Bulletin of Monetary Economics and Banking, Vol. 22, No. 3 (2019), pp. 311 - 350

p-ISSN: 1410 8046, e-ISSN: 2460 9196

EVIDENCE ON MONETARY POLICY TRANSMISSION DURING TRANQUIL AND TURBULENT PERIODS*

Chioma Peace Nwosu**, Afees A. Salisu***, Margaret Johnson Hilili****, Izuchukwu

Ifeanyi Okafor*****, Izuchukwu Oji-Okoro******, Idis Adediran*******

**Research Department, Central Bank of Nigeria, Abuja, Nigeria. Email: cpnwosu@cbn.gov.ng

***Department for Management of Science and Technology Development, Ton Duc Thang University, Ho Chi Minh City, Vietnam ; Faculty of Business Administration, Ton Duc Thang University, Ho Chi Minh City, Vietnam; Centre for Econometric & Allied Research,

University of Ibadan, Ibadan, Nigeria. Email: afees.adebare.salisu@tdtu.edu.vn

****Research Department, Central Bank of Nigeria, Abuja, Nigeria. Email: mjhilili@cbn.gov.ng

*****Research Department, Central Bank of Nigeria, Abuja, Nigeria. Email: iiokafor@cbn.gov.ng

******Research Department, Central Bank of Nigeria, Abuja, Nigeria. Email: ioji-okoro@cbn.gov.ngd

*******Centre for Econometric & Allied Research, University of Ibadan, Ibadan, Nigeria.

ABSTRACT

This paper evaluates monetary policy transmission in both tranquil and turbulent periods for Mexico, Indonesia, Nigeria, and Turkey. Using a structural vector autoregressive model, we find that the effect of structural shocks from supply, demand, and financial sources tend to fizzle out faster for Nigeria and Mexico compared to Indonesia and Turkey. Another important finding is that while monetary authorities in Indonesia and Turkey are more responsive to inflation those in Mexico and Nigeria are more influenced by the exchange rate. We also observe differences in the conduct of monetary policy between the tranquil and turbulent periods.

Keywords: Monetary policy transmission; Structural Vector Autoregressive model;

MINT countries; Global financial crisis.
JEL Classification: E44; G21.
Article history:
Received	: July 23, 2019
Revised	: September 23, 2019
Accepted	: September 26, 2019

Available online : October 15, 2019 https://doi.org/ 10.21098/bemp.v22i3.1111

*The authors thank the anonymous reviewers and the managing editor – Professor Paresh Kumar Narayan for their useful comments. The technical support received from the econometric workshops of the Central Bank of Nigeria is graciously acknowledged.

Evidence on Monetary Policy Transmission During Tranquil and Turbulent Periods	313

Following the motivation for this study, the remainder of the paper is organized as follows. Section II discusses the data and methodology. Section III discusses the results. Section IV checks the robustness of the results. Section V concludes the study.

II. METHODOLOGY AND DATA

We construct a structural VAR-X model with four endogenous variables for the interest rate channel of monetary policy transmission.2 We formerly argued that the interest rate channel is theoretically implied in the Taylor rule of monetary policy. In this framework, central banks take account of the level of economic activity and prices in their objective functions while setting their policy rates.3 Thus, to reflect the international macroeconomy, the ensuing empirical model incorporates output growth, inflation, the exchange rate, and the interest rate as endogenous variables (Taylor, 2001). Our motivation also implies that the monetary policy transmission mechanism is sensitive to structural shifts, such as the effects of the GFC, as well as global financial market fluctuations. Hence, we account for global oil prices and the global (US) interest rate in the interest rate equation. Thus, we justify a distinct analysis for the periods before and after the GFC and account for other structural changes posed by seasonal effects in the data.

We first specify the following standard VAR model as the baseline for the SVAR model:

	(1)
where	is a 4×1 vector of endogenous variables (growth rate of

industrial output, inflation, log differences in the exchange rate, and changes in

the nominal interest rate); is a matrix of contemporaneous effects;	is a 4×1
vector of constants; is a 4×4 matrix of coefficients of lagged series for all		;

is a vector of disturbances or structural innovation, serially uncorrelated, with a mean of zero; and p is the optimal lag, obtained using the Schwartz Information Criterion (SIC). The variables are transformed so as to circumvent nonstationarity in their level series.

