Automation risk and subjective wellbeing in the UK

Abstract

The personal well-being of workers may be influenced by the risk of job automation brought about by technological innovation. Here we use data from the Understanding Society survey in the UK and a fixed-effects model to examine associations between working in a highly automatable job and life and job satisfaction. We find that employees in highly automatable jobs report significantly lower job satisfaction, a result that holds across demographic groups categorised by gender, age and education, with higher negative association among men, higher degree holders and younger workers. On the other hand, life satisfaction of workers is not generally associated with the risk of job automation, a result that persists among groups disaggregated by gender and education, but with age differences, since the life satisfaction of workers aged 30 to 49 is negatively associated with job automation risk. Our analysis also reveals differences in these associations across UK industries and regions.

Full text

Automation risk and subjective wellbeing in the UK
Jiyuan Zhenga, Bertha Rohenkohlb, Mauricio Barahonaa, Jonathan M Clarkea1
aCentre for Mathematics of Precision Healthcare, Department of Mathematics, Imperial College
London, London
b Institute for the Future of Work, London
ABSTRACT
The personal well-being of workers may be influenced by the risk of job automation brought about
by technological innovation. Here we use data from the Understanding Society survey in the UK
and a fixed-effects model to examine associations between working in a highly automatable job
and life and job satisfaction. We find that employees in highly automatable jobs report
significantly lower job satisfaction, a result that holds across demographic groups categorised by
gender, age and education, with higher negative association among men, higher degree holders and
younger workers. On the other hand, life satisfaction of workers is not generally associated with
the risk of job automation, a result that persists among groups disaggregated by gender and
education, but with age differences, since the life satisfaction of workers aged 30 to 49 is
negatively associated with job automation risk. Our analysis also reveals differences in these
associations across UK industries and regions.
Keywords: Job automation risk, Life satisfaction, Job satisfaction, Understanding Society survey.
1Corresponding author email: j.clarke@imperial.ac.uk (Jonathan Clarke)
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint
NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.

2
1. Introduction
Work is a major part of the lives of people of working age across the world. It has been estimated
that people spend between 30 and 50% of their waking hours at work and the strong relationship
between job satisfaction and overall wellbeing is well known (Layard & de Neve, 2023a,
Rohenkohl & Clarke 2023).
For those in employment, job satisfaction can be profoundly affected by the characteristics of the
workplace. Good interpersonal relationships with colleagues, having interesting work, and job
security may lead to higher job satisfaction, while jobs that are difficult, stressful or offer a poor
work-life balance may lower job satisfaction (de Neve, 2018). Furthermore, unemployment is
consistently associated with lower wellbeing, and the experience of being unemployed can have
lasting negative effects on wellbeing even after finding a new job (Layard & de Neve, 2023b).
Therefore, changes to labour markets that alter the availability or characteristics of work across
industries may in turn have important and widespread consequences the wellbeing of societies as a
whole.
The emergence of new automation technologies, including artificial intelligence and robotics in
recent years, have the potential to significantly change the world of work (Acemoglu & Restrepo,
2018). These technologies are already being applied across industries and may both substitute and
complement existing occupations. Their introduction may lead to the loss of jobs in some
occupations and the creation of entirely new occupations, and may also change the tasks an
employee performs, such as the automation of more routine tasks, leaving employees to focus on
more creative and rewarding tasks (Nazareno & Schiff, 2021).
While extensive research has been conducted on how labour markets respond to new technologies,
much of this has focussed on their impact on employment, wages and productivity. These studies
indicate a complex array of effects on labour markets in which both increases and decreases in
unemployment and wages may occur differently across industries and over time (Acemoglu &
Restrepo, 2018, Rohenkohl & Clarke, 2023).
Less is known about the impact of the emergence of automation technologies on the job
satisfaction and overall life satisfaction of workers. Recent studies have examined various aspects
of the relationship between the risk of job automation and the subjective job satisfaction or life
satisfaction of workers across different countries (Johnson et al., 2020; Tamers et al., 2020;
Schwabe & Castellacci, 2020; Lordan & Stringer, 2022; Liu, 2022; Gorny & Woodard, 2020;
Nazareno & Schiff, 2021; Stankevi
č
i
ū
t
ė
, Staniškien
ė
& Ramanauskait
ė
, 2021), albeit not focussed
on the United Kingdom.
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

3
Just as there is no consensus on the impact of automation technologies on labour markets, there is
no consensus on how job automation may affect the subjective wellbeing of workers. Research has
demonstrated that fear of robot technologies is associated with lower levels of life satisfaction
(Hinks, 2021), and fear of replacement by smart machines has a negative impact on job satisfaction
(Schwabe & Castellacci, 2020). Existing literature has also suggested that lower job satisfaction is
related to higher awareness of automation technologies (Brougham & Haar, 2018). On the other
hand, automation may enhance wellbeing by reducing workload, lowering stress and improving
work-life balance (Lordan & Stringer, 2022; Makridis & Han, 2021). Indeed, a rise in
technological growth has been shown to correlate with increased current and expected future life
satisfaction (Makridis & Han, 2021).
In the UK, around 30% of occupations are at high risk of being automated, with jobs in retail,
manufacturing, administration and logistics thought to face the highest job automation risk
(Berriman & Hawksworth, 2017). Here, we investigate the relationship between job automation
risk and the reported job and life satisfaction of UK employees using data collected over
successive waves of the Understanding Society (USoc) survey (University of Essex, Institute for
Social and Economic Research, 2022) between 2009 and 2022. Life satisfaction and job
satisfaction responses from the survey are used as measures of subjective wellbeing, whereas the
automation risk is quantified by the probability of computerisation of a worker's current
occupation, as estimated by Frey & Osborne (2017). We then apply a fixed-effects model with
individual-, wave- and region-specific fixed effects to establish he presence of significant
associations.
We find that employees in highly automatable occupations report significantly lower job
satisfaction, a finding confirmed for both men and women, with a stronger negative association
among men. Age-wise, this negative association is present in those aged below 50 years of age, but
no association is observed for older employees. Additionally, workers with a degree or other
higher degree in highly automatable jobs report lower levels of job satisfaction compared with
those without a degree or other higher degree. On the other hand, our analysis finds no significant
relationship between job automation risk and the overall life satisfaction of employees, a finding
reproduced across groups disaggregated by gender and education levels. Age-wise, life
satisfaction is negatively associated with the risk of job automation only for people aged 30-49.
Finally, our results disaggregated by industries indicate that working in a highly automatable
occupation is negatively related to job satisfaction in the agriculture, manufacturing, transport and
service industries.
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

