国际清算银行-加密交易和比特币价格：来自新零售采用数据库的证据（英）-2022.11-28正式版.doc

《国际清算银行-加密交易和比特币价格：来自新零售采用数据库的证据（英）-2022.11-28正式版.doc》由会员分享，可在线阅读，更多相关《国际清算银行-加密交易和比特币价格：来自新零售采用数据库的证据（英）-2022.11-28正式版.doc（28页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、BIS Working PapersNo 1049Crypto trading and Bitcoinprices: evidence from a newdatabase of retail adoptionby Raphael Auer, Giulio Cornelli, Sebastian Doerr,Jon Frost and Leonardo GambacortaMonetary and Economic DepartmentNovember 2022JEL classification: E42, E51, E58, F31, G28, L50, O32.Keywords:Bitc

2、oin,cryptocurrencies,cryptoassets,regulation,decentralisedfinance,DeFi,retailinvestment.BIS Working Papers are written by members of the Monetary and Economic Department of the Bank for International Settlements, and from time to time by other economists, and are published by the Bank. The papers ar

3、e on subjects of topical interest and are technical in character. The views expressed in them are those of their authors and not necessarily the views of the BIS.This publication is available on the BIS website (www.bis.org).Bank for International Settlements 2022. All rights reserved. Brief excerpt

4、s may be reproduced or translated provided the source is stated.ISSN 1020-0959 (print)ISSN 1682-7678 (online)Crypto trading and Bitcoin prices: evidence from a new database of retail adoptionRaphael Auer, Giulio Cornelli, Sebastian Doerr, Jon Frost and Leonardo Gambacorta1November 2022AbstractPrices

5、 for cryptocurrencies have undergone multiple boom-bust cycles, together with ongoing entry by retail investors. To investigate the drivers of crypto adoption, we assemble a novel database (made available with this paper) on retail use of crypto exchange apps at daily frequency for 95 countries over

6、 201522. We show that a rising Bitcoin price is followed by the entry of new users. About 40% of these new users are men under 35, commonly identified as the most “risk-seeking” segment of the population. To establish a causal effect of prices on adoption, we exploit two exogenous shocks: the crackd

7、own of Chinese authorities on crypto mining in mid-2021 and the social unrest in Kazakhstan in early 2022. During both episodes price changes have a significant effect on the entry of new users. Results from a PVAR model corroborate these findings. Overall, back of the envelope calculations suggest

8、that around three-quarters of users have lost money on their Bitcoin investments.JEL classification: E42, E51, E58, F31, G28, L50, O32.Keywords: Bitcoin, cryptocurrencies, cryptoassets, regulation, decentralised finance, DeFi, retail investment.1 Raphael Auer is with the Bank for International Settl

9、ements (BIS) and is a research fellow of the Centre for Economic Policy Research (CEPR). Giulio Cornelli is with the BIS and the University of Zrich. Sebastian Doerr is with the BIS. Jon Frost is with the BIS and is a research affiliate of the Cambridge Centre for Alternative Finance (CCAF). Leonard

10、o Gambacorta is with the BIS and is a research fellow of CEPR. The authors thank Priscilla Koo Wilkens, Marina Sanchez del Villar, Vatsala Shreeti and Luciano Somoza for helpful comments, and Emma Claggett, Nicola Faessler, Monica Mauron and Alan Soughley for editorial support. The views expressed h

11、ere are those of the authors and not necessarily those of the BIS.11. IntroductionOver the past 13 years, cryptocurrencies have evolved from a niche technological proposal for peer-to-peer payments to a financial asset class traded by millions of users around the world. The largest cryptocurrency by

12、 market capitalisation remains Bitcoin, introduced in 2009 by an anonymous developer under the pseudonym Satoshi Nakamoto (2008). The price of Bitcoin rose from $1 in February 2011 to a peak of $69,000 in November 2021. Globally, it was estimated that over 220 million people owned a cryptocurrency i

13、n June 2021 up from 5 million in 2016.2To date, the volatile price of cryptocurrencies prevents them from becoming widely used as a means of payment. Nor is crypto used as a unit of account; the same volatility makes it impractical to set a fixed price in a specific cryptocurrency, or to use cryptoc

