Anticompetitive effects of common ownership by blockholder bnstitutional investors

1

Anticompetitive Effects of Common Ownership by

Blockholder Institutional Investors:

Pierre Borst (10000352)

May 2017

MSc Economics

Thesis Supervisor: Prof. Dr. J. Seldeslachts

ABSTRACT

This thesis contributes to the existing research on the anticompetitive influence from common ownership, by theorizing that investors with stakes in multiple firms within an industry need to own a share higher than a certain threshold in at least one of their firms to exercise anticompetitive influence. By separating investors at the 5% ownership threshold, I find that the large investors indeed exercise a significant anticompetitive influence on competition, while the effects by small investors are negligible.

2

TABLE OF CONTENTS

8 SHORTCOMINGS AND SUGGESTIONS FOR FUTURE RESEARCH 22

3

1 INTRODUCTION

A long series of research states that product market outcomes can be negatively influenced by common ownership (the ownership of competing firms by the same investors)1_.

The studies theorize that in case of common ownership, a firm’s desired gain from aggressive competition – an increased market share – cannibalizes the market share of “competitors” who are part of the same investors’ portfolio and internalizes the adverse effects it inflicts on the other firms (Gilo et al., 2006). Therefore, shareholders will exert pressure to reduce each firm’s

incentives of pursuing a competitive strategy and push the market towards a monopolistic outcome. This implies an equilibrium with higher prices and lower output for consumers and a dead weight loss for the economy compared to the natural Cournot equilibrium (Azar et al., 2015).

Because of their active involvement in most of their investments’ management,

“institutional investors” are most commonly studied within the context of common ownership’s anticompetitive influence. Institutional investors are all professional investment agencies such as banks, investment funds, insurance funds and pension funds, who manage and invest other people’s (often individual investors) money (Gillan & Starks, 2003; Ferreira & Matos, 2006). Their increasing share in today’s economy is mentioned in almost every paper that studies their anticompetitive influence on competition. In the US, big institutional investors like BlackRock, Vanguard, Fidelity, and State Street owned around 80% of all stock in S&P 500 corporations in 2015 (Elhauge, 2016). Within Europe, the Netherlands hold the 4th _{place in the institutional}

assets ranking behind the UK, Germany and France, with € 1.631 Billion in assets in 20162_.

Institutional investors can pursue a blockholder status among their investments, which means that they acquire a share that is large enough to actively influence the firm’s

management (Brickley et al., 1988). The ownership threshold above which an institutional investor is considered to be a blockholder varies among the studied literature between 3% and 10%, but the most commonly used threshold is 5% (Thomsen, 2000). However, even though active interference in portfolio firms’ management by large investors in particular is at the groundwork of the common ownership anticompetitive influence theory, existing empirical research within this context does not distinguish between the anticompetitive pressure by blockholders and small investors. By my knowledge, in this field of research only Azar (2012)

1 _{See: Rotemberg (1984); Gordon (1990); Bolle & Güth (1992); Gilo (2000); Salop & O’Brien (2000); Gilo}

et al. (2006); Azar (2012); Azar et al. (2015); Azar et al. (2016); Azar (2016); Elhauge (2016)

4 mentions the existence of blockholders. He studies the number of common ownership linkages between any pair of firms (the number of common investors between any pair of competing firms) in his data by blockholders only, under 3 blockholder thresholds: a share in both firms larger than 3%, 5% or 7%. Yet, the models in all of his papers are regressed on a measurement of common ownership without any threshold. Here, all common ownership linkages by any size of investor are accounted for3_.

By using a single measurement for common ownership by all investors, thus not

separating the blockholders from smaller investors, the existing literature implicitly assumes that all investors proportionally exercise the same anticompetitive influence. The concept of

blockholders is an existing topic in economic literature, but in a separate field of research that focuses on blockholders’ influence on individual firm value and performance. Here, shareholders are found to be distinguishable and the distribution of share ownership is relevant for a firm’s performance (McConnell & Servaes, 1990). However, research in this field fails to look beyond the role of blockholder presence for the performance of individual firms, where it could elaborate on an industry scale to analyze the collusive effects of blockholders’ common ownership.

This thesis builds upon the mentioned theoretical and empirical research, and contributes to the existing literature by studying a new geographical market, by means of an analysis that questions the current implicit assumption that every investor of any size (thus not only

blockholders) exercises proportionally the same anticompetitive influence on the firm’s

management. The following four Dutch industries are used as research subjects: Chemicals and Allied Products, Rubber and Miscellaneous Plastic Products, Electronic and other Electrical Equipment, and Transportation Equipment4_{. I look at the influence of common ownership by}

blockholders and small investors on competition. This study will be dedicated to test the following hypothesis:

“Common ownership’s anticompetitive effects are driven by blockholders only, as small investors have a negligible impact”

3 _{Azae (2012); Azar et al. (2015); Azar et al. (2016) empirically study the relation between common}

ownership and industry markups.

5

2 LITERATURE REVIEW

Anticompetitive pressure by shareholders arises when there exist “common ownership linkages”: the situation where two or more competing firms share the same owners or investors (Merlone, 2001). The controlling/investing entity that “links” the competing firms through

shareholding can either be an (institutional) investor (Elhauge, 2016) or one of the competing firms itself (Gilo, 2000). Flath (1992) proves that common ownership may serve as a vehicle for collusion, by restricting production in Cournot industries towards a monopolistic outcome in order to maximize profit for the firms’ owners.

When there exists common ownership, the common investors are not necessarily required to have a dominant, active (influential) saying in every firm of their portfolio to establish an anticompetitive outcome. In his paper on competing firms who invest in one another, Gilo (2000) demonstrates that also passive investments (where the investing firm solely gains

financial interest and no influential role in the other firm’s management) can act as a facilitator of price increases and output decreases5_{. Having a financial stake in rival firms makes price cuts or}

other competitive measures less appealing for the investing firm, as any adverse effects on its rivals will be (partly) internalized: a price war will hurt both firms’ profits and any gain in market share will be cannibalized from the firms it has invested in (Gilo et al., 2006). The mentioned competition discouraging effects also turn passive investments into a credible non-price cutting commitment towards competitors who may expect price cuts (for any reason) and potentially start a price war on their own (Gilo, 2000). This makes passive investments a successful strategy to prevent future price wars by competitors and maintain (non-competitive) prices.

When all firms within an industry are partially owning one another, each individual firms’ total profits do not only depend on its own performance, but on the performance of the whole industry. These reciprocal financial interests among competitors facilitate tacit collusion (where firms implicitly collide, e.g. by avoiding any opportunity to cut prices): a price-cut would make every firm worse of in the long run, including the firm who has started the price war, whereas the latter might not be the case in the situation without cross-ownership ties (Gilo et al., 2006). Passive and active (where the investor gains an influential role in the invested firm’s

management) investments differ in their impact on competition. Passive investments affect the

5 _{Passive (or silent) investments among industry rivals are omnipresent in today’s economy. Some}

examples covered by the literature are the US and Japanese automobile industries (Alley, 1997), the Dutch financial sector (Dietzenbacher et al., 2000), the global steel industry (Gilo et al., 2006) and some cases of well-known firms: Microsoft’s $150 million investment in non-voting Apple stock and Gilette’s acquisition of 22.9% non-voting stock and 13.6% debt of its competitor Wilkinson Sword (Gilo et al., 2006)

6 incentives of the acquiring firm or institution (as described above), whereas active investments affect the incentives of the acquired firm: active investors with a market portfolio will discourage competitive moves by the acquired firm (Salop & O'Brien, 2000).

