|
on Forecasting |
| By: | José García-Montalvo; Omiros Papaspiliopoulos; Timothée Stumpf-Fétizon |
| Abstract: | We propose a new methodology for predicting electoral results that combines a fundamental model and national polls within an evidence synthesis framework. Although novel, the methodology builds upon basic statistical structures, largely modern analysis of variance type models, and it is carried out in open-source software. The methodology is largely motivated by the specific challenges of forecasting elections with the participation of new political parties, which is becoming increasingly common in the post-2008 European panorama. Our methodology is also particularly useful for the allocation of parliamentary seats, since the vast majority of available opinion polls predict at the national level whereas seats are allocated at local level. We illustrate the advantages of our approach relative to recent competing approaches using the 2015 Spanish Congressional Election. In general, the predictions of our model outperform the alternative specifications, including hybrid models that combine fundamental and polls' models. Our forecasts are, in relative terms, particularly accurate to predict the seats obtained by each political party. |
| Keywords: | multilevel model, Bayesian machine learning, inverse regression, evidence synthesis, elections |
| JEL: | C11 C53 C63 D72 |
| Date: | 2018–12 |
| URL: | http://d.repec.org/n?u=RePEc:bge:wpaper:1065&r=for |
| By: | Pablo Montero-Manso; George Athanasopoulos; Rob J Hyndman; Thiyanga S Talagala |
| Abstract: | We propose an automated method for obtaining weighted forecast combinations using time series features. The proposed approach involves two phases. First, we use a collection of time series to train a meta-model to assign weights to various possible forecasting methods with the goal of minimizing the average forecasting loss obtained from a weighted forecast combination. The inputs to the meta-model are features extracted from each series. In the second phase, we forecast new series using a weighted forecast combination where the weights are obtained from our previously trained meta-model. Our method outperforms a simple forecast combination, and outperforms all of the most popular individual methods in the time series forecasting literature. The approach achieved second position in the M4 competition. |
| Keywords: | time series feature, forecast combination, XGBoost, M4 competition, meta-learning. |
| JEL: | C10 C14 C22 |
| Date: | 2018 |
| URL: | http://d.repec.org/n?u=RePEc:msh:ebswps:2018-19&r=for |
| By: | Havranek, Tomas; Zeynalov, Ayaz |
| Abstract: | In this paper, we examine the usefulness of Google Trends data in predicting monthly tourist arrivals and overnight stays in Prague during the period between January 2010 and December 2016. We offer two contributions. First, we analyze whether Google Trends provides significant forecasting improvements over models without search data. Second, we assess whether a high-frequency variable (weekly Google Trends) is more useful for accurate forecasting than a low-frequency variable (monthly tourist arrivals) using Mixed-data sampling (MIDAS). Our results stress the potential of Google Trends to offer more accurate prediction in the context of tourism: we find that Google Trends information, both two months and one week ahead of arrivals, is useful for predicting the actual number of tourist arrivals. The MIDAS forecasting model that employs weekly Google Trends data outperforms models using monthly Google Trends data and models without Google Trends data. |
| Keywords: | Google trends,mixed-frequency data,forecasting,tourism |
| JEL: | C53 L83 |
| Date: | 2018 |
| URL: | http://d.repec.org/n?u=RePEc:zbw:esprep:187420&r=for |
| By: | José Garcia Montalvo; Omiros Papaspiliopoulos; Timothée Stumpf-Fétizon |
| Abstract: | We propose a new methodology for predicting electoral results that com- bines a fundamental model and national polls within an evidence synthesis framework. Although novel, the methodology builds upon basic statistical structures, largely modern analysis of variance type models, and it is car- ried out in open-source software. The methodology is largely motivated by the speci c challenges of forecasting elections with the participation of new political parties, which is becoming increasingly common in the post-2008 European panorama. Our methodology is also particularly useful for the al- location of parliamentary seats, since the vast majority of available opinion polls predict at the national level whereas seats are allocated at local level. We illustrate the advantages of our approach relative to recent competing approaches using the 2015 Spanish Congressional Election. In general the predictions of our model outperform the alternative speci cations, including hybrid models that combine fundamental and polls' models. Our forecasts are, in relative terms, particularly accurate to predict the seats obtained by each political party. |
