nep-for New Economics Papers
on Forecasting
Issue of 2015‒02‒22
fifteen papers chosen by
Rob J Hyndman
Monash University

  1. Forecasting in a DSGE Model with Banking Intermediation: Evidence from the US By Roberta Cardani; Alessia Paccagnini; Stefania Villa
  2. Forecasting German key macroeconomic variables using large dataset methods By Pirschel, Inske; Wolters, Maik
  3. Evaluating Non-Linear Approaches in Forecasting Tourist Arrivals By Andrea Saayman and Ilse Botha
  4. Model Pooling and Changes in the Informational Content of Predictors: an Empirical Investigation for the Euro Area By Tim Schwarzmüller
  5. Long term probabilistic load forecasting and normalization with hourly information By Tao Hong; Jason Wilson; Jingrui Xie
  6. Fluctuations of the Real Exchange Rate, Real Interest Rates, and the Dynamics of the Price of Gold in a Small Open Economy By Rohloff, Sebastian; Pierdzioch, Christian; Risse, Marian
  7. Bivariate GARCH models for single asset returns By Tomasz Skoczylas
  8. Comparing Linear and Non-linear Benchmarks of Exchange Rate Forecasting By SJ Retief, M Pretorius and I Botha
  9. Global Energy Forecasting Competition 2012 By Tao Hong; Pierre Pinson; Shu Fan
  10. Finding SPF Percentiles Closest to Greenbook By Tae-Hwy Lee; Yiyao Wang
  11. Modeling and predicting the market volatility index: The case of VKOSPI By Han, Heejoon; Kutan, Ali M.; Ryu, Doojin
  12. Stock-Flow Dynamic Projection By LI, XI HAO; Gallegati, Mauro
  13. Exchange Rate Determination and Out of Sample Forecasting: Cointegration Analysis By Hina, Hafsa; Qayyum, Abdul
  14. Why prediction markets work : the role of information acquisition and endogenous weighting By Siemroth, Christoph
  15. Government Forecasts of Budget Balances Under Asymmetric By Rülke, Jan-Christoph; Pierdzioch, Christian

  1. By: Roberta Cardani; Alessia Paccagnini; Stefania Villa
    Abstract: This paper examines the forecasting performance of DSGE models with and without banking intermediation for the US economy. Over the forecast period 2001-2013, the model augmented with a banking sector leads to an improvement of point and density forecasts for inflation and the short term interest rate, while the better forecast for output depends on the forecasting horizon/period. To interpret this finding it is crucial to take into account parameters instabilities showed by a recursive-window estimation. Moreover, rolling estimates of point forecasts show that a banking sector helps improving the forecasting performance of output and inflation in the recent period.
    Keywords: Bayesian estimation, Forecasting, Banking sector
    JEL: C11 C13 C32 E37
    Date: 2015–02
  2. By: Pirschel, Inske; Wolters, Maik
    Abstract: We study the forecasting performance of three alternative large scale approaches for German key macroeconomic variables using a dataset that consists of 123 variables in quarterly frequency. These three approaches handle the dimensionality problem evoked by such a large dataset by aggregating information, yet on different levels. We consider different factor models, a large Bayesian VAR and model averaging techniques, where aggregation takes place before, during and after the estimation of the different models, respectively. We find that overall the large Bayesian VAR provides the most precise forecasts compared to the other large scale approaches and a number of small benchmark models. For some variables the large Bayesian VAR is also the only model producing unbiased forecasts at least for short horizons. While a Bayesian factor augmented VAR with a tight prior also provides quite accurate forecasts overall, the performance of the other methods depends on the variable to be forecast.
    JEL: C53 E37 E47
    Date: 2014
  3. By: Andrea Saayman and Ilse Botha
    Abstract: Quantitative methods to forecasting tourist arrivals can be sub-divided into causal methods and non-causal methods. Non-causal time series methods remain popular tourism forecasting tools due to the accuracy of their forecasting ability and general ease of use. Since tourist arrivals exhibit seasonality, SARIMA models are often found to be the most accurate. However, these models assume that the time-series is linear. This paper compares the baseline seasonal Naïve and SARIMA forecasts of a seasonal tourist destination faced with a structural break in the data, with alternative non-linear methods, with the aim to determine the accuracy of the various methods. These methods include the unobserved components model, smooth transition autoregressive model (STAR) and singular spectrum analysis (SSA). The results show that the non-linear forecasts outperform the other methods. The linear methods show some superiority in short-term forecasts when there are no structural changes in the time series.
