Content uploaded by Júlio Lobão

Author content

All content in this area was uploaded by Júlio Lobão on Dec 28, 2019

Content may be subject to copyright.

Psychological barriers in the

cryptocurrency market

Vítor Fonseca

Faculty of Economics, Universidade do Porto, Porto, Portugal

Luís Pacheco

Department of Economics and Management,

Universidade Portucalense Infante D. Henrique, Porto, Portugal, and

Júlio Lobão

Faculty of Economics, Universidade do Porto, Porto, Portugal

Abstract

Purpose –The purpose of this paper is to study the existence of psychological barriers in cryptocurrencies.

Design/methodology/approach –To detect psychological barriers, the authors perform a uniformity test,

a barrier hump test, a barrier proximity test and conditional effects test to a sample comprised by the daily

closing quotes of six of the most liquid cryptocurrencies.

Findings –The results evidence the existence of psychological barriers in four of the cryptocurrencies under

scrutiny, namely, Bitcoin, Dash, NEM and Ripple.

Practical implications –The fact that the cryptocurrency market has a high share of unexperienced

investors and presents several cases of psychological barriers is consistent with the hypothesis that that class

of investors is particularly prone to the behavioral biases which cause psychological barriers.

Originality/value –This paper studies, for the first time, the existence of psychological barriers in the

market of cryptocurrencies.

Keywords Market efficiency, Bitcoin, Psychological barriers, Cryptocurrencies

Paper type Research paper

1. Introduction

Psychological barriers are currently one of the most important topics of research in the field

of behavioral finance, with several empirical studies published since the 1990s on a variety

of financial assets, such as stock indices, single stocks, bonds, derivatives and gold, among

others. However, there is a lack of studies regarding the existence of psychological barriers

in the cryptocurrency market. This gap in the literature is surprising given the size of the

market. According to CoinMarketCap (http://coinmarketcap.com), the market capitalization

of the ten largest cryptocurrencies totaled more than US$140bn in March 2019.

Although Philip et al. (2018) defended that cryptocurrencies exhibit patterns identical to

those detected in stock markets, a question yet to be answered is whether there are

psychological barriers in those assets. In this paper we fill this gap in the literature by

examining for the first time the existence of psychological barriers on the emerging market

of cryptocurrencies. We include in our sample six of the most liquid cryptocurrencies and

apply the methodology suggested by Aggarwal and Lucey (2007) for the detection of

psychological barriers, which includes a uniformity test, a barrier proximity test, a barrier

hump test and a conditional effects test.

The results evidence the existence of psychological barriers in four of the cryptocurrencies

under scrutiny: Bitcoin, Dash, NEM and Ripple. The fact that the cryptocurrency market has a

high share of unexperienced investors and presents several cases of psychological barriers is

consistent with the hypothesis that unexperienced investors are more prone to the behavioral

biases which cause psychological barriers than professional traders. Review of Behavioral Finance

© Emerald Publishing Limited

1940-5979

DOI 10.1108/RBF-03-2019-0041

Received 8 March 2019

Revised 6 April 2019

Accepted 21 April 2019

The current issue and full text archive of this journal is available on Emerald Insight at:

www.emeraldinsight.com/1940-5979.htm

JEL Classification —G14, G15, G41

Cryptocurrency

market

The remainder of this paper is organized as follows: Section 2 presents the literature

review, and Section 3 addresses the methodological aspects, namely, the data used in the

study and the various methodological steps. Section 4 discusses the empirical results.

Section 5 concludes the paper.

2. Literature review

2.1 Causes of psychological barriers

2.1.1 Behavioral biases. Psychological barriers can be seenaslimitstoarbitrageprovoked

by mental biases. For example, Mitchell (2001) pointed out that a psychological barrier can

be viewed as an impediment to an individual’s mental outlook, that is, an obstacle created

by the mind, barring advance or preventing access. Accordingly, Hirshleifer (2001)

claimed that investors are affected by judgment and decision biases caused by heuristic

simplification, self-deception and emotional loss of control. Anchoring is one of those

biases and it can be defined as the phenomenon that different starting points yield

different estimates, since people tend to make estimates by starting from an initial value

and then adjust, most of the times insufficiently (Tversky and Kahneman, 1974).

Westerhoff (2003) developed a model in which investors’perception of the fundamental

value is anchored to the nearest round number. This model predicts excessive volatility,

alternation between periods of turbulence and periods of tranquility, and fluctuation of the

exchange rate around its perceived fundamental value. Because of anchoring, exchange

rates are persistently misaligned, which establishes support and resistance levels in the

limits of the fluctuation band. In other words, the perceived fundamental value acts as a

psychological barrier.

2.1.2 Aspiration levels. Sonnemans (2006) indicated that some investors, when

buying a stock, already have an idea for what price they will be able to sell the stock in the

future. This notion can be linked to aspiration levels, a concept of the psychological theory

which was introduced to the economic theory by Simon (1955), alongside the notion that

the economic agent might not pursue the optimal solution but actually settle for a

satisfactory solution.

Additionally, some financial analysts also use target prices for individual stocks and

these are also typically round numbers (Sonnemans, 2006), which will lead to many limit sell

offers being posted at round whole numbers.

2.1.3 Odd pricing. According to Sonnemans (2006, p. 1938), “odd pricing is the tendency

of consumers to consider an odd price like 19.95 as significantly lower than the round price

of 20.00.”This tendency could be originated by the limited amount of memory that people

have, which leads them to attach more significance to the first digits of a price as they

contain more significant information than the last digits (Brenner and Brenner, 1982).

Odd pricing is very common in consumer goods, with a number of studies (e.g.

Holdershaw et al., 1997; Folkertsma, 2002) showing that prices tend to have 9 as the last

significant digit. Additionally, Kahn et al. (1999) argued that, due to this tendency,

financial institutions would profit by quoting retail loan rates with odd-ending yields and

deposit rates with even-ending yields. The theory proposed by the authors predicts that

banks tend to set deposit rates at integers and that rates are sticky at those levels. Also,

when banks set non-integer rates, those are more likely to be just above, rather than just

below integers.

2.1.4 Option exercise prices. Dorfleitner and Klein (2009) stated that psychological

barriers could also be caused by the fact that option exercise prices are usually round

numbers. Delta hedgers are frequently most active when the price of the underlying is close

to the exercise price –in other words, when the option is at the money –so, purely technical

reasons can also cause additional trading activity in the underlying asset.

RBF

2.2 Previous empirical studies

Our study adds to the literature on psychological barriers by studying the phenomenon for

the first time in the market of cryptocurrencies. The first studies about psychological

barriers in financial markets took place in the 1990s. The initial research focused mainly on

stock indices of developed markets, concluding in most cases by the existence of

psychological barriers. These conclusions are shared by the studies conducted by

Donaldson and Kim (1993), Ley and Varian (1994), Koedijk and Stork (1994) and Cyree et al.

