Michiel De Pooter

Predicting the Daily Covariance Matrix for S&P 100 Stocks using Intraday Data - But Which Frequency to Use?

This article investigates the merits of high-frequency intraday data when forming mean-variance efficient stock portfolios with daily rebalancing from the individual constituents of the S&P 100 index. We focus on the issue of determining the optimal sampling frequency as judged by the performance of these portfolios. The optimal sampling frequency ranges between 30 and 65 minutes, considerably lower than the popular five-minute frequency, which typically is motivated by the aim of striking a balance between the variance and bias in covariance matrix estimates due to market microstructure effects such as non-synchronous trading and bid-ask bounce. Bias-correction procedures, based on combining low-frequency and high-frequency covariance matrix estimates and on the addition of leads and lags do not substantially affect the optimal sampling frequency or the portfolio performance. Our findings are also robust to the presence of transaction costs and to the portfolio rebalancing frequency.

Modeling and Forecasting S&P 500 Volatility: Long Memory, Structural Breaks and Nonlinearity

The sum of squared intraday returns provides an unbiased and almost error-free measure of ex-post volatility. In this paper we develop a nonlinear Autoregressive Fractionally Integrated Moving Average (ARFIMA) model for realized volatility, which accommodates level shifts, day-of-the-week effects, leverage effects and volatility level effects. Applying the model to realized volatilities of the S&P 500 stock index and three exchange rates produces forecasts that clearly improve upon the ones obtained from a linear ARFIMA model and from conventional time-series models based on daily returns, treating volatility as a latent variable.

Term Structure Forecasting Using Macro Factors and Forecast Combination

We examine the importance of incorporating macroeconomic information and, in particular, accounting for model uncertainty when forecasting the term structure of U.S. interest rates. We start off by analyzing and comparing the forecast performance of several individual term structure models. Our results confirm and extend results found in previous literature that adding macroeconomic information, through factors extracted from a large number of individual series, tends to improve interest rate forecasts. We then show, however, that the predictive power of individual models varies over time significantly. Models with macro factors are the more accurate in and around recession periods. Models without macro factors do particularly well in low-volatility subperiods such as the late 1990s. We demonstrate that this problem of model uncertainty can be mitigated by combining individual model forecasts. Combining forecasts leads to encouraging gains in predictability, especially for longer-dated maturities, and importantly, these gains are consistent over time.

Predicting the Term Structure of Interest Rates: Incorporating Parameter Uncertainty, Model Uncertainty and Macroeconomic Information

We forecast the term structure of U.S. Treasury zero-coupon bond yields by analyzing a range of models that have been used in the literature. We assess the relevance of parameter uncertainty by examining the added value of using Bayesian inference compared to frequentist estimation techniques, and model uncertainty by combining forecasts from individual models. Following current literature we also investigate the benefits of incorporating macroeconomic information in yield curve models. Our results show that adding macroeconomic factors is very beneficial for improving the out-of-sample forecasting performance of individual models. Despite this, the predictive accuracy of models varies over time considerably, irrespective of using the Bayesian or frequentist approach. We show that mitigating model uncertainty by combining forecasts leads to substantial gains in forecasting performance, especially when applying Bayesian model averaging.

Bayesian near-boundary analysis in basic macroeconomic time series models

Several lessons learnt from a Bayesian analysis of basic macroeconomic time series models are presented for the situation where some model parameters have substantial posterior probability near the boundary of the parameter region. This feature refers to near-instability within dynamic models, to forecasting with near-random walk models and to clustering of several economic series in a small number of groups within a data panel. Two canonical models are used: a linear regression model with autocorrelation and a simple variance components model. Several well-known time series models like unit root and error correction models and further state space and panel data models are shown to be simple generalizations of these two canonical models for the purpose of posterior inference. A Bayesian model averaging procedure is presented in order to deal with models with substantial probability both near and at the boundary of the parameter region. Analytical, graphical and empirical results using U.S. macroeconomic data, in particular on GDP growth, are presented.

Examining the Nelson-Siegel Class of Term Structure Models: In-Sample Fit versus Out-of-Sample Forecasting Performance

In this paper I examine various extensions of the Nelson and Siegel (1987) model with the purpose of fitting and forecasting the term structure of interest rates. As expected, I find that using more flexible models leads to a better in-sample fit of the term structure. However, I show that the out-of-sample predictability improves as well. A four-factor model, which adds a second slope factor to the three-factor Nelson-Siegel model, forecasts particularly well. Especially with a one-step state-space estimation approach the four-factor model produces accurate forecasts and outperforms competitor models across maturities and forecast horizons. Subsample analysis shows that this outperformance is also consistent over time.

On the Practice of Bayesian Inference in Basic Economic Time Series Models using Gibbs Sampling

Several lessons learned from a Bayesian analysis of basic economic time series models by means of the Gibbs sampling algorithm are presented. Models include the Cochrane-Orcutt model for serial correlation, the Koyck distributed lag model, the Unit Root model, the Instrumental Variables model and as Hierarchical Linear Mixed Models, the State-Space model and the Panel Data model. We discuss issues involved when drawing Bayesian inference on regression parameters and variance components, in particular when some parameter have substantial posterior probability near the boundary of the parameter region, and show that one should carefully scan the shape of the posterior density function. Analytical, graphical and empirical results are used along the way.

The design and production of new retirement savings products: a note

An earlier contribution to this journal implied that protective floor investment strategies in defined-contribution pension investments beat investments in traditional asset mixes, with respect to both the probability of achieving a minimum target level of future pension income and the expected level of future pension income. These commentators show that protective floor strategies may be attractive to loss-averse investors, but they do not beat traditional investments simultaneously on expected pension income and on the probability of achieving the target pension income.