U.S. Consumers’ Adoption and Use of Bitcoin and other Virtual Currencies ∗ Scott Schuh† Federal Reserve Bank of Boston Oz Shy‡ MIT Sloan School of Management December 3, 2015 Very preliminary and incomplete Abstract Bitcoin has gained notoriety as a speculative financial asset, vehicle for criminal activity, alternative monetary policy instrument, and a low-cost payment instrument for merchants to accept. However, relatively little attention has been devoted to measuring its adoption and use by consumers for payments—the main purpose for which Bitcoin was designed (Nakamoto, 2008). The 2014-2015 Survey of Consumer Payment Choice (SCPC) indicates that about half of U.S. consumers had heard of Bitcoin or the other roughly 700 virtual currencies by the end of 2015. The vast majority of consumers who are aware of virtual currencies are unfamiliar with them, and they struggle to answer survey questions accurately. Less than 1 percent of U.S. consumers (2 percent of consumers aware of virtual currency) have ever owned (adopted) virtual currency, but most adopters of virtual currency used it to pay a person (most commonly) or merchant in the past month. Awareness, adoption, and use of virtual currencies are correlated with various demographic and economic characteristics of consumers. A typical Bitcoin owner is more likely to be a younger, non-white male with lower education who expects Bitcoin to appreciate, has adopted a higher number of other payment instruments, and has most of the responsibility for household shopping. Keywords: Digital currency; virtual currency; cryptocurrency; Bitcoin; cash; money demand JEL Classification Number: TBD (Draft = bitcoin-14.tex 2015/12/03 18:25) ∗ We thank Hanna Halaburda for helpful comments on an earlier draft and David Zhang for excellent research assistance. The views expressed in this paper are those of the authors and do not necessarily represent the views of the Federal Reserve Bank of Boston or the Federal Reserve System. † E-mail: scott.schuh@bos.frb.org. Research Department, Federal Reserve Bank of Boston, 600 Atlantic Avenue, Boston, MA 02210, U.S.A. ‡ E-mail: ozshy@ozshy.com. MIT Sloan School of Management, 100 Main Street, Building E62, Cambridge, MA 02142, U.S.A. 1. Introduction Since its inception (Nakamoto, 2008), Bitcoin has grown steadily in stature and garnered increasing attention. As of March 2015, about two-thirds (14 million) of the intended long-run supply of bitcoins had been distributed worldwide, held in 109 million accounts and used for 62.5 million ¨ total transactions or about 200,000 per day (Bohme et al., 2015). Most attention toward Bitcoin has focused on its exchange rate with the U.S. dollar, which has exhibited breathtaking swings over short periods of time. It reached a peak of about $1,400 in 2013 (market capitalization of $12 billion, equal to 4.5 percent of the U.S. money stock M1) before falling back to $200-450 during the past year (market capitalization of $3-6.5 billion). During its relatively short life, Bitcoin has struggled to gain credibility with the general public for many reasons. For most consumers, Bitcoin is hard to understand because it is new, unfamiliar, and technologically complex. It is also risky to hold because its exchange rate with sovereign currencies like the U.S. dollar is extremely volatile and not suitable for risk averse consumers. Like many new and volatile assets, Bitcoin has attracted considerable speculative investment while its financial infrastructure was being developed, as well as fraud, theft, and related activity causing financial losses for some owners. One of Bitcoin’s key advantages—cryptographic security based on blockchain technology—is alleged to have attracted criminal and terrorist payment activity that requires anonymity in financial transactions. All of these challenges increasingly have motivated governments to regulate, restrict, or even ban virtual currencies, which have buffeted the value of Bitcoin further and hindered its goal of becoming a viable alternative to official government currencies. Nevertheless, its potential as a speculative investment has fueled not only Bitcoin’s rise to prominence but also naturally attracted market competition, as described in Gandal and Halaburda (2014). Since Bitcoin’s value peaked in 2013, about 700 diverse other virtual currencies have emerged to compete in the market for private digital money characterized by cryptographic security. Bitcoin’s competitors collectively represent less than 10 percent of the industry’s market cap, and some may have been created primarily to score short-term speculative profits. However, others are making serious efforts build better digital payment systems, such as Ripple’s meta net1 work on which Bitcoin and other currencies of all kinds could be used. The emergence of the virtual currency market also has stimulated attempts to develop cryptographic versions of official government currencies such as the Canadian and U.S. dollars.1 Virtual currency poses fascinating questions about its potential role as a form of private money, which is an unfamiliar concept to modern U.S. consumers. Economists disagree about whether Bitcoin and other virtual currencies exhibit the core features of money (store of value, unit of account, and medium of exchange) despite being supplied by the private sector.2 The U.S. Internal Revenue Service (IRS) ruled that virtual currency is not a “currency” per se but property, on which capital gains taxes must be paid. Thus, virtual currency is an asset (store of value) but one with an expected return that so far has been unusually volatile for money. Like most government money today, virtual currencies have no intrinsic value but rather a fiat declaration and promise to sustain its usability. For Bitcoin, the fiat emerges from the democratic social consensus of the open-source community that owns and operates the system, but it does not have any legal or public authority. Virtual currencies are very divisible and precise (16 decimal places) units of account (typically a “coin”) with variable exchange rates against the U.S. dollar similar to those of foreign sovereign currencies. Finally, virtual currencies such as Bitcoin are designed to serve as a medium of exchange—a form of payment instrument—that can be used at relatively low social cost and (so far) little or no transaction cost for either the payer or payee. Although it is too early to tell whether the foothold gained thus far by private virtual currency will last, or grow to an economically significant magnitude, the future of private virtual currency depends crucially on the extent to which it is used and accepted for payment by agents. Much data are readily available on the Internet for the aggregate supply, market value, and transaction use (in volume and value) of virtual currencies. However, to date there has been very little data and research produced on particular consumers and merchants who use and accept virtual currencies for payments. Polasik et al. (2015) provides interesting evidence about merchants and their acceptance of bitcoin for payment obtained from a recent international survey, but it does not contain details about the actual consumers who paid these merchants in virtual currency. Another recent 1 2 For examples, see MintChip, IBM, JP Morgan Chase, Tibado, etc. For example, Lo and Wang (2014) argue that Bitcoin is not money. 2 study by CoinDesk (2015) reports demographic features of Bitcoin users, but it is a proprietary study that does not reveal details about its statistical and sampling properties.3 To best of our knowledge, this paper provides the first nationally representative evidence on consumer adoption and use of virtual currencies, in this case for the United States. We use the annual (2008-2015) Survey of Consumer Payment Choice (SCPC) from the Federal Reserve Bank of Boston to estimate the diffusion of virtual currencies among U.S. consumers. The SCPC collects data on the adoption and use of all bank accounts and payment instruments for a longitudinal panel of consumers, as well as consumer preferences and assessments of those instruments. In 2014, the SCPC began to add questions about Bitcoin and other virtual currencies, which were supplemented by additional surveys during 2015. A key advantage of our analysis is that the SCPC contains comprehensive information on each consumer’s payment behavior. These data allows us not only to characterize the types of consumers who adopt and use virtual currencies, but also to investigate their reasons and determine whether it produced any substitution for other payment instruments (such as cash) or practices (in-person versus online). As of October 2015, still less than half (47 percent) of all U.S. consumers had heard of (were aware of) any virtual currency, up from about 39 percent the year before. Among consumers who are aware of virtual currencies, nearly nine in ten of them report being “slightly” or “not at all” familiar with them. In October 2014, nearly 4 percent of all consumers (or nearly 10 percent of those aware of virtual currency) misidentified the general term “other virtual currency” as referring to things that are not virtual currencies (e.g., sovereign currencies, PayPal, and other payment services). A majority of virtual currency owners were not able to consistently and accurately report the number of coins and dollar values of their virtual currency holdings. In light of low awareness and familiarity with virtual currency, and its inherent challenges, it is not surprising that less than 1 percent of U.S. consumers (less than 2 percent of aware consumers) have ever owned (adopted) any virtual currency; even fewer currently own it because they have discarded what they had. However, most adopters have used their virtual currency to make a payment in the past month to a person (most common), merchant, or both. This result suggests 3 Add other consumer studies here. 3 that Bitcoin and other virtual currencies are not being held primarily or exclusively for investment, but also for their originally intended function as an electronic-cash means of payment. Awareness, adoption, and use of virtual currencies are correlated with various demographic and economic characteristics of consumers. Consumers who are male, white, and high income or highly educated are more likely to be aware of virtual currency than consumers without these characteristics. Conditional on awareness, however, a typical Bitcoin owner (adopter) is more likely to be a younger, non-white male with lower education who expects Bitcoin to appreciate and bears relatively more responsibility for household shopping. Far fewer consumer characteristics are correlated with the decision to use virtual currency to make payments (of any type), but nearly all self-reported reasons for adopting virtual currency are significant determinants of use. Interestingly, consumers who have adopted a higher number of other payment instruments are more likely to be aware of virtual currency, own it, and use it—further evidence that consumers view virtual currency as a viable payment instrument. The remainder of the paper proceeds as follows. Section 2 defines terms and concepts, briefly reviews the nascent literature on virtual currency, and discusses the theories underlying the paper’s econometric models. Section 3 describes the SCPC and other data used to report statistics and estimate econometric models. Section 4 analyzes the diffusion of information about virtual currency among U.S. consumers. Section 5 reports results for adoption and holdings of virtual currency, and Section 6 reports results for the use of virtual currency. Section 7 offers some tentative conclusions. 