The Network Law Review is pleased to present you with a Dynamic Competition Initiative (“DCI”) symposium. Co-sponsored by UC Berkeley, EUI, and Vrije Universiteit Amsterdam’s ALTI, the DCI seeks to develop and advance innovation-based dynamic competition theories, tools, and policy processes adapted to the nature and pace of innovation in the 21st century. The symposium features guest speakers and panelists from DCI’s first annual conference held in April 2023. This contribution is co-authored by Benoît Durand, Partner at RBB Economics, and Apolline Jaoui, Associate Principal at RBB Economics.
1. A new conventional wisdom
Over the last couple of years, a “new conventional wisdom” (Shapiro, 2018)1Shapiro, Carl, Antitrust in a Time of Populism (October 24, 2017). Available at SSRN: https://ssrn.com/abstract=3058345 has emerged: there is evidence of a widespread increase in market concentration in both the U.S. and the EU economies over the last few decades. Because higher level of market concentration may reflect greater firm market power, this has fuelled concern that competition authorities have not done enough to prevent this development, notably through tighter merger control and tougher enforcement of competition law against exclusionary practices.
The finding of a systematic and widespread increase in market concentration has been documented based on studies using industry level statistics (e.g., the North American Industry Classification System (NAICS) or the International Standard Industrial Classification of All Economic Activities (ISIC)).
In the U.S., several studies such as Peltzman (2014),2Peltzman, Sam. “Industrial Concentration under the Rule of Reason.” The Journal of Law & Economics 57, no. S3 (2014): S101–20. https://doi.org/10.1086/675719 Grullon et al. (2019),3Gustavo Grullon and others, Are US Industries Becoming More Concentrated?, Review of Finance, Volume 23, Issue 4, July 2019, Pages 697–743, https://doi.org/10.1093/rof/rfz007 Gutierrez and Philippon (2017),4Gutierrez Gallardo, German and Philippon, Thomas, Declining Competition and Investment in the U.S (July 2017). NBER Working Paper No. w23583, Available at SSRN: https://ssrn.com/abstract=3003721 Autor et al. (2020)5Autor D, Dorn D, Katz LF, Patterson C, Reenen JV. The Fall of the Labor Share and the Rise of Superstar Firms. Quarterly Journal of Economics. 2020;135 (2) :645-709. or the Council of Economic Advisers brief (2016)6https://obamawhitehouse.archives.gov/sites/default/files/page/files/20160414_cea_competition_issue_brief.pdf have reported an increasing trend in concentration. All of them show that over the last two decades in a majority of industries, the share of total turnover earned by top firms has increased. This result is robust to using different concentration measures (e.g., HHI or the share of the top firms) and/or to adjusting the aggregation level at which the analysis is performed (e.g., using NAICS industry at the two-digit level, which corresponds to 13 NAICS sectors, or the four-digit level, which corresponds to about 700 industries).
On this side of the Atlantic, the European Commission conducted a similar study in 2021 to assess industry concentration in France, Germany, Italy, Spain and the UK using ISIC industry classification (corresponding to 156 categories).7https://competition-policy.ec.europa.eu/system/files/2021-12/Competition%20Policy%20Brief%202-2021_Industry%20concentration%20and%20competition%20policy.pdf Based on the estimated share of the four largest firms’ turnover value (C4) in each country-year-industry combination, this study shows a moderate but widespread increase in industry concentration for the period 1998-2019. In addition, the study reports that this trend has led to a significant increase in the share of highly concentrated industries (i.e., where the four largest firms account for at least 50% of the industry). This finding is corroborated by Bajgar et al (2019)8Matej Bajgar & Giuseppe Berlingieri & Sara Calligaris & Chiara Criscuolo & Jonathan Timmis, 2019. “Industry concentration in Europe and North America,” CEP Discussion … Continue reading that find evidence of a rise in industry concentration in Europe (both at the country and the regional level) using different data sources.9This study, using the OECD MultiProd database, examines how the share of the 10% largest companies has evolved in 10 European economies between 2001 and 2012. It also uses the Orbis-Worldscope-Zephyr … Continue reading
2. Is rising market concentration always a competition concern?
Put bluntly, the answer is no.