4
2. Data and Methods
2.1. Data
This paper uses UK data from the Understanding Society survey (USoc) led by the University of
Essex, Institute for Social and Economic Research. The sample is nationally representative of the
population in the UK2, and respondents are interviewed yearly if they live in the UK and agree to
take part (Institute for Social and Economic Research, 2021). The Understanding Society survey
contains rich information on a wide range of socio-demographic characteristics, as well as work
and wellbeing across the UK (University of Essex, Institute for Social and Economic Research,
2022).
We use data from wave 1 (January 2009-March 2011) to wave 12 (January 2020-May 2022). The
sample is made up of individuals aged 16 years and above who are in employment at the time of
the survey. Self-employed respondents are excluded due to missing data on job-related
characteristics for this group of workers. To measure the automation risk of each occupation, we
focus on employees whose three-digit Standard Occupational Classification (SOC) 2000 code can
be matched to occupations whose probabilities of job automation have been estimated by Frey and
Osborne (2017). We include only respondents with non-missing data for all measures of subjective
wellbeing including life satisfaction and job satisfaction resulting in a dataset of 43471 workers
across 12 waves.
2.1.1. Subjective Wellbeing
Subjective wellbeing is a synoptic reflection of how individuals feel about their own lives, and thus
does not rely on analysts deciding what specific factors should be used to quantify wellbeing
(Hicks, Tinkler & Allin, 2013). Here, we use a widely adopted measure of subjective wellbeing:
life satisfaction. National life satisfaction statistics have been adopted as a measure of wellbeing of
citizens to inform policy in the UK (Diener, Inglehart & Tay, 2013). The life satisfaction variable
is created here from the question within USoc about an individual’s overall satisfaction with life,
and measured on a seven-point scale, ranging from 1 (“completely dissatisfied”) to 7 (“completely
satisfied”). The variable is dichotomized into “not satisfied” (comprising scales 1-4: “completely
dissatisfied”, “mostly dissatisfied”, “somewhat dissatisfied”, “neither satisfied nor dissatisfied”)
and “satisfied” (comprising scales 5-7: “somewhat satisfied”, “mostly satisfied”, “completely
satisfied”).3 We then explore whether job automation risk is linked with specific facets of life
2 The representativeness of the sample used in empirical analysis is checked in Section 3.1.
3 Using the original categorical variables on the 7-point Likert type scale as outcome variables, the results are
similar to those obtained using dichotomous (‘not satisfied’/’satisfied’) job satisfaction as outcome variables.
The same applies to the dichotomous job satisfaction variable below.
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

5
satisfaction, including satisfaction with health, income and the amount of leisure time, each of
which are also included as similar ordinal variables within USoc.
As the second measure of subjective wellbeing, we use job satisfaction. Job satisfaction is typically
positively correlated with the subjective wellbeing of those in employment (Bowling, Eschleman
& Wang, 2010) and may thus be understood as a component of overall life satisfaction. In USoc,
the survey question concerning job satisfaction asks how satisfied the respondent is with the
present job on a seven-point scale, which is dichotomized between ‘satisfied’ (1-4) or ‘not
satisfied’ (5-7). We adopt the dichotomized variable for life satisfaction and job satisfaction for
clarity of exposition but the results hold without change for the seven-point variable (see section
2.2.1).
2.1.2. Automation Risk
As a measure of automation risk, we use the probability of future computerisation for 702
occupations from Frey and Osborne (2017) (hereafter FO). Based on the 2010 version of O*Net
data, which includes characteristics of 903 occupations4 in the US, FO classify 70 occupations as
fully automatable or non-automatable and implement a classification algorithm to derive the
probability of computerisation of all occupations based on features such as perception and
manipulation, creativity and social intelligence required to carry out job tasks. The FO measure
thus indicates the automatability of jobs beyond the computerisation of routine tasks (Frey &
Osborne, 2017). To match the FO probabilities of computerisation to the three-digit SOC 2000
occupation codes in USoc, the probability of job automation corresponding to four-digit UK SOC
2010 codes is obtained from ONS (White, Lacey & Ardanaz-Badia, 2019) and crosswalked to
four-digit UK SOC 2000 codes. Then, we assign the mean probability of job automation to the
corresponding three-digit SOC 2000. To make our results directly comparable to other work (e.g.,
Adamczyk, Monasterio & Fochezatto (2021), Albuquerque et al. (2019)), we create a binary
variable that equals 1 when the probability of automation is in the top quartile (i.e., larger than the
75th percentile of all jobs) and zero otherwise. However, our results do not change if the threshold
is varied, or if we use the probability of automation as a continuous variable.
2.2. Estimation Strategy
Our data captures not only variation across individuals, but also changes within individuals over
time, which can help identify unobserved heterogeneity affecting outcomes of interest (Hsiao,
1985). Hence, we apply the following fixed effects model to estimate the association between
4Jobs under the category of “all other” which do not correspond to the Labour Department’s six-digit
Standard Occupational Classification are excluded by Frey and Osborne (2017).
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

6
wellbeing and job automation risk, which allows us to remove bias due to possible confounders:

௜௝௧

௜௝௧
·

௜௝௧

௜


௧

௝

௜௝௧
(1)
where

௜௝௧
denotes the wellbeing variable (job satisfaction or life satisfaction) of individual

in
area

during wave

;

௜௝௧
is a binary variable which equals one if an employee is working in a
highly automatable job and zero otherwise (as described above);

௜
denotes individual fixed
effects;

௧
is the wave fixed effects;

௝
represents the region fixed effects and

௜௝௧
is the
idiosyncratic error term. The vector of control variables

௜௝௧
includes socio-demographic
characteristics (i.e., age, gender, marital status5, education levels6, ethnicity7, the presence of
children, housing tenure and rural/urban residence) as well as job-related attributes (i.e., industry
classifications8, size of workplace9, contract type and public/private sector).
The coefficient of interest is

, which indicates the relationship between working in a highly
automatable job on wellbeing outcomes. The individual fixed effects capture personal attributes
which remain unchanged over time. The wave fixed effects represent factors which could vary
over time, such as changes in macroeconomic conditions as well as national health and wellbeing
policies. The region fixed effects indicate region-specific features, such as regional economic
development, local labour market conditions and local public policies about employment. Standard
errors were clustered at the individual level. The model thus removes unobserved individual-, time-
and region-specific heterogeneity which may otherwise bias our results.
2.2.1 Robustness
We have confirmed the robustness of our results to the definition of the variables. First, we check
that the results do not change for different thresholds when dichotomising the job automation risk
5The marital status is categorised into married/civil partnership/cohabitation, widowed/divorced/separated
and single.
6 We classify the highest educational and vocational qualification into the following categories: degree and
other higher degree, A-level or equivalent qualifications, GCSE or equivalent qualifications, other
qualification and no qualifications.
7 Based on Government Statistical Service (GSS) harmonised standards on ethnic group from the Office for
National Statistics (n.d.), ethnicity is classified into white, mixed ethnicity, Asian/Asian British,
black/African/Caribbean/black British, Arab and other ethnic groups.
8 The industry of firms classified using two-digit Cross-National Equivalent File (CNEF) industrial
classification is aggregated to one-digit CNEF industrial classification.
9 Based on Organisation for Economic Co-operation and Development (2022), organisations are classified
into micro and small organisations (1-49 employees), medium-sized organisations (50-499 employees) and
large organisations (>=500 employees).
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

