PowerPoint-Präsentation

On the exposure of the BRIC
countries to global economic
shocks
by Ansgar Belke, Christian Dreger, Irina Dubova*
2017 ASSA Annual Meeting
06-08 January 2017
*Ansgar Belke: University of Duisburg-Essen
Christian Dreger: German Institute for Economic Research (DIW Berlin)
Irina Dubova: University of Duisburg-Essen, RGS Econ
1
Outline
1.
Emerging Markets and BRICS:
– Before and just after GFC
– Slowdown
– Worldwide role of BRICS
2.
3.
4.
Motivation & Research Questions
Literature Overview
Estimation:
– Variables
– Methodology
– Issues
5.
6.
Results
Conclusions
2
Emerging Markets and BRICS:
Before and just after GFC
•
•
•
Source: IMF *weighted by share of GDP at PPP
The catching-up of the BRIC countries (Brazil,
Russia, India, and China) has been a primary
source for global GDP growth in the period
before and the first years after the financial
crisis
EM economies surprised by weathering the
global financial crisis better than advanced
economies and bouncing back rapidly in 2010
EMs successful performance during the crisis
combined with their newfound business cycle
independence led some even to argue that
emerging market economies had decoupled
from AE
3
Emerging Markets and BRICS:
Slowdown
Difference between weighted average EM growth and
weighted average AE growth
•
•
•
•
•
Despite the gradual recovery in the industrial
countries in most recent years, the evolution started
to weaken in emerging markets (EMs)
Growth in EMs has been slowing, from 7.6% in 2010,
to 3.7% in 2015 and is now below its long-run average
This slowdown has been highly synchronized across
EMs, with significant declines in growth in most EM
regions
In the largest EMs heterogeneous group of BRICS
growth has slowed from almost 9% in 2010 to about
4% in 2015, on average, with India being a notable
exception
This slowdown reflects both easing growth in China,
persistent weakness in South Africa, and steep
recessions in Russia since 2014 and in Brazil since
2015
4
Emerging Markets and BRICS:
Worldwide role of BRICS
The BRICS
–
–
–
–
are the largest and most regionally integrated heterogeneous emerging market group in their
respective regions
have been the main source of EM growth and integration into the global economy
during 2010-14, the BRICS contributed about 40% to global growth (all EMs – 60%), up from about
10% during the 1990s
now account for two-thirds of EM activity and more than one-fifth of global activity—as much as the
US and more than the EA—compared with less than one-tenth in 2000
Share of the BRICs in the world economy
Economic growth in the BRICs and China
5
*The economic size of BRICS is much larger in terms of PPP adjusted GDP. BRICS constitute about 30% of global activity while the US constitutes only about 16%*
Emerging Markets and BRICS:
Worldwide role of BRICS
•
•
•
Given the size and integration with the global economy of the largest EMs - the BRICS - a
synchronous slowdown in these economies could have significant spillovers to the rest of the
world
Growth spillovers can operate via
– trade (reduced import demand from BRICS would weaken trading partner exports)
– financial linkages
– the confidence channel (consumer and business sentiment, e.g Levchenko and PandalaiNayar (2015))
The empirical literature finds sizeable spillovers from China for countries with close trade ties
(EAP region, Japan and Germany among the AE), and commodity exporters. Growth in Russia
and Brazil tends to affect growth of their neighbors and those with whom they have strong
trade and remittance linkages
6
Motivation & Research Questions
•
The recent slowdown in EM has been a source of a lively debate:
–
–
–
•
•
Some economists argue - impressive growth performance of EM prior to the crisis was driven by
temporary commodity booms and rapid debt accumulation, and will not be sustained
Others emphasize cyclical and structural factors as the slowdown drivers: weakening macroeconomic
fundamentals after the crisis; prospective tightening in financial conditions; resurfacing of
deep‐rooted governance problems in EM; difficulty adjusting to disruptive technological changes
Still others highlight differences across EM - some of them are in a better position to weather the
slowdown and will likely register strong growth in the future
This paper analyzes the relative role of external factors in GDP growth for the individual
BRIC economies in the period before, during, and after GFC and more recently
The following questions are addressed:
–
–
–
How have external conditions typically affected BRIC economies’ growth over the past decade and a
half?
