nep-for New Economics Papers
on Forecasting
Issue of 2019‒02‒25
seven papers chosen by
Rob J Hyndman
Monash University

  1. Cross-validation based forecasting method: a machine learning approach By Pinto, Jeronymo Marcondes; Marçal, Emerson Fernandes
  2. The effects of a mixed approach toward management earnings forecasts: evidence from China By Huang, Xiaobei; Li, Xi; Tse, Senyo; Tucker, Jennifer Wu
  3. On the external validity of experimental inflation forecasts : a comparison with five categories of fields expectations By Camille Cornand; Paul Hubert
  4. DCC-HEAVY: a multivariate GARCH model with realized measures of variance and correlation By Xu, Yongdeng
  5. New testing approaches for mean-variance predictability By Fiorentini, Gabriele; Sentana, Enrique
  6. Deep Adaptive Input Normalization for Price Forecasting using Limit Order Book Data By Nikolaos Passalis; Anastasios Tefas; Juho Kanniainen; Moncef Gabbouj; Alexandros Iosifidis
  7. Non-Stationary Dividend-Price Ratios By Vassilis Polimenis; Ioannis Neokosmidis

  1. By: Pinto, Jeronymo Marcondes; Marçal, Emerson Fernandes
    Abstract: Our paper aims to evaluate two novel methods on selecting the best forecasting model or its combination based on a Machine Learning approach. The methods are based on the selection of the ”best” model, or combination of models, by crossvalidation technique, from a set of possible models. The first one is based on the seminal paper of Granger-Bates (1969) but weights are estimated by a process of cross-validation applied on the training set. The second one selects the model with the best forecasting performance in the process described above, which we called CvML (Cross-Validation Machine Learning Method). The following models are used: exponential smoothing, SARIMA, artificial neural networks and Threshold autoregression (TAR). Model specification is chosen by R packages: forecast and TSA. Both methods – CvML and MGB - are applied to these models to generate forecasts from one up to twelve periods ahead. Frequency of data is monthly. We run the forecasts exercise to the following to monthly series of Industrial Product Indices for seven countries: Canada, Brazil, Belgium, Germany, Portugal, UK and USA. The data was collected at OECD data, with 504 observations. We choose Average Forecast Combination, Granger Bates Method, MCS model, Naive and Seasonal Naive Model as benchmarks.Our results suggest that MGB did not performed well. However, CvML had a lower mean absolute error for most of countries and forecast horizons, particularly at longer horizons, surpassing all the proposed benchmarks. Similar results hold for absolute mean forecast error.
    Date: 2019–02
  2. By: Huang, Xiaobei; Li, Xi; Tse, Senyo; Tucker, Jennifer Wu
    Abstract: Chinese regulators mandate management earnings forecasts when managers’ earnings expectations meet bright-line thresholds and allow voluntary forecasts in other circumstances. We examine the effects of this mixed approach. We find that Chinese mandatory forecasts have significant information content. Moreover, we observe a learning effect: mandatory forecasts appear to stimulate voluntary forecasts in subsequent periods as managers become familiar with the forecasting and disclosing procedures through forced experience. We find one negative consequence of the mixed approach, however: managers appear to manipulate earnings to avoid the forecast threshold of large earnings decreases. Overall, we document the pros and cons of a mixed approach toward management earnings forecasts in a major emerging market.
    Keywords: China; forecast mandate; management earnings forecast; voluntary disclosure
    JEL: F3 G3
    Date: 2018–03–01
  3. By: Camille Cornand (Centre National de la Recherche Scientifique (CNRS)); Paul Hubert (Observatoire français des conjonctures économiques)
    Abstract: Establishing the external validity of laboratory experiments in terms of inflation forecasts is crucial for policy initiatives to be valid outside the laboratory. Our contribution is to document whether different measures of inflation expectations based on various categories of agents (participants to experiments, households, industry forecasters, professional forecasters, financial market participants and central bankers) share common patterns by analyzing: the forecasting performances of these different categories of data; the information rigidities to which they are subject; the determination of expectations. Overall, the different categories of forecasts exhibit common features: forecast errors are comparably large and autocorrelated, forecast errors and forecast revisions are predictable from past information, which suggests the presence of information frictions. Finally, the standard lagged inflation determinant of inflation expectations is robust to the data sets. There is nevertheless some heterogeneity among the six different sets. If experimental forecasts are relatively comparable to survey and financial market data, central bank forecasts seem to be superior.
