Common Ownership in America: 1980-2017

Matthew Backus is Assistant Professor at Columbia Business School; Christopher Conlon is Assistant Professor at the New York University Stern School of Business; and Michael Sinkinson is Assistant Professor at the Yale School of Management. This post is based on their recent paper. Related research from the Program on Corporate Governance includes Index Funds and the Future of Corporate Governance: Theory, Evidence, and Policy by Lucian Bebchuk and Scott Hirst (discussed on the forum here); The Agency Problems of Institutional Investors by Lucian Bebchuk, Alma Cohen, and Scott Hirst (discussed on the Forum here); and New Evidence, Proofs, and Legal Theories on Horizontal Shareholding by Einer Elhauge (discussed on the Forum here).

The classic profit-maximizing model of the publicly-traded firm has underpinned every aspect of economics for a century, from antitrust and regulation to theories of taxes and trade. According to this model, a public firm’s shareholders hire the management to maximize the firm’s profits, and thereby maximize the value of those shares. Trends emerging in the past few decades, including the rise of indexing among investors have led some to question whether the classic model is still applicable. Unsurprisingly, this has led to a forceful debate, centered on the implications of common ownership, and the reliability of the research purporting to show its effects. Our new working paper, Common Ownership in America: 1980–2017 provides a new analytical framework for this debate, by comprehensively analyzing the theoretical and empirical implications of the common ownership hypothesis among all S&P 500 firms from 1980–2017. Our paper identifies why common ownership presents such a tremendous challenge to markets and regulators if prevailing hypotheses are true, but we also identify important data problems, misconceptions and erroneous assumptions that call into question the reliability of existing research. Our goal in doing so is to pave the way for more rigorous research and discussion on this important question.

The literature on this subject has been spurred by two trends: diversification of holdings and concentration among institutional asset managers. In particular, the past four decades have seen the rise in indexing investment strategies: instead of investors placing bets on which stocks will rise or fall, many investors instead choose to “buy the market”, perhaps by investing in funds that mirror the composition of, say, the S&P 500. These “indexed” funds are attractive for many reasons: they generally have low fees, many are exchange-traded, and they typically perform well relative to actively-managed funds.

The amount of securities held by such funds has grown tremendously over time. The largest firms in this asset management space—BlackRock, Vanguard, and State Street—control more than $13T in securities. To put this in context, on average, each of those three firms now holds between 5% and 7% of every firm in the S&P 500, and so these three diversified firms combined hold almost 20% of the index that bills itself as a reflection of the US economy. As a result, most of the largest firms in the United States count those three firms as some of their largest shareholders. This begs the question: if your largest shareholders also hold large stakes in your competitors, should you be trying to maximize your own firm’s profits? Or, might your objective to be to maximize your investors’ portfolio values, which may imply that you value profits at your competitors to some degree? This alternative notion has become known as the “common ownership hypothesis.”

This underlying idea is not new: writings on this concept go back to Rotemberg (1986). Other important early contributions to this literature include Bresnahan and Salop (1986), which develops a first measure of this concept, the “MHHI”, and O’Brien and Salop (2000), which establishes a formal connection between common ownership and prices under different forms of competition (Cournot and Bertrand). What has really ignited this literature is the rich availability of data on common ownership that has resulted in empirical papers attempting to measure the effects of common ownership in product markets. The most relevant examples are Azar, Schmalz, and Tecu (2018) and Azar, Raina, and Schmalz (2016) which examine prices among airlines and banks as functions of common ownership. In a companion piece, we discuss the existing literature in depth, and why some approaches seem problematic.

Data to study this question are nominally public: all firms managing $100M or more in “13(f)” securities must file a quarterly report with their holdings to the Securities and Exchange Commission (SEC). However, as noted in our companion piece and elsewhere, there are significant issues with both the raw filings, and the commercial datasets that aggregate them.

Our paper explores the implications of the common ownership hypothesis among all S&P 500 firms from 1980-2017. We use a commercial dataset on institutional holdings for 1980-1999 and gather a novel dataset of institutional holdings for 2000-2017 based on concerns expressed by other parties as well as our own investigations of the available commercial datasets for this data. We first compute the implied “profit weights”—a measure of how much a dollar of my rival’s profits is worth to me once I internalize our overlapping investors—implied by the common ownership hypothesis for all S&P 500 firms for 1980-2017, excluding multi-class stock firms. The time-series trend is profound: from an average of roughly 0.2 in 1980 to 0.7 in 2017. To put this in perspective, the classic model says a profit weight on a competitor should be zero; if two firms were to merge, the profit weight would go to one. This analysis shows that the common ownership hypothesis predicts a tripling in the “integration” of the US economy over 37 years. If these incentives are reflected in firm behavior, they would imply a substantial lessening of competition among publicly-traded firms. Moreover, as we discuss in that paper, they may also give rise to the expropriation of undiversfied shareholders, a result which hearkens to the literature on “tunneling” (Johnson, La Pora, Lopez-de-Silanes and Shleifer 2000) and is particularly surprising in the world of the “widely-held firm” (Berle and Means, 1932).

From our research, we take away three additional, more nuanced findings that bear on how we have the discussion moving forward: 1) Existing research using the “MHHI” measure is misguided. These papers confound common ownership with market shares (and, therefore, market definitions). Market shares are functions of many things, including underlying demand patterns that have nothing directly to do with common ownership patterns, and so should not be conflated with them. The underlying object of interest is common ownership, and it should be studied directly. 2) Existing studies that make culprits of BlackRock, Vanguard and State Street ignore that the trend toward “indexed” investing both predates the rise of those firms, and has contributed greatly to the incentives predicted by the common ownership hypothesis. Profit weights rose from 0.2 to 0.5 from 1980–2000, before these asset management firms experienced their massive growth. 3) Existing studies that regress (mostly) prices on measures of common ownership interacted with market shares (e.g. “MHHID”) are tainted by the strong assumptions required: Cournot competition being the strongest example. In differentiated product markets, spurious correlations should exist between prices and measures such as “MHHID”, meaning that some of the results of existing empirical studies could be obtained whether or not the common ownership hypothesis is true. However, our own work shows that the incentives posited by the common ownership hypothesis are large: in our paper, we show that the implied (simulated) increase in markups attributable to the rise in common ownership is comparable to the rise in markups measured in De Loecker and Eeckhout (2017), which measures economy-wide markups using accounting data over a similar time horizon.

In conclusion, we find that the common ownership hypothesis presents an interesting challenge to economists and regulators. It is undoubtedly “big if true.” At the same time, studying it requires careful industry-by-industry analysis. Finally, we hope to see more work on the mechanisms by which investor wishes could be transmitted to managers. The rise of “index” investing has created phenomenal value for many middle-class investors, and any attempt to regulate these investment firms must trade off the potentially negative effects of these funds in product markets with the potentially positive effects they have in investment markets.

The complete paper is available for download here.

Both comments and trackbacks are currently closed.