**5. Conclusions**

One of the most widespread strategies for combining individual forecasts is to take a simple average of the forecasts. Empirically, many studies have shown that the mean outperforms complex combining strategies. Theoretically, the use of the simple arithmetic mean could be justified when all the forecasters have shown the same forecasting ability or when the available information about their ability seems to be not enough to calibrate the forecasters di fferently. This paper proposes the use of an entropy-based technique estimator to obtain an a ffine transformation of the equal weighted forecast combination by using the small available information, a data-weighted prior (DWP) estimator.

We tested the validity of the proposed model by a simulation exercise and compared its ex-ante forecasting performance with other combining methods. The benchmarks for comparing the competing method were the arithmetic mean of the forecasters, a restricted least squares, and weight scheme forecasts based on Bayesian model averaging (where the weights are determined on the basis of the Bayesian information criterion).

We set three di fferent values for the number of individual forecasts to be combined (6, 12, and 24) and we have divided our set of forecasters in two di fferent subsets, which can be classified as "good" or "bad" predictors. The obtained results of the simulation indicate that the proposed DWP estimator seems to beat the competing combination techniques, given that it takes the weighting scheme as the arithmetic mean and only departs from these weights if the sample contains information providing strong enough empirical evidence to weigh di fferently than equal. The most relevant advantage of this estimator is that, even in situations characterized by a large number of forecasters, the DWP estimator generates a better set of recovered forecasters´ weights than the arithmetic mean which is capable to identify groups of forecasters into groups of "good" and "bad" forecasts. Additionally, the empirical application could be extended by comparing the forecasting performance of the proposed method with other combining methods based on an information-theoretic approach [6].

**Author Contributions:** Conceptualization, E.F.-V., B.M. and G.J.D.H.; Methodology, E.F.-V.; Validation, E.F.-V. and B.M.; Formal Analysis, E.F.-V.; Resources, B.M.; Writing-Original Draft Preparation, E.F.-V., B.M. and G.J.D.H.; Writing-Review & Editing, E.F.-V., B.M. and G.J.D.H.; Funding Acquisition, E.F.-V. and B.M.

**Funding:** This research was partially funded by the research project "Integrative mechanisms for addressing spatial justice and territorial inequalities in "Europe (IMAJINE)" in the EU Research Framework Programme H2020.

**Acknowledgments:** The authors acknowledge the support of the gues<sup>t</sup> editors of this special issues and the comments received by two anonymous reviewers.

**Conflicts of Interest:** The authors declare no conflict of interest.
