A recent NBER working paper by Gutiérrez & Philippon attempts to link differences in U.S. and EU antitrust enforcement and product market regulation to differences in market concentration and corporate profits. The paper’s abstract begins with a bold assertion:
Until the 1990’s, US markets were more competitive than European markets. Today, European markets have lower concentration, lower excess profits, and lower regulatory barriers to entry.
The authors are not clear what they mean by lower, however its seems they mean lower today relative to the 1990s.
This blog post focuses on the first claim: “Today, European markets have lower concentration …”
At the risk of being pedantic, Gutiérrez & Philippon’s measures of market concentration for which both U.S. and EU data are reported cover the period from 1999 to 2012. Thus, “the 1990s” refers to 1999, and “today” refers to 2012, or six years ago.
The table below is based on Figure 26 in Gutiérrez & Philippon. In 2012, there appears to be no significant difference in market concentration between the U.S. and the EU, using either the 8-firm concentration ratio or HHI. Based on this information, it cannot be concluded broadly that EU sectors have lower concentration than the U.S.
Gutiérrez & Philippon focus on the change in market concentration to draw their conclusions. However, the levels of market concentration measures are strikingly low. In all but one of the industries (telecommunications) in Figure 27 of their paper, the 8-firm concentration ratios for the U.S. and the EU are below 40 percent. Similarly, the HHI measures reported in the paper are at levels that most observers would presume to be competitive. In addition, in 7 of the 12 sectors surveyed, the U.S. 8-firm concentration ratio is lower than in the EU.
The numbers in parentheses in the table above show the change in the measures of concentration since 1999. The changes suggests that U.S. markets have become more concentrated and EU markets have become less concentrated. But, how significant are the changes in concentration?
A simple regression of the relationship between CR8 and a time trend finds that in the EU, CR8 has decreased an average of 0.5 percentage point a year, while the U.S. CR8 increased by less than 0.4 percentage point a year from 1999 to 2012. Tucked in an appendix to Gutiérrez & Philippon, Figure 30 shows that CR8 in the U.S. had decreased by about 2.5 percentage points from 2012 to 2014.
A closer examination of Gutiérrez & Philippon’s 8-firm concentration ratio for the EU shows that much of the decline in EU market concentration occurred over the 1999-2002 period. After that, the change in CR8 for the EU is not statistically significantly different from zero.
A regression of the relationship between HHI and a time trend finds that in the EU, HHI has decreased an average of 12.5 points a year, while the U.S. HHI increased by less than 16.4 points a year from 1999 to 2012.
As with CR8, a closer examination of Gutiérrez & Philippon’s HHI for the EU shows that much of the decline in EU market concentration occurred over the 1999-2002 period. After that, the change in HHI for the EU is not statistically significantly different from zero.
Readers should be cautious in relying on Gutiérrez & Philippon’s data to conclude that the U.S. is “drifting” toward greater market concentration while the EU is “drifting” toward lower market concentration. Indeed, the limited data presented in the paper point toward a convergence in market concentration between the two regions.
At issue is the growth in the incidence of common-ownership across firms within various industries. In particular, institutional investors with broad portfolios frequently report owning small stakes in a number of firms within a given industry. Although small, these stakes may still represent large block holdings relative to other investors. This intra-industry diversification, critics claim, changes the managerial objectives of corporate executives from aggressively competing to increase their own firm’s profits to tacitly colluding to increase industry-level profits instead. The reason for this change is that competition by one firm comes at a cost of profits from other firms in the industry. If investors own shares across firms, then any competitive gains in one firm’s stock are offset by competitive losses in the stocks of other firms in the investor’s portfolio. If one assumes corporate executives aim to maximize total value for their largest shareholders, then managers would have incentive to soften competition against firms with which they share common ownership. Or so the story goes (more on that in a later post.)
Elhague and Posner, et al., draw their motivation for new antitrust offenses from a handful of papers that purport to establish an empirical link between the degree of common ownership among competing firms and various measures of softened competitive behavior, including airline prices, banking fees, executive compensation, and even corporate disclosure patterns. The paper of most note, by José Azar, Martin Schmalz, and Isabel Tecu and forthcoming in the Journal of Finance, claims to identify a causal link between the degree of common ownership among airlines competing on a given route and the fares charged for flights on that route.
Measuring common ownership with MHHI
Azar, et al.’s airline paper uses a metric of industry concentration called a Modified Herfindahl–Hirschman Index, or MHHI, to measure the degree of industry concentration taking into account the cross-ownership of investors’ stakes in competing firms. The original Herfindahl–Hirschman Index (HHI) has long been used as a measure of industry concentration, debuting in the Department of Justice’s Horizontal Merger Guidelines in 1982. The HHI is calculated by squaring the market share of each firm in the industry and summing the resulting numbers.
The MHHI is rather more complicated. MHHI is composed of two parts: the HHI measuring product market concentration and the MHHI_Delta measuring the additional concentration due to common ownership. We offer a step-by-step description of the calculations and their economic rationale in an appendix to our paper. For this post, I’ll try to distill that down. The MHHI_Delta essentially has three components, each of which is measured relative to every possible competitive pairing in the market as follows:
A measure of the degree of common ownership between Company A and Company -A (Not A). This is calculated by multiplying the percentage of Company A shares owned by each Investor I with the percentage of shares Investor I owns in Company -A, then summing those values across all investors in Company A. As this value increases, MHHI_Delta goes up.
A measure of the degree of ownership concentration in Company A, calculated by squaring the percentage of shares owned by each Investor I and summing those numbers across investors. As this value increases, MHHI_Delta goes down.
