How much value can blockholders tunnel? Evidence from the Bulgarian mass privatization auctions

ARTICLE IN PRESS

Journal of Financial Economics 76 (2005) 191–234

0304-405X/$

doi:10.1016/j

$I would

numerous su

improved the

Gatchev, Stu

Nenova, Tim

Corporate G

Doctoral Stu

and Queen’s

Center for M

study. Janice�Fax: +1

E-mail ad

www.elsevier.com/locate/econbase

How much value can blockholders tunnel?Evidence from the Bulgarian mass

privatization auctions$

Vladimir Atanasov�

Babson College, Babson Park, MA 02457-310, USA

Received 22 April 2003; received in revised form 8 April 2004; accepted 24 May 2004

Available online 6 January 2005

Abstract

This study examines the emergence of the Bulgarian stock market and the role of controlling

blockholders. A new approach using mass privatization auction data measures the premium

for control and demonstrates that, in the absence of legal constraints, majority owners extract

more than 85% of firm value as private benefits of control. Institutional investors form

portfolios of predominantly controlling positions or participate in majority coalitions.

Ownership stakes cluster at 51%. After the privatized companies begin trading on the

Bulgarian Stock Exchange, majority-owned firms trade at 40–60% discounts. Overall, the

results support the Fama and Jensen (J. law Ecorg 26 (1983) 301) view that majority-owned

- see front matter r 2004 Elsevier B.V. All rights reserved.

.jfineco.2004.05.005

like to thank Clifford Holderness, Ian Domowitz, Chris Muscarella, and Dennis Sheehan for

ggestions and helpful discussions. The comments of two anonymous referees greatly

paper. Thanks are also due to Audra Boone, Conrad Ciccotello, Andrei Evtimov, Vladimir

art Gillan, Gordon Hanka, Oliver Hansch, Laurie Krigman, Harold Mulherin, Tatiana

Simin, David Stolin, Nikola Zikatanov, and participants at the Conference on International

overnance at The Tuck School of Business, Dartmouth College, the Penn State Finance

dent Seminar, and seminars at Babson College, Baruch College, The Federal Reserve Board,

University. I am grateful to Rumen Sokolov, Borislava Nedelcheva, Kamen Dikov, and the

ass Privatization, Sofia, Bulgaria for providing me with important parts of the data for this

Willett provided expert editorial assistance. All remaining errors are mine.

781 239 5004.

dress: [email protected] (V. Atanasov).

www.elsevier.com/locate/econbase

ARTICLE IN PRESS

V. Atanasov / Journal of Financial Economics 76 (2005) 191–234192

firms cannot persist as publicly traded corporations if the expropriating activities of

controlling blockholders are not legally restricted.

r 2004 Elsevier B.V. All rights reserved.

JEL classification: G31; G34; D44; L33

Keywords: Private benefits of control; Large shareholders

1. Introduction

A majority shareholder can have two very different effects on firm value andminority shareholder wealth. The first effect arises because, compared to smallshareholders, a majority shareholder has stronger incentives and more opportunitiesto monitor management and improve firm performance. The monitoring activities ofthe large shareholder can lower agency costs in the firm and increase cash flows forall shareholders. Shleifer and Vishny (1986) were among the first to discuss the roleof the large shareholder as a monitor who creates shared benefits for allequityholders. This favorable view of blockholders is further developed byGrossman and Hart (1988) and Harris and Raviv (1988), among a long line ofpapers.The second and opposite effect occurs when a controlling shareholder can capture

monetary or other private benefits of control, sometimes at the expense of smallershareholders. As Fama and Jensen (1983) note, a majority shareholder is insulatedfrom the market for corporate control, and commands both decision management(initiation and implementation of decisions) and decision control (monitoring andratification). Fama and Jensen argue that firms with a majority shareholder cannotexist as publicly traded corporations in equilibrium. The rationale is that anunconstrained majority owner can fully expropriate all assets and cash flows of thecompany. Small investors will not purchase shares in such companies, which willultimately be 100% owned by the majority holder. There are many theoretical papersin the corporate governance literature that adopt this view of blockholders (Burkartet al., 1998; Bennedson and Wolfenzon, 2000).Due to the proliferation of concentrated firm ownership and controlling

shareholders in most countries (La Porta et al., 1999), it is important in the studyof corporate governance to determine which effect dominates in a particular setting.Previous empirical studies have examined this question, but have come up withconflicting results. On one hand, Holderness and Sheehan (1988, 2000) studymajority shareholders in U.S. publicly traded corporations and provide empiricalevidence that broadly supports the monitoring view of blockholders and contradictsthe expropriating view. DeAngelo et al. (1984) also find that large investors andminority shareholders share most of the gains in LBO transactions, and Mitton(2002) shows that outside blockholders improve firm performance during the Asiancrisis. On the other hand, a large literature focusing on the value of voting rightsdocuments that large investors enjoy private benefits of control. In particular, studies

ARTICLE IN PRESS

V. Atanasov / Journal of Financial Economics 76 (2005) 191–234 193

of large block transactions and dual-class share companies report a substantialcontrol premium (Dyck and Zingales, 2004; Nenova, 2002), which is more consistentwith expropriation by controlling shareholders than with a monitoring role.The conflicting evidence arises in part because the net effect of majority

shareholders on firm value can heavily depend on the legal structure in theparticular market. The Fama and Jensen (1983) conclusion directly relies on theassumption that there are no legal constraints on the expropriating activities of amajority shareholder. This premise is violated in most markets, because there istypically at least some level of minority shareholder protection. Adequate legalconstraints are a sufficient condition to derive the results in Shleifer and Vishny(1986) and the other monitoring models, although other mechanisms like reputationcan also motivate large shareholders to increase value for all shareholders, even inmarkets with no legal protection.This paper presents new evidence as to which view of majority shareholders better

describes the reality of a developing market that fits the Fama and Jensen (1983)setting. The specific market studied is the Bulgarian stock market, because the legalsystem in Bulgaria imposes few constraints on the behavior of controllingshareholders. Also, in contrast to the U.S. treatment of institutional investorsdocumented by Roe (1990), Bulgarian law does not prohibit large institutionalinvestors from taking controlling positions in the firms in their portfolios.A lack of legal protection is not unique to the Bulgarian market. Black et al. (2000)

and Glaeser et al. (2001) show that the Russian and Czech markets fit thatdescription as well. What makes the Bulgarian stock market particularly well suitedfor a study of the effect of controlling blockholders is the existence of rich data. In1996, the Bulgarian government created more than 1,000 corporations through aprocess of mass privatization. In the majority of these firms, a controlling block wasauctioned off by the government to a large number of potential investors. Thisprocess produced wide cross-sectional variation in ownership structures. Regardlessof their ownership structures after the mass privatization auctions, however, all firmswere ultimately listed as public companies and traded on the Bulgarian StockExchange (BSE). The Bulgarian mass privatization process thus affords an out-of-equilibrium setting to analyze the emergence and behavior of controlling share-holders in public companies.The mechanism of the Bulgarian mass privatization auctions generated different-

sized bids at different prices for shares in the same company. The availability of suchdata provides a natural means of estimating the premium that investors are willing topay to acquire control. This paper presents a new methodology based on theseindividual auction bids for measuring the control premium, while accounting forsome of the selection biases in previous studies. The magnitude of the controlpremium is then a direct indicator of whether investors expect majority shareholdersto increase firm value through monitoring or to expropriate minority shareholderwealth for their own private benefit.The results reveal that the behavior of large shareholders in Bulgaria closely

resembles the predictions of Fama and Jensen (1983). Shares in controlling blocksare priced up to ten times higher than shares in small blocks. A control premium of

ARTICLE IN PRESS


this magnitude suggests that controlling shareholders can expropriate or ‘‘tunnel’’ asmuch as 85% of firm value. After the completion of the mass privatization process,concentrated firm ownership with a majority owner (a single investor or a coalition)becomes the norm. Moreover, the initial holdings of the majority owners cluster at51%, which is the minimum percentage of firm equity required for absolute control.Virtually all institutional investors in Bulgaria either form portfolios of controllingblocks or participate in majority coalitions. The conclusions about the effect ofcontrolling shareholders on firm value are further supported by stock marketvaluations. After the privatized companies start trading on the Bulgarian StockExchange, shares of companies with a majority shareholder have 40–60% lowermarket-to-book ratios than government-owned companies or companies withcontested control.The remainder of the paper is structured as follows. The next section describes the

institutional details of the Bulgarian mass privatization scheme and the subsequentdevelopment of the Bulgarian stock market. Section 3 presents the data and reportsunivariate statistics. Section 4 introduces the econometric model and providesestimates of the value of control using auction bids. Section 5 analyzes the effect ofdifferent ownership structures on stock market valuations and presents anecdotes oftunneling. Section 6 concludes. Details about the econometric methodology arepresented in the Appendix.

2. The Bulgarian mass privatization

2.1. Design of the mass privatization process

In 1996, seven years after the fall of the Socialist regime, the Bulgariangovernment still controlled more than 90% of the assets in the economy. Thegovernment was under heavy political pressure to privatize these assets and tocomplete the transition to a market economy. Following the example of the CzechRepublic, the Bulgarian authorities concluded that mass privatization not onlywould be a way to privatize assets quickly, but would also distribute these assets to alarge number of individual investors, which would possibly result in a more liquidstock market. A liquid stock market would be the best advertisement for the successof the transition and would attract foreign direct and portfolio investors. Perhapsmore important, the transfer of ownership from state to private hands wouldimprove the governance of the enterprises and make them more profit-oriented andthereby more efficient, as claimed universally by economists around the globe.1 Thegovernment concluded that distributing assets essentially for free would be a smallprice to pay relative to the political and economic merits of mass privatization, andcommenced the process in mid-1996. To ensure that the scheme for mass

1See Atanasov (1997) for a direct exposition of the goals of mass privatization from the executive

director of the Center for Mass Privatization, Sofia, Bulgaria. For outside analysis of the reasons for mass

privatization see Mitrev (1997) and Pamouktchiev et al. (1997).

ARTICLE IN PRESS


privatization would be designed with the highest technical expertise, the Bulgarianauthorities employed the Barents Group/KPMG under the advisory and financialsupport of USAID (Lloyd and Nenov, 1997).The mass privatization was initiated by allowing every Bulgarian voter to acquire

25,000 vouchers for a nominal fee. These vouchers could not be traded for cash andwere the only acceptable currency for bidding on the shares of 1,040 privatized firmsoffered in three auction rounds. The government adhered to the precept of moderneconomic theory that an auction is an efficient and fair mechanism for allocatingownership. It went even further in ascertaining the fairness of the process bychoosing a discriminatory auction with a reserve price, because theory suggests thatsuch auctions are less prone to collusion. Each auction round was held as a set ofsimultaneous (one for each firm) sealed-bid, discriminatory (‘‘pay-your-bid’’)auctions. Any unsold shares at the end of a round were offered in the next roundwith a lower reserve price. All vouchers submitted in unsuccessful bids were returnedto investors and could be used in subsequent rounds. In the first round, 968 firmswere offered; 64 additional firms were offered in the second round, and another eightfirms were introduced in the third. Iliev (1997) provides further description of theauction scheme and motivation of the choice of mechanism and implementationdetails.Close to 80% of the firms offered more than 51% of their equity for mass

privatization. The remaining 20% of the firms – some of the largest and mostimportant Bulgarian state-owned enterprises – offered from 10% to 45% of theirshares to the public. The remaining equity in these firms was retained by thegovernment to be sold for cash at a later date. This process of partial privatization ofthe most valuable Bulgarian enterprises matched the framework of Zingales (1995)that prescribes how ownership in companies with large control benefits should bedivested. The overall book value of the 1,040 firms was 210 billion Bulgarian levs, ornearly half of the state-owned assets eligible for privatization (Center for MassPrivatization, 1997). The book value of the shares offered for sale in the auctions was90 billion Bulgarian levs. If all shares were sold, the mass privatization process wouldhave resulted in the government transferring to private ownership 12% to 14% of allstate-owned assets (Miller and Petranov, 2000).To gain support for the mass privatization process, 10% of each firm’s shares

allocated for mass privatization were offered directly to employees at a 50% discountto face value (Article 23a, Law for Transformation and Privatization of State andMunicipal Enterprises). More than 530,000 employees became shareholders in theprivatized companies via this offer, effectively reducing the supply of shares availableat auction. Another three million Bulgarian citizens, out of the 6.5 million eligible,acquired vouchers in the main mass privatization scheme, obtaining 75 billionvouchers in aggregate.The original design of the mass privatization would have generated dispersed

ownership, because an individual’s allotment of 25,000 vouchers could buy no morethan a handful of shares. It was commonly believed at the World Bank and otherinternational advisory bodies that dispersed ownership would lead to weakgovernance of the privatized companies (Pistor and Spicer, 1997); at the time, the

ARTICLE IN PRESS


monitoring view of large shareholders was prevalent and the potential forshareholder expropriation was essentially disregarded. The solution for thegovernance problem was found in the creation of large institutional investors whowould collect vouchers from individuals and then use this voucher capital to acquiresignificant equity blocks in privatized companies. In the Czech Republic, theseinstitutional investors were called Investment Privatization Funds; the World Bankcalled them voucher funds.2

The mass privatization scheme was thought to have the best features of both theGerman and U.S. systems of corporate governance, which were the two establishedsystems in the West (Coffee, 1996). It combined the control features of the largeinstitutional investors from the German system and the liquid stock market withdispersed ownership from the U.S. system. In support of the belief that this hybridcorporate governance system could be effective, early studies of the post-privatizationperformance of companies in transition economies found that ownership byprivatization funds was associated with better performance (Claessens et al., 1997).Given the experience of the Czech Republic and the policy recommendations of the

World Bank and other economic organizations, the Bulgarian government furtherrefined the privatization scheme with a special Privatization Funds Act. The governmentenvisioned that privatization funds would serve two functions (Prohaska, 1996;Atanasov, 1997). The first was to create large shareholders that would actively engage inmonitoring managers and restructuring firms. This function would help achieve one ofthe main goals of the mass privatization process, which was the more efficient use offormerly state-owned assets. The second function, also motivated by modern financetheory, was to create diversification opportunities for individual investors. This functionwas viewed as necessary for achieving the second main goal of privatization, which wasthe emergence of a liquid stock market with a wide investor base.Besides participating on their own in the auctions, individual investors could

deposit their vouchers in a privatization fund and receive shares in the portfolioeventually acquired by the fund at the auctions. A total of 81 privatization fundswere registered. The funds had several months during which they could advertise,employ distribution agents, or use other channels to market themselves to the public.When the marketing campaign was over, the funds had collected about 80% of thetotal number of vouchers distributed to individuals (Atanasov, 1997).The government was not unaware of the expropriation view of dominant

shareholders, however, and was concerned that the emergence of majority ownerswho might expropriate firm value from small investors could undermine the creationof a liquid stock market. To address this concern, the Privatization Funds Actprohibited privatization funds from acquiring more than 34% of the shares of asingle enterprise during the auction rounds (Article 30, Privatization Funds Act). Onone hand, the government wanted to provide strong incentives for monitoring; on

2The notion of Investment Privatization Funds dates back to 1990 and was proposed by Dusan Triska,

the Deputy Minister of Finance in charge of privatization in the Czech government (Coffee, 1996). For

references to the World Bank policy recommendation to set voucher funds see Anderson (1994) and

Ellerman (1998).

