This research uses a dynamic conditional correlation model to examine the optimal hedge ratios of crypto token markets. For this purpose, we aggregate 598 crypto tokens 28/04/2014 to 01/07/2018 in three market-cap weighted indexes, by clustering the price series in order to minimize the differences in terms of a dispersion measure, in particular, the Gini coefficient of the market capitalizations. As a result, three indices emerged: a large-cap token index(30 most capitalized crypto tokens), a medium-cap token index (411 most capitalized tokens, excluding top 30), and a small-cap token index (residual 157 tokens). All the series were found to have positive non-zero mean but very high variance, positive skewness (from moderate to intense moving to smaller cap tokens), and excess kurtosis. A t-student distribution with ν= 4 for large and mid-cap tokens, and ν=3 for small-cap tokens is found to be more appropriate to model the returns series than the normal distribution. Moreover, all the series shows autocorrelation in the return (with intensity rising from large to small cap index), volatility clustering but non-variance asymmetric response with respect to the error sign, when tested for the whole sample period. In some subperiods instead, we found empirical evidence of inverse asymmetric volatility phenomenon which may suggest the tokens act as a safe haven in different time period, similarly to Gold (D. G. Baur, 2012) and Bitcoin (Bouri, Azzi, & Dyhrberg, 2016). Finally, just medium-cap tokens were found positive to risk aversion effect. This result may underly possible similarities in small and large-cap token investors’ utility functions. For example, small- and large- cap investors may act “bravely” in a similar way when investing in large capitalized and very famous small capitalized and highly unknown tokens. The specifications used to model conditional means and variances are ARMA, GARCH, APARCH, APARCH-in-Mean with normal and t-student unconditional distributions. To model the conditional correlation six dynamic models were proposed (DCC-MVN, DCC-MVL, DCC-MVT, aDCC-MVN, aDCC-MVL and aDCC-MVT) with a pool of 15 other financial assets (S&P 500, FTSE 100, DAX 30, Nikkei 225, Shanghai A-share, MSCI World, MSCI EU, MSCI Asia Pacific, US dollar index, SPGS commodity index, Pimco Investment Grade Corporate Bond Index ETF, Brent Crude Oil, Gold and Corn spot prices, and CRIX). We find that hedging strategies involving either large, medium, or small-cap tokens reduce a portfolio’s risk (variance), as compared to the risk of a portfolio composed of 100% of the asset only. However, diversifier ability worst off by moving from large- to small-cap tokens. Moreover, risk reduction effectiveness(RRE) ratio is still too low even when compared to CRIX (top 30 cryptocurrencies traded in the market index) making the crypto tokens not a valid hedging alternative to mainstream cryptocurrencies, at the moment.
Questa tesi utilizza un modello di correlazione condizionale dinamico per esaminare gli optimal hedge ratios dei mercati dei crypto token. A tal fine, la ricerca ha aggregato 598 crypto tokens dal 28/04/2014 al 01/07/2018 in tre indici ponderati dalla capitalizzazione di mercato (market-cap weighted index), al fine di ridurre al minimo (per quanto possibile) le differenze in termini di dispersione del campione. In particolare è stato guardato il Coefficiente di Gini delle capitalizzazioni di mercato dei tokens. Come conseguenza di ciò sono emersi tre indici: un indice “large-cap” (con i 30 tokens più capitalizzati), un indice a capitalizzazione media (con i 411 token più capitalizzati, esclusi i primi 30) e un indice token small-cap ( comprensivo dei residui 157 tokens) ). Tutte le serie sono risultate avere una media positiva diversa da zero ma una varianza molto alta, un'asimmetria positiva (da moderata a intensa passando a token con minore capitalizzatione) e una curtosi in eccesso. Una distribuzione t-student con ν = 4 per i token large e mid-cap, e ν = 3 per i token small-cap si è rivelata più appropriata per modellare le serie dei ritorni rispetto alla distribuzione normale. Inoltre, tutte le serie mostrano autocorrelazione nel rendimenti (con l'aumento dell'intensità passando da big- a small-cap), volatility clustering effect ma non risposta asimmetrica della volatilità rispetto al segno degli errori, quando testata per l'intero periodo campionario. In alcuni sottoperiodi, invece, abbiamo trovato evidenze empiriche di fenomeni di “volatilità asimmetrica inversa” che potrebbero suggerire che i token possano aver agito come un “safe-haven asset” in diversi periodi temporali, analogamente al “Gold” (DG Baur, 2012) e al “Bitcoin” (Bouri, Azzi, & Dyhrberg, 2016) . Infine, solo i token a media capitalizzazione sono risultati positivi al “risk aversion effect”. Questo risultato potrebbe suggerire somiglianze nelle