Do investors feedback trade in the Bitcoin—and why?

Abstract

We empirically examine whether feedback traders are active in the Bitcoin and the extent to which their presence is affected by a series of noise-related factors (sentiment; volume; liquidity) at three different frequencies (hourly; daily; weekly) for the April 2013–July 2019 period based on Bitstamp data. Our findings suggest that positive feedback trading grows stronger for higher (hourly; daily) frequencies, with its presence manifesting itself mainly during periods of high/improving sentiment and high/rising volume/liquidity. Additional tests reveal that the significance of hourly feedback trading is identified during hours corresponding to the trading hours of major European/North American markets. Overall, our results confirm extant literature evidence on the prevalence of noise trading in cryptocurrencies, while further showcasing that the factors motivating feedback trading in other asset classes (equities; ETFs; futures) exhibit similar effects over the presence of feedback traders in the cryptocurrency market.

KEYWORDS:

• Feedback trading
• Bitcoin
• sentiment
• volume
• liquidity

JEL Classifications:

• G15
• G40
• G41

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 See the review of Koutmos (2014) which surveys some of that evidence.

2 Fund managers, for instance, have been found to feedback trade more strongly when trading small capitalization stocks (Lakonishok, Shleifer, and Vishny 1992; Wermers 1999; Sias 2004), motivated by the opacity of those stocks’ informational environment (due to limited analyst following, little is known about them). Additionally, rational speculators may resort to feedback trading in order to exploit the trading conduct of their noise counterparts (DeLong et al. 1990).

3 Research on Bitcoin (Takaishi 2018; Vidal-Tomás 2020) has often relied on multiple frequencies, in view of the coin’s differential time-series properties across frequencies; as these studies suggest, inefficiencies in Bitcoin’s series tend to rise in magnitude as one examines higher frequencies. The latter is confirmed in our study, though, of course, it is not possible to ascertain causality (i.e., whether feedback traders cause these inefficiencies or whether inefficiencies motivate their presence, even though both possibilities are theoretically valid). For more on cryptocurrencies’ inefficiencies at high frequencies, see Eross et al. (2019), Baur et al. (2019) and Zargar and Kumar (2019).

4 The role of sentiment in the Bitcoin-market has been explored in Baig, Blau, and Sabah (2019) and Eom et al. (2019), who find that investor sentiment interacts significantly with price-clustering and volatility in the Bitcoin-market.

5 i.e. the moving average value of the previous 30-hours/-days/-weeks, depending on the frequency examined.

6 The presence of high volatility in daily Bitcoin-returns has been established by a large array of studies, including Katsiampa (2017), Phillip, Chan, and Peiris (2018), Chaim and Laurini (2018), Troster et al. (2019) and Katsiampa, Corbet, and Lucey (2019). Evidence from these studies suggests that Bitcoin-volatility exceeds several times that of equity markets, with cross-cryptocurrency studies (Baur and Dimpfl 2018; Katsiampa, Corbet, and Lucey 2019) broadly confirming this for other cryptocurrencies as well. Although cryptocurrencies’ volatility is high, it tends to grow stronger following positive shocks, contrary to extant evidence (e.g. Bollerslev, Engle, and Nelson 1994) from financial time series on the volatility leverage effect (which is reflected via higher volatility following negative shocks).

7 Research has established the presence of positive feedback trading during periods with negative return autocorrelation and high volatility (see e.g., Sentana and Wadhwani 1992).

8 These results are relevant to extant research (Takaishi 2018; Vidal-Tomás 2020), where Bitcoin is found to generate amplified inefficiencies for intraday frequencies.

9 However, note the limited evidence of positive feedback trading reported in Antoniou, Koutmos, and Pericli (2005) for index futures and Chau, Holmes, and Paudyal (2008) for single-stock futures.

10 Of course excluding here the near-future unveiling of a centralized cryptocurrency by the People’s Bank of China.

11 2018 alone witnessed 1253 ICOs (source: https://www.icodata.io/stats/2018), with 2017 entailing a further 343.

12 Source: https://coinmarketcap.com/.

13 Source: https://coinmarketcap.com/.

14 See, for example, Fantazzini et al. (2016).

15 Some studies (Anghel 2021; Hitam, Ritahani Ismail, and Saeed 2019; Torres et al. 2020) have demonstrated how machine-learning techniques can be utilized to exploit information hidden in cryptocurrencies’ time series of prices; machine-learning techniques have traditionally been used in technical analysis (Nazário et al. 2017) and, as such, could be viewed as capable of contributing to feedback trading in this asset class.

16 We opted for BTC/USD values from Bitstamp, considering the fact that it is the largest cryptocurrency exchange internationally (Zargar and Kumar 2019) and its wide popularity among relevant research in this asset class (see e.g. Brandvold et al. 2015; Baur et al. 2019; Choi 2020).

