Lower Forecast Errors Research Articles

Weighting schemes in global VAR modelling: a forecasting exercise

We provide a comprehensive analysis of the out-of-sample predictive accuracy of different global vector autoregressive (GVAR) specifications based on alternative weighting schemes to address global spillovers across countries. In addition to weights based on bilateral trade, we entertain schemes based on different financial variables and geodesic distance. Our results indicate that models based on trade weights, which are standard in the literature, are systematically outperformed in terms of predictive accuracy by other specifications. We find that, while information on financial linkages helps improve the forecasting accuracy of GVAR models, averaging predictions by means of simple predictive likelihood weighting does not appear to systematically lead to lower forecast errors.

IFRS non‐GAAP earnings disclosures and fair value measurement

AbstractWe investigate IFRS non‐GAAP earnings adjustments for fair value remeasurements made by companies and analysts and the usefulness of these disclosures for analysts. Examining Australian listed (ASX 200) companies during 2008–2010 (576 firm‐years), we find that companies disclosing non‐GAAP earnings are more likely to have a higher incidence and magnitude of profit or loss items reflecting asset remeasurements and impairment in their financial statements. We find non‐GAAP disclosing companies are more likely to have analyst adjustments to earnings for these items and lower forecast error and dispersion in the following year, suggesting usefulness rather than opportunism in the adjustments.

Money and monetary policy in Israel during the last decade

This study examines how money and monetary policy have influenced output and inflation during the past decade in Israel by comparing two New Keynesian DSGE models. One is a baseline separable model (Gali, 2008) and the other assumes non-separable household preferences between consumption and money (Benchimol & Fourçans, 2012). We test both models by using rolling window Bayesian estimations over the last decade (2001–2013). The results of the presented dynamic analysis show that the sensitivity of output with respect to money shocks increased during the Dot-com, Intifada, and Subprime crises. The role of monetary policy increased during these crises, especially with regard to inflation, even though the effectiveness of conventional monetary policy decreased during the Subprime crisis. In addition, the non-separable model including money provides lower forecast errors than the baseline separable model without money, while the influence of money on output fluctuations can be seen as a good predictive indicator of bank and debt risks. By impacting and monitoring households’ money holdings, policy makers could improve their forecasts and crisis management through models considering monetary aggregates.

The accuracy of alternative GDP growth forecasts: Do they represent a credible alternative to the official ones?

Abstract The paper deals with the accuracy of the real GDP growth forecasts produced by two Czech non-governmental institutions: the Czech-Moravian Confederation of Trade Unions (CMKOS) and the Czech Banking Association (CBA) in the years 2007-2014 and 2011-2014 respectively. Utilizing a research method composed of simple (AFE), scale-dependent (RMSE) as well as relative (MASE) error measures, we found out that (i) CMKOS predictions achieved a lower forecasting error on average, beginning with a notable overestimation in the first turnover point from growth to decline (2008-2009), yet followed by gradual improvement resulting in superior accuracy over set benchmarks (Ministry of Finance, Czech National Bank, OECD) in later years (2010-2014). The CBA predictions, on the other hand, exhibited (ii) a high level of interconnection with official bodies (MF, CNB), but with overall inferior forecasting accuracy, despite the shorter time frame (2011-2014). Overall, the study suggests that of the two surveyed non-governmental bodies, only CMKOS forecasts represent a viable alternative to the official predictions published by the Ministry of Finance or the Czech National Bank, as CBA forecasts were found to be a less accurate satellite of these bodies.

