New research suggests that forecasting models which adapt as economic relationships change could improve predictions for inflation and interest rates, particularly during periods of disruption.
Forecasts produced by the Bank of England carry considerable weight. They influence monetary policy, shape market expectations and inform decisions made by businesses, households and government.
However, an analysis by Costas Milas and Georgios Papapanagiotou argues that the Bank may be able to improve forecast performance by making greater use of models with time-varying parameters.
Unlike models that assume relationships between economic variables remain broadly stable, time-varying models allow those relationships to change as conditions evolve. This can be particularly valuable when an economy is exposed to repeated shocks or structural breaks.
Responding to a changing economic environment
The UK economy has faced a succession of major disruptions over the past two decades, including the global financial crisis, Brexit, the COVID-19 pandemic, energy price shocks and wider geopolitical and trade uncertainty.
These events have affected inflation, growth, employment and interest rates in different ways. They have also changed how quickly shocks move through the economy and how businesses and consumers respond.
For Operational Research practitioners, this raises a familiar modelling challenge: how can a forecasting system remain reliable when the environment it represents is continually changing?
Milas and Papapanagiotou’s approach uses time-varying parameters to allow the effects of economic shocks to shift over time. Their model combines domestic and international indicators, including measures of geopolitical risk, financial stress, wage pressure and economic and trade uncertainty.
It also incorporates Divisia M4, a measure of liquidity that assigns different weights to forms of money according to how readily they are likely to be spent.
Comparing forecast performance
According to the authors, their probabilistic forecasts generally outperform those published by the Bank of England for Consumer Prices Index inflation.
They report that the model more effectively captures several major changes in inflation, including the sharp decline during the 2008–09 financial crisis, the increase in 2022 and the subsequent movement in late 2024.
The model also performs strongly when forecasting Bank Rate, particularly from the middle of 2024 onwards. However, the authors acknowledge that it did not fully anticipate the scale of the interest-rate reductions made during 2009.
The findings support the broader principle that forecasting models should be tested not only for accuracy under normal conditions, but also for their ability to adapt when economic structures and behavioural relationships change.
Communicating uncertainty
The research also highlights the importance of communicating uncertainty clearly.
The Bank of England’s forecasting framework does not rely on a single model. It uses a combination of structural, semi-structural and statistical approaches to inform its projections and policy decisions.
However, Milas and Papapanagiotou argue that published scenario analysis does not always provide enough information about the likelihood of different outcomes.
Scenario planning can help decision-makers understand a range of possible futures, but scenarios without probabilities may leave businesses, investors and the public uncertain about which outcome is considered most plausible.
This is an important issue for Operational Research. Effective decision support requires more than presenting possible outcomes. It also involves communicating risk, uncertainty and confidence in a way that helps people compare options and act.
Building models that remain useful
The study does not suggest that the Bank of England should replace its existing forecasting framework with a single alternative model.
Instead, it makes the case for incorporating more adaptive approaches into a wider modelling system. Models that reflect structural change may be better equipped to support decisions in volatile and uncertain conditions.
The wider lesson extends beyond economic forecasting. In any complex system, models can quickly lose value when they rely on relationships that no longer hold.
For analysts and decision-makers, the challenge is therefore not simply to predict the next result. It is to build forecasting systems that remain robust, transparent and useful as the world around them changes.
References
https://blogs.lse.ac.uk/businessreview/2026/06/18/how-to-improve-the-bank-of-englands-forecasts/
https://www.bankofengland.co.uk/working-paper/2017/a-time-varying-parameter-structural-model-of-the-uk-economy