Financial Econometrics
Problems, Models, and Methods
Financial
econometrics is the adaptation of statistical and time series methods originally
pioneered for economics (the field of econometrics) to financial applications.
It is largely an academic—as opposed to practitioner—field, with applications
found in studies of market efficiency, enhanced capital asset pricing models,
and studies of market microstructure. The field has grown tremendously over the
past 25 years, with important results making their way into practitioners'
toolkits. Most notable of these are ARCH and GARCH models for time-varying
volatility. These originated in econometric work of the early 1980s.
Some
texts on financial econometrics focus on modeling techniques. Others focus on
empirical results obtained by applying those techniques to financial data. This book by Gourieroux and Jasiak focuses on modeling techniques
presented in the context of financial models such as the capital asset pricing
model (CAPM). Hence, it is
a time series analysis text that emphasizes techniques that are applicable to
finance. One thing that makes this book stand out from other books on time
series analysis is its in-depth treatment of continuous-time processes and
stochastic calculus. In practice, financial engineers need to approximate
continuous-time processes with discrete-time processes and conversely. The
integrated treatment offered by this book provides excellent information on
doing so.
Compared to other
books on time series analysis, this is an intermediate level text. It is
rigorous without being tedious. Proofs are provided if they are simple and offer
insights. Otherwise, results are just stated. More than anything else, this book
offers a bridge for researchers or quantitative professionals between
introductory texts and current research in
time series analysis
financial engineering, and
market risk measurement.
The book touches
on many topics in financial econometrics. Most chapters are relevant to
finance. A few are more applicable to economics, including Chapters 4, 7 and 10.
Those chapters can easily be skipped without loss of continuity.
After a
poorly written
introductory chapter, two chapters cover moving average, autoregressive, and
autoregressive moving average (MA, AR and ARMA) models in one and multiple
dimensions. The discussions are excellent for someone who is already familiar
with these models but wants deeper practical insights into topics such as
estimation and stability.
Chapter 5 covers
persistence and cointegration. Both topics are important for finance, and the
discussions are excellent, although brief. This chapter offers the most
accessible treatment of cointegration I have seen anywhere.
Contents
1. Introduction
2. Univariate
Linear Models
3. Multivariate Linear Models: VARMA
Representation
4. Simultaneity, Recursivity, and Causality
Analysis
5. Persistence and Cointegration
6. Conditional Heteroscedasticity
7. Expectation and Present Value Models
8. Intertemporal Behavior and the Method of
Moments
9. Dynamic Factor Models
10. Dynamic Qualitative Processes
11. Diffusion Models
12. Estimation of Diffusion Models
13. Econometrics of Derivatives
14. Dynamic Models for High-Frequency data
15. Market Indexes
16. Management of Extreme Risks
Chapter 6
covers models for conditional heteroskedasticity, including ARCH, GARCH,
nonlinear autoregressive, and stochastic volatility models. There is plenty of
practical information here, but prior familiarity with this topic will be
helpful.
Chapter 8
covers intertemporal models, which are primarily applicable to certain
multi-period versions of the CAPM but may also have relevance for pricing
path-dependent derivatives.
Several chapters
cover continuous time models, stochastic calculus, and applications to
derivatives pricing. These are excellent. Readers will benefit from having some
familiarity with the subject, perhaps obtained from Chriss (1997),
Baxter and Rennie (1996), and Seydel
(2002). I like these chapters because they are
reasonably rigorous without depending on measure theoretic arguments. They
closely link discrete time methods discussed elsewhere in the book with the
continuous time methods. This aids intuition and is important for model
specification—say, specifying a continuous time model based upon discrete time
data.
Other nice chapters discuss the analysis of
high-frequency data and extreme value distributions.
Be aware that this is a theoretical text written by
academics who work for economics departments as opposed to finance departments
at their respective universities. While they offer some wonderful insights into how the
mathematics relates to financial issues, they also reveal a lack of
practical experience. They insist on detailing time series methods that might be
used to predict future prices. The techniques are borrowed from economics
practice, where future values of time series can often be predicted to some
extent. In the context of finance, they fly in the face of the weak efficient
market theory. Particularly unfortunate are discussions that appear to
endorse technical analysis—proposing how chartistry might be "improved" upon.
The book does present a lot of cutting-edge material in time series analysis,
but it does rather leave it up to the reader to sort out what is useful and what
is not.
Who is this book
for? It will appeal to researchers, financial engineers, and
theoretically-inclined risk managers. It will also appeal to sophisticated
quantitative professionals working in investment management. Readers should have
strong quantitative skills. They should have some prior experience with time
series analysis and knowledge of statistics. The necessary background can be
obtained from Franses (1994), Harvey (1993),
and Casella and Berger (2001).
.gif)
time series analysis