Book Read Free

The Economics of Artificial Intelligence

Page 46

by Ajay Agrawal


  cult or too costly to hire a high- skill worker;

  furthermore, and perhaps more important, the low- occupation employee

  is expected to stay longer in the fi rm than higher- skill employees, which in

  turn encourages the fi rm to invest more in trust- building and fi rm- specifi c

  human capital and knowledge. Overall, such low- occupation employees can

  make a big diff erence to the fi rm’s performance.

  This alternative view of AI and fi rms is consistent with the work of theo-

  rists of the fi rm such as Luis Garicano. Thus in Garicano (2000) down-

  stream, low- occupation employees are consistently facing new problems;

  among these new problems they sort out are those they can solve them-

  selves (the easier problems) and the more diffi

  cult questions they pass on

  to upstream—higher- skill—employees in the fi rm’s hierarchy. Presum-

  ably, the more innovative or more AI- intensive the fi rm is, the harder it

  is to solve the more diffi

  cult questions, and therefore the more valuable

  the time of upstream high- occupation employees becomes; this in turn

  makes it all the more important to employ downstream, low- occupation

  employees with higher ability to make sure that less problems will be passed

  on to the upstream, high- occupation employees within the fi rm so that

  these high- occupation employees will have more free time to concentrate

  on solving the most diffi

  cult tasks. Another interpretation of the higher

  complementarity between low- occupation and high- occupation employees

  in more innovative (or more AI- intensive) fi rms, is that the potential loss

  from unreliable low- occupation employees is bigger in such fi rms: hence

  the need to select out those low- occupation employees that are not reliable.

  268 Philippe Aghion, Benjamin F. Jones, and Charles I. Jones This higher complementarity between low- occupation workers and other

  production factors in more innovative (or more AI- intensive) fi rms in turn

  increases the bargaining power of low- occupation workers within the fi rm

  (it increases their Shapley Value if we follow Stole and Zwiebel [1996]).

  This in turn explains the higher payoff for low- occupation workers. It also

  predicts that job turnover should be lower (tenure should be higher) among

  low- occupation workers who work for more innovative (more AI- intensive)

  fi rms than for low- occupation workers who work for less innovative fi rms,

  whereas the turnover diff erence should be less between high- occupation

  workers employed by these two types of fi rms. This additional prediction is

  also confronted to the data in Aghion et al. (2017).

  Note that so far R&D investment has been used as the measure of the

  fi rm’s innovativeness or frontierness. We would like to test the same predic-

  tions, but using explicit measures of AI intensity as the RHS variable in the

  regressions (investment in robots, reliance on digital platforms). Artifi cial

  intelligence and fi rm organizational form: recent empirical studies (e.g., see

  Bloom et al. 2014) have shown that the IT revolution has led fi rms to elimi-

  nate middle- range jobs and move toward fl atter organizational structure.

  The development of AI should reinforce that trend, while perhaps also

  reducing the ratio to low- occupation to high- occupation jobs within fi rms

  as we argued above.

  A potentially helpful framework to think about fi rms’ organizational

  forms is Aghion and Tirole (1997). There, a principal can decide whether or

  not to delegate authority to a downstream agent. She can delegate author-

  ity in two ways: (a) by formally allocating control rights to the agent (in

  that case we say that the principal delegates formal authority to the agent);

  or (b) informally through the design of the organization, for example, by

  increasing the span of control or by engaging in multiple activities: these

  devices enable the principal to commit to leave initiative to the agent (in

  that case we say that the principal delegates real authority to the agent).

  And agents’ initiative particularly matters if the fi rm needs to be innova-

  tive, which is particularly the case for more frontier fi rms in their sectors.

  Whether she decides to delegate formal or only real authority to her agent,

  the principal faces the following trade- off : more delegation of authority to

  the agent induces the agent to take more initiative; on the other hand, this

  implies that the principal will lose some control over the fi rm, and there-

  fore face the possibility that suboptimal decisions (from her viewpoint) be

  taken more often. Which of these two counteracting eff ects of delegation

  dominates, will in turn depend upon the degree of congruence between the

  principal’s and the agent’s preference, but also about the principal’s ability

  to reverse suboptimal decisions.

  How should the introduction of AI aff ect this trade- off between loss of

  control and initiative? To the extent that AI makes it easier for the princi-

  pal to monitor the agent, more delegation of authority will be required in

  Artifi cial Intelligence and Economic Growth 269

  order to still elicit initiative from the agent. The incentive to delegate more

  authority to downstream agents, will also be enhanced by the fact that with

  AI, suboptimal decision- making by downstream agents can be more easily

  corrected and reversed: in other words, AI should reduce the loss of control

  involved in delegating authority downstream. A third reason for why AI may

  encourage decentralization in decision- making has to do with coordination

  costs: namely, it may be costly for the principal to delegate decision- making

  to downstream units if this prevents these units from coordinating within the

  fi rm (see Hart and Holmstrom 2010). But here again, AI may help overcome

  this problem by reducing the monitoring costs between the principal and

  its multiple downstream units, and thereby induce more decentralization

  of authority.

