Finally, the effective rate of protection (ERP) measures the proportionate increase in per
unit value added of a sector due to the complete system of tariffs. More specifically, it
takes into account the protection on output and the cost-raising effects of protection on
inputs. Our hypothesis is that firm turnover is lower in highly protected sectors,
although Arthukorola (2006) notes that much of the ERP levels and changes reflect
levels and increases in import duties on intermediates rather than on final goods.
3. Empirical approach
3.1 Estimating efficiency and TFP
A broad range of methodologies have been developed for the purpose of estimating
productivity,9 and choices have to be made in identifying the appropriate approach.
Measurement error in inputs is common in most firm level data, particularly for
developing countries. Parametric methods that calculate productivity from a
stochastically estimated production function will be less vulnerable to measurement
errors than their non-parametric alternatives.10 While this will come at the cost of a less
flexible technology specification, appropriate testing procedures can be used to ensure
that the production function is correctly specified. An additional issue is the
simultaneity of productivity and firm input choices. When firms choose inputs they may
be aware of their own productivity but the econometrician is not. As such inputs will be
correlated with the unobserved error term which captures productivity. One way of
dealing with this is the stochastic frontier approach.11 This involves the calculation of
productivity from a parametrically estimated production function which imposes
assumptions on the distribution of the unobserved productivity component to separate
productivity from the deterministic part of the production function and the statistical
noise term.
9 See Van Biesebroeck (2003) for an overview of the various methodologies that have been proposed in
the literature.
10 For example, index number approaches or data envelopment analysis (DEA).
11 See Kumbhakar and Lovell (2000) for an overview. Alternative approaches, not used here due to data
limitations, are instrumental variables estimation, for example the approach developed by Blundell and
Bond (1998; 2000) and semi-parametric estimation, for example the approaches developed by Olley and
Pakes (1996) and Levinsohn and Petrin (2003).