ARTICLE IN PRESS


the other hand, however, it did not want to allow investors, at least initially, toaccumulate an uncontested majority (Pamouktchiev et al., 1997). The 34% was acompromise. In the Czech scheme the limit was set at 20%, but Coffee (1996) arguesthat a 20% limit is too low to provide incentives for monitoring and suggests insteada limit between 30% and 35%, which is consistent with the choice of the Bulgariangovernment and its U.S. advisors. The 34% limit also protected a privatization fundfrom large-scale expropriation by a potential majority shareholder, because theCommercial Code required a two-thirds supermajority approval for major changesin the corporate charter.One other important difference from the Czech scheme was that the privatization

funds in Bulgaria were not viewed as permanent entities. Bulgarian authorities gaveprivatization funds an option to change their structure after the mass privatizationprocess was complete: the funds could become either holding companies orinvestment companies. If a fund chose to become a holding company, it had toown more than 25% of the shares in each of its subsidiary investments and it couldextend its ownership beyond the 34% limit. Holding companies were treated in theCommercial Code as financial groups similar to holding companies in WesternEurope and business groups in East Asia. In contrast, the investment companyoption resembled typical closed-end mutual funds in the United States. Privatizationfunds that chose this option could not own more than 10% of the equity in acompany. They had to forgo any active governance and focus on trading andinvestment profits.In summary, the blueprint of the mass privatization process was entirely driven by

what seemed to be sound economic and finance principles. The scheme was designedin large part by a U.S.-sponsored consulting company. The chosen allocationmechanism was an auction in which millions of investors were allowed to participatein order to assure liquidity in the subsequent stock market. The process alsoencouraged the creation of large shareholders, as recommended by proponents of themonitoring shareholder view. Finally, absolute control through the accumulation ofa majority ownership stake was inhibited.Unfortunately, the international research on the relationship between law and

finance, pioneered by La Porta et al. (1998), had not yet become well known. Forexample, Nellis (2000) asserts that the World Bank economists after 1996 graduallybecame aware of the importance of the law in privatization. This slow processculminated in 1999, when Joseph Stiglitz, the World Bank Chief Economist at thetime, argued that privatization without strong legal protection of ownership rightscould lead to ‘‘more asset stripping than wealth creation’’ (Nellis, 2000). Unaware ofthe research, the Bulgarian government inadvertently overlooked a key element inthe design of their stock market ‘‘big bang’’ – strong legal protection of minorityshareholders.

2.2. The Bulgarian securities law at the time of the mass privatization process

The Bulgarian government proceeded speedily through mass privatization withoutensuring that the legal framework for a well-functioning stock market was in place.

ARTICLE IN PRESS


If we look at the comparison between the Czech and Polish approach in Glaeseret al. (2001), Bulgaria more closely resembles the Czech Republic in that extantsecurities law was inadequate. Joint stock companies were regulated by theBulgarian Commercial Code, which was originally passed in 1897 and then updatedand reinstated in 1991. The Commercial Code, which was based on German law, wasthe backbone of securities law and was designed with privately held limited liabilitycompanies in mind. It did not provide sufficient protection of minority shareholdersin publicly traded companies. The other relevant law was the 1995 Law on Securities,Stock Exchanges, and Investment Companies, which included certain limitedprovisions for equity issuers and created the Securities and Stock ExchangesCommission to oversee stock market participants.The Center for Study of Democracy (1999a) outlines several major deficiencies of

the Commercial Code that in the period 1998–1999 were quickly exploited bymajority shareholders in the newly privatized companies. These deficiencies concernthe protection of shareholder ownership rights and fit exactly into the La Porta et al.(1998) antidirectors index, the Pistor (2000) shareholder protection VOICE andEXIT indices, and the Glaeser et al. (2001) comparison of shareholder rights in theCzech and Polish Commercial Codes and regulation of listed companies.The first category of insufficient protections involves the lack of minority

shareholder rights at the shareholder assembly. The Commercial Code set theminimum number of votes required to convene a general shareholder meeting at10% of company capital. This minimum level is exactly the average of the countriesstudied in La Porta et al. Combined with the next deficiency of the CommercialCode, however, which was its failure to provide a mechanism for shareholder accessto the company shareholder register and communication with other similar-mindedowners, this shortcoming made it extremely hard for minority shareholders to voicetheir interests by summoning an extraordinary meeting and initiating a proxy fight.Another deficiency was that the law did not require companies to hold shareholdermeetings at the city of incorporation or another easily accessible place. Majorityshareholders frequently summoned meetings abroad or at remote locations, ensuringthat they would be the only shareholders in attendance. A final deficiency was thelack of cumulative voting for directors at the shareholder assembly. A shareholderowning 50% of the shares, or even less if no other shareholders attended the generalmeeting, could elect all the directors on the company board. All of these deficienciescreated the potential for majority shareholders to vote shareholder assemblydecisions in their own interest.The second category of deficiencies concerns direct mechanisms for large-scale

expropriation that can be implemented through control of the shareholder assemblyvote. The Commercial Code and the Law on Securities, Stock Exchanges, andInvestment Companies did not provide sufficient pre-emptive rights to shareholdersto prevent a majority shareholder from diluting their stakes. Given that shareholdermeetings could be summoned in remote locations and that voting by proxy wasinhibited, the supermajority provision that required two-thirds of the votes at ameeting to make changes in company capital did not prevent shareholders holdingas little as 50% of company capital from diluting everyone else’s holdings. Some

ARTICLE IN PRESS


pre-emptive rights were implemented in the 1999 Securities Law, but these provedto be largely ineffective. The pre-emptive rights were improved and made effectiveonly in early 2002, more than four years after the mass privatization process. Thelaw also failed to provide for any mandatory buyout of small shareholders bythe majority owner after the majority shareholder’s stake was increased throughdilution or open market purchases. Thus, small investors could be forced to holdworthless, diluted shares without any opportunity to obtain value from theinvestment via exit or otherwise. Last, the laws did not include sufficient appraisalrights and allowed majority shareholders to freeze out minority owners at greatlydepressed prices.In addition to inadequate written laws, the legal system had no experience with the

implementation of these laws, and individual investors were neither educated norknowledgeable enough to enforce their rights in court. In Slavova (2000), Bulgariascores well below the average among 24 transition economies both on extensivenessof stock market laws and on enforcement. The lack of disclosure and enforcement ofshareholder rights is confirmed in the Fall 1998 survey of the Center for Study ofDemocracy (1999b) and the study of Petranov and Miller (1999).

2.3. Concentration of ownership before the auctions

As Bebchuk (1999) suggests and La Porta et al. (1999) demonstrate, a legalenvironment that does not protect minority shareholders facilitates concentratedownership. In the Bulgarian market, the forces for majority ownership were sostrong that the accumulation of majority blocks was negotiated even before theprivatization auctions started. The first fallen defense against majority ownership atthe mass privatization auctions was the 34% maximum block rule. The response ofthe privatization funds to this limitation was the formation of bidding coalitions inwhich two or more funds agreed to bid for a firm and jointly acquire a 51% or largerblock of its shares.Conversations with fund managers revealed a common coalition bidding scheme,

also referred to in Miller and Petranov (2000) and Prohaska and Tchipev (2000). Afund interested in securing majority control in a firm would submit a bid for 34% ofthe firm’s capital. The fund would also sign a preliminary contract with a secondfund that committed to acquire an additional 17% on behalf of the first fund. Theywould agree to transfer the 17% block to the 34% holder at a later date. Suchcontracts were often agreed between the same two funds for different firms, with thetwo funds interchanging roles as the 34% and 17% buyer. The repeated interactionsguaranteed that the obligations under the contracts would be honored. Othercoalition agreements included two funds submitting equal-sized bids in long-termstrategic alliances to share control of a firm, or a 34% bidder using several otherfunds to buy smaller blocks on its behalf. By May 1998, about nine months after theend of the last auction, all firms that were offered in the mass privatization auctionswere registered on the Bulgarian Stock Exchange (BSE). This allowed the funds tofreely exchange blocks of shares acquired in the auctions and thereby fulfill theircoalition contract agreements.

ARTICLE IN PRESS


3. Data and summary statistics of the privatization fund bids

The main empirical tests in this paper are performed on a unique dataset ofprivatization fund bids from the second auction round. The Bulgarian governmentwas the only Eastern European government to choose a discriminatory auctionmechanism for its mass privatization process.3 This mechanism generated a set ofbids, potentially of different sizes and at different prices, for shares in the sameprivatized firm. The existence of such data makes it possible to measure the size ofthe control premium. The data also include the identity of the bidders, which is usedto document any regional bias of Bulgarian institutional investors and the coalitionarrangements among them. I first introduce the dataset and discuss the sampleselection. The distribution of privatization fund bids is analyzed next, and theempirical regularities evident in these bids as well as in subsequent stock price dataare interpreted in relation to the two competing views (monitoring versusexpropriation) about controlling shareholders. Table 1 outlines the main facts ofthe mass privatization process and the subsequent developments of the BulgarianStock Exchange in comparison with the findings of existing studies of the U.S.markets. The table also summarizes the empirical results from this paper.

3.1. Data sources and sample selection

The mass privatization auction data are from the Center for Mass Privatization(CMP) in Bulgaria. The data include all individual bids of each of the 81privatization funds in the second round of the Bulgarian mass privatization auctions.Each bid contains the fund code, the firm code, the bid price, and the number ofshares requested. The reserve price and the number of shares offered for each firm inthe second auction round are also available in electronic form from the CMP. Otherfirm-specific data are obtained from the Catalogue of All Firms Offered for Mass

Privatization, a publication distributed at no charge to all participants in the massprivatization process. This publication provides data on the total number of shares,the percentage of shares offered for mass privatization, and the city and the regionalcenter where the firm is incorporated. The CMP created 28 regional centers roughlycorresponding to the administrative districts of Bulgaria. Fund-specific informationcomes from the fund prospectuses submitted to the Bulgarian Securities andExchange Commission. Besides fund capital, the prospectuses contain informationabout the city of incorporation of each fund. This information is used to allocateeach fund to one of the 28 regional centers created by the CMP.A total of 820 firms had shares available for purchase on the second round. Table

2 reports the distribution of these firms across different categories depending on theamount of equity offered for purchase on the second round and the type of fund bids

3For example, the prices in the Czech privatization auctions were chosen by the government and

adjusted each round depending on supply and demand on the previous auction around. Investors in these

auctions could choose only the quantity of their bids. For an analysis of this data see Hingorani et al.

(1997) and the papers cited therein.

ARTICLE IN PRESS

Table 1

Comparison of the U.S. and Bulgaria relative to the two views of dominant shareholders

Bulgaria United States

Theoretical view Dominant shareholders expropriate

all firm value (Fama and Jensen,

1983)

Dominant shareholders monitor on

behalf of all shareholders (Shleifer

and Vishny, 1986)

Majority

shareholders

More than 75% of listed firms have a

majority shareholder

About 8% of listed firms have a

majority shareholder (Holderness and

Sheehan, 1988)

Fractional stake Majority stakes cluster at 51% Majority stakes are usually larger

than 51% (Holderness and Sheehan,

1988)

Coalitions among

blockholders

More than 50% of all auction bids are

coalitions

No evidence that coalitions are

common

Mutual funds Form portfolios of predominantly

controlling blocks

Seldom own large equity blocks (Roe,

1990)

Investors in mutual

funds

Small investors receive almost nothing

compared to fund sponsors

Some evidence of discounts due to

managers obtaining private benefits of

control (Barclay et al., 1993)

Control premium 10 times 20% (Barclay and Holderness, 1989)

Stock market

valuation

Market-to-book ratio of companies

controlled by a large shareholder is

40–60% lower

No significant difference in market-

to-book ratio of majority-owned

versus minority-owned companies

(Holderness and Sheehan, 1988)

Role of blockholders Dominant blockholders extract all

value; private benefits of control

approximately equal to the value of

the firm

Both shared and private benefits

(Barclay and Holderness, 1989)

Equilibrium Minority shareholders are extinct Large and small shareholders co-exist

(Holderness and Sheehan, 1988)


submitted for each firm. Throughout most of the paper, two types of bids will beconsidered: small bids, which will be independent bids for less than 25% of firmequity, and large bids, which will be bids for more than 25% of firm equity or bids inmajority coalitions. The choice of 25% as the cutoff level is motivated by the factthat, as previously discussed, holding companies must own at least 25% of theirsubsidiaries. This forced the privatization funds that intended to become holdingcompanies to acquire blocks above 25% in the companies where they had strategicinterests.Firms in which control cannot be obtained because less than 50% of their shares

are available on the second round are much less likely to attract bids by privatizationfunds. Out of 157 firms with less than 50% of their capital initially allocated for mass

ARTICLE IN PRESS

Table 2

Distribution of all firms available on the second auction round across firm types

A total of 820 firms have shares available for purchase on the second auction round. The auction bid data

include only bids by privatization funds and are obtained from the Center for Mass Privatization (CMP).

