Academic journal article Journal of Money, Credit & Banking

Inflation Targeting under Asymmetric Preferences

Academic journal article Journal of Money, Credit & Banking

Inflation Targeting under Asymmetric Preferences

Article excerpt

THIS PAPER DEVELOPS and estimates a game-theoretical model of monetary policy where the central banker's preferences are asymmetric around the targeted inflation rate. The preference specification permits different weights for positive and negative inflation deviations from the target and includes as a special case the quadratic loss function employed in previous literature. The symmetry of the quadratic form implies that the loss associated with an inflation deviation from the target depends solely on its magnitude. In contrast, under asymmetric preferences both the magnitude and sign of a deviation matter to the central banker.

Arguments in favor of the quadratic loss function include that it is tractable, yields simple analytical results, and might provide a reasonable approximation to the central banker's preferences. On the other hand, recent anecdotal and empirical evidence appears consistent with the notion of asymmetric preferences. For example, Clarida and Gertler (1997) estimate a reaction function for the Bundesbank and find that it raises the day-to-day interest rate when inflation is above its steady-state trend value but barely responds when it is below. Ruge-Murcia (2000) estimates implicit bounds for the Canadian inflation target zone using data on market-determined nominal interest rates. Results indicate that financial markets perceive the band to be of approximately the same width as announced but asymmetrically distributed around the official target.

The study of central bank preferences is framed here in a specific institutional setup. Publicly announced inflation targets have been adopted recently by several countries (for example, New Zealand, Australia, Canada, the UK, and Sweden). Under this arrangement, the central bank commits itself to gear monetary policy toward keeping a measure of inflation close to an explicit target. Following Svensson (1997), Beetsma and Jensen (1998), and Muscatelli (1999), the central banker' s loss function is defined around this target. Because the target is observable, it is possible to compare inflation realizations and the stated policy goal. This simplifies the estimation strategy and reduces the number of parameters to be estimated and means that the model generates testable empirical implications, regardless of whether the target is the socially optimal rate or not.

Previous game-theoretical models with quadratic preferences predict a positive linear relationship between inflation and unemployment, and an average rate of inflation strictly larger than the inflation target (see Green, 1996, and Svensson, 1997). In contrast, the model with asymmetric preferences predicts a positive but nonlinear relationship between inflation and unemployment, and inflation can be on average above or below the target depending on the central banker's preference parameters. To understand this result, recall that relaxing the assumption of quadratic preferences means that certainty equivalence no longer holds. Then, the expected marginal cost of departing from the inflation target is nonlinear in inflation. When the central banker associates a larger loss to positive inflation deviations from the target than to negative deviations, uncertainty raises the expected marginal cost and induces prudent behavior on the part of the central banker.

The asymmetric model also implies that the conditional variance of inflation is helpful in forecasting its mean. The asymmetry preference parameter is the coefficient of the conditional variance. Since the quadratic model corresponds to the special case where this coefficient is zero, one can test the null hypothesis of quadratic preferences against the well-defined alternative of asymmetric preferences. In an empirical application to Canada, Sweden, and the UK, results support the notion of asymmetric preferences in the form of a positive and statistically significant estimate of the asymmetry preference parameter. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.