RDP 2015-09: Inflation Targeting: A Victim of Its Own Success? 3. The Next 25 Years

As we have seen, there is strong evidence that 25 years of inflation targeting have delivered inflation processes that are much better anchored and much less affected by the business cycle than they were before the advent of inflation targeting. Although we focused on Australia above, its experience is illustrative of the experience of many countries around the world.[7] These changes mean that the challenges facing central banks are likely to be of a quite different character to the challenges dealt with over the past 25 years. Moreover, the financial crisis has stimulated a renewed debate about whether inflation targeting is the most appropriate way to conduct monetary policy. We suggest that it is – subject to some evolutionary changes. But, before we get there, it is useful to review some of the criticisms that have been directed at inflation targeting since the financial crisis. We discuss the alternative monetary policy frameworks that have been suggested in light of these criticisms and how these criticisms are a natural consequence of the changed behaviour of inflation over the past 25 years. It is, ultimately, the fact that the changed behaviour is a reflection of successful inflation targeting that argues against wholesale change.

3.1 A Flavour of the Debate

The financial crisis has been the catalyst for much criticism of inflation targeting. Wren-Lewis states

Whatever the causes, there is now a clear conflict between what a sensible UK monetary policy would be doing and what is actually happening. Monetary policy is not providing enough stimulus to the UK economy, because it is focusing on the inflation target, and not the output gap. Inflation targeting in the UK is not working, and something needs to change. Wren-Lewis 2013

Joe Stiglitz (2011) put it thus, ‘[t]he idea that targeting inflation will lead to financial stability or that focusing on only price and financial stability is sufficient for maintaining a low output gap and stable and robust growth is fundamentally flawed’.[8] Jeffrey Frankel (2012a) has already prepared an obituary for inflation targeting, writing that ‘[t]he monetary regime, known affectionately as “IT” to its friends, evidently passed away in September 2009’.

These criticisms stem from a view that, given depressed economic conditions, central banks should be running very stimulatory monetary policy, pretty much regardless of the rate of headline inflation. While central banks have generally been running stimulatory policy, the criticism is that they have not been aggressive enough because of fears of breaching their inflation targets. For example, it is suggested that the European Central Bank delayed lowering interest rates because it was overly concerned about headline inflation rates that were being boosted by temporary oil and commodity price increases. In the United Kingdom, as alluded to by the quote from Wren-Lewis above, the suggestion is that persistently high inflation outcomes and rising inflation expectations constrained the stimulus that the Bank of England provided.

In short, in the view of many critics, current monetary policy frameworks place too much weight on CPI inflation. The solutions that have been proposed address the perceived shortcomings in two main ways. One strand of suggestions has been to focus on inflation measures other than the consumer price index – in particular, to focus on measures that respond more closely to domestic cyclical conditions. For example, targets could be defined in terms of the rate of increase in labour earnings net of productivity gains (unit labour costs). Monetary policy would thus be tightened when abnormal increases in wages signal bottlenecks in the labour market. Another suggestion is to give asset price inflation more prominence in monetary policymaking, given the large asset price rises that occurred during the first decade of the 2000s and their role in the financial crisis. Asset price developments may signal changes in financial stability and, thus, inform judgements on the risks to output. To the extent that monetary policy can influence asset prices, sacrificing near-term output by ‘leaning against the wind’ could in some circumstances more than offset the expected future output cost of a financial crisis. While both labour earnings net of productivity gains and asset price changes are still measures of inflation, the ideas have at their heart the goal of choosing targets that are more in line with output fluctuations. If the economy is booming, it is argued, it is more likely to be showing up in wage measures or asset price rises than in headline inflation.

The other main strand of suggestions is to target output fluctuations more directly. For some, this would be an explicit mandate to stabilise output – similar to the Federal Reserve's so-called dual mandate. In this dual-mandate framework, central banks' decisions would be based not only on their views about inflation, but also on direct measures of output and unemployment gaps. Central banks would thus have more discretion to allow inflation fluctuations if addressing them would exacerbate cyclical downturns. Alternative approaches would incorporate output into the framework by making nominal GDP the target of policy.

3.2 What are the Options?

As discussed above, there are two broad suggestions for how to ‘fix’ inflation targeting given the tensions revealed in the aftermath of the financial crisis: (i) modify the particular definition of inflation that is being used or (ii) incorporate output into the target more explicitly. There is, also, a third option to maintain the current framework. We discuss these general suggestions next.

