The Great Reversal Read online

  The CPI can be slow at incorporating new products: automobiles only entered the index in 1935, air conditioners in 1964, and cell phones in 1998, long after their prices had fallen significantly. The issue is that the CPI misses the (big) initial introduction effect and overestimates inflation.

  When BLS data collectors cannot obtain a price for an item in the CPI sample (for example, because the outlet has stopped selling it), they look for a replacement item that is closest to the missing one. The BLS then adjusts for changes in quality and specifications. It can use manufacturers’ cost data or hedonic regressions to compute quality adjustments. Hedonic regressions are statistical models to infer consumers’ willingness to pay for goods or services. When it cannot estimate an explicit quality adjustment, the BLS imputes the price change using the average price change of similar items in the same geographic area. Finally, the BLS has specific procedures to estimate the price of housing (rents and owners’ equivalent rents) and medical care. (The medical care component of the CPI covers only out-of-pocket expenses.)

  Where are we today, a little more than two decades after the publication of the Boskin Commission report? We can probably say that the glass is half full. The BLS has improved its process of quality adjustments and pays close attention to new goods, but some of the methods that the BLS uses to assess replacement goods work well with relatively small changes in quality or characteristics but less well when an innovative new product significantly disrupts consumer behavior. It is therefore possible that we still overestimate inflation.

  This issue is particularly severe in the case of “free” goods and services provided by internet platforms such as Google and Facebook. Fortunately, top-notch economists like Chad Jones (2017) from the University of Chicago have looked into that problem and have given us some ways to think about it. Estimation of inflation is tightly connected to the measurement of productivity growth, and we will continue the discussion in Chapter 5. For now, let’s just note that the mismeasurement is not large enough to materially change our conclusions about the slowdown in productivity.

  Clearly, Amazon is great for busy, high-earning households.

  Another relevant point is that Amazon’s profit margins do not appear to be particularly high. Amazon’s profit margin is in line with the average in retail. If anything, it appears to be a bit lower. And more important, Amazon invests a lot. We will discuss investment in greater detail later, but it’s important to point out immediately that when competition is weak, investment incentives are also weak. The fact that Amazon invests a lot is clearly a good sign as far as competition is concerned.

  Does this mean that there is nothing to worry about? Not quite. Amazon’s pricing decisions are dynamic, not static. Amazon charges low prices to build its market share. The fear, as Lina M. Khan of the Open Markets Institute has explained, is that it is willing to forego profits only to establish its dominance in many markets and then exploit its market power (Khan, 2017). But do we have any evidence that this is happening?

  Two concepts from industrial economics shed light on the issue: loss leader pricing and predatory pricing. Loss leader pricing is the strategy of a firm that sells a product at a loss in order to attract customers and stimulate the sales of other, more profitable goods and services. It is a common strategy in retail. Predatory pricing, on the other hand, happens when a firm sets low, unsustainable prices in order to drive its competitors out of business. If you think about it, you quickly realize that, in practice, it is hard to distinguish one from the other.

  Khan discusses the case of the e-commerce company Quidsi, owner of Diapers.com and a few other sites. Amazon expressed interest in acquiring Quidsi in 2009, but Quidsi rejected the offer. Soon after, Amazon engaged in a price war with Quidsi by cutting the price of diapers and other baby products on Amazon.com and by offering steep discounts in its new Amazon Mom program. Quidsi did not have a war chest to match Amazon’s, and eventually Quidsi’s owners decided to sell the company. They received offers from both Walmart and Amazon, and accepted Amazon’s offer at the end of 2010. The Federal Trade Commission reviewed the deal under Section 7 of the Clayton Act. The deal clearly raised some red flags, but the FTC decided not to pursue the case. A year later, Amazon rolled back its discounts.

  This can plausibly be described as predatory pricing. But is it harmful for consumers? It is not obvious since we do not know if prices are now higher than they would have been without the acquisition. This illustrates the difficulty of judging cases of predatory pricing. It also clearly shows that regulators must remain vigilant.