Since the analysis of the structural VAR requires that the model be identified,

we impose restrictions that satisfy the condition and a recursive approach. The study assumes a simple recursive structure of the model to identify structural shocks. The recursive restrictions derive from the intuition that:

•The interest rate does not have an immediate impact on the output gap.

•There is no contemporaneous relation between current interest rates and prices (inflation and exchange rate).

•The exchange rate does not have a direct impact on aggregate demand; in other words, there is no contemporaneous relation between the exchange rate and output.

2The work is limited to the interest rate channel for brevity, and the analyses can therefore be replicated for other channels for potential generalization of the research findings.

3Other channels of monetary policy transmission (e.g., the credit channel, the balance sheet channel, the asset price channel) have been noted, and strong motivation has been established for the interest rate channel.

314	Bulletin of Monetary Economics and Banking, Volume 22, Number 3, 2019

•There is an interest rate rule (i.e., the Taylor rule) that states that central banks take into account current output growth and inflation in setting monetary policy rates.

•Central banks adopt an active stabilizing role in the foreign exchange markets.

From the foregoing, the matrix of contemporaneous effect ( ) is defined

in Table 1. The standard SVAR model described in the analysis is augmented, based on the motivation, by some inherent effects (seasonal effects and structural changes) in the data sets, and the reconstructed model is described as an SVAR-X model. We demonstrate that including these effects in the analysis of the SVAR model can offer useful insights for policy decision purposes.

Table 1.

Restrictions on the Matrix

This table shows the recursive restrictions imposed on the matrix of contemporaneous effects. gt, πt, rt and Δi represent industrial production growth, inflation, exchange rate returns and changes in interest rate, respectively.

Endogenous Variable		Contemporaneous Variable
*		0	0	0
1		*	0	0
1		1	*	0
1		1	1	*

To estimate the standard SVAR model, we specify the reduced-form of equation (1) as

		(2)
where	and	, with constants

suppressed for notational convenience. Having justified the recursive restrictions theoretically with the Taylor rule, we see the reduced-form residuals could follow a Cholesky decomposition-type orthogonalization. Thus, the matrix of initial impulse responses can be constructed as follows:

(3)

The shocks in the SVAR system are called the supply shock () emanating from the output, demand shock () is reflected in the general price level, financial shocks () are transmitted via the exchange rate, and interest rate shock () arises from monetary policy actions. In addition to the Taylor rule, the intuition for the

Evidence on Monetary Policy Transmission During Tranquil and Turbulent Periods	315

absence of contemporaneous effects from the output or the price on the interest rate consists of the operational rigidities in the production process that create a lag between policy pronouncements and production. On the other hand, demand and supply shocks will evoke unanticipated policy responses from the monetary policy authority to remedy their long-term consequences on the economy.

The paper employs monthly data series for the Industrial Production Index (IPI), the Consumer Price Index (CPI), the Nominal Effective Exchange Rate (NEER), and the Interest Rate (INTR) covering the period from January 2000 to December 2018 for all the MINT countries, except Nigeria, where quarterly data spanning 2000Q1 to 2018Q4 were employed. The data for the other MINT countries were sourced from the International Financial Statistics database, and the data for Nigeria were sourced from the National Bureau of Statistics and the Central Bank of Nigeria databases. The data for the global factors, namely, the oil prices and US interest rates, were obtained from the Federal Reserve database.

III. RESULTS AND DISCUSSION

We attempt to briefly describe the variables of interest in Table 2 and Figures 1 through 3. In the data description, all four MINT countries record average interest rates sufficiently larger than that of the United States (1.95%), with 7.28% for Mexico, 14.02% for Indonesia, 18.17% for Nigeria, and the lowest, 4.81%, for Turkey. Notably, average inflation appears to be the highest in Nigeria (2.99%), where the rate of interest is also the highest, although the reverse cannot be inferred. On the other hand, the IPIs and the nominal effective exchange rates of the countries appear to be close. Figures 1 to 3 show that the variables are associated with the interest rate over time. Evaluation of the time series properties reveals that they are all stationary in the level expressed (for details of the transformation of the series, see the notes for Table 2). These findings are exhaustive, in that the results of both the conventional augmented Dickey–Fuller tests and the generalized autoregressive conditional heteroskedasticity-based Narayan–Liu unit root tests corroborate each other.