funzioni di utilità degli investitori di token di piccole e grandi dimensioni. Ad esempio, gli investitoridi large e small-cap tokens potrebbero agire "coraggiosamente" in modo simile dinannzi alla prospettiva di investire in progetti famosi e molto capitalizzati o largamente sconosciuti e poco capitalizzati. Le specifiche utilizzate per modellare la media e la varianza condizionali sono ARMA, GARCH, APARCH e APARCH-in-Mean con distribuzioni incondizionate normali e t-student. Per modellare la correlazione condizionale sono stati proposti sei modelli dinamici (DCC-MVN, DCC-MVL, DCC-MVT, aDCC-MVN, aDCC-MVL e aDCC-MVT) con un pool di 15 financial assets (S&P 500, FTSE 100, DAX 30, Nikkei 225, Shanghai A-share, MSCI World, MSCI EU, MSCI Asia Pacific, US dollar index, SPGS commodity index, Pimco Investment Grade Corporate Bond Index ETF, Brent Crude Oil, Gold and Corn spot prices, e l’indice CRIX) . La nostra anlisi trova evidenze empiriche che adottare hedging strategies che coinvolgono token large, medium o small cap riducono il rischio (varianza) del portafoglio rispetto al rischio di un portafoglio composto solo dal 100% del patrimonio. Tuttavia, le abilità di diversificatore peggiorano passando da token di dimensioni grandi a small cap. Inoltre, il rapporto RRE (risk reduction effectiveness) risulta essere ancora troppo basso anche solo se confrontato con CRIX (che comprende le prime 30 criptovalute scambiate globalmente) rendendo i crypto token ancora troppo giovani e per il momento non una valida alternativa di hedging rispetto alle criptovalute mainstream.
Multivariate dynamic GARCH hedge ratio and hedge effectiveness in post-ICO crypto tokens market
SANTEUSANIO, VITTORIO
2017/2018
Abstract
This research uses a dynamic conditional correlation model to examine the optimal hedge ratios of crypto token markets. For this purpose, we aggregate 598 crypto tokens 28/04/2014 to 01/07/2018 in three market-cap weighted indexes, by clustering the price series in order to minimize the differences in terms of a dispersion measure, in particular, the Gini coefficient of the market capitalizations. As a result, three indices emerged: a large-cap token index(30 most capitalized crypto tokens), a medium-cap token index (411 most capitalized tokens, excluding top 30), and a small-cap token index (residual 157 tokens). All the series were found to have positive non-zero mean but very high variance, positive skewness (from moderate to intense moving to smaller cap tokens), and excess kurtosis. A t-student distribution with ν= 4 for large and mid-cap tokens, and ν=3 for small-cap tokens is found to be more appropriate to model the returns series than the normal distribution. Moreover, all the series shows autocorrelation in the return (with intensity rising from large to small cap index), volatility clustering but non-variance asymmetric response with respect to the error sign, when tested for the whole sample period. In some subperiods instead, we found empirical evidence of inverse asymmetric volatility phenomenon which may suggest the tokens act as a safe haven in different time period, similarly to Gold (D. G. Baur, 2012) and Bitcoin (Bouri, Azzi, & Dyhrberg, 2016). Finally, just medium-cap tokens were found positive to risk aversion effect. This result may underly possible similarities in small and large-cap token investors’ utility functions. For example, small- and large- cap investors may act “bravely” in a similar way when investing in large capitalized and very famous small capitalized and highly unknown tokens. The specifications used to model conditional means and variances are ARMA, GARCH, APARCH, APARCH-in-Mean with normal and t-student unconditional distributions. To model the conditional correlation six dynamic models were proposed (DCC-MVN, DCC-MVL, DCC-MVT, aDCC-MVN, aDCC-MVL and aDCC-MVT) with a pool of 15 other financial assets (S&P 500, FTSE 100, DAX 30, Nikkei 225, Shanghai A-share, MSCI World, MSCI EU, MSCI Asia Pacific, US dollar index, SPGS commodity index, Pimco Investment Grade Corporate Bond Index ETF, Brent Crude Oil, Gold and Corn spot prices, and CRIX). We find that hedging strategies involving either large, medium, or small-cap tokens reduce a portfolio’s risk (variance), as compared to the risk of a portfolio composed of 100% of the asset only. However, diversifier ability worst off by moving from large- to small-cap tokens. Moreover, risk reduction effectiveness(RRE) ratio is still too low even when compared to CRIX (top 30 cryptocurrencies traded in the market index) making the crypto tokens not a valid hedging alternative to mainstream cryptocurrencies, at the moment.File | Dimensione | Formato | |
---|---|---|---|
thesissss.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: Tesi_final
Dimensione
2.76 MB
Formato
Adobe PDF
|
2.76 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/141366