The central banking community has made significant steps to improve its communication strategy since the days of myth and mystique criticised by Alan Blinder in 1998:
“Greater openness is not a popular case in central banking circles, where mystery is sometimes argued to be essential to effective monetary policy…[but] a more open central bank, by contrast, naturally conditions expectations by providing the markets with information about its own view of the fundamental forces guiding monetary policy.”
The challenges posed by the financial crisis and specifically from the need to manage the exit from an ultra-loose monetary stance have led to further improvements in the way central banks speak to the market and the wider economy. Forward guidance about the future path of policy has become a staple of central bank communication, although the world's major central banks are yet to adopt the fuller disclosure model pioneered by the Reserve Bank of New Zealand (and subsequently adopted in Scandinavia). In this column we make the case for completing the forward guidance provided by those major central banks. We believe central banks should describe how they expect policy to evolve in the form of probabilistic statements which reflect their understanding of the scale of uncertainty about the future, and correspondingly present density functions that encompass their likely responses.
Back to basics: Distributions matter
The idea that distributions should take centre stage in the internal policy debate is not controversial – and so neither should be our suggestion that central banks publish probabilistic statements about the current and future stance of policy. There is a consensus in the academic literature that the stance of monetary policy should reflect the range of possible outcomes, not just the most likely outcome (the mode). Under certain strong assumptions it is proper for the central bank to act as if it is certain about the future – that the economy will follow the path implied by the mean (Tinbergen 1952, Theil 1958). But even to calculate the mean we need to know the entire distribution, and richer features of the distribution – the spread and the skew – start to matter too. In short, central banks need to examine distributions of possible outcomes when setting policy, and for a given view about the optimal strategic response to particular circumstances that implies a distribution of outcomes for the policy rate. That internal analysis allows policymakers to reach a conclusion on how the optimal policy stance is most likely to evolve; it also quantifies the uncertainty around that path. By focusing on the range of possible outcomes policymakers are better placed to design a state contingent strategy – how policy should respond in off-central case scenarios – which could feed back into the calibration of the current stance. The question is the extent to which this internal debate should be open to scrutiny.
Mission possible: Forecasting distributions
Seasoned forecasters may have a reasonable idea about the most likely conditional path for any particular macroeconomic variable. What is required is a model of how key macroeconomic variables behave jointly, and forecasts for the most likely outcome for the exogenous inputs to that model that can then be combined with expert judgement on where the economy is heading. Forecasting distributions – constructing density functions for the future value of macroeconomic variables – is a more complex task because we have to take our uncertainty and ignorance about how the economy behaves more seriously.
The process of building coherent forecasts of the range of outcomes involves combining outputs from a wide variety of sources and methods. Analysis of the data and forecast errors provides a crude guide to the potential variance of individual series and the scale of uncertainty around projections. Theoretical models indicate the likelihood of different scenarios – the joint occurrence of particular outcomes – given repeated simulations based on assumptions about the range of possible outcomes for those exogenous inputs. One would have to average across a diverse suite of theoretical models, covering different features of the economy (and diverse calibrations of each particular model), to capture our uncertainty about how the economy behaves. Likewise, empirical models can generate forecasts for joint distributions (Wallis 2005).
The output of these models can also be supplemented with information extracted from options markets, where prices reveal the views of market participants about the distribution of possible outcomes. Finally, the producers (central bank staff) and consumers (the policymakers) can apply judgement to these distributions to address the limitations of the forecast process – instances where the past is unlikely to provide a reliable guide to the future. For example, one might reasonably expect to observe a structural break in the dynamics of the financial cycle after the crisis, given that the rule book of microprudential regulation has been re-written and a new macroprudential policy regime has been created to dampen the financial cycle. That will have implications for the future dynamics of variables that tend to dominate the monetary policy debate, as well as the transmission mechanism of monetary policy.
One critical point is that the information used to calibrate these distributional forecasts implicitly or explicitly assume that the stance of monetary policy will respond to changing economic circumstances, which should stabilise the system to some extent. The data used to train theoretical and empirical models will incorporate previous monetary policy responses and as long as market participants anticipate a policy response in the future, so too will financial market prices. It is actually harder to construct distributional forecasts of what could happen to output and inflation in the future on the extreme assumption that policy will not respond, given the lack of theory and data on what would happen to the economy when the nominal anchor is removed.
