Automation, digitalisation and platforms: Implications for ...€¦ · processes, such as Kuhn’s The structure of scientific revolutions (Kuhn, 1962) or the punctuated equilibrium

RESEARCH REPORT

Automation, digitalisation and platforms:Implications for work and employment

Digital age

Automation, digitalisation and platforms:Implications for work and employment

European Foundationfor the Improvement ofLiving and WorkingConditions

Europe Direct is a service to help you find answers to your questions about the European Union.

Freephone number*: 00 800 6 7 8 9 10 11*Certain mobile telephone operators do not allow access to 00 800 numbers or these calls may be billed.

When citing this report, please use the following wording:Eurofound (2018), Automation, digitalisation and platforms: Implications for work and employment, PublicationsOffice of the European Union, Luxembourg.

Author: Enrique Fernández-Macías (Eurofound)

Research manager: Enrique Fernández-Macías

Eurofound project: The employment impact of the automation of services (170901)

The author would like to thank Brian Fabo, Vincenzo Maccarone, Rafael Muñoz de Bustillo Llorente, Martijn Poel,Lea Schmidlechner, Gérard Valenduc, Esperanza Vera-Toscano and Christopher Warhurst, as well as the colleaguesof the Digital age activity working group in Eurofound, for their very useful input to earlier versions of this paper.

Luxembourg: Publications Office of the European Union

Print: ISBN: 978-92-897-1651-2 doi:10.2806/385749 TJ-04-18-316-EN-CPDF: ISBN: 978-92-897-1652-9 doi:10.2806/13911 TJ-04-18-316-EN-N

This report and any associated materials are available online at http://eurofound.link/ef18002

© European Foundation for the Improvement of Living and Working Conditions, 2018

Reproduction is authorised provided the source is acknowledged.

For any use or reproduction of photos or other material that is not under the Eurofound copyright, permission mustbe sought directly from the copyright holders.

Cover image: © Zapp2photo/Shutterstock.com

The European Foundation for the Improvement of Living and Working Conditions (Eurofound) is a tripartiteEuropean Union Agency, whose role is to provide knowledge in the area of social, employment and work-relatedpolicies. Eurofound was established in 1975 by Council Regulation (EEC) No. 1365/75 to contribute to the planningand design of better living and working conditions in Europe.

European Foundation for the Improvement of Living and Working Conditions

Telephone: (+353 1) 204 31 00 Email: [email protected] Web: www.eurofound.europa.eu

Contents

Introduction 1

1. The foundations: Technological and socioeconomic change 5Division of labour 5Role of institutions 6

2. Attributes of the digital economy 9Flexibility of production 9Availability of information 10Zero marginal costs 10Network effects 12Conclusions 13

3. Implications for work and employment: Three vectors of change 15Automation of work 16Digitalisation of processes 18Coordination by platforms 19

4. Commentary 23

References 25

iii

Glossary3D printers: machines that can create physical objects from three-dimensional digital models, generally by layingdown successive layers of material.

Algorithm: a set of precisely defined steps and rules to accomplish a task.

Automation of work: the replacement of (human) labour input by machine input for some types of tasks withinproduction and distribution processes.

Coordination by platforms: the use of digital networks to coordinate economic transactions in an algorithmic way.

Digital age: an historical period marked by the widespread use of digital technologies in different aspects of humanactivity.

Digital goods: strings of bits (digital information) that have economic value.

Digital revolution: a general acceleration in the pace of technological change in the economy, driven by a massiveexpansion of our capacity to store, process and communicate information using electronic devices.

Digitalisation of processes: the use of sensors and rendering devices to translate (parts of) the physical productionprocess into digital information (and vice versa).

Division of labour: the separation and allocation of tasks to different persons cooperating in an economic process.

Economic institutions: rules, structures and mechanisms of social coordination of the economic process.

Employment conditions: contractual and statutory conditions of the work relation that have an impact on the well-being of the worker.

Industrial relations: the relatively institutionalised ways in which workers and employers organise their relations andsettle their disputes.

Intellectual property rights: monopoly rights given to the creators of informational goods over their use andreproduction, for a given number of years, backed and imposed by the state.

Long-tail markets: massively large markets with near-perfect information, where there is economic value in theprovision of even extremely rare goods or services.

Massive Open Online Courses (MOOCs): free or very low-cost courses available on the internet that use online videosand texts, together with interactive exercises and algorithmic monitoring of progress, to provide an alternative to face-to-face education.

Network effects (also demand-side economies of scale): a situation in which the value for consumers of a particulartype of good increases with the number of users.

Occupations: coherent bundles of tasks that require specific skills, corresponding to different positions within thedivision of labour in society.

Tasks: units of work activity that produce output and which are coherently bundled into occupations.

Technology: in a general sense, the tools and methods used for carrying out the economic transformation process.

Internet of Things (IoT): sensors attached to outputs, inputs, components, materials or tools used in production.

Winner-take-all markets: markets in which a single provider of a particular type of good or service tends toconcentrate the vast majority of economic activity.

Working conditions: the physical and psychological requirements and attributes of work and its environment thathave an impact on the well-being of workers.

Zero marginal costs (as applied to digital goods): no marginal costs for non-rival and infinitely expandable digitalgoods.

iv

1

From the digital revolution to thedigital ageThe digital revolution can be defined as a generalacceleration in the pace of technological change in theeconomy, driven by a massive expansion of our capacityto store, process and communicate information usingelectronic devices. Although some of its key underlyingtechnologies and scientific foundations were developedbetween the 1950s and 1970s, the ‘big bang’ ofinnovations and applications of digital technologies wastriggered by the invention of the microprocessor in theearly 1970s – a general-purpose programmableelectronic device capable of processing digitalinformation. The continuous increase in performanceand decrease in the cost of microprocessors over thenext four decades facilitated a very rapid spread ofdifferent digital technologies, such as the personalcomputer, the internet and mobile phones.

The digital age can be defined as a historical periodmarked by the widespread use of digital technologies indifferent aspects of human activity, including theeconomy, politics and most forms of human interaction.This widespread use of digital technologies implies aprofound transformation of social, economic andpolitical systems, in the same way as the steam engineor electricity transformed past societies. This paper willset out a conceptual and analytical framework to assessthe implications of the digital age on work andemployment.

In order to understand why it is important now to studythe implications of the digital age, this introduction willprovide some historical context and a broadinterpretation of the significance of the digitalrevolution, synthesised in three main contentions thatunderlie the approach of the rest of this paper. Thesearguments derive from the work of economists ChrisFreeman and Francisco Louçã (Freeman and Louçã,2001), and Carlota Pérez (Pérez, 2003).

The first contention is that changes in the methods andtools used in the economy tend to cluster aroundperiodic ‘revolutions’, rather than following linear andincremental trends. The reasons are both technological

and socioeconomic. From a technological perspective,since each new technology is essentially arecombination of previous ones, the introduction of anew general-purpose key technology, such as themicrochip, opens up a myriad of new possibilities ofrecombination and applications. This generates aself-reinforcing process of fast technological change,with each ‘new’ technology opening up furtherpossibilities until they are eventually exhausted.

From a socioeconomic perspective, since productiontechnologies are embedded in social structures, theintroduction of new technologies will initially struggleagainst the existing organisational forms, culturalattitudes, vested interests and institutional settings(consistent with the pre-existing productiontechnologies). However, when such resistance isovercome, the same organisational forms, interests andinstitutions can foster the diffusion and furtherdevelopment of these new technologies. Thesetechnological and socioeconomic factors givetechnological change a ‘syncopated rhythm’ similar toother and in some ways related to evolutionaryprocesses, such as Kuhn’s The structure of scientificrevolutions (Kuhn, 1962) or the punctuated equilibriumof biological evolution.

Thus, the digital revolution represents the most recentof a long sequence of periodic bursts of innovation andchange in the tools and methods used in the economy.This is all due, as already mentioned, to the invention ofthe microprocessor and microchip – a general-purposetechnology that has seen a steady reduction inproduction costs and an equally steady increase incapabilities. It has created a whole new set of productsand industries with massive investment opportunities,but it has also created socioeconomic imbalances.Indeed, the microchip has facilitated new forms ofeconomic organisation that have slowly spread to moreand more sectors and activities – a process that isongoing.

As with previous technological revolutions, the digitalrevolution requires a paradigm change in theorganisation of the economy, which in turn will bringabout new social structures and the need for newinstitutions.

Introduction

2

The second contention is that there is a time lagbetween the initial big bang of innovation provoked bya technological revolution and its full transformation ofthe socioeconomic structure. As previously mentioned,productive technologies are embedded insocioeconomic structures, and their change, on a largescale, requires a transformation of infrastructures,organisational practices and institutional frameworks,overcoming the explicit (or implicit) resistance of theexisting dominant actors and industries.

A typical sequence, from technological revolution tosocioeconomic transformation, starts with theappearance of new products and industries, initially atthe margins of the economy, but then growing very fast.This rapid growth would attract investment, providingleverage for further innovation and growth, as well asthe necessary funding for the installation of newinfrastructures and the development of furtherapplications.

