Revisiting the Effect of Common Ownership on Pricing in the Airline Industry

The common ownership hypothesis suggests that large, institutional investors who own equity stakes in firms that compete in the same industry have an incentive to reduce competition by, for example, increasing prices, lowering production, or increasing barriers to entry. The theory behind the hypothesis is relatively straightforward. Firms are tasked with maximizing shareholder value, which typically coincides with simply maximizing their own profits. However, if shareholders also own equity in direct competitors, then maximizing shareholder value implies that firms have an incentive to maximize some combination of their own profits as well as those of their competitors. But if firms consider their competitors’ profits when making business decisions then competition may decline and consumer welfare may decrease.

The hypothesis began to garner significant attention from academic researchers, regulators, and practitioners with a highly influential and provocative paper by Azar, Schmalz, and Tecu (2018) (hereafter AST) that claimed to find evidence of anti-competitive effects from increased institutional common ownership. The AST paper provided empirical evidence from the airline industry showing that markets with higher levels of common ownership concentration are characterized by higher average ticket prices.

The AST findings have led legal scholars and policymakers to pressure antitrust authorities to examine regulations that may hinder the ability of institutions to act in the best interests of their clients. For example, Posner, Morton, and Weyl (2016) proposed that Congress pass legislation to withdraw tax advantages for retirement funds that invest in any mutual fund owning shares in multiple firms in the same industry, and that investors “limit their holdings of an industry to a small stake (less than 1% of the total size of the industry), or hold shares in only a single ‘effective firm’ per industry.” Posner (2021) proposes that regulators forbid common owners to communicate with, or exert control over, managers in any way, and that when common ownership of a firm exceeds a threshold, the compensation plan must be reviewed by the Securities and Exchange Commission. Elhauge (2015), citing the AST evidence, argued for enforcement of the Clayton Act, specifically §7, to challenge stock acquisitions that result in a common set of investors owning shares in corporations that are horizontal competitors. Furthermore, the Department of Justice and the Federal Trade Commission are investigating whether the rise of common ownership violates antitrust laws. Specifically, in question 12.h of their request for information in January 2022, they solicit input to the effect of common ownership on mergers. In response, Schmalz (2022) has proposed that regulators account for common ownership when approving mergers, which will make them more difficult. The effects of implementing these proposals would, de facto, limit institutional investors’ ability to offer well-diversified portfolios to their clients. This may result in harm to institutional clients, many of which are individual investors served via defined contribution and defined benefit retirement plans, in the form of reduced returns for a given level of risk.

Our paper, “Common Ownership Does Not Have Anti-Competitive Effects in the Airline Industry”, revisits the AST evidence and argues that the positive correlation between the measure of common ownership concentration used in the paper and airline ticket prices is likely spurious and thus, should not be used to argue for increased regulation of institutional investors.

In the paper we make a number of critiques of the AST empirical analysis. Our first, and primary criticism focuses on the measure of common ownership concentration that AST use, which we refer to as HHI∆. HHI∆ is a measure of common ownership that varies at the market level, where a market in this context is a route between two end-point airports. The measure is a non-linear function of both the market shares of all of the airlines that operate in the market and the extent of institutional ownership and control of those airlines, where control is proxied for by voting rights and ownership by shares held. The AST paper documents a strong positive correlation between HHI∆ and airfares, and argues that the correlation is likely causal in nature. We argue that for such an interpretation to be valid, one must disentangle whether it is the market share or the ownership and control components of the common ownership measure that drives the positive correlation with ticket prices. If the positive correlation with prices is mainly driven by variation in airline market shares, due to endogeneity bias for example, then it is unlikely that common ownership is causing anti-competitive effects in the market.

To shed light on this issue, we construct two version of HHI∆: The first, uses placebo ownership and control and true airline market shares, and the second uses true ownership and control and placebo market shares. We repeat the AST analysis for each version of HHI∆. If ownership and control is driving the positive correlation between HHI∆ and ticket prices, we expect that the measure using placebo ownership and control and true market shares will yield insignificant results, while the measure constructed with true ownership and control and placebo market shares would yield the positive correlation with prices documented in AST. Our results show no significant correlation between ticket prices and the measure of HHI∆ that uses placebo market shares and true ownership and control. However, we find a positive, large and significant correlation between ticket prices and the version of HHI∆ that uses true market shares and placebo ownership. We argue that this finding casts doubt on a causal interpretation of the AST results, namely, that common ownership leads to anti-competitive effects.

In the remainder of the paper, we show that reasonable alternative empirical assumptions for
measuring investor control and for treating the extent of investor ownership and control of airlines operating in bankruptcy lead to highly attenuated correlations between prices and common
ownership.

We argue that while measuring an institutional investor’s ownership in a firm is relatively straightforward, as it equals the percentage of shares the institution owns out of the firm’s total outstanding shares, measuring the extent of investor control is not as simple. The literature on corporate governance has identified two channels through which equity holders exert control. The first is direct intervention or “voice” (Hirschman (1970)). Examples of the voice mechanism include communication between shareholders and firm managers, discussions between shareholders and board members outside management, and shareholders’ vote to implement profitable projects or terminate under-performing managers. The second channel, recently revisited theoretically by Edmans (2009), Admati and Pfleiderer (2009), and Edmans and Manso (2011), involves the threat of exit, whereby shareholders discipline managers by “voting with their feet” and selling their shares if dissatisfied with management performance. The AST paper assumes that shareholders exert control exclusively through voting and equate control with the number of shares designated as having voting rights. We take a more agnostic approach and consider both mechanisms. After replicating the AST measure of control by vote, we explore how the results are affected if shareholders instead exert control through the threat of exit or through any of the alternative voice mechanisms available to all equity owners, regardless of the voting designations associated with the shares owned. We find that the positive correlation between HHI∆ and ticket prices in AST becomes statistically insignificant when control is assumed to operate through these alternative mechanisms. Further, we show that voting designations are classified inconsistently both across institutions and within institutions over time, likely due to the vague definitions and unclear reporting instructions provided by the Securities and Exchange Commission (SEC). These results cast doubt on the accuracy of capturing the extent of an institution’s control using reported voting designations.

Finally, we show that the positive relationship between airline prices and common ownership is particularly sensitive to assumptions about investor control and ownership in a bankrupt carrier. Bankruptcy is a significant issue in this context as at least one major airline operated under bankruptcy protection in half of the quarters in the sample period (2001:Q1 to 2014:Q4). The AST paper assumes that equity holders retain full control and cash-flow rights over insolvent airlines. We show that alternatively assuming that equity holders of bankrupt airlines either lose cash-flow rights or retain cash flow rights but lose control significantly attenuates the positive correlation between AST’s measure of common ownership concentration and airline ticket prices.

To summarize, we find that the AST results are driven by the market share, not the ownership, component of the common ownership measure. Furthermore we find that the AST results are sensitive to how governance and and bankruptcy are accounted for. Given the harm that investors could suffer from constraining institutions with additional regulation, lawmakers and regulators should be skeptical of the claimed negative effects of common ownership.