The relationship between market concentration and competition is ambiguous. Whilst increased concentration may reflect higher market power, it may also be the outcome of competition. Indeed, some firms may become larger than their rivals because they are more efficient or have brought to market innovative products or services. A mere increase in market concentration is therefore not sufficient to conclude that competition is not working.
Demsetz (1973)10Demsetz, Harold. “Industry Structure, Market Rivalry, and Public Policy.” The Journal of Law & Economics 16, no. 1 (1973): 1–9. http://www.jstor.org/stable/724822 was amongst the first ones to develop this view, when he presented his critique of the work pioneered by Bain (1951)11Joe S. Bain “Relation of Profit-Rate to Industry Concentration: American Manufacturing, 1936-1940”, Quarterly Journal of Economics, August 1951, pp. 293-324. that showed that firms in more highly concentrated industries12Industries in which the share of turnover earned by the 8 biggest firms was larger than 70%. had a higher profit rate, suggesting a positive association between concentration and market power in the U.S.13For further discussion on trends in market power, see for the U.S. Jan De Loecker and others, The Rise of Market Power and the Macroeconomic Implications, The Quarterly Journal of Economics, Volume … Continue reading Alternatively, Demsetz (1973) suggested that large firms record superior performance because of higher cost efficiencies. In other words, concentration can also be the result of competition. More recently, Bighelli et al. (2020)14https://cepr.org/voxeu/columns/increasing-market-concentration-europe-more-likely-be-sign-strength-cause-concern documented that the observed increase in market concentration across Europe coincided with an increase in productivity. This finding was based on the CompNet micro-aggregated firm-level dataset, which evidences that in Europe, larger firms are relatively more productive than smaller ones. This again suggests that concentration can be the result of competition, which lead to the reallocation of revenue from less productive to more productive firms.
Van Reenen (2018)15https://www.kansascityfed.org/Jackson%20Hole/documents/6974/VanReenenPaper_JH2018.pdf has put forward another possible explanation regarding the observed rise in market concentration: if more markets are becoming “winner takes all” (a feature that may characterise not only markets with digital platforms but also lower tech markets such as retail and wholesale), then higher concentration levels may reflect changes in the nature of competition. In a “winner takes all” market, more competition means that firms with a competitive advantage (whether it consists in lower costs or higher quality products) will capture a larger share of the market (up to the whole market). Author et al (2020) also find that an increase in competition is consistent with an increase in concentration when “superstar firms” (the few most productive firms) capture a larger share of the market.16Autor D, Dorn D, Katz LF, Patterson C, Reenen JV. The Fall of the Labor Share and the Rise of Superstar Firms. Quarterly Journal of Economics. 2020;135 (2) :645-709.
Notwithstanding the above, a steady rise in market concentration over several decades could still be concerning. Even if firms that have achieved a stronger position may have done so because of business acumen, innovative investments or lower costs, this does not change the fact that if they remain unchallenged, potentially because of high barriers to entry, this would be consistent with few and large firms enjoying high returns in concentrated markets, without the threat of disruptive entry.
3. Industry level concentration does not say much about market concentration
Even though the trend in industry concentration is well-documented, it does not provide much insight about how market concentration has evolved over the last couple of decades. This is because industry categories do not correspond to antitrust markets.17See Shapiro, Carl, Antitrust in a Time of Populism (October 24, 2017). Available at SSRN: https://ssrn.com/abstract=3058345 or Werden, Gregory J. and Froeb, Luke M., Don’t Panic: A Guide to … Continue reading
On the one hand, industry classifications such as NAICS18The North American Industry Classification System (NAICS) was developed joint by the U.S.A, Canada and Mexico. The principle is to group economic units that have similar production processes in the … Continue reading or ISIC19The International Standard Industrial Classification (ISIC) is the international reference classification of production activities. have been developed to allocate economic units with a similar production process or similar activities to a single industry. These standards provide a systematic classification of productive activities that can be used for the collection of statistics on businesses, without any consideration for competition.
