The Effects of Automation
Rapid development of digital technologies affects the demand for work in various occupations. Some professions are becoming increasingly popular but many others are less needed due to automation and greater efficiency of workers through the use of new technologies. In this paper we examine the dynamics of occupational structure in Poland, analyzing how changes in employment in the various professions depend on the risk of computerization. To this end we use the probabilities of computerisation of different occupations that were estimated by Frey and Osborne (2013). Using longitudinal data from Social Diagnosis study, i.e. a large survey conducted every two years from 2003 to 2015 with over 26000 of respondents in each wave, we make an empirical verification of the effects of work automation in different occupation predicted by Frey and Osborne. We verify whether the probability of automation of a given profession explains the risk of the job loss, as well as how large was the risk of unemployment associated with automation in recent years. Finally, we discuss the situation of people who lost their jobs estimating the scale of technological unemployment in Poland. The results demonstrated in this paper suggests that the automation is an important factor in the dynamics of the labor market.
Asymmetry in the atmosphere of the ultra-hot Jupiter WASP-76 b
The Effects of Automation. How the development of new technologies affects the change in the popularity of various professions.
1. Dominik Batorski
Marek Błażewicz
University of Warsaw
The Effects of
Automation
How the development of new
technologies affects the change in
the popularity of various professions.
3. Technology is likely to dramatically reshape
labour markets in the long run and to cause
reallocations in the types of skills that the
workers of tomorrow will need.
4. The battle between man
and machines goes back
centuries.
In the early 1800s it was
the Luddites smashing
weaving machines.
5. Automation
• These days retail staff
worry about automatic
checkouts.
• Narrative Science, in
Chicago automate the
writing of reports and is
used by Forbes, a
business magazine, to
cover basic financial
stories.
• Many firms use
computers to answer
telephones.
6. Sooner or later taxi drivers will be
fretting over self-driving cars.
7. The end of work
The end of work was predicted
20 years ago (Rifkin 1995)
“We are entering a new phase in
world history—one in which fewer
and fewer workers will be needed
to produce the goods and services
for the global population.”
8. Analogies to the Industrial
Revolution
“The role of humans as the most important factor of
production is bound to diminish in the same way that
the role of horses in agricultural production was first
diminished and then eliminated by the introduction
of tractors.”
Wassily Leontief (1983)
9. Why people still matters?
Frank Levy i Richard Murnane
(2004) The New Division of Labor
• Levy and Murnane argued that
pattern recognition and complex
communication were the two
broad areas where humans would
continue to hold the high ground
over digital labor.
• However, this has not always
proved to be the case.
10. Background
• McAfee and Brynjolfsson
predict dramatic economic
shifts to result from the coming
of the ‘Second Machine Age’.
• The technology will dramatically
reshape the kind of skills
required by workers.
• The automation of jobs threatens
not just routine tasks with rule-
based activities but also,
increasingly, jobs defined by
pattern recognition and non-
routine cognitive tasks.
11. Background
• Stewart, De and Cole (2015) argue that the debate
has been skewed towards the job-destroying
effects of technological change, which are more
easily observed than its creative aspects.
• Others think that we are entering a new era of low
economic growth where new technological
developments will have less impact than past ones
(Gordon).
12. Questions
• Are computers and machines taking our jobs?
or are they merely easing our workload?
• Who will win and who will lose from the impact of
new technology onto old areas of employment?
13. The Future of Work
• Frey and Osborne (2013) combine elements from the labour
economics literature with techniques from machine learning to
estimate how ‘computerisable’ different jobs are.
• The gist of their approach is to modify the theoretical model of
Autor et al. (2003) by identifying three engineering bottlenecks
that prevent the automation of given jobs – these are creative
intelligence, social intelligence and perception and manipulation
tasks. They classify 702 occupations according to the degree to
which these bottlenecks persist.
• Using these classifications, they estimate the probability (or risk)
of computerisation – this means that the job is “potentially
automatable over some unspecified number of years, perhaps a
decade or two”.
• According to their calculations, 33% of working American
population are located in the low risk occupations, while 19% of
them are at medium and 47% at high risk of automation.
14. Probability of automation
• Frey and Osborne (2013) used detailed job characteristics
(including skills and abilities required, and tasks performed)
from O*NET database. The variables were chosen, according
to their importance in the automation process (perception
and manipulation, creativity, social intelligence).
• 70 randomly chosen occupations (out of 702 with an unique
6 - digit SOC code) were labeled as "automatable" or not.