Are all BRIC countries equally exposed to external shocks, or are some economies more vulnerable?
Within BRIC economies, how has China’s growth influenced global variables?
7
Literature overview
•
•
Both external and domestic as well as cyclical and structural factors have
contributed to the slowdown in emerging markets (Didier et al. 2015)
Before slowdown - the large and sustained increase in commodity prices raised
investment and GDP in most commodity-exporting economies. The higher growth
reflected a combination of improved fundamentals and strong tailwinds that
boosted demand and raised productivity in most countries
– Cubeddu et al. (2014): strong foreign demand, facilitated by advances in trade
liberalization, lower global interest rates, and the acceleration in commodity
prices accounted for one half of the acceleration in growth in the 2000s
– Fayad and Perrelli (2014): lower demand from trading partners plays a key role
in explaining the current slowdown, besides an increase in risk aversion of
international investors
– Anand et al. (2014): both China and India exhibited a slowdown in potential
growth related to a decline in TFP growth
8
Estimation: Variables
•
•
Here foreign variables are captured by commodity prices (real oil price – OIL), World Trade
Index (source: CPB Netherlands Bureau for Economic Policy Analysis, WT) and international
financial conditions (US interest rate, VIX)
Global shocks are disseminated through different channels, like
– the fiscal policy stance (e.g. lower prices of raw materials put increasing pressure on the public
budget)
– tighter monetary policy (e.g. to combat capital outflows due to a higher risk perception of
investors)
– and the real exchange rate, as the real depreciation of the BRIC currencies generates inflation
pressure through higher import prices
•
Domestic variables:
– Real government expenditure(GSpend)
– Real GDP (GDP)
– Short-Term Interest Rate (further as difference with US interst rate - IR)
– Real Effective Exchange Rate (REER)
9
Estimation: Variables
10
Estimation: Methodology
*Estimations are performed with RATS software; cointegration analysis with CATS software
11
Estimation: Methodology
•
•
Structural shocks are identified by imposing contemporaneous restrictions (apart from
exogeneity assumptions). There are no further restrictions on the lagged variables in order to
let the data reveal the patterns of the responses and the transmissions
Within “domestic” block (GSpend, GDP, REER, IR) we make the standard assumption of the
following ordering of the variables:
–
–
•
GSpend is the most inertial variable, IR is the most fast-moving one
MP transmission to the economy is lagged
In terms of cross-linkages - “global” block (OIL, WT, VIX) has a simultaneous impact on
domestic variables for each country
𝜀
𝑢
1
0
0
0
𝑎
𝑎
𝑎
𝜀
𝑎
1
0
0
𝑎
𝑎
𝑎
𝑢
(GSpend, GDP, REER, IR, OIL, WT, VIX)
𝜀
𝑎
𝑎
1
0
𝑎
𝑎
𝑎
𝑢
The robustness check to the alternative
𝜀
𝑎
𝑎
1
𝑎
𝑎
𝑎
𝑢
= 𝑎
.
orderings was plausible
𝜀
0
0
0
0
0
0
𝑢
1
𝐺𝑆𝑝𝑒𝑛𝑑
•
•
𝐺𝐷𝑃
21
𝑅𝐸𝐸𝑅
31
32
𝐼𝑅
41
42
43
15
16
17
𝐺𝑆𝑝𝑒𝑛𝑑
25
26
27
𝐺𝐷𝑃
35
36
37
𝑅𝐸𝐸𝑅
45
46
47
𝐼𝑅
𝑂𝐼𝐿
𝑂𝐼𝐿
𝜀𝑊𝑇
0
0
0
0
𝑎65
1
0
𝑢𝑊𝑇
𝜀𝑉𝐼𝑋
0
0
0
0
𝑎75
𝑎76
1
𝑢𝑉𝐼𝑋
12
Estimation: Methodology
•
•
•
•
Bayesian methods provide an explicit, straightforward means of incorporating uncertainty
into modelling and forecasting
We perform MCMC (Markov Chain Monte Carlo) analysis of a combination of a near-VAR for
the lag coefficients and a structural VAR for the covariance matrix
The use of a near-VAR (rather than a flat prior full VAR) complicates the Monte Carlo
integration. For the full VAR, the structural coefficients can be drawn using only the
covariance matrix from the OLS estimates; here, we need to use the recomputed covariance
matrix at each Gibbs sweep
Technically, Metropolis-within-Gibbs is used. The structural (contemporaneous) model is
drawn using Random Walk Metropolis. Given the values of the structural parameters, the
variances of the shocks can be drawn directly, and given the covariance matrix, the
coefficients can be drawn using Gibbs Sampling methods for linear SUR systems of modest
size
13
Estimation: Methodology
14
Estimation: Issues
•
Stationarity issues – VAR in levels or differences? VECM?
