|
on Forecasting |
| By: | Afees A. Salisu (Centre for Econometric & Allied Research, University of Ibadan, Ibadan, Nigeria); Rangan Gupta (Department of Economics, University of Pretoria, Pretoria, South Africa); Won Joong Kim (Department of Economics, Konkuk University, Seoul, Republic of Korea) |
| Abstract: | We examine exchange rate predictability using time-varying and constant parameter models that are conditioned on three variants of Taylor rules as well as six additional alternative models, namely, monetary model (MM); purchasing power parity (PPP); uncovered interest rate parity (UIRP) and three different factor (F1, F2 and F3) models, for BRICS countries. Monthly consumer price index, industrial production index, interest rate, broad money and exchange rates were used to construct the alternative fundamentals for exchange rate predictability for the period of January 1999 and March 2020. The out-of-sample forecast performances of the contending models were evaluated at the forecasting horizons of 1, 4, 8 and 12 using RMSFE and DM statistics, under the full, pre-GFC and post-GFC sample periods. We find that models conditioned on the Taylor rule fundamentals with homogeneous coefficients without interest rate smoothing as well as PPP- and UIRP-based fundamentals offer better exchange rate predictability of the BRICS than the random walk model across the forecast horizons. In addition, constant parameter models offer superior forecasting ability relative to the time-varying parameter models. Our results are sensitive to the data sample, frequency and the choice of fundamentals captured in the predictive model of exchange rate. |
| Keywords: | Exchange Rate Predictability, BRICS, time-varying parameter (TVP) model, Taylor rule, random walk |
| JEL: | F31 F37 |
| Date: | 2021–02 |
| URL: | http://d.repec.org/n?u=RePEc:pre:wpaper:202116&r=all |
| By: | Alexandre Bonnet R. Costa; Pedro Cavalcanti G. Ferreira; Wagner P. Gaglianone; Osmani Teixeira C. Guillén; João Victor Issler; Yihao Lin |
| Abstract: | The purpose of this paper is to explore machine learning techniques to forecast the oil price. In the era of big data, we investigate whether new automated tools can improve over traditional approaches in terms of forecast accuracy. Oil price point and density forecasts are built from 22 methods, including regression trees (random forest, quantile regression forest, xgboost), regularization procedures (elastic net, lasso, ridge), standard econometric models and forecast combinations, besides the structural factor model of Schwartz and Smith (2000). The database contains 315 macroeconomic and financial variables, used to build high-dimensional models. To evaluate the predictive power of each method, an extensive pseudo out-of-sample forecasting exercise is built, in monthly and quarterly frequencies, with horizons from one month up to five years. Overall, the results indicate a good performance of the machine learning methods in the short run. Up to six months, the lasso-based models, oil future prices, and the Schwartz-Smith model provide the best forecasts. At longer horizons, forecast combinations also become relevant. In several cases, the accuracy gains in respect to the random walk forecast are statistically significant and reach two-digit figures, in percentage terms, using the R2 out-of-sample statistic; an expressive achievement compared to the previous literature. |
| Date: | 2021–02 |
| URL: | http://d.repec.org/n?u=RePEc:bcb:wpaper:544&r=all |
| By: | Andrew B. Martinez (Dept of Economics and Institute for New Economic Thinking at the Oxford Martin School, University of Oxford); Jennifer L. Castle (Magdalen College, Climate Econometrics and Institute for New Economic Thinking at the Oxford Martin School, University of Oxford); David F. Hendry (Nuffield College, Climate Econometrics and Institute for New Economic Thinking at the Oxford Martin School, University of Oxford) |
| Abstract: | We investigate whether smooth robust methods for forecasting can help mitigate pronounced and persistent failure across multiple forecast horizons. We demonstrate that naive predictors are interpretable as local estimators of the long-run relationship with the advantage of adapting quickly after a break, but at a cost of additional forecast error variance. Smoothing over naive estimates helps retain these advantages while reducing the costs, especially for longer forecast horizons. We derive the performance of these predictors after a location shift, and confirm the results using simulations. We apply smooth methods to forecasts of U.K. productivity and U.S. 10-year Treasury yields and show that they can dramatically reduce persistent forecast failure exhibited by forecasts from macroeconomic models and professional forecasters. |
