Monetary policy rules in practice: evidence from Turkey and Israel

Abstract

Forward looking monetary policy rules are estimated for Israel and Turkey. When variable inflation targets are taken into consideration, as opposed to the fixed targets used in prior research that use data from developed countries, forward looking Taylor rules seem to provide reasonable description of Central Bank behaviour in both countries. In general, it can be said that monetary policy appears to be quite strong in these countries, and especially so in Turkey, when compared with developed countries.

Notes

¹ This terminology belongs to Svensson (2003).

² See Clarida et al . (1999). Berument and Tasci (2004) estimates monetary policy rules for Turkey over the period from 1990 to 2001. They adopt the framework of CGG. This approach can be criticized on two grounds. As stated above, targets are treated as fixed and inflation targeting was adopted as a policy at the beginning of 2001.

³ It should be noted that r is an ‘approximate’ real rate since forecast horizon for inflation will generally differ from the maturity of the short-term nominal rate used as a monetary policy instrument. As noted by Clarida et al . (2000), in practice the presence of high correlation between short-term rates and at maturities associated with the target horizon (1 year) prevents this from being a problem.

⁴Clarida et al . (2001, 2002) indicates that openness only affects the magnitude of the coefficients in the policy rule.

⁵ It is assumed that is identically and independently distributed.

⁶ These results are available upon request.

⁷ The usual Two Stage Least Square estimators yield exactly the same results. For LIML estimators, one modifies the GAUSS code originally used by Stock and Wright (2000).

⁸ For the sake of the comparison that we make between Turkey and Israel, it is important to choose the instruments the same for both countries. By choosing these instruments, one implicitly assumes that these two variables are strong instruments for predicting output gap.

⁹ As suggested by Kleibergen (2002), the AR-test and the K-test statistics are calculated by interpreting all data matrices in the test as residuals from the projection on exogenous variables.

¹⁰ In this case, the estimation method is nonlinear LIML with continuous updating algorithm. It is again important to note that, for the sake of comparison between Turkish and Israeli data, the set of instruments is the same for all of the equations estimated, for both countries.