|
on Forecasting |
| By: | Peter Reinhard Hansen (European University Institute and CREATES); Allan Timmermann (UCSD and CREATES) |
| Abstract: | Out-of-sample tests of forecast performance depend on how a given data set is split into estimation and evaluation periods, yet no guidance exists on how to choose the split point. Empirical forecast evaluation results can therefore be difficult to interpret, particularly when several values of the split point might have been considered. When the sample split is viewed as a choice variable, rather than being ?xed ex ante, we show that very large size distortions can occur for conventional tests of predictive accu- racy. Spurious rejections are most likely to occur with a short evaluation sample, while conversely the power of forecast evaluation tests is strongest with long out-of-sample periods. To deal with size distortions, we propose a test statistic that is robust to the effect of considering multiple sample split points. Empirical applications to predictabil- ity of stock returns and in?ation demonstrate that out-of-sample forecast evaluation results can critically depend on how the sample split is determined. |
| Keywords: | Out-of-sample forecast evaluation, data mining, recursive estimation, predictability of stock returns, in?ation forecasting. |
| JEL: | C12 C53 G17 |
| Date: | 2012–02–07 |
| URL: | http://d.repec.org/n?u=RePEc:aah:create:2012-43&r=for |
| By: | Torsten Schmidt; Simeon Vosen |
| Abstract: | Information about special events can improve economic forecasts substantially. However, due to the lack of timely quantitative data about these events, it has been difficult for professional forecasters to utilise such information in their forecasts. This paper investigates whether Internet search data can improve economic predictions in times of special events. An analysis of “cash for clunkers” programs in four selected countries exemplifies that including search query data into statistical forecasting models improves the forecasting performance in almost all cases. However, the challenge to identify irregular events and to find the appropriate time series from Google Insights for search remains. |
| Keywords: | Forecast adjustment; Google Trends; private consumption |
| JEL: | C53 E21 E27 |
| Date: | 2012–11 |
| URL: | http://d.repec.org/n?u=RePEc:rwi:repape:0382&r=for |
| By: | Peter Reinhard Hansen (European University Institute and CREATES); Allan Timmermann (UCSD and CREATES) |
| Abstract: | We establish the equivalence between a commonly used out-of-sample test of equal predictive accuracy and the difference between two Wald statistics. This equivalence greatly simpli?es the computational burden of calculating recursive out-of-sample tests and evaluating their critical values. Our results shed new light on many aspects of the test and establishes certain weaknesses associated with using out-of-sample forecast comparison tests to conduct inference about nested regression models. |
| Keywords: | Out-of-sample Forecast Evaluation, Nested Models, Testing. |
| JEL: | C12 C53 G17 |
| Date: | 2012–10–10 |
| URL: | http://d.repec.org/n?u=RePEc:aah:create:2012-45&r=for |
| By: | Nowotarski, Jakub; Tomczyk, Jakub; Weron, Rafal |
| Abstract: | When building stochastic models for electricity spot prices the problem of uttermost importance is the estimation and consequent forecasting of a component to deal with trends and seasonality in the data. While the short-term seasonal components (daily, weekly) are more regular and less important for valuation of typical power derivatives, the long-term seasonal components (LTSC; seasonal, annual) are much more difficult to tackle. Surprisingly, in many academic papers dealing with electricity spot price modeling the importance of the seasonal decomposition is neglected and the problem of forecasting it is not considered. With this paper we want to fill the gap and present a thorough study on estimation and forecasting of the LTSC of electricity spot prices. We consider a battery of models based on Fourier or wavelet decomposition combined with linear or exponential decay. We find that all considered wavelet-based models are significantly better in terms of forecasting spot prices up to a year ahead than all considered sine-based models. This result questions the validity and usefulness of stochastic models of spot electricity prices built on sinusoidal long-term seasonal components. |
| Keywords: | Electricity spot price; Long-term seasonal component; Robust modeling; Forecasting; Wavelets |
| JEL: | C53 C45 Q47 C80 |
| Date: | 2012–11–11 |
| URL: | http://d.repec.org/n?u=RePEc:pra:mprapa:42563&r=for |
| By: | Alessandro Flamini (Department of Economics and Management, University of Pavia) |
| Abstract: | Forecast accuracy in macroeconomics is based on statistical techniques for extrapolating time series. This paper takes a new theoretical route studying the relation between forecast accuracy of macroeconomic variables and alternative monetary policies. Considering optimal policy with model-parameter uncertainty in a small open-economy, the paper shows that Domestic Inflation Targeting (DIT) leads to more forecast accuracy than Consumer Price index Inflation Targeting (CPIIT). Furthermore, forecast accuracy and policy aggressiveness turn out to be inversely related, and the trade-o¤ is more severe under CPIIT. These results are obtained in a New-Keynesian model measuring forecast accuracy by the volatility of simulated fan-charts. |
| Keywords: | Multiplicative uncertainty; Markov jump linear quadratic systems; small open-economy; optimal monetary policy; inflation index. |
| JEL: | E52 E58 F41 |
| Date: | 2012–11 |
| URL: | http://d.repec.org/n?u=RePEc:pav:demwpp:020&r=for |
