Drivers of automation: who or what leads the way?

A new electronic road toll gate system in Indonesia

This week, we continue a special series of blogs based on our new paper on the future of economic development, work, and wages in developing countries.

Last week we asked what actually is automation? In this blog we take a look at the drivers of automation. In the following blogs we’ll assess whether 1.8bn jobs in the developing world will really be lost by automation; consider the public policy responses to automation, and finally conclude with a wrap up blog.

So, what drives automation?

Under what conditions might such a transformation or revolution take place? Technological feasibility is just one condition. There are multiple criteria which the decision to automate involves: can a task be automated in a way that reliably produces a good or service at a specified level of quality? Is it profitable to automate that task? Is it legally possible for a firm to replace workers with machines? How do relevant stakeholders such as political groupings, particularly trade unions, and society at large, particularly consumers, respond to automation (and the potentially ensuing lay-offs)?

Corresponding to these criteria, one could split the literature on automation into different theoretical approaches. Much recent research has focused on the technological feasibility of automation. Yet, automatable tasks do not necessarily or instantly get automated: one can observe a set of tasks currently being carried out both by humans and machines in different contexts and places. Consider, for instance, subway drivers and autonomous subways, supermarket cashiers, and self-checkout machines, university lecturers, and online courses. The coexistence of automated and non-automated modes of operation of the same task suggests that a narrowly technologically deterministic view is insufficient. There are less tangible but powerful—economic, political, social, and cultural—reasons to be factored in. Such factors up until now often seem to have been neglected in research on automation, but could be particularly important in the context of developing countries. They not only determine if automation occurs but the terms of automation vis-a-vis governing institutions.

An example from Indonesia

Consider, for example, the case of Indonesia. In Indonesia, there have been numerous media reports related to automation and employment impacts. The McKinsey Global Institute estimates that around half of all jobs in Indonesia are automatable using existing technologies. One example is that motorway toll booths are being automated to an e-payment system which has placed a question over 20,000 jobs, leading the Minister of Finance to announce at the annual meeting of the International Monetary Fund and the World Bank that automation might create a case for a future universal basic income in Indonesia.

While formerly each toll gate required five employees working in shifts to ensure vehicles had paid the road toll, the cashless system which is being rolled out runs entirely without human operators, thus speeding up the transaction process and reducing traffic congestion. Yet, as of early 2018, the toll road operator PT Jasa Marga asserts that “former tollgate keepers would instead be relocated to different positions within the company (…) and would keep their permanent employee status”.

There have, indeed, so far, been no reports of mass lay-offs despite the electronic system being implemented. What could be the reason? First, it could be that, as implementation is still in an early stage, lay-offs may be a matter of time, and could happen in a gradual manner. The company may also reduce its future intake of new employees as a result. Second, it could be that, in line with the quote above, PT Jasa Marga, which is currently expanding its business, truly has the capacity to absorb 20,000 people in other sectors of its operation. If that is the case, this raises the important question as to whether by raising overall productivity and competitiveness, automation somehow allowed the company to expand. The latter would mean that automation has the double effect of reducing labor demand per unit of capital in one domain (e.g. manual toll collection) while raising labor demand in complementary domains (e.g. administrative or construction tasks).

Finally, there is a set of institutional reasons that could be an important explanatory factor as to why PT Jasa Marga—a state-owned enterprise and thus facing potential developmental obligations—has not laid off workers: political and social-norms pressures as well as legal constraints could be preventing the toll road operator from firing employees. One could imagine the political backlash of a state-owned enterprise making 20,000 people unemployed. There may be also concerns over strikes, attacks on the new toll-booth machinery, political interventions (including fears of the political replacement of senior management making such decisions) or negative media reports which demonstrably influence business decisions in part of wholly owned SOEs and to some extent in private companies too.

In short, all of this suggest much current debate focuses too much on technological capabilities, and not enough on the economic, political, legal, and social factors that will profoundly shape the way automation affects employment. Questions like profitability, labour regulations, unionization, and corporate-social expectations will be at least as important as technical constraints in determining which jobs get automated, especially in developing countries.

We’ll continue the series of blogs with some discussion next of whether 1.8bn jobs in the developing world will really be lost by automation.

Lukas Schlogl is a Research Associate with the ESRC GPID Research Network at King’s College London. He works on structural change, digital transformation, and political behaviour in developing countries.

Andy Sumner is a Reader in International Development in the Department of International Development, King’s College London. He is Director of the ESRC Global Poverty & Inequality Dynamics (GPID) Research Network.