3.2.1 Modify the target definition

During the Great Moderation there was an unusual correspondence between stabilisation of CPI inflation and output: cost-push shocks were short-lived and typically small. But as the Bank of England's experience illustrates, this correspondence has broken down. Confronted with persistent imported inflationary pressures, it has been argued that the CPI inflation target restricted its ability to accommodate non-domestically generated variation in inflation. In contrast, the rise of China and other emerging market economies as low-cost producers of manufactured goods in the 1990s and early 2000s restrained tradeable inflation and allowed central banks to tolerate relatively high rates of non-tradeable inflation. Put this way, it seems natural to consider adjusting the target inflation measure to allow for a greater degree of flexibility.

Adopting an inflation measure that corresponds more closely to domestic economic conditions reduces the potential conflict between output and inflation stabilisation, while maintaining a credible nominal anchor for monetary policy. A target inflation measure that abstracts from idiosyncratic variation is attractive because doing so holds the central bank responsible only for the prices under its influence.

Replacing CPI inflation with non-tradeable inflation as the target measure, for example, would largely abstract from commodity price and exchange rate movements. As Bharucha and Kent (1998) explain, targeting non-tradeable rather than CPI inflation allows the central bank to tolerate relatively large movements in the exchange rate. They draw attention to the exchange rate channel of monetary policy transmission, and show using a small open economy model that it is optimal for a central bank with a non-tradeable inflation target to respond relatively aggressively to supply and demand shocks, at the expense of exchange rate and CPI inflation variability. Targeting a non-tradeable inflation measure does not hold central banks responsible for cross-country spillover effects of export price inflation, but neither does current practice: Inflation-targeting central banks use consumer rather than producer price target measures.

A complication associated with adopting non-tradeable inflation as the target measure is that non-tradeable inflation has consistently exceeded tradeable inflation. Because non-tradeable inflation is a biased measure of average CPI inflation, consumer inflation expectations might become anchored at this higher level because it was the target of policy. If so, policymakers would have to either tolerate a higher level of average inflation, or engineer a costly disinflation to align the pace of non-tradeable inflation with the existing inflation target.

As mentioned earlier, another alternative is to adopt a measure of labour earnings net of productivity as the target measure, potentially providing a better indication of the trend pace of inflation than a consumer price measure. A drawback is the notorious difficulty in estimating productivity growth: reliable productivity estimates are only available for the market sector, and the data are often substantially revised. Changes in the composition of employment over the business cycle would also complicate the use of a labour cost target measure to guide monetary policy. Furthermore, such a measure would abstract from the important role that changes in margins play in the inflation process.

3.2.2 Target output more explicitly

Rather than change the target inflation measure, central banks could adopt an explicit output stabilisation objective, to complement the inflation target. A dual mandate would provide flexibility to accommodate persistent commodity price or exchange rate shocks that push inflation above target during times of economic slack. In contrast, a strict CPI inflation objective requires monetary policy tightening, exacerbating the fall in output. The flattening of the Phillips curve would also suggest that a re-evaluation of the trade-off between inflation and output would be in order as offsetting even relatively minor cost-push shocks requires a larger fall in output than in the past.

One mechanism to increase the importance of output relative to inflation is to replace inflation targeting with nominal GDP growth targeting, an old idea that has gained prominence since the financial crisis. A nominal GDP growth target implicitly places equal weight on output and inflation stabilisation, which to its proponents achieves a better balance of objectives than inflation targeting. But targeting nominal GDP growth does more than reweight the inflation and output stabilisation objectives: it changes the target inflation measure. The consumer price inflation measure used by inflation-targeting central banks includes the price of imports and excludes the price of exports, while the GDP price measure does the reverse. Excluding import prices automatically accommodates imported inflation, such as oil price shocks, as would adopting non-tradeable inflation as the target inflation measure. However, the desirability of adopting a target measure that includes export prices is less clear. Frankel (2012b) argues that producer price targeting has the beneficial effect of stabilising export prices in local currency terms. But for a small open economy such as Australia, the inclusion of export prices in the target inflation measure would expose the non-resources economy to large, and mostly exogenous, monetary policy changes. This is potentially problematic when there are level shifts in the terms of trade that the central bank must seek to identify in real time.