7
into a binary variable, and when we use directly the original automation risk probabilities as a
continuous variable. Second, the results remain unchanged when considering the original
categorical variables of job satisfaction and life satisfaction on the 7-point Likert type scale
analysed through a fixed effect ordered logit model (Baetschmann et al., 2020), which considers
unobservable fixed effects in the ordered logit model.
3. Results
3.1. Descriptive Statistics
Figure A-1 in the Appendix shows the distribution of the probability of job automation over the
whole sample. Overall, the probability of automation ranges between 0.2 and 0.7, and we classify
the top quartile (21 out of 81 occupations) as highly automatable, so that 16217 out of 43471
respondents work in highly automatable occupations. Table 1 presents descriptive statistics of the
sample. Statistically significant differences between those working in highly automatable or less
automatable jobs were found for all descriptors, with the exception of the presence of children
aged 15 years or under in the household, as shown in Table 1. Notable differences between groups
include 22.6% of those working in a highly automatable job having a university degree, compared
to 52.6% of those working in a less automatable job. Those working in highly automatable jobs
were also more likely than those in less automatable jobs to have worked in the private sector
(89.9% vs. 68.8%) and less likely to have worked in the service sector (50.8% vs. 66.8%).
Table A-1 in the Appendix compares the distribution of gender, ethnicity, age groups and
qualifications between the USoc sample in our empirical study and the population of the UK
according to the 2011 UK Census (UK Data Service, 2014). Overall, the distribution of our sample
is broadly similar to that of the Census with respect to ethnicity, gender and age groups, yet there is
a modest overrepresentation in our sample of individuals with higher qualification, including those
holding a degree or other higher degree (level 4+) and GCSE (level 3), while there is an
underrepresentation of people with qualifications at level 1 and without qualifications.
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

8
Table
1
. Descriptive statistics.
The characteristics of those working in
highly automatable
and
less automatable jobs were compared using chi-squared tests for each of the categorical variables
and a t-test for age, the only continuous variable.
Categorical variables
Full sample
Highly
automatable job
Less automatable job
Frequency
Percent
10 Frequency Percent Frequency Percent P value
Panel A. Individual
characteristics
Gender
Women 23648 54.40 9321 57.48 17365 53.52 < 0.001
Men 19841 45.64 6900 42.55 15094 46.52
De facto marital status
Married/Civil partner 22573 51.93 5869 36.19 18347 56.55 < 0.001
Living as couple
9353
21.52
2800
17.27
7261
22.38
Widowed/Surviving
civil partner 663 1.53 238 1.47 463 1.43
Divorced/Dissolved civil
partner 3046 7.01 921 5.68 2331 7.18
Separated 1587 3.65 444 2.74 1195 3.68
Never married 14920 34.32 7890 48.65 9329 28.75
Education
Degree or other higher
degree
19196 44.16 3712 22.89 17051 52.55
< 0.001
A-level 12002 27.61 5699 35.14 7859 24.22
GCSE 10146 23.34 5258 32.42 6040 18.62
Other
qualification
3140
7.22
1503
9.27
1922
5.924
No qualification 2253 5.18 1369 8.44 1017 3.13
Ethnicity
White 35410 81.46 13154 81.11 26659 82.16 < 0.001
Mixed ethnic group 947 2.18 395 2.44 675 2.08
Asian 4509 10.37 1747 10.77 3178 9.80
Black/African/Caribbean 2146 4.94 747 4.61 1603 4.94
Arab
111
.26
28
.17
92
.28
Any other ethnic group 348 .80 146 .9 239 .74
Presence of children
aged 15 or under
Yes 20737 47.70 7159 44.15 15323 47.23 0.510
No
30700
70.62
10870
67.03
22985
70.84
Housing tenure
Owner 32046 73.72 10464 64.53 24930 76.84 < 0.001
Non-owner 15421 35.47 6636 40.92 10454 32.22
Urban/Rural area
Rural area 10633 24.46 3693 22.77 8050 24.81
< 0.001
Urban area 34938 80.37 12915 79.64 26002 80.14
10 The percent and frequency refer to between percent and between frequency. The between percent denotes
the percent of individuals who have ever belonged to a certain category of the variable.
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

9
Table 1
. (Continued).
Categorical variables
Full sample
Highly
automatable job
Less automatable job
Frequency Percent Frequency Percent Frequency Percent P value
Panel B. Job-related
characteristics
Working in an
automatable job
or not
Yes 16217 37.31
No
32446
74.64
Size of workplace
Micro- and small-sized 25767 59.274 10553 65.07 17664 54.44 < 0.001
Medium-sized 18008 41.43 5714 35.24 13404 41.31
Large-sized 10151 23.35 2069 12.76 8483 26.15
Private/public sector
Private sector 33076 76.09 14572 89.86 22317 68.78 < 0.001
Public sector 17988 41.38 3196 19.71 15492 47.75
Contract type
Temporary contract 9175 21.11 3941 24.30 5704 17.58 < 0.001
Permanent contract
40975
94.26
14353
88.51
30895
95.22
Industry
Agriculture 280 .64 63 .39 223 .69 < 0.001
Energy
481
1.11
112
.69
390
1.20
Mining 183 .42 27 .17 165 .51
Manufacturing 5403 12.43 1828 11.27 3896 12.01
Construction 2155 4.96 383 2.36 1842 5.68
Trade
8249
18.98
5912
36.46
2906
8.96
Transport
2132
4.90
735
4.53
1495
4.61
Bank/Insurance 3510 8.07 646 3.98 3047 9.39
Services
27749
63.83
8236
50.79
21664
66.77
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