On the other hand, market definition in antitrust is used to identify “the boundaries of competition between firms” along both a product and a geographic dimension.20See European Commission’s Guidelines at https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:31997Y1209(01)&from=EN As firms compete to attract customers, demand-side substitution, which is the most immediate and effective disciplinary force on a firm’s market power, is used to determine the boundaries of markets. So-called antitrust markets are therefore defined following a product-centric approach, to include products and/or services that are sufficiently substitutable such that the elimination of competition between firms supplying those would give rise to an increase in market power.
Unsurprisingly, standardised industry classifications, even at the most granular level, can be very different to antitrust markets. For example, NAICS 212230, which includes “Copper, Nickel, Lead and Zinc Mining”, groups these activities together because of the similarity in the extracting process. However, copper, nickel, lead and zinc do not belong to the same antitrust market: they are used for different applications, meaning that there would not be any immediate substitution between these different metals in response to a small but significant and non-transitory price increase. Typically, antitrust markets involving metals would be much narrower. For example, the Commission has defined an antitrust market for the supply of copper cathodes, which is used to make copper rods and copper shapes.21[ref]See for example Case M9409 – Aurubis/Metallo Group Holding, 4/5/2020. In this case, the Commission has even considered whether to segment further copper cathodes by grades, between so-called … Continue reading This example illustrates how a production-based approach can lead to a very different product grouping than the product-based approach used by competition authorities.
That said, even if industry classifications and antitrust markets are different, it might be tempting to consider that trends in concentration observed in industry segments can provide some insights about trends in antitrust markets. This is because industries are typically wider than antitrust markets: as set out above, industries usually include many more products (and therefore potentially more firms) than antitrust markets. Furthermore, the geographic scope of antitrust markets may be defined more narrowly than under industry classifications (e.g., NAICS industries are systematically defined nationally whereas antitrust markets can be defined locally). One could expect concentration to be systematically lower in broad industry segments than in narrow antitrust markets, and on this basis, conclude that trends in industry concentration may be informative about trends in antitrust markets.
This view is short sighted. An increase in concentration at the industry level is compatible with a decrease in concentration in corresponding antitrust market(s). Notably, the number of firms operating in an industry might be reduced, raising industry concentration level while more and more firms might be competing in each of the antitrust markets that are part of this industry, reducing concentration in the markets in question. Consider again the above example, in which firms may have started producing one type of metal, which constitutes a separate antitrust market, but gradually expanded in extracting other metals, thereby entering other antitrust markets. At the same time, the number of firms active in the wider “Copper, Nickel, Lead and Zinc Mining” industry may have declined either because some firms stopped their activity or became relatively smaller (e.g., because they are just active on a single product-market).
The same logic applies with geographic markets. Industry classifications are defined at the national level, but antitrust markets can be local. In particular, this is the case of retail markets. Let us imagine an industry with 3 retailers (A, B, C) and 3 local markets (each market being of the same size). Each retailer initially holds a monopoly position in its own local area. Consider that retailer C closes down while retailers A and B expand in each of the remaining markets. These developments correspond to an increase in industry concentration at the national level (going from 3 to 2 firms) but a fall in market concentration at the local level (going from a monopoly to a duopoly in each local market). In this hypothetical case, an increase in concentration at the national level does not therefore imply an increase in concentration in the relevant local markets.
In summary, there is no systematic connection between the evolution of concentration measured at the industry classification level and that for antitrust markets. In other words, trends in concentration in a given industry may not say much about how concentration develops in antitrust markets.
