by Ajay Agrawal
already discussed in section 8.2, the history of technology of the last two cen-
turies is full of examples of automation, ranging from weaving and spinning
machines to the mechanization of agriculture, as discussed in the previous
section. Even with capital accumulation and the deepening of automation,
if there were no other counteracting force, we would see the share of labor
in national income declining steadily. Our conceptual framework highlights
a major force preventing such a decline—the creation of new tasks in which
labor has a comparative advantage.
This can be seen by putting together equations (7) and (10), which yields
R
W
(11)
d ln W = ln
ln
dN
( N
1)
( N )
M
L
W
R
+ ln
ln
dI +
1
( dN dI ),
( I )
( I )
N
I
L
M
and also from equation (6),
ds = dN
dI.
L
For the labor share to remain stable and for wages to increase in tandem
with productivity, as has been the case historically, we need I—capturing
the extensive margin of automation—to grow by the same amount as the
range of new tasks, N. When that happens, equilibrium wages grow propor-
tionately with productivity, and the labor share, s , remains constant, as can
L
be seen from the fact that the fi rst line of equation (11) is in this case equal
to the increase in productivity or gross domestic product (GDP) per worker.
Indeed, rewriting equation (11) imposing dN = dI, we have
( N )
( I )
d ln W = ln
L
ln
L
dI > 0,
( N
1)
( I )
M
M
which is strictly positive because of assumption (A1).
8.4.6 A False Dichotomy: Recap
With our conceptual framework exposited in a more systematic manner,
we can now briefl y revisit the false dichotomy highlighted in the introduc-
tion. Our analysis (in particular equation [7]) highlights that there is always
a negative displacement eff ect on labor resulting from automation. Equa-
tion (11) reiterates that there is no presumption that this displacement eff ect
could not reduce overall demand for labor.
However, several countervailing eff ects imply that a negative impact from
automation on labor demand is not a forgone conclusion. Most important,
the productivity eff ect could outweigh the displacement eff ect, leading to an
expansion in labor demand and equilibrium wages from automation. The
Artifi cial Intelligence, Automation, and Work 219
presence of the productivity eff ect as counterweight to the displacement
created by automation highlights an important conceptual issue, however.
In contrast to the emphasis in the popular discussions it is not the brilliant,
superproductive automation technologies that threaten labor, but the “so-
so” ones that create the displacement eff ect as they replace labor in tasks
that it previously performed, but do not engender the countervailing pro-
ductivity eff ect.
The productivity eff ect is supplemented by the capital accumulation
that automation sets in motion and the deepening of automation, which
increases the productivity of machines in tasks that have already been auto-
mated. But even with these countervailing eff ects, equation (9) shows that
automation will always reduce the share of labor in national income. All the
same, this does not signal the demise of labor either, because the creation
of new tasks in which labor has a comparative advantage could counterbal-
ance automation, which is our interpretation of why the demand for labor
has kept up with productivity growth in the past despite several rapid waves
of automation.
Our framework suggests that the biggest shortcoming of the alarmist
and the optimist views is their failure to recognize that the future of labor
depends on the balance between automation and the creation of new tasks.
Automation will often lead to a healthy growth of labor demand and wages
if it is accompanied with a commensurate increase in the set of tasks in
which labor has a comparative advantage—a feature that alarmists seem to
ignore. Even though there are good economic reasons for why the economy
will create new tasks, this is neither a forgone conclusion nor something
we can always count on—as the optimists seem to assume. Artifi cial intel-
ligence and robotics could be permanently altering this balance, causing
automation to pace ahead of the creation of new tasks with negative conse-
quences for labor, at the very least in regard to the share of labor in national
income.
8.4.7 Generalizations
Many of the features adopted in the previous subsection are expositional
simplifi cations. In particular, the aggregate production function (1) can be
taken to be any constant elasticity of substitution aggregate. One impli-
cation of this would be that aggregate output in equation (3) would be a
constant elasticity aggregate itself. This does not aff ect any of our main
conclusions, including the negative impact of automation on the labor share
(see Acemoglu and Restrepo 2016).10
We also do not need assumption (A1) for any of the results. If the second
10. Recent work by Aghion, Jones, and Jones (2017) points out, however, that if the elasticity of substitution between tasks is less than one and there is an exogenous and high saving rate, the labor share might asymptote to a positive value even with continuously ongoing auto mation.
