The future of cryptofinance: An Empirical Analysis of the Adoption of Bitcoin
1. 252 FINANCIEEL FORUM / BANK- EN FINANCIEWEZEN 2017/3 LARCIER
VARIA / DIVERS
Liesje DEGRAEVE
Postgraduate in Business
Economics, IÉSEG School
of management
Niki DE KEGEL
Advanced Master of
Taxation, KU Leuven
The future of cryptofinance:
an empirical analysis of the
adoption of Bitcoin
The impact of country determinants on
the adoption rate of Bitcoin
In the existing literature, little research has been conducted on the adoption of Bitcoin and the factors
driving or preventing this adoption. This paper adds to the literature by using country-level determi-
nants to investigate the different drivers and barriers that influenced the adoption rate of Bitcoin
during the period 2011-2015. Our findings demonstrate that the cryptocurrency Bitcoin is used to a
lesser extent in corrupt countries. Furthermore, we find that both the inflation rate and the efficiency
of the banking system in a country act as a barrier to the adoption of Bitcoin while the occurrence of
many non-cash transactions tends to drive the adoption rate. When we divide the sample into four
subsets based on the level of gross national income of a country, our findings reveal similar results for
certain country determinants but for most of them, we find some surprising results. We now establish
that the better the accessibility to the internet and the more open a country is to globalization, the
higher the adoption rate of the cryptocurrency. Previously, these determinants did not seem to be
important drivers since no significant impact on the adoption rate of Bitcoin could be found. In addi-
tion, we find unexpected results for the impact of financial inclusion and the degree of non-cash
utilization on the adoption rate since the sign of the beta estimate of the variables changes over dif-
ferent subsets. It is therefore not clear what the true relationship is between these country determinants
and Bitcoin’s adoption rate1
.
1. Introduction
Cryptocurrencies, and more specific
Bitcoins, have gained popularity over the
years (Schuh et al., 2015). Nevertheless,
many people have heard about the concept
of these digital currencies but do not know
that it is more than an alternative to the
national currencies we use every day.
Bitcoin is the first completely decentralized
cryptocurrency in the world and has the
potential to revolutionize online payment
systems in such a way that it benefits
consumers and merchants (Brito et al.,
2013). The currency and its underpinning
technology blockchain offer many inter-
esting features.
One of the features of the Bitcoin block-
chain is that it works with a peer-to-peer
model. The technology doesn’t require
intermediaries, which makes it possible to
go from an “access economy” with compa-
nies such as AirBnB and Uber to a real
sharing economy where people can just
transfer money from A to B without going
to a bank or needing a platform such as
AirBnB to connect (Lundy, 2016).
Another feature is that Bitcoin allows
people to do microtransactions, which
might be interesting for the sharing
economy as well. Nowadays, you cannot
do a micropayment with your credit card
without being subject to higher fees. In
addition, it is not possible to do transac-
tions below one euro cent. In the Bitcoin
world, however, you can. The blockchain
technology is developed with the intention
1. This article is a summary of our master thesis, written in cooperation with Prof. Dr. H. Degryse. The full text is available
upon request, please email liesje.degraeve@hotmail.be.
2. LARCIER FORUM FINANCIER / REVUE BANCAIRE ET FINANCIÈRE 2017/3 253
LIESJE DEGRAEVE & NIKI DE KEGEL THE FUTURE OF CRYPTOFINANCE: AN EMPIRICAL ANALYSIS OF THE ADOPTION OF BITCOIN
to allow you to transfer for example one euro cent at a
very low fee (Dramaliev, 2016). Unfortunately, those low
transaction costs are just a utopian dream at the moment
since the current average transaction fee is about 1.8 euro,
which is more than the daily income of one person in
several (developing) countries (Blockonomics, 2017). In
order to prevent these unreasonably high costs, several
scaling solutions are suggested. These new layers are being
built as an extension of the Bitcoin protocol. One of them
is called the Lightning Network. This network will
become operational this year and will facilitate one billion
settled transactions per second at a cost of almost zero.
When we make a comparison with Visa, we see that its
capacity is 30,000 transactions per second. In addition,
they charge very high transaction fees making microtrans-
actions impossible (Francis, 2017).
Unfortunately it is not all roses, Bitcoin is often associated
with the dark web and more specific the online market-
place previously known as Silk Road. Unlike other online
marketplaces, this black market is used to trade illicit
drugs, weapons and other illegal goods and services. The
platform ensures anonymity of buyers and sellers by using
Bitcoin to perform transactions. Due to this anonymity,
the US government had no control over the criminal busi-
nesses taking place in this deeper layer of the internet
(Christin, 2012; Raeesi, 2015). In October 2013, the FBI
succeeded in shutting down Silk Road. Federal and
national governments are aware of the advantages and
possible future role of cryptocurrencies such as Bitcoin in
becoming a worldwide monetary payment system.
However, scandals as Silk road indicate that there is still a
long way ahead for governments to regulate and control
digital currencies (Kleiman, 2013). Next to the issues
regarding regulation, people might be deterrent to adopt
cryptocurrencies (Barski et al., 2014).
Bitcoin has become a market leader in the world of
cryptocurrencies thanks to its first mover advantage. The
value of the currency has shown a significant and steady
exponential growth over time with periodical peaks. The
volatile price has fluctuated substantially, ranging from
$13 at the beginning of January 2013 up to $1,708 at the
beginning of May 2017 (see Figure 1). The number of
Bitcoin users doubles each year and in 2016 there were
already over ten million Bitcoin wallets (Coindesk.com,
2016). Users of Bitcoin and other cryptocurrencies are
scattered across the world. In some countries, this digital
currency is used to a lesser extent than in others due to
factors driving this adoption. In Belgium for example, the
adoption rate is rather low compared to that of its neigh-
bouring country the Netherlands. Belgium’s most Bitcoin-
friendly city Ghent only registered 110 transactions in
2015 (Gentbitcoinstad.be, 2015). It is on the basis of our
research question that we might be able to find out why
Belgium has such a low adoption rate at the moment.
