Skip to main content
European Commission logo print header

New Methods For Forecast Evaluation

Article Category

Article available in the following languages:

The science of forecasting reaches new heights

Many forecasting methods for assessing the economy and evaluating its robustness have fallen short of expectations. A new family of forecasting tools has uncovered these shortcomings and improved the science of forecasting.

Industrial Technologies icon Industrial Technologies

Many disciplines and sciences – from business to economics – rely on forecasting as a pivotal tool to support decision-making and policymaking. The EU-funded NEW FORECAST METHODS (New methods for forecast evaluation) looked at how to improve forecasting by developing new tools and methods. It achieved this through five different pillars of research or subprojects, yielding much new information and a number of new tools on the topic. Through its first sub-project called ‘Out-of-sample forecast tests robust to the choice of window size’ the team articulated a novel method for assessing the forecasting performance of economic models. Taking the choice of the estimation window size in forecasting as the main issue to address, it proposed new testing procedures to evaluate forecasting performance in different instances. Another sub-project called ‘Conditional predictive density evaluation in the presence of instabilities’ helped the project to establish new tests to evaluate the correct specification of predictive densities in the presence of instabilities. It showed how these new tests can effectively detect misspecification in the predictive distribution, pointing out where other tests can fail. The third sub-project was about ‘Evaluating predictive densities of US output growth and inflation in a large macroeconomic data set’. Here the team looked at different models to assess the limitations of each and propose more effective methods to achieve its aims. With the fourth sub-project, the team developed ‘Forecast rationality tests in the presence of instabilities, with applications to federal reserve and survey forecasts’. These tests identified shortcomings in Federal Reserve Greenbook forecasts and many survey-based forecasts. Lastly, ‘Alternative tests for correct specification of conditional forecast densities’ were developed under the fifth subproject to evaluate the correct specification of density forecasts, successfully detecting misspecifications. Attesting to the project’s success, the European Central Bank has shown interest in adding some of these new tests to their forecast evaluation procedures. The project also led to the launch of workshops that in Barcelona, Spain, on the topic, solidifying the science of forecasting and creating strong networks in the field. The work carried out during this project has shown admirable progress in improving forecasting, a development which is bound to have a positive impact on economic development.

Keywords

Forecasting, forecast methods, economics, forecast evaluations, misspecifications

Discover other articles in the same domain of application