Of course, the implicit policy response embedded in these distributional forecasts of output and inflation may not accord with policymakers’ understanding of their loss function – and hence their views on optimal policy. That is the ultimate prize of this work agenda, as it forces policymakers to take a stand on monetary strategy so they can make internally consistent statements about the range of outcomes for the macroeconomy and the policy stance.
Revealed preference and the costs and benefits of clarity
For a central bank trying to implement optimal policy the question is simply whether to publish this information and explicitly communicate how the stance of policy might evolve in the future. There are three standard arguments against publishing this information, all of which relate to the risk that private sector agents will misinterpret a probabilistic statement about the future path of rates as a promise to deliver the most likely path.
- Agents may take rash decisions leading to financial instability at the micro and macro level because they will not understand the scale of rate uncertainty.
- The central bank's reputation (and hence its capacity to perform its core functions) may be damaged when rates deviate from the perceived 'promise'.
- Agents will herd on the signal in an environment in which higher order beliefs matter (Morris and Shin 2000).
These concerns apply to all forms of central bank communication about the future path of policy. By revealed preference, central bankers do not find these arguments overpowering given their increased use of forward guidance, including narrow ranges for the level of the neutral rate and point estimates for the future path of rates. It therefore follows that complete forward guidance – nesting statements about the future path of policy within a rigorous depiction of the uncertainty about that path – should be less vulnerable to these concerns. Emphasising uncertainty should discourage agents from treating the modal optimal path as a promise. Better still, central banks could encourage the revelation of information from the private sector by explicitly saying that published forecasts do not have a clear understanding of the risks to Factor X or Y. Unsubtle hints, narrow interval estimates, and certainly point forecasts are far more prone to misunderstanding and herding.
If we can dismiss the costs of complete forward guidance, what are the benefits? Communicating the extent and sources of uncertainty around the future path of policy should help to remove the sources of volatility in asset prices in markets – and behaviour in the real economy – that reflect Knightian uncertainty about the central bank's strategy. Internally consistent and transparent distributional statements about the outlook and the policy response would reveal the central bank's assessment of the economy and reaction function. It is then up to private sector agents to optimise given their private information and beliefs based on this comprehensive description of the central bank's strategy.
In addition to this timeless case for complete forward guidance there is a more powerful case for enhanced communication about the future path of policy given the heightened uncertainty about the exit strategy from emergency policy settings. The considerations that shape the normalisation of policy rates and the likely end point of that process are unclear. To make matters worse, it is unclear how that process will interact with the unwinding of the asset purchase programmes that some central banks undertook during the crisis. Finally, there is added complexity created by the scope for micro and macroprudential interventions during the exit process which might have macroeconomic consequences and should therefore be reflected in the monetary policy stance. All things considered, the outlook for policy – and in particular the underlying strategy of central banks – is unusually uncertain when approaching exit. In the absence of a clear statement of how policy could evolve along different paths, there is a risk that seemingly trivial comments by policymakers or pieces of news (e.g. the so-called Taper Tantrum in May and June 2013) could have a disproportionate impact on markets – and the real economy – as investors radically revise their priors on the central bank's strategy.
What we propose
We explain the communication reforms we have in mind in the context of the United Kingdom (Barwell and Chadha 2013). The Bank of England has already made considerable progress on this front, at present the publishing fan charts – which illustrate the range of possible outcomes for each quarter of the forecast – for key macro variables (McKeown and Paterson 2014). These are conditional on a particular assumption about the policy stance (i.e., the path of Bank Rate and stock of purchased assets). We believe that the Bank should instead publish a set of internally consistent fan charts for both key macro variables and the policy stance which are conditioned on the Committee's view of the monetary policy reaction function over the forecast horizon.
