Transcript Document

TOTAL FACTOR PRODUCTIVITY AND THE WIDE SPREAD CROSSCOUNTRY GROWTH DISPARITIES IN AFRICA
Paper Presented during the 10th ORSEA annual International Conference held at the
University of Nairobi School Of Business, Lower Kabete Road, Nairobi, October 16-18,
2014
Authors
Dickson Turyareeba
Makerere University Business School,
Email: [email protected], [email protected]
Joyce Abaliwano,
Makerere University Business School,
Email: [email protected]
7/20/2015
1
Outline
•
•
•
•
•
•
•
Introduction
Motivation
Literature Review
Methodology
Results
Conclusion
Policy implications
7/20/2015
2
Introduction
• Cross‐country differences in income growth
are widely known to be enormous (Kanczuk,
et al., 2009,….)
• Increasing and stable economic growth is a
fundamental policy objective in nearly all
modern societies (Malmaeus, 2010; ….)
• High income growth improves income
distribution- Kuznets curve (Kuznets, 1955)
7/20/2015
3
Introduction(cont)
• Economic growth is positively associated with
reductions in poverty (Roemer & Gugerty,
1997; …)
• Citizens’ welfare increases if national income
grows bigger (Stevenson & Wolfers, 2010)
• ….
• To sum it up: Most other goals in society will
be more effectively achieved if the economy
gets bigger (Malmaeus,2010)
7/20/2015
4
Introduction(cont)
• Persistence of pervasive cross-country growth
disparities perpetuates inequalities between
countries, perpetuates dependency and puts the
low income economies at risk of insurmountable
impoverishment and miserable welfare standards
• On African continent, growth disparities are
indeed pronounced
• The so called ‘lion economies’ of Africa have
outpaced the rest of countries in the rate at
which they generate their economic wealth.
7/20/2015
5
Table1:Average Annual growth rates of selected
African countries (2001-2012)
Country
'fast growing economies'
'slow growing economies'
Average annual growth
Average annual growth in
in GDP
Country
GDP
Angola
11.3
C. African Rep.
1.0
Chad
9.8
Guinea-Bissau
1.3
Sierra Leone
9.7
Comoros
2.1
Liberia
8.6
Gabon
2.3
Ethiopia
8.5
Seychelles
2.3
Rwanda
8.0
Swaziland
2.3
Mozambique
7.8
Guinea
2.6
Uganda
7.5
Cameron
3.3
Tanzania
7.0
Burundi
3.3
Nigeria
6.5
Djibouti
3.6
Country group average 8.5
Country group average
2.4
Source: Based on World Bank data bank (last updated December 2014)
7/20/2015
6
Motivation
Until recently, there has been a debate on the drivers of the impressive growth
recorded by the ‘fast growing economies’ on the African continent (see Radelet,
2007; McKinsey & Company, 2010; AfDB, OECD, UNDP & UNECA, 2011: 2012;
Aryeetey, Devarajan, Kanbur & Kasekende, 2012).
Although a number of growth drivers have been identified as sources of rapid
growth in a handful of countries in Africa (natural resources, Agricultural
commodity booms, diversification, better governance and stability, structural
reforms, FDI,…), the empirical growth literature largely misses the recognition
of the role of variations in TFP in explaining the sources of growth differences
existing in Africa.
Yet TFP has gained credible importance in explaining cross country growth
differences (see Harrigan, 1995; Comin and Hobijn 2004; Hafiz, Mohammad &
Mohammad, 2010 ).
The current study intends to fill this knowledge gap
7/20/2015
7
Methods and procedures
Measurement of TFP
• Common methods: Growth accounting and
frontier analysis.
But our study adopts:
• Growth accounting, based on the estimation
of aggregate production function to produce a
measure that approximates technological
progress; Solow residual
7/20/2015
8
Methods and procedures (ctd)
• The objective of this method is to determine
how much economic growth can be attributed
to advances in technological and
organizational competences.
7/20/2015
9
Methods and procedures
• Data and the sample
. Panel data- Secondary
. A total of 20 African countries
. 10 ‘fast growing’ economies and 10 ‘slow
growing countries’
. Study period: 2001-2012
• Sample selection
. Purposive: 10 ‘lion economies’ ( The
Economist, 2012); bottom10 slowest growing
economies (according to growth averages:20012012)
7/20/2015
10
Methods and procedures
• Data sources
. WB Data Bank ( last updated, 2014)
• Research design: Quantitative
• Empirical model(s):
Two econometric models, each specified for a
particular country group:
(i) a pooled cross-section model (ii) fixed effects
panel model
7/20/2015
11
Methods and procedures (ctd)
• Pre-estimation diagnostic tests (To a panel model)
(i) Panel unit root tests to avoid possibility of spurious
regression estimates
. Two panel unit root tests adopted: IPS (2003) and the
Fisher-type (Maddala and Wu,1999).(Why?)
(ii) The cointegration test
. Study adopts the Pedroni (2004) cointegration test
(ii) Causality test- Only for the panel model
Study adopts the Granger (1969) causality test- on
‘stacked data’ that disregards space but not time.
7/20/2015
12
Methods and procedures (ctd)
• Estimation techniques
(i) Panel FE estimator on panel growth specifications, not
RE because the later is appropriate for large
panels(T>20, n=?) and the former largely exploits the
advantages of panel data and controls for unobserved
heterogeneity.