2. Theoretical Discussion 2.1 Literature review Although some predecessors existed at the time of of the seminal Bitcoin publication, Bitcoin, (Nakamoto, 2008), can be credited with introducing and popularizing the concept we call “virtual currency” that has achieved widespread notoriety and recognition of potential broad-based appeal. As is often the case with innovations, the early stage of development of Bitcoin and other virtual currencies has been characterized by a lack of consensus about terminology and defini- 4 tions. So, before reviewing the academic literature, we offer some perspective on these fundamentals and explain our choice of terminology.4 These issues are central to consumers’ ability to understand the questions asked in the SCPC. 2.1.1 Terminology and concepts The title of (Nakamoto, 2008) “Bitcoin: A Peer-to-Peer Electronic Cash System,” provides a starting point for discussing terminology and concepts. A central focus of that paper is creating a network (“system”) for people to pay other people (“peer-to-peer” or P2P) using an online only (“electronic”) form of cash that does not require the involvement of a financial institution as a third-party to validate the transaction. In essence, this proposal advances a form of private payment instrument, or money, that is not sovereign currency but has characteristics similar to cash— especially privacy or anonymity for the payer and payee. Nakamoto proposes relying on cryptographic proof instead of trusted third-party financial institutions to verify the transaction via a public ledger that tracks and records all transactions, which all members of the system can read and verify as correct. Economists and other experts who study Bitcoin and similar currencies have proposed a number of terms to refer to these types of currencies. The ECB (European Central Bank, 2012) first proposed the term “virtual currency” (see especially Table 1, “A money matrix”), which also was used by Greene and Shy (2014). A more recent report by the Bank for International Settlements (BIS 2015) favored the term “digital currency,” with digital intended to be synonymous to electronic, although it acknowledged the essential interchangeability of the term with virtual currency (see BIS footnote 2). As emphasized by the ECB, digital currency is a broader term that can also be applied to electronic forms of sovereign currencies like the U.S. dollar, such as a debit card or the Automated Clearing House (ACH). Furthermore, in all countries, digital sovereign currency can be transformed into physical currency (cash, or coins and notes), whereas virtual currency conveys the notion that it remains digital and cannot be converted to the physical realm. For this reason, we prefer the term virtual currency. However, both terms (virtual or digital) are imperfect short-hand descriptors because they do 4 We thank Hanna Halaburda for suggesting the need for this discussion. 5 not adequately reflect all important characteristics of these currencies. For example, neither one explicitly reflects the cryptographic security, which is why some prefer the term “cryptocurrency,” yet not all virtual or digital currencies have cryptography. Neither term reflects the ownership of the currency (public or private), even though the expected returns vary dramatically, or the intrinsic value (fiat or asset-backed), even though it varies considerably as well. The ECB classification provides a related but different distinction: regulated versus unregulated, which seems to imply a mapping to public versus private currency, respectively. One can further distinguish between type of fiat: centralized (as in sovereign currencies) versus decentralized (as in Bitcoin). As one can see from this discussion, there are multiple dimensions along which one can classify Bitcoin and similar currencies so there is considerable ambiguity not only about how Bitcoin works but also about how to think about what it is. This ambiguity is even more severe among the general public of consumers and respondents to surveys. Therefore, in the SCPC, we define multiple terms (virtual currency, digital currency, and cryptocurrency) in hopes of reaching all consumers. We further ask respondents about Bitcoin, which is likely the most common “brand name,” and a short list of the most valuable and well know competitors to Bitcoin (e.g., Ripple and Dogecoin), to maximize respondent understanding. Nevertheless, if economists have not proposed and settled on a consensus term, it is likely that many consumers may be confused about the concept. 2.1.2 Other research TO BE COMPLETED • Virtual currency as technology, payments, and money: Velde (2013), Lo and Wang ¨ (2014), Bohme et al. (2015), Greene and Shy (2014) • Virtual currency as asset and pricing: Glaser et al. (2014), Hencic and Gouriroux (2015), Donier and Bouchaud (2015), Dwyer (2015), Bolt and van Oordt (2015). ¨ • Network economics of virtual currency industry: Halaburda and Gandal (2014), Moser ¨ and Bohme (2015). 6 • Virtual currency acceptance and use: (merchants) Polasik et al. (2015), (consumers) Tsanidis et al. (2015), Saito (2015), Christin (2013), CoinDesk (2015). 2.2 Underpinnings of econometric results The first step in modeling consumer adoption and use of virtual currencies involves handling the incomplete diffusion of knowledge about about them among consumers. Unlike traditional payment instruments like credit cards, which have existed since the 1950s and reached steady state adoption rates of around 70 percent by the 1980s, virtual currencies are relatively new and not widely disseminated. By late 2015, not all consumers were even aware that Bitcoin had been created seven years earlier, much less understood the conceptual similarities and differences between e-money and virtual currency. Even many who had heard of virtual currency via the media and had read about the concept of virtual currency were unfamiliar with the basic mechanics of Bitcoin described in the primer by Velde (2013), much less experts in blockchain technology underlying virtual currency. For these reasons, we first estimate the diffusion of awareness of virtual currency. Next we model the ownership of virtual currency by consumers. To do so, we restrict our population of study to consumers who are aware of virtual currency and exclude those who are not. This restriction is logical because a consumer who does not know virtual currency exists cannot be reasonably expected to have made a rational decision to adopt it or not. There are at least two potential ways to model demand for virtual currency, depending on whether it is viewed as money or not. If we ignore the potential monetary aspect of virtual currency, we can follow the literature on adoption of payment instruments and treat the decision as a discrete choice and estimate the latent probability of adoption using logit regressions.5 Alternatively, if we assume that virtual currency is private money, it may be more appropriate to specify a model of the demand for money in terms of the dollar value of holdings. Absent a structural model of money demand that distinguishes between sovereign and private currencies, we follow the traditional literature that uses reduced-form econometric models in the spirit of Baumol (1952) and Tobin (1956); in particular, we adopt the specification in Briglevics and Schuh (2013), which also uses SCPC data to estimate the demand for U.S. currency (or “cash”). 5 For examples of this approach, see Schuh and Stavins (2010, 2012, 2015) and the references therein. 7 A key differences between the demand for sovereign and virtual currencies is their opportunity costs. Demand for sovereign currency depends on the nominal interest rate—a real interest rate and expected inflation (the devaluation of the currency)—available to holder of sovereign currency. For this measure, we can use interest provided on bank checking or savings accounts, or short-term investments like money market mutual funds. However, banks are not able to hold or transmit virtual currencies because their cryptographic security prevents them from complying with the Bank Secrecy Act (1970), which requires adherence to anti-money laundering (AML) and know-your-customer (KYC) regulations. New types of depository institutions have emerged to hold virtual currency stocks (e.g., CoinBase, Circle, etc.), but so far they do not fractionally reserve balances and issue loans to earn interest. However, there appear to be some innovative ways of earning interest on Bitcoins, such as margin trading exchanges.6 Demand for virtual currency depends on exchange (or principal) risk due to its floating exchange rate with the U.S. dollar. Therefore, expectations of future exchange rates (the “price” of virtual currency) should influence the demand for virtual currency in a way that it does not influence the demand for sovereign currency. Finally, for the population of consumers who have adopted virtual currencies, we model the use of them on the extensive and intensive margins. We model the incidence of using virtual currency (extensive margin) as a discrete decision and estimate the latent probability of use with logit regressions. These results provide an estimate of the extent to which consumers hold virtual currency just for investment purposes or for payments as well. Also following the literature on consumer payment choice cited earlier, we model the virtual currency share of consumer payments (intensive margin) using OLS regressions. These regressions include consumers’ assessments of virtual currency characteristics (cost, convenience, security, etc.), relative to the characteristics of other payment instruments, as explanatory variables; for details, see Schuh and Stavins (2010). 