Time-varying industry beta in Indian stock market and forecasting errors

Purpose – The purpose of this paper is to empirically estimate industry beta in Indian stock market with three alternative models and compare the accuracy of forecasting error to find the most suitable model for time-varying beta estimation. Design/methodology/approach – The paper applies the standard regression model, Kalman filter model, other statistical approaches and secondary material. Findings – The paper finds that the existence of dynamic beta in Indian market. The results also indicate systematic risk or beta of Indian industries is susceptible to the global economic effect. Finally, the Kalman filter generates the lower forecasting error compared to the other method for almost all the industries. Practical implications – The accurate estimation of beta which is a measure of systematic risk helps investors to make investment decision easier. The implication of this result is important for finance practitioners such as portfolio managers, investment advisors and security analysts. This study will help to determine the country risk with respect to the global index and analyze the global financial market integration effect on India. Originality/value – This paper reliably estimate industry portfolio beta for India. The time-varying beta is estimated using Kalman filter method which is rarely applied in Indian literature. This paper contributes by extending the knowledge of existing literature by introducing a new data set with Indian data which is not affected by the “data snooping” bias. This study will also help to determine the country risk with respect to the global index and analyze the global financial market integration effect on India.

Wind Energy Predictions of Small-Scale Turbine Output Using Exponential Smoothing and Feed-Forward Neural Network

This article presents the comparisons of energy production predictions of a small-scale 40 kW wind turbine using an exponential smoothing technique and multilayer feed-forward neural network. For wind energy predictions, the developed mathematical model based on exponential smoothing was used to smoothen any seasonality arising in the time series data obtained at the site. This model was developed using three smoothing constant values of 0.20, 0.65, and 0.90, as well as a combination of a smoothing constant value of 0.90 with a seasonal adjustment factor for prediction of a small-scale wind turbine output for a period of 12 months. In addition, an energy model based on a multilayer feed-forward neural network was used to compute the energy generation of the turbine. The seasonally adjusted forecast model accurately predicted the wind energy output with the lowest forecast errors when compared to the chosen three smoothing constants. The energy forecasts obtained from the seasonal adjusted forecast model and multilayer feed-forward neural network were compared to the actual energy generation of the turbine at the considered tower height in terms of their forecast erroneous values. KeywordsTime Series Data (TSD); Smoothing and Seasonal Factor; Exponential Smoothing; Feed-Forward Neural Network (FNN), Small-Scale Wind Turbine

ESG Rating in Investment Risk Analysis of Companies Listed on the Public Market in Poland

This study verifies the hypothesis that greater transparency of public companies in disclosing non‐financial (ESG) data results in lower volatility of return rates on securities issued by them, thereby reducing investment portfolio risk understood as return rate volatility. Non‐financial data reporting contributes to increased transparency, predictability of companies' operations and hence to significant reduction in information asymmetry on the capital market, and ultimately considerably reduces forecasting errors in risk–return profile of investment portfolios. The research conducted has shown that there is a large information gap on the Polish market, especially as regards ESG reporting. The overall level of reporting on non‐financial data is low. In the analysed period, the shares issued by companies with higher ESG rating were distinguished by an over‐average return rate and lower return rate volatility as well as lower forecasting error in return rates, which is indicated by the standard error parameter (alpha and beta coefficients).

Improving forecasts using equally weighted predictors

The usual procedure for developing linear models to predict any kind of target variable is to identify a subset of most important predictors and to estimate weights that provide the best possible solution for a given sample. The resulting “optimally” weighted linear composite is then used when predicting new data. This approach is useful in situations with large and reliable datasets and few predictor variables. However, a large body of analytical and empirical evidence since the 1970s shows that such optimal variable weights are of little, if any, value in situations with small and noisy datasets and a large number of predictor variables. In such situations, which are common for social science problems, including all relevant variables is more important than their weighting. These findings have yet to impact many fields. This study uses data from nine U.S. election-forecasting models whose vote-share forecasts are regularly published in academic journals to demonstrate the value of (a) weighting all predictors equally and (b) including all relevant variables in the model. Across the ten elections from 1976 to 2012, equally weighted predictors yielded a lower forecast error than regression weights for six of the nine models. On average, the error of the equal-weights models was 5% lower than the error of the original regression models. An equal-weights model that uses all 27 variables that are included in the nine models missed the final vote-share results of the ten elections on average by only 1.3 percentage points. This error is 48% lower than the error of the typical, and 29% lower than the error of the most accurate, regression model.

