The skill premium effect of technological change: New evidence from United States manufacturing

• Date: 01 March 2017
• Author: Sushanta K. MALLICK,Ricardo M. SOUSA
• Published date: 01 March 2017
• DOI: http://doi.org/10.1111/j.1564-913X.2015.00047.x

International Labour Review, Vol. 156 (2017), No. 1

Copyright © The authors 2017

Journal compilation © International Labour Organization 2017

The skill premium effect

of technological change: New evidence

from United States manufacturing

Sushanta K. MALLICK* and Ricardo M. SOUSA**

Abstract. Using the NBER-CES Manufacturing Industry Database, the authors

identify a positive relationship between total factor productivity and the skilled-

to-unskilled labour and wage ratios. Highlighting the skill premium for skilled

workers, they nd that technology has become more favourable to skilled labour

since the 1980 s. The productivity differentials between skilled and unskilled labour

increase relative demand for the former when they are imperfect substitutes. The

authors show that the relationships between technology and both ratios are posi-

tive in science-based and production-intensive industries, and negative in supplier-

dominated industries, suggesting industry heterogeneity in technological knowledge.

From a policy perspective, governments should promote science-based innovation.

Wage inequality between skilled and unskilled workers in the United

States has increased substantially since the early 1980s, with product-

ivity differences at industry level explaining the dynamics of the skill pre-

mium.1 In exploring these dynamics and the drivers of technological change,

some studies have focused on the role played by nancial constraints (Cabral

and Mata, 2003; Mallick and Yang, 2011), institutions (Nelson, 1988 and 2008),

knowledge networks (Giuliani and Bell, 2005), uncertainty and risk aversion

(Appelbaum, 1991), the increasing diffusion of new organizational practices

within rms (Piva, Santarelli and Vivarelli, 2005), the impact of external/in-

ternal hirings (Martins and Lima, 2006), or the learning-by-exporting process

(Martins and Yang, 2009; Mallick and Yang, 2013).

*

Queen Mary University of London, School of Business and Management, email: s.k.mallick@

qmul.ac.uk. ** University of Minho, Braga, Portugal, email: rjsousa@eeg.uminho.pt. Ricardo

Sousa wishes to acknowledge that this research was nanced by the Operational Programme for

Competitiveness Factors (COMPETE) and by national funds through the Portuguese Foundation

for Science and Technology (FCT) within the remit of project FCOMP-01-0124-FEDER-03726 8

(PEst-C/EGE/UI3182/2013)..

Responsibility for opinions expressed in signed articles rests solely with their authors, and

publication does not constitute an endorsement by the ILO.

1 See, for example, Pavitt (1984), Kortum (1993), Nickell and Bell (1995), Machin and van

Reenen (1998), Agnello, Mallick and Sousa (2012), Sila (2012) and Agnello and Sousa (2014).

International Labour Review114

Other studies have investigated the sources of economic growth by

decomposing changes in output into changes in factors of production and

changes in total factor productivity (TFP) (Fethi, Pasiouras and Zopounidis,

2013), or by considering that TFP changes can be further disaggregated be-

tween technical efciency changes and technological changes (Jones, 2005;

Pasiouras, 2013; Pasiouras and Sifodaskalakis, 2010). In this context, Carreira

and Teixeira (2008) highlight the role played by internal restructuring vis-

à-vis external restructuring in industry productivity growth. These authors

suggest that while the share of external restructuring is stronger during re-

cessions, internal restructuring dominates in economic expansions. Along

the same lines, Escribano and Stucchi (2014) nd that there is productivity

convergence in recessions because, during such episodes, productivity grows

more in the case of followers than it does for leaders. This result can be ex-

plained by the exogenous increase in competition that causes demand to fall

and threatens followers more strongly. Lozano-Vivas and Pasiouras (2014)

also emphasize the importance of variation in business conditions and vari-

ation in productivity as drivers of sectoral performance. These authors show

that changes in productivity can be decomposed further into changes in best

practices and changes in (in)efciency.

Another strand of the literature on productivity has accounted for the

degree of substitution between skilled and unskilled labour. Nelson and

Winter (1982) argue that the non-neutrality of technological change bene-

ts skilled labour more than unskilled workers. Greenwood and Yorukoglu

(1997) suggest that if skilled workers have a comparative advantage in tech-

nology implementation, the acceleration in investment-specic technological

change will affect productivity growth and increase wage inequality. Caselli

(1999) shows that a biased technology revolution affects wage inequality if

the workforce is heterogeneous in terms of training costs. Caselli and Cole-

man (2001a and 20 01b) and Caselli and Wilson (2004) nd that factor en-

dowments impact on the diffusion of R&D-intensive technologies across

countries. Acemoglu (2002) also shows that technology shifts favour skilled

labour and provide a better understanding of the sources of growth in the

United States. Caselli and Coleman (2006) study cross-country differences

in skilled and unskilled labour efciencies when skilled and unskilled labour

are imperfect substitutes. These authors emphasize that the skill-biased

nature of technology helps to explain the dramatic change in the relative

supply of skills, as well as the skill premium.

In this article, we look at the potential non-neutrality of technology using

industry-level data from the NBER-CES Manufacturing Industry Database.

Our empirical estimation is based on data from the United States manufac-

turing sector over the period 1958–20 09. While assessing the skill premium

effect of technological change, we improve upon the existing literature in var-

ious dimensions. First, in contrast with a vast amount of work in this eld re-

lying on macroeconomic data, we use industry-level data. Therefore, we are

able to investigate the non-neutrality (or skill-bias) effect of technology for a