RBA Annual Conference – 2004 Discussion
1. Warwick J McKibbin[1]
The paper by Rick Mishkin is a clear and timely overview of the state of the debate on the transparency of central banks. The paper asks three supplementary and related questions on the way to answering the main question – ‘Can central bank transparency go too far?’ The first question is whether central banks should publish forecasts of a range of variables and not just forecasts of inflation. In particular, the paper considers whether central banks should announce a forecast for real output and a ‘no change interest rate’ or a future path for interest rates. The second question is whether central banks should announce their objective function, so that the private sector understands how they will react to a variety of possible future events. The third question is how central banks should talk about output given that most have both output and inflation in their objective functions.
The paper gives a good overview of current practice, which varies substantially across countries. Almost all inflation-targeting central banks publish forecasts of inflation. Many central banks do not publish output forecasts. Most central banks do not publish forecasts of interest rates and no central banks publish their objective function.
The first practical issue that needs to be considered is what ‘too far’ means? Is the issue being ‘too far’ relative to the optimal amount of transparency, or is it relative to the current status quo? Figure 1 illustrates this point. It shows the extent of transparency of central banks expressed as an index which is zero when non-transparent and unity when completely transparent. The vertical axis shows the social welfare obtained from being transparent. The frontier in the figure shows the pay-offs to society of various degrees of transparency from none to full. Now suppose we pick three arbitrary central banks – central bank A, central bank B and central bank C. Clearly central bank B is at the optimal degree of transparency where any further revelation of information leads to a reduction in social welfare. Thus central bank C has gone too far relative to central bank B. However, it has not gone too far relative to central bank A, which generates lower social welfare because of low transparency. In reality most central banks are at point A and therefore even if they ‘go too far’ it could be welfare improving relative to the status quo.
This example is completely arbitrary because in fact the pay-off frontier may not have the shape in Figure 1, but it does illustrate the point that we need to be precise about going ‘too far’ in the context of how far central bank transparency has already moved. Most central banks are down near point A, in which case greater transparency from their current positions would be beneficial for the wider community, even though it might make life more difficult for central bankers.
A further issue which emerges from the paper is that there is a great deal of ambiguity about how actual central bank behaviour fits into the theoretical literature on monetary regimes. Much of the theoretical and empirical literature in the early 1990s (see Bryant, Hooper and Mann 1993 and the discussion in McKibbin 1997) focused on nominal income targeting, rather than inflation targeting, as a regime that handled the uncertainty of possible shocks the best. The emergence of the inflation-targeting literature in the mid 1990s was mainly generated by the desire of economists to better understand actual central bank behaviour in their inflation-targeting rhetoric, rather than a natural extension of the literature on evaluation of alternative regimes. The notion of flexible inflation targeting, which is now popular in the literature, is really closer to the concept of nominal income growth targeting in the sense that, as also pointed out by Rick, real output considerations enter into the decisions on how quickly central banks respond to inflation changes. In my opinion, the Reserve Bank of Australia's concept of ‘inflation targeting over the cycle’ is a form of flexible inflation targeting which is closer to a nominal income growth-targeting regime than a pure inflation-targeting regime. This ambiguity in the mapping of the theoretical to the actual regimes is illustrated in Figure 2.
As Rick points out, there are two parts to an inflation-targeting regime. The first is the central bank's objective function, which usually contains both inflation and output. The objective function is defined for all monetary regimes. The other part of the inflation-targeting regime is the policy rule in which the interest rate adjusts in response to either actual or expected inflation over some future period. The form of the interest rate rule determines whether the central bank is an inflation targeter or following some other regime.
There are a number of streams of literature which bear on the theoretical aspects of Rick's paper. The analytical example in the paper of the role of transparency, which is adapted from the Svensson (1997) approach, assumes rational expectations. There is a new emerging literature, such as in Preston (2004), in which learning can change the nature of the interaction between central banks and the private sector in important ways. Indeed in a learning model, inflation targeting can be destabilising, depending on the forecasting process.
The other relevant literature is the evaluation of regimes based on the ability to offset shocks. There is a clear theoretical and strong empirical result that nominal income targeting dominates pure inflation targeting in the case of supply shocks or shocks to country risk (see Henderson and McKibbin 1993). This is particularly true for developing countries in which a majority of shocks are likely to be supply shocks due to events such as weather outcomes or underlying structural changes (see McKibbin and Singh 2003 for India).