10
Notes: Differences in population characteristics between those working in highly automatable and less automatable jobs
are calculated using two-tailed chi-squared tests for categorical variables and using two-tailed t-tests for age.
Categorical variables
Full sample Highly automatable job Less automatable job
Between
frequency Between
percent Between
frequency Between
percent Between
frequency Between
percent P value
Panel C.
Wellbeing
Life
satisfaction
Completely
dissatisfied 2250 5.176 872 5.377 1420 4.377 < 0.001
Mostly
dissatisfied 6641 15.277 2055 12.672 4742 14.615
Somewhat
dissatisfied
9290 21.371 2854 17.599 6781 20.899
Neither
satisfied nor
dissatisfied
11082 25.493 3758 23.173 7807 24.062
Somewhat
satisfied 19965 45.927 6087 37.535 14928 46.009
Mostly
satisfied 30497 70.155 9594 59.16 23228 71.59
Completely
satisfied 9885 22.739 3547 21.872 6784 20.909
Job
satisfaction
Completely
dissatisfied 3009 6.922 1066 6.573 2000 6.164 < 0.001
Mostly
dissatisfied 5001 11.504 1492 9.2 3589 11.061
Somewhat
dissatisfied
10176 23.409 2865 17.667 7553 23.279
Neither
satisfied nor
dissatisfied
11275 25.937 3761 23.192 7822 24.108
Somewhat
satisfied
22480 51.713 7035 43.38 16571 51.073
Mostly
satisfied
29192 67.153 8754 53.98 22313 68.77
Completely
satisfied 15722 36.167 4903 30.234 11459 35.317
Continuous variables
Full sample Highly automatable job Less automatable job
Overall
mean Overall
standard
deviation
Overall
mean Overall
standard
deviation
Overall
mean Overall
standard
deviation
P value
Age at the
time of
interview
41.86 13.00 38.73 15.17 42.91 12.00 < 0.001
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

11
3.2. Automation Risk and Subjective Wellbeing
Table 2 shows the results of the fixed effects model (1) for the full sample. We find that working in
a highly automatable job is associated with a reduction in the level of job satisfaction. On the other
hand, we find no evidence of a significant relationship between the life satisfaction of workers and
automation risk. Yet we find significant associations with specific dimensions of life satisfaction;
as shown in Table 3, working in a job at high risk of automation is associated with a small decline
in levels of satisfaction with health, and a larger reduction in the level of satisfaction with income;
but an increase in the level of satisfaction with the amount of leisure time.
Table 2. The relationship between working in a highly automatable job and life satisfaction/job satisfaction.
(1)
(2)
Life
satisfaction
Job
satisfaction
Full sample
Highly automatable job -0.00420 -0.0490***
(0.00528)
(0.00625)
Mean of Dep. Var. 0.767 0.787
Number of Obs
187630
187630
Number of Individuals 32468 32468
Notes: The fixed effect model in Equation 1 is applied, which includes all sociodemographic and job-related control
variables as well as individual, region and wave fixed effects. Standard errors in parentheses are clustered at the
individual level. *** p < 0.01, ** p < 0.05, * p < 0.1.
Table 3. The relationship between working in a highly automatable job and different aspects of life
satisfaction.
Different dimensions of life satisfaction
(1)
(2)
(3)
Satisfaction with
health Satisfaction with
income
Satisfaction with
amount of leisure
time
Highly
automatable job
-
0.00988
*
-
0.0269
***
0.0220
***
(0.00565) (0.00611) (0.00629)
Mean of Dep. Var. 0.682 0.616 0.552
Number of Obs 187518 187476 187557
Number of Individuals 32454 32438 32457
Notes: The fixed effect model in Equation 1 is applied, which includes all sociodemographic and job-related control
variables as well as individual, region and wave fixed effects. Standard errors in parentheses are clustered at the
individual level. *** p < 0.01, ** p < 0.05, * p < 0.1.
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

12
3.3. Subgroup Analysis
In Table 4, we investigate the heterogeneous relationship between job automation risk and life and
job satisfaction, by examining subsamples classified by gender, age and education. The results for
the gender and education subgroups are fully consistent with the associations found for the entire
sample (see Table 2). However, some differences are found when the sample is divided into four
age groups (16-29, 30-49, 50-64 and 65+). In contrast to the full sample, workers aged 30-49 in
highly automatable jobs experience lower life satisfaction; and no significant (negative)
relationship between job automatability and job satisfaction is observed for those aged 50-64 and
65+. Hence the negative associations between risk of automation and job and life satisfaction are
stronger in the younger segments of the population.
Table 4. Heterogeneous relationship between working in a highly automatable job and life satisfaction/job
satisfaction.
(1) (2)
Life satisfaction Job satisfaction
Gender
Men
Highly automatable job -0.00459 -0.0660***
(0.00842) (0.0101)
Mean of Dep. Var. 0.772 0.772
Number of Obs 82845 82845
Number of Individuals 14540 14540
Women
Highly automatable job -0.00391 -0.0373
***
(0.00677) (0.00789)
Mean of Dep. Var. 0.764 0.799
Number of Obs 104772 104772
Number of Individuals 17933 17933
Age Group
16-29
Highly automatable job -0.00274 -0.0740
***
(0.00866) (0.0103)
Mean of Dep. Var. 0.782 0.780
Number of Obs 35085 35085
Number of Individuals 9325 9325
30-49
Highly automatable job -0.0225
**
-0.0475
***
(0.00914) (0.0109)
Mean of Dep. Var. 0.768 0.788
Number of Obs 90038 90038
Number of Individuals 16880 16880
50-64
Highly automatable job 0.00315 -0.0145
(0.0134) (0.0146)
Mean of Dep. Var. 0.751 0.781
Number of Obs 53813 53813
Number of Individuals 10569 10569
>=65
Highly automatable job 0.0616 0.0169
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

13
(0.0503) (0.0385)
Mean of Dep. Var. 0.840 0.914
Number of Obs 4417 4417
Number of Individuals 1138 1138
Education level
People with a degree or
other higher degree
Highly automatable job -0.0132 -0.0647
***
(0.00938) (0.0116)
Mean of Dep. Var. 0.795 0.793
Number of Obs 88548 88548
Number of Individuals 14698 14698
People without a degree
or other higher degree
Highly automatable job -0.00391 -0.0427
***
(0.00668) (0.00783)
Mean of Dep. Var. 0.743 0.782
Number of Obs 97929 97929
Number of Individuals 18458 18458
Notes: The fixed effect model in Equation 1 is applied, which includes all sociodemographic and job-related control
variables as well as individual, region and wave fixed effects. Standard errors in parentheses are clustered at the
individual level. *** p < 0.01, ** p < 0.05, * p < 0.1.
To further quantify the differences in associations between the subgroups in Table 4, we added
interaction terms between job automation and particular variables for gender, education or age (see
Table 5). The most salient differences are as follows. Women exhibit a significantly less negative
association between job satisfaction and job automatability, as compared to men. In contrast,
workers with a degree or other higher degree report a significantly more negative association
between job satisfaction and risk of job automation, as compared to people without a degree.
Relative to the age group 30-49, workers aged 16-29 report a significantly more negative
association between job satisfaction and risk of job automation, but the opposite applies to older
workers (above 50). The negative association between risk of job automation and life satisfaction
found in the age group 30-49 is found to be significantly more negative than in any of the other age
groups.
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