6 See “How to Earn Interest on Bitcoin 5 Different Ways,” Cryptorials, June 27, 2015, http://cryptorials.io/how-to-earninterest-on-bitcoin-5-different-ways/ 8 3. Survey Data 3.1 Survey of Consumer Payment Choice We use the Survey of Consumer Payment Choice (SCPC) from the Federal Reserve Bank of Boston to report statistics and estimate econometric models. The SCPC is a 30-minute annual (2008-2015) recall-based survey of U.S. consumers that measures their adoption and use of payment instruments, including cash, and their associated transaction accounts (bank and non-bank); see Schuh and Stavins (2012, 2014) for more details. Each year except 2008, the SCPC included about 2,000 respondents who form an unbalanced longitudinal panel. The SCPC was administered to members of the RAND Corporation’s American Life Panel (ALP) from 2008-2014 and the University of Southern California’s (USC) Understanding America Study (UAS) from 2014-2015.7 These Internet survey panels are designed to be approximately representative of U.S. consumers but may exhibit unobserved selection effects. Aggregate survey results are weighted to match the U.S. population as measured by the Current Population Survey. Questions about virtual currency were first introduced in the 2014 SCPC (ALP and UAS versions) and then expanded in three additional surveys in 2015, all of which are summarized in Table 1. Because the adoption rate of virtual currency by U.S. consumers is so low, the 2014 SCPC yielded only 26 adopters out of 3,047 total respondents (ALP plus UAS) and it was not feasible to do statistically precise analysis. Consequently, the Boston Fed used Qualtrics to administer the 2014 SCPC to a targeted over-sample of 611 virtual currency adopters in the summer of 2015.8 Although this over-sample version of the 2014 SCPC was completed in July 2015, nine months after the official SCPC (October 2014), we combine the two samples and refer to them as the 2014 SCPC. Finally, the 2015 SCPC (UAS version only) was conducted in October and included an oversample of about 125 virtual currency adopters from Qualtrics. [NOTE: All aggregate 2015 survey results appearing in tables are not weighted yet.] Internet survey panelists (ALP and UAS) and Qualtric survey participants exhibit some poten7 By and large, the ALP and UAS panels have similar composition. However, the ALP was originally constructed Qualtrics works with a number of partners who maintain, or can gain access, to a large number of potential respondents. Initially, Qualtrics administered a brief screening question to a large and broad sample of respondents that asked whether the respondent had (owned) virtual currency. If so, the respondent was offered the opportunity to take the entire 2014 SCPC. This pre-screen sampling strategy signficantly reduces the survey costs. 8 9 tially important differences. First, the composition and recruitment methodology of the Internet survey panelists are well-known and designed to be as representative as possible. However, some Qualtrics respondents may come from survey vendors who do not invest in same quality of survey and sampling methodologies as the ALP and UAS, and some of their respondents may come from convenience samples or marketing firms. Thus, in general, it is not possible to estimate precisely the Qualtrics sampling frame or its relation to the U.S. population; hence it is difficult to construct accurate sampling weights. Also, the Internet survey panelists are paid a much higher incentive ($20 for 30 minutes) than the Qualtrics respondents (typically $5 or even less), so participation and response rates are generally better for the ALP and UAS members. 3.2 Survey questions While most of the SCPC questions and structure were very similar between 2014 and 2015, Table 2 shows that the virtual currency questions varied across surveys. The table reports the survey questions in the first column and then indicates with an “X” which questions were included in the questionnaires in the next four columns: (1) 2014 SCPC (ALP and UAS); (2) 2014 over-sample (Qualtrics); (3) 2015 SCPC (UAS) and 2015 over-sample (Qualtrics).9 The official 2014 SCPC (October 2014) included the fewest questions. The 2014 and 2015 SCPC questionnaires first ask respondents whether they had heard of Bitcoin, then they asked whether they had heard of any other virtual currencies. If a respondent answered “yes” to either question, he is defined to be “aware” of the respective virtual currency. In the case of other virtual currencies, the 2014 SCPC respondents also had to indicate specifically which other virtual currency(ies) they had heard about, and we are able to verify whether their answers correctly identified virtual currencies. Because we found a high frequency of errors in the 2014 SCPC, we added a specific list of the top 5 virtual currencies (in terms of market capitalization) after Bitcoin before asking the open-ended question about other virtual currencies.10 In the 2015 SCPC, we added a question for everyone who was aware of virtual currency about their degree of familiarity (1-5 scale) with virtual currency. 9 The third questionnaire (2015) was the same for the UAS and Qualtrics respondents because USC recruited the 2015 Qualtrics over-sample respondents to take their online UAS survey. 10 The top 5 other virtual currencies were Ripple, Litecoin, Dash, Stellar, and Dogecoin. 10 Unfortunately, the pre-screening questions for the Qualtrics over-samples made it infeasible to ask awareness questions because only one question could be asked and the question was whether the respondent currently owned (adopted) any virtual currency. For this reason, the over-samples use a longer list of other virtual currencies (top 14 by market capitalization).11 However, given that it is very difficult to identify the composition of the full Qualtrics sample, it is not really feasible to estimate awareness (as a percentage of the population) from the over-samples anyway. So, the over-samples are primarily used to supplement the 2014 and 2015 SCPC samples for the regression analyses; when they are used in tables, the SCPC sample weights are adjusted for the over-sampling. In both years, the SCPC asked basic questions about adoption, holdings, and use of virtual currencies (Bitcoin and each specific other ones). The questionnaires ask about current adoption (“Do you have...”) and historical adoption (If not, “Have you ever had...”) because holdings of virtual currency, like cash (and unlike payment cards), can be depleted through spending and must be replenished. The questionnaire also asked how much virtual currency adopters have (own), measured by both the number of coins and the U.S. dollar-equivalent value of their holdings to increase the quality of measurement and implicitly test for consistency relative to market exchange rates. The questionnaires also asked whether adopters had used virtual currency to make a payment during the past 12 months. However, the 2015 SCPC and both over-sample questionnaires contain some additional questions to improve understanding of the initial 2014 results. These questionnaires asked the primary and second reasons for owning virtual currency. They also asked about the intensity of use of virtual currency (number of payments) and asked about payments in the past 30 days as well as past 12 months, and expectations of price ($/bitcoin exchange rate) changes over the next week, month, and year. Finally, because the 2014 results for virtual currency holdings raised concerns about possible measurement error (described in detail later), respondents to the 2015 SCPC were also asked whether they used web sites or their financial records in reporting the estimates of coins, dollar 11 The top 14 other virtual currencies included the top 5 mentioned earlier plus Bitshares, Nxt, BanxShares, Peercoin, MaidSafeCoin, NameCoin, ByteCoin, Monero, BlackCoin. 11 values, or both. For the consumers who reported both the number of coins and the dollar value of holdings, the SCPC offers the chance to test the internal consistency of their data by comparing the implied exchange rate (dollar value divided by number of coins) of the two estimates with the market exchange rate on the day of the survey. For Bitcoin, approximately 79 percent of respondents reported both the number of coins and dollar value.12 4. Awareness of Virtual Currency This section examines the degree to which information about virtual currency has disseminated among consumers using data on Internet search activity and the 2014-2015 SCPC. 4.1 Internet search data One potential proxy for consumer awareness of Bitcoin over time is the number of Google search hits in the United States on the term (“Bitcoin”), which is plotted in Figure 1.13 By this measure, it appears that some consumers first began to show awareness of Bitcoin in mid-2011, about two years after Bitcoin was first made operational. The intensity of searching for Bitcoin surged again in early 2013 and early 2014. Interestingly, Figure 1 shows that the intensity of searches for Bitcoin is highly correlated (ρ = .80) with movements in the dollar value of Bitcoin. This result suggests that early awareness of Bitcoin may have been primarily related to store of value and speculative investment rather than its use for payment. In any case, if search intensity proxies for information accumulation, it appears that awareness of Bitcoin is roughly following the standard S-shaped diffusion pattern, as depicted by the cumulative intensity of search for the term “Bitcoin” shown in Figure 2. In principle, we can treat awareness of other virtual currencies the same way as awareness of Bitcoin, though there are many more of them and each is much less well known than Bitcoin. The total number of virtual currencies in existence, also shown in Figure 2, appears to follow a remarkably similar S-shaped diffusion pattern as well, lagging Bitcoin awareness by roughly a year. 12 Another 11 percent reported only coins, 2 percent reported only dollar values, and 8 percent reported neither (conditional on reported having adopted Bitcoin). 13 The Google search intensity is an index for which units are not available. 