Journal of Economic Literature Vol. XXXIX (June 2001) pp. 321–389 Megginson and Netter: From Journal of Economic State Literature, Vol.to Market XXXIX (June 2001) From State to Market: A Survey of Empirical Studies on Privatization W ILLIAM L. M EGGINSON and JEFFRY M. N ETTER 1 1. Introduction state-owned enterprises (SOEs) or assets to private economic agents, is now in use worldwide. Since its introduction by T HE POLITICAL AND economic policy of privatization, broadly defined as the deliberate sale by a government of Britain’s Thatcher government in the early 1980s to a then-skeptical public (that included many economists), privati- 1 Megginson: University of Oklahoma. Netter: zation now appears to be accepted as a University of Georgia. This paper was developed with financial support from the SBF Bourse de legitimate—often a core—tool of state- Paris and the New York Stock Exchange, and the craft by governments of more than 100 assistance of George Sofianos, Bill Tschirhart, and countries. Privatization is one of the Didier Davidoff is gratefully acknowledged. We appreciate comments received on this paper from most important elements of the continu- Geert Bekaert, Anthony Boardman, Bernardo Bor- ing global phenomenon of the increasing tolotti, Narjess Boubakri, Jean-Claude Cosset, use of markets to allocate resources. Kathryn Dewenter, Alexander Dyck, Oleh Havrylyshn, Ivan Ivanov, Jonathan Karpoff, Ranko It is tempting to point to the spread Jelic, Claude Laurin, Marc Lipson, Luis López- of privatization programs around the Calva, John McMillan (the editor), Sandra Sizer- world during the past two decades and Moore, Harold Mulherin, Rob Nash, John Nellis, David Newberry, David Parker, Enrico Perotti, conclude that the debate on the eco- Annette Poulsen, Ravi Ramamurti, Susan Rose- nomic and political merits of govern- Ackerman, Nemat Shafik, Mary Shirley, Mike ment versus private ownership has been Stegmoller, Aidan Vining and three anonymous referees. We appreciate comments from partici- decided. But such a conclusion is pants at the NYSE/Paris Bourse Global Equity flawed, since 25 years ago proponents Markets conference (Paris, Dec. 1998), Harvard of state ownership could just as easily Institute for International Development Privatiza- tion Workshop (June 2000), International Federa- have surveyed the postwar rise of state- tion of Stock Exchanges’ Third Global Emerging owned enterprises and concluded that Markets Conference (Istanbul, April 2000), World their model of economic organization Bank and/or IFC meetings, OECD conferences (Paris and Beijing), 1999 Conference on Privatiza- was winning the intellectual battle with tion and the Kuwaiti Economy in the Next Cen- free-market capitalism. Instead of tury, 1998 Financial Management Association pointing to the spread of privatization meeting, 1999 European Financial Management Association meeting, Fondazione ENI Enrico and calling it destiny, our goal is to as- Mattei (FFEM), Swiss Banking Institute and sess the findings of empirical research on Credit Suisse, and seminars at City University the effects of privatization as a policy. Business School (London), London Guildhall Uni- versity, and University of Oklahoma. All remaining Therefore, this paper surveys the rap- errors are the authors’ alone. idly growing literature on privatization, 321 322 Journal of Economic Literature, Vol. XXXIX (June 2001) attempts to frame and answer the key tion’s overall impact in the transition questions this stream of research has economies. Section 6 asks whether do- addressed, and then describes some of mestic and international investors who its lessons on the promise and perils of purchase privatizing share offerings ex- selling state-owned assets. Throughout perience positive initial and long-term this survey, we adopt the perspective of investment returns, and section 7 evalu- an advisor to a government policymaker ates the impact of privatization on the who is wrestling with the practical development of non-U.S. capital mar- problems of whether and how to imple- kets over the past two decades. Finally, ment a privatization program. The poli- section 8 discusses how privatization cymaker asks “What does the research programs have impacted the develop- literature have to tell us about these as- ment of—and interest in—corporate pects of privatization as an economic governance practices around the world. policy?” We attempt to answer these Section 9 concludes and summarizes important questions. our survey. This paper is organized as follows. Section 2 provides a brief historical overview of privatization. We examine 2. How Large Has Privatization’s the impact that privatization programs Impact Been to Date? have had in reversing SOE involvement Given the attention the press has in the economic life of developed and given to the global movement toward developing countries. Section 3 briefly markets, especially the privatization of surveys the recent theoretical and em- state-owned enterprises, some might pirical research on the relative eco- conclude that privatization has almost nomic performance of state-owned and ended the involvement of state-owned privately owned firms. Section 4 details enterprises in global economic activity. 2 the different types of transactions that This is a significant overstatement. To are labeled “privatization” in different understand the impact of privatization regions. We draw particular attention to on the state’s role in different econo- the structure and pricing selected for mies, we must first briefly review the share issue privatizations. We also history behind both privatization and its evaluate the various forms of “voucher” precursor, nationalization. or “mass” privatizations that have been Throughout history, there has been a implemented. This section also exam- mixture of public (often including reli- ines whether less radical methods of im- gious institutions) and private ownership proving the performance of SOEs, such of the means of production and com- as deregulation and allowing greater merce. Robert Sobel (1999) writes that competition (or more routine steps such state ownership of the means of produc- as using management performance con- tion, including mills and metal working, tracts), can effectively substitute for was common in the ancient Near East, outright privatization. In section 5, we while private ownership was more com- examine the issue of whether, and by mon in trading and money lending. In how much, privatization programs have actually improved the economic and fi- 2 Throughout this paper, we will use the World nancial performance of divested firms. Bank’s definition of state-owned enterprises, as Our discussion first evaluates privatiza- described in World Bank (1995): “government- owned or government-controlled economic enti- tion in industrialized and developing ties that generate the bulk of their revenues from countries, and then assesses privatiza- selling goods and services.” Megginson and Netter: From State to Market 323 ancient Greece, the government owned production. In many countries, state- the land, forests, and mines, but con- owned banks were also given either mo- tracted out the work to individuals and nopoly or protected positions, as dis- firms. In the Ch’in dynasty of China, cussed in Rafael La Porta, Florencio the government had monopolies on salt López-de-Silanes, and Andrei Shleifer and iron. Sobel notes that in the Roman (2000a). Republic the “publicani (private indi- Rondinelli and Iacono (1996) argue viduals and companies) fulfilled virtu- that government ownership grew in the ally all of the state’s economic require- developing world for slightly different ments.” Dennis Rondinelli and Max reasons, primarily that government own- Iacono (1996) note that by the time of ership was perceived as necessary to the Industrial Revolution in the western promote growth. In the post-colonial industrialized societies and their colo- countries of Asia, Africa, and Latin nies, the private sector was the most im- America, governments sought rapid portant producer of commercial goods growth through heavy investment in physi- and was also important in providing cal facilities. Another reason for govern- public goods and services. This pattern, ment ownership, often through nation- with more government involvement in alization, was a historical resentment of some countries and less in others, con- the foreigners who had owned many of tinued into the twentieth century in the largest firms in these countries (see western Europe and its colonies and also Roger Noll 2000). former colonies. In the United States, Thus there had been tremendous there was less government involvement growth in the use of SOEs throughout than in many other countries. much of the world, especially after The Depression, World War II, and World War II, which in turn led to pri- the final breakup of colonial empires vatizations several decades later. 