Читать книгу Automation and the Future of Work - Aaron Benanav - Страница 10

CHAPTER 2

Labor’s Global Deindustrialization

IF TECHNOLOGICALLY INDUCED JOB destruction is to have widespread social ramifications, it will have to eliminate employment in the service sector, which has absorbed 74 percent of workers in high-income countries and 52 percent worldwide.¹ Purveyors of the automation discourse therefore focus on “new forms of service-sector automation” in retail, transportation, and food services, where “robotization” is said to be “gathering steam” with a growing army of machines that take orders, stock shelves, drive cars, and flip burgers. Many more service sector jobs, including some that require years of education and training, will supposedly be rendered obsolete in the coming years due to advances in artificial intelligence.² Of course, these claims are mostly predictions about the effects that technologies will have on future patterns of employment. Such predictions can go wrong—as, for example, in the first week of January 2020, when three espresso-and-burger-slinging robotics firms in the Bay Area either closed or were forced to cut their losses.³

In making their case, automation theorists often point to the manufacturing sector as the precedent for what they imagine is beginning to happen in services. In manufacturing, the employment apocalypse has already taken place.⁴ To evaluate these theorists’ claims, it therefore makes sense to begin by looking at what role automation has played in that sector’s fate. After all, manufacturing is the area most amenable to automation, since on the shop floor it is possible to “radically simplify the environment in which machines work, to enable autonomous operation.”⁵ Industrial robotics has been around for a long time: the first robot, the “Unimate,” was installed in a General Motors plant in 1961. Still, until the late 1960s, scholars studying this sector were able to dismiss out of hand Luddite fears of long-term technological unemployment. Manufacturing employment grew most rapidly precisely in those lines where technical innovation was happening at the fastest pace, because it was in those lines that prices fell the fastest, stoking the growth of demand for products.⁶ That era is long over. Over the past fifty years, industrialization has given way to deindustrialization, and not just in any one line, but across the manufacturing sectors of most countries.⁷

The Productivity Paradox

In the scholarly literature, deindustrialization is “most commonly defined as a decline in the share of manufacturing in total employment.”⁸ That share fell first of all across the high-income world, starting in the late 1960s and early 1970s. Manufacturing employed 22 percent of all workers in the United States in 1970, a share that declined to just 8 percent in 2017. Over the same period, manufacturing employment shares fell from 23 percent to 9 percent in France, and from 30 percent to 8 percent in the UK. Japan, Germany, and Italy experienced smaller but still-substantial declines: in Japan, from 25 percent to 15 percent; in Germany, from 29 percent to 17 percent; and in Italy, from 25 percent to 15 percent. In all cases, the declines were eventually associated with substantial falls in the total number of people employed in manufacturing. In the US, Germany, Italy, and Japan, the overall number of manufacturing jobs fell by approximately a third from postwar peaks; in France, it fell by 50 percent, and in the UK, by 67 percent.⁹

It is commonly assumed that deindustrialization in these high-income countries must be the result of production facilities moving offshore. Offshoring has certainly contributed to deindustrialization in the United States and UK, which boast the world’s largest trade deficits. Yet in none of the countries named above, including the Unites States and UK, has manufacturing job loss been associated with declines in absolute levels of manufacturing output. On the contrary, the volume of manufacturing production, as measured by real value added, more than doubled in the United States, France, Germany, Japan, and Italy between 1970 and 2017. Even the UK, whose manufacturing sector fared worst of all among this group, saw a 25 percent increase in manufacturing real value added over this period. To be sure, low- and middle-income countries are producing more and more goods for export to high-income countries; however, deindustrialization in the latter cannot simply be the result of productive capacity moving to the former, since the high-income countries produced more manufactured goods at the end of the 2010s than they had anytime in the past. In line with automation theorists’ core expectations, more goods are being produced but by fewer workers.

