The Great Reversal Read online

  Persistence of Market Shares

  Economists who specialize in industrial organization and antitrust rightfully complain about the use of industry-level HHIs to measure concentration. Indeed, we have pointed out the limitations of HHI in Chapter 2 and discussed cases where they can lead to misleading conclusions about the state of an industry. On the other hand, it makes no sense to completely ignore industry-level market shares because there are many cases in which they do provide useful information.

  Matias Covarrubias, Germán Gutiérrez, and I consider an alternative, more dynamic measure of competition: instead of looking at the concentration of market shares at a point in time, we look at the persistence of market shares over time.

  Our intuition is that in a competitive industry, the leaders should be challenged. To make the point, imagine an industry with five firms. In any given year, one firm dominates and has a market share of 60 percent, while the other four firms have only 10 percent each. This looks like a concentrated industry. It has an HHI of 4,000, clearly above the “highly concentrated” threshold of 2,500. But now imagine that every two years or so, the leader is replaced by one of the followers. This might be because these five firms are constantly trying to innovate and outsmart each other, and one succeeds on average every two years. This dynamic turnover would radically change the picture. We would say that this industry is in fact quite competitive, because the dominance of the leader is temporary. Its large market share is transient. There is turnover at the top.

  Gutiérrez and I computed two measures to get at this idea: one measure captures turnover at the top, and the other measure captures the reshuffling of market shares. Our first measure works as follows: given that a firm is at the top of its industry now—among the top four by profits or by market value—how likely is it that it will drop out over the next three years? Figure 3.2 shows the results of our calculations. You can see that the likelihood of being replaced was about 45 percent in the 1990s. At that time, the chance of remaining a top dog for more than three years was barely more than fifty-fifty. Today the likelihood of being replaced within three years is only 30 percent. Leaders have less to worry about today than twenty years ago.

  Our second measure captures a similar idea, that of reshuffling. We rank firms by market value or by revenue in a particular year. We compute their rank again five years later. Then we compute the correlation of the two rankings. If the correlation between the two ranking series is one, it means that the relative position of firms has not changed at all over five years. If it is zero, it means that there has been a complete reshuffling within the industry. We can therefore define reshuffling as one minus the rank correlation. Figure 3.3 shows that reshuffling has decreased over the past twenty years. Market shares have become more persistent. You can more easily predict who will be on top five years from now. The answer is: the same firms as today.

  The conclusions we draw from Figures 3.2 and 3.3 are not consistent with what most people think of as competition. If we go back to our six hypotheses, we can say that Figures 3.2 and 3.3 rule out lower search costs as an explanation because this hypothesis predicts the opposite pattern. Under the Lower Search Costs hypothesis, small changes in productivity or innovation lead to large swings in market shares. Instead we see increased persistence and stability of market shares.

  FIGURE 3.2  Turnover at the top. See text for details.

  Figures 3.2 and 3.3 are consistent with decreasing domestic competition. They are also consistent with the hypotheses of the rise of superstar firms and the role of intangible assets if we assume that the comparative advantages of leaders have become more persistent. Why that would be the case is unclear, however. I have often heard arguments that intangible assets are subject to higher increasing returns to scale than tangible assets, but I have not seen convincing evidence that this is the case. In fact, I will show later that standard estimates of returns to scale have not changed much over the past twenty years. But at this point we can simply acknowledge that Figures 3.2 and 3.3 could be consistent with a rising persistence of stardom.

  Let us continue our investigation by looking at profits. The Rise of Superstar Firms and Decreasing Domestic Competition hypotheses predict rising profits. Globalization predicts decreasing profits for firms exposed to global competition. The Much Ado hypothesis predicts no systematic relationship between concentration and profits. Looking at profit margins can thus help us to parse out these hypotheses.

  FIGURE 3.3  Reshuffling. See text for details.

  Profits Margins and Payouts

  Let us now scrutinize the profits of US firms. As usual, there are several measures and several sources to construct them: profit margins versus profit rates, and national accounts versus firm-level accounts. Box 3.1 explains the key concepts.

  Figure 3.4 computes the ratio of after-tax corporate profits to GDP using the US national accounts. The profit share of GDP varies with the business cycle, and you will notice a trough during most recessions, such as in the fourth quarter of 2000 or the fourth quarter of 2008. But you can also see that over fifty years the profit share remains stable—stationary, to use the technical term—around 6 percent or 7 percent, from the end of World War II to the end of the twentieth century. Over the past two decades, however, profits have outpaced economic growth, and the after-tax profit share has increased to around 10 percent. This suggests that something fundamental has changed. We reach a similar conclusion if we compute profit margins from firm-level data. Using merged data from Compustat and the Center for Research in Security Prices, Gustavo Grullon, Yelena Larkin, and Roni Michaely (forthcoming) show that the ratio of operating income after depreciation (but before tax) over sales was around 10 percent from 1970 to 2000, and then increased to 12 percent after 2000.