Rick raises the dilemma of how central banks should reveal their concern over fluctuations in real output while maintaining an inflation target. Presumably by being more explicit about monetary policy responses to real output, a central bank provides a clear view to the private sector about how it will respond to certain shocks. This should enhance the credibility of the nominal anchor associated with an inflation target, even while the central bank responds to some shocks in a way that temporarily moves inflation away from the target. Perhaps the answer to this dilemma can be found in the early theoretical literature on the desirability of some form of explicit nominal income growth targeting. Rick argues that bringing greater transparency about output fluctuations into the monetary debate might violate the KISS (Keep It Simple Stupid) principle, because talking about both inflation and output targets would be too complicated. However, an announced nominal income target is a single variable (although it does contain both inflation and real output growth as components). The general public could focus on the nominal income target and the central bank could announce the underlying inflation target for aficionados. The advantages of a nominal income growth target over a pure inflation target are many. Nominal income growth is as forecastable as inflation for the United States and more forecastable for some countries such as Korea (see Ponomareva 2004). The critical issue for forecastability is the covariance of forecast errors between real GDP and inflation, which Ponomareva (2004) finds is negative for Korea. Nominal income growth is a single variable as is inflation. It provides a nominal anchor through the inflation component of nominal income growth. Nominal income is easy to understand for the wider population (especially since it is not a complex price index but a measured variable). Nominal income targeting also requires less explicit exemptions than inflation targeting for a wide variety of shocks. How many times can a central bank make an exemption for a negative supply shock before the nominal anchor of inflation targeting is seriously undermined?
The key issue in the debate about transparency is not where central banks are relative to the extreme but where they are relative to the optimal. Surely, most central banks have not gone far enough when it comes to transparency – in terms of Figure 1 they are located closer to central bank A than central bank B. Inflation targeting has worked well in the 1990s. How much of this is good luck in terms of the types of shocks experienced, rather than good management, is an open question. There is a reasonable argument that the current form of inflation targeting will not necessarily survive the types of shocks that might buffet the world economy in the future. A case can even be made that in the event of large shocks, such as the Asian crisis, some inflation-targeting central banks outside those countries directly impacted were lucky with the way they adapted their inflation targets to deal with the shock while others were unlucky. Looking forward, there are a wide range of potential negative shocks – such as oil price shocks, terrorism, weather events and other unpredictable events – which may really test the forms of inflation targeting that some central banks have adopted. The extent to which a nominal anchor such as ‘inflation targeting with exemptions’ can withstand these shocks will inevitably be tested. A case can be made that it is better to improve a policy regime when it is not under stress rather than wait until the flaws are exposed during a period of serious systemic stress. Perhaps nominal income growth targeting in some form will be the monetary regime that evolves over the coming decade, when central banks become more transparent and Rick's revelation of the ‘dirty little secret’, that central banks also care about output, will be more clearly evident.
Footnote
The views in this note do not necessarily reflect the views of any of the institutions with which the author is affiliated. Useful comments and discussions with Renee Fry, David Gruen and Adrian Pagan are gratefully acknowledged. [1]
References
Bryant RC, P Hooper and CL Mann (eds) (1993), Evaluating policy regimes: new research in empirical macroeconomics, Brookings Institution, Washington DC.
Henderson DW and WJ McKibbin (1993), ‘A comparison of some basic monetary policy regimes for open economies: implications of different degrees of instrument adjustment and wage persistence’, Carnegie-Rochester Conference Series on Public Policy, 39, pp 221–317.
McKibbin WJ (1997), ‘Which monetary policy regime for Australia?’ in P Lowe (ed), Monetary policy and inflation targeting, Proceedings of a Conference, Reserve Bank of Australia, Sydney, pp 166–173.
McKibbin WJ and K Singh (2003), ‘Issues in the choice of a monetary regime for India’, in K Kalirajan and U Sankar (eds), Economic reform and liberalisation of the Indian economy, Edward Elgar, Cheltenham, pp 221–274.
Ponomareva N (2004), ‘What is the appropriate monetary policy regime for a developing country?’, draft PhD dissertation, Research School of Pacific and Asian Studies, Australian National University.
Preston B (2004), ‘Adaptive learning and the use of forecasts in monetary policy’, paper presented at the ANU Centre for Applied Macroeconomic Analysis (CAMA) seminar series, Canberra, 27 July.
