\chapter*{Introduction}         % ne pas numéroter
\label{chap-introduction}       % étiquette pour renvois
\phantomsection\addcontentsline{toc}{chapter}{\nameref{chap-introduction}} % inclure dans TdM

% BROUILLON À RÉVISER PAR L'AUTEUR — introduction générale de la thèse
% (rédigée à partir des trois articles ; les renvois bibliographiques utilisent
% la bibliographie consolidée de la thèse).

Futures markets are where much of the world's price discovery takes place.
Crude oil, natural gas, gold, equity indexes, Treasury notes: for each of
these assets, the futures contract is typically the most liquid instrument,
the first to react to news, and the reference point from which spot prices,
inventories, and investment decisions are set. Understanding how information
is impounded into futures prices --- and how that process shapes volatility
--- is therefore central to asset pricing, to market design, and to a series
of policy debates that have accompanied the transformation of these markets
over the past two decades: the growth of speculative and index-based trading
in commodities \citep{tang2012index, cheng2014financialization}, the rise of
exchange-traded funds (ETFs) as the retail gateway to commodity exposure, and
the increasing weight of central bank communication as a market-moving event
in its own right.

A common empirical obstacle runs through these debates. Information arrives
in minutes, but much of the literature has measured its effects in days.
Daily data conflate the announcement of interest with everything else that
happens the same day, attenuate estimates through microstructure noise and
intraday reversals, and are silent about the dynamics --- amplification,
absorption, resolution of uncertainty --- that unfold in the first minutes
and hours after an information event \citep{andersen2003micro,
kothari2007econometrics}. The high-frequency literature pioneered by
\citet{andersen1998deutsche} and extended by realized-measure econometrics
\citep{andersen2001distribution, barndorff2004power, corsi2009simple} showed
that intraday data are not merely more precise: they change what can be
identified. Event windows measured in minutes isolate the causal effect of a
single release; realized variance turns latent volatility into an observable;
and the separation of continuous and jump components distinguishes gradual
information diffusion from discrete repricing.

This dissertation applies that identification strategy to three questions
about commodities and financial futures markets. Each essay constructs a new
high-frequency dataset, each exploits a well-defined information event or
transmission channel, and each documents effects whose magnitude, direction,
or even sign would be invisible at the daily frequency. The common thread is
the same throughout: \emph{who} transmits information to prices, through
\emph{which channel}, and at \emph{what horizon}.

\section*{Essay 1: Speculative trading and macroeconomic surprises in energy
  markets}

The first essay addresses one of the most persistent controversies in
commodity market policy: whether speculative trading destabilizes energy
markets. The commodity price run-up of 2004--2008 spawned an influential
narrative --- the ``Masters hypothesis'' --- according to which the growing
presence of financial traders amplifies price movements and volatility, with
harmful consequences for the real economy. The theoretical literature shows
that such distortions are possible \citep{basak2016model,
goldstein2022commodity}, but the empirical evidence, largely based on daily
data and Granger-causality designs, is inconclusive \citep{fattouh2013role,
irwin2012testing}.

The essay takes a different route. Macroeconomic announcement releases
provide sharply identified, exogenous information shocks whose surprise
component can be standardized following \citet{balduzzi2001economic}. Using
5-minute futures data for two energy commodities (crude oil and natural gas)
and four metals (gold, silver, copper, and palladium) from April 2007 to
February 2024, together with a time-varying, commodity-specific measure of
speculative intensity built from the CFTC's disaggregated Commitments of
Traders data, the essay asks whether the reaction of returns, volatility, and
bid-ask spreads to macroeconomic surprises is amplified or dampened when
speculative trading is more intense. The answer is unambiguous: the
interaction between surprises and speculative intensity is consistently
stabilizing. Prices and conditional volatility react \emph{less} to surprises
when speculative positions are larger, and bid-ask spreads narrow rather than
widen. Disaggregating trader categories shows that these beneficial effects
are driven by money managers --- informed traders in the sense of
\citet{cheng2015convective} --- rather than swap dealers. For energy markets,
where volatility feeds directly into household energy costs and investment
under uncertainty, the finding that speculation dampens rather than amplifies
news-driven volatility speaks directly to the design of position limits and
to the broader regulatory debate.

\section*{Essay 2: The iNAV and volatility transmission between commodity
  ETFs and their underlying assets}

The second essay turns from the futures market itself to the fastest-growing
channel through which investors access commodities: exchange-traded funds. An
ETF is an equity-like wrapper around an underlying basket --- physical
bullion for gold and silver funds, futures positions for oil and natural gas
funds --- and the arbitrage mechanism that keeps the wrapper aligned with its
contents is also a conduit for volatility \citep{ben2018etfs}. Measuring that
conduit precisely requires observing the value of the basket at the same
frequency as the ETF price. The essay's central contribution is the
construction of a novel minute-level dataset of indicative net asset values
(iNAV) for four commodity ETFs (GLD, SLV, USO, UNG), validated against
official NAVs and exchange-disseminated values, over more than a decade.

