Systemic Thinking for Policy Making

10. Social Protection in the Face of Digitalisation and Labour Market Transformations

Herwig Immervol

Duncan MacDonald

Elena Rovenskaya

Leena Ilmola

Abstract

It is shown how systems analysis can facilitate policy responses to complex labour-market changes, including from digitalisation and technological progress. A simple systems map covering linkages and interactions between technology adoption, alternative work arrangements, wages, and social protection illustrates how qualitative system mapping can provide a holistic perspective and enable a systematic exploration of connections between elements of a complex system. The map illustrates how systems analysis facilitates consideration of the behaviour of a complex system by decomposing it into sub-processes that can be verbally described in a straightforward and relatively simple way. From this, policy makers can obtain a broad view of the drivers of policy outcomes. Agent-based modelling (ABM) can complement a systems mapping approach. These models, including one maintained by IIASA, provide a “sandbox” for exploring the consequences of linkages and feedback effects for individual agents and for the system as a whole.

Introduction

Automation and digitalisation are driving labour market transformations across OECD countries. These transformations bring about opportunities for increased productivity, new products and novel ways of organising production (Graetz and Michaels, 2015[1]; Acemoglu and Restrepo, 2017[2]). However, there are concerns that this transformation is occurring too fast for societies to be able to adapt (Brynjolfsson and McAfee, 2014[3]; Ford, 2015[4]). In particular, a faster pace of technology adoption creates risks that job losses will outpace the creation of new employment opportunities.

Main Messages

Systems analysis facilitates policy responses to complex labour-market changes, including from digitalisation and technological progress.

A rise in alternative work arrangements has accompanied the digitalisation of work. Responding to the resulting labour-market challenges requires considering both the direct and indirect effects of digital transformations on labour markets, including linkages between non-standard work, social protection systems, and technological progress.

For example, digitalisation promotes the further adoption of technology. However, it also facilitates alternative work arrangements with lower labour costs, which, in turn, may discourage further technology adoption in some sectors.

Likewise, less stable, non-standard forms of work create new demands on social protection, which existing policy configurations may not be ready to meet. A greater availability of non-standard employment contracts can also create strong incentives for workers or employers to reduce costs by opting out of social support provisions. Opt-out opportunities, in turn, can undermine the foundations of risk sharing and lead to a declining reach of social protection.

Qualitative system mapping can provide the required holistic perspective. It enables a systematic exploration of connections between elements of a complex system to help understand the policy challenges. From this, policy makers can obtain a broad view of the drivers of policy outcomes.

Agent-based modelling (ABM) can complement a systems mapping approach. These models, including one maintained by IIASA, provide a “sandbox” for exploring the consequences of linkages and feedback effects for individual agents and for the system as a whole.

Fears regarding the social and economic consequences of innovation and technological change are not new. Since the Industrial Revolution, there have been concerns about technology-induced job losses. In the 1930s, John Maynard Keynes (1931[5]) warned of technological unemployment, and similar concerns have remained present ever since, although some have suggested that the present bout of technological disruption may be different (Brynjolfsson and McAfee, 2011[6]; Mokyr, Vickers and Ziebarth, 2015[7]). Technology and digitalisation have lowered transaction costs, allowing firms to outsource or automate not just jobs, but individual tasks (Nedelkoska and Quintini, 2018[8]). As a result, work via online platforms has increased rapidly in recent years, although it still accounts only for a small share of workers in OECD countries (Katz and Krueger, 2016[9]).

While past innovations certainly destroyed some jobs, in the long-term they have created more than they destroyed (Autor, 2015[10]). However, newly created jobs are by their nature different from those destroyed, and they may be of lower quality. For example, non-standard work and alternative work arrangements, such as temporary employment, own-account work, and “gigs”, are more likely to be low-quality jobs, and they have been on the rise in OECD countries (OECD, 2018[11]).1

The rise in alternative work arrangements has implications for the development of countries’ social protection systems. On the one hand, more-dynamic labour markets with less-stable employment strengthen the case for social protection. But on the other hand, existing social protection provisions, which were typically designed around traditional full-time employee-employer work relationships, may be less effective or accessible for non-standard workers, e.g. if entitlements are conditional on regular employment over prolonged periods of time. In addition, alternative working arrangements create strong financial incentives for workers or employers to bypass social risk-sharing mechanisms and their associated short-term costs, such as social insurance contributions. These opt-out opportunities can undermine the foundations of risk sharing, and ultimately lead to a cycle of declining reach of social protection and rising costs for those requiring insurance (Rothschild and Stiglitz, 1976[12]; Akerlof, 1970[13])

