Chinese New Year (Spring Festival) is a festival celebrated at scale, by 2 billion Ethnic Chinese worldwide (including 1.3 billion of our friends in China).
- To give money and or expensive gifts
- Settle debts (it is important to display model behaviour on the first day of Chinese New Year or else you get bad luck the entire year)
Since young, a crucial errand my parents have to make every year without fail is a trip down to the bank: to get crisp, clean and new dollar bills for ‘hongbaos’ (red packets) to give out during Chinese New Year. While it may be still a tradition in Singapore and Hong Kong, this custom is slowly moving to the digital realm in China. Since Tencent’s WeChat released its red envelope feature in 2014, followed by Alibaba through its digital wallet Alipay in 2015, the record number of red envelopes being sent online via WeChat and Alipay has been breathtaking.
Over the six-day Chinese Spring Festival period last year, 46 billion digital red envelopes were sent or received on just WeChat itself, up from 32 billion in 2016 and 3 billion in 2015. The figures highlight the exponential upwards trend of P2P mobile payment transactions and also the seasonality of these transaction volumes.
Hope this provides some context for the analysis of the relationship between the most widely celebrated festival in the world and price movement in cryptocurrency markets. At the same time, shed some light for our media friends in the West that Chinese New Year has already shifted towards being a digital money fest. Dear Forbes, we don’t burn ‘fake money’ to the deceased for good luck during Spring Festival, actually.
Does the price volatility of Bitcoin have anything to do with major holidays?
There are countless literatures that discuss ‘Holiday Effects’ and stock market returns with studies conducted across many markets: US (NYSE, AMEX, NASDAQ), UK (FTSE), Hong Kong, Japan, Germany, Australia. Some of these literatures suggested the presence of ‘Holiday Effects’ where stock returns tend to be abnormally high, with spikes in trading volumes on the last trading day prior to major holidays such as Thanksgiving and Christmas. However, this effect on price movements tend to be temporary.
This lay down an important assumption of this analysis: drivers of short-term price changes are rooted in human behaviour and traditions.
What about BTC market?
As we know, Bitcoin is still likely to be controlled by Chinese despite all the regulations. Indeed, Chinese are unable to buy Bitcoin anymore using Chinese Yuan (CNY) on exchanges but it opened doors for ‘Over-The-Counter’ (OTC) trading and buying BTC using USDT since ‘USDT/BTC’ trading pair is available in all major exchanges. Let’s not forget about the scale of mining operations in China before the crackdown by Chinese government. And we also know the Chinese celebrate Chinese New Year with all that tradition with giving red packets and increased spending during the period.
Connecting the dots, it is not difficult to see this relationship:
BTC is controlled by Chinese -> Chinese celebrates Chinese New Year -> Increased demand for fiat for spending and red packets -> BTC holders cash out some BTC for Chinese New Year -> Increase sell positions in OTC markets (P2P through WeChat, Alipay, bank transfers) in weeks leading up to Chinese New Year -> Sell > Buy -> Potentially translate to ‘sell-off’ events (market cap of BTC is very small and Chinese market is massive) -> slump in BTC prices 2–4 weeks before Chinese New Year
This could the explanation for an observed calendar-related anomaly known as the ‘Reverse January Effect’ or the ‘Chinese New Year Effect’ where BTC prices tend to dip every January, which is coincidentally always 4–6 weeks before Chinese New Year.
TL;DR: Chinese New Year might be one of the driving factors of the dip in BTC prices every January and we might be able to continue to anticipate the dip as long as the Chinese (i) control bitcoins and (ii) celebrate Chinese New Year.
With my rudimentary R skills, I attempted to unravel, if any, the relationship between Chinese New Year and BTC price movement.
In order to examine the presence of ‘sell-off’ events weeks prior to Chinese New Year, the main datasets required would be historical BTC price data and trading volume generated by Chinese BTC holders. Historical price data is tricky to obtain simply because there are sizeable price spreads between exchanges and historical data obtained from price aggregator such as Coinmarketcap is not 100% reliable. A reasonable approach would be to aggregate price data from major exchanges offering BTC/USD trading pair and assign weights according to their market share in terms of trading volume. Again, trading volumes provided by exchanges are not 100% reliable but I guess this is a limitation that we have to work with. My data was extracted from Quandl using the Quandl R package and cross-checked against data from Bitcoinity and Coinmarketcap.
Data for trading volume generated by Chinese BTC holders is an aggregation of trading volume in CNY markets (before the ban) and OTC trading volume (for transactions denominated in CNY and/or off-ramp to Alipay/WeChat/Chinese Banks) from platforms such as localbitcoins, Paxful and CoinCola.
Time frame: Jan 2012 — Jan 2019 (that’s when we started to have reliable price data since its inception)
Examining seasonality of BTC price
An important assumption for seasonality in markets is that sentiments and behaviour of investors tend to be affected by scheduled events in the calendar (such as vacation days, official holidays and even tax season). After the price rallies that happened in 2013 and 2017, we know that cryptocurrency markets are very reliant on the sentiment of people and barely anchored to any particular fundamental value (except perhaps the cost of mining). Therefore, we can assume that calendar-related anomalies are likely to be present and BTC prices are likely to exhibit seasonal element.
