|
on Forecasting |
| By: | Jing Zeng (Department of Economics, University of Konstanz, Germany) |
| Abstract: | European Monetary Union (EMU) member countries' forecasts are often combined to obtain the forecasts of the Euro area macroeconomic aggregate variables. The aggregation weights which are used to produce the aggregates are often considered as combination weights. This paper investigates whether using different combination weights instead of the usual aggregation weights can help to provide more accurate forecasts. In this context, we examine the performance of equal weights, the least squares estimators of the weights, the combination method recently proposed by Hyndman et al. (2011) and the weights suggested by shrinkage methods. We find that some variables like real GDP and GDP deflator can be forecasted more precisely by using flexible combination weights. Furthermore, combining only forecasts of the three largest European countries helps to improve the forecasting performance. The persistence of the individual data seems to play an important role for the relative performance of the combination. |
| Keywords: | Forecast combination, aggregation, macroeconomic forecasting, hierarchical time series, persistence in data |
| JEL: | C22 C43 C53 |
| Date: | 2015–05–13 |
| URL: | http://d.repec.org/n?u=RePEc:knz:dpteco:1511&r=for |
| By: | Mammadov, Fuad; Shaig Adigozalov, Shaiq |
| Abstract: | This paper has attempted to construct leading indicator systems and based on that to predict future contraction period of the Azerbaijan non-oil economy using more than 100 publicly available economic and financial data. Our results show plausible and significant performance of composite leading indicator system with average leading time of 7.2 months. We found that between January of 2000 and May of 2014, there were 6 turning points in Azerbaijan non-oil economy, consisting of three peaks and three troughs corresponding three expansion and four contraction periods. It turns out that the average duration of expansion and contraction phases is 43 and 10 month, respectively. Based on selected leading indicators we constructed composite indicator is found to be able to predict all the six turning points. Using dynamic probit model we estimated contraction probability of non-oil output gap for the future period. Out-of-sample as well as in-sample forecast performance suggest that the leading indicator systems have significant predictive power and could be used as a useful tool for economic forecasting. |
| Keywords: | Business cycles, Dating, Turning points, Forecasting, Probit Model |
| JEL: | C25 C53 E32 |
| Date: | 2014–10–10 |
| URL: | http://d.repec.org/n?u=RePEc:pra:mprapa:64367&r=for |
| By: | Korobilis, Dimitris |
| Abstract: | Bayesian model averaging (BMA) methods are regularly used to deal with model uncertainty in regression models. This paper shows how to introduce Bayesian model averaging methods in quantile regressions, and allow for different predictors to affect different quantiles of the dependent variable. I show that quantile regression BMA methods can help reduce uncertainty regarding outcomes of future inflation by providing superior predictive densities compared to mean regression models with and without BMA. |
| Keywords: | Bayesian model averaging; quantile regression; inflation forecasts; fan charts |
| JEL: | C11 C22 C52 |
| Date: | 2015–04 |
| URL: | http://d.repec.org/n?u=RePEc:pra:mprapa:64341&r=for |
| By: | Sandra Stankiewicz (Department of Economics, University of Konstanz, Germany) |
| Abstract: | I use the adaptive elastic net in a Bayesian framework and test its forecasting performance against lasso, adaptive lasso and elastic net (all used in a Bayesian framework) in a series of simulations, as well as in an empirical exercise for macroeconomic Euro area data. The results suggest that elastic net is the best model among the four Bayesian methods considered. Adaptive lasso, on the other hand, shows the worst forecasting performance. Lasso is generally better then adaptive lasso, but worse than adaptive elastic net. The differences in the performance of these models become especially large when the number of regressors grows considerably relative to the number of available observations. The results point to the fact that the ridge regression component in the elastic net is responsible for its improvement in forecasting performance over lasso. The adaptive shrinkage in some of the models does not seem to play a major role, and may even lead to a deterioration of the performance. |