A narrowing reach of social protection systems raises both equity and efficiency concerns, especially during periods of labour market change and elevated uncertainty. Responding to this emerging challenge is made more difficult by its novelty - a new phenomenon about which policy makers have little intuition. For example, they may examine non-standard work, social protection systems, or technological progress in isolation, but not the systematic linkages between them. More generally, there are concerns that policymakers have difficulty applying a systems thinking approach to devising decisions by not always considering relevant interactions in full (Levy, Lubell and McRoberts, 2018[14])

A number of tools are available to aid policymakers in examining complex systems. Qualitative systems maps can help give an overview of a system as a collection of interacting components and illuminate key feedback loops, while agent-based models can simulate quantitative scenarios emerging from these dynamic interrelationships. The remainder of this chapter presents a proof-of-concept of the systems analysis approach applied to the social protection policy challenges. It discusses qualitative systems mapping as a concept and provides an illustrative example relating to technological progress, alternative work arrangements, and social protection systems. The chapter also briefly outlines linkages to agent-based models.

Systems Mapping

A qualitative systems map helps its users to understand the nature of the system’s boundaries, and understand system’s elements and relations between them. The process consists of three steps: identification of the system’s key elements; identification of key interrelations between them; and identification of key feedback loops that define the behaviour of the system and identification of actions that can lead to desired outcomes via these feedback loops. This analysis can be performed based on input from experts or evidence from literature.

Ultimately, a systems map can provide insight into indirect effects between system’s elements relevant for a given policy question and helps anticipate effects of certain policy interventions. The insights gleaned from the maps aid in identifying policy issues or interventions that policymakers can miss when focusing attention on individual components.

The system’s behaviour requires understanding of its feedback loops. A feedback loop is a sequence of interactions within the dynamics of a system that begins and ends with one component. These loops can be either reinforcing or balancing. Reinforcing loops compound the change of previous iterations of the loop, either positively or negatively, while balancing loops resist forces that pull components away from their initial state.

Building and analysing a qualitative systems map with the involvement of decision makers can lead to unexpected, counter-intuitive results. This makes them useful for addressing so-called “wicked problems” (Churchman, 1967[15]; Rittel and Webber, 1973[16]). These problems typically arise when addressing novel and unique problems, or when the system is so complex that it is impossible to oversee all critical factors and their interrelations.

The development of a systems map can be a first step for further exploration and, in particular, quantitative modelling. It can also be a tool for consensus building amongst stakeholders to explore possible further actions.

Technology and alternative working arrangements: An illustrative systems map

This chapter presents a simple systems map, covering interactions between technological adoption, alternative work arrangements, wages, and social protection. The impact of technological adoption is a suitable subject for a systems map, as it can have diverse and complexly interacting disruptions in labour markets. Further, the pace of technological adoption is likely to continue to increase in the coming years.

This example is not intended to represent a full-scope systems map. The components and linkages are based on research outlined in the most recent OECD Employment Outlook (OECD, 2019[17]). The aim is thus illustrative: to provide a graphical mapping of the drivers featured in the Outlook to highlight some benefits of qualitative systems mapping. The next section briefly summarises the drivers, and the following section presents the systems map.

Figure 10.1. Robots hold the potential to disrupt labour markets

Note:* = forecast

Source: International Federation of Robotics (IFR), https://ifr.org/

Emerging linkages between technology and work

With technology adoption continuing at a steady pace, many studies, with varying degrees of urgency, have predicted that these technological disruptions will lead to job displacement and technological unemployment (Nedelkoska and Quintini, 2018[8]; Frey and Osborne, 2017[18]; Brynjolfsson and McAfee, 2011[6]). At the same time, technology can increase workers’ productivity and increase their wages (Autor and Salomons, 2018[19]; Acemoglu and Restrepo, 2018[20]; Acemoglu and Restrepo, 2017[2]; Bessen, 2017[21]). These two trends need not be mutually exclusive. With the rise of digital platforms, economies are experiencing the automation of certain tasks and the reorganisation of others. This has given rise to alternative work arrangements, which notably includes the “gig economy” (European Commission, 2017[22]; Katz and Krueger, 2016[9]; Huws, Spencer and Syrdal, 2017[23]). Under these various factors, the overall trend of employment is highly uncertain. Some workers will lose their jobs, while new types of work will be created. Notably, so far, the net effect for most economies has been positive (OECD, 2018[24]).