14
Table 5. The comparison of the relationship between working in a highly automatable job and life
satisfaction and job satisfaction across subgroups
(1) (2)
Life satisfaction Job satisfaction
Comparison between genders:
Women × Highly automatable
job
-0.00241 0.0367
***
(0.0103) (0.0123)
Mean of dep. var. 0.767 0.787
Number of obs
187630 187630
Number of individuals
32468 32468
Comparison between age
groups:
16-29 × Highly automatable job
0.0166
*
-0.0442
***
(0.00928) (0.0105)
50-64 × Highly automatable job
0.0235
**
0.0164
*
(0.00963) (0.00987)
>=65 × Highly automatable job
0.0328 0.0230
(0.0208) (0.0206)
Mean of dep. var. 0.767 0.787
Number of obs
187630 187630
Number of individuals
32468 32468
Comparison between levels of
qualification:
Degree or other higher degree ×
Highly automatable job -0.0121 -0.0303
**
(0.0102) (0.0122)
Mean of dep. var. 0.767 0.787
Number of obs
187630 187630
Number of individuals
32468 32468
Notes: An interaction term between the dummy variable for working in a highly automatable job and
gender/education/age is added in the fixed effect model. In the table, the variable “Gender” equals 1 for women and 0 for
men. The variables “16-29”, “50-64” and “>=65” represent dummy variables which equal 1 if the age an individual lies
within a certain range. “Degree or other higher degree” indicates a dummy variable which equals 1 if a person holds a
degree or other higher degree and zero otherwise. Men, workers aged between 30 and 49, and employees without a
degree or other higher degree are considered the referenced group. * p<=0.1, ** p<=0.05, *** p<=0.01.
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

15
Finally, Table 6 explores variability by industry sectors. The results on life satisfaction are
consistent with the full sample, i.e., there is no significant relationship between automation risk
and life satisfaction across all industries. On the other hand, while some industries exhibit the same
negative association between job automation risk and job satisfaction as in the full sample (i.e.,
agriculture, manufacturing, transport and services), there is no significant (negative) association for
the other sectors (i.e., energy, mining, construction, trade, bank/insurance). For regional effects,
see Table A-2 in the Appendix.
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

16
Table 6.
The relationship between working in a highly automatable job and life satisfaction/job satisfaction across different industries in the UK.
Life satisfaction
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
Agriculture Energy Mining Manufacturing Construction Trade Transport Bank/Insurance Services
Highly
automatable job
-
0.124
-
0.0686
-
0.00477
-
0.0319
0.0590
-
0.00914
-
0.0490
-
0.0314
-
0.00176
(0.132) (0.0595) (0.0638) (0.0208) (0.0446) (0.0160) (0.0361) (0.0270) (0.00810)
Mean of Dep. Var.
0.792
0.800
0.837
0.775
0.789
0.734
0.739
0.795
0.771
Number of Obs 717 1686 645 19659 6581 23248 6793 12279 111530
Number of Individuals 161 333 132 3732 1385 5078 1364 2362 20100
Job satisfaction
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Agriculture
Energy
Mining
Manufacturing
Construction
Trade
Transport
Bank/Insurance
Services
Highly
automatable job
-
0.345*
-
0.0766
0.0745
-
0.0745***
-
0.0627
-
0.0229
-
0.130***
-
0.0359
-
0.0406***
(0.188) (0.0905) (0.0796) (0.0226) (0.0583) (0.0175) (0.0380) (0.0310) (0.00940)
Mean of Dep. Var.
0.900
0.795
0.816
0.776
0.797
0.762
0.771
0.785
0.797
Number of Obs 717 1686 645 19659 6581 23248 6793 12279 111530
Number
of Individuals
161
333
132
3732
1385
5078
1364
2362
20100
Notes: The fixed effect model in Equation 1 is applied, which includes all sociodemographic and job-related control variables as well as individual, region and wave fixed effects. *** p < 0.01,
** p < 0.05, * p < 0.1.
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

17
3.4. Robustness Checks
We have conducted additional analyses to confirm the robustness of our results. Firstly, our
findings are robust to different definitions of jobs at high risk of automation. Table A-3 shows
that the results are consistent when we use the first or second quartile as cut-offs to
dichotomise the variable into high- and low-risk, instead of the third quartile. Furthermore,
Table A-4 confirms the results if we use the probability of automation as a continuous
variable. Secondly, the results are also consistent if we use a seven-point Likert-type scale to
define job and life satisfaction and apply a fixed-effects ordered logit model (Baetschmann et
al., 2020), as shown in Table A-5.
4. Discussion
Using data from the Understanding Society survey from 2009-22, we find a significant
association between working in a job at high risk of automation and lower job satisfaction in
UK workers, but no consistent relationship to their overall life satisfaction. These findings
agree with results for the USA (Nazareno & Schiff, 2021; Liu, 2022), Australia (Lordan &
Stringer, 2022) and 29 European countries (Gorny & Woodard, 2020).
The negative relationship we find between job automation risk and job satisfaction may be
interpreted as the result of job insecurity induced by the pre-existing or impending
introduction of automation technologies perceived by workers to pose a risk to their jobs.
Echoing findings from Nazareno & Schiff, 2021 and Lordan & Stringer, 2022, job insecurity
may lead to lower job satisfaction through increased stress at work and uncertainty about
future employment. Job insecurity is an established cause of lower work-related and general
wellbeing, which if chronic may in turn lead to persistent lower wellbeing and a deterioration
in mental and physical health. (de Witte, Pienaar & de Cuyper, 2016).
On the other hand, the lack of a significant association between working in a highly
automatable job and life satisfaction can be qualified by the observed differences across the
three dimensions of life satisfaction (Table 3), where we find a significant negative
association with satisfaction with income (a job-related dimension) compensated by a
significant positive association with satisfaction with the amount of leisure time. Where this is
the result of existing implementation of automation technologies, this finding may indicate a
transition to shorter working hours for lower pay. Conversely, where this is in anticipation of
the introduction of automation technologies, these findings may indicate a greater emphasis
on time spent outside of work driven by employees ‘quietly quitting’ in the face of job
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

18
insecurity.
Our analysis of demographic subgroups also reveals differential trends. The negative
relationship of risk of job automation with both life satisfaction and job satisfaction is only
observed among age groups 16-29 and, especially, 30-49. This broadly aligns with Schwabe
& Castellacci (2020), who suggest that younger employees may consider job replacement by
automation technologies a potential threat to future employment prospects while older
workers may expect these changes to occur beyond their working lifetimes. Additionally,
younger workers are more likely to be aware of the potentially negative consequences of the
adoption of new technologies for their future employment (Brougham & Haar, 2018). The
relatively greater magnitude of the negative relationship between automation risk and job
satisfaction among employees with a degree or other higher degree could reflect higher
expectations for their jobs given the greater investment of time and energy in their human
capital (Schwabe & Castellacci, 2020).
We also find evidence of lower job satisfaction among employees working in highly
automatable jobs in agriculture, manufacturing, transport and services. Occupations in these
industries are particularly susceptible to automation according to Frey & Osborne (2017),
which may lead to greater collective awareness among workers of the potential for
automation technologies to transform their work and leading to lower job satisfaction
primarily for the reasons of job insecurity described above (Brougham & Haar, 2018).
There are several limitations in our study. Firstly, we use automation risk, as defined by Frey
& Osborne (2017), to understand the extent to which automation technologies may have
influenced the working lives of respondents. Despite being a widely used reference measure
of automation risk, FO cannot differentiate employees working in roles already influenced by
automation technologies from those where such technologies have not been implemented. As
such, our measure of occupational automation risk unavoidably captures a mixed picture of
existing, anticipated and unanticipated exposure to automation technologies by workers.
The FO measures, in being limited to occupations that can be automated by computer-
controlled equipment, capture a narrower range of occupations than may be affected by
automation technologies. Furthermore, the six years since the publication of the FO measures,
and the ten years since the original primary data collection (Frey and Osborne, 2013),
represent a long time when considering technological change. Our sample also excludes self-
employed individuals due to the unavailability of job-related data, an area that deserves
further research. Finally, our use of retrospective survey data and estimates of occupational
automation risk by necessity exclude more recent technological advances including the advent
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