220 Daron Acemoglu and Pascual Restrepo
inequality in this assumption does not hold, changes in automation tech-
nology have no impact on the equilibrium because it is not cost eff ective to
adopt all available automation technologies (for this reason, in the general
case, Acemoglu and Restrepo [2016] distinguish technologically automated
tasks from equilibrium automation). Given our focus here, there is no loss
of generality in making this assumption.
A fi nal feature that is worth commenting on is the fact that in the aggregate
production function (1), the limits of integration are N – 1 and N, ensuring that the total measure of tasks is one. This is useful for several reasons. First,
when the introduction of new tasks expands the total measure of tasks, it
becomes more challenging to obtain a balanced growth path (see Acemoglu
and Restrepo 2016). Second, in this case some minor modifi cations are nec-
essary so that an expansion in the total measure of tasks leads to productiv-
ity improvements. In particular, consider the general case where the elastic-
ity of substitution between tasks is not necessarily equal to one. If it is
greater than one, an increase in N leads to higher productivity, but not nec-
essarily when it is less than or equal to one. In this latter case, we then need
to introduce direct productiv
ity gains from task variety. For example, in
the present case where the elasticity of substitution between tasks is equal
to one, we could modify (1) to ln Y = (1 / N)
N
0 ln[ N 1+ y( i )], where ≥ 0
represents these productivity gains from task variety and ensures that the
qualitative results explicit here continue to apply.
8.4.8 Employment
and
Unemployment
An additional generalization concerns the endogenous adjustment of
employment in the face of new automation technologies. We have so far
taken labor to be supplied inelastically for simplicity. There are two ways in
which the level of employment responds to the arrival of new technologies.
The fi rst is via a standard labor supply margin. Acemoglu and Restrepo
(2016) show that the endogenous adjustment of labor supply, including
income eff ects and the substitution of consumption and leisure, links the
level of employment to the share of labor in national income.
The second possibility is through labor market frictions, for example, as
in Acemoglu and Restrepo (2018a). Under appropriate assumptions, the
endogenous level of employment in this case is also a function of the share
of labor in national income. Though both models with and without labor
market frictions endogenize employment as a function of the labor share,
their normative implications are potentially diff erent, as we discuss below.
For now, however, the more important implication of such extensions
is to link the level of employment (or unemployment) to labor demand.
Automation, when it reduces labor demand, will also reduce the level of
employment (or increase the level of unemployment). Moreover, because the
supply of labor depends on the labor share, in our framework automation
results in a reduction in employment (or an increase in unemployment). As
such, our analysis so far also sheds light on (and clarifi es the conditions for)
Artifi cial Intelligence, Automation, and Work 221
the claims that new automation technologies will reduce employment. It also
highlights, however, that the fact that automation has been ongoing does not
condemn the economy to a declining path of employment. If automation is
met by equivalent changes in the creation of new tasks, the share of labor in
national income can remain stable and ensure a stable level of employment
(or unemployment) in the economy.
8.5 Constraints and Ineffi
ciencies
Even in the presence of the countervailing forces limiting the displace-
ment eff ect from automation, there are potential ineffi
ciencies and con-
straints limiting the smooth adjustment of the labor market and hindering
the productivity gains from new technologies.
Here we focus on how the mismatch between skills and technologies not
only increases inequality, but also hinders the productivity gains from auto-
mation and new tasks. We then explore the possibility that, concurrent with
rapid automation, we are experiencing a slowdown in the creation of new
tasks, which could result in slow productivity growth. Finally, we examine
how a range of factors leads to excessive automation, which not only creates
ineffi
ciency but also hinders productivity.
8.5.1 Mismatch of Technologies and Skills
The emphasis on the creation of new tasks counterbalancing the potential
negative eff ects of automation on the labor share and the demand for labor
ignores an important caveat and constraint: the potential mismatch between
the requirements of new technologies (tasks) and the skills of the workforce.
To the extent that new tasks require skilled employees or even new skills to
be acquired, the adjustment may be much slower than our analysis so far
suggests.