Technology fosters the creation of new jobs, and their destruction. It can reduce relative investment prices and spur the accumulation of capital, which in turn replaces labour and leads to a decline in the labour share (OECD, 2018[24]; Schwellnus, Pak and Pionnier, forthcoming[25]). Recent technology has encouraged market concentration and winner-takes-most dynamics, leading to a few large firms in some sectors (Autor et al., 2017[26]). Beyond reducing demand for labour due to their capital intensity, these concentrated industries can further supress wage growth owing to a lack of competition for workers (Azar, Marinescu and Steinbaum, 2017[27]; Benmelech, Bergman and Kim, 2018[28]).

At the same time, many people are turning to alternative work arrangements. For example, online platforms help to reduce job search frictions, allowing potential workers to more easily find jobs. With these platforms, firms can easily find workers with unique skill sets, and workers can gain increased flexibility and access a wider range of opportunities, thereby enlarging the labour market (European Commission, 2017[22]). By reducing job search frictions, these new digital platforms can reduce time in unemployment (Manyika et al., 2015[29]).

These alternative work arrangements are not without downsides. While workers gain increased flexibility, this comes with increased risk of low and uncertain earnings. As technology removes barriers to finding a job, more workers enter the labour market, driving wages down. The digital nature of these platforms often means that employers can find workers worldwide, and the variability in labour standards and living costs across countries can often mean a “race-to-the-bottom” for all workers. Those countries with higher rates of non-standard work also have lower wages, less employment protection, less access to social protection, and low bargaining power (OECD, 2014[30]).

Downward pressure on wages can have a balancing effect on technology adoption. The apparently inexorable advance of technology is not certain if workers can compete with machines on cost. Lower wages can slow the adoption of automation and the related decline in the labour share. Consequently, countries with relatively low labour costs have not seen the same hollowing out of routine jobs as countries with higher wages (OECD, 2017[31]).

Beyond direct wages, social policy and tax-benefit systems can provide a safety net for workers with low earnings, or who suffer unfortunate circumstances. However, in many countries, social protection is either optional or unavailable for those in alternative work arrangements. When given such a choice, these workers often undervalue the safety net provisions and choose either the minimum amount or not to participate at all in the schemes. For example, in Latvia and Spain, two countries where the self-employed can choose their level of commitment to the unemployment insurance program, nine out of ten self‑employed workers choose the minimum contribution (Arriba and Moreno-Fuentes, 2017[32]; Rajevska, 2017[33]).

Thus, growing forms of non-standard work places pressure on the financing of social protection, as it largely relies on contributions or taxes levied on incomes from work. Without public subsidies, this incentivises further declines in social protection membership and, ultimately, a cycle of escalating costs and declining coverage (Rothschild and Stiglitz, 1976[12]; Akerlof, 1970[13]). In turn, unequal financing burdens or social protection entitlements can promote certain forms employment while discouraging others (OECD, 2019[34])

Technology and alternative working arrangements: An illustrative systems map

The trends outlined above provide only a partial picture of the interactions and effects brought on by technological progress. However, here we focus on the labour market, and more specifically, on alternative work arrangements, and present a simple systems map. Even from this simple system, there emerge some interesting feedback loops, illustrated below.

Figure 10.2. Example systems map of labour market interactions with technology

Note: Blue arrows indicate positively reinforcing links, while red arrows indicate negatively reinforcing effects. The direction of the arrow indicates the direction of the effect.

Source: OECD and IIASA analysis

One example of a simple positively reinforcing loop is the link between technology adoption, labour productivity, and wages. As indicated in the systems map, increased adoption of technology can increase labour productivity, which then feeds into higher wages for workers. These higher wages then provide incentives to replace labour with capital and so adopt more technology. However, a caveat is needed. Technology adoption increases productivity only for those workers with compatible skills, while often displacing those workers with substitutable (that is, automatable) skills. This displacement can be observed in the balancing loop that connects technology adoption, labour’s share of production, labour demand, and wages. When technology adoption decreases the amount of labour needed, the total sum of wages is decreasing which in turn decreases the motivation for investment in automation.

