Transcript Determinants of economic growth in Africa

Grace Alinaitwe
Makerere University Business School
10th ORSEA-15-17October 2014
Motivation
Literature
review
Methodology
Conclusion
Results

Growth theories do not
clearly specify the
explanatory variables to
include in the "true"
regression.

A few studies have looked
at determinants of economic
growth using a Bayesian
averaging of classical
estimates

The debate of whether
finance leads or follows
economic growth

Negative, positive and none
relationships have been
found between economic
growth and financial
intermediaries.
To find the true determinants
of economic growth in Africa
Research Questions
To determine whether financial
intermediaries affect growth.
What are the true determinants of
economic growth in Africa
To find out if Beta convergence
exists in Africa
Do financial intermediaries cause
economic growth in Africa?
Does Beta convergence exists in Africa?
1
• This study has
contributed to the
debate of whether
finance causes or
follows economic
growth by finding that
in Africa, financial
development is not a
significant determinant
of economic growth
2
• It has improved upon
other studies by using
a new method called
Bayesian Averaging
of Classical estimates
which takes into
account all the
possible models.
3
• Many financial
intermediary
indicators: Liquid
liabilities/GDP
(llgdp), Central bank
assets/GDP (cdagdp)
and Private credit by
deposit money
bank/GDP have
been used.
4
• Used most of the
African countries
Relationship between economic growth and finance: Pagano, M (1993)
Finance causes growth: King, R. G. and Levine, R. (1993), Spiegel, M. M. (2001), Fritzer, F.
(2004), Odhiambo, N. M. (2009) McKINNON, R. I. (1989) Arestis, P. et al. (2001) and
Ghani, E. (1992)
Growth causes finance: Robinson (1952)
The causal link between growth and finance is determined by the nature and operation
of the financial institutions and policies pursued in each country: Demetriades, P. O. and
Hussein, K. A. (1996) and Arestis, P. and Demetriades, P. (1997). Odhiambo, N. M.
(2009)
Data
• Cross-section data of 37 countries over a period of
1986-2007
• 14 variables
Maddison
data set
Penn world
tables
•
•
•
•
•
GDP per capita,
GDP per capita growth rate (gdpg),
real GDP per capita in current prices (cgdp)
Population (Pop) and
population growth rate (popg)
• Price level of investment (pi),
• Investment share of real GDP (ki),
• Openness in current prices (openc)
• Liquid liabilities/GDP (llgdp),
• Central bank assets/GDP (cdagdp)
Financial structure • Private credit by deposit money bank/GDP (pcrdbgdp)
dataset
World bank
data base
•
•
•
•
•
fertility rate(fert),
inflation rate (inf),
life expectancy (life),
years of schooling (scho) and
oil availability (oil)

Bayesian Averaging of Classical Estimates
y   0  1 x1       n xn  
pi y  
1
2k
j 1
 1 y, m j  pm j
i
y

Posterior inclusion probability of a variable shows the
importance of a certain variable in explaining the
dependent variable

Important variables must have a higher posterior
inclusion probability than their prior one.
p m j
y 
p m j T
2k

i 1

p m j T  K I
SSEJT
2
2
SSEiT
2
BIC weights penalize large models and helps address
the problem of colinearity in large models.

k
P ( M j)  
k



k j 2





kj


k

1 

k




k k j
Expected model size equals 5, the prior inclusion
probability is 5/14 = 0.3571
The posterior model weights in the above equation are
equal to the prior model weights times the Bayesian
Information Criterion (BIC) developed by Schwarz (1978)
divided by the sum of prior weights times the Bayesian
Information Criterion of all possible models.
Similar variables usually explain relatively less variation in
the dependent variable and (BIC) implies less weight on
such models.

BACE combines the averaging of estimates across
models with classical ordinary least-squares (OLS)
estimation.

Its advantages over model-averaging
◦ requires the specification of only one prior hyper-parameter
the expected model size k
◦ estimates are calculated using only repeated OLS
◦ This method takes into account all the possible models
Variable
FDI
posterior prob
Posterior unconditional
posterior conditional
Mean
Mean
st. dev.
st. dev
1
0.0021
0.0002
0.0021
0.0002
Llgdp
0.4108
0.0136
0.0197
0.0332
0.0173
Lcgdp
0.2995
-0.0071
0.0142
-0.0239
0.0166
Popg
0.2792
-0.2317
0.4688
-0.83
0.5391
Fert
0.2608
-0.0016
0.0038
-0.0063
0.0051
INFL
0.1929
0.0001
0.0002
0.0003
0.0002
pcrdbc
0.1873
0.006
0.0172
0.0321
0.0272
Oil
0.1498
-0.0014
0.0048
-0.0091
0.0092
Scho
0.1214
0
0.0001
0.0001
0.0002
Lpop
0.1026
0.0002
0.0016
0.0015
0.0047
cbagdp
0.1004
0.0002
0.0086
0.0021
0.0271
Lpi
0.0872
0.0002
0.0043
0.0025
0.0144
Open
0.0855
-0.0001
0.0029
-0.0007
0.0099
Life
0.0844
0
0.002
-0.0001
0.0069
variable
Kbar=3
Kbar=5
Kbar=7
Kbar=9
Kbar=11
prior inclusion
probalility
0.2143
0.3571
0.5
0.6429
0.7857
1
1
1
1
1
Llgdp
0.3369
0.4108
0.4733
1
1
Popg
0.2071
0.2792
0.3352
1
1
Fert
0.1704
0.2608
0.3618
1
1
Lcgdp
0.1492
0.2995
0.4936
1
1
Pcrdbc
0.1197
0.1873
0.2656
1
1
INFL
0.1003
0.1929
0.2955
1
1
Oil
0.0743
0.1498
0.215
1
1
Scho
0.0706
0.1214
0.1702
1
1
Lpop
0.0557
0.1026
0.1573
1
1
Cbagdp
0.0509
0.1004
0.1514
1
1
Life
0.0492
0.0844
0.151
1
1
Lpi
0.0454
0.0872
0.1489
1
1
Open
0.0443
0.0855
0.1452
1
1
FDI
In Africa financial markets are
positively correlated with
economic growth but for most
indicators, this relationship is
very weak.
Poor countries grow
relatively faster than richer
ones hence beta
convergence.
Strongest evidence in Africa
is found for foreign direct
investment.
Positive but
non-significant
relationship
between
growth and
financial
intermediaries
is probably
due to:
• Africa has not yet reached the required
minimum development level of financial
markets.
• Africa has banks which lack transparency
and good management
• Poor policies could be in place.
I thank you for your kind attention