To illustrate these ideas in the simplest possible fashion, we follow Acemo-
glu and Restrepo (2016) and assume that there are two types of workers,
low- skill with supply L and high- skill with supply H , both of them supplied inelastically. We also assume that low- skill workers can only perform
tasks below a threshold S ∈ ( I, N), while high- skill workers can perform all tasks. For simplicity, we assume that the productivity of both low- skill
and high- skill workers in the tasks that they can perform is still given by
( x).11 Low- skill workers earn a wage W and high- skill workers earn a L
L
wage W .
H
11. We can also introduce diff erential comparative advantages and also an absolute productivity advantage for high- skill workers, though we choose not to do so to increase transparency (see Acemoglu and Restrepo 2016). The more restrictive assumption here is that automation happens at the bottom of the range of tasks. In general, automation could take place in the middle range, and its impact would depend on whether automated tasks are competing pre-dominantly against low- skill or high- skill workers (see Acemoglu and Autor 2011; Acemoglu and Restrepo 2018b).
222 Daron Acemoglu and Pascual Restrepo
In this simple extension of the framework presented so far, the threshold
S can be considered as an inverse measure of the mismatch between new
technologies and skills. A greater value of S implies that there are plenty
of additional tasks for low- skill workers, while a low value of S implies the presence of only a few tasks left that low- skill workers can perform.
Assuming that in equilibrium W > W ,12 which implies that low- skill H
L
workers will perform all tasks in the range ( I, S), equilibrium wages satisfy W = Y ( N
S ) and W = Y ( S
I ).
H
H
L
L
Thus, the impact of automation on inequality—defi ned here as the wage
premium between high- and low- skill workers—is given by
d ln W / W
H
L =
1
> 0.
dI
S
I
This equation shows that automation increases inequality. This is not sur-
prising, since the tasks that are automated are precisely those performed by
low- skill workers. But in addition, it also demonstrates that the impact of
automation on inequality becomes worse when there is a severe skill mis-
match—the threshold S is close to I. In this case, displaced workers will be squeezed into a very small range of tasks, and hence, each of these tasks will
receive a large number of workers and will experience a substantial drop in
price, which translates into a sharp decline in the wage of low- skill workers.
In contrast, when S is large, displaced workers can spread across a larger set
of tasks without depressing their wage as much.
A severe mismatch also aff ects the productivity gains from automation.
In particular, we have
d ln( Y / L)
W
R
= ln
L
ln
> 0.
dI
( I )
( I )
L
M
This equation shows that the productivity gains from automation depend
positively on W / R: it is precisely when displaced workers have a high oppor-L
tunity cost that automation raises productivity. Using the fact that R =
( Y/ K)( I – N + 1), we obtain
W
K
L =
S
I
.
R
I
N + 1 L
A worse mismatch (a lower S) reduces the opportunity cost of displaced
workers further, and via this channel, it makes automation less profi table.
This is because a severe mismatch impedes reallocation, reducing the pro-
ductivity gains of freeing workers from automated tasks.
12. This is equivalent to [( N – S)/ ( S – I)] > ( H/ L), so that high- skill workers are scarce relative to the range of tasks that only they can produce.
Artifi cial Intelligence, Automation, and Work 223
Equally important are the implications of a skill mismatch for the pro-
ductivity gains from new tasks. Namely,
d ln( Y / L)
R
W
= ln
ln
H
> 0,
dN
( N
I )
( N )
M
H
which depends negatively on W / R: it is precisely when high- skill workers H
have a relatively high wage that the gains from new tasks will be limited. With
similar arguments to before, we also have
W
K
H
= N S
,
R
I
N + 1 L
which implies that in the presence of a worse mismatch (a lower S), the
productivity gains from new tasks will be limited. This is because new tasks
require high- skill workers who are scarce and expensive when S is low.
An important implication of this analysis is that to limit increasing
inequality and to best deploy new tasks and harness the benefi ts of auto-
mation, society may need to simultaneously increase the supply of skills. A
balanced growth process requires not only automation and the creation of
new tasks to go hand- in-hand, but also the supply of high- skill workers to
grow in tandem with these technological trends.
8.5.2 Automation at the Expense of New Tasks
As discussed in section 8.2, a puzzling aspect of recent macroeconomic
developments has been the lack of robust productivity growth despite the
bewildering array of new technologies. Our conceptual framework provides
three novel (and at least to us, more compelling) reasons for slow produc-
tivity growth. The fi rst was the skill mismatch discussed in the previous