| Keywords: | Location Shifts; Long differencing; Productivity forecasts; Robust forecasts. JEL codes: C51, C53 |
| Date: | 2021–01–14 |
| URL: | http://d.repec.org/n?u=RePEc:nuf:econwp:2101&r=all |
| By: | Robert W. Rich; Joseph Tracy |
| Abstract: | This paper examines data from the European Central Bank’s Survey of Professional Forecasters to investigate whether participants display equal predictive performance. We use panel data models to evaluate point- and density-based forecasts of real GDP growth, inflation, and unemployment. The results document systematic differences in participants’ forecast accuracy that are not time invariant, but instead vary with the difficulty of the forecasting environment. Specifically, we find that some participants display higher relative accuracy in tranquil environments, while others display higher relative accuracy in volatile environments. We also find that predictive performance is positively correlated across target variables and horizons, with density forecasts generating stronger correlation patterns. Taken together, the results support the development of expectations models featuring persistent heterogeneity. |
| Keywords: | professional forecasters; survey data; forecast accuracy; point forecasts; density forecasts; persistent heterogeneity |
| JEL: | C12 C33 C53 |
| Date: | 2021–02–25 |
| URL: | http://d.repec.org/n?u=RePEc:fip:fedcwq:90000&r=all |
| By: | Boussios, David; Skoriansky, Sharon Raszap; MacLachlan, Matthew |
| Abstract: | This report examines the potential for statistical forecast models to improve the performance of the U.S. Department of Agriculture’s (USDA) long-term agricultural baseline projections for harvested area for U.S. corn, soybeans, and wheat. After-the-fact analysis for years 1997 to 2017 reveals the baseline projections have, historically, consistently overestimated the harvested area of wheat and underestimated soybean area. The baseline projections also tend to underestimate the corn area, though to a lesser degree. Part of the difference between the projections and realized values is likely attributable to policy, program, weather, and other unforeseen changes when USDA developed the projections. Still, the results of quantitative forecast models show there may be substantial potential for improvement on the existing methodology. Forecasts generated using 3 econometric time-series models did not improve performance relative to the current baseline approach for nearer forecast horizons but improved performance for projection horizon lengths of 8-10, 2-10, and 4-10 years for harvested area of corn, soybeans, and wheat, respectively, when using 1 of our statistical measures. The forecasts generated using the econometric models produce predictions with an average absolute forecasting error 10 years out that is between 26 percent to 60 percent smaller than those provided by baseline projections. The results suggest that econometric models offer the potential to improve the performance of forecasting long-term trends in agricultural markets. As of 2020, USDA begun using statistical forecast models such as these when developing its long-term agricultural projections as complements to the existing process. USDA is also in the process of testing these models for additional commodities to improve the long-term projections for all commodities. |
| Keywords: | Agribusiness, Agricultural and Food Policy, Agricultural Finance, Crop Production/Industries, Farm Management, Land Economics/Use |
| Date: | 2021–02–10 |
| URL: | http://d.repec.org/n?u=RePEc:ags:usdami:309619&r=all |
| By: | Boussios, David; Skorbiansky, Sharon Raszap; Maclachlan, Matthew |