A more radical proposal is the adoption of nominal GDP as a level rather than growth target. Like price level targeting, a nominal GDP target does not let ‘bygones be bygones’: past deviations from target must be corrected in the future. During his time as Governor of the Bank of Canada, Mark Carney (2012) argued that nominal GDP targeting has particularly attractive properties at the zero lower bound. In an economic slump nominal GDP falls, and inflation expectations must rise for the central bank to maintain its nominal GDP target; any rise in inflation expectations lowers the real interest rate and stimulates demand. In essence, a nominal GDP target might endogenously generate countercyclical inflation expectations. The success of nominal GDP targeting crucially depends on the speed with which consumers' and firms' inflation expectations adjust. Following the adoption of inflation targeting, inflation expectations remained substantially above target for several years. Imperfect inflation credibility is likely to have been important, but so was sluggish adjustment of inflation expectations. Supporting this, a growing literature argues that information frictions are an important source of inertia in the monetary policy transmission mechanism (see, for example, Mankiw and Reis (2002)). If inflation expectations adjust sluggishly, a nominal GDP target may only raise inflation expectations marginally in an economic slump, undermining one of its key features.

Conflict between output and inflation stabilisation in the post-financial crisis period should not be overemphasised: inflation has remained close to its target for most inflation-targeting central banks, despite substantial economic slack and highly accommodative monetary policy. With the exception of the United Kingdom, the potential relevance of a dual mandate policy is clearer in the lead-up to the crisis, during the sustained rise in oil prices. Jeffrey Frankel (2012a) argues that ‘… it is widely suspected that the reason for the otherwise-puzzling decision of the European Central Bank to raise interest rates in July 2008, as the world was sliding into the worst recession since the 1930s, was that oil prices were just then reaching an all-time high’. Regardless of whether Frankel's assessment is correct, it is this type of conflict between output and inflation stabilisation that a dual mandate policy is designed to avoid.

3.2.3 Maintain current targets

An alternative to changing the target measure or adopting a dual mandate is to retain CPI inflation as the target. CPI inflation is perhaps the simplest and most relevant inflation target to consumers: it measures consumers' average inflation experience, is a key input to wage negotiations, and is used for indexation purposes in contracts. However, this is not a ‘no change’ option. For a start, inflation targeting has evolved over time and generally takes account of activity in practice. Furthermore, even if the ultimate target does not change, that does not mean that further evolution of the framework is precluded. Indeed, there are a couple of changes in practice that may be worth considering in light of the changes identified above.

First, the communication of the central bank may need to become much more nuanced. Some inflation shocks, those that reflect fluctuations in domestic economic activity, may have to be addressed aggressively, while it may be better to look through others, such as exchange rate shocks. Of course, not all exchange rate shocks are alike, and the appropriate degree of monetary policy accommodation depends on the source of the shock. The challenge for central banks' communication strategies is to explain why certain shocks are being ignored, while others are being addressed. Similarly, there would be communications challenges to the extent that financial stability concerns motivate any leaning against the wind. Second, central banks' internal analysis may need to improve. While the flattening of the Phillips curve and anchoring of inflation expectations might seem like good news, it has an important drawback. Inference about the state of the economy based upon the behaviour of inflation is now much more difficult. Previously, capacity constraints would show up in inflation relatively clearly and induce an appropriate tightening of policy. Now, with the effect muted, it can be hard to identify a structural tightness in the economy, which can lead to persistence of that tightness that may have undesired effects. A prime example would be the experience of many euro area countries that saw property booms in the lead-up to the financial crisis. Contained inflation was taken as evidence that output gaps were smaller than they actually were and allowed stimulatory policy to go on for longer than it otherwise would have. Compounding these analytical challenges are the difficulties of forecasting the highly non-linear effects of financial instability. In short, the flattening of the slope of the Phillips curve and greater anchoring of expectations means that the separation of systematic movements in inflation from random noise is now much harder – NAIRU-based forecasts of inflation are now much less reliable and new techniques will need to be developed. And changes in the processes governing inflation identified above mean that the Lucas Critique applies with great force. Models which fail to take account of this are likely to make systematic errors.

3.3 Discussion of the Options

In thinking through the options, it is worth emphasising that the effects of inflation targeting evident in the data are twofold. First, there has been a flattening of the Phillips curve, whereby the linkages between inflation rates and output gaps have weakened. Second, there has been an increase in the anchoring of inflation around long-run expectations, which are invariably the same as the stated targets. Although we have documented these changes in the inflation process for Australia, the same qualitative changes have been documented elsewhere for other countries (see IMF (2013)), so the implications we draw apply generally.