This is the initial period (the ‘installation’) of atechnological revolution, which in the model ofFreeman and Pérez generally lasts about three decades.It is a period marked by growing imbalances betweenthe old and new industries, and the firms and workersthat benefit from the new technologies. It is also oftenassociated with a speculative frenzy that ends in afinancial crisis (Pérez, 2003). In this installation period,the transformational power of the new technologiesremains mostly limited to the associated industries andthose most directly related, such as the building ofassociated infrastructures. The financial crisis serves as

a turning point and a cleansing mechanism for thepossible excesses of the installation period,consolidating the structures of the new industries andreducing any excessive levels of expectations.

After this crisis, the new technologies are mature, thenew infrastructures have been installed, and the skillsand knowhow required for the new tools and methodsare widely diffused. Then, the new technologies canspread to other industries and activities where their fullpotentials can be realised and put into practice. Thissecond period of the technological revolution, whichgenerally takes another three decades after the turningpoint of the crisis, is what Pérez calls the ‘deploymentperiod’. Over this period, the possibilities afforded bythe technological revolution are slowly depleted; thisleads to a period of stagnation that prepares thegroundwork for the next technological revolution,which then starts the process all over again. This longcycle theory of technological revolution, the most well-known of which was the Fordism of automobiles, oil andmass production (starting around 1908 and finishing inthe early 1970s), fits surprisingly well with thedevelopment of the digital revolution.

The invention of the microprocessor in the early 1970sfacilitated the creation of many new products andcustomer markets in the margins of the economy, suchas videogames and microcomputers. These marketsexperienced very fast growth rates and developed toolsand methods that were subsequently applied to othernew and fast-growing products and markets, especiallyin the development of the internet and mobile phones.

Automation, digitalisation and platforms: Implications for work and employment

This paper roughly follows the approach of Chris Freeman, Francisco Louçã and Carlota Pérez, who interpret thedigital revolution as the fifth technological revolution of capitalism over the last 200 years (Freeman and Louçã,2001; Pérez, 2003). The four previous technological revolutions were: the initial Industrial Revolution (circa 1771);the steam and railways revolution (circa 1829); the steel, electricity and heavy engineering revolution (circa 1875);and the oil, automobile and mass production revolution (circa 1908). Each of those revolutions triggered aparadigm shift of the economy, and the cycle of installation-crisis-deployment-stagnation (Pérez, 2003),discussed in this chapter.

However, in the literature there are also very different arguments about the historical significance of the digitalrevolution. The most extreme and opposing view has been propounded by the American economic historianRobert J. Gordon (2016), who interprets digital technologies as a

peripheral set of innovations mostly relevant for leisure industries, but with very little effects on growth in thelong run (in fact, the digital age would coincide with a period of secular stagnation, since the fruits of theIndustrial Revolution have been already reaped).

Others speak about a third Industrial Revolution (Rifkin, 2011) or even a fourth (Schwab, 2017); such argumentsare closer to the Freeman-Pérez framework, although more loosely constructed. The very influentialMassachusetts Institute of Technology (MIT) researchers Erik Brynjolfsson and Andrew McAfee interpret it as‘the second machine age’, giving it a much larger historical significance, equivalent to that of the originalIndustrial Revolution (Brynjolfsson and McAfee, 2014). Furthermore, some scholars believe the digital revolutionto be the trigger of an evolutionary leap in humankind equivalent to that of the emergence of Homo sapiens(Kurzweil, 2005; Harari, 2016).

Box 1: The historical significance of the digital revolution

3

The huge profits of these new industries attracted ever-increasing levels of investment with very highexpectations, which ultimately led to the burst of thedotcom bubble in 2001 (perhaps extending to thefinancial crash of 2008). According to this cyclicalmodel, the deployment period of the digital revolutionshould be starting now, when the new tools andmethods have diffused throughout the entiresocioeconomic structure, and the real economictransformation takes place.

This leads to the third contention: for a technologicalrevolution to produce valued and shared benefits tosociety, the institutional framework has to significantlychange in order to deal with the broad socioeconomicimplications of the new forms of economic activity.Again, this is a corollary of the social embeddedness ofproductive technologies. The institutional framework ofmarket economies has to deal with the externalities andcontradictions created by economic activity, forinstance, providing employment insurance or incomeredistribution, but it also performs some importantregulatory functions, such as employment regulations,competition policies, demand stimulation, educationand R&D policies. It is clear that a technologicalrevolution that implies a transformation in the tools andmethods used in the economy will also require asignificant change in the institutional framework thatregulates and helps to coordinate such an economy.

Indeed, the history of previous technologicalrevolutions shows that they have been associated withprofound changes in economic regulation and stateintervention in response to increasing socioeconomicimbalances and contradictions in the installation phase– following the sequence previously presented.

For instance, the Keynesian welfare state andemployment regulation model can be interpreted as aninstitutional response to the imbalances andcontradictions created by the Fordist mass productionsystem (Boyer, 1990). By regulating industrial conflictand employment relations, redistributing income andstimulating demand, the Keynesian welfare modelfacilitated the full deployment of the Fordist massproduction system and ensured that its benefits weremore widely shared by the population.

The Keynesian welfare model is an example of asuccessful reorganisation of the institutional frameworkto deal with the imbalances and contradictions createdby a technological revolution, in this case the Fordistmass production system. However, it is important toemphasise that a successful reorganisation of theinstitutional framework cannot automatically beassumed. Institutional reorganisations are the result ofpolitical processes which have their own logic, which isbeyond the scope of this paper.

Technological revolutions tend to generatesocioeconomic imbalances and contradictions whichthe existing institutional framework (developed in andfor a different context) cannot resolve. This is likely tolead to some form of political crisis with anindeterminate outcome. It can, for example, be asuccessful reorganisation of the institutionalframework, or perhaps an unsuccessful one, or even nomajor change at all. This historical argument can also beapplied to the subject of this paper. The digitalrevolution has created significant imbalances andcontradictions over the last few decades that are (atleast partly) the result of an increasing incongruencebetween the underlying economic structure and theinstitutional framework; this is highlighted by increasesin income inequality, as well as economic and politicalinstability.

As digital technologies and the associatedorganisational changes – automation, digitalisation andplatforms (discussed later) – extend to more and moresectors of the economy, the contradictions are likely tobecome even greater. That is why now, at this historicalconjecture, it is particularly important to improve ourunderstanding of how the digital revolution changes thenature of economic activity, work and employment.This knowledge should assist the democratic politicalprocess in redesigning and reorganising theinstitutional framework of the economy, ensuring thatthe digital age is one of prosperity and progress for all –the ultimate aim of Eurofound’s programme of work onthe digital age and its implications for work andemployment.

The rest of this report is divided in four chapters.Chapter 1 discusses how technology, the division oflabour and institutions interact with each other as theytransform socioeconomic structures.

Chapter 2 presents some of the key attributes of thedigital economy, inferred from the observation of theindustries and sectors that are at the forefront of thedigital revolution.

Chapter 3 introduces three key vectors of change –automation, digitalisation and platforms. In theauthor’s view, these are of greatest significance forunderstanding the implications of the digital age forwork and employment.

Finally, Chapter 4 presents some final remarks andconsiderations to help guide a research programmebased on this conceptual and analytical framework.

Introduction

5

From a material perspective, the economy can bedefined as the process of combining and transforminginputs into outputs in order to produce goods andservices for human needs. Thus, technology can bedefined as the tools and methods used for carrying outthese processes. Analysed in greater detail, atechnology can be understood as a domesticatednatural phenomenon: a device that allows thereproduction and control of a mechanism observed innature. Most significantly, once a technology has beenperfected, it becomes a building block that can becombined with others to form more advancedtechnologies. Furthermore, technologies can facilitatethe discovery and domestication of new naturalphenomena (the way microscopes opened up thecellular and microbial levels, and with that a myriad ofnew possibilities for the life sciences and associatedtechnologies). The possibilities inherent in technology –of recombinations and of uncovering further naturalphenomena – lead some authors to describetechnological change as an ‘autopoietic’ self-maintaining process, which builds on itself (Arthur,2009). The more technologies that become available,the more the possibilities for the recombination anddiscovery of new phenomena and, therefore, thepossibility for further technological development.The result is a self-reinforcing process of technologicalprogress – ultimately, the accumulation of (applied)knowledge.

It is important to note that even from this purelytechnological perspective, technological change is notseen as a continuous process, but one that ispunctuated by periodic bursts of innovation.Technologies tend to cluster in domains, groups oftechnologies that share a family of effects, a commonpurpose or underlying theory. These become toolboxesfor the assembly of new technologies or applications.Bursts of innovation often happen when a new domainis opened up by the discovery or domestication of a newtype of natural phenomenon, making a new toolbox oftechnologies available for further recombination andapplications. Also, innovation often results from‘redomaining’ (Arthur, 2009), where the application ofan existing solution to a different problem allows thetranslation of the entire toolbox associated with theexisting domain into a new one. This is purely a‘technological’ account of technological change, onethat helps in understanding its peculiar self-reinforcing,punctuated and accelerating nature. However, for thepurposes of this report, it is problematic because thereis obviously something missing. The economy may be

seen as a transformational process, but it still needsagents to enact such a transformation – human beings,whose input into the production process must becoordinated. In this respect, it is useful to distinguishbetween the two types of mechanisms for thecoordination of human input in production processes –technical (the division of labour) and social (thesocioeconomic institutions).

