Methodological Approach Analysis II

47 stationarity conditions, should not go unacknowledged (Caporin and McAleer, 2013). Therefore, Caporin and McAleer suggest that the model should only be used with care to forecast returns but serves well as a means to extract the DCCs. This is in line with how this thesis utilizes the DCC GARCH model.

With respect to the regression model, it is noteworthy that the extracted DCCs are, in fact, predicted correlations generated by the DCC GARCH forecast model, which is based on the inserted historical data. Consequently, an element of uncertainty in the estimated DCCs used for the regression analysis is present, despite being predicted from the DCC GARCH model with a high degree of accuracy.

Moreover, it would be negligent not to point to the critique of the quantile regression analysis provided by Reboredo (2013), who advocates that the model is insufficient in describing the dependence structure, as the marginal effects do not fully account for the joint extreme market movements. However, this is accommodated by substantiating the regression with a graphical analysis of the returns of the asset pairs that present negative correlations to detect the precise reason for the correlations over time. Lastly, it is recognized that amid the considered COVID-19 period, several other factors, i.a., the US presidential election and Brexit disputes, have affected financial markets, why it cannot be refuted that these have impacted the marginal effects of the COVID-19 dummy variables.

48 5.2.2.1. Implicit Costs of Trading

The daily bid-a k e cen age ead f m Bi c in bid and a k ice f he e i d f m Oc be 2013 through August 2020 are generated. In addition, the bid-ask percentage spreads of gold, Apple Inc (hereafter Apple), and Twitter Inc (hereafter Twitter) are computed for comparative reasons. The choice of these three assets is reasoned for in data section 5.3.2. This allows for the comparison of Bitcoin with gold, which has already been established to present safe haven capabilities in the existing literature, as well as one relatively volatile stock, Twitter, and one relatively less volatile stock, Apple.

The bid-ask percentage spread is calculated as follows:

𝐵𝑖𝑑 𝑎𝑠𝑘 𝑠𝑝𝑟𝑒𝑎𝑑 % 𝑎𝑠𝑘 𝑝𝑟𝑖𝑐𝑒 𝑏𝑖𝑑 𝑝𝑟𝑖𝑐𝑒

𝑎𝑠𝑘 𝑝𝑟𝑖𝑐𝑒 100 9

Organizing the spreads in a graph over the entire time period as well as during a shorter and more recent time frame from September 2019 through August 2020 enables a comparative trend analysis f Bi c in li idi cha ac e i ic i h e ec gold, Apple, and Twitter. To allow for a more precise comparison of the liquidity characteristics of Bitcoin, gold, Apple, and Twitter, the mean of each a e bid-ask spreads for 1) the entire sample ranging from October 2013 through August 2020, 2) the more recent sub-period ranging from September 2019 through August 2020, 3) a sub-period ranging from February 24^th, 2020 to April 10^th, 2020 - the same period of high COVID-19 related market stress as utilized in Analysis I - are computed. To test whether the differences in means be een he a e d ing he h ee e i d a e ignifican l diffe en f m e , Welch mean-comparison two-sample t-tests with unequal variances are run (Agresti and Franklin, 2014; Stata, 2020). This test is performed in Stata (see Appendix 5) and evaluated on the basis of the p-value, which is the probability summary of the evidence against the null hypothesis that the difference of the means is zero. The smaller the p-value, the stronger the evidence that the null hypothesis can be rejected. This thesis follows the academically accepted 5% significance level (Agresti & Franklin, 2014). Furthermore, the bid-ask percentage spread of Bitcoin is graphed against two financial stress indice f m Se embe 2019 h gh A g 2020 e amine Bi c in li idi de el men during COVID-19.

5.2.2.2. Explicit Costs of Trading

As outlined in section 2.1.1., every Bitcoin transaction must be added to the blockchain, the official public ledger of all Bitcoin transactions, in order for the transaction to be successfully completed and

49 valid. Bitcoins cannot exist or be held independently of the blockchain. The validation of all transactions occurs through the process of mining, which takes care of including transactions in the limited space of a 1 MB block. When a block is filled up with transactions, it is added to the blockchain, which occurs circa every 10 minutes. Transaction fees are charged for this process, which make up the most substantial share of the overall fees charged when trading Bitcoins on exchanges.

While smaller, additional fees might be charged by the exchanges at which Bitcoins are bought and sold, this analysis solely focuses on the transaction costs of using the Bitcoin network and disregards the additional fees applied by exchanges, which differ across exchanges (CoinDesk, 2020a). The explicit transaction cost characteristics of trading Bitcoins are explored by computing the mean and maximum of the average transaction fees per day during the same (sub-)periods as for the bid-ask spread analysis. Moreover, daily median transaction fee data is graphed against the number of daily transactions from October 1^st, 2013 to August 31^st, 2020, to assess the relation between investment demand and transaction costs. Since investors flee to safe haven assets during crises, demand often rises, why it is essential to know if the transaction fees increase when safe havens are needed the most (Schmitz & Hoffmann, 2020).

5.2.2.3. Methodological Limitations II

The me h d l gical a ach ch en anal e Bi c in li idi is subject to criticism.

Commencing with the implicit costs, it is acknowledged that the significance test of the differences in means of the bid-ask spreads can be impaired due to possible Type I or Type II errors. Whereas Type I errors imply the rejection of the null hypothesis, i.e., that the difference in means is zero, when it, in fact, cannot be rejected, Type II errors imply that the null hypothesis is not rejected even when it, in fact, should be. This study minimizes Type I errors by applying the academically accepted 5%

significance level. Type II errors are reduced by utilizing large samples of 1,805 observations for the whole period, 261 observations for sub-period 1, and 35 observations for sub-period 2 (Agresti and Franklin, 2014). Moreover, it is acknowledged that the statistical significance of the results does not necessarily mean practical significance (Ibid).

Turning to the explicit costs, it is important to note that the approach only provides evidence for Bi c in an ac i n fee related to mining, which does not allow for a comparison to the transaction costs of other assets. The latter is also a consequence of difficulties associated with measuring the transaction costs of traded assets, as these consist of several components of which especially the variable component consisting of commissions charged by brokers, taxes, and transfer fees are hard

50 to measure (Collins and Fabozzi, 1991; Lv, Liu and Wang, 2012). Hence, this has been omitted due to the scope of the thesis and the difficulty related to arriving at comparable measures.

Moreover, it is acknowledged that liquidity is not only associated with bid-ask spreads and transaction costs but also concerns the market depth of assets. The latter refers to the ability of the market to sustain relatively large market orders without impacting the price of an asset. However, Scharnowski (2020) a g e ha hile bid-ask spreads typically matter most for retail investors, institutional in e a e m e c nce ned ab hei ice im ac (p. 2). Consequently, as this thesis is delimited to primarily provide practical implications for retail investors (see section 1.1.), it is not of utmost importance to investigate market depth, why it has been disregarded in this thesis.