-FGLS
(ii) OLS on pooled model
7/20/2015
13
Methods and procedures (ctd)
Model variables
• The study uses the growth rate in GDP per worker (ggdpw) as
the dependent variable. The study includes growth in Total
Factor Productivity per worker (gtfpw) and lags of growth rate
in GDP per worker as control variables.
Model1:
Model2:
7/20/2015
14
Results
• Unit roots ( Model 2): ggdpw and gtfw were
stationary in Levels i.e. all variables were I(0).
• Cointegration: No evidence of cointegration
in all specifications.
• Causality (Model2): gtfpw was not an
endogenous regressor
7/20/2015
15
Results(ctd)
• Model1 Regression estimates
‘Fast growing economies’ (n=110a); Method: OLS
‘Slow growing economies’(n=110a); Method: OLS
Dep. Var.= ggdpw
Dep. Var.= ggdpw
Indep.vars.
coef.
St. Error p-value
gtfgw
Cons
R2 = 0.072
0.209 0.0723
14.417 1.5718
0.005*
0.000*
Indep.vars.
coef.
St. Error p-value
gtfpw
Cons
R2 = 0.031
0.133 0.0714
7.621 1.1804
R 2 = 0.063
R 2 = 0.022
Prob(F) = 0.0046*
Prob(Breusch-Pagan Chi2)= 0.74
Prob(F) =0.0663
Prob(Breusch-Pagan Chi2)= 0.52
7/20/2015
0.066**
0.000*
16
Results (ctd)
• Heteroscedasticity test after Model 1 Regression did
not show evidence of heteroscedastic residulas from
the regression: Prob(B.Pagan stastistic) > the level of
significance (0.05)
7/20/2015
17
Results (ctd)
Model2 Regression results: The Fixed effects
The Fixed effects results were unreliable because
there were plunged with heteroscedasticity(Table
of results not shown here).
To fix the problem, we adopted the FGLS Estimator
7/20/2015
18
Results(ctd)
Model2 Regression estimates: FGLS
‘Fast growing economies’' (n=10,T=11) , Method: FGLS
‘Slow growing economies’(n=10,T=11), Method: FGLS
Dep. var.= ggdpw
Dep. var.= ggdpw
Indep.vars.
coef.
ggdpw(-1)
gtfpw
Cons
0.280 0.0921
0.156 0.0748
10.761 2.0628
Prob(Chi2)=0.0003*
St. Error p-value
0.002*
0.037**
0.000*
Indep.vars.
coef.
St. Error p-value
ggdpw(-1)
gtfpw
Cons
-0.081 0.0989
0.149 0.0751
8.902 1.513
0.413
0.048**
0.000*
Prob(Chi2)=0.1072
• The heteroscedasticity test after the FGLS did not
detect heteroscedastic residuals
7/20/2015
19
Discussion of the results
• Models 1&2 regression results agree: growth in TFP
per worker has a positive causal effect on growth in
GDP per worker in both country groups.
• Implication: GTFP is an important predictor of growth
in Africa, regardless of country groupings.
However,
• Higher and stronger gains from growth in TFP
separates the ‘fast growing economies’ from the
‘slow growing economies’ in terms of the pace at
which the two country groups generate their
economic wealth.
7/20/2015
20
Discussion of the results (ctd)
But also: The first lag of the dep. Variable is statistically
significant for the ‘fast growing economies’ BUT NOT for the
‘slow growing economies’.
Implication: Current good growth performance in the ‘fast
growing economies’ is influenced by the past good growth
performance. This is why fast growing economies continue to
do better: Important results but it does not address the study
problem holistically.
Caveat!
Our study results only provide a precursor to the prominence
of TFP in growth accounting in Africa but do not provide a
sufficient explanation for the existing widespread crosscountry growth differences on the continent. Already
highlighted by IMF (2013).
7/20/2015
21
Conclusion
• TFP matters for growth (both country groups)
• Larger and ‘stronger’ gains from growth in TFP
has put the ‘ fast growing economies’ a head;
only a partial explanation for the existing
growth disparities.
7/20/2015
22
Policy Implications
• To accelerate rates of income growth of countries in Africa, it is important
for the governments, especially those of low income and slow income
growth, to put in place policies that raise TFP; and provide incentives that
nurture the determinants of TFP.
Policy interventions that may raise TFP include:
 infrastructural development (Inmaculada, Osvaldo & Laura, 2011;
Mushtaq, Ali, Ashfaq, Abedullah & Dawson, 2012);
 industrial development (Conway & Meehan, 2013);
 enhancement of human capital development (Jorgenson and Fraumeni,
1993);
 Economic diversification (Zilibotti,1997)
 increased participation in international trade [Enrique & Ouattara, 2011);
promoting institutional efficiency especially in the public sector
(Acemoglu, Johnson & Robinson, 2004);
 promotion of female labor participation [for instance McGuckin and Van
Ark (2005);
 enactment of priorities to develop the financial sector (Isaksson, 2007) ;
 adequate investments in Information and Communications Technology
(Kretschmer & Strobel, 2013), among others.
7/20/2015
23
Thank You All
for
Lending me your ears
7/20/2015
24