ABSTRACT
\We analyse the relationship between the endowment of Key Enabling Technologies (KETs) and the demand for occupations, tasks, and skills in the local labour market areas (LLMAs) of Emilia-Romagna, Italy. We merge three data sources, and we compute the share of highly educated employees, of employees accomplishing low- versus high-routine tasks, and three novel indicators measuring the complexity of occupations, tasks, and skills. Our panel estimates show that a larger share of KETs not only corresponds to a higher demand for workers holding a tertiary education degree, or accomplishing less routinary tasks, but also to a higher demand for a wider, and more exclusive, set of occupations, tasks, and skills. These results are also robust to unobserved heterogeneity and reverse causality.
Acknowledgments
We thank the participants to the 17th European Network on the Economics of the Firm (ENEF) Workshop, and Jacopo Staccioli in particular, for their valuable comments. The access to the SILER dataset has been provided by ART-ER; the treatment and processing of data for research purposes has been authorized by resolution no. 554, by the Emilia-Romagna region (Regional Employment Agency) as part of the research project COME - Skills for Manufacturing in Emilia-Romagna (Emilia Lab). All the errors remain ours.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Correction Statement
This article has been corrected with minor changes. These changes do not impact the academic content of the article.
Notes
4 We intend skills as the set of competences and capabilities needed to competently perform the constituent task of an occupation. In this vein, they represent the ‘trait-d’union’ between demanded tasks and workers’ knowledge and abilities. This notion of skills is grounded on labour demand and is still associated with the concept of ‘routine-biased technical change’ to the extent that a change in the demanded tasks will also entail a corresponding change of the demanded skills. Such notion of skills must not be confused with the one used by Acemoglu (Citation2002), which is supply-side based and uses the terms ‘skilled’ and ‘unskilled’ to describe workers’ educational attainment. To avoid this confusion, we referred to the concept of ‘Skill-Biased Technical Change’ à la Acemoglu as ‘Educational-biased technical change’ in this paper.
5 Using the 2007 edition, we are sure that occupation, task, and skill classifications are not affected by the state of technology from 2008 onwards.
6 Compared to Lo Turco and Maggioni (Citation2020), who use the O*NET to inspect the knowledge and skill content of complex goods, we focus on occupations without combining the jobs content with the level of product complexity.
7 According to these measures, for instance, the LLMAs of Parma and Bologna report the highest levels of SC in 2016 because their occupational structure combines a wide range of highly demanded skills (19 and 20 skill titles respectively) with a relative concentration of skills that are less ubiquitous in the region, namely ‘active learning’, ‘critical thinking’ and ‘learning strategies’.
8 We should stress that our measure cannot fully capture the actual adoption of KETs by companies located in the region. Rather, what we are measuring is the total endowment of KETs in the region, as a mix of adoption, generation, and development of these technologies.
9 Nanotechnology-related patents are not present in Emilia-Romagna in the period of reference.
10 The results of RE estimates are provided in the Appendix A3, . The results do not change with respect to FE estimates. We have also re-estimated EquationEquation 6(6)
(6) using the unweighted KETS share as main regressor. The results are shown in . The results remain similar: the only differences are that the estimated coefficients of (unweighted) KETS on HEDU and on OC (only at t-2) are no more statistically significant.
11 We have also computed the ratio between RTI-L and RTI-H and we have re-estimated EquationEquation 6(6)
(6) using it as a dependent variable. The regressions show that the estimated coefficients of KETS (up to three-year lagged) are positive, but never statistically significant. The results are not shown here but are available on request.
12 Results of the panel unit root tests are provided in Appendix, Table A3.5.
13 The statistics are not reported but available on request.
14 In Column 1, the Hansen J statistic strongly rejects the null hypothesis of absence of overidentification. We reduce the number of instruments considering up to 3, and even 2, lags of the regressors (in levels), but the test still rejects H0 at the 1% level. However, since the results are in line with all the others, we consider the Granger causality running from KETS to HEDU as reliable.
15 The related question of the ICP survey reads ‘How much this skills/tasks/knowledge/activity is important in accomplishing your current occupation?’
16 The related question of the ICP survey reads ‘What is the required level of this skills/tasks/knowledge/activity to accomplish your current occupation?’
17 See https://www.onetcenter.org/content.html for details.
18 Results do not change when dropping Bologna, the regional capital, from the observations.