2. Literature review
2.1. Factors driving bitcoin adoption
At first, especially early adopters and innovators showed
interest in cryptocurrencies. Over the years, major key
players have started to adopt these digital currencies,
which led to a user group consisting of multinational
corporations, retailers and financial leaders (Yosupov,
2015). Trust, habit, price and performance expectancy
play a significant role in the decision or intention to adopt
Bitcoin in small and medium sized businesses. Moreover,
consumers who expect that the currencies will appreciate
in value are more likely to demand them suggesting that
they are requested as a financial investment (Roos, 2015).
Bohr et al. (2014) adds to the findings of Yosupov (2015)
and Roos (2015) that younger and older users are less
optimistic towards Bitcoin than users in their late 30s.
Puri (2016) is one of the first to study Bitcoin adoption
rates at countrywide level. He analysed the impact of
public interest in Bitcoin on countrywide adoption rates
and Bitcoin prices using the number of Bitcoin clients
downloaded as a proxy for adoption rates and the amount
of Google searches for the keyword “Bitcoin” as a proxy
for public interest. The findings of this study indicate that
the number of search volumes at country level have a
significant impact on the number of client downloads per
capita for that country. However, the magnitude of this
impact is decreasing every month.
Google Trends is used in many studies concerning the
adoption of Bitcoin and other cryptocurrencies. Yelowitz
et al. (2014) collects Google Trends query data with the
aim of relating the interest in Bitcoin and its possible
users. He divides Bitcoin users into four segments: crimi-
nals, computer programming enthusiasts, speculative
investors and libertarians. His results show that the first
two segments are positively correlated with Bitcoin
interest while there is no or limited support for investment
and political motives. Other studies show that libertarians
associate the decentralized structure of Bitcoin with indi-
vidualism. This structure enables them to escape from the
power of financial organizations and institutions. They
state freedom as their main reason for their positive atti-
tude towards Bitcoin (Bohr et al., 2014). According to
Bashir et al. (2016) libertarians are more likely to own
3. 254 FINANCIEEL FORUM / BANK- EN FINANCIEWEZEN 2017/3 LARCIER
THE FUTURE OF CRYPTOFINANCE: AN EMPIRICAL ANALYSIS OF THE ADOPTION OF BITCOIN LIESJE DEGRAEVE & NIKI DE KEGEL
Bitcoins compared to centrist and people who have beliefs
that are at the left of the political spectrum. While
anonymity is associated with individualism by libertar-
ians, it is seen as a threat by financial regulators and law
enforcement authorities. However, individuals trying to
escape financial surveillance consider it as an important
driver (Bashir et al., 2016). This raises the question
whether Bitcoin could be a financial solution for devel-
oping economies. Since it might be a way of avoiding
corrupt government influence in the financial institutions.
As stated by Polasik et al. (2015) merchants in countries
with a low GDP per capita are more open to accept
cryptocurrencies as a payment method, which suggests
that it is considered as a medium to circumvent govern-
ment restrictions. This paper however does not contain
details about consumers who paid these merchants in
Bitcoins.
Furthermore, Darlington (2014) argues that countries
with a high corruption and counterfeiting rate, unstable
monetary policies and many inhabitants with no access to
safe financial institutions are the ones that are more likely
to adopt Bitcoin. However, these struggling economies are
technologically underdeveloped and not all of their inhab-
itants have access to the internet through a personal
computer, which makes it difficult for them to adopt the
cryptocurrency. Still, the perks of electronic payment
systems lead to greater financial inclusion and social bene-
fits in these developing countries. Due to the very low
transaction fees, Bitcoin is considered as affordable to the
people in these nations (Clegg, 2014).
2.2. Barriers to bitcoin adoption
Schuh et al. (2015) provides the first representative
evidence on consumer adoption and the use of cryptocur-
rencies in the United States. The findings of the study indi-
cate that barely half of the US adult population is aware
of the existence of Bitcoin and other cryptocurrencies.
Moreover, consumers who are aware of these currencies
report that they are unfamiliar with them, which might
explain the low adoption rate of Bitcoin and other crypto-
currencies.
Bitcoin is perceived as a virtual currency with a lack of
user-friendliness (Spenkelink, 2015). Baur et al. (2015)
adds that the virtual currency has future potential as a
payment method partly thanks to its decentralised struc-
ture. However, the perceived ease of use among stake-
holders is still relatively low making it necessary to
educate people about the use of Bitcoin and its advan-
tages. Besides a low perceived ease of use, Conolly et al.
(2015) finds that the lack of market trust acts as another
barrier to Bitcoin adoption. Every user is able to observe
the Bitcoin balance of every other user, whereby he is
unable to observe the identity of that user. This informa-
tion, however, is not observable in the case of private bank
accounts. Due to this visibility, the accounts may be
targets of thieves. In his paper, he refers to the example of
the Mt. Gox Bitcoin exchange that went bankrupt after
more than 700,000 Bitcoins worth nearly half a billion
dollars went missing.
Furthermore, Bitcoin prices have fluctuated considerably
since its inception and the market does not show much
interest in unstable currencies (Scott, 2014). Due to its
extreme volatility, Bitcoin cannot (yet) be perceived as a
reliable store of value. Merchants who accept Bitcoin as a
payment method immediately exchange the coin for a
traditional currency, because of its riskiness (Harvey,
2014).