Reynolds & Snapp (1986) highlight the anticompetitive effect that arises in the situation where firms have financial interest in one another, but are initially operating in distinct product or geographical markets. Thus, these firms will only be actual rivals in case one of both firms decides to expand to the adjacent (product or geographical) market. In their model, reciprocal financial interests may serve as a successful deterrence strategy by prohibiting both firms from entering each other’s market, with monopoly prices and outputs in both markets as a result.

Malueg (1992) acknowledges the forces that facilitate collusion from common ownership, but he also theorizes a force with an opposing effect: he argues that punishments after cheating will we softer and less credible if the other firms hold interest in the cheater, therefore making cheating more tempting and collusion less likely. As a result, a high level of common ownership could theoretically lead to more competitive outcomes but he stresses that these forces’ net result is ambiguous and depends on the situation per industry. Malueg (1992) solely advocates the recognition of this potential anti-collusive force from common ownership, to spark further research and analysis. In response, Gilo et al. (2006) specifically contradict Malueg’s theory, when discussed in their introduction: “We believe that in practice, the anticompetitive effect is likely to dominate the anti-collusive effect, otherwise firms would have no incentive to invest in rivals” (Gilo et al., 2006, p. 83). Gilo et al.’s (2006) n-firm oligopoly model appears to be a more accurate representation of reality: Their model, in which firms do not have similar stakes in one another, allows for them to neutralize the anti-collusive effect and focus on the competition distorting effects, as opposed to Malueg’s (2012) symmetrical duopoly model where firms hold identical stakes in one another.

With the exception of Malueg (1992), economic literature agrees on the welfare reducing and cartelizing effects of common owenership in symmetric/asymmetric and

homogenous/heterogeneous Bertrand and Cournot markets6_{. Other papers elaborate on how}

anticompetitive mechanisms could even be reinforced7_{. Dietzenbacher et al. (2000) build upon}

the work of Flath (1992) on indirect ownership: the situation in which firm A holds stock in firm B, who holds stock in firm C, which makes firm A an indirect shareholder of firm C. Both papers

6 _{See: Reynolds & Snapp (1986); Farrell & Shapiro (1990); Flath (1992); Gilo, Moshe & Spiegel (2006);}

Shelegia & Spiegel (2011); Gilo et al. (2013).

7 show that the cartelizing effects of common ownership are even greater when firms are attentive of their indirect links. Gilo (2000) finds that anticompetitive effects of rival ownership may be even stronger when firm A’s controller (a firm or investor with a share larger than 50% in firm A) passively invests in competitor firm B, rather than when firm A itself passively invests in firm B. In particular, when the controller dilutes its stake in the controlled firm A (e.g. from 100% to 51%) while acquiring or maintaining a passive financial stake in firm B, it will place even more weight on its stake in firm B. Therefore, the smaller the stake of the controller in the firm it controls, the less aggressively it would cause the controlled firm to compete.

To study the specific effect of institutional ownership on competition, Fershtman & Judd (1987) have set up a principal-agent model in which a firm’s shareholders (principals) incentivize managers (agents) from their controlled firms according to their industry portfolio. This model replicates the reality in which shareholders elect their firm’s managers that best suit their strategic interest (managers who promise a reticent competitive policy) and where there exist questionable management compensations based on industry performance, rather than on firm performance (Azar et al., 2015). As a result, the outcomes from Freshtman & Jude’s (1987) model do not always correspond with the optimal outcome for each individual firm when it would be operating independently. This paradox in which some firms are worse off in a collusive outcome arises when some firms’ production is (partly) “sacrificed” by the common owners, in order to have joint profit maximization. Bolle and Güth (1992) provide an empirical example, by proving that a complicated hierarchy of mother-daughter relationship in the West German wholesale gas market permits the strategical closedown of certain firms or the withdrawal of supply from the market: “This finding stresses that one must be very skeptical when being told that some factories are unproductive and therefore will be closed down, as higher compensated gains from the competing firms’ output may be the true reason.” (Bolle and Güth,1992; p.232)

On the same subject, Azar (2016) states that when all shareholders hold identical market portfolios, they unanimously agree on joint profit maximization, thus an economy-wide monopoly outcome in a Cournot industry. Here, common ownership is used as a vehicle by which all investors simultaneously pursue the same agenda among their portfolio firms to establish the collusive outcome. Azar (2012; 2016) builds upon the issue of anticompetitive incentives that arise from investors’ market portfolios by developing oligopoly models with shareholder voting and utility functions for shareholders and managers. He shows that firms’ objectives are derived by aggregating their owners’ objectives (value maximization of the investors’ industry portfolios) through majority voting, rather than the commonly assumed objective of profit maximization for

8 the individual firm (Azar, 2012). Azar et al. (2015) show that large institutional investors’

executives often serve on the boards of their portfolio firms, which makes it even more plausible that these managers – elected by and representing the largest shareholders – are able to preserve cooperative arrangements and prevent the incidence of price wars between the commonly owned firms.

Azar (2012) shows the positive relationship between common ownership levels and markups in the US, that the density of shareholder networks in the US have more than doubled over the period 2000-2011 and that high levels of common ownership correlate with a higher probability that firm pairs share directors. The latter suggests that institutional investors have a significant influence on a firm’s management. Azar et al. (2015) and Azar et al. (2016) elaborate on the subject of institutional ownership and examine its exact effects on industry prices, where they find the anticompetitive effects of common ownership in respectively the US Airline and US Banking industry.

Azar et al. (2015) state that as investors’ size and the level of involvement in the firm’s management grow, the distinction between passive investment and active ownership becomes increasingly important. Shareholders are distinguishable and the distribution of share ownership is relevant for firm performance (McConnell & Servaes, 1990). A 49% stakeholder remains powerless if there exists a majority shareholder with a 51% share. Meanwhile, as ownership in many large firms is largely dispersed, a stake smaller than 50% can be sufficient to be effectively in control of the company. Courts have often deemed that with as little as 20% ownership,

shareholders can already be in control of the company (Ackman, 2016). In the context of

common ownership, it is relevant that whenever there exist multiple minority shareholders with a considerable share in more than one firm within an industry, they can form voting coalitions that can jointly control an anticompetitive outcome (Salop & O'Brien, 2000).

A frequently used definition for large investors who are actively involved in the management of their portfolio firms is “blockholders”. There is no consensus in economic literature on the exact ownership threshold from where an investor could be considered a

blockholder, but it ranges mostly from 3% to 10% with the most commonly found threshold being 5% (Thomsen, 2000). The concept of blockholders is an existing topic in economic literature but it focusses on blockholders’ role within individual firm performance rather than on

(anti-)competitive effects. When different types of investors are analyzed, blockholder investors are found to either have a positive or negative influence on individual firm value (Thomsen, 2000). Blockholders have greater power and more incentives to ensure value maximization for

9 shareholders, which causes them to actively align management objectives and reduce agency problems8_{. On the other hand, blockholder ownership may lead to short-term focused}

self-entrenchment and the expropriation of minority shareholders’ wealth by dominant blockholders9_.

Brickley et al. (1988) provide evidence that blockholders vote more actively on antitakeover amendments and that active opposition is greater among blockholders when a proposal from the management appears to harm them. Also, the monitoring of management increases

proportionally with the size of blockholders, as large investors have the strongest incentives to intervene with management despite the effort and costs involved (Gillan & Starks, 2003; Edmans & Manso, 2010).

However, nowhere does the literature (in the context of blockholders’ influence on firm performance) attribute any source of potential misalignment between firm objectives and shareholder interests to blockholders’ market portfolios, nor does it scale-up to industry level to analyze the anticompetitive consequences of blockholder activism.

