Conference Agenda

This paper is concerned with the nature of jobs created by the expansion or penetration of digital economic activity in lower income contexts, and what this may mean for economic or structural transformations for countries in the global South. We ask what possibilities new jobs and forms of labour in the digital economy hold for the transformation of economies in ways that contribute to achieving the goals of human, inclusive and sustainable development. The key question thus concerns how digital work or technologically driven changes in the nature of work can be part of or contribute to a productive transformation of low-income economies and thus to economic development. Can a ‘digital transformation’ spur development, and if so how, and to whose benefit? What are the impacts on work and workers in this process? This paper provides a synthesis of literature and debates - conceptual, historical and empirical - linking digital jobs with ideas of ‘structural transformation’ and development. We explore the relevance of these ideas to contemporary development conditions, including how digitalisation is transforming the way countries are integrated in global supply chains, to identify what opportunities digital transformation offers low income countries, and what actions will be needed to shape these towards a sustainable, fair and inclusive social and economic development.

The surge in social media and e-commerce usage during the COVID-19 pandemic increased the demand for business process outsourcing (BPO) services, especially data labelling, in the Global South. Recent literature on BPO firms catering to technology and artificial intelligence (AI) companies has highlighted their employment potential and conditions of work, but little is known about the everyday mechanisms by which such firms generate value and increase productivity.

Indian BPOs in AI data labelling form distinct supply chains with domestic or international data-driven companies at the helm. Within these supply chains, BPO workers are often put into intra-firm, target-based competitions during working hours. BPOs have historically followed similar tactics to increase productivity, yet data labelling as work offers AI supply chain BPOs something different. Partially accurate datasets (produced by workers who do not win competitions) can still be used to train machine learning models for the firm’s clients. As a result, unlike in traditional BPOs, the worker’s skills, strategies, and accuracy are less valuable than speed, as other studies have found. Further, the firms’ use of automated evaluation tools (with human-in-the-loop quality analysts) to ‘judge’ data labelling work constrains the scope for feedback and worker growth.

In this paper, we present primary data with Indian data labellers’ perceptions and experiences of their working conditions, including competitions, reasons for continuing work, choices to leave, internal mobility and skill requirements in the post-pandemic Indian economy. Responses from survey data (120) and in-depth interviews (9) indicate that Indian data labellers do not see their work as particularly helpful for mobility in the labour market, while foregrounding the wage theft that accompanies the seeping in of ‘playbour’ or gamified behavioural design in their work.

This has significance since the current policy environment regards BPOs as drivers of inclusive growth and job generators across urban and rural India, especially for women workers. There are public subsidies for the sector through programmes such as the India BPO Promotion Scheme (IBPS). BPOs in the AI value chain will receive further attention after the announcement of national AI Centres of Excellence in India’s 2023 Budget. Our data highlights how BPO firms with different business models that use competitions and output targets create poor quality and stagnant employment where productivity is not adequately compensated for. This study highlights the need for worker-facing studies to look closely at intra-firm governance and incentive structures to measure socially productive and meaningful employment growth.

Content moderation is a key practice for maintaining the public discourse on social media platforms, but is driven by profit interests of technology firms (lead firms) operating them. For reasons of labor cost arbitrage and to access local skills and knowledge, most of this work is outsourced to the Global South, including to the IT-Services sector in India. Lead firms design the organization of work in manners that allow them direct control over the labor process, so as to ensure quality service delivery, which refers to the accurate and context-sensitive content moderation decisions. Informed by an interdisciplinary study of content moderation value chains using GVC and GPN constructs together with the labor process theory, this paper shows how lead firms use their automated work software for directing tasks to outsourced workers, monitoring them and codifying their knowledge in the software. While lead firms continue to rely on workers for their tacit knowledge, as all labor knowledge cannot be codified, the use of automated technology results in high degrees of work standardization and limited skill development. Automation of certain managerial tasks (as mentioned above) also has implications for the inter-firm governance relationships, which although relying on suppliers for on-floor supervision of workers, allow for limited possibilities for functional upgrading.

The paper focuses on the case of a supplier exit from its project with a large lead firm, and the subsequent project transfer, including workers, to another supplier in India. Taking place during the COVID-19 led lockdown in India, where all moderation work was taking place from home, this case of GVC restructuring shows on one hand, the replaceability of one large supplier firm with another, and on the other, the reliance on experienced workers for supporting the transfer of work and continuing work from home during the lockdown. From this perspective, the study here provides granularity to the existing analyses of labor flexibility in service value chains by showing how direct control of lead firms over the labor process, using automation, has implications for both workers and suppliers. This sheds new light on the competitive dynamics in the global economy and positions the role of public policy in securing the interests of workers and upgrading possibilities for low-end service suppliers. The analysis is informed by a mixed-methods inquiry with 65 interviews and 99 surveys with workers, management at supplier firms, one lead firm, trade union representatives, civil society experts and other experts.

The Kaldor-Verdoorn law is a cornerstone of Keynesian economics. It considers a dynamic relationship between the growth rates of output and labour productivity with a direction of causality from the former to the latter. This relationship has been investigated by several generations of economists from both a theoretical and empirical perspective. The law can be thought of as a stylized fact generally confirmed for developed as well as developing countries. Yet, the structural change from a manufacturing to service economy led scholars to reconsider its overall validity. The lack of economies of scale and the different division of labour in the service sector are believed as key mechanisms behind the detachment between increases of demand and gains in productivity experienced since the Eighties. In this paper we want to investigate if, and in which way, the increasing robotisation experienced by many industries impacts upon the supposed channels through which the Kaldor-Verdoorn law shapes the dynamics of labour-productivity growth and employment. We first offer a simple and evolutionary interpretation of the law that combines both the traditional Kaldorian and Post-Keynesian arguments with the evolutionary literature on innovation and technical change. We then collect data on labour productivity, capital-labour ratio, GDP, and robot adoption for a panel of 17 industries in 25 OECD countries for the period 1990-2018. After checking for the presence of mechanisms à la Kaldor-Verdoorn in the dynamics of productivity and employment, we analyse through a GMM approach in which way the increasing robotisation affects the channels behind the Kaldor-Verdoorn law. The results are contrasting and somewhat puzzling. Industries whose robot density is lower than cross-country sector average still benefit from a cumulative-causation process in their productivity dynamics. Conversely, industries with a robot density higher than the corresponding cross-country sector average suffer from a weakening of the meso-economic channel that links productivity achievements to the growth in mechanisation. At the same time, their higher robot density sustains the macro-level channel that associates changes in labour productivity with GDP growth. The evidence of technological unemployment depends on the specification of the underlying econometric model. This overall evidence agrees with the empirical literature that finds a positive relationship between robots adoption and employment at firm and industry level.