Table 2.

Preliminary Tests

The variables INT, IPI, INF NEER, US_INT, WTI and BRENT denote domestic interest rate, industrial production index, inflation, nominal effective exchange rate, US interest rate, West Texas Intermediate and Brent oil prices. Inflation is computed as log in differences of consumer prices (CPI) as follows: 100*Δlog(CPI). The descriptive statistics of the series (mean, standard deviation (SD) and Jarque-Bera (JB) statistics) are computed on the level forms of the series. The unit root tests (Augmented Dickey Fuller [ADF] and Narayan & Liu [NL] (2015) tests) are conducted on the growth rate of IPI and returns series of NEER, WTI and BRENT, the differenced series of INTR and US_INTR computed as: Δ(INTR) and Δ(US_INTR). Superscripts a & b represent 1% and 5% significance levels, respectively. Finally, * and ^ represent the inclusion or exclusion of trend component in the underlying unit root test equation, respectively.

					Description
Variables			Mexico					Indonesia
	Mean	SD	JB	ADF	NL	Mean	SD	JB	ADF	NL
INTR	7.278	3.33	160.4a	-12.9^	-15.8*	14.02	2.56	23.0a	-5.5^	-9.1*
IPI	102.4	6.50	11.6a	-8.24^	-27.4^	118.2	13.0	1.57	-21.4^	-43.2*
INF	0.365	0.34	12.8a	-10.4*	-12.9^	0.544	0.74	27947a	-12.3*	-16.1*
NEER	106.8	23.6	6.6b	-11.9^	-14.3^	98.19	18.5	9.46	-11.1^	-12.6*

316Bulletin of Monetary Economics and Banking, Volume 22, Number 3, 2019

Table 2.

Preliminary Tests (Continued)

			Nigeria					Turkey
	Mean	SD	JB	ADF	NL	Mean	SD	JB	ADF	NL
INTR	18.17	2.56	38.3a	-9.2^	-19.9^	4.81	1.90	46.4a	-6.42*	-75.1^
IPI	107.6	10.6	6.33b	-10.3^	-362*	106.2	34.6	12.5a	-11.5^	-22.4*
INF	2.993	2.26	15.4a	-8.2*	-7.6*	1.107	1.36	903a	-3.12^	-10.8*
NEER	109.4	32.9	1.80	-8.0^	-8.9*	105.0	57.3	529a	-10.0^	-12.7^
		Global factors
	Mean	SD	JB	ADF	NL
US_INT	1.952	1.96	37.7a	-10.2*	-18.1^
WTI	62.16	26.7	11.3a	-11.1*	-10.9*
BRENT	64.73	30.6	15.3a	-12.0*	-11.1^

Figure 1. Trends in Industrial Production Index and Interest Rate in the MINT

Countries, 2000-2018

The plots are arranged in the order of MINT countries from the top left to down right. The MINT acronym denotes Mexico, Indonesia, Nigeria and Turkey.

Trends in Industrial Production Index and Interest Rate in Mexico, 2000M1 to 2018M12
Industrial Production Index	Interest Rate

				IPI	INTR				20
									16
									12
									8
115									4
									0
110
105
100
95
90
85
2000	2002	2004	2006	2008	2010	2012	2014	2016	2018
					Period

Evidence on Monetary Policy Transmission During Tranquil and Turbulent Periods	317

Figure 1. Trends in Industrial Production Index and Interest Rate in the MINT

Countries, 2000-2018 (Continued)

Trends in Industrial Production Index and Interest Rate in Indonesia, 2000M1 to 2018M12

Industrial Production Index				IPI	INTR			Interest Rate		22
										20
										18
										16
										14
160										12
										10
140
120
100
80
60
2000	2002	2004	2006	2008	2010	2012	2014	2016	2018
					Period

	Trends in Industrial Production Index and Interest Rate in Nigeria, 2000Q1 to 2018Q4
Industrial Production Index								Interest Rate		28
				IPI	INTR					24
										20
130										16
120										12
110
100
90
80
2000	2002	2004	2006	2008	2010	2012	2014	2016	2018
					Period

