spb/phd_thesis Public

1\chapter*{Abstract}             % ne pas numéroter2\label{chap-abstract}           % étiquette pour renvois3\phantomsection\addcontentsline{toc}{chapter}{\nameref{chap-abstract}} % inclure dans TdM45% BROUILLON À RÉVISER PAR L'AUTEUR.6\begin{otherlanguage*}{english}7  This dissertation consists of three essays that use high-frequency data to8  study return and volatility dynamics in commodities and financial futures9  markets. The common thread is the minute-level identification of the10  mechanisms through which information --- macroeconomic announcements, trader11  positioning, and central bank communications --- is transmitted to prices12  and volatility, mechanisms that daily data cannot disentangle.1314  The first essay asks whether the intensity of speculative trading in energy15  futures markets amplifies or dampens the reaction of prices to macroeconomic16  news. Using 5-minute data on six U.S. futures contracts (crude oil, natural17  gas, gold, silver, copper, and palladium) from April 2007 to February 2024,18  standardized surprises from 26 macroeconomic announcements, and a19  time-varying measure of speculative intensity built from disaggregated CFTC20  positions, the essay shows that higher speculative activity \emph{dampens}21  the response of returns and volatility to surprises and narrows bid-ask22  spreads. These stabilizing effects --- stronger for procyclical commodities23  than for gold, a safe haven --- are driven by money managers rather than24  swap dealers, suggesting that speculative trading improves liquidity and25  price discovery.2627  The second essay builds a novel dataset of minute-level indicative Net Asset28  Value (iNAV) observations for four commodity exchange-traded funds (ETFs)29  --- gold, silver, crude oil, and natural gas --- and measures volatility30  transmission between each ETF and its underlying basket. Decomposing31  realized variance into continuous and jump components and combining HAR32  models with Bayesian vector autoregressions, the essay establishes that33  transmission runs primarily through jumps rather than diffusion, that it is34  substantially larger in 1-minute than in 30-minute data, and that its35  direction depends on the arbitrage mechanism: unidirectional from iNAV to36  ETF for physically backed precious metals, bidirectional and asymmetric for37  futures-based energy ETFs.3839  The third essay decomposes FOMC statements into two semantic dimensions ---40  policy tone (hawkish/dovish) and informational novelty relative to the41  previous statement --- using an ensemble of language models (MiniLM and42  BERT) trained on Federal Reserve communications with data-driven, PCA-based43  reference selection. Across 148 FOMC announcements (2008--2025) and 1-minute44  data for seven futures contracts, the essay shows that tone predicts45  directional returns while novelty predicts volatility, with the46  tone--novelty interaction on VIX being the most robust effect. The two47  channels display opposite dynamics --- rapid uncertainty resolution for48  volatility, gradual two-hour repricing for returns --- a pattern49  inconsistent with a single information channel.5051  Together, the essays show that intraday granularity is not a technical52  refinement but a condition for identification: the magnitude, direction, and53  even sign of the documented effects would be invisible or biased at the54  daily frequency. The results inform policy debates on speculative position55  limits, index-product design, and central bank communication strategy by56  identifying who transmits information to prices, through which channel, and57  at what horizon.58\end{otherlanguage*}59