Another balancing loop links technology adoption, alternative work arrangements, and wages. Here, new technologies encourage alternative work arrangements, which can lead to lower wages that, in turn, disincentivises further adoption of technology.

A final example of a loop in this systems map relates to the interaction between alternative work arrangements, social protection systems, and taxes. As alternative work becomes more common, associated voluntary opt-in provisions can erode membership in social protection and can lead to increases in contribution rates to cover funding shortfalls. All of this increases the tax wedge differential between standard and non-standard workers, encouraging more workers to take up alternative work arrangements. Further iterations of this loop have negative impacts on overall wages and household income, regardless of the type of work arrangement, either standard or non-standard.

Figure 10.3. Labour market feedback loops

Note: Blue arrows indicate positively reinforcing links, while red arrows indicate negatively reinforcing effects. The direction of the arrow indicates the direction of the affect.

Source: OECD and IIASA analysis

The above examples reveal one of the benefits of systems analysis: it makes us think about the behaviour of a complex system by decomposing it into sub-processes, which can be verbally described in a rather straightforward and relatively simple way. For example, What is the net impact of technology adoption on labour? To assess that, we analyse all feedback loops and their relative strengths in order to identify the dominating loop and analyse the dynamics that define the behaviour of this system.

Systems maps as an output, or an input

As shown above, a systems mapping exercise can provide clarity on key elements and interactions in a system. The identification of feedback loops is essential to understanding a system, and these loops are not always evident when examining components in isolation. The high-level perspective of the system allows policymakers to gain a conceptual understanding of key interdependencies. In the context of labour markets, this understanding then assists in the design of social protection systems.

Social protection systems provide support during episodes of low earnings capacity and economic difficulties. These systems seek to prevent the deterioration of human capital, which can lead to long-term disadvantages and exclusion. Some components of social protection systems can be seen as “triggers” that policymakers can use to influence a system. For example, income protection measures redistribute resources to groups with elevated needs and provide support during periods of low earnings, unemployment, or other types of non-employment. Additionally, promotion measures can strengthen or re-establish self-sufficiency through incentives, and by tackling individual employment and social-inclusion barriers.

Importantly, national context is a key determinant of the operation of social protection systems. Both the strength of the linkages between components within a system and the configuration of the social protection infrastructure are country specific, and these specifics will dictate the behaviour of a system

These national circumstances can be notable. For instance, some countries have voluntary insurance schemes for some types of employment. Unless these schemes achieve a high coverage rate, they risk falling into a negatively reinforcing feedback loop of rising insurance premiums and falling coverage. Unfortunately, voluntary schemes suffer from adverse selection, with the most in need of insurance the most likely to enrol. For example, opt-ins for the Canadian Special Benefits for Self-employed Workers, a maternity and parental benefit scheme, were found to be mostly women of child-bearing age with significantly lower income than those who did not opt in (OECD, 2018[11]; Employment and Social Development Canada, 2016[35]). Likewise, an increase in voluntary unemployment premiums in Sweden in 2007/08 led to one in eight participants opting-out of the program (OECD, 2018[11]). Those who left the fund were those least likely to benefit, either older workers with little unemployment risk, or younger workers with low unemployment durations.

Gaining a meaningful understanding of a system is a useful result itself. As shown with the systems map presented above, even a simple systems map can reveal useful insights. Further specification of the system in a national context, and with particular reference to social protection systems, can yield even further insights.

Additional detail and specification comes with both costs and benefits. Systems are complex by nature, and a systems map can provide clarity. Adding additional detail jeopardises that clarity. In this sense, the development of a systems map can be the key output of a research project.

At the same time, a systems map can provide a platform for further analytical development. With a clear view on the components and linkages of a system, it is possible to develop a “policy playground” by specifying the magnitude of the interactions between components. Researchers can use the map to formulate working simulations, such as agent-based models, which allows policymakers to pose “what if” type questions.

Agent-based models

Agent-based models can quantify relationships outlined in a systems map

In examining complex systems, a natural extension to qualitative systems mapping is to use them to inform a simulation model. One such modern modelling paradigm, agent-based modelling, consists of the simulation of a number of heterogeneous agents, according to empirically based decisions rules. These models have been garnering increased attention in the aftermath of the 2008 financial crisis and Great Recession, due to their ability to provide alternative perspectives from traditional macroeconomic modelling techniques (Blanchard, 2018[36]; Stiglitz et al., 2017[37]).