| Abstract: | This report examines the potential for statistical forecast models to improve the performance of the U.S. Department of Agriculture’s (USDA) long-term agricultural baseline projections for the harvested area for U.S. corn, soybeans, and wheat. After-the-fact analysis for years 1997 to 2017 reveals the baseline projections have, historically, consistently overestimated the harvested area of wheat and underestimated soybean area. The baseline projections also tend to underestimate the corn area, though to a lesser degree. Part of the difference between the projections and realized values is likely attributable to policy, program, weather, and other unforeseen changes when USDA developed the projections. Still, the results of quantitative forecast models show there may be substantial potential for improvement on the existing methodology. Forecasts generated using 3 econometric time-series models did not improve performance relative to the current baseline approach for nearer forecast horizons but improved performance for projection horizon lengths of 8-10, 2-10, and 4-10 years for harvested area of corn, soybeans, and wheat, respectively, when using 1 of our statistical measures. The forecasts generated using the econometric models produce predictions with an average absolute forecasting error 10 years out that is between 26 percent to 60 percent smaller than those provided by baseline projections. The results suggest that econometric models offer the potential to improve the performance of forecasting long-term trends in agricultural markets. As of 2020, USDA begun using statistical forecast models such as these when developing its long-term agricultural projections as complements to the existing process. USDA is also in the process of testing these models for additional commodities to improve the long-term projections for all commodities. |
| Keywords: | Agribusiness, Agricultural Finance, Crop Production/Industries, Demand and Price Analysis, Farm Management, Land Economics/Use |
| Date: | 2021–02–10 |
| URL: | http://d.repec.org/n?u=RePEc:ags:usdami:309616&r=all |
| By: | Matteo Bonato (Department of Economics and Econometrics, University of Johannesburg, Auckland Park, South Africa; IPAG Business School, 184 Boulevard Saint-Germain, 75006 Paris, France); Oguzhan Cepni (Copenhagen Business School, Department of Economics, Porcelaenshaven 16A, Frederiksberg DK-2000, Denmark; Central Bank of the Republic of Turkey, Haci Bayram Mah. Istiklal Cad. No:10 06050, Ankara, Turkey); Rangan Gupta (Department of Economics, University of Pretoria, Pretoria, 0002, South Africa); Christian Pierdzioch (Department of Economics, Helmut Schmidt University, Holstenhofweg 85, P.O.B. 700822, 22008 Hamburg, Germany) |
| Abstract: | We use an international dataset on 5-minutes interval intraday data covering nine leading markets and regions to construct measures of realized volatility, realized jumps, realized skewness, and realized kurtosis of returns of international Real Estate Investment Trusts (REITs) over the daily period of September, 2008 to August, 2020. We study out-of-sample the predictive value of realized skewness and realized kurtosis for realized volatility over and above realized jumps, where we also differentiate between measures of ``good" realized volatility and ``bad" realized volatility. We find that realized skewness and realized kurtosis significantly improve forecasting performance at a daily, weekly, and monthly forecast horizon, and that their contribution to forecasting performance outweighs in terms of significance the contribution of realized jumps. Our results have important implications for investors and policymakers. |
| Keywords: | REITs, International data, Realized volatility, Forecasting |
| JEL: | C22 C53 G15 |
| Date: | 2021–02 |
| URL: | http://d.repec.org/n?u=RePEc:pre:wpaper:202114&r=all |
| By: | Afees A. Salisu (Centre for Econometric and Allied Research, University of Ibadan, Ibadan, Nigeria); Rangan Gupta (Department of Economics, University of Pretoria, Private Bag X20, Hatfield 0028, South Africa); Ahamuefula E. Ogbonna (Centre for Econometric and Allied Research and Department of Statistics, University of Ibadan, Ibadan, Oyo State, Nigeria) |
