The UK Government’s consultation on employment classification and control: a response

A response to the UK Government’s consultation on employment classification and control.

Wood A.J and Graham M (2018) The government consultation on employment classification and control a response

The UK government is currently undertaking a consultation on employment status. One element under consideration is what ‘control’ means in the contemporary labour market and to what degree it is relevant for determining a worker’s entitlement to labour rights. This distinction is especially important because most firms in the ‘gig economy’ tend not to classify workers as employees: thus leaving workers without any entitlement to labour rights. The current consultation follows a suggestion by the government’s independent review of modern working practices that greater emphasis should be placed on control when determining worker status (Taylor et al. 2017).

Our empirical research on the remote gig economy is useful in highlighting the complexities of work in the contemporary world of work (Graham et. al. 2017a, Graham et. al. 2017b). We interviewed 107 gig economy workers and conducted a survey with a further 679 in order to investigate the nature of platform-based algorithmic control (Wood et al. 2017).

As is common on gig economy platforms (Lee et al., 2015; Rosenblat and Stark, 2016; Shapiro, 2017) our interviews made clear how control was achieved through rating and reputation systems. Workers are rated by their clients following the completion of tasks, with the average rating being attached to the worker’s profile. Platforms then algorithmically rank workers so that work can be funneled to workers with the best reputations. Such reputation systems are usually understood not as mechanisms of control but rather as an efficient manner in which information regarding quality can be communicated. Indeed, while reputation systems make sense as a means for consumers to rate products (e.g. on Netflix or Amazon), and thus signal quality to others, the same reputation systems operate quite differently when applied to people. When humans rate other humans we tend not to provide the same honest feedback that we would when rating a commodity such as a book. Instead social norms and empathy pressure us to be polite give other people high ratings. As born out by analysis of rating distributions, most people follow the saying ‘If you can’t say anything nice, say nothing at all’ (Slee, 2016). Thus reputation systems convey very little information when applied to labour as most people will give another person around a five-star rating regardless of the quality of the service. In this context, low scores are only purposefully given when a client wishes to punish the worker for what they perceive as bad behaviour rather than as genuine comment on the worker’s quality. Workers thus operate in a climate of fear that displeasing a client will result in a punishment via a negative review. The head of policy for one major platform admitted to us that such reputation systems constituted a form of control not dissimilar to the performance management which traditional employees experience.

Our interviews demonstrated that algorithmic control was effective but also afforded workers freedom to work however they wished as long as the end product was accurate or satisfactory to the client. In effect, this form of control afforded significant autonomy and discretion. Using our survey, we found that 72% of respondents felt able to choose and change the order in which they undertook online tasks, and 74% were able to choose or change their methods of work. The autonomy of this work extended to the freedom to choose which clients to connect with and how much to charge them as well as the ability to choose when and where to work. What these findings highlight is how control exists even while workers experience significant autonomy and discretion over when, where, how, for whom and for how much they work. In fact, in economic sociology, it is widely recognised that control is central to all types of exchange unless embedded within high-trust personal relations. Algorithmic control then is central to the gig economy because it is precisely this enabling of exchange in low-trust environments which platforms permit. No longer does a client need to build a relationship with a local freelancer that a friend referred. Platforms provide the architecture to hire people on the other side of the planet who clients have never before been in contact with. Therefore, control must be understood as the existence of a disciplinary mechanism which effectively regulates behaviour regardless of the level of autonomy workers experience.

Such an understanding is clearly broad and thus raises the question of whether control is suitable for determining labour rights? In answering this question, it useful to consider what the purpose of labour law is, essentially we have labour laws in recognition that ‘employees’ are in a vulnerable position relative to the person hiring them due to their dependence on that person or firm for their livelihood and are thus in need of special protections which limit exploitation such as minimum wages. This contrasts with the situation of the genuinely ‘self-employed contractor’ who can be considered an equal party to those with whom they are entering into contracts and whose terms and conditions can thus be left to the market.

Therefore, it is clear that what matters is not the level of control (which will necessarily be significant in low-trust environments, regardless of the level of autonomy experienced by workers) but the degree of dependence. Labour law exists to protect those who would be in a vulnerable position due to the potential cost of replacing their current engagement if it were to end. In determining whether labour protections should cover an individual, the law should seek to determine the contribution which a particular source of paid work makes to a worker’s income and the likely difficulty they would experience in replacing that source of income. The significance of a particular job to a worker’s livelihoods will be effected by whether there are other companies operating locally in the sector which the worker could access as a viable alternative, and whether there are specific impediments to them doing so. In reaching a decision, the level of competition which exists in the local market and the ease with the worker could switch their loyalty should be looked at. For instance, whether a company retains workers’ data and client information or legally bars workers from making use of alternatives are all important barriers to an individual’s ability to find replacement sources of income.

The above considerations have important implications for the gig economy due to the monopolistic tendencies that platforms have due “network effects.” A network effect is a phenomenon whereby each additional user increases the value of the platform for all users. The network effect can make it difficult for new platforms to compete with established ones, as a new platform is of little value unless everyone switches platform at the same time. Moreover, many platforms include exclusivity clauses in their terms of services which can hinder workers and clients doing business outside of the platforms, and prevent workers from taking their platform profiles and reputations with them to another platform. In sum, a platform that controls the market for any given service might not exert much direct control over workers, but does place all of them into a highly dependent position from which workers will require labour protections.

In summary, gig economy platforms enable economic exchanges to take place in low trust environments, they do so by provide algorithmic control mechanisms by which workers can be disciplined. However, this control does not preclude gig economy workers experiencing high levels of autonomy over when, where, how, for whom and for how much they work. Therefore, control alone is not suitable for determining worker status. Instead greater emphasis should be placed on dependence when determining worker status. And to understand dependence in the gig economy, we need to develop nuanced understandings of the role that each platform seeks to play in their local social and economic environments.


Graham, M., Hjorth, I., Lehdonvirta, V. 2017. Digital labour and development: impacts of global digital labour platforms and the gig economy on worker livelihoods. Transfer: European Review of Labour and Research. 23 (2) 135-162.

Graham, M., Lehdonvirta, V., Wood, A., Barnard, H., Hjorth, I., and Simon, D. P. 2017. The Risks and Rewards of Online Gig Work At the Global Margins. Oxford: Oxford Internet Institute.

Lee, M. K., Kusbit, D., Metsky, E., & Dabbish, L. 2015. Working with machines: The impact of algorithmic, data-driven management on human workers. In Proceedings of the 33rd Annual ACM SIGCHI Conference, Seoul, South Korea (pp. 1603–1612). New York, NY: ACM Press.

Rosenblat, A. and Stark, L. 2016. Algorithmic Labor and Information Asymmetries: A Case Study of Uber’s Drivers. International Journal of Communication. 10: 3758-3784.

Shapiro, A. 2017. Between autonomy and control: Strategies of arbitrage in the “on-demand” economy. New Media and Society. 00: 1-18 (accessed 3 March 2018).

Slee, T. 2016. Platforms and Trust: Beyond Reputation. In Scholz, T and Schneider, N (eds) Ours to Hack and to Own: The Rise of Platform Cooperativism, A New Vision for the Future of Work and a Fairer Internet. New York: OR Books

Taylor, M. Marsh, G., Nicole, D., Broadbent, P. 2017. Good Work: The Taylor Review of Modern Working Practices. Department for Business, Energy & Industrial Strategy. Available at:

Wood, A.J., Graham, M., Lehdonvirta, V. and Hjorth, I., 2017. Good gig, bad gig: autonomy and algorithmic control in the global gig economy. International Labour Process Conference. 5th April 2017. Sheffield.