Modelling volatility in bitcoin a GARCH approach

[Jack3d]

To all those that are interested i am currently in the middle of my undergraduate dissertation. My topic is modelling the volatility of bitcoin. The outcome of the project will hopefully be some well thought out policy recommendations to decrease the volatility of bitcoin at a faster rate than would naturally occur hence speeding up the adoption of bitcoin and increasing usage. Since my supervising professor does not know a lot about crypto currencies i was hoping for some feedback on here about the project. Do you think the project is viable? what problems do you think I might encounter? I've written a brief background on bitcoin that i will use at the beginning of the paper. Have a read any feedback would be greatly appreciated as would any credible literature/sources 

The value of Bitcoin has exploded over the past year prompting a surge in both public interest and skepticism; continued growth in both usage and market value will have important implications for both business and government. Bitcoin is a new, decentralized digital currency that enables instant transactions to occur between individuals across the globe. Bitcoin is based on money cowry, however unlike  money cowry there is a fixed supply of Bitcoins, they are impossible to counterfeit, they can be divided into a small a piece as needed and are transferrable across vast distances almost instantly through the internet. This builds upon the notion that the four functions of money are summarized in the couplet: "Money's a matter of functions four, a Medium, a Measure, a Standard, a Store of value (Jevons (1875) and (1919) . Bitcoin began in November 2008 when the paper Bitcoin: A Peer-to-Peer Electronic Cash System (Nakomoto:2008) was published through a cryptography mailing list under the pseudonym Satoshi Nakamoto. The paper describes the process by which a peer-to-peer network can be used to generate and support "a system for electronic transactions without relying on trust" (Wallace:2013). In January 2009, the Bitcoin network came into existence with the first open source Bitcoin client and the first Bitcoins were created, with Nakamoto mining the first block of Bitcoins known as the "genesis block" (Nakamoto:2009). Bitcoin is one of the first successful implementations of a distributed crypto-currency, described in part by Dai (1998) on the ‘Cypherpunks’ mailing list. Bitcoin is designed around the idea of using cryptography to control the creation and transfer of money, rather than relying on central authorities. Bitcoin uses peer-to-peer technology to operate with no central authority: management of transactions and issuance of money are carried out collectively by the network without government intervention.

[1714570511]

1714570511

[1714570511]

1714570511

[1714570511]

1714570511

[1714570511]

1714570511

[1714570511]

1714570511

[1714570511]

1714570511

[jimhsu]

Oooh, quantitative finance ... haven't seen that in a while. Note my exposure to this is basically internet blog posts and probably some random textbooks, so probably not up to date. But...

1. Is bitcoin actually experiencing less volatility on a percentage basis as adoption goes up? This could be argued both ways -
On one hand, 30-day rolling annualized volatility has not shown a decrease; rather the process seems to be cyclical (apologies for the hasty graph and badly labeled data -- this is a btc-e timeseries):



30-day autocorrelation is also not that interpretable:


On the other hand, the magnitude of moves and drawdowns seem to have decreased - thus longer term volatility measures may indicate a decrease. Depends on your measurement.

2. Should lower volatility be a policy goal?
On one hand, for bitcoin to be adpoted as a currency, current volatility is clearly unacceptable. But would anti-vol measures ("plunge protection teams", circuit breakers, derivatives) trade short term stability for catastrophic long-term consequences (in other words, would they increase kurtosis)? Haven't done the math, but I suspect bitcoin is even more Levy-distributed (aka not normal distributed) than stocks. Also, due to decentralization, would any anti-vol measures actually be effective? You can't exactly trade stocks the same way you can trade bitcoin.

Just my thoughts.

[Jack3d]

Thanks for the response. The volatility certainly does seem to be cyclical with a nice peak trough structure, as expected volatility is higher when there are large price increases/decreases. Theory would suggest so, as more people adopt bitcoin and it becomes more widespread in its usage volatility should decrease, at least according to macro economic theory. A futher aim of the research is to compare Bitcoin to another newish FIAT/traditional currency, i imagine if i get this far it will be the most difficult bit to do successfully. The goal is to see if there are any similarities I'm thinking an eastern European country however this may not be ideal as I'd prefer a left wing/liberal country as this would fit in with the ideology or thinking behind Bitcoin. The charts are great which program did you use to produce them? I'm using Gretl at the minute. BTC-e is that from a specific exchange or just a variation of traditional bitcoin. Is the X scale years/months/days?

In terms of auto-correlation you would expect quite a lot but that isn't necessarily a bad thing and logically makes sense; the price this period is dependent upon previous periods. I'm going to go slightly beyond this though. GARCH stands for (Generalised AutorRegressive Condiotnal Heteroskedasticity). The variance of financial series is not constant over time like you said its not bell shaped/normally distributed it is 'fat tailed' there is also quite a lot of volatility clustering in financial markets. This is where high volatility is clustered or pooled. This is what i intend to model if there is a pattern of pooling or if pooling follows certain news announcements this would be helpful when trading the currency.