4. Lessons from studies relying on antitrust markets
A couple of studies have attempted to evaluate trends in concentration in antitrust markets, or at least based on a classification that is closer to antitrust markets. Whilst these studies are not exhaustive, they provide on balance no clear indication that horizontal concentration has increased systematically over the last couple of decades.
Benkard, Yurukoglu & Zhang (2021)22C. Lanier Benkard & Ali Yurukoglu & Anthony Lee Zhang, 2021. “Concentration in Product Markets,” NBER Working Papers 28745, National Bureau of Economic Research, Inc. examine the evolution of market concentration in multiple U.S. markets defined from a product-centric approach, hence closer to an antitrust market definition. Specifically, this analysis is based on consumer survey data about consumers’ brand choices from MRI-Simmons and covers around 350 product markets, clearly a subset of the whole U.S. economy (for instance, the data does not cover intermediate goods). The data is sufficiently granular to define local markets at the state level. This paper reports that since 1994, concentration has been declining in these U.S. markets, especially in highly concentrated product markets. This finding, which holds at the national and local levels, clearly contradicts the “new conventional wisdom”.23The authors note that when product markets are aggregated into broader industries, they do find an increase in concentration over time.
Affeldt et al. (2021)24Affeldt, Pauline, Duso, Tomaso, Gugler, Klaus and Piechucka, Joanna, Market Concentration in Europe: Evidence from Antitrust Markets (January 2021). DIW Berlin Discussion Paper No. 1930, Available at … Continue reading rely on antitrust markets from 2,000 merger decisions adopted by the European Commission and find that market concentration has increased. More specifically, they assess the evolution of market concentration in over 20,000 antitrust markets over the 1995-2014 period. They show that concentration levels in these antitrust markets have on average gone up over time, especially in service markets where concentration used to be lower. Importantly, they note that the results are heavily heterogeneous across geographic markets and industries, which at the very least mitigates the finding of a systematic increase in market concentration.
While the results of Affeldt et al. (2021) and Benkard, Yurukoglu & Zhang (2021) differ,25We note that both articles find that concentration levels (as measured by HHI levels) are materially higher when relying on such a micro-approach than when assessing concentration at industry levels. they are nonetheless compatible. This is because each of these studies assess the evolution of market concentration in possibly very different markets. For instance, Affeldt et al. (2021) focus on markets where mergers took place, which is clearly not representative of the whole economy. Most importantly their sample of market suffers from a selection bias: there has been at least one merger in each of the markets under analysis, potentially giving rise to higher concentration.26Most mergers reviewed by the European Commission are unconditionally cleared.
Notwithstanding the fact that antitrust markets constitute the right level to perform the analysis, it is difficult to develop a systematic analysis for the entire economy. This is because defining antitrust markets requires a careful case-by-case assessment of demand-side substitution and supply-side substitution, which is difficult and time consuming.
So what else can be done?
5. A novel approach to delineate markets and measure concentration systematically
Hoberg and Phillips (2016)27Gerard Hoberg & Gordon Phillips, 2016. “Text-Based Network Industries and Endogenous Product Differentiation,” Journal of Political Economy, University of Chicago Press, vol. 124(5), … Continue reading have developed a text-based approach which, even though imperfect from an antitrust perspective, might be a promising alternative to define markets systematically in a way closer to antitrust markets than standard industry classifications. Hoberg and Phillips’ approach is product-centric, and as such more in line with how antitrust markets are defined.
Specifically, this new method relies upon firms’ own description of their activities. In the U.S., publicly listed companies must file a 10-K document which contains a business description section. This section mandated by the Security and Exchange Commission (SEC) regulations requires firms to describe the significant products they offer to customers. Hoberg and Phillips run a clustering algorithm on the product descriptions provided by firms in this section to define markets. The algorithm can be tightened or loosened to have wider or narrower markets: it can go from defining as many markets as the number of firms in the dataset28More specially, Hoberg and Phillips (2016) start from the subsample of N single-segment firms in 1997 (multiple-segment firms are identified using the Compustat segment database). to a target number of markets (in their study, around 300 to match existing industry classifications), depending on the degree of product similarity selected They then assign each firm’s activities to these markets. Naturally a firm can be active in several markets.