As these models are capable of incorporating various behaviour rules of agents without relying on oversimplified assumptions, they are particularly suited to quantitatively model the feedback loops discovered during a systems mapping exercise. These models add to the researcher’s toolbox and they allow relationships outlined in a systems map to be calibrated to observed data in an economy. This can be important, as the strength of the linkages between concepts can heavily influence the dynamics of a feedback loop. A weak link at one point of the loop can short-circuit a loop, just as strong connections elsewhere can amplify relationships. Examining the quantitative aspects of these connections can lead to an increased understanding of the feedback loops and their most crucial parameters. However, these models come at the cost of increased complexity and computational resource requirements.

A number of government institutions and international organisations have developed or are developing agent-based models. These include the EURACE model developed with funding from the European Union (Dawid et al., 2011[38]), and a model of the Austrian economy developed by researchers at IIASA (Poledna, Miess and Hommes, 2019[39]). which have been used to examine financial fragility (Cincotti, Raberto and Teglio, 2010[40]) and worker skill upgrading (Dawid et al., 2009[41]).

Social protection in an agent-based model

Crucially, agent-based models allow policy makers to introduce or alter policies within the model, and then evaluate the distributional effects of the changes. Governments’ tax-benefit systems rely on the redistribution of cash and of risks, and so the additional distributional insights gleaned from agent-based models can be illuminating.

For instance, the feedback loop explained previously, between alternative forms of work, membership in social protection schemes, and labour taxation, could be explored in detail. An agent-based model could highlight those workers who are likely to find themselves in non-standard work (through their own choice or the choice of their employer) and policymakers could explore methods of ensuring adequate social protection for these workers.

Policymakers can vary incentives for various forms of work by changing the contribution rates that finance social protection. Although financing methods vary across countries, governments largely fund protection systems with contributions or taxes levied on work income. Using agent-based models, policymakers can explore alternative funding structures that promote certain types employment while making others less attractive. Alternatively, they could examine complementary policies that encourage firms to hire more workers using standard contracts.

With a set of realistic decision rules that mimic reality and a high-level view of the system informed by a systems map, agent-based models can be an aid to policymakers in developing optimal policies. With agent-based models, policymakers can amplify or dampen feedback loops by introducing innovative and targeted policy solutions. An added benefit of these models is that by modelling the entire economy, it is easy to observe any unintended consequences of any policy reform within the model. Further refinement can reduce these externalities, and ensure that policies have only the intended consequences.

Conclusion

Technological advances will continue to transform the world of work. Social protection systems need to be prepared to support those workers with outdated skills who are ill equipped to compete in tomorrow’s job market. With an increasing variety of work arrangements, policymakers need to ensure social protection systems provide adequate risk pooling to help smooth negative outcomes for all workers. A first step in achieving this is to understand all of the ways that technology influences labour markets. Qualitative systems mapping methods provide a means of achieving this and they are especially useful when a policy problem is relatively new, when data availability is limited, or when potential interactions between different elements of the system are powerful and complex. The resulting map can itself form an insightful research output, or it can be the basis of more extensive quantitative research, such as the development of an agent-based model.

References

[20] Acemoglu, D. and P. Restrepo (2018), “Artificial Intelligence, Automation and Work”, No. 24196, National Bureau of Economic Research, Cambridge, MA, http://www.nber.org/papers/w24196 (accessed on 10 August 2018).

[2] Acemoglu, D. and P. Restrepo (2017), Robots and Jobs: Evidence from Us Labor Markets, https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2941263 (accessed on 19 November 2018).

[13] Akerlof, G. (1970), “The Market for ’’Lemons’’: Quality Uncertainty and the Market Mechanism”, The Quarterly Journal of Economics, Vol. 84/3, p. 488, http://dx.doi.org/10.2307/1879431.

[32] Arriba, R. and Moreno-Fuentes (2017), “ESPN Thematic Report on Access to social protection of people working as self-employed or on non-standard contracts Spain”, in ESPN (ed.), Access to social protection of people working as self-employed or on non-standard contracts, European Commission.