  Box 3.1. Profits, Margins, Dividends, and Share Buybacks

  Consider the following stylized example of a firm. The firm starts the year with a capital stock (assets) of $100. Its annual revenues are $150 and its gross operating profits (income) are $15. In the course of doing business, the capital depreciates by 5 percent. The firm invests $7, $5 of which serve to replace the depreciated capital.

  Assets

  Revenues

  Income

  Depreciation

  Taxes

  Gross Investment

  Dividends

  $100

  $150

  $15

  $5

  $3

  $7

  $5

  The gross profit margin of the firm is 10 percent. It is the ratio of two flows: the flow of income ($15) over the flow of revenues ($150). Income net of depreciation is $10. The net margin is 6.67 percent.

  Gross profit margin

  15 / 150 = 10%

  Net profit margin

  (15–5) / 150 = 6.67%

  Net profit rate

  (15–5) / 100 = 10%

  The profit rate of the firm is 10 percent. It is the ratio of income net of depreciation (10) over the stock of capital at the beginning of the year (100). The firm pays a 30 percent tax rate on net income. It is thus left with $7 of income after taxes and depreciation.

  After paying taxes and replacing the depreciated asset, the firm must decide what to do with the leftover money. It can either invest it to grow its capital stock, or it can pay it out to its owners, the shareholders. The firm decides to pay out $5. The payout rate of the firm is 5 percent. It is the ratio of the flow of dividends (5) over the stock of capital (100). Shareholders can be paid in one of two ways. They can receive a dividend—basically, the firm writes a check to each shareholder. If our notional firm has 100 shares outstanding, each share receives 5 cents. Shareholders often prefer to avoid cash payments. For various tax and accounting issues, they tend to favor capital gains. Instead of sending 100 checks for 5 cents each, the firm could spend $5 to buy back its own shares. The value of the shares would rise and shareholders would get exactly the same payouts. In our simple example there is no difference be
tween dividends and share buybacks. In more complex and realistic cases (with stock options granted to managers, for instance), there are some differences, but it is always useful to start from the premise that dividends and buybacks are basically equivalent.

  Finally, the firm invests $2 in addition to the replacement of depreciated assets of $5. Its gross investment is $7 and its net investment is $2. The net investment rate of the firm is 2 percent. This means that the capital stock of this firm will grow by 2 percent and its assets at the beginning of next year will be $102. With more assets, it should be able to hire and produce more next year. This net investment rate is clearly crucial for real economic growth, as we discuss in Chapter 4.

  The other natural way to look at profits is to compare them to assets. We can further refine our profit measure by looking at the share of profits that is paid out to investors. Figure 3.5 shows the total payout rates for US-incorporated firms included in our Compustat sample, both dividends and share buybacks. The payout rate has increased substantially, primarily driven by share buybacks. The increase is so large that firms are now repurchasing as much as 3 percent of the book value of their assets each year.

  We now have two sets of facts: market shares have become more concentrated and more persistent, and profits have increased. The natural next question to ask is whether the two sets of facts are connected: do we see higher profits precisely in industries where we see more concentration? The answer is yes. The increase in profits is systematically linked to the increase in concentration, as shown by Grullon, Hund, and Weston (2018) and Gutiérrez and Philippon (2017): firms in concentrating industries experience rising profit margins; firms in stable industries do not. This means that our concentration measures, despite all their flaws and limitations, are capturing something real, and it rules out the Much Ado about Nothing hypothesis.

  What about international trade? Figure 3.5 is inconsistent with globalization being the dominant force for the entire US economy. Simply put, businesses struggling with foreign competition and forced to consolidate would not increase payouts to their shareholders. The fact that payouts have increased suggests that many firms feel like they have a lot of cash to spare. This of course does not mean that globalization is not crucial in some industries, particularly those heavily exposed to foreign competition from China.

  FIGURE 3.4  Corporate profits over GDP. Corporate profits after tax with inventory valuation adjustment and capital consumption adjustment, quarterly, seasonally adjusted. Data source: FRED

  FIGURE 3.5  Share buybacks and payouts. Annual data for all US-incorporated firms in our Compustat sample. Results are similar when including foreign-incorporated firms. The SEC instituted in 1982 rule 10b-18, which allows companies to repurchase their shares on the open market without regulatory limits. It was followed by a large increase in buybacks.

  The China Shock

  Industry HHIs are national measures focused on domestic firms. They can be criticized both for being too broad or too narrow. We have already discussed the critique that industry HHIs are too broad when markets are in fact local. A more interesting critique is that domestic HHIs are too narrow in a globalized world. When foreign competitors wipe out domestic firms, competition clearly increases, but domestic measures of concentration, computed with surviving firms, may very well increase. This is a serious issue.