The strongest critics of inflation targeting argue that wholesale change is required: either adopt explicit dual mandates or change target inflation measures. Both these proposals share the common objective of minimising the chance of conflict between output and inflation stabilisation. But, as we have argued, these arguments for change are, in part, a consequence of the success of inflation targeting. With inflation expectations now firmly anchored at target and the Phillips curve flatter, the non-tradeable component of inflation has been stabilised, and the relative importance of the idiosyncratic and uncontrollable component of CPI inflation has risen. Inflation is now much more affected by shocks where the inflation and output stabilisation objectives are in conflict than in the past. And, in such an environment, a pure CPI inflation target risks destabilising output to offset idiosyncratic shocks. The adoption of a dual mandate minimises the possibility of conflict by permitting inflation to be above target when output is depressed, as does changing the target to an inflation measure more closely associated with economic activity. A difficulty, however, with proposals to down-weight the inflation target is that, even if it does not affect the slope of the Phillips curve relationship, it risks undermining the anchoring of expectations. And it is only because expectations are so strongly anchored that idiosyncratic shocks appear to be so important.

Furthermore, the characterisation of inflation-targeting central banks as caring exclusively about CPI inflation is something of a straw man. The practice of inflation targeting has evolved in conjunction with monetary policy frameworks. For example, the Reserve Bank of New Zealand's inflation target band was widened from 0–2 per cent to 0–3 per cent in December 1996 to provide additional flexibility, and the 2010 ‘Statement on the Conduct of Monetary Policy’ between the Reserve Bank of Australia and the Australian Government officially recorded the Reserve Bank's responsibility for financial system stability. More generally, underlying inflation measures are now routinely used as a guide for policy, abstracting from sharp idiosyncratic variation in inflation that is unrelated to domestic economic conditions. Central banks have also become more forward looking, setting monetary policy based on forecasts of inflation and output/unemployment, rather than contemporaneous estimates. These forecasts are typically guided by a Phillips curve relationship and, because idiosyncratic changes in inflation more than a couple of quarters ahead are essentially unforecastable, inflation forecast-targeting central banks implicitly set monetary policy based on a measure of inflation that reflects domestic economic activity. (Although, as the simulation above showed, if those forecasts are premised on an unchanged Phillips curve, they may prove to be misleading.) As Ryan and Thompson (2000) explain, the benefit from adopting a non-tradeable inflation target is unclear when monetary policy is guided by inflation forecasts that abstract from exchange rate shocks.

There is, thus, a middle ground between the wholesale change envisaged by the sharpest critics of inflation targeting and a ‘do-nothing’ position: remove barriers to the practice of ‘flexible inflation targeting’ where they exist by lengthening the target horizon and continuing the evolution that has been occurring over the past 25 years. A long horizon, such as the Reserve Bank of Australia's ‘over the cycle’ criterion, maintains CPI inflation as a clear, transparent, medium-term nominal anchor, but minimises the likelihood of conflict between output and inflation objectives. A lengthening of the target horizon is a natural consequence of changes in the inflation process we have documented over the inflation-targeting period. Now that inflation credibility has been established, there is greater scope than in the early years of inflation targeting to tolerate meaningful deviations from target: consumers and firms are less likely to interpret deviations from target as revisions to the implicit inflation target than when inflation targeting was in its infancy. How much central banks can leverage their credibility to tolerate persistent deviations in inflation from target is an unknown empirical question. Clearly, there is a limit: expectations adjust, even if only sluggishly. Nevertheless, the potential for inflation expectations to become ‘unanchored’ should not be overemphasised: a defining feature of the past decade has been the constancy of long-term inflation expectations through large swings in commodity prices and a deep economic slump.

Our suggestion should not be mistaken for complacency. Indeed, we cannot forget that the benign inflationary outcomes during the 2000s masked the build-up of imbalances that contributed to the financial crisis. Rather, central banks must be increasingly vigilant in identifying changes in the trend pace of inflation, and at the same time willing to tolerate commodity price or exchange rate shocks that push CPI inflation away from target for a time. Clear communication will be required to explain changes in the stance of policy. Policy tightening may be required when the trend pace of inflation is forecast to rise even if CPI inflation remains close to the target. Conversely, in the presence of idiosyncratic shocks, monetary policy may often remain accommodative. Because the appropriate policy response to an inflation surprise crucially depends on its cause, structural models that can identify the source of shocks are needed. Furthermore, the breakdown in the forecasting performance of the Phillips curve suggests that near-term forecasting will need to make use of a broad range of economic indicators.


See IMF (2013). [7]

Notably, however, he acknowledges the following in a parenthetical comment immediately after his criticism ‘(In extreme cases, of course, where the issue is not 3, 4, or 5 percent inflation but more like 10 percent inflation, central banks must focus on inflation as well. But in places like the United States and Europe, where inflation has been controlled, this is not the issue.)’ (emphasis added). We believe this really is the issue and discuss it further below. [8]