[10] Autor, D. (2015), “Why Are There Still So Many Jobs? The History and Future of Workplace Automation”, Journal of Economic Perspectives, Vol. 29/3, http://dx.doi.org/10.1257/jep.29.3.3.

[26] Autor, D. et al. (2017), “The Fall of the Labor Share and the Rise of Superstar Firms”, http://www.iza.org (accessed on 27 June 2017).

[19] Autor, D. and A. Salomons (2018), Is automation labor-displacing? Productivity growth, employment, and the labor share, Brookings Institution, Washington, DC, https://www.brookings.edu/wp-content/uploads/2018/03/1_autorsalomons.pdf (accessed on 13 August 2018).

[27] Azar, J., I. Marinescu and M. Steinbaum (2017), Labor Market Concentration, National Bureau of Economic Research, Cambridge, MA, http://dx.doi.org/10.3386/w24147.

[28] Benmelech, E., N. Bergman and H. Kim (2018), Strong Employers and Weak Employees: How Does Employer Concentration Affect Wages?, National Bureau of Economic Research, Cambridge, MA, http://dx.doi.org/10.3386/w24307.

[21] Bessen, J. (2017), “Automation and jobs: When technology boosts employment”, Law & Economics Paper, http://www.bu.edu/law/faculty-scholarship/working-paper-series/ (accessed on 6 September 2018).

[36] Blanchard, O. (2018), “On the future of macroeconomic models”, Oxford Review of Economic Policy, http://dx.doi.org/10.1093/oxrep/grx045.

[3] Brynjolfsson, E. and A. McAfee (2014), The second machine age : work, progress, and prosperity in a time of brilliant technologies, WW Norton & Company, http://books.wwnorton.com/books/the-second-machine-age/ (accessed on 19 November 2018).

[6] Brynjolfsson, E. and A. McAfee (2011), Race against the machine : how the digital revolution is accelerating innovation, driving productivity, and irreversibly transforming employment and the economy, Digital Frontier Press.

[15] Churchman, C. (1967), “Wicked Problems”, Management Science, Vol. 14/4, https://punkrockor.files.wordpress.com/2014/10/wicked-problems-churchman-1967.pdf (accessed on 19 April 2019).

[40] Cincotti, S., M. Raberto and A. Teglio (2010), “Credit Money and Macroeconomic Instability in the Agent-based Model and Simulator Eurace”, Economics: The Open-Access, Open-Assessment E-Journal, Vol. 4/2010-26, p. 1, http://dx.doi.org/10.5018/economics-ejournal.ja.2010-26.

[41] Dawid, H. et al. (2009), “On the Effects of Skill Upgrading in the Presence of Spatial Labor Market Frictions: An Agent-Based Analysis of Spatial Policy Design”, Journal of Artificial Societies and Social Simulation, Vol. 12/4, p. 5, http://jasss.soc.surrey.ac.uk/12/4/5/5.pdf (accessed on 10 July 2019).

[38] Dawid, H. et al. (2011), “The Eurace@Unibi Model: An Agent-Based Macroeconomic Model for Economic Policy Analysis”, Working paper, Bielefeld University., https://pdfs.semanticscholar.org/f899/dd5b606a01b32dd7621830cf9f399b745170.pdf (accessed on 25 April 2019).

[35] Employment and Social Development Canada (2016), Evaluation of the Employment Insurance Special Benefits for Self-employed Workers, http://canada.ca/publicentre-ESDC.

[22] European Commission (2017), COLLEEM (Collaborative Economy) research project, https://ec.europa.eu/jrc/en/colleem (accessed on 4 April 2018).

[4] Ford, M. (2015), Rise of the Robots: Technology and the Threat of a Jobless Future, Basic Books, https://scholar.google.fr/scholar?hl=en&as_sdt=0%2C5&q=Ford%2C+Martin+%282015%29+The+Rise+of+the+Robots%2C+Basic+Books%2C+New+York.&btnG= (accessed on 19 November 2018).

[18] Frey, C. and M. Osborne (2017), “The future of employment: How susceptible are jobs to computerisation?”, Technological Forecasting and Social Change, Vol. 114, pp. 254-280, http://dx.doi.org/10.1016/J.TECHFORE.2016.08.019.

[1] Graetz, G. and G. Michaels (2015), Robots at Work, https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2575781 (accessed on 19 November 2018).

