Automation and Its Macroeconomic Consequences

Theory, Evidence, and Social Impacts

Klaus Prettner, David E. Bloom

• Publisher: Academic Press
• Year: 2020
• Print ISBN: 9780128180280
• eBook ISBN: 9780128180297
• Topic: Economics
• Subtopic: Business Development
• Index: Economics

Automation and Its Macroeconomic Consequences reveals new ways to understand the economic characteristics of our increasing dependence on machines. Illuminating technical and social elements, it describes economic policies that could counteract negative income distribution consequences of automation without hampering the adoption of new technologies. Arguing that modern automation cannot be compared to the Industrial Revolution, it considers consequences of automation such as spatial patterns, urbanization, and regional concerns. In touching upon labor, growth, demographic, and policy, Automation and its Macroeconomic Consequences stands at the intersection of technology and economics, offering a comprehensive portrait illustrated by empirical observations and examples.- Introduces formal growth models that include automation and the empirical specifications on which the data-driven results rely- Focuses on formal modeling, empirical analysis and derivation of evidence-based policy conclusions- Considers consequences of automation, such as spatial patterns, urbanization and regional concerns

Chapter 1

Introduction

Abstract

This chapter provides an overview of the potential economic and social consequences of technological progress in general and automation and artificial intelligence (AI) in particular. First, we discuss the effects of technological progress in the past and analyze the extent to which historical insights can be a guide to the future in the age of automation. Then, we describe some of recent years’ path-breaking developments in automation and AI and the political, social, and economic challenges they present. Finally, we contemplate the extent to which humans can compete with robots and AI by racing against, or—as others have suggested—racing with the machine.

Keywords

Technological change; historical effects of technological progress; industrial robots; artificial intelligence (AI); structural change; autonomous driving; urbanization; education; economic policy; human augmentation; political; social and economic consequences of automation

1.1 Technological progress and its economic consequences

The exceptionally rapid pace of technological progress since the Industrial Revolution has allowed many countries and their inhabitants to reach historically unique levels of economic and social prosperity.¹ In the 18th century, global life expectancy hovered around 30 years, almost every other child died before the age of 5 years, average income levels were barely above subsistence level, and close to 90% of the population was illiterate (Maddison, 2010; Roser, 2018). The situation could not be more different a bit more than 200 years later. Global average life expectancy is above 70 years, more than 95% of children survive their fifth birthday, average real incomes have increased by a factor of 11 on a global scale, and illiteracy is below 15% (Maddison, 2010; Roser, 2018). Because these numbers are collective averages, they imply that some richer countries such as Australia, Canada, France, Germany, Japan, the United Kingdom, and the United States have experienced substantially larger gains and more pronounced transformations.

Notwithstanding the positive effects of technological progress on material well-being and quality-of-life in the aggregate, some have expressed fears about the distribution of the gains among different segments of the population—in terms of high unemployment, the impoverishment of workers, and the evolving nature of work in general (see, e.g., Frey, 2019; Scott & Gratton, 2020; Geiger, Prettner, & Schwarzer, 2018; Prettner, Geiger, & Schwarzer, 2018). As early as 1776, Adam Smith expressed some concerns in his book An Inquiry into the Nature and Causes of the Wealth of Nations (1776). While generally optimistic about the quantitative employment effects of technological progress, the accompanying increase in the division of labor could, in his view, result in a greater monotony of work and thus less pleasant working conditions.

The most prominent resistance to new technologies from a historical perspective emerged in the Luddite uprisings, which took place between 1811 and 1816 in England. These violent protests were directed against the introduction of the mechanical loom, which raised weaving productivity by a factor of more than three. The Luddites destroyed weaving machines as their main form of protest in uprisings that the military ultimately suppressed. In response to the riots, David Ricardo revised his previous view that the introduction of a new technology would, without fail, be advantageous to all. In the newly appended chapter “On Machinery” in the third edition of his book On the Principles of Political Economy and Taxation, Ricardo (1821) examines the conditions under which technological unemployment can arise in the wake of technological changes. He claims that advances in the form of new machines could lead to higher unemployment over time, given that capitalists spend heavily on the new labor-saving machines and reduce the overall outlays for wages. This in turn could lead to competition among workers, reducing their wages. These changes could even lead to starvation in some parts of the population as the price of food could be put out of reach for some households. Ricardo’s analysis, which remains influential to this day, led to significant contributions by well-known economists such as Wicksell (1906), Hicks (1973), and Samuelson (1988).

