Response of the Bank of England to the Bernanke review of forecasting for monetary policy making and communication at the Bank of England

Forecasting and related processes during times of significant uncertainty

The global economy has experienced an unprecedented series of shocks in recent years, including substantial supply disruptions, resulting in high inflation and elevated uncertainty.

In an environment where “shocks have been of a different nature, and their impact has been more uncertain”^{footnote [1]} forecasting naturally became more difficult. The Review provides a thorough evaluation of how the Bank’s forecasting framework has performed during a period of large and unprecedented shocks.^{footnote [2]} It notes that the Monetary Policy Committee’s (MPC's) forecasting performance has been similar to that of other central banks and UK private-sector forecasters.

Even so, as inflation falls towards the MPC’s target, it is appropriate to reflect – and act – on the lessons of recent years. The Review presents a unique and valuable opportunity to develop and strengthen the Bank’s processes in support of the MPC’s forward-looking approach to the formulation and communication of monetary policy.^{footnote [3]} Designing and delivering change will require detailed planning and careful implementation.

Two broad themes underpin many of the Review’s assessments and recommendations and are likely to guide the design of future changes:

• Substantial investment is needed, beyond that already underway, to develop key parts of the data, modelling, forecasting and evaluation infrastructure and the expert staff to support them.
• Within an overall approach that assigns more prominence to risks and welcomes challenge to underlying assumptions, the role of the central projection and the MPC’s discussions surrounding it should be reconsidered.

Infrastructure investment

As noted in the Review, the Bank has already embarked on considerable investment in infrastructure.^{footnote [4]} In early 2023 the Bank began a Future Platform programme to manage the move to a new product for accessing, analysing and visualising data. This is part of a wider data portfolio that will provide the Bank with a new environment for working with data on the cloud and will modernise the production of the monetary and financial statistics, which are themselves an important input into the monetary policy process. This will not only address limitations of the existing infrastructure but also enable staff more easily to automate processes, as well as to work with the increasingly large and complex data sets that are now commonly utilised by the central banking community. These improvements will make routine tasks more efficient thereby freeing up staff time for more productive activities. As this initiative predated the Review, the Bank will explore the extent to which further expansion of these activities is required.

The unusual shocks over the recent period have also revealed some challenges to the Bank’s use of forecasting models. The Bank’s approach has long recognised the value of using a suite of models so that insights can be drawn from a range of alternative approaches. A range of models is used to assess the implications of different shocks and to inform the application of judgement to forecasts. However, through the recent period of unprecedented shocks, it has been particularly difficult to incorporate systematically and coherently these assessments into a central projection. In particular, when constructing the central projection it has been challenging to achieve a good balance between model-based insights and human judgements.

Ensuring that the Bank’s modelling toolkit and forecast approach – and the MPC discussions that they support – are robust to the future economy will require a reconsideration of how models and analytical work are used. There are several possible options for a core modelling framework, as the Review notes,^{footnote [5]} as well as possibilities for how other analytical work is incorporated. The appropriate choice of core modelling framework will require careful consideration, including an assessment of whether any of several models currently under development at the Bank would be suitable (including a redevelopment of COMPASS^{footnote [6]} and a new semi-structural model). An equally important consideration will be the design of a strategy that allows staff to draw insights from a suite of models and analytical work as efficiently and flexibly as possible. More broadly, the Bank recognises that continual improvement of tools and models in light of experience should be a priority.

The role of the central projection

The second key theme of the Review is that the role of the central projection within the processes for discussing and communicating monetary policy should be reconsidered. A forecasting framework that performed well during periods in which inflationary shocks were relatively small may be less robust during periods of large and unusual shocks and the associated heightened uncertainty, or in an economy transformed by structural changes, for example the emergence of artificial intelligence, or a realignment of trading relations.

Forecasts have played a central role in the MPC's policy strategy and communications since its inception in 1997. As noted in the Review, the lags between monetary policy decisions and their effects on the economy mean that it is the prospects for inflation, rather than its current rate, that are most relevant to the MPC. Monetary policy actions have larger effects on inflation over the subsequent one or two years than in the very near term.

In more stable periods, when shocks are small and the behaviour of the economy is well understood, these insights can be used to formulate a simple monetary policy strategy in which the policy rate is adjusted up or down when the medium-term inflation forecast is expected to be above or below target (Haldane (1998)). An important property of such a strategy is that its formulation within the central bank is mirrored in the public communication of its policy decisions. The central bank's inflation forecast is a powerful vehicle for rationalising and explaining policy decisions: they are a "joint product".^{footnote [7]} This property is a strength of such an approach, when it is working well, because it simplifies communications by allowing them to be presented within an internally consistent framework.

The Bank's forecast and monetary policy making processes have been designed to support this strategy: the forecast meetings and the processes that underpin them play a central role in supporting the MPC's policy discussions during Monetary Policy Report rounds.^{footnote [8]} Despite this prominent role in the process, forecasts are ultimately a means to an end: the formulation and communication of the appropriate monetary policy to achieve the inflation target (King (2002)).

