By Klaus Prettner and David E. Bloom
Contents
1. Introduction 1
1.1 Technological progress and its economic consequences 1
1.2 The economic consequences of automation: could this time be different? 5
1.3 The social impacts of automation 7
1.4 The race against, or the race with, the machine? 11
1.5 Summary 13
References 14
2. The stylized facts 21
2.1 Adoption of automation technology 21
2.2 Dynamics of economic growth and welfare 22
2.3 Dynamics of the labor force and of unemployment 30
2.4 The evolution of inequality 34
2.5 Summary 41
References 42
3. Empirical evidence on the economic effects of automation 47
3.1 Occupations, jobs, and tasks susceptible to automation 47
3.2 Cross-country evidence on the economic consequences of automation 51
3.3 Summary 62
References 63
4. A simple macroeconomic framework for analyzing automation 67
4.1 Preliminaries and definitions 67
4.1.1 Growth rates in discrete and in continuous time 67
4.1.2 Representative individuals and representative firms 69
4.1.3 Aggregate production function 70
4.2 The simplest version of the standard 1956 Solow model in discrete time 75
4.3 The Solow model in continuous time with technological progress and with population growth 82
4.4 The Solow model with automation 90
4.5 Endogenization of the share of investment in traditional physical capital 98
4.6 Automation and wage inequality 101
4.7 The tradeoff between growth and inequality 104
4.8 Summary 107
References 109
5. Endogenous savings and extensions of the baseline model 113
5.1 Introduction 113
5.2 Cookbook procedures for static and dynamic optimization 114
5.2.1 The method of Lagrange 115
5.2.2 The method of Karush_Kuhn_Tucker 123
5.2.3 Dynamic optimization in discrete time in the case of two time periods 124
5.2.4 Dynamic optimization in continuous time 126
5.3 Endogenous savings in the Ramsey_Cass_Koopmans model 131
5.4 Automation in the Ramsey_Cass_Koopmans model 138
5.5 Endogenous savings in the OLG model 140
5.6 Automation in the OLG model 147
5.7 Discussion of extensions 150
5.7.1 Endogenous technological progress and automation 150
5.7.2 Technological unemployment 152
5.7.3 International trade, foreign direct investment, and automation 154
5.8 Summary 156
References 158
6. Automation as a potential response to the challenges
of demographic change 163
6.1 Introduction and stylized facts 163
6.2 Demographic change and its economic consequences 168
6.3 How robots can help 172
6.4 Future employment projections based on automation 175
6.5 The reverse channel: could robots affect demography? 176
6.6 Summary 177
References 179
7. Policy challenges 187
7.1 The challenges 187
7.2 Education as a strategy to cope with the negative effects of
automation 188
7.3 Labor market policies 190
7.4 Taxation in the age of automation 192
7.5 Social security in the age of automation 197
7.6 Demand-side policies 201
7.7 Summary 202
References 205
8. Peering into the future: long-run economic and social
consequences of automation; with an epilogue on
COVID-19 209
8.1 Joblessness, misery, and deaths of despair or “the happy
leisure society”? 210
8.2 Spatial and regional implications: the future of cities 212
8.3 The question of how we care for each other 214
8.4 The meaning of being human 216
8.5 Epilogue on COVID-19 217
References 218
Index 223
Preface
News reports abound that robots, three-dimensional (3D) printers, and algorithms based on machine learning are outperforming humans in many different tasks (see, e.g., Davison, 2017; Financial Times, 2019; The Economist, 2014, 2017a, 2019). Industrial robots have long been familiar sights on assembly lines in many businesses—especially in the automotive industry.
Recent advances, however, relate to tasks and activities that were seen as being nonautomatable just a few years ago. For example, because driving cars and trucks requires many instinctual moves and reactions to unforeseen events, it has been regarded as an activity that an algorithm cannot perform.
Yet driverless cars are now being tested on public roads, and self-driving cars and trucks are widely expected to become increasingly common within the next 10_20 years—and much safer than human drivers (Chen, Kuhn, Prettner, & Bloom, 2019; The Economist, 2018). Producing customized parts, prototypes, and medical implants without any human labor input has become feasible by advances in 3D printing (Abeliansky, Algur, Bloom, & Prettner, 2020; The Economist, 2016, 2017b). Algorithms can diagnose illness— particularly rare diseases—and reliably interpret X-rays based on advances in machine learning, the availability of large datasets to inform the algorithms, and the secular rise in computing power over the preceding decades (Ford, 2015). While programs have written simple reports and straightforward newsflashes for quite some time, more surprising is that computers have surmounted more complex challenges, such as writing novels and conducting experiments to uncover some laws of nature (Barrie, 2014; National Science Foundation, 2009; Schmidt & Lipson, 2009).
These unprecedented technological advances have raised fears that society is unprepared for the attendant—potentially dramatic—economic and social changes that may be looming. The most pressing concern from an economic perspective is that the replacement of labor by automation technologies will lead to high rates of technological unemployment. In addition, because less-educated persons are likely more susceptible to displacement by automation, some fear that wage inequality might rise. Moreover, in contrast to the wages of workers, the revenue that robots and other automation technologies generate flows to the owners of these devices, who are predominantly wealthy. Thus, functional income distribution might change such that capital’s share of income rises further and the labor income share correspondingly decreases. Because labor income is much more equally distributed than wealth and the returns thereof, such changes would mechanically lead to a rise in overall inequality.
The fear that new technologies might create technological unemployment and raise inequality—sometimes referred to as “automation anxiety” and a potential source of political resistance—is not a recent phenomenon (LeVine, 2018; see Frey, 2019 for an overview). The emergence of new types of machines that threaten to replace workers has always sparked resistance in different parts of the population (e.g., the Luddite uprisings in the 19th century).
Also the economic literature analyzing the effects of technological changes from different angles (Ricardo, 1821; Wicksell, 1906) concedes that the course of technological progress does not only produce winners. However, past dire predictions of mass unemployment and widespread poverty due to the emergence of modern machines did not come to pass.
The crucial question to ask in the age of automation is whether this time will be different (Ford, 2015). Industrial robots, 3D printers, and devices based on machine learning are not only increasing labor productivity and thereby allowing workers to produce more output with less labor input, but are also performing their tasks with full autonomy—i.e., they are replacing workers altogether. For example, while the assembly line raised the productivity of those workers who were still required to staff the lines in factories (and thereby their wages), an industrial robot does not require direct labor input and thus replaces workers without raising their productivity (see Growiec, 2019 for a discussion of the important distinctions among mechanization, automation, and artificial intelligence (AI)). To what extent this difference between machines’ historical effects and robots’ contemporaneous effects changes the likely outcomes of technological progress in the future remains an open and highly important question.