David Dorn’s research covers globalization, labor markets, technological change, innovation and inequality in society. In 2016, he and his co-authors David Autor (MIT) and Gordon Hanson (University of California, San Diego) published the highly acclaimed paper "The China Shock". In this and subsequent scientific papers, he uses empirical data to show that China's entry into world trade had not only economic but also political and social consequences for the USA.
We spoke with him about the impact of these developments on the labor market and society.
How has digitalization changed the labor market?
Since the 1970s, computer-controlled machines and robots increasingly perform repetitive activities within the industrial production process and replacing human labor. Since the 1990s, office computers have become commonplace and the internet has entered most workplaces. In this process, an increasing number of jobs involved in information processing and transmission, such as accounting and archiving, have been automated.
The factory and office jobs affected by automation often lie in the middle of the income distribution. The loss of employment in these middle tiers has led to a polarization of employment, with the remaining occupations being at either the top or the bottom end of the income scale. These remaining roles include managers, engineers and scientists on the one hand, and waiters, childcare workers and caretakers on the other.
Is income distribution becoming increasingly unequal?
Automation has indeed led to an increase in income inequality. The distribution of total income is shifting away from earned income to capital income, which is much more concentrated in the hands of a few people. However, we also see increasing inequality within earned income: the wages of well-qualified employees are rising faster than those of less qualified workers. For the latter, we have seen declining income levels in some countries during recent decades.
... so there is little hope for the middle class?
The recent technological developments have clearly hurt the middle class. There are, however, political opportunities through the tax and transfer system to compensate some of the undesirable side effects of technological change.
Artificial intelligence is the next step of technological change. How do these two developments and their impact on the labor market differ?
The main difference between artificial intelligence and automation is the logic by which these two technologies work. In the case of automation, computer programs govern computers and robots. A programmer defines exactly how the robot or machine should react in a given situation. This requirement for precise programming limits the use of computers and robots. It is often difficult to define all the potential situations a machine could encounter and how it should react in each case. In the case of artificial intelligence, which is based on machine learning technologies, the computer does not need detailed instructions on what to do in a step-by-step program. Instead, a large amount of data is given to the machine, allowing it to calculate the best course of action by itself based on the provided data. In practice, we already see great improvements in speech and image recognition, where computers can be trained using libraries of images and sound recordings, so that they become increasingly better at recognizing such content. However, a wide-ranging use of such technologies in the workplace has yet to take place. It is currently still unclear in which areas these methods will actually be good enough so that they surpass human performance, and can be applied widely in the labor market.
So in future, machines could make the kind of decisions humans make today?
This is one of the major challenges. Think of medical treatments for example. Do you really want to rely on medical diagnoses generated by a computer? Or should a computer take on the decision of a human resource manager and select candidates for a job? A major issue to overcome is that the computer’s decisions are based on the libraries they have been given for training. This means they could propagate unwanted patterns from these libraries into the future. For example, if a computer observes that in the past, men were much more likely to move up into management positions than women, the computer may suggest only men to be promoted.
What kind of work will computers not be able to replace in the near future?
Despite the somewhat misleading term ‘artificial intelligence’, machines do not develop the kind of creative intelligence we humans possess. A machine trained on data from the past cannot develop a novel idea that is completely outside the spectrum of previous expectations or experiences. Machines also have great difficulty in all areas in which a response to a fundamentally new situation is required. Such situations are not only typical for scientific research or management tasks, but they occur also in many less well qualified professions, i.e. when dealing with individual customer needs. These kinds of jobs are difficult for machines to master.
You have done a lot of research on the impact of job losses in the US manufacturing sector. What did you find?
In the US, the number of jobs in the manufacturing sector has declined rapidly since 2000. This was due to advances in automation, but also because of increased trade competition, especially from China. The disappearance of relatively well-paid factory jobs led to a significant drop in employment and incomes in some regions of the USA. Men in particular have been strongly affected by these developments, as they traditionally provided the economic livelihood of families. As a result, we now see a whole series of consequences: an increase in the proportion of young men who are in neither education nor in the active workforce, declining marriage rates, and an elevated mortality among young men. The latter is accompanied by an increase in drug abuse and violent crime. These are real, long-term social consequences of economic decline.
Is this a US-specific problem, or will it also affect other regions and countries?
The extent of the problem seems to be much greater in the US than in other countries. Different from most European countries, the welfare system of the U.S. provides a relatively weak protection against negative economic developments. It can happen that a person becomes homeless and lacks adequate medical care after losing a job.
Switzerland has the advantage of being a small country. No region will be completely left behind economically, since there is always spatial proximity to more prosperous areas. Even though cities such as St. Gallen or Winterthur have historically suffered from the decline in their manufacturing sectors, they are located near enough to other booming cities so that employees can easily access those labor markets.
One of the surprising results of my research is that the majority of the U.S. factory workers who lost their jobs do not move to other areas where there are more jobs. Instead, most of them remain in declining regions even once most jobs are gone.
You mentioned that there are political measures that could soften the negative side effects of technological innovations. Can you expand on this?
The response to this development should be twofold. On the one hand, there is the question of how we should deal with people who have already lost their jobs. This is a big challenge. Older workers have difficulties finding new employment after losing a job. Governments should ensure that basic human needs are provided for in the event of unemployment and support these workers in finding work. The second, forward-looking aspect of the adaptation process lies in the realm of our education system. Children should learn skills that computers cannot replace. Creativity, problem-solving and communication skills are key. Memorization, which traditionally was very important in education, has much less value these days as we have machines that are clearly superior to us in this regard.
Is there potential for technology to counter the negative effects?
In general, technological progress allows economy to produce goods and service at a lower cost, which creates greater prosperity. Looking back at the last century, we see dramatic developments based on innovation: New means of transport such as cars or planes, improvements in medicine and major developments in communication technologies, such as television and the internet. All these advances have increased the prosperity of society as a whole.
Since the 1970s, however, macroeconomic growth has slowed significantly. Contrary to frequent newspaper headlines, there is much evidence to suggest that the pace of technological change has decelerated considerably during the last several decades.
With less technological change, there is also less growth in total income. When this modestly increasing total income has an ever more unequal distribution in the population, then we will face a situation where the income of many individual people declines. If, as some people predict, we instead have some major technological breakthroughs ahead of us, then that is a desirable scenario. It would generate more overall wealth, thus giving us more opportunities to address the problem of stagnating middle-class incomes.
Which digitized function would you miss most on a daily basis?
As an employee in an information-based sector, the internet is crucial for researching information and for communication and collaboration with my international co-authors.
What is the most unnecessary?
That is a difficult question. If it is unnecessary, I don't need it.
David Dorn ist Professor für Globalisierung und Arbeitsmärkte, gestiftet vom UBS International Center of Economics in Society. Er befasst sich mit technologischem Wandel, Innovation und sozialer Ungleichheit. Seine wissenschaftlichen Arbeiten gehören zu den meistzitierten Publikationen in wissenschaftlichen Zeitschriften. 2016 veröffentlichte er mit seinen Co-Autoren David Autor (MIT) und Gordon Hanson (University of California, San Diego) das viel beachtete Paper «The China Shock». In dieser und nachfolgenden wissenschaftlichen Arbeiten zeigt er aufgrund von empirischen Daten, dass der Eintritt Chinas in den Welthandel nicht nur wirtschaftliche, sondern auch politische und soziale Konsequenzen für die USA hatte.