318Bulletin of Monetary Economics and Banking, Volume 22, Number 3, 2019

Figure 1. Trends in Industrial Production Index and Interest Rate in the MINT

Countries, 2000-2018 (Continued)

Trends in Industrial Production Index and Interest Rate in Turkey, 2000M1 to 2018M12

Industrial Production Index

Interest Rate

				IPI	INTR				10
									8
									6
200									4
160									2
120
80
40
2000	2002	2004	2006	2008	2010	2012	2014	2016	2018
					Period

Figure 2. Trends in Inflation and Interest Rate in the MINT Countries, 2000-2018

The plots are arranged in the order of MINT countries from the top left to down right. The MINT acronym denotes Mexico, Indonesia, Nigeria and Turkey.

Trends in Inflation and Interest Rate in Mexico, 2000M1 to 2018M12

Inflation								Interest Rate
				INF	INTR				20
									16
									12
2,0									8
1,5									4
1,0									0
0,5
0,0
-0,5
-1,0
2000	2002	2004	2006	2008	2010	2012	2014	2016	2018
					Period

Evidence on Monetary Policy Transmission During Tranquil and Turbulent Periods	319

Figure 2. Trends in Inflation and Interest Rate in the MINT Countries, 2000-2018

(Continued)

Trends in Inflation and Interest Rate in Indonesia, 2000M1 to 2018M12

Inflation								Interest Rate
				INF	INTR				22
									20
									18
10									16
8									14
6									12
4									10
2
0
-2
2000	2002	2004	2006	2008	2010	2012	2014	2016	2018
					Period
		Trends in Inflation and Interest Rate in Nigeria, 2000Q1 to 2018Q4

Inflation								Interest Rate
				INF	INTR				28
									24
12									20
8									16
4									12
0
-4
-8
2000	2002	2004	2006	2008	2010	2012	2014	2016	2018
					Period

320	Bulletin of Monetary Economics and Banking, Volume 22, Number 3, 2019

Figure 2. Trends in Inflation and Interest Rate in the MINT Countries, 2000-2018

(Continued)

		Trends in Inflation and Interest Rate in Turkey, 2000M1 to 2018M12
Inflation								Interest Rate
				INF	INTR				10
									8
10
8									6
6									4
4
2									2
0
-2
2000	2002	2004	2006	2008	2010	2012	2014	2016	2018
					Period

Figure 3. Trends in Exchange Rate and Interest Rate in the MINT Countries,

2000-2018

The plots are arranged in the order of MINT countries from the top left to down right. The MINT acronym denotes Mexico, Indonesia, Nigeria and Turkey.

Trends in Exchange Rate and Interest Rate in Mexico, 2000M1 to 2018M12

Exchange Rate								Interest Rate
				NEER		INTR			20
									16
									12
160									8
140									4
120									0
100
80
60
2000	2002	2004	2006	2008	2010	2012	2014	2016	2018
					Period

322	Bulletin of Monetary Economics and Banking, Volume 22, Number 3, 2019

Figure 3. Trends in Exchange Rate and Interest Rate in the MINT Countries,

2000-2018 (Contined)

Trends in Exchange Rate and Interest Rate in Turkey, 2000M1 to 2018M12

Exchange Rate	Interest Rate
	10
NEER	INTR
400
	8
300
	6
200
	4
100

2

0

2000

2002

2004

2006

2008

2010

2012

2014

2016

2018

Period

As argued previously, we explore an SVAR-X model that accounts for seasonal effects, with distinct analyses for the periods before and after the GFC. The underlying optimal VAR model for the SVAR-X is determined using the information provided by the SIC for selection among competing models. The SIC selects an optimal lag of one for all the models across the four countries. The resulting model is evaluated for the coefficients of the contemporaneous effects, impulse responses, and variance decomposition functions. In the analysis of the contemporaneous effects (see Table 3), the relevant coefficients are C3, C5, and C6, representing the immediate effects of the IPI and inflation and exchange rates, respectively, on the interest rate. The impulse response graphs are constructed such that the responses lie within +2 standard deviations. The forecast horizons for the shock analysis are set at 10, but the monetary policy variance decompositions for periods 1, 5, and 10 are selected for a summary. The results are highlighted as follows.