With ETF and basket volatilities observable minute by minute, the essay
measures transmission using HAR-type cascade models \citep{corsi2009simple}
extended with cross-asset terms, decomposes realized variance into continuous
and jump components, and complements the analysis with Bayesian vector
autoregressions. Three results stand out. First, the direction of
transmission mirrors the arbitrage technology: for physically backed precious
metal funds, volatility flows one way, from the basket to the ETF; for
futures-based energy funds, transmission is bidirectional and asymmetric.
Second, transmission operates primarily through jumps rather than through the
continuous component --- a channel that standard connectedness measures,
built on daily total volatility, largely obscure. Third, sampling frequency
matters quantitatively: 1-minute data reveal transmission up to twice as
large as 30-minute estimates. Beyond the substantive findings, the essay
demonstrates that the informational content of the iNAV --- a series
disseminated in real time but rarely archived or studied --- provides a
sharper image of the ETF--underlying relationship than price-based proxies.

\section*{Essay 3: Policy tone, informational novelty, and the
  high-frequency reaction to FOMC announcements}

The third essay studies the information events that arguably matter most for
financial futures: announcements of the Federal Open Market Committee. A
large literature measures monetary policy surprises through interest-rate
futures \citep{kuttner2001, gurkaynak2005} or through the joint reaction of
rates and equities \citep{nakamura2018}. Yet FOMC statements are texts, and
their market impact depends not only on what they imply for the policy rate
but on how they say it. The essay decomposes each statement into two semantic
dimensions: \emph{policy tone} --- the hawkish-versus-dovish orientation of
the language --- and \emph{informational novelty} --- the distance between a
statement and its predecessor. The measurement apparatus is a dual-model
ensemble of transformer language models (MiniLM and BERT) fine-tuned on
Federal Reserve communications, with reference statements selected by a
data-driven, PCA-based procedure that removes researcher discretion.

Combining these measures with 1-minute data for seven futures contracts ---
equities, VIX, Treasuries, the dollar, crude oil, and gold --- across 148
FOMC announcements between 2008 and 2025, the essay documents a clean
division of labour between the two dimensions. Tone predicts directional
returns: a one-standard-deviation dovish shift is associated with equity
gains that accumulate to roughly twelve basis points over two hours. Novelty
predicts volatility, with a sign opposite to the naive prediction: statements
that depart more from their predecessor \emph{reduce} uncertainty, consistent
with communication acting as clarification rather than noise. The
tone--novelty interaction on VIX futures is the most robust effect in the
paper, and pre-announcement placebo tests together with five independent
inference methods support a causal reading. The two channels also operate on
different clocks --- volatility effects resolve within roughly twenty
minutes, return effects build for two hours --- a pattern inconsistent with
any single-channel account of central bank communication.

\section*{Contributions and structure of the dissertation}

Taken together, the three essays make three kinds of contributions. First,
they build and document new high-frequency datasets --- standardized
macroeconomic surprises matched to 5-minute commodity futures, a minute-level
iNAV panel for commodity ETFs, and a semantically decomposed corpus of FOMC
communications aligned with 1-minute futures data --- each of which has value
beyond the questions asked here. Second, they deliver identified answers to
long-standing questions: speculation stabilizes rather than destabilizes
energy futures around news; ETF--underlying volatility transmission is
jump-driven and direction-dependent; and central bank communication moves
returns through tone but volatility through novelty. Third, they carry policy
implications --- for the calibration of speculative position limits, for the
design and monitoring of commodity index products, and for the communication
strategy of central banks.

The remainder of the dissertation is organized as follows.
Chapter~\ref{chap-chapitre1} presents the first essay, on speculative trading
and macroeconomic surprises in energy futures markets.
Chapter~\ref{chap-chapitre2} presents the second essay, on iNAV-based
volatility transmission between commodity ETFs and their underlying baskets.
Chapter~\ref{chap-chapitre3} presents the third essay, on the high-frequency
effects of the tone and novelty of FOMC statements. A general conclusion
summarizes the findings, discusses their limitations, and outlines avenues
for future research. Appendices~\ref{ch3:sec:proofs}
and~\ref{ch3:sec:app_additional} collect the mathematical proofs and
additional results of Chapter~\ref{chap-chapitre3}.