  More delegation of authority in turn can be achieved through various

  means: in particular, by eliminating intermediate layers in the fi rm’s hier-

  archy, by turning downstream units into profi t centers or fully independent

  fi rms, or through horizontal integration that will commit the principal to

  spending time on other activities. Overall, one can imagine that the develop-

  ment of AI in more frontier sectors should lead to larger and more horizon-

  tally integrated fi rms, to fl atter fi rms with more profi t centers, which out-

  source an increasing number of tasks to independent self- employed agents.

  The increased reliance on self- employed independent agents will in turn

  be facilitated by the fact that, as well explained by Tirole (2017), AI helps

  agents to quickly develop individual reputations. This brings us to the third

  aspect of AI and organizations on self- employment. Artifi cial intelligence

  and self- employment: as stressed above, AI favors the development of self-

  employment for at least two reasons: fi rst, it may induce AI intensive fi rms

  to outsource tasks, starting with low- occupation tasks; second, it makes it

  easier for independent agents to dev
elop individual reputations. Does that

  imply that AI should result in the end of large integrated fi rms with individu-

  als only interacting with each other through platforms? And which agents

  are more likely to become self- employed?

  On the fi rst question: Tirole (2017) provides at least two reasons for why

  fi rms should survive the introduction of AI. First, some activities involve

  large sunk costs and/or large fi xed costs that cannot be borne by a single indi-

  vidual. Second, some activities involve a level of risk- taking that also may

  not be borne by one single agent. To this we should add the transaction cost

  argument that vertical integration facilitates relation- specifi c investments in

  situations of contractual incompleteness: Can we truly imagine that AI will

  by itself fully overcome contractual incompleteness?

  On the second question: our above discussion suggests that low- skill ac-

  tivities involving limited risk and for which AI helps develop individual rep-

  utations (hotel or transport services, health assistance to the elder and/or

  handicapped, catering services, house cleaning) are primary candidates for

  increasingly becoming self- employment jobs as AI diff uses in the economy.

  270 Philippe Aghion, Benjamin F. Jones, and Charles I. Jones And indeed recent studies by Saez (2010), Chetty et al. (2011), and Kleven

  and Waseem (2013) point to low- income individuals being more respon-

  sive to tax or regulatory changes aimed at facilitating self- employment.

  Natural extensions of these studies would be to explore the extent to which

  such regulatory changes have had more impact in sectors with higher AI

  penetration.

  The interplay between AI and self- employment also involves potentially

  interesting dynamic aspects. Thus it might be worth looking at whether

  self- employment helps individuals accumulate human capital (or at least

  protects them against the risk of human capital depreciation following the

  loss of a formal job), and the more so in sectors with higher AI penetra-

  tion. Also interesting would be to look at how the interplay between self-

  employment and AI is itself aff ected by government policies and institu-

  tions, and here we have primarily in mind education policy and social or

  income insurance for the self- employed. How do these policies aff ect the

  future performance of currently self- employed individuals, and are they

  at all complemented by the introduction of AI? In particular, do currently

  self- employed individuals move back to working for larger fi rms, and how

  does the probability of moving back to a regular employment vary with

  AI, government policy, and the interplay between the two? Presumably, a

  more performing basic education system and a more comprehensive social

  insurance system should both encourage self- employed individuals to bet-

  ter take advantage of AI opportunities and support to accumulate skills

  and reputation and thereby improve their future career prospects. On the

  other hand, some may argue that AI will have a discouraging eff ect on

  self- employed individuals, if it lowers their prospects of ever reintegrating

  a regular fi rm in the future, as more AI- intensive fi rms reduce their demand

  for low- occupation workers.

  9.6 Evidence on Capital Shares and Automation to Date

  Models that conceptualize AI as a force of increasing automation suggest

  that an upswing in automation may be seen in the factor payments going to

  capital—the capital share. In recent years, the rise in the capital share in the

  United States and around the world has been a central topic of research.

  For example, see Karabarbounis and Neiman (2013), Elsby, Hobijn, and

  S¸ahin (2013), and Kehrig and Vincent (2017). In this section, we explore this

  evidence, fi rst for industries within the United States, second for the motor

  vehicles industry in the United States and Europe, and fi nally by looking

  at how changes in capital shares over time correlate with the adoption of

  robots.

