![](https://csdnimg.cn/release/download_crawler_static/88109176/bg7.jpg)
AI & The Future of Work 7
Over the last three decades, European economies
have been most successful in the automation and
offshoring of low-paid, low-skilled and low-status
occupations such as farm workers and plant
operators, data-entry clerks and sales assistants.
In parallel, job expansion was most vigorous
among higher-paid and better-skilled positions in
management and the professions (Oesch 2013).
As a result, labor-market opportunities expanded
for the salaried (upper) middle class, whereas
the core of the traditional working class and
subordinate white-collar employees lost ground.
For the polarization thesis to hold, we should also
have observed job growth in low-skilled services
among interpersonal service workers. However,
this mostly female class did not substantially in-
crease its employment share in Western Europe
(Oesch and Piccitto, forthcoming).
While the doom scenario of polarization and
middle-class erosion may make newspaper
headlines, it is not in line with the secular trends
in Western Europe’s employment structure.
Given the extent of educational expansion over
the last few decades, this is good news. As
technical colleges and universities were sending
out highly educated workers in greater numbers,
the economy was also creating more jobs in
occupations requiring higher education. However,
it is an altogether different question whether
upgrading trends from the past translate into
the future.
Disruption versus continuous change
Adopting the framework we sketched out
earlier, Frey and Osborne (2017) questioned
a group of machine learning experts and
examined job descriptions to identify comput-
erization bottleneck tasks that would preclude
an occupation from being computerized. They
estimated the share of jobs that are exposed
to automation to account for 47% of total
employment in the USA.
However, other scholars present very different
assessments of how technology will affect the
labor market of the future. Two recent OECD
studies expect that 9%–14% of all jobs could be
easily automated in the near future (Arntz et al.
2016, Nedelkovska and Quintini 2018). Since
the OECD studies focus on individual tasks
rather than entire occupations, the estimates
therefore diverge widely and are very sensitive
to assumptions. They point to two very distinct
future scenarios: a rst scenario predicts a deep
break with sudden and disruptive changes, while
a second scenario holds continuous and progres-
sive change as more likely.
Automation and employment:
The scenario of disruptive change
According to the scenario of disruptive change,
the next wave of automation should replace em-
ployment in a wide range of activities, including
agriculture, industrial production and logistics
(through robotization), ground transportation
(through unmanned machinery), secretarial and
administrative support (through new software
and Big Data), sales (through self-checkout),
cleaning (through robots) and even construction
(through prefabrication).
Most studies suggest that the degree of au-
tomation is negatively correlated with the level
of training in a given profession. Low-skilled
occupations would therefore be more exposed to
technological change than occupations requiring
a high level of training. More specically for
Switzerland, a recent analysis suggests that
vocational education and training (VET) graduates
are more threatened by automation than university
graduates. While 65% of the jobs held by the
former would be threatened, this is only the case
for 25% of the jobs held by academics.
The disruption scenario predicts de-industrial-
ization and therefore a decreasing demand for
production workers employed in manufacturing.
Through its impact on the secretariat, payment
management, IT maintenance and e-com-
merce, digitalization would also jeopardize the
jobs of commercial employees, sales assistants
and retail managers. More generally, structural
changes in the labor market should mainly affect
the weakest areas – both in terms of educational
level and geographical setting, with the peripheral
regions of countries being more affected than
the service centers of large cities.
Automation and employment:
The scenario of continuous change
The second scenario predicts a much more gradual
change. Widespread automation will not take years,
but decades in the second scenario – and will
depend on various factors such as the feasibility
and costs of new technological solutions as well
as their legal, social and political acceptance.
The technical feasibility of automating a process
does not necessarily mean that an activity will be
automated and employment will disappear any
time soon. Moreover, workers continuously adapt
to technological change and the multiple tasks
that make up an occupation will constantly evolve
over time.
Some tasks appear difcult to automate,
especially if they involve social skills (negoti-
ating, coordinating, teaching or care-giving)
or creative skills (inventing new products and
services, creating art and culture). Therefore,
automation does not seem to threaten the bulk
of employment in areas such as management,