• Then, a discriminant analysis using Gaussian process
classifiers was performed, resulting with remaining 632
being assigned a probability of automation (Frey, Osborne,
2013).
• They provide a table of job’s probability of computerisation
and the Standard Occupational Classification (SOC) code
associated with the job.
17. SOC to ISCO
• The computerisation risks we use are exactly the same
as in Frey and Osborne paper but we needed to
translate them to an International Standard
Classification of Occupations (ISCO) classification
commonly used in most European countries.
• We used a crosswalk proposed by the Bureau of Labor
Statistics (BLS, 2012).
• Finally, we tested the relation between the original
probabilities and those computed by aggregating
categories by ISCO groups. The correlation was r = .94
(weighted by the total number of employed in the US)
18. Probabilities: from SOC to ISCO
Relation between
probabilities form the
original paper (Frey &
Osborne, 2013) and
those determined by
the weighted mean of
SOC groups which are
a part of the same
ISCO category.
19. Data: BKL dataset
• A BKL study (Study of Human Capital in Poland) is an
annual research on the labor market in Poland.
• It provides the detailed description of the participants
occupation, which we reduced to a 4 digit ISCO code.
• In case of unemployed people, a previously performed
job was ascribe to them as their occupation for the
purpose of analysis.
• Data collected over the period of 5 years (2010, 2011,
2012, 2013, 2014) provided us with 83809
observations.
• Before further steps, farmers were excluded from the
sample.
20. The risk of automation in Poland
Frey and Osborne (2013)
proposed a cut - off points
for determining low, medium
and high risk of automation
at p = .3 and p = .7 risk
respectively.
24. Occupations and skills
• ISCO classification is constructed
to reflect the increasing level of
skills required to perform the jobs.
• Widely used method proposes the
division of ISCO occupations into
three groups (Neujobs, 2014):
• high-skills jobs (ISCO beginning with
1, 2 or 3): professionals, managers,
technicians and associate
professionals,
• medium-skills jobs (ISCO 4, 5 ,7 and
8): clerks, service workers, craft and
trades, plant and machine
operators,
• low-skills jobs: elementary
occupations (9).
High skills jobs are usually those
which are not easy to automate.
25. Predicting unemployment
Multiple logistic regression models using:
• skills required (models 1-3),
• probability of automation (models 2 - 3)
• and their interaction (model 3)
as predictors of unemployment,
with other factors as controlling variables:
• size of the city of residence,
• respondent age and squared age,
• education level,
• whether respondent is a university student,
• gender,
• country region,
• year of completing the survey (2010, 2011, 2012, 2013 or 2014).
26. Results
Status:
unemployed = 1
Model 1 Model 2 Full model
B S.E. B S.E. B S.E.
(Intercept) 0.30 (0.08) -0.19 (0.09) -3.25 (0.28)
medium skills job -0.56 (0.04) -0.53 (0.04) 2.55 (0.28)
high skills job -0.65 (0.06) -0.39 (0.06) 2.95 (0.28)
probability of automation 0.62 (0.06) 4.54 (0.33)
medium skills * probability -3.90 (0.34)
high skills * probability -4.53 (0.35)
… … … …
Nagelkerke 𝑅2 .119 .123 .131
Note: The effect of skills is positive in the 3rd model, but it is only because the
interaction effect is included in the model.
Those who worked
in the low skill jobs
were more likely to
be unemployed.
Probability of automation
positively predicted
unemployment.
27. Interaction between automation and
skills in predicting unemployment
Probabilities
on y axis are
fitted by the
model,
meaning that
the effect of
other factors
is accounted
for.
28. Conclusions
• Probability of automation is a significant predictor of
unemployment,
• but the effect depended on the level of skills required
for the job,
• if a job required low skills, chance of being unemployed rises
dramatically with the probability of automation.
• For the high skill group, there was no relation between
required skills and automation.
• Workers with high skills are not affected by
computerization. It doesn’t mean that their job is not
affected but they can more easily adapt or move to
another sector, where their skills are still required.
29. Consequences of technological
unemployment – an anecdote
The consequences of the problem of technological
unemployment are summarized in a classic though
possibly apocryphal story:
Ford CEO Henry Ford II and United Automobile Workers
president Walter Reuther are jointly touring a modern auto
plant.
Ford jokingly jabs at Reuther: “Walter, how are you going to
get these robots to pay UAW dues?”
Not missing a beat, Reuther responds: “Henry, how are you
going to get them to buy your cars?”