Bayesian Statistical Computations of Nonlinear Financial Time Series Models: A Survey with Illustrations

A survey of Bayesian statistical computations of quadratureformula, Laplace approximation, and Markov Chain Monte Carlo algorithms ispresentedand their applications to nonlinear financial time series models arediscussed.     

Asian Financial Crisis.Prologue and the Case of Thailand

We suggest that the Asian financial crisis began because of theinconsistent exchange rate system and miss-management within Thailand. We showthat prior tothe Asian financial crisis the Thai Baht against the Yen followed the theoryofpurchasing power parity but the Baht against the U.S. Dollar did not. We givebriefsummaries of the papers included in this issue.     

The relationships between stress,health, and physically active leisure as a function of life‐cycle

This article examines relationships between stress, health, and participation in physically active leisure using data collected as part of the 1985 and 1990 U.S. National Health Interview Surveys (NHIS). In particular, the article analyzes: (a) the effects of life‐cycle progression (controlled for gender) on subjectively perceived stress and self‐reported health; (b) the role of physically active leisure in enhancing health or buffering adverse health outcomes of stress; and (c) the effects of selected social‐economic factors, such as education and income, on the relationships between life‐cycle, stress, health, and physically active leisure. According to our analyses: (1) stress is unevenly distributed across the life span; (2) stress‐reducing effects of physically active leisure are present among retired older people more so than in the middle‐aged life‐cycle groups; and (3) direct health‐enhancing effects of participation in physically active leisure are more evident than the buffering effects of leisure on stress‐health relationships. A critical discussion of methodological and substantive issues facing researchers interested in the relationships between leisure, stress, and health is provided.     

New product introduction under demand uncertainty in competitive industries

A game-theoretic model is employed to examine the conditions under which firms in a duopoly, faced with a new product introduction opportunity of uncertain profitability (because of uncertainty in demand), choose to enter the market immediately or, alternatively, decide to wait, thereby avoiding the risk of failure. The implications for a firm's strategic behavior are contrasted with situations in which the competitor is believed to be (1) passive, implying that the firm expects to enjoy indefinite monopoly status if it introduces the new product, and (2) committed to a waiting strategy, implying monopoly status for a limited time period, if the new product is successful.     

Use of the principal component method in the maximum likelihood estimation procedure of the logit model

The maximum likelihood estimates of a qualitative response market are solved by an iterative procedure. When severe multicollinearity exists among the explanatory variables, this procedure may fail to converge. In this note, using a bank-failure model, we demonstrate how the principle component method can make the iterative procedure converge when it fails to do so in the untransformed model.     

A Bayesian test of the product cycle hypothesis applied to Japanese crude steel production

It is postulated that the product cycle of crude steel production may be observed as a gradual shift in the time trend parameter in the family of the CES production functions. Quarterly data from 1957 to 1973 are used for the empirical test. Posterior pdf's for the join point as well as for the adjustment parameter are derived, and they indicate that Japanese crude steel production passed from the new to standardized product stage between 1968 and 1972. With the arrival of the standardized state it is noted that the industry has become less competitive.     

On statistical inference on the mean-variance efficient set

A proof of the asymptotic distribution of the estimated mean-variance frontier is given. A Bayesian prediction interval is derived for the capital asset pricing model. Numerical illustrations show that the prediction intervals for the CAPM are smaller than those for the constant mean model, if the fit of the CAPM is better than that of the constant mean model.