Division of labour ‘Division of labour’ refers to the separation andallocation of tasks to different persons cooperating inan economic process. It is an attribute of economicactivity that is as important and universal as technologyitself. It acts as a mechanism for coordinating the inputof different individuals towards a common (productive)goal, which can enormously increase the efficiency ofany type of cooperative production. This increase inefficiency is the result of specialisation (which increasesthe dexterity of workers) and a better coordination ofthe labour input (which reduces time between tasks,facilitates standardisation, as well as other efficiencyprocedures).

Division of labour can be understood as a very generalpurpose organisational method (and therefore – insome ways – a technology). However, it is so universaland important on its own that it is better to consider itseparately. A more difficult question to answer iswhether the division of labour is a technical attribute ofthe economy or a social one. On one hand, it isobviously a social attribute, because it is a form of socialcoordination, as already argued. On the other hand,division of labour can be considered to be a technicalattribute of economic activity, since it is a way ofincreasing the efficiency of a productive processindependently of the interests and values of the peopleinvolved.

In this last sense, division of labour can be neatlydifferentiated from economic institutions, which (aswill be discussed later) aim to coordinate workers associal beings rather than as inputs into a productiveprocess. Thus, the division of labour is a technicalattribute of economic activity: it is a method that allowsfor better coordination and efficiency of labour inputinto the economic process. And it is also a socialattribute, since the division of labour is a form of socialcoordination that gives rise to a social structure. Therelationship between the division of labour andtechnology is fundamental to how economicdevelopment works in both directions.

1 The foundations: Technologicaland socioeconomic change

6

Technological change shapes the evolution of thedivision of labour, by directly changing the productionprocess and the types of labour input necessary. Theintroduction of new technology into a productionprocess that is organised in a particular way will alwaysrequire some kind of reorganisation of work: some tasksmay change or become unnecessary, while others maybe created anew. Consequently, the skills, positions andconditions of workers within the process will alsochange.

But the division of labour is also a key enabler oftechnological change. First, the breakdown ofproduction processes into separate tasks facilitates abetter identification of problems and potentialtechnological solutions. Second, the specialisation ofworkers increases their knowledge of the economicprocess and therefore their capacity to develop newtools and methods. In general terms, the division oflabour expands human knowledge of the productionprocess and therefore facilitates innovation andtechnological change. A good example of how thedivision of labour and technological change feed intoeach other is automation – the replacing of human inputby machine input for certain production tasks.

Historically, a specific division of labour (andspecialisation) has been a precondition of automation,but only if the processes are broken down into verysimple, specific tasks that can be automated. Theautomation of certain tasks has been, in the long run, akey determinant of the opening up of the division oflabour: for instance, the importance of routine manualtasks as forms of human labour has decreaseddramatically in modern economies.

Technology and the division of labour form the materialfoundation of the economy as a transformative process.However, the coordination of human input intoproduction is not only a technical problem, but also asocial one. Humans have different needs, interests andvalues, and their input in production requires rules,structure and mechanisms of social coordination –namely, institutions.

Role of institutionsInstitutions support the functioning of economicprocesses by providing stability and social coordination,and by dealing with their external effects. Institutionsmake the economic process socially sustainable,allowing the material process of economictransformation to proceed while respecting the fabric ofsociety. Given that institutions are necessary for theeconomy to function, why not recognise them as acategory of technical solution to the issue of socialcoordination? For instance, the coordination ofeconomic activity by the mechanisms of markets andfirms anchored in institutions such as property rights,contract regulations and enforcement could be seen asan organisational technology that facilitates a moreefficient coordination of economic activity. There is aninherent logic to this: any set of defined rules andprinciples of behaviour is a method, an algorithm and,therefore, a technology (in this case, one of socialcoordination). However, it is important to differentiatebetween organisational methods that explicitly aim atthe social coordination of human interaction(institutions) and organisational methods thatostensibly aim at the technical coordination of humaninput into a productive process (the division of labour orwork organisation). The latter may also have socialimplications (giving rise to power structures, forinstance); however, this effect is unintended (in thesame way that a technology can have socialimplications). The whole point of an institution, incontrast, is its social implications.

This may seem like an unnecessary distinction, but itcan be important for discussions on economic policy:technology and the division of labour form thetechnological substrata of the socioeconomic system,which can be associated with very different institutionalframeworks. This explains the wide institutionalvariations that can exist between economies whoseoverall use of technology and division of labour (theireconomic development) is very similar. Perhaps thebest example is that for a long time a similar underlyingeconomic process (similar technologies and division oflabour) existed in two radically different institutionalforms – capitalism and state socialism. It is theinstitutional framework that determines most directlythe distribution of life chances across the population,even if this distribution is constrained by productivepossibilities set by the underlying economic structure.In summary, the benefits of technological change andeconomic development can be differently distributed,depending on the institutional framework that eachsociety sets for itself.


7

As with the division of labour, the relationship betweentechnology and economic institutions is fundamentaland works both ways. Technological change anddevelopments in the division of labour are continuouslychanging the nature and structure of economic activity.This changes the needs, interests and values ofeconomic agents and erodes the stabilising andcoordinating role of economic institutions. Sooner orlater, economic institutions have to reorganise andadapt to new technologies used in production. This alsoapplies to the need to reorganise the institutionalframework of the economy in the digital age.

For example, the Internet of Things (IoT) promises a bigleap in the efficiency of industrial processes, but it canalso transform a factory into an invasive surveillancesystem.1 The existing regulations in industrial labourcannot deal with such developments: they were notdesigned for such a factory. Hence, regulations need tobe changed to ensure that production is carried out inaccordance with an employee’s expectation of privacyand personal autonomy.

However, the relationship between technology andeconomic institutions works in both directions:institutions also shape technological development.First, because human agency, the ultimate driver oftechnological change is fundamentally structured byinstitutions. For instance, the ownership rights ofmarket economies place most investment decisions inthe hands of capital owners, who can steertechnological development towards their particularinterests, unlike a system in which investment decisionsare democratically made. Second, institutions can alsobe explicitly and directly tasked with redirectingtechnological change, because of the (expected) effectsof such change. For instance, some types of technologycan be prohibited by law, if their expected effectviolates societal norms (as has happened with sometypes of genetic engineering).

The foundations: technological and socioeconomic change

On the basis of the arguments discussed in this chapter, four different aspects of the implications of technologicalchange for work and employment can be differentiated.

£ Tasks and occupations: the distribution of tasks in the economy and the occupational structure that aredirectly and continuously changing as a result of technological advances (every new technology involvessome new way of carrying out a particular process, and therefore a change in the associated tasks).

£ Conditions of work: the physical, psychological and environmental requirements and conditions of work(also directly affected by the technology used).

£ Conditions of employment: the contractual and social conditions of the work, including issues such asstability, opportunities for development and pay (these mostly depend on the institutional framework andlabour regulation, with the effect of technology being more indirect).

£ Industrial relations: the relatively institutionalised ways in which workers and employers organise theirrelations and settle their disputes; the effect of technological change on this domain is also indirect (affectingthe three previous aspects in the areas of interests, power and organisational capacity of workers andemployers).

Tasks and occupations and conditions of work are two aspects of the division of labour and part of the materialattributes of the economy where the effect of technological change is direct and immediate (it can changedirectly the types of tasks needed in production and the conditions in which work takes place). In contrast,conditions of employment and industrial relations are part of the social and institutional attributes of theeconomy; the effect of technological change on them is indirect and more indeterminate.

Box 2: Implications of technological change on work and employment

1 In terms of manufacturing, the Internet of Things relies on the use of cheap digital sensors to digitally monitor every single object in a factory.

9

The key technology behind the digital revolution is themicroprocessor. It is the quintessential general-purposetechnology, since it can be applied to any type ofprocess that involves information. Microprocessor-based technologies and devices have been developedfor the processing, storage and communication ofinformation of all kinds. The possibilities forrecombinations and new applications are growingrapidly. The steady reduction in production costs andincrease in capabilities of microprocessor-basedtechnologies further leverages their applicability andcombinatory possibilities.

In terms of its general applicability, the microprocessorcan only be compared with such historical innovationsas steam power and electricity. This comparisonsuggests that it takes a significant amount of time foreconomic agents to grasp the full possibilities of a newgeneral-purpose technology, and to transformeconomic processes accordingly. Historically, keyinnovations often start out as curiosities. They are thenslowly applied to the most obvious and directly relatedindustries and activities (such as artificial light in thecase of electricity). Only after a significant time lag arethey rolled out to all types of industries and activities,reaching their full transformational potential. Forinstance, the use of steam power in industrial processesrequired factories to be organised around one or severalcentral large engines. This type of organisation wasretained for a period after the introduction of electricity,despite the fact that electricity allowed for the use ofsmaller motors and therefore a more flexible andefficient modular organisation of production. It tookengineers some time to notice this possibility andreorganise factories accordingly; the debate inengineering schools about the relative benefits of thetwo systems lasted several decades (McAfee andBrynjolfsson, 2017).

The diffusion and application of digital technologiesfollowed a similar process. They appeared mostsignificant firstly in the information and communicationtechnologies (ICT) sector itself, transforming a marginalactivity into a massive industry. Digital technologiesthen spread into related activities, such as media,leisure industries and telecommunications. They arenow diffusing to (and transforming) all types ofeconomic activity, including retail, manufacturing,health and education.

The diffusion of digital technologies across all types ofeconomic activity also involves a diffusion of the skillsand work methods of the ICT sector itself. In fact, it cansometimes involve a direct colonisation of other typesof economic activity by the big players of the ICTindustry. This can be seen in the examples of Amazon in

retail, and Google and Facebook in advertising andmedia.

