The Empirical Economics Letters, 15(8): (August 2016) ISSN 1681 8997 Relationship between Overnight Interest Rates in the US and in Turkey: Some Evidence from Symmetric and Asymmetric Granger Causality Tests Umit Bulut * Department of Economics, Ahi Evran University Kirsehir, Turkey EkremErdem ** Department of Economics, Erciyes University, Turkey Abstract: This paper aims at examining the causal relationships between shadow rates in the US as the right measure of the stance of monetary policy of the FED and overnight interest rates in Turkey by employing monthly data from 2009:1 to 2015:11. To this end, the paper, first, conducts a unit root test with two structural breaks. Then, the paper performs symmetric and asymmetric Granger causality tests. According to the findings obtained from these causality tests, there is no causal relationship between shadow rates in the US and overnight interest rates in Turkey. Eventually, upon its findings, the paper explores that the CBRT does not directly consider shadow rates in the US while it is adjusting interest rates. Keywords: The Federal Reserve, Shadow Rate, Central Bank of the Republic of Turkey, Granger Causality Test JEL Classification Number: C32, E52, E58. 1. Introduction After the global crisis that began in September 2008, the FED first set federal funds rate near the zero lower bound. Then, the FED applied forward guidance aiming at reducing future overnight interest rate expectations of the market actors and thus affecting longterm interest rates and implemented three quantitative easing (large-scale asset purchases) programmes during the period November 2008-October 2014. The first quantitative easing programme lasted during the period November 2008-March 2010. The second quantitative easing programme began in November 2010 and finished in June 2011. Finally, the last quantitative easing programme began in September 2012 and ended in October 2014.As a result of these quantitative easing programmes, with regard to the FED data, the FED s balance sheet doubled during the period December 2008-December 2015 (from 2.23 trillion USD to 4.48 trillion USD). After the zero lower bound, forward guidance and these asset purchase programmes, the expectations that the FED would increase the interest rates * Corresponding author. Email: ubulut@ahievran.edu.tr; ** Email: ekremerdem@erciyes.edu.tr
The Empirical Economics Letters, 15(8): (August 2016) 788 in the forthcoming periods became extensive in the US, and the FED increased federal funds rate by 25 basis points in December 2015. During the period late 2008-late 2015, the federal funds rate can not capture the actions in monetary policy and thus can not present the stance of monetary policy in the US due to forward guidance policy and quantitative easing programmes. Therefore, some recent papers use the shadow federal funds rates (shadow rates) as the right measure of the stance of monetary policy in the US (Bullard, 2012; Krippner, 2013; Wu and Xia, 2014). Figure 1: Federal funds rate and shadow rate (%, last business day of the month) 6 4 2 0-2 -4 Federal funds rate Wu-Xia (2014) shadow rate Source: Wu and Xia (2014) Figure 1 shows the federal funds rate and the shadow rate developed by Wu and Xia (2014). Wu and Xia (2014) produce the shadow rate using one-month Treasury forward rates for maturities of 3 and 6 months, 1, 2, 5, 7, and 10 years by following Gurkaynak et al. (2007). After Wu and Xia (2014) construct the shadow rate, they yield that the shadow rate interacts with key macroeconomic variables, such as industrial production index, consumer price index, capacity utilization, unemployment, and housing starts, similarly as the federal funds rate did historically. Therefore, Wu and Xia (2014) remark that the shadow rate can be used instead of federal funds rate to depict the stance of monetary policy in the US. 1 As seen from Figure 1, the federal funds rate hit near the zero lower bound from late 2008 to late 2015. In this period, the shadow rate developed by Wu and Xia (2014) was negative due to the forward guidance and asset purchase programmes. After the last asset purchase programme ended, the shadow rate began to converge to the federal funds rate. In late 2015, the shadow rate was equal to the federal funds rate. To the best of the authors knowledge, there are not any papers investigating whether a central bank directly follows another central bank while it is adjusting short-term interest rates. After unconventional/extraordinary monetary policy of the FED, one may argue that 1 SeeWuandXia (2014) forthecalculationdetails of theshadow rate.
