The effectiveness of Central Bank intervention: evidence from Turkey

Abstract

This article investigates the effectiveness of Central Bank intervention on the conditional variance and the mean of the exchange rate returns in Turkey during the float period. The daily exchange rates are studied within an Exponential General Autoregressive Conditional Heteroskedastic (EGARCH) framework. Little evidence is found for the effectiveness of intervention operations. Empirical results suggest that foreign exchange (FX) selling auctions increase exchange rate volatility. However, a reverse causality relationship detected between one-day past sales of FX auctions and exchange rate returns which is supportive of leaning-against-the-wind behaviour of the Central Bank contradicts its announcements. Also the Central Bank of Turkey (CBT) tends to intervene through FX selling auctions when one-day past volatility is higher.

Notes

¹ See Dominguez (1998), Baillie and Osterberg (1997) and Beine et al. (2002).

² See Dominguez and Frankel (1993) for detailed explanation.

³ See Dominguez (1998) for detailed explanation.

⁴ Dominguez (1998) discusses the possible scenarios of the effects of intervention transactions on the conditional mean and variance of the exchange rates in detail depending on the type of commitment of Central Banks and the quality of signals.

⁵ The Turkish Central Bank Press Announcement, 2001-8, 22.2.2001.

⁶ Ibid., 2002-1, 2.1.2002.

⁷ These banks were extended loans in collateral of the State Domestic Borrowing Securities (DİBS) by the Central Bank.

⁸ The Turkish Central Bank Press Announcement, 2004-2, 2.1.2004.

⁹ Ibid., 2002-10, 11.2.2002.

¹⁰ Ibid., 2001-14, 13.4.2001.

¹¹ Ibid., 2001-12, 29.3.2001, 2001-20, 17.5.2001, 2001-32, 11.7.2001, 2001-67, 1.11.2001.

¹² This value was merely $5.4 billions in 2004.

¹³ Ibid., 2003-52, 10.9.2003.

¹⁴ Ibid., 2003-64, 22.10.2003.

¹⁵ According to the press announcements of the CBT, the appreciation of the Turkish lira is due to the reversal of the expectations to a positive perspective about the Turkish economy which led people to gain confidence in the Turkish lira raising the demand to hold securities nominated in the domestic currency.

¹⁶ Hsieh (1988, 1989) and Diebold and Nerlove (1989).

¹⁷ The choice is based on Akaike's Information Criterion (AIC), Bayesian Information Criterion (BIC) and Ljung–Box Q* statistics for the conditional mean equation when modelling the restricted model without intervention and policy variables.

¹⁸ Other dummy variables for the week days are found to be statistically insignificant.

¹⁹ The amounts of intervention transactions are announced by the CBT on the first business day followed after 3 months of the actual intervention date.

²⁰ Under the null hypothesis, the Q-statistics asymptotically follow a chi-square distribution with k degrees of freedom.

²¹ The unit root test results are not shown here to save space.

²² Intervention analysis in the literature usually faces with endogeneity bias as Central Banks may actually be intervening into the foreign exchange market due to the volatility of the exchange rates and to influence the level of its currency.

²³ This means the Central Bank attempts to reverse the direction of the trend movement of its currency.

²⁴ The Turkish Central Bank Press Announcements, 2001-33, 16.7.2001.

²⁵ Funabashi (1989) computed this target rate in his study of central bankers estimation for the nominal exchange rates.

²⁶ The explanatory variables could cause multicollinearity problem but their correlations are found to be very low.

²⁷ Other specifications with deviations from 5 and 10 day moving averages of both explanatory variables are also estimated. However, they were insignificant and not reported.