|
on Forecasting |
| By: | Baumeister, Christiane; Kilian, Lutz |
| Abstract: | The U.S. Energy Information Administration regularly publishes short-term forecasts of the price of crude oil. Traditionally, such out-of-sample forecasts have been largely judgmental, making them difficult to replicate and justify, and not particularly successful when compared with naïve no-change forecasts, as documented in Alquist et al. (2013). Recently, a number of alternative econometric oil price forecasting models has been introduced in the literature and shown to be more accurate than the no-change forecast of the real price of oil. We investigate the merits of constructing real-time forecast combinations of six such models with weights that reflect the recent forecasting success of each model. Forecast combinations are promising for four reasons. First, even the most accurate forecasting models do not work equally well at all times. Second, some forecasting models work better at short horizons and others at longer horizons. Third, even the forecasting model with the lowest MSPE may potentially be improved by incorporating information from other models with higher MSPE. Fourth, one can think of forecast combinations as providing insurance against possible model misspecification and smooth structural change. We demonstrate that over the last 20 years suitably constructed real-time forecast combinations would have been more accurate than the no-change forecast at every horizon up to two years. Relative to the no-change forecast, forecast combinations reduce the mean-squared prediction error by up to 18%. They also have statistically significant directional accuracy as high as 77%. We conclude that suitably constructed forecast combinations should replace traditional judgmental forecasts of the price of oil. |
| Keywords: | Forecast combination; Model misspecification; Oil price; Real-time data; Structural change |
| JEL: | C53 E32 Q43 |
| Date: | 2013–07 |
| URL: | http://d.repec.org/n?u=RePEc:cpr:ceprdp:9569&r=for |
| By: | Granziera, Eleonora; Hubrich, Kirstin; Moon, Hyungsik Roger |
| Abstract: | In this paper we introduce Quasi Likelihood Ratio tests for one sided multivariate hypotheses to evaluate the null that a parsimonious model performs equally well as a small number of models which nest the benchmark. We show that the limiting distributions of the test statistics are non standard. For critical values we consider two approaches: (i) bootstrapping and (ii) simulations assuming normality of the mean square prediction error (MSPE) difference. The size and the power performance of the tests are compared via Monte Carlo experiments with existing equal and superior predictive ability tests for multiple model comparison. We find that our proposed tests are well sized for one step ahead as well as for multi-step ahead forecasts when critical values are bootstrapped. The experiments on the power reveal that the superior predictive ability test performs last while the ranking between the quasi likelihood-ratio test and the other equal predictive ability tests depends on the simulation settings. Last, we apply our test to draw conclusions about the predictive ability of a Phillips type curve for the US core inflation. JEL Classification: |
| Keywords: | direct multi-step forecasts, fixed regressors bootstrap., multi-model comparison, Out-of sample, point-forecast evaluation, predictive ability |
| Date: | 2013–08 |
| URL: | http://d.repec.org/n?u=RePEc:ecb:ecbwps:20131580&r=for |
| By: | Jonas Dovern; Ulrich Fritsche; Prakash Loungani; Natalia T. Tamirisa |
| Abstract: | We examine the behavior of forecasts for real GDP growth using a large panel of individual forecasts from 30 advanced and emerging economies during 1989–2010. Our main findings are as follows. First, our evidence does not support the validity of the sticky information model (Mankiw and Reis, 2002) for describing the dynamics of professional growth forecasts. Instead, the empirical evidence is more in line with implications of "noisy" information models (Woodford, 2002; Sims, 2003). Second, we find that information rigidities are more pronounced in emerging economies than advanced economies. Third, there is evidence of nonlinearities in forecast smoothing. It is less pronounced in the tails of the distribution of individual forecast revisions than in the central part of the distribution. |
| Keywords: | Economic growth;Developed countries;Emerging markets;Economic forecasting;forecast, economic, information, expectations |
| Date: | 2013–02–27 |
| URL: | http://d.repec.org/n?u=RePEc:imf:imfwpa:13/56&r=for |
| By: | Kenneth Bruninx; Erik Delarue; William D’haeseleer |
| Abstract: | As the share of wind power in the electricity system rises, the limited predictability of wind power generation becomes increasingly critical for operating a reliable electricity system. In most operational & economic models, the wind power forecast error (WPFE) is often assumed to have a Gaussian or so-called ï¢-distribution. However, these distributions are not suited to fully describe the skewed and heavy-tailed character of WPFE data. In this paper, the Lévy ï¡-stable distribution is proposed as an improved description of the WPFE. Based on 6 years of historical wind power data, three forecast scenarios with forecast horizons ranging from 1 to 24 hours are simulated via a persistence approach. The Lévy ï¡-stable distribution models the WPFE better than the Gaussian or so-called ï¢-distribution, especially for short term forecasts. In a case study, an analysis of historical WPFE data showed improvements over the Gaussian and ï¢-distribution between 137 and 567% in terms of cumulative squared residuals. The method presented allows to quantify the probability of a certain error, given a certain wind power forecast. This new statistical description of the WPFE can hold important information for short term economic & operational (reliability) studies in the field of wind power. |
| Keywords: | Error analysis, Lévy a-stable distribution, Statistical analysis, Stable process, Wind power forecasting, Wind power generation |
| Date: | 2013–07 |
| URL: | http://d.repec.org/n?u=RePEc:rsc:rsceui:2013/50&r=for |
| By: | Mark Rosenzweig; Christopher R. Udry |
| Abstract: | We use newly-available Indian panel data to estimate how the returns to planting-stage investments vary by rainfall realizations. We show that the forecasts significantly affect farmer investment decisions and that these responses account for a substantial fraction of the inter-annual variability in planting-stage investments, that the skill of the forecasts varies across areas of India, and that farmers respond more strongly to the forecast where there is more forecast skill and not at all when there is no skill. We show, using an IV strategy in which the Indian government forecast of monsoon rainfall serves as the main instrument, that the return to agricultural investment depends substantially on the conditions under which it is estimated. Using the full rainfall distribution and our profit function estimates, we find that Indian farmers on average under-invest, by a factor of three, when we compare actual levels of investments to the optimal investment level that maximizes expected profits. Farmers who use skilled forecasts have increased average profit levels but also have more variable profits compared with farmers without access to forecasts. Even modest improvements in forecast skill would substantially increase average profits. |
| JEL: | D24 D81 O12 O13 Q12 Q54 |
| Date: | 2013–08 |
| URL: | http://d.repec.org/n?u=RePEc:nbr:nberwo:19334&r=for |