3 Most pushed government into a more active people associate modern privatization role, including ownership of production programs with Thatcher’s government. and provision of all types of goods and However, the Adenauer government in services, in much of the world. In west- the Federal Republic of Germany ern Europe, governments debated how launched the first large-scale, ideologi- deeply involved the national govern- cally motivated “denationalization” pro- ment should be in regulating the na- gram of the postwar era. In 1961, the tional economy and which industrial German government sold a majority sectors should be reserved exclusively stake in Volkswagen in a public share for state ownership. Until Margaret offering heavily weighted in favor of Thatcher’s conservative government small investors. 4 Four years later, the came to power in Great Britain in 1979, the answer to this debate in the United 3 The historical overview of postwar privatiza- Kingdom and elsewhere was that the tions is based on a longer historical discussion in Megginson, Robert Nash, and Matthias van Ran- government should at least own the denborgh (1994). Other discussions of the histori- telecommunications and postal services, cal evolution of privatization include Timothy electric and gas utilities, and most Jenkinson and Colin Mayer (1988), Shirley and John Nellis (1991), World Bank (1995), Josef forms of non-road transportation (espe- Brada (1996), Paul Bennell (1997), and Daniel cially airlines and railroads). Many poli- Yergin and Joseph Stanislaw (1998). 4 Using a broader definition of privatization— ticians also believed the state should one that encompassed reactively changing the poli- control certain “strategic” manufactur- cies of an immediate predecessor government— ing industries, such as steel and defense the Churchill government’s denationalization of 324 Journal of Economic Literature, Vol. XXXIX (June 2001) government launched an even larger of- It was not until the successful British fering for shares in VEBA. Both offer- Telecom initial public offering in No- ings were initially received favorably, vember 1984 that privatization became but the appeal of share ownership did established as a basic economic policy not survive the first cyclical downturn in the United Kingdom. A series of in- in stock prices, and the government was creasingly massive share issue privatiza- forced to bail out many small sharehold- tions (SIPs) during the last half of the ers. It was almost twenty years before 1980s and the early 1990s reduced the another major western nation chose to role of SOEs in the British economy pursue privatization as a core economic to essentially nothing after the Tories or political policy. 5 left office in 1997, from more than 10 Although the Thatcher government percent of GDP eighteen years earlier. may not have been the first to launch a We note that the objectives set for large privatization program, it is with- the British privatization program by the out question the most important histori- Conservatives were virtually the same cally. Privatization was not a major cam- as those listed by the Adenauer govern- paign theme for the Tories in 1979, but ment twenty years before—and almost the new conservative government em- every government since. These goals, braced the policy. Thatcher adopted the as described in Price Waterhouse label “privatization,” which was origi- (1989a,b), are to (1) raise revenue for nally coined by Peter Drucker and the state, (2) promote economic effi- which replaced the term “denationaliza- ciency, (3) reduce government interfer- tion” (Yergin and Stanislaw 1998, p. ence in the economy, (4) promote wider 114). Early sales were strenuously at- share ownership, (5) provide the oppor- tacked by the Labour opposition, which tunity to introduce competition, and (6) promised that if it were reelected it subject SOEs to market discipline. The would renationalize divested firms such other major objective mentioned by the as British Aerospace and Cable and Thatcher and subsequent governments Wireless. 6 was to develop the national capital mar- ket. 7 We note these goals can be con- the British steel industry during the early 1950s flicting and we discuss the trade-offs could well be labeled the first “privatization.” We thank David Parker for pointing this out to us. further in the paper. 5 Pan Yotopoulos (1989) describes and assesses The perceived success of the British the Chilean programs, which began before the privatization program helped persuade program in the U.K. The Pinochet government of Chile, which gained power after the ouster of Sal- many other industrialized countries to vador Allende in 1973, attempted to privatize com- begin divesting SOEs through public panies that the Allende government had national- share offerings. Jacques Chirac’s gov- ized. However, the process was poorly executed and required very little equity investment from ernment, which came to power in purchasers of assets being divested. Thus, many of France in 1986, privatized 22 compa- these same firms were renationalized once Chile nies (worth $12 billion) before being entered its debt and payments crisis in the early 1980s. Chile’s second privatization program, which ousted in 1988. The returning socialist was launched in the mid-1980s and relied more on government did not execute any further public share offerings than direct asset sales (in sales, but neither did it renationalize which the government often acted as creditor as well as seller) was much more successful. the divested firms. Beginning in 1993, 6 Ironically, a labor government partially priva- tized an SOE just before Thatcher came to power. 7 Kojo Menyah, Krishna Paudyal, and Charles In 1977, the Labour government sold a relatively Inganyete (1995) and Menyah and Paudyal (1996) small fraction of the government’s shares in Brit- have more detailed discussions of the goals of the ish Petroleum as a means of raising cash. British privatization program. Megginson and Netter: From State to Market 325 the Balladur government launched a nese economy. While there have been new and even larger French privatiza- numerous small privatizations, there tion program, which has continued un- have been relatively few outright sales der the Jospin administration. The So- of SOEs, thus the overall impact of pri- cialists, in fact, launched the two largest vatization has been limited. Though the French privatizations ever, the $7.1 bil- government recently (1999) reaffirmed lion France Telecom initial public of- its commitment to privatizing all but fering (IPO) in October 1997 and the the very largest state enterprises, the subsequent $10.5 billion seasoned fact that Chinese SOEs are burdened France Telecom issue in November 1998. with so many social welfare responsi- Several other European governments, bilities suggests that it will be extraordi- including those of Italy, Germany, and, narily difficult to implement a privatiza- most spectacularly, Spain, also launched tion program large enough to seriously large privatization programs during the undermine the state’s economic role 1990s. These programs typically relied (Cyril Lin 2000 ; Justin Lin, Fang Cai, on public share offerings, and were and Zhou Li 1998; and Chong-en Bai, often launched by avowedly socialist David Li, and Yijaiang Wang 1997). The governments. Privatization spread to other special Asian case is India, which the Pacific Rim, beginning in the late adopted a major economic reform and 1980s. Japan has sold only a relative liberalization program in 1991, after be- handful of SOEs during the past fifteen ing wedded to state-directed economic years (usually relying on SIPs), but development for the first 44 years of its many of these have been truly enor- independence. India’s reform program mous. The three Nippon Telegraph and shares two key features with China’s: it Telephone share offerings executed be- was adopted in response to highly dis- tween February 1987 and October 1988 appointing SOE performance (Sumit raised almost $80 billion, and the $40 Majumdar 1996), and privatization has billion NTT offer in November 1987 re- thus far not figured prominently in the mains the largest single security offering reform agenda. in history. Elsewhere in Asia, govern- On the other hand, Latin America has ments have taken an opportunistic ap- truly embraced privatization. Chile’s proach to SOE divestment, selling pieces program is particularly important, both of large companies when market condi- because it was Latin America’s first and tions are attractive, or when money is because the 1990 Telefonos de Chile needed to plug budget deficits. It is un- privatization, which used a large Ameri- clear how the economic difficulties that can depository receipt (ADR) share gripped the region during the late 1990s tranche targeted toward U.S. investors, will impact privatizations in the future. opened the first important pathway for Two Asian countries deserve special developing countries to directly tap attention. These