It is on this basis that commentators typically cite rapidly rising labor productivity, rather than an influx of low-cost imports from abroad, as the primary cause of industrial job loss in advanced economies.¹⁰ On closer inspection, however, this explanation also turns out to be inadequate. Manufacturing productivity has been growing at a sluggish pace for decades, leading economist Robert Solow to quip, “We see the computer age everywhere, except in the productivity statistics.”¹¹ Automation theorists discuss this “productivity paradox” as a problem for their account—explaining it in terms of weak demand for products, or the persistent availability of low-wage workers—but they understate its true significance. This is partly due to the appearance of steady labor-productivity growth in US manufacturing, at an average rate of around 3 percent per year since 1950. On that basis, Erik Brynjolfsson and Andrew McAfee suggest, automation could show up in the compounding effects of exponential growth, rather than an uptick in the growth rate.¹²

However, official US manufacturing growth-rate statistics are vastly overinflated, since they log the production of computers with higher processing speeds as equivalent to the production of more computers.¹³ For that reason, government statistics suggest that productivity levels in the computers and electronics subsector rose at a galloping average annual rate of over 10 percent per year between 1987 and 2011, even as productivity growth rates outside of that subsector fell to around 2 percent per year over the same period.¹⁴ Starting in 2011, trends across the manufacturing sector worsened: real output per person employed in the sector as a whole was lower in 2017 than in 2010. Productivity growth rates in manufacturing collapsed precisely when, according to automation theorists, they were supposed to be rising rapidly due to advancing technologies.

Correction of US manufacturing-productivity statistics brings them more into line with trends in countries like Germany and Japan, where manufacturing-productivity growth rates have fallen dramatically since their postwar peaks. In Germany, manufacturing productivity grew at an average annual rate of 6.3 percent per year in the 1950s and ’60s, falling to 2.4 percent from 2000 to 2017. This downward trend was to some extent an expected result of the end of an era of rapid catch-up growth. However, it should still be surprising to the automation theorists, since Germany and Japan have raced ahead of the United States in the field of industrial robotics. Indeed, the robots used in Tesla’s largely automated car factory in California were made by a German robotics company.¹⁵ As of 2016, German and Japanese firms deployed about 60 percent more industrial robots per manufacturing worker, compared to the US.¹⁶

Yet deindustrialization has continued to take place in all these countries, despite lackluster manufacturing-productivity growth rates; that is, it has taken place as the automation theorists expect, but not for the reasons they offer. To explore the causes of deindustrialization in more detail, I rely on the following definitions. Output, as used both above and below, is a measure of the volume of production (how much is produced), in terms of real or inflation-adjusted “value added” in a given economic sector.¹⁷ Gross domestic product, or GDP, is just value added for the economy as a whole. Employment, as I use it here, is a measure of the number of workers rather than of hours worked—the latter are typically unavailable outside of wealthier countries—while productivity is the ratio of output to employment: the more output is produced per worker, the higher that worker’s productivity level. For any economic sector, the rate of growth of output (ΔO) minus the rate of growth of labor productivity (ΔP) equals the rate of growth of employment (ΔE). Thus, ΔO – ΔP = ΔE.¹⁸ This equation is true by definition. If the output of automobiles grows by 3 percent per year, and productivity in the automotive industry grows by 2 percent per year, then employment in that industry must have risen by one percent per year (3 – 2 = 1). Contrariwise, if output grows by 3 percent per year and productivity grows by 4 percent per year, employment will have contracted by 1 percent per year (3 – 4 = –1).

Disaggregation of manufacturing-output growth rates in France provides us with a sense of the typical pattern playing out across the high-income countries (Figure 2.1).¹⁹ During the so-called golden age of postwar capitalism, productivity growth rates in French manufacturing were much higher than they are today—5.2 percent per year, on average, between 1950 and 1973—but output growth rates were even higher than that—5.9 percent per year. As a result, employment had to have grown steadily, at a pace of 0.7 percent per year. Since 1973, both output and productivity growth rates have declined, but output growth rates fell much more sharply than productivity growth rates. By the early years of the twenty-first century, productivity was rising at a much less rapid pace than it had during the postwar era, at 2.7 percent per year. However, slower productivity growth rates were now faster than their corresponding industrial output growth rates, at 0.9 percent. The result was that manufacturing employment contracted rapidly, by 1.7 percent per year. Even before that contraction got going, deindustrialization had already technically begun: as soon as the rate of growth of manufacturing employment consistently fell below the rate of growth of the total workforce, the manufacturing employment share started its downward trend.

Figure 2.1. French Manufacturing Sector, 1950–2017

Source: Conference Board, International Comparisons of Productivity and Unit Labour Costs, July 2018 edition.