But how do digital technologies transform economicprocesses? How does a digital economy differ from ananalogue (pre-digital) economy? To a large extent, thisis still an open question, since the transformativepotential of digital technologies has by no means beenexhausted. However, by looking at the sectors andindustries where digital technologies have already had amajor impact, it might be possible to gauge how digitaltechnologies can transform economic processes.

This report will emphasise four key aspects of digitaltechnologies that – in the author’s view – havesignificant transformative potential for economicactivity:

£ flexibility of production £ availability of information £ zero marginal costs£ network effects

Flexibility of productionUntil recently, machines applied in any productiveprocess tended to be relatively rigid. The functionalityof the machine was physically encoded in itsmechanical design: a change of function or operationrequired a physical change in the design of the machine.With this use of mechanically-assisted productionprocesses (the classic example being the assembly lineof Fordism), human operators were the factor providingflexibility to the system, dealing with unforeseencircumstances or giving final touches to the finalproduct, including any necessary customisation. Incontrast, however, digitally-enabled productionprocesses are programmed, so the process is notembodied in fixed mechanisms; instead, the process iscontrolled by algorithms that can be recalibrated asneeded. This applies to any type of digitally-enabledproduction process, whether it be informational (suchas administrative processes controlled with databasesoftware) or that of physical goods (for instance, anindustrial robotic arm that can be programmed toperform different types of operations).

The programmability and algorithmic control ofproduction processes makes them intrinsically muchmore flexible than previous methods of mechanically-controlled devices. But how far can this flexibility go?Ultimately, it depends on the processing poweravailable to the algorithms. Since it has beenexponentially growing in the last few decades, thedegree of programmability and flexibility inherent in

2 Attributes of the digital economy

10

digitally controlled processes has also grown at thesame rate. Artificial intelligence and deep learningalgorithms, for instance, can directly observe theirenvironment and learn whatever tasks they areassigned, with minimal human intervention.Theoretically, therefore, an algorithm could ultimatelybe as flexible and adaptable as a human being – maybeeven more so (even if it is impossible for us to imaginehow). This is why the digital revolution could take theautomation of labour input in production to theextreme, making human labour redundant. Algorithmsthat can do anything that a human being can do couldmake human labour unnecessary.

Availability of informationDigital technologies make information more available atall levels and points of the economic process. Thisreduces transaction costs, facilitates more complexorganisational structures, expands marketopportunities and makes location increasinglyirrelevant.

Over 80 years ago, British economist Ronald Coaseargued that firms exist because some types oftransactions are too costly to coordinate by markets(Coase, 1937). Most of the costs of those internalisedtransactions are, in fact, associated with limited orimperfect information. The increasing accessibility andubiquity of information associated with digitaltechnologies, therefore, predictably leads to asignificant increase in the outsourcing of specific tasksand functions to other companies, even outside thenational boundaries. This has deepened and expandedmarkets to an unprecedented extent, significantlycontributing to globalisation. The global value chains ofmultinational corporations would not be possiblewithout the information and communicationcapabilities of digital technologies.

In recent years, the combination of instant and almostunlimited availability of information with the principleof algorithmic control, discussed in the previous point,has given rise to an even more radical challenge to theargument of Coase on the boundaries between marketsand firms. Digital platforms such as Uber perform someof the functions of markets (providing a space wheresuppliers and consumers of certain services can meet);however, they also perform some of the functions offirms (coordinating, monitoring and disciplining thesupply of services through algorithms). In fact, it isprobably correct to say that platforms transcend bothmarkets and firms: they provide functions of both butcan do even more than either (they facilitate economictransactions that neither markets nor firms couldcoordinate).

Defying the distinction between firms and markets,platforms also defy existing forms of labour and marketregulations, as attested by some recent court cases in

Europe. As argued in previous chapters, platforms are anew form of economic activity that probably requiresnew regulations and institutions.

Another important effect of the massive expansion anddeepening of markets enabled by digital technologies isthe creation of long-tail markets, where the demand forlow-market products and services can collectivelyexceed that of large mainstream goods, subject toeffective distribution channels. In massively big marketswith near perfect information, there is economic valuein the provision of even extremely rare goods or nicheservices. This is reinforced by the possibilities ofcustomised digitally-enabled production processes, aspreviously mentioned. The contrast with the massproduction technologies of the 20th century is stark:instead of homogeneous national markets for mass-produced goods, digital technologies enable highlyspecialised long-tail markets on a global scale.

But easier access to digital information can also createwinner-take-all markets, in which a single provider of aparticular type of good or service tends to concentratethe vast majority of economic activity. The increasedlevel of information on the quality of goods and servicesavailable in digital markets removes one of the keytraditional advantages of local markets – the trustprovided by short-range transactions. If a global onlineretailer provides detailed and reliable information on aproduct (including buyers’ reviews), secures thetransaction and provides fast and real-time trackabledelivery, why buy it in the local store at a higher price?Furthermore, the long-tail effect ensures that big onlineproviders will have a massively wider range of productsto choose from. Thus, big online global providers arelikely to take a very significant share of the market, withpotentially damaging effects in terms of marketcompetition and inequality.

Zero marginal costsThe third important aspect of digital technologiesconcerns digital goods rather than digital technologiesdirectly. Digital goods can be defined as strings of bits(digital information) that have economic value. Thegeneralised use of digital technologies in productiontends to make digital goods more central for theeconomy with low or even zero marginal costs.

In economics, the marginal cost is the increase in totalcosts associated with the production of an additionalunit of good or service: in a textbook competitivemarket, prices would tend to equal marginal costs.Above marginal costs, producers would increase thesupply of the product, bringing down the price; whereasif below marginal costs, the product is not profitable.

Digital goods tend to have zero marginal costs becausethey are non-rival and infinitely expandable. They arenon-rival because their use by someone does not makethem less useful for anyone else: a piece of music does


11

not lose value if someone listens to it, whereas asandwich loses all its value if someone eats it. They arealso infinitely expandable because they can be infinitelyreproduced at (virtually) no cost – a digitised piece ofmusic can be freely copied an infinite number of times.Therefore, in a competitive market, non-rival andinfinitely expandable goods would have zero marginalcosts and therefore a price of zero.

But although the use and reproduction of a digital goodhas no cost, its production (creation) does. Thisgenerates an incentive problem in an economy whereproduction is driven by profit: nobody would producegoods that are costly to produce and that generate norevenue, even if there were a demand for them. Inmarket economies, different institutions have beencreated to deal with this incentive problem, whichapplies to any kind of informational good (includingideas, many forms of art and communication), not justdigital goods. The most important of these institutionsis that of intellectual property rights (IPR).

In principle, IPR give the creators of informational goodsmonopoly rights over their use and reproduction for agiven number of years, backed and enforced by thestate. The two most important types of IPR are patents(for inventions with industrial applicability) andcopyright (for creative, intellectual and artistic works).Most digital goods are protected by copyright, althoughpatents can also play an important role. For instance, inthe case of software, patents are often used to restrictthe use of generic ideas or procedures (such as aprogress bar for displaying how much of a task has beencompleted; The Guardian, 2005); copyright is used forthe particular form in which such ideas are expressed ina commercialised piece of software. It is important tonote that despite the similarity in their names, IPR differsignificantly from ordinary property rights (OPR) – thesocially-enforced rules that determine the use andownership of goods that are rival and not infinitelyexpandable. This difference arises from two key factors.

First, OPR defends the rights of the owner of a good,whereas IPR defends the rights of the producer of thegood. OPR primarily restrict the capacity of thirdpersons to use a good they do not rightfully own; incontrast, IPR primarily restrict the capacity of therightful owner of a good to make certain uses of it, suchas sharing or reproduction. An unintended effect of thisIPR-based restriction is that the enforcement of IPRrequires a much more intrusive surveillance, since itsfocus is on the private use of goods by their rightfulowners. In the case of digital goods, cheap andpervasive computers and online tools such as peer-to-peer trackers make the sharing of intellectual propertyextremely easy. This has led to increasingly intrusivemeasures of tracking and monitoring the private use ofdigital goods, such as the controversial use of digital

rights management (DRM) systems for e-books, whichscan users’ entire libraries and send the information tocorporate producers (Electronic Frontier Foundation,2014).

Second, while OPR does not in itself restrict thepotential benefits derived from the use of a good, IPRdoes. Since digital goods are non-rival and infinitelyexpandable, their potential use is infinite. For example,a piece of digital music can be shared any number oftimes without any deterioration; therefore, IPR reducesinfinitely the potential use of a good (that is, frominfinite to one). This contrasts with OPR, which concernsonly those entitled to use the good, but it does not limitits use otherwise: as long as someone eats a sandwich,all its potential benefits are realised. This effect isparticularly problematic in the case of patents, whichconcern inventions with industrial applicability: therestriction of potential uses of an idea may also restrictmany potential recombinations and further possibleapplications of that idea. Patents may incentiviseproduct creation, but they drastically limitrecombination, which is one of the most importantmechanisms of innovation, as previously discussed. IPRsystems may solve the incentive problem of zeromarginal costs for digital goods, but at the expense ofcreating two perhaps bigger issues – the need forintrusive enforcement and the drastic limitation of thepotential uses (including combinatorial innovation) ofdigital goods.

