Ownership, Governance and Investment *

Transcription

1 Ownership, Governance and Investment * Germán Gutiérrez and Thomas Philippon March 2017 Preliminary Abstract The US business sector has under-invested relative to Tobin's Q since the early 2000s; and the under-investment appears to be explained by decreasing competition (due to both increasing concentration and common ownership) and tight or short-termist governance [Gutiérrez and Philippon, 2016]. In this paper, we use a combination of natural experiments and instrumental variables to establish a causal relationship between increased quasi-indexer institutional ownership and decreased investment. In particular, we use the Russell index threshold as a natural experiment, and lagged quasi-indexer ownership as an IV. We nd that higher quasi-indexer ownership leads to higher buybacks and less investment. We then study the interaction between governance and competition in causing under-investment and nd contrasting results. At the rm-level, governance matters most for rms in non-competitive industries: they tend to buyback more shares and invest less. At the industry-level, anti-competitive eects of common ownership disproportionately aect industries that `appear' competitive according to traditional measures but actually are not (due to common ownership). In Gutiérrez and Philippon [2016], we show that investment is weak relative to measures of profitability and valuation, and that this weakness starts in the early 2000s. We argue that investment is not low because Q is low, but rather despite high Q; which rules out a long list of potential explanations such as low expected growth. We then use a mix of industry- and rm-level data to test whether under-investment relative to Q is driven by (i) nancial frictions, (ii) measurement error (due to the rise of intangibles, globalization, etc), (iii) decreased competition (due to technology, regulation or common ownership), or (iv) tightened governance and/or increased short-termism. We nd that proxies for competition and ownership explain the bulk of the investment gap, across * We are grateful to Janice Eberly, Olivier Blanchard, René Stulz, Boyan Jovanovic, Tano Santos, Charles Calomiris, Glenn Hubbard, Holger Mueller, Alexi Savov, Philipp Schnabl, Ralph Koijen, Ricardo Caballero, Emmanuel Farhi, Viral Acharya, and seminar participants at Columbia University and New York University for stimulating discussions. New York University New York University, CEPR and NBER 1

2 industries and across rms. Controlling for current market conditions, industries with less entry and more concentration (traditional or due to common ownership) invest less. Within each industryyear, the investment gap is driven by rms owned by quasi-indexers and located in industries with more concentration/more common ownership. These rms spend a disproportionate amount of free cash ows buying back their shares. The conclusions in Gutiérrez and Philippon [2016] are supported by discussions in policy and industry circles. 1 However, they are based on simple regressions and therefore cannot establish causality between competition, ownership and investment. In this paper, we (i) test the causality of increased quasi-indexer ownership on investment; and (ii) study the interaction between governance and competition in causing under-investment. A companion paper studies the causality between increased competition and investment [Gutiérrez and Philippon, 2017]. From a theoretical perspective, ownership can aect management incentives in three main ways: (i) by aecting the investment horizon of managers (i.e., by inducing short-termism); (ii) by tightening governance, leading to less over-investment; and (iii) by inducing anti-competitive eects through common ownership. We develop a simple model of Dynamic Competition to clarify these dynamics, to show that these three mechanisms aect investment in very dierent ways: improved governance aligns the (manager's) maximization problem with that of the shareholder's, thereby increasing the focus on long term value. Increased short-termism shifts the objective function of the maximization towards short-term value. Increased common ownership reduces incentives to compete. We are able to dierentiate the rst two from the third empirically; as common ownership diers from institutional ownership in the cross-section. However, we are unable to dierentiate between short-termism and governance. Improvements in governance reduce managerial entrenchment and require managers to continuously demonstrate strong performance, just as increased short-termism would. We simply test whether increases in (passive) institutional ownership lead to higher payouts and lower investment. We use largely the same dataset as Gutiérrez and Philippon [2016], which includes a mixture of rm- and industry-level data. These data are sourced (primarily) from Compustat, SEC (13f) lings, and the Bureau of Economic Analysis (BEA). We aim to explain aggregate-level underinvestment, which requires clean identication as well as external validity. As a result, we start by discussing Crane et al. [2016], which exploits a discontinuity in institutional ownership around the Russell index thresholds to establish a causal relationship between ownership and payouts. Crane et al. [2016] provides clean identication, but applies only to a limited set of (smaller) rms close to 1 For instance, the decline in investment and associated decrease in competition is discussed in Furman [2015] and CEA [2016], among others. In particular, CEA [2016] argues that several indicators suggest that competition may be decreasing in many economic sectors, including the decades-long decline in new business formation and increases in industry-specic measures of concentration. The rise of buybacks and associated implications for investment have been discussed by senior executives of major asset management companies, as well as the media. For instance, in his March 2016 letter to the executives of S&P 500 rms, Laurence Fink (BlackRock CEO) states that in the wake of the nancial crisis, many companies have shied away from investing in the future growth of their companies. Too many companies have cut capital expenditure and even increased debt to boost dividends and increase share buybacks. See also As Activism Rises, U.S. Firms Spend More on Buybacks Than Factories, The Wall Street Journal, May 26, 2015; The repurchase revolution and Corporate cocaine, The Economist, September 13, 2014, among others. 2