  Concentration Is Not Always Bad

  We have seen some of the standard tools that economists use to study competition: market shares, concentration, profits, and prices. We have seen that none of them is perfect. Concentration raises legitimate concerns of market dominance, but it can also reflect the increasing efficiency of market leaders. Efficient firms are profitable, but sustained abnormal profits are a bad sign. Low prices are almost always a good sign, unless they involve predatory pricing.

  When we look at antitrust actions, we hope to see lower prices. Can we then conclude that the antitrust actions were beneficial? The answer is that it’s likely, but it’s not entirely obvious because competition can be excessive, at least in theory. This can happen when investment and innovation decline after competition increases. When we look at episodes of deregulation, we hope to see entry by new firms. Does this mean deregulation was a good idea? Again, the answer is probably yes, but we can also construct examples where there are too many entrants, and the decline in profits exceeds the gain for consumers. These are delicate empirical questions, and we will need to look at a broad set of economic indicators. You guessed it: we need data, more data!

  In Chapters 3, 4, and 5 we will review the broad trends in the US economy over the past twenty years, looking at entry and exit of businesses, market shares, mergers, profits, stock buybacks, and investment.

  * * *

  a  Olivier Blanchard (2003) explains in his discussion of Basu et al. (2003), “fully one-third of the increase in TFP [total factor productivity] growth from the first to the second half of the 1990s in the United States came from the retail trade sector.” A study by the McKinsey Global Institute (Lewis et al., 2001) focused on the factors behind US TFP growth in the 1990s. In general merchandise (representing 16 percent of the TFP growth acceleration), the study found that “Wal-Mart directly and indirectly caused the bulk of the productivity acceleration through ongoing managerial innovation that increased competitive intensity and drove the diffusion of best practice.” Similarly, the wholesale trade sector contributed a lot to productivity growth after 1995. In pharmaceuticals wholesaling, the study found that “half of the acceleration was driven by warehouse automation and improvements in organization.”

  b  Suzanne Kapner, “Sears reshaped America, from Kenmore to Allstate.” Wall Street Journal, October 15, 2018.

  CHAPTER 3

  The Rise in Market Power

  IN 1998 JOEL KLEIN, who ran the antitrust operation at the US Department of Justice, declared in a January 29 address before the New York State Bar Association that “our economy is more competitive today than it has been in a long, long time.” That statement was true, but unfortunately for US households, it was not prescient. This is not a critique of Klein. He could not have foreseen the evolution of US markets. And the history of economics is replete with more embarrassing predictions. The well-known Yale economist Irving Fisher gave a speech at the monthly dinner of the American Purchasing Agents Association where he argued that “stock prices have reached what looks like a permanently high plateau.” As a general statement about stock prices, it is rather silly. The more unfortunate point, however, is that he made that claim on October 15, 1929.a

  In Chapter 2, we described various ways to measure competition. We understand their meaning, usefulness, and limitations. Let’s put them to work.

  Concentration of Market Shares


  It is natural to start with concentration. There are two basic measures of concentration: one is the market share of the top firms, either of a single firm (concentration ratio CR1) or of the top five or top eight firms in the industry (CR5 or CR8); the other is the Herfindahl-Hirschman index (HHI) that we dicussed in Chapter 2. Using both measures, we are going to show that concentration has increased in most US industries.