A third, institutionally troubling effect can be added.Because IPR are essentially government-sanctionedmonopolies, they immediately create a big incentive forproducers to lobby the government in order tostrengthen and expand IPR rights, which results infurther suboptimal outcomes. Some of the benefits ofinnovation are deployed in the intrinsicallyunproductive activity of lobbying, and to the extent thatlobbyists are successful, they simply expand rentswithout any benefit to society (see, for example,Depoorter, 2004).

IPR is not the only institution used for solving theincentive problem of informational goods. Historically,alternative methods such as procurement andpatronage have been used as well. The maincharacteristic of these alternatives is that they ‘divorcethe ex-ante incentive of an innovator from the ex poststream of rents generated by the innovation’ (Quah,2002, p. 27). In other words, they incentiviseinnovation/creation directly, by providing grants orawards for innovators/creators, making the resultingdigital goods or works generally available in the publicdomain, thus ensuring their potential benefits arefulfilled, even if authorship is fully recognised. These arethe systems most widely used for academic, space,military and basic research and development, whichunderlie most of the key innovations of the digitalrevolution (Mazzucato, 2015).

Attributes of the digital economy

12

Network effectsDigital technologies in economic processes tend tocreate demand-side economies of scale, or networkeffects. This means that the value for consumers ofmany types of digital goods and services increases withthe number of users. This effect is typical ofcommunication-related goods and services, a goodexample being telephony: the more users in thenetwork, the more people can be called, and thereforethe more value the service has for the users. A goodexample of network effects in today’s digital economy issocial networks, but it also applies to many other digitalgoods, services and technologies such as softwaresystems and tools, digital industrial applications (IoT)and industry standards.

Network effects lead to increasing returns in economicactivity, favouring market concentration. However, it isimportant to note that this effect is much stronger thanin traditional supply-side economies of scale (typical ofFordism) in which costs tend to decrease with moreoutput because high fixed costs are more diffused.Whereas supply-side economies of scale generally havelimits after which additional production impliesdiminishing returns, the limits of demand-sideeconomies of scale are much larger or even

non-existent, as with peer-to-peer systems used even inlarge commercial social networks, such as Facebook.

Perhaps most importantly, network effects can createconsumer lock-in, because the cost of switchingproduct or service also grows with the size of thenetwork, to the extent that it can effectively makecustomers entirely dependent on a particular vendor.For instance, it is nearly impossible to use the type ofsocial networking service provided by Facebook withoutusing Facebook itself, simply because (nearly) everyoneuses it. Switching to another service provider wouldrequire all of a user’s contacts to move simultaneously,which would then require their contacts to move aswell.

Other examples of lock-in resulting from network effectsis the dominance of Microsoft in the market for desktopoperating systems: again, since the value of anoperating system also depends on the number ofpeople using it – because we want to be able tocollaborate and share information – switching to adifferent solution provider would involve high costs interms of learning and require that many other peoplealso switch in order to maintain functionality. This iswhy the dominance of Microsoft only ended with theappearance of other computing devices beyond the


A more radical alternative to the IPR system, whose origins go back to the very beginning of the digital revolutionin the hacking culture of the first software programmers, is the open-source model. This is a model ofdecentralised production of digital goods (originally software, although it has been extended to many other typesof digital goods), where authorship, or rather its contribution, is recognised, but there are explicitly no limitationswith respect to the use, reproduction or modification of the good in question. The incentive to contribute toopen-source development is reputation rather than money, although that reputation can lead to monetary gainseventually – for instance, through better employment opportunities (Fernández-Macías, 2002).

In the open-source model, the creation of digital goods does not generate any direct monetary benefits for thedeveloper/producer. In this sense, it is a model that on its own cannot entirely replace the IPR system in a marketeconomy. However, it can be easily combined with a patronage or ‘spoils’ system to make it perfectly sustainable,solving the incentive problem, which is how the open-source model has (in practice) been operating since itbegan.

As previously mentioned, contributions to open-source projects can generate a reputation that can be latermonetised by the opportunity to access better jobs. It could therefore be argued that companies hiring respectedopen-source programmers are subsidising (providing patronage to) open-source development. Many softwarecompanies go one step further by explicitly allowing their employees to spend part of their working time onopen-source projects. Open-source development is also widely subsidised by public money, since a significantproportion of developers work in universities or publicly funded research centres.

What is clear from the history of open-source software development is that it can be a very powerful system ofinnovation and work organisation. Despite being entirely voluntary, decentralised and with no (direct) monetarycompensation, it has overtaken commercial software development for many types of applications – fromoperating systems to web servers. The open-source model is extremely interesting for discussions on the widersocioeconomic implications of digital technologies because it shows their enormous transformative potential, if afavourable institutional framework is in place. With instant and pervasive communication, a digitally skilleduser-base, and the advantages of a decentralised algorithmic coordination, the possibilities for recombinationand innovation can grow in a truly exponential way without generating troubling distributional effects.

Box 3: The open-source alternative

13

desktop, such as the technology used in smartphones,in which vendors other than Microsoft have achievedtheir own lock-in (mostly Google with the Androidoperating system and, to a lesser extent, Apple’s ownsystem).

The very strong concentration effect of demand-sideeconomies of scale tends to create large monopolies,which is a cause for concern. On top of antitrust andcompetition policies, solutions such as the use ofpublicly available open standards and interoperabilityhave been proposed. It should be noted that this topic islinked with the problems of IPR, as previouslydiscussed. Large digital companies often resist or try tocontrol open standards and interoperability, claimingthat it challenges their profitability and their own IPR. Inthis respect, the open source systems discussed in Box 3can provide a viable alternative.

ConclusionsThis chapter discussed four key attributes of the digitaleconomy: flexibility of production; readily availableinformation; zero marginal costs; and strong networkeffects. These attributes can already be observed in thesectors and industries where digital transformation ismore advanced; foremost in the IT sector itself, and inthe broader communication and leisure industries.

As digital technologies become more widespread andproduction and work process become more digitised,the attributes listed will also be observed inmanufacturing, retail and social services, graduallytransforming economic processes in these sectors. Ofcourse, these sectors have their own specific issues andwill probably not be entirely reshaped in the image ofthe IT industry.

Attributes of the digital economy

15

How can the digital revolution transform work andemployment? As in the previous chapter, discussing thissubject inevitably requires some speculation becausethe process is still unfolding. However, potentialdevelopments can be explored on the basis of howthings are already changing in some sectors andindustries where the use of digital technologies is moreadvanced.

On the basis of a review of the literature in this area,2

this chapter discusses three vectors of change. Thesecorrespond to three broad categories of combinedapplications of digital technologies in economicprocesses, with different implications for work andemployment.

Automation of work: the replacement of (human)labour input by (digitally-enabled) machine input forsome types of tasks within production and distributionprocesses. Although machine automation predateseven the Industrial Revolution, the use of digitaltechnologies allows the algorithmic control ofmachinery and, therefore, many more possibilities forautomation. With digitally enabled machines andartificial intelligence, all kinds of tasks can bepotentially automated.

Digitalisation of processes: the use of sensors andrendering devices to translate (parts of) the physicalproduction process into digital information (and viceversa), and thus take advantage of the greatly enhancedpossibilities of processing, storage and communicationof digital information. This is the main way in which theattributes of the digital economy have spread to sectorsand industries beyond ICT, as discussed in the previouschapter.

Coordination by platforms: the use of digital networksto coordinate economic transactions in an algorithmicway.

These three vectors of change rely on digitalinfrastructures, technologies and skills already widelyavailable in the economy. In that sense, they are clearlyattributes of the deployment rather than the installationphase of the digital revolution, according to the schemaof Freeman and Pérez presented in the Introduction.They presuppose a certain degree of maturity and

diffusion of digital technologies, and involve the kind ofprofound transformation of socioeconomic structuresthat characterises the second phase of technologicalrevolutions. Each of these three vectors of change hasthe potential to fundamentally transform work andemployment in a technological and in a social way.However, each of these vectors has particularly strongeffects on one of the domains of work and employmentintroduced in Box 2 (p. 7): tasks and occupations,working conditions, employment conditions, industrialrelations.

Automation has particularly strong implications for theevolution of the types of task input necessary for theproduction process, and therefore the structure ofemployment by occupation and sector, as well as theskill levels required. However, it also has directimplications for working conditions (since theautomation of certain tasks eliminates some types ofwork and creates others) and indirect implications foremployment conditions and industrial relations (forinstance, it can alter the balance of power withinworkplaces).

The effect of digitalisation is most direct and clear onworking conditions, since it involves a change in theenvironment and nature of work processes. But, for thesame reasons, it also involves changes in tasks andoccupations, and has an indirect effect on employmentconditions and industrial relations.

Finally, platforms represent most directly a change inthe social organisation of production, since they arethemselves a new type of economic institution:therefore, their most obvious and direct impact is interms of the conditions and regulation of employment.However, they can also change the division of labour(for instance, they enable a much more detailedbreakdown of tasks) and affect industrial relations.

Before looking more in depth at each of the threevectors of change, it must be acknowledged that thedistinction between them is (to some extent) moreanalytical than actual. Very often, digitalisation,automation and platforms will be implementedsimultaneously, because there are strong synergiesbetween them. For instance, the use of advanced robots

3 Implications for work andemployment: Three vectors ofchange

2 For more details, please see the three separate literature reviews carried out by Eurofound in parallel with this report: Peruffo, 2017; Peruffo andSchmidlechner, 2017; Schmidlechner and Peruffo, 2017.