3 the index threshold. To gain external validity, we then use pre-2000 quasi-indexer ownership as an instrumental variable for post-2000 buybacks and investment, controlling for rm fundamentals. We nd that higher (pre-2000) quasi-indexer ownership indeed leads to higher (post-2000) buybacks. This approach is supported by the very high persistence of quasi-indexer ownership, even after controlling for rm characteristics such as market capitalization, protability, Q, etc. Next, we study the interaction between competition, ownership and investment, which yields two conclusions. First, anti-competitive eects of common ownership disproportionately aect industries that `appear' competitive according to traditional measures (e.g., have low Herndahls) but actually are not (because of common ownership). In other words, common ownership has a limited eect on industries that are noncompetitive according to traditional measures; but a substantial eect on industries that appear competitive according to traditional measures but have large amounts of common ownership. Second, the impact of ownership on governance and short-termism is most material for rms in noncompetitive industries (consistent with Giroud and Mueller [2010, 2011], among others). The remainder of this paper is organized as follows. Section 1 presents four recent trends in ownership and investment that drive our analysis. Section 2 discusses the mechanisms through which ownership may aect investment; as well as the interaction between ownership and competition. In doing so, we also review the relevant literature for the interaction of ownership, competition and investment. For a complete literature review, please refer to Gutiérrez and Philippon [2016]. Section 3 introduces a simple model of competition, with governance, short-termism and common ownership. We use this model to discuss and clarify the discussion in Section 2. Section 4 discusses the dataset used for empirical tests. Section 4.4 discusses the relevant identication strategies and test results to establish causality between ownership and investment. Section 5 discusses our results on the interaction between competition, ownership and investment. Section 6 concludes. 1 Four Facts about ownership and investment 1.1 Fact 1: Investment has Decreased Gutiérrez and Philippon [2016] provide ample evidence that investment is low relative to measures of protability and valuation, and that this weakness starts in the early 2000's. Gutiérrez and Philippon [2016] also use industry-level and rm-level data to test whether under-investment relative to Q is driven by (i) nancial frictions, (ii) measurement error (due to the rise of intangibles, globalization, etc), (iii) decreased competition (due to technology or regulation), or (iv) tightened governance and/or increased short-termism. They nd that proxies for competition and ownership explain the bulk of the investment gap, across industries and across rms. Controlling for current market conditions, industries with less entry and more concentration (traditional or due to common ownership) invest less. Within each industry-year, the investment gap is driven by rms owned by quasi-indexers and located in industries with more concentration/more common ownership. These 3

4 rms spend a disproportionate amount of free cash ows buying back their shares Fact 2: Institutional Ownership Has Increased Figure 1 shows the average share of institutional ownership across all US-incorporated rms in Compustat, by type. As shown, there has been a substantial increase in institutional ownership since 2000, primarily driven by growth in transient and quasi-indexer institutions. This is not shown in the gure, but the increase is particularly pronounced for smaller rms: since 2000, the dollar-weighted share of quasi-indexer institutional ownership increased from ~30% to ~45%, while the median share increased from ~15% to ~50%. That is, while the dollar-weighted quasi-indexer ownership increased by about 50%, it more than doubled for the median rm. Figure 1: Institutional ownership Average share of institutional ownership, by type year All institutions Dedicated Quasi Indexer Transient Notes: Annual data for all US incorporated rms in Compustat. Results are similar when including foreignincorporated rms. The vertical line in the rst graph highlights the passing of SEC rule 10b-18, which allows companies to repurchase their shares on the open market without regulatory limits. The increase in institutional ownership is linked to long term trends in stock ownership, as shown in Figure 2. The household share of ownership has declined drastically from more than 85% 2 Several papers study related issues. The decline in investment has been discussed in policy papers [Furman, 2015], especially in the context of a perceived decrease in competition in the goods market [CEA, 2016]; as well as academic papers (see, for example, Hall [2015]). Lee et al. [2016] nd that industries that receive more funds have a higher industry Q until the mid-1990s, but not since then. The change in the allocation of capital is explained by a decrease in capital expenditures and an increase in stock repurchases by rms in high Q industries since the mid-1990s. Alexander and Eberly [2016] study the implications of the rise of intangibles on investment. Last, Jones and Philippon [2016] explore the macro-economic consequences of decreased competition in a DSGE model with timevarying parameters and an occasionally binding zero lower bound. They show that the trend decrease in competition can explain the joint evolution of investment, Q, and the nominal interest rate. Absent the decrease in competition, they nd that the U.S. economy would have escaped the ZLB by the end of 2010 and that the nominal rate today would be close to 2%. Kose et al. [2017] study weak investment growth in emerging markets. 4

5 in 1960 to 35%. The decline was driven by private pension funds in the 1970s and 1980s; and the rise of mutual funds and ETFs since the 1990s. The latter part of the trend is due to the transition from dened-benet pension plans to dened contribution pension plans. Foreign ownership has also increased since Figure 2: Institutional ownership Notes: Annual data from Flow of Funds 1.3 Fact 3: Common Ownership has Increased If owners were fully disaggregated, institutional ownership would aect investment only through its impact on governance and short-termism. However, given the increased concentration in the asset management industry, common ownership of natural competitors by the same investors may introduce substantial anti-competitive eects. This is the subject of recent work linking common ownership to competition (see Section 2 for additional discussion of the literature). Figure 3 shows the mean Herndahl and Modied-Herndahl across all BEA industries in Compustat, where the Modied-Herndahl includes an adjustment to account for anti-competitive eects of common ownership. As shown, the mean Herndahl starts relatively high in the 1980s; decreases in the early 1990s as more rms go public and enter Compustat; and increases rapidly there after. The modied Herndahl increases even faster, from 25% to nearly 50%. The rise is primarily due to both the growth and concentration of institutional ownership. 5