3 THEORETICAL FRAMEWORK

Competition plays a central role in the field of industrial organization. A commonly used statistical measure of competition is the Herfindahl-Hirschman Index (𝐻𝐻𝐼), which measures market concentration as the sum of all firms’ squared market shares: 𝐻𝐻𝐼 = ∑𝑛𝑗=1𝑠𝑗, with 𝑠𝑗 as

the market share of firm 𝑗. Scholars and antitrust authorities use the index to analyze potential collusive product market outcomes and to (dis)approve proposed mergers and acquisitions (Rhoades, 1993). The underlying assumption of the 𝐻𝐻𝐼 is that all firms are fully separate and independent entities that are engaged in an oligopoly. By definition, the incompatibility of the 𝐻𝐻𝐼 with various ownership structures makes it an unsuitable statistical measure in the research context of common ownership.

Bresnahan & Salop (1986) developed a Modified Herfindahl-Hirschman Index (𝑀𝐻𝐻𝐼) for this reason, when studying the effects of several joint venture control and financial interest arrangements on industry structure, where the parent companies are incumbent industry firms. They derive 8 formulae to calculate the 𝑀𝐻𝐻𝐼 under a corresponding number of joint venture ownership structures10_{. Bresnahan & Salop (1986) present the 𝑀𝐻𝐻𝐼 as a useful tool for}

8 _{See: Jensen & Meckling (1976); Zeckhouser & Pound (1990); Burkart et al. (1997); Burkart et al. (1998)} 9 _{See: Fama & Jensen (1983); Morck et al. (1988); Shleifer & Vishny (1997).}

10 _{Bresnahan and Salop (1986) have derived MHHI formulae for the following ownership structures:}

10 analyzing the potential competitive effects of a proposed joint venture, analogous to how the 𝐻𝐻𝐼 is used for evaluating proposed mergers and acquisitions.

Salop & O’Brien (2000) have extended Bresnahan & Salop’s (1986) 𝑀𝐻𝐻𝐼 to a broader range of corporate control scenarios and multiple, overlapping ownership structures. Salop & O’Brien’s (2000) 𝑀𝐻𝐻𝐼 shows that actual market concentration could be significantly higher when common ownership linkages are accounted for, compared to the 𝐻𝐻𝐼 calculations that assume independent firms. Azar et al. (2015) use Salop & O’Brien’s (2000) 𝑀𝐻𝐻𝐼 to study the anticompetitive effects of common ownership in the US airline industry and show that MHHI levels are found to be 10 times larger than the HHI-limit beyond which the burden of proof would have shifted from the regulator to the involved firms to show that the implied concentration does not lead to the abuse of market power.

In this thesis, I will also use Salop & O’Brien’s (2000) 𝑀𝐻𝐻𝐼 to measure common ownership: 𝑀𝐻𝐻𝐼 = 𝐻𝐻𝐼 + ∑ ∑ (∑ γ_{∑ γ}𝑖 𝑖𝑗β𝑖𝑘

𝑖𝑗β𝑖𝑗 𝑖 )

𝑘 ≠𝑗

𝑗 𝑠𝑘𝑠𝑗. Here, γ is a measure of control and β

measures financial interest by investor 𝑖 in firms 𝑗 and 𝑘. Given that my data does not distinguish between vote (control) and non-vote (financial interest) shares, I equate γ𝑖𝑗 to β𝑖𝑗, as the number

of shares per investor 𝑖 in firm 𝑗 divided by the total amount of shares outstanding by the same firm: γ𝑖𝑗 = β𝑖𝑗=

𝑠ℎ𝑎𝑟𝑒𝑠 ℎ𝑒𝑙𝑑𝑖𝑗

𝑡𝑜𝑡𝑎𝑙 𝑠ℎ𝑎𝑟𝑒𝑠_𝑗. 𝑠𝑗 and 𝑠𝑘 are the market shares for firms 𝑗 and 𝑘, which are

calculated by taking the firm’s revenue as a share of total industry revenue: 𝑠𝑗 =

𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑗

𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝑅𝑒𝑣𝑒𝑛𝑢𝑒.

The numerator in the 𝑀𝐻𝐻𝐼 formula represents the sum of common ownership linkages by all investors that simultaneously hold shares in competing firms 𝑗 and 𝑘, for all possible firm pairs. To make the formula more intuitive, I name the numerator the Concentration of Ownership Across Firms (𝐶𝑂𝐴𝐹). The sum of 𝐶𝑂𝐴𝐹𝑗𝑘 values for all possible firm pairs (𝑗 and 𝑘) measures

how diversified the investors’ portfolios across the industry are, hence how strong the investors’ incentives are to pursue an “industry wide monopoly outcome”. I name the denominator the Concentration of Ownership Within Firm (𝐶𝑂𝑊𝐹), to measure the density of ownership within the firm. Analogous to how the 𝐻𝐻𝐼 for market concentration is higher when there are only a few large firms, the density of ownership is higher when there are only a few large investors. A high density of ownership within firm 𝑗, relatively to the 𝐶𝑂𝐴𝐹, makes it less likely that the involved

joint control, Partial merger, Competitor-based output formula, Transfer-price formula with control by distributing parent, Transfer-price formula with limited joint control.

11 investors would “sacrifice” this firm’s output for the sake of maximizing the value of their total industry portfolio, and vice versa.

By subtracting the 𝐻𝐻𝐼, I am left with the part of 𝑀𝐻𝐻𝐼 that measures the common ownership linkages, thus my variable of interest: delta MHHI (𝑑𝑀𝐻𝐻𝐼):

𝑑𝑀𝐻𝐻𝐼 = ∑ ∑ ∑ β𝑖 𝑖𝑗β𝑖𝑘 ∑ 𝛽𝑖 _𝑖𝑗2 𝑘 ≠𝑗 𝑗 𝑠𝑘𝑠𝑗 = ∑ ∑ 𝐶𝑂𝐴𝐹𝑗𝑘 𝐶𝑂𝑊𝐹𝑗 𝑘 ≠𝑗

𝑗 𝑠𝑘𝑠𝑗. First, I derive the 𝑑𝑀𝐻𝐻𝐼 by accounting for

all possible common ownership linkages as described above, which is the conventional method in this field of research: 𝑑𝑀𝐻𝐻𝐼𝑎𝑙𝑙. To test my hypothesis “Common ownership’s anticompetitive

effects are driven by blockholders only, as small investors have a negligible impact”, I compare

this method to the method in which I derive separate measurements for common ownership by blockholders and by small investors. As mentioned in section 2. Literature Review, the 5% ownership threshold is the most common level from where an investor is considered a blockholder among the studied literature. Also, I am able to find many common ownership linkages where investors have an ownership share of at least 5% in one of both firms per firm pair, so I will use the same 5% threshold to distinguish blockholders (investors with an

ownership share of at least 5% in one of both firms per firm pair) from small investors (investors with a share of less than 5% in both firms). This gives: 𝑑𝑀𝐻𝐻𝐼𝑏𝑙 = ∑ ∑

𝐶𝑂𝐴𝐹𝑗𝑘 𝐶𝑂𝑊𝐹𝑗|_𝛽≥0.05 𝑘 ≠𝑗 𝑗 𝑠𝑘𝑠𝑗 and 𝑑𝑀𝐻𝐻𝐼𝑠𝑖= ∑ ∑ 𝐶𝑂𝐴𝐹𝑗𝑘 𝐶𝑂𝑊𝐹𝑗|_𝛽<0.05 𝑘 ≠𝑗

𝑗 𝑠𝑘𝑠𝑗. Here, I expect to find an insignificant 𝑑𝑀𝐻𝐻𝐼𝑠𝑖 and a

significant 𝑑𝑀𝐻𝐻𝐼𝑏𝑙 that is also more efficient than the 𝑑𝑀𝐻𝐻𝐼𝑎𝑙𝑙 from my first set of regressions.