16

both requires and generates the digitalisation ofproduction – a massive amount of digital data on therobots’ environment.

Platforms also require and generate vast amounts ofdata on the economic processes they coordinate, andthey can facilitate automation by breaking up theseprocesses into ever smaller tasks. However, it is usefulto distinguish between these three vectors of changebecause they are distinct processes and have differentpotential implications. Digitalisation of production cancertainly proceed without any automation if everyprocess is transformed into bits, but all labour input stillrequires human interaction. A good example is theprovision of psychological services in a virtual realityenvironment, with a real psychologist and a realcustomer behind digital avatars. And automation cantake place without platforms.

Automation of workThis report understands automation as the replacementof labour input by machine input for some types of tasksin production and distribution processes. The focus ontasks in this definition emphasises the link betweenautomation and the unfolding of the division of labour.

Automation presupposes a relatively advanced divisionof labour into highly differentiated tasks, since it isthose detailed tasks that can be encoded andimplemented by machines. And by replacing labour bymachine input (in certain tasks), automation directlyalters the division of labour. A significant part of recentresearch about the implications of automation hasfocused on how automation has altered the structure ofemployment – in terms of different categories of taskand worker, and how it may alter the employmentstructure in in the future.

It is also important to emphasise that, following theabove definition, it is tasks that are automated ratherthan occupations or jobs. In human labour, tasks veryrarely appear in isolation, being instead bundled intooccupations or jobs. Consequently, all occupations orjobs involve many different types of task (Fernández-Macías and Bisello, 2016). Until human-level artificialgeneral intelligence (AGI) exists, automation will bealways focused on the replacement of particular tasks(or a set of related tasks): technology will never be ableto replace all the tasks involved in a particularoccupation. Successive rounds of automation mayindeed eliminate the entire bundle of tasks associatedwith a particular occupation, although to date this hasbeen relatively rare. In most cases, automation changesthe task content of occupations and perhaps therelative importance of some occupations with respectto others, but it rarely eliminates occupations entirely.A good example is that of how the occupation of bankteller has changed with the introduction of automatedteller machines, or ATMs (Bessen, 2015).

Defined in general terms, automation is as old as theuse of machinery in production. In the sectors ofagriculture and manufacturing, automation has beenvery significant over the last 200 years, which is whythese two sectors nowadays account for a fraction oftheir historical employment, and yet production hasincreased in both considerably. What is new aboutautomation in the digital age is that the use ofalgorithmic control of machinery and digital sensors,with ever-increasing computing power, expandsenormously the range of tasks that machines can carryout. The tasks framework proposed (Fernández-Macíasand Bisello, 2016) is useful for differentiating those tasksthat can, more or less, be automated using digitaltechnologies. Routine tasks (repetitive andstandardised, generally as a result of a particular workorganisation strategy and a detailed division of labour)are relatively easy to automate. In fact, physical routinetasks had already been automated (to a large extent) inadvanced market economies before the digitalrevolution; today, these are just a marginal category ofaggregate labour input (Fernández-Macías and Bisello,2016, Figure 2). The automation of intellectual routinetasks, which grew with the bureaucratic control of theeconomy in the first half of the 20th century, is a muchmore recent phenomenon that has been directlyenabled by the digital revolution.

Although it still has some way to go, such change seemsinevitable since digital technologies are much moreefficient than human labour at routine intellectualtasks, at a much lower cost. According to some authors,it is the decline of these two categories of labour input(routine physical tasks and routine intellectual tasks)that is associated with job polarisation (Autor, 2010);others have argued that it is neither the main driver nornecessarily linked to a decline of mid-skilled jobs(Fernández-Macías and Hurley, 2016). Other types oftask are still relatively free from automation, althoughdigital technologies have made considerable progress inthis area in recent years.

Physical non-routine tasks that require mostlyhand–eye coordination and manual dexterity, typical ofmany service activities such as cleaning, serving anddriving, seemed nearly impossible to automate.However, recent advances in machine learning, sensorsand big data are making this prospect increasinglyfeasible. Soon, the limits of automation for such tasksare more likely to be determined by social norms, andsuch considerations as regulations, safety concerns andhuman labour costs, than technological feasibility.

Intellectual non-routine tasks involving creativity,problem-solving and pattern recognition are oftenconsidered as the most advanced expression of humanactivity; even these types of task, however, arebecoming increasingly open to automation. Deeplearning techniques, such as artificial neural networks,are allowing computers to perform creative,


17

problem-solving and pattern-recognition tasks thatproduce results often impossible to distinguish fromthose arrived at by humans. Whether such digitalnetworks are ‘creative’ in the same way as humans is acomplicated philosophical discussion, beyond thescope of this report. But it is important to realise thatthe results of such ‘creative’ work could eventually besufficiently similar to replace human labour.

Non-routine physical and intellectual tasks now accountfor a very significant share of the total labour input inadvanced market economies; automation of these taskswould therefore significantly affect the employmentstructure. Such non-routine tasks are more likely to befound at the bottom and the top of the skillsdistribution (respectively, physical non-routine andintellectual non-routine). Hence, their automation mayhave a centripetal rather than polarising effect onoccupational structures, moving employment towardsthe middle of the skills spectrum.

There is, however, one big category of tasks which so farhas not been discussed. Social tasks that inherentlyrequire human interaction – education, health, leisureand social services (routine or non-routine) – areintrinsically more difficult to automate. To the extentthat human interaction essentially defines what a taskis, by definition machines cannot perform it unless theythemselves become indistinguishable from humans,which is still some way in the future, even in the mostradical forecasts.3 Following this line of argument, it isplausible that all employment lost by automation wouldbe displaced into social tasks (routine or non-routine).

The image of a future in which robots carry out allphysical and intellectual work, while humans occupythemselves in entertaining and looking after each othermay not appear so threatening. However, recentadvances in human–robot interaction in the areas ofcaring assistance and companionship suggest that evenif they are far from being fully human, social robots maybe able to fulfil human needs for some basic types ofsocial interaction and companionship (Breazeal, 2017).

It is important to note that although the automation ofsocial tasks seems unlikely for the foreseeable future,digital technologies can still have a significant effect onthe demand for such tasks by increasing verysignificantly labour productivity. An example in the fieldof education is the increasing availability of MassiveOpen Online Courses (MOOCs). These free (or verylow-cost) courses available over the internet use onlinevideos and texts, together with interactive exercises andalgorithmic monitoring of progress, to provide analternative to face-to-face education. In this case, no

tasks are automated since a human being (the professorwho designed the course and whose lessons have beenrecorded) still provides the educational service.However, this model can obviously reduce verysignificantly the demand for human labour ineducation, which highlights a fundamental issue in thevery concept of automation. In the understanding ofautomation as the replacement of human involvementby machine input, what does ‘replacement’ actuallymean? Even the most advanced industrial robotrequires the human intervention to enable it tofunction. Someone must design and maintain the robot.When something unexpected happens that has notbeen encoded in the control algorithms, a humanoperator must take control. In other words, machinescannot entirely replace human labour for theperformance of any task, not at least until artificialintelligence comparable with that of humans isdeveloped. If this is the case, what is the differencebetween a robot and any other tool that increases theproductivity of workers? Is it fundamentally erroneousto use the term ‘automation’? Instead, should we simplytalk about technological changes that increase theproductivity of some workers, and therefore reduce theamount of labour input which is necessary for someother types of tasks?

This section on automation cannot be closed withoutsome words about the future. In the recent literature onthe subject, there have been several attempts atforecasting how many jobs will disappear in the face ofautomation, and how fast. These forecasts havegenerated a lot of attention, as well as anxiety. But issuch anxiety justified? Assuming that the current roundof automation is not fundamentally different fromprevious periods of productivity-enhancingtechnological change, perhaps history can providesome answers.

Previous technological revolutions did reduce thedemand for some types of labour. In some cases, thiscreated considerable difficulties for the displacedworkers. Perhaps the most dramatic example is thepopulations displaced by the agricultural revolution andthe land enclosures that preceded the IndustrialRevolution in the UK (Polanyi, 1957). Over aconsiderable period of time, displaced surplus labourwas absorbed by the expansion of demand in othertypes of jobs and activities as a result of growing incomelevels. Agricultural automation continued todramatically reduce the need for human labour in thesector, with marginally more labour in industry, andmost now in services; 200 years ago, the situation wasthe opposite.

Implications for work and employment: three vectors of change

3 In fact, the moment in which machines become indistinguishable from humans would probably be of such evolutionary significance that its potentialeffects on employment would be irrelevant as the problem for humans would be existential, not economic.

18

But although history suggests that (in the long term) theemployment effects of automation will probably beabsorbed by the economy (although in unpredictableways), it also shows that the large-scale processes ofeconomic restructuring associated with technologicalrevolutions can be (in the short and medium term)socially and politically catastrophic. The terribleconditions of the working class in England during theIndustrial Revolution, or the terrifying politicalconsequences of the crisis that followed the FordistRevolution in the 1930s, testify to that.

Going back to the argument presented in theIntroduction, the potential effects of automation onemployment and society highlight the need to assessand redesign economic institutions to deal with thesocial and political tensions that can be expected as aresult of the digital revolution.