Non-GAAP Earnings Disclosures and IFRS

We investigate IFRS non-GAAP earnings adjustments for fair value remeasurements made by companies and analysts and the usefulness of these disclosures for analysts. Examining Australian listed (ASX 200) companies during 2008-2010 (576 firm-years), we find that companies disclosing non-GAAP earnings are more likely to have a higher incidence and magnitude of profit or loss items reflecting asset remeasurements and impairment in their financial statements. We find non-GAAP disclosing companies are more likely to have analyst adjustments to earnings for these items and lower forecast error and dispersion in the following year, suggesting usefulness rather than opportunism in the adjustments.

Volatility Forecasting Using Financial Statement Information

ABSTRACTThis paper examines whether financial statement information can predict future realized equity volatility incremental to market-based equity volatility forecasts. I use an analytical framework to identify accounting-based drivers of realized volatility. My main hypothesis is that accounting-based drivers can be used to forecast future realized volatility incremental to either past realized volatility or option-implied volatility. I confirm this empirically and document abnormal returns to an option-based trading strategy that takes a long (short) position in firms with financial statement information indicative of high (low) future realized volatility. These results suggest that accounting-based volatility drivers may serve as useful indicators of variance risk. Finally, I demonstrate that the incorporation of accounting-based fundamental information into forecasting models yields lower forecast errors relative to models based solely on past realized volatility.

ANALYZING THE DYNAMICS OF GROSS DOMESTIC PRODUCT GROWTH. A MIXED FREQUENCY MODEL APPROACH

Frequency mismatch has been a problem in time series econometrics. Many monthly economic and financial indicators are normally aggregated to match quarterly macroeconomic series such as Gross Domestic Product when performing econometric analysis. However, temporal aggregation, although widely accepted, is prone to information loss. To address this issue, mixed frequency modelling is employed by using state space models with time-varying parameters. Quarter-on-quarter growth rate of GDP estimates are treated as monthly series with missing observation. Using Kalman filter algorithm, state space models are estimated with eleven monthly economic indicators as explanatory variables. A one-step-ahead forecast for GDP growth rates is generated and as more indicators are included in the model, the predicted values became closer to the actual data. Further evaluation revealed that among the group competing models, using Consumer Price Index (CPI), growth rates of Philippine Stock Exchange Index (PSEi), Exchange Rate, Real Money Supply, Wholesale Price Index (WPI) and Merchandise Exports are the more important determinants of GDP growth and generated the most desirable forecasts (lower forecast errors).

Reference Class Forecasting: Resolving Its Challenge to Statistical Modeling

Statisticians generally consider statistical modeling superior (or at least a useful supplement) to experience-based intuition for estimating the outputs of a complex system. But recent psychological research has led to an enhancement of experience-based intuition known as reference class forecasting. The reference class forecasting approach has been championed as a superior alternative to statistical modeling and is already well-regarded in the planning community. This presents a challenge to statistical modeling. To address this challenge, this article uses a Bayesian approach for combining the reference class forecast and the model-based forecast. The Bayesian prior is informed by the reference class information. A likelihood function was constructed to reflect the model’s information. This approach was used to estimate healthcare costs under a voluntary employee benefit association (VEBA). The resulting Bayesian posterior forecast had lower variance (and lower forecast error) than either the model-based forecast or the reference-class forecast.

Judgmental Forecasting: Cognitive Reflection and Decision Speed

This research analyzes how individual differences affect performance in judgmental time‐series forecasting. Decision makers with the ability to balance intuitive judgment with cognitive deliberation, as measured by the cognitive reflection test, tend to have lower forecast errors. This relationship holds when controlling for intelligence. Furthermore, forecast errors increase for very fast or very slow decisions. We provide evidence that forecast performance can be improved by manipulating decision speed.