14、urrencies as a yardstick for valuing real economy flows. Moreover, the system is largely self-referential and does not finance real-world investments (Aramonte et al (2022).But why do people invest in cryptocurrencies? In advanced economies, there is evidence that distrust of domestic financial inst

15、itutions or the domestic fiat currency is not a key driver.3 As they fluctuate widely in value and can sustain only a limited volume of transactions,4 cryptocurrencies are also not useful to date for payments in real transactions (purchases) or cross-border money transfers. Some users may however se

16、e cryptocurrencies as a store of value and safe haven (ie “digital gold”) that cannot be appropriated. And certainly, cryptocurrencies could be seen as a speculative investment asset.5In this paper, we shed further light on the role of speculative and safe haven considerations as drivers of cryptocu

17、rrency adoption. For this, we investigate the relationship between the use of crypto trading apps, Bitcoin prices and other macroeconomic variables. We assemble a novel cross-country database on retail downloads and use of crypto exchange apps at daily frequency for 95 countries over 201522.Our main

18、 findings are as follows.First, we show that a rise in the price of Bitcoin is associated with a significant increase in new users, ie entry of new investors. This positive correlation remains robust when we control for other potential drivers, such as overall financial market conditions, uncertaint

19、y or country characteristics. In particular, the price of Bitcoin remains the most important factor when we control for global uncertainty or volatility, contradicting explanations based on Bitcoin as a safe haven. Likewise, when controlling for variables that proxy institutional quality or trust, a

20、s well as the level of economic development, the Bitcoin price still has an economically and statistically significant effect on the number of new users and explains the lions share of the variation in the entry of new users.2345See Blandin et al (2021) and de Best (2022). This is a lower-bound esti

21、mate of identity-verified users.The estimates are subject to uncertainty given the potential for users to have multiple accounts.See Auer and Tercero-Lucas (2022) and FCA (2021).See Boissay et al (2022).See Foley et al (2019), Hileman (2015), Knittel et al (2019) and Swartz (2020).2Second, analysing

22、 the demographic composition of app users we find that 40% of users are men under 35, commonly identified as the most “risk-seeking” segment of the population. These users are more sensitive to changes in the price of Bitcoin than female users and older men. We also find a user sensitivity for Andro

23、id users, who tend to have lower incomes than iOS users.Taken together, these patterns are consistent with the speculative motive being caused by feedback trading considerations, ie users being drawn to Bitcoin by rising prices rather than a dislike for traditional banks, the search for a store of v

24、alue or distrust in public institutions.A concern for our estimation strategy is that the entry of new users could also lead to price increases, raising concerns about reverse causality. To address this issue, we perform two complementary analyses. First, we focus on two specific episodes: the crack

25、down of Chinese authorities on crypto mining activities and the social unrest in Kazakhstan. During both episodes, structural changes affected the global price of Bitcoin, independently of changes in the number of users in other countries. We find that the exogenous change in the Bitcoin price durin

26、g both episodes had a strong and significant effect on the entry of new users. Second, we estimate a panel vector autoregression (PVAR) model, tackling endogeneity issues by means of a Cholesky decomposition which orders the Bitcoin price last.Our contribution to the literature is to provide cross-c

27、ountry evidence that retail investors enter the market following Bitcoin price increases. We speak to papers that seek to explain Bitcoin pricing, from a theoretical and empirical perspective (Garratt and Wallace, 2018; Bolt and van Oordt, 2019; Schilling and Uhlig, 2019; Shams, 2020; Liu and Tsyvin

28、ski, 2021; Biais et al, 2022). We complement recent evidence on investors decision to buy cryptocurrencies and stocks, which helps to explain the recent positive correlation in price movements (Somoza and Didisheim, 2022). With our novel new dataset, we are able to assess retail trading adoption at

29、the country level over time, thus better understanding the link between prices and the entry of new retail investors. Moreover, we are able to show how feedback trading, by which past price changes drive buying and selling (Koutmos, 1997; Danelsson and Love, 2006) is present in crypto markets.Our fi

30、ndings also have relevance for policy discussions on the regulation of cryptocurrencies for consumer and investor protection and financial stability reasons. Indeed, simple simulations suggest that, at the time of writing, 73-81% of users had likely lost money on their investments in cryptocurrencie

31、s. Analysis of blockchain data finds that, as prices were rising and smaller users were buying Bitcoin, the largest holders (the so-called “whales” or “humpbacks”) were selling making a return at the smaller users expense. Our findings raise concerns that individual decisions are backward-looking an