According to Luther (2015) one of the main explanations
for the non-acceptance of Bitcoin is the existence of
switching costs and network effects. These switching costs
are the transition costs that occur while switching from
traditional monetary payment systems to Bitcoin. One
example of these transition costs are the costs associated
with the retooling of vending machines and automatic
teller machines. Miles (2014) argues that central and
commercial banks spend loads of money on maintenance
costs at the moment. These maintenance costs can be
saved when governments start to use cryptocurrencies,
which frees up money to cover the switching costs. Plas-
saras (2013) refers to network effects as positive external-
ities whereby the more people start accepting Bitcoin, the
more valuable the virtual currency becomes. However, at
the moment, these network effects for Bitcoin are not
large enough in comparison to those of the traditional
currencies. When all other barriers are resolved,
consumers and merchants still need to be persuaded to use
Bitcoin as their main method of payment for Bitcoin to
become a globally used currency.
In the past, successful monetary transitions did occur in
several countries, but all of these transitions occurred as a
consequence of government involvement and/or economic
instability (hyperinflation). Currently, these two compo-
nents are not present on a large scale, which might explain
why the currency is not accepted worldwide even when
people consider Bitcoin as a possible alternative to the
traditional monetary system (Luther, 2013). Hauschildt
(2012) states that the major barrier for the adoption of
Bitcoin as a worldwide currency is the lack of regulation,
4. LARCIER FORUM FINANCIER / REVUE BANCAIRE ET FINANCIÈRE 2017/3 255
LIESJE DEGRAEVE & NIKI DE KEGEL THE FUTURE OF CRYPTOFINANCE: AN EMPIRICAL ANALYSIS OF THE ADOPTION OF BITCOIN
which makes it unstable when challenges pop up. Plas-
saras (2013) supports this and puts the International
Monetary Fund forward as the potential regulator in
order to avoid an economic disaster. In addition, the
Bitcoin system does not facilitate commercial loans or
borrowing options, which renders the Bitcoin system less
attractive.
However, Bitcoin provides solutions to many problems. A
decentralised structure, low transaction costs, microtrans-
actions, security and no counterfeiting make Bitcoin a
virtual currency with a lot of potential. However, this will
not guarantee a successful future for Bitcoin. The digital
currency will have to change to be able to survive. But
there is no doubt that the current financial structure will
be disrupted in the future (Harvey, 2014).
3. Methodology
3.1. Hypotheses
In this section, we develop a proper hypothesis for all
explanatory variables included in our regression.
H1: The corruption rate is positively related to the adop-
tion rate of Bitcoin. We argue that, in countries with a
high corruption rate, people might be more willing to use
Bitcoin as an alternative to their domestic currency whose
value is impacted by the policies of the domestic national
bank.
H2: Citizens of countries with a low financial inclusion
are more likely to adopt Bitcoin to a larger extent. We
assume that citizens of countries with a low financial
inclusion rate will be more eager to adopt Bitcoin on a
larger scale since it avoids trusted third parties as banks or
clearing houses (Nakamoto, 2008; Darlington, 2014).
H3: If the inhabitants of a country can not easily access
the internet, they will adopt Bitcoin to a lesser extent.
Since Bitcoin is a digital currency, you need internet access
to be able to buy coins and transfer them to others. This
is why citizens of countries where access to internet is rare
or very expensive, might not be as motivated to adopt
Bitcoin as a currency as citizens of countries where
internet can be accessed easily.
H4: The degree to which a country is engaged in interna-
tional trade is positively associated with the adoption rate
of Bitcoin among the inhabitants of that country. More
and more companies are engaging in international trade,
which means that they import and export goods and serv-
ices from and to companies, customers or others in
foreign countries. Using financial institutions to transfer
money from one country to another can be relatively
expensive and risky at the same time. When companies
are in need of cash, they might use factoring in order to
receive the payment of the buyer of the goods or services
before the payment was executed. In return, the supplier
of the goods or services needs to pay a factoring fee, which
is higher than the interest rates charged by commercial
banks since the factor bears the default risk. Factoring
financing is often provided by unconventional lenders,
which makes it an expensive and risky alternative.
Making use of the blockchain of Bitcoin, however, reduces
the risk of fraud and duplication of payments and allows
users to transfer money at low transaction fees (Harris,
2016).
H5: Residents of countries regulated by unstable mone-
tary policies have a greater likelihood of choosing Bitcoin.
This might result in a higher adoption rate of the crypto-
currency. In countries with high inflation rates, inhabit-
ants look for a safe-haven asset since the country’s
currency becomes worthless. Another option is to buy
dollars or other strong currencies. Since you can only
purchase those through financial institutions, they might
not be eager to give you access to these currencies in order
to prevent the situation in the country from getting worse.
This is why purchasing Bitcoins might be a better alterna-
tive (Darlington, 2014).
H6: The fraction of the population in a country that
makes use of debit cards is positively associated with the
fraction of the population in that country that makes use
of Bitcoin as a means of payment. People using debit cards
instead of cash to pay for goods or services, might be more
likely to adopt the cryptocurrency Bitcoin since they are
already familiar with doing online transactions. This
might facilitate the transition to Bitcoin for them.
H7: The efficiency of a country's banking system is nega-
tively related to the extent to which inhabitants of that
country have adopted Bitcoin as a means of payment.
Banks that work in an efficient way might be more likely
to gain the trust of the clients they are serving. These
clients might have no incentives to switch to another
payment system such as Bitcoin.