(1999), and also by the more recent study authored by Woodhouse et al. (2016).

Donaldson and Kim (1993) tested whether DJIA’s movements around key reference

points exhibited psychological barriers during the period 1974-1990. The main finding is

that those movements were indeed restrained by support and resistance levels at multiples

of 100. After breaking through a 100-level, the DJIA moved by more than otherwise

warranted. Ley and Varian (1994) confirmed these findings considering a wider timer

interval (1952–1993). They found fewer observations around 100-levels than what is

predicted by a uniform distribution. Additionally, they argued that there was no predictive

power on the daily closing prices resulting from psychological barriers.

Koedijk and Stork (1994) expanded the research to a number of indices, considering the

stock markets from Belgium, Germany, Japan, USA and UK. The results corroborate the

conclusions from the previous studies: psychological barriers are real, but they do not

necessarily imply predictability of stock returns.

Cyree et al. (1999) concluded that the last two digits of the DJIA, the S&P 500, the

Financial Times UK Actuaries (London) and the DAX are not equally distributed. Moreover,

their findings support the existence of psychological barriers and show that these effects are

particularly pronounced when the barrier is approached in an upward move.

Finally, Woodhouse et al. (2016) found evidence of psychological barriers at the 100-level

in the NASDAQ Composite index in the period 1990–2012.

Some other studies focusing on European stock markets like those conducted by

Dorfleitner and Klein (2009) and Shawn and Kalaichelvan (2012) found mixed results.

Dorfleitner and Klein (2009) analyzed the German DAZ 30, the French CAC 40, the

British FTSE 50 and the Euro-zone related DJ EURO STOXX 50 for different periods until

2003. They found “fragile traces of psychological barriers”in all indices at the 1,000s barrier

level and also indications of barriers at the 100 barrier level, except in the CAC 40. However,

the authors did not find any systematic barrier effect for any barrier level, thus concluding

that there are no consistent barriers in European stock indices.

Shawn and Kalaichelvan (2012) examinedfive indices (ATX, CAC, DAX, FTSE, SMI) in the

period 2001–2011, having found evidence for barriersin only one index (SMI) at the 1,000-level

but no significant evidence of barriers at the 100 and 1,000 levels in the other indices.

Bahng (2003) and Lobão and Fernandes (2018) carried out two of the few studies

regarding the existence of psychological barriers in emerging stock markets.

Bahng (2003) conducted the first study about the topic on Asian stock markets, using the

daily prices of seven indices (South Korea, Taiwan, Hong Kong, Thailand, Malaysia,

Singapore and Indonesia) from the beginning of 1990 to the end of 1999. The Taiwanese

index and the Indonesian index exhibited some signs of psychological barriers.

Lobão and Fernandes (2018) found no consistent psychological barriers in individual

stock prices near round numbers in the markets of Taiwan, Brazil and South Africa in the

period 2000–2014. The authors also documented that the relationship between risk and

return tended to be weaker in the proximity of round numbers for about half of the stocks

under study.

Different studies concluded that price barriers or at least significant deviations from

uniformity also exist in other asset classes such as bonds (Burke, 2001), commodities

(Aggarwal and Lucey, 2007) and derivatives (Schwartz et al., 2004).

Cryptocurrency

market

The present study also contributes to the emerging literature about the price dynamics of

cryptocurrencies. For example, Dwyer (2015) documented a significant difference between

the pattern of volatility in Bitcoin and that observed in gold and other currencies while

Urquhart (2016) and Barriviera (2017) found signs of inefficiency in the market of that

cryptocurrency. Fry and Cheah (2016) showed that cryptocurrency markets contain a

considerable speculative component, and Cheah and Fry (2015) and Corbet et al. (2018)

concluded that Bitcoin and Ethereum exhibited speculative bubbles.

Finally, our paper relates to the literature on price clustering in cryptocurrencies.

Mitchell (2001) draws a distinction between psychological barriers and price clustering.

While price barriers regard some numbers as support or resistance levels, price clustering is

the concentration of prices on some numbers rather than others.

Urquhart (2017) found significant evidence of clustering of daily Bitcoin closing prices on

round numbers and Hu et al. (2019) concluded that the price clustering also existed in the

intraday prices of cryptocurrencies Bitcoin, Litecoin and Ripple. Xin et al. (2019) found

significant clustering for open, high and low Bitcoin prices at various time frames, suggesting

that that can be explained by the existence of psychological barriers. Some authors have tried

to shed light on the causes of price clustering. For example, Mbanga (2019) showed that the

clustering of Bitcoin was not driven by any day-of-the-week and Baig et al. (2019) documented

the existence of a strong positive association between sentiment and price clustering.

3. Data and methodology

3.1 Data

The sample considered in the study includes six of the most liquid cryptocurrencies, namely,

Bitcoin, Dash, Ethereum, Litecoin, NEM and Ripple, which, at the end of March 2019, and

according to CoinMarketCap, account for nearly 85 percent of the cryptocurrencies’market

capitalization. We collected the daily closing quotes for each asset for the period between the

first trading day of the asset and December 31, 2017 from CoinMarketCap.

Table I presents the descriptive statistics of the assets included in the sample. The table

shows that all the studied cryptocurrencies yielded positive mean returns in the period

under analysis.

3.2 Methodology

3.2.1 Definition of barriers. Following Dorfleitner and Klein (2009), we will use the so-called

band technique and define barriers as an interval between two numbers at the same

distance from the number which constitutes the actual barrier. The main reason for this

technique is the idea that market players will become more active at a certain level before

the price touches a round number. Dorfleitner and Klein (2009) defined the barrier level las

the number of zeros that a barrier has; we will follow the same definition, also introducing

Return series Level series

Asset Start date End date nMean SD Skewness Kurtosis Min. Max.

Bitcoin April 28, 2013 December 31,

2017

1709 0.2727 4.4018 −0.1384 8.9278 68.43 19 497.40

Dash February 14,

2014

1417 0.5608 8.5917 3.1572 41.6881 0.31 1550.85

Ethereum August 7, 2015 878 0.6397 8.5247 −3.7296 64.9273 0.43 826.82

Litecoin April 28, 2013 1709 0.2328 6.9101 1.8842 26.9030 1.16 358.34

NEM April 1, 2015 1006 0.8315 9.3127 2.0248 16.8800 0.00 1.06

Ripple August 4, 2013 1611 0.3707 7.9535 2.2587 29.2155 0.00 2.30

Table I.

