New Statistical Method Helps Policymakers Adjust Economic Forecasts More Reliably

Researchers from the European Central Bank, European Stability Mechanism, and Universität Bonn propose a new forecasting method called parametric tilting that helps economists incorporate new information into economic forecasts more reliably. The technique improves on traditional methods by producing stable, realistic probability distributions, especially when new data differs sharply from existing model predictions.

Economic forecasts guide decisions at central banks, governments, and international institutions. Yet these forecasts are often built on historical data, which means they may react slowly when sudden shocks hit the economy. During crises such as financial crashes or pandemics, policymakers frequently have new information, such as expert surveys, real-time indicators, or policy insights, that suggests the models are already outdated.

A recent study by researchers from the European Central Bank, the European Stability Mechanism, and Universität Bonn proposes a new way to address this problem. The economists, Carlos Montes-Galdón, Joan Paredes, and Elias Wolf, have developed a statistical technique that helps policymakers adjust economic forecasts more realistically when new information appears.

Their work focuses on improving how economists update probabilistic forecasts, which show not just one prediction but a range of possible outcomes and their likelihood.

The Limits of Traditional Forecast Adjustments

Modern economic models often generate thousands of simulated scenarios about the future. Together, these scenarios form what economists call a density forecast, a probability distribution that describes how likely different economic outcomes are.

However, when new information becomes available, such as survey expectations about growth or inflation, economists need a way to incorporate it into these forecasts. For years, a technique known as entropic tilting has been used for this purpose.

This method works by adjusting the probability weights assigned to each simulated outcome. If outside information suggests weaker growth than the model predicts, the algorithm shifts probability toward lower growth outcomes while keeping the forecast close to the original model results.

In theory, the method allows economists to blend model results with new insights. In practice, it often runs into trouble when the new information differs sharply from the model’s expectations.

Why Forecasts Can Become Unstable

The biggest challenge with the traditional approach is that it relies entirely on the model’s original simulations. If the model never considered certain outcomes, the method cannot easily assign probability to them.

When policymakers try to push the forecast far away from the model’s predictions, the algorithm may end up giving almost all the probability weight to just a few simulated observations. This creates unstable forecasts and can produce strange probability shapes.

In some cases, the resulting distribution may even show multiple peaks or unrealistic patterns. These shapes are not driven by economic logic but by the mechanics of the statistical method.

For policymakers who rely on forecasts to evaluate risks and communicate economic outlooks, such distortions can make the results harder to interpret.

A New Method: Parametric Tilting

To overcome these issues, the researchers propose a new approach called parametric tilting. Instead of simply reweighting the model’s simulated outcomes, the method searches for a new probability distribution that closely resembles the original forecast while also reflecting the new information.

The technique relies on a flexible statistical model called the skew-t distribution. Unlike the familiar bell-shaped curve, this distribution can tilt to one side and allow for heavier tails. This makes it well-suited for economic data, which often shows asymmetry and extreme events.

The distribution is defined by four parameters that control its location, spread, skewness, and tail thickness. By adjusting these parameters, economists can shift the forecast toward the desired outcome while keeping the overall distribution smooth and realistic.

The process involves finding the set of parameters that best matches the original forecast while satisfying the conditions imposed by policymakers, such as a new expected growth rate or level of uncertainty.

Lessons from the COVID-19 Crisis

The researchers demonstrate the value of their approach using a real-world example from the early months of the COVID-19 pandemic.

At the start of 2020, many economic forecasting models had not yet captured the severity of the coming downturn. At the same time, participants in the Survey of Professional Forecasters were already predicting a sharp decline in economic activity.

When economists tried to incorporate these survey expectations using traditional tilting methods, the forecasts often became unstable or unrealistic. The new parametric tilting method, however, was able to integrate the survey information smoothly.

The resulting forecasts reflected stronger downside risks without producing strange or distorted probability shapes.

A More Reliable Tool for Policymakers

The researchers argue that parametric tilting could become an important addition to the forecasting toolkit used by central banks and policymakers.

By allowing forecasts to incorporate external insights without destabilizing the underlying distribution, the method helps economists adapt more quickly to changing conditions. It also produces clearer probability distributions, making it easier for policymakers to assess risks and communicate economic outlooks.

As global economies face increasingly unpredictable shocks, from pandemics to geopolitical tensions, the ability to adjust forecasts quickly and reliably is becoming more important than ever. The new approach offers a promising step toward making economic forecasting more flexible, transparent, and resilient.