Journal of Forecasting最新文献

The subset selection of individual prediction methods is gradually becoming a hot topic. Among numerous forecasts, identifying the optimal subset approach has become a major focal point of research. To address this issue, the paper introduces a novel method based on information theory, which is called common mutual information (CMI) selection algorithm. This optimal subset selection method not only simultaneously considers the relationships of three factors, which include the candidate feature set, the selected feature set, and the actual time series, but also provides a more precise treatment of these relationships. Therefore, CMI algorithm employs the mutual information (MI) shared among the three factors as the criterion for selection and improves the accuracy of the redundancy or correlation measure for existing algorithms. Furthermore, it overcomes the deficiency of calculating MI between the candidate subset and the actual time series. Existing algorithms use the average MI values between individual elements within the subset and the actual sequence; this paper takes the selected subset as a multidimensional input for MI computation, thus reducing computational errors. Finally, the proposed algorithm is compared with two other approaches of the MI algorithm, the Max-Relevance and Min-Redundancy (mRMR) algorithm in both theoretical and empirical aspects. The experiments are illustrated to show the effectiveness and superiority of CMI algorithm.

This paper evaluates the roles of jump and sign-asymmetry spillovers in forecasting the realized volatility in a large sample of 20 stock markets. We compare for the first time whether controlling for either the jumps or asymmetric spillovers into the heterogeneous autoregressive–realized volatility (HAR-RV) model improves the forecasts over 1, 5 and 22 days. Before doing so, the spillovers predictors are generated. In analyzing the spillover process, we find that the US stock market remains the main net transmitter of shocks, and while China is relatively detached from the spillover linkages, such effects may be transmitted through Hong Kong, which is a significant receiver of shocks. The out-of-sample results reveal that the incorporation of jump spillovers improves forecast performance the most across a range of measures. This is more clearly demonstrated at the 22-day forecasting horizon more notably in Europe, France, Germany, India, and the United Kingdom. Lastly, irrespective of the forecasting horizon, performing the predicting stability test uncovers significant improvements in the jump spillover–based model during periods of notable market stress such as the 2014–2016 oil price crash and COVID-19. Overall, results suggest paying more attention to jump spillover while constructing international portfolios based on the realized volatility.

Both the Federal Reserve (Fed) and the European Central Bank (ECB) have been criticized for not having perceived that the outbreak of Covid at the beginning of 2020 would lead to a structural change in inflation in the early 2020s. Both central banks viewed the initial inflation surge in 2021 as temporary and delayed monetary tightening until 2022. We argue that the existing literature on structural breaks could not have been useful to policymakers because it identifies the breaks in an arbitrary way. The tests used to identify breaks do not incorporate prior knowledge that a break may have occurred so that the tests have very little power to detect a break that occurs at the end of the sample. We show that, in the event of a major shock, such as Covid, using knowledge that a break may have occurred and testing for a break in a recursive way as new data become available could have alerted policymakers to the break in inflation. We conduct Monte Carlo simulations suggesting that our method would have identified that a break had occurred in inflation by the end of 2020, well before policymakers had perceived the break.

We investigate mean squared forecast error (MSE) accuracy improvements for linear–convex combination forecasts, whose components are pretreated by a postprocessing procedure called “vector autoregressive forecast error modeling” (VAFEM). Assuming that the forecast error series of the individual forecasts are governed by a stable VAR process under classic conditions, we obtain the following results: (i) VAFEM treatment bias corrects all individual and linear–convex combination forecasts. (ii) Any VAFEM-treated combination has a smaller theoretical MSE than its untreated analog, if the VAR parameters are known. (iii) In empirical applications, VAFEM gains depend on (1) in-sample sizes, (2) out-of-sample forecast horizons, and (3) the biasedness of the untreated forecast combination. We demonstrate the VAFEM capacity in simulations and for realized-volatility forecasting, using S&P 500 data.

Motivated by the increasing role of trade in global economic developments, this paper derives novel econometric methods to forecast global trade by exploiting the relationship between economic activity and trade itself. We empirically document that the relation between trade and economic activity changes along the business cycle—the stronger the cycle, the larger their elasticity. Consistently with theoretical predictions, such cyclicality depends on two key factors: (i) the high pro-cyclicalilty of the demand for intensively traded items and (ii) the presence of low-frequency (“trend”) components in trade and GDP series. We show that the latter is key to generate a cyclical income elasticity of trade and that a linear relationship holds once those components are filtered out. These empirical findings are exploited in two original empirical approaches to map GDP forecasts, for which rather accurate and timely projections are available, into world trade forecast. In an out-of-sample real-time forecasting exercise, with both the proposed methods, we obtain predictions that are vividly more accurate than naive linear models and nearly halve the forecast error of the IMF-WEO.

This paper evaluates the informative value of the ECB inflation forecasts vis-à-vis other institutional and model-based forecasts in the euro area using ex post optimal combinations of forecasts and nonnegative weights. From a methodological perspective, we adapt the corresponding forecast encompassing test to the constrained parameter space, showcasing its superior performance over traditional encompassing tests in both size and power properties. Empirically, the combining weights and the forecast encompassing test reveal that the ECB was the most informative forecaster of euro area inflation over the 2009–2021 period. This changed in 2022: The ECB lost its position as the most informative forecaster, and when using rolling windows to estimate the combining weights using a rolling window, we find an important decline in the ECB's weight over time. This time dependency can be associated with the economic environment and, in particular, the level of uncertainty, the monetary policy, and the macro-financial conditions in which the ECB operates.