  The US Census Bureau provides estimates of revenue concentration by industry. In April 2016, the Council of Economic Advisers, chaired by President Obama’s chief economist Jason Furman, pointed out that “the majority of industries have seen increases in the revenue share enjoyed by the 50 largest firms between 1997 and 2012” (Council of Economic Advisers, 2016). Figure 3.1 replicates and extends these results. It shows the rise in concentration in the US economy using CR8 computed from Census data separately for manufacturing and nonmanufacturing industries. We have more granular data for the manufacturing sector, where we can perform the analysis with 360 manufacturing industries (NAICS level 6; NAICS levels are explained in the first section of the Appendix). In manufacturing, the top eight firm concentration ratio (CR8) increased from 50 percent to 59 percent. For nonmanufacturing industries, we can perform the analysis at NAICS level 3, which has a few more than seventy industries. With this wider definition the CR8s are smaller, but the increase is large, from 15 percent to 25 percent.b

  The Census includes the universe of US firms, so it represents the most comprehensive data source. It has some limitations, however. It does not contain financial information, and it is based on establishment accounting. An alternate data source is Compustat, a database of firm-level financial information from S&P Global Market Intelligence. Compustat, unlike the Census, gives a partial coverage of the economy: it includes only large firms that are (or have been) publicly traded. Its coverage is smaller than that of the Census, but it has more historical depth, and it contains a wealth of financial information consolidated at the firm level. Compustat allows us to check the robustness of our results and to expand them. With my co-authors Matias Covarrubias and Germán Gutiérrez (2019) we show that the rise in concentration is similar in Compustat and in Census data, and whether we measure concentration using HHI scores or CR8.c Gustavo Grullon, Yelena Larkin, and Roni Michaely (forthcoming) were the first to point out the increase in concentration in the Compustat data set. They found that concentration had increased in more than three-quarters of US industries. In addition, they showed that firms in concentrating industries experience rising profit margins. We are going to study profits later in the chapter.

  FIGURE 3.1  Concentration using top eight firm Census shares, cumulative change in CR8. Annual data. The concentration ratio is defined as the market share (by sales) of the eight largest firms in each industry. See Autor et al. (2017) for concentration time-series under a consistent segmentation, which exhibit similar trends. Data sources: US Concentration Ratios (CRs) from Economic Census, based on SIC segments before 1997 and NAICS segments after 1997. Data for manufacturing are reported at NAICS level 6 (SIC 4 for 1992) because it is available only at that granularity in 1992. Data for nonmanufacturing are based on NAICS level 3 segments (SIC 2 for 1992).

  In Chapter 2 we discussed some caveats when using industry concentration as an indicator of competition. One caveat is that the relevant market concentration might not be correctly captured by industry-level measures (recall our example of routes versus national average for airlines). A second caveat is that concentration might signal changes in industry dynamics that are not directly related to market power, such as an increasing efficiency gap between industry leaders and laggards (Walmart in the 1990s) or consolidation in declining industries.

  Spelling Out the Hypotheses

  Although the rise in concentration has been well documented, there is little agreement about its causes and even less about its consequences. Jason Furman, when he was chair of the Council of Economic Advisers, argued that the rise in concentration suggested “economic rents and barriers to competition.” David Autor, David Dorn, Lawrence Katz, Christina Patterson, and John Van Reenen (2017) have argued almost exactly the opposite, namely that concentration reflects “a winner take most feature” explained by the fact that “consumers have become more sensitive to price and quality due to greater product market competition.”

  Measures of concentration are suggestive, but they cannot by themselves tell us that competition has indeed decreased. To continue our inquiry and deepen our understanding, it is useful to articulate several hypotheses. This is, after all, the usual process in scientific research. So here is a list of six hypotheses to interpret the data:

  Hypothesis of Much Ado about NothingIndustry concentration measures are meaningless because industry codes are too coarse and because markets are local (an argument of antitrust specialists).

  Hypothesis of Decreasing Domestic CompetitionCompetition has declined in many US industries (the argument of this book).

  Hypothesis of the Rise of Superstar FirmsConcentration reflects the increasing productivity of industry leaders.

  Hypothesis of Lower Search CostsThe internet makes price comparisons easier, and this leads to winner-take-all outcomes.

  Hypothesis of GlobalizationForeign competition leads to domestic consolidation.

  Hypothesis of Intangible AssetsThe growth of intangible assets explains the evolution of concentration, profits, and investment.

