Tobias Burggraf, Toan Luu Duc Huynh, Markus Rudolf, Mei Wang
Purpose This study examines the prediction power of investor sentiment on Bitcoin return. Design/methodology/approach We construct a Financial and Economic Attitudes Revealed by Search (FEARS) index using search volume from Google's search engine to reveal household-level (âbankruptcyâ, âunemploymentâ, âjob searchâ, etc.) and market-level sentiment (âbankruptcyâ, âunemploymentâ, âjob searchâ, etc.). Findings Using a variety of quantitative methodologies such as the transfer entropy model as well as threshold regression and OLS, GLS and 2SLS estimations, we find that (1) investor sentiment has strong predictive power on Bitcoin, (2) household-level sentiment has larger effects than market-level sentiment and (3) the impact of sentiment is greater in low sentiment regimes than in high sentiment regimes. Based on these information, we build a hypothetical trading strategy that outperforms a simple buy-and-hold strategy both on an absolute and risk-adjusted basis. The results are consistent across cryptocurrencies and regions. Research limitations/implications The findings contribute to the ongoing debate in the literature on the efficiency of cryptocurrency markets. The results reveal that the Bitcoin market is not efficient in the sense of the efficient market hypothesis â asset prices do not fully reflect all available information and we were able to âbeat the marketâ. In addition, it sheds further light on the debate whether Bitcoin can be considered a medium of exchange, i.e. a currency or an investment product. Because investors are reallocating their Bitcoin holdings during times of increased market sentiment due to liquidity needs, they obviously consider bitcoin an investment product rather than a currency. Originality/value This study is the first to examine the impact of investor sentiment measured by FEARS on Bitcoin return.
We study the problem of the intraday short-term volume forecasting in cryptocurrency exchange markets. The predictions are built by using transaction and order book data from different markets where the exchange takes place. Methodologically, we propose a temporal mixture ensemble, capable of adaptively exploiting, for the forecasting, different sources of data and providing a volume point estimate, as well as its uncertainty. We provide evidence of the outperformance of our model by comparing its outcomes with those obtained with different time series and machine learning methods. Finally, we discuss the predictions conditional to volume and we find that also in this case machine learning methods outperform econometric models.
Purpose This paper aims to elaborate on the optimization of two particular cryptocurrency portfolios in a mean-variance framework. In general, cryptocurrencies can be classified to as coins and tokens where the first can be thought of as a medium of exchange and the latter accounts for security or utility tokens depending upon its design. Design/methodology/approach Against this backdrop, this empirical study distinguishes, in particular, between pure coin and token portfolios. Both portfolios are optimized by maximizing the Sharpe ratio and, subsequently, compared with alternative portfolio strategies. Findings The empirical findings demonstrate that the maximum utility portfolio of coins, with a risk aversion of Îť = 10, outweighs alternative frameworks. The portfolios optimized by maximizing the Sharpe ratio for both coins and tokens indicate a rather poor performance. Testing the maximized utility for different levels of risk aversion confirms the findings of this empirical study and confers them more robustness. Research limitations/implications Further investigation is strongly recommended as tokens represent a new phenomenon in the cryptocurrency universe, for which only a limited amount of data are available, which restricts the sampling. Furthermore, future study is to include more sophisticated optimization models using different constraints in portfolio creation. Practical implications In light of the persistently substantial volatility in cryptocurrency markets, the empirical findings assert that portfolio managers are advised to construct a global minimum variance portfolio. In the absence of sophisticated optimization models, private investors can invest according to the market values of cryptocurrencies. Despite minor differences in the risk and reward ratios of the portfolios tested, tokens tend to be more speculative, especially, if the Tether token is excluded, which may require enhanced supervision and investor protection by regulating authorities. Originality/value As the current literature investigates on diversification effects of blended cryptocurrency portfolios rather than making an explicit distinction, this paper reflects one of the first to explore the investability and role of diversifying coins and tokens using a classic Markowitz approach.
Huaigang Long, Adam Zaremba, Ender Demir, Jan Jakub Szczygielski ¡ 5 authors
This study presents the first attempt to examine the cross-sectional seasonality anomaly in cryptocurrency markets. To this end, we apply sorts and cross-sectional regressions to investigate daily returns on 151 cryptocurrencies for the years 2016 to 2019. We find a significant seasonal pattern: average past same-weekday returns positively predict future performance in the cross-section. Cryptocurrencies with high same-day returns in the past outperform cryptocurrencies with a low same-day return. This effect is not subsumed by other established return predictors such as momentum, size, beta, idiosyncratic risk, or liquidity.
