Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
External Shocks and Monetary Policy Responses: The Central Bank of Trinidad and Tobago and its Operational Sphere
UWI Conference On The Economy – October 2018
Avinash Ramlogan, Stefan Edwards, and Alon Dhanessar
Outline
2
1. Introduction
2. Literature Review
3. Stylized Facts
4. Methodology and Estimation
5. Econometric Results and Discussion
6. Conclusion and Policy Considerations
Introduction
3
• Small economies are defined by their vulnerability to exogenous shocks, theimpacts of which can be severe.
• T&T experienced 3 major energy price shocks over the past four decades.– Post‐oil boom (mid‐1980s); Global financial crisis (2008); Global oversupply (2015).
• A Central Bank’s Operational Sphere allows it to influence short term financialmarkets.
• The Central Bank policy using indirect instruments must first affect the short‐termbehaviour of the financial sector before transmitting to any other economic sector.
• The paper seeks to determine if the CBTT exerts significant influence in itsoperational sphere ‐ short term financial markets – by assessing the effects of anenergy price shock, and evaluating the monetary policy responses.
Literature Review
4
‐ Bejarano, Hamann, and Rodriguez (2015 and 2016).‐ Koh (2016) ‐ Ferrero and Seneca (2015) ‐ Tang, Wu, and Zhang (2009) ‐Watson (2003)
‐ Nchor, Klepáč, and Adamec (2016)‐ Jbir and Zouari‐Ghorbel (2009)
‐ Degiannakis, Filis and Arora (2017)‐ Rizvi and Masih (2014)‐ Arouri and Fouquau (2009) ‐ Arouri, Lahiani, and Nguyen (2011)
Stylized Facts
5
0.0
20.0
40.0
60.0
80.0
100.0
120.0
140.0
‐150.0
‐100.0
‐50.0
0.0
50.0
100.0
US$
TT$ Mn
Average Quarterly Change in Govt Balance (NDFIs) ($M) Average Quarterly WTI (Right Axis)
Chart 1: Trends in NDFIs and WTI
Source: Central Bank of Trinidad and Tobago (CBTT).
Stylized Facts
6
Chart 2: Trends in Foreign Exchange Market
Source: Central Bank of Trinidad and Tobago (CBTT).
‐2.0
0.0
2.0
4.0
6.0
8.0
10.0
12.0
14.0
16.0
18.0
0
20
40
60
80
100
120
140
160
Jan‐07 Jan‐08 Jan‐09 Jan‐10 Jan‐11 Jan‐12 Jan‐13 Jan‐14 Jan‐15 Jan‐16 Jan‐17
US$ M
n
Purchases from CBT
T (US$ M
n); W
TI (U
S$)
Purchases from CBTT (US$ Mn) ‐ 3 Month Rolling Average
Net Sales of Foreign Exchange (US$) ‐ 3 Month Rolling Average (Right Axis)
WTI (US$)
Stylized Facts
7
Chart 3: Trends in Inter Bank Activity and Commercial Bank Excess Reserves
Source: Central Bank of Trinidad and Tobago (CBTT).
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
‐
50
100
150
200
250
300
350
400
450
500
Dec‐06 Dec‐07 Dec‐08 Dec‐09 Dec‐10 Dec‐11 Dec‐12 Dec‐13 Dec‐14 Dec‐15 Dec‐16 Dec‐17
TT $MnIBV
TT $Mn
Average Resverve Excess ($M) (Right Axis)Average Daily Inter Bank Volume (IBV) ($M)
‐
20
40
60
80
100
120
140
160
WTI US$
Average Monthly WTI
Stylized Facts
8
Chart 4: Trends in the Sovereign Term Spread and Stock Market Trading
Source: Central Bank of Trinidad and Tobago (CBTT).
0.0
1.0
2.0
3.0
4.0
5.0
6.0
0
20
40
60
80
100
120
140
160
Dec‐06 Dec‐07 Dec‐08 Dec‐09 Dec‐10 Dec‐11 Dec‐12 Dec‐13 Dec‐14 Dec‐15 Dec‐16 Dec‐17
Percen
tage Points
WTI (U
S$)
Term Spread (Percentage Points) (Right Axis)WTI (US$)
0
50
100
150
200
250
300
350
400
450
500
ATI V
olum
e (Tho
usands)
ATIVOL ‐ 3 Month Rolling Average
9
Data
• A dataset of 10 variables is used in populating Yt.
• The period spanned by the data is January 2007 – December 2017, a total of 2709 daily observations per series.
• The data series were individually subject to transforms and scaling in order to minimise model instability.
• They are utilised in an order that represents our interpretation of the sequence in which they are affected by an external shock.