A 4DVAR System for the Navy Coastal Ocean Model. Part I: System Description and Assimilation of Synthetic Observations in Monterey Bay*

Abstract A 4D variational data assimilation system was developed for assimilating ocean observations with the Navy Coastal Ocean Model. It is described in this paper, along with initial assimilation experiments in Monterey Bay using synthetic observations. The assimilation system is tested in a series of twin data experiments to assess its ability to fit assimilated and independent observations by controlling the initial conditions and/or the external forcing while assimilating surface and/or subsurface observations. In all strong and weak constraint experiments, the minimization of the cost function is done with both the gradient descent method (in the control space) and the representer method (observation space). The accuracy of the forecasts following the analysis and the relevance of the retrieved forcing correction in the case of weak constraints are evaluated. It is shown that the assimilation system generally fits the assimilated and nonassimilated observations well in all experiments, yielding lower forecast errors.

Impact of the Royal Decree-Law 16/2012 on the number of prescriptions and pharmaceutical expenditures

this research aims to understand if the consequences on drug expenditures and number of prescriptions of Royal Decree-Law 16/2012 as estimated by the Ministry of Health, Social Services and Equality (MHSSE) are similar to those found by using common statistical approaches. In addition, several models have been built to forecast the evolution of both variables for the period September 2013-December 2014. the Box-Jenkins methodology and the Box-Tiao intervention analysis were applied to data of the period 2003-13 to forecast the monthly values of the number of prescriptions and pharmaceutical expenditures. Forecasts were used in a counter-factual analysis to be compared to the actual values of prescriptions and drug expenditures. Moreover, forecasts for the period September 2013 to December 2014 were obtained to observe the impact of the policy in the future. the counterfactual analysis estimated a decrease in the number of prescriptions of 12.18% and 12.83% in the pharmaceutical expenditure; these figures were 12,75% and 14,03% respectively, when the intervention analysis was used. the estimated reduction in the number of prescriptions for the period June 2012-August 2013 was similar to the figure offered by the MHSSE, while the reduction in the drug expenditure series was smaller. The Box-Jenkins methodology generated low forecast errors (less than 3%) what makes this procedure useful to reliably anticipate future consumptions.

Estimating and Forecasting APARCH‐Skew‐t Model by Wavelet Support Vector Machines

ABSTRACTThis paper concentrates on comparing estimation and forecasting ability of quasi‐maximum likelihood (QML) and support vector machines (SVM) for financial data. The financial series are fitted into a family of asymmetric power ARCH (APARCH) models. As the skewness and kurtosis are common characteristics of the financial series, a skew‐t distributed innovation is assumed to model the fat tail and asymmetry. Prior research indicates that the QML estimator for the APARCH model is inefficient when the data distribution shows departure from normality, so the current paper utilizes the semi‐parametric‐based SVM method and shows that it is more efficient than the QML under the skewed Student's‐t distributed error. As the SVM is a kernel‐based technique, we further investigate its performance by applying separately a Gaussian kernel and a wavelet kernel. The results suggest that the SVM‐based method generally performs better than QML for both in‐sample and out‐of‐sample data. The outcomes also highlight the fact that the wavelet kernel outperforms the Gaussian kernel with lower forecasting error, better generation capability and more computation efficiency. Copyright © 2014 John Wiley & Sons, Ltd.