two countries are al- western capital markets. ready the world’s second and fifth larg- Mexico’s program was both vast in est economies on a purchasing-power- scope and remarkably successful at re- parity basis, and promise to become ducing the state’s role in what had been even more important over time. The an interventionist economy. La Porta People’s Republic of China launched a and López-de-Silanes (1999) report that major economic reform and liberaliza- in 1982, Mexican SOEs produced 14 tion program in the late-1970s that has percent of GDP, received net transfers transformed the productivity of the Chi- and subsidies equal to 12.7 percent of 326 Journal of Economic Literature, Vol. XXXIX (June 2001) GDP, and accounted for 38 percent of eastern Europe. These countries began fixed capital investment. By June 1992, privatizing SOEs as part of a broader the government had privatized 361 of its effort to transform themselves from roughly 1,200 SOEs, and the need for command to market economies. There- subsidies had been virtually eliminated. fore, they faced the most difficult chal- Several other countries in Latin lenges and had the most restricted set America have also executed large di- of policy choices. After the collapse of vestment programs (Pablo Gottret communism in 1989–91, all of the 1999). For example, Bolivia’s innovative newly elected governments of the re- “capitalization” scheme has been widely gion were under pressure to create acclaimed. However, the most impor- something resembling a market econ- tant program in the region is Brazil’s. omy as quickly as possible. However, Given the size of Brazil’s economy and political considerations essentially re- its privatization program, and the fact quired these governments to significantly that the Cardoso government was able limit foreign purchases of divested assets. to sell several very large SOEs (CVRD Since the region had little financial in 1997 and Telebras in 1998) in spite savings, these twin imperatives com- of significant political opposition, this pelled many—though not all—govern- country’s program is likely to remain ments throughout the region to launch very influential. “mass privatization” programs. These Privatization in sub-Saharan Africa programs generally involved distrib- has been something of a stealth eco- uting vouchers to the population, which nomic policy. Few governments have citizens could then use to bid for shares openly adopted an explicit SOE divest- in companies being privatized. The ment strategy, but Bennell (1997) programs resulted in a massive reduc- shows that there has been substantially tion of state ownership and were ini- more privatization in the region than is tially popular politically, but became commonly believed. For example, unpopular in many countries (especially Steven Jones, Megginson, Robert Nash Russia ) because of the largely correct and Netter (1999) show that Nigeria has perception that they were robbery by been one of the most frequent sellers of the old elite and the new oligarchs. The SOEs, using public share offerings, al- net effects have been disappointing in though they were very small. The expe- some cases, but have varied widely. We rience of the African National Congress discuss the empirical evidence on after it came to power in South Africa voucher privatization in section 5. also shows the policy realities that gov- Although different regions have em- ernments with interventionist instincts braced privatization at varying speeds, face in this new era. Though nationali- governments have found the lure of zation and redistribution of wealth have revenue from sales of SOEs to be at- been central planks of ANC ideology for tractive—which is one reason the policy decades, the Mandela and Mbeki gov- has spread so rapidly. According to Pri- ernments have almost totally refrained vatisation International (Henry Gibbon from nationalization and have even sold 1998, 2000), the cumulative value of off several SOEs (though use of the proceeds raised by privatizing govern- word “privatization” remains taboo). ments exceeded $1 trillion sometime The last major region to adopt priva- during the second half of 1999. As an tization programs comprises the former added benefit, this revenue has come to Soviet-bloc countries of central and governments without raising taxes or Megginson and Netter: From State to Market 327 $US Billions 180 160 140 120 100 80 60 40 20 0 1988 1990 1992 1994 1996 1998 Figure 1. Annual Privatization Revenues for Divesting Governments, 1988–99 Source: Privatisation International. cutting other government services. An- the role of state-owned enterprises in nual proceeds grew steadily before peak- the economies of high-income (industri- ing at over $160 billion in 1997. Since alized) countries has declined signifi- then, proceeds seem to have leveled off cantly, from about 8.5 percent of GDP at an annual rate of about $140 billion. in 1984 to less than 6 percent in Figure 1 shows the annual revenues 1991. Data presented in James Schmitz governments have received from privati- (1996), Mahboobi (2000), and Bernado zations from 1988 through 1999. Ladan Bortolotti, Marcella Fantini, and Mahboobi (2000) reports similar figures Domenico Siniscalco (1999a), as well as classified by privatizations in OECD our own empirical work on share issue and non-OECD countries. He reports privatizations suggests that the SOE that since 1990 privatization in OECD share of industrialized-country GDP countries has raised over $600 billion, has continued to decline since 1991, approximately two-thirds of global pri- and is now probably below 5 percent. vatization activity. Western Europe has The low-income countries show an accounted for over half of these pro- even more dramatic reduction in state ceeds. Finally, Jeffry Davis, Rolando Os- ownership. From a high point of almost sowski, Thomas Richardson, and Steven 16 percent of GDP, the average SOE Barnett (2000) report for a sample of share of national output dropped to transition and non-transition countries barely 7 percent in 1995, and has prob- that privatization proceeds were an average ably dropped to about 5 percent since of one and three-quarters percent of GDP. then. The middle-income countries also The historical discussion suggests that experienced significant reductions in state ownership has been substantially state ownership during the 1990s. Since reduced since 1979, and in most coun- the upper- and lower-middle-income tries this has in fact occurred. Using data groups include the transition economies from Eytan Sheshinski and Luis Felipe of central and eastern Europe, this de- López-Calva (1999), figure 2 demonstrates cline was expected, given the extremely 328 Journal of Economic Literature, Vol. XXXIX (June 2001) SOE as % of GDP 18 16 14 12 Low income 10 Lower-middle income 8 Upper-middle income High income 6 4 2 0 1980 1985 1990 1995 Figure 2. SOE Share of GDP by State of National Development, 1979–96 Source: World Bank, as reported in Sheshinski and López-Calva (1999). high beginning levels of state ownership. economists, this debate now spans many For example, Nemat Shafik (1995) reports areas, including welfare economics, that the Czechoslovakian government public choice, public finance, industrial owned 98 percent of all property in 1989. organization, law and economics, corpo- rate finance, and macroeconomics. In 3. Why Have Governments Embraced this section, we summarize some of the Privatization? important theoretical issues that arise in the study of privatization and that are 3.1 Efficiency of State vs Private needed to analyze the empirical evi- Ownership: Theory dence we review in the rest of the pa- Throughout history, scholars, includ- per. We concentrate on empirical evi- ing economists, have debated the role dence because, as Jean-Jacques Laffont of government in the economy. 