Is Taiwan the Next Hong Kong?
China Tests the Limits of Impunity
In the first decade of this century, the U.S. manufacturing sector shed jobs at an alarming and unprecedented rate. That coincided with a surge in imports, weak growth in exports, and a yawning trade deficit. The sharp job losses in manufacturing helped keep employment growth weak and labor force participation low.
During the 2016 presidential election, both Republican candidate Donald Trump and Senator Bernie Sanders blamed trade and globalization. That message helped propel Trump to the presidency. But most economists dismissed it, arguing that automation was to blame. Media reports often take this view as fact. In late 2016, for example, New York Times reporter Binyamin Appelbaum wrote, “From an economic perspective . . . there can be no revival of American manufacturing, because there has been no collapse. Because of automation, there are far fewer jobs in factories.”
Yet this view reflects a misreading of the data. As I discuss in a recent paper, although automation is occurring in manufacturing, as it is in other sectors of the economy, the evidence does not support the idea that automation was the main cause of the sudden decline in manufacturing employment after 2000. Although it’s difficult to precisely pin down what was to blame, a large body of research suggests that economists and pundits are wrong to so quickly dismiss trade and to blame machines.
Employment in U.S. manufacturing peaked at over 19 million in 1979, and although it fell by about seven percent in the 1980s it remained fairly stable throughout the 1990s. That changed after the turn of the century. Between the economic peaks of 2000 and 2007, manufacturing employment dropped by 20 percent, or 3.4 million people. It was hit hard again by the Great Recession of 2008–09 and rebounded only slightly during the recovery. In total, since 2000 manufacturing employment has fallen by over 28 percent, or by nearly five million jobs. There are some 22 percent fewer factories in the United States today than there were in 2000.
As employment in manufacturing has fallen, so has its share of the wider economy. Figure 1 shows manufacturing’s share of U.S. employment and GDP over the last 70 years. In 1953, manufacturing accounted for 35 percent of private sector employment. By 2016, that figure had fallen to under ten percent. Manufacturing’s share of private sector GDP has experienced a parallel decline: it peaked at 33 percent in 1953 and by 2016 was just 13 percent. These trends suggest that for decades manufacturing has performed less well than the rest of the economy.
Yet when adjusted for inflation, the apparent relative decline in manufacturing seems to disappear. Figure 2 shows inflation-adjusted output (known as real GDP) for manufacturing and private industry overall. By controlling for price changes, real GDP captures growth in the quantity of goods produced in factories and the wider economy. For decades, the two grew at roughly the same pace. Only since the Great Recession has manufacturing started growing significantly more slowly than the rest of the economy.
How can these apparently contradictory trends be reconciled? The prevailing narrative goes something like this. If real GDP growth in manufacturing has kept pace with that in the private sector, but manufacturing’s share of private sector GDP has fallen, then prices of manufactured goods must have grown much more slowly than inflation. Similarly, if real GDP growth in manufacturing is about the same as in the private sector overall, but employment growth is much slower (or even negative), manufacturing’s labor productivity must be rising considerably faster than the private sector average. Workers in factories are somehow producing much more stuff per person much more cheaply than workers everywhere else.
Many economists assume that this higher productivity growth reflects greater rates of automation. So they conclude that automation has caused most of the decline in manufacturing employment. Even those who do acknowledge some role for trade argue that the effect is small.
There are two big problems with this conclusion. First, the evidence is misleading. Manufacturing’s apparently low price growth, strong output growth, and big productivity rises are largely driven by one industry: computers and electronic products.
That’s because, since the 1980s, statistical agencies have tried to account for rapid improvements in computers, semiconductors, and other IT products by adjusting their prices downward. The computers and smartphones that consumers and businesses buy today are much more powerful than those they bought in the past. If, for example, this year’s model sells for the same dollar price as last year’s but has more features that consumers want, its true price has, in effect, dropped. So government agencies produce statistics showing that output, adjusted for price, has risen. Although this kind of adjustment for improvements in quality make sense, the result has been that official reports show extraordinarily high levels of growth in real output and productivity in the computer industry.
That extraordinary growth distorts the overall numbers. Although the computer industry accounts for less than 15 percent of the value created by manufacturing as a whole, it has had an outsized effect on the sector’s measured real growth numbers since the 1980s. That has given a misleading impression of how healthy U.S. manufacturing really is.
Recall that from 1979 to 2000, real GDP growth in manufacturing was about the same as the private sector average (Figure 2). But Figure 3 shows that when the computer industry is dropped from both sets of data, manufacturing’s real GDP growth rate is less than half that of the private sector. Between 2000 and 2016, manufacturing GDP grew 63 percent as fast as the private sector average. Omitting the computer industry, manufacturing grew just 12 percent as fast as the private sector over that period, and over the last decade has still not recovered from its losses in the Great Recession. In addition, without computers, labor productivity grew little faster in manufacturing than it did in the private sector overall. Once the anomalous effects of computer industry are excluded, the data offer no obvious evidence that automation has driven the decline in manufacturing employment.
The second big problem with the prevailing narrative that automation has cost manufacturing jobs is that data on productivity by themselves cannot shed light on the causes of manufacturing’s employment collapse. Productivity growth alone does not destroy jobs and it can be caused by many things besides automation. Within the computer industry, for example, strong productivity growth largely stems from product improvements driven by R & D, not from automating production.
Rises in productivity can be caused by globalization, as well. Manufacturers have outsourced many activities to domestic and foreign suppliers and many have shifted to buying parts from lower-cost providers, often foreign ones. As research I carried out with three other economists shows, shifting to lower-cost sources artificially raises measured productivity.
Increased competition from foreign companies can have the same effect. That’s because the factories and industries displaced by competition from low-wage countries are likely to be the most labor-intensive and least automated. When they close, average labor productivity in U.S. manufacturing as a whole rises, even if productivity in each individual factory and industry that remains hasn’t changed.
Manufacturing has declined so much—it now accounts for less than ten percent of U.S. employment—that it’s reasonable to ask if it still matters. It does. Because its supply chains run throughout the U.S. economy and because it accounts for a disproportionate share of R & D spending, manufacturing continues to have outsized effects on employment, output, and innovation in the wider economy. Economists and policymakers need to understand the causes of manufacturing’s decline if they are to develop sensible policy responses. That needs to rest on a realistic understanding of the data and the findings from rigorous research.
Studies have failed to find a clear connection between automation and job losses in manufacturing. Industrial robots may yet displace large numbers of workers, but so far they have had little effect on manufacturing employment.
In contrast, a large and growing body of research examining the effects of trade on domestic manufacturing since 2000 has found that trade did hurt manufacturing employment, output, and investment. Economists and politicians who deny the effects of globalization on U.S. manufacturing are standing in the way of a much-needed, better informed debate over trade policies.