12 4.2 Consumer survey data Turning to the SCPC results, we find that about half (47 percent, unweighted) of U.S. consumers were aware of any virtual currency by October 2015, up from about 39 percent (weighted) in October 2014, as shown in Table 3. By far (and not surprisingly), the vast majority of consumers who are aware of virtual currency are aware of Bitcoin, not other virtual currencies. Note that the individual estimates of awareness for Bitcoin and other virtual currencies do not sum to the total virtual currency number because some respondents were aware of both Bitcoin and other virtual currencies. While it has taken seven years since (Nakamoto, 2008) for half of the U.S. population at least to have heard of virtual currency, not everyone who is aware of virtual currency is intimately familiar with the complicated new technology. Table 3 shows that 87 percent of respondents to the 2015 SCPC reported being “not at all” or “slightly” familiar with virtual currency. Thus, our discrete measure of awareness (aware versus not aware) overlooks important variation in the extent to which consumers actually understand virtual currency—many are aware but don’t really understand. In the future, it would be better to measure the specific degree of consumer’s understanding of virtual currency. Nevertheless, it is interesting and feasible to estimate the pace of diffusion of information about virtual currency in the United States. Given our two observations on awareness (October of 2014 and 2015) and a reliable estimate of the initial condition (zero percent of the population prior to the Bitcoin paper in October 2008), we use the Bass (1969) diffusion model to fit and forecast Bitcoin awareness. Figure 3 plots the estimated process, which suggests that it could take a total of 15 years (2008 to 2023) to reach nearly complete awareness of Bitcoin in the United States. Another sign that consumers lack full and accurate information about virtual currency can be found in the row of Table 3 labeled “incorrectly identified,” which indicates that about 3 percent of consumers (or nearly 10 percent of those aware) do not truly know what virtual currency is. Table 3 contains the types of other virtual currencies respondents identified when asked what they were. Some consumers correctly named virtual currencies, even the closed-loop Linden Dollar, which is quite different from Bitcoin and other open-loop virtual currencies. However, some 13 consumers mistake online payments (PayPal, Google Wallet, etc.), sovereign currencies (such as the Euro, Peso, etc.), and other things for virtual currency. Henceforth, we exclude erroneous responses from definitions of awareness, adoption, and other measures from our analysis. To determine the types of consumers who are more likely to be aware of virtual currency, we use a dummy variable for awareness (aware = 1, otherwise = 0) to run logit regressions and estimate the latent probability of being aware of Bitcoin or other virtual currencies (corrected for erroneous responses). The results appear in Table 4. Estimates for Bitcoin are most significant; estimates for other virtual currencies are qualitatively similar but much less significant given the small number of non-zero observations.14 Bitcoin awareness does not depend on age, but it is increasing in income and education, and decreasing in household size. Men and white consumers also are much more likely to be aware of Bitcoin. Interestingly, awareness of Bitcoin (but not other virtual currencies) increases with the number of payment instruments adopted by consumers, suggesting that a preference for having more ways to make payments is correlated with knowledge of new payment methods. We characterize the types of consumers who are more likely to erroneously classify something as a virtual currency using a logit regression (dependent variable is a dummy variable with error = 1). Table 5 shows a somewhat paradoxical result: consumers with higher education (college degree or post-graduate study) are less likely to erroneously classify virtual currency, but lower income consumers also are less likely to make a classification error. Older consumers are more likely to make a classification error, but the effect is an order of magnitude smaller than for income and education. To determine the types of consumers who more familiar with virtual currency, we use the degree of familiarity (1 to 5 scale) to run OLS regressions and estimate the effects of consumer characteristics on familiarity with virtual currency. The results appear in Table ??. Only two characteristics are significant: older consumers are less likely to be familiar with virtual currency, and male consumers are more likely to be familiar. 14 The dependent variables in these regressions are fewer and simplified to conserve on degrees of freedom in the other virtual currency regressions. The Bitcoin regression dependent variables are the same for comparability. 14 5. Demand for Virtual Currency This section reports empirical and econometric results on two measures of consumer demand for virtual currency. First, the extensive margin of demand is measured as the discrete state of holding (owning) virtual currency or not. Second, the intensive margin of demand is measured as the amount of virtual currency holdings, a continuous variable in the U.S. dollar value of virtual currency.15 5.1 Adoption of Virtual Currency Table 7 shows that adoption of virtual currency by U.S. consumers is quite low. In 2014, only .52 percent of consumers reported having virtual currency (current adoption; more had Bitcoin (.47 percent) than other virtual currencies (.16 percent).16 Historical adoption of virtual currency, which includes consumers who previously owned virtual currency but did not have any at the time of the survey, was .82 percent; thus .32 percent of consumers had discarded virtual currency at some point. A reasonable case may be made for measuring adoption as a percentage of only consumers who are aware of virtual currency, but even then only 1.3 percent of consumers may be said to have adopted virtual currency (or 2.1 percent historically). A key issue for consumers adopting virtual currency is their expectation of future exchange rates (price), which determines the principal (or exchange rate) risk of virtual currency holdings.17 The SCPC asks respondents for their qualitative (decrease, stay the same, or increase) expectations of virtual currency exchange rates over the next week, month, and year. Expectation questions were asked only of respondents who were aware of virtual currency, and Table 8 contains the results. Perhaps not surprisingly, there is a marked difference in expectations between adopters and non-adopters of virtual currency. Well more than half of non-adopters expect the exchange rate to stay the same, with a reasonably distribution of expected increases and decreases, though slightly more seeing a decline. In sharp contrast, a large majority of Bitcoin and non-Bitcoin adopters ex15 Alternatively, virtual currency holdings could be expressed in units of the currency (“coins”), but the two measures should be essentially the same due to the existence of an exchange rate ($/coin). 16 Note that the 2015 estimates of adoption rates in the table are not weighted yet and thus may not be accurate estimates of the population. 17 Expectations of future exchange rates also is related to consumer awareness and familiarity with virtual currency, or consumer financial literacy more broadly. 15 pect the exchange rate to stay the same, at worst, but most expect the exchange rate to increase at all horizons.18 To determine the types of consumers who have more optimistic price expectations, we use the degree of optimism (-2 to 2 scale) to run OLS regressions and estimate the effects of consumer characteristics on virtual currency price expectations at each horizon (week, month, year). The results appear in Table 9. Three characteristics are significant explanatory variables for expectations. Older consumers have lower price expectations for some reason(s), but consumers who have more responsibility for making investment decisions in their household have higher price expectations. Consumers who have adopted relatively more other payment instruments also have higher price expectations. To determine the types of consumers who are more likely to adopt virtual currency, we use a dummy variable for adoption (adopt = 1, otherwise = 0) to run logit regressions and estimate the latent probability of adopting virtual currency of any type, and individually for Bitcoin and other virtual currencies.19 The results, which appear in Table 10, are highly significant and generally similar for Bitcoin and other virtual currencies. Unlike awareness, age is correlated with adoption as older consumers are considerably less likely to adopt virtual currencies. Likewise, although white consumers are notably more likely to be aware of virtual currencies they are somewhat less likely to adopt them, and men are not more likely to adopt virtual currencies conditional on being aware of them. Consumers who have adopted more payment instruments, and consumers who have higher shares of cash payments, are more likely to adopt virtual currencies as well, even though virtual currency may be a substitute for other payment instruments. Perhaps the most interesting—but certainly not surprising—adoption results are the large and highly significant coefficients on price expectations. All three horizons are important independently, so the shape of the forecast matters. A higher expected change in price over the next week 18 Future drafts of the paper will explore the role of exchange rate expectations in the adoption and holdings of virtual currency. 19 This regression is based on a pooled sample of the July 2015 oversample and the 2014 SCPC sample, excluding consumers who are not aware of virtual currency (and therefore cannot be expected to adopt something they don’t know exists). Although the virtual currency adoption measure was different in the oversample because it did not ask awareness questions first, we believe that directly asking respondents if they had any Bitcoin is likely to produced similar results. Furthermore, the oversample questions asked about 14 specific types of other virtual currencies and thus is likely to have measured adoption better than the 2014 SCPC questions. 16 raises the probability of adoption by 9 to 10 percent. Because consumers’ expected changes over each horizon are almost surely correlated, next two expectation variables are the differences from the the immediately earlier horizon, which reduce problems of multicollinearity. Both changes in price expectations (week to month and month to year) also are highly significant and of similar magnitude. Although illuminating, the adoption regressions do not reveal the specific reasons for consumers adopting or not adopting virtual currency. To obtain this information, the SCPC directly asked consumers to identify their primary reasons for adopting or not adopting virtual currency, and their secondary reasons for adopting (only in the 2014 SCPC oversample). The results appear in Table 11. The single most common reason for consumers owning virtual currency is that more than one in five consumers are “interested in new technologies” which corresponds well with the technological sophistication of the product. However, combining answers into common groups reveals that more than one in three consumers adopts virtual currencies for reasons related to making payments (i.e., to buy goods and services, to make remittances, or to make payments anonymously). In contrast, less than one in five consumers holds virtual currency for investment. Furthermore, only one in ten consumers holds virtual currency because they distrust banks or sovereign currency. Consumers who do not adopt virtual currency cite a wide range of reasons with roughly similar frequencies. 