Group 4 Report
Group 4 Project - A Tale of Bitcoin
|
|
|
|
|
|
|
|
Contents
Analytics
1. Time Series
A simple line plot of different daily close price against time shows price changes over a period of 5 years. From first observation (chart below), it is expected that the price volatility from 2016 onwards will be high. Further discussion on this volatility is set out at a later part of this research. From this chart, we can observe massive price movements in 2014 and 2017. The trend is close to an exponential curve see in 2017.
Zooming into the period between 1 September 2017 and 15 September 2017 during which the price of bitcoin dropped from about $4,900 to about $3,200 in a matter of 10 business days, the following is chart is what we observe:
The candle stick chart was helpful to highlight large drops in prices in minute blocks. Further investigation reveals that during this period, various negative news was released. The following is a list of them :
- 14th Sept: Elvira Nabiullina, Governor Bank of Russia: “China doesn’t recognize cryptocurrency as payment and forbids ICOs … Our views are absolutely similar. In our view, it's a sort of a financial pyramid that may collapse at any moment.”
- 12th Sept: Jamie Dimon, CEO of JPMorgan Chase: “It’s a fraud.” If a JPMorgan trader began trading in bitcoin, he said “I'd fire them in a second. For two reasons: It’s against our rules, and they’re stupid. And both are dangerous.”
- 8th Sept: Howard Marks, Co-chairman and co-founder Oaktree Capital Group: “So my initial bottom line is that I see no reason why bitcoin can’t be a currency … But I still don’t feel like putting my money into it, because I consider it a speculative bubble. I'm willing to be proved wrong.”
- 5th Sept: Robert Shiller, Nobel Prize winning professor of economics, Yale University: “The best example [of a speculative bubble] right now is bitcoin”
(Source:Bloomberg, Bitcoin bulls and bears <https://www.bloomberg.com/features/bitcoin-bulls-bears/>
The price seems to be reactive to news released which may be a sign of speculative investors looking for gains in bitcoin investment. Another tell tale sign that investor may be pouring money into bitcoin was the historic low volatility levels in the financial markets . Low volatility in financial markets means lower money making opportunity, thus, making bitcoin a potential alternative.
2. Cyclical Effects
The technique applied to account for seasonality is through the use of Auto Correlation Function (ACF). How this works is as follow:
- the data is replicated and shifted down the time series known as lags;
- correlation is then computed between the lag against the original data points; and
- finally, it is plotted on an ACF graph.
The sample table below shows the closing price and its respective lags in R:
The purpose of performing the lags and calculating the correlation is to identify whether there is any repeating trends in the underlying data as time passes. If trends do change, we will observe a drop in correlation and vice versa.
We will analyse this topic in four parts, being:
- entire data point and observe for seasonality up to lag 90;
- data between 2014 and 2016 with 360 day lag;
- the year 2017; and
- specific months in 2017.
2.1 Entire data point and observe for seasonality up to lag 90
There is strong correlation even up to 90 day lag. This is expected because we have a massive spike in price between 2016 and 2017. The correlation value would have been heavily influenced by the bitcoin prices towards the tail end. Thus, unless we drag the lag into extreme number of days, we are unlikely to see any meaningful trend from the above.
2.2 Data between 2014 and 2016 with 180 day lag
Next, we explore auto-correlation in the period between 2014 and 2016. This period was selected because there was some price activity towards end of 2013 and right before the massive price increase in 2017. We dragged the lag to the extreme of 180 days (half a year).
The strong correlation gradually tapered off towards the 6 months. There is no strong indication of seasonality seen from this chart.
2.3 Year 2017
The above encompasses all data points in the year 2017. No seasonality is observed up to for 90 day lags. This chart is likely skewed due to the 700% increase in bitcoin value from the start of the year. Thus, the upward trend is apparent.
2.4 Specific Months in 2017
The above is charted based on bitcoin price between August 2017 and October 2017. Trends begin to change at 10 day period intervals, which show some form of cyclical effect. This trend is not apparent in other periods of 2017. The fluctuations may be purely coincidental and could be a consequence of a series of market news released during this period which drove the prices up and down at specific intervals.
2.5 Conclusion for this sub-segment
Due to the high volatility in recent prices and a strong tendency upwards, a longer period auto-correlation does not reveal much information apart from what the price chart has already done. However, analysis at specific shorter period, such as the three month period between August 2017 and October 20 17, reveals some form of cyclical effect (10 day cycle).
