Before ChatGPT: the economic impact of technological improvement since the 1970s
Last week I went to a business event and unsurprisingly, everyone was discussing ChatGPT and big data. While investors are trying to seize new profitability and students are developing new cheating methods, some people express fear of technological improvement – it is not new and has existed for centuries. In 1811, the Luddites who thought the new machinery would cause unemployment and wage reductions sent threatening letters to employers, and broke new machines in the factories; in the 20th century, the fear of AI was one of the top American fears. The productivity improvement has also caught economists’ attention. A century after the Luddite movement, John Keynes first used the concept of ‘technological unemployment’ and expressed his concern over it in his article, ‘Economic Possibilities for our Grandchildren’. He wrote: ‘We are being afflicted with a new disease of which some readers may not yet have heard the name, but of which they will hear a great deal in the years to come—namely, technological unemployment.’ Has Keynes’ prediction come true? In this article, I will first review the neoclassical arguments for technological progress and discuss them with data since the 1970s.
Neoclassical Economics: machines may not destroy jobs
The declining trend of labour share in the agricultural and manufacturing sector while a rising share of which in the service industry has been witnessed across developed countries. Therefore, Lacalle concluded: ‘Technology does not destroy employment, what it does is free capital from obsolete sectors to new sectors and, thereby, improve the quality of life of all and, in addition, create much more direct and indirect employment.’ There are five mechanisms which prevent long-term technological unemployment from happening.
First, there is a rising demand for labour to maintain the operation of the technology. Two economists, Wyatt and Hecker, from the U.S Bureau of Labour Statistics found that due to greater mechanization and the number of machinery increased dramatically, demand for mechanics repairing these machines thrived as well. Between 1910 and 1950, as a proportion of the labour force, mechanics and repairers grew 10.9 times in the USA. Second, technological progress is always accompanied by the creation of new industries, for example, Online shopping (also known as E-commerce) was generated by the emergence of the internet, which provided 380,100 people with jobs in 2016 in America ranging from skilled occupations like market research analysts, operation managers, software developers and web developers, to relatively unskilled occupations like packers and consumer service representatives. Third, there is a demand for researchers to improve technologies. As firms are more aware of the benefits of improved technology resulting from increased profits, the demand for technology may rise as well, thus generating more job opportunities in research and development.
Apart from direct channels, indirect mechanisms create jobs by altering other factors like aggregate supply and aggregate demand. When more advanced technology is adopted, productivity rises and the cost of production falls. Therefore, firms see higher profitability and may expand their plants and hire more workers to produce more. Besides, workers usually earn higher wages when productivity rises, as economic theories suggest that real wages should be equal to the marginal revenue product, which is positively correlated with productivity. Aggregate demand shifts rightward as a result and production expands to keep up.
Overall, these mechanisms functioned well in the past few decades. There were 24,764,881 jobs in England and Wales in 1992 and the figure expanded to 30,537,415 in 2014. The trend that job creation outnumbers job destruction may keep in existence in the coming decades: a report published by the World Economic Forum in 2018 estimated that ‘75 million jobs may vanish, while 133 million additional new roles may emerge concurrently’ in the next coming decade; the Institute for the Youth and Dell Technologies estimated that 85% of the jobs in 2030 haven’t been invented yet.
However, the above analysis missed the story of inequality.
According to the compensation factors, the real income of all workers should rise resulting from an increase in productivity during technical progress. However, in the USA, GDP per capita rose exponentially since 1975, indicating that productivity was rising, while real median household income experienced a stagnation (figure 1) and even fell in the last decade if only working-aged people were considered, according to Brynjolfsson and McAfee. Therefore, we can deduce that in terms of purchasing power, some workers did not benefit economically from such progress.
An interesting fabricated story helps to explain the theory behind the divergence between GDP per capita and real median household income: ‘If 50 construction workers are drinking at a bar and Bill Gates walks in as the poorest customer walks out, the mean wealth of the customers would soar to $1 billion. However, the wealth of the median customer, the one exactly in the middle of the distribution, wouldn’t change at all.’
Figure 1: Index of growth in U.S. real GDP per capita and real median household income. (Brynjolfsson & McAfee, 2012: p105)
Besides, it is not a unique ‘American’ phenomenon. From reviewing the Gini index in other countries, being more unequal is a trend across the globe North since the 1970s and 1980s, though the extent differs from country to country. Not everyone has benefitted from technological progress and the less privileged are even worse off.
The enlarged wealth gap could be explained by the classical theory of the labour market. As technology progresses, the relative demand for skilled workers rises simultaneously and therefore, the relative wage of skilled workers increases. We can see strong evidence of it from the increasing premium of a college or a high school degree and the growing proportion of college graduates in employment (Goldin & Katz, 2009). Although there were more skilled workers in the labour force, the premium of skill was still rising (figure 2), implying that the supply of skill was not keeping up with the demand.
Figure 2：Relative demand and relative supply of skill. (Goldin & Katz, 2009: p97)
However, the driver of the higher relative demand for skilled workers could be controversial, as technological change was not the only disruptive phenomenon in the 1970s. Globalization, on the other hand, also sent jobs in developed countries to economies with lower labour costs. Nevertheless, China was responsible for only 25% of the 5.5 million lost manufacturing jobs in America during the 2000s (Hernandez, 2018). Hence, describing globalization as the greatest job-destroying machine and the biggest threat of inequality is biased; Technology has played an important role as well and the impact of it might be even more influential. ‘A large empirical literature concludes that trade has played a relatively minor role, relative to skill-biased technical change, in the rise of the skill premium.’ Ethier (2005) suggested.
The solution provided by many economists sounds simple – education! Once the supply of skilled labour catches the demand, the wage gap shrinks and the merits of technology will be shared more equally. This analysis would only be correct if technologies are tools of humans so the values are still added to human workers. However, in the era of ChatGPT and advanced AI, are there possibilities that machines will substitute human workers eventually? In that case, not only the recommended policy responses but also the economic theories that we are used to nowadays will be changed dramatically. For me, the fear of technology is never irrational under future uncertainties.