8 Among and Jean Tirole (1993) say after pre- 8 For example, Friedrich von Hayek’s (1994) senting their model analyzing trade-offs passionate critiques of the welfare state and col- between government and private own- lectivism, exemplified in the 1944 book The Road ership in promoting efficiency, “theory to Serfdom, had a direct impact on policymakers in alone is thus unlikely to be conclusive developing a motive for privatization. Yergin and Stanislaw (1998, pp. 98–107) discuss how Hayek’s in this respect.” There are also several work was the intellectual basis for Keith Joseph excellent articles that discuss the the- and then Thatcher and the Tory politicians who ory of privatization and review the lit- began the intellectual campaign against statism in the U.K. that triggered the worldwide privatiza- erature, including Anthony Boardman tion movement. and Aidan Vining (1989), John Vickers Megginson and Netter: From State to Market 329 and George Yarrow (1991), Shleifer Intellectual arguments for govern- (1998), Oleh Havrylyshyn and Donald ment intervention based on efficiency McGettigan (2000), John Nellis (1999, considerations have been made in many 2000), Sheshinski and López-Calva areas. Governments perceive the need (1999), Simeon Djankov and Peter to regulate (or own) natural monopolies Murrell (2000a,b) and Shirley and or other monopolies, intervene in the Patrick Walsh (2000). case of externalities (such as regulating The economic theory of privatization pollution), and help provide public is a subset of the large literature on the goods (such as providing national de- economics of ownership and the role fense and education, or in areas where for government ownership (or regula- there is a public good aspect to provid- tion) of productive resources. An initial ing information). The arguments for question to be asked is “what is the government intervention become more proper role of government?” Implicitly, complicated when they extend to distri- we assume that the goal of government butional concerns. For example, some is to promote efficiency. Thus, we dis- argue that the role of government is to cuss the efficiency implications of gov- act as a “welfare state” (A. Briggs 1961), ernment ownership and, more impor- using state intervention in the market tantly, the movement from government economy to modify the actions of the ownership to privatization. To a large market. 9 Thus, the arguments for state extent we ignore the arguments regard- ownership or control rest on some ac- ing the importance of equitable con- tual or perceived market failure, and cerns such as income distribution, be- countries have often responded to mar- cause they are beyond the scope of this ket failure with state ownership. Privati- review. The effects of privatization on zation, in turn, is a response to the productive efficiency, or at least observ- failings of state ownership. Some able variables that are proxies for pro- theoretical arguments that have arisen in ductive efficiency, is the focus of most of the privatization debate are discussed the empirical literature we review here. next. The theoretical arguments for the ad- 3.1.1. The impact of privatization de- vantages of private ownership of the pends on the degree of market failure. means of production are based on a As noted above, welfare theory (ignor- fundamental theorem of welfare eco- ing the theory of second best) argues nomics: Under strong assumptions, a that privatization tends to have the competitive equilibrium is pareto opti- greatest positive impact in cases where mal. However, the assumptions include the role for government in lessening requirements that there are no exter- market failure is the weakest, i.e., for nalities in production or consumption, SOEs in competitive markets or mar- that the product is not a public good, kets that can readily become competi- that the market is not monopolistic in tive. Sheshinski and López-Calva structure, and that information costs (1999), in summarizing the theoreti- are low. Thus, a theoretical argument cal literature, argue that there should for government intervention based on be “. . . important efficiency gains efficiency grounds rests on an argument from changes to private ownership in that markets have failed in some way, 9 I. Gough (1989) notes that Briggs (1961) one or more of these assumptions do claims that Archbishop Temple first used the term not hold, and that the government can in wartime Britain to differentiate Britain from the resolve the market failure. “warfare” state of Nazi Germany. 330 Journal of Economic Literature, Vol. XXXIX (June 2001) competitive structures.” In fact, the effects the government’s goals can be inconsis- of competition can be so strong that tent with efficiency and maximizing SOEs, in an increasingly global environ- social welfare, or even malevolent ment, may be forced to respond to pres- (see Laffont and Tirole 1993; Shleifer sures that maximize productive effi- 1998). ciency without the ownership change of In addition, even if the government and privatization. (Shirley and Walsh 2000 the nation’s citizens agree that profit provide additional discussion of the ef- maximization is the goal of the firm, it fects of competition on the privatization is difficult to write complete contracts decision.) that adequately tie managers’ incentives In contrast, the justification for priva- to that goal. Shleifer (1998) argues that tization is less compelling in markets the owners of public firms (the nation’s for public goods and natural monopo- citizens) are less able to write complete lies where competitive considerations contracts with their managers because are weaker. However, Shleifer (1998) of diffuse ownership, making it difficult and others have argued that even in to tie the managers’ incentives to the those markets, government-owned firms returns from their decisions. This is a are rarely the appropriate solution, for subset of the broader arguments, based many of the reasons discussed below. on property rights and agency costs, 3.1.2. Contracting ability impacts the that there will be differences in perfor- efficiency of state and private owner- mance between government and pri- ship. Government ownership of firms vately held firms because there is a results in problems in defining the goals broader range of monitoring devices of the firm. While the shareholder- under private ownership.11 wealth-maximizing model of corporate 3.1.3. Ownership structure affects the organization is becoming increasingly ease with which government can inter- dominant in part because of the advan- vene in firm operations. Governments tages of having a well-defined corporate can intervene in the operations of any goal (see Henry Hansmann and Reinier firm, either public or private. However, Kraakman 2000), governments have other the government’s transaction costs of objectives than profit or shareholder- intervening in production arrangements wealth maximization. Further, these ob- and other decisions of the firm are jectives can change from one adminis- greater when firms are privately owned. tration to the next. Government’s in- Thus, to the extent that government in- ability to credibly commit to a policy tervention has greater costs than benefits, can significantly reduce the efficiency private ownership is preferred to public of an SOE’s operations and governance. ownership (see David Sappington and Even if the government does attempt to Joseph Stiglitz 1987). maximize social welfare, for example, 3.1.4. A major source of inefficiency in welfare is a difficult thing to measure public firms stems from less-prosperous and use in guiding policy. 10 In addition, firms being allowed to rely on the 10 Stiglitz (1998) provides an insightful analysis, 11 Armen Alchain (1977, p. 36) notes, “behavior based on personal experience, of the difficulty under [public and private] ownership is different, governments face in implementing pareto- not because the objectives sought by organizations efficient improvements due to information costs under each form are different, but, instead, be- and the problems of commitment and dynamic cause even with the same explicit organization bargaining. These arguments apply to both gov- goals, the costs-rewards system impinging on the ernment regulation (the main case Stiglitz ana- employees and the ‘owners’ of the organization are lyzes) and to state ownership. different.” Megginson and Netter: From State to Market 331 government for funding, leading to tiveness of privatization programs and “soft” budget constraints. The state is markets themselves are simultaneously unlikely to allow a large SOE to face determined. It has been clear in the bankruptcy. Thus the discipline en- transition economies that the success of forced on private firms by capital mar- privatization depends on the strength of kets and the threat of financial distress the markets within the economies, and is less important for state-owned firms. vice versa. Thus, the impact of privati- János Kornai (1988, 1993, 2000), Eric zation will differ across countries de- Berglof and Gérard Roland (1998), and pending on the strength of the existing Roman Frydman, Cheryl Gray, Marek private sector. Similarly, evidence sug- Hessel, and Andrzej Rapaczynski (2000) gests that the effectiveness of privatiza- all suggest that soft budget constraints tion depends on institutional factors were a major source of inefficiency in such as the protection of investors. How- communist firms. They also note that ever, privatization can also stimulate supposedly “hard” budget constraints the development of institutions that imposed on SOEs by government are improve market operations. not very effective either. 