These new fan charts would provide the Monetary Policy Committee's best estimate of both how the Bank Rate and the stock of purchased assets are likely to evolve, and the uncertainty around those paths that reflects the Committee's assessment of the economic outlook and the optimal response. The nature of that uncertainty could be explained by incorporating scenario analysis into the communication strategy. To be fair, the Bank has started to do this – describing the sensitivity of the economic forecasts to particular assumptions – but the all-important policy response in these scenarios is absent. What we have in mind is a detailed discussion of precisely how policy responds in these scenarios to demystify monetary strategy. Earlier work by Chadha and Nolan (2001) does not suggest that such attempts increase market perceptions of volatility. Indeed, these fan charts could form the basis of a broader discussion about central bank strategy. Consider the following two examples.
Example 1: The Bank could publish fan charts for policy variables which stretch over a longer time horizon than the economic forecasts (three years) to provide information about the entirety of the exit strategy. It could publish fan charts for the level of the Bank Rate further into the future to shed light on the uncertainty around the level of the neutral rate, the time taken to get there, and the distance the Bank Rate might be from neutral at that point given the potential for shocks to hit the economy. Likewise, the Bank could publish fan charts illustrating the uncertainty around the expected run-off date for the stock of purchased assets to provide information to the gilt market on the net supply of bonds.
Example 2: The Bank could use the fan charts as a vehicle to explain the interaction between macroprudential and monetary policy in a possible future upswing of the financial cycle – for example, the emergence of a bubble in the residential property market. The macroprudential remit has been drawn to allow policymakers to intervene in the pursuit of broader economic stability, beyond safeguarding the resilience of core financial institutions. Exactly how policymakers will intervene, with what instruments, to what effect, and to what end is less clear given the absence of clear framework for macroprudential policy (Barwell 2013). That in turn leads to uncertainty around the circumstances in which monetary policy will be used as a first, intermediate, or last line of defence to tame the financial cycle.
Communication by a committee of experts
Institutional arrangements for the conduct of monetary policy vary from country to country. In almost all cases policy is set by committee, and in most of those cases the committee members are encouraged to express their individual views. Many central banks already publish the pattern of votes within the policy committee, albeit with a lag. In our view that delay creates needless speculation and volatility. From time to time members of those committees are likely to disagree about more than just the current policy stance. Differences of view about the overall strategy – for example the relative merits of an early but gradual exit from the lower bound versus a late and rapid – and the potential for those views to evolve are a genuine source of uncertainty about the outlook for policy. This is information which could be put into the public domain alongside the fan charts representing the views of the committee as a whole. Rather than asking policymakers to make point forecasts of the policy stance several years into the future (the FOMC dots) we propose publication of probabilistic statements which reflect – and respect – uncertainty. Given that uncertainty or news itself is increasingly thought to be a possible driver of the business cycle (see, for example, Schmitt-Grohe and Uribe 2012), it is time for central banks to come clean on the various interest rate strategies that are under active consideration.
Barwell, R (2013), Macroprudential Policy: Taming the wild gyrations of credit flows, debt stocks and asset prices, Palgrave.
Barwell, R and J S Chadha (2013), “Complete forward guidance”, in den Haan, W (ed.) Forward guidance, VoxEU.
Blinder, A. (1998), Central Banking in Theory and Practice, Cambridge: MIT Press.
Chadha, J S and C Nolan (2001), “Inflation Targeting, Transparency and Interest Rate Volatility: Ditching Monetary Mystique in the U.K”, Journal of Macroeconomics, 23(3), 349-366.
McKeown, J and L Paterson (2014), “Enhancing the transparency of the Bank of England's Inflation Report”, VoxEU
Morris, S and H Shin (2000), “Rethinking multiple equilibria in macroeconomic modelling”, NBER Macroeconomics Annual, 15, 139-182.
Schmitt-Grohe, S and M Uribe (2012), “What's News in Business Cycles”, Econometrica 80, 2733-2764.
Theil, H (1958), Economic Forecasts and Policy, North Holland.
Tinbergen, J (1952), On the Theory of Economic Policy, North Holland.
Wallis, K (2005), “Combining density and interval forecasts: A modest proposal”, Oxford Bulletin of Economics and Statistics 67, 983-994.