    Keywords: Inflation expectations; Experimental forecasts; Survey forecasts; Market -based forecasts; Central bank forecasts
    JEL: E3 E5
    Date: 2019–02
  4. By: Xu, Yongdeng (Cardiff Business School)
    Abstract: This paper proposes a new class of multivariate volatility model that utilising high-frequency data. We call this model the DCC-HEAVY model as key ingredients are the Engle (2002) DCC model and Shephard and Sheppard (2012) HEAVY model. We discuss the models' dynamics and highlight their differences from DCC-GARCH models. Specifically, the dynamics of conditional variances are driven by the lagged realized variances, while the dynamics of conditional correlations are driven by the lagged realized correlations in the DCC-HEAVY model. The new model removes well known asymptotic bias in DCC-GARCH model estimation and has more desirable asymptotic properties. We also derive a Quasi-maximum likelihood estimation and provide closed-form formulas for multi-step forecasts. Empirical results suggest that the DCC-HEAVY model outperforms the DCC-GARCH model in and out-of-sample.
    Keywords: HEAVY model, Multivariate volatility, High-frequency data, Forecasting, Wishart distribution
    JEL: C32 C58 G17
    Date: 2019–02
  5. By: Fiorentini, Gabriele; Sentana, Enrique
    Abstract: We propose tests for smooth but persistent serial correlation in risk premia and volatilities that exploit the non-normality of financial returns. Our parametric tests are robust to distributional misspecification, while our semiparametric tests are as powerful as if we knew the true return distribution. Local power analyses confirm their gains over existing methods, while Monte Carlo exercises assess their finite sample reliability. We apply our tests to quarterly returns on the five Fama-French factors for international stocks, whose distributions are mostly symmetric and fat-tailed. Our results highlight noticeable differences across regions and factors and confirm the fragility of Gaussian tests.
    Keywords: Financial forecasting; Misspecification; Moment tests; robustness; volatility
    JEL: C12 C22 G17
    Date: 2019–01
  6. By: Nikolaos Passalis; Anastasios Tefas; Juho Kanniainen; Moncef Gabbouj; Alexandros Iosifidis
    Abstract: Deep Learning (DL) models can be used to tackle time series analysis tasks with great success. However, the performance of DL models can degenerate rapidly if the data are not appropriately normalized. This issue is even more apparent when DL is used for financial time series forecasting tasks, where the non-stationary and multimodal nature of the data pose significant challenges and severely affect the performance of DL models. In this work, a simple, yet effective, neural layer, that is capable of adaptively normalizing the input time series, while taking into account the distribution of the data, is proposed. The proposed layer is trained in an end-to-end fashion using back-propagation and can lead to significant performance improvements. The effectiveness of the proposed method is demonstrated using a large-scale limit order book dataset.
    Date: 2019–02
  7. By: Vassilis Polimenis; Ioannis Neokosmidis
    Abstract: Dividend yields have been widely used in previous research to relate stock market valuations to cash flow fundamentals. However, this approach relies on the assumption that dividend yields are stationary. Due to the failure to reject the hypothesis of a unit root in the classical dividend-price ratio for the US stock market, Polimenis and Neokosmidis (2016) proposed the use of a modified dividend price ratio (mdp) as the deviation between d and p from their long run equilibrium, and showed that mdp provides substantially improved forecasting results over the classical dp ratio. Here, we extend that paper by performing multivariate regressions based on the Campbell-Shiller approximation, by utilizing a dynamic econometric procedure to estimate the modified dp, and by testing the modified ratios against reinvested dividend-yields. By comparing the performance of mdp and dp in the period after 1965, we are not only able to enhance the robustness of the findings, but also to debunk a possible false explanation that the enhanced mdp performance in predicting future returns comes from a capacity to predict the risk-free return component. Depending on whether one uses the recursive or population methodology to measure the performance of mdp, the Out-of-Sample performance gain is between 30% to 50%.
    Date: 2019–02

This nep-for issue is ©2019 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.