5.2 Holdings of Virtual Currency The 2014–2015 SCPC asked all current adopters of virtual currency to estimate the amounts of their holdings, which are reported in Table 12. Unfortunately, the results shows signs that there may be significant measurement error in the responses. The first panel of the table shows that in 2014 the median Bitcoin adopter held about 3–7 bitcoins and the mean holding was about 75 bitcoins.20 The second panel shows that median Bitcoin adopter held about $150-500 in bitcoin, with a mean of $800-1,600. Individually, the coin and value numbers seem plausible, but viewed together there appears to be an inconsistency in the reporting. With an approximate exchange rate 20 The mean number of coins for other virtual currency seems implausibly large and may require data cleaning so we ignore them for now. 17 of $300 per bitcoin, the median coin holdings suggest considerably higher value holdings than the owners’ reported estimates. To verify this conjecture, the third panel of the table reports the implied value of coin holdings using the actual official exchange rates at the time of the survey. Indeed, these value estimates are much larger than the owner estimates. The wide disparity in dollar-value estimates of holdings can be further illuminated by looking at the implied exchange rates of consumers who reported both components of Bitcoin. The bottom panel of Table 12 shows that the 2014 SCPC sample estimates of Bitcoin exchange rate (mean $193, median $333) are reasonably close to the actual market exchange rate during this time (roughly $250 to $400). However, the 2014 oversample estimates of the Bitcoin exchange rate (mean $4,151 and median $9) are much further from the actual range in July 2015. Apparently, many consumer estimates of virtual currency holdings are not particularly trustworthy. However, to refine the estimates a bit, we identified and reported a set of “best” estimates in the fourth panel of the table. The best estimates use information about whether respondents checked their financial records when reporting coin and value estimates, and limited the exchange rate error (difference between implied rate and actual market rate) to 10 percent in absolute value. Using these “best” responses the mean and median estimates of Bitcoin value fall between the full sample of owner reported holdings and the coin estimate evaluated at actual market exchange rates.21 In light of considerable uncertainty about the quality of the data on virtual currency holdings (both coins and dollar values), we do not estimate models of money demand (holdings of virtual currency) at this time. If we are able to better verify and clean these data, we will report these econometric results in future versions of the paper for comparison with models of consumer demand for U.S. currency. 6. Use of Virtual Currency This section explores the propensity of virtual currency adopters to use the currency to make payments and, if so, who they pay and what they buy. Like adoption, there are two margins of 21 In future work, we will benchmark these estimates of coin and value holdings to online data on virtual currency holdings by account. 18 virtual currency use. First, the extensive margin of use is measured as the discrete state of having made a payment with virtual currency during the past 12 months (past month or otherwise). Second, the intensive margin of use is measured by the number of payments made with virtual currency during a typical month, which can also be expressed as the share of monthly payments for each consumer, the latter having econometric advantages demonstrated in previous research. Although the SCPC clearly asks respondents about virtual currency payments for goods and services, it is possible that some respondents may include strictly financial payments made with virtual currency, such as exchange virtual currency for sovereign currency (buying dollars, for example). Furthermore, knowing the payee (merchant versus person) does not necessarily reveal what consumers paid for with virtual currency or the type of payment (bill versus non-bill). 6.1 Incidence of Use Many virtual currency adopters used it to make a payment in the past 12 months, as shown in Table 13. The incidence of use estimates vary widely between the 2014 SCPC (28 percent) and the 2014 oversample (75 percent), and we have not yet constructed proper sampling weights for the combined sample yet. Given the dominant size of the oversample of adopters (about 500 compared to less than 30 in the SCPC), it is likely that more than half of them used virtual currency to pay. The 2014 oversample also indicates that the vast majority of adopters (about nine of ten) who made payments in the past 12 months also made a payment within the past month, so paying with virtual currency appears to be the norm, not a rarity, among adopters. Note again, however, that some of these reported payment could have been financial payments. To determine the types of adopters who are more likely to make payments with virtual currency, we use a dummy variable for incidence of use (use = 1, otherwise = 0) to run logit regressions and estimate the latent probability of paying with Bitcoin or other virtual currencies.22 The results, which appear in Table 14, yield many fewer significant results than the adoption regressions for the demographic and other common explanatory variables. Men are more likely to use Bitcoin for payments, but not other virtual currencies. Consumers who have adopted more pay22 This regression is based on a pooled sample of the July 2015 oversample and the 2014 SCPC sample. The results for any virtual currency (Bitcoin and other virtual currency together) were qualitatively and quantitatively quite similar and not reported separately. 19 ment instruments are more likely to make payments with all types of virtual currency. But the other common variables do not contribute significantly to the probably of paying with virtual currency. However, the incidence of use regressions also add reasons for adopting virtual currency as explanatory variables, and these are quite significant. All of the payment and investment reasons are economically and statistically significant correlates with use. However, the magnitudes of the coefficients are remarkably similar so no specific reason dominates. Payment security is very significant for Bitcoin, which is based on blockchain technology, but it is not for other virtual currencies, which use a variety of security technologies. Distrust of the government is marginally significant. 6.2 Type of Payee From the perspective of two-sided market analysis, it is important to know who the payees are that receive virtual currency payments from consumers. Table 13 also shows the extent to which consumers used virtual currency to pay a person, a merchant, or both types of payees. More than half (56 percent) of virtual currency users paid another person, and two of five (39 percent) paid a merchant, while one in four (25 percent) made payments to both in the past 12 months. To determine the types of payers who are more likely to make payments with virtual currency to each type of payee, we use a dummy variable for each payee (paid = 1, otherwise = 0) to run logit regressions and estimate the latent probability of paying each type of payee with virtual currency.23 The results, which appear in Table 15, have very few significant coefficients. Only consumers who have adopted more payment instruments have consistent, significant correlation with payee choice. Men are more likely to pay a merchant. But in general we do not find many explanations for the choice of payee by consumers who pay with virtual currency. 6.3 Frequency of Use In this section we report estimates of two-stage Heckman selection model of the adoption and use of virtual currency in the spirit of the literature exemplified by Schuh and Stavins (2010). The first23 This regression is based on a pooled sample of the July 2015 oversample and the 2014 SCPC sample. 20 stage adoption model is analogous to, but simpler than, the model reported in earlier results. The second-stage model’s dependent variable is the share of monthly consumer payments made with virtual currency, which implicitly controls for each consumer’s heterogeneous adoption pattern. However, our preliminary results do not include the relative characteristics of virtual currency (cost, security, convenience, etc.), which have been shown to be important determinants of both adoption and use. Table 16 reports the results of this preliminary estimation. The first-stage adoption estimates (bottom panel) are broadly similar to results reported earlier and most variables are statistically and economically significant. However, the second-stage use estimates (top panel) reveal far fewer significant variables—similar to the results reported earlier for the extensive margin of use (incidence). Perhaps the most notable estimates are for the two coefficients on reasons for adoption related to payments (buy goods and services, and make remittance payments) in the Bitcoin equation. The estimates are positive and large (3 percentage point effect on the share of Bitcoin payments), providing support for the hypothesis that Bitcoin is used as a means of payment. However, these results, like others throughout the paper, are very preliminary and subject to change after further refining and development. 