Data on the percent of firm capital are from the Catalogue of All Firms Offered for Mass Privatization, a

publication distributed by the CMP. Small bids are single (not participating in a coalition) bids for less

than 25% of firm equity; large bids are bids for more than 25% of firm equity or bids by funds

participating in majority coalitions.

Firm type Number of

firms receiving

no bids

Number of

firms receiving

only large bids

Number of

firms receiving

only small bids

Number of

firms receiving

both small and

large bids

Total

Firms offering less than 50%

of their shares for mass

privatization

93 4 60 0 157

Firms offering more than 50%

of their shares for mass

privatization, but with less

than 50% of their shares left

for purchase on the second

auction round

200 21 106 15 342

Firms with more than 50% of

their shares available for

purchase on the second

auction round but more than

1% of their shares have been

purchased by funds on the first

auction round

9 4 1 8 22

Firms with more than 50% of

their shares available for

purchase on the second

auction round and less than

1% of their equity has been

purchased by funds on the first

auction round

104 92 19 34 249

Firms that are offered for the

first time on the second

auction round and have more

than 50% of their equity

allocated for mass

privatization

15 24 1 10 50


privatization, 93 firms or 59% receive no bids. Similarly, no funds bid for the 58% offirms that initially offered more than 50% of their equity for mass privatization buthad less than 50% left unsold on the second auction round; only 3–10% of thesefirms receive at least one large bid. In contrast, more than 50% of the firms in whicha majority block is still unclaimed receive a large bid. The distribution of large and

ARTICLE IN PRESS


small bids across the 820 firms implies that large bidders avoid firms with establishedsignificant ownership and prefer firms in which they have a chance of being thedominant blockholder. Conversely, small bidders are most likely to bid for firms inwhich the government retains a majority block and to avoid companies in whichanother investor can acquire a majority stake on the second auction round (theseresults are confirmed by unreported multinomial logit and Poisson models).The main empirical test of the paper is the estimation of the premium for control.

This estimation requires data on the valuations of investors who expect to becomecontrolling shareholders. Such valuations can be observed only for firms in which amajority block can be acquired on the second auction round (firms with 50% ormore of their vouchers still available for purchase on that round). This requirementreduces the 820-firm population to a sample of 321 firms.An additional filter is imposed that firms should not have more than 1% of their

shares bought by privatization funds on the first round. The reason for imposing thisfilter is to rule out dynamic block purchases or coalition agreements, where one fundobtains a partial stake in a company on the first round and then on the second roundeither increases its stake by buying more shares or contracts with another fund to doso on its account. A dynamic bidding and coalition formation model that can fullycontrol for such events is beyond the scope of this paper. Furthermore, the mainresults of the paper remain unchanged if the 22 firms are included in the sample.Other filters like 0%, 5%, or 10% do not affect the results either. There are 299 firmsthat satisfy this condition. Funds submitted bids for 180 of the 299 eligible firms andthe empirical analysis will concentrate on this subset of firms. Of these firms, 35 areintroduced for the first time on the second round. The remaining firms wereintroduced on the first round, but still have most of their shares available on thesecond round. Compared to the rest of the firms that receive at least one auction bidby a privatization fund, the sample firms are smaller, receive more large bids, andattract fewer small bids. The sample selection process is summarized in Table 3.Another key statistic is the variation in the percent of equity offered for

privatization across the sample firms. Of the 180 firms in the estimation sample, 100firms offer 67%. Effectively, 60.3% of the shares in most of these firms are availablefor purchase, because 6.7% (or 10% of 67%) were previously acquired byemployees. The second-largest group contains 47 firms that offer 90% for massprivatization, and effectively 81% of their capital is available on the second auctionround. Of the remaining 33 firms, 22 offer 70%, two firms offer 75%, eight offer80%, and one 85%.The percent of equity offered for mass privatization could be important in the

determination of the expected voting power of each block of shares. If only 60% ofthe shares are available for purchase, the winner of a 34% bid will beunconditionally the largest shareholder in the company, because all remainingbidders can win only 26%. If more than 67% of the equity is auctioned off, a 34%winner can potentially face another 34% blockholder. In such scenarios, smallblockholders will have large control power because they will decide which 34%blockholder will control the company. Based on the above discussion, the controlvalue per share of bids in firms offering 67% for mass privatization is strictly

ARTICLE IN PRESS

Table 3

Summary of sample selection steps for the econometric tests

The sample selection criteria are: (1) a firm is offered on the second round of the mass privatization

auctions and has more than 50% of its shares available for purchase on that round; (2) a firm has less than

1% of its shares purchased by privatization funds on the first auction round; and (3) a firm receives at least

one bid by a privatization fund on the second auction round. These criteria reduce the population of 1,040

firms to the final sample of 180 firms used in the paper.

Description of sample selection step Number of firms in

the sample before

selection step

Number of firms in

the sample after

selection step

Step 0 The government chooses a set of firms for

mass privatization from the total population

of government-owned firms

n.a. 1,040

Step 1 Some firms are dropped because they are

completely sold out at the first auction round

and are not offered on the second round

1,040 820

Step 2 A number of firms are dropped because the

government offers less than a majority block

for mass privatization

820 663

Step 3 A number of firms are dropped because

privatization fund bids on the first auction

round reduce below 50% the available firm

equity at the second auction round

663 321

Step 4 Some firms are dropped because

privatization funds acquire toeholds of more

than 1% at the first auction round

321 299

Step 5 Some firms are dropped because they do not

receive a bid from a privatization fund in the

second auction round

299 180


increasing in bid size, while this is not necessarily so for firms offering 90%.Measures of the control premium based on the difference between the prices of smalland large bids will have more power in the group of 67% firms versus the group of90% firms. Nevertheless, all empirical tests below will be estimated for bothsubsamples. The firms with 70% of shares for mass privatization are added to the67% subsample, while the 75%, 80%, and 85% firms are added to the 90%subsample.

3.2. Bids by privatization funds

Out of the universe of 81 funds, 72 funds submit at least one bid for a firm in the180-firm sample. The remaining nine funds are dropped from further analysis. Atotal of 615 bids are submitted by the 72 funds. As discussed in Section 2, many

ARTICLE IN PRESS


funds submit coalition bids to jointly acquire a large and, in most cases, majoritystake in a company. The empirical analysis of the value of control has to account forthese coalition bids because the voting power of the participants in coalitions isdifferent from the voting power of single bidders. The identification of coalition bidsis based on finding pairs of funds that repeatedly either bid jointly at the second andthird auction round, submit bids at the same price that complement to a majoritystake, or own equity stakes in the same companies after the end of the massprivatization auctions. A more detailed description of the method identifyingcoalition bids can be found in the Appendix. An alternative method that classifiesfunds as a coalitional pair if they own shares in more than five firms together isimplemented in Prohaska and Tchipev (2000). The two methods produce largelysimilar results. A total of 393 of the 615 bids are submitted by coalition partners(coalition bids), while the remaining 222 bids are independently submitted (singlebids).The distribution of bid size, measured as a percent of total firm equity for single

and coalition bids, is shown in Fig. 1. Focusing on Panel A, more than 50% of allsingle bids, 118 out of 222, are for the maximum allowed 34% stake. Only 82 of thesingle bids are for less than 25% of firm equity and are classified as small bids. Thelow frequency of small bids suggests that the majority of privatization funds intendto become the dominant shareholder in the companies they bid for. A question mightarise as to why a small bid would be submitted at all if expropriation by a controllingshareholder were expected. In the majority of cases, the privatization funds acquiredsmall positions with the intention to sell these quickly on the Bulgarian StockExchange and generate short-term cash flows (Prohaska and Tchipev, 2000). Ofcourse, some privatization funds intended to become investment companies, and bylaw such companies could not own more than 10% of the shares in a singlecompany. These privatization funds concentrated most of their voucher capital insmall bids. Another reason for some small bids may be a binding budget constraint,where a privatization fund has some capital left over after making its strategic bids(Tchipev, 1997). Last, some of the small bids could be misclassified one-shotcoalition agreements.The striking difference between Panel A and Panel B of Fig. 1 is that 17% bids are

very common in a coalition setting, and quite rare as a single-bid choice. This is dueto the frequent use by the privatization funds of the ‘‘34% plus 17%’’ coalitionbidding scheme described earlier. To complete the analysis of the choice of bid size,Fig. 2 presents the sum of the bid sizes of all participants in a coalition bid in a firm.The vast majority of coalitions seek to achieve uncontested control by bidding jointlyfor a majority equity position, with a significant clustering around 51%. In fact, 56%of the coalitional bids are for equity blocks between 50% and 52% and only 13% arefor less than 50% of firm equity.Controlling shareholders interested in monitoring will increase their equity

position in the company beyond minimum majority, because for the same amount ofeffort they will capture more value from their monitoring activities. The opposite istrue if majority shareholders are interested mostly in securing private benefits ofcontrol, in which case they will have an incentive to acquire just the minimum block

ARTICLE IN PRESS

0.5

0.4

0.3

0.2

0.1

17 25 34

Bid size (percent of equity)

17 25 34


Fre

quen

cy

0.5

0.4

0.3

0.2

0.1

Fre

quen

cy

(A)

(B)

Fig. 1. Distribution of bids. On the x-axis is bid size measured as a percent of firm equity. By law, the

maximum allowed bid is for 34% of firm shares. On the y-axis is the relative frequency of bids. A total of

222 single bids and 393 coalition bids are submitted for the 180 firms in the sample. The frequency

distribution of single bids is in Panel A. The distribution of coalition bids is in Panel B.


necessary for control because they will have more outside equity holdings toexpropriate. Based on Holderness and Sheehan (1988, 2000), clustering at 51% is notobserved in the U.S. This finding suggests that legal constraints on the activities ofcontrolling shareholders in the U.S. prevent these shareholders from expropriatingvalue from minority owners. In such an environment, monitoring is more efficientand controlling blockholders hold stakes above the minimum majority. Theclustering at 51% in the Bulgarian market indicates that, in the absence of legalconstraints, controlling blockholders attempt to maximize the private benefits ofcontrol by purchasing the minimum equity stake necessary for control.

ARTICLE IN PRESS

0.5

0.4

0.3

0.2

0.1

17 34 51 68


Fre

quen

cy

Fig. 2. Distribution of coalition bids. On the x-axis is the sum of the bid size all fund bids submitted by a

coalition for a particular firm. On the y-axis is the relative frequency of such bids. The 393 separate

coalition bids are reduced to a total of 190 coalition bids, keeping one coalition bid per firm.


Table 4 provides additional support for the claim that institutional investors preferto be controlling shareholders rather than minority shareholders. The table showsthat the intended composition of the privatization fund portfolios is heavily biasedtoward large equity blocks. Even the smallest one-third of funds bid more than 66%of their capital for large blocks. As the average size of the fund increases, they investa higher percentage of their capital in large equity positions, with the largest one-third of funds investing as much as 84% of their capital in such positions. Theconcentrated portfolio composition of the Bulgarian institutional investors impliesthat the benefits of control outweigh the costs of lower diversification.

4. Estimates of the control premium

Under the assumptions of Shleifer and Vishny (1986), the shares held by amonitoring majority shareholder should have the same pro rata equity valuation asthe holdings of the remaining small shareholders, because the large shareholderprovides a public good unrestricted to all equity investors in the firm. According toFama and Jensen (1983), however, the value of shares held by a majority shareholdershould be essentially equal to the value of the whole firm, while the value of minorityshareholdings should be close to zero. Thus, the two models have conflictingpredictions about the control premium, measured as the difference between thevaluations of small and large shareholdings. Fama and Jensen predict an extremelylarge control premium, while Shleifer and Vishny predict a zero premium. The mainempirical evidence of this paper regarding the validity of the monitoring versus

ARTICLE IN PRESS

Table 4

Average proportion of privatization fund capital invested in large versus small bids

In the computation of capital invested, only bids for sample firms are considered. Fund capital is

denominated in thousands of vouchers. Small bids are single bids for less than 25% of firm equity; large

bids are bids for more than 25% of firm equity or bids by funds participating in majority coalitions. Small

funds are funds in the bottom third of funds ranked by voucher capital, medium funds are in the middle

third, and large funds are in the top third.

Fund size Average fund capital

invested in the sample firms

Proportion of capital

invested in small bids

Proportion of capital

invested in large bids

Small 19,867 0.338 0.662

Medium 81,339 0.212 0.788

Large 552,869 0.163 0.837


expropriation views will involve an analysis of the control premium at the Bulgarianmass privatization auctions.

