econometrica, 55(2), 251. doi: 10.2307/1913236 Europe Brent Spot Price FOB (Dollars per Barrel). (2020). Retrieved 12 Mars 2020, from
28 Ewing, B., & Thompson, M. (2007). Dynamic cyclical comovements of oil prices with industrial production, consumer prices, unemployment, and stock prices. Energy Policy, 35(11), 5535-5540. doi: 10.1016/j.enpol.2007.05.018
Ghosh S, Kanjilal K (2014) Oil price shocks on Indian Economy; evidence form Toda Yamamoto and Markov regime-switching VAR. Macroecon Finance Emerg Mark Econ 7(1):122–139
Gisser, M., & Goodwin, T. (1986). Crude Oil and the Macroeconomy: Tests of Some Popular Notions: Note. Journal Of Money, Credit And Banking, 18(1), 95. doi: 10.2307/1992323
Granger, C. (1969). Investigating Causal Relations by Econometric Models and Cross-spectral Methods. Econometrica, 37(3), 424. doi: 10.2307/1912791
Gärtner, M. (2016). Macroeconomics (5th ed.). Harlow: Pearson.
Hamilton, J. (1983). Oil and the Macroeconomy since World War II. Journal Of Political Economy, 91(2), 228-248. doi: 10.1086/261140
Jarque, C., & Bera, A. (1980). Efficient tests for normality, homoscedasticity and serial independence of regression residuals. Economics Letters, 6(3), 255-259. doi: 10.1016/0165-1765(80)90024-5
Jo, S. (2014). The Effects of Oil Price Uncertainty on Global Real Economic Activity. Journal Of Money, Credit And Banking, 46(6), 1113-1135. doi: 10.1111/jmcb.12135 Kim Karlsson, H., Li, Y., & Shukur, G. (2018). The Causal Nexus between Oil Prices, Interest
Rates, and Unemployment in Norway Using Wavelet Methods. Sustainability, 10(8), 2792. doi: 10.3390/su10082792
Kisswani, A., & Kisswani, K. (2019). Modeling the employment–oil price nexus: A non-linear cointegration analysis for the U.S. market. The Journal Of International Trade & Economic Development, 1-17. doi: 10.1080/09638199.2019.1608461
Kisswani, K. (2015). Does oil price variability affect ASEAN exchange rates? Evidence from panel cointegration test. Applied Economics, 1-9. doi: 10.1080/00036846.2015.1109040
Kocaarslan, B., Soytas, M., & Soytas, U. (2020). The asymmetric impact of oil prices, interest rates and oil price uncertainty on unemployment in the US. Energy Economics, 86, 104625. doi: 10.1016/j.eneco.2019.104625
Kocaaslan, O, K. (2019). Oil price uncertainty and unemployment. Energy Economics, 81, 577-583. doi: 10.1016/j.eneco.2019.04.021
Lardic, S., & Mignon, V. (2008). Oil prices and economic activity: An asymmetric cointegration approach. Energy Economics, 30(3), 847-855. doi: 10.1016/j.eneco.2006.10.010