The reason why I require investors only to be a blockholder in one of the firms per pair among their portfolio firms is that they can exercise their influence in this firm, while the anticompetitive incentives come from their stake in the other firm where, however, they do not have a sufficient share of ownership to have any influence. This situation is analogous to Salop & O’Brien’s (2000) narration in the context of competing firms who actively (by acquiring a controlling share) and passively (by acquiring only a financial share) invest in one another. According to their paper, passive investments affect the incentives of the acquiring firm or institution, whereas active investments affect the incentives of the acquired firm. Here, active investors with a market portfolio will discourage competitive moves by the acquired firm. Whenever an investor is a blockholder in both firms in a firm pair, the common ownership linkages are accounted for in both firms (divided by the respective within firm concentration of ownership), as the investor asserts influence in both competitors.

12 After establishing the method for measuring and comparing common ownership, I require a dependent variable that captures competition. In industrial organization, competition is often measured by prices, marginal revenues and marginal costs. My dataset lacks these specific data, but I do have data on revenues and gross profit margins per firm. This margin consists of the firm’s gross profit (revenue – costs of goods sold) divided by its revenue. Under the

assumption that the goal of collusion is to have high profits above competitive level, I use profit margins as dependent variable to indicate if there might be collusion (while other factors are accounted for). Since I analyze the outcomes on industry level, I use the average gross profit margin per industry on a given date as dependent variable per observation. Average gross profit margins in industry 𝛼 are calculated as the sum of each firms’ (𝑗) market share 𝑠𝑗 and gross

profit margin 𝜋𝑗 product: 𝜋𝛼 = ∑𝑛𝑗=1𝑠𝑗 𝜋𝑗

4 DATA

In this thesis, I use a dataset from Thomson Reuters Global Dataset11 _and

macroeconomic data from the Centraal Bureau voor de Statistiek (CBS)12_{. The Thomson}

Reuters Global Dataset contains information of many Dutch companies in various industries and

about their investors: company names, industry sic codes, the largest 100 investors per firm and the amount of shares they hold on each date, each firm’s shares outstanding, number of

employees and performance indicators such as net sales, gross profit and close price per share. The total dataset contains information on a quarterly basis from 2003Q1 to 2015Q4. From CBS, I retrieved quarterly macro-economic data on GDP growth (against the same period in the previous year) and the level of unemployment in the Netherlands.

With this raw data, the variables of interest are derived. Unfortunately, some of the data points have missing observations. I had to drop the data points where observations were missing on gross profit, net sales and investor ownership, since these data are needed for the dependent variable average profit and the variable of interest 𝑑𝑀𝐻𝐻𝐼. Given that the

comparison between all investors and blockholders only – in their respective potential

anticompetitive effects – is the focus of this thesis, I require the presence of common ownership linkages by blockholders. This β𝑖𝑗 ≥ 0.05 restriction makes that I am left with fewer observed

11 _{Prof. Dr. Jo Seldeslachts shared the data for the purpose of this thesis} 12 _{http://statline.cbs.nl/}

13 common ownership linkages, compared to the situation where I would only look at the

unrestricted 𝑑𝑀𝐻𝐻𝐼𝑎𝑙𝑙. This leaves me with 1108 data points by 26 firms in 4 industries.

The data that will be used in the regressions consist of 125 aggregated observations from the mentioned data points on quarterly dates on industry level: Chemicals and Allied Products (52 observations), Rubber and Miscellaneous Plastic Products (36 observations), Electronic and other Electrical Equipment (26 observations), and Transportation Equipment (11 observations). The summary statistics for the variables are presented in Table 1.

Table 1: summary statistics

Variable Obs Mean Std. Dev. Min Max Variable definition avg_profit 125 .2493422 .0864617 .1187939 .5391032 Dependent variable: Average

gross profit margin per firm dMHHI_all 125 321.1256 316.8206 .1572316 2024.076 Common ownership by all

investors dMHHI_bl 125 263.3286 246.7578 .1095434 1227.399 Common ownership by

blockholders only dMHHI_si 125 57.79693 112.7394 .0452021 796.6773 Common ownership by small

investors HHI 125 3855.728 1087.307 2178.766 7898.588 Herfindahl-Hirschman

Index

lag_avg_profit 121 .2489413 .0866276 .1187939 .5391032 Lagged dependent variable

avg_employees 125 62477.79 30242.11 19697.5 129614 Average number of employees per firm avg_closeprice 125 45.48073 44.92859 8.125244 132.2543 Average share close price per

firm

gdp_growth 52 2.469231 2.533272 -4.1 6.7 GDP growth against the same period last year (%) unemployment 52 5.453846 1.178311 3.4 8.1 National level of

unemployment (%)

All 𝑑𝑀𝐻𝐻𝐼 values show large standard deviations in relation to their mean, which indicate substantial variability in common ownership levels across my data. Industry specific variables 𝑎𝑣𝑔_𝑐𝑙𝑜𝑠𝑒𝑝𝑟𝑖𝑐𝑒, and to a lesser extent 𝑎𝑣𝑔_𝑒𝑚𝑝𝑙𝑜𝑦𝑒𝑒𝑠, also show substantial variability. The same applies to 𝑔𝑑𝑝_𝑔𝑟𝑜𝑤𝑡ℎ, which however is unrelated to variability across

14 industries since it is a macroeconomic control variable. The variability in 𝑔𝑑𝑝_𝑔𝑟𝑜𝑤𝑡ℎ can be explained by the series of negative values in the crisis period 2009Q1-2010Q1 and the low values that follow, after high growth figures prior to the crisis of around 6%.

Table 2 zooms into the variables’ means and standard deviations on industry level and allows to make a point that is relevant for my hypothesis. Here, all standard deviations of 𝑑𝑀𝐻𝐻𝐼𝑠𝑖 are larger than their corresponding mean, except for industry 28. This suggests that

ownership by small investors is very volatile, hence most likely short-term oriented. Elaborating on this assumption of short term profit-maximization, common ownership linkages by small investors are more likely to be mere coincidences, rather than strategic positions. In contrary, 𝑑𝑀𝐻𝐻𝐼𝑏𝑙 is far less volatile across industries (again, with the exception of industry 28). This

allows to conclude that blockholders are more long-term orientated and specifically take strategic positions. This apparent different approach could be explained by the fact that

blockholders have an influential role that is sufficiently valuable not to abandon their shares for short-term profit maximization, as opposed to small investors. This potential explanation of influential blockholders and uninfluential small investors would be in line with my hypothesis Table 2: Variable means per industry

Industry Chemicals and Allied _Products Rubber and Miscellaneous _{Plastic Products Electrical Equipment} Electronic and other Transportation _Equipment Siccode 28 30 36 37 Industry obs # Firms included 52 11 36 4 26 8 11 3 Mean .1793803 Std. Dev. .0224761 Mean .35437 Std. Dev. .0603792 Mean .2133214 Std. Dev. .03345 Mean .3214836 Std. Dev. .0173971 avg_profit dMHHI_all 108.9418 85.16188 507.8949 314.4596 329.653 323.34 692.7753 297.2881 dMHHI_bl 82.02663 82.81668 416.8905 205.1605 249.8029 181.2019 649.7965 264.687 dMHHI_si 26.91522 16.45436 91.00438 150.545 79.85006 159.3891 42.97875 54.35845 HHI 3625.757 446.3299 4606.608 906.5907 2506.219 107.9291 5675.179 238.1135 lag_avg_profit .1797814 .022511 .3547515 .0612167 .21358 .0341133 .3197241 .0172758 avg_employees 94211.71 12763.72 50397.34 11838.79 24925.45 3757.932 40758.97 6067.53 avg_closeprice 13.45268 2.118352 115.1407 6.368826 19.22348 2.73303 30.97064 4.419508

5 ESTIMATION MODEL

The methodology of this study aims at comparing anticompetitive effects from common ownership between blockholders and small investors. For this reason, 𝑑𝑀𝐻𝐻𝐼 (in its 3 possible forms) is the explanatory variable of interest in my models and its effect will be measured on dependent variable 𝑎𝑣𝑔_𝑝𝑟𝑜𝑓𝑖𝑡. Using OLS and Panel regressions (random effects and fixed

15 effects), the anticompetitive effects of common ownership are first tested with the conventional variable 𝑑𝑀𝐻𝐻𝐼𝑎𝑙𝑙: OLS1, RE1 and FE1. Next, in OLS2 RE2 and FE2, the common ownership

measurement will be separated into 𝑑𝑀𝐻𝐻𝐼𝑏𝑙 and 𝑑𝑀𝐻𝐻𝐼𝑠𝑖 so either’s effect can be isolated.