I decomposed the time series using 365-days moving average (because I wanted to examine the seasonal component in annual data). BTC prices seemed to exhibit seasonality with repeated price patterns during a fixed time period. Interestingly, we also note that residuals (‘remainder’) had deviated sharply away from zero in recent years suggesting the high level of noise in BTC markets, possibly as a result of increased interest and entry of more speculators into the market due to the price rally in late 2017. Therefore, BTC price is seasonal and non-stationary and you would have to adjust the data for its seasonal effect and perform first differencing if you want to forecast price movement of BTC.
In this study, I am accounting for the presence of calendar-related anomalies and so I am going to leave the data as it is.
Is there really a ‘sell-off’ event?
In attempt to validate the presence of the ‘Chinese New Year Effect’, we examine:
- Is fall in price persistent during the 4–6 weeks prior to Chinese New Year?
- Is there any (causal) relationship between trading activity and price movements during the same period?
Some simple visualisations:
Since 2014 as the market gradually matures with increased adoption of Bitcoin after the first rally in 2013, there seemed to be a consistent pattern where a dip in BTC price is observed at the start of the year and continues towards Chinese New Year (denoted by area highlighted in gray) before a slight recovery takes place thereafter. I will treat 2017 as an ‘outlier year’ because the country-wide clamping down of Bitcoin exchanges by the Chinese government in Jan 2017 caused trading volumes in China to decrease drastically and might have introduced some anomalies in BTC prices during that period. We also observe a countercyclical relationship between BTC prices and trading volumes (both OTC and exchange) in recent years: spikes in Chinese Yuan-denominated BTC trading volume are followed by persistent fall in prices in the period before Chinese New Year.
Recall my thought process:
- Increase sell positions in Exchange/OTC markets (P2P through WeChat, Alipay, bank transfers) in weeks leading up to Chinese New Year -> Increase in trading volumes -> Sell > Buy -> Prices fall by the logic of pricing mechanisms
Hence, this seems to suggest some sort of relationship between fall in BTC prices and the collective behaviour of cashing out of Bitcoin by Chinese BTC holders for spending during Chinese New Year.
Looking at the price changes 30 days before and after Chinese New Year from 2012 to 2018, we observe that price changes tend to be negative prior to Chinese New Year from 2014 onwards and turn positive shortly after Chinese New Year (exhibiting slight recovery of the market). Also, magnitude of price changes tend to be smaller just a few days before and after Chinese New Year, possibly signalling weaker trading activity. Coupled with decrease in trading volumes just right before Chinese New Year exhibited from the charts above, it seemed that the sell-offs tend to slow down or cease during then. Possibly because the Chinese are too busy spending quality time with their family during the festive period — is this a coincidence? ;)
Although we observe countercyclical relationship between BTC prices and trading volumes, we are unable to prove that the fall in BTC prices are indeed caused by spike in Chinese Yuan-trading volumes during Chinese New Year. There could be other events happening in markets outside China. An example would be a conspiracy theory that I have heard of: institutional investors (or ‘whales’) are building a short position before the CBOE bitcoin futures contract expiration which is coincidentally during the same period. Some people are speculating that aggressive selling activity could have been used to drive down the price of bitcoin on the exchange and turn the futures contracts into winning bets.
One way to prove that the slump in BTC price is caused by Chinese sell-offs would be to compare the behaviour similarity between user activity in payment services such as Alipay and WeChat (which act as off-ramps to cash) and transactional signal (BTC price or transaction volumes). Percy Venega created a network graph which visualises the similarity of behaviour of the users of payment services to BTC prices using a signal processing technique known as canonical time warping, for the study of alignment of human behaviour between different subjects. The network graph revealed that user activity of the Hong Kong based P2P OTC platform, CoinCola, has the highest similarity to BTC price movement. More insights from his study here.
Are the signs there?
We’ve seen some potentially interesting relationships (though not proven causal) between BTC price movement and trading activities during Chinese New Year. Are the signs compelling? Maybe not.
Or else if that’s the case, the ‘Chinese New Year’ effect should be visible in other markets, notably in mutual funds and gold since they make up a significant proportion of risky asset allocation of private households in China. (Note: Chinese still prefer holding non-risky assets such as cash and real estate.) However, we do not see any downturns in these markets in January due to a sell-off. Additionally, the cost of celebrating Chinese New Year need to increase exponentially to make the math fit.
Thus for making it thus far and Happy Chinese New Year!
Thank you Percy Venega for providing some ideas for my preliminary research.
Bitcoin Historical Events and Price Movement During Spring Festival Holiday-Average Annual Growth…
At the beginning of 2011, bitcoin was 0.6 USD /BTC and at at the end of 2017, price increased to over 10k USD, an…