  It might seem strange at first to entertain the Much Ado about Nothing hypothesis as an explanation, but I think it is healthy to keep in the back of our mind the possibility that we might simply be creating more background noise. I am reminded of this issue every time I listen to pundits discuss the stock market. They always have a view about where the market is going and why. They can argue passionately about the meaning of a particular pattern and come up with theories to rationalize fluctuations that are essentially random. The lack of knowledge has rarely prevented human beings from discussing the news, so let us entertain this hypothesis as a way of keeping our overconfidence in check. Under it, measures of industry concentration should be treated as noise, and they would neither predict nor explain real outcomes.

  The remaining five hypotheses are not mutually exclusive. It is obvious that foreign competition (from Mexico, China, and Japan) has affected some industries. It is also obvious that some firms have amazing intangible assets. We have already discussed Amazon. We will study Apple, Facebook, Google, and Microsoft in Chapters 13 and 14. In all likelihood, therefore, the truth is a mix of these hypotheses with varying relevance across industries and time periods. There could be other hypotheses, but I think they would boil down to a combination of these five.

  The Rise of Superstar Firms hypothesis is the story of Walmart in the 1990s, which is discussed in Chapter 2. According to this view, concentration is good news and should be linked to faster productivity growth. The hypothesis of Lower Search Costs is related to the superstar firms hypothesis, but it is conceptually distinct. It argues that consumers have become more price-elastic thanks to online shopping tools. Notice that this hypothesis implies that ex post competition has increased and that profit margins (earnings over sales) have decreased. Sales concentration increases because, with lower margins, firms need to be larger in order to recoup their fixed entry costs. In their well-known 2017 paper, Autor, Dorn, Katz, Patterson, and Van Reenen argue that the rise of superstar firms explains the fall in the labor share of income that we see in most US industries and in the aggregate. We will come back to this point.

  The Decreasing Domestic Competition hypothesis argues the opposite position, that barriers to entry have increased and that this has given incumbents more market power, thereby decreasing domestic competition. This is the interpretation that I am promoting in this book for most—but not all—industries. If correct, it invites the next question: why?

  Globalization is not really a hypothesis. Globalization is a fact. Rather, the question is
a quantitative one: Does it explain much of what we see or only a little? Are there industries where globalization is the main driving force, and others where it is not? Under the Globalization hypothesis, we expect foreign competition to put downward pressure on profit margins, forcing domestic firms to exit or to merge. It is clear that some manufacturing industries such as textile manufacturing have followed this pattern. Trade economists Robert Feenstra and David Weinstein estimated in a 2017 paper the impact of globalization on mark-ups and concluded that mark-ups generally decrease in industries affected by foreign competition. Another important fact to keep in mind is that globalization is a two-way street. The flip side of foreign competition is that successful domestic firms can expand globally and become large relative to their home countries. A perfect example is the rise of the Finnish telecom company Nokia in the 2000s. At its peak, it accounted for about two-thirds of Helsinki’s stock exchange market capitalization, almost half of corporate research and development, and about 20 percent of Finnish exports.d We should be careful, then, when comparing consolidated firm revenues (including foreign sales) to domestic GDP.

  Finally, the Intangible Assets hypothesis contains several ideas. Intangible assets are nonphysical in nature. They include intellectual property, like patents and copyrights, but extend to vague or fuzzy assets, such as brand recognition. Economists Nicolas Crouzet and Janice Eberly (2018) argue that industry leaders are often firms that are very good at producing intangible assets. In fact, they argue that this is how they became leaders in the first place. The attractive feature of a theory of intangible assets is that it can explain concentration both through increasing productivity (superstar firms) and through decreasing domestic competition since intangible assets can create barriers to entry. To test this idea, we will look carefully at intangible investments across firms and industries in Chapter 5. Network effects and increasing differences in the productivity of information technology could also increase the efficient scale of operation of the top firms, leading to higher concentration.