RDP 9701: Inflation Regimes and Inflation Expectations 1. Introduction and Summary

Average inflation rates in industrial countries have fallen substantially since the 1980s. In several cases, countries that experienced higher inflation than the industrial-country average during the 1970s and 1980s have achieved lower than average inflation in the 1990s. In most cases, these recent low-inflation countries have not experienced a commensurate reduction in their long-term interest rates relative to the industrial-country average, suggesting that long-term inflation expectations have not moved in proportion to recent inflation outcomes.^[1]

To provide concrete examples, examine two bilateral comparisons: the United States versus Canada and Australia versus New Zealand. According to the OECD (1996), consumer price inflation (measured by the consumption deflator) was lower in Canada than in the United States for every one of the past five years. The average inflation rate over this period was 1.3 per cent in Canada and 2.6 per cent in the United States. Despite this consistent record of lower inflation, the yield on 10-year government bonds was identical across the two countries at 6.7 per cent at year-end 1996. Similarly, the inflation rate was lower in New Zealand than in Australia for every one of the past five years, and the average over this period was 1.7 per cent in New Zealand and 2.0 per cent in Australia.^[2] Nevertheless, at year-end 1996, 10-year interest rates were identical in the two countries at 7.7 per cent.

One explanation for these findings is that long-term inflation expectations depend on a long history of past inflation – more than just the past five years. Indeed, during the 1980s, inflation averaged 6.2 per cent in Canada versus 5.3 per cent in the United States, and inflation averaged 11.9 per cent in New Zealand versus 8.2 per cent in Australia. In each case the country with lower recent inflation experienced higher inflation over a long period in the past. The effect of past inflation over a long horizon can also explain the higher 10-year interest rates in Australia and New Zealand versus the United States and Canada.^[3]

More generally, there is evidence documented by Gagnon (1996) that nominal long-term interest rates are strongly correlated with both recent inflation and past inflation over a long horizon. This correlation holds both across countries and within countries over time. One explanation for this correlation is that long-term inflation expectations are influenced by a long history of past inflation. Gagnon (1996) also presents direct evidence for this hypothesis from the spread between nominal and indexed bond yields.

This paper develops a theoretical framework to explain these empirical findings. The basic idea is that since the collapse of the gold standard earlier this century, central banking in most countries has been characterised by periodic changes in policy regime. At the most basic level, regime changes are associated with changes in the central bank governor or political party in power, depending on the institutional independence of the central bank. Other factors may give rise to regime changes: evolving theories about economic behaviour may lead to new views on the optimal conduct of monetary policy. Or, extreme social or economic shocks may necessitate a persistent shift in monetary policy. However, in general, it is not useful to think of the regime changing with every shock. Instead, regimes are viewed as implicit or explicit rules governing the behaviour of monetary policy in response to ordinary shocks.

One important outcome of different monetary regimes is different average inflation rates across regimes. When agents are considering expected inflation over a long future horizon, they must factor in the possibility that the current regime will not survive over the horizon in question. Recent inflation rates may provide a good forecast of future inflation rates if the current regime survives, but they may not provide a good forecast if the current regime is replaced. To factor in the effect of a potential new regime, agents may base their forecasts on their experience of past monetary regimes over a long horizon.

To take the example of New Zealand once again, the Reserve Bank of New Zealand's previous central target of 1 per cent inflation was lower than the inflation rate in every year but one since World War II. Thus, if agents were considering the possibility of a new inflation regime in the future, it seems likely that they would expect any new regime to have average inflation greater than 1 per cent. Even if agents believed that the Reserve Bank of New Zealand would achieve its target of 1 per cent inflation in the current regime, they would have to factor the possibility of a change to a higher-inflation regime into their expectations, thereby raising expected future inflation above 1 per cent. The importance of regime changes for expectations of future inflation was highlighted in the context of the recent New Zealand election. Some of the parties argued for a higher-inflation regime and no party argued for lower inflation. In the event, the central inflation target has been raised slightly, from 1 to 1.5 per cent. Moreover, there was a possibility that an even greater increase in the inflation target might have resulted after the election.

In addition to explaining long-run inflation expectations in bond markets, a model with regime changes can explain the peculiar time-series properties of actual inflation over the postwar period. For most industrial countries, it is difficult to reject a unit root in the inflation rate. Yet, recent studies have found some evidence of weak mean reversion of inflation rates over long horizons. It is well known that structural breaks in an otherwise stationary series can induce apparent unit roots into the series. If inflation has undergone a small number of regime shifts in the postwar period, it would be difficult to reject a unit root. However, if the regime shifts themselves were around a constant average inflation rate, one would expect to find some evidence of mean reversion in inflation. Moreover, within relatively long-lasting regimes it should be possible to reject a unit root, which may explain the apparent stationarity of inflation over certain subsamples.

Finally, the possibility of regime shifts leads to highly asymmetric distributions of future inflation rates. The asymmetric distribution of future inflation may explain the asymmetric distribution of survey responses on future inflation expectations. Moreover, the asymmetric distribution of future inflation may explain the frequently large discrepancies between surveys of inflation expectations and implied inflation expectations in bond yields. If survey respondents report the most likely outcome, and bondholders care about the average outcome, then the discrepancy between different measures of inflation expectations would be resolved.

Footnotes

See Ammer and Freeman (1995) and Freeman and Willis (1995). [1]

Although the difference in recent inflation outcomes between these countries has been relatively small, the Reserve Bank of New Zealand has a target band for inflation of 0 to 3 per cent (formerly 0 to 2 per cent) whereas the midpoint of the Reserve Bank of Australia's objective is 2½ per cent. Thus, one might expect to find lower inflation expectations in New Zealand since the center of its target band is one percentage point lower than Australia's. For more on the specification of inflation targets in different countries see Reserve Bank of New Zealand (1996) and International Monetary Fund (1996). [2]

Another explanation for different nominal long-term interest rates is that real long-term interest rates may differ across countries. Real rates may differ due to different risk premia or to different levels of real exchange rates relative to their long-run equilibria. Both of these factors may be contributing to the interest rate differentials observed here, especially the differential between Australia and New Zealand on the one hand and the United States and Canada on the other hand. [3]