4 Skill-Technology Complementarities
Improvements in workforce skills can increase firm productivity by raising the productivity of other firm inputs. The focus of this section is the way skilled labour complements capital investments, particularly in supporting the introduction and effective use of new technologies, the implementation of R&D findings, and the assimilation of external knowledge. Various measures of skills are used. These include the proportions of production and non-production workers, education level, or the proportions of graduates. These proxies have limitations in measuring skill levels, but are required by the availability of data. Consequently, only broad inferences on the types of skills required for innovation can be drawn from these studies.
Section 4.1 begins with some of the stylised facts about technical change, and the effects on the demand for and supply of more skilled labour. Sections 4.2 and 4.3 look at firm level evidence from the perspective of firm dynamics. This describes the dispersion and persistence of firm productivity, and the relationship between skill levels and technology adoption. Section 4.4 then explores firm level evidence on the role of work re-organisation in accessing the productivity gains from technology adoption. Section 4.5 points to more recent evidence on the effect of the diffusion of information and communication technologies (ICT) on productivity performance, first at an aggregate and then at an industry level.
4.1 Skill biased technical change
There is general agreement that new technologies have increased the relative demand for skilled labour, in what has been described as skill biased technical change (SBTC). Two broad explanations have been provided as to how technology is having this effect. One is that the introduction of technology has increased the productivity of skilled labour. It has enabled tasks to be undertaken by existing skilled workers with a higher level of productivity. Alternatively, it may be the case that new technologies mainly replace labour intensive tasks, thereby altering the structure of employment. Here new technologies may require certain tasks to be carried out by high skilled rather than low skilled workers. There is a need to distinguish between worker productivity and workforce composition effects. Both these effects appear to be present. They may well yield aggregate evidence of skill-technology complementarities, though they have very different employment effects.
Earlier evidence suggested that the adoption of new technologies, often involving ICT, has required a more skilled workforce or higher levels of formal education. Productivity effects received less attention. For instance, Machin and van Reenen (1998) compare technical change (as measured by R&D intensity) and skill structure (as measured by the proportions of production and non-production workers) across 7 OECD countries. They find a positive association between R&D intensity and higher levels of skills in all 7 countries over time, and conclude that the relative demand for skilled workers has increased with technical change. Further, they observe that this upgrading has occurred within rather than between industries, and has occurred in the same industries in the 7 countries. Haskel and Heden (1999) provide similar evidence for the UK on how computerisation (as well as R&D) has increased the demand for skilled relative to manual workers. Using firm and industry panel data containing information on computerisation and skills, they too find that most skill upgrading is occurring within continuing establishments, and involves a movement away from manual workers irrespective of their education level. Chennels and van Reenen (1999) review the econometric evidence on the association between measures of technology and skills, and the wages and employment of less skilled workers. They note that there is considerable evidence of the skill levels and skilled/unskilled wage margins being positively correlated with R&D investment and use of advanced technology. The evidence on total employment is more mixed, with there being a positive association with measures of technology diffusion, and a negative association with measures of R&D.
Autor, Katz, and Krueger (1998) document the widening education-wage differential in the US over the past 5 decades, and its relationship to technological change as measured by computerisation. This widening differential is attributed to strong and persistent growth in the demand for skills, in spite of a continuing growth in the supply of college graduates. They note that most of this skill upgrading is occurring within industries, which supports a skill-biased technical change rationale. These and similar findings suggest that physical capital and new technologies are complements with more skilled workers. They also find that skill upgrading has been higher in a range of industries with more a more intensive use of computers. They model changes in workforce skills as a function of changes in industry capital intensity and industry-level investment in computer equipment.
Another aspect of this story is the interaction, at an aggregate level, between technical change and the supply of skilled workers. Acemoglu (2002) explores the connections between SBTC, the demand for and supply of skills, and the widening earnings inequality in recent decades. He argues that technical change has affected the demand for skills, and, in turn, the type of technologies adopted has been influenced by the availability of skills. Overall, in recent decades the skill bias has accelerated. One piece of evidence for this is that US skill premiums have continued or increased during the ‘80s and ‘90s, in spite of sharp increases in the supply of educated workers. Acemoglu (1998, 2002) also argues that technology development and adoption is driven by profit incentives. As a result, the availability of skills induces the development of skill biased technologies: “When there are more skilled workers, the market for skill complementary technologies will be larger.” Acemoglu (1998, p1056). That is, technology is skill biased by design, and not by nature. He considers that the alternative arguments, that technological change is exogenous to labour supply and driven by advances in science, Government funded R&D, or non-profit behaviour by entrepreneurs, cannot account for the concurrent growth in the demand for and supply of skills since the 1970s.
Thus, the impact on technology of an increase in the supply of skills, initiated by rising skill premia, is determined by two competing forces. First, there is a substitution effect between skill types, as increasing skilled labour becomes available. Second, there is the directed technology effect which, in response to skill availability, causes the demand for skills to increase. Whether the final premium for skills is above or below the starting point will depend on the size of these two effects. It is possible, even with an increase in the supply of skills, for earnings inequality to grow, as has happened in the US over the past two decades.
Card and DiNardo (2002) review the evidence in favour of the SBTC hypothesis (which they take to be an increase in the relative demand for higher skilled workers). They focus on economy wide trends in wage inequality. The main problem they see with the SBTC hypothesis is that wage inequality stabilised in the 1990's, in spite of continuing advances in and diffusion of computer technologies. This suggests that some of the acceleration in skill premia in the late ‘80s and early ‘90s may have been an episodic effect. They conclude that the SBTC has limitations in explaining the changes that are occurring in the US wage structure, and that other institutional factors such as changes in the real minimum wage also seem to be having an effect.
More recently, Autor, Levy and Murnane (2003) have explored micro-evidence on how computerisation alters the demand for skills, and the tasks that computers are best suited to doing. This provides a more detailed understanding of the nature of SBTC. They use detailed US data on occupation characteristics, and how they have changed over time, in order to identify shifts in job task composition. These are then matched with industry or occupational worker characteristics. They find that over the past three decades computer capital has been substituted for workers following routine or repetitive manual and cognitive tasks, and complemented workers undertaking non-routine problem solving and communication tasks. As computer prices have fallen, both substitution and complementarity effects have been occurring. These changes in job content are occurring within a range of industries, occupations, and education groups. Thus, computerisation is associated with a reduced demand for labour input into routine tasks and an increased labour input into non-routine cognitive tasks. Task shifts have been occurring at all educational levels. Nevertheless, the net effect is an increased demand for better qualified workers (graduates) who have an advantage in undertaking analytical non-routine tasks.
