A new paper, Firming up Inequality, has been receiving substantial attention in the media for its claim that wage inequality is not occurring within firms but only occurs between firms. The authors claim that their results disprove the claim made by me and others, such as Thomas Piketty and Emmanuel Saez, that the growth of top 1 percent incomes was driven by the pay of executives and those in the financial sector. Though the authors present valuable new data, which offers the possibility of great insights, their current analysis does not disprove that executive pay has fueled top 1 percent income growth. In fact, the study neither examines nor rebuts claims about executive pay.

The authors also offer a “we live in the best possible world” interpretation of their findings—inequality is due to high productivity growth of “superfirms.” This is pure speculation and is completely disconnected from their actual empirical work. A similar study examined productivity trends and contradicts their narrative about superfirms.

Last, there are reasons to be skeptical of their findings because they imply huge wage disparities have opened up between median workers across firms within an industry that are implausible.

1. The paper neither examines nor rebuts the claim that executive pay was a major factor in the doubling of the income share of the top 1 percent.

The paper characterizes itself as a critique of the findings that executive pay (and financial sector pay) has fueled the growth of top 1 percent incomes, citing my work with Natalie Sabadish, as well as Piketty’s:

In the absence of comprehensive evidence on wages paid by firms, it is frequently asserted that inequality within the firm is a driving force leading to an increase in overall inequality. For example, according to Mishel and Sabadish (2014), “a key driver of wage inequality is the growth of chief executive officer earnings and compensation.” Piketty (2013) (p. 315) agrees, noting that “the primary reason for increased income inequality in recent decades is the rise of the supermanager.”

Look directly at the top 1 percent

The bottom line is that this paper does not actually challenge the basis for these claims even though the authors have the data that could contribute to such an analysis. Both Piketty and my claim rest on the Bakija et al. (2012) analysis of tax returns that show that executives account for roughly 40 percent of the growth of top 1.0 and top 0.1 percent household incomes. Financial sector employees account for another 23 percent of the growth. Bakija et al.’s data show the top 1 percent’s share of income rising from 9.7 percent in 1979 to 21.0 percent in 2005: executives and financial sector workers account for 6.5 of the 11.2 percentage point rise in the income share of the top 1 percent and 4.7 of the 7.0 rise in the top 0.1 percent’s share of income. Josh Bivens and I argue that the higher incomes of executives and the finance sector do not reflect an increased productive contribution (see our paper for the American Economic Association’s Journal of Economic Perspectives’ forum on the top 1 percent).

One need only compute the amount of executive pay in the top 1 percent now and in the past to assess whether executive pay fueled top wage growth. The authors of Firming up Inequality could have done so for all wage income as they have data on firms and the workers in those firms. For instance, they could identify the wages, including realized stock options, of the top five earners in every firm and then compute their share of top 1 percent wages now and in an earlier year. They could break out their sample by size of firm, profiling the very largest firms who employ a large share of the workforce. They have data on all firms, publicly traded and privately held, and this would be a very useful contribution. Absent such an analysis, their evidence does not overturn our findings.

Look at the largest firms

Nothing in Firming up Inequality focuses on the compensation of CEOs or executives in the very largest firms, which has been the focus of much research and policy discussion. The authors do provide a breakdown by firm size but the group with the largest firms provides no real insights since it includes all firms with at least one hundred employees: this category, according to Census reports, covers roughly two-thirds of all employment (the median worker in their sample works at a firm of roughly a thousand employees).

It is well established that the pay of executives in the largest firms grew tremendously over the last few decades. Nothing in Firming up Inequality challenges or even examines these trends, which are not a matter of dispute among economists across the political spectrum. (The issue is why this occurred and what if anything to do about escalating CEO pay.) In our research (see methodology) we examine the CEO pay of the 350 largest firms (by revenue) and show that the average compensation ($2013) grew from $1.5 million in 1978 to $5.4 million in 1993 and then to $18.5 million in 2007 but fell in the financial crisis so our latest measurement, for 2013, was $15.2 million. CEO pay in these large firms grew 937 percent from 1978 to 2013, while the compensation of a typical worker grew by 10 percent (see the table below).

CEO-to-worker pay ratios

Given the escalation of CEO pay and the slow growth of a typical worker’s pay, one would suspect a widening of the pay gap between CEOs and the workers in their firms. There are no publicly available data which provides CEO and worker pay data within a particular firm. Consequently, our research identifies the CEO-to-worker pay ratio by comparing a CEO’s pay to the pay of the ‘typical worker’ (the average production/nonsupervisory worker) in the CEO’s firm’s main industry. We found that the CEO-to-worker pay ratio across large firms rose from 111.8 in 1993 to 295.9 in 2013.