4.1. Previous approaches to measuring the value of control

Extant empirical evidence about the control premium is offered by two broad setsof studies that use the existence of more than one price for shares in the samecompany.4 The first set documents the premium of negotiated trades of large blocksof equity over the market price of publicly traded companies. Barclay andHolderness (1989, 1991) and Bebchuk (1994) show that large blocks trade at apremium on average in the U.S., which implies that the private benefits of controldominate the costs of holding large blocks in public corporations. Based on Barclayand Holderness (1989), the control premium is 20% of the stock price of publiclytraded U.S. companies. A recent cross-country study by Dyck and Zingales (2004)finds that the premium for control in 39 countries ranges from �4% to 65%. Thesecond set of studies examines the price differential between the share classes of dual-class firms. Some examples include Lease et al. (1983) and Zingales (1994). Nenova(2002) provides a survey and a cross-country study of dual-class share firms. Basedon those results, the value of control in different countries ranges from as low as 0%to as high as 50% of the stock price.Both approaches have their shortcomings, however. The premium for control

measured in negotiated block transactions can be affected by selection bias becauseonly the valuations of major shareholders that decide to sell their stakes areobserved. The disadvantage of the dual-class share approach is that it measures thevoting power of the marginal shareholder multiplied by the probability of a controlcontest. If ownership is concentrated or the market for corporate control is weak, the

4There is an alternative approach to detect private benefits of control that does not rely on the existence

of more than one price, but can be applied only to companies in business groups. This approach is

developed by Bertrand et al. (2002) for Indian business groups and has been also applied to study Korean

groups by Bae et al. (2002), and European groups by Faccio and Stolin (2003).

ARTICLE IN PRESS


price differential of dual-class shares will underestimate the benefits of control of themajor existing shareholders. Both dual-class share and negotiated block transactionstudies take the existing firm ownership structure as given and measure controlbenefits as a function of ownership. A violation of the assumption that ownership isexogenous to control could lead to incorrect inferences about the size of the privatebenefits of control. Several studies have indeed shown that ownership isendogenously determined by the level of private benefits of control in a company(Demsetz and Lehn, 1985; Bebchuk, 1999).The Bulgarian mass privatization process provides a natural solution for the

endogeneity of ownership structure and control, because the privatization constitutesan exogenous shock to ownership. The ownership structure of all privatized firms iscreated from scratch as a result of the investor bidding at the auctions. Due to therichness of the data, an econometric model can also address some of the potentialselection bias in existing approaches to measuring the control premium.

4.2. Unconditional statistics of the control premium

The first estimate of the control premium is a comparison of the average prices bidfor large and small blocks. This price differential can then be used to compute thepercentage of firm value that majority shareholders expect to capture as privatebenefits of control, as in Barclay and Holderness (1989) and other block transactionstudies. However, instead of observing a block trade at a particular point in time andcomparing its price to the market price, this study directly compares the prices ofsimultaneous blocks of different sizes for the same firm.Table 5 contains the mean prices of small and large bids and the percentage

control premium for control calculated using those prices. As can be seen in Panel A,which shows firms with 67% or 70% of their capital allocated for mass privatization,large bids are submitted at higher prices. As discussed in Section 3.1, thecomputation of control premiums using the difference between large and small bidprices will be more precise in this subsample. The premium for control using rawprices is 26% and is highly statistically significant. Large bidders also pay a 143%higher markup over the reserve price relative to small bidders. However, conclusionsabout the difference in prices using sample averages could be flawed because, as thediscussion in Section 3.1 indicates, large and small bids can be submitted fordifferent types of firms. Table 5 further computes the average prices, markups, andcontrol premiums only for firms receiving both small and large bids; large bids againhave on average 26% higher prices, and large bidders pay a 90% higher markup overthe reserve price. Another way to compare the prices of large and small bids is tocompute the premium of large bids over small bids for each firm that receives bothtypes of bids and then compute the average premium across all such firms. Thecomputation of the price difference between large and small bids on a firm-by-firmbasis alleviates the effect of firm-specific unobservable factors and affords a cleanermeasure of the control premium. The average premium computed using this method,reported in the last column of Panel A, is 32.3%, which is also highly statisticallysignificant.

ARTICLE IN PRESS

Table 5

Mean (median) prices of different-sized bids

Small bids are single bids for less than 25% of firm equity; large bids are bids for more than 25% of firm

equity or bids by funds participating in majority coalitions. Panel A includes 122 firms with capital

allocated for mass privatization (masspriv) equal to 67% or 70%. Panel B includes 58 firms with

masspriv ¼ 75%, 80%, 85%, or 90%. See Section 3.1 for discussion of why the means and medians are

estimated separately for each panel. Prices are measured in vouchers. The markup over the reserve price is

computed as (bid price�reserve price)/reserve price. The premium is computed as (price of large bids �

price of small bids)/(price of small bids). For raw prices and markups over the reserve price, the average

prices (markups) are substituted in the premium calculation. In the last column of each panel, the premium

based on raw prices is computed on a firm-by-firm basis and the average of all firm-level premiums is

reported. P-values for the hypothesis that each premium is less than zero are in parentheses. The P-values

for the premium based on raw prices and markups over the reserve price are computed using

bootstrapping (see the Appendix).

Raw prices Markup over reserve price Firm-by-firm

calculation

All firms Firms with both

small and large

bids

All firms Firms with both

small and large

bids

Firms with both

small and large

bids

Panel A. Capital for privatization equals 67% or 70% of firm equity

Small bids 609.85 (504) 623.52 (503) 0.467 (0.083) 0.606 (0.119) 623.52 (503)

Large bids 769.55 (611) 788.25 (672) 1.132 (0.515) 1.154 (0.737) 788.25 (672)

Premium 0.2619 0.2642 1.4272 0.9043 0.3237

(0.008) (0.026) (0.000) (0.002) (0.000)

Panel B. Capital for privatization equals 75%, 80%, 85%, or 90% of firm equity

Small bids 1364.98 (655) 1504.76 (902) 2.158 (0.281) 2.565 (0.925) 1504.76 (902)

Large bids 1000.36 (603) 1637.82 (1091) 1.294 (0.484) 2.897 (1.372) 1637.82 (1091)

Premium �0.2671 0.0884 �0.4004 0.1292 �0.0382

(0.807) (0.410) (0.848) (0.408) (0.5942)


For the group of firms in Panel B, which had 75% or more of their capitalallocated for mass privatization, the price difference between small and large bidsswitches signs depending on the method of computation and is never statisticallysignificant. This confirms the conjecture expressed in Section 3.1 that the per-sharepower of small bidders might not necessarily be smaller than the power of largebidders if more than one large owner can emerge in a firm.In summary, the unconditional control premium is very large for the subsample of

firms in which only one large shareholder can emerge. Depending on the method, theestimated prices of large bids are between 26% and 143% higher than the prices ofsmall bids.

4.3. Econometric model and estimation

The unconditional measures of the control premium in the previous subsectionmight not provide the best estimates of the value of private benefits of control

ARTICLE IN PRESS


because they do not control for firm and fund characteristics. At first glance, anordinary least squares model of bid prices regressed on firm and fund characteristicsand a dummy variable for bid size would provide the required multivariate controls.5

However, funds submit small- or large-sized bids only when doing so is optimalamong all alternatives. A comparison of bid prices using a classic regression modelignores the underlying decision process of the privatization funds and suffers fromselection bias. Because of this selection bias, the OLS coefficient on the large-biddummy will understate the magnitude of the private benefits of control.The selection bias of an OLS bid price model can be addressed using a

methodology developed in Lee (1983) that involves a two-stage estimator. In the firststage, every privatization fund chooses among several mutually exclusive alternativesfor each of the 180 firms in the sample. At a minimum, these alternatives can be toabstain from bidding (choice ¼ 0), to submit a bid for a small block (choice ¼ 1), orto submit a bid for a large block (choice ¼ 2). The first-stage selection process is thenestimated as a multinomial logit model.The second-stage estimation involves running for each of the positive bid decisions

of the investors a separate OLS regression of the following form:

yij ¼ b0jX ij þ jjlij þ �ij ; (1)

where the i subscript denotes the observation and the j subscript denotes the choiceat the first stage. The dependent variable yij is the bid price, Xij is a vector of firm-and fund-specific regressors, and lij is a selection correction term. This selection termis a function of the predicted values from the first-stage multinomial logit model andcontrols for the endogenous decision of the funds to submit a bid of a particular sizefor the firms in the sample.Last, the estimated coefficients and the means of the right-hand-side variables are

used to compute the conditional expected prices for bids in each of the investmentchoice categories. The difference between these conditional expected prices is anestimate of the premium for control. The two-stage methodology is described inmore detail in the Appendix.The implementation of the methodology in this paper is as follows. The chosen set

of alternatives faced by the privatization funds in their investment decision for eachfirm is the most parsimonious specification and includes three investment choices –no bid, small bid, or large bid. After choosing the set of investment alternatives at thefirst-stage selection equation, a critical decision is the choice of instruments that arecorrelated with the propensity of a privatization fund to submit a large or small bidin a particular firm. In order to estimate the two-stage model, at least one instrumentuncorrelated with bid prices is needed. A theoretical analysis of the determinants ofinvestment choice of large institutional investors when private benefits of control andmonitoring can increase returns is beyond the scope of this paper, although someprogress towards such a model can be found in Atanasov (2002). That paper shows,under certain assumptions, which firm- and investor-specific characteristics will

5An OLS regression of bid price regressed on bid size and controls was estimated in a previous draft.

The coefficient on bid size was positive and statistically significant.

ARTICLE IN PRESS


affect the probability that an investor will acquire a large block of equity in a firm.Three instruments expected to be correlated with the investment choice can bederived from that model, and will be included in the first-stage investment selectionequation.The first of the three instruments is a measure of firm size. If an investor is

considering the tradeoff between the higher returns from owning a large block ofequity and the increased diversifiable risk associated with a large position, theinvestor will prefer to accumulate large percentage blocks in smaller firms. Forexample, if the increase in return due to private benefits of control or monitoring isindependent of firm size, then an investment in majority blocks in ten one-million-levfirms will have the same return as an investment in a majority block in one ten-million-lev firm, but will have a much smaller contribution to portfolio risk,assuming that the ten firms are not perfectly correlated. The particular measure offirm size that will be used in the estimation is the total number of firm shares(firmshr). The shares of all privatized firms have the same par value (1,000 Bulgarianlevs). The number of shares is therefore equivalent to the book value of equity. As anad hoc check on whether this variable qualifies as an instrument, it was included onthe right-hand side in the second-stage regression of bid prices. Its coefficient wasnegative but not statistically different from zero.The second instrument is a measure of investor capital. For a fixed firm size, an

investor with more capital will find it more profitable to invest in a large block thanan investor with less capital, because the same lev investment will have much asmaller weight in the portfolio of the large investor. Also, if the bid size is constant, alarge investor will be able to submit more bids. Thus, a measure of investor capitalwill be correlated with the propensity of the investor to bid for large blocks and tosubmit more bids. Such a measure is directly available as the number of vouchersthat each privatization fund possesses at the beginning of the second auction round(capital). When included in the second-stage regressions of bid prices, fund capitalhas a positive sign, but its coefficient is not significant. Still, there could be anargument that capital should also be included in the second-stage equation. This isdone in Section 4.5 below.The last instrument is geographical proximity. Everything else equal, an investor

would prefer to bid for firms that have high benefits of control or are easy tomonitor. Geographical proximity is an observable measure of these traits, becausefirms that are located closer to a privatization fund might be easier to manage (or toexpropriate). Coval and Moskowitz (1999) show that U.S. institutional investorsexhibit a strong preference for holding positions in nearby firms, which providessome empirical evidence that geographical proximity can be used as an instrument inan investment selection equation. Gompers and Lerner (1999) also find that venturecapitalists have strong regional biases. Additional empirical evidence that regionalproximity is an important factor in the formation of privatization fund portfolios isprovided by the fact that 14 Bulgarian funds explicitly state in their prospectuses thatthey are going to invest exclusively in firms in their geographic region; a further 16funds plan to invest the majority of their capital in regional firms (Prohaska, 1996).Geographic proximity is measured by a dummy variable (region) equal to one if a

ARTICLE IN PRESS


fund is located in the same regional center as a firm. This variable is conservativebecause a fund can be only 50–60 miles away from a firm in another regional center(the average size of a regional center is less than 1,500 square miles). When used as apredictor of bid prices, region enters with a small negative effect that is notstatistically significant.Overall, the three instruments in the first-stage selection equation are motivated in

part by theory and in part by empirical evidence. An informal test of whether theyare correlated with bid prices beyond the controls in the second-stage regressionrejects significant correlation and lends further support for their selection. Besidesthe three instruments, the other variable in the multinomial logit equation is thereserve price. It is strongly correlated with bid prices in the second-stage regressionsand therefore does not qualify as a pure instrument, but it is included to control forthe fact that only bid prices above the reserve price for each firm can be submitted.The estimated version of the first-stage selection equation is as follows:

Prob

large_bid

small_bid

no_bid

8><>:

9>=>;

¼ f ðfirmshr; capital; region; reserve_priceÞ; (2)

where the functional connection between the probability of a bid type and the right-hand-side variables is given by the multinomial logit specification.The second stage of the model consists of the estimation of two separate OLS

regressions, one for small-bid and one for large-bid observations. The dependentvariable is the log of bid price. Auction prices have only positive values and Lee(1984) gives a detailed explanation why in this case a log transformation is required.Four independent variables are used. The first is the reserve price, which controls forthe fact that the funds have to submit a price equal to or larger than the reserve priceannounced by the government. The second variable is a dummy equal to one if thefirm has been introduced for the first time on the first auction (A1firm). This variablemeasures the effect of potentially negative information about firms that have beenintroduced on the previous auction round and that have attracted no interest fromany privatization funds on that round. The third variable is the number of seriousbids submitted for a firm (numbids). Serious bids are defined as bids for more than15% of the firm’s shares. This variable is a proxy for the level of competition in anauction and for unobservable firm quality that generates high demand from largebidders; using the total number of bids or the ratio of shares demanded to sharesoffered as alternative proxies for firm quality and auction competition does notsubstantially change the results. The last variable is the selection term lambda,created using the estimates from the first-stage selection equation. Details about itscomputation are included in the Appendix. The estimated versions of the twosecond-stage regression equations are as follows:

LogðbidpriceÞs ¼ b0;s þ b1;sreserveps þ b2;sA1firms þ b3;sNumbidss

þ jslambdas þ �s; ð3:SÞ

ARTICLE IN PRESS


LogðbidpriceÞl ¼ b0;l þ b1;lreservepl þ b2;lA1firml þ b3;lNumbidsl

þ jllambdal þ �l ð3:LÞ

where the first equation is estimated using only observations with small bids, and thesecond one uses observations with large bids.