324	Bulletin of Monetary Economics and Banking, Volume 22, Number 3, 2019

For all the countries, the supply shock (i.e., the C3 coefficient) exerts positive effects on the interest rate in the calm period (i.e., pre-GFC). The results are consistent with economic theory, although the coefficient proves to be rather nonsignificant, except in Turkey. However, a sign reversal takes place for Mexico and Nigeria, whereas the weight of the negative impacts is greater in Nigeria, affecting the full sample. Hence, although a coefficient variation exists across periods for Mexico and Nigeria, it is absent for Turkey and Indonesia. A plausible reason for its absence in Turkey and Indonesia is that, in turbulent times, the monetary authorities tend to use unconventional measures that might not reflect a priori expectations. We also observe contrasting results for the two data samples in terms of the contemporaneous effect of inflation on the interest rate (i.e., the C5 coefficient) for Mexico, Nigeria, and Turkey, but not Indonesia. Given the empirical results, the monetary policy authorities in Mexico, Nigeria, and Turkey tend to respond to the prevailing macroeconomic dynamics, as evidenced in the contrasting signs between the pre- and post- GFC periods. However, such a contrast between the pre- and post GFC periods is not observed for the immediate impact of financial shock (the C6 coefficient) on monetary policy across the countries.

Beyond the contemporaneous relations, we employ impulse response functions to show the responses of the interest rate to the three macroeconomic shocks, that is, supply, demand, and financial shocks. By design, impulse response functions also determine whether the effects of these shocks are transient or permanent. The impulse response functions in Figures 4a to 4d are arranged in rows and columns such that the rows show the results of the analysis for the pre-GFC period, the post-GFC period, and the full sample, while the columns shows the responses of the interest rate to supply, demand, and financial shocks respectively. Hence, the figure in the ith row and jth column represents the response of monetary policy to the jth macroeconomic shock for the ith data sample.

The analysis shows that all three shocks are transient in the system, although it took different time horizons for this to be achieved in each country. The transitory nature of the shocks is shown in Figures 4a to 4d by the blue line in between the +2 standard deviations (upper and lower dotted red lines) being tangential to the horizontal line at the specified period. Just as the findings revealed under the contemporaneous effects, the plots show evidence of contrasting relations, shown by the impulse response functions for pre- and post-GFC (in support of Jannsen et al., 2015; Mishkin, 2017; Egea and Hierro, 2019). However, the clear exception is Indonesia, for which no such effect is observed across the data samples or across the macroeconomic shocks considered. Further, Indonesia and Turkey take the longest time for the blue line to meet the zero line. Hence, the impact of macroeconomic shocks dies out more quickly in Nigeria and Mexico, whether we consider turbulent or calm periods.

Evidence on Monetary Policy Transmission During Tranquil and Turbulent Periods	325

Figure 4a. Impulse Response Functions (Mexico)

The plots are arranged such that rows represent analysis for pre-GFC, post-GFC and full sample. The columns represent the impulse responses of interest rate to IPI, inflation and exchange rate respectively.

Response of D(INTR) to 100*DLOG(IPI)

.8
.4
.0
-.4
1	2	3	4	5	6	7	8	9	10

Response of D(INTR) to 100*DLOG(CPI)

.8
.4
.0
-.4
1	2	3	4	5	6	7	8	9	10

326	Bulletin of Monetary Economics and Banking, Volume 22, Number 3, 2019

Figure 4a. Impulse Response Functions (Mexico) (Continued)

Response of D(INTR) to 100*DLOG(NEER)

.8
.4
.0
-.4
1	2	3	4	5	6	7	8	9	10

.3			Response of D(INTR) to 100*DLOG(IPI)
.2
.1
.0
-.1
1	2	3	4	5	6	7	8	9	10

Evidence on Monetary Policy Transmission During Tranquil and Turbulent Periods	327

Figure 4a. Impulse Response Functions (Mexico) (Continued)

