Get Complete Project Material File(s) Now! »

**Empirical Results**

** Unit Root Test Results**

We perform all ADF tests at level by using 5 years data. Here, LBIT, LOMXS30, LNIKKEI225, LMOSCOW, LS_P500, LKOSPI are the logs of bitcoin and OMX Stockholm 30 index, Nikkei 225 index, Moscow Exchange index, Standard & Poor´s 500 indexes and Korea Composite Stock Price index respectively. According to the Augmented Dickey-Fuller test (1979) for testing unit roots, we know that the null hypothesis is existing unit roots in series and the alternative hypothesis is no unit roots in series. By looking at the results, it appears that the p-values for all the included variables in our research are greater than the critical value (5%). So, we cannot reject the null hypothesis and we must, therefore, conclude that all six variables which are growing are non-stationary, meaning that those variables follow a random walk with drift and no time trend. This implies that we need to take the first difference of those variables before they can be run in the regression model.

We perform all ADF tests at first difference by using 5 years data. As we can see from the results, all the p-values are smaller than the critical value (5%), we can reject the null hypothesis which is there exist unit roots and series are non-stationary. Therefore, the results of ADF test at first difference implies that all variables are stationary. Since stationary properties are proved by ADF test, Johansen cointegration test can be performed to test if the variables exhibit long-run relationships in their first differences.

We perform all LM tests by using 5 years data. We use Serial Correlation LM test for diagnostic checking autocorrelation. The null hypothesis is no autocorrelation up to the specified lag for variables, the alternative hypothesis is autocorrelation up to the specified lag for variables. From this results table, all the p-values are higher than the critical value (5%). Based on this situation, we cannot reject the null hypothesis. Therefore, there is no autocorrelation between variables. According to this test results, there is no serial correlation in the error term, we can perform the Johansen cointegration test for the next step.

The time span last 5 years is from 1^{st} January 2013 to 31^{st} December 2017. The second time period last 1 year is in 2017 and the last three months is from 1^{st} October 2017 to 31^{st }December 2017. If the p-values are higher than 0.05 (critical value), the null hypothesis cannot be rejected. Therefore, one variable does not Granger causality another variable need to be accepted.

After checking for autocorrelation, we perform all Johansen Cointegration tests by using 5 years data. The null hypothesis is no cointegration between variables and the alternative hypothesis is existing cointegration between variables. The results are shown on the above table, we can see that the statistic value of both Trace test (14.00) and Maximum eigenvalue test (13.78) are greater than 12.32 and 11.22 (5 percent critical value). We can reject the null hypothesis based on the requirements, so there is existing cointegration between bitcoin and Nikkei 225 index in Japan. This indicates a long-run relationship between these two variables. In addition, Toda Yamamoto model works no matter variables are cointegrated or not, it will not affect the next step.

In Russia, we find the statistic value of Maximum Eigenvalue test is 28.19 which is higher than 11.22 (5 percent critical value). Moreover, the statistic value of the Trace test is 29.02 that higher than 12.32(5 percent critical value) as well. The null hypothesis of no cointegration can be rejected, there is existing cointegration between bitcoin and Moscow Exchange in Russia. It is worth to mention that Toda Yamamoto model works no matter variables are cointegrated or not. Hence, this will not affect our analysis.

In this case, all the statistic values of Trace test (28.83) and Maximum Eigenvalue test (27.27) are higher than 25.87 and 19.39 (5 percent critical value) respectively. Hence, we can reject the null hypothesis of no cointegration between variables. We can conclude that there is cointegration between bitcoin price and KOSPI stock index in South Korea. This implies a long-run relationship for bitcoin and KOSPI stock index. As we mentioned before, Toda Yamamoto model works regardless of variables are cointegrated or not.

Based on the results, we find that statistic value of Maximum Eigenvalue test (12.27) is greater than 11.22 (5 percent the critical value), and the statistic value of Trace test (12.28) is higher than 12.12(5 percent the critical value). So, the null hypothesis can be rejected in the case of Sweden, we can find cointegration between bitcoin and OMXS 30 stock index This implies a long-run relationship for bitcoin and OMXS 30 stock index.

The cointegration results of United States imply that no cointegration can be found between bitcoin price and S&P 500 stock index. Due to both statistic values of Maximum Eigenvalue test (14.27) and Trace test (16.90) is less than 25.87 and 19.39 (5 percent the critical value) respectively. We are failing to reject the null hypothesis of no cointegration, so there is no cointegration between these two variables.