Yt=f[WTI, NETSALE, GOVBAL, OMO, INT, REPORATE, DXSL, TERM_S, ATIVOL, DSELR]+ εt
10
Methodology and Estimation
Yt= c + A1Yt‐1 + … + AiYt‐I + εt (Sims, 1980)
Sample pdf
Posterior pdf g(α│y) = [f(y│α) g(α)]/ f(y) (Bayes, 1763)Prior pdf
f(yt │Yt‐1, α) = N(yt ; ΣApyt‐p,ψ) (Karlsson, 2012)
11
Methodology and Estimation• Paper uses the Minnesota‐Litterman (M‐L) prior developed in Litterman (1979). It
is based on three stylised facts about macroeconomic time series:
i) they are mainly characterised by trends (dominance of sample vs. prior info) i.e., ‘overall tightness’,
ii) the lags nearest to the present time affect the variable most i.e., ‘lag decay’ and,
iii) a variables own lags affect it more than those of other variables i.e., ‘relative tightness’.
• Edit covariance matrix ψ to reflect these prior assumptions.
• The M‐L prior imputes ‘Bayesian shrinkage’ toward a univariate random walk for each variable in the VAR.
12
Econometric Results and Discussion
• The unrestricted model was estimated with 25 lags and provided satisfactory results pertaining to stationarity and autocorrelation.
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5
Inverse Roots of AR Characteristic Polynomial VAR Residual Serial Correlation LM Tests Null Hypothesis: no serial correlation at lag order h Date: 03/09/18 Time: 11:33Sample: 1/03/2007 12/29/2017Included observations: 2684
Lags LM-Stat Prob Lags LM-Stat Prob1 121.4988 0.0708 16 104.8713 0.3498 2 88.57024 0.7863 17 104.5113 0.35893 96.86539 0.5701 18 102.7618 0.40494 92.47525 0.6910 19 114.8839 0.14665 111.6161 0.2009 20 130.1347 0.0231 6 104.0598 0.3706 21 148.5324 0.00127 101.0039 0.4531 22 139.6501 0.00548 91.38994 0.7189 23 127.2684 0.03429 105.3017 0.3389 24 87.24214 0.8149 10 77.09969 0.9568 25 113.8464 0.162611 112.4434 0.186112 130.4769 0.022013 102.0022 0.4255 14 103.2882 0.390915 100.1022 0.4783
Probs from chi-square with 100 df.
13
Econometric Results – Impulse Responses
0.0
0.2
0.4
0.6
0.8
1.0
5 10 15 20 25 30 35 40 45 50
NETSALE_N NETSALE_A
-.5
-.4
-.3
-.2
-.1
.0
.1
5 10 15 20 25 30 35 40 45 50
GOVBAL_N GOVBAL_A
14
Econometric Results
-200
-160
-120
-80
-40
0
5 10 15 20 25 30 35 40 45 50
OMO_N OMO_A
0.0
0.4
0.8
1.2
1.6
2.0
2.4
5 10 15 20 25 30 35 40 45 50
INT_N INT_A
-.036
-.032
-.028
-.024
-.020
-.016
-.012
-.008
-.004
.000
.004
5 10 15 20 25 30 35 40 45 50
REPORATE_N REPORATE_A
-5
-4
-3
-2
-1
0
5 10 15 20 25 30 35 40 45 50
OMO_N
.00
.01
.02
.03
.04
.05
.06
.07
5 10 15 20 25 30 35 40 45 50
INT_N
-.0004
-.0002
.0000
.0002
.0004
.0006
.0008
5 10 15 20 25 30 35 40 45 50
REPORATE_N
15
Econometric Results
-.05
.00
.05
.10
.15
.20
.25
.30
5 10 15 20 25 30 35 40 45 50
DXSL_N DXSL_A
-.004
-.002
.000
.002
.004
.006
.008
.010
.012
.014
5 10 15 20 25 30 35 40 45 50
TERM_S_N TERM_S_A
-1.4
-1.2
-1.0
-0.8
-0.6
-0.4
-0.2
0.0
5 10 15 20 25 30 35 40 45 50
ATIVOL_N ATIVOL_A
-.0015
-.0010
-.0005
.0000
.0005
.0010
.0015
.0020
.0025
5 10 15 20 25 30 35 40 45 50
DSELR_N DSELR_A
16
Discussion Recap of Results: Response to Negative Oil Price Shock
With No Accommodative Policy Adjustment
Net Sales of Foreign Exchange
NDFIs
OMO maturities
Central Bank Intervention
Repo rate
Liquidity
Term Spread
ATI volume
FX apprecia on → FX equilibrium
Negative Oil Price Shock +
Accommodative Policy Adjustment
Net Sales of Foreign Exchange
NDFIs
OMO maturities
Central Bank Intervention
Repo rate
Liquidity
Term Spread
ATI volume
FX depreciation
17
Discussion (Cont’d)
• Deliberate monetary policy action by the Central Bank can significantly affect short term financial markets under a large commodity price shock i.e., NOT a fine tuner.
• But why the marked difference in the results for the two scenarios?