32、d that many retail investors are not fully informed of the risk or volatility of the crypto sector. As recent events have made clear, rising interest rates and other shocks can lead to a persistent fall in prices, as the dynamics that buoyed the market move into reverse.The paper is organised as fol

33、lows. Section 2 introduces our dataset and empirical approach. Section 3 presents our key empirical findings on crypto app use and Bitcoin prices. Section 4 presents a number of extensions that further underscore the causal nature of the results. Finally, section 5 concludes.32. Data descriptionOur

34、data on adoption of crypto apps come from Sensor Tower, a proprietary app intelligence data provider. Sensor Tower collects data on various app statistics, among which downloads and active use, for apps from the Apple and the Google Play store. These statistics are available for up to 95 countries,

35、where the country refers to the location of the downloading users. The data are at daily frequency. Additionally, we collect information on the operating system of the downloading device Apple iOS vs Android users, whereby the former is a common proxy for relatively higher-income individuals (see Be

36、rg et al (2020). 6 We also have information on the gender (men vs women) and age group (young vs old) of the user downloading the app. The latter are only available at the app-quarter level. For our empirical analysis, we draw on more than 200 crypto exchange apps at monthly frequency over August 20

37、15 June 2022. To select the sample of apps, we take the list of crypto exchanges from the CryptoCompare “All Exchanges General Info” application programming interface (API) endpoint. We find a match with the Sensor Tower database for 187 of these exchanges (out of 296). We complement this selection

38、with a list of 26 apps identified as crypto exchange apps by Sensor Tower directly.Sensor Tower gauges unique downloads per iOS or Google Play account. This methodology avoids double-counting due to re-downloads, ie if a user installs, deletes, then reinstalls the same app on the same device or a ne

39、w device from the same iOS or Google Play account. Active users are defined as any user that has at least one session on an app over a specific time period (eg day, week or month). If a user has more than one session over the selected time period, they will still only count as one active user for th

40、at time period. The active user metric is estimated by Sensor Tower based on a representative sample of users. Bearing this caveat in mind, these data offer the unique possibility of measuring real user-adoption directly rather than through a proxy.Data on Bitcoin prices are obtained from CryptoComp

41、are, a leading source of data on cryptocurrency prices.7 In addition to the price and volume data, CryptoCompare, in collaboration with IntoTheBlock, collects statistics on the distribution of Bitcoin holdings at daily frequency. This dataset provides both the number of addresses and the total volum

42、e, broken down by various buckets ranging from balances smaller than 0.001 up to more than 100,000 Bitcoin.We further collect data on stock market prices (MSCI indices), volumes and turnover (Datastream indices), consumer price index (CPI) inflation and foreign exchange (FX) volatility for the count

43、ry in which the app is downloaded. We also use global gold prices and economic policy uncertainty, as measured by the Global Economic Policy Uncertainty (GEPU) Index of Baker et al (2016). In addition, we collect information on commercial bank branches per 100,000 adults, regulatory quality, total67

44、Of course, it is possible that the phone operating system captures other user characteristics such as a preference for a more competitive ecosystem of app developers relative to Apples iOS. In the absence of income data, we do not attempt to distinguish between these possible explanations.While Bitc

45、oin and other cryptocurrency markets are in principle borderless, there can be differences in the prices quoted on exchanges in different countries, eg due to regulation. See Auer and Claessens (2018). These price differences are generally small. As such, we use global price indicators.4population,

46、and real GDP at the country-year level.8 Data on payment app active users and downloads come from Cornelli et al (forthcoming). In this paper the authors collect the top 25 finance apps in each of the countries covered by Sensor Tower and manually tag those apps which are used mainly for payments. F

47、or instance, a stock trading app would not be classified as a payment app, while an app like Venmo would be classified as a payment app.Our final panel includes 95 countries at monthly frequency over the period August 2015 June 2022. Table 1 provides descriptive statistics for our main variables.Descriptive statisticsTable 1No observationsMeanStandardMinMaxdeviationLn(monthly average daily active6,6779.012.581.1315.99users)Ln(monthly average downloads)7,1705.712.353.4312.86Ln(Bitcoin price)7,2428.701.605.4711.04Ln(MSCI equity index price)15