19
of large language models such as Chat GPT, for which no data are yet available.
Collectively, our findings indicate a complex, multifactorial relationship between
occupational automation risk, job satisfaction and wellbeing. Whether through anticipated or
existing exposure to automation technologies at work, it is younger workers and those
working in the most automatable industries who have the most negative associations between
occupational automation risk and job satisfaction. More broadly, life satisfaction appears less
sensitive to occupational automation risk, pointing to varying perceptions of the importance
of work in life. Further empirical research is required to determine the specific mechanisms
underlying the negative relationship between job automation and subjective wellbeing. These
could include higher levels of job insecurity, higher cognitive demands arising from the
automation of less cognitively demanding tasks, loss of meaningfulness of work, or increased
surveillance and control in the workplace.
Irrespective of the cause, our analysis identifies existing differences according to the risk of
job automation across industries and workers in how satisfying they find their work and to a
lesser extent in how this goes on to influence their wider satisfaction with their lives.
Directives and institutional policies introduced to ensure a fair transition to the widespread
use of automation technologies at work must thus consider differences between individuals
and industries, and whether the actual degree of job automation can influence wellbeing
outcomes among employees.
Contributorship statement
JZ: Conceptualisation, Formal Analysis, Writing - Original Draft
BR: Conceptualisation, Formal Analysis, Writing - Review & Editing
MB: Conceptualisation, Writing - Review & Editing, Supervision
JC: Conceptualisation, Writing - Review & Editing, Supervision
Financial disclosures
None.
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

20
Declaration of competing interest
The authors declare no competing interests in relation to the work.
Data availability
Understanding Society data used in this study were obtained from the UK Data Service
https://beta.ukdataservice.ac.uk/datacatalogue/series/series?id=2000053
Occupational automation risk scores were obtained from the Office for National Statistics
https://www.ons.gov.uk/employmentandlabourmarket/peopleinwork/employmentandemploye
etypes/datasets/probabilityofautomationinengland
Acknowledgements
This paper is part of the Pissarides Review of the Future of Work and Wellbeing, based at the
Institute for the Future of Work and funded by the Nuffield Foundation. Bertha Rohenkohl,
Jiyuan Zheng and Mauricio Barahona acknowledge support from the Nuffield Foundation.
Mauricio Barahona also acknowledges support by the EPSRC under grant EP/N014529/1
funding the EPSRC Centre for Mathematics of Precision Healthcare at Imperial. Jonathan
Clarke acknowledges support from the Wellcome Trust (215938/Z/19/Z). We would like to
thank Professor Christopher Pissarides for his helpful comments on earlier versions of this
manuscript.
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

Appendix A Tables and Figures
Figure A-1. The histogram of the probability of job automation for the entire sample.
Notes: The figure displays the histogram of the probability of job automation using measures of job automation
risk from Frey & Osborne (2017).
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

Table A-1. Sample representativeness
Percent (%) from Census Percent (%) from the sample
Ethnicity
White 87.166 81.46
Asian 6.922 10.37
Black/African/Caribbean 3.015 4.94
Mixed ethnic group 1.979 2.18
Arab and any other ethnic group 0.919 1.06
Gender
Women 50.891 54.40
Men 49.109 45.64
Age groups
16-29 23.048 19.99
30-49 34.281 33.54
50-64 22.446 24.57
>=65 20.225 21.89
NVQ Qualification
Level 1 15.374 7.22
Level 2 16.558 23.34
Level 3 13.236 27.61
Level 4+ 29.510 44.16
No qualifications 25.322 5.18
Notes: This table shows the percentage of population by ethnicity, gender, age group and the highest level of
qualifications across the UK based on UK 2011 Census data compared with the sample used in this study.
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

Table A-2. The relationship between working in a highly automatable job and life satisfaction and job satisfaction across regions in the UK.
Life satisfaction
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
North East North
West Yorkshire
and the
Humber
East
Midlands West
Midlands East of
England London South East South
West Wales Scotland Northern
Ireland
Highly
automatable
job
0.0162 0.0112 -0.000787 -0.0140 -0.0344
*
0.0188 -0.0382
**
-0.000936 0.00392 0.0168 -0.00791 -0.0343
(0.0253)
(0.0169)
(0.0181)
(0.0181)
(0.0191)
(0.0198)
(0.0183)
(0.0153)
(0.0185)
(0.0207)
(0.0164)
(0.0232)
Mean of
Dep. Var. 0.768 0.769 0.765 0.756 0.739 0.772 0.738 0.777 0.777 0.762 0.778 0.822
Number of
Obs 7044 18970 14875 14158 15081 16712 20176 23424 15575 12352 17281 11052
Number of
Individuals 1161 3306 2637 2488 2684 2921 4138 4116 2585 2208 2936 1981
Job satisfaction
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
North East North
West Yorkshire
and the
Humber
East
Midlands West
Midlands East of
England London South East South
West Wales Scotland Northern
Ireland
Highly
automatable
job
-0.0694
**
-0.0389
**
-0.0354 -0.00862 -0.0599
**
-0.0642
***
-0.0986
***
-0.0645
***
-0.0319 -0.0696
***
-0.0298 -0.0391
(0.0324)
(0.0189)
(0.0241)
(0.0201)
(0.0239)
(0.0215)
(0.0212)
(0.0167)
(0.0216)
(0.0234)
(0.0210)
(0.0287)
Mean of
Dep. Var.
0.760 0.779 0.790 0.785 0.766 0.788 0.776 0.797 0.801 0.794 0.786 0.822
Number of
Obs
7044 18970 14875 14158 15081 16712 20176 23424 15575 12352 17281 11052
Number of
Individuals 1161 3306 2637 2488 2684 2921 4138 4116 2585 2208 2936 1981
Notes: The fixed effect model in Equation 1 is applied, which includes all sociodemographic and job-related control variables as well as individual, region and wave fixed effects. *** p
< 0.01, ** p < 0.05, * p < 0.1.
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