2.This is a much more difficult question. The way i see things panning out it should be a policy goal but to what extent is the real question. In order for bitcoin to become part of the fabric of society volatility needs to decrease, people will become  more computer literate and accepting off a non-physical currency however peoples attitude to volatility is unlikely to change. We as humans like stability and thinking that we know whats round the corner so to speak. By decreasing volatility it then becomes more mainstream more likely that my Mum and Dad would want to use bitcoins, this is a good thing as it will increase the depth of Bitcoins impact into peoples lives as well as the scale. However there seems to be a lot of big money coming bitcoins way in terms of wall street types ( a good example would the Winklevoss twins soon to be launched ETF) i think they would prefer even less volatility in the market, get involved drive up the price of bitcoin more comission etc. (this is a very synical view). I think you are right in hypothesizing that traditional volatility measures wouldn't work, the question becomes what would.

thanks again for responding to the post i'm sure if other realised the potential value of what i am attempting they would also get involved. Apologies for the short essay hope its not too boring.   

[aminorex]

The volatility of bitcoin is not important for its use as a money transmission mechanism, because it transmits instantly.  There is no time window for volatility to affect the transaction.

[jimhsu]

The volatility of bitcoin is not important for its use as a money transmission mechanism, because it transmits instantly.  There is no time window for volatility to affect the transaction.

Not exactly true as you are limited by block-level resolution (10 minutes, give or take). A lot can happen then (as we've seen, for example, April 2013). Furthermore, someone needs to hold the funds -- exchangers, merchant services such as Bitpay, etc.

However the bigger issue is that the "money transmission" part of BTC is a very limited market compared to its perception as a store of value. Again part of the problem is the chicken and egg problem; how do you get merchants to accept BTC, and ordinary people to use BTC?

[jimhsu]

Thanks for the response. The volatility certainly does seem to be cyclical with a nice peak trough structure, as expected volatility is higher when there are large price increases/decreases. Theory would suggest so, as more people adopt bitcoin and it becomes more widespread in its usage volatility should decrease, at least according to macro economic theory. A futher aim of the research is to compare Bitcoin to another newish FIAT/traditional currency, i imagine if i get this far it will be the most difficult bit to do successfully. The goal is to see if there are any similarities I'm thinking an eastern European country however this may not be ideal as I'd prefer a left wing/liberal country as this would fit in with the ideology or thinking behind Bitcoin. The charts are great which program did you use to produce them? I'm using Gretl at the minute. BTC-e is that from a specific exchange or just a variation of traditional bitcoin. Is the X scale years/months/days?

In terms of auto-correlation you would expect quite a lot but that isn't necessarily a bad thing and logically makes sense; the price this period is dependent upon previous periods. I'm going to go slightly beyond this though. GARCH stands for (Generalised AutorRegressive Condiotnal Heteroskedasticity). The variance of financial series is not constant over time like you said its not bell shaped/normally distributed it is 'fat tailed' there is also quite a lot of volatility clustering in financial markets. This is where high volatility is clustered or pooled. This is what i intend to model if there is a pattern of pooling or if pooling follows certain news announcements this would be helpful when trading the currency.

2.This is a much more difficult question. The way i see things panning out it should be a policy goal but to what extent is the real question. In order for bitcoin to become part of the fabric of society volatility needs to decrease, people will become  more computer literate and accepting off a non-physical currency however peoples attitude to volatility is unlikely to change. We as humans like stability and thinking that we know whats round the corner so to speak. By decreasing volatility it then becomes more mainstream more likely that my Mum and Dad would want to use bitcoins, this is a good thing as it will increase the depth of Bitcoins impact into peoples lives as well as the scale. However there seems to be a lot of big money coming bitcoins way in terms of wall street types ( a good example would the Winklevoss twins soon to be launched ETF) i think they would prefer even less volatility in the market, get involved drive up the price of bitcoin more comission etc. (this is a very synical view). I think you are right in hypothesizing that traditional volatility measures wouldn't work, the question becomes what would.

thanks again for responding to the post i'm sure if other realised the potential value of what i am attempting they would also get involved. Apologies for the short essay hope its not too boring.    

It's just excel (2013) for the charts. X-axis is in years (decimal) since inception of btc-e (a major bitcoin exchange). This is a small part of my long-term planning platform (which I won't be posting here).

I'd advise you to look into commodities also to find something similar; BTC has many characteristics of a commodity (relatively stable production rate relative to value, limited supply, cyclical scarcity) as well as a currency (fungible, divisible, no "intrinsic value"). Might fit a commodity better than any other "traditional" currency (which is either pegged (to a commodity, no less) or can be printed infinitely). I can see why regulators such as FinCEN are confused -- as far as I know, bitcoin is the first asset (to ever exist) with properties intrinisic to both currencies and commodities.

[Jack3d]

The volatility of bitcoin is not important for its use as a money transmission mechanism, because it transmits instantly.  There is no time window for volatility to affect the transaction.

As pointed out your statement is flawed i think you might have missed the point of my rteserach. Volatility is extremely important if you want to trade bitcoin, or are a business that uses Bitcoin as pointed out by Jimshu. If you took the time to read my original post you would also see that my theory is that when Bitcoin volatility decreases there will be more wide spread adoption, this widespread adoption will have consequences.