| Keywords: | Multilevel models, Bayesian machine learning, inverse regression, evidence synthesis, elections |
| Date: | 2018–12 |
| URL: | http://d.repec.org/n?u=RePEc:upf:upfgen:1624&r=for |
| By: | Luigi Grossi (University of Verona); Fany Nan (European Commission's Joint Research Centre (JRC)) |
| Abstract: | In this paper a robust approach to modelling electricity spot prices is introduced. Differently from what has been recently done in the literature on electricity price forecasting, where the attention has been mainly drawn by the prediction of spikes, the focus of this contribution is on the robust estimation of nonlinear SETARX models (Self-Exciting Threshold Auto Regressive models with eXogenous regressors). In this way, parameters estimates are not, or very lightly, influenced by the presence of extreme observations and the large majority of prices, which are not spikes, could be better forecasted. A Monte Carlo study is carried out in order to select the best weighting function for Generalized M-estimators of SETAR processes. A robust procedure to select and estimate nonlinear processes for electricity prices is introduced, including robust tests for stationarity and nonlinearity and robust information criteria. The application of the procedure to the Italian electricity market reveals the forecasting superiority of the robust GM-estimator based on the polynomial weighting function respect to the non-robust Least Squares estimator. Finally, the introduction of external regressors in the robust estimation of SETARX processes contributes to the improvement of the forecasting ability of the model. |
| Keywords: | Electricity Price, Nonlinear Time Series, Price Forecasting, Robust GM-Estimator, Spikes, Threshold Models |
| JEL: | C13 C15 C22 C53 Q47 |
| Date: | 2018 |
| URL: | http://d.repec.org/n?u=RePEc:ieb:wpaper:doc2018-10&r=for |
| By: | Gunnhildur H. Steinbakk; Alex Lenkoski; Ragnar Bang Huseby; Anders L{\o}land; Tor Arne {\O}ig{\aa}rd |
| Abstract: | Accurate forecasts of electricity spot prices are essential to the daily operational and planning decisions made by power producers and distributors. Typically, point forecasts of these quantities suffice, particularly in the Nord Pool market where the large quantity of hydro power leads to price stability. However, when situations become irregular, deviations on the price scale can often be extreme and difficult to pinpoint precisely, which is a result of the highly varying marginal costs of generating facilities at the edges of the load curve. In these situations it is useful to supplant a point forecast of price with a distributional forecast, in particular one whose tails are adaptive to the current production regime. This work outlines a methodology for leveraging published bid/ask information from the Nord Pool market to construct such adaptive predictive distributions. Our methodology is a non-standard application of the concept of error-dressing, which couples a feature driven error distribution in volume space with a non-linear transformation via the published bid/ask curves to obtain highly non-symmetric, adaptive price distributions. Using data from the Nord Pool market, we show that our method outperforms more standard forms of distributional modeling. We further show how such distributions can be used to render `warning systems' that issue reliable probabilities of prices exceeding various important thresholds. |
| Date: | 2018–12 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:1812.02433&r=for |
| By: | Carlo Fezzi; Luca Mosetti |
| Abstract: | Electricity price forecasting models are typically estimated via rolling windows, i.e. by using only the most recent observations. Nonetheless, the current literature does not provide much guidance on how to select the size of such windows. This paper shows that determining the appropriate window prior to estimation dramatically improves forecasting performances. In addition, it proposes a simple two-step approach to choose the best performing models and window sizes. The value of this methodology is illustrated by analyzing hourly datasets from two large power markets with a selection of ten different forecasting models. Incidentally, our empirical application reveals that simple models, such as the linear regression, can perform surprisingly well if estimated on extremely short samples. |