Descriptive statistics

RBF

the barrier levels l¼−1 and −2 for the 0.1-level and 0.01-level barriers, respectively. We

define as potential barriers the multiples of 0.01, 0.1, 1, 10, 100 and 1,000 and define intervals

with an absolute length of 2, 5 and 10 percent to the corresponding barriers, thus

considering the following restriction bands:

Barrier level l¼31;000sðÞ:9801;020;9501;050;9001;100

Barrier level l¼2 100sðÞ:98102;95105;90110

Barrier level l¼1 10sðÞ:9:810:2;9:510:5;9:011:0

Barrier level l¼01sðÞ:0:981:02;0:951:05;0:901:10

Barrier level l¼10:1sðÞ:0:0980:102;0:0950:105;0:0900:110

Barrier level l¼20:01sðÞ:0:00980:0102;0:00950:0105;0:00900:0110:

For each cryptocurrency, we then examine the barrier levels which are susceptible of

constituting psychological barriers.

3.2.2 M-values. The concept of M-values was introduced by Donaldson and Kim (1993),

who considered potential barriers at the levels …, 300, 400, …, 3,400, 3,500, …, i.e. at:

kX100;k¼1;2; ::: (1)

De Ceuster et al. (1998) found two problems with this approach: not only it was too narrow,

as the series was not multiplicatively regenerative, leading to 3,400 being considered a

barrier while 340 was not, for instance, but also the gap between barriers, as defined by

Equation (1), would tend to zero as the price series increased, intuitively reducing the

probability of those levels to represent psychological barriers. Thus, the authors claim we

should consider the possibility of barriers at the levels …, 10, 20, …, 100, 200, …, 1,000,

2,000, …, i.e. generally at:

kX10l;k¼1;2;...;9;l¼:::; 1;0;1;...;(2)

and also at the levels …, 10, 11, …, 100, 110, …, 1,000, 1,100, …, i.e. generally at:

kX10l;k¼10;11;...;99;l¼:::; 1;0;1;... (3)

The M-values we will use in our study can now be defined according to these barriers:

Mk¼Pt100

k

mod 100;k¼0:01;0:1;1;10;100;1;000;(4)

where Pt100

ðÞ

=k

ðÞ

is the integer part of Pt100

ðÞ

=k

ðÞ

, and mod 100 is the reduction

modulo of 100.

Illustrating this with a purely theoretical quote of 1,234.56789, the M0.01 is 78, the M0.1 is

67, the M1 is 56, the M10 is 45, the M100 is 34 and the M1000 is 23.

3.2.3 Uniformity test. After defining the M-values, the next step is to examine if they

follow a uniform distribution, through the uniformity test introduced by Ley and Varian

(1994), which consists of a Kolmogorov–Smirnov Z-statistic test where we will be testing: H

0

–uniform distribution –against H1 –non-uniform distribution. In the presence of

psychological barriers, it is expected to reject the null hypothesis. However, it is important to

underline that this rejection does not by itself confirm the existence of such barriers.

Cryptocurrency

market

Additionally, De Ceuster et al. (1998) stated that, as the series grows, the interval between

barriers widens and, as a result, the distribution of digits and their frequency of occurrence

tends to stop being uniform.

3.2.4 Barrier tests. We will perform two barrier tests, a barrier proximity test and a

barrier hump test. The purpose of these tests is to assess if the series observations on or

near a barrier occur less frequently than what would be predicted by a uniform distribution,

examining the shape of the M-values distribution.

A barrier proximity test examines the frequency of M-values in the proximity of

potential barriers, applying the following equation:

fMðÞ¼aþbDþe;M¼00;01;...;99;(5)

where f(M) is defined as the frequency with which a quote closes with its last two digits in

cell M, minus 1 percentage point, and Dis a dummy variable which takes the value 1 if the

price of the asset is at the potential barrier and 0 elsewhere. Besides the strict dummy, which

takes the value 1 if M¼00 and takes the value 0 otherwise, we will study 3 dummies for

each potential barrier level:

D9802 ¼1ifMX98 or Mp02;¼0 otherwise;

D9505 ¼1ifMX95 or Mp05;¼0 otherwise;

D9010 ¼1ifMX90 or Mp10;¼0 otherwise:

The βcoefficients are expected to be negative and statistically significant in the presence of

psychological barriers.

The barrier hump test examines the entire shape of the distribution of M-values and is

broader than the barrier proximity test as it does not focus solely on the proximity of the

potential barriers. We will implement this test using the following equation, which was

introduced by Bertola and Caballero (1992):

fMðÞ¼aþgMþdM2þe;M¼00;01;...;99;(6)

where f(M) is once again defined as the frequency with which a quote closed with its last

two digits in cell M, minus 1 percentage point, and the independent variables are the M-

value and its square.

In the presence of psychological barriers, the M-values are expected to follow a hump-

shape distribution, which will be reflected in Equation (6) through negative and statistically

significant δ, whereas under the null hypothesis of no barriers, δshould be zero and the M-

values should follow a uniform distribution.

3.2.5 Conditional effects test. The final test of our methodology was introduced by Cyree

et al. (1999) and is designed to detect changes in the conditional mean and variance of the

distribution of returns during the sub-periods before and after crossing a barrier, either from

above or below. We use a five-day window before and after crossing a barrier.

In order to identify if a barrier is crossed in an upward or downward movement and

examine the difference in returns between the five-day periods before and after the barrier is

crossed, we will use four dummy variables: UB for the five-day period before prices cross a

barrier on an upward movement, UA for the five-day period after prices cross a barrier on

an upward movement, DB for the five-day period before prices cross a barrier on a

downward movement and DA for the five-day period after prices cross a barrier on a

downward movement. Each of these dummies will take the value 1 on the identified days

and the value 0 elsewhere. Taking into account, as stated by Cyree et al. (1999), that the

distributional shifts implied by psychological barriers invalidate the basic assumptions of

RBF

OLS, we will then regress the following equations using a GARCH (1,1) model:

Rt¼b1þb2UBtþb3UAtþb4DBtþb5DAtþet;(7)

eteN0;Vt

ðÞ;(8)

Vt¼a1þa2UBtþa3UAtþa4DBtþa5DAtþa6Vt1þa7e2

t1þZt:(9)

In the absence of barriers, it is expected that the coefficients of the indicator variables will take the

value zero both in the mean and variance equations, whereas any coefficient significantly

different from zero (either positive or negative) will indicate the presence of psychological barriers.

The four null hypotheses to be tested using a Wald test are the following:

H1. There is no significant difference in the conditional mean return before and after an

upwards crossing of a barrier.

H2. There is no significant difference in the conditional mean return before and after a

downwards crossing of a barrier.

H3. There is no significant difference in the conditional variance before and after an

upwards crossing of a barrier.

H4. There is no significant difference in the conditional variance before and after a

downwards crossing of a barrier.

