Is job polarization path-dependent? Evidence from Korea

ABSTRACT

What is the source of job polarization in Korea? In this paper, we empirically examine if two competing hypotheses, a path-dependency hypothesis and an ageing hypothesis, can explain patterns of job polarization in Korea. By exploiting regional variations, we find that between 2008 and 2019, job polarization was more evident in regions in which routine workers were historically more important than non-routine workers (path-dependency) while ageing is not associated with the structural change in the labour market. We further show that job polarization is mainly driven by female workers.

KEYWORDS:

• Job polarization
• routine worker
• ageing
• Korea

JEL CLASSIFICATION:

• J20
• J23
• R23

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

¹ See Acemoglu and Autor (2011), Autor (2010), Jaimovich and Siu (2020), Mazzolari and Ragusa (2013), and Shim and Yang (2018) among others that study job polarization using the U.S. data. In this paper, we use the terms high-skilled/cognitive, middle-skilled/routine, and low-skilled/manual interchangeably. For detailed classification, refer to Table A.1.

² Authors’ calculation using the Local Area Labour Force Survey. We closely follow Acemoglu and Autor (2011) to classify workers into three groups. Table A1 in Appendix shows how we classify workers.

³ According to Cheon (2007), job polarization in Korea started during the early 1990s. Given that job polarization emerged in the mid-1980s in the U.S. (Acemoglu and Autor (2011) and Shim and Yang (2018)), the emergence of job polarization in Korea seems to follow the trend of other advanced economies.

⁴ For more related discussions on path-dependency and structural changes in the Korean labour market, see Perez (2016) and Focacci (2021).

⁵ According to OECD, Korea is one of the world’s fastest ageing countries (https://www.oecd.org/economy/ageing-inclusive-growth/).

⁶ This paper is not the first to study job polarization in Korea: Cheon (2007) and Kim (2015) are a few examples that also analyse job polarization using Korean data. Our work is differentiated from them by utilizing a local labour market analysis to identify the role of initial conditions in job polarization.

⁷ This is because only the first half-year of 2020 is available, and the labour market in 2020 might be heavily affected by COVID-19.

⁸ Results are available upon request.

⁹ We follow Acemoglu and Restrepo (2022) for constructing this variable. Our findings do not alter if we instead use (1) the initial level of ageing to be consistent with the other main regressor – the initial routine share – and (2) different definitions of middle-aged and older workers (e.g. middle-aged from 21–50, older from 51 onwards; middle-aged from 21–60, older from 61 onwards; middle-aged from 35–55, older from 56 onwards). Results are available upon request.

Additional information

Funding

This work was supported by the Bank of Korea; Yonsei University.