The model is specified as follows:

𝑎𝑣𝑔_𝑝𝑟𝑜𝑓𝑖𝑡𝑖,𝑡 = 𝛽0+ 𝛽1 𝑑𝑀𝐻𝐻𝐼𝑖,𝑡+ 𝛽2 𝐻𝐻𝐼𝑖,𝑡+ 𝛽3 𝑎𝑣𝑔_𝑝𝑟𝑜𝑓𝑖𝑡𝑖,𝑡−1+ 𝛽4 𝑎𝑣𝑔_𝑒𝑚𝑝𝑙𝑜𝑦𝑒𝑒𝑠𝑖,𝑡

+ 𝛽5 𝑎𝑣𝑔_𝑐𝑙𝑜𝑠𝑒𝑝𝑟𝑖𝑐𝑒𝑖,𝑡 + 𝛽6 𝑔𝑑𝑝_𝑔𝑟𝑜𝑤𝑡ℎ𝑡+ 𝛽7 𝑢𝑛𝑒𝑚𝑝𝑙𝑜𝑦𝑒𝑛𝑡𝑡+ 𝜀𝑖,𝑡

Here, 𝑎𝑣𝑔_𝑝𝑟𝑜𝑓𝑖𝑡 is regressed on 𝑑𝑀𝐻𝐻𝐼 (𝑑𝑀𝐻𝐻𝐼𝑎𝑙𝑙 in version 1 of the every regression

; 𝑑𝑀𝐻𝐻𝐼𝑏𝑙 and 𝑑𝑀𝐻𝐻𝐼𝑠𝑖 simultaneously in each version 2), the Herfindahl-Hirschman index

(𝐻𝐻𝐼𝑖,𝑡), a lagged value of the dependent variable to account for autoregression

(𝑎𝑣𝑔_𝑝𝑟𝑜𝑓𝑖𝑡𝑖,𝑡−1), the average number of employees per firm (𝑎𝑣𝑔_𝑒𝑚𝑝𝑙𝑜𝑦𝑒𝑒𝑠𝑖,𝑡) and average

close price per share (𝑎𝑣𝑔_𝑐𝑙𝑜𝑠𝑒𝑝𝑟𝑖𝑐𝑒𝑖,𝑡) to control for industry characteristics in industry 𝑖, the

GDP growth versus the same period last year (𝑔𝑑𝑝_𝑔𝑟𝑜𝑤𝑡ℎ𝑡) and level of unemployment

(𝑢𝑛𝑒𝑚𝑝𝑙𝑜𝑦𝑒𝑛𝑡𝑡) to account for macroeconomic factors, and the error term 𝜀𝑖,𝑡. Except for the

macroeconomic variables, all variables are measurements or averages on industry level for industry 𝑖 on date 𝑡.

Table 3 demonstrates the correlations between the explanatory variables. By definition, 𝑑𝑀𝐻𝐻𝐼𝑎𝑙𝑙 shows a high correlation with 𝑑𝑀𝐻𝐻𝐼𝑏𝑙 and 𝑑𝑀𝐻𝐻𝐼𝑠𝑖, but these two

variables will not simultaneously be used with the former in any regression. The table shows no strong correlation between other variables, so we can safely assume that the model will not suffer from multicollinearity. This assumption is confirmed by the mean VIF score of 2.45 with all values ranging below 5, as threshold for potential multicollinearity problems is 10 (Dormann, 2013).

Table 3: explanatory variable correlations

dMHHI_all dMHHI_bl dMHHI_si HHI avg_employees avg_closeprice gdp_growth unemployment dMHHI_all 1 dMHHI_bl 0.9501 1 dMHHI_si 0.7307 0.4812 1 HHI 0.2859 0.3952 -0.0615 1 avg_employees -0.4451 -0.4859 -0.1872 0.0362 1 avg_closeprice 0.4169 0.4556 0.1745 0.5014 -0.3322 1 gdp_growth -0.0512 0.0731 -0.3038 0.1279 -0.3603 0.0019 1 unemployment -0.2172 -0.2115 -0.1473 0.1823 0.3026 -0.1404 -0.1911 1

16 A potential issue for the accuracy of the model might be the inclusion of lagged

dependent variable 𝑎𝑣𝑔_𝑝𝑟𝑜𝑓𝑖𝑡𝑖,𝑡−1 in the Panel regressions. Including an autoregressive term

as a control puts the remaining independent variables at risk of being biased towards small and insignificant coefficients, occasionally even with the wrong sign (Achen, 2001). In general, the necessity for a method to control for autoregression would favor a System GMM estimation, rather than an OLS or Panel regression, to obtain more efficient and less biased coefficients. System GMM estimates the parameters in a dynamic panel data model with unobserved individual specific heterogeneity by transforming the model into first differences. Then,

sequential moment conditions are used, with the lagged levels of the variables as instruments for the endogenous differences and the parameters estimated by System GMM (Bun & Windmeijer, 2009). The results and direction of the bias in OLS, Panel and System GMM depend on the level of persistency of the dependent variable (Soto, 2009). The autocorrelation of 𝑎𝑣𝑔_𝑝𝑟𝑜𝑓𝑖𝑡 in my data is 0.8956, which indicates a high persistency. In Soto’s (2009)

experiment, System GMM’s estimator is found to outperform OLS’ and Panel’s estimators in terms of bias and efficiency when persistency is high. However, the bias in the lagged dependent variable of the OLS estimator is considerably reduced in this context.

System GMM is specifically designed for panels with ‘Small T, large N’, since this method is able mitigate the bias from the lagged dependent variable’s coefficient by increasing N

(contrary to panel fixed effects). However, while my data contains a comfortable number of 52 dates (T), I have a low N of 4 industries. My low number of observations disqualify System GMM as a suitable estimator for my model, which is confirmed by the insignificant results for all

explanatory variables when experimenting with this method. Because of System GMM’s

unsuitability for my dataset, I will stick with the OLS and Panel regressions with the notation that System GMM might be preferred in future research with a more extensive dataset.

Another potential issue for the validity of my results could be the “reversed causality argument” (Azar et al., 2015). As the term suggests, the reversed causality argument states a direction of causality where (correctly foreseen) profitable markets attract investors, rather than that investors impose the anticompetitive dynamics as described in this thesis. The investors who are studied in this field of research are professional institutions who spend a lot of time, money and effort into modeling and forecasting the product demand in potentially emerging markets, which could explain the correlation between profitable industries and investor density. The reversed causality argument suggests that investors are simply drawn to markets in which

17 they correctly foresee an increase in product demand and profit margins, rather than that they assert pressure to enhance the profit margins themselves.

In previous research, this problem has successfully been mitigated by including

information on product or service quantities sold or delivered. Here, a decreased quantity in the product market outcome specifically counters the argument of a correctly foreseen increase in product demand (as quantity should be higher for this argument to hold), while the combination between a decreased quantity and increased profit margins indicate the existence of collusive forces (possibly by common ownership). An alternative method to oppose this argument is by performing a 2SLS regression with an instrumental variable for 𝑑𝑀𝐻𝐻𝐼 (Azar, 2015).

