account not only shifts of labour, but also shifts of capital. In Section 6, a variety of possible sources for biases in the conventional shift-share analysis are discussed,
focusing in particular on the implications of the Verdoorn Law. We propose a modified shift-share analysis in Section 7 that explicitly takes into the account
dynamic scale effects of Verdoorn. This decomposition formula uses branch-specific estimates of the Verdoorn effect to measure the importance of structural change for
aggregate productivity growth.
2. The structural-bonus hypothesis for manufacturing
Explicitly or implicitly, the structural-bonus hypothesis is present in many models of industrial development. Since the empirical work of Hoffmann 1958 and Chenery
et al. 1986, the standard perception of industrial development is a general shift in relative importance from light to heavy industries. Light industries have relatively
low ratios of capital to labour, while heavy industries have relatively high ratios. As more capital-intensive activities have normally higher levels of labour productivity,
this shift from light to heavy industries will generate extra labour productivity growth at the aggregate level. Another often used characterisation of industrial development
is in terms of a sequence of early, middle and late industries, where early industries are associated with low productivity activities and low levels of technological
sophistication Chenery and Taylor, 1968; Syrquin, 1988. Here, the emphasis is on demand factors, with early industries catering to basic domestic needs such as
foodstuffs and textiles, middle industries focusing on intermediate inputs, such as non-metallic minerals and chemicals, and late industries producing investment goods
and sophisticated consumer durables. The shift from early to middle and late industries is also referred to as a process of technological upgrading and is supposed
to generate a bonus for aggregate productivity growth in the manufacturing sector.
1
More recently, Harberger 1998 presented a general vision of the growth process in which economic growth is driven by a ‘mushroom-process’. In a ‘mushroom-pro-
cess’, economic growth is characterised by continuous shifts of resources into specific dynamic sectors. This is contrasted with a ‘yeast-process’ in which econ-
omy-wide growth tendencies predominate. Harberger’s analysis implies that pro- ductivity growth rates vary considerably across industries and factor inputs move to
faster growing industries. Structural change is also a fundamental component of the model presented by Nelson and Pack 1999 to describe the evolution of countries
such as South Korea and Taiwan. In their two-sector model, aggregate productivity growth is driven by expansion of the modern sector. The modern sector uses more
productive technologies and has a higher profitability than the traditional craft sector, but its growth is determined by the effectiveness of entrepreneurial response
1
The classical argument primarily refers to labour productivity, arguing that the advanced sectors of the economy have more scope for capital accumulation than traditional sectors. The modern argument
is couched in terms of both labour productivity and total factor productivity. Given a constant capital per worker ratio, increases in TFP go hand in hand with increases in labour productivity.
to these opportunities. According to Nelson and Pack 1999, the response in East Asia was much higher than in most other developing countries due to, among others,
a more stimulating policy environment and a greater availability of educated labour. More formal models of industrial development, either from the supply side as
in Lucas 1993 or from the demand side as in Verspagen 1993, also tend to stress the importance of structural change for productivity growth. Specialisation in a
limited number of expanding, technologically dynamic branches boosts aggregate productivity additional to any intra-industry growth. Another line of reasoning in
favour of the structural-bonus hypothesis refers to the beneficial effects of structural changes induced by liberalisation of the economy. When a country liberalises its
domestic markets and opens up to international trade and foreign direct investment, neo-classical theory predicts that factor inputs move towards more productive and
efficient activities. South Korea and Taiwan gradually liberalised domestic and international trade markets in the 1960s and 1970s and India and Indonesia in the
1980s. Hence, we expect to find evidence in favour of the structural-bonus hypothesis, if not for labour productivity then at least for total factor productivity growth.
3. Data description