4. Empirical results

4.1 Uniformity test

Table II shows the results of the uniformity tests for each cryptocurrency, using a

Kolmogorov–Smirnov Z-test. Overall, the studied financial assets show signs of psychological

M0.01 M0.1 M1 M10 M100 M1000

Bitcoin

Z-stat. ––1.900118 1.095154 3.199988 6.327789

p-value ––0.0015*** 0.1815 0.0000*** 0.0000***

Dash

Z-stat. –0.111838 0.026889 0.172266 0.186074 0.264230

p-value –0.0000*** 0.2538 0.0000*** 0.0000*** 0.0000***

Ethereum

Z-stat. –5.269663 2.057020 3.405512 7.523218 –

p-value –0.0000*** 0.0004*** 0.0000*** 0.0000*** –

Litecoin

Z-stat. ––3.569276 14.03169 15.61957 –

p-value ––0.0000*** 0.0000*** 0.0000*** –

NEM

Z-stat. 8.549875 13.32748 17.56190 –––

p-value 0.0000*** 0.0000*** 0.0000*** –––

Ripple

Z-stat. 0.215958 0.631059 0.810789 –––

p-value 0.0000*** 0.0000*** 0.0000*** –––

Notes: The results of a Kolmogorov–Smirnov test for uniformity. Z-stat. stands for the value of the test

statistic; p-value shows the marginal significance of this statistic. H

0

: uniformity; H1: non-uniformity.

*,**,***Significant at the 10, 5 and 1 percent levels, respectively

Table II.

Uniformity test results

Cryptocurrency

market

barriers, as there are statistically significant evidence at a 1 percent significance level that M-

values do not follow a uniform distribution at, at least, three barrier levels for all the six assets.

Additionally, the Ethereum, Litecoin, NEM and Ripple cryptocurrencies reject uniformity at a

5 percent significance level for every potential barrier level.

4.2 Barrier tests

4.2.1 Barrier proximity test. Table III–VI show the results of the barrier proximity

tests performed on the selected cryptocurrencies. As previously mentioned, in the

presence of a psychological barrier, it is expected that βis negative and significant,

which means that there is a lower frequency on the M-values which constitute the

potential barrier.

Testing for psychological barriers in cryptocurrencies, when we consider a barrier to

be in the exact zero modulo point (Table III), we find no negative and significant

βestimates for any asset or any barrier level, even though we find negative (but not

significant) estimates for all assets except NEM. When we assume a barrier in the 98-02

interval (Table IV), Bitcoin and Dash present negative estimates for βfor the 1,000-level

potential barrier, both significant at a 10 percent level. Widening the interval to 95-05

(Table V), we find 9 negative and significant βestimates: Bitcoin at the 100 and 1,000-level

barriers, Dash at the 1-, 10-, 100- and 1,000-level barriers, Litecoin at the 10- and 100-level

barriers and Ripple at the 0.01-level barrier. Finally, considering a barrier to be in the 90-10

interval (Table VI), we find the same negative and significant estimates for βof the

previous table except for Dash at the 1-level barrier, leading to a total of eight negative

and significant βestimates. Once again, as we widen the barrier intervals some estimates

change signal, becoming positive with the enlargement of the intervals, namely, for

Ethereum at the 10- and 100-level barriers, Litecoin at the 1-level barrier and Ripple at the

0.1-level barrier.

Summing up, we are not able to reject the no-barrier hypothesis for Ethereum and NEM

at any potential barrier level; Bitcoin presents some signs of the existence of a psychological

barrier around the 100- and 1,000-level round numbers; Dash presents some signs of the

existence of a psychological barrier around the 1-, 10-, 100- and 1,000-level round numbers;

Litecoin presents some signs of the existence of a psychological barrier around the 10 and

100-level round numbers and the same happens with Ripple at the 0.01-level round numbers.

4.2.2 Barrier hump test. The barrier hump test examines the entire shape of the

distribution of M-values. As mentioned before, in the presence of barriers, these values are

assumed to follow a hump-shape distribution, and thus δis expected to be negative and

significant in the presence of such barriers. Table VII shows the results of this test on the

selected cryptocurrencies, which seem to corroborate most of the results obtained from the

barrier proximity test.

We find negative and significant δestimates for Bitcoin at the 100- and 1,000-level,

Dash at the 10-level, Litecoin at the 10- and 100-level and Ripple at the 0.01-level, leading to

a total of six potential barriers. This means that the barrier hump test does not detect

signs of psychological barriers in three of the nine barrier levels which, according to the

proximity test, exhibited some signs of psychological barriers, namely, Dash at the 1-, 100-

and 1,000-level. The results also corroborate the absence of psychological barriers for

Ethereum and NEM.

Summing up, at this point of our battery of tests, we have found consistent signs of the

existence of psychological barriers around round numbers for four of the six selected

cryptocurrencies –Bitcoin, Dash, Litecoin and Ripple. We also observe that, as we widen the

barrier intervals, the existence of evidence supporting psychological barriers tends to

become more frequent and also more significant.

RBF

M0.01 M0.1 M1 M10 M100 M1000

Cryptocurrency β

p-

value R²β

p-

value R²β

p-

value R²β

p-

value R²β

p-

value R²β

p-

value R²

Bitcoin –––– – –0.8221 0.0017*** 0.0963 −0.1235 0.6000 0.0028 −0.1826 0.5887 0.0039 −0.6555 0.4233 0.0066

Dash –––8.5420 0.0030*** 0.0861 −0.1546 0.5953 0.0029 −0.3685 0.6301 0.0024 −0.6536 0.4320 0.0063 −0.9388 0.3446 0.0091

Ethereum –––1.9811 0.0000*** 0.2638 0.0253 0.9440 0.0001 −0.0897 0.8709 0.0003 −0.6649 0.6899 0.0016 –––

Litecoin –––– – –−0.4782 0.1363 0.0225 −0.8919 0.4467 0.0059 −0.8919 0.4781 0.0051 –––

NEM 2.3033 0.3506 0.0089 33.2289 0.0000*** 0.8285 69.3757 0.0000*** 0.9918 –––––––––

Ripple −0.0069 0.9926 0.0001 −0.6966 0.8064 0.0006 64.9506 0.0000*** 0.9596 –––––––––

Notes: The results of a barrier proximity test using the regression f(M)¼α+βD+ε, where the dependent variable is the frequency of appearance of M-values, minus 1 percentage

point, and ð·is a dummy variable that takes the value 1 if M¼00 and 0 otherwise (see Section 3.2.4. for further details). H

0

:β¼0; H1:βo0. *,**,***Significant at the 10, 5 and 1

percent levels, respectively

Table III.