3.2 Summary of Privatization Theory 3.1.5. Privatization can impact effi- ciency through its effect on government Theoretical work that examines priva- fiscal conditions. As noted in section 1, tization offers many reasons why, even governments have raised huge amounts in the case of market failure, state own- of money by selling SOEs. Such sales ership has important weaknesses. As have helped reduce the fiscal deficit in Shleifer (1998) sums up much of the lit- many countries. Though important, ex- erature, “. . . a good government that amining the efficiency effects of reduc- wants to further ‘social goals’ would rarely ing government deficits is beyond the own producers to meet its objectives.” scope of this paper. Davis et al. (2000) A question for the post-privatization review the evidence on the macro- world is the role of the public sector in economic effects of privatization, dis- the economy and in the regulation of cuss the difficulties of using macro- firms. The alternative to state owner- economic privatization data and report ship is rarely purely private, unregu- some evidence on the effects from lated firms. State ownership is only one eighteen developing countries. They find form of the continuum of governance evidence that the proceeds from privati- structures that reflect the level of state zation are saved by governments and not regulation of public and privately used to increase government spending. owned firms (Laffont and Tirole 1993). 3.1.6. At a macroeconomic level, pri- Many of the theoretical arguments for vatization can help develop product and privatization are based on the premise security markets and institutions. One that the harmful effects of state inter- important motivation for privatization is vention have a greater impact under to help develop factor and product mar- state ownership than under state regu- kets, as well as security markets. As dis- lation, not that the harmful effects can cussed above, welfare economics argues be eliminated through privatization. that efficiency is achieved through com- However, in this paper we leave to oth- petitive markets. Thus, to the extent ers the continuing debate on the proper that privatization promotes competi- role of regulation in a market-oriented tion, privatization can have important economy. Instead, we analyze recent efficiency effects. Inevitably, the effec- empirical literature examining the 332 Journal of Economic Literature, Vol. XXXIX (June 2001) relative effectiveness of state versus gree of perceived market failure within private ownership. 12 the particular industry. These factors that determine whether the firm is pub- 3.3 Efficiency of State vs Private licly or privately owned likely also have Ownership: Empirical Evidence significant effects on performance. Thus, it is difficult to evaluate the ef- Comparing the performance of fects of government ownership where government-owned to privately owned the ownership structure is itself en- firms is one method through which the dogenous to the system that includes impact of government ownership on both political and performance goals. firm performance can be analyzed. 13 In Despite these problems, researchers section 5 we present a more complete have compared SOE and private firm discussion of the potential problems in performance in several cases with some all empirical work in this area, which in- success. We summarize the papers cludes lack of data and bad data, omit- included here in table 1. ted variables, endogeneity, and selection Given the above noted limitations, Is- bias. There are two methodological dif- sac Ehrlich, George Gallais-Hamonno, ficulties that are especially pronounced Zhiqiang Liu, and Randall Lutter in attempts to isolate the impact of (1994) provide good evidence on pro- ownership on performance. First, in ductivity differences between state- comparing SOEs to privately owned owned and privately owned firms. They firms, it is difficult, if not impossible, to use a sample of 23 comparable interna- determine the appropriate set of com- tional airlines of different (and in some parison firms or benchmarks, especially cases changing) ownership categories in developing economies with limited over the period 1973–83 for which they private sectors. Second, there are gen- are able to obtain good and comparable erally fundamental reasons why certain cost, output, and ownership data. They firms are government owned and others develop a model of endogenous, firm- are privately owned, including the de- specific productivity growth as a func- 12 The opinions of policymakers throughout the tion of firm-specific capital and use the world have been moving closer to those expressed model as a basis for their fixed-effects by Ronald Coase in his classic 1960 article, “The Problem of Social Cost.” In analyzing market fail- regressions estimating a cost function in ure, Coase says, “All solutions have costs, and a simultaneous framework with input- there is no reason to suppose that governmental demand equations. They argue that regulation is called for simply because the prob- lem is not handled well by the market or the firm.” they are able to separate the impact of James Brickley, Clifford Smith, and Jerold Zim- ownership changes on short-term levels merman (2001, p. 54), in a more recent analysis, of productivity changes from the long- say markets have worked better because, “First, the price system motivates better use of knowl- term effects on the rate of productivity edge and information in economic decisions. Sec- growth, improving on earlier studies ond, it provides stronger incentives for individuals that concentrated on static rather than to make productive decisions.” 13 A related literature that we do not review ana- dynamic effects of, and changes in, lyzes the relative performance of nonprofit firms state ownership. Further, they suggest and for-profit firms. James Brickley and R. they are able to isolate the effects of Lawrence Van Horn (2000), in an analysis of large hospitals, argue that the evidence suggests there is ownership from other factors impact- little distinction between the behavior of nonprofit ing the productivity growth rate, includ- and for-profit hospitals. Their results suggest the ing market conditions and exogenous similarities in behavior are due to the effects of competition and not identical objective functions technical changes. of the managers. Ehrlich et al. (1994) find a significant TABLE 1 RECENT EMPIRICAL STUDIES ON PUBLIC VS PRIVATE OWNERSHIP Study Sample description, study period, and methodology Summary of findings and conclusions Boardman Examines economic performance of 500 largest non- SOEs and MEs are significantly less profitable and and Vining US firms in 1983, classified by ownership structure productive than private firms. MEs are no more prof- 1989 as SOE, private, or mixed (ME). Employs 4 itable than pure SOEs—so full private ownership is re- profitability ratios and 2 measures of X-efficiency. quired to gain efficiency. Vining and Asks whether ownership “matters” in determining After controlling for size, market share and other fac- Boardman efficiency of SOEs, or if only the degree of com- tors, private firms are significantly more profitable 1992 petition is important. Estimates performance and efficient than MEs and SOEs, though now find model using 1986 data from 500 largest nonfinan- that MEs outperform SOEs. Thus, ownership has an cial Canadian firms, including 12 SOEs and 93 effect separable from competition alone. MEs. Pinto, Belka, Tests whether privatization is required to improve Significant performance improvement due to macro- and Krajewski performance of SOEs by examining how Polish state economic stabilization package, even without priva- 1993 sector responded in the 3 years after “Big Bang” re- tization; mostly due to hard budget constraints, tight forms of Jan. 1990, which liberalized prices, tight- bank lending policies, enhanced credibility of govern- ened fiscal/monetary policy and introduced competi- ment’s “no bailout” pledge. tion, without privatization. Ehrlich, Examines impact of state ownership on long-run rate State ownership can lower long-run annual rate of Gallais- of productivity growth and/or cost decline for 23 in- productivity growth by 1.6–2.0% and rate of unit cost Hamonno, ternational airlines during 1973–83. by 1.7–1.9%. Ownership effects not affected by de- Liu, Lutter gree of competition. 1994 Majumdar Using industry-level survey data, compares perfor- Documents efficiency scores averaging 0.975 for pri- 1996 mance of SOEs, MEs, and private Indian firms for vate firms, significantly higher than averages of 0.912 1973–89. SOEs and MEs account for 37% of employ- for MEs and 0.638 for SOEs. State sector efficiency ment and 66% of capital investment in India in 1989. improves during “efficiency drives” but declines afterwards. Kole and Tests whether postwar performance of 17 firms partly Though these firms experience abnormally high turn- Mulherin 1997 owned by US government due to seizure of “enemy” over among boards of directors, manager tenure is property during WWII differs significantly from per- stable, and SOE performance is not significantly dif- formance of private US firms. ferent from private firms. Dewenter and Tests whether profitability, labor intensity, and debt After controlling for business cycles, finds private Malatesta levels of SOEs listed among 500 largest non-US firms significantly (often dramatically) more profitable, 2001 firms in 1975, 1985, and 1995 differ from private have significantly less debt, and less labor intensive firms on same lists. production processes than SOEs. LaPorta, Using data from 92 countries, examines whether Extensive state ownership, especially in poorest coun- Lopez-de- state ownership of banks impacts financial system de- tries, retards financial system development and re- Silanes, and velopment and growth rates of economy and produc- stricts economic growth rates, mostly due to impact Shleifer 2000a tivity. on productivity. Tian 2000 Studies relation between state shareholding and firm Performance of “private” enterprises significantly su- performance of 825 publicly traded Chinese firms in perior to “mixed” enterprises. Corporate value gen- 1998. 