Robustness Checks
Table A-3. The relationship between working in a highly automatable job with different cut-offs and
life satisfaction/job satisfaction.
(1) (2)
Life satisfaction
Job satisfaction
Probability of
automation>=1st quartile
Working in a highly
automatable job -0.00527
(0.00476) -0.0252
***
(0.00543)
Mean of dep. var. 0.767 0.787
Number of obs 187630 187630
Number of individuals 32468 32468
Probability of
automation>=2nd quartile
Working in a highly
automatable job
-0.00453
(0.00463) -0.0364
***
(0.00543)
Mean of dep. var.
0.767
0.787
Number of obs 187630 187630
Number of individuals
32468
32468
*** p<=0.01.
Table A-4. The relationship between the probability of job automation and life satisfaction/job
satisfaction.
(1) (2)
Life
satisfaction
Job satisfaction
Probability of job
automation -0.0352 -0.266***
(0.0244) (0.0285)
Mean of Dep. Var. 0.767 0.787
Number of Obs 187630 187630
Number of Individuals 32468 32468
*** p<=0.01.
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

Table A-5. The marginal effect of working in a highly automatable job on life satisfaction and job
satisfaction based on the fixed effects ordered logit model.
(1) (2)
Life satisfaction
Job satisfaction
Working in a highly
automatable job
Completely
dissatisfied
0.0000534
0.00478
***
(0.000428) (0.000673)
Mostly dissatisfied
0.000173
0.00810
***
(0.00138) (0.00114)
Somewhat dissatisfied
0.000249
0.0171
***
(0.00200) (0.00240)
Neither satisfied nor
dissatisfied 0.000234 0.0150
***
(0.00188)
(0.00211)
Somewhat satisfied 0.000262 0.0206
***
(0.00210)
(0.00291)
Mostly satisfied -0.000675 -0.0333
***
(0.00542) (0.00469)
Completely satisfied -0.000296 -0.0322
***
(0.00238) (0.00454)
Number of Obs 185475 183556
Number of Individuals 31748 31262
Notes: The table shows estimates of the marginal effect of working in a highly automatable job on life satisfaction
and job satisfaction using the fixed effects ordered logit model. *** p<=0.01.
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

References
Acemoglu, D. & Restrepo, P. (2018) The Race between Man and Machine: Implications of
Technology for Growth, Factor Shares, and Employment. American Economic Review. 108
(6), 1488–1542. doi:10.1257/aer.20160696.
Adamczyk, W.B., Monasterio, L. & Fochezatto, A. (2021) Automation in the future of public
sector employment: the case of Brazilian Federal Government. Technology in Society. 67,
101722. doi:10.1016/j.techsoc.2021.101722.
Albuquerque, P.H.M., Saavedra, C.A.P.B., de Morais, R.L. & Peng, Y. (2019) The Robot
from Ipanema goes Working: Estimating the Probability of Jobs Automation in Brazil. Latin
American Business Review. 20 (3), 227–248. doi:10.1080/10978526.2019.1633238.
Autor, D.H. & Dorn, D. (2013) The Growth of Low-Skill Service Jobs and the Polarization of
the US Labor Market. American Economic Review. 103 (5), 1553–1597.
doi:10.1257/aer.103.5.1553.
Autor, D.H., Katz, L.F. & Kearney, M.S. (2006) The Polarization of the U.S. Labor Market.
American Economic Review. 96 (2), 189–194. doi:10.1257/000282806777212620.
Baetschmann, G., Ballantyne, A., Staub, K.E. & Winkelmann, R. (2020) feologit: A new
command for fitting fixed-effects ordered logit models. The Stata Journal. 20 (2), 253–275.
doi:10.1177/1536867X20930984.
Berriman, R. & Hawksworth, J. (2017) Will Robots Steal Our Jobs? The Potential Impact of
Automation on the UK and Other Major Economies.
Bowling, N.A., Eschleman, K.J. & Wang, Q. (2010) A meta-analytic examination of the
relationship between job satisfaction and subjective well-being. Journal of Occupational and
Organizational Psychology. 83 (4), 915–934. doi:10.1348/096317909X478557.
Brougham, D. & Haar, J. (2018) Smart Technology, Artificial Intelligence, Robotics, and
Algorithms (STARA): Employees’ perceptions of our future workplace. Journal of
Management & Organization. 24 (2), 239–257. doi:10.1017/jmo.2016.55.
Clément S. Bellet, Jan-Emmanuel De Neve, George Ward (2023) Does Employee Happiness
Have an Impact on Productivity?. Management Science 0(0).
Cobb-Clark, D.A. (2015) Locus of control and the labor market. IZA Journal of Labor
Economics. 4 (1), 3. doi:10.1186/s40172-014-0017-x.
Cubel, M., Nuevo

Chiquero, A., Sanchez

Pages, S. & Vidal

Fernandez, M. (2016) Do
Personality Traits Affect Productivity? Evidence from the Laboratory. The Economic Journal.
126 (592), 654–681. doi:10.1111/ecoj.12373.
De Neve, J. E. (2018). Work and well-being: A global perspective. Global Happiness and
Well-Being Policy Report 2018.
De Witte, H., Pienaar, J. & De Cuyper, N. (2016) Review of 30 Years of Longitudinal Studies
on the Association Between Job Insecurity and Health and Well

Being: Is There Causal
Evidence? Australian Psychologist. 51 (1), 18–31. doi:10.1111/ap.12176.
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