6 Figure 3: Firm entry, exit and concentration Mean Herfindahl across industries (Compustat) Herfindahl year Mod Herfindahl Herfindahl Mod Herfindahl Note: Annual data from Compustat. 1.4 Fact 4: Payouts Have Increased Figure 4 shows the total buybacks and payouts for all US-incorporated rms in Compustat. As shown, there has been a substantial increase in total payouts, primarily driven by an increase in share buybacks. The increase starts soon after 1982, when SEC Rule 10b-18 was instituted (noted by the vertical line). Rule 10b-18 allows companies to repurchase their shares on the open market without regulatory limits. But it is most pronounced from approximately 2000 onward; where buybacks and payouts shift from 2-3% of assets to nearly 5% of assets. 6

7 Figure 4: Firm entry, exit and concentration Share Buybacks and Payouts year Payouts/Assets Buybacks/Assets Note: Annual data. 2 Ownership and Investment: Theoretical Background 2.1 Three mechanisms Ownership can aect management incentives in three main ways: (i) by aecting the investment horizon of managers (i.e., by inducing short-termism); (ii) by tightening governance, leading to less over-investment; and (iii) by inducing anti-competitive eects through common ownership. Short-termism. Regarding short-termism, some have argued that equity markets can put excessive emphasis on quarterly earnings, and that higher stock-based compensation incentivizes managers to focus on short term share prices rather than long term prots (see Martin [2015], Lazonick [2014], for example). Almeida et al. [2016] show that the probability of share repurchases is higher for rms that would have just missed the EPS forecast in the absence of a repurchase; and Jolls [1998] shows that rms which rely heavily on stock-option-based executive compensation are signicantly more likely to repurchase their stock. Given the rise of institutional ownership, an increase in market-induced short-termism may lead rms to increase buybacks and cut long term investment. 3 Governance. Regarding governance, a large literature following Jensen [1986] argues that conicts of interest between managers and shareholders can lead rms to invest in ways that do not maximize shareholder value. This does not necessarily imply that managers invest too much; they might invest in the wrong projects. The general view, however, is that managers are reluctant 3 In an unreported test, we used ExecuComp data to test whether rms with more stock-based executive compensation reduced investment more. We nd some, albeit limited support this hypothesis beyond our measures of competition and ownership. 7

8 to return cash to shareholders, and that they might over-invest. This view is supported byharford et al. [2008] and Richardson [2006], who nd evidence that poor governance associates with greater industry-adjusted investment. Thus, improvements in governance driven by changes in institutional ownership may lead to lower investment levels. Common ownership. As discussed above, the share and concentration of ownership has increased drastically since the 1960s. Institutions owned less than 10% of stocks in 1960, compared to nearly 65% today. And with the shift from dened benets to dened contribution plans, the asset management has become more concentrated. Azar [2012], Fichtner et al. [2016], Anton et al. [2016] discuss these trends in detail. Fichtner et al. [2016] show that the Big Three asset managers (BlackRock, Vanguard and State Street) together constitute the largest shareholder in 88 percent of the S&P500 rms, which account for 82% of market capitalization. This is the subject of a long theoretical literature in industrial organization, which argues that common ownership of natural competitors may reduce incentives to compete (see, for example, Salop and O'Brien [2000]). Azar [2012] shows that industries with higher levels of common ownership have higher markups on average. Azar et al. [2016b] and Azar et al. [2016a] use alternate identication strategies to show that this eect is empirically important for the U.S. Airline and the U.S. Banking industries. Azar et al. [2016b] shows that airline ticket prices increased as much as 10 percent because of common ownership of airlines; and Azar et al. [2016a] show that the increase in the concentration of bank ownership has led to higher fees and lower returns on savings. It is worth noting that the exact mechanisms through which common ownership reduces competition remain to be identied; but they need not be explicit directions from shareholders. They may result from lower incentives for owners to push rms to compete aggressively if they hold diversied positions in natural competitors; or from the ability of board members elected by and representing the largest shareholders to minimize breakdowns of cooperative arrangements and undesirable price wars between their commonly owned rms. See Salop and O'Brien [2000] and Azar et al. [2016b] for additional details. 2.2 Interaction with Activism The rise of institutional ownership, and associated implications for investment are intricately linked to the rise of activism. In fact, some have argued that activists ll a governance void created by the growth of `lazy' investors (i.e., index funds, exchange-traded funds and more broadly quasi indexers). 4 Brav et al. [2008] and Brav et al. [2010] 5 study hedge fund activism and nd that activist targets have signicantly higher institutional ownership. This is because activists typically hold minority stakes on their targets, hence they depend on the understanding and support of other shareholders to implement their desired changes. Figure 5 shows the number of activist events in the US from 1994 to Events were fairly 4 Capitalism's unlikely heroes, The Economist, Feb 7th, See also the updated results using a more recent sample [ ], but the same data collection procedure and estimation methods ( 8