413 had some government ownership, 312 had erally declines with state ownership, then increases none. after state share passes 45%. Karpoff 2001 Examines 35 government financed and 57 privately Private expeditions performed better using several funded expeditions to the Arctic from 1819–1909. measures of performance. More major discoveries were made by private expeditions; most tragedies oc- curred on government-sponsored expeditions. Ro- bust results in regressions explaining expedition outcomes. 334 Journal of Economic Literature, Vol. XXXIX (June 2001) link between ownership and firm-specific between ownership and productivity, rates of productivity growth. Their re- and find that causality goes from owner- sults suggest that private ownership ship to productivity, and not vice versa. leads to higher rates of productivity The weakness in the work is that it is growth and declining costs in the long based on one industry with relatively run, and these differences are not af- old data. The authors also note that fected by the degree of market compe- they make the implicit assumption that tition or regulation. Their estimates all firms are cost minimizing, but if suggest that the short-run effects of state-owned enterprises have other ob- changes from state to private ownership jectives, it is difficult to interpret the on productivity and costs are ambigu- meaning of differences in costs. ous, providing a possible explanation Sumit Majumdar (1996) examines for some of the anomalous results in differences in efficiency between studies. However, their point estimates government-owned, mixed, and private- indicate that the change from complete sector firms in India. He finds support state to private ownership in the long for the superior efficiency of private run would increase productivity growth and mixed-sector firms over SOEs. by 1.6 to 2 percent a year, while costs Using aggregate, industry-level survey would decline by 1.7 to 1.9 percent. data, Majumdar finds that SOEs owned Their empirics also suggest that a par- by the central and state governments tial change from state to private owner- have average efficiency scores of 0.658 ship has little effect on long-run pro- and 0.638, respectively, over the period ductivity growth—the benefits are 1973–89. Mixed enterprises score 0.92, based on complete privatization of the and private enterprises score 0.975. A firm. concern with Majumdar’s study is that This paper has advantages over much the aggregated nature of the data, along of the other work in the area due to the with problems arising from the reliance good data, as well as guidance from a on survey data, limits his ability to iden- well-developed literature in estimating tify any specific areas where private the determinants of productivity. The versus state ownership works best, and authors perform some of the more so- whether there are simultaneity and se- phisticated econometric analysis of pa- lection bias problems in trying to esti- pers in this area. For example, they rep- mate the effects of ownership and pro- licate their results with a subset of firms ductivity. In addition, he can provide that did not experience any within-firm little insight into the reasons for the changes in ownership, enabling the efficiency differences between the authors to be sure that their time- sectors. ownership interaction term captures George Tian (2000) offers another only between-firm variations in owner- country-specific study. He examines 825 ship. Ehrlich et al. also perform various companies listed on the Shanghai Stock other robustness checks using different Exchange, with 513 mixed-ownership specifications and subsamples, as well firms and 312 private firms. He finds that as controlling for the special charac- private firms perform better than mixed teristics of their sample period (oil ownership firms. In addition, he exam- price shocks and deregulation in the ines the valuation of the companies and United States), and find that their re- finds that corporate value with small sults are robust. Finally, they consider government shareholdings decreases the potential for simultaneity effects with the fraction of state shareholding Megginson and Netter: From State to Market 335 but rises when the government is a U.S. firms until 1995, the data are large shareholder. mainly international. After controlling Another approach to studying the ef- for firm size, location, industry, and fects of government ownership on effi- business-cycle effects, Dewenter and ciency relies on a multi-industry, multi- Malatesta find robust evidence that pri- national, time-series methodology. While vate companies are significantly (often cross-sectional time series studies suffer dramatically) more profitable than from methodological problems we dis- SOEs, and also have lower levels of in- cuss later, they are able to capture dif- debtedness and fewer labor-intensive ferences that are not apparent in single- production processes than do their country or single-industry studies. An state-owned counterparts. influential paper taking this approach is Finally, Frydman, Gray, Hessel, and Anthony Boardman and Aidan R. Vining Rapaczynski (1999) compare the perfor- (1989) who examine the economic per- mance of privatized and state firms in formance of the 500 largest non-U.S. the transition economies of Central industrial firms in 1983. Using four Europe, and explicitly try to control for profitability ratios and two measures of selection bias. 14 Using survey data for X-efficiency, they show that state- 506 midsize manufacturing firms in the owned and mixed (state and private) Czech Republic, Hungary, and Poland ownership enterprises are significantly in 1994, they compare four measures of less profitable and productive than are firm performance—sales revenues, em- privately owned firms. They also find ployment, labor productivity (revenue that mixed enterprises are no more per employee) and material costs per profitable than SOEs, suggesting that unit of revenue. They compare the pri- full private control, not just partial own- vatized group to the nonprivatized ership, is essential to achieving perfor- group with panel data, controlling for mance improvement. In a later study, potential pre-privatization differences Vining and Boardman (1992) use a sam- between the two groups. Frydman, ple of Canadian firms to re-examine the Gray, Hessel, and Rapaczynski find that state versus private ownership question. the average effect of privatization is Their results are qualitatively similar to that it works—privatized firms perform their earlier findings. In addition, the better than the state owned firms. How- Canadian study finds that mixed enter- ever, the performance improvement is prises are more profitable than SOEs, concentrated in revenue improvement though they fall far short of private-firm (not cost reduction) in firms privatized levels. to outside owners. Kathryn Dewenter and Paul Malatesta Frydman et al. (1999) make two im- (2001) follow the general approach of portant contributions. First, they show Boardman and Vining (1989) using more that while privatization improves per- recent data. They test whether the prof- formance, the effect is limited to cer- itability, labor intensity, and debt levels tain measures of performance and cases of SOEs in the 500 largest international where the SOE is sold to outside owners. companies, as reported in Fortune for Second, they attempt to control for the 1975, 1985, and 1995, differ from pri- effects of selection bias in examining vately owned firms in the same samples. 14 Frydman et al. also compare the performance Their data have 1,369 total firm years, of the privatized firms to that of the firms when of which 147 represent government- they were SOEs. Thus, we also discuss the paper owned firms. Since Fortune excluded in section 5 and it is summarized in table 5. 