| Keywords: | electricity price forecasting, day-ahead market, parameter instability, bandwidth selection, artificial neural networks |
| JEL: | C22 C45 C51 C53 Q47 |
| Date: | 2018 |
| URL: | http://d.repec.org/n?u=RePEc:trn:utwprg:2018/10&r=for |
| By: | Susan Wahab (Department of Accounting, University of Hartford); Karen Teitel (Department of Economics, College of the Holy Cross); Bernard Morzuch (University of Massachusetts - Amherst) |
| Abstract: | We examine the relative efficiency of whisperers’ and analysts’ forecasts of one-quarter-ahead earnings per share (EPS) and identify commonalities and differences in their use of fundamentals to forecast earnings. Results suggest that (a) fundamentals that focus on sales and cost of sales are relevant in explaining one-quarter-ahead EPS changes; (b) whisperers focus on cash flow fundamentals and accrual-based earnings measures in their one-quarter-ahead forecasts, whereas analysts focus on only cash flow fundamentals; and (c) although neither analysts nor whisperers fully incorporate information contained in fundamentals and accrual-based earnings measures in their forecasts, whisperers’ earnings forecast model (forecast errors model) exhibits higher (lower) explanatory power than that of analysts. We also examine robustness of our results by reestimating the models using a two-way random-effects panel data estimator. Although our conclusions remain the same, more statistically significant fundamentals emerge in panel regression results. Evidence presented in this article is consistent with (a) whisperers being different from analysts and (b) whisper forecasts containing unique incremental information beyond that of analysts’ forecasts. Market participants may want to consider using both forecasts when making investment decisions. |
| Keywords: | Accounting, Earnings |
| JEL: | M41 |
| Date: | 2017–12 |
| URL: | http://d.repec.org/n?u=RePEc:hcx:wpaper:1707&r=for |
| By: | Antonio Pacifico |
| Abstract: | The paper develops empirical implementations of the standard time-varying Panel Bayesian VAR model to deal with confounding and latent effects. Bayesian computations and mixed hierarchical distributions are used to generate posteriors of conditional impulse responses and conditional forecasts. An empirical application to Eurozone countries illustrates the functioning of the model. A survey on policy recommendations and business cycles convergence are also conducted. The paper would enhance the more recent studies to evaluate idiosyncratic business cycles, policy-making, and structural spillovers forecasting. The analysis confirms the importance to separate common shocks from propagation of country- and variable-specific shocks. |
| Keywords: | Hierarchical Mixture Distributions in Normal Linear Model; Bayesian Model Averaging; Panel VAR; Forecasting; Structural Spillovers; MCMC Implementations. |
| JEL: | A2 D1 D2 |
| Date: | 2018–12–15 |
| URL: | http://d.repec.org/n?u=RePEc:eei:rpaper:eeri_rp_2018_15&r=for |
| By: | Nikolaos Kourentzes; George Athanasopoulos |
| Abstract: | Key to ensuring a successful tourism sector is timely policy making and detailed planning. National policy formulation and strategic planning requires long-term forecasts at an aggregate level, while regional operational decisions require short-term forecasts, relevant to local tourism operators. For aligned decisions at all levels, supporting forecasts must be `coherent', that is they should add up appropriately, across relevant demarcations (e.g., geographical divisions or market segments) and also across time. We propose an approach for generating coherent forecasts across both cross-sections and planning horizons for Australia. This results in significant improvements in forecast accuracy with substantial decision making benefits. Coherent forecasts help break intra- and inter-organisational information and planning silos, in a data driven fashion, blending information from different sources. |
| Keywords: | cross-sectional aggregation, temporal aggregation, forecast combinations, spatial correlations. |
| Date: | 2018 |
| URL: | http://d.repec.org/n?u=RePEc:msh:ebswps:2018-24&r=for |