The Empirical Economics Letters, 15(8): (August 2016) 789 it is worthwhile to examine direct interactions between the FED and other central banks in the world. Based on these explanations, some questions become to be important for Turkey: Does extraordinary monetary policy of the FED have a direct impact on monetary policy of the CBRT beyond having impact on expected inflation and output?more clearly, does the CBRT directly consider shadow rates in the US while it is adjusting overnight interest rates in Turkey? Then, this paper tries to find an answer for these questions. In other words, this paper examines the causal relationships between the shadow rates in the US developed by Wu and Xia (2014) and the overnight interest rates in Turkey by using monthly data covering the period from 2009:1-2015:11. How this paper contributes to the literature lies in three points. First, to the best of the authors knowledge, in the monetary economics literature, this is the first paper examining whether a central bank directly follows another central bank while it is adjusting shortterm interest rates. Second, this paper employs causality methods to examine the causal relationships between variables. Third, this paper employs recently developed causality tests based on bootstrapping to obtain more reliable findings. The first causality test developed by Hacker and Hatemi-J (2012) lets researchers examine symmetric causal relationships between variables, and the second causality test produced by Hatemi-J (2012) lets researchers investigate asymmetric causal relationships between variables. The rest of the paper is organized as follows: Section 2 gives data. Section 3 reports estimation results. Section 4 concludes the paper. 2. Data This paper conducts a time series analysis for Turkey. The data are monthly and cover the period 2009:1-2015:11. The variables are shadow rate in the US and overnight TRlibor in Turkey. While SR represents shadow rate, INT represents overnight TRlibor. While SR is obtained from Wu and Xia (2014), INT is extracted from the Banks Association of Turkey. 3. Estimation results In this section, the paper first presents the results of the unit root test. Then, the paper presents the results of the symmetric and asymmetric causality tests. Table 1 depicts the results of the Narayan and Popp (2010) unit root test with two structural breaks. As seen from the table, the null hypothesis of a unit root can be rejected at first differences for both variables. In other words, both variables are integrated of order one with regard to the Narayan and Popp (2010) unit root test.
The Empirical Economics Letters, 15(8): (August 2016) 790 Table 1: Narayan and Popp (2010) Unit Root Test 1% Variable a Test statistic Break dates M1 M2 M1 M2 SR -0.19-0.85 2011:3, 2012:7 2011:3, 2012:7 INT -2.97-4.34 2011:6, 2013:10 2011:6, 2013:10 ΔSR -6.97 c -8.32 c ΔINT -10.04 c -9.93 c Critical -4.95-5.76 values b 5% -4.31-4.93 10% -3.98-4.60 Note: a is the first difference operator. b Critical values are obtained from Table 3 in Narayan and Popp (2010). c Illustrates 1% statistical significance. The break dates obtained from the Narayan and Popp (2010) unit root test correspond to the considerable periods for the US and Turkey. Accordingly, while the second quantitative easing programme of the FED may account for the break in March 2011, the break in July 2012 corresponds to the period before the third quantitative easing programme of the FED in the US. Besides, the first break in June 2011 corresponds to the period when the debt crisis in the Euro Area began and the national currency of Turkey (TL) depreciated while the rise in inflation and the depreciation of the TL may account for the break in October 2013 for Turkey. Table 2 reports the Hacker and Hatemi-J (2012) symmetric causality test. Accordingly, the null hypothesis of no causal relationship running from SR to INT can not be rejected. Besides, the null hypothesis of no causal relationship running from INT to SR can not be rejected, too. In other words, the findings of the Hacker and Hatemi-J (2012) symmetric causality test point out that there are no causal relationships between INT and SR. Table 2: Hacker and Hatemi-J (2012) Symmetric Causality Test a Null hypothesis Test statistic Critical values b 1% 5% 10% SR does not Granger cause INT 0.22 7.44 4.02 2.86 INT does not Granger cause SR 1.12 7.18 4.11 2.81 Note: a Maximum lag length is 4, and the Hatemi-J criterion (HJC) that is developed by Hatemi-J (2003, 2008) is used to determine the optimal lag length. b Critical values are obtained through 10000 bootstrap replications. Table 3 presents results of the Hatemi-J (2012) asymmetric causality test. According to the findings, the null hypothesis that a positive (negative) shock in SR does not Granger cause