9 low until the mid-2000s but rose extremely quickly thereafter. The pattern is strikingly similar to the rise of institutional ownership shown in Figure 1. Figure 5: Number of Activist Events in the US Source: JP Morgan (February 12, 2014) The implications of shareholder activism and institutional ownership on governance, shorttermism and payouts are studied in several papers. Regarding activism, Brav et al. [2008] and Brav et al. [2010] nd that target rms have lower capital expenditures pre-activism, and reduce them even further following the intervention. They also nd evidence for increased payouts in the year following the intervention, but this reverts back to normal levels two years after. Appel et al. [2016b] nd that larger ownership stakes of passive institutional investors make rms more susceptible to activists (increasing the ambitiousness of activist objectives as well as the rate of success). Regarding governance, Appel et al. [2016a] nd that passive owners inuence rms' governance choices (they lead to more independent directors, lower takeover defenses, and more equal voting rights; as well as more votes against management). And McCahery et al. [2016] survey institutional investors to better understand their role in corporate governance; and conclude that investors actively inuence governance choices. Related to quasi-indexers, they nd that long-term investors intervene more aggressively. 6 Crane et al. [2016] show that higher (total and quasi-indexer) institutional ownership causes rms to increase their payouts. In the end, it is unclear whether the higher payouts and the increased susceptibility to activist investors are evidence of tighter governance or increased short-termism. Some papers provide qualitative support for governance (e.g., Crane et al. [2016] refer to Chang et al. [2014] which argues that increasing passive institutional ownership leads to share price increases), but it is inconclusive. And other studies such as Asker et al. [2014] show that public rms invest substantially less and are less responsive to changes in investment opportunities than private rms. In the end, we are unable to dierentiate between these two hypotheses empirically. We simply test whether increases 6 It is worth highlighting that the evidence is not clear-cut: Schmidt and Fahlenbrach [2016] nd opposite eects for some governance measures (including the likelihood of CEO becoming chairman and appointment of new independent directors), and an increase in value-destructing M&A linked to institutional ownership. 9

10 in (passive) institutional ownership lead to higher payouts and lower investment. 2.3 Interaction between competition and ownership Since at least Adam Smith, economists have argued that managerial slack is primarily an issue for rms in non-competitive industries. Managers in such rms tend to enjoy the quiet life and therefore avoid generating excessive competition. By contrast, managers of rms in highly competitive industries are under constant pressure to improve eciency and are therefore less aected by managerial slack. They must constantly innovate to remain protable. Attempts to formalize the notion that competition mitigates managerial slack have proven dicult, but several recent papers have demonstrated it empirically (see, for example, Giroud and Mueller [2010, 2011]). The literature suggests that ownership, competition and investment can interact in two ways. On one hand, common ownership may aect industries dierently depending on their level of competition based on `traditional' measures. For instance, common ownership may materially reduce incentives to compete in industries that `appear' competitive according to traditional measures (e.g., have low Herndahls); yet have a limited eect on on industries that are already noncompetitive according to traditional measures. On the other hand, the impact of ownership on governance, short-termism and investment may be most material for rms in noncompetitive industries (as highlighted by Giroud and Mueller [2010, 2011], among others). We test these hypotheses in section 5. 3 A model of competition, governance, short-termism and ownership [TO BE ADDED] 4 Data We use the same dataset as in Gutiérrez and Philippon [2016]. The dataset includes a wide range of aggregate-, industry- and rm-level data. The data elds and data sources are summarized in Table 1. Sections 4.1 and 4.2 discuss the aggregate and industry datasets, respectively. Section 4.3 discusses the rm-level dataset, including key denitions. Firm- and industry-data are not readily comparable; they dier in their denitions of investment and capital, and in their coverage. As a result, we spent a fair amount of time simply reconciling the various data sources. We refer the reader to Gutiérrez and Philippon [2016] for details on the reconciliation and validation exercises. 10

11 Table 1: Data sources Primary datasets Additional datasets Data elds Source Granularity Aggregate investment and Q Flow of Funds US Industry-level investment and BEA ~NAICS L3 operating surplus Firm-level nancials Compustat Firm Institutional ownership Thomson Reuters 13F Firm Bushee's institutional investor Brian Bushee's website Institutional classication Investor 4.1 Aggregate data Aggregate data on funding costs, protability, investment and market value for the US Economy and the non nancial sector is gathered from the US Flow of Funds accounts through FRED. These data are used in the aggregate analyses discussed in Section 1; and to reconcile and ensure the accuracy of more granular data. 4.2 Industry data Industry-level investment and protability data including measures of private xed assets (currentcost and chained values for the net stock of capital, depreciation and investment) and value added (gross operating surplus, compensation and taxes) are gathered from the Bureau of Economic Analysis (BEA). Note that BEA I and K now include intangible assets (including software, R&D, and some intellectual property), not just tangible capital. Investment and protability data are available at the sector (19 groups) and detailed industry (63 groups) level, in a similar categorization as the 2007 NAICS Level 3. We start with the 63 detailed industries and group them into 47 industry groupings to ensure investment, entry and concentration measures are stable over time. In particular, we group detailed industries to ensure each group has at least ~10 rms, on average, from and it contributes a material share of investment (see Gutiérrez and Philippon [2016] for details on the investment dataset). We exclude Financials and Real Estate; and also exclude Utilities given the inuence of government actions in their investment and their unique experience after the crisis (e.g., they exhibit decreasing operating surplus since 2000). Last, we exclude Management because there are no companies in Compustat that map to this category. This leaves 43 industry groupings for our analyses. All other datasets are mapped into these 43 industry groupings using the NAICS Level 3 mapping provided by the BEA. We dene industry-level gross investment rates as the ratio of `Investment in Private Fixed Assets' to lagged `Current-Cost Net Stock of Private Fixed Assets'; depreciation rates as the ratio of `Current-Cost Depreciation of Private Fixed Assets' to lagged `Current-Cost Net Stock of Private Fixed Assets'; and net investment rates as the gross investment rate minus the depreciation rate. 11