Diener, E., Inglehart, R. & Tay, L. (2013) Theory and Validity of Life Satisfaction Scales.
Social Indicators Research. 112 (3), 497–527. doi:10.1007/s11205-012-0076-y.
Frey, C.B. & Osborne, M.A. (2017) The future of employment: How susceptible are jobs to
computerisation? Technological Forecasting and Social Change. 114, 254–280.
doi:10.1016/j.techfore.2016.08.019.
Georgellis, Y., Tsitsianis, N. & Yin, Y.P. (2009) Personal Values as Mitigating Factors in the
Link Between Income and Life Satisfaction: Evidence from the European Social Survey.
Social Indicators Research. 91 (3), 329–344. doi:10.1007/s11205-008-9344-2.
Ghislieri, C., Molino, M. & Cortese, C.G. (2018) Work and Organizational Psychology Looks
at the Fourth Industrial Revolution: How to Support Workers and Organizations? Frontiers in
Psychology. 9. https://www.frontiersin.org/articles/10.3389/fpsyg.2018.02365.
Gorny, P.M. & Woodard, R.C. (2020) Don’t Fear the Robots: Automatability and Job
Satisfaction. MPRA Paper. https://ideas.repec.org/p/pra/mprapa/103424.html.
Hicks, S., Tinkler, L. & Allin, P. (2013) Measuring Subjective Well-Being and its Potential
Role in Policy: Perspectives from the UK Office for National Statistics. Social Indicators
Research. 114 (1), 73–86. doi:10.1007/s11205-013-0384-x.
Highhouse, S. & Becker, A.S. (1993) Facet measures and global job satisfaction. Journal of
Business and Psychology. 8 (1), 117–127. doi:10.1007/BF02230397.
Hill, T.D., Davis, A.P., Roos, J.M. & French, M.T. (2020) Limitations of Fixed-Effects
Models for Panel Data. Sociological Perspectives. 63 (3), 357–369.
doi:10.1177/0731121419863785.
Hinks, T. (2021) Fear of Robots and Life Satisfaction. International Journal of Social
Robotics. 13 (2), 327–340. doi:10.1007/s12369-020-00640-1.
Hsiao, C. (1985) Benefits and limitations of panel data. Econometric Reviews. 4 (1), 121–174.
doi:10.1080/07474938508800078.
Institute for Social and Economic Research (2021) Understanding Society: Waves 1-11,
2009-2020 and Harmonised BHPS: Waves 1-18, 1991-2009, User Guide.
Johnson, A., Dey, S., Nguyen, H., Groth, M., Joyce, S., Tan, L., Glozier, N. & Harvey, S.B.
(2020) A review and agenda for examining how technology-driven changes at work will
impact workplace mental health and employee well-being. Australian Journal of
Management. 45 (3), 402–424. doi:10.1177/0312896220922292.
Khubchandani, J. & Price, J.H. (2017) Association of Job Insecurity with Health Risk Factors
and Poorer Health in American Workers. Journal of Community Health. 42 (2), 242–251.
doi:10.1007/s10900-016-0249-8.
Layard R, De Neve J-E. (2023a) The Quality of Work. In: Wellbeing: Science and Policy.
Cambridge University Press; 2023:178-201.
Layard R, De Neve J-E. (2023b) Unemployment. In: Wellbeing: Science and Policy.
Cambridge University Press; 2023:166-177.
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

Liu, L. (2022) Job quality and automation: Do more automatable occupations have less job
satisfaction and health? Journal of Industrial Relations. 00221856221129639.
doi:10.1177/00221856221129639.
Lordan, G. & Stringer, E.-J. (2022) People versus Machines: The Impact of Being in an
Automatable Job on Australian Worker’s Mental Health and Life Satisfaction. 101.
Makridis, C.A. & Han, J.H. (2021) Future of work and employee empowerment and
satisfaction: Evidence from a decade of technological change. Technological Forecasting and
Social Change. 173 (C).
https://ideas.repec.org/a/eee/tefoso/v173y2021ics0040162521005953.html.
Nazareno, L. & Schiff, D.S. (2021) The impact of automation and artificial intelligence on
worker well-being. Technology in Society. 67, 101679. doi:10.1016/j.techsoc.2021.101679.
Nieß, C. & Zacher, H. (2015) Openness to Experience as a Predictor and Outcome of Upward
Job Changes into Managerial and Professional Positions. PLOS ONE. 10 (6), e0131115.
doi:10.1371/journal.pone.0131115.
Office for National Statistics (n.d.) Ethnic group, national identity and religion - Office for
National Statistics.
https://www.ons.gov.uk/methodology/classificationsandstandards/measuringequality/ethnicgr
oupnationalidentityandreligion [Accessed: 11 December 2022].
Organisation for Economic Co-operation and Development (2022) Entrepreneurship -
Enterprises by business size - OECD Data. http://data.oecd.org/entrepreneur/enterprises-by-
business-size.htm [Accessed: 1 December 2022].
Oswald, A.J., Proto, E. & Sgroi, D. (2015) Happiness and productivity. Journal of labor
economics. doi:10.1086/681096.
Pehkonen, J., Viinikainen, J., Böckerman, P., Pitkänen, N., Lehtimäki, T. & Raitakari, O.
(2019) Health endowment and later-life outcomes in the labour market: Evidence using
genetic risk scores and reduced-form models. SSM - Population Health. 7, 100379.
doi:10.1016/j.ssmph.2019.100379.
Rohenkohl B. & Clarke J. (2023) What do we know about automation at work and workers’
wellbeing? Literature Review Working Paper, Institute for the Future of Work.
doi:10.5281/zenodo.8074436.
Schwabe, H. & Castellacci, F. (2020) Automation, workers’ skills and job satisfaction. PLOS
ONE. 15 (11), e0242929. doi:10.1371/journal.pone.0242929.
Stankevi
č
i
ū
t
ė
, Ž., Staniškien
ė
, E. & Ramanauskait
ė
, J. (2021) The Impact of Job Insecurity
on Employee Happiness at Work: A Case of Robotised Production Line Operators in
Furniture Industry in Lithuania. Sustainability. 13 (3), 1563. doi:10.3390/su13031563.
Tamers, S.L., Streit, J., Pana-Cryan, R., Ray, T., Syron, L., Flynn, M.A., Castillo, D., Roth,
G., Geraci, C., Guerin, R., Schulte, P., Henn, S., Chang, C.-C., Felknor, S. & Howard, J.
(2020) Envisioning the future of work to safeguard the safety, health, and well-being of the
workforce: A perspective from the CDC’s National Institute for Occupational Safety and
Health. American Journal of Industrial Medicine. 63 (12), 1065–1084.
doi:10.1002/ajim.23183.
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint

UK Data Service (2014) InFuse. 29 May 2014. https://infuse.ukdataservice.ac.uk/ [Accessed: 22 June
2023].
University of Essex, Institute for Social and Economic Research (2022) Understanding
Society: Waves 1-12, 2009-2021 and Harmonised BHPS: Waves 1-18, 1991-2009.
10.5255/UKDA-SN-6614-18.
Viinikainen, J., Böckerman, P., Hakulinen, C., Kari, J.T., Lehtimäki, T., Raitakari, O.T. &
Pehkonen, J. (2022) Schizophrenia polygenic risk score and long-term success in the labour
market: A cohort study. Journal of Psychiatric Research. 151, 638–641.
doi:10.1016/j.jpsychires.2022.05.041.
Wang, H., Cheng, Z., Zhe Wang, B. & Chen, Y. (2021) Childhood left-behind experience and
labour market outcomes in China. Journal of Business Research. 132, 196–207.
doi:10.1016/j.jbusres.2021.04.014.
White, S., Lacey, A. & Ardanaz-Badia, A. (2019) The probability of automation in England:
2011 and 2017. 25 March 2019.
https://www.ons.gov.uk/employmentandlabourmarket/peopleinwork/employmentandemploye
etypes/articles/theprobabilityofautomationinengland/2011and2017 [Accessed: 1 March 2023].
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 18, 2024. ; https://doi.org/10.1101/2024.01.18.24301484doi: medRxiv preprint