4.4. Results

Table 6 reports the results from the estimation of the first-stage investment choicemodel. The model is first estimated using the whole 180-firm sample. As a robustnesscheck, the model is also estimated using a restricted sample of only firms that receiveboth small and large bids. This alleviates a potential problem if firms that receiveonly large bids are systematically different from firms that receive only small bidsand these differences are not captured by the independent variables.Focusing on Panel A, the coefficient on the book value of equity (firmshr) is

negative and statistically significant for large blocks, and positive and statisticallysignificant for small blocks. Based on the coefficients, a one-standard-deviationincrease in book equity decreases the odds of a large bid by 86% in favor of a smallbid and by 64% in favor of no bid. Conversely, a one-standard-deviation increase inbook equity increases the odds of a small bid by 13% compared to no bid. Thedifference in the effect of firm size on the probabilities of receiving both small andlarge bids is statistically significant based on a Wald test.Given the estimated coefficients on firm size, it appears that small bids and large

bids are submitted for different types of firms and it is definitely the case that smallbidders prefer not to bid for firms in which a large owner is likely to emerge. Thisresult is not consistent with the monitoring view, because if it were true that largeinvestors increase firm value, minority investors would prefer to bid for firms inwhich a majority investor is likely to emerge. Another reason why small biddersmight prefer larger firms is that such firms could be more liquid after they starttrading on the Bulgarian Stock Exchange and the acquired shares could be easierto sell.Interpreting the other results in Table 6, larger investors are more likely to submit

large bids. A one-standard-deviation increase in fund size (capital) increases the oddsof a large bid by 42% relative to a small bid. This finding supports the theoreticalmodel in Atanasov (2002) and confirms the results from the univariate test in Table 4that funds with more capital allocate a higher percentage of that capital to biddingfor large blocks.Last, regional proximity dramatically increases the likelihood of both large and

small bids. The effect of regional proximity is almost the same for large and small bidsubmission for the subsample of firms offering 67% of their capital in the massprivatization. A zero-to-one change in the dummy region increases the odds of asmall bid by 151% and the odds of a large bid by 117% (versus not bidding). Apossible interpretation of the large magnitude of regional bias in privatization fundbids is that in an economy with poor disclosure and a lack of other publicly available

ARTICLE IN PRESS

Table 6

First-stage multinomial logit estimation results

The dependent variable, choice, is equal to zero if a fund does not submit a bid for a firm, one if the bid is

small (single bid for less than 25%), and two if the bid is large (larger than 25% or participating in a

majority coalition). Choice ¼ 0 (no bid) is the comparison group. The independent variables are capital,

which equals the fund voucher capital, firmshr, which equals number of firm shares, region, a dummy equal

to one if a fund is located in the same region as the firm, and reserve price. The multinomial logit model is

estimated using maximum likelihood. Panel A includes 122 firms with capital allocated for mass

privatization (masspriv) equal to 67% or 70%. Panel B includes 58 firms with masspriv ¼ 75%, 80%, 85%,

or 90%. See Section 3.1 for a discussion of why the models are estimated separately for each panel. Each

panel includes two reported models. The first model is estimated over all firms (122 firms in Panel A and 58

in Panel B), while the second model is estimated only over firms that receive both small and large bids (37

firms in Panel A and 7 in Panel B). P-values for each coefficient are in parentheses. The odds-ratio

measures the change in the probability of submitting a large bid versus a small bid after increasing the

continuous variables by one standard deviation, or changing the dummy from zero to one. The Wald test

is a test of the restriction that the coefficients on each independent variable are equal across the equation

for small bids and the equation for large bids.

All firms Firms with small and large bids

Variable Small bids Large bids Small bids Large bids


Capital Coefficient 0.00111 0.00220 0.00104 0.00192

(0.000) (0.000) (0.004) (0.000)

Odds-ratio 1.4192 1.3564

Wald test P-value 0.0009 0.0282

Firmshr Coefficient 0.00025 �0.00096 0.00005 �0.00043

(0.086) (0.006) (0.735) (0.108)

Odds-ratio 0.5379 0.6620


Region Coefficient 0.92132 0.77193 0.64020 0.83548

(0.006) (0.000) (0.110) (0.001)

Odds-ratio 0.8612 1.2157


Reserve

price

Coefficient �0.00054 �0.001744 �0.00096 �0.00120

(0.558) (0.000) (0.445) (0.130)

Constant Coefficient �5.14216 �3.27240 �3.76631 �3.06729

(0.000) (0.000) (0.000) (0.000)

Observations ¼ 8784 Observations ¼ 2072

Pseudo R2¼ 0.0968 Pseudo R2

¼ 0.0715

Likelihood ratio

X2¼ 347.68

Likelihood ratio

X2¼ 99.15

Prob

4X2(20) ¼ 0.0000

Prob

4X2(20) ¼ 0.0000


Capital Coefficient 0.00035 0.00136 0.00040 0.00156

(0.615) (0.000) (0.654) (0.002)

Odds-ratio 1.3886 1.4893


Firmshr Coefficient �0.00692 0.00028 0.00029 �0.03254

(0.239) (0.808) (0.992) (0.082)


ARTICLE IN PRESS

Table 6 (continued )



Odds-ratio 1.5844 0.5784


Region Coefficient 2.48296 1.13925 1.98903 0.93567

(0.000) (0.000) (0.013) (0.096)

Odds-ratio 0.2609 0.3488


Reserve

price

Coefficient �0.00042 0.00094 0.00013 �0.00036

(0.739) (0.002) (0.951) (0.780)

Constant Coefficient �5.49028 �4.13251 �4.18579 �2.45572

(0.000) (0.000) (0.013) (0.013)

Observations ¼ 4176 Observations ¼ 392

Pseudo R2¼ 0.0588 Pseudo R2

¼ 0.0919

Likelihood ratio

X2¼ 98.19

Likelihood ratio

X2¼ 24.55

Prob

4X2(20) ¼ 0.0000

Prob

4X2(20) ¼ 0.0019


firm-level information, a value-maximizing strategy is to invest in geographicallyproximate firms.A puzzling result in Panel B of Table 6 is that in the subsample of firms offering

75% or more of their equity for mass privatization, investors in the same region aremuch more likely to submit a small bid than a large bid. The firms in this sample arevery small and it is unlikely that they will be frequently traded on the stock exchange.Either this result is an artifact of the small number of observations in the small-bidcategory, or the regional funds are allocating some of their capital to small bids inthe expectation of selling later to a large owner (or perhaps in a control contestbetween two large owners).The estimation output from the second-stage regression equations is reported in

Table 7. All control variables have the expected signs. One finding that has to bestressed is that the coefficient on lambda is negative in the regression for largebids and positive in the regression for small bids. This finding supports theearlier suggestion that the possibility of selection bias will in general understatethe value of the private benefits of control. Both regressions have significantexplanatory power, which gives confidence in the estimates of control benefitspresented below.The last stage of the econometric methodology requires the computation of the

expected values of bid prices corresponding to bids for different-sized blocks. Theexpected bid prices are evaluated at the means of the right-hand-side continuousvariables and for each value of the dummy A1firm (first-round auction firms). Lee(1984) follows a similar approach and provides an explanation of why the correctionterm lambda should not be used in the computations of the expected value of the

ARTICLE IN PRESS

Table 7

Second-stage regression results

The dependent variable is the log of bid price. The independent variables are reserve price, numbids (the

number of bids submitted for a firm), and A1firm (a dummy equal to one if a firm was introduced on the

first auction round). Lambda is the selection correction term (details about its formulation are in the

Appendix). In each panel, two regressions (one for each investment choice) are estimated. Panel A includes

122 firms with capital allocated for mass privatization (masspriv) equal to 67% or 70%. Panel B includes

58 firms with masspriv ¼ 75%, 80%, 85%, or 90%. See Section 3.1 for a discussion of why the models are

estimated separately for each panel. Each panel includes two reported models. The first model is estimated

over all firms (122 firms in Panel A and 58 in Panel B), while the second model is estimated only over firms

that receive both small and large bids (37 firms in Panel A and 7 in Panel B). Lee (1983) t-statistics are in

parentheses.




Reserve price 0.0020 0.0021 0.0028 0.0020

(3.46) (10.66) (4.20) (5.85)

Numbids 0.0384 0.0364 0.0317 0.0240

(1.35) (3.67) (1.06) (1.78)

A1firm �0.2328 �0.2322 �0.2424 �0.2794

(�1.22) (�3.65) (�1.42) (�2.46)

Lambda 0.7278 �0.2226 1.0759 �0.0982

(2.97) (�4.05) (2.50) (�1.06)

Constant 3.4959 6.0235 2.7291 5.9530

(6.03) (36.16) (3.13) (20.30)

R2 0.5003 0.3857 0.5495 0.4148

Observations 64 334 47 125


Reserve price 0.0031 0.0014 �0.0003 0.0002

(1.35) (6.58) (�0.11) (0.25)

Numbids 0.1217 0.0408 0.2550 0.23404

(1.83) (1.63) (2.39) (6.59)

A1firm 0.4329 �0.2083 2.5621 1.4542

(0.52) (�1.36) (1.55) (3.43)

Lambda 0.2256 �0.1681 �0.7548 �0.9041

(0.45) (�0.84) (�0.96) (�2.53)

Constant 3.9656 6.0728 5.7003 6.1363

(3.17) (13.63) (2.82) (8.48)

R2 0.4127 0.3857 0.7130 0.7789

Observations 18 173 10 23


dependent variables. The results from this conditional mean computation for each ofthe two investment choices are included in Table 8. The estimated value of control isenormous. The prices of large blocks are about ten times higher than the prices ofsmall blocks. This estimate of the control premium implies that majorityshareholders expect to extract more than 85% of firm value as a private benefit ofcontrol.

ARTICLE IN PRESS

Table 8

Conditional expected bid prices

The conditional mean bid prices for each bid choice are evaluated at the means of the continuous

independent variables reserve price and numbids (the number of bids submitted for a firm) using the

parameter vector estimates in Table 7. The means in the first two columns in each panel are computed

setting the dummy A1firm equal to zero (indicating that the firm was not introduced in the first round); the

means in the second two columns are computed setting it equal to one. The definitions of Panel A and

Panel B are the same as in Table 7. Boostrapped P-values for H0: E[bidprice of large blocks]� E[bidprice

of small blocks] p 0 are in parentheses. Details about the bootstrapping procedure are in the Appendix.

Due to lack of sufficient number of observations in 24% of the runs in the last model in Panel B,

bootstrapped P-values could not be estimated.

Round 2 firms (A1firm ¼ 0) Round 1 firms (A1firm ¼ 1)

Log value Voucher equivalent Log value Voucher equivalent


All firms (122 firms)

Large blocks, 334 bids 7.0787 1186 6.8465 941

(0.006) (0.008)

Small blocks, 64 bids 4.5225 92 4.2897 73

(n.a.) (n.a.)

Only firms receiving small and large bids (37 firms)


(0.039) (0.051)

Small blocks, 47 bids 3.9934 54 3.7510

(n.a.) (n.a.) 43


All firms (58 firms)

Large blocks, 173 bids 6.9667 1,061 6.7584 861

(0.233) (0.400)


(n.a.) (n.a.)

Only firms receiving small and large bids (7 firms)

Large blocks, 23 bids 7.9668 2,884 9.4210 12,346

(0.293) (0.462)

Small blocks, 10 bids 7.5006 10.0627

(n.a.) 1,809 (n.a.) 23,451


The premium computed in Section 3 as an unconditional average price differentialbetween large and small bids, although substantial, is well below what would oneexpect given the lack of shareholder protection. The econometric model estimated inthis section adjusts for selection bias and produces a control premium that is closelyaligned with the ex post outcomes when the privatized firms are listed on theBulgarian Stock Exchange, which are analyzed in the next section. The model notonly generates economically accurate estimates of the private benefits of control inBulgaria, but it also proposes a new way of computing the value of private benefits ofcontrol using auction data as an alternative to the existing dual-class share andnegotiated block trade approaches.