12 Investment rates are computed across all asset types, as well as separating intellectual property from structures and equipment. 4.3 Firm-level data Dataset Firm-level data is primarily sourced from Compustat, which includes all public rms in the US. Data is available from 1950 through 2016, but coverage is fairly thin until the 1970s. We exclude rm-year observations with assets under $1 million; with negative book or market value; or with missing year, assets, Q, or book liabilities. 7 In order to more closely mirror the aggregate and industry gures, we exclude utilities (SIC codes 4900 through 4999), real estate (SIC codes 5300 through 5399) and nancial rms (SIC codes 6000 through 6999); and focus on US incorporated rms (see Section Gutiérrez and Philippon [2016] for more details on the sample selection). Firms are mapped to BEA industry segments using `Level 3' NAICS codes, as dened by the BEA. When NAICS codes are not available, rms are mapped to the most common NAICS category among those rms that share the same SIC code and have NAICS codes available. Firms with an `other' SIC code (SIC codes 9000 to 9999) are excluded from industry-level analyses because they cannot be mapped to an industry. Firm-level data is used for two purposes: rst, we aggregate rm-level data into industrylevel metrics and use the aggregated quantities to explain industry-level investment behavior. We consider the aggregate (i.e., weighted average), the mean and the median for all quantities, and use the specication that exhibits the highest statistical signicance. Second, we use rm-level data to analyze the determinants of rm-level investment through panel regressions. We compute a wide range of nancial measures, including investment, cash ow, operating surplus, etc. The main variables are discussed in the following section; with additional details on the sample selection, variable denitions and data quality tests available in Gutiérrez and Philippon [2016] Denitions Investment. We consider two main denitions of investment. First, the `traditional' gross investment rate is dened as in Rajan and Zingales [1998] specically, as capital expenditures (Compustat item CAPX) at time t scaled by net Property, Plant and Equipment (item PPENT) at time t 1. Net investment rate is calculated by imputing the industry-level depreciation rate from BEA gures. In particular, note that the depreciation gures available in Compustat include only the portion of depreciation that aects the income statement, and therefore exclude depreciation included in Cost of Goods Sold. For consistency, and because we are interested in aggregate quantities, we assume all rms in a given industry have the same depreciation rate, and compute the net investment rate as the gross investment rate minus the BEA-implied depreciation rate. We use 7 These exclusion rules are applied for all measures except rm age, which starts on the rst year in which the rm appears in Compustat irrespective of data coverage 12

13 the industry-level depreciation rate for structures and equipment only since it is applied to CAPX. Second, we proxy investment in intangibles with the ratio of R&D expenses to assets (Compustat XRD / AT) 8. We consider only the gross investment rate (i.e., do not subtract depreciation) since a good proxy for R&D depreciation is not available. Q. Firm-level Q is dened as the book value of total assets (AT) plus the market value of equity (ME) minus the book value of equity scaled by the book value of total assets (AT). The market value of equity (ME) is dened as the total number of common shares outstanding (item CSHO) times the closing stock price at the end of the scal year (item PRCC_F). Book value of equity is computed as AT - LT - PSTK. The resulting aggregate and mean Q from Compustat closely mirror the Flow of Funds Q. Buybacks. Last, we compute the ratio of share repurchases (item PRSTKC) to assets as a measure of buybacks. 9 Ownership. We gather data on institutional ownership from Thomson-Reuters' Institutional Holdings (13F) Database. This dataset includes investments in all US publicly traded stocks by institutional investors managing more than $100 million. We use these data to compute the MHHI (see below); as well as the share of institutional ownership; and the share of quasi indexer ownership. Quasi indexer ownership is computed following Brian Bushee's permanent classication of institutional investors, available on his website. Quasi indexers are institutional investors with diversied holdings and low portfolio turnover consistent with a passive, buy-and-hold strategy of investing portfolio funds in a broad set of rms. This category includes `pure' index investors as well as actively managed investors that hold diversied portfolios. It is therefore heavily inuenced by index position and participation (see, for example, Wurgler [2011] for more details). Quasi-indexers are the largest category, and account for ~60% of total institutional ownership. Appel et al. [2016a,b] and Crane et al. [2016] all use Bushee's classications when studying the implications of institutional ownership on governance and payouts. The classication is available from 1981 to Competition. Measures of competition typically aim to measure business dynamism (e.g., entry and exit), concentration (e.g., Herndahls) and/or market power (e.g., price-cost ratios). We consider versions of all such measures, 10 but use the `modied Herndahl' as our primary measure. The Modied Herndahl described in Salop and O'Brien [2000] and Azar et al. [2016b] is dened as 118 XRD set to zero if missing 9 We also considered total payouts to assets and found very similar results 10 In particular, we compute (i) the log-change in the number of rms in a given industry as a measure of entry and exit; (ii) the share of sales and market value held by the top 4, 8 and 20 rms in each industry; and (iii) the price-cost ratio (also known as the Lerner index). 11 According to the theory, it would better to compute MHHI = HHI + j k j sjs k i γ ij β ik i γ ij β ij, where γ ij denotes the control share of investor i in rm j. However, because data on the total number of voting shares per company is 13