3.2. Data
To conduct our research, we need country-level data on all
our dependent and independent variables. In order to
obtain a complete dataset with the necessary information
for our analysis, we combine data from different data
sources. In our study, we focus on the period 2011 to
5. 256 FINANCIEEL FORUM / BANK- EN FINANCIEWEZEN 2017/3 LARCIER
THE FUTURE OF CRYPTOFINANCE: AN EMPIRICAL ANALYSIS OF THE ADOPTION OF BITCOIN LIESJE DEGRAEVE & NIKI DE KEGEL
2015. However, we make use of lagged explanatory vari-
ables, which implies that we lag the data of our inde-
pendent variables by one year. For the independent varia-
bles, we use data on the years 2011 to 2014 and for the
dependent variable, we look at the period 2012 to 2015.
These lagged explanatory variables are used because their
effect on the dependent variable might not be realised
immediately. Thus, we take into account the possibility of
reverse causality.
The complete sample for the stated time period consists of
country-level data on 237 countries. However, we make
some modifications to ensure a reliable sample, which
leaves us with a smaller final sample. Countries for which
only limited data is available, are removed from the
sample. Furthermore, we use linear interpolation to
replace most of our missing values. When data is available
for at least two out of four years, the missing values for
the other year(s) are replaced. To make sure our results are
not driven by this extrapolation, we will run the regres-
sion with the missing values and check whether our results
remain more or less the same. Finally, outliers are filtered
in the original dataset, meaning that we delete the outliers
from the 1st
and 99th
percentile of the distribution for all
our variables. Our final sample consists of 468 observa-
tions and contains information for 117 countries.
For our dependent variable, we use the number of client
downloads as a proxy for the adoption rate of Bitcoin.
Before someone is able to start trading Bitcoins, he either
must download a client on his computer or mobile phone
that implements the Bitcoin protocol or create an account
on a website that runs the client for him. The client is used
for saving the user’s bitcoins in a file called a wallet (Grin-
berg, 2011).
In this paper, we focus on the clients that are downloaded
on a computer or mobile phone as this is considered the
most secure option. By installing the software, Bitcoins
can be stored offline. This protects users from anything
malicious (Watson, 2014). We assume that the proportion
of people who prefer safety, in other words who prefer the
safer desktop client or the mobile client over the web
client, is similar in all countries. Therefore, this is likely to
be a good proxy for Bitcoin usage. Data for this variable
is obtained from Sourceforge. This is an Open Source
community resource that helps people with open source
software development and distribution. Sourceforge
provides us a sample of 236 countries for which the
number of client downloads is available. The number of
client downloads across all operating forms such as
Android, Macintosh and Windows is stated as a number
per year. Since the client needs to be downloaded on the
device once, this number can be seen as the increase in the
number of users in a certain country per year. We are,
however, interested in the adoption rate and not in the
increase in the number of users per year. Therefore, we
accumulate the number of client downloads each year to
get a clear image of the real adoption rate. In the end, we
divide the cumulative numbers by the population of the
country and multiply by hundred in order to have a
percentage relative to the size of a country in terms of
population.
Data on the explanatory variables “non-cash utilization”,
“banking system efficiency”, “inflation”, “internet
access”, “international trade” and the control variables
“GDP per capita”, “unemployment”, “GNI per capita”
and “population” are obtained from different databases
and surveys made available by the World Bank.
Information on the inflation rate, the possibility to access
internet, international trade, the unemployment rate, the
GDP per capita, the GNI per capita and the population of
a certain country is retrieved from the World Develop-
ment Indicators database (WDI). This database provides
accurate global development data including global,
national and regional estimates.
Statistics on our variable “financial inclusion” are
obtained from the Financial Access Survey (FAS) offered
by the IMF. This database contains information on access
to and use of financial services by households and firms.
The last variable is the variable 'corruption', which is
measured by observations from the Corruption Percep-
tion Index (CPI) of Transparency International.
In the second part of our analysis, we will focus on
different subsamples of our original sample. We divide
our primary dataset into four subsets: high income coun-
tries, upper middle income countries, lower middle
income countries and low income countries.
3.3. Model
In order to test our hypotheses, we based ourselves on the
model constructed by Puri (2016) since this was the first
paper that studied the impact of country determinants on
the adoption rate. Our own modified model looks as
follows:
Adoption rateit = β0 + β1 Corruptioni(t-1) + β2 Financial
Inclusioni(t-1) + β3 Inflationi(t-1) + β4 Internet Accessi(t-1)
+ β5 Non Cash Utilizationi(t-1) + β6 International Tradei(t-
1) + β7 Bank System efficiencyi(t-1) + β8 GDP per
capitai(t-1) + β9 Populationi(t-1) + β10 Unemploymenti(t-
1) + εi(t-1)
6. LARCIER FORUM FINANCIER / REVUE BANCAIRE ET FINANCIÈRE 2017/3 257
LIESJE DEGRAEVE & NIKI DE KEGEL THE FUTURE OF CRYPTOFINANCE: AN EMPIRICAL ANALYSIS OF THE ADOPTION OF BITCOIN
The dependent variable in our model is the adoption rate
it. It measures the adoption rate of Bitcoin in a country i
at time t. We assume that this adoption rate can be
captured for each country by the number of Bitcoin
desktop or mobile client downloads across all operating
platforms such as Android, Macintosh and Windows.