Svensson L (1997), ‘Inflation forecast targeting: implementing and monitoring inflation targets’, European Economic Review, 41(6), pp 1111–1146.
2. General Discussion
There was general agreement with Mishkin's conclusion that central bank transparency can go ‘too far’. One participant noted that this followed from the recognition of the limits of central bankers' knowledge. In other words, many of the arguments in Mishkin's paper against greater transparency reflected the complexities involved in determining the exact details of the extra information being provided, for example, with respect to central banks' objective functions or the future path of policy interest rates. In that context, one participant suggested that the Brazilian experience referred to by Mishkin was an example where central bank transparency had gone too far. A possible effect of the Banco Central do Brasil's decision to present alternative policy paths was an excessive focus on output stabilisation, which was an argument Mishkin had raised for why a central bank should not release its objective function.
A number of participants noted that, while McKibbin's Figure 1 did help to elucidate some of the issues involved with central bank transparency, it did not represent developments accurately enough. Echoing a wider sentiment, one participant suggested that, 10 to 15 years ago, all central banks were to the left of central bank ‘A’, with very low levels of transparency. Transparency had since increased significantly, bringing most central banks to a point somewhere in the vicinity of central bank ‘B’, at the optimal level of transparency. Furthermore, the curve was significantly flatter in this area than indicated by McKibbin's figure. As such, while earlier increases in transparency were of first-order importance for welfare, any further refinements were likely to be of second-order importance.
A number of participants noted that, when considering the desirable level of central bank transparency, one size does not fit all. The appropriate level of transparency was, to some extent, dependent on a country's institutional structure. For instance, one participant noted that the Reserve Bank of New Zealand's high level of transparency was common across the entire New Zealand government. The legislative environment, and the Official Information Act in particular, would make it difficult for the RBNZ to argue against releasing its interest rate projections. In a similar vein, one participant noted that the appropriate size and make-up of the central bank's policy committee will also reflect a country's structural characteristics. For example, a larger, geographically diverse committee is considered appropriate in the US, given the consolidation of financial power (in New York) and political power (in Washington DC). Similarly, a large committee of central bank governors was appropriate for the European Central Bank.
In terms of specific elements of central bank transparency, discussion focused on a number of matters. While some participants argued against the publication of any endogenous policy interest rate projections, others contended that financial markets had taken a reasonably sophisticated approach to the presentation of such projections in New Zealand. Opinion also diverged on the desirability of the central bank publishing inflation forecasts based on an endogenous interest rate or a constant interest rate. One participant questioned whether releasing forecasts based on an endogenous interest rate was necessarily the more transparent option; as such a forecast was less easily understood by the non-specialist than a forecast based on a constant interest rate. Another participant noted that the decision to increase transparency – including the release of detailed information regarding projections – was largely irreversible. This was especially relevant for emerging market economies, given the importance of establishing and then sustaining central bank credibility.
There was also some discussion about the desirability of publishing the minutes of the central bank's policy committee. A number of participants noted that the release of detailed minutes, which attributed comments to individual committee members, had the potential to substantially alter the operation of the policy-making body: the power of the chairperson may be enhanced and greater deliberation may occur outside the formal meeting. Both of these changes could compromise the integrity of the decision-making process. One participant suggested that the release of detailed minutes may stop committee members playing devil's advocate during policy deliberations, which can often advance the policy discussion.
Following McKibbin's comments, a number of participants raised the issue of nominal income targeting, with most rejecting the approach. Some participants noted the difficulties that nominal income targeting would pose for developing countries – and also countries such as Australia, Canada and New Zealand – as they can experience significant terms of trade shocks, both supply and demand induced. With a nominal income target, a positive terms of trade shock may lead the central bank to tighten policy, which might not be desirable. Other participants focused on the fact that nominal income targeting is more complicated to explain to the general public than inflation targeting. In that sense, the introduction of inflation targeting from the late 1980s, at a time when the academic literature was focused on nominal income targeting, was a reflection of Mishkin's ‘KISS’ principle. With an inflation target the central bank can also nail down inflation expectations with a specific number. This is not possible with a nominal income target unless the central bank is also explicit about a target for potential growth, which some participants believed to be fraught with difficulties. Finally, one participant pointed to the US economic performance in the second part of the 1990s – strong growth combined with low inflation – and contended that a nominal income target would have led the Federal Reserve to tighten monetary policy more than was necessary.