The Empirical Economics Letters, 15(8): (August 2016) 791 a positive (negative) shock in INT can not be rejected in the same way the null hypothesis that a negative (positive) shock in SR does not Granger cause a positive (negative) shock in INT can not be rejected. Besides, the null hypothesis that a positive (negative) shock in INT does not Granger cause a positive (negative) shock in SR can not be rejected in the same way the null hypothesis that a negative (positive) shock in INT does not Granger cause a positive (negative) shock in SR can not be rejected. That is to say, the findings obtained from the Hatemi-J (2012) asymmetric causality test indicate that there are no causal relationships between INT and SR. Table 3: Hatemi-J (2012) Asymmetric Causality Test a Null hypothesis Test statistic Critical values b 1% 5% 10% SR + does not Granger cause INT + 0.02 19.08 4.94 2.47 SR - does not Granger cause INT - 0.11 8.18 4.21 2.86 SR - does not Granger cause INT + 0.04 8.45 4.39 2.91 SR + does not Granger cause INT - 0.03 8.03 4.26 2.87 INT + does not Granger cause SR + 0.12 18.98 4.92 2.47 INT - does not Granger cause SR - 0.88 7.86 4.27 2.95 INT - does not Granger cause SR + 0.82 15.54 5.58 3.01 INT + does not Granger cause SR - 0.03 10.98 4.28 2.71 Note: a Maximum lag length is 4, and the HJC is used to determine the optimal lag length. b Critical values are obtained through 10000 bootstrap replications. Both symmetric and asymmetric causality tests indicate that extraordinary monetary policy of the FED does not have a direct impact on monetary policy of the CBRT beyond having impact on expected inflation and output. In other words, the findings of the causality tests denote that the CBRT does not directly consider shadow rates in the US while it is adjusting overnight interest rates in Turkey. 4. Conclusion Beginning with the global financial crisis, the FED set federal funds rate near the zero lower bound, applied forward guidance and implemented three quantitative easing programmes during late 2008-October 2014. Since federal funds rate can not present the stance of monetary policy of the FED because of forward guidance policy and the quantitative easing programmes, some papers produce the shadow rates as the right measure of the monetary policy stance in the US. This paper investigates the causal relationships between shadow rates in the US suggested by Wu and Xia (2014) and overnight interest rates in Turkey using monthly data from 2009:1-2015:11. After
The Empirical Economics Letters, 15(8): (August 2016) 792 conducting the Narayan and Popp (2010) unit root test, the paper employs the Hacker and Hatemi-J (2012) symmetric Granger causality test and the Hatemi-J (2012) asymmetric Granger causality test. Both tests indicate that there is no causal relationship between shadow rates in the US and overnight interest rates in Turkey. Based on the empirical findings, the paper yields that extraordinary monetary policy of the FED based on forward guidance and quantitative easing programmes does not have a direct impact on monetary policy of the CBRT. In other words, the paper explores that the CBRT does not directly consider shadow rates in the US while it is adjusting overnight interest rates in Turkey. References Banks Association of Turkey, https://www.tbb.org.tr/, Accessed on May 4, 2016. Bullard, J., 2012, Shadow Interest Rates and the Stance of US Monetary Policy. https://www.stlouisfed.org/~/media/files/pdfs/bullard/remarks/bullard_cfar_stlouis8 Nov2012final.pdf, Accessed on November 10, 2015. Federal Reserve, http://www.federalreserve.gov/, Accessed on May 4, 2016. Gurkaynak, R.S., B. Sack and J.H. Wright, 2007, The U.S. Treasury Yield Curve: 1961 to the Present, Journal of Monetary Economics, 54, 2291-2304. Hacker, S. and A. Hatemi-J, 2012, A Bootstrap Test for Causality with Endogenous Lag Length Choice: Theory and Application in Finance, Journal of Economic Studies, 39, 144-160. Hatemi-J, A., 2003, A New Method to Choose Optimal Lag Order in Stable and Unstable VAR Models, Applied Economics Letters, 103, 135-137. Hatemi-J, A., 2008, Forecasting Properties of a New Method to Determine Optimal LagOrder in Stable and Unstable VAR models, Applied Economics Letters, 15,239-43. Hatemi-J, A., 2012, Asymmetric Causality Tests with an Application, Empirical Economics, 43, 447-456. Krippner, L., 2013, Measuring the Stance of Monetary Policy in Zero Lower Bound Environments, Economics Letters, 118, 135-138. Narayan, P.K. and S. Popp, 2010, A New Unit Root Test with Two Structural Breaks in Level and Slope at Unknown Time, Journal of Applied Statistics, 37, 1425-1438. Wu, J.C. and Xia, F.D., 2014, Measuring the Macroeconomic Impact of Monetary Policy at the Zero Lower Bound, NBER Working Paper No. 20117.