    Keywords: forecasting, tourism demand, SARIMA, STAR, Spectrum analysis, basic structural model (BSM)
    Date: 2015
  4. By: Tim Schwarzmüller
    Abstract: I study the performance of single predictor bridge equation models as well as a wide range of model selection and pooling techniques, including Mallows model averaging and Cross-Validation model averaging, for short-term forecasting euro area GDP growth. I explore to what extend model selection and model pooling techniques are able to outperform a simple autoregressive benchmark model in the periods before, during and after the Great Recession. I find that single predictor bridge equation models suffer a great variation in the forecast performance relative to the benchmark model over the analysed sub-samples. Moreover, model selection techniques turn out to produce quite poor forecasts in some sub-samples. On the contrary, model pooling based on the Cross-Validation and the Mallows criterion provide a very stable and accurate forecast performance
    Keywords: Short-term forecasting, Great Recession, mixed frequency data, model selection and model pooling
    JEL: C53 E37
    Date: 2015–01
  5. By: Tao Hong; Jason Wilson; Jingrui Xie
    Abstract: The classical approach to long term load forecasting is often limited to the use of load and weather information occurring with monthly or annual frequency. This low resolution, infrequent data can sometimes lead to inaccurate forecasts. Load forecasters often have a hard time explaining the errors based on the limited information available through the low resolution data. The increasing usage of Smart Grid and Advanced Metering Infrastructure (AMI) technologies provides the utility load forecasters with high resolution, layered information to improve the load forecasting process. In this paper, we propose a modern approach that takes advantage of hourly information to create more accurate and defensible forecasts. The proposed approach has been deployed across many US utilities, including a recent implementation at North Carolina Electric Membership Corporation (NCEMC), which is used as the case study in this paper. Three key elements of long term load forecasting are being modernized: predictive modeling, scenario analysis and weather normalization. We first show the superior accuracy of the predictive models attained from hourly data, over the classical methods of forecasting using monthly or annual peak data. We then develop probabilistic forecasts through cross scenario analysis. Finally, we illustrate the concept of load normalization and normalize the load using the proposed hourly models.
    Keywords: Load forecasting; Load normalization; Weather normalization; Multiple linear regression model
    JEL: C22 C53 Q41 Q47
    Date: 2013–12–31
  6. By: Rohloff, Sebastian; Pierdzioch, Christian; Risse, Marian
    Abstract: Economic theory predicts that, in a small open economy, the dynamics of the real price of gold should be linked to real interest rates and the rate of change of the real exchange rate. Using data for Australia, we use a real-time forecasting approach to analyze whether real interest rates and the rate of change of the real exchange rate help to forecast out-ofsample the rate of change of the real price of gold. We study the economic value-added of out-of-sample forecasts using a behavioral-finance approach that takes into account that a forecaster may have an asymmetric loss function.
    JEL: C53 E44 G12
    Date: 2014
  7. By: Tomasz Skoczylas (Faculty of Economic Sciences, University of Warsaw)
    Abstract: In this paper an alternative approach to modelling and forecasting single asset returns volatility is presented. A new, bivariate, flexible framework, which may be considered as a development of single-equation ARCH-type models, is proposed. This approach focuses on joint distribution of returns and observed volatility, measured by Garman-Klass variance estimator, and it enables to examine simultaneous dependencies between them. Proposed models are compared with benchmark GARCH and range-based GARCH (RGARCH) models in terms of prediction accuracy. All models are estimated with maximum likelihood method, using time series of EUR/PLN spot rate quotations and WIG20 index. Results are very encouraging especially for foreasting Value-at-Risk. Bivariate models achieved lesser rates of VaR exception, as well as lower coverage tests statistics, without being more conservative than its single-equation counterparts, as their forecasts errors measures are rather similar.