14 MHHI = j s 2 j + j k j i s j s β ijβ ik k i β2 ij = HHI + HHI adj where s j and s k denote the share of sales for rms j, k in a given industry; and β ik denotes the ownership share of investor i in rm j.we refer to the rst term (HHI) as the `traditional' Herndahl; and to the second term (HHI adj ) as the `common ownership adjustment'. The second term follows a long theoretical literature in industrial organization that recognizes that common ownership of natural competitors by the same investors reduces incentives to compete (see, for example, Salop and O'Brien [2000]). Theoretical justication for the M HHI can be derived in a Cournot setting with common ownership. 12 See Salop and O'Brien [2000] and Azar et al. [2016b] for additional details. We focus on this measure because it exhibits robust correlations with investment, as discussed in Gutiérrez and Philippon [2016]. We consider the combined MHHI in most of our tests; but also separate HHI and HHI adj to assess their impact independently in some cases. We make two assumptions to compute this measure empirically: rst, because ownership data is only available for institutional investors, we compute β ik as the ownership share of investor i in rm j relative to total institutional ownership reported in the 13F database, not total ownership. This is not expected to substantially inuence the results because ownership by non-institutional investors is likely limited and restricted to a single rm, which does not induce common ownership links. Second, following Azar et al. [2016b], we restrict the data to holdings of at least 0.5% of shares outstanding. We use Compustat item SALE for measures of sales concentration and market value as dened in the computation of Q above for measures of market value concentration.we also compute age (from entrance into Compustat) and size (log of total assets) as rm-level controls. 4.4 Empirical tests and results Consider a rm with a given amount of free cash ow. It can either invest it, pay it out to shareholders (or continue to hold it as cash). 13 Thus, a rm that repurchases more shares, invests less almost by denition. We want to show that higher quasi indexer institutional ownership leads to higher buybacks and as a result lower investment. The results in Gutiérrez and Philippon [2016] establish a correlation between these eects i.e., that rms with higher quasi indexer ownership spend a disproportionate not readily available, we assume γ ij = β ik (i.e., we consider total ownership rather than voting and non-voting shares separately). 12 Azar et al. [2016b] and Azar et al. [2016a] use versions of the modied Herndahl and show that the eect of common ownership is empirically signicant for the U.S. Airline and the U.S. Banking industries as test-cases. 13 Formally, the nancing decit can be split into dividends, investment, changes in working capital and internal cash ow (see Frank and Goyal [2003]) 14

15 amount of free cash ows buying back their shares; and they also invest less. However, those results do not establish causality. The main problem with those regressions is that ownership, buybacks and investment all depend on current and expected nancial performance and investment opportunities. Investors even quasi indexers have some discretion on which rms to invest in. They may have requirements to track and/or benchmark against particular indices, but beyond that, their investment decisions are aimed at maximizing alpha (see, for example, Wurgler [2011]). Indeed, we can infer investor preferences by studying the characteristics of stocks with higher quasi indexer ownership. For instance, rms with lower leverage seem to have higher quasi indexer ownership, even after controlling for other rm- and industry- characteristics. Similarly, managers choose how much to invest and/or buyback depending on the nancial performance and investment opportunities of their rms (and, arguably, other private incentives that aect governance and short-termism, such as compensation). We therefore need to isolate the portion of buybacks that is driven by ownership, but is exogenous to nancial performance; and then study its implications for investment. We follow a two-pronged approach that relies on a mix of Natural Experiments and Instrumental Variables Natural Experiment: Russell Index threshold In particular, we start by discussing Crane et al. [2016], who exploit a discontinuity in institutional ownership around the Russell index thresholds to establish a causal relationship between ownership and payouts. In particular, note that Russell indices are re-constituted annually; and they result in dierential weights for rms around the 1000/2000 cuto (e.g., in 2005, the ten smallest rms in the Russell 1000 had a combined index weight of 0.004%, and the next ten largest rms were in the Russell 2000 with a combined index weight of 2.3%). These dierences in weights lead to sharp exogenous variation in institutional ownership; which can be used as an instrument for institutional ownership. 14 Crane et al. [2016] use a two-stage least-squares specication, where the rst stage models ownership as a function of index inclusion; and the second stage tests the eect of instrumented ownership on dividends, buybacks and payouts. They nd that more institutional ownership (and more quasi indexer ownership) indeed leads to higher payouts; and that this eect is economically important. Rather than replicate their analysis, we simply validate their conclusions for our sample. They report a coecient of 4.57 when regressing log-payouts on the percent of institutional ownership and 2.53 when regressing log-buybacks. 15 We run a simple OLS regression of log-payouts and logbuybacks on institutional ownership across all rms in our sample and obtain coecients of Indeed, this discontinuity has been exploited in a variety of papers such as Appel et al. [2016a,b] and Schmidt and Fahlenbrach [2016]. 15 Results for the ±100 rms around the threshold. They nd larger coecients (4.39 and 4.13, respectively) when considering ±750 rms around the threshold 15