Barrier proximity test

results for the strict

dummy

Cryptocurrency

market

M0.01 M0.1 M1 M10 M100 M1000

Cryptocurrency β

p-

value R²β

p-

value R²β

p-

value R²β

p-

value R²β

p-

value R²β

p-

value R²

Bitcoin ––––––0.3271 0.0065*** 0.0731 −0.0455 0.4290 0.0064 −0.1657 0.2815 0.0118 −0.6584 0.0762* 0.0317

Dash –––0.9382 0.4850 0.0050 −0.1166 0.3798 0.0079 −0.5029 0.1481 0.0212 −0.5178 0.1714 0.0190 −0.8001 0.0760* 0.0318

Ethereum –––0.5539 0.0014*** 0.0990 0.0024 0.9884 0.0001 0.2182 0.3858 0.0077 −0.1175 0.8773 0.0002 –––

Litecoin ––––––−0.1287 0.3815 0.0078 −0.6708 0.2089 0.0161 −0.8555 0.1344 0.0227 –––

NEM 5.4138 0.0000*** 0.2358 8.6157 0.0000*** 0.2672 14.705 0.0000*** 0.2138 ––––––– ––

Ripple −0.4515 0.1820 0.0181 −0.8174 0.5283 0.0041 16.223 0.0000*** 0.2873 ––––––– ––

Notes: The results of a barrier proximity test using the regression f(M)¼α+βD+ε, where the dependent variable is the frequency of appearance of M-values, minus 1 percentage point,

and ð·is a dummy variable that takes the value 1 if M⩾98 or M⩽02 and 0 otherwise (see Section 3.2.4. for further details). H

0

:β¼0; H1:βo0. *,**,***Significant at the 10, 5 and 1

percent levels, respectively

Table IV.

Barrier proximity test

results for the 98-02

dummy

RBF

M0.01 M0.1 M1 M10 M100 M1000

Cryptocurrency βp-value R²βp-value R²βp-value R²βp-value R²βp-value R²βp-value R²

Bitcoin ––––––0.1196 0.1589 0.0201 −0.0717 0.3379 0.0094 −0.2271 0.0326** 0.0458 −0.6813 0.0078*** 0.0700

Dash –––−0.1216 0.8967 0.0002 −0.2009 0.0282** 0.0482 −0.4892 0.0423** 0.0414 −0.5613 0.0321** 0.0460 −0.8208 0.0083*** 0.0689

Ethereum –– –0.2143 0.0821* 0.0305 −0.0184 0.8726 0.0003 0.1910 0.2752 0.0121 −0.0416 0.9374 0.0001 –––

Litecoin ––––––−0.0418 0.6839 0.0017 −0.6454 0.0814* 0.0307 −0.9742 0.0134** 0.0607 –––

NEM 2.5418 0.0009*** 0.1071 5.5067 0.0000*** 0.2250 6.6845 0.0023*** 0.0911 ––––––– ––

Ripple −0.4705 0.0446* 0.0405 1.5457 0.0849* 0.0300 7.5502 0.0003*** 0.1282 ––––––– ––

Notes: The results of a barrier proximity test using the regression f(M)¼α+βD+εwhere the dependent variable is the frequency of appearance of M-values, minus 1 percentage point, and ð·is

a dummy variable that takes the value 1 if M⩾95 or M⩽05 and 0 otherwise (see section 3.2.4. for further details). H

0

:β¼0; H1:βo0. *,**,***Significant at the 10, 5 and 1 percent levels,

respectively

Table V.

Barrier proximity test

results for the 95-05

dummy

Cryptocurrency

market

M0.01 M0.1 M1 M10 M100 M1000

Cryptocurrency βp-value R²βp-value R²βp-value R²βp-value R²βp-value R²βp-value R²

Bitcoin ––––––−0.0243 0.7108 0.0014 −0.0455 0.4290 0.0064 −0.1830 0.0247** 0.0504 −0.6698 0.0006*** 0.1147

Dash –––0.6314 0.3795 0.0079 −0.0620 0.3831 0.0078 −0.5342 0.0035*** 0.0836 −0.4576 0.0227** 0.0518 −0.6575 0.0058*** 0.0750

Ethereum –––0.0454 0.6336 0.0023 0.0730 0.4069 0.0070 0.4093 0.0019*** 0.0944 0.4642 0.2528 0.0133 –––

Litecoin ––––––0.1097 0.1620 0.0199 −0.7932 0.0048*** 0.0785 −0.9872 0.0010*** 0.1057 –––

NEM 1.2327 0.0392** 0.0424 3.3898 0.0001*** 0.1445 3.4257 0.0446** 0.0405 –––––––––

Ripple −0.6073 0.0006*** 0.1144 3.0857 0.0000*** 0.2025 3.8714 0.0166** 0.0571 –––––––––

Notes: shows the results of a barrier proximity test using the regression f(M)¼α+βD+εwhere the dependent variable is the frequency of appearance of M-values, minus 1 percentage point,

and ð·is a dummy variable that takes the value 1 if M⩾90 or M⩽10 and 0 otherwise (see Section 3.2.4. for further details). H

0

:β¼0; H1:βo0. *,**,***Significant at the 10, 5 and 1 percent levels,

respectively

Table VI.

Barrier proximity test

results for the 90-10

dummy

RBF

M0.01 M0.1 M1 M10 M100 M1000

Cryptocurrency δp-value R²δp-value R²δp-value R²δp-value R²δp-value R²δp-value R²

Bitcoin ––––––−1.11E-05 0.7567 0.0052 −3.78E-05 0.2298 0.0154 −8.72E-05 0.0398** 0.1445 −0.0004 0.0001*** 0.1985

Dash –––6.67E-05 0.8660 0.0031 −2.53E-05 0.5153 0.0105 −0.0002 0.0624* 0.1104 −2.55E-05 0.8136 0.0636 −0.0001 0.3873 0.1311

Ethereum ––––– –2.62E-05 0.5644 0.1177 0.0004 0.0000*** 0.3347 0.0006 0.0038*** 0.0939 –––

Litecoin ––––––0.0001 0.0006*** 0.2150 −0.0005 0.0003*** 0.2485 −0.0006 0.0001*** 0.2898 –––

NEM 0.0007 0.0225** 0.1302 0.0019 0.0000*** 0.2641 0.0022 0.0145** 0.1046 –––––––––

Ripple −0.0005 0.0000*** 0.3389 0.0013 0.0001*** 0.3081 0.0024 0.0050*** 0.1312 –––––––––

Notes: The results of a barrier hump test using the regression f(M)¼α+γM+δM

2

εwhere the dependent variable is the frequency of appearance of M-values, minus 1 percentage point, regressedto the said

M-values and the respective squares (see Section 3.2.4. for further details). H

0

:δ¼0; H1:δo0. *,**,***Significant at the 10, 5 and 1 percent levels, respectively

Table VII.

Barrier hump

test results

Cryptocurrency

market

4.3 Conditional effects test

Tables VIII–X present the results of the conditional effects test, where we examine the

behavior of the selected cryptocurrencies’prices in the five-day periods before and after

crossing a barrier from below (thus, constituting a potential resistance level), and also in the

five-day periods before and after crossing a barrier from above (thus, constituting a

potential support level). We assess if the return series exhibit a pattern on these days which

is significantly different from that when prices are not in the proximity of any barrier.