7. Conclusions This paper provides very preliminary empirical and econometric results that characterize U.S. consumers’ experience with Bitcoin and other virtual currencies. The evidence is based on a respected, nationally representative survey that measures consumer adoption and use of all major U.S. payment instruments. However, the data and analysis are not complete and all results should be viewed circumspectly. Nevertheless, at this stage several basic results emerge. First, information about Bitcoin and other virtual currencies still had only reached about half of the U.S. adult population by the end of 2015, and it may take another decade before the vast majority of consumers are aware. Moreover, most consumers who are aware of virtual currencies report being largely unfamiliar with them, and even those who own them exhibit errors in reporting their holdings. Men and consumers with high income and education appear to be more aware 21 of virtual currencies. In any case, incomplete and imperfect information about virtual currencies contributes to their limited dissemination in the U.S. economy. Second, adoption of Bitcoin and other virtual currencies is remarkably low even among consumers who are aware of them. This finding suggests that virtual currencies still do not appeal to consumers enough to obtain, hold and use them. One key exception is that consumers who expect virtual currencies to appreciate in value are much more likely to demand them. This result may suggest that consumers demand virtual currencies as a financial investment. At the very least, high volatility in virtual currency values likely predisposes consumers toward owning them only if they do not pose a serious exchange rate (principal) risk. Third, among those consumers who own virtual currency there appears to be considerable use of them to make payments for goods and services and to other consumers, roughly in equal proportions. This finding raises further questions about the extreme view that virtual currencies are merely speculative investment, especially given the relatively limited acceptance for payments among merchants and, presumably, consumers thus far. 22 Table 1: Description of surveys. 2014 Respondents Time period of implementation Vendor Awareness Adoption Summary of other questions 2015 SCPC 3,047 Oct-14 Oversample 611 Jul-15 SCPC 1800 (est) Oct-15 Oversample 125 (est) Oct-15 ALP + UAS Bitcoin, other Same Ownership and records (1), Use (1) Qualtrics NA Bitcoin, top 14 Ownership and records (2), Use (7), Reasons (2), Expectations (2) UAS Bitcoin, top 5, other Bitcoin, top 5 Ownership and records (2), Use (7), Reasons (3), Expectations (1), Familiarly (1) Qualtrics NA Bitcoin, top 5 Ownership and records (2), Use (7), Reasons (3), Expectations (1), Familiarly (1) 23 Table 2: Virtual currency questions in the 2014 SCPC (Oct), 2014 SCPC oversample (OS, Jul 2015), and 2015 SCPC (official and oversample, Oct 2015). 2014 SCPC OS Awareness Have you heard of Bitcoin? Have you heard of any other virtual currencies [Top 5]? Have you heard of any other virtual currency? (if yes, please specify) How familiar are you with Bitcoin and how it works? Adoption (conditional on awareness) Do you have or own any Bitcoin? Do you have or own any of these other virtual currencies? [Top 5] Do you have or own any of these other virtual currencies? [Top 14] Do you have or own any other virtual currencies? Historical adoption (conditional on awareness) Have you ever had or owned any of these virtual currencies? [Bitcoin] Have you ever had or owned any of these virtual currencies? [Top 5] Have you ever had or owned any of these virtual currencies? [Other VCs] Reasons for (non-)adoption (conditional on (non-)adoption) What is the main reason that you do not own any virtual currency? Please tell us your primary reason for owning virtual currency [9 options] Please tell us your secondary reason for owning virtual currency [9 options] Amount owned (conditional on adoption) How much virtual currency do you have or own? [coins, USD equivalent; Bitcoin, other] Did you have to refer to records or websites to know the number of coins or the equivalent value in U.S. dollars? Incidence and frequency of use (conditional on adoption) In the past 30 days, have you used virtual currency to make a payment or transaction? [Bitcoin, other] In the past 12 months, have you used virtual currency to make a payment or transaction? [Bitcoin, other] In the past 30 days, how many payments did you make using a virtual currency? [Bitcoin, other] In the past 12 months, how many payments do you make using a virtual currency? [Bitcoin, other] Location of use (conditional on adoption) Have you used virtual currency to pay a merchant (store, company, or other business)? (If yes) Please list up to three merchants you have paid using virtual currency. Have you used virtual currency to pay a person (somebody who is not a merchant)? Price expectations and assessments (conditional on awareness) How do you expect the value of one bitcoin to change over the following time periods. [week, month, year] Assessments of bitcoin/virtual currency [security, acceptance, cost, convenience, getting & setting up, payment records]24 X X X X X X X 2015 SCPC/OS X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X SCPC July 2015 - PRELIMINARY AND UNOFFICIAL Table 3: Virtual currency awareness and familiarity. Table A Virtual Currencies Awareness and Familiarity Percentage of consumers. 2014 SCPC Awareness.............................................................................................................. Bitcoin............................................................................................................... Other virtual currency*...................................................................................... Incorrectly identified.................................................................................. Other payment services (eg. Apple, Paypal, Google Wallet)................ Sovereign Currencies (eg. Euro, Pesos, Pound)................................... Cannot remember/other........................................................................ Familiarity with Bitcoin (among those that are aware)..................................... Not at all familiar.............................................................................................. Slightly familiar................................................................................................. Somewhat familiar............................................................................................. Moderately familiar........................................................................................... Extremely familiar............................................................................................. Full UAS 2015 SCPC** UAS 39.5 39.5 1.1 2.9 1.1 0.9 0.9 39.6 39.6 0.9 3.5 1.4 1.3 0.7 46.6 46.2 3.5 1.2 0.2 0.3 0.7 100.0 57.5 29.6 8.3 3.9 0.8 * Excluding incorrectly identified. Correctly identified VCs include Dogecoin, Litecoin, Linden dollars, Ripple, Namecoin, Eucador, game based coins, Isracoin, Darkcoin, Blackcoin, BAMstorm and Feathercoin. ** Preliminary and unweighted. 25 Table 4: Awareness of virtual currencies. Age Agesq (x 1000) edu c pgs hh size White male log income num pi adopt csh sh cc debt revolver cc rewards adopt smartphone adopt Observations Pseudo R-squared N Positive 2014 Full 0.00196 -0.0155 0.158∗∗∗ -0.0403∗∗∗ 0.117∗∗∗ 0.201∗∗∗ 0.0852∗∗∗ 0.0339∗∗∗ 0.0978∗∗ -0.000181 0.0410∗∗ 0.00944 2786 .14 1385 Bitcoin 2014 UAS -0.00121 0.0194 0.168∗∗∗ -0.0342∗∗∗ 0.106∗∗∗ 0.188∗∗∗ 0.119∗∗∗ 0.0292∗∗ 0.0427 0.00961 0.0151 0.0141 1081 .18 551 2015 UAS 0.00128 -0.0177 0.154∗∗∗ -0.0474∗∗∗ 0.160∗∗∗ 0.207∗∗∗ 0.0476∗∗ 0.0503∗∗∗ -0.0829 -0.0103 0.0295 0.0285 988 .17 543 Marginal effects ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01 26 2014 Full -0.00161 0.00843 0.00720 -0.00129 0.00512 0.0164∗∗∗ 0.00822∗∗ 0.00459∗ 0.00208 0.00727 0.00408 -0.00338 2786 .18 37 Other VCs 2014 UAS 2015 UAS -0.00168 -0.00195 0.00917 0.0141 -0.00372 0.000553 -0.000301 -0.00337 0.0115 -0.00591 0.0116 0.0206∗ 0.0312∗∗ -0.00623 0.00526 -0.00381 0.00592 -0.0212 0.00614 0.0102 -0.00648 0.0105 0 0.0253 828 986 .17 .04 13 41 Table 5: Erroneous responses to other virtual currency. male age log income edu c pgs cc sh num pi adopt csh sh cc debt revolver cc rewards adopt smartphone adopt Observations Pseudo R-squared N Positive vc error -0.0276 0.0110∗∗∗ -0.122∗∗ -0.159∗∗ -0.275∗ -0.0241 0.158 -0.138∗ 0.0324 -0.0348 118 .46 101 Marginal effects Sample: 2014 SCPC oversample and 2015 SCPC ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01 27 Table 6: OLS on familiarity (1-5 scale). age agesq (x 1000) edu c pgs hh size White male log income num pi adopt csh sh cc debt revolver cc rewards adopt smartphone adopt Constant Observations Rsquared Mean SD familiarity -0.0408∗∗ 0.336∗ 0.0824 -0.0263 0.172 0.402∗∗∗ -0.0368 0.0659 -0.212 -0.00974 -0.0924 0.141 2.427∗∗∗ 459 .09 1.599 .847 Sample: 2015 SCPC. ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01 28 SCPC July 2015 - PRELIMINARY AND UNOFFICIAL Table 7: Virtual currency adoption. Table A2 Virtual Currencies Adoption Percentage of consumers. 2014 SCPC 2015 SCPC** UAS Full UAS Current adoption................................................................................................... Bitcoin............................................................................................................... Other virtual currency*...................................................................................... Incorrectly identified.................................................................................. 0.52 0.47 0.16 0.37 0.56 0.56 0.00 0.73 0.44 0.44 0.00 Historical adoption................................................................................................ Bitcoin............................................................................................................... Other virtual currency*...................................................................................... Incorrectly identified.................................................................................. 0.84 0.71 0.24 0.75 0.92 0.77 0.15 1.01 0.96 0.88 0.09 Discarding.............................................................................................................. Bitcoin Other virtual currency*...................................................................................... Incorrectly identified.................................................................................. 0.32 0.24 0.07 0.38 0.36 0.21 0.15 0.28 0.52 0.44 0.09 * Excluding incorrectly identified. ** Preliminary and unweighted. 29 SCPC July 2015 - PRELIMINARY AND UNOFFICIAL Table 8: Expectations of Bitcoin exchange value. Table F Expectations of Bitcoin exchange value Percentage of those who are aware of Bitcoin, except as noted. Decrease a lot Decrease some Stay about the same Increase some Increase a lot Bitcoin (week) Bitcoin Adopters.................. VC Adopters (no Bitcoin).... VC Non-adopters................. 6.2 13.8 5.8 7.2 12.6 8.8 41.4 36.8 79.8 28.4 25.3 4.7 16.8 11.5 0.9 Bitcoin (month) Bitcoin Adopters.................. VC Adopters (no Bitcoin).... VC Non-adopters................. 2.2 6.9 5.2 9.6 10.3 13.5 35.9 40.2 71.8 35.5 23.0 8.8 16.8 19.5 0.5 Bitcoin (year) Bitcoin Adopters.................. VC Adopters (no Bitcoin).... VC Non-adopters................. 3.4 8.0 9.9 5.6 9.1 16.9 28.3 36.4 56.4 30.3 28.4 14.4 32.3 18.2 2.3 * All respondents in oversample are VC adopters. ** Preliminary and unweighted. 