Massimo La Morgia, Alessandro Mei, Francesco Sassi, Julinda Stefa
In the last years, cryptocurrencies are increasingly popular. Even people who are not experts have started to invest in these securities and nowadays cryptocurrency exchanges process transactions for over 100 billion US dollars per month. However, many cryptocurrencies have low liquidity and therefore they are highly prone to market manipulation schemes. In this paper, we perform an in-depth analysis of pump and dump schemes organized by communities over the Internet. We observe how these communities are organized and how they carry out the fraud. Then, we report on two case studies related to pump and dump groups. Lastly, we introduce an approach to detect the fraud in real time that outperforms the current state of the art, so to help investors stay out of the market when a pump and dump scheme is in action.
Mohammad Hashemi Joo, Yuka Nishikawa, Krishnan Dandapani
Cryptocurrencies have gained popularity as new economic investment assets globally in recent years. This study examines market reactions to major news events associated with cryptocurrencies. Abnormal returns as well as cumulative abnormal returns (CARs) around major news announcements, both positive and negative, are investigated for three primary cryptocurrencies: Bitcoin, Ethereum, and Ripple. High abnormal returns are observed on the event day (Day 0), and CARs typically diverge during event windows of (â3, 6) and (0, 6), indicating that the information is not fully reflected in prices immediately after the news events. The CARs that linger for six days after an event suggest that the information flow in the cryptocurrency market is visibly slow. The magnitudes of CARs are larger for negative events than for positive events, implying that the market reaction to negative events is stronger than to positive announcements. The findings of this study may have crucial implications for investors, arbitragers and practitioners as we document evidence of potential trading opportunities for investors who initiate a trading position even after announcements.
Bitcoin prices have fluctuated greatly, and news media have warned investors about a possible price bubble, arguing that the fluctuation arises mainly from people's blind pursuit of short-term trends. Despite these increasing concerns, however, only a few studies have addressed them. This article examines the problem using agent-based modeling. In our model, agents are designed to interact with one another in two ways: Price and social interactions. In their price interactions, agents adopt one of three strategies (fundamentalist, momentum trading, and contrarian trading) in investing at each time and adopt a strategy to maximize their expected benefits. In their social interactions, uninvolved agents become involved at various times based on their network properties (word-of-mouth effect). To examine the distinctive properties of Bitcoin from a comparative perspective, two representative currencies (Euro and Turkish lira) and two financial assets (Nasdaq and Nasdaq leverage index) are used in the agent-based model. The results show that the fraction of fundamentalists and price volatility have mutual Granger causal relationships overall. Also, no significant differences are found in the parameters of social interaction. These results are contrary to commonly held beliefs that Bitcoin prices are merely a result of blind pursuit and herding behavior.
Economic theory suggests that introduction of derivative contracts can improve the informational efficiency of the underlying asset prices (Danthine, 1978). In this study, we examine the impact of the introduction of Bitcoin futures on price clustering in Bitcoin. Our findings suggest that price clustering in Bitcoin meaningfully decreases post the introduction of its futures contracts.
Abstract This study explores whether Bitcoin constitutes as a hedging instrument whilst seeking portfolio diversification opportunities among sustainable, conventional and Islamic asset classes since Bitcoin emerges as a distinct alternative investment and asset class across the world. We apply multivariate generalised autoregressive conditional heteroscedasticâdynamic conditional correlation and continuous wavelet transforms based on the recent data set ranging from August 18, 2011, to September 10, 2018. First, our findings show that Bitcoin returns are meanâreverting which implies that its value tends to come down to mean value in the long run and not completely crushed to zero irrespective of price changes suggesting Bitcoin as a sustainable asset class. Second, the timeâinvariant model shows that Bitcoin offers portfolio diversification opportunities with almost all equity indices, in particular, Dow Jones Islamic followed by FTSE 4 Good index. Finally, the timeâvariant analysis reconfirms that Bitcoin offers portfolio diversification benefits both in the short and long run. These findings carry meaningful policy considerations for fund managers and crossâcountry investors.