Unfortunately, my data lacks information on prices and quantities and the industries that are subjected to my research provide no suitable instruments for a 2SLS regression.

However, recall that the assumption of investors correctly foreseeing an increase in product demand and profit margins forms the basis of the reversed causality argument. In order for this argument to hold, (potentially) emerging industry should be proportionally attractive for ownership of any size, not just from the threshold above which one can assert influence himself (which I assume to be 5%, as explained in section 3. Theoretical Framework), as the argument specifically states that investors do not exercise any (anticompetitive) influence. Given that every investor from my data is a professional institution that analyzes the industries according to similar methods as its competitors, one would expect all investors to have access to comparable industry forecasts. Some investors just happen to be more actively involved in some industries than other investors, but in the end, all investors are supposed to act on the same incentives (profit maximization) and with access to the same information. This implied uniformity in

incentives and access to information among institutional investors – who happen to have either small or large ownership shares in any given industry – has led me to personally develop a new method to test the reversed causality argument.

I have already introduced the separate common ownership variables for blockholders and small investors (𝑑𝑀𝐻𝐻𝐼𝑏𝑙 and 𝑑𝑀𝐻𝐻𝐼𝑠𝑖) to test my main hypothesis, which will now also be

used as instruments to test for reversed causality. Taking the assumption that all investors act on the same incentives with similar information and the reversed causality argument (that influence on firm management size is irrelevant) together, one should expect to find the same coefficient for common ownership by small investors and by influential blockholders on 𝑎𝑣𝑔_𝑝𝑟𝑜𝑓𝑖𝑡. After all, the only difference between what is measured by these variables is

18 common ownership with (𝑑𝑀𝐻𝐻𝐼𝑏𝑙) and without (𝑑𝑀𝐻𝐻𝐼𝑠𝑖) the ability to influence the firm’s

management, which should be irrelevant according to the reversed causality argument.

Whenever the effect of 𝑑𝑀𝐻𝐻𝐼𝑏𝑙 on 𝑎𝑣𝑔_𝑝𝑟𝑜𝑓𝑖𝑡 (or, according to the reversed causality

argument, the other way around) turns out to be more significant and larger than the effect from 𝑑𝑀𝐻𝐻𝐼𝑠𝑖, there is reason to believe that the cause lies in the intrinsic difference between both

variables. If the results show that only blockholders are correlated with collusive outcomes, and if these results are allowed to be fully attributed to blockholders’ ability to influence the firms’ management, then they implicitly indicate that that this influence by investors is the source of causality. Hence, larger and more significant coefficients for 𝑑𝑀𝐻𝐻𝐼𝑏𝑙 on 𝑎𝑣𝑔_𝑝𝑟𝑜𝑓𝑖𝑡 than for

𝑑𝑀𝐻𝐻𝐼𝑠𝑖 give reason to reject the reversed causality argument.

6 RESULTS

The results for the estimations are presented in Table 4. A first look at the R-squared values shows positive feedback for the goodness of fit for all models, with values ranging from 0.7808 to 0.8481. Also, all models show high F-values and Chi2 statistics for overall

significance.

As mentioned before, the purpose of this thesis is to study the influence of common ownership on competition, while accounting for common ownership with different criteria. Models OLS1, RE1 and FE1 present the results for the regression of 𝑎𝑣𝑔_𝑝𝑟𝑜𝑓𝑖𝑡 on the conventional 𝑑𝑀𝐻𝐻𝐼𝑎𝑙𝑙 variable that measures all possible common ownership linkages, including small

investors. I recall that Azar (et. al) was able to find significant test statistics for a similar 𝑑𝑀𝐻𝐻𝐼 without distinguishing between small investors and blockholders in previous empirical research (2012, 2015, 2016). However, regressions with my data show no significant results for 𝑑𝑀𝐻𝐻𝐼𝑎𝑙𝑙

with respective p-values of 0.2601, 0.1192 and 0.6434. Here, a pooled measure of all common ownership linkages by any size of investor leads to an inefficient estimator.

Only when I split 𝑑𝑀𝐻𝐻𝐼 up into separate common ownership measurements for blockholders and small investors in OLS2, RE2 and FE2, I find significant results. OLS2, RE2 and FE2 each give a significant test statistic for 𝑑𝑀𝐻𝐻𝐼𝑏𝑙 at the 0.05 level with respective

p-values of 0.0199, 0.0000 and 0.0450. Meanwhile, 𝑑𝑀𝐻𝐻𝐼𝑠𝑖 complements OLS2 RE2 and FE2 as

representative variable for common ownership by small investors. This variable only shows a significant test statistic in model RE2, which is highly significant at the 0.01 level (p = 0.0005). However, the inconsistency with the other two models’ estimations, together with the Hausman

19 test results that suggest fixed effects (Chi2 = 20.59, p = 0.0044), raise doubts about the validity of this model’s outcomes. OLS2 and FE2 show insignificant test statistics for 𝑑𝑀𝐻𝐻𝐼𝑠𝑖, which

means that common ownership by small investors has no significant influence on 𝑎𝑣𝑔_𝑝𝑟𝑜𝑓𝑖𝑡 in these models.

Table 4: regression results Dep. Variable:

avg_profit

OLS1 RE1 FE1 OLS2 RE2 FE2

dMHHI_all .2092825 .2092825 .0277608 (1.13) (1.56) (0.51) dMHHI_bl .53402 .53402 .1167834 (2.36)** (7.00)*** (3.32)** dMHHI_si -.4896082 -.4896082 -.1394501 (-1.54) (-3.47)*** (-1.63) HHI .066292 .066292 -.1059572 .0360999 .0360999 -.1050325 (1.05) (0.68) (-3.50)** (0.56) (0.44) (-3.30)** lag_avg_profit .48764921 .48764921 .39020095 .48260724 .48260724 .3947287 (2.43)** (5.81)*** (6.71)*** (2.38)** (6.21)*** (6.85)***

avg_employees -2,03E-07 -2,03E-07 -5,62E-08 -1,49E-07 -1,49E-07 -7,10E-08

(-1.09) (-1.91)* (-0.33) (-0.79) (-1.85)* (-0.41) avg_closeprice .00055755 .00055755 .00037263 .00056397 .00056397 .00032488 (2.70)*** (2.64)*** (0.53) (2.75)*** (3.27)*** (0.45) gdp_growth -.0000269 -.0000269 .00002867 -.00071982 -.00071982 -.00013793 (-0.02) (-0.03) (0.07) (-0.49) (-0.50) (-0.37) unemployment -.0018696 -.0018696 .00091696 -.00153 -.00153 .00088149 (-0.50) (-0.47) (0.30) (-0.41) (-0.42) (0.29) _cons .09050959 .09050959 .17136962 .09561924 .09561924 .17234056 (1.94)* (5.33)*** (5.52)** (2.03)** (6.12)*** (5.49)** N 121 121 121 121 121 121 F Chi2 92.229125 190.77 16.27 111.48615 202.60 16.82 r2 .8417 .8417 .7808 .8481 .8481 .7898

legend: Coefficient / (t-value) / *p<.1; **p<.05; *** p<.01

Hence, isolating and separately regressing the common ownership linkages by blockholders and small investors shows an efficient estimator 𝑑𝑀𝐻𝐻𝐼𝑏𝑙 and an inefficient

estimator 𝑑𝑀𝐻𝐻𝐼𝑠𝑖, which confirms my hypothesis: the anticompetitive effects of common

ownership come from influential blockholders and the effect of small investors is negligible. Including common ownership linkages by small investors in the 𝑑𝑀𝐻𝐻𝐼 leads to a less efficient