–
–
–
–
–
•
Lag length selection
–
•
Due to the shortness of the data set, we set the lag length of the SVARX to 2 –also in accordance with Schwartz and HannanQuinn’s information criteria and the tests autocorrelations of residuals
Stability
–
•
Most of the empirical studies that identify the impact of MP did not test the stationary condition of their variables (Cushman &
Zha (1997); Brischetto & Voss (1999); Kim & Roubini(2000))
Sims et al. (1990) reported that the VAR model with these non-stationary variables although might incur some loss in the
estimator’s efficiency but not its consistency
According to Brooks (2002), differencing the variables might lead to losing a lot of information – such as long-run relationships
between variables
Sims (1980) recommended against differencing the variables since the main objective of the VAR approach is to analyze the
inter-relationships not the coefficients
An alternative would be to estimate a VECM. However, since it is hard to identify with any degree of accuracy the underlying
structural parameters of a VECM which includes a large number of variables, for practical reasons we derive impulse responses
from the model in levels
According to Luthkepol (2005) and Hamilton (1994), the estimated SVAR model is stable if the absolute value of all eigenvalues
or the largest value is less than one
Residual tests
–
–
Serial correlation test (LM tests)
Normality tests
15
Results: Impulse responses for model with global variables
and China’s GDP
•
•
Impulse Responses
Response to Cholesky One S.D. Innov ations ± 2 S.E.
Response of WT to WT
Response of WT to OIL
Response of WT to VIX
Response of WT to GDP
.03
.03
.03
.03
.02
.02
.02
.02
.01
.01
.01
.01
.00
.00
.00
.00
-.01
-.01
-.01
-.01
-.02
-.02
-.03
-.02
-.03
1
2
3
4
5
6
7
8
9
10
2
Response of OIL to WT
-.1
4
5
6
7
8
5
6
7
8
9
10
-.03
1
2
9
3
4
5
6
7
8
9
10
1
.0
.0
-.1
-.1
1
2
Response of VIX to WT
3
4
5
6
7
8
9
2
Response of VIX to OIL
3
4
5
6
7
8
9
10
1
.3
.3
.3
.2
.2
.2
.1
.1
.1
.1
.0
.0
.0
.0
-.1
-.1
-.1
-.1
-.2
2
3
4
5
6
7
8
9
10
-.2
1
2
Response of GDP to WT
3
4
5
6
7
8
9
10
2
3
4
5
6
7
8
9
10
1
.04
.04
.04
.02
.02
.02
.00
.00
.00
.00
-.02
-.02
-.02
-.02
-.04
3
4
5
6
7
8
9
10
-.04
1
2
3
4
5
6
7
8
9
10
7
8
9
10
3
4
5
6
7
8
9
10
3
4
5
6
7
8
9
10
9
10
Response of GDP to GDP
.02
2
2
Response of GDP to VIX
.04
1
6
-.2
1
Response of GDP to OIL
-.04
5
Response of VIX to GDP
.2
1
2
Response of VIX to VIX
.3
-.2
4
-.1
1
10
3
Response of OIL to GDP
.0
10
2
Response of OIL to VIX
.1
.0
3
4
.1
.1
2
3
Response of OIL to OIL
.1
1
-.02
-.03
1
Variance Decomposition
-.04
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
16
Results: GDP impulse responses
•
Brazil
•
Russia
17
Results: GDP impulse responses
•
India
•
China
18
Results: GDP variance decompositions
Brazil
Russia
India
China
19
Conclusions
•
The Bayesian VAR analysis suggests that the BRIC are heavily affected by the global economy*,
albeit to a different degree:
–
–
–
•
However, in contrast to other countries, the relationship for China is bidirectional
–
–
•
Commodity prices can explain the downturn in Brazil and Russia to a huge extent
India is less affected by the price evolution, but the slower expansion of world trade reduced GDP growth
Prices for raw materials and the expansion of world trade are both relevant to explain output growth in
China
China plays a crucial role in determining oil prices and global trade.