ARTICLE IN PRESS


4.5. Robustness checks

The results are robust to estimating the model over the much smaller subsample offirms receiving both small and large bids. The only change is the explosive behaviorof expected prices in the restricted model in Panel B of Table 8. This is due to thesmall sample size (ten small- and 23 large-bid observations), which is not sufficient toprecisely estimate the parameters of the model. Additional robustness checks arereported in Table 9, where the model is estimated over six subsamples of the mainsample. Conclusions about the size of the premium for control are unchanged afterestimating the model separately for exporting firms, large firms, or firms introducedat the second auction round. An additional insight from the subsample estimations isthat the premium for control is largest for smaller firms and for firms introduced atthe second auction round.The main weakness of the two-stage model is the choice of instruments for the

first-stage selection equation.6 Even though firm size, fund capital, and geographicalproximity are convincingly correlated with the type of fund to submit a bid for aspecific firm, some of these variables could be correlated with bid prices in thesecond- stage equation. This is unlikely for firm size and regional proximity, becausewhen put as stand-alone variables in the second stage their effect is negative andclose to zero in both economic and statistical terms. On the other hand, fund capitalcan be expected to affect bid prices. After all, a fund with greater capital can affordto pay a higher price per share. This potential problem can be illustratedmathematically when we decompose fund capital into a product of (1) the numberof submitted bids, (2) the average bid size in terms of the number of shares, and (3)the average price per share. If we take logs on both sides of the equation we get:

lnðcapitalÞ ¼ lnðnumbidsÞ þ lnðmean_bidsizeÞ þ lnðmean_bidpriceÞ: (4)

After this decomposition it is easy to see the two effects of fund capital on theestimated model in Section 4.4. On one hand, greater capital increases the number ofbids and the probability of submitting a large bid; on the other hand, it can increasethe average bid price per share. The decomposition suggests a way to deal with thisproblem. We can either use log(number of bids) and log(average bid size) asinstruments in the first-stage selection model and add log(fund capital) to the second-stage equation, or use an approach that resembles instrumental variable regression.This approach is to regress log(capital) on log(number of bids) and log(average bidsize), use the estimated coefficients to compute predicted values for log(capital), anduse that as an instrument in the first-stage selection equation, putting rawlog(capital) in the second-stage bid price regressions.The results from the estimation of the pseudo instrumental variable model for the

subsample of firms offering 67% for mass privatization are reported in Table 10. Theintuition that fund capital has a positive effect on bid prices is correct; however,conclusions about the premium for control remain unchanged. The results are alsounchanged by using the number of bids and the percentage of large bids separately in

6I thank the referee for pointing out this fact and for suggesting the econometric approach to address it.

ARTICLE IN PRESS

Table 9

Robustness checks – subsample estimation

The conditional mean bid prices for each bid choice are computed the same way as in Table 8. The models

are estimated over six subsamples: exporting firms, non-exporting firms, large firms, small firms, Round 1

firms, and Round 2 firms. Exporting firms are firms that have nonzero exports in 1995. Large firms are

above median size, while small firms are below median size. Round 1 firms are firms that were offered for

privatization in the first auction round. Round 2 firms are firms that were offered for privatization in the

second auction round. Due to the small number of observations in the category of firms offering more

than 70% of their capital for mass privatization, in three of the subsamples there are not enough bid

observations to estimate the two-stage model. Consequently, only results for firms offering 67% or 70% of

their capital are reported.



Panel A. Results for exporters, non-exporters, large, and small firm subsamples

Exporting firms (67 firms)



Non-exporting firms (55 firms)



Large firms (73 firms)



Small firms (49 firms)



Panel B. Results for Round 1 and Round 2 firm subsamples

Log value Voucher equivalent

Round 1 firms (100 firms)

Large blocks, 219 bids 6.6152 746

Small blocks, 54 bids 5.4340 229

Round 2 firms (22 firms)

Large blocks, 115 bids 7.7218 2257

Small blocks, 10 bids 4.1060 61


the first-stage selection model, or by just adding capital in both the first stage andsecond stage.

5. Developments after the mass privatization auctions

The empirical tests in Sections 3 and 4 document the premium for control and theexpected effect of a controlling blockholder on firm value by directly observing thevaluations of large blockholders revealed in prices of auction bids. These valuationsare all ex ante, before the ownership structure of the firms has been formed andbefore the firms have started trading on the Bulgarian Stock Exchange. Another

ARTICLE IN PRESS

Table 10

Robustness checks – instrumenting fund capital

All models are estimated only on the subsample of firms with capital allocated for mass privatization

(masspriv) equal to 67% or 70%. The estimated multinomial model in Panel A is the same as in Table 6,

with the only difference that fund capital has been replaced by LcapitalIV. LcapitalIV is computed as the

predicted value of Log(capital) after this variable has been regressed on Log(number of bids submitted by

a fund) and Log(percentage of large bids submitted by a fund). The estimation in Panel B is the same as in

Table 7 with the inclusion of Log(capital) in the set of right-hand-side variables. The conditional expected

prices of large and small bids in Panel C are computed the same way as in Table 8.

Variable Small bids Large bids

Panel A. First-stage multinomial logit model

LcapitalIV 0.4597 1.0114

(0.000) (0.000)

Firmshr 0.0002 �0.0009

(0.085) (0.006)

Region 0.95382 0.8285

(0.005) (0.000)

Reserve price �0.0005 �0.0017

(0.562) (0.000)

Constant �13.4068 �21.6112

(0.000) (0.000)

Observations ¼ 8784

Pseudo R2¼ 0.1008

Likelihood ratio X2¼ 362.04

Prob 4X2(20) ¼ 0.0000

Panel B. Second-stage regressions

All firms

Variable Small bids Large bids

Lcapital 0.0088 0.0590

(0.19) (2.38)

Reserve price 0.0021 0.0020

(3.58) (10.05)

Numbids 0.0435 0.0381

(1.53) (3.82)

A1firm �0.1630 �0.2231

(�0.85) (�3.50)

Lambda 0.6546 �0.0614

(2.71) (�0.81)

Constant 3.4359 4.5671

(2.96) (7.62)

R2 0.4955 0.3821

Observations 64 334

Panel C. Conditional expected bid prices






ARTICLE IN PRESS


concern is that the premium for control reflects not lower minority shareholdervaluations because of potential extensive tunneling but higher majority shareholdervaluations because of future monitoring and restructuring benefits. In order tocomplete the analysis of the effect of large shareholders on firm value and to addressthese concerns, this section examines ownership structures following the auctions aswell as the stock market valuations of the privatized firms after they are listed on theBulgarian Stock Exchange.

5.1. Ownership structures after the mass privatization auctions

The ownership structure for all firms is available in electronic form from theCenter for Mass Privatization. The approach to quantifying ownership structure isto sort all firms participating in the mass privatization auctions into five categoriesbased on their ownership structure immediately after the last auction round. The fiveownership categories are as follows: (1) firms in which the government keeps amajority stake; (2) firms that do not have any ownership by privatization funds; (3)firms with contested control, defined as either two large unaffiliated privatizationfunds that each own 25% of more of firm equity, or two or more unaffiliatedprivatization funds that own less than 25% of firm equity with no large owner; (4)firms with uncontested control, where only one privatization fund has more than25% of firm equity; and (5) firms owned by a majority coalition, where two or moreprivatization funds that are identified as a coalition own more than 50% of firmequity.Table 11 reports the number of firms in each of the five ownership categories and

some summary statistics of interest. The resulting ownership structures are highlyconcentrated, with roughly 75% of the non-government-owned firms categorized aseither majority-owned or with a single large blockholder enjoying uncontestedcontrol. The analysis of the size- and auction-related variables suggests that thegovernment kept control in the largest and most valuable companies; thesecompanies are six times larger than the rest, and both their reserve and averageauction prices are higher. The next most valuable companies have contested control,and the least valuable have no privatization fund ownership.

5.2. Market valuation on the Bulgarian Stock Exchange

A common approach to documenting the effect of a controlling blockholder onfirm value is to examine the marginal shareholder valuation of companies withdifferent ownership structures. For example, Morck et al. (1988) and McConnell andServaes (1990) relate Tobin’s q to the level of insider ownership, while Holdernessand Sheehan (1988) compare the market-to-book ratios of majority-owned anddispersely owned companies. Claessens (1999) directly implements this approach inhis analysis of the outcome of the Czech voucher privatization scheme. In thissection, I estimate a similar regression model of the effect of ownership structure onstock market valuation and provide additional evidence that majority owners were

ARTICLE IN PRESS

Table 11

Ownership structures resulting from the mass privatization auctions

The data on ownership are recorded at the end of the last auction round. All firms are divided into five

possible ownership structures. The first ownership structure is when the government keeps a majority

stake; this structure is the baseline and its effect is captured by the regression constant. The second

ownership structure is no privatization fund ownership. The third ownership structure is contested

ownership, when a firm has two unaffiliated large privatization fund owners or two or more unaffiliated

small owners but no large owner. A large owner is defined as a privatization fund or coalition that controls

more than 25% of firm equity. The fourth ownership structure is uncontested ownership when a firm has

one large owner. The last ownership structure is when a firm is owned by a majority coalition. Coalitions

are identified by searching for repeated interactions between two privatization funds that submit auction

bids for the same companies in the second and third auction rounds or that own equity in the same firms at

the end of the auctions (see the Appendix). The numbers reported in the last four columns are means of the

respective variables for the firms of each ownership type. Firm size is measured by the number of shares

(firmshr), which is a direct measure of the book value of equity. Reserve price is the price chosen by the

government at the first auction round. Weighted-average prices are computed by the Center for Mass

Privatization based on all auction bids in the first round. Demand ratios are computed as the number of

shares demanded in the first auction round divided by the number of shares offered in that round.

Firm ownership type Number of

firms

Firm size Reserve price Weighted

average price

Demand ratio

Government

ownership

218 652,962 1031.62 1821.36 0.875

No privatization fund

ownership

120 112,310 859.76 1226.34 0.684

Contested ownership 93 139,414 972.91 1587.43 1.115

Uncontested

ownership

337 101,520 898.50 1448.38 1.628

Majority coalition 272 107,509 908.27 1566.28 1.531


expected to expropriate a large amount of firm value after the Bulgarian massprivatization.The basic equation relates a measure of stock market valuation, in this case stock

prices, to firm-level controls and measures of ownership structure. One convenientfeature of the Bulgarian mass privatization, also present in the Czech voucherscheme as pointed out by Claessens (1999), is that the number of shares of eachprivatized company was chosen in proportion to its book value of equity at thebeginning of the privatization. Thus, the book value per share of all companies listedon the Bulgarian Stock Exchange is the same and their stock price is also a measureof their market-to-book ratio. As a result, using stock prices as the dependentvariable in the regressions is largely equivalent to the use of market-to-book ratiosand Tobin’s q in the extant literature. Due to the fact that on any particular day lessthan a handful of companies trade on the Bulgarian Stock Exchange, the stock pricesare measured as the equally weighted average price of all trades during the calendarquarter. Stock price data are taken directly from the tapes of the Bulgarian StockExchange, which include information on the date, price, volume, and broker ID forall trades in all companies listed on the exchange from 1998 to 2002.

ARTICLE IN PRESS


The independent variable of primary interest will be the measure of ownershipstructure. This variable is created as a set of dummies for four out of the fiveownership categories described in Section 5.1. Government ownership is chosen asthe baseline case and its effect will be captured by the regression constant. Includingadditional firm-level controls in the regressions is necessary because firm ownershipis endogenous and could be correlated with firm performance, not because particulartypes of owners affect performance but because these owners prefer to invest inbetter-performing firms. This concern is consistent with the summary statistics inTable 11. Any of the price or demand variables from the auctions can provide acontrol for unobservable firm quality. The weighted-average auction price(wa_price1)is by far the most correlated with stock prices on the Bulgarian StockExchange and it will be the control variable included in the regressions. Data on thisvariable are available from the Center for Mass Privatization, but only for the firstauction round. This is not a major problem since 968 out of the 1,040 firmsparticipated in this round.Size and industry controls are included in the regressions because they are

expected to contribute to firm performance or growth opportunities captured by themarket-to-book ratio. Firm size is measured by the number of shares, and industrydummies for textiles, alcohol and tobacco, and tourism are included because theseindustries are expected to perform best in Bulgaria based on their historical record.Last, the number of trades during the calendar year is included in the regressionsbecause liquidity could reflect the beliefs of small investors about firm performanceor the probability of expropriation.The final estimated equation is:

Stockprice ¼ aþ b1 firmshr þ b2Numtrades þ b3wa_price1þ b4textile

þ b5tobacco þ b6alchohol þX4j¼1

djownership_typej : ð5Þ

Four regressions are estimated, one for each calendar quarter from June 1998 toJune 1999. The output from these regressions is reported in Table 12. The evidencefrom the regressions overwhelmingly suggests that controlling blockholders areassociated with large discounts in equity valuation. Firms with a majority ownertrade at significantly lower market-to-book ratios than government-owned firms orfirms with dispersed ownership or contested control. The magnitude of the discountsis large. Majority-owned firms trade from five to ten Bulgarian levs belowgovernment-owned firms and two to five levs below firms with contested control.Since the average stock price for the sample companies is in the neighborhood of tenlevs, these results imply discounts of at least 40% to 60% when a firm has a majorityowner. All of the controls except for the industry dummies are highly significant andhave the expected signs.The regressions in Table 12 conform strongly with the results in Section 4 and

show that controlling blockholders in Bulgaria are expected to expropriate most offirm value. Even if blockholders created any monitoring benefits, these wereevidently not shared with minority owners and were not manifested in stock market

ARTICLE IN PRESS

Table 12

Firm valuations on the Bulgarian Stock Exchange

The dependent variable is the equally weighted average stock price of all trades during the respective

calendar quarter. Four regressions, one for each of the quarters between June 1998 and June 1999, are

estimated. The independent variables are the number of firm shares (firmshr); the total number of trades in

the firm’s stock in 1998 for the first two regressions and in 1999 for the second two regressions; the

weighted average price of all Round 1 auction bids for a particular firm (wa_price1); industry dummy

variables for textiles, alcohol, and tourism; and dummy variables for each possible ownership structure.

All firms are divided into five possible ownership structures after the end of the mass privatization

auctions. The first ownership structure is when the government keeps a majority stake; this structure is the

baseline and its effect is captured by the regression constant. The second ownership structure is no

privatization fund ownership. The third ownership structure is contested ownership, when a firm has two

unaffiliated large privatization fund owners or two or more unaffiliated small owners but no large owner.