This disaggregation helps explain why automation theorists falsely perceive productivity to be growing at a rapid pace in manufacturing. Productivity growth rates have been high relative to output growth rates, but not because productivity has been growing more rapidly than before—which would be a sure sign of accelerating automation. On the contrary, the key to this trend is that output has been growing much more slowly than before. The same pattern can be seen in the statistics of other countries: no absolute decline in levels of manufacturing production took place—more and more was produced—but the rate at which output grows declined, so output growth came to be consistently slower than productivity growth (Table 2.1). As industrial output growth rates fell below corresponding productivity growth rates in country after country, quantitative declines in economic indicators became qualitative in their effects: manufacturing employment shares fell progressively. Worsening economic stagnation thus combined with a limited technological dynamism to generate labor’s global deindustrialization.

Table 2.1. Manufacturing Growth Rates, 1950–2017

		Output	Productivity	Employment
USA	1950–73	4.4%	3.1%	1.2%
1974–2000	3.1%	3.3%	-0.2%
2001–17	1.2%	3.2%	-1.8%
Germany	1950–73	7.6%	5.7%	1.8%
1974–2000	1.3%	2.5%	-1.1%
2001–17	2.0%	2.2%	-0.2%
Japan	1950–73	14.9%	10.1%	4.3%
1974–2000	2.8%	3.4%	-0.6%
2001–17	1.7%	2.7%	-1.1%

Source: Conference Board, International Comparisons of Productivity and Unit Labour Costs, July 2018 edition.

Such “output-led” deindustrialization is impossible to explain in purely technological terms.²⁰ In their search for alternative perspectives, economists have mostly preferred to describe this trend as a harmless evolutionary feature of advanced economies.²¹ However, that perspective is itself at a loss to explain extreme variations in the GDP per capita levels at which this supposedly evolutionary economic shift has taken place. Deindustrialization unfolded first in high-income countries in the late 1960s and early 1970s, at the tail end of a period in which levels of income per person had converged across the United States, Europe, and Japan. In the decades that followed, deindustrialization then spread “prematurely” to middle-and low-income countries, with larger variations in incomes per capita (Figure 2.2).²² In the late 1970s, deindustrialization arrived in southern Europe; much of Latin America, parts of East and Southeast Asia, and southern Africa followed in the 1980s and ’90s. Peak industrialization levels in many poorer countries were so low that it may be more accurate to say that they never industrialized in the first place.²³

Figure 2.2. Global Waves of Deindustrialization, 1950–2010

Source: Groningen Growth and Development Centre, 10-Sector Database, January 2015 edition.

By the end of the twentieth century, it was possible to speak of a global wave of deindustrialization: worldwide manufacturing employment rose in absolute terms by 0.4 percent per year between 1991 and 2016, but that was much slower than the overall growth of the global labor force, with the result that the manufacturing share of total employment declined by 3 percentage points over the same period.²⁴ China is a key exception, but only a partial one (Figure 2.3). In the mid 1990s, Chinese state-owned enterprises shed millions of workers, sending manufacturing-employment shares on a steady downward trajectory.²⁵ China reindustrialized, in employment terms, starting in the early 2000s, but then it began to deindustrialize once again in the mid 2010s. Its manufacturing-employment share has since dropped significantly, from 19.3 percent in 2013 to 17.2 percent in 2018. If deindustrialization cannot be explained by either automation or the internal evolution of advanced economies, what could be its source?

Figure 2.3. Deindustrialization in China, India and Mexico, 1980–2017

Source: Conference Board, International Comparisons of Productivity and Unit Labour Costs, July 2018 edition.

Blight of Manufacturing Overcapacity

What the economists’ accounts fail to register in their explanations of deindustrialization is also what is missing from the automation theorists’ accounts. The truth is that rates of output growth in manufacturing have tended to decline, not only in this or that country, but worldwide (Figure 2.4).²⁶ In the 1950s and ’60s, global-manufacturing production expanded at an average annual rate of 7.1 percent per year, in real terms. That rate fell progressively to 4.8 percent in the 1970s and to 3.0 percent between 1980 and 2007. From the 2008 crisis up to 2014, manufacturing output expanded at just 1.6 percent per year, on a world scale—that is, at less than a quarter of the pace achieved during the post-war “golden age.”²⁷ It is worth noting that these figures include the dramatic expansion of manufacturing productive capacity in China.