In this model, we include country-specific and time-
varying variables and analyze what their influence is on
the adoption of Bitcoin. When we lag these country-
specific variables by one year, we account for reverse
causality. Corruption i(t-1)is an explanatory variable for
the degree of corruption for a specific country i at time t-
1. Financial inclusion i(t-1)measures the extent to which
financial services and products that fit the needs of the
people are available in a country i at time t-1. The variable
inflation i(t-1) is expressed as an annual percentage change
in consumer prices as compared to consumer prices of the
previous year in a country i at time t-1. Internet Access i(t-
1) reflects the percentage of the population in a country i
at time t-1 that has access to the internet, which is an
essential requirement in order to be able to use Bitcoin.
The amount of people that possess a debit card is captured
by the variable Non-cash Utilization i(t-1). The inde-
pendent variable International trade i(t-1) is a country-
specific variable that captures the number of products and
services imported and exported, expressed as a percentage
of a country’s GDP at time t-1. The last explanatory vari-
able included in our regression is Bank System Efficiency
i(t-1). How efficient a bank works can be captured by the
cost-to-income ratio of that bank. In our research, the
bank system efficiency of a country i at time t-1 is the
average of the cost-to-income ratios of all the banks oper-
ating in that particular country.
Besides the explanatory variables explained above, we
include different country-specific control variables: GDP
per capita, Population and Unemployment. GDP per
capita i(t-1)is a measure of the standard of living in a
particular country i at time t-1. It tells us how prosperous
a country is. We include this variable in our regression to
control for differences in prosperity levels across coun-
tries. The variable Population i(t-1) is included to control
for the size of a country. When we would not do this, our
results would be biased since countries with more citizens
and thus more potential users are able to reach a higher
adoption rate. In addition, we also control for Unemploy-
ment i(t-1).
At first, we run the baseline regression stated at the begin-
ning of this section. Afterwards, we divide our primary
dataset into four subsets, which are datasets containing
solely observations of respectively high income countries,
upper middle income countries, lower middle income
countries and low income countries. In order to make the
distinction between the subsets, we use income thresholds
based on the GNI per capita that are provided by the
World Bank. Since one threshold is given for each year
and our independent variables cover a period of four years
(2011 to 2014), the average of the thresholds for the four
years is taken to simplify the model. We run the baseline
regression for these four subsamples. By doing this, we
can verify whether different drivers and barriers to Bitcoin
adoption are regarded as important in each of the four
subsamples.
4. Regression results
4.1. Baseline regression results
The results of the baseline regression can be found in
appendix C. In this regression, the total sample consisting
of 468 observations from 117 countries is included. The
p-value of the overall F-test indicates that the null hypoth-
esis, stating that the coefficients of the independent varia-
bles are jointly equal to zero, can be rejected. The overall
model is statistically significant at the one percent level.
The R-squared equals 0.5839 meaning that 58.39% of the
variance in the adoption rate is explained by the inde-
pendent variables, which is rather high. Since we apply a
multiple regression, we also take into account the adjusted
R-squared. The adjusted R-squared is equal to 0.5748 or
57.48%, which is not very different from the R-squared
value.
To see whether our independent variables have a signifi-
cant effect on the dependent variable adoption, we
conduct a t-test on a significance level of 0.05 for all our
variables individually. In addition, we take into account
the p-value. The variables inflation, efficiency, noncash
and gdp are highly statistically significant at the 1% level.
The variable corruption is statistically significant at the
10% level since its p-value is 0.059, which is below 0.10.
The variables finincl, internet, trade, population and
unemp are not statistically significant and thus have no
significant impact on the adoption rate of Bitcoin. This
result looks counterintuitive, because internet access is a
crucial asset to be able to use Bitcoin. The fact that the
impact of financial inclusion, which refers to ability to
access financial institutions, is not significant might be
explained by the existence of other financial alternatives
such as Paypal or other digital currencies next to Bitcoin.
So even though people are not able to access a branch of
7. 258 FINANCIEEL FORUM / BANK- EN FINANCIEWEZEN 2017/3 LARCIER
THE FUTURE OF CRYPTOFINANCE: AN EMPIRICAL ANALYSIS OF THE ADOPTION OF BITCOIN LIESJE DEGRAEVE & NIKI DE KEGEL
a commercial bank because the number of branches of
commercial banks in their country is rather low, they do
not necessarily choose for Bitcoin as their payment
system. Furthermore, international trade does not seem to
have a significant impact on the adoption of the crypto-
currency. One possible explanation could be that interna-
tional trade is a more important determinant for the adop-
tion of Bitcoin for companies engaging in globalization
rather than for individuals.
Subsequently, we take a look at the individual coefficients
of the different regressors in the baseline regression that
are statistically significant to study the impact of an
increase in the independent variables on the adoption rate.
These coefficients can be verified in Appendix C. When
we look at the results, we observe a negative sign for the
beta estimate of the variable corruption (-0.0334). In
other words, a one unit increase in the modified Corrup-
tion Perception Index of a certain country leads to a
decrease in the adoption rate of Bitcoin of 0.0334
percentage points. Hence, countries with a lower corrup-
tion rate tend to have more inhabitants using the crypto-
currency Bitcoin.
For the variable inflation, we find a negative relationship
with the dependent variable whereby the beta coefficient
is equal to -0.0764. This indicates that a one percentage
point increase in the inflation rate results in a 0.0764
percentage point decrease of the adoption rate of the
cryptocurrency. This result is not in line with our expecta-
tions. However, since there are no countries included in
our sample that are subject to severe hyperinflation, we
cannot conclude that hyperinflation does not have any
effect on the adoption rate of Bitcoin.