3. Risk
The panels below contain both the price chart at the top 2 rows and the corresponding standard deviation at the bottom. The red reference line is the average for the year Towards the end of 2013, there was a surge in standard deviation which extended into 2014 but gradually died down. The years 2016 and 2017 represent a whole new paradigm of standard deviations. There were massive spikes and it remained consistent so from middle of 2017 towards October 2017.
Conclusion: The uptrend or downtrend of bitcoin price is one that is accompanied by spurts in standard deviation, i.e. not a smooth growth.
4. Comparative
4.1 Similar Events in the Past?
If the observations and conclusions above ring any bell, one may recall the dot-com bubble that occurred not long ago in the year 2000. The Nasdaq index climbed steeply in the span of one year before collapsing the year after. At the start of 1999, the index was 2,208. It continued to rally into the year 2000 and reached its peak on 10 March 2000 at a record high (at that time) of 5,048.62. The increment was in excess of 200% before the collapse.
“Markets can be something less than efficient in immediately distilling new information and that investors, driven by emotion, can indeed lead markets awry” (https://www.mckinsey.com/business-functions/strategy-and-corporate-finance/our-insights/do-fundamentalsor-emotionsdrive-the-stock-market). This is what happened during the dot-com bubble – emotions rather than technicality and fundamentals drove the index up.
To illustrate the point above, see the three year time horizon on bitcoin’s return and four year time horizon for NASDAQ’s return plotted on a same timeseries below. We observe a similar trend, except that bitcoin price climb is much steeper.
4.2 Comparative performance against long term growth and safe haven products
The following index and commodity have been chosen for the following reasons:
- Gold: safe haven (debateable) investment. For the interest of this research, we have included the same as a comparative; and
- SP500: as a benchmark for long term growth.
Bitcoin return displays trends in both 5 year and 2 year return that is dissimilar compared to both gold and SP500, which means it does not have the characteristics of a long term growth or safe haven investment. We will be expecting some sort of ‘high risk high return’ type scenario for bitcoin. The following return on risk is being plotted:
The impact on bitcoin’s return ratio on risk is muted in the 5 year chart. This is largely contributed by lower volatility for periods to 2016. The 2 year chart clearly shows the impact of return on high risk assets. SP500 (green) showed a stable upward growth. As for gold, the trend moves around and about the long-term growth level. The ups downs of gold are typically reflective of periods when markets are either inclined towards risk taking or risk aversion position. Meanwhile, bitcoin’s return on risk was dragged all the way to the bottom.
4.3 Conclusion for this sub-segment
Bitcoin does not show any characteristics of long term growth or safe haven. Moreover, due to unequal volatility values for the window period that we are analysing, different outcomes are shown for the short term and long term.
5. Forecast
Time series forecasting has been a data science task that is aimed to predict future trend based on the analysis of historical data. It assumes that historical patterns and trends will repeat itself. Through forecasting, we will be able to uncover the following three attributes of time series data: Trend, Periodicity, Randomness. These are the general component found in most forecasting models. The R package that helps us to achieve this is “Prophet”. The model used in prophet is also known as the additive model, where the sum of the different components will result in expected outcome. Here is the formula for Prophet:
The components are explained as follows:
- g(t) is the trend function which models non-periodic changes in the value of the time series;
- s(t) represents periodic changes (e.g. weekly and yearly seasonality);
- h(t) represents the effects of holidays which occur on potentially irregular schedules over one or more days; and
- error term represents any changes that are not accommodated by the model.
Plugging the data into the function yielded the result below, with the red line being the forecast. Based on the 30 day forecast, the value of bitcoin could appreciate close to $8,000.
Conclusion and Future Work
This application could be developed further to include more comparative instruments and even develop an API to update new prices.
We have explored the various aspects of bitcoin’s value from price volatility, cyclical effects, comparative performance to forecasting. Each of these analyses demonstrate the risk that bitcoin poses. In our forecasting section, we projected that value of bitcoin could reach close to $8,000. If our prediction were to be realised, we have every reason to be worried. What makes the value of bitcoin reasonable at that level? Given the mixed reactions from different parties, we know for certain that no one knows the strategic direction and application of bitcoin. Besides, as observed in earlier sections, the price moves wildly whenever negative news is being published – a sign that investors are pouring in money for the possible lucrative gains. The message as shown from the charts above should be clear. Precaution needs to be taken by the different financial regulators to safeguard the financial system. We have seen the pain that was inflicted on the society as a result of the dot-com bubble and more recently the global financial crisis. These events could repeat themselves if we are not careful. And we already know which party will suffer the most.
