The impact of AI on international trade and labour market inequalities
Structural transformations as a result of artificial intelligence could level out inequalities between different types of workers, writes Bart Los, professor of economics at University of Groningen.
The world seems to be a more dynamic place than ever. Artificial intelligence has made huge leaps forward over the past few years. This has made many people think not only about the opportunities that new technologies offer, but also about the extent to which their jobs might be at risk. Increased geopolitical tensions have made multinational firms reconsider the configurations of their supply chains, which will have implications for employment around the world.
The GI-NI research project, funded by the European Union’s Horizon 2020 programme, analyses the effects of these structural ‘transformations’ on inequality across categories of workers. These categories are defined by the ‘business function’ that workers perform. A ‘value chain’ is defined as all the activities required to bring new products to the market. The ‘smile’ production curve splits value chains into three types of business functions. These are pre-fabrication functions, fabrication functions and post-fabrication functions.
Workers in fabrication functions (which generally involve routinised activities, both in manufacturing and services industries) generally capture a small part of the value created in the chain. The pre-fabrication and post-fabrication functions (like research and development/design and marketing activities, often labelled ‘high quality’ functions) tend to command much higher remuneration (Figure 1).
Figure 1. Fabrication workers receive lower remuneration than pre- or post-fabrication
Source: US Grain Council (2020)
The globalisation of the world economy due to trade liberalisation, falling costs of transportation and reductions in the costs of coordinating activities over long distances thanks to rapid improvements in information and communication technology have led to a deepening of the smile curve. The large numbers of workers with low wages compared to their productivity in countries in the East (of which China is a prime example) became competitors to fabrication workers in western Europe, North America and Japan.
Falling costs of locating elements of value chains in different parts of the world allowed firms in advanced countries to offshore fabrication functions, leading to the proliferation of global value chains. Fabrication workers in China, India and Mexico did not threaten HQ workers in countries like Germany and the US, and the ‘additional supply’ of HQ workers from countries entering the global economy was much smaller. The smile curve became deeper and inequality between fabrication and HQ workers increased.
For a long time, technological change also led to a deepening of the smile curve. Jobs that require a lot of routinised activities are not only easily offshored but can also be done relatively easily by computers or robots. Fabrication workers in Germany and the US began to compete not only with workers in other countries, but also with capital.
Although research has revealed that not all ‘easy-to-offshore’ types of jobs are also ‘easy-to-automate’, it is clear that many workers in the middle part of the smile curve faced a double whammy. At the same time, the productivity of HQ workers often increased due to affordable new technology and new types of jobs requiring creativity rather than routine. The two structural transformations contributed to increasing inequality in advanced countries.
Will AI increase labour market inequalities?
The labour market consequences of this set of new technologies will most probably save many jobs, like robots and computers (and the steam machine and electricity in the more distant past) did. The types of jobs that are at risk of being substituted by AI are very different, however.
The launch of ChatGPT suggested that AI might soon perform the activities of many HQ workers in cheaper and more effective ways. Writing software code and designing websites are activities currently performed by HQ workers and in which AI is progressing rapidly. The extent to which AI can replace routine workers might be smaller, suggesting that AI might flatten the smile curve.
Further innovations might affect the location of business functions in GVCs. Currently, some activities are most effectively performed if done in close proximity. Firms can improve the effectiveness of their R&D and design activities if the workers involved communicate frequently with employees involved in marketing and after sales services. AI does not require this type of face-to-face exchange, so opportunities to offshore larger fractions of HQ jobs increase. AI might reverse the tendency towards increasing inequality between workers in fabrication and HQ-functions via this indirect channel.
Firms are not the only agents shaping the world economy and its inequalities. Public policies have an impact on the pace at which AI can be introduced. As the US-China trade war and the tensions between the European Union and China have shown, such policies can also affect the rate of globalisation.
It is impossible to predict how AI will affect inequalities in labour markets. Still, it is important to understand the underlying mechanisms, if only to evaluate the effects of policy alternatives.