Indeed, in many cases a single forecast is not sufficient to provide a mechanical link to the appropriate policy. For example, the future paths of a number of variables (including Bank Rate) are ‘conditioned’ on particular conventions that, on average, are likely to be plausible forecasts while also avoiding the risk of being misinterpreted as an endorsement by the MPC. However, they will not at every moment in time represent the paths for those variables that members of the MPC judge to be most likely. Moreover, it may be appropriate for monetary policy to respond to risks that are not captured by a single forecast. In practice, therefore, the Bank’s approach includes several elements of flexibility in the connections between the forecast, the policy decision and their communication. These include the regular production of forecasts under alternative paths for Bank Rate (based on market expectations and a constant rate assumption), the use of fan charts to communicate the balance of risks and occasional adjustments in the assumptions underlying the conditioning paths.^{footnote [9]}

For much of the period since the creation of the MPC, when inflationary shocks were relatively small, this approach worked well. Inflation was relatively stable and close to target (Carney (2020)), inflation forecast errors were small and similar to peer central banks^{footnote [10]} and the framework was a well-understood means of communication with external audiences. Over recent years, however, large and unprecedented global shocks have made forecasting more difficult and revealed some constraints within the Bank's existing approach,^{footnote [11]} and the MPC needed to use all aspects of its "limited discretion" in the construction and communication of its forecasts.^{footnote [12]}

An approach that places relatively less weight on a central projection may be more robust in such an environment in the future. This could also offer an opportunity to systematise the application of insights from a wide range of staff analysis and research to the monetary policy making process. The Review offers a number of options for adjusting the relative focus on a central projection in internal processes and external communications, including the use of scenarios, consideration of alternative conditioning assumptions and adjustments to the mix between quantitative and qualitative information in communications.

Design considerations

The Bank will consider the implications of the recommendations carefully since the consequences could affect many aspects of the monetary policy making process including infrastructure, models, staffing, the internal processes for production and discussion of the forecast and policy strategy, and the number and nature of external communications.

For example, as noted in the Review, the production of scenarios alongside a central projection could have several distinct benefits including the ability to consider the key risks to the macroeconomic outlook and communication of the monetary policy reaction function.^{footnote [13]} The wide-ranging nature of these benefits also implies that a greater role for scenarios could have large effects on internal processes, model development activities and communication. Some of these connections are described below.

Scenarios have the potential to offer a better means of articulating the balance of risks around the outlook. To implement scenarios the Bank will need to ensure that the forecasting system is sufficiently flexible and efficient to support the production of multiple alternative views of the outlook. In addition to updates to infrastructure this will require adjustments to the staffing model including the relevant incentives and development opportunities for staff. Another important design consideration will be the process by which the most policy relevant risks are selected: identifying an appropriate number and type and when a scenario should be published externally. Internal processes will also need to provide sufficient time within MPC discussions to consider scenarios and their policy implications, potentially reducing the time available to formulate and discuss the central projection. Finally, communications will need to adapt to ensure that the roles of the central forecast and of different scenarios in explaining the policy outlook are clear.

In principle, fan charts can be used to communicate prevailing risks in the context of past forecast errors and emphasise that those risks may span a wide range of possible outcomes. However, in practice fan charts have not proved to be an effective way of communicating with the public about risks to the macroeconomic outlook.^{footnote [14]} The extent to which scenarios can serve as a sufficient substitute for fan charts in the communication of risks will require careful consideration, as will whether there are alternative communication strategies that convey this information, as noted in the Review.

An approach that draws more on scenarios also has the potential to articulate better differences of view among Committee members. This could relieve some of the challenges in agreeing a ‘best collective judgement’ forecast.^{footnote [15]} However, internal processes will need to adapt to ensure that significant differences of view are well captured by scenarios, and that the process continues to support the development of a clear collective narrative about the most important factors influencing monetary policy.

These types of wider-ranging effects will similarly apply to other recommendations in the Review.

Acting on the recommendations therefore requires the Bank to consider substantial changes to key components of the existing forecasting and policy processes. Some of those changes will systematise and prioritise the way that existing types of research and analysis are used to inform the MPC’s approach to setting and communicating monetary policy. Others will require more fundamental changes. Careful thought will be needed to ensure that an updated approach builds on existing strengths. For example, it will be important to maintain continuity in several key elements, including the focus on transparency and accountability, and the forward-looking approach to monetary policy.

Next steps

The intertwined nature of many of the Review’s recommendations means that it will take some time to develop detailed plans as well as to manage their implementation. The Bank will initiate a substantial upgrade programme, the complete scope of which is to be determined, but that is likely to include at least the following workstreams:

• Continuing and extending existing infrastructure and modelling investment projects essential to provide the foundations for a future approach.
• Embedding a state-of-the-art approach to forecast error evaluation and a process to ensure that the results of such evaluations lead to improvements in models and techniques.
• Designing and implementing a more efficient forecasting approach – including the role of forecasts in MPC discussions and decision making – that reconsiders the role of the central projection, augmenting it with scenarios.
• Enhancing the assimilation of alternative approaches, staff research, Bank surveys and cross checks using internal and external analysis.
• Evaluating communications products, ensuring that they are focused on the key determinants of the monetary policy outlook.
• Developing incentives for expert staff to lead model development and infrastructure investment activities and providing opportunities to progress through building expertise.

These workstreams will build on further investigation of the lessons from recent years, some of which is already underway. For example, staff work is exploring the sources and nature of recent forecast errors using methods that help to identify parts of the forecasting approach that could particularly benefit from further development and improvement.^{footnote [16]}

As noted in the Review, a phased approach to implementing changes is likely to be appropriate for a programme of this scale.^{footnote [17]} The Bank therefore intends to introduce discrete ‘packages’ of changes over time, as the relevant stages of the workstreams are completed.

The MPC welcomes Dr Bernanke’s review and endorses the Bank’s commitment to action as outlined in the workstreams above. It will remain involved closely throughout the design and implementation phases.

The Court of the Bank, which commissioned the Review, also supports the Bank’s commitment to action on the recommendations and will continue to ensure that resources are in place to allow for the effective and efficient discharge of the Bank’s monetary policy function, in line with the Bank of England Act (1998).

The Bank is committed to reporting on progress in a transparent way. It will update the Court on progress against the workstreams above at appropriate intervals, consistent with the Court’s oversight function, and will provide an update externally by the end of the year.