nep-for New Economics Papers
on Forecasting
Issue of 2014‒09‒25
six papers chosen by
Rob J Hyndman
Monash University

  1. Forecasting Value-at-Risk Using High Frequency Information By Tae-Hwy Lee; Huiyu Huang
  2. Forecasting Copper Prices with Dynamic Averaging and Selection Models By Buncic, Daniel; Moretto, Carlo
  3. A Bayesian MIDAS Approach to Modeling First and Second Moment Dynamics By Davide Pettenuzzo; Rossen Valkanov; Allan Timmermann
  4. Adaptive Order Flow Forecasting with Multiplicative Error Models By Wolfgang Karl Härdle; Andrija Mihoci; Christopher Hian-Ann Ting;
  5. Data-based priors for vector autoregressions with drifting coefficients By Korobilis, Dimitris
  6. New Monthly Estimation Approach for Nowcasting GDP Growth: The Case of Japan By Naoko Hara; Shotaro Yamane

  1. By: Tae-Hwy Lee (Department of Economics, University of California Riverside); Huiyu Huang (GMO Emerging Markets)
    Abstract: In prediction of quantiles of daily S&P 500 returns we consider how we use high-frequency 5-minute data. We examine methods that incorporate the high frequency information either indirectly through combining forecasts (using forecasts generated from returns sampled at different intra-day interval) or directly through combining high frequency information into one model. We consider subsample averaging, bootstrap averaging, forecast averaging methods for the indirect case, and factor models with principal component approach for both cases. We show, in forecasting daily S&P 500 index return quantile (VaR is simply the negative of it), using high-frequency information is beneficial, often substantially and particularly so in forecasting downside risk. Our empirical results show that the averaging methods (subsample averaging, bootstrap averaging, forecast averaging), which serve as different ways of forming the ensemble average from using high frequency intraday information, provide excellent forecasting performance compared to using just low frequency daily information.
    Keywords: VaR, Quantiles, Subsample averaging, Bootstrap averaging, Forecast combination, High-frequency data.
    JEL: C53 G32 C22
    Date: 2014–09
  2. By: Buncic, Daniel; Moretto, Carlo
    Abstract: We use data from the London Metal Exchange (LME) to forecast monthly copper returns using the recently proposed dynamic model averaging and selection (DMA/DMS) framework, which incorporates time varying parameters as well as model averaging and selection into one unifying framework. Using a total of 18 predictor variables that include traditional fundamental indicators such as excess demand, inventories and the convenience yield, as well as indicators related to global risk appetite, momentum, the term spread, and various other financial series, we show that there exists a considerable predictive component in copper returns. Covering an out-of-sample period from May 2002 to June 2014 and employing standard statistical evaluation criteria we show that the out-of-sample R2 (relative to a random walk benchmark) can be as high as 18.5 percent for the DMA framework. Time series plots of the cumulative mean squared forecast errors and time varying coefficients show further that firstly, a large part of the improvement in the forecasts is realised during the peak of the financial crisis period at the end of 2008, and secondly that the importance of the most relevant predictor variables has changed substantially over the out-of-sample period. The coefficients of the SP500, the VIX, the yield spread, the TED spread, industrial production and the convenience yield predictors are most heavily affected, with the TED spread and yield spread coefficients even changing signs over this period. Our predictability results remain valid for forecast horizons up to 6 months ahead, but are weaker and smaller than at the one month horizon.
    Keywords: Copper forecasting, time varying parameter model, state-space modelling, dynamic model and selection models
    JEL: C11 C52 C53 G17
    Date: 2014–09
  3. By: Davide Pettenuzzo (International Business School, Brandeis University); Rossen Valkanov (University of California San Diego); Allan Timmermann (University of California San Diego)
    Abstract: We propose a new approach to predictive density modeling that allows for MI- DAS e¤ects in both the ?rst and second moments of the outcome and develop Gibbs sampling methods for Bayesian estimation in the presence of stochastic volatility dy- namics. When applied to quarterly U.S. GDP growth data, we ?nd strong evidence that models that feature MIDAS terms in the conditional volatility generate more accurate forecasts than conventional benchmarks. Finally, we ?nd that forecast combination methods such as the optimal predictive pool of Geweke and Amisano (2011) produce consistent gains in out-of-sample predictive performance.
    Keywords: MIDAS regressions; Bayesian estimation; stochastic volatility; out- of-sample forecasts; GDP growth.
    JEL: C53 C11 C32 E37
    Date: 2014–07
  4. By: Wolfgang Karl Härdle; Andrija Mihoci; Christopher Hian-Ann Ting;
    Abstract: A flexible statistical approach for the analysis of time-varying dynamics of transaction data on financial markets is here applied to intra-day trading strategies. A local adaptive technique is used to successfully predict financial time series, i.e., the buyer and the seller-initiated trading volumes and the order flow dynamics. Analysing order flow series and its information content of mini Nikkei 225 index futures traded at the Osaka Securities Exchange in 2012 and 2013, a data-driven optimal length of local windows up to approximately 1-2 hours is reasonable to capture parameter variations and is suitable for short-term prediction. Our proposed trading strategies achieve statistical arbitrage opportunities and are therefore beneficial for quantitative finance practice.
    Keywords: multiplicative error models, trading volume, order flow, forecasting
    JEL: C41 C51 C53 G12 G17
    Date: 2014–07
  5. By: Korobilis, Dimitris
    Abstract: This paper proposes full-Bayes priors for time-varying parameter vector autoregressions (TVP-VARs) which are more robust and objective than existing choices proposed in the literature. We formulate the priors in a way that they allow for straightforward posterior computation, they require minimal input by the user, and they result in shrinkage posterior representations, thus, making them appropriate for models of large dimensions. A comprehensive forecasting exercise involving TVP-VARs of different dimensions establishes the usefulness of the proposed approach.
    Keywords: TVP-VAR, shrinkage, data-based prior, forecasting,
    Date: 2014
  6. By: Naoko Hara (Bank of Japan); Shotaro Yamane (Bank of Japan)
    Abstract: This paper proposes a new approach for nowcasting as yet unavailable GDP growth by estimating monthly GDP growth with a large dataset. The model consists of two parts: (i) a few indicators that explain a large part of the variation in GDP growth, and (ii) principal components, which are orthogonal to those indicators and are extracted from a number of GDP source data, capturing the rest of the variation. The approach relies on a static factor model comprising a number of indicators that have a simultaneous relationship with GDP. Applying this approach to data for Japan, we find that our model produces more precise estimates of recent GDP growth at an earlier stage of nowcasting than the nowcasts of professional forecasters.
    Keywords: Factor Models; Forecasting; Nowcasting; Monthly GDP; Real-time Data
    JEL: C53 C82 E37
    Date: 2013–10–15

This nep-for issue is ©2014 by Rob J Hyndman. It is provided as is without any express or implied warranty. It may be freely redistributed in whole or in part for any purpose. If distributed in part, please include this notice.
General information on the NEP project can be found at For comments please write to the director of NEP, Marco Novarese at <>. Put “NEP” in the subject, otherwise your mail may be rejected.
NEP’s infrastructure is sponsored by the School of Economics and Finance of Massey University in New Zealand.