As the Trump administration pushes ahead with its effort to renegotiate NAFTA, we must never miss an opportunity to remind people that the globalization of US economic activity has, by design, shifted the balance of class power away from working people. A commonly cited indicator of class power is labor’s share of output (or income), which, as shown below, dramatically fell after the turn of the 21st century after decades of slow decline.
Michael W. L. Elsby, Bart Hobijn, and Aysegül Sahin, writing in the Fall 2013 Brookings Papers on Economic Activity, tested several hypotheses about the cause of labor’s declining share of output. They concluded, based on their econometric work, that “increases in the import exposure of U.S. businesses” was key, accounting for approximately 85 percent of the decline in the U.S. payroll share over the period 1987 to 2011. This finding led them to suggest “that a particularly fruitful avenue for future research will be to delve further into the causal channels that underlie this statistical relationship, in particular the possibility that the decline in the U.S. labor share was driven by the offshoring of the labor-intensive component of the U.S. supply chain.”
Labor’s share of income
It is important to be clear about how the labor share is estimated and how well it captures class dynamics. The starting point is simple: labor’s share of output is calculated by dividing the labor compensation earned during a given period by the economic output produced over the same period. Things quickly get more complicated, however, because the labor compensation used in the calculation is actually the sum of the labor earnings of two different groups of workers: those who work for others and those who work for themselves.
The compensation of the first group includes the sum of all employee pay and benefits: wages and salaries; commissions; tips; bonuses; severance payments; early retirement buyout payments; exercised stock options; and employer contributions to employee pension and insurance funds, and to government social insurance. Calculating the employee share of output, known as the payroll share, is relative straightforward thanks to employer fillings.
Things are not so simple when it comes to the second group, since their earnings reflect “both returns to their work effort and returns to the business property they invested in” and there is no simple way to separate their earnings into those two components. The Bureau of Labor Statistics (BLS) handles this problem by assuming that the self-employed receive an hourly labor compensation similar to that earned by employees who work in the same sector of the economy.
The figure below, from the Brookings Papers article, shows the division of the labor share into its two component parts, the payroll share and the self-employed share. As we can see, the payroll share is significantly greater than the self-employed share. In fact, the share of hours of the self-employed in total work hours “has declined steadily from about 14 percent in 1948 to 8.5 percent in 2012.” However, as Elsby, Hobijn, and Sahin point out, “In spite of the relatively small share of self-employment hours, the treatment of self-employment income plays an important role in the recent behavior of the evolution of the labor share.”
A number of economists have raised concerns about the methodology used by the BLS to divide the compensation of the self-employed into its labor and capital returns components. One example: the BLS methodology ends up crediting the self-employed with more labor compensation than their total reported earnings for much of the 1980s and early 1990s, a highly unlikely outcome.
Alternative methodologies have been suggested, and the authors of the Brookings Papers article calculate labor’s share using the two most often cited. The one they call the “asset basis” assumes that the return on self-employed capital is the same as the return on capital in the non-farm business sector, with the remaining earnings credited to labor. The other, called the “economy-wide basis,” assumes that the division between labor compensation and capital income is the same for the self-employed as it is for the non-farm business sector. As we see below, the two alternatives generally produce labor share trends that are relatively close together, and significantly lower than that published by the Bureau of Labor Statistics from the start of the series until the late 1990s, when all three series generally converge.
Because of its methodological shortcoming, Elsby, Hobijn, and Sahin prefer either of the two alternative measures, which leads them to the conclusion that use of the BLS series overstates the actual decline in the labor share. As they explain:
The upshot of these comparisons is that around one third of the decline in the headline measure of labor’s share appears to be a by-product of the methods employed by the BLS to impute the labor income of the self-employed. Alternative measures that have less extreme implications regarding the return to capital among proprietors are more consistent with one another and indicate a more modest decline.
The fact that the difference between the BLS and the alternative measures of labor’s share largely disappeared beginning in the late 1990s suggests that the average hourly earnings of the self-employed have grown much faster than that of the employed. This, in turn, suggests a significant transformation in the make-up of the self-employed; in particular an increase in the number of individuals engaged in highly lucrative professional work. In this regard it is important to recall that labor compensation includes not just wage and salary earnings but also things like bonuses and stock options, rewards that became increasingly popular for a select few starting in the late 1990s thanks to the run-up in the stock market.
And in fact, this transformation is confirmed by the authors, who disaggregated the structure of the labor share for employees and total earnings for the self-employed. The results are illustrated in the following figure, which shows that “the share of income accounted for by both payroll wages and salaries and by proprietors’ income [the sum of their labor and nonlabor earnings] has been buoyed up since the 1980s by substantial rises in the shares accounted for by the very top fractiles of households in the United States.”
As the authors point out:
This rise in inequality is even more striking for proprietors’ income than it is for payroll income. In 1948 the bottom 90 percent of employees earned 75 percent of payroll compensation. By 2010 this had declined to 54 percent. For entrepreneurial income, however, this fraction declined from 42 percent in 1948 to 14 percent in 2010. Even more starkly, over the same period the share of proprietors’ income accounted for by the bottom 99 percent fell from 74 percent to 45 percent. This suggests that the sharp rise in the average hourly compensation of proprietors relative to the payroll-employed since the late 1980s is related to substantial increases in income inequality among proprietors that dominate even the considerable rise in inequality witnessed among the payroll-employed. Moreover, this has been driven by extreme rises in proprietors’ income at the very top of the income distribution—the top 1 percent in particular.