30 Table 9: OLS on Expectations (-2 to 2 scale). age 1825 age 2534 age 4554 age 5564 age o65 edu c pgs hh size white male inc lt25 inc 2549 inc 7599 inc gt100 investing res num otherpi adopt csh sh cc debt revolver cc rewards adopt smartphone adopt Constant Observations Rsquared Mean SD Week -0.177 -0.00741 -0.176∗∗ -0.217∗∗ -0.273∗∗∗ -0.0421 -0.0356 -0.0743 -0.0655 -0.173∗ -0.0452 0.0755 -0.0380 0.0767∗∗∗ 0.149∗∗∗ 0.110 0.0619 -0.129 0.0254 -0.670∗∗∗ 1017 .13 .124 .918 Month -0.259∗∗ 0.0716 -0.134 -0.251∗∗∗ -0.319∗∗∗ -0.0891 0.00389 -0.0828 -0.00608 -0.0628 -0.0290 0.0398 -0.0811 0.0896∗∗∗ 0.139∗∗∗ 0.209 -0.00684 -0.0818 0.0439 -0.763∗∗∗ 1018 .15 .202 .903 Sample: 2014 SCPC oversample and 2015 SCPC. ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01 31 Year 0.00198 0.0673 -0.304∗∗∗ -0.385∗∗∗ -0.497∗∗∗ -0.0359 -0.0206 -0.103 -0.0481 -0.0871 -0.0236 0.0528 -0.196∗∗ 0.102∗∗∗ 0.192∗∗∗ 0.364∗∗ -0.0183 -0.137 0.00951 -0.798∗∗∗ 1016 .18 .304 1.089 Table 10: Adoption of virtual currency. expectations week expectations week month expectations month year age 1825 age 2534 age 4554 age 5564 age o65 edu c pgs hh size white male inc lt25 inc 2549 inc 7599 inc gt100 billpay res shopping res budgeting res investing res num pi adopt csh sh cc debt revolver cc rewards adopt smartphone adopt Observations Pseudo R-squared N Positive AnyVC 0.0992∗∗∗ 0.128∗∗∗ 0.0842∗∗∗ 0.0158 0.0140 -0.109∗∗∗ -0.155∗∗∗ -0.356∗∗∗ -0.0356 0.0213∗∗∗ -0.116∗∗∗ 0.0426∗ 0.0481 -0.000761 -0.0260 -0.0432 -0.0124 0.0572∗∗∗ 0.00608 -0.0182 0.100∗∗∗ 0.159∗∗∗ -0.0874∗∗∗ -0.0927∗∗∗ 0.0134 1010 .5 592 Marginal effects Sample: 2014 SCPC oversample and 2015 SCPC ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01 32 Bitcoin 0.0934∗∗∗ 0.0875∗∗∗ 0.0911∗∗∗ 0.0462 0.0281 -0.0872∗∗ -0.112∗∗∗ -0.290∗∗∗ -0.0703∗∗∗ 0.0339∗∗∗ -0.0863∗∗∗ 0.0682∗∗∗ 0.0428 -0.0260 -0.0162 -0.0365 0.00205 0.0663∗∗∗ -0.000765 -0.00375 0.0847∗∗∗ 0.0574 -0.0957∗∗∗ -0.0781∗∗ 0.0506 1013 .4 502 Other 0.0901∗∗∗ 0.102∗∗∗ 0.0367∗∗ -0.0283 -0.00621 -0.108∗∗∗ -0.211∗∗∗ -0.420∗∗∗ -0.0315 0.0164∗∗ -0.0744∗∗∗ 0.0202 0.0871∗∗ 0.0456 0.0178 -0.0119 -0.00600 0.0156 0.000450 -0.0194 0.102∗∗∗ 0.152∗∗ -0.0482∗ -0.0606∗ -0.0195 1010 .38 483 SCPC July 2015 - PRELIMINARY AND UNOFFICIAL Table 11: Reasons for owning and not owning virtual currency. Table B Reasons for owning and not owning virtual currency Percentage of adopters. Oct 2014 SCPC 2015 SCPC Oversample Oct** Primary Secondary Primary Why Virtual Currency...................................................................... 100.0 100.0 I am interested in new technologies..................................................... It is an investment................................................................................ Payments............................................................................................. I use it to buy goods and services in the United States.................. It allows me to make payments anonymously................................ I use it to make remittances or other international payments......... It uses secure blockchain technology to prevent loss and fraud........... Distrust................................................................................................ I do not trust banks......................................................................... I do not trust the government or U.S. dollar................................... Other.................................................................................................... 23.8 19.2 37.5 16.1 13.7 7.7 6.2 10.3 5.7 4.5 2.9 21.3 14.0 38.6 13.7 16.4 8.5 12.9 12.4 7.2 5.3 0.8 Why Not Virtual Currency............................................................... I do not understand the technology...................................................... Not accepted for payment very often................................................... My current payment methods meet all of my needs............................. The U.S. dollar value of the virtual currency varies too much............. It is not guaranteed by the U.S. government........................................ It is not easy to acquire or use.............................................................. Other.................................................................................................... ** Preliminary and unweighted. 33 100.0 16.2 13.4 15.1 16.2 13.8 18.7 6.6 SCPC July 2015 - PRELIMINARY AND UNOFFICIAL Table 12: Virtual currency holdings. Table C Virtual Currency Holdings Dollars or coins per adopter, as specified. SCPC Tables 14-15. 2014 SCPC Oct Oversample Mean Median Mean Median Number of coins owned........................................................... 75 8 1180 9 Bitcoin.................................................................................. Other virtual currency*........................................................ Incorrectly identified..................................................... 74 1 112 7 1 20 75 1446 3 7 1551 1500 1425 321 Bitcoin.................................................................................. 1638 Other virtual currency*........................................................ 553 Incorrectly identified..................................................... 29 500 30 30 793 729 169 69 23539 629 26755 919 41 2 Value of VC owned (official exchange rate, "best")***....... 13794 604 Bitcoin.................................................................................. Other virtual currency*........................................................ Incorrectly identified..................................................... 15195 285 616 5 Value of virtual currency owned (owner estimate)............... Value of coins owned (official exchange rate)**.................... Bitcoin.................................................................................. 25345 2510 Other virtual currency*........................................................ Incorrectly identified..................................................... 2015 SCPC Oct Mean Median Addendum Implied reported exchange rates............................................ Bitcoin.................................................................................. 193 333 4151 9 -45 -9 1233 -97 % Exchange rate error............................................................ Bitcoin.................................................................................. * Excluding incorrectly identified. ** In oversample, includes only those that used a reference for number of coins. *** The criteria for "best" includes: (i) owner estimates of value with records, (ii) owner estimates of coins with records, converted using the official exchange rate, (iii) if no records were used, owner estimates of value if exchange rate error is less than 10%. If exchange rate error were greater than 10%, the observation is not kept. **** All numbers, other than the "best", are cut off at the 98 percentile. Use of records Percentage of adopters, except as noted. Oct 34 Used records Used records for the number of coins................................................ Used records for the value in U.S. dollars.......................................... Used records for both......................................................................... 2014 SCPC Jul 2015 90.4 33.3 27.4 29.7 2015 SCPC Oct SCPC July 2015 - PRELIMINARY AND UNOFFICIAL Table 13: Incidence of use and payee for virtual currency. Table D Incidence of use and payee for virtual currency Percentage of adopters. SCPC Tables 19-27. Incidence of use and payee Used in last 12 months.............................................................................................. Bitcoin.................................................................................................................. Other virtual currency*......................................................................................... Used in last month.............................................................................................. Bitcoin............................................................................................................ Other virtual currency.................................................................................... Payee Merchant.............................................................................................................. Person................................................................................................................... 2014 SCPC Oct Jul 2015 28.4 26.2 21.6 75.1 65.7 51.8 69.3 61.4 44.8 2015 SCPC Oct 39.0 56.3 Both...................................................................................................................... 25.1 * Excluding incorrectly identified Table E Number ad share of payments using virtual currency Number or percentage of payments Number of payments 2014 SCPC Oct (USC) Number of payments per month (if used last month) Bitcoin.................................................................................................................. Other virtual currency.......................................................................................... Number of payments per 12 months (if used last 12 months) Bitcoin.................................................................................................................. Other virtual currency.......................................................................................... Number of payments per month, Median (if used last month) Bitcoin.................................................................................................................. Other virtual currency.......................................................................................... 