16 and 3.68, respectively. These coecients are of the same magnitude; and indeed very close to those of Crane et al. [2016]. Considering only quasi-indexer ownership yields slightly larger coecients. The results in Crane et al. [2016] provide robust identication, but apply only to a small subset of rms/industries. Indeed they focus on the ±100 rms around the threshold; and provide results for ±750 rms cautioning that widening the bands can introduce bias. We therefore move to an instrumental variable approach that has weaker identication but applies to all rms Instrumental Variables: Lagged Ownership Empirical Strategy. To gain external validity, we use pre-2000 quasi-indexer ownership as an instrumental variable for post-2000 buybacks, controlling for industry, protability, Q and rm characteristics. The exclusion restriction is as follows. Exclusion restriction: Controlling for industry, protability, Q and rm characteristics, pre-2000 quasi-indexer ownership is uncorrelated with post-2000 expected growth and nancial performance This is supported by three facts: First, quasi indexer ownership is highly persistent. For instance, a regression of ownership at t on ownership ve years prior yields a coecient of 0.9, even after controlling for rm characteristics (including market value of equity). This suggests that quasi indexer investment decisions made before 2000 are likely to persist and inuence buybacks post- 2000; but are not heavily inuenced by changes in expected growth and nancial performance. This is particularly critical because of the second fact: the dot-com bubble crash substantially impacted expected growth and nancial performance. Investors made pre-2000 decisions at the height of the bubble, under optimistic growth prospects. Those growth prospects changed with the burst of the bubble, yet ownership remained generally stable. Last, one of primary mechanisms through which ownership aects buybacks is activism. Yet as Figure 5 shows, activism increased substantially only after Pre-2000 investment decisions could not have foreseen (or timed) the rise of Activism in the mid-2000s. Together, these facts suggest that pre-2000 quasi indexer ownership is indeed uncorrelated with post-2000 expected growth and nancial performance. It can therefore be used as an instrument for post-2000 buybacks, which in turn aects investment. Results. 16 Table2 shows our regression results. All regressions are based on the post-2000 period. The rst two columns instrument buybacks with pre-2000 quasi indexer ownership; and Q with the industry median stock Q to mitigate some of the measurement error. Column 3 then uses the portion of buybacks that is explained by ownership to predict investment. All regressions include industry and year xed eects; as well as a wide range of controls including measures of pre-2000 size, market capitalization, protability, cash ow, dividend payments, sales growth and leverage, 16 The limited time-period in our regression introduces a limitation to our results as they pertain to a particular period of time, over which investor preferences for buybacks may have been particularly high. 16

17 among others. As shown, higher pre-2000 quasi indexer ownership causes higher buybacks, which in turn cause lower investment. Note that we use the ratio of buybacks to assets as the dependent variable, instead of log-buybacks as used by Crane et al. [2016]. This explains the substantially lower coecient. The next regression (columns 4-6) adds the interaction between pre-2000 quasi indexer ownership and average buybacks across all rms in our sample. It shows that rms with higher quasi indexer ownership are more sensitive to aggregate trends in the level of buybacks. And given that the level of buybacks has increased drastically since 2000 (see Figure 1), this implies they increased their share buybacks even more. The last regression (columns 7-9) adds rm-level xed eects, but omits all pre-2000 controls because they would be absorbed into the xed eect. We do keep log-age as a control, and the interaction between quasi indexer ownership and aggregate buybacks. As shown, even including rm-level xed eects, rms with higher quasi indexer ownership exhibit higher sensitivity to aggregate buyback trends. Figure 6 supports these results: it shows the mean buyback rates across all rms, by quasi indexer ownership tercile (assigned within each industry and year). As shown, rms with higher ownership payout more; and exhibit far more sensitivity to aggregate trends. Figure 6: Buyback rate by quasi indexer ownership Mean Buyback/Assets year Low QIX High QIX Med QIX Notes: Annual data for all US incorporated rms in Compustat. Firm nancials from Compustat; ownership from Thomson Reuters and Brian Bushee's website. Figure 7 shows rst stage for the last regression graphically (columns 7-9). In particular, it plots the residual from regressing Buybacks/Assets on industry medianq, log-age, rm and year xed eects (as in the last regression), against the residual from regressing the interaction term QIX 96 99(i) BBA(t) on all other variables included in the columns 7-9. As shown, the slope is strongly positive and is not heavily inuenced by outliers. More quasi-indexer ownership increases the sensitivity of rms to aggregate trends in buybacks. 17