Figure 2.4. World Manufacturing and Agricultural Production, 1950–2014

Source: World Trade Organization, International Trade Statistics 2015, Table A1a, World Merchandise Exports, Production and GDP, 1950–2014.

Again, it is the incredible degree of slowdown in the rate at which manufacturing production expands, visible on the world scale, that explains why manufacturing-productivity growth appears to have advanced at a rapid clip, even though it was actually much slower than in previous eras. More and more is produced with fewer workers, as the automation theorists claim, but not because technological change has given rise to high rates of productivity growth. Far from it—productivity growth in manufacturing has appeared rapid only because the yardstick of output growth, against which it is measured, has been shrinking.

Following economist Robert Brenner, I argue that global waves of deindustrialization find their origins not in runaway technical change, but first and foremost in a worsening overcapacity in world markets for manufactured goods.²⁸ The rise in overcapacity developed stepwise after World War II. In the immediate postwar period, the United States hosted the most dynamic economy in the world, with the most advanced technologies: in 1950, output per hour worked in the US economy was more than twice as high as output per hour in European countries.²⁹ Under the threat of Communist expansion within Europe, as well as in East and Southeast Asia, the US proved willing to share its technological largesse with its former imperial competitors Germany and Japan, as well as with other “frontline” countries, in order to bring them all under the US security umbrella.³⁰ In the first few decades of the post–World War II era, these technology transfers were a major boost to economic growth in European countries and Japan, opening up opportunities for rapid export-led expansion. This strategy was supported by the devaluation of their currencies against the dollar in 1949, which improved these countries’ international competitiveness at the expense of domestic, working-class buying power (a move that in many European countries led to the eviction of left political parties from government).³¹ However, as Brenner has argued, rising manufacturing capacity across the globe quickly generated overcapacity, issuing in a “long downturn” in manufacturing-output growth rates.

What mattered here was not only the later build-out of manufacturing capacity in the global South, but the earlier creation of such capacity in countries like Germany, France, Italy, and Japan. These countries hosted the first low-cost producers in the postwar era to succeed, first, in taking shares in global markets for industrial goods and, second, in invading the previously impenetrable US domestic market. Due to rising competition with lower-cost producers, rates of industrial output growth in the US began to decline starting in the late 1960s, issuing in deindustrialization in employment terms. As the US responded to heightened import penetration in the early 1970s by breaking up the Bretton Woods order and devaluing the dollar—which increased US firms’ international competitiveness—these same problems spread from North America and northwestern Europe to the rest of the European continent and Japan.³²

Intensifying competition among firms in these high-income regions did not dissuade more countries from building up manufacturing capacity, adopting export-led growth strategies, and entering global markets for manufactured goods. As additional manufacturing capacity appeared and entered the fray of international competition, falling rates of manufacturing-output growth and consequent labor deindustrialization spread to more regions: Latin America, the Middle East, Asia, and Africa, as well as to the global economy taken as a whole. Deindustrialization came to most global South regions in the aftermath of the 1982 Third World debt crisis, amid the imposition of IMF-led structural adjustment programs. As trade liberalization opened the borders of poorer countries to imports, while financial liberalization brought hot money flowing into “emerging markets,” their currencies revalued sharply. Unit labor costs in these regions rose just as markets were becoming more overcrowded, with the result that firms found themselves able neither to compete with imports nor to export their wares abroad.³³

Deindustrialization was a matter not only of technological advance, but also of global redundancy of productive and technological capacities. In more crowded international markets, rapid rates of industrial expansion became more difficult to achieve.³⁴ The mechanism transmitting this problem across the world was depressed prices in global markets for manufactured goods (which also explains why shifting currency valuations played such a major role in determining competitiveness).³⁵ As Harvard economist Dani Rodrik notes, “Developing countries ‘imported’ deindustrialization from the advanced countries” because they were “exposed to the relative price trends” coming from the capitalist core.³⁶

Everywhere, depressed prices for manufactures led to falling income-per-unit capital ratios (falling capital productivity), then to falling rates of profit, then to lower rates of investment, and finally to lower output growth rates.³⁷ In this environment, firms faced heightened competition for market share: as overall growth rates slowed, the only way for new firms to grow quickly was to steal market shares from established firms. The latter responded by retreating to the apex of global value chains. Overcapacity explains why, from the early 1970s, productivity growth rates fell less severely than output growth rates. Firms either raised their productivity levels as best they could—in an effort to keep up with their competitors despite the slower growth of the demand for their products—or else went under, disappearing from statistical averages.³⁸ The implementation of technological innovations, although occurring at a slower pace than before, generated sector-wide job loss.³⁹ As output growth rates fell toward (and in many cases below) productivity growth rates, in one country after another, deindustrialization spread worldwide.