A large owner is defined as a privatization fund or coalition that controls more than 25% of firm equity.

The fourth ownership structure is uncontested ownership when a firm has one large owner. The last

ownership structure is when a firm is owned by a majority coalition. P-values for the null hypothesis that

each coefficient equals zero are in parentheses.

Variable 3Q1998 4Q1998 1Q1999 2Q1999

firmshr �1.8258 �1.0039 �1.0267 �0.9216

(0.000) (0.014) (0.025) (0.051)

Number of trades 0.0331 0.0185 0.0130 0.0123

(0.000) (0.000) (0.002) (0.004)

wa_price1 0.0045 0.0032 0.0021 0.0031

(0.000) (0.000) (0.000) (0.000)

Textile �1.0346 �0.7772 �1.4375 0.3615

(0.521) (0.597) (0.429) (0.855)

Alcohol and tobacco �1.4948 1.3922 1.3222 �0.8711

(0.549) (0.551) (0.595) (0.760)

Tourism �4.4229 �2.7626 �1.0782 �1.3636

(0.047) (0.148) (0.647) (0.569)

Ownership dummies

No privatization fund ownership �6.2437 �3.8429 �3.3590 0.2411

(0.029) (0.235) (0.317) (0.940)

Contested ownership �7.3643 �1.8349 �0.7292 �1.3945

(0.000) (0.305) (0.735) (0.553)

Uncontested ownership �7.4580 �2.0784 �3.9827 �2.8425

(0.000) (0.133) (0.013) (0.098)

Majority coalition �10.5506 �4.9045 �6.6614 �6.1530

(0.000) (0.002) (0.000) (0.005)

Constant 8.6224 4.4563 7.3243 5.2578

(0.000) (0.003) (0.000) (0.005)

R2 0.4321 0.2420 0.1880 0.1812

Obs. 319 270 217 198


valuations. The evidence in Table 12 is also consistent with recent studies by Weissand Nikitin (2002) and Pivovarsky (2003) documenting that block ownership byprivatization funds is associated with poor firm performance in the Czech Republicand Ukraine, respectively. Still, some concerns could arise about the power of theseregressions. The sample size is only 200–300 companies out of the 1,000 companiesoriginally listed on the Bulgarian Stock Exchange, which suggests the potential for

ARTICLE IN PRESS


selection bias. The reason that there are only 200–300 companies in each regression isthat the remaining companies on the exchange did not trade in a calendar quarter. Itis reasonable to expect that expropriation will be more severe in less tradedcompanies because small investors will trade less frequently in the companies wherethey expect to have worse protection. Thus, the selection bias most likely worksagainst finding any evidence of expropriation.The regressions could also suffer from an endogeneity-of-ownership bias even

after including firm-level controls. This endogeneity bias is likely to be small becausethe auction data should largely control for it. Nonetheless, a major problem is thepossibility that the government kept majority blocks in the best firms and that this iswhy there are such large discounts for firms with a non-government majority owner.Yet this concern is not consistent with the finding that majority-owned firmsunderperform not only government-owned firms but also all other firms.

5.3. Shareholder expropriation

Overall, an examination of equity valuations on the Bulgarian Stock Exchangeconvincingly refutes the notion that controlling blockholders in Bulgaria wereexpected to monitor and increase firm value. The regression results are much moreconsistent with blockholders expropriating wealth from the minority owners of theircompanies. These results and the estimates of the private benefits of control fromSection 4 are clearly supported by the developments subsequent to the privatizationauctions and the anecdotal evidence of tunneling presented next. In contrast to theexamples in Johnson et al. (2000), who focus on tunneling through firm operationslike transfer pricing, the tunneling examples below all involve financial transactions,which are probably more severe and less costly to implement. The developmentsfurther show that it is unlikely that the premium for control is due to expectedmonitoring benefits. As described below, even if any monitoring or restructuringbenefits were created, in most cases these benefits were not shared with minorityowners, who were frozen out at depressed prices.The first anecdotal support for the results that the premium for control is enormous

comes from the companies that offered less than 50% of their equity at the massprivatization auctions. After the auctions, the government sold the controlling blocksin many of these companies to strategic investors for cash. Markov (2000) reports theratio of the stock market price of a few of these companies to the prices per share paidby the majority owners. For example, in 2000 the stock price of Neftochim, thenational oil refinery, was 24% of the price paid by Lukoil for the majority block; forBalkanfarma (a holding of three pharmaceutical companies) this ratio was 21%; andfor Sodi, the second largest producer of soda ash in the world, this ratio was only10.8%. These ratios translate into premiums for control of three to nine times thetraded price, very much in line with the estimates from Section 4, and suggest thatmajority owners are able to extract 75–90% of firm value as private benefits of control.These companies were the largest privatized companies at the time and it is reasonableto believe that it is probably harder to tunnel a large and well-known company thanthe small to medium-sized privatized companies in my sample.

ARTICLE IN PRESS


How could majority shareholders tunnel as much as 90% of firm value? Startingfrom mid-1999, controlling blockholders in most majority-owned firms diluted theownership rights of the remaining shareholders. A notable example is Sopharma,one of the most profitable Bulgarian companies and a member of SOFIX, theBulgarian Stock Exchange ten-most-liquid-stocks index. In late 2000 the majorityshareholder of Sopharma initiated an equity issue that increased the number ofoutstanding shares by more than six times. The new shares were issued at a pricemore than ten times lower than the current stock price. Most minority shareholderswere not able to subscribe to the new issue and the majority owner increased hisstake from 67% to 85%. After the capital increase, the stock price dropped from 13to 2 Bulgarian levs and the P/E ratio reached an incredibly low 0.5. For moreinformation on tunneling techniques in Bulgaria and the Sopharma case seeAtanasov et al. (2004) and the sources cited within.Many majority shareholders used another expropriation technique: taking a

company private through a freeze-out of the outstanding minority shares. Anexample is Sodi, another member of SOFIX. Solvey, a Belgian chemicals company,purchased from the government a 67% controlling block in Sodi for roughly US$36U.S. per share. Through market purchases at depressed stock prices ranging fromtwo to five U.S. dollars, Solvey increased its equity block to 92% and then at the endof 2001 made a tender offer for the remaining shares at about six U.S. dollars. Bymid-2001, more than two-thirds of the 1,040 originally listed companies hadsimilarly frozen out their minority shareholders, and the out-of-equilibriumcoexistence of majority and minority shareholders had been replaced with 100%private ownership by the original majority owner. The majority shareholders in thesefirms essentially tunneled the value of the whole firm, while minority shareholdersreceived close to nothing, which is exactly what Fama and Jensen (1983) predict.The dilution and freeze-out of privatized firms affected mostly minority investors

that had purchased shares directly at the auctions. A large number of individualinvestors had deposited their vouchers in a privatization fund and most of thesefunds were the majority shareholders that secured enormous private benefits ofcontrol. Did the individual investors in these funds fare well? When the time came totransform the privatization funds into investment or holding companies, 76 of the 81funds became holding companies and only one of the top 20 largest funds by assetsdecided to become an investment company (Prohaska and Tchipev, 2000). Soonafterwards, many of the fund managers diluted all remaining shareholders using thedeficiencies in the Commercial Code outlined in Section 2.2, the same way as wasdone in the privatized companies. When the process was over, the fund managersgained a majority of the voting shares and secured complete control over thedistribution of fund cash flows.

6. Conclusion

The Bulgarian mass privatization offers an opportunity to observe the behavior ofinstitutional investors in an extreme case with no mechanisms to protect the rights of

ARTICLE IN PRESS


minority shareholders and constrain majority owners. The results show that in sucha setting, institutional investors form portfolios of controlling blocks that are valuedten times more highly than minority equity stakes. The control premium computedin this study is orders of magnitude larger than the control premium documented inthe US, and several times larger than the premium found in other developedmarkets. The observed behavior of institutional investors and the magnitude of thecontrol premium in Bulgaria imply a strong preference on the part of majorityshareholders for expropriating value rather than adding value through monitoring.This effect is further manifested in stock market valuations. Companies with amajority owner trade at 40% to 60% discounts.These findings add further support to the international law and finance literature

pioneered by La Porta et al. (1998). Most of the studies in this extensive literaturehave focused on firm- or macro-level effects of the legal environment on differentaspects of capital markets. A country-level study by Johnson et al. (2000) suggeststhat poor legal protection can lead to extensive tunneling in times of economicdecline, and then shows that this can explain the stock market and foreign currencycollapse during the Asian crisis. The firm-level evidence suggests that weak investorprotection leads to concentrated ownership (La Porta et al., 1999), lower firmvaluation (La Porta et al., 2002), and lower firm growth (Demirguc-Kunt andMaksimovic, 1998). This paper extends the literature along another dimension byshowing that the legal framework not only results in predictable patterns of firmcharacteristics, but also affects the behavior of investors. In contrast to blockholdersin the US, unconstrained large blockholders in Bulgaria expect to take the easier andmore profitable way of expropriating value from minority shareholders, rather thanexpending effort on monitoring and increasing firm value.A relevant research topic might be an examination of the types of constraints on

majority shareholders that are most important in providing incentives for theseshareholders to have a positive effect on both firm value and the wealth of minorityequity owners. Investigating the micro-level effect of particular legal statutes, taxcode provisions, or reputation mechanisms is beyond the scope of this paper and isleft for future research. Nevertheless, one sound conclusion can be made afterpresenting the findings of this paper. A stock market with dispersed ownership andhigh minority shareholder valuations will not emerge if there is no legal control overcontrolling blockholders.

Appendix. Description of the methodology to measure control premiums and expected

private benefits of control

For each of the 180 firms in the sample, 72 observations are created for the 72funds that submitted at least one bid. An indicator variable choiceik is created to beequal to zero if fund k does not make a bid for firm i or is not the participant with thelargest bid in a coalition seeking less than 50%; choiceik is equal to one if fund k

submits a bid for firm i and the size of the bid is less than 25%; and it equals two ifthe bid is larger than 25% or the bid is part of a majority coalition.

ARTICLE IN PRESS


The empirical methodology requires the identification of bids where two or morefunds agree before the auction to jointly acquire a block in a firm. This is not astraightforward task because the coalition contracts are not publicly available. Adescription of the algorithm used to perform this identification is as follows. First,the ownership data of the 1,040 privatized firms immediately after the last auctionround are screened to find pairs of funds with an abnormal number of occurrenceswhere the two funds own blocks of shares in the same firm. These potential suspectedcoalitions are then further explored to find in how many instances the two funds owncomplementary blocks that sum to a majority block or else are the only two owners.Forty-nine fund pairs jointly control five or more firms and the owners of the largeand supporting blocks change across the controlled firms. These pairs are classifiedas suspected coalitions.Next, the bid data from the second and third auction round are filtered to find

instances where bids at the same price are submitted. The probability of identicalbids on a firm’s shares is very low. If on more than one occasion the same two fundssubmit bids at the same price, the two funds are also qualified as a suspectedcoalition. After the coalition suspects are identified, the 615 bids in the data set aremanually examined. A bid is classified as a coalition if both members of a suspectedcoalition pair bid for the same firm and their bids are complementary and/orsubmitted at the same or a very close price. Additionally, coalition bids of threefunds are identified if these funds are connected through suspected coalition pairsand their bids are at the same price or complementary to a large block. If more thanthree coalition partners bid for a firm, and the sum of their bids is greater than thetotal amount of available shares, coalition pairs are identified based on closenessbetween the partners in terms of the number of interactions and same-price biddingoccasions.After identifying coalitions, I estimate a multinomial logit model with a dependent

variable choice ¼ j, j ¼ 0, 1, 2 (no bid, small bid, large bid). After estimating theparameter vector of the model ½g1; g2�; the predicted probabilities for each choice j arecomputed using the following formulae:

Pj ¼ prob choice ¼ j; j40ð Þ ¼exp g0jZ

1þ exp g01Z

þ exp g02Z (A.1)

P0 ¼ prob choice ¼ 0ð Þ ¼1

1þ exp g01Z

þ exp g02Z (A.2)

where Z is the matrix of right-hand-side variables of the logit equation.I then calculate the terms lj ; which are vectors with a typical element lij given by

lij ¼fðF�1ðPijÞÞ

Pij

(A.3)

where fðxÞ and FðxÞ are the p.d.f. and the c.d.f. of the standard normal distributionand F�1ðxÞ denotes the inverse function of FðxÞ: The lj are parallel to what

ARTICLE IN PRESS


Heckman (1976) calls the ‘‘inverse of the Mill’s ratio.’’ The difference in the Lee(1983) model is that it requires one further transformation involving F�1ðxÞ:After forming lj ; at the second stage, two separate OLS regressions are estimated

(one for each j40) of the following form:

LogðbidpriceÞj ¼ b0jX j þ j0jlj þ �j (A.4)

where Xij is a matrix of right-hand-side regressors, jj ¼ sjrj ; sj is the standarddeviation of �j ; and rj is the correlation between ej and a transformation of the errorsfrom the logit equation. Lee (1983) proves that under certain normality assumptions,OLS provides consistent estimates of the parameter vector yj ¼ ½bj ;jj�:The only complication is that the errors are heteroskedastic and the variance-

covariance matrix Vj of the parameters yj should be estimated using the followingexpression:7

Vj ¼ ð ~X0

j~X jÞ

�1XNj

i¼1~�2i ~x

0i ~xi þ ~X

0

jDnCOVnD0 ~X j

j kð ~X

0

j~X jÞ

�1 (A.5)

where Nj is the number of observations in the regression corresponding to choice j, D

is the matrix of derivatives of lj with respect to the logit parameters g; COV is thevariance-covariance matrix of these parameters, � is the vector of regressionresiduals, and ~X j ¼ ½X j ; lj�:After the parameter vectors bj are estimated in Eq. (A.4), the conditional mean

prices for each choice are computed as

EðpricejÞ ¼ b0jX (A.6)

where X is the vector of average values for the continuous variables and equal tozero or one for the dummy variables. There is only one dummy variable in thespecification estimated in the paper and the prices are computed twice: once settingthe dummy equal to zero and once setting it equal to one.The P-values of the hypothesis that the conditional mean price for choice ¼ 2

(large bid) is greater than the conditional mean price for choice ¼ 1 (small bid) arecomputed using bootstrapping. The exact details of the bootstrapping procedure areas follows. A sample of size N ¼ 180n72 is drawn from the original dataset. The two-stage estimates of bj are computed for the drawn sample and then the conditionalmean prices are computed using Eq. (A.6). This process is repeated for 1,000 drawnsamples. At the end, the number of samples in which E(bidprice | choi-

ce ¼ 2)4E(bidprice| choice ¼ 1) is counted. One minus this proportion gives theP-value of a one-sided test for equality between the conditional bid prices ofchoice ¼ 2 (large bids) and choice ¼ 1 (small bids).The value of monetary private benefits of control can be estimated after making

the assumption that bid prices are a linear function of the share valuation of

7The original expression is found in Lee (1982), who cites White (1980) as the source of the result and

proof. Details for the actual implementation into a GAUSS program are available from the author.