[23] Huws, U., N. Spencer and D. Syrdal (2017), Work in the European Gig Economy, University of Hertfordshire, http://www.feps-europe.eu/assets/9d13a6d2-5973-4131-b9c8-3ca5100f92d4/work-in-the-european-gig-full-report-pppdf.pdf.

[9] Katz, L. and A. Krueger (2016), The Rise and Nature of Alternative Work Arrangements in the United States, 1995-2015, National Bureau of Economic Research, Cambridge, MA, http://dx.doi.org/10.3386/w22667.

[5] Keynes, J. (1931), “Economic Possibilities for our Grandchildren”, http://www.econ.yale.edu/smith/econ116a/keynes1.pdf (accessed on 26 July 2018).

[14] Levy, M., M. Lubell and N. McRoberts (2018), “The structure of mental models of sustainable agriculture”, Nature Sustainability, Vol. 1/8, pp. 413-420, http://dx.doi.org/10.1038/s41893-018-0116-y.

[29] Manyika, J. et al. (2015), A Labor Market that Works: Connecting talent with opportunity in the digital age, McKinsey Global Institute, http://www.mckinsey.com/mgi. (accessed on 14 August 2018).

[7] Mokyr, J., C. Vickers and N. Ziebarth (2015), “The History of Technological Anxiety and the Future of Economic Growth: Is This Time Different?”, Journal of Economic Perspectives, Vol. 29/3, pp. 31-50, http://dx.doi.org/10.1257/jep.29.3.31.

[8] Nedelkoska, L. and G. Quintini (2018), “Automation, skills use and training”, OECD Social, Employment and Migration Working Papers, No. 202, OECD Publishing, Paris, http://dx.doi.org/10.1787/2e2f4eea-en.

[34] OECD (2019), OECD Employment Outlook 2019: The Future of Work, OECD Publishing, Paris, https://dx.doi.org/10.1787/9ee00155-en.

[17] OECD (2019), “The future of work: What do we know?”, in OECD Employment Outlook 2019, OECD Publishing, Paris.

[24] OECD (2018), OECD Employment Outlook 2018, OECD Publishing, Paris, http://dx.doi.org/10.1787/empl_outlook-2018-en.

[11] OECD (2018), The Future of Social Protection: What Works for Non-standard Workers?, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264306943-en.

[31] OECD (2017), OECD Employment Outlook 2017, OECD Publishing, Paris, https://dx.doi.org/10.1787/empl_outlook-2017-en.

[30] OECD (2014), OECD Employment Outlook 2014, OECD Publishing, Paris, http://dx.doi.org/10.1787/empl_outlook-2014-en.

[39] Poledna, S., M. Miess and C. Hommes (2019), “Economic Forecasting with an Agent-Based Model”, SSRN Electronic Journal, http://dx.doi.org/10.2139/ssrn.3484768.

[42] Poledna, S. and S. Thurner (2016), “Elimination of systemic risk in financial networks by means of a systemic risk transaction tax”, Quantitative Finance, Vol. 16/10, pp. 1599-1613, http://dx.doi.org/10.1080/14697688.2016.1156146.

[33] Rajevska, F. (2017), “ESPN Thematic Report on Access to social protection of people working as self-employed or on non-standard contracts Latvia”, in ESPN (ed.), Access to social protection of people working as self-employed or on non-standard contracts, European Commission.

[16] Rittel, H. and M. Webber (1973), “Dilemmas in a General Theory of Planning”, Policy Sciences, Vol. 4/2, pp. 155-169, http://www.sympoetic.net/Managing_Complexity/complexity_files/1973%20Rittel%20and%20Webber%20Wicked%20Problems.pdf (accessed on 19 April 2019).

[12] Rothschild, M. and J. Stiglitz (1976), “Equilibrium in Competitive Insurance Markets: An Essay on the Economics of Imperfect Information”, The Quarterly Journal of Economics, http://dx.doi.org/10.2307/1885326.

[25] Schwellnus, C., M. Pak and P. Pionnier (forthcoming), “Labour share developments over the past two decades: The role of public policies”, OECD Economics Department Working Papers, No. forthcoming, OECD, Paris.

[37] Stiglitz, J. et al. (2017), “Where modern macroeconomics went wrong”, NBER Working Paper Series, No. 23795, National Bureau of Economic Research, http://www.nber.org/papers/w23795.

Note

← 1.  Own-account workers are those self-employed who have no employees.