آیا ترجیحات بانک مرکزی نامتقارن برای توضیح نتایج تورم مشاهده شده کمک می کند؟

When the central banker’s loss function is asymmetric, changes in the volatility of inflation and/or unemployment affect equilibrium inflation. This suggests that changing macroeconomic volatilities may be an important driving force behind trends in observed inflation. Previous evidence, which has offered support for this idea, suffers from a spurious regression problem. Once this problem is controlled for, the evidence suggests that the volatility of unemployment does not help explain inflation outcomes. There is some evidence of a relationship between inflation and its volatility, but overall the data does not support the view that changing economic volatility, as filtered through asymmetric central bank preferences, is an important driver of inflation trends.

It is well known that US inflation was low in the early 1960s, rose through the late 1960s and 1970s before falling through the 1980s, and remaining low thereafter. A similar pattern of rising then falling inflation occurred in many other OECD countries.1 What caused this rise and fall of inflation is an open question that has attracted a great deal of recent attention. In this paper we ask to what extent the common observed inflation trend in OECD countries is the result of the interaction of time inconsistency problems in monetary policy interacting with changes in the volatility of shocks to inflation and/or unemployment. This is a promising candidate explanation of inflation trends for a number of reasons. First, it is well known that the degree of macroeconomic volatility has fallen along with the level of inflation in many OECD countries in recent years.2 Thus the decline in OECD inflation rates is roughly coincident with the so-called Great Moderation. Second, standard time inconsistency models of monetary policy, extended to allow for asymmetric central bank preferences, provide a clear theoretical channel through which these factors would affect trend inflation. Since these models abstract away from country specific institutional details, the theory naturally extends from the US case to the cross country setting. Third, the comments of policy insiders3, empirical work on monetary policy reaction functions4, as well as direct estimates of central bank preference parameters based on more structural models5 support the key feature of these models: that central bank preferences are asymmetric. Finally, existing research explicitly examining the relationship between asymmetric central bank preferences and inflation outcomes appears to support the hypothesis.6 While the literature to date has largely focused on the US experience, the existence of a common pattern in inflation suggests that any successful explanation ought to be robust across OECD countries. The existing empirical work on this topic, however, covers only a handful of the countries that experienced a rise and fall of inflation. Surico, 2006 and Ruge-Murcia, 2003a, for example, only study the US, while Ruge-Murcia (2004) examines the US and four other countries. While these results are intriguing, the question of whether the theory works for a broad set of countries has not been answered. Addressing this question is the main objective of our paper. We begin the paper by documenting the existence of a common trend in the inflation rates of OECD other than the US. Fig. 1, Fig. 2, Fig. 3, Fig. 4, Fig. 5, Fig. 6, Fig. 7, Fig. 8, Fig. 9, Fig. 10, Fig. 11 and Fig. 12 plot inflation rates, measured by annualized quarterly percentage changes in the Consumer Price Index, for 12 OECD countries. For comparison, a 2-year centered moving average of US inflation is included on each plot (the dashed red line) A common pattern is visible in the raw data (the light blue line), but more transparent in the 2-year centered moving averages also displayed in the figures (the heavy blue line), and is as follows: inflation starts out low in the early 1960s in most countries. This is followed by a period of rising inflation lasting until the late 1970s or early 1980s in all countries except Germany and Japan (where inflation peaks in the early and mid 1970s, respectively). After this period of rising inflation, inflation rates then fall until the present, and are generally as low or lower by the end of the 1990s than they were in the early 1960s. The commonality of inflation outcomes over the past four decades suggests that a successful explanation of long run inflation trends ought to be applicable across OECD countries. Full-size image (16 K) Fig. 1. Australia. Figure options Full-size image (16 K) Fig. 2. Austria. Figure options Full-size image (16 K) Fig. 3. Canada. Figure options Full-size image (16 K) Fig. 4. Denmark. Figure options Full-size image (17 K) Fig. 5. Finland. Figure options Full-size image (16 K) Fig. 6. France. Figure options Full-size image (18 K) Fig. 7. Germany. Figure options Full-size image (15 K) Fig. 8. Italy. Figure options Full-size image (14 K) Fig. 9. Japan. Figure options Full-size image (15 K) Fig. 10. Norway. Figure options Full-size image (16 K) Fig. 11. Sweden. Figure options Full-size image (15 K) Fig. 12. UK. Figure options Theories of time inconsistent monetary policy based on asymmetric central bank preferences are a plausible candidate explanation of this common trend as these models are general enough to encompass the differing institutional arrangements across OECD countries. While early versions of these models required the policy makers target an unattainable unemployment rate, recent theoretical innovations show that monetary policy may suffer from time inconsistency even when central bankers target the NAIRU. When central banks have asymmetric preferences, policy makers care about the sign as well as the magnitude of deviations of unemployment and inflation from target. In this case, monetary policy suffers from a time inconsistency problem which causes equilibrium inflation rates to depend on the variance of the shocks to inflation and unemployment. 7 Consider, by way of illustration, a policy maker who dislikes above NAIRU unemployment more than below NAIRU unemployment, and suppose that the variance of shocks to unemployment increases. With a higher variance, the probability of an episode of very high (and, due to the asymmetry of preferences, strongly disliked) unemployment increases. The central banker will respond with expansionary monetary policy, in an attempt to drive the average unemployment rate down, to reduce the likelihood of an episode of very high unemployment. This policy, however, results in an increase in equilibrium inflation. Given the asymmetry in the loss function, the policy maker would be willing to pay this price to avoid a more distasteful episode of very high unemployment.8 A natural test of this theory is to use a GARCH model to estimate the conditional variance of shocks to unemployment and/or inflation and then regress inflation on this conditional variance to measure any correlation. The results of this simple exercise support the proposition that asymmetric preferences and changing volatility can explain inflation trends in at least some OECD countries.9 We argue that findings, based on this test, that changes in the conditional volatility of unemployment have statistically significant effects on changes in inflation are likely spurious, as these results are found