Digitalisation of processesThe definition of digitalisation used in this report refersto the use of sensors and rendering devices to translateparts of the physical production process into digitalinformation (strings of bits), and vice versa. Sensors aremachines that translate analogue into digitalinformation, such as a scanner or digital camera.Rendering devices do the opposite, translating digitalinto analogue information – for example, a printer. Thekey advantage of digitalisation is that the processing,storage and communication of digital information isvastly cheaper and more efficient than the analogueequivalent.4 By digitalising a process, it can beunderstood, controlled and manipulated moreeffectively. To better illustrate this idea, the focus ofdiscussion will turn to three of the key technologiesdriving the digitalisation of economic processes:

£ Internet of Things (IoT)£ 3D printing£ virtual and augmented reality

The processes creating the Internet of Things attachsensors to outputs, inputs, components, materials ortools used in production. These feed into a real-timedigital model of the entire process. In turn, this can beanalysed, monitored and controlled using algorithms,to an extent that would be impossible in the physicalworld.

3D printers literally create physical objects from three-dimensional digital models, generally by laying downsuccessive layers of material. Although they arecurrently mostly used for prototyping and specialisedapplications, 3D printers have the potential of

transforming all industrial production from beginning toend into a digital process. In such a model, most of thevalue would reside in the ideas (digital models); thephysical objects would have only very limited value.

Finally, virtual reality can move entire economicprocesses to the digital realm – for example, theprovision of some types of face-to-face service. Andaugmented reality can blend the digital and physicalworlds by superimposing digital information overhuman perception of physical reality.

By digitalising economic processes, these threetechnologies expand the four attributes of the digitaleconomy previously discussed in Chapter 2 into(potentially) all sectors of the economy: productiveflexibility, fast and pervasive information availability,zero marginal costs, and strong network effects. Butwhat are the potential implications for work andemployment – specifically, tasks and occupations,working conditions, and employment conditions andindustrial relations?

Tasks and occupationsThe increased efficiency of digitalised processmanagement and control is likely to be associated withlabour-saving productivity growth, especially in areassuch as logistics, quality control and administration.Digitalisation facilitates the algorithmic automation ofmany of those tasks although – as previously discussed– the distinction between automation and labour-savingproductivity growth is somewhat artificial.

Another crucial effect of digitalisation, in terms of thedivision of labour, is the increasing irrelevance of thephysical location of labour input in the productionprocess; this could contribute to a further and perhapsfinal round of globalisation. Richard Baldwin (2016),argued that telepresence (virtual reality technology)and virtual and augmented reality can facilitate thedelivery of face-to-face services from any distance,breaking the final boundary that has protected manyservice activities (and jobs) from globalisation.

Working conditionsThe digitalisation of economic processes raises someserious concern for the autonomy and privacy ofworkers. If every single object in the workplace is asensor that feeds real-time information to a centralisedmanagement algorithm, workers may legitimately feelthat their autonomy and privacy are beingcompromised. The other side of the equation is thatimproved intelligence and information on workprocesses can reduce accidents, and dispense with theneed for certain isolated, repetitive tasks. For instance,


4 Thanks to Moore’s Law (processing power for computers will double every two years), it is becoming ever cheaper and more efficient.

19

quality control largely means repeatedly checking thatan item or process meets certain standards, somethinga sensor can easily do in real time. Digitalisation couldalso diffuse the methodology and skills of ICT into othersectors of the economy, such as manufacturing, retailand other services.

Employment conditions and industrialrelations Digitalisation makes possible more complexorganisational forms of production; it may facilitate thebreakdown and subcontracting of an increasing numberof tasks, even in traditional production processes.Subcontracting and outsourcing, even crowdsourcing,can result in less favourable conditions of employmentfor workers in terms of stability, income and workinghours. By blurring company boundaries and disruptingunion solidarity, such forms of work can also makecollective representation more difficult. On the otherhand, the digitalisation of all types of economicprocesses opens them up to alternative methods ofcollaborative decentralised production, such as thosediscussed in Box 3 (p. 12). A good example is the‘makers’ movement of some 3D printing enthusiastsand ‘artisan-hackers’, who use open source licences fordigital designs and hardware, and defend asocioeconomic model of cooperative, non-hierarchicaland sustainable production (Anderson, 2012).

Coordination by platformsPlatforms are digital networks that coordinatetransactions in an algorithmic way. There are twoimportant elements in this definition. First, the networkis a structured digital ‘space’ where goods or servicescan be offered or requested. These online spacessystematically collect, organise and store large amountsof data about the platform users and transactions.Some of these data are fed back to users as records ofsuccessful transactions or evaluations, which serve boththe purpose of facilitating trust between users andincentivising good behaviour.

The second key element of platforms is a set ofalgorithms for matching and coordinating transactionsin an automated way. The algorithms provide agovernance structure to the platforms, incorporatingencoded rules as well as automated monitoring andenforcement mechanisms. Platforms are hybrids ofmarkets and firms: the network and algorithmiccomponents of platforms perform the functions of eachof those basic economic institutions. Whereas the

structured online space (network) provided byplatforms make them similar to markets as spaceswhere supply and demand can meet, the governingalgorithms make them similar to firms as structures ofcommand. The algorithms of platforms are essentiallyautomated forms of management.

What distinguishes platforms from the other twovectors of change previously discussed – automationand digitalisation – is that platforms are at least asmuch a form of institutional innovation as a form ofproductive innovation. There is some debate aboutwhether platforms really enable a more efficientorganisation of production, or just simply facilitate theexploitation of labour and competitors. From a purelytechnical perspective, platforms enable a very efficientand transparent distribution of information across alarge numbers of users, and algorithmic matching andcoordination is much more cost-effective than humancoordination.

It has been shown that platforms enable a moreefficient use of capacity and resources (Cramer andKrueger, 2016), and facilitate transactions of loweconomic value that were not previously viable.However, at least part of the success of some well-known platforms can probably be attributed to theirsuccess in circumventing regulation in the markets inwhich they operate, hence profiting from unfaircompetition. Another reason for their success is theweakened position of workers in such platformscompared with traditional firms. In this sense, the keypolicy question concerning platforms may be howcitizens can benefit from their superior coordinatingefficiency while avoiding their potentially negativesocial outcomes; this question relates to theirinstitutional design and regulation.

Platforms are (to a large extent) a new form of economicactivity that does not fit well in existing regulatoryframeworks. To ensure that these regulatoryframeworks continue to fulfil their social coordinationand protection functions, they may need to be adapted.Alternatively, innovative policy approaches could betried, such as promoting the expansion of differentforms of platform governance that provide moredesirable social outcomes. For example, open sourcealgorithms – with rules and enforcement mechanismsdemocratically agreed by the users in peer-to-peernetworks – can (in principle) be at least as technicallyefficient as proprietary commercial models whilegenerating fairer distributional outcomes, and a moreeven ground for exchange.


20

But what are the potential implications of platforms forwork and employment? The most immediate and directimplications of platforms are in employmentconditions, since they are a new form of economicorganisation that does not fit neatly into the existingcategories of dependent employment and self-employment. Concern has been expressed that thesituations of some platform workers could combine theworst of both worlds: the more limited social andcontractual protection of self-employed workers withthe dependence and lack of autonomy of employees.However, the diverse nature of platforms can beassociated with very different situations in terms ofemployment conditions.

The same ambiguity in the classification of platformworkers as independent contractors suggestsdifficulties for collective representation andparticipation. As independent contractors, platformworkers are not entitled to collective bargaining inrelation to their platforms or clients; and althoughunions do represent self-employed workers in somecountries, they tend to play a marginal role.Furthermore, the very nature of the tasks and workorganisation in platforms makes collective organisationless likely than in traditional companies: the manager isan algorithm, co-workers are independent contractors(potentially geographically dispersed and incompetition with each other) and the work is oftencarried out in isolation or in contact only with the client.


As there are many different types of platforms, it is important to classify them. The definition of platforms used inthis report refers to the coordination of economic transactions (involving the exchange of goods and services); ittherefore excludes online spaces that are sometimes also considered as platforms – most importantly, socialnetworks. Different authors use different criteria for classifying platforms; key criteria are set out below.

Platform ownership: A key distinction is drawn between privately owned platforms (generally for-profitbusinesses such as Uber and Airbnb) and platforms commonly owned by their users, such Blockchain. In mostcases, private platforms generate revenue by charging a fee or percentage of the value of each transaction; insome cases, however (especially if transactions are not commercial), they may charge entry fees or generaterevenue by displaying ads.

Economic nature of transactions: Both commercial and non-commercial transactions (if the services contractedare paid for or not) may be facilitated through platforms. The category of platforms for non-commercialtransactions corresponds most closely with the original idea of the ‘sharing economy’, where goods and servicesare shared or exchanged rather than purchased. If goods are simply shared without any expectation ofreciprocation (beyond recognition) it is a pure gift economy, for example the homestay app Couchsurfing. But ifgoods are exchanged despite no financial involvement it is a barter economy, such as the service exchangeplatform Simbi. It should be noted that even if the transactions are non-commercial, the platforms themselvescan be for-profit businesses, generally generating revenue by subscription fees or advertisement, as withCouchsurfing.

Content of transactions: The main difference here is between platforms for the exchange of goods (such as eBayand Amazon Marketplace) and platforms for the exchange of services, (including Uber, Airbnb and TaskRabbit).This report focuses on this second type, service platforms. Service platforms can be further subdivided.