20 estimator, which is demonstrated by comparing the insignificant 𝑑𝑀𝐻𝐻𝐼𝑎𝑙𝑙 with the significant

𝑑𝑀𝐻𝐻𝐼𝑏𝑙. A suest test allows to elaborate on the differences between 𝑑𝑀𝐻𝐻𝐼𝑎𝑙𝑙 and 𝑑𝑀𝐻𝐻𝐼𝑏𝑙,

by testing for a significant difference in their coefficient on 𝑎𝑣𝑔_𝑝𝑟𝑜𝑓𝑖𝑡. The test shows a chi2 value of 4.45 with a p-value of 0.0331 against the null hypothesis H0 that both coefficients are

the same. Therefore, I can conclude that the variables have significantly different coefficients at the 0.05 level. Also, the F-test and Chi2 values for overall significance are higher in the second version of every model. This suggests that the increased efficiency of the model comes from splitting the apparent inefficient unit of measure for common ownership (𝑑𝑀𝐻𝐻𝐼𝑎𝑙𝑙) into the

suspected relevant source of anticompetitive effects (𝑑𝑀𝐻𝐻𝐼𝑏𝑙) and a variable with negligible

impact (𝑑𝑀𝐻𝐻𝐼𝑠𝑖).

The coefficients of the significant 𝑑𝑀𝐻𝐻𝐼𝑏𝑙 should be interpreted as follows: 𝑎𝑣𝑔_𝑝𝑟𝑜𝑓𝑖𝑡

is positively correlated with 𝑑𝑀𝐻𝐻𝐼𝑏𝑙, so an increase in common ownership linkages within an

industry will result in a higher 𝑎𝑣𝑔_𝑝𝑟𝑜𝑓𝑖𝑡. Since the dependent variable 𝑎𝑣𝑔_𝑝𝑟𝑜𝑓𝑖𝑡 is the instrument by which I (and other scholars in a similar way) try to detect the existence of collusive forces, I interpret the significantly positive relationship as a proof that blockholders have an anticompetitive impact on product market outcomes. Higher 𝑑𝑀𝐻𝐻𝐼𝑏𝑙 levels can come from

higher ownership shares of an existing blockholder in its current portfolio firms, new investments by an existing blockholder in industry rivals of its portfolio firms, or by small investors acquiring a sufficiently large ownership share among their portfolio firms to be classified as a blockholder. All of these situations are expected to lead to anticompetitive outcomes.

Looking at the other explanatory variables, I find counterintuitive signs for the 𝐻𝐻𝐼

coefficients in FE1 and FE2, that are both negative and significant at the 5% level: -.1059572 (t= -3.50, p=0.0395) and -.1050325 (t= -3.30, p=0.0456). These result contradict traditional

economic theory that suggests a positive correlation between industry concentration and

markups. However, Azar et al. (2016) have also found a significant negative correlation between the prices for certain services in the US banking industry and the traditional HHI in models where the HHI is not corrected for ownership structures. Other notable signs are found in the negative coefficient for the average number of employees per firm (𝑎𝑣𝑔_𝑒𝑚𝑝𝑙𝑜𝑦𝑒𝑒𝑠) in all regressions, the negative coefficients for GDP (𝑔𝑑𝑝_𝑔𝑟𝑜𝑤𝑡ℎ) in all models except for FE1 and the positive

coefficients for the level of unemployment (𝑢𝑛𝑒𝑚𝑝𝑙𝑜𝑦𝑚𝑒𝑛𝑡) in the fixed effects models. Yet, except for the coefficients of 𝑎𝑣𝑔_𝑒𝑚𝑝𝑙𝑜𝑦𝑒𝑒𝑠 that are only significant at the 10% level in the random effects models, these coefficients are all insignificant.

21 In section 5. Estimation Model, I have already assumed that all institutional investors act on the same profit maximizing incentives and have access to the same forecasts in product demand, whether they are small investors among their portfolio firms or have a blockholder status. Taking this assumption together with the reversed causality argument, one should expect to find equal coeffiecients and significance among common ownership by blockholders (thus who – like any investor – do not exercise anticompetitive pressure, according to the reversed causality argument) and small investors, for the latter to be true. The fact that I have a positive, significant 𝑑𝑀𝐻𝐻𝐼𝑏𝑙 and a negative, insignificant 𝑑𝑀𝐻𝐻𝐼𝑠𝑖 sign tells me that the correlations

between common ownership by blockholders and 𝑎𝑣𝑔_𝑝𝑟𝑜𝑓𝑖𝑡 on the one hand and common ownership by small investors and 𝑎𝑣𝑔_𝑝𝑟𝑜𝑓𝑖𝑡 on the other, are not the same. The most sensible explanation is that blockholders’ are able to influence product market outcome through their portfolio firms, contrary to small investors. This does not only validate my main hypotheses, but it also counters the reversed causality argument which states that investors are simply drawn to correctly foreseen profitable markets and do not influence the product market outcome

themselves.

7 CONCLUSION

Economic theory leaves little doubt about the anticompetitive influence that institutional investors can have on industries. However, the empirical research on product market outcomes within this subject is limited, focusses on US markets and implicitly assumes that investors of any size influence product market outcomes, by not distinguishing between larger and smaller investors. By separating blockholders (investors with a share of at least 5% in one of their portfolio firms) from smaller investors, this thesis studies whether investor size at this specific threshold is a relevant factor for exercising the mentioned anticompetitive influence. The hypothesis “Common ownership’s anticompetitive effects are driven by blockholders only, as

small investors have a negligible impact” is tested on the following Dutch industries: Chemicals

and Allied Products, Rubber and Miscellaneous Plastic Products, Electronic and other Electrical Equipment, and Transportation Equipment.

The results confirm my hypothesis that blockholders are the source behind the collusive forces from common ownership and that the influence by small investors is negligible. The separate measurement of common ownership by small investors gives an insignificant estimator for its effect on average profits, correspondingly to the measurement of all common ownership linkages that includes these small investors. Meanwhile, the measurement of common

22 ownership by only blockholders gives an efficient estimator and shows significant positive effects on average profits. This indicates that only from a certain size, investors are able to exercise anticompetitive influence and including smaller investors in the measurement for common ownership leads to a less efficient estimator (even though other studies have managed to come up with significant results, despite including small investors). Also, the coefficient of the seperate common ownership measurement for blockholders only is significantly different from the

measurement for common ownership linkages by all investors.

The differences in sign and significance between the separate common ownership measurement for blockholders and small investors suggest that the reversed causality argument can be rejected: the coefficients for the influential and uninfluential investors should be similar to be in line with this argument’s reasoning (investors are drawn to – correctly foreseen – profitable markets and do not influence the product market outcomes themselves) and with the

assumption that all investors act on the same profit-maximizing incentives and with the same information.

Regarding policy implications, this study’s findings advocate that competition authorities should take a closer look at the density and distribution of blockholder investors across

companies within an industry. Actual industry concentration might be denser than the official firm boundaries suggest when active blockholders are found in multiple companies, with the

customary anticompetitive outcomes from a concentrated industry as a result.

8 SHORTCOMINGS AND SUGGESTIONS FOR FUTURE RESEARCH

This study’s major shortcomings are the limited amount of observations from only 4 industries and the fact that the data only covers Dutch industries, given that the Netherlands have an open economy with an important role for international trade. Many of the Dutch companies from my data make a share of their net sales from international sales, while the domestic market also consists of foreign companies that are absent from the data. These concerns impact the accuracy of the size and concentration of the Dutch domestic markets as I have defined them. Future research could mitigate these problems by studying a broader, more closed market, such as the European Economic Area (EEA). Studying a broader would market also increase the probability of finding horizontal interlocks between firms, which would provide more relevant common ownership levels. Subsequently, this would make a higher number of industries potentially suitable research subjects, rather than the 4 industries (of which 3 with a

23 limited amount of observations) that were available for this study, which in turn would allow for an analysis using System GMM.