35 Number of payments per 12 months, Median (if used last 12 months) Bitcoin.................................................................................................................. Other virtual currency.......................................................................................... Jul 2015 7.5 4.7 7.6 3.1 3.2 2.8 0.5 0.1 0.4 0.2 0.2 0.3 2015 SCPC Oct Table 14: Use of virtual currency in last 12 months. Bitcoin -0.00893 0.0000641 0.0175 0.0734∗∗ 0.0462 -0.0166 0.0650 0.0623∗∗∗ -0.0478 -0.0134 -0.0304 0.0594 0.148∗∗∗ 0.128∗∗ 0.134∗∗∗ 0.0997∗∗ 0.126∗∗ 0.0549 0.127∗ 479 .18 399 age agesq edu c pgs white male log income employed num otherpi adopt csh sh cc debt revolver cc rewards adopt smartphone adopt whyvc buygands whyvc remitintl whyvc anonymous whyvc investment whyvc blockchain whyvc notrustinbanks whyvc notrustingovt Observations Pseudo R-squared N Positive OtherVC 0.00609 -0.0000701 0.0153 -0.0552 0.0542 0.0157 0.00427 0.0929∗∗∗ -0.0303 0.0265 -0.0344 0.155∗∗ 0.135∗∗∗ 0.259∗∗∗ 0.196∗∗∗ 0.106∗∗ 0.0310 0.0830 0.120∗ 447 .2 326 Marginal effects Sample: 2014 SCPC with oversample, 2015 SCPC ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01 36 Table 15: Respondent virtual currency payees. age u25 age 2534 age 4554 age 5564 age o65 edu c pgs white male inc lt25 inc 2549 inc 7599 inc gt100 employed num otherpi adopt csh sh cc debt revolver cc rewards adopt smartphone adopt whyvc buygands whyvc remitintl whyvc anonymous whyvc investment whyvc blockchain whyvc notrustinbanks whyvc notrustingovt Observations Pseudo R-squared N Positive Paid merchant -0.139∗ 0.0000548 0.000474 -0.160 -0.140 0.0269 -0.0285 0.0718∗ -0.0584 -0.0712 -0.0905 -0.0549 -0.00330 0.0975∗∗∗ 0.153 -0.0204 -0.0179 0.134∗ 0.112∗∗ 0.132∗∗ 0.0596 0.0813 0.0286 -0.0363 0.0288 578 .12 246 Marginal effects Sample: 2014 SCPC with oversample, 2015 SCPC ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01 37 Paid person 0.103 0.0481 -0.0483 -0.0658 -0.0291 0.0186 0.0468 0.0268 0.00620 -0.0391 0.131∗∗ -0.0214 -0.0145 0.0855∗∗∗ 0.0291 0.122∗∗∗ -0.0582 0.0730 0.0856∗ 0.162∗∗∗ 0.137∗∗∗ 0.0300 0.0768 0.0851 0.0418 577 .16 332 Paid both -0.0600 0.00365 -0.0469 -0.108 -0.182 0.0428 -0.0161 0.0269 -0.0145 -0.0634 0.0220 -0.0428 -0.0230 0.107∗∗∗ 0.140 0.0698∗ 0.0951 -0.00376 0.0352 0.0576 0.0276 0.0510 0.0241 -0.0336 0.00426 572 .23 146 Table 16: Share of use Heckman regression. main male age log income num pi adopt csh sh cc debt revolver cc rewards adopt smartphone adopt whyvc buygands whyvc remitintl whyvc anonymous whyvc investment whyvc blockchain whyvc notrustinbanks whyvc notrustingovt Constant adopt coin expectations week expectations week month expectations month year male age log income num pi adopt csh sh cc debt revolver cc rewards adopt smartphone adopt Constant mills lambda Observations vc sh bitcoin sh othervc sh 0.0187 -0.00193∗∗ -0.00114 0.00497 -0.129∗∗∗ 0.0449∗∗∗ -0.0223 0.0227 0.0427∗∗ 0.0346 0.0144 -0.0174 0.0200 -0.0237 0.0257 0.119 0.00676 -0.000953 -0.000866 -0.0137∗∗ -0.146∗∗∗ 0.0317∗∗ -0.00493 0.00785 0.0343∗∗ 0.0332∗∗ -0.00369 -0.0119 0.00280 -0.00668 0.00714 0.215∗∗∗ 0.0321∗∗ -0.00169∗ 0.00428 0.0147 -0.0378 0.0426∗∗∗ -0.0199 0.00438 0.0356∗ 0.0182 0.0284 -0.00864 0.0192 -0.00843 0.0348 -0.0752 0.565∗∗∗ 0.716∗∗∗ 0.384∗∗∗ 0.110 -0.0348∗∗∗ -0.139∗∗ 0.538∗∗∗ 0.832∗∗∗ -0.370∗∗∗ -0.568∗∗∗ 0.0433 -0.00235 0.441∗∗∗ 0.479∗∗∗ 0.386∗∗∗ 0.148 -0.0267∗∗∗ -0.0499 0.401∗∗∗ 0.348 -0.345∗∗∗ -0.508∗∗∗ 0.194 -0.892 0.403∗∗∗ 0.458∗∗∗ 0.132∗∗ 0.0345 -0.0297∗∗∗ -0.134∗∗ 0.438∗∗∗ 0.710∗∗∗ -0.179∗ -0.320∗∗∗ -0.0158 -0.167 0.0103 1079 -0.0112 1082 0.0281 1084 Sample: 2014 SCPC oversample and 2015 SCPC ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01 38 Bitcoin Price (USD/coin) Figure 1: Bitcoin price and search interest (Google trends). 1200 1200 1000 1000 800 800 Bitcoin Bitcoin search search interest interest (right (right axis) axis) 600 Bitcoin Bitcoin price price (left axis) (left axis) 600 400 400 200 200 0 2009 2009 2010 2010 2011 2011 2012 2012 20132013 2014 2014 YearYear Note: Bitcoin search interest is based on US Google Trends, scaled to overlay with prices. 39 2015 2015 0 2016 2016 Number of Virtual Currencies (including 112 defunct) Figure 2: Number of virtual currencies and cumulative search interest. 800 800 700 700 600 600 500 500 400 Number Number of of Virtual Virtual Currencies Currencies 400 (left axis) (left axis) Cumulative Cumulative search search interest interest (right (right axis) axis) 300 300 200 200 100 100 0 2009 2009 2010 2010 2011 2011 2012 2012 2013 2013 20142014 20152015 0 2016 2016 Year Year Note: Cumulative search interest is based on US Google Trends for Bitcoin, scaled to overlay with number of coins. 40 Percent of consumers aware of Bitcoin Figure 3: Bitcoin awareness in a diffusion model. 100 100 90 90 80 80 70 70 60 60 50 50 Data Data on awareness on awareness (SCPC) (SCPC) 40 40 Diffusion Diffusion(Bass) (Bass) 30 30 20 20 10 10 0 2007 0 2010 2010 2013 2013 2016 2016 Year Year 41 2019 2019 20222022 2025 2025 A. Description of variables Table 17: Description of independent variables. Variable age agesq age u25 age 2534 age 4554 age 5564 age o65 edu c pgs hh size white male log income inc lt25 inc 2549 inc 7599 inc gt100 employed num pi adopt csh sh cc debt revolver cc rewards adopt smartphone adopt expectations week num expectations month num expectations year num Description Age of respondent Age of respondent squared Indicator variable for age of respondent <25 Indicator variable for age of respondent >= 25 and <= 34 Indicator variable for age of respondent >= 45 and <= 54 Indicator variable for age of respondent >= 55 and <= 64 Indicator variable for age of respondent >= 65 Indicator variable for education of respondent at the college or postgrad level Respondent household size Indicator variable for respondent race (white) Indicator variable for respondent gender (male) Log of respondent family income Indicator variable for respondent family income <$25k Indicator variable for respondent family income >=$25k and <=49k Indicator variable for respondent family income >=$75k and <=99k Indicator variable for respondent family income >=$100k Indicator variable for whether the respondent is employed or not The respondent’s number of payment instruments adopted (out of 9) The share of respondent’s cash payments (out of all payments) by number Indicator variable for whether the respondent revolves on credit card debt Indicator variable for whether the respondent adopted a rewards credit card Indicator variable for whether the respondent adopted a smartphone Respondent’s expectations of Bitcoin exchange rate growth/decline over the next week (1-5) Respondent’s expectations of Bitcoin exchange rate growth/decline over the next month (1-5) Respondent’s expectations of Bitcoin exchange rate growth/decline over the next year (1-5) 42 Table 18: Description of dependent variables. Variable vc error Paid merchant Paid person Paid both vc sh bitcoin sh othervc sh Description Indicator variable for respondent mis-identifying virtual currency Indicator variable for whether the respondent used virtual currency to pay a merchant Indicator variable for whether the respondent used virtual currency to pay another person Indicator variable for whether the respondent used virtual currency to pay both a merchant and another person The share of respondent’s virtual currency payments (out of all payments) by number The share of respondent’s bitcoin payments (out of all payments) by number The share of respondent’s other virtual currency payments by number 43 References Bass, Frank. 1969. “A new product growth for model consumer durables.” Management Science 15: 215–227. Baumol, William J. 1952. “The transactions demand for cash: An inventory theoretic approach.” The Quarterly Journal of Economics 545–556. ¨ Bohme, Rainer, Nicolas Christin, Benjamin Edelman, and Tyler Moore. 2015. “Bitcoin: Economics, Technology, and Governance.” The Journal of Economic Perspectives 29(2): 213–238. Bolt, Wilko, and Maarten van Oordt. 2015. “On the Value of Virtual Currencies.” Working paper. Briglevics, Tam´as, and Scott Schuh. 2013. “U.S. Consumer Demand for Cash in the Era of Low Interest Rates and Electronic Payments.” Working paper series. Federal Reserve Bank of Boston. Christin, Nicolas. 2013. “Traveling the Silk Road: A Measurement Analysis of a Large Anonymous Online Marketplace.” In Proceedings of the 22Nd International Conference on World Wide Web, WWW ’13, 213–224. International World Wide Web Conferences Steering Committee. ISBN 978-1-4503-2035-1. CoinDesk. 2015. “Who really uses Bitcoin?” Tech. rep. Donier, J., and J.-P. Bouchaud. 2015. “Why Do Markets Crash? Bitcoin Data Offers Unprecedented Insights.” ArXiv e-prints. Dwyer, Gerald P. 2015. “The economics of Bitcoin and similar private digital currencies.” Journal of Financial Stability 17: 81 – 91. ISSN 1572-3089. Special Issue: Instead of the Fed: Past and Present Alternatives to the Federal Reserve System. European Central Bank. 2012. “Virtual currency schemes.” Tech. rep. Glaser, Florian, Kai Zimmermann, Martin Haferkorn, Moritz Christian Weber, and Michael Siering. 2014. “Bitcoin-Asset or Currency? Revealing Users’ Hidden Intentions.” SSRN Working Paper. 44 Greene, Claire, and Oz Shy. 2014. “E-cash and virtual currency as alternative payment methods.” Journal of Payments Strategy and Systems 8(3): 274–288. Halaburda, Hanna, and Neil Gandal. 2014. “Competition in the Cryptocurrency Market.” Working paper. Hencic, Andrew, and Christian Gouriroux. 2015. “Noncausal Autoregressive Model in Application to Bitcoin/USD Exchange Rates.” In Econometrics of Risk, eds. Van-Nam Huynh, Vladik Kreinovich, Songsak Sriboonchitta, and Komsan Suriya, vol. 583 of Studies in Computational Intelligence, 17–40. Springer International Publishing. ISBN 978-3-319-13448-2. Lo, Stephanie, and J Christina Wang. 2014. “Bitcoin as money?” Working paper series. Federal Reserve Bank of Boston. ¨ ¨ Moser, Malte, and Rainer Bohme. 2015. “Trends, Tips, Tolls: A Longitudinal Study of Bitcoin Transaction Fees.” In Financial Cryptography and Data Security, eds. Michael Brenner, Nicolas Christin, Benjamin Johnson, and Kurt Rohloff, vol. 8976 of Lecture Notes in Computer Science, 19–33. Springer Berlin Heidelberg. ISBN 978-3-662-48050-2. Nakamoto, Satoshi. 2008. “Bitcoin: A Peer-to-Peer Electronic Cash System.” Polasik, Michal, Anna Iwona Piotrowska, Tomasz Piotr Wisniewski, Radoslaw Kotkowski, and Geoffrey Lightfoot. 2015. “Price Fluctuations and the Use of Bitcoin: An Empirical Inquiry.” International Journal of Electronic Commerce 20(1): 9–49. Saito, Tetsuya. 2015. “Chapter 12 - A Microeconomic Analysis of Bitcoin and Illegal Activities.” In Handbook of Digital Currency, ed. David Lee Kuo Chuen, 231 – 248. San Diego: Academic Press. ISBN 978-0-12-802117-0. Schuh, Scott, and Joanna Stavins. 2010. “Why are (some) consumers (finally) writing fewer checks? The role of payment characteristics.” Journal of Banking & Finance 34(8): 1745–1758. Schuh, Scott, and Joanna Stavins. 2012. “How Consumers Pay: Adoption and Use of Payments.” Working paper series. Federal Reserve Bank of Boston. 45 Schuh, Scott, and Joanna Stavins. 2014. “The 2011 and 2012 Surveys of Consumer Payment Choice.” Research data report. Federal Reserve Bank of Boston. Schuh, Scott, and Joanna Stavins. 2015. “How Do Speed and Security Influence Consumers’ Payment Behavior?” Public policy discussion paper. Federal Reserve Bank of Boston. Tobin, James. 1956. “The interest-elasticity of transactions demand for cash.” The Review of Economics and Statistics 241–247. Tsanidis, Charalambos, Dafni-Maria Nerantzaki, George Karavasilis, Vasiliki Vrana, and Dimitrios Paschaloudis. 2015. “Greek consumers and the use of Bitcoin.” The Business & Management Review 6: 295–302. Velde, Franc¸ois. 2013. “Bitcoin: A primer.” Chicago fed letter. Federal Reserve Bank of Chicago. 46