    Keywords: bivariate volatility models, joint distribution, range-based volatility estimators, Garman-Klass estimator, observed volatility, volatility modelling, GARCH, leverage, Value-at-Risk, volatility forecasting
    JEL: C13 C32 C53 C58 G10 G17
    Date: 2015
  8. By: SJ Retief, M Pretorius and I Botha
    Abstract: Throughout the past 3 decades, the random walk model served as exchange rate forecasting benchmark to verify that a model is able to outperform a random process. However, its application as forecasting benchmark is contradictory. Rather than serving as a benchmark that explains exchange rate behaviour, it serves as a benchmark of what we do not understand in exchange rate forecasting – the random component. In order to accommodate for the observed mean reverting and non-linear patterns in exchange rate information, this study considers various univariate models to serve as linear or non-linear benchmarks of exchange rate forecasting. The results of forecasting performance indicate that the random walk model is an insufficient benchmark to explain exchange rate movements for non-static models. As linear alternative, an autoregressive model performed best to explain the mean reverting patterns in exchange rate information for quarterly, monthly and weekly forecasts of the exchange rate. As non-linear alternative, a Kernel regression was best able to explain volatile exchange rate movements associated with daily forecasts of the exchange rate.
    Keywords: random walk, autoregressive (AR), moving average (MA), Kernel regression
    Date: 2015
  9. By: Tao Hong; Pierre Pinson; Shu Fan
    Abstract: The Global Energy Forecasting Competition (GEFCom2012) attracted hundreds of participants worldwide, who contributed many novel ideas to the energy forecasting field. This paper introduces both tracks of GEFCom2012, hierarchical load forecasting and wind power forecasting, with details on the aspects of the problem, the data, and a summary of the methods used by selected top entries. We also discuss the lessons learned from this competition from the organizers’ perspective. The complete data set, including the solution data, is published along with this paper, in an effort to establish a benchmark data pool for the community.
    Keywords: Energy forecasting; Load forecasting; Forecasting competition; Wind power forecasting
    JEL: Q41 Q47
    Date: 2013–12–31
  10. By: Tae-Hwy Lee (Department of Economics, University of California Riverside); Yiyao Wang (Booth School of Business, University of Chicago)
    Abstract: To find forecasts that are closest to Greenbook forecast from the Survey of Professional Forecasters, this paper looks for SPF cross-sectional percentile forecasts that are not encompassed by Greenbook forecast under Greenbook's loss preference, which exhibits time-varying asymmetry. To evaluate SPF percentile forecasts under Greenbook's loss function, we introduce the forecast encompassing test for the asymmetric least square regression of conditional expectiles. From the analysis of the U.S. quarterly real output and inflation forecasts over the past four decades, we find that almost all SPF percentiles are encompassed by Greenbook forecast in full data period. However there is evidence in sub-periods that many SPF percentiles are not encompassed by Greenbook. Among those not-encompassed SPF percentiles, the best SPF percentile closest to Greenbook for real output growth forecast is near the median, while the best SPF percentile for inflation forecast is far below the median in the left tail of the SPF cross-sectional distribution.
    Keywords: Greenbook, Survey of Professional Forecasters, estimation of ‡exible loss function, SPF cross-sectional distribution, SPF percentiles, encompassing test, asymmetric least squares.
    JEL: C53 E37 E27
    Date: 2015–02
  11. By: Han, Heejoon; Kutan, Ali M.; Ryu, Doojin
    Abstract: The KOSPI 200 options are one of the most actively traded derivatives in the world. This paper empirically examines (a) the statistical properties of the Korea's representative implied volatility index (VKOSPI) derived from the KOSPI 200 options and (b) macroeconomic and financial variables that can predict the implied volatility process of the index, using augmented heterogeneous autoregressive (HAR) models with exogenous covariates. The results suggest that the dynamics of the VKOSPI is well described by the elaborate HAR framework and that some Korea's macroeconomic variables significantly explain the VKOSPI. In addition, we find that the stock market return and implied volatility index of the US market (i.e., the S&P 500 spot return and the VIX from S&P 500 options) play a key role in predicting the level of VKOSPI and explaining its dynamics, and their explanatory power dominates that of Korea's macro-finance variables. Further, while Korea's stock market return does not predict the VKOSPI, US stock market return well predicts the future VKOSPI level. When both US stock market return and US implied volatility index are incorporated into the HAR framework, the model's both in-sample fitting and out-of-sample forecasting ability exhibits the best performance.