| Abstract: | This study examines the out-of-sample predictability of market risks measured as tail risks for stock returns of eight (Canada, France, Germany, Japan, Italy, Switzerland, the United Kingdom (UK), and the United States (US)) advanced countries using a long-range monthly data of over a century. We follow the Conditional Autoregressive Value at Risk (CAViaR) of Engle and Manganelli (2004) to measure the tail risks since it utilizes the tail distribution rather the whole distribution. Consequently, we produce results for both 1% and 5% VaRs across four variants (Adaptive, Symmetric absolute value, Asymmetric slope and Indirect GARCH) of the CAViaR. Thereafter, we use relevant model diagnostics such as the Dynamic Quantile test (DQ) test and %Hits to determine the model that best fits the data. The results obtained are then used in the return predictability following the Westerlund and Narayan (2012, 2015) method which allows us to account for some salient features such as persistence, endogeneity and conditional heteroscedasticity effects. We consequently partition our models into three variants (one-predictor, two-predictor and three-predictor models) and examine their forecast performance in contrast with a driftless random walk model. Three findings are discernible from the empirical analysis. First, we find that the choice of VaR matters when determining the “best†fit CAViaR model for each return series as the outcome seems to differ between 1% and 5% VaRs. Second, the predictive model that incorporates both stock tail risk and oil tail risk produces better forecast outcomes than the one with own tail risk indicating the significance of both domestic and global risks in the return predictability of advanced countries. |
| Keywords: | Stock returns, Tail risks, Forecasting, Advanced equity markets |
| JEL: | C22 G15 G17 Q02 |
| URL: | http://d.repec.org/n?u=RePEc:pre:wpaper:202117&r=all |
| By: | Manuela Pedio |
| Abstract: | The Great Financial Crisis of 2008 – 2009 has raised the attention of policy-makers and researchers about the interconnectedness among the volatility of the returns of financial assets as a potential source of risk that extends beyond the usual changes in correlations and include transmission channels that operate through the higher order co-moments of returns. In this paper, we investigate whether a newly developed, forward-looking measure of volatility spillover risk based on option implied volatilities shows any predictive power for stock returns. We also compare the predictive performance of this measure with that of the volatility spillover index proposed by Diebold and Yilmaz (2008, 2012), which is based on realized, backward-looking volatilities instead. While both measures show evidence of in-sample predictive power, only the option-implied measure is able to produce out-of-sample forecasts that outperform a simple historical mean benchmark. |
| Keywords: | connectedness, volatility networks, implied volatility, realized volatility, equity return predictability, spillover risk |
| JEL: | G12 G17 |
| Date: | 2021 |
| URL: | http://d.repec.org/n?u=RePEc:baf:cbafwp:cbafwp20154&r=all |
| By: | Svatopluk Kapounek; Evžen Kocenda; Zuzana Kucerová |
| Abstract: | We analyze the exchange rate forecasting performance under the assumption of selective attention. Although currency markets react to a variety of different information, we hypothesize that market participants process only a limited amount of information. Our analysis includes more than 100,000 news articles relevant to the six most-traded foreign exchange currency pairs for the period of 1979–2016. We employ a dynamic model averaging approach to reduce model selection uncertainty and to identify time-varying probability to include regressors in our models. Our results show that smaller sizes models accounting for the presence of selective attention offer improved fitting and forecasting results. Specifically, we document a growing impact of foreign trade and monetary policy news on the euro/dollar exchange rate following the global financial crisis. Overall, our results point to the existence of selective attention in the case of most currency pairs. |
| Keywords: | exchange rate, selective attention, news, forecasting, dynamic model averaging |
| JEL: | F33 G41 C11 |
| Date: | 2021 |
| URL: | http://d.repec.org/n?u=RePEc:ces:ceswps:_8901&r=all |
| By: | Olsson, Henrik |
| Abstract: | We present a new Bayesian bootstrap method for election forecasts that combines traditional polling questions about people’s own intentions with their expectations about how others will vote. It treats each participant’s election winner expectation as an optimal Bayesian forecast given private and public evidence available to that individual. It then infers the independent evidence and aggregates it across participants. The bootstrap forecast outperforms aggregate national polls in the 2020 U.S. election, as well as the forecasts based on traditional polling questions posed on large national probabilistic samples before the 2018 and 2020 U.S. elections. The bootstrap forecast puts most weight on people’s expectations about how their social contacts will vote, which might incorporate information about voters who are difficult to reach or who hide their true intentions. Beyond election polling, the new method is expected to improve the validity of other social science surveys. |