Complexity is costly: a meta‐analysis of parametric and non‐parametric methods for short‐term population forecasting

Short‐term forecasts based on time series of counts or survey data are widely used in population biology to provide advice concerning the management, harvest and conservation of natural populations. A common approach to produce these forecasts uses time‐series models, of different types, fit to time series of counts. Similar time‐series models are used in many other disciplines, however relative to the data available in these other disciplines, population data are often unusually short and noisy and models that perform well for data from other disciplines may not be appropriate for population data. In order to study the performance of time‐series forecasting models for natural animal population data, we assembled 2379 time series of vertebrate population indices from actual surveys. Our data were comprised of three vastly different types: highly variable (marine fish productivity), strongly cyclic (adult salmon counts), and small variance but long‐memory (bird and mammal counts). We tested the predictive performance of 49 different forecasting models grouped into three broad classes: autoregressive time‐series models, non‐linear regression‐type models and non‐parametric time‐series models. Low‐dimensional parametric autoregressive models gave the most accurate forecasts across a wide range of taxa; the most accurate model was one that simply treated the most recent observation as the forecast. More complex parametric and non‐parametric models performed worse, except when applied to highly cyclic species. Across taxa, certain life history characteristics were correlated with lower forecast error; specifically, we found that better forecasts were correlated with attributes of slow growing species: large maximum age and size for fishes and high trophic level for birds.SynthesisEvaluating the data support for multiple plausible models has been an integral focus of many ecological analyses. However, the most commonly used tools to quantify support have weighted models’ hindcasting and forecasting abilities.For many applications, predicting the past may be of little interest. Concentrating only on the future predictive performance of time series models, we performed a forecasting competition among many different kinds of statistical models, applying each to many different kinds of vertebrate time series of population abundance. Low‐dimensional (simple) models performed well overall, but more complex models did slightly better when applied to time series of cyclic species (e.g. salmon).

A Regional GSI-Based Ensemble Kalman Filter Data Assimilation System for the Rapid Refresh Configuration: Testing at Reduced Resolution

Abstract A regional ensemble Kalman filter (EnKF) system is established for potential Rapid Refresh (RAP) operational application. The system borrows data processing and observation operators from the gridpoint statistical interpolation (GSI), and precalculates observation priors using the GSI. The ensemble square root Kalman filter (EnSRF) algorithm is used, which updates both the state vector and observation priors. All conventional observations that are used in the operational RAP GSI are assimilated. To minimize computational costs, the EnKF is run at ⅓ of the operational RAP resolution or about 40-km grid spacing, and its performance is compared to the GSI using the same datasets and resolution. Short-range (up to 18 h, the RAP forecast length) forecasts are verified against soundings, surface observations, and precipitation data. Experiments are run with 3-hourly assimilation cycles over a 9-day convectively active retrospective period from spring 2010. The EnKF performance was improved by extensive tuning, including the use of height-dependent covariance localization scales and adaptive covariance inflation. When multiple physics parameterization schemes are employed by the EnKF, forecast errors are further reduced, especially for relative humidity and temperature at the upper levels and for surface variables. The best EnKF configuration produces lower forecast errors than the parallel GSI run. Gilbert skill scores of precipitation forecasts on the 13-km RAP grid initialized from the 3-hourly EnKF analyses are consistently better than those from GSI analyses.

An Analysis of the Predictive Ability of Direct Cash Flow Information Under IFRS

In this paper, we examine the predictive ability of direct cash flow information under IFRS. Employing a combination of in- and out-of-sample cross sectional models, we provide the first empirical evidence on the predictive ability of direct cash flow information in an IFRS environment. Under IFRS, our models show the greatest explanatory power and lowest forecast errors when utilising information found only in a direct cash flow statement. Crucially, after controlling for size, we note that while our direct cash flow models report lower forecast errors for large firms, they produce no better forecasts for small firms than a parsimonious operating cash flow model. While confirming the predictive ability of direct cash flow information in an IFRS setting, our results question the benefit of the IASB and FASB mandating direct cash flow statements for small firms.

Application of Genetic Algorithm and Neural Network in Construction Cost Estimate

Genetic algorithm optimizing BP has been proposed to aim at handling locality minimum and low convergence speed. The method based on analyzing the basic fundamental states that how to use genetic algorithm to improve the ability of BP. After optimizing, the GA-BP model has been built up. The result of GA-BP model can get lower forecast error and iterations. For these reason, GA-BP model is appropriate for construction cost estimation.