Using this novel approach in a separate study, Hoberg and Phillips (2022)29Hoberg, Gerard and Phillips, Gordon M., Scope, Scale and Concentration: The 21st Century Firm (November 2022). NBER Working Paper No. w30672, Available at SSRN: https://ssrn.com/abstract=4282522 examine how U.S. firms’ scope of activities have changed over time, but their findings have implications on trends in market concentration.
First, using this new text-based measure of product market scope (instead of traditional industry classification), they indicate that the increase in the average firm size in the U.S. over the last three decades might partly be the result of firms expanding their activities in related markets. They estimate that, between 1989 and 2017, firms’ scope increased on average by 71%. Importantly, they note that this increase in scope is driven by firms serving industries that are related (for example selling computers and cell phones) rather than diversified and unrelated (such as selling gasoline and pet food).
This increase in the average firm’s scope may well explain why concentration measured at the industry level has trended up. Indeed, if a firm expands its activities by entering product markets adjacent to its historical market, and if all these markets belong to the same industry segment, the size of that firm will grow, thereby increasing concentration at the industry level. Yet, this expansion has possibly stimulated competition by giving customers more choice in adjacent product markets. This study therefore adduces evidence that the increase in firms’ size stems at least partly from the fact that firms are expanding their activities in multiple markets.
Further, this study also revisits trends in horizontal concentration at the industry level using this novel text-based method, with results that challenge the conventional wisdom. In particular, the study estimates an aggregate HHI for the U.S. economy, taking into account firms’ scope by allocating their sales across the multiple markets in which they operate.30As noted by Hoberg and Phillips, one limitation of this approach is that they do not have the sales weights by segment and therefore use textual intensity weights instead. Hoberg and Phillips (2022) find that the aggregated HHI has not increased between 1989 and 2017. In fact, their results suggest that horizontal concentration even declined between 1989 and 1998, and has since then remained stable. Even if this analysis is limited to publicly traded firms (due to data availability), this clearly questions the widely accepted results of increasing concentration based on standard industry classification.
To sum up, the “new conventional wisdom” (i.e., the alleged widespread increase in market concentration) is not based on any robust findings. First, the finding of this very increase is challenged by new studies such as Hoberg and Phillips (2022), which have developed a text-based approach to group firms instead of relying on traditional industry classifications. Second, and importantly, as discussed here and by others,31See Shapiro, Carl, Antitrust in a Time of Populism (October 24, 2017). Available at SSRN: https://ssrn.com/abstract=3058345 or Werden, Gregory J. and Froeb, Luke M., Don’t Panic: A Guide to … Continue reading the finding that horizontal concentration is trending up using industry classifications is simply not meaningful from an antitrust perspective. The new conventional wisdom should not be used to justify a change in antitrust policy, notably to abandon the consumer welfare standard in favor of a “big is bad” approach to antitrust regulation.
Notwithstanding the above, it is impossible to develop a comprehensive evaluation of horizontal concentration in the economy using antitrust markets. While imperfect from an antitrust perspective, Hoberg and Phillips’ new text-based approach could be pursued for that purpose. When comparing firms’ scope based upon their new method with that based on standard industry classifications, Hoberg and Phillips had to define around 300 markets. This is, from an antitrust perspective, a too low number for the U.S. economy. However, a major feature of the clustering algorithm is that it can generate a larger number of segments or markets (leveraging similarity in product description). This new method represents a potentially powerful alternative to using standard industry classification to evaluate trends in market concentration.
Benoît Durand & Apolline Jaoui
|Citation: Benoît Durand & Apolline Jaoui, “A Rising Trend in Market Concentration: Myth or Reality?”, Network Law Review, Summer 2023.