Optimized Taylor rules with domestic bond yields in emerging market economies

Abstract

We study optimized Taylor rules with the appropriate lag structure, which has been little done for emerging market economies. Setting the policy interest rate in response to movements in domestic long-term bond yields, in addition to the output gap, the inflation gap and the exchange rate, can make monetary policy more effective. But a more complex rule can reduce monetary policy effectiveness, notably in the presence of uncertainty about the effects of capital flows on domestic monetary conditions.

Keywords:

• emerging market economies
• Taylor rule
• monetary policy
• exchange rate
• bond yields

JEL Classification:

• E43
• E52
• E58
• F31
• F36
• F42

Acknowledgements

This article was prepared with assistance from Tracy Chan, Sonja Fritz, Emese Kuruc and Lillie Lam. Comments from Claudio Borio, Andrew Filardo, Madhusudan Mohanty, Kostas Tsatsaronis, Philip Turner, James Yetman and BIS seminar and meeting participants are gratefully acknowledged.

The views expressed in this publication are those of the authors and not necessarily the views of the BIS. A previous version of this article was published as “How might EME central banks respond to the influence of global monetary factors?” in BIS Papers No. 78 “The transmission of unconventional monetary policy to the emerging markets” (2014).

Notes

¹ Equation (3) captures the notion that exchange rate dynamics are difficult to predict using macroeconomic variables such as output and inflation. The exchange rate is assumed to respond to signals created by policy rate actions about the future change in bond yields (Eichengreen, 2002). We estimated the parameters in a panel regression with fixed effects on a set of 14 EMEs using quarterly data for 2000–2007: Brazil, Chile, Czech Republic, Hungary, India, Indonesia, Israel, Malaysia, Mexico, Philippines, Poland, South Africa, Thailand and Turkey. We set α₁₁ = 0.60, α₁₃ = 0.05, α₁₄ = −0.10, α₁₅ = −0.20, α₂₁ = 0.15, α₂₂ = 0.75, α₂₃ = 0.10, α₃₃ = 0.85, α₃₅ = −0.50, α₄₁ = 0.10, α₄₃ = 0.10, α₄₄ = 0.70, α₄₅ = 0.20 (Gadanecz et al. 2014).

² In practice, when the central bank augments its reaction function with additional variables, the parameters of the underlying economy may change. Our results should be interpreted with this Lucas critique in mind.

³ It is not required to get a loss reduction significant at the 95% confidence level when moving from R3(1,1) to R3(1,2) because both models have the same complexity (in terms of the number of parameters to estimate). Alternative simulations reported in the Appendix show that more complex rules, which react to the sum of lags of each variable, rather than, as above, to one specific lag, do not improve monetary policy effectiveness further.

⁴ The coefficient is raised by 0.2, which corresponds to one side of the 95% confidence interval of the parameter estimates in Equations (1)–(4).