Furthermore, our results indicate that the less efficient the
banking system in a country, the higher the adoption rate
of Bitcoin in that particular country. This is pointed out
by the positive sign of efficiency (0.0751), which means
that a one percentage point increase in the cost-to-income
ratio results in an increase in the adoption of Bitcoin of
0.0751 percentage points. This is rather intuitive, as the
higher the cost-to-income ratio, the lower the banking
system efficiency and thus the more likely people will look
for other alternatives such as Bitcoin.
Finally, we find a positive correlation between the varia-
bles noncash and adoption. The estimated coefficient of
noncash equals 0.0368 meaning that a one percentage
point increase in the usage of non-cash payment methods
in a particular country, increases the adoption rate of
Bitcoin with 0.0368 percentage points. This might be
explained by the familiarity with doing online transac-
tions, which makes the transition to Bitcoin less radical.
After running the baseline regression, we check for multi-
collinearity by computing the variance inflation factors
for all our independent variables. None of the VIFs exceed
the threshold meaning that there is no sign of severe multi-
collinearity. The VIFs can be found in Appendix A. In
addition, we construct a correlation matrix of the esti-
mated regression coefficients, which leads to the same
conclusion. The extent to which collinearity among
predictors may cause problems in the estimation of the
regression coefficients is rather low, as can be verified in
Appendix B. The results of our initial analysis can thus be
considered as representative.
4.2. Further analysis for low income, lower middle
income, upper middle income and high
income countries
In the second part of our analysis, we focus on different
subsamples, which are derived from our initial sample by
taking into account the gross national income of a
country. We repeat the original baseline regression for all
four subsets. The results are presented in Appendix D.
After running the regressions, we test for overall signifi-
cance of each of the four models and notice that the vari-
able population is dropped every time during the test,
which is why we will run the regressions once more
without the variable population. This results in four
overall significant models with an R-squared of 0.5225
for low income countries, 0.3909 for lower middle
income countries, 0.3986 for upper middle income coun-
tries and 0.4331 for high income countries. Thus, the
percentage of the variance in the adoption rate explained
by the independent variables is considerably lower than
for the original model except for the subset containing
low income countries whereby the R-squared is only
slightly lower.
When comparing the effects of the variables with our
earlier conclusions, we find similar results for our signifi-
cant variable inflation. In all four subsets, the results indi-
cate that the inflation rate is negatively correlated with the
adoption rate of Bitcoin. For low income countries, the
estimated coefficient of inflation equals -0.0016, which is
statistically significant at the 5% level. For high income
countries, the estimated coefficient of inflation equals -
1.0748, which is statistically significant at the 1% level.
This indicates that the higher the gross national income of
a country, the stronger the negative impact on the adop-
tion of Bitcoin. In contrast to previous findings, the
8. LARCIER FORUM FINANCIER / REVUE BANCAIRE ET FINANCIÈRE 2017/3 259
LIESJE DEGRAEVE & NIKI DE KEGEL THE FUTURE OF CRYPTOFINANCE: AN EMPIRICAL ANALYSIS OF THE ADOPTION OF BITCOIN
impact of the variables corruption and efficiency on the
adoption rate of Bitcoin are no longer significant in the
low, lower middle and upper middle income subsets. For
high income countries, on the other hand, the beta esti-
mate of the variable corruption, which is statistically
significant at the 5% level, equals -0.0603. In other
words, a one unit increase in the modified Corruption
Perception Index results in a decrease in the adoption rate
of Bitcoin of 0.0603 percentage points. Moreover, we
observe a positive sign for the estimated coefficient of the
variable efficiency (0.1562) in high income countries. This
variable is statistically significant at the 1% level. The
variables internet (0.0051 for the low income sample,
0.0088 for the lower middle income sample and 0.0122
for the upper middle income) and trade (0.0005 for the
low income sample, 0.0045 for the upper middle income
sample and 0.0103 for the high income sample) are statis-
tically significant and positively correlated with the adop-
tion rate of Bitcoin. The impact of internet access on the
adoption rate of Bitcoin increases with the level of income
and is more than doubled when you compare the low and
upper middle income model. In addition, the import and
export of goods and services seems to have a bigger
impact on the adoption of the cryptocurrency when the
income level in a country is higher.
In the low income model, the variable noncash (-0.0024),
which is statistically significant at the 5% level, no longer
has a positive effect on the adoption rate. The more non-
cash transactions are performed in low income countries,
the lower the adoption of the cryptocurrency. The results
for the variable noncash in the other models are similar to
the ones we find when analysing the full sample. In other
words, the beta coefficient of the variable noncash equals
0.0180 for upper middle income countries and 0.0761 for
high income countries, indicating a positive relationship
with the adoption of Bitcoin. These variables are statisti-
cally significant at the 1% level. However, this relation-
ship is not significant for the lower middle income subset.
Lastly, the variable finincl does not have a significant
impact on the adoption of the cryptocurrency in the low
and high income model. In the results of the lower middle
income model, we find a negative impact of finincl (-
0.0143) on Bitcoin’s adoption rate. But if we look at the
upper middle income model, we find contradicting
results. The beta coefficient of finincl in the upper middle
income model, which is significant at the 1% level, equals
0.0087. This implies that the greater the degree of finan-
cial inclusion, the higher the adoption rate. So, further
research is needed to study the real impact of financial
inclusion on the adoption rate of Bitcoin.
5. General conclusion
This empirical paper studies the impact of country deter-
minants on the adoption rate of Bitcoin. The paper
contributes to the existing literature as little research has
been conducted on the adoption of Bitcoin and the factors
driving this adoption. Moreover, this paper is the first to
study Bitcoin adoption rates at countrywide level on such
a large scale. Our sample consists of 117 countries and
covers the period 2011 to 2015. We analyse the entire
sample followed by an additional analysis whereby we
divide our sample into subsamples. By dividing the sample
into different subsets based on the income level of a
country, we can add some meaningful insights to the
discussion on the effect of country determinants on the
adoption rate of the cryptocurrency Bitcoin. For both
analyses, we use pooled OLS regression models.