  A prime example is what economists now refer to as the China shock. China became a member of the World Trade Organization on December 11, 2001. It marked the end of lengthy negotiations as well as a significant step toward the integration of China into the world economy. Daron Acemoglu and his co-authors (2016) estimate that import competition from China was a major force behind reductions in US manufacturing employment during the 2000s.

  Imports from China to the US had risen since the early 1990s and experienced a very rapid rise in the 2000s. This growth affected different US industries in different ways. One effect is particularly interesting for us. Before 2000, China was not considered to be a market economy. Under the Smoot-Hawley Tariff Act of 1930, nonmarket economies are subject to a relatively high tariff, known as a nonnormal trade relations (non-NTR) tariff. From 1980 onward, US presidents began granting NTR tariff rates to China, but these waivers had to be reapproved each year by Congress. If Congress failed to renew the waiver, the tariffs would jump back up to levels set in the 1930s. This introduced substantial uncertainty around future tariff rates that limited investment by both US and Chinese firms, as explained in an influential paper by Justin R. Pierce and Peter K. Schott (2016). Some industries were more affected because their 1930s reset tariffs were particularly high. We say that these industries faced high NTR gaps—gaps between the rolled-over NTR tariffs and the reset values.

  In 2000, permanent normal trade relations (PNTR) were granted to China, which took effect in December 2001. The granting of PNTR removed uncertainty around tariffs, which was particularly beneficial for industries with large NTR gaps. Indeed, Pierce and Schott (2016) show that industries with larger NTR gaps experienced larger increases in Chinese imports and larger decreases in US employment.

  Let us now study the impact of China’s import competition across US manufacturing industries. We split our industries according to their exposure to the China shock, defined by import penetration from China. Figure 3.6 shows the normalized number of firms in industries with high and low Chinese import penetration. Chinese competition leads to a strong replacement effect. Both groups have the same pre-existing trends, including during the dot-com boom, but start to diverge after 2000. By 2015 the number of firms in manufacturing industries with low exposure to China is about the same as it was in 1991. In manufacturing industries with high exposure to China, it is 40 percent lower.

  Figure 3.6 focuses only on the manufacturing sector, which employs a relatively small (and decreasing) share of the population. But it clearly shows that Globalization is a valid hypothesis in this limited field. We therefore need to take it into account.

  Our ideal competition measure should cover the whole economy and should consider foreign competition.e Once we control for foreign competition, we find that concentration has remained stable in US manufacturing (see Covarrubias, Gutiérrez, and Philippon, 2019). For the economy as a whole, we still find an increase in concentration, but it is more muted when we take into account foreign competition.

  FIGURE 3.6  The China shock: The number of active US firms in manufacturing, by exposure to China, normalized to 1 in 1991. Annual data. Manufacturing industries only are split into “high” (above-median) and “low” (below-median) exposure based on import penetration from 1991 to 2011. Data sources: Firm data from Compustat; import data from UN Comtrade

  Trade and competition interact in many fascinating ways. In Chapter 5 we will see how foreign competition is sometimes used to justify dubious domestic mergers.

  Concentration, Entrenchment, and Profits

  We have shown two important and related sets of facts. In most US industries, market shares have become more concentrated and more persistent. Industry leaders are less likely to be challenged and replaced than they were twenty years ago. At the same time, their profit margins have increased.

  We have classified the various theories into several broad hypotheses. The data that we have analyzed so far allow us to narrow our focus down to three hypotheses: consolidation driven by foreign competition, increasing efficiency of leaders, perhaps driven by intangible assets, or decreasing domestic competition. We have shown that globalization is a powerful explanation of the trends observed in the manufacturing sector, and in industries exposed to foreign competition more broadly. In the rest of the economy, however, we are left to consider star firms, decreasing domestic competition, and intangible assets as the leading theories.

  Two hypotheses can explain increasing concentration and increasing profit margins—the Rise of Superstar Firms and Decreasing Domestic Competition. They have opposite predictions for growth and welfare. How
can we tease them apart? We are going to do it in two steps. In Chapter 4 we will look at investment and employment. In Chapter 5 we will look at entry, exit, and mergers to understand how concentration took place.

  * * *

  a  The account of the dinner discussion published in the New York Times on October 16, 1929, is well worth reading.

  b  Figure 3.1 uses data from the US Census Bureau (available from their website). David Autor and his co-authors (2017) extend these data by accessing detailed files from the US Economic Census, which is conducted every five years and surveys all establishments in selected sectors. They collect data over the period 1982–2012 for six sectors (manufacturing, retail trade, wholesale trade, services, finance, and utilities and transportation) and perform a consistent analysis of 388 manufacturing and 288 nonmanufacturing industries, computing the CR4 and CR20 across these industries as well as HHI. They document a clear increase in concentration in all industries.

  c  The broad increase in concentration after 1995 is clear in both data sets, but the timing varies. In the Census data it occurs mostly in the 1990s, but in Compustat it occurs mostly in the 2000s. HHI declines in the early 1990s in Compustat. This reflects the quick increase in the number of listed firms.