336 Journal of Economic Literature, Vol. XXXIX (June 2001) the effects of privatization in several firms and state “the preceding results ways. They use a fixed effects model to stand in contrast to the typical results control for selection bias caused by un- regarding the inefficiency of govern- observed firm characteristics correlated ment enterprise.” The authors argue with performance outcomes that are that the fact that these firms were oper- fixed over time. Further, they contrast ating in competitive industries forced the performance of firms privatized in them to operate efficiently. one period with those privatized in an- The Kole and Mulherin (1997) re- other for two different time periods to sults are evidence that in a competitive compare the privatized firms with how environment, where the government they would have performed without has no agenda other than as a passive privatization. Finally, to control par- investor, factors other than ownership tially for the possibility that better determine firm performance. Many of firms are selected for privatization, they the firms were involved in the war ef- contrast the pre-privatization perfor- fort, so the government had an incen- mance of managerially controlled firms tive to run them efficiently. In addition, with those controlled by other owners. all the firms were eventually repriva- Thus, the paper does an excellent tized, so the government was also con- job of controlling for potential biases, cerned with running them efficiently to though it necessarily depends on survey maximize the later sale value. Kole and data. Mulherin admit that their sample and We conclude this section with two the period they study is novel, limiting studies that use unique situations to its generality. Further, their results are analyze the effects of government ver- based on only five firms. Still, their find- sus private ownership. Stacey Kole and ings do illustrate the importance of factors J. Harold Mulherin (1997) set out to an- other than ownership in determining swer the basic question in the public firm performance. versus private debate as posed by Sam In a paper featuring a very interest- Peltzman (1971), “If a privately owned ing natural experiment, Jonathan Kar- firm is socialized, and nothing else hap- poff (2001) studies a comprehensive pens, how will the ownership alone af- sample of 35 government-funded and fect the firm’s behavior?” Kole and 57 privately-funded expeditions to the Mulherin study seventeen firms with Arctic from 1818 to 1909 seeking to lo- significant German or Japanese owner- cate and navigate a northwest passage, ship when the United States entered discover the North Pole, and make World War II. The U.S. government as- other discoveries in arctic regions. Kar- sumed ownership of the foreign stock in poff finds that the private expeditions these firms and ended up holding be- performed better using several measures tween 35 and 100 percent of the com- of performance. He shows most major mon stock for up to 23 years during and arctic discoveries were made by private after World War II. Kole and Mulherin expeditions, while most tragedies (lost find industry controls for five firms, ships and lives) were on publicly comprising 61 percent of the book funded expeditions. He notes the fact value of the seventeen firms, and com- that the public expeditions had greater pare the performance of the government- losses could mean the public expedi- owned firms. They find no significant tions took greater risks, but then the difference between the performance of public expeditions would have had a their sample with the private-sector greater share of discoveries, which did Megginson and Netter: From State to Market 337 not occur. He also estimates regressions been implemented instead of, or prior explaining outcomes in several ways to, full privatization. 16 (crew deaths, ships lost, tonnage of Brian Pinto, Merek Belka, and Stefan ships lost, incidence of scurvy, level of Krajewski (1993) examine the way in expedition accomplishment), control- which the Polish state sector responded ling for exploratory objectives sought, in the three years following Poland’s country of origin, the leader’s previous “Big Bang” reforms of January 1990. arctic experience, or the decade in These reforms deregulated prices, in- which the expedition occurred. In es- troduced foreign competition to many sentially every regression, the dummy industries, and signaled that tight variable for private expedition is signifi- monetary and fiscal policies would be cant with a sign indicating that the pri- pursued. However, the Polish govern- vate expedition performed better. Kar- ment did not immediately launch a poff concludes that the incentives were large-scale privatization program. The better aligned in the private expeditions, authors document significant perfor- leading to systematic differences in the mance improvements on the part of ways public and private expeditions were most manufacturing firms. They con- organized. While the uniqueness of the clude that these improvements were sample limits its generality, he provides due to the imposition of hard budget an interesting illustration of the impact constraints reinforced by tighter bank of ownership on the performance of an lending behavior, consistency in the organization. 15 government’s “no bailout signal,” im- port competition, and reputational 3.4 Policy Alternatives to Privatization concerns of SOE managers. The use of incentive contracts for As discussed earlier, some argue that management and workers is potentially competition and deregulation are more the best way to improve performance in important than privatization or gover- SOEs (Leroy Jones 1991). The World nance changes in improving perfor- Bank endorsed these contracts in the mance of firms (George Yarrow 1986; 1980s. China has undergone widespread John Kay and D. J. Thompson 1986; economic reform with minimal privati- Matthew Bishop and Kay 1989; John zation through the use of these incentive Vickers and Yarrow 1991; Franklin contracts and offers a natural setting in Allen and Douglas Gale 1999). which to study their impact. Others maintain that privatization is Theodore Groves, Yongmiao Hong, necessary for significant performance John McMillan, and Barry Naughton improvements (Vining and Boardman (1995) discuss the ways incentives were 1992; Maxim Boycko, Shleifer, and added to the Chinese managerial labor Robert Vishny 1994, 1996a,b; John Nel- market by the late 1980s, including lis 1994; Josef Brada 1996; and Shleifer replacement after poor performance 1998). Although much of this debate is and linking managerial pay to profits. outside the scope of this paper, there Further, managers were selected by are a few empirical studies that examine auctions, where the auction process countries where economic reform has 15 Kelly Olds (1994) also uses data from the 16 Majumdar (1996) also suggests that reform 1800s to show that after the privatization of the can improve SOE performance by showing that tax-supported Congregationalist churches in New the gap between the private and public firms’ per- England, demand for preachers and church mem- formance partly closes during those periods when bership rose dramatically. governments are pushing reform agendas. 338 Journal of Economic Literature, Vol. XXXIX (June 2001) revealed information about the manag- especially in light of the evidence from ers that in a market economy could the studies of Chinese firms. have come from observations of their The evidence from China suggests performance. Groves, Hong, McMillan, that enterprise restructuring, concen- and Naughton (1994) show that after trating on improving the allocation of 1978, when Chinese firms were given property rights and incentives can yield more autonomy and allowed to retain large benefits even without privatiza- more profits and to increase workers’ tion. 17 Naturally, this begs the question incentives through bonuses and differing whether economic reform coupled with work contracts, there were increases privatization could lead to even greater in workers’ incomes (though not of performance improvements. Unfortu- managers’) and in investment in the nately, this is little evidence on this firms. question and it would be very difficult Wei Li (1997) documents marked im- to develop such evidence. Note also that provements in the marginal and total the evidence on the benefits of reform factor productivity of 272 Chinese without privatization comes primarily SOEs over the period 1980–89 as a re- from one country where country-specific sult of economic reforms in China, in- factors may play an important but un- cluding the increased use of incentives. identified role. One thing we can say is He finds evidence of substantial in- that, as we note later in the paper, the creases in productivity over the reform evidence demonstrating the benefits of period, much of which can be attrib- privatization is weakest for countries in uted to the reform. In addition, his evi- eastern Europe, where privatization was dence suggests that 87 percent of the implemented rapidly. This may suggest growth in productivity was due to im- that