We perform this test for one potential barrier level only for each asset, chosen as the most

likely to constitute an actual barrier level, according to the results from the previous tests.

Therefore, this test is applied to the 0.1-level barrier for NEM and Ripple, to the 10-level barrier

for Ethereum and Litecoin and to the 100-level barrier for Bitcoin and Dash. The results of the

mean return equation are shown in Table VIII. We may observe that the mean return after

crossing a barrier from below is positive for all the six cryptocurrencies –and significant at a 5

percent level for all of them except Litecoin –while before crossing a barrier in such movement

it is positive for all cryptocurrencies but only significant at a 5 percent level for two of them:

Bitcoin and Ripple. Still regarding the upward movements, the results show that the

magnitude of returns is higher after crossing a barrier for all assets except Litecoin. The

crossing of a barrier from below does not originate a change in the signal of the mean returns

and regarding crossings from above we observe that the mean return is negative for all

assets after crossing a barrier in such movement –significant at a 5 percent level for all

cryptocurrencies but Ripple –and also negative before crossing the barrier for four

assets –Dash and Ripple being the exceptions. Curiously, there are no negative and significant

mean returns for any of the six studied cryptocurrencies in thefive-day periods before crossing

CUBUADBDA

Bitcoin

Coefficient 0.0473 0.6532 1.4683 −0.1983 −1.7376

p-value 0.5462 0.0131** 0.0000*** 0.5782 0.0000***

Dash

Coefficient 0.1795 0.8628 2.5630 0.6653 −3.0760

p-value 0.2156 0.4316 0.0268** 0.0559* 0.0000***

Ethereum

Coefficient 0.3179 0.2525 1.3925 −0.1249 −1.6333

p-value 0.1601 0.6747 0.0049*** 0.8291 0.0023***

Litecoin

Coefficient −0.1418 2.2870 1.3822 −0.6813 −1.8873

p-value 0.3025 0.0908* 0.1519 0.5085 0.0086***

NEM

Coefficient 0.3632 1.0636 5.9505 −1.2022 −5.1457

p-value 0.1155 0.1095 0.0106** 0.0937* 0.0062***

Ripple

Coefficient −0.4388 3.4611 3.9798 1.9683 −1.8759

p-value 0.0000*** 0.0475** 0.0154** 0.4140 0.1406

Notes: The results of the mean equation of a GARCH estimation of Rt¼b1þb2UBtþb3UAtþb4DBtþ

b5DAtþet;eteN0;Vt

ðÞVt¼a1þa2UBtþa3UAtþa4DBtþa5DAtþa6Vt1þa7e2

t1þZt, where UB,UA,

DB and DA are dummy variables. UB takes the value 1 in the 5 days before crossing a barrier from below; UA

takes the value 1 in the 5 days after crossing a barrier from below; DB takes the value 1 in the 5 days

before crossing a barrier from above; DA takes the value 1 in the 5 days after crossing a barrier from above.

*,**,***Significant at the 10, 5 and 1 percent levels, respectively

Table VIII.

Conditional effects

test results –return

equation

RBF

a barrier in a downward movement. The magnitude of returns is higher in the five-day periods

after crossing a barrier for all assets and the crossing of a barrier in a downward movement

promotes a sign change in the mean return of only two cryptocurrencies: Dash and Ripple.

Table IX shows the results of the variance equation. In the presence of psychological

barriers, we should find positive variance indicators before crossing a barrier –meaning that

the market is turbulent –and negative indicators after crossing a barrier –meaning that the

market is calmer. Regarding upward movements, we find positive variance indicators before

crossing a barrier for five cryptocurrencies and negative indicators after crossing a barrier for

also five cryptocurrencies. As for downward movements, we find positive indicators before

crossing barriers for two cryptocurrencies and negative indicators after crossing barriers for

six cryptocurrencies. We observe that volatility tends to increase after crossing a barrier from

below and decrease after crossing a barrier from above which, considering the results obtained

for the mean return equation, is in line with the efficient market hypothesis and the reasoning

that higher returns tend to compensate higher volatility levels. The GARCH term is positive

and significant at a 1 percent level for every asset, indicating significant GARCH effects.

Finally, Table X exhibits the results of the Wald test to the hypotheses listed in Section

3.2.5. We find significant (at a 5 percent significance level) changes in the conditional mean

returns after crossing a barrier in an upwards movement for two cryptocurrencies (Bitcoin

and NEM), while for downwards movements, we observe that changes in the conditional

mean returns are significant at 5 percent for three cryptocurrencies (Bitcoin, Dash and

NEM). As for differences in the conditional variance, we observe significant results for four

cryptocurrencies –Bitcoin, Ethereum, Litecoin and NEM –concerning upwards movements

CRESID(−1)

2

GARCH(−1) UB UA DB DA

Bitcoin

Coefficient 0.3054 0.1192 0.8413 2.0988 −1.3859 2.8976 −0.9813

p-value 0.0000*** 0.0000*** 0.0000*** 0.0000*** 0.0000*** 0.0000*** 0.0051***

Dash

Coefficient 2.1703 0.2091 0.7567 11.277 8.5335 −5.2712 −6.5721

p-value 0.0000*** 0.0000*** 0.0000*** 0.0021*** 0.0214** 0.0405** 0.0073***

Ethereum

Coefficient 3.1468 0.2749 0.6608 7.4373 −0.3394 −4.4360 −0.4347

p-value 0.0000*** 0.0000*** 0.0000*** 0.0000*** 0.8561 0.0000*** 0.6087

Litecoin

Coefficient 1.8237 0.0916 0.8484 42.705 −22.821 −1.3866 −1.9194

p-value 0.0000*** 0.0000*** 0.0000*** 0.0000*** 0.0000*** 0.3691 0.1763

NEM

Coefficient 20.019 0.4368 0.3565 −2.7009 83.051 −19.261 −8.6145

p-value 0.0000*** 0.0000*** 0.0000*** 0.3486 0.0000*** 0.0000*** 0.4938

Ripple

Coefficient 6.5235 0.7574 0.3392 26.087 −1.3252 58.358 −0.7711

p-value 0.0000*** 0.0000*** 0.0000*** 0.1615 0.9389 0.0669* 0.9360

Notes: The results of the variance equation of a GARCH estimation of Rt¼b1þb2UBtþb3UAtþ

b4DBtþb5DAtþet;eteN0;Vt

ðÞVt¼a1þa2UBtþa3UAtþa4DBtþa5DAtþa6Vt1þa7e2

t1þZt,where

UB,UA,DB and DA are dummy variables. UB takes the value 1 in the 5 days before crossing a barrier

from below; UA takes the value 1 in the 5 days after crossing a barrier from below; DB takes the value 1 in the

5 days before crossing a barrier from above; DA takes the value 1 in the 5 days after crossing a barrier

from above. V

t−1

refers to the moving average parameter and e2

t1stands for the GARCH parameter.

