But Blundell and Bond (1998) found that this has poor finite sample properties in terms of
bias and precision, when the series are persistent and the instruments are weak predictors of
the endogenous changes. Arellano and Bover (1995) and Blundell and Bond (1998)
proposed a systems based approach to overcome these limitations in the dynamic panel data
models. This method uses extra moment conditions that rely on certain stationarity
conditions of the initial observation. The systems GMM estimator (SGMM) combines the
standard set of equations in first differences with suitably lagged levels as instruments, with
an additional set of equations in the levels with lagged first differences as instruments; see
for further details on the advantages of SGMM Rao, Tamazian and Singh (2009) and Rao,
Tamazian and Kumar (2009). We shall use this estimation method in this paper.
Our data covers the period 1970-2005 for 21 African countries. The list of these countries is
in the appendix. The average per capita incomes during our study period range from a low of
U$ 122 of Burundi to a high of US$ 765 of Cote d'Ivoire. It is estimated by the World Bank
that about 46.4% of the population in Africa lives under US$ 1.0 per day (WDI, 2005). In
contrast to other developing nations, the number of extremely poor people in African region
has almost doubled from 1981 to 2005, from 200 to 380 million people and is likely to
increase to 404 million in 2015 (WDI, 2005). Furthermore, most of the countries in the
region have poverty rate of over 50% to 70%. For example, the percentage of people living
below poverty line in Mali, one of the low income African countries, is about 73%. Many
agree that if Africa is to achieve its millennium development goal of reducing poverty, then
the best strategy is high and sustained economic growth.
We first estimated the standard specifications of the production function in equations (1) and
(2) with three alternative methods viz., OLS (pooled data), Generalized Least Squares (GLS)
and the standard GMM and the results are in Table 1. All the estimated coefficients are
significant at the 5% level. However, the OLS estimate of profit share (α) in column 1 at
___2
0.858 seems to be too high and the R of GMM estimate in column 3 is very low. This
leaves the GLS estimate in column 2 at this stage as more reliable. This estimate implies that
the profit share is at about 0.2 and is also close to the GMM estimate. Both the OLS and