• Outcome resembles the ‘Keynesian Trilemma’ for small open economies (Obstfeld and Taylor, 1997).
• Capital account is open and exogenous, thus the Bank has the choice between exchange rate stability and domestic monetary policy.
• In adopting statistically large shifts to domestic monetary policy, control over the exchange rate in the model was ceded and depreciation occurred.
18
Conclusion and Policy Considerations
• The important consideration for the policymaker is that significant accommodative action can sow the seeds of its own currency depreciation, a situation shown in Ferrerro and Seneca (2015) above as one that is best avoided.
• Considering the main ‘small‐open‐economy constraint’, i.e., that the official reserves of a central bank are always and everywhere finite, the model suggests stability of the market for foreign currency seems to be an appropriate operational choice in the very short run.
• This however does not preclude the need for longer run flexibility in the foreign exchange market as per Rodrik (2008).
19
20
Technical Appendix
• Normality taken to be asymptotic.
• Typical tests become overpowered with large datasets leading to failure i.e. rejection of the normality null.
• This is true even with Box‐Cox and nonparametric (LOESS) transforms of the data or bootstrapping.
• Nevertheless, maximum likelihood estimation gives consistent estimates of coefficients and covariances asymptotically, and standard errors for coefficients can be based on OLS (Hamilton,1994 pp. 298‐299).
• Main requirement to interpret IRF’s is mean‐zero errors.
21
Appendix (Cont’d)
-2,000
-1,000
0
1,000
2,000
3,000
4,000
5,000
250 500 750 1000 1250 1500 1750 2000 2250 2500
GOVBAL gb_t‐60
‐40
‐20
0
20
40
60
80
100
T‐Stat Value
T‐Statistics
dWTI
GOVBAL
REPORATE
dIBVOL
dXSL
OMO
TERM_S
dATIVAL
Mean‐Zero ErrorsRESID01 ‐0.000000000000000153RESID02 0.000000000000007050RESID03 ‐0.000000000000000379RESID04 ‐0.000000000000055300RESID05 0.000000000000013000RESID06 ‐0.000000000000000004RESID07 ‐0.000000000000005430RESID08 0.000000000000000000RESID09 ‐0.000000000000008280RESID10 ‐0.000000000000000034
ATIVOL_BOXCOX DSELR_BOXCOX NETSALE_BOXCOX INT_BOXCOXMean ‐0.518616 ‐0.499921 0.041345 ‐9.065274Median ‐0.175509 ‐0.499985 0.178476 ‐11.11111Maximum 5.640256 ‐0.497901 3.530936 4.689169Minimum ‐11.11111 ‐0.500000 ‐11.11111 ‐11.11111Std. Dev. 2.575859 0.000162 0.900530 4.648307Skewness ‐2.920442 4.288492 ‐1.832433 1.840839Kurtosis 12.89377 29.09738 16.03532 4.414687
Jarque‐Bera 11451.22 65464.83 15905.69 1349.491Probability 0.000000 0.000000 0.000000 0.000000
22
Appendix (Cont’d)
0.0
0.5
1.0
1.5
2.0
2.5
5 10 15 20 25 30 35 40 45 50
Accumulated Response of NET_LD_TREND to Shock1
-1.2
-0.8
-0.4
0.0
0.4
5 10 15 20 25 30 35 40 45 50
Accumulated Response of GOVBAL_LD_TREND to Shock1
-280
-240
-200
-160
-120
-80
5 10 15 20 25 30 35 40 45 50
Accumulated Response of OMO_LD_TREND to Shock1
1.0
1.5
2.0
2.5
3.0
3.5
4.0
5 10 15 20 25 30 35 40 45 50
Accumulated Response of INT_LD_TREND to Shock1
-.07
-.06
-.05
-.04
-.03
-.02
-.01
5 10 15 20 25 30 35 40 45 50
Accumulated Response of REPO_LD_TREND to Shock1
-.2
-.1
.0
.1
.2
.3
5 10 15 20 25 30 35 40 45 50
Accumulated Response of DXSL_LD_TREND to Shock1
.00
.01
.02
.03
.04
.05
5 10 15 20 25 30 35 40 45 50
Accumulated Response of TERM_S_LD_TREND to Shock1
-6
-5
-4
-3
-2
-1
0
5 10 15 20 25 30 35 40 45 50
Accumulated Response of ATI_LD_TREND to Shock1
-.002
.000
.002
.004
.006
5 10 15 20 25 30 35 40 45 50
Accumulated Response of DSELR_LD_TREND to Shock1
Accumulated Response to User Specif ied Innov ations
LOESS filtered impulse responses for the accommodation scenario
23
Appendix (Cont’d)
• The hyperparameters of the priors are set up to reflect that i) the effect of the prior dominates the effect of sample information (λ1=0.1), ii) the relative cross variable weights are low but existent (λ2=0.99) and iii) lag decay is linear (λ3=1).