.3			Response of D(INTR) to 100*DLOG(CPI)
.2
.1
.0
-.1
1	2	3	4	5	6	7	8	9	10

Response of D(INTR) to 100*DLOG(NEER)

.3
.2
.1
.0
-.1
1	2	3	4	5	6	7	8	9	10

328	Bulletin of Monetary Economics and Banking, Volume 22, Number 3, 2019

Figure 4a. Impulse Response Functions (Mexico) (Continued)

Response of D(INTR) to 100*DLOG(IPI)

.6
.4
.2
.0
-.2
1	2	3	4	5	6	7	8	9	10

Response of D(INTR) to 100*DLOG(CPI)

.6
.4
.2
.0
-.2
1	2	3	4	5	6	7	8	9	10

Evidence on Monetary Policy Transmission During Tranquil and Turbulent Periods	329

Figure 4a. Impulse Response Functions (Mexico) (Continued)

Response of D(INTR) to 100*DLOG(NEER)

.6
.4
.2
.0
-.2
1	2	3	4	5	6	7	8	9	10

Figure 4b. Impulse Response Functions (Indonesia)

The plots are arranged such that rows represent analysis for pre-GFC, post-GFC and full sample. The columns represent the impulse responses of interest rate to IPI, inflation and exchange rate respectively.

Response of D(INTR) to 100*DLOG(IPI)

.20
.15
.10
.05
.00
-.05
1	2	3	4	5	6	7	8	9	10

330	Bulletin of Monetary Economics and Banking, Volume 22, Number 3, 2019

Figure 4b. Impulse Response Functions (Indonesia) (Continued)

Response of D(INTR) to 100*DLOG(CPI)

.20
.15
.10
.05
.00
-.05
1	2	3	4	5	6	7	8	9	10

Response of D(INTR) to 100*DLOG(NEER)

.20
.15
.10
.05
.00
-.05
1	2	3	4	5	6	7	8	9	10

Evidence on Monetary Policy Transmission During Tranquil and Turbulent Periods	331

Figure 4b. Impulse Response Functions (Indonesia) (Continued)

Response of D(INTR) to 100*DLOG(IPI)

.08
.04
.00
1	2	3	4	5	6	7	8	9	10

Response of D(INTR) to 100*DLOG(CPI)

.08
.04
.00
1	2	3	4	5	6	7	8	9	10

332	Bulletin of Monetary Economics and Banking, Volume 22, Number 3, 2019

Figure 4b. Impulse Response Functions (Indonesia) (Continued)

Response of D(INTR) to 100*DLOG(NEER)

.08
.04
.00
1	2	3	4	5	6	7	8	9	10

Response of D(INTR) to 100*DLOG(IPI)

.12
.08
.04
.00
1	2	3	4	5	6	7	8	9	10

Evidence on Monetary Policy Transmission During Tranquil and Turbulent Periods	333

Figure 4b. Impulse Response Functions (Indonesia) (Continued)

Response of D(INTR) to 100*DLOG(CPI)

.12
.08
.04
.00
1	2	3	4	5	6	7	8	9	10

Response of D(INTR) to 100*DLOG(NEER)

.12
.08
.04
.00
1	2	3	4	5	6	7	8	9	10

334	Bulletin of Monetary Economics and Banking, Volume 22, Number 3, 2019

Figure 4c. Impulse Response Functions (Nigeria)

The plots are arranged such that rows represent analysis for pre-GFC, post-GFC and full sample. The columns represent the impulse responses of interest rate to IPI, inflation and exchange rate respectively.

Response of D(INTR) to 100*DLOG(IPI)

1.0
0.5
0.0
-0.5
1	2	3	4	5	6	7	8	9	10

Response of D(INTR) to 100*DLOG(CPI)

1.0
0.5
0.0
-0.5
1	2	3	4	5	6	7	8	9	10

Evidence on Monetary Policy Transmission During Tranquil and Turbulent Periods	335

Figure 4c. Impulse Response Functions (Nigeria) (Continued)

Response of D(INTR) to 100*DLOG(NEER)

1.0
0.5
0.0
-0.5
1	2	3	4	5	6	7	8	9	10

Response of D(INTR) to 100*DLOG(IPI)

.6
.4
.2
.0
-.2
1	2	3	4	5	6	7	8	9	10