|
on Forecasting |
| By: | Karol Szafranek (Narodowy Bank Polski, Warsaw School of Economics) |
| Abstract: | Accurate inflation forecasts lie at the heart of effective monetary policy. By utilizing a thick modelling approach, this paper investigates the out-of-sample quality of the short-term Polish headline inflation forecasts generated by a combination of thousands of bagged single hidden-layer feed-forward artificial neural networks in the period of systematically falling and persistently low inflation. Results indicate that the forecasts from this model outperform a battery of popular approaches, especially at longer horizons. During the excessive disinflation it has more accurately accounted for the slowly evolving local mean of inflation and remained only mildly biased. Moreover, combining several linear and nonlinear approaches with diverse underlying model assumptions delivers further statistically significant gains in the predictive accuracy and statistically outperforms a panel of examined benchmarks at multiple horizons. The robustness analysis shows that resigning from data preprocessing and bootstrap aggregating severely compromises the forecasting ability of the model. |
| Keywords: | inflation forecasting, artificial neural networks, principal components, bootstrap aggregating, forecast combination |
| JEL: | C22 C38 C45 C53 C55 |
| Date: | 2017 |
| URL: | http://d.repec.org/n?u=RePEc:nbp:nbpmis:262&r=for |
| By: | Gilles Mourre; Caterina Astarita; Anamaria Maftei |
| Abstract: | This paper provides an assessment of the uncertainty surrounding revenue predictions, through an ex post analysis of European Commission’s forecasts over the last 15 years. It estimates the forecast errors affecting revenue for all 28 Member States, using the different vintages of the autumn and spring Commission forecasts. It analyses both the direction and magnitude of errors, using standard summary statistics. The paper looks into the various components of forecast errors to better understand their drivers (forecasting error related to real GDP, inflation or revenue-to-GDP ratio) and which types of revenues (direct tax, indirect tax or social security contributions) are particularly affected. The paper also examines the pattern of revenue errors over time and in particular how revenue forecasts perform before, during and after the crisis. To further deepen the analysis, a set of tests are carried out on the quality of the prediction (serial correlation, unbiasedness, weak and informational efficiency). The estimator-based tests confirm the sound track record of the European Commission’s forecasts. This is also shown by a comparison with the OECD’s revenue forecasts. Lastly, the paper reviews various possible determinants of forecast errors and examines their significance by means of a pooled time series technique. The econometric study allows for the identification of factors which increase or reduce the risk of over-forecasting revenue. |
| JEL: | C1 E60 E66 |
| Date: | 2016–12 |
| URL: | http://d.repec.org/n?u=RePEc:euf:dispap:039&r=for |
| By: | Asai, M.; McAleer, M.J. |
| Abstract: | For forecasting volatility of futures returns, the paper proposes an indirect method based on the relationship between futures and the underlying asset for the returns and time-varying volatility. For volatility forecasting, the paper considers the stochastic volatility model with asymmetry and long memory, using high frequency data for the underlying asset. Empirical results for Nikkei 225 futures indicate that the adjusted R2 supports the appropriateness of the indirect method, and that the new method based on stochastic volatility models with the asymmetry and long memory outperforms the forecasting model based on the direct method using the pseudo long time series. |
| Keywords: | Forecasting, Volatility, Futures, Realized Volatility, Realized Kernel, Leverage Effects, Long Memory |
| JEL: | C22 C53 C58 G17 |
| Date: | 2017–01–15 |
| URL: | http://d.repec.org/n?u=RePEc:ems:eureir:99517&r=for |
| By: | Jorge A. Chan-Lau |
| Abstract: | Model selection and forecasting in stress tests can be facilitated using machine learning techniques. These techniques have proved robust in other fields for dealing with the curse of dimensionality, a situation often encountered in applied stress testing. Lasso regressions, in particular, are well suited for building forecasting models when the number of potential covariates is large, and the number of observations is small or roughly equal to the number of covariates. This paper presents a conceptual overview of lasso regressions, explains how they fit in applied stress tests, describes its advantages over other model selection methods, and illustrates their application by constructing forecasting models of sectoral probabilities of default in an advanced emerging market economy. |
| Date: | 2017–05–05 |
| URL: | http://d.repec.org/n?u=RePEc:imf:imfwpa:17/108&r=for |
| By: | MIYAKAWA Daisuke; MIYAUCHI Yuhei; Christian PEREZ |
| Abstract: | The goal of this paper is to forecast future firm performance with machine learning techniques. Using data on over one million Japanese firms with supply-chain linkage information provided by a credit reporting agency, we show high performance in the prediction of exit, sales growth, and profit growth. In particular, our constructed proxies far outperform the credit score assigned by the credit reporting agency based on a detailed survey and interviews of firms. Against such baseline score, our models are able to ex-ante identify 16% of exiting firms (baseline: 11%), 25% of firms experiencing growth in sales (baseline: 8%), and 22% of firms exhibiting positive profit growth (baseline: 13%). The proof of concept of this paper provides practical usage of machine learning methods in firm performance prediction. |
| Date: | 2017–05 |
| URL: | http://d.repec.org/n?u=RePEc:eti:dpaper:17068&r=for |