in those countries for which the conditional variance is most persistent. We employ Monte Carlo methods to show that, when inflation is persistent, as it is in most OECD countries, simple regressions of inflation on the conditional volatility of unemployment over-reject a correct null hypothesis of no relationship if the conditional volatility of unemployment also exhibits persistence. To correct for this problem, we estimate the model as a cointegrating relationship. The results from this exercise suggest that changes in the conditional variance of unemployment do not explain changes in inflation, as we find support for the model in three countries out of a sample of 13. To control for the possibility of changes in central banks’ inflation targets, we repeat the exercise for a more recent subsample of data, where a constant inflation target is a more plausible assumption. This does not change the results. Overall the implication is that the combination of asymmetric central bank preferences and changes in the volatility of unemployment is not a promising explanation of time series inflation trends in OECD countries. We then examine the relationship between inflation and its conditional variance in time series data for OECD countries to allow for the possibility that central bank preferences are asymmetric in inflation, rather than unemployment. Again we estimate the model using a cointegration framework to reduce the likelihood of spurious results.10 The results here are more mixed than the unemployment results. Overall there is evidence of a relationship between inflation and its own conditional variance. It is less clear whether this relationship is due to asymmetric central bank preferences, however, as the sign of the correlation between inflation and its conditional variance is positive in all but one country, which is not attractive from the perspective of the theory. We interpret these results to mean that the theoretical channel by which inflation volatility affects inflation through asymmetric central bank preferences is not the correct explanation for the co-movement of inflation and its volatility at co-movements. Our conclusion is that, while it appears at first to be a promising candidate, a careful look at the data suggests that the time inconsistency story coupled with asymmetric central bank preferences and time varying variance of economic shocks is not a likely cause of the rise and fall of inflation in OECD countries in recent years. Our paper is most closely related to papers by Ruge-Murcia, 2004 and Surico, 2006 who investigate the possibility that economic volatility coupled with asymmetric central bank preferences interact to explain inflation trends. Ruge-Murcia has much the same approach as taken here, though he does not control for spurious results generated by persistence in inflation and the conditional variances and he looks at a much smaller set of countries. Surico takes a different approach to the question, estimating the degree of asymmetry in preferences from the policy reaction function for the US Federal Reserve in different periods. His research question differs from ours in that he investigates the possible effect of changes in the degree of asymmetry in preferences, rather than changes in the variances of the shocks, on inflation and finds that such changes may account for a sizable fraction of the decline in US inflation. Furthermore, he only considers US inflation. More broadly, our paper fits into the literature investigating the causes of the rise and fall of US inflation. Loosely speaking, this literature bifurcates around the question of whether the Great Inflation was caused by monetary policy errors11 or by adverse shocks.12 The monetary mistakes literature has produced a number of theories of why the US Fed may have performed poorly including misleading real time data13 problems with learning about key parameters of the economy14 and changing fundamentals, such as the NAIRU, exacerbating time inconsistency problems.15 Our paper falls into this last category, where the contribution is the use of the common international experience as a way of disciplining our empirical work, and the use of changes in economic volatility, rather than the NAIRU, as the source of the time inconsistency problem. Given the importance of asymmetric central bank preferences for the theory we are testing, our paper is also related to the empirical literatures that investigate asymmetries in central bank preferences (see Footnote 3) and monetary policy reaction functions (see Footnote 2). Our paper is only weakly related to these literatures, however, in that we do not attempt to estimate deep structural parameters, either of the central bank’s loss function or the broader macroeconomy, nor do we directly estimate the monetary policy rule. Relative to this literature, our paper represents an indirect and somewhat inefficient test of whether central bank preferences exhibit asymmetry as, for example, our statistical approach will only detect asymmetric monetary policy in the case where there are variations in economic volatility over time. For our purpose, which is to test whether this model can explain long run inflation trends, this is an advantage. Our approach, however, would not the best approach to take in order to answer questions like whether central bank preferences or monetary policy rules are asymmetric. The paper proceeds as follows: Section 2 presents a model of time inconsistent monetary policy with asymmetric central bank preferences. Section 3 introduces the spurious regression problem and presents our estimates of the effect of changing volatility of unemployment shocks on inflation, correcting for this problem. Section 4 examines the relationship between inflation and its own conditional variance, as well as the general case in which inflation may depend on the conditional variance of both inflation and unemployment. Section 5 concludes.

The results presented in this paper provide little support for the view that the interaction of asymmetric central bank preferences and changing macroeconomic volatility is an important determinant of long term inflation trends in OECD countries. The data does not support the view that the volatility of unemployment helps explains inflation trends. Estimated coefficients on measures of this volatility are generally not statistically significantly different from zero (and frequently possess the wrong sign). In the best case (the cointegration estimates in the more recent sample) the theory seems capable of explaining inflation outcomes in only one third of the countries in the sample. The results concerning the relevance of asymmetric preferences in inflation are more difficult to interpret. There is evidence of a relationship between inflation and its own volatility in more than half of the countries in the sample. However, in each of these cases, the sign of the coefficient is of the theoretically unattractive sign, suggesting that, perhaps, the asymmetric preferences hypothesis is not the correct explanation of the observed relationship. Overall, our conclusion is that while it appears promising on the surface, a careful investigation of the data does not provide much support for the view that the combination of asymmetric central bank preferences and changing economic volatility is an important driver of recent inflation trends in OECD countries.