18 Table 2: Ownership IV: Regression results Table shows the results of rm-level 2SLS regressions of industry-level investment. Q instrumented by median industry Q; and buybacks instrumented by pre-2000 quasi indexer ownership. Firm-level controls include market capitalization, leverage, sales growth, dividends, protability, size...). Data sourced from Compustat. T-stats in brackets. + p<0.10, * p<0.05, ** p<.01. 1st Stage 2nd 1st Stage 2nd 1st Stage 2nd (1) (2) (3) (4) (5) (6) (7) (8) (9) Stock Q Buyb/Ass Net I/K Stock Q Buyb/Ass Net I/K Stock Q Buyb/Ass Net I/K Industry Median Q (t-1) 0.650** ** ** [21.46] [-0.56] [21.22] [-0.7] [25.47] [-0.33] % QIX owners(96-99) 0.279** 0.013** 0.771** ** [3.03] [4.32] [4.97] [-8.26] QIX96 99(i) BBA(t) ** 4.140** * 3.969** [-3.7] [13.05] [-2.36] [14.85] Stock Q (t-1) 0.048** 0.046** 0.046** [2.99] [3.15] [2.86] Buyback/Assets (t-1) * ** ** [-1.98] [-3.16] [-6.08] Pre-2000 rm-level controls Yes Yes No Year FE Yes Yes Yes Industry FE Yes Yes No Firm FE No No Yes Observations Between/OverallR % / 4.6% 28.4 / 10.0% 8.1% / 4.0% Only log-age is included as control. 18

19 Figure 7: IV rst stage regression results Notes: Figure shows rst stage for the last regression graphically (columns 7-9). Annual data for all US incorporated rms in Compustat. Firm nancials from Compustat; ownership from Thomson Reuters and Brian Bushee's website. See text for additional details. The above industry- and rm-level regression results show that higher quasi indexer ownership causes higher buybacks and lower investment. And Gutiérrez and Philippon [2017] shows that higher concentration also causes lower industry- and rm-level investment. These analyses consider each hypothesis independently, and therefore do not account for the interaction between ownership, competition and investment (beyond the inclusion of anti-competitive eects of common ownership in the modied Herndahl). This is the subject of the next section. 5 Interaction between competition and ownership As discussed above, the literature suggests that ownership, competition and investment can interact in two ways. First, common ownership may aect industries dierently depending on their level of competition based on `traditional' measures. Second, the impact of ownership on governance, shorttermism and investment may be most material for rms in noncompetitive industries (as highlighted by Giroud and Mueller [2010, 2011], among others). 5.1 Test results We start by testing the rst hypothesis at the industry-level. In particular, we regress industrylevel investment on the two components of the modied Herndahl the traditional Herndahl and the common ownership-adjustment as well as their interaction. The traditional Herndahl measures the `traditional' level of competition in an industry, while the common ownership-adjustment measures the anti-competitive eects of common ownership. Results are shown in Table 3. Column 1 is based on OLS, while columns 2-7 instrument for the 19

20 Herndahl using excess entry as dened in Gutiérrez and Philippon [2017]. The interaction term is not instrumented instead, we report reduced form results considering the Herndahl, excess entry and China import exposure as measures of competition for the interactions (columns 2, 3 and 4, respectively). As shown, the interaction coecient is nearly signicant in the rst test and signicant in the next two. Columns 5 to 7 instrument for common ownership using the common ownership adjustment as of Again, coecients are signicant and exhibit the right sign. To gain intuition about this result, consider the eect of substantial common ownership (75th percentile) on a competitive (25th percentile) and non-competitive (75th percentile) industry, as measured by the traditional Herndahl. The 25th and 75th percentile of the traditional Herndahl are 0.07 and 0.2, respectively; while the 75th percentile of the common ownership adjustment is The interaction term for the competitive industry is therefore 0.02, while for the non-competitive industry it is Applying the coecients in the rst IV regression (column 2), traditional competition implies a reduction in net investment of 1.4% for the competitive industry and 3.8% for the non-competitive industry; and the common ownership adjustment implies an additional 4.08% reduction for both industries. The interaction term reduces the impact of common ownership by only 1.2% in the competitive industry but by 3.2% in the non-competitive one. That is, the reductions amount to roughly 30% and 80% of the impact of common ownership, respectively. Thus, common ownership has a dampened eect on industries that are non-competitive to begin with; but a substantial eect on industries that appear competitive according to traditional measures but have large amounts of common ownership. Let's move on to the second hypothesis: that ownership (and therefore governance and shorttermism) aect investment primarily in noncompetitive industries. We test this at the rm-level, by adding interactions between the Herndahl and pre-2000 ownership to the rm-level regressions reported in Table 2 above. We use the point-in-time Herndahl rather than an IV because we are mainly interested in the eect of ownership on buybacks for a given level of competition irrespective of how that level of competition aroused. Table 4 shows the results. As shown, the interaction term is positive and signicant in column 2 suggesting that quasi indexer ownership disproportionately aects rms in non-competitive industries. The next regression (columns 4-6) adds the interaction between pre-2000 quasi indexer ownership and average buybacks across all rms in our sample. It adds rm-level xed eects, but omits all pre-2000 variables because they would be absorbed into the xed eect. It shows that rms with higher quasi indexer ownership are more sensitive to aggregate trends in the level of buybacks. And given that the level of buybacks has increased drastically since 2000 (see Figure 1), this implies they increased their share buybacks by more. This is true even controlling for time invariant rm-level characteristics through xed eects. We conclude that ownership has an outsized eect on investment for rms in non-competitive industries. OLS results (not reported) are consistent with those using IVs. 20