ARTICLE IN PRESS


investors. Given this assumption, the private benefits of control (PBC) are equal to

PBC ¼ 1�ðEðprice1Þ ownership1Þ

ðEðprice1Þ ownership1 þ Eðprice2Þ ownership2Þ

1

ownership1

(A.7)

where E(price1) is the conditional expected price for small blocks (choice ¼ 1),E(price2) is the conditional expected price for large blocks (choice ¼ 2), andownership1 and ownership2 are the percentage ownership of small shareholders andthe large shareholder, respectively.Eq. (A.7) is based on the definition that private benefits of control equal the value

of the firm that is not distributed pro rata to all shareholders but is secured only bythe controlling shareholder. Under this definition, PBC equal one minus theproportion of firm value distributed pro rata. The share valuation of smallshareholders is based only on the pro rata distribution. The pro rata valuedistribution can then be computed as the percentage of small shareholder value ofoverall firm value rescaled by one over their ownership stake. With the assumptionthat bid prices are a linear function of value, conditional expected prices can besubstituted for values and Eq. (A.7) follows directly.

References

Anderson, R., 1994. Voucher funds in transitional economies: the Czech and Slovak experience. World

Bank Policy Research Working Paper No. 1324.

Atanasov, P., 1997. Mass privatization in Bulgaria (in Bulgarian). Report presented at the Conference and

Workshop Mass Privatization in Bulgaria: Program Framework, First Results and Future Tasks,

Sponsored by USAID through Barents Group/KPMG, April 2, 1997.

Atanasov, V., 2002. Optimal portfolios with monitoring, private benefits of control, and budget

constraints. Unpublished working paper, Babson College.

Atanasov, V., Ciccotello, C., Gyoshev, C., 2004. How does law affect finance: an empirical examination of

tunneling in an emerging market. Unpublished working paper, Babson College.

Bae, K.H., Kang, J.K., Kim, J.M., 2002. Tunneling or value added? Evidence from mergers by Korean

business groups. Journal of Finance 57, 2695–2740.

Barclay, M., Holderness, C., 1989. Private benefits from control of public corporations. Journal of

Financial Economics 25, 371–395.

Barclay, M., Holderness, C., 1991. Negotiated block trades and corporate control. Journal of Finance 44,

861–877.

Barclay, M., Holderness, C., Pontiff, J., 1993. Private benefits from block ownership and discounts on

closed-end funds. Journal of Financial Economics 33, 263–291.

Bebchuk, L., 1994. Efficient and inefficient sales of corporate control. Quarterly Journal of Economics

109, 109–993.

Bebchuk, L., 1999. A rent-protection theory of corporate ownership and control. NBER Working Paper

No. 7203.

Bennedsen, M., Wolfenzon, D., 2000. The balance of power in closely held corporations. Journal of


Bertrand, M., Mehta, R., Mullainathan, S., 2002. Ferreting out tunneling: an application to Indian

business groups. Quarterly Journal of Economics 117, 121–148.

Black, B., Kraakman, R., Tarassova, A., 2000. Russian privatization and corporate governance: what

went wrong. Stanford Law Review 52, 1731–1808.

ARTICLE IN PRESS


Burkart, M., Gromb, D., Panunzi, F., 1998. Why higher takeover premia protect minority shareholders.

Journal of Political Economy 106, 172–204.

Center for Mass Privatization, 1997. Policy choices in Bulgaria’s mass privatisation programme. Memo,

Sofia, Bulgaria.

Center for Study of Democracy, 1999a. Policy for corporate governance development in joint-stock

companies in Bulgaria: action plan. Policy recommendation paper, Sofia, Bulgaria.

Center for Study of Democracy, 1999b. Corporate governance and control in Bulgaria: A summary

report. Memo, Sofia, Bulgaria.

Claessens, S., 1999. Corporate governance and equity prices: evidence from the Czech and Slovak

Republics. Journal of Finance 52, 1641–1657.

Claessens, S., Djankov, S., Pohl, G., 1997. Ownership and corporate governance: Evidence from the Czech

Republic. World Bank Policy Research Paper No. 1737.

Coffee, J., 1996. Institutional investors in transitional economies: Lessons from the Czech experience. In:

Frydman, R., Gray, C., Rapaczynski, A. (Eds.), Corporate Governance in Central Europe and Russia,

Vol. 1. Banks, Funds, and Foreign Investors. Central European Press, London, pp. 111–186.

Coval, J., Moskowitz, T., 1999. Home bias at home: local equity preference in domestic portfolios. Journal

of Finance 54, 2045–2073.

DeAngelo, H., DeAngelo, L., Rice, E., 1984. Going private: minority freezeouts and stockholder wealth.

Journal of Law and Economics 27, 367–401.

Demirguc-Kunt, A., Maksimovic, V., 1998. Law, finance and firm growth. Journal of Finance 53,

2107–2137.

Demsetz, H., Lehn, K., 1985. The structure of corporate ownership: Causes and consequences. Journal of

Political Economy 93, 1155–1177.

Dyck, A., Zingales, L., 2004. Private benefits of control: an international comparison. Journal of Finance

59, 537–600.

Ellerman, D., 1998. Voucher privatization with investment funds: an institutional analysis. World Bank

Working Paper No. 1924.

Faccio, M., Stolin, D., 2003. Expropriation vs. proportional sharing in corporate acquisitions.

Unpublished working paper, Vanderbilt University.

Fama, E., Jensen, M., 1983. Separation of ownership and control. Journal of Law and Economics 26,

301–325.

Glaeser, E., Johnson, S., Shleifer, A., 2001. Coase versus the Coasians. Quarterly Journal of Economics

116, 853–899.

Gompers, P., Lerner, J., 1999. The Venture Capital Cycle. The MIT Press, Cambridge.

Grossman, S., Hart, O., 1988. One share one vote and the market for corporate control. Journal of


Harris, M., Raviv, A., 1988. Corporate governance: voting rights and majority rules. Journal of Financial

Economics 20, 203–235.

Heckman, J., 1976. The common structure of statistical models of truncation, sample selection and limited

dependent variables and a simple estimator for such models. Annals of Economic and Social

Measurement 5, 475–492.

Hingorani, A., Lehn, K., Makhija, A., 1997. Investor behavior in mass privatization: the case of the Czech

voucher scheme. Journal of Financial Economics 44, 349–396.

Holderness, C., Sheehan, D., 1988. The role of majority shareholders in publicly held corporations: an

exploratory analysis. Journal of Financial Economics 20, 317–346.

Holderness, C., Sheehan, D., 2000. Constraints on large shareholders. In: Morck, R. (Ed.),

Concentrated Corporate Ownership, NBER Conference Report. University of Chicago Press,

Chicago, pp. 139–168.

Iliev, Y., 1997. Centralized auctions: procedures and technology (in Bulgarian). Paper presented at the

Conference and Workshop Mass Privatization in Bulgaria: Program Framework, First results and

Future tasks, Sponsored by USAID through Barents Group/KPMG, April 2, 1997.

Johnson, S., Boone, P., Breach, A., Friedman, E., 2000. Corporate governance in the Asian financial crisis.

Journal of Financial Economics 58, 141–186.

ARTICLE IN PRESS


Johnson, S., LaPorta, R., Lopez-de-Silanes, F., Shleifer, A., 2000. Tunneling. American Economic Review

90, 22–27.

LaPorta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 1998. Law and finance. Journal of Political

Economy 106, 1113–1155.

LaPorta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 1999. Corporate ownership around the world.

Journal of Finance 54, 129–152.

LaPorta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 2002. Investor protection and corporate

valuation. Journal of Finance 57, 1147–1170.

Lease, R., McConnell, J., Mikkelson, W., 1983. The market value of control in publicly traded

corporations. Journal of Financial Economics 11, 439–471.

Lee, L.F., 1982. Some approaches to the correction of selectivity bias. Review of Economic Studies 49,

355–372.

Lee, L.F., 1983. Generalized econometric models with selectivity. Econometrica 51, 507–512.

Lee, L.F., 1984. Technical training and earnings: a polychotomous choice model with selectivity. The

Review of Economics and Statistics 66, 151–156.

Lloyd, R., Nenov, Z., 1997. USAID support for the Bulgarian mass privatization program. Report

presented on the Conference and Workshop Mass Privatization in Bulgaria: Program

Framework, First Results and Future Tasks, sponsored by USAID through Barents Group/KPMG,

April 2, 1997.

Markov, T., 2000. How much are the shares in Solvey–Sodi worth. Newspaper Capital 124, 29 June 2000

(in Bulgarian).

McConnell, J., Servaes, H., 1990. Additional evidence on equity ownership and corporate value. Journal

of Financial Economics 25, 263–290.

Miller, J., Petranov, S., 2000. The first wave of mass privatization in Bulgaria and its immediate aftermath.

Economics of Transition 8, 225–250.

Mitrev, K., 1997. Bulgaria. In: Lieberman, I., Nestor, S., Desai, R. (Eds.), Between State and Market:

mass Privatization in Transition Economies. Studies of Economies in Transition, World Bank,

Washington.

Mitton, T., 2002. A cross-firm analysis of the impact of corporate governance on the East Asian financial

crisis. Journal of Financial Economics 64, 215–241.

Morck, R., Shleifer, A., Vishny, R., 1988. Management ownership and market valuation: an empirical

analysis. Journal of Financial Economics 20, 293–316.

Nellis, J., 2000. Privatization in transition economies: What happened? What is next? Unpublished

working paper, World Bank, Washington.

Nenova, T., 2002. The value of corporate votes and control benefits: a cross-country analysis. Journal of


Pamouktchiev, H., Parvoulov, S., Petranov, S., 1997. Process of privatization in Bulgaria. In: Jones, D.,

Miller, J. (Eds.), The Bulgarian Economy: Lessons from Reform during Early Transition. Ashgate

Publishing, Brookfield.

Petranov, S., Miller, J., 1999. Bulgaria’s capital markets in the context of EU accession: a status report.

Center for Study of Democracy and The International Bank for Reconstruction and Development/

World Bank, Sofia, Bulgaria.

Pistor, K., 2000. Patterns of legal change: Shareholder and creditor rights in transition economies.

European Business Organization Law Review 1, 59–108.

Pistor, K., Spicer, A., 1997. Investment funds in mass privatization and beyond: evidence from the Czech

Republic and Russia. In: Lieberman, I., Nestor, S., Desai, R. (Eds.), Between State and Market: Mass

Privatization in Transition Economies. World Bank, Washington.

Pivovarsky, A., 2003. Ownership concentration and performance in Ukraine’s privatized enterprises. IMF

Staff Papers 50, 10–42.

Prohaska, M., 1996. Privatization funds – the Bulgarian model. Memo, Center for Study of Democracy,

Sofia, Bulgaria.

Prohaska, M., Tchipev, P., 2000. Establishing corporate governance in an emerging market: Bulgaria.

Memo, Center for Study of Democracy, Sofia, Bulgaria.

ARTICLE IN PRESS


Roe, M., 1990. Political and legal restraints on ownership and control of public companies. Journal of


Shleifer, A., Vishny, R., 1986. Large shareholders and corporate control. Journal of Political Economy 94,

461–488.

Slavova, S., 2000. Law and finance in transition economies. Unpublished working paper, World Bank,

Washington.

Tchipev, P., 1997. Bulgarian mass privatization scheme: Implications on corporate governance.

Unpublished working paper, DP 01/97. Bulgarian Academy of Sciences.

Weiss, A., Nikitin, G., 2002. Effects of ownership by investment funds on the performance of Czech firms.

In: Meyendorff, A., Thakor, A. (Eds.), Designing Financial Systems in Transition Economies:

Strategies for Reform in Central and Eastern Europe. MIT Press, Cambridge.

White, H., 1980. A heteroskedasticity-consistent covariance matrix estimator and a direct test for

heteroskedasticity. Econometrica 48, 817–838.

Zingales, L., 1994. The value of the voting right: a study of the Milan Stock Exchange. Review of

Financial Studies 7, 125–148.

Zingales, L., 1995. Insider ownership and the decision to go public. Review of Economic Studies 62,

425–448.

Documents

How much value can blockholders tunnel? Evidence from the Bulgarian mass privatization auctions