References and Literature
Models of directed technological change were developed in Acemoglu (1998, 2002a, 2003a,b, 2007a), Kiley (1999), and Acemoglu and Zilibotti (2001). These papers use the term directed technical change, but here we used the related term directed technological change, to emphasize continuity with the models of endogenous technological change studied in the previous chapters.
The framework presented here builds on Acemoglu (2002a). A somewhat more general framework, with much less functional form restrictions, is presented in Acemoglu (2007a).Other papers modeling the direction of technological change include Caselli and Coleman (2004), Xu (2001), Gancia (2003), Thoenig and Verdier (2003), Ragot (2003), Duranton (2004), Benabou (2005), and Jones (2005).
Models of directed technological change are closely related to the earlier literature on induced innovation. The induced innovation literature was started indirectly by Hicks, who in The Theory of Wages (1932), argued
“A change in the relative prices of the factors of production is itself a spur to invention, and to invention of a particular kind—directed to economizing the use of a factor which has become relatively expensive.” (pp. 124-5).
Hicks’ reasoning, that technical change would attempt to economize on the more expensive factor, was criticized by Salter (1960) who pointed out that there was no particular reason for saving on the more expensive factor—firms would welcome all cost reductions. Moreover, the concept of “more expensive factor” did not make much sense, since all factors were supposed to be paid their marginal product. An important paper by Kennedy (1964) introduced the concept of “innovation possibilities frontier” and argued that it is the form of this frontier— rather than the shape of a given neoclassical production function—that determines the factor distribution of income.
Kennedy, furthermore, argued that induced innovations would push the economy to an equilibrium with a constant relative factor share (see also Samuelson, 1965, and Drandakis and Phelps, 1965). Around the same time, Habakkuk (1962) published his important treatise, American and British Technology in the Nineteenth Century: the Search for Labor Saving Inventions, where he argued that labor scarcity and the search for labor saving inventions were central determinants of technological progress. The flavor of Habakkuk’s argument was one of induced innovations: labor scarcity increased wages, which in turn encouraged labor-saving technical change. Nevertheless, neither Habakkuk nor the induced innovation provided any micro-founded model of technological change or technology adoption. For example, in Kennedy’s specification the production function at the firm level exhibited increasing returns to scale because, in addition to factor quantities, firms could choose “technology quantities,” but this increasing returns to scale was not taken into account 602in the analysis. Similar problems are present in the other earlier works as well. It was also not clear who undertook the R&D activities and how they were financed and priced. These shortcomings reduced the interest in this literature, and there was little research for almost 30 years, with the exception of some empirical work, such as that by Hayami and Ruttan (1970) on technical change in American and Japanese agriculture.
The analysis in Acemoglu (1998) and the subsequent work in this area, instead, starts from the explicit micro-foundations of the endogenous technological change models discussed in the previous two chapters. The presence of monopolistic competition avoids the problems that the induced innovations literature had with increasing returns to scale.
Acemoglu (2002 and 2003b) show that the specific way in which endogenous technological change is modeled does not affect the major results on the direction of technological change.
This is also illustrated in Exercises 15.19 and 15.29. In addition, even though the focus here has been on technological progress, Acemoglu (2007a) shows that all of the results generalize to models of technology adoption as well. Acemoglu (2007a) also introduces the alternative concept of weak absolute bias and strong absolute bias, which look at the marginal product of a factor rather than relative marginal product. He shows that there are even more general theorems on weak and strong absolute bias. I sometimes referred to weak relative bias and strong relative bias to distinguish the results here from the absolute bias results. The results in Acemoglu (2007a) also show that the constant elasticity of substitution aggregator of the output of the two sectors is unnecessary for the results. Nevertheless, I have kept the constant elasticity of substitution structure to simplify the exposition.Changes in the US wage inequality over the past 60 years are surveyed in Katz and Autor (1999), Autor, Katz and Krueger (1998) and Acemoglu (2002b). The latter paper also discusses how models on directed technological change can provide a good explanation for changes in wage inequality over the past 100 years and also changes in the direction of technological change in the US and UK economies over the past 200 years. There are many studies estimating the elasticity of substitution between skilled and unskilled workers. The estimates are typically between 1.4 and 2. See, for example, Katz and Murphy (1992), Krusell, Ohanian, Rios-Rull and Violante (1999), and Angrist (1995). A number of estimates are summarized and discussed in Hamermesh (1993) and Acemoglu (2002b).
Evidence that 19th century technologies were generally labor-complementary (unskilled- biased) is provided by, among others, James and Skinner (1985) and Mokyr (1990), while Goldin and Katz (1998) argued the same for a range of important early 20th century technologies.
Blanchard (1997) discusses the persistence of European unemployment and argues that the phase during the 1990s can only be understood by changes in technology reducing demand for high-cost labor.
This is the basis of Exercise 15.27 below. Caballero and Hammour (1999) provide an alternative and complementary explanation to that suggested here.Acemoglu and Zilibotti (2001) discuss implications of directed technological change for cross-country income differences. We have not dwelled on this topic here, since this will be discussed in greater detail in Chapter 18.4 in the context of appropriate technologies.
Acemoglu (2003b) suggested that increased international trade can cause endogenous skill-biased technological change. Exercise 15.21 is based on this idea. Variants of this story have been developed by Xu (2001), Gancia (2003), Thoenig and Verdier (2003).
The model of long-run purely labor-augmenting technological change presented in Section 15.6 was first proposed in Acemoglu (2003a), and the model presented here is a simplified version of the model in the paper. Similar ideas were discussed informally in Kennedy (1964) and a recent paper by Funk (2002) provides microfoundations for the ideas by Kennedy. Section 15.8 builds on Jones (2005), which in turn uses ideas that first appeared in Houthakker (1955). A related paper to Jones (2005) is Lagos (2001), who also uses an approach similar to Houthakker’s (1955) and shows how aggregation of productivity across heterogeneous production units can determine aggregate total factor productivity in the economy.
The assumption that the elasticity of substitution between capital and labor is less than 1 receives support from a variety of different empirical strategies. The evidence is summarized in Acemoglu (2003a).
The Habakkuk hypothesis has been widely debated in the economic history literature. It was first formulated by Habakkuk (1962), though Rothbarth (1946) had anticipated these ideas almost two decades earlier. David (1975) contains a detailed discussion of the Habakkuk hypothesis and potential theoretical explanations. Recent work by Allen (2005) argues for the importance of the Habakkuk hypothesis for understanding the British Industrial Revolution. Exercise 15.20 shows how models of directed technological change can clarify the conditions necessary for this hypothesis to apply.
15.11.