  Figure 9.6 reports capital shares by industry from the US KLEMS data of

  Jorgenson, Ho, and Samuels (forthcoming); shares are smoothed using an

  HP fi lter with smoothing parameter 400 to focus on the medium- to long-

  Artifi cial Intelligence and Economic Growth 271

  run trends. It is well- known that the aggregate capital share has increased

  since at least the year 2000 in the US economy. Figure 9.6 shows that this

  aggregate trend holds up across a large number of sectors, including agricul-

  ture, construction, chemicals, computer equipment manufacturing, motor

  vehicles, publishing, telecommunications, and wholesale and retail trade.

  The main place where one does not see this trend is in services, including

  education, government, and health. In those sectors, the capital share is

  relatively stable or perhaps increasing slightly since 1990. But the big trend

  one sees in these data from services is a large downward trend between 1950

  and 1980. It would be interesting to know more about what accounts for

  this trend.

  While the facts are broadly consistent with automation (or an increase

  in automation), it is also clear that capital and labor shares involve many

  other economic forces as well. For example, Autor et al. (2017) suggest that a

  composition eff ect involving a shift toward superstar fi rms with high capital

  shares underlies the industry trends. That paper and Barkai (2017) propose

  that a rise in industry concentration and markups may underlie some of

  the increases in the capital share. Changes in unionization over time may

  be another contributing factor to the dynamics of factor shares. This is all

  to say that a much more careful analysis of factor shares and automation is

  required before any conclusions can be drawn.

  Keeping that important caveat in mind, fi gure 9.7 shows evidence on the

  capital share in the manufacturing of transportation equipment for the

  United States and several European countries. As Acemoglu and Restrepo

  (2017) note (more on this below), the motor vehicles industry is by far the

  industry that has invested most heavily in industrial robots during the past

  two decades, so this industry is particularly interesting from the standpoint

  of automation.

  The capital share in transportation equipment (including motor vehicles,

  but also aircraft and shipbuilding) shows a large increase in the United

  States, France, Germany, and Spain in recent decades. Interestingly, Italy

  and the United Kingdom exhibit declines in this capital share since 1995.

  The absolute level diff erences in the capital share for transportation equip-

  ment in 2014 are also interesting, ranging from a high of more than 50 per-

  cent in the United States to a low of around 20 percent in recent years in

  the United Kingdom. Clearly it would be valuable to better understand

  these large diff erences in levels and trends. Automation is likely only a part

  of the story.

  Acemoglu and Restrepo (2017) use data from the International Federation

  of Robots to study the impact of the adoption of industrial robots on the

  US labor market. At the industry level, this data is available for the decade


  2004 to 2014. Figure 9.8 shows data on the change in capital share by indus-

  try versus the change in the use of industrial robots.

  Two main facts stand out from the fi gure. First, as noted earlier, the motor

  Fig. 9.6 US capital shares by industry

  Source: The graph reports capital shares by industry from the U.S. KLEMS data of Jorgenson, Ho, and Samuels (2017).

  Note: Shares are smoothed using an HP fi lter with smoothing parameter 400.

  Fig. 9.7 The capital share for transportation equipment

  Sources: Data for the European countries are from the EU- KLEMS project (http:// www

  .euklems .net/ ) for the “transportation equipment” sector, which includes motor vehicles, but also aerospace and shipbuilding; see Jägger (2016). US data are from Jorgenson, Ho, and Samuels (2017) for motor vehicles.

  Note: Shares are smoothed using an HP fi lter with smoothing parameter 400.

  Fig. 9.8 Capital shares and robots, 2004– 2014

  Sources: The graph plots the change in the capital share from Jorgenson, Ho, and Samuels (2017) against the change in the stock of robots relative to value added using the robots data from Acemoglu and Restrepo (2017).

  274 Philippe Aghion, Benjamin F. Jones, and Charles I. Jones vehicles industry is by far the largest adopter of industrial robots. For example, more than 56 percent of new industrial robots purchased in 2014

  were installed in the motor vehicles industry, the next highest share was

  under 12 percent in computers and electronic products.

  Second, there is little correlation between automation as measured by

  robots and the change in the capital share between 2004 and 2014. The

  overall level of industrial robot penetration is relatively small, and as we

  discussed earlier, other forces including changes in market power, unioniza-

  tion, and composition eff ects are moving capital shares around in a way that

  makes it hard for a simple data plot to disentangle.

  Graetz and Michaels (2017) conduct a more formal econometric study

  using the EU- KLEMS data and the International Federation of Robotics

  data from 1993 until 2007, studying the eff ect of robot adoption on wages

  and productivity growth. Similar to what we show in fi gure 9.8, they fi nd

  no systematic relationship between robot adoption and factor shares. They

  do suggest that adoption is associated with boosts to labor productivity.

 

‹ Prev