Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

In this thesis, we aim to fill the lack of scientific literature that covers the field of cryptocurrency portfolio management by analyzing the link between the risk profile of a cryptocurrency portfolio and the market provenance of its components. In fact, despite the fact that more and more investors consider cryptocurrencies as a speculative tool, the lack of knowledge that covers this topic still procures a lot of uncertainty. In order to do find an answer to our research question, we firstly define the market provenance of a digital-asset by stating that investors can construct their portfolio from two different types of components: cryptocurrencies traded on a decentralized exchange platform and cryptocurrencies traded on a centralized exchange platform. Through the reviewing of existent literature, we make the emphasis of two key features that will constitute the basis of our hypothesis: (1) decentralized exchange platforms are less liquid than centralized exchange platforms (Barbon and Ranaldo, 2021) and (2) there is a negative correlation between liquidity and expected returns in the cryptocurrency market (Zhang and Li, 2020). By linking these two statements with our research question, we take as a main hypothesis that market provenance effects exist and affect the returns and the risk profile of a portfolio. We aim to verify this hypothesis by conducting the risk-return analysis of two portfolios constructed from two different market provenance: a centralized exchange platform (Binance) and a decentralized exchange platform (Uniswap). In order to have a wider range of results, we follow the work of Mazanec (2021) who studies cryptocurrencies and compares three different portfolio strategies. By extracting the historical data from January 2020 to June 2022, we allocate the weights of the components according the Equally-weighted model and two models following Markowitz’s mean-variance framework: the Maximum Sharpe ratio model and the Minimum Volatility model. To carry out the empirical study, we use the software Python following the work of Lewinson (2020). At the end of this thesis, we confirmed by reviewing the results of our analysis that the market provenance of a digital-asset has an impact on the returns and the risk profile of a portfolio. In fact, our calculations showed that the portfolio exclusively constructed from components bought on a decentralized exchange platform has higher returns and a riskier profile than the portfolio exclusively constructed from components bought on a centralized exchange platform. 
After positioning our results according existent scientific literature, we found out that despite several differences related to the sample period of the data, the outcomes of our study were relevant and aligned with the methodology of Mazanec (2021) we followed.

cryptocurrency --- portfolio --- returns --- liquidity risk --- risk profile --- market provenance --- centralized exchange platform --- decentralized exchange platform --- Sciences économiques & de gestion > Méthodes quantitatives en économie & gestion

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

This master thesis deals with portfolio optimisation under prospect theory, which aims to better reflect reality in terms of purchasing behaviour than modern utility theory. 3 optimisation algorithms are implemented in this work, namely the genetic algorithm, the simulated annealing algorithm and the ant colony algorithm. Their performances are evaluated and compared on several criteria, and in several datasets (CAC40, S&P500 and Nasdaq).

Sciences économiques & de gestion > Finance

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

The objective of the thesis is to investigate the usefulness of popular investment algorithms and go beyond the traditional mean-variance optimization approach, through the use of a realistic investment universe of stocks and mutual funds. The subject of this thesis would consider asset allocation strategies generated from machine learning and robo-advisors, and compare their performance using state-of-the-art statistical approaches.

Asset Allocation, Machine Learning --- Sciences économiques & de gestion > Finance

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

This thesis aims at designing a credit risk rating methodology for the Luxembourg banking sector. More precisely, the objective of this thesis is to develop a model to estimate the probability of default for investment funds, the latter representing a type of low default portfolio particularly concentrated in the Grand Duchy of Luxembourg.
Using an international database featuring 720 exchange traded funds over 10 years, this thesis develops a model based on funds’ intrinsic variables to estimate the probability of default for investment funds. By combining relevant performance measures with additional variables such as the age of a fund, the model constructed using a logistic regression computes reliable probability of default estimates. Indeed, the mean of the estimated probabilities of default for the surviving funds amounts to 0.83 percent while the mean of the estimated probabilities of default for the defaulted funds amounts to 74.14 percent. Besides, the model turns out to significantly differentiate surviving funds from defaulted ones and correctly classifies 97.09 percent of the funds.

Probability of default --- Low default portfolios --- Investment funds --- Sciences économiques & de gestion > Finance

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

This research thesis is based on a theoretical model that analysed how a CRAs’ incentives to
provide ratings of high quality vary depending on which economic cycle (boom or recession)
exists. From the results it appears that ratings quality is lower during booms than during
recessions. This is known as the counter-cyclicality of ratings and is due to incentives to milk
a reputation. A distinction is also made between a monopoly and a duopoly and it is clear that
in both cases the counter-cyclical ratings quality holds.
The aim of this master’s thesis is to analyse by means of a stress test how this model is sensitive
to a modification of three parameters. The three parameters are gamma which is the parameter
that captures labor-market conditions, lambda that is the probability that an investment is good
and tau that is the probability of a transition from the current state to another one. Again, a
distinction is made between a monopoly and a duopoly.

Sciences économiques & de gestion > Finance