£ Online vs local: commercial service online platforms (such as Amazon Mechanical Turk, or MTurk)correspond to the widely used concept of crowd work; in contrast, commercial platforms providing personallocal services (such as TaskRabbit) are often referred to as the ‘gig economy’.

£ Types of tasks involved: namely physical (Taskrabbit), intellectual (MTurk) and social (Bubble). A distinctioncan be drawn between platforms aimed at routine repetitive and standardised tasks (such as MTurk) andnon-routine tasks which are more complex and/or require creative skills; micro tasks (again MTurk) andlarger projects such as the exchange of freelance services (for which Freelancer would be more appropriate)and the associated level of skills required of the worker – low, medium or high.

The heterogeneity of platforms is increasing as they spread across different sectors and activities; this impliesthat more criteria are needed for a comprehensive classification. Eurofound is currently working on such adetailed classification, specifically aimed at platform work – paid work organised through online platforms.

Box 4: Classifying platforms

21

However, there have been some recent examples ofmobilisation of platform workers (Tassinari andMaccarrone, 2017), especially in the category ofcommercial platforms providing personal local services(the gig economy). New forms of online collectiveorganisation are also emerging for crowd workers –for instance, through the use of internet forums andplatforms such as Turker Nation (Martin et al, 2014).Such mobilisation may become more frequent asplatforms grow, perhaps giving rise to new forms ofindustrial relations.

The impact of platforms on the division of labour can besubstantial. The organisational efficiency of platformsallows for the division of labour into very small tasks;this can result in those tasks being tedious andrepetitive (on top of their often being carried out inisolation). These are not ideal psychosocial conditionsfor work and can often be associated with feelings ofalienation. At the same time, some categories ofplatform work can provide autonomy and flexibility,allowing people who may otherwise find it difficult toparticipate in certain types of employment. Thedifferent categories of platform work are veryheterogeneous, and can have very different implicationsin terms of employment and working conditions.


23

Embarking on a new programme of work for socialresearch can be very exciting but inherently risky,especially if the subject is as broad and ambitious as theimplications for work and employment of atechnological revolution, which is still unfolding.

The risks comprise unwarranted optimism, unduepessimism and mistargeted insights. It is easy to beovertaken by visions of an ideal world based on thetransformative potential of new technologies, whichmay never be realised. It is equally possible to fall victimto an overly pessimistic viewpoint, assuming the worstpossible uses for new technologies, or even attributinghumanlike motivations and effects to them, giving riseto fears of robots stealing jobs. It is also easy toovergeneralise, to be inconsistent, and to focus ontrivialities while missing important underlying trends;because the subject itself is fascinating (newtechnologies giving us glimpses of a possible future),even this kind of research may generate some interest.Such research is, however, unlikely to be of much use inhelping the democratic process create better policies –ultimately the aim of this report.

This report seeks to minimise the above-cited risks andestablish a solid base for Eurofound’s research on theimplications of the digital age. To that end, this reporthas looked to provide clear demarcations for the keyconcepts in this area, and make explicit theassumptions that underlie this research from itsinception. Of course, the different research strands thatwill be carried out within this body of work over thecoming years will necessarily require somereadjustment of these concepts and assumptions;indeed continuously updating our knowledge about theworld on the basis of new evidence is the purpose ofresearch. But interpreting this evidence requires clearconcepts and analytical tools, and this report has triedto provide this, for a subject that is continuouslyevolving.

4 Commentary

25

Anderson, C. (2012), Makers: The New IndustrialRevolution, Random House, New York.

Arthur, W. B. (2009), The nature of technology: What it isand how it evolves, Simon & Schuster, New York.

Autor, D. (2010), The polarization of job opportunities inthe US labor market – Implications for employment andearnings, Center for American Progress and theHamilton Project, Washington DC.

Baldwin, R. (2016), The Great Convergence, HarvardUniversity Press, Cambridge, Massachusetts.

Bessen, J. (2015), Learning by doing: The real connectionbetween innovation, wages, and wealth, Yale UniversityPress, Boston.

Boyer, R. (1990), The regulation school: A criticalintroduction, Columbia University Press, New York.

Breazeal, C. (2017), ‘Social robots: From research tocommercialization’, in Proceedings of the 2017 ACM/IEEEInternational Conference on Human-Robot Interaction,6–9 March, Vienna.

Brynjolfsson, E. and McAfee, A. (2014), The secondmachine age: Work, progress, and prosperity in a time ofbrilliant technologies, W.W. Norton & Company, NewYork.

Coase, R. H. (1937), ‘The nature of the firm’, Economica,Vol. 4, No. 16, pp. 386–405.

Cramer, J. and Krueger, A. B. (2016), ‘Disruptive changein the taxi business: The case of Uber’, AmericanEconomic Review, Vol. 106, No. 5, pp. 177–182.

Depoorter, B. (2004), ‘The several lives of Mickey Mouse:The expanding boundaries of intellectual property law’,Virginia Journal of Law & Technology, Vol. 9, No. 1.

Electronic Frontier Foundation (2014), Adobe spywarereveals (again) the price of DRM: Your privacy andsecurity, 7 October.

Fernández-Macías, E. (2002), ‘Una aproximaciónsociológica al fenómeno del software libre’, RevistaInternacional de Sociología, No. 31, pp. 167–184.

Fernández-Macías, E. and Bisello, M. (2016),‘A framework for measuring tasks across occupations’,Vox.eu.

Fernández-Macías, E. and Hurley, J. (2016), ‘Routine-biased technical change and job polarization in Europe’,Socio-Economic Review, Vol. 15, No. 3, pp. 563–585.

Freeman, C. and Louçã, F. (2001), As time goes by: Fromthe industrial revolutions to the information revolution,Oxford University Press, UK.

Gordon, R. J. (2016), The rise and fall of Americangrowth: The US standard of living since the civil war,Princeton University Press, New Jersey.

Harari, Y. N. (2016), Homo deus: A brief history oftomorrow, Random House.

Kuhn, T. S. (1962), The structure of scientific revolutions,University of Chicago Press.

Kurzweil, R. (2005), The singularity is near: When humanstranscend biology, Penguin Books, London.

Martin, D., Hanrahan, B. V., O’Neill, J. and Gupta, N.(2014), ‘Being a turker’, in Proceedings of the 17th ACMconference on Computer supported cooperative work andsocial computing (pp. 224–235), 15–19 February,Baltimore, US.

Mazzucato, M. (2015), The entrepreneurial state:Debunking public vs. private sector myths, Anthem Press,London.

McAfee, A. and Brynjolfsson, E. (2017), Machine platformcrowd, W.W. Norton & Company, New York.

Pérez, C. (2003), Technological revolutions and financialcapital: The dynamics of bubbles and golden ages,Edward Elgar Publishing, Cheltenham, UK.

Peruffo, E. (2017), A literature review on the implicationsof digitisation for work and employment, Eurofound,Dublin.

Peruffo, E. and Schmidlechner, L. (2017), A literaturereview on the implications of automation for work andemployment, Eurofound, Dublin.

Polanyi, K. (1957), The great transformation, BeaconPress, Boston.

Quah, D. (2002), Digital goods and the new economy, CEPdiscussion paper (563), London School of Economicsand Political Science, London.

ReferencesAll Eurofound publications are available at www.eurofound.europa.eu

26

Rifkin, J. (2011), The third industrial revolution: Howlateral power is transforming energy, the economy, andthe world, Palgrave Macmillan, New York.

Schmidlechner, L. and Peruffo, E. (2017), A literaturereview on the implications of platforms for work andemployment, Eurofound , Dublin.

Schwab, K. (2017), The fourth industrial revolution,Crown Publishing, New York.

Tassinari, A. and Maccarrone, V. (2017), ‘Organising theunorganisable? Gig economy couriers’ mobilisations inItaly and the UK’, paper presented at the PlatformLabour in the Digital Economy workshop, 17 May,University of Milan.

The Guardian (2005), Patent absurdity, 20 June.


EF/18/002

The onset of the digital revolution has resulted intechnological advances that are constantlyevolving. A key element of concern to policymakersis the impact that these changes will have on theworld of work and employment. This reportreviews the history of the digital revolution to date,placing it in the context of other periods of markedtechnological advances and examining howtechnological change interacts with changes ininstitutions. Digital technologies have considerabledisruptive potential, including making productionmuch more flexible and information more readilyavailable. While the information technology sectorhas been most affected to date, other sectors arerapidly changing with the diffusion of newtechnology. The report also examines three keyvectors of change: automation of work, theincorporation of digital technology into processes,and the coordination of economic transactionsthrough the digital networks knowns as‘platforms’.

The European Foundation for the Improvement ofLiving and Working Conditions (Eurofound) is atripartite European Union Agency, whose role isto provide knowledge in the area of social,employment and work-related policies.Eurofound was established in 1975 by CouncilRegulation (EEC) No. 1365/75, to contribute to theplanning and design of better living and workingconditions in Europe.

TJ-04-18-316-EN

-N

ISBN: 978-92-897-1652-9doi:10.2806/13911

Automation, digitalisation and platforms: Implications for ...€¦ · processes, such as Kuhn’s The structure of scientific revolutions (Kuhn, 1962) or the punctuated equilibrium

Documents