Optimizations given historical data unsurprisingly produce sizeable allocations to Bitcoin (XBT). But further analyses of risks raise questions, even abstracting from expected returns. GARCH-based measures of dynamic XBT volatility and covariance suggest that optimal weights change over time. Also, quantile regressions indicate that conditional XBT returns with respect to the S&P 500 are modestly positively skewed. Yet benevolent symmetry is hardly stable or consistent along the distribution. Spectral analysis shows that the XBT volatility primarily owes to higher-frequency cycles. Nonetheless, XBT betas are substantially greater, and notably positive, over longer cycles compared with shorter cycles, which implies that XBT has been a much less effective strategic hedge. Dynamic principal components analysis indicates that individual coinsâ exposures to the âcrypto market factorâ have likely increased meaningfully enough over time to diminish diversification benefits. <b>TOPICS:</b>Currency, portfolio construction, portfolio theory <b>Key Findings</b> ⢠Standard mean-variance portfolio optimizations given historical data unsurprisingly produce sizeable allocations to Bitcoin (XBT). But further analyses of risks raise questions, especially for passive investors and abstracting from expected returns. For example, GARCH-based measures of dynamic XBT volatility and covariance suggest that optimal portfolio weights change substantially over time. ⢠Quantile regressions indicate that conditional XBT returns with respect to the S&P 500 are modestly positively skewed, arguably unlike even safe-haven assets such as US Treasuries. However, this comparatively benevolent symmetry is hardly stable or consistent along the distribution. ⢠Spectral analysis shows that the XBT volatility primarily owes to higher-frequency cycles, much like common asset classes. Nonetheless, XBT betas are substantially greater, and notably positive, over longer cycles compared with shorter cycles, which implies that XBT has been a much less effective strategic hedge. Also, dynamic principal components analysis indicates that individual coinsâ exposures to the âcrypto market factorâ have likely increased meaningfully enough over time to diminish diversification benefits for passive investors.
The main goal of this study is to examine whether the cryptocurrency market impacts the stock market returns in the Gulf countries. Understanding this impact is quite interesting to clarify whether the cryptocurrency market and the stock market are substitutes or complements for investors. The author compiles the data on the stock market of the Gulf countries with the cryptocurrency data on a daily basis over the period 2014-2019. Generalized Method of Moments with Instrumental Variable (IV - GMM) approach has been implemented as the main strategy to fulfill the objective of the paper. The results of this paper show that the Stock market and the cryptocurrency market are substitutes for investors in Gulf countries. In fact, each 10 percent increase in the cryptocurrency returns is associated with a decline in the stock market returns by 0.17 percent. The cryptocurrency market hampers the stock market indices in the Gulf countries. Having agreed upon in the literature that the stock market is affected by fundamental factors, market sentiment, technical factors, and anomalies, this study offers robust evidence that the cryptocurrency should be introduced as one of the main determinants of stock market prices and returns.
We utilise principal component analysis to determine whether a (small) set of factors can explain cryptocurrency returns and whether this varies over time. We find that a substantial proportion of cryptocurrency return variation is explained by a single principal component that is highly correlated with bitcoin returns. The explanatory power of this factor is greatest for larger cryptocurrencies and increases markedly in the most recent part of the sample. Our results have implications for investors determining optimal portfolio decisions and for policyâmakers wary of systemic risk.
This study uses a smooth transition autoregressive model with exogenous variables (STARX) to investigate whether there is a nonlinear relationship between Bitcoin and Taiwanâs stock market taking into account Taiwanâs monetary policy threshold during 2 February 2012 to 31 August 2019. The statistical results show there is a threshold effect and confirm a nonlinear relationship between Taiwanâs stock market and Bitcoin, with variations over time and across Bitcoin and Taiwanâs stock market. Specifically, we find that Bitcoin responds asymmetrically to Taiwanâs stock market according to the threshold value. Furthermore, the return on the closing price of TAIEX with a lag of two periods under Taiwanâs monetary policy threshold has a nonlinear impact on the return on the closing price of Bitcoin.
There has been a recent surge in interest in the application of artificial intelligence to automated trading. Reinforcement learning has been applied to single- and multi-instrument use cases, such as market making or portfolio management. This paper proposes a new approach to framing cryptocurrency market making as a reinforcement learning challenge by introducing an event-based environment wherein an event is defined as a change in price greater or less than a given threshold, as opposed to by tick or time-based events (e.g., every minute, hour, day, etc.). Two policy-based agents are trained to learn a market making trading strategy using eight days of training data and evaluate their performance using 30 days of testing data. Limit order book data recorded from Bitmex exchange is used to validate this approach, which demonstrates improved profit and stability compared to a time-based approach for both agents when using a simple multi-layer perceptron neural network for function approximation and seven different reward functions.
Shuyu Zhang, Dunli Zhang, Jianming Zheng, Walter Aerts
Abstract Policy uncertainty created by regulatory authorities regarding the blockchain matters for the initial coin offering (ICO) market. Using an ICO dataset from four major cryptocurrency exchanges, we find that higher policy uncertainty regarding the blockchain leads to a lower return and a lower trading volume of an ICO on its first trading day. Although ICOs are decentralised by design and less vulnerable to direct regulatory intervention, we conclude that blockchain policy uncertainty constitutes an important component of the information environment of ICO ventures.