| Keywords: | Elastic net, Lasso, Bayesian, Forecasting |
| JEL: | C11 C22 C53 |
| Date: | 2015–05–13 |
| URL: | http://d.repec.org/n?u=RePEc:knz:dpteco:1512&r=for |
| By: | Zhiwei Shen; Matthias Ritter; ; |
| Abstract: | The increasing share of wind energy in the portfolio of energy sources highlights its uncertainties due to changing weather conditions. To account for the uncertainty in predicting wind power production, this article examines the volatility forecasting abilities of different GARCH-type models for wind power production. Moreover, due to characteristic features of the wind power process, such as heteroscedasticity and nonlinearity, we also investigate the use of a Markov regime-switching GARCH (MRS-GARCH) model on forecasting volatility of wind power. The realized volatility, which is derived from lower-scale data, serves as a benchmark for the latent volatility. We find that the MRS-GARCH model significantly outperforms traditional GARCH models in predicting the volatility of wind power, while the exponential GARCH model is superior among traditional GARCH models. |
| Keywords: | Wind energy, volatility forecasting, GARCH models, Markov regime-switching, realized volatility |
| JEL: | C22 Q42 Q47 |
| Date: | 2015–05 |
| URL: | http://d.repec.org/n?u=RePEc:hum:wpaper:sfb649dp2015-026&r=for |
| By: | Manabu Asai (Faculty of Economics, Soka University); Michael McAleer (Econometric Institute, Erasmus School of Economics, Erasmus University Rotterdam and Tinbergen Institute, The Netherlands, Department of Quantitative Economics, Complutense University of Madrid, and Institute of Economic Research, Kyoto University.) |
| Abstract: | The paper investigates the impact of jumps in forecasting co-volatility, accommodating leverage effects. We modify the jump-robust two time scale covariance estimator of Boudt and Zhang (2013) such that the estimated matrix is positive definite. Using this approach we can disentangle the estimates of the integrated co-volatility matrix and jump variations from the quadratic covariation matrix. Empirical results for three stocks traded on the New York Stock Exchange indicate that the co-jumps of two assets have a significant impact on future co-volatility, but that the impact is negligible for forecasting weekly and monthly horizons. |
| Keywords: | Co-Volatility; Forecasting; Jump; Leverage Effects; Realized Covariance; Threshold Estimation. |
| JEL: | C32 C53 C58 G17 |
| Date: | 2015–02 |
| URL: | http://d.repec.org/n?u=RePEc:ucm:doicae:1502&r=for |
| By: | Harry Vander Elst (Université libre de Bruxelles) |
| Abstract: | We introduce the class of FloGARCH models in this paper. FloGARCH models provide a parsimonious joint model for low frequency returns and realized measures and are sufficiently flexible to capture long memory as well as asymmetries related to leverage effects. We analyze the performances of the models in a realistic numerical study and on the basis of a data set composed of 65 equities. Using more than 10 years of high-frequency transactions, we document significant statistical gains related to the FloGARCH models in terms of in-sample fit, out-of-sample fit and forecasting accuracy compared to classical and Realized GARCH models. |
| Keywords: | Realized GARCH models, high-frequency data, long memory, realized measures. |
| JEL: | C22 C53 C58 G17 |
| Date: | 2015–04 |
| URL: | http://d.repec.org/n?u=RePEc:nbb:reswpp:201504-280&r=for |
| By: | Henri Nyberg (University of Helsinki); Harri Pönkä (University of Helsinki and CREATES) |
| Abstract: | We study the directional predictability of monthly excess stock market returns in the U.S. and ten other markets using univariate and bivariate binary response models. Our main interest is on the potential benefits of predicting the signs of the returns jointly, focusing on the predictive power from the U.S. to foreign markets. We introduce a new bivariate probit model that allows for such a contemporaneous predictive linkage from one market to the other. Our in-sample and out-of-sample forecasting results indicate superior predictive performance of the new model over the competing models by statistical measures and market timing performance, suggesting gradual diffusion of predictive information from the U.S. to the other markets. |
| Keywords: | Excess stock return, Directional predictability, Bivariate probit model, Market timing |
| JEL: | C22 G12 G17 |
| Date: | 2015–05–05 |
| URL: | http://d.repec.org/n?u=RePEc:aah:create:2015-20&r=for |
| By: | Wanfeng Yan; Edgar van Tuyll van Serooskerken |