A measure of the degree of product market power exerted by Company A and Company -A, calculated by multiplying the market shares of the two firms. As this value increases, MHHI_Delta goes up.
This process is repeated and aggregated first for every pairing of Company A and each competing Company -A, then repeated again for every other company in the market relative to its competitors (e.g., Companies B and -B, Companies C and -C, etc.). Mathematically, MHHI_Delta takes the form:
where the Ss represent the firm market shares of, and Betas represent ownership shares of Investor I in, the respective companies A and -A.
As the relative concentration of cross-owning investors to all investors in Company A increases (i.e., the ratio on the right increases), managers are assumed to be more likely to soften competition with that competitor. As those two firms control more of the market, managers’ ability to tacitly collude and increase joint profits is assumed to be higher. Consequently, the empirical research assumes that as MHHI_Delta increases, we should observe less competitive behavior.
And indeed that is the “blockbuster” evidence giving rise to Elhauge’s and Posner, et al.,’s arguments For example, Azar, et. al., calculate HHI and MHHI_Delta for every US airline market–defined either as city-pairs or departure-destination pairs–for each quarter of the 14-year time period in their study. They then regress ticket prices for each route against the HHI and the MHHI_Delta for that route, controlling for a number of other potential factors. They find that airfare prices are 3% to 7% higher due to common ownership. Other papers using the same or similar measures of common ownership concentration have likewise identified positive correlations between MHHI_Delta and their respective measures of anti-competitive behavior.
Problems with the problem and with the measure
We argue that both the theoretical argument underlying the empirical research and the empirical research itself suffer from some serious flaws. On the theoretical side, we have two concerns. First, we argue that there is a tremendous leap of faith (if not logic) in the idea that corporate executives would forgo their own self-interest and the interests of the vast majority of shareholders and soften competition simply because a small number of small stakeholders are intra-industry diversified. Second, we argue that even if managers were so inclined, it clearly is not the case that softening competition would necessarily be desirable for institutional investors that are both intra- and inter-industry diversified, since supra-competitive pricing to increase profits in one industry would decrease profits in related industries that may also be in the investors’ portfolios.
On the empirical side, we have concerns both with the data used to calculate the MHHI_Deltas and with the nature of the MHHI_Delta itself. First, the data on institutional investors’ holdings are taken from Schedule 13 filings, which report aggregate holdings across all the institutional investor’s funds. Using these data masks the actual incentives of the institutional investors with respect to investments in any individual company or industry. Second, the construction of the MHHI_Delta suffers from serious endogeneity concerns, both in investors’ shareholdings and in market shares. Finally, the MHHI_Delta, while seemingly intuitive, is an empirical unknown. While HHI is theoretically bounded in a way that lends to interpretation of its calculated value, the same is not true for MHHI_Delta. This makes any inference or policy based on nominal values of MHHI_Delta completely arbitrary at best.
We’ll expand on each of these concerns in upcoming posts. We will then take on the problems with the policy proposals being offered in response to the common ownership ‘problem.’
Yesterday the hearing in the DOJ’s challenge to stop the Aetna-Humana merger got underway, and last week phase 1 of the Cigna-Anthem merger trial came to a close.
The DOJ’s challenge in both cases is fundamentally rooted in a timeworn structural analysis: More consolidation in the market (where “the market” is a hotly-contested issue, of course) means less competition and higher premiums for consumers.
Following the traditional structural playbook, the DOJ argues that the Aetna-Humana merger (to pick one) would result in presumptively anticompetitive levels of concentration, and that neither new entry not divestiture would suffice to introduce sufficient competition. It does not (in its pretrial brief, at least) consider other market dynamics (including especially the complex and evolving regulatory environment) that would constrain the firm’s ability to charge supracompetitive prices.
Aetna & Humana, for their part, contend that things are a bit more complicated than the government suggests, that the government defines the relevant market incorrectly, and that
the evidence will show that there is no correlation between the number of [Medicare Advantage organizations] in a county (or their shares) and Medicare Advantage pricing—a fundamental fact that the Government’s theories of harm cannot overcome.
The trial will, of course, feature expert economic evidence from both sides. But until we see that evidence, or read the inevitable papers derived from it, we are stuck evaluating the basic outlines of the economic arguments based on the existing literature.
Our paper challenges these claims. As we summarize:
This white paper counsels extreme caution in the use of past statistical studies of the purported effects of health insurance company mergers to infer that today’s proposed mergers—between Aetna/Humana and Anthem/Cigna—will likely have similar effects. Focusing on one influential study—Paying a Premium on Your Premium…—as a jumping off point, we highlight some of the many reasons that past is not prologue.
In short: extrapolated, long-term, cumulative, average effects drawn from 17-year-old data may grab headlines, but they really don’t tell us much of anything about the likely effects of a particular merger today, or about the effects of increased concentration in any particular product or geographic market.
While our analysis doesn’t necessarily undermine the paper’s limited, historical conclusions, it does counsel extreme caution for inferring the study’s applicability to today’s proposed mergers.
By way of reference, Dafny, et al. found average premium price increases from the 1999 Aetna/Prudential merger of only 0.25 percent per year for two years following the merger in the geographic markets they studied. “Health Insurance Mergers May Lead to 0.25 Percent Price Increases!” isn’t quite as compelling a claim as what critics have been saying, but it’s arguably more accurate (and more relevant) than the 7 percent price increase purportedly based on the paper that merger critics like to throw around.
Moreover, different markets and a changed regulatory environment alone aren’t the only things suggesting that past is not prologue. When we delve into the paper more closely we find even more significant limitations on the paper’s support for the claims made in its name, and its relevance to the current proposed mergers.