| By: | Alexios Ghalanos (Faculty of Finance, Cass Business School); Eduardo Rossi (Department of Economics and Management, University of Pavia); Giovanni Urga (Faculty of Finance, Cass Business School and University of Bergamo) |
| Abstract: | In this paper, we propose a novel Independent Factor Autoregressive Conditional Density (IFACD) model able to generate time-varying higher moments using an independent factor setup. Our proposed framework incorporates dynamic estimation of higher comovements and feasible portfolio representation within a non elliptical multivariate distribution. We report an empirical application, using returns data from 14 MSCI equity index iShares for the period 1996 to 2011, and we show that the IFACD model provides superior VaR forecasts and portfolio allocations with respect to the CHICAGO and DCC models. |
| Keywords: | Independent Factor Model, GO-GARCH, Independent Component Analysis, Timevarying Co-moments |
| JEL: | C13 C16 C32 G11 |
| Date: | 2012–11 |
| URL: | http://d.repec.org/n?u=RePEc:pav:demwpp:021&r=for |
| By: | Alberto Locarno (Bank of Italy) |
| Abstract: | The applied literature on adaptive learning has mostly focused on small, linear models, with homogenous expectations. In non-linear models heterogeneous expectations prevail and the process through which agents select (and change) a forecasting model becomes a necessary ingredient of the analysis; moreover, the temporary equilibrium of the learning process approaches an asymptotic limit that may be affected by the communication strategies of the monetary policymaker. The objective of this paper is to assess whether in such a model economy the optimal monetary policy exhibits properties that are similar to those found in the literature for small, linear models. The main results are the following: (1) expectations heterogeneity is an intrinsic feature of the economy: no PLM succeeds in ruling out all the other forecasting models; (2) contrary to previous findings, the monetary policymaker has no incentive to adopt highly inflation-averse policies: too strong a reaction to price shocks increases both inflation and output volatility; (3) partial transparency seems to enhance somewhat welfare (but fully transparent policies do not); (4) a higher degree of transparency calls for stronger inflation aversion. |
| Keywords: | Bounded rationality, generalised stochastic gradient learning, transparency. |
| JEL: | E52 E31 D84 |
| Date: | 2012–10 |
| URL: | http://d.repec.org/n?u=RePEc:bdi:wptemi:td_888_12&r=for |
| By: | Søren Johansen (University of Copenhagen and CREATES); Marco Riani (Dipartimento di Economia, Università di Parma); Anthony C. Atkinson (Department of Statistics, London School of Economics) |
| Abstract: | We develop a $C_{p}$ statistic for the selection of regression models with stationary and nonstationary ARIMA error term. We derive the asymptotic theory of the maximum likelihood estimators and show they are consistent and asymptotically Gaussian. We also prove that the distribution of the sum of squares of one step ahead standardized prediction errors, when the parameters are estimated, differs from the chi-squared distribution by a term which tends to infinity at a lower rate than $\chi _{n}^{2}$. We further prove that, in the prediction error decomposition, the term involving the sum of the variance of one step ahead standardized prediction errors is convergent. Finally, we provide a small simulation study. Empirical comparisons of a consistent version of our $C_{p}$ statistic with BIC and a generalized RIC show that our statistic has superior performance, particularly for small signal to noise ratios. A new plot of our time series $C_{p}$ statistic is highly informative about the choice of model. On the way we introduce a new version of AIC for regression models, show that it estimates a Kullback-Leibler distance and provide a version for small samples that is bias corrected. We highlight the connections with standard Mallows $C_{p}$. |
| Keywords: | AIC, ARMA models, bias correction, BIC, $C_{p}$ plot, generalized RIC, Kalman filter, Kullback-Leibler distance, state-space formulation |
| JEL: | C22 |
| Date: | 2012–11–08 |
| URL: | http://d.repec.org/n?u=RePEc:aah:create:2012-46&r=for |
| By: | Timothy J. Garrett |
| Abstract: | For those concerned with the long-term value of their accounts, it can be a challenge to plan in the present for inflation-adjusted economic growth over coming decades. Here, I argue that there exists an economic constant that carries through time, and that this can help us to anticipate the more distant future: global economic wealth has a fixed link to civilization's overall rate of energy consumption from all sources; the ratio of these two quantities has not changed over the past 40 years that statistics are available. Power production and wealth rise equally quickly because civilization, like any other system in the universe, must consume and dissipate its energy reserves in order to sustain its current size. One perspective might be that financial wealth must ultimately collapse as we deplete our energy reserves. However, we can also expect that highly aggregated quantities like global wealth have inertia, and that growth rates must persist. Exceptionally rapid innovation in the two decades following 1950 allowed for unprecedented acceleration of inflation-adjusted rates of return. But today, real innovation rates are more stagnant. This means that, over the coming decade or so, global GDP and wealth should rise fairly steadily at an inflation-adjusted rate of about 2.2% per year. |
| Date: | 2012–11 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:1211.3102&r=for |