In short, there are a lot of moving parts to the calculation of and evaluation of trends in the labor share of income. The BLS measure may have overstated the decline, but the explosion of inequality means that the measure’s two components mask an even greater fall in the share of income going to the great majority of working people.
Globalization and the decline in the payroll share of output
Although the labor share is the “headline” statistic, the authors decided to narrow their focus to the payroll share. As we saw above, it is no simple matter to determine the labor compensation of the self-employed. In contrast, the payroll share is relatively easy to measure and, as a bonus, can be disaggregated by industry. Moreover, it is the largest component of the labor share, which means that its movement is most responsible for changes in the overall labor share.
Elsby, Hobijn, and Sahin begin with a standard neoclassical aggregate production model and the most common neoclassical explanations for the decline, which rest on investment and technological change: the growth in the capital/labor ratio and skill-biased technical change. The basic neoclassical argument is that growing investment shifts income away from labor in the first case and unskilled workers in the second. However, in both cases the authors found that the movement in relevant variables was not consistent with the actual movement in the payroll share.
Recognizing the limitations inherent in a simple aggregate production function model of the economy, the authors decided to take advantage of their industry data to see whether a more micro/industry perspective yielded better results. More specifically, they econometrically tested whether investment specific technological change, declines in unionization, or increases in import competition can explain the decline in the payroll share. They found that “Our data yield one robust correlation: that declines in payroll shares are more severe in industries that face larger increases in competitive pressures from imports.”
In the case of investment specific technical change, the authors looked to see whether those industries which enjoyed the lowest price increases for investment goods had the largest declines in payroll share, with the assumption being that these industries would be the most likely to replace workers with capital. In fact, it turned out that there was a weak negative relationship between the change in equipment prices and the change in payroll shares across industries, the opposite of what was expected “if capital deepening due to the decline in price of equipment were the driving force of the decline in the payroll share.” This result reinforced the conclusion from their aggregate analysis that investment activity does not explain the decline in the payroll share of output.
The test of unionization was more straight forward. The authors looked to see if there was a positive relationship between changes in union density in an industry and changes in payroll shares. While they did find “a positive correlation between the change in unionization and the change in payroll shares across industries,” the relationship was weak. “The weighted least squares regression indicates that cross-industry variation in changes in unionization rates explains less than 5 percent of the variation in changes in payroll shares across industries.”
Last was the test of globalization, or more specifically a test of whether the import-caused hollowing out of US industry was a primary cause of the decline in the payroll share. Elsby, Hobijn, and Sahin assumed two possible channels for a rise in imports to cause a fall in the payroll share. The first involved trade-generated capital deepening. In this case, the outsourcing of production by US firms would lead to a reduction in labor, a rise in the capital-labor ratio, and a decline in the payroll share of income. However, as the authors noted, they had already tested capital deepening as a potential cause of the decline and found no support for the hypothesis.
The second trade channel relied on wage differentials rather than shifts in capital intensity. Industries with high labor shares likely have high labor costs, making them vulnerable to import competition. The greater the competition the more likely firms in these industries were to take actions to lower those costs, including offshoring segments of their production process, thereby producing a decline in their payroll share.
The authors pursued this possibility by computing the import exposure of each industry. They did so by asking the following question:
If the United States were to produce domestically all the goods that it imports, how much additional value added would each industry have to produce? For example, if all U.S. imports of clothes were produced domestically, how much would value added increase in sectors like retail, textile manufacturing, and so on.
To be able to calculate this measure of import exposure we use the annual input-output matrices that are available for the years 1993 to 2010 from the BLS. Import exposure is expressed as the percentage increase in value added needed to satisfy U.S. final demand if the United States would produce all its imports domestically.
The figure below shows the relationship between changes in import exposure and changes in the payroll share for each industry. As we can see, import exposure increased for almost all industries—reflecting the growing hollowing out of the US economy–and the larger the exposure the greater the decline in payroll share. A simple regression showed that the import exposure variable was significant in explaining changes in the payroll share, with cross-industry variation in changes in import exposure explaining 22 percent of the variation in changes in payroll share.
The authors then ran a regression which included all three possible explanations for the decline in the payroll share. The globalization variable remained highly significant and was the only variable to do so. With the import exposure valuable included in the regression, the unionization variable became insignificant. “This suggests that those sectors where deunionization was most prevalent are also sectors that saw the biggest increase in import exposure.”
Elsby, Hobijn, and Sahin conclude:
our results indicate a cross industry link between the increases in import exposure and the decline in the labor share. While this result cannot be interpreted as causal, it is worth noting that the statistical relationship between import exposure and payroll shares across industries is large enough to account for a substantial fraction of the aggregate trend decline in the labor share. In particular, aggregating the results of the weighted-least-squares regression across industries suggests that increases in the import exposure of U.S. businesses can account for 3.3 percentage points of the 3.9 percentage point decline in the U.S. payroll share over the past quarter century.
We know that trade agreements are about a lot more than lowering tariffs to promote trade. Foremost, they are about strengthening corporate power and profitability. And despite mainstream economic theorizing to the contrary, there is strong evidence that these corporate gains come, as designed, at the expense of majority well-being.
Studies of the effect on US workers from imports from China (see Autor, Dorn, and Hanson) and Mexico (see Hakobyan and McLaren), most of which are produced within US transnational corporate-controlled production networks, show that US workers pay a steep price in terms of job loss and lost earnings from corporate driven globalization. And, as we have seen, Elsby, Hobijn, and Sahin’s work strongly suggests that this process is also the main factor behind the decline in the payroll share of output. This is class power at work–unfortunately theirs, not ours.