Another possibility would be to conduct this research on a firm level. Here, one could study the effects of sharing common investors by competitors on each individual firm’s productivity and profit. This would leave more observations, since individual observations for every firm would be used here instead of industry averages. Also, this would allow to study the specific effects of across- and within firm density of investors more closely.

Another shortcoming is the lack of information on prices and the share of sales that comes from industries other than the company’s primary industry. Prices as the dependent variable act as a stronger competitive indicator than profit margins, as increased profit margins could also result from improved efficiency and better monitoring by the involved blockholders13_.

Information on secondary and tertiary (etc.) industry would result in more accurate estimations for industry size and concentration.

In further research with a more extensive dataset, one could study if there exists a non-proportional relationship between the size of the largest blockholders across multiple firms and the anticompetitive effects. Also, one could look for a way to prove that investors actively focus on and fortify a strategic blockholder position in multiple firms within their core investment industries. In general, applying this study to more geographical markets would help to validate the theory of common ownership’s anticompetitive effects, while it also allows to compare for differences in the effects among these markets. Potential variation in the magnitude of

anticompetitive effects from common ownership could be the result from differences in antitrust supervision, in investment legislation, or in blockholders’ density, role, and influence within and among firms. Analyzing the variation in anticompetitive effects that comes from differences in the mentioned characteristics would help us to isolate these characteristics and understand their individual (positive or negative) influence on competition even better. Finally, one could experiment with blockholder thresholds other than 5% to study if there exists a general size of ownership from which investors start to excersize influence, or if this might be industry or country specific.

13 _{See: Jensen & Meckling (1976); Zeckhouser & Pound (1990); Burkart et al. (1997, 1998); Gillan &}

24

REFERENCES

Achen, C. (2001). Why lagged dependent variables can suppress the explanatory power of other independent variables. University of Michigan.

Ackman, W. (2016). Pershing Square 2015 Annual Letter. Pershing Square.

Alley, W. (1997). Partial Ownership Arrangements and Collusion in the Automobile Industry. The

Journal of Industrial Economics, 191-205.

Azar, J. (2012). A New Look at Oligopoly: Implicit Collusion. Princeton University. Azar, J. (2016). Portfolio Diversification, Market Power, and the Theory of the Firm. Azar, J., Raina, S., & Schmalz, M. (2016). Ultimate Ownership and Bank Competition. Azar, J., Schmalz, M., & Tecu, I. (2015). Anti-Competitive Effects Of Common Ownership. Bolle, F., & Güth, W. (1992). Competition Among Mutually Dependent Sellers. Journal of

Institutional and Theoretical Economics, 209-239.

Bresnahan, T., & Salop, S. (1986). Quantifying the Competitive Effects of Production Joint Ventures. International Journal of Industrial Organization, 155-175.

Brickley, J., Lease, R., & Smith, C. (1988). Ownership Structure and Voting on Antitakeover Amendments. Journal of Financial Economics, 267-291.

Bun, M., & Windmeijer, F. (2009). The Weak Instrument Problem of the System GMM.

Universiteit van Amsterdam.

Burkart, M. (1997). Large shareholders, monitoring, and the value of the firm. The Quarterly

Journal of Economics, 693-729.

Burkart, M., Gromb, D., & Panunzi, F. (1998). Why Takeover Premia Protect Minority Shareholders. Journal of Political Economy, 172-204.

Dietzenbacher, E., Smid, B., & Volkerink, B. (2000). Horizontal integration in the Dutch financial sector. International Journal of Industrial Organization, 1223–1242.

Dormann, C. (2013). Collinearity: a review of methods to deal with it and a simulation study evaluating their performance . Ecography, 27-46.

25 Edmans, A., & Manso, G. (2010). Governance Through Trading and Intervention: A Theory of

Multiple Blockholders. The Review of Financial Studies, 2395-2428. Elhauge, E. (2016). Horizontal Shareholding. Harvard Law Review, 1267-1317.

Fama, E., & Jensen, M. (1983). Agency problems and residual claims. Journal of Law and

Economics, 327-349.

Farrell, J., & Shapiro, C. (1990). Asset Ownership and Market Structure in Oligopoly. The RAND

Journal of Economics, 275-292.

Ferreira, M., & Matos, P. (2006). The Colors of Investors’ Money: Which Firms Attract

Institutional Investors From Around the World? ISCTE Business School-Lisbon CEMAF. Fershtman, C., & Judd, K. (1987). Equilibrium Incentives in Oligopoly. The American Economic

Review, 927-940.

Flath, D. (1989). Vertical integration by means of shareholding interlocks. International Journal of

Industrial Organization, 369-380.

Flath, D. (1992). Horizontal shareholding interlocks. Managerial and Decision Economics. Gillan, S., & Starks, L. (2003). Corporate Governance, Corporate Ownership, and the Role of

Institutional Investors: a Global Perspective. University of Delaware.

Gilo, D. (2000). The Anticompetitive Effect of Passive Investment. Tel Aviv University Law

Faculty Papers.

Gilo, D., Moshe, Y., & Spiegel, Y. (2006). Partial cross ownership and tacit collusion. RAND

Journal of Economics, 81-99.

Gilo, D., Spiegel, Y., & Temurshoev, U. (2013). Partial Cross Ownership and Tacit Collusion

Under Cost Assymetries. Henry Crown Institute of Business Research.

Gordon, R. (1990). Do publicly traded corporations act in the public interest? Cambridge: National Bureau of Economic Research.

Hansen, R., & Lott, J. (1996). Externalities and Corporate Objectives in a World with Diversified Shareholder/Consumers. The Journal of Financial and Quantitative Analysis, 43-68. Jensen, M., & Meckling, W. (1976). Theory of the firm: Managerial behavior, agency costs and

26 Malueg, D. (1992). Collusive behavior and partial ownership of rivals. International Journal of

Industrial Organization, 27-34.

McConnell, J., & Servaes, H. (1990). Additional evidence on equity ownership and corporate

value. North-Holland: ournal of Financial Economics.

Merlone, U. (2001). Cartelizing Effects of Horizontal Interlocks.

Morck, R., Shleifer, A., & Vishny, R. (1988). Management ownership and firm value: an empirical analysis. Journal of Financial Economics, 293-315.

Reynolds, R., & Snapp, B. (1986). The competitive effects of partial equity interests and joint ventures . International Journal of Industrial Organization, 141-153.

Rhoades, S. (1993). The Herfindahl-Hirschman Index. Fed. Res. Bull.

Rotemberg, J. (1984). Financial Transaction Costs and Industrial Performance. Massachesetts: Massachesetts Institute of Technology.

Salop, S., & O'Brien, D. (2000). Competitive Effects of Partial Ownership:.

Shelegia, S., & Spiegel, Y. (2011). Bertrand Competition When Firms Hold Passive Ownership Stakes in One Another.

Shleifer, A., & Vishny, R. (1997). A survey of corporate governance. Journal of Finance, 737-783.

Soto, M. (2009). System GMM estimation with a small sample. Barcelona Economics Working

Paper Series.

Thomsen, S. (2000). Blockholder Ownership, Dividends and Firm Value In Continental Europe. Copenhagen: Department of International Economics and Management.

Yi, S.-S. (1992). Asset ownership and market structure in oligopoly: Further results. Economics

Letters, 437-442.

Zeckhouser, R., & Pound, J. (1990). Are large shareholders effective monitors? An investigation