Explaining global waves of deindustrialization in terms of global overcapacity rather than industrial automation allows us to understand a number of features of this phenomenon that otherwise appear paradoxical. For example, rising overcapacity explains why deindustrialization has been accompanied not only by ongoing efforts to develop new labor-saving technologies, but also by the build-out of gigantic, labor-intensive supply chains—usually with a more damaging environmental impact.⁴⁰ A key turning point in that story came in the 1960s, when low-cost Japanese and German products invaded the US domestic market, sending the US industrial import penetration ratio soaring from less than 7 percent in the mid ’60s to 16 percent in the early ’70s.⁴¹ From that point forward, it became clear that high levels of labor productivity would no longer serve as a shield against competition from lower-wage countries. The firms that did best in this context were the ones that responded by globalizing production. Facing competition on prices, US multinational corporations (MNCs) built international supply chains, shifting the more labor-intensive components of their production processes abroad and playing suppliers against one another to achieve the best prices.⁴² In the mid ’60s the first export-processing zones opened in Taiwan and South Korea. Even Silicon Valley, which formerly produced its computer chips locally in the San Jose area, shifted its production to low-wage areas, using lower grades of technology while benefiting from laxer laws around pollution and workers’ safety.⁴³ MNCs in Germany and Japan adopted similar strategies, which were everywhere supported by new transportation and communication infrastructures.⁴⁴ The globalization of production allowed the world’s wealthiest economies to retain manufacturing capacity, but it did not reverse the overall trend toward labor deindustrialization. As supply chains were built out across the world, firms in more and more countries were pulled into the swirl of world market competition. In some countries, this move was accompanied by shifts in the location of new plants: rust belts, oriented toward production for domestic markets, went into decline; sun belts, integrated into global supply networks, expanded dramatically. Chattanooga grew at the expense of Detroit, Juárez at the expense of Mexico City, Guangdong at the expense of Dongbei.⁴⁵ Yet given the overall slowdown in rates of world market expansion, this reorientation toward the world market resulted in lackluster outcomes: the rise of sun belts failed to balance out the decline of rust belts, resulting in global deindustrialization.

At the same time, global manufacturing overcapacity explains why the countries that have succeeded in attaining a high degree of robotization are not those that have seen the worst degree of deindustrialization. Measured in terms of robots deployed per thousand workers in manufacturing, South Korea (63), Germany (31), and Japan (30) had advanced much further along the road to full automation, as compared to the United States (19) and UK (7), in 2016. Yet manufacturing employment shares in that same year were significantly higher in South Korea (17 percent), Germany (17 percent), and Japan (15 percent) than in the US (8 percent) and UK (8 percent). In the context of intense global competition, high degrees of robotization translate into international competitive advantages, helping firms win larger shares of world markets for the goods they produce. Unlike workers in the United States, workers in European and East Asian firms believe that automation helps preserve their jobs.⁴⁶ Chinese firms have also been major players in global markets for manufactured goods, providing China’s industrial sector with a gigantic boost in terms of both output growth and employment growth, yet Chinese firms advanced on this front not due to high levels of robotization—in 2016, China deployed just 7 robots per thousand workers in manufacturing—but rather due to a mix of low wages, moderate to advanced technologies, and strong infrastructural capacities. Still, the result was the same: in spite of system-wide over-capacity and slow growth rates, China has industrialized rapidly because its firms have been able to take market share away from other firms—not only in the United States, but also in countries like Mexico and Brazil. It could not have been otherwise. In an environment where average growth rates are low, firms can only achieve high rates of growth by taking market share from their competitors. Whether China will be able to retain its competitive position as its wage levels rise remains an open question; Chinese firms have been robotizing to try to head off this possibility.⁴⁷