In contrast to Ricardo (1821), Wicksell (1906) views labor supply as inelastic and wages as elastic, such that reducing wages could avert technological unemployment. Wages only become inelastic when they are at the subsistence level and cannot fall further. Only from that point onward might unemployment rise, in which case Wicksell (1906) recommends a social security system financed by the profits of capital owners—who gain from the introduction of new machines—as a remedy for technological unemployment (see also Hagemann, 1995; Humphrey, 2004). On top of these arguments, Wicksell (1906) questions whether the new technologies introduced in the 19th century are always labor saving and reduce the marginal value product of workers; in fact, he expected the opposite to be true because the advances were at least somewhat complementary to the production factor of labor. This argument, however, cannot be transferred unconditionally to today’s technological debate in the area of automation, as we will see later.

This brief discussion already demonstrates that the expected economic effects of technological change in general and of automation in particular depend crucially on the underlying modeling assumptions. Thus, from a theoretical perspective, the likely outcomes of technological changes in terms of employment, inequality, and overall economic growth and well-being are rather sensitive to variations in the underlying framework. Consequently, complementing theoretical considerations with empirical analyses to gauge the total effects of automation is particularly important.

If we depart from theoretical considerations that have shaped historical debates and look more closely at the data, clearly some of the fears regarding the economic effects of technological changes were overblown. As far as technological unemployment is concerned, a record number of persons are employed worldwide today and unemployment rates are comparatively low.² Furthermore, at least in industrialized countries, mass starvation has been eradicated.

The negative economic consequences of technological changes that many feared did not materialize primarily for the following three reasons: first, technological developments triggered strong growth in income levels and declines in the relative price of the goods that were produced with the advanced technologies. As a result, aggregate demand increased to such an extent that, despite the increase in labor productivity due to technological progress, the volume of work did not decrease; on the contrary, it often increased. Take, for example, the power loom that raised productivity in weaving (by a bit more than a factor of three). If the overall increase in wages and the relative price decline of textiles due to better technology lead to a rise in demand by a factor of five, then even more workers are needed in the new long-run equilibrium for weaving. Of course, this mechanism only works when labor remains an essential input in operating the looms. In general, the described mechanism is relevant for technologies that are, to some extent, complementary to labor. However, this mechanism might not be operative in the age of automation because—if we take the definition of automation seriously—it implies a perfect and complete substitution of capital for labor (see, e.g., Merriam-Webster, 2017, for the formal definition of automation and Acemoglu & Restrepo, 2018b; Growiec, 2019; Hémous & Olsen, 2018; Prettner, 2019; and Prettner & Strulik, 2020, for the differences in the economic effects of automation versus other forms of technological progress such as mechanization).³

The second reason why past technological changes did not result in mass unemployment is the structural transformation of modern industrialized economies (cf. Vermeulen, Kesselhut, Pyka, & Saviotti, 2018; Vermeulen, Pyka, & Saviotti, 2020). Two hundred years ago, most of the population was employed in agriculture producing food. Today, the employment share of agriculture in developed countries is below 5% (see, e.g., Herrendorf, Rogerson, & Valentinyi, 2014). Where did all that labor go? Of course, entirely new employment possibilities emerged in manufacturing due to technological changes. More important, a whole new sector emerged: labor-intensive services. Two factors drove the shift in employment away from agriculture and later from manufacturing:

• • The first factor was the increasing demand for services because the income elasticity of many services is well above unity since they often represent a kind of luxury good. By contrast, the income elasticity of agricultural production, in particular of food, is typically well below unity because consumers are subject to satiation. With this structure of nonhomothetic preferences, rising income levels imply a mechanical rise in the demand for services and thus, a rising employment share of the service sector.
• • The second factor is that technological progress raised productivity in agriculture and manufacturing such that fewer workers were needed to satisfy the limited demand of goods that exhibit an income elasticity below unity. Thus the higher productivity set workers free and drove them into services, where productivity did not rise that much (cf. Autor & Dorn, 2013; Baumol & Bowen, 1966).

Based on these arguments, Cowen (2013, p. 23) claims that demand will always exist for increasingly more personal services by high-income earners or by the wealthy. The reason is that opportunities always exist “to make them feel better. Better about the world. Better about themselves. Better about what they have achieved.” Two hundred years ago, when most people lived close to the subsistence level, the modern labor structure—wherein the majority of the population is employed in the service sector in occupations such as care assistant, nursery school teacher, nutrition adviser, marketing expert, manicuris...

Table of contents

• Cover image
• Title page
• Table of Contents
• Copyright
• Dedication
• Preface
• Chapter 1. Introduction
• Chapter 2. The stylized facts
• Chapter 3. Empirical evidence on the economic effects of automation
• Chapter 4. A simple macroeconomic framework for analyzing automation
• Chapter 5. Endogenous savings and extensions of the baseline model
• Chapter 6. Automation as a potential response to the challenges of demographic change
• Chapter 7. Policy challenges
• Chapter 8. Peering into the future: long-run economic and social consequences of automation; with an epilogue on COVID-19
• Index