Our results indicate that a higher corruption rate has a
negative effect on the adoption rate of Bitcoin. Hence,
people living in countries with more corrupt public sectors
will use the cryptocurrency Bitcoin as their monetary
payment system to a lesser extent than people living in
countries with a lower corruption rate. This is not in line
with findings of comparable studies (Darlington & James
(2014) and Bashir, Strickland & Bohr (2016)) where
results show that more corruption in the public sectors of
a country leads to a higher adoption rate of Bitcoin in that
country. Our results thus add a valuable insight to the
discussion on the link between the corruption rate and the
adoption rate of the cryptocurrency Bitcoin. Furthermore,
we find that inhabitants of countries with a higher infla-
tion rate tend to adopt the cryptocurrency Bitcoin in
lower amounts. However, since there are no countries
included in our sample that are subject to severe hyperin-
flation, we cannot conclude that hyperinflation does not
have any effect on the adoption rate of Bitcoin. Next, our
results indicate that the more efficient the banking system
in a country is, the less likely it is that people switch to
another currency such as Bitcoin. According to Luther
(2015), the switching costs, i.e. the transition costs of
switching from traditional monetary payment systems to
Bitcoin, are considered too high. At last, in countries
where people are used to perform online transactions
instead of cash transactions, inhabitants are more likely to
adopt the digital currency Bitcoin to a larger extent. The
extent to which a country's inhabitants have access to
internet, the extent to which a country is involved in inter-
national trade and the extent of financial inclusion among
a country's inhabitants, on the contrary, do not have a
statistically significant impact on the countrywide adop-
tion rate of Bitcoin.
9. 260 FINANCIEEL FORUM / BANK- EN FINANCIEWEZEN 2017/3 LARCIER
THE FUTURE OF CRYPTOFINANCE: AN EMPIRICAL ANALYSIS OF THE ADOPTION OF BITCOIN LIESJE DEGRAEVE & NIKI DE KEGEL
When focussing on different subsets of countries, we
notice the connection between the income of a country
and the adoption rate of Bitcoin. Subsets of higher income
countries tend to have a higher adoption rate. Low
income countries are more likely to be extremely corrupt.
However, the degree of corruption is not regarded as an
important driver for Bitcoin adoption for three of the four
subsets. Only the high income countries, also the ones
with the lowest corruption rate, regard corruption in
public sectors as a barrier to the adoption of the crypto-
currency. This is in line with our previous findings.
However, in contrast to our previous findings, the effi-
ciency of the banking system in a country does not have a
significant impact on the adoption of Bitcoin for the low
income, lower middle income and upper middle income
subsets. For the high income subset, on the other hand,
our findings are similar to the results of our first analysis:
the higher the efficiency of the banking system, the lower
the adoption rate of Bitcoin.
Furthermore, we see that the fraction of people that make
use of debit cards in upper middle income countries and
high income countries are positively associated with the
fraction of the population in that country that makes use
of Bitcoin as a means of payment. This is not the case for
inhabitants of low income countries. Results for this
subset indicate that the use of debit cards acts as a barrier
to Bitcoin adoption. For lower middle income countries,
the use of debit cards does not have a statistically signifi-
cant impact on the adoption of Bitcoin.
While the access to internet and the level of international
trade in a country are not regarded as important drivers
for the adoption of Bitcoin when we take into account the
total sample, this does not hold for the different subsam-
ples. Results indicate that the better the accessibility to
internet and the more open a country is to globalization,
the more likely it is that people are adopting the crypto-
currency. Finally, we find that the degree of financial
inclusion is negatively related to the adoption rate of the
cryptocurrency for the upper middle income sample and
positively related to the adoption rate for the lower
middle income sample. The results for this variable point
out that there still is room for further research.
Appendices
Appendix A: VIF Baseline Regression
Appendix B: Correlation matrix of the estimated coefficients
Variables VIF 1/VIF
Gdp 5.16 0.194
Noncash 5.09 0.197
Internet 4.98 0.2
Corruption 3.84 0.261
Finincl 1.47 0.682
Inflation 1.25 0.799
Trade 1.25 0.802
Unemp 1.16 0.859
Population 1.10 0.907
Effficiency 1.07 0.934
Mean VIF 2.64
Corruption Finincl Inflation Internet Trade Efficiency Noncash Gdp Population Unemp Constant
Corruption 1.0000
Finincl 0.3075 1.0000
Inflation -0.3316 0.0804 1.0000
Internet -0.3185 -0.3589 -0.0081 1.0000
Trade 0.0882 0.3159 0.1678 -0.0388 1.0000
Efficiency 0.0759 0.1889 -0.1243 -0.2778 0.0342 1.0000
Noncash 0.0836 -0.3752 -0.3458 -0.2508 -0.2892 0.1161 1.0000
Gdp 0.5578 0.3399 0.0243 -0.5825 0.0005 0.1813 -0.2961 1.0000
Population 0.1494 0.4259 0.0658 -0.1803 0.3916 0.5108 -0.3349 0.2443 1.0000
Unemp 0.0838 0.1475 0.0292 -0.2165 -0.0667 0.0710 -0.0471 0.1625 0.2100 1.0000
Constant -0.7895 -0.4283 0.2184 0.3847 -0.2774 -0.6206 -0.0651 -0.5116 -0.4964 -0.1881 1.0000
10. LARCIER FORUM FINANCIER / REVUE BANCAIRE ET FINANCIÈRE 2017/3 261
LIESJE DEGRAEVE & NIKI DE KEGEL THE FUTURE OF CRYPTOFINANCE: AN EMPIRICAL ANALYSIS OF THE ADOPTION OF BITCOIN
Appendix C: Regression output of the baseline regression
This table presents results for the baseline regression. Model (1) represents the baseline regression for our full sample
including all variables of interest and all control variables. Model (2), (3) and (4) represent the same regression, but each of
them excludes one variable that showed high correlation values (see appendix B). Model (2) excludes GDP per capita and
internet access, Model (3) excludes internet access and Model (4) removes non-cash utilization and population. In all four
models Bitcoin’s adoption rate is used as the dependent variable. All results below are rounded off to four significant figures.
