nep-for New Economics Papers
on Forecasting
Issue of 2018‒04‒23
seven papers chosen by
Rob J Hyndman
Monash University

  1. Forecasting external demand using BVAR models By Lajos Tamás Szabó
  2. Does a financial accelerator improve forecasts during financial crises?: Evidence from Japan with Prediction Pool Methods By Hasumi, Ryo; Iiboshi, Hirokuni; Matsumae, Tatsuyoshi; Nakamura, Daisuke
  3. How News and Its Context Drive Risk and Returns Around the World By Charles W. Calomiris; Harry Mamaysky
  4. International transmissions of aggregate macroeconomic uncertainty in small open economies: An empirical approach By Jamie L. Cross; Chenghan Hou; Aubrey Poon
  5. Predicting International Equity Returns: Evidence from Time-Varying Parameter Vector Autoregressive Models By Rangan Gupta; Florian Huber; Philipp Piribauer
  6. Choice of Benchmark When Forecasting Long-term Stock Returns By Ioannis Kyriakou; Parastoo Mousavi; Jens Perch Nielsen; Michael Scholz
  7. “A regional perspective on the accuracy of machine learning forecasts of tourism demand based on data characteristics” By Oscar Claveria; Enric Monte; Salvador Torra

  1. By: Lajos Tamás Szabó (Magyar Nemzeti Bank (Central Bank of Hungary))
    Abstract: As the impact of monetary policy decisions manifests itself with a lag, decision-makers also need economic forecasts when they make decisions. In this paper, we present a method that may facilitate the integration of incoming data in the external demand forecast faster than is currently possible. The external demand forecast helps to forecast exports and, through that, developments in GDP. In the current practice, for the imports of Hungary’s key trading partners we use the forecasts of international institutions as a starting point. Data received in the meantime can be included in the forecast using expert judgements. With the method described in this paper, we forecast the imports of Hungary’s key trading partners – and with the help thereof – their external demand, relying on BVAR models and using monthly time series (confidence indices, industrial production, orders). Based on the literature, we use the Kalman filter to eliminate the differences in the publication lags of the individual time series. The missing variable is then forecast using the other variables. The forecasts thus obtained perform better than the best ARMA models, and the model containing global imports and the oil price. With one exception, the forecast of the imports of the individual countries is more accurate when prepared on the whole sample, rather than on the rolling sample. The forecast of external demand is also more accurate if we use the whole sample. The most accurate BVAR model used to forecast external demand provides an unbiased forecast and also yields a better forecast of turning points than the models used for comparison. Compared to the forecasts of international institutions, the BVAR forecast performs better when actual import data from the respective year are already available. Thus, compared to previous practice, the novelty is represented by the BVAR methodology and the monthly time series, which can be integrated into the forecast in a formalised manner. Looking ahead, it may also be worthwhile to forecast GDP components using this method.
    Keywords: BVAR, forecast of external demand.
    JEL: C11 F17 F47
    Date: 2018
  2. By: Hasumi, Ryo; Iiboshi, Hirokuni; Matsumae, Tatsuyoshi; Nakamura, Daisuke
    Abstract: Using a Markov-switching prediction pool method (Waggoner and Zha, 2012) in terms of density forecasts, we assess the time-varying forecasting performance of a DSGE model incorporating a financial accelerator a la Bernanke et al. (1999) with the frictionless model by focusing on periods of financial crisis including the so-called "Bubble period" and the "Lost decade" in Japan. According to our empirical results, the accelerator improves the forecasting of investment over the whole sample period, while forecasts of consumption and inflation depend on the fluctuation of an extra financial premium between the policy interest rate and corporate loan rates. In particular, several drastic monetary policy changes might disrupt the forecasting performance of the model with the accelerator. A robust check with a dynamic pool method (Del Negro et al., 2016) also supports these results.
    Keywords: Density forecast, Optimal prediction pool, Markov-switching prediction pool, Dynamic prediction pool, Bayesian estimation, Markov Chain Monte Carlo, Financial Friction.
    JEL: C3 C32 C53 E3 E32 E37
    Date: 2018–03
  3. By: Charles W. Calomiris; Harry Mamaysky
    Abstract: We develop a classification methodology for the context and content of news articles to predict risk and return in stock markets in 51 developed and emerging economies. A parsimonious summary of news, including topic-specific sentiment, frequency, and unusualness (entropy) of word flow, predicts future country-level returns, volatilities, and drawdowns. Economic and statistical significance are high and larger for year-ahead than monthly predictions. The effect of news measures on market outcomes differs by country type and over time. News stories about emerging markets contain more incremental information. Out-of-sample testing confirms the economic value of our approach for forecasting country-level market outcomes.
    JEL: G12 G14 G15
    Date: 2018–03
  4. By: Jamie L. Cross; Chenghan Hou; Aubrey Poon