*,**,***Significant at the 10, 5 and 1 percent levels, respectively

Table IX.

Conditional effects

test results –variance

equation

Cryptocurrency

market

and regarding downwards movements we find significant differences for three

cryptocurrencies, namely, Bitcoin, Ethereum and Ripple.

Summing up, the conditional effects test indicates that the magnitude of the mean returns

is higher in the five-day period after crossing a barrier, both for upward and downward

movements. Also, we observe that markets tend to be turbulent before crossing a barrier and

calmer after that barrier is crossed, but results also show that volatility tends to stay aligned

with returns, as predicted by the efficient markets hypothesis. Finally, analyzing the results of

the Wald test, we observe significant signs of the existence of psychological barriers in all

cryptocurrencies. Clearly, Bitcoin presents the stronger case for the existence of psychological

barriers, as all four null hypotheses are rejected at a 5 percent significance level.

We are now in position to summarize the results of each test, reaching the global results

shown in the last column of Table XI. The disagreement across tests in some of the

H1 H2 H3 H4

Bitcoin

χ

2

4.2549 10.061 69.024 33.263

p-value 0.0391** 0.0015*** 0.0000*** 0.0000***

Dash

χ

2

1.0170 27.605 0.1577 0.0897

p-value 0.3132 0.0000*** 0.6913 0.7645

Ethereum

χ

2

1.3740 1.9733 5.3936 6.8673

p-value 0.2411 0.1601 0.0202** 0.0088***

Litecoin

χ

2

0.2872 0.9152 150.6705 0.0404

p-value 0.5920 0.3387 0.0000*** 0.8407

NEM

χ

2

4.2216 4.0788 20.1286 0.7282

p-value 0.0399** 0.0434** 0.0000*** 0.3935

Ripple

χ

2

0.0400 2.3157 0.9561 3.1178

p-value 0.8415 0.1281 0.3282 0.0774**

Notes: The results of a Wald test to four hypotheses. H1: there is no significant difference in the conditional

mean return before and after an upwards crossing of a barrier; H2: there is no significant difference in the

conditional mean return before and after a downwards crossing of a barrier; H3: there is no significant

difference in the conditional variance before and after an upwards crossing of a barrier; H4: there is

no significant difference in the conditional variance before and after a downwards crossing of a barrier.

*,**,***Significant at the 10, 5 and 1 percent levels, respectively

Table X.

Conditional effects

test results

Uniformity

test

Barrier proximity

test

Barrier hump

test

Conditional effects

test

Psychological

barriers?

Bitcoin Yes Yes Yes Yes Yes

Dash Yes Yes Yes Yes Yes

Ethereum Yes No No Yes No

Litecoin Yes Yes Yes Yes Yes

NEM Yes No No Yes No

Ripple Yes Yes Yes Yes Yes

Table XI.

Summary of results

from the various tests

RBF

cryptocurrencies stem from the different features of psychological barriers that were captured

by the tests applied to our sample. For example, while uniformity tests only examine whether

M-values follow a uniformdistribution,barrier proximity tests assess whether observations in

the vicinity of a potential barrier are relatively rarer than would be expected. Overall and

considering the four tests we had conducted in this study, we can conclude that there are

strong signs of psychological barriers in four of the cryptocurrencies under scrutiny: Bitcoin,

Dash, Litecoin and Ripple. It is also noteworthy that the cryptocurrencies that present the least

signs of psychological barriers are among the ones that exhibited the highest return and

volatility (measured by the standard deviation of the returns).

5. Conclusion

In this paper, we conducted the first study on psychological barriers in the

cryptocurrencies market.

We considered a sample of six of the most liquid cryptocurrencies. After analyzing the

range of each asset’s quotes and defining all potential psychological barriers, we started by

performing a uniformity test, observing that all assets rejected the null hypothesis, which

claimed that the respective M-values followed a uniform distribution. Then, we conducted a

barrier proximity test using several intervals to each of the previously defined potential

barrier levels, finding signs of the existence of psychological barriers in four cryptocurrencies.

The following test was a barrier hump test, which focused on the whole shape of the M-values

distribution, assessing if they follow a uniform distribution or a hump-shape distribution –as

should be the case in the presence of psychological barriers –and it confirmed the majority of

theprevioustest’s results. Finally, we conducted a conditional effects test, in its three

modalities: mean return equation, variance equation and hypotheses test. The mean return

equation indicated that in all financial markets the magnitude of the mean returns tended to

be higher in the five-day period after crossing a barrier, both for upward and downward

movements; the variance equation led to the conclusion that markets were significantly more

volatile before crossing a barrier; and, through the hypotheses test, we observed signs of the

existence of psychological barriers in four cryptocurrencies.

Overall, we found evidence of the existence of psychological barriers in four of the

cryptocurrencies under study: Bitcoin, Dash, Litecoin and Ripple. Among all the

cryptocurrencies, Bitcoin –the quasi-monopolist leader of the cryptocurrencies market –is

by far the one which presents stronger signs of psychological barriers.

The results of our study may potentially be used by investors to build more profitable

strategies when in presence of psychological barriers. Moreover, our results seem difficult to

reconcile with the efficient market paradigm, since one of the chief features of an efficient

capital market is that prices should not exhibit any particular patterns (Fama, 1970).

Our results are also relevant to the debate about the prevalence of decision-making

biases in different categories of investors. Given that the cryptocurrency market is

essentially inhabited by unexperienced investors (Yermack, 2015; Kow, 2017), our results

suggest that this category of investors is particularly susceptible to the decision-making

biases usually associated with the formation of psychological price barriers.

Our study presents several limitations which may lead to future research on this topic:

for instance, studies with broader samples, namely, a larger set of cryptocurrencies, could

lead to stronger results; it could be fruitful to analyze the prevalence of psychological

barriers in different time periods according to the price trend prevailing in the market at the

time; it would be interesting to investigate the characteristics of cryptocurrencies (in terms

of liquidity, volatility, transaction volume, etc.) that tend to lead to a higher prevalence of

psychological barriers; and, finally, the finding of significant psychological barriers in some

cryptocurrencies implies the need to address its practical implications, namely, the

possibility to earn extraordinary profits exploiting that anomaly.

Cryptocurrency

market

References

Aggarwal, R. and Lucey, B.M. (2007), “Psychological barriers in gold prices?”,Review of Financial

Economics, Vol. 16 No. 2, pp. 217-230.

Bahng, S. (2003), “Do psychological barriers exist in the stock price indices? Evidence from Asia’s

emerging markets”,International Area Studies Review, Vol. 6 No. 1, pp. 35-52.

Baig, A., Blau, B.M. and Sabah, N. (2019), “Price clustering and sentiment in Bitcoin”,Finance Research

Letters, Vol. 29, pp. 111-116.

