Notes on economic time series analysis: system theoretic perspectives (Q789283)

During the last two decades time series analysis techniques have become an important part in applied econometrics. Besides its useful and fruitful supplementation of standard econometric techniques, time series analysis has a justification of its own. But most of its applications in economics have been restricted to the time domain approach. Only in the recent past, with the introduction of Kalman filtering into the area of economic research, an alternative view of interpreting time series analysis techniques - the system theoretic perspective - became more important. For economists the system theoretic perspective of time series analysis is still uncommon; so it is the attempt of Aoki's book to help bridge the gap between the vector valued time series approach and the techniques which are available in recent system literature. Due to this intention, the main interest of the book consists of interpreting system theoretic results and perspectives in time series by filtering out non-essential details. So many results on various estimation methods and their statistical properties are excluded from consideration in the book. The book is divided into three parts: Chapters 1 through 6 preparatory to the main part of the note, while chapters 7 through 10 constitute the main body of the book and chapters 11 through 13 deal with some special topics such as intertemporal optimization, identification of closed-loop systems, and rational expectation models. The final part, chapter 14, discusses some numerical examples. The book contains 248 pages, 64 of which constitute the very detailed appendix. Referring to the first part of the book, after the introduction, the notion of state is given in chapter 2. Time-invariant linear dynamics are discussed in the following chapter, which includes continuous time systems, inverse systems, and discrete-time sequences. The time series representation is presented in chapter 4, and chapter 5 shows the equivalence of ARMA and state space models. The final chapter of the first part is concerned with the decomposition of data into cyclical and growth components. The main part of the book starts with the prediction of time series (chapter 7). After introducing and discussing the prediction space, which is the main part of the chapter, the problem of equivalence of two Markovian models is discussed briefly. - The next chapter 8 introduces the spectra and covariances and presents a factorization of the spectral density matrices. Computational aspects of calculating the sample covariance matrices end the chapter. Chapter 9 deals with the initial phase of estimating system matrices. An important aspect of estimation refers to well-behaved data and thus to the appropriate scalings of the variables. To construct state space models which are numerically well- behaved, ''internally balanced''-models are considered, (section 9.4.). Their properties are discussed and the relationship to principal component analysis is shown. Additional topics of chapter 9 are an example of a state space model and an ARMA model. As a final aspect the method of estimation is connected to the canonical correlation method of Akaike. - Innovation processes build up chapter 10. Within this context, orthogonal projections and Kalman filters are presented first, followed by a discussion of causal invertibility of innovation models. In the Kalman filter calculations noise covariance information is used. An alternative to this approach is given in section 10.4. - Chapter 11 is concerned with time series from intertemporal optimization. The chapter starts with example of dynamic resource allocation problems. The following section presents discrete time systems and analyzes the optimal solutions by analyzing the choice of weighting matrices. - Closed-loop systems are presented in chapter 12. The first section defines closed- loop systems and the next paragraph discusses the identifiability of a closed-loop system. - An application of system theoretic methods to rational expectations models is given in chapter 13. - The final part of the book discusses a two vector-valued time series model of the Japanese economy. The presented model is an innovation model of the foreign sector. The calculations are very detailed and the data matrix is also given so that the interested reader can go through the single steps of calculations very easily. - As already mentioned, the appendix is very extensive and covers various technical aspects such as Fourier- and z- transforms, solutions of difference equations, spectral decomposition representation, singular value decomposition theorem, dual relations, Hankel matrices etc. Although very briefly, the definitions and descriptions given in the appendix are completely sufficient to understand the system theoretic methods used in the main part of the book. Summing up, the book provides a useful overview over various methods used in system theory without complicating the basic issues by highly technical aspects, and it can also serve as a guideline for several economic applications using system theoretic methods.