17 According to traditional market microstructure literature (e.g., O’Hara 1995), order-flows reflect innovations in information signals dispersed across market participants. By construction, OFIB varies with the heterogeneity in investors’ beliefs, contingent on whether these beliefs are dominated by bullish (buy-side volume prevails) or bearish (sell-side volume prevails) sentiments. Its usage here as a control variable hinges on the fact that we wish to gauge whether feedback trading varies contingent on whether it is optimistic or pessimistic sentiment that prevails in the market each period.

18 For brevity purposes, statistical significance is defined here at the 10 percent level of significance.

19 In line with Baur and Dimpfl (2018), who find that cryptocurrencies tend to accommodate reverse-asymmetric volatility effects (their volatility grows stronger following positive shocks) or insignificant asymmetric volatility altogether.

20 Examples of such inefficiencies at high frequencies are the intraday patterns documented (Eross et al. 2019; Baur et al. 2019; Zargar and Kumar 2019) for several cryptocurrencies; the presence of those patterns can both be motivated as well as exploited by feedback traders.

21 By definition, negative feedback traders buy when prices fall and sell when they rise.

22 By definition, positive feedback traders sell when prices fall and buy when they rise and returns during low/deteriorating sentiment periods are expected to be, on average, negative.

23 Our results support earlier evidence (Baur and Dimpfl 2018) on noise trading in cryptocurrencies being stronger during positive return periods; showing that volatility grows, on average, stronger among cryptocurrencies following positive shocks, the authors attribute this to noise investors buying aggressively into cryptocurrencies when the latter exhibit price-rallies.

24 Weakly significant (10% level) positive feedback trading is also observed for declining volume periods at the daily frequency.

25 We have repeated our estimations using 15-, 45- and 60-period moving averages for high versus low sentiment/ volume / liquidity. Our results are qualitatively and quantitatively similar to the results with 30-period moving average values and support our original findings. We thank an anonymous referee for making this suggestion.

26 By ‘open' here we refer to each major market’s trading activity; not all markets in Europe/North America (Asia-Pacific) will be simultaneously open within (outside) the 08:00 – 21:00 UTC interval.

27 Eross et al. (2019) state in p. 75 that the volume patterns documented in their paper suggest that ‘[…] European and North American investors are the main drivers of the volume traded of USD denominated Bitcoin', yet also state in the same page (footnote 10) that ‘Although investors can trade outside the usual trading hours of stock markets, consistent with the literature we assume that they will conduct most of their trading during normal stock market trading hours'. We repeat here that, in the absence of transaction data and the possible presence of overnight trading, it is impossible to be assertive as per the geographical origin of feedback trading.

Additional information

Notes on contributors

Rabaa Karaa

Rabaa Karaa is Teaching Follow in Finance at the institute of Higher Commercial Studies (IHEC) of Sousse in Tunisia. She obtained her PhD in Finance in 2017 from the IHEC of Carthage. Her research interests include behavioral finance, financial econometrics, market microstructure, technical analysis, and volatility. Her works include publication in Research in International Business and Finance.

Skander Slim

Skander Slim is Associate Professor of finance at Dubai Business School – University of Dubai. He holds a PhD in Quantitative finance from the University of Paris X-Nanterre (France). His research areas include risk management, option pricing, and market microstructure. His research papers appear in leading refereed journals, including Economic Modelling, Journal of Forecasting, Quantitative Finance, and International Review of Economics and Finance.

John W. Goodell

John W. Goodell is a Professor of Finance at The University of Akron. His research, focusing on the impact of institutional differences on financial systems, has recently been highlighted in numerous media outlets including the Washington Post, PBS NewsHour, and Bloomberg Businessweek, as well as the blogs of the Columbia University and Duke University law schools. He is currently Editor-in-Chief of Elsevier's Research in International Business and Finance.

Abhinav Goyal

Abhinav Goyal (Ph.D. Banking and Finance, University College Dublin) is Professor of Corporate Finance at the University College Cork, Ireland. His primary research interests are in IPOs, corporate governance, emerging markets, IPOs, and privatization. He has presented his research in more than 50 academic seminars, workshops, and conferences worldwide, including Reserve Bank of India and De Nederlandsche Bank, and published in top-tier international finance journals—Journal of Financial Economics, Journal of Financial and Quantitative Analysis, Journal of Corporate Finance, Journal of Econometrics, Journal of Financial Research, Accounting Horizons, Financial Review, and British Accounting Review among others. Since Sept. 2017, he has been an Area Editor of Frontier Markets for (Journal) Research in International Business and Finance.

Vasileios Kallinterakis

Dr. Vasileios Kallinterakis is Senior Lecturer in Corporate Finance at the University of Liverpool Management School. His research interests focus on behavioral finance, institutional investors, market volatility, emerging/frontier markets and high frequency trading. To date, he has published a large number of academic articles in high-quality peer-reviewed journals and edited works. He has served as referee for a multitude of internationally acclaimed, peer-reviewed journals and reviewed research projects for institutions. He is currently a member of the editorial board of several peer-reviewed journals.