Barriviera, A.F. (2017), “The inefficiency of Bitcoin revisited; a dynamic approach”,Economics Letters,

Vol. 161, pp. 1-4.

Bertola, G. and Caballero, R.J. (1992), “Target zones and realignments”,American Economic Review,

Vol. 82 No. 3, pp. 520-536.

Brenner, G.A. and Brenner, R. (1982), “Memory and markets, or why are you paying $2.99 for a

widget?”,Journal of Business, Vol. 55 No. 1, pp. 147-158.

Burke, S. (2001), “Barriers in US benchmark bonds”, Unpublished manuscript, Vancouver.

Cheah, E. and Fry, J. (2015), “Speculative bubbles in Bitcoin markets? An empirical investigation into

the fundamental value of Bitcoin”,Economics Letters, Vol. 130, pp. 32-36.

Corbet, S., Lucey, B. and Yarovaya, L. (2018), “Datestamping the Bitcoin and Ethereum bubbles”,

Finance Research Letters, Vol. 26, pp. 81-88.

Cyree, K.B., Domian, D.L., Louton, D.A. and Yobaccio, E.J. (1999), “Evidence of psychological barriers in

the conditional moments of major world stock indices”,Review of Financial Economics, Vol. 8

No. 1, pp. 73-91.

De Ceuster, M.J.K., Dhaene, G. and Schatteman, T. (1998), “On the hypothesis of psychological barriers

in stock markets and Benford’s law”,Journal of Empirical Finance, Vol. 5 No. 3, pp. 263-279.

Donaldson, R.G. and Kim, H.Y. (1993), “Price barriers in the Dow Jones industrial average”,Journal of

Financial and Quantitative Analysis, Vol. 28 No. 3, pp. 313-330.

Dorfleitner, G. and Klein, C. (2009), “Psychological barriers in European stock markets: where are

they?”,Global Finance Journal, Vol. 19 No. 3, pp. 268-285.

Dwyer, G.P. (2015), “The economics of Bitcoin and similar private currencies”,Journal of Financial

Stability, Vol. 17, pp. 81-91.

Fama, E.F. (1970), “Efficient capital markets: a review of theory and empirical work”,Journal of

Finance, Vol. 25 No. 2, pp. 383-417.

Folkertsma, C.K. (2002), “The euro and psychological prices: simulations of the worst-case scenario”,De

Economist, Vol. 150 No. 1, pp. 19-40.

Fry, J. and Cheah, E. (2016), “Negative bubbles and shocks in cryptocurrency market”,International

Review of Financial Analysis, Vol. 47, pp. 343-352.

Hirshleifer, D. (2001), “Investor psychology and asset pricing”,Journal of Finance, Vol. 56 No. 4,

pp. 1533-1597.

Holdershaw, J., Gendall, P. and Garland, R. (1997), “The widespread use of odd pricing in the retail

sector”,Marketing Bulletin-Department of Marketing Massey University, No. 8, pp. 53-58.

Hu, B., McInish, T. and Miller, J. (2019), “Intraday price behavior of cryptocurrencies”,Finance Research

Letters, Vol. 28, pp. 337-342.

Kahn, C., Pennacchi, G. and Sopranzetti, B. (1999), “Bank deposit rate clustering: theory and empirical

evidence”,Journal of Finance, Vol. 54 No. 6, pp. 2185-2214.

Koedijk, K.G. and Stork, P.A. (1994), “Should we care? Psychological barriers in stock markets”,

Economics Letters, Vol. 44 No. 4, pp. 427-432.

Kow, Y.M. (2017), “Cryptocurrencies and their potential for large-crowd, cost-effective transactions in

peer production”,First Monday, Vol. 22 No. 8.

RBF

Ley, E. and Varian, H.R. (1994), “Are there psychological barriers in the Dow-Jones index?”,Applied

Financial Economics, Vol. 4 No. 3, pp. 217-224.

Lobão, J. and Fernandes, J. (2018), “Psychological barriers in single stock prices: evidence from three

emerging markets”,Revista Brasileira de Gestão de Negócios, Vol. 20 No. 2, pp. 248-272.

Mbanga, C.L. (2019), “The day-of-the-week pattern of price clustering in Bitcoin”,Applied Economic

Letters, Vol. 26 No. 10, pp. 807-811.

Mitchell, J. (2001), “Clustering and psychological barriers: the importance of numbers”,Journal of

Futures Markets, Vol. 21 No. 5, pp. 395-428.

Philip, A., Chan, J. and Peiris, S. (2018), “A new look at cryptocurrencies”,Economics Letters, Vol. 163,

pp. 6-9.

Schwartz, A.L., Van Ness, B.F. and Van Ness, R.A. (2004), “Clustering in the futures market: evidence

from S&P 500 futures contracts”,Journal of Futures Markets, Vol. 24 No. 5, pp. 413-428.

Shawn, L.K.J. and Kalaichelvan, M. (2012), “A critical evaluation of the significance of round numbers

in European equity markets in light of the predictions from Benford’s law”,International

Research Journal of Finance and Economics, No. 95, pp. 196-210.

Simon, H.A. (1955), “A behavioral model of rational choice”,Quarterly Journal of Economics, Vol. 69

No. 1, pp. 99-118.

Sonnemans, J. (2006), “Price clustering and natural resistance points in the Dutch stock market: a

natural experiment”,European Economic Review, Vol. 50 No. 8, pp. 1937-1950.

Tversky, A. and Kahneman, D. (1974), “Judgment under uncertainty: heuristics and biases”,Science,

Vol. 185 No. 4157, pp. 1124-1131.

Urquhart, A. (2016), “The inefficiency of Bitcoin”,Economics Letters, Vol. 148, pp. 80-82.

Urquhart, A. (2017), “Price clustering in Bitcoin”,Economics Letters, Vol. 159, pp. 145-148.

Westerhoff, F. (2003), “Anchoring and psychological barriers in foreign exchange markets”,Journal of

Behavioral Finance, Vol. 4 No. 2, pp. 65-70.

Woodhouse, S.A., Singh, H., Bhattacharya, S. and Kumar, K. (2016), “Invisible walls: do psychological

barriers really exist in stock index levels?”,North American Journal of Economics and Finance,

Vol. 36, pp. 267-278.

Xin, L., Shenghong, L. and Chong, X. (2019), “Price clustering in Bitcoin market –an extension”,

Finance Research Letters, forthcoming.

Yermack, D. (2015), “Is Bitcoin a real currency? An economic appraisal”, in Chuen, D.L.K. (Ed.),

Handbook of Digital Currency, chapter 2, Academic Press, Cambridge, MA, pp. 31-43.

Corresponding author

Luís Pacheco can be contacted at: luisp@upt.pt

For instructions on how to order reprints of this article, please visit our website:

www.emeraldgrouppublishing.com/licensing/reprints.htm

Or contact us for further details: permissions@emeraldinsight.com

Cryptocurrency

market