Firming up Inequality challenges whether these observed trends reflect widening pay disparities within individual firms—between the CEOs and workers in the same firm. The authors suggest that the inequality is occurring across firms but not within firms. No data are presented which compare the wages of the top earner or earners in a firm to wages earned by the median worker in the same firm, for any size firm let alone the largest firms. There are some comparisons of the wages of the top 1percent of earners to the average pay of workers in the same firm. That’s not the same as a comparison to the median or typical worker. There was certainly no analysis of CEO-to-median worker pay in very large firms.

Some skepticism: do the pieces really fit together?

I am skeptical of the authors’ claim that there was not a large rise in the gap between pay of the median worker and the CEOs in large firms. We know that CEO pay grew tremendously in these large firms. I would like to see evidence that median pay in these firms grew correspondingly. Given that a typical worker’s pay in that same industry grew far less than that of the CEOs (this is what our results show) their claim can only be true if the median wage in the firms where CEOs pay escalated grew far faster than the median wage in other firms in the same industry. To believe their claim is to believe that a tremendous pay gap (comparable to what we show between CEOs and workers!) has grown between the median worker in large firms and the median workers in other firms in the same industry. I find this implausible. If it were true it could only have resulted from a huge change in the character of the median worker, such that the median worker used to be a factory operative but now is a software engineer.

One of the authors, Fatih Guvenen, reports in a slide presentation that, “The pay of workers in the top 0.01 percent increased by 500 percent from 1982 to 2012. The pay gap between these top earners and the average employee at the same firm has increased by only 20 percent during the same time (p. 18).” Guvenen is reporting log wage growth (wonky note: wage earnings are capped at the 99.999th percentile, so top earner growth is actually understated). If the top 0.01percent earners had 500 percent wage growth—and the average worker fell only 20 percent behind the top 0.01 percent over the same period—then the other 99.99 percent of workers had average log wage growth of 480 percent. But, in the workforce overall, wage growth for the bottom 99 percent was far less. Between 1982 and 2012, for example, the bottom 90 percent of all workers saw a cumulative increase in wages of only 20 log-points. Even the 90th-99th percentile workers experienced real wage growth of only 41 log-points. It is painfully obvious that implicit in Guvenen’s calculations is an enormous gap in the wage growth of regular workers in superfirms relative to regular workers in other firms. This is especially true given that the data I presented for the overall workforce averages include workers in and out of superfirms.

My skepticism could be answered by seeing computations showing wage data for the firms with the faster and the slower wage growth: median wage growth; the ratio of top earnings to median earnings; and, especially, computations for very large firms.

2. Living in the best of all possible worlds: the rise of superfirms

The paper provides no evidence to explain why inequality across firms has risen. In press reports, however, they speculate that it is the rise of superfirms with higher productivity. See Bloomberg:

“There’s this view out there that the main reason inequality is rising is because of super managers,” Fatih Guvenen, a University of Minnesota economist who is among the paper’s four co-authors, said in an interview. “We’re arguing that it’s the rise of super firms.” It’s unclear what makes some companies stand-outs while others fall behind, Guvenen said. The researchers hypothesize that diverging productivity could be among the reasons. Highly skilled workers might also be concentrating at certain businesses.

It needs to be made clear that any speculation about superfirms and their productivity is not based on any evidence presented in the paper. It is simply the default view of those who consider that markets automatically provide optimal outcomes. This claim that the fast wage growth in superfirms corresponds to their superlative productivity growth is contradicted by a very similar study by Erling Barth, Alex Bryson, James C. Davis, and Richard Freeman released in 2014 which examined earnings inequality across establishments (not firms).

Barth et al. examined a proxy for productivity, revenue per worker, and found:

“With an impact on earnings of 0.163 the increased revenue per worker adds about 5-6 percent to the variance of earnings among establishments and thus falls far short of explaining the increased variance of establishment effects and increased inequality of individual earnings. Factors beyond demand-driven rent-sharing would seem to be needed to account for the divergence of establishments in earnings space.” (p. 21)

They concluded:

“Our results suggest the value of renewed analysis of establishment pay setting and hiring policies on the demand side and on establishment-level mobility on the supply side, and on factors beyond establishment demand shocks, such as productivity shocks associated with the introduction of innovative products or processes, in producing the divergence of establishment earnings. The huge role of establishment factors in the trend rise in inequality documented in this study is a signpost to pay attention to the places where people work as well as to their skills in studies of inequality.” (underlining added)

So, the speculation is unwarranted. There are many reasons why the divergence in earnings across firms took place, including the “fissuring” of the workplace that David Weil (of Boston University) documented—firms establishing a core scope of work and contracting out all other functions, where the subcontractors end up paying less and bearing the risks. There’s offshoring and other explanations. There is no basis for jumping to some conclusion that we live in the best of all possible worlds. The speculation by the authors of Firming up Inequality about highly productive superfirms is not empirically grounded and, in fact, remains totally unexamined by them using their data.