Hence, the change in the Chinese growth strategy puts reform pressure on countries with abundant natural
resources
To be continued..
–
–
Considering other potential determinats – e.g Policy Uncertainty (DoF problem  FAVAR?)
Explicit analysis of spillovers (from China to BRI; within BRIC; BRIC and other EMs; BRIC and AEs  GVAR?)
*In line with Almansour et al. (2014) who concluded that the global development accounted for one half of the variance of
emerging markets growth
20
Thank you!
21
I*
Annex 1.0 Weak exogeneity tests
r
0
1
2
3
4
II***
Test
WE**
r
0
1
2
3
4
III****
Test
WE
Test
WE
r
0
1
2
3
4
Brazil
Trace
P-Value
76.38
0.01
50.07
0.03
26.47
0.12
10.32
0.26
1.27
0.26
𝜒 2 (2) =0.114,
[0.944]
Trace
P-Value
67.81
0.07
44.91
0.09
24.15
0.20
6.50
0.64
1.51
0.22
𝜒 2 (2) =3.027,
[0.220]
Trace
P-Value
69.94
0.05
46.49
0.07
26.43
0.12
11.75
0.17
0.61
0.44
𝜒 2 (2) =2.252,
[0.324]
China
Trace
P-Value
98.12
0.01
64.13
0.05
39.73
0.01
20.23
0.22
8.03
0.26
𝜒 2 (3) =11.044,
[0.011]
Trace
P-Value
86.72
0.07
47.80
0.52
28.51
0.60
11.73
0.83
3.94
0.75
𝜒 2 (1) =5.251,
[0.022]
Trace
P-Value
94.19
0.02
54.12
0.25
30.14
0.50
13.65
0.69
4.20
0.71
𝜒 2 (1) =5.293,
[0.021]
India
Trace
P-Value
95.15
0.02
44.06
0.69
20.51
0.95
5.02
1.00
0.92
1.00
𝜒 2 (1) =1.004,
[0.316]
Trace
P-Value
96.82
0.01
53.19
0.29
22.46
0.89
6.58
0.99
1.72
0.97
𝜒 2 (1) =1.173,
[0.279]
Trace
P-Value
97.73
0.01
56.90
0.17
29.36
0.55
8.27
0.97
4.18
0.72
𝜒 2 (1) =0.063,
[0.802]
Russia
Trace
P-Value
96.61
0.01
64.17
0.05
34.29
0.28
15.93
0.51
5.45
0.54
𝜒 2 (2) =2.251,
[0.324]
Trace
P-Value
90.93
0.03
59.18
0.12
36.28
0.20
15.80
0.52
7.87
0.27
𝜒 2 (1) =0.527,
[0.468]
Trace
P-Value
86.73
0.07
51.17
0.37
27.07
0.68
13.57
0.70
4.36
0.69
𝜒 2 (1) =1.118,
[0.290]
* The model I for each county includes domestic variables (GSpend, GDP, IR, REER) and real oil prices
** LR test for weak exogeneity performed based on obtained cointegrating rank, P-values in brackets
*** The model II for each county includes domestic variables (GSpend, GDP, IR, REER) and world trade
**** The model III for each county includes domestic variables (GSpend, GDP, IR, REER) and VIX
Restricted linear trend specifications have been chosen in order to allow the cointegrating relationships
to be trend-stationary and have non-zero intercepts, the lag length of two was chosen according to the
autocorrelation tests.
22
Annex 2.0
24 countries
23