Our campuses are closed but thesis defences are proceeding normally, albeit remotely. Refer to our COVID-19 FAQs for more information.
When studying for a doctoral degree (PhD), candidates submit a thesis that provides a critical review of the current state of knowledge of the thesis subject as well as the student’s own contributions to the subject. The distinguishing criterion of doctoral graduate research is a significant and original contribution to knowledge.
Once accepted, the candidate presents the thesis orally. This oral exam is open to the public.
This thesis organized in three chapters, essentially covers two main fields: finance and econometric theory with an application to macroeconomics. The first chapter proposes a methodology to uniquely measure price discovery, the mechanism by which the price of a security or an asset cross-listed on multiple markets is determined. The second chapter develops an information criterion that remains robust in presence of weaker instruments. Finally, the third chapter illustrates the benefits of optimal instruments' selection in assessing the impact of an example of monetary policy.
Being a process that allows market participants to uncover the real worth of an asset in a timely manner, the price discovery may lead to arbitrage opportunities. As such, the Information Share (IS) commonly used to measure it, needs to be as accurate as possible to help mitigate related market inefficiencies. In the first chapter of this thesis, we investigate the identification issues encountered by the IS due to its sensitivity to price ordering. This translates to price innovations vectors leading to a serious lack of robustness of the IS metric. Exploiting some statistical features of price innovations, we propose to use Independent Component Analysis (ICA) in order to decompose the residuals into independent signals. Compared to leading measures in the literature, our approach is shown to perform well in the standard two-market data framework. We also obtain consistent results while extending our simulations to larger number of markets framework, notably the three-market set-up. We finally confirm our findings by studying the mechanism of price discovery in two analogous empirical applications. The first analyzes futures and spot prices in the European Union Allowances (EUAs) market for CO2 emissions, and the second concentrates on three Exchange Traded Funds (ETFs) tracking the performance of the Russell 2000 index. Our evidence suggests that futures prices and the IWM (ETF issued by iShares), respectively dominate their companions in contribution to price discovery.
The second chapter is motivated by the fact that the usual exogeneity assumption is essential to the least squares estimator as it guarantees its consistency. However, when this condition fails, the explanatory variable is said to be endogenous and Instrumental Variable (IV) regressions is one of the methods available to the researcher to obtain consistent estimates. In response to the importance of the instruments selection step in the construction of a good IV estimator, we propose the alternative Relevant Moment Selection Criterion (aRMSC). This information criterion improves model selection when instruments are only weakly correlated with the endogenous variable. Through Monte Carlo simulations, we first illustrate that existing information criteria are not robust to these types of issues; naively selecting the larger models. We benefit from recent development on the importance of the strength of identification in achieving efficient estimation, and leverage it to evaluate how this may affect instruments selection when the candidate instruments available to the researcher are equally weakened or a pool of instruments with various strengths. Our evidence suggests that despite their weakness some instruments still contribute to improving the estimator's efficiency, in such a way that the selection of the most parsimonious model is possible.
In the final chapter of this thesis, we first illustrate the performance of our proposed information criterion in a macroeconomic application. Moreover, we study empirically the relationship between news from forward guidance and monetary policy. We account for interactions both between various macroeconomic variables while considering their own lagged values using a structural vector autoregressive (VAR) model including interest rates, consumer price index, industrial production and excess bond premium. Our analysis relies on Gertler and Karadi’s (2015) high frequency identification (HFI) approach for monetary policy shocks to extend monetary policy indicators to the 2 year government bond rate even though authors initially considered it as facing weak instruments issues. The aRMSC allows us to identify relevant instruments in the VAR model with the 2 year government bond rate and compare our results to those predicted with the 1 year rate, a stronger instrument. We also consider the limited information maximum likelihood estimator (LIML) to improve the instruments' selection. All together, our results highlight that the model based on the optimal set of instruments in comparison to the model with the naive inclusion of all instruments from the candidate set, produces more accurate impulse responses for economic and financial variables regardless of the estimator used to obtain the alternative information criterion.