The compilation of articles in this book gives a wide overview of the many techniques that are used in economic forecasting and each article compares the virtues of its approach with others that are used. Some readers, especially those who are just beginning their careers in econometrics and financial modeling may believe that economic models form a hierarchy, with the best ones being on top while the marginal ones are at the bottom. In addition, some may believe that more modern approaches are better than the ones taken several decades ago or conversely that no improvements have been made in economic modeling since it was established as a profession. As an example, recently a former chairman of the Federal Reserve stated that the economic forecasting tools of today are no better than ones in existence fifty years ago. He offered no evidence for this claim, no doubt because of the gargantuan amount of effort it would take to establish it. Indeed, to compare econometric models requires agreement on what constitutes a "valid" or "good" model, and once this is settled one frequently requires another model to do the comparisons. In addition, models are built for particular situations, domains, and contexts, and it is very common for a model to work much better than another in one context but fail miserably when compared to the other in another context. Ordinary linear regression for example can be better than more "sophisticated" approaches like neural networks in some areas of application. A more complex model is therefore not necessarily better than one that is relatively simple. So economic models do not form a hierarchy under any nontrivial classification of merit, nor can it be said that no progress has been made in economic modeling over the past fifty years.

The approaches to economic modeling as outlined in this book definitely support the idea that there is no free lunch when it comes to forecasting. It is the context that governs the efficacy of one model over another, and it might be said with fairness that the ability to select the proper model for this context comes with experience. This experience is definitely reflected in the authors that have composed the articles in this book. Readers will probably not read every article in the book, but will instead select those that interest them or those that show the most practical promise.

For this reviewer, some of the highlights of this work include:
* The discussion of the two principles behind Bayesian forecasting: the principle of explicit formulation and the principle of relevant conditioning. The later principle is always violated by non-Bayesian forecasting techniques
* A view of economic models of being "means", not "ends". This distinction is one to be kept in mind especially at the present time where financial modeling and economic forecasting is being blamed for much, if not all, of the turmoil in the financial markets.
* The discussion of the importance and need for discarding irrelevant information when doing simulations of joint distributions. As discussed in the book the justification for this omission can be given a sound, quantitative foundation.
* The importance of doing simulations rather than finding analytical solutions in economic modeling.
* More in-depth discussion on how to choose the prior distribution in Bayesian economic forecasting, thus removing some of the objections of this selection always being "purely subjective."
* The discussion on `hyperparameters' and their connection to latent variables and `hierarchical prior distributions.' These notions have recently been applied to forecasting of housing prices. The book mentions many other applications.
* The discussion on the `Bayes factor' and its use in assessing the evidence in favor of one economic model versus another. Recently the notion of a Bayes factor has been generalized in the field of artificial intelligence, wherein it is START TRANSACTION WITH CONSISTENT SNAPSHOT; /* 9d3ac415194ff8f960c4f613efd5a453