    Abstract: We estimate the effects of domestic and international sources of macroeconomic uncertainty in three commonly studied small open economies (SOEs): Australia, Canada and New Zealand. To this end, we propose a common stochastic volatility in mean panel VAR (CSVM-PVAR), and develop an efficient Markov chain Monte Carlo algorithm to estimate the model. Using a formal Bayesian model comparison exercise, our in-sample results suggest that foreign uncertainty spillovers shape the macroeconomic conditions in all SOEs, however domestic uncertainty shocks are important for Australia and Canada, but not New Zealand. The general mechanism is that foreign uncertainty shocks reduce real GDP and raise inflation in all SOEs, however the interest rate responses are idiosyncratic; being positive in Australia and New Zealand, and negative in Canada. Conversely, domestic uncertainty shocks tend to raise all three macroeconomic variables. Finally, in a pseudo out-of-sample forecasting exercise, the proposed model also forecasts better than traditional PVAR and CSV-PVAR benchmarks.
    Keywords: Bayesian VARs, International Spillovers, State-Space Models, Stochastic Volatility in Mean, Uncertainty
    JEL: C11 C32 C53 E37
    Date: 2018–04
  5. By: Rangan Gupta (Department of Economics, University of Pretoria, Pretoria, South Africa); Florian Huber (Department of Economics, WU Vienna University of Economics and Business); Philipp Piribauer (Austrian Institute of Economic Research (WIFO))
    Abstract: In this paper, we forecast monthly stock returns of eight advanced economies using a time varying parameter vector autoregressive model (TVP-VAR). Compared to standard TVP-VARs, our proposed model automatically detects whether time-variation in the parameters is needed through the introduction of a latent threshold process that is driven by the absolute size of parameter changes. The advantage of this framework is that it can dynamically detect whether a given regression coefficient is constant or time-varying during distinct time periods. We moreover compare the performance of this model with a wide range of nested alternative time-varying and constant parameter VAR models. Our results indicate that the threshold TVP-VAR outperforms its competitors in terms of point and density forecasts. A portfolio allocation exercise confirms the superiority of our proposed framework. In addition, a copula-based analysis also indicates that it pays off to adopt a multivariate modeling framework, especially during periods of stress, like the recent financial crisis.
    Keywords: International equity markets, Time-varying vector autoregression, Point and density forecasts, Portfolio allocation
    JEL: C32 G10 G17
    Date: 2018–04
  6. By: Ioannis Kyriakou (Cass Business School, City, University of London, UK); Parastoo Mousavi (Cass Business School, City, University of London, UK); Jens Perch Nielsen (Cass Business School, City, University of London, UK); Michael Scholz (University of Graz, Austria)
    Abstract: Recent advances in pension product development seem to favour alternatives to the risk free asset so often used in financial theory. In this paper, we investigate other benchmarks of the financial model than the classical risk free asset; for example, a benchmark following an inflation index is just one important case. Modelling extra returns above the inflation of risky assets is important, for example, for actuarial applications aiming at providing real income forecasts for pensioners. We study market timing when three alternative benchmarks are considered: the risk free rate, inflation and the long bond yield. We conclude that forecasting future stock returns is of similar complexity for all three considered benchmarks. From a pension fund modelling perspective, it therefore seems that one should use the most convenient benchmark for the problem at hand.
    Keywords: Benchmark; Cross validation; Prediction; Stock returns
    JEL: C14 C53 C58 G22
    Date: 2018–04
  7. By: Oscar Claveria (AQR-IREA AQR-IREA, University of Barcelona (UB). Tel. +34-934021825; Fax. +34-934021821. Department of Econometrics, Statistics and Applied Economics, University of Barcelona, Diagonal 690, 08034 Barcelona, Spain); Enric Monte (Department of Signal Theory and Communications, Polytechnic University of Catalunya (UPC)); Salvador Torra (Riskcenter-IREA, Department of Econometrics and Statistics, University of Barcelona (UB))
    Abstract: In this work we assess the role of data characteristics in the accuracy of machine learning (ML) tourism forecasts from a spatial perspective. First, we apply a seasonal-trend decomposition procedure based on non-parametric regression to isolate the different components of the time series of international tourism demand to all Spanish regions. This approach allows us to compute a set of measures to describe the features of the data. Second, we analyse the performance of several ML models in a recursive multiple-step-ahead forecasting experiment. In a third step, we rank all seventeen regions according to their characteristics and the obtained forecasting performance, and use the rankings as the input for a multivariate analysis to evaluate the interactions between time series features and the accuracy of the predictions. By means of dimensionality reduction techniques we summarise all the information into two components and project all Spanish regions into perceptual maps. We find that entropy and dispersion show a negative relation with accuracy, while the effect of other data characteristics on forecast accuracy is heavily dependent on the forecast horizon.
    Keywords: STL decomposition, non-parametric regression, time series features, forecast accuracy, machine learning, tourism demand, regional analysis. JEL classification:C45, C51, C53, C63, E27, L83.
    Date: 2018–03

This nep-for issue is ©2018 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.