Standard errors in parentheses
*** p <0.01 , ** p <0.05 , * p < 0.1
Appendix D: Regression output of additional analysis
This table presents the results for the additional analysis where we divide our sample in four subsets. Model (1) repre-
sents the baseline regression (excluding the variable population) for the low income countries. Model (2) includes the
lower middle income countries, Model (3) includes the upper middle income countries and Model (4) includes the high
income countries. All results below are rounded off to four significant figures.
Standard errors in parentheses
*** p <0.01 , ** p <0.05 , * p < 0.1
Variables/Models (1) (2) (3) (4)
Corruption -0.0334*
(0.0176)
-0.0714***
(0.0137)
-0.0335*
(0.0176)
-0.0519***
(0.0174)
Finincl -0.0016
(0.0132)
-0.0018
(0.0141)
-0.0019
(0.0128)
0.0051
(0.0129)
Inflation -0.0764***
(0.0226)
-0.0761***
(0.0254)
-0.0757***
(0.0227)
-0.0597***
(0.0197)
Internet -0.0029
(0.01)
0.0334***
(0.0113)
Trade 0.0061
(0.0041)
0.0063
(0.0043)
0.006
(0.0041)
0.0068
(0.0041)
Efficiency 0.0751***
(0.0159)
0.08***
(0.0164)
0.075***
(0.0158)
0.0668***
(0.0153)
Noncash 0.0638***
(0.0129)
0.0843***
(0.0097)
0.0623***
(0.0126)
Gdp 0.00009***
(0.00003)
0.00009***
(0.00003)
0.0001***
(0.00003)
Population 4.78 x 10-11
(5.28 x 10-10
)
-1.79 x 10-10
(5.28 x 10-10
)
5.12 x 10-11
(5.31 x 10-10
)
Unemp 0.0349
(0.0244)
0.0071
(0.0265)
0.0342
(0.024)
0.047*
(0.0239)
Constant -3.1613*
(1.6611)
-0.6838
(1.4644)
-3.1839*
(1.6934)
-1.544
(1.5942)
Observations 468 468 468 468
R-squared 0.5839 0.5664 0.5839 0.5603
Variables/Models (1) (2) (3) (4)
Corruption 0.0003
(0.0005)
0.0045
(0.0046)
-0.0048
(0.0063)
-0.0603**
(0.0267)
Finincl 0.0015
(0.0014)
-0.0143**
(0.0072)
0.0087***
(0.003)
0.002
(0.0308)
Inflation -0.0016**
(0.0007)
-0.014**
(0.0057)
-0.0279***
(0.0097)
-1.0748***
(0.2701)
Internet 0.0051***
(0.0015)
0.0088**
(0.0041)
0.0122***
(0.0044)
0.0094
(0.0574)
Trade 0.0005*
(0.0003)
0.0019
(0.0013)
0.0045**
(0.002)
0.0103**
(0.0044)
Efficiency -.0001
(0.0004)
-0.0009
(0.005)
0.0046
(0.0047)
0.1562***
(0.0347)
Noncash -0.0024**
(0.0011)
0.0096
(0.0066)
0.018***
(0.0054)
0.0761***
(0.0279)
Gdp -0.00002
(0.00002)
0.0001**
(0.00005)
0.00009***
(0.00003)
3.52 x 10-6
(0.00004)
Unemp -0.0041**
(0.002)
-0.0074*
(0.0042)
0.0212
(0.0133)
0.0732
(0.092)
Constant 0.0125
(0.0385)
-0.4473
(0.3734)
-1.5**
(0.7128)
-3.7277
(4.3153)
Observations 77 103 130 158
R-squared 0.5225 0.3909 0.3986 0.4331
11. 262 FINANCIEEL FORUM / BANK- EN FINANCIEWEZEN 2017/3 LARCIER
THE FUTURE OF CRYPTOFINANCE: AN EMPIRICAL ANALYSIS OF THE ADOPTION OF BITCOIN LIESJE DEGRAEVE & NIKI DE KEGEL
Important sources
DARLINGTON III, J. K., 2014, The Future of Bitcoin: Mapping the Glo-
bal Adoption of World’s Largest Cryptocurrency Through Benefit
Analysis. University of Tennessee. Honors Thesis Projects. Retrieved
from: http://trace.tennessee.edu/cgi/viewcontent.cgi?arti-
cle=2741&context=utk_chanhonoproj.
NAKAMOTO, S., 2008, Bitcoin: A Peer-to-Peer Electronic Cash System.
Retrieved from: https://bitcoin.org/bitcoin.pdf.
PURI, V., 2016, Decrypting Bitcoin Prices and Adoption Rates using
Google Search. CMC Senior Theses, Claremont McKenna College.
Retrieved from: http://scholarship.claremont.edu/cgi/viewcon-
tent.cgi?article=2379&context=cmc_theses.