    Keywords: heterogeneous autoregressive (HAR) model,implied volatility index,VKOSPI,VIX,KOSPI 200 options
    JEL: C22 C50 G14 G15
    Date: 2015
  12. By: LI, XI HAO; Gallegati, Mauro
    Abstract: Borrowing from our experience in agent-based computational economic research from `bottom-up', this paper considers economic system as multi-level dynamical system that micro-level agents' interaction leads to structural transition in meso-level, which results in macro-level market dynamics with endogenous fluctuation or even market crashes. By the concept of transition matrix, we develop technique to quantify meso-level structural change induced by micro-level interaction. Then we apply this quantification to propose the method of dynamic projection that delivers out-of-sample forecast of macro-level economic variable from micro-level big data. We testify this method with a data set of financial statements for 4599 firms listed in Tokyo Stock Exchange for the year of 1980 to 2012. The Diebold-Mariano test indicates that the dynamic projection has significantly higher accuracy for one-period-ahead out-of-sample forecast than the benchmark of ARIMA models.
    Keywords: economic forecasting, dynamic projection, multi-level dynamical system, transition matrix
    JEL: C53 C63 E27
    Date: 2015–01–01
  13. By: Hina, Hafsa; Qayyum, Abdul
    Abstract: Forecasting the nominal exchange rate has been one of the most difficult exercises in economics. This study employs the Frankel (1979) monetary model of exchange rate to examine the long run behavior of Pakistan rupee per unit of US dollar over the period 1982:Q1 to 2012:Q2. Johansen and Juselious (1988,1992) likelihood ratio test indicates one long-run cointegrating vector among the fundamentals. Cointegrating vector is uniquely identified as Dornbusch (1976) monetary model by imposing plausible economic restrictions. Finally, the short-run dynamic error correction model is estimated on the bases of identified cointegrated vector. Out of sample forecasting analysis of parsimonious short run dynamic error correction model is able to beat the naïve random walk model on the basis of root mean square error, Theil’s U coefficient and Diebold and Mariano (1995) test statistics.
    Keywords: Exchange rate determination; Unit root; Cointegration; Error correction model; Forecasting; Random walk model
    JEL: C32 C4 C58 F3 F31
    Date: 2015
  14. By: Siemroth, Christoph
    Abstract: In prediction markets, investors trade assets whose values are contingent on the occurrence of future events, like election outcomes. Prediction market prices have been shown to be consistently accurate forecasts of these outcomes, but we don't know why. I formally illustrate an information acquisition explanation. Traders with more wealth to invest have stronger incentives to acquire information about the outcome, thus tend to have better forecasts. Moreover, their trades have larger weight in the market. The interaction implies that a few well-situated traders can move the asset price toward the true value. One implication for institutions aggregating information is to put more weight on votes of agents with larger stakes, which improves on equal weighting, unless prior distribution accuracy and stakes are negatively related.
    Keywords: Information Acquisition , Information Aggregation , Forecasting , Futures Markets , Prediction Markets
    JEL: D83 D84 G13
    Date: 2014
  15. By: Rülke, Jan-Christoph; Pierdzioch, Christian
    Abstract: We study the loss function of 15 European governments as implied by their budget balance forecasts. Results suggest that the shape of the loss function varies across countries. The loss function becomes more asymmetric as the forecast horizon increases and in advance of parliamentary election. Compared to that, government ideology does not affect the shape of the loss function. Under a fiscal rule, government agencies experience a higher loss when overpredicting the fiscal balance compared to an underprediction of the same size. We also document that under an asymmetric loss function government forecasts look more rational compared to a symmetric loss function. This may explain why government agencies' forecasts have been found to be too optimistic (Frankel 2012).
    JEL: E62 H50 E27
    Date: 2014

This nep-for issue is ©2015 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.