| Date: | 2021–02–18 |
| URL: | http://d.repec.org/n?u=RePEc:osf:osfxxx:nqcgs&r=all |
| By: | Andrea Fronzetti Colladon (University of Perugia); Stefano Grassi (University of Rome Tor Vergata); Francesco Ravazzolo (Free University of Bozen—Bolzano and CAMP, BI Norwegian Business School); Francesco Violante (CREST, GENES, ENSAE Paris, Institut Polytechnique de Paris and CREATES - Aarhus University) |
| Abstract: | This paper uses a new textual data index for predicting stock market data. The index is applied to a large set of news to evaluate the importance of one or more general economic related keywords appearing in the text. The index assesses the importance of the economic related keywords, based on their frequency of use and semantic network position. We apply it to the Italian press and construct indices to predict Italian stock and bond market returns and volatilities in a recent sample period, including the COVID—19 crisis. The evidence ShOWS that the index captures the different phases of financial time series well. Moreover, results indicate strong evidence of predictability for bond market data, both returns and volatilities, Short and long maturities, and stock market volatility. |
| Date: | 2021–03–03 |
| URL: | http://d.repec.org/n?u=RePEc:crs:wpaper:2021-06&r=all |
| By: | Ricardo Crisóstomo |
| Abstract: | This document shows that disentangling sentiment-induced biases from fundamental expectations significantly improves the accuracy and consistency of probabilistic forecasts. Using data from 1994 to 2017, 15 stochastic models and risk-preference combinations are analyzed and in all possible cases a simple behavioral transformation delivers substantial forecast gains. The results are robust across different evaluation methods, risk-preference hypotheses and sentiment calibrations, demonstrating that behavioral effects can be effectively used to forecast asset prices. Further analyses confirm that the real-world densities outperform densities recalibrated to avoid past mistakes and improve predictive models where risk aversion is dynamically estimated from option prices. |
| Keywords: | Sentiment, density forecasts, pricing kernel, options data, behavioral finance |
| JEL: | C14 C52 C53 G12 G13 |
| Date: | 2021 |
| URL: | http://d.repec.org/n?u=RePEc:cnv:wpaper:dt_73en&r=all |
| By: | Manfred M. Fischer; Niko Hauzenberger; Florian Huber; Michael Pfarrhofer |
| Abstract: | Time-varying parameter (TVP) regressions commonly assume that time-variation in the coefficients is determined by a simple stochastic process such as a random walk. While such models are capable of capturing a wide range of dynamic patterns, the true nature of time variation might stem from other sources, or arise from different laws of motion. In this paper, we propose a flexible TVP VAR that assumes the TVPs to depend on a panel of partially latent covariates. The latent part of these covariates differ in their state dynamics and thus capture smoothly evolving or abruptly changing coefficients. To determine which of these covariates are important, and thus to decide on the appropriate state evolution, we introduce Bayesian shrinkage priors to perform model selection. As an empirical application, we forecast the US term structure of interest rates and show that our approach performs well relative to a set of competing models. We then show how the model can be used to explain structural breaks in coefficients related to the US yield curve. |
| Date: | 2021–02 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2102.13393&r=all |
| By: | Martin Bodenstein; Pablo Cuba-Borda; Jay Faris; Nils Gornemann |
| Abstract: | The global collapse in economic activity triggered by individual and policy-mandated responses to the spread of COVID-19 is unprecedented both in scale and origin. At the time of writing, U.S. GDP is expected by professional forecasters to contract a staggering 6 percent over the course of 2020 driven by its 32 percent collapse in the second quarter (measured at an annual rate). |
| Date: | 2021–02–01 |
| URL: | http://d.repec.org/n?u=RePEc:fip:fedgfn:2021-02-01-2&r=all |