Understanding and forecasting financial time series depend crucially on identifying any non-linearity which may be present. Recent developments in tests for non-linearity very commonly display low power, most likely because of over-smoothing and discarding pertinent information. In this...

Evaluation of forecast optimality in economics and finance has almost exclusively been conducted under the assumption of mean squared error loss. Under this loss function optimal forecasts should be unbiased and forecast errors should be serially uncorrelated at the single period horizon with...

Under the squared error loss, the optimal forecast is the conditional mean, and the one-step forecast error is a martingale difference (MD). The one-step forecast error forms the conditional moment condition obtained from the loss derivative with respect to the forecast. Similarly, under a...

Recently there have been much discussion of the theory and applications of long memory processes. In this paper we consider the standard linear model y=X*b+u and assume that the variance covariance matrix of the errors being generated from an ARFIMA(0,d,0) model. Following Banerjee and Magnus...

A common problem in out-of-sample prediction is that there are potentially many relevant predictors that individually have only weak explanatory power. We propose bootstrap aggregation of pre-test predictors (or bagging for short) as a means of constructing forecasts from multiple regression...

This paper analyzes the application of the Markov-switching ARCH model (Hamilton and Susmel, 1994) in improving value-at-risk (VaR) forecast. By considering a mixture of normal distributions with varying variances over different time and regimes, we find that the “spurious high...

When a price limit regime exists for all of the stocks involved in an index, the index return is an aggregate of limited variables and thereby it is restricted to the same limits. We argue that neither a censored nor a truncated distribution model is appropriate for the aggregate return. The...

Two approaches dominate the time series literature for modeling expected value models. The first one is based on observable variables and includes ARMA and GARCH models, while the second one is based on latent variables and includes state space and stochastic volatility (or SV) models. The first...

This paper proposes bootstrap versions of the seasonal unit root tests of, inter alia, Hylleberg, Engle, Granger and Yoo (1990,Journal of Econometrics 55, 305-328)[HEGY]. We report a simulation study of the properties of both the conventional and bootstrapped seasonal unit root tests when...

In real time forecasting, the sample is usually split into an estimation period of R observations and a prediction period of P observations, where T=R+P. Parameters are often estimated in a recursive manner, initially using R observations, then R+1 observations and so on until T-1 observations...