**Granger Causality Results**

*Denotes the unidirectional causality.*

According to the requirements, we find that no Granger causality between NIKKEI 225 stock index and bitcoin price for last 5 years and for last 3 months. The reason is p-values are higher than 0.05, we cannot reject the null hypothesis. We find no Granger Causality relationship between these two variables. However, a unidirectional Granger Causality which is bitcoin price Granger-cause NIKKEI 225 stock index but not vice versa for last 1 year.

*Denotes the unidirectional causality.*

South Korea: The empirical results find no Granger causality between KOSPI stock index and bitcoin price for last 5 years and last 1 year. Nevertheless, the results find unidirectional Granger Causality between two of these variables. That is bitcoin price does Granger Cause

In the case of Sweden, the empirical results find no Granger causality between OMXS 30 stock index and bitcoin price for last 5 years and for last 1 year. However, the results find a unidirectional Granger causality which is OMXS 30 stock index granger cause bitcoin price but not vice versa for last 3 months.

USA: The empirical results find unidirectional Granger Cause for these two variables. The causality direction is bitcoin price Granger Cause S&P 500 stock index but not vice versa for last 5 years, last 1 year and last 3 months. It means that bitcoin price is leading S&P 500 stock index in the United States for the analysis time-period.

**Conclusion and Discussion**

In our thesis, we analyze the causality relationship between bitcoin and 5 different stock market indexes which are Japan, Russia, South Korea, Sweden and the United States. We apply four tests to see any causality. With the Granger Causality test, we can conclude results and you can see the next paragraph.

From last 5 years, the empirical results find no Granger causality between stock index and bitcoin price in Japan, Sweden, Russia and South Korea, except the United States. For last 1 year, the empirical results find no Granger causality between stock index and bitcoin price in Sweden, Russia and South Korea. Moreover, the results show that unidirectional Granger Cause between two variables in Japan and United States. First causality direction is bitcoin price Granger Cause stock price but not vice versa in Japan. However, the second causality direction is stock index Granger Cause bitcoin price but not vice versa in the United States. For last 3 months, the empirical results find no Granger causality between stock index and bitcoin price in Japan and Russia. In addition, the results exhibit unidirectional Granger Cause between two variables in Sweden, South Korea, and the United States. One causality direction is bitcoin price Granger Cause stock index but not vice versa in South Korea and the United States. Another causality direction is stock index Granger Cause bitcoin price but not vice versa in Sweden.

From these results, the most reasonable interpretation is that there is no any causality in the long-term except the United States. Bitcoin is on the currency market in 2009, but we start to hear its name at the beginning of 2017 and become popular instantaneously. Bitcoin has not an original currency like dollar or euro, but the transactions are made by dollar currency this will cause the bitcoin prices have a cause on S&P 500 index. In medium-run, when we looked for the year of 2017, Japan and USA are affected by bitcoin prices for their indexes NIKKEI 225 and S&P 500. With the development of cryptocurrencies, Japan one of the three big markets for trading Bitcoins and there are some speculations about decreasing the usage of cash and starting the using of new cryptocurrencies called J-Coin and Monacoin. These events cause the change of stock market index movement of Japan and actively being affected. In USA stock market continued to be caused by bitcoin volatility prices. Lastly, in the last 3 months of 2017, in the short run, except for Russia each stock market indexes affected the Bitcoin price or being affected. Sweden is the only example that its stock market index affected the Bitcoin prices. Sweden is one of the countries that decreasing its cash usage and want to implement a cashless economy in the country.

Riksbank, Swedish central bank, present the project **“**e-krona**”** in its website that Swedish government wants to use this new digital currency in the economy. The impacts of this currency may create a movement of the economy to cause Bitcoin. prices. Japan also continued to be affected but Japanese government makes some actions against cryptocurrencies usage. They are afraid so dependent on cryptocurrency market that Japanese stock market responds this action negatively. However, they are also affected by bitcoin prices but not statistically significant to its last 1-year index prices. Like Japan, South Korea is also frightened by the popularity of cryptocurrencies especially Bitcoin. Bitcoin prices have a causality on the stock market index of South Korea which is the same situation applies to their economy like the USA.

1. Introduction

1.1. Background

1.2. Problem and Purpose

1.3. Research Question

2. Literature Study

2.1. Cryptocurrencies and Bitcoin

2.2. Stock Exchanges

3. Data and Methodology

3.1. Data Collection

3.2. Augmented Dickey-Fuller Test

3.3. Lagrange Multiplier-Autocorrelation Test

3.4. Johansen Cointegration Test

3.5. Granger Causality Test

4. Empirical Results

4.1. Unit Root Test Results

4.2. Serial Correlation LM Test Result

4.3. Johansen Cointegration Test Results

4.4. Granger Causality Results

5. Conclusion and Discussion

6. References

7. Appendix

GET THE COMPLETE PROJECT

Bitcoin and Stock Market Indexes Causality