| Abstract: | Investors in stock market are usually greedy during bull markets and scared during bear markets. The greed or fear spreads across investors quickly. This is known as the herding e?ect, and often leads to a fast movement of stock prices. During such market regimes, stock prices change at a super-exponential rate and are normally followed by a trend reversal that corrects the previous over reaction. In this paper, we construct an indicator to measure the magnitude of the super-exponential growth of stock prices, by measuring the degree of the price network, generated from the price time series. Twelve major international stock indices have been investigated. Error diagram tests show that this new indicator has strong predictive power for?nancial extremes, both peaks and troughs. By varying the parameters used to construct the error diagram, we show the predictive power is very robust. The new indicator has a better performance than the LPPL pattern recognition indicator. |
| Date: | 2015–05 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:1505.04060&r=for |
| By: | J.D. Daron; David A. Stainforth |
| Abstract: | The Lorenz-63 model has been frequently used to inform our understanding of the Earth's climate and provide insight for numerical weather and climate prediction. Most studies have focused on the autonomous (time invariant) model behaviour in which the model's parameters are constants. Here we investigate the properties of the model under time-varying parameters, providing a closer parallel to the challenges of climate prediction, in which climate forcing varies with time. Initial condition (IC) ensembles are used to construct frequency distributions of model variables and we interpret these distributions as the time-dependent climate of the model. Results are presented that demonstrate the impact of ICs on the transient behaviour of the model climate. The location in state space from which an IC ensemble is initiated is shown to significantly impact the time it takes for ensembles to converge. The implication for climate prediction is that the climate may, in parallel with weather forecasting, have states from which its future behaviour is more, or less, predictable in distribution. Evidence of resonant behaviour and path dependence is found in model distributions under time varying parameters, demonstrating that prediction in nonautonomous nonlinear systems can be sensitive to the details of time-dependent forcing/parameter variations. Single model realisations are shown to be unable to reliably represent the model's climate; a result which has implications for how real-world climatic timeseries from observation are interpreted. The results have significant implications for the design and interpretation of Global Climate Model experiments. Over the past 50 years, insight from research exploring the behaviour of simple nonlinear systems has been fundamental in developing approaches to weather and climate prediction. The analysis herein utilises the much studied Lorenz-63 model to understand the potential behaviour of nonlinear systems, such as the 5 climate, when subject to time-varying external forcing, such as variations in atmospheric greenhouse gases or solar output. Our primary aim is to provide insight which can guide new approaches to climate model experimental design and thereby better address the uncertainties associated with climate change prediction. We use ensembles of simulations to generate distributions which 10 we refer to as the \climate" of the time-variant Lorenz-63 model. In these ensemble experiments a model parameter is varied in a number of ways which can be seen as paralleling both idealised and realistic variations in external forcing of the real climate system. Our results demonstrate that predictability of climate distributions under time varying forcing can be highly sensitive to 15 the specification of initial states in ensemble simulations. This is a result which at a superficial level is similar to the well-known initial condition sensitivity in weather forecasting, but with different origins and different implications for ensemble design. We also demonstrate the existence of resonant behaviour and a dependence on the details of the \forcing" trajectory, thereby highlighting 20 further aspects of nonlinear system behaviour with important implications for climate prediction. Taken together, our results imply that current approaches to climate modeling may be at risk of under-sampling key uncertainties likely to be significant in predicting future climate. |
| JEL: | J1 |
| URL: | http://d.repec.org/n?u=RePEc:ehl:lserod:61890&r=for |