privatization should have pro- proved incentives and compensation. Li ceeded along a more gradual path. We notes, however, the potential for selec- address that question later on. tion bias in his study both in the firms selected for the survey and in the 4. How Do Countries Privatize? responses to the survey. A key decision to be made by the pri- Shirley and Lixin Xu (1998) come to vatizing government is on the method the opposite conclusion concerning the of transferring the state-owned asset to ability of incentive contracts to improve private ownership. This decision is diffi- firm performance. They analyze the ef- cult because, in addition to the eco- fects of these contracts in twelve mo- nomic factors such as valuing the assets, nopoly SOEs, and find that the incentive privatizations are generally part of an contracts have no effect on profitability ongoing, highly politicized process. or labor productivity; they also find Some of the factors that influence the some evidence of negative effects on privatization method include: (1) the growth in total factor productivity. They history of the asset’s ownership, (2) attribute the failure of the contracts to the financial and competitive position the inability of governments to follow of the SOE, (3) the government’s ideo- through on promised actions and the in- logical view of markets and regulation, ability of supervisory agencies to negoti- ate and monitor the contracts effec- 17 This is consistent with the findings of Brickley tively. It must be noted, however, that and Van Horn (2000) that the managers of non- profit hospitals face similar incentives to the man- the study is based on a small sample, agers of for-profit hospitals and behave in a similar limiting the ability to draw conclusions, manner. Megginson and Netter: From State to Market 339 (4) the past, present, and potential fu- ated in years past can be returned to ture regulatory structure in the country, either the original owners or to their heirs. (5) the need to pay off important inter- This method is rarely observed outside est groups in the privatization, (6) the eastern Europe, though it has been im- government’s ability to credibly commit portant there. For example, Brada (1996) itself to respect investors’ property rights reports that up to 10 percent of the after divestiture, (7) the capital market value of state property in the Czech Re- conditions and existing institutional public consisted of restitution claims. The framework for corporate governance in major difficulty with this method is that the country, (8) the sophistication of the records needed to prove ownership potential investors, and (9) the govern- are often inadequate or conflicting. ment’s willingness to let foreigners own The second method is privatization divested assets. through sale of state property, where a The complexity of goals means that government trades its ownership claim countries have used various methods for for an explicit cash payment. This cate- privatizing different types of assets. Al- gory takes two important forms. The though financial economists have first is direct sales (or asset sales) of learned much about selling assets in state-owned enterprises (or some parts well-developed capital markets, we still thereof) to an individual, an existing have a limited understanding of the de- corporation, or a group of investors. terminants and implications of the pri- The second form is share issue privati- vatization method for state-owned as- zations (SIPs), in which some or all of a sets. Theoreticians have modeled some government’s stake in an SOE is sold to aspects of the privatization process, but investors through a public share offer- to be tractable, their models must ig- ing. These are similar to IPOs in the nore important factors. Empirical evi- private sector, but where private IPOs dence on the determinants of privatiza- are structured primarily to raise reve- tion is also limited by the complexity of nue, SIPs are structured to raise money the goals of the privatization process. and to respond to some of the political factors mentioned earlier. 4.1 Methods of Privatization Brada’s third category is mass or Brada (1996) presents an excellent voucher privatization, whereby eligible taxonomy of privatization methods. Al- citizens can use vouchers that are dis- though the context of his paper is cen- tributed free or at nominal cost to bid tral and eastern Europe, his classifica- for stakes in SOEs or other assets. This tion of four principal divestment method has been used only in the tran- methods is quite general. In addition, sition economies of central and eastern he provides a review of the successes Europe, where it has brought about and failures of each of these general ap- fundamental changes in the ownership proaches in central and eastern Europe. of business assets in those countries, al- Of course, there are many variations though it has not always changed effec- within each of his categories, and he tive control. Longer descriptions of the shows that many privatizations use issues that these governments have con- combinations of different methods. fronted when designing voucher privati- Brada’s first category is privatization zation programs are provided in Morris through restitution. This method is ap- Bornstein (1994, 1999), Melinda Alex- propriate when land or other easily androwicz (1994), Bernard Drum (1994) identifiable property that was expropri- and Shafik (1995). 340 Journal of Economic Literature, Vol. XXXIX (June 2001) The final method is privatization the government. Others who examine from below, through the startup of new non-pricing issues relating to actual private businesses in formerly socialist divestment contracts include Carliss countries. Havrylyshyn and McGettigan Baldwin and Sugato Bhattacharya (2000) stress the importance of this (1991), Rondinelli and Iacono (1996), type of economic growth in the transi- Klaus Schmidt (1996), Shafik (1996), tion economies. Although privatization and Francesca Cornelli and David Li from below has progressed rapidly in (1997). many regions (including China, the Two empirical papers analyze the transition economies of central and choice of privatization method. One ex- eastern Europe, Latin America, and plicitly studies the choice between an as- sub-Saharan Africa), a survey of this set sale and a share issue privatization. phenomenon is beyond the scope of our Using a sample of 1,992 privatizations paper. that raised $720 billion in 92 countries, There are many other methods be- Megginson, Nash, Netter, and Annette sides the four described above that gov- Poulsen (2000) examine why 767 firms ernments can use to increase private- were divested using share offerings (in sector participation. For example, the public capital markets), but 1225 com- term “privatization” in the United States panies were privatized via direct sales means something different from any of (in private markets). They find robust these strategies. As López-de-Silanes, results that the choice is influenced by Shleifer, and Vishny (1997) show, the pri- capital market, political, and firm-specific vatization debate in the United States re- factors, and report that SIPs are more fers to the choice between provision of likely to be used when capital markets goods and services by (state and local) are less developed, presumably as a way government employees and the contract- to develop capital markets, and when ing out of that production to private there is less income inequality. SIPs are firms. Their empirical study finds that also more likely the larger the size of the more binding are state fiscal con- the offering and the more profitable the straints and the less powerful are public- SOE. On the other hand, governments sector unions, the greater the likelihood with greater ability to commit to prop- of privatization. erty rights are more likely to privatize via asset sales. Perhaps the most inter- 4.2 The Choice of Sale Method esting result is that governments choose Henry Gibbon (1997) provides one of to privatize the more profitable SOEs the most helpful delineations of the deci- through SIPs—evidence supporting the sions facing a government that wants to possibility of sample selection bias in privatize through cash sales. Gibbon dis- studies of performance of privatized cusses the steps such a government must firms. In the second paper, Bernardo take in developing a divestment program. Bortolotti, Marcella Fantini, and These include setting up a structure for Domenico Siniscalco (1999a) estimate privatization (including legislation, if nec- the determinants of the fraction of pri- essary), providing adequate performance vatization revenues that come from records for SOEs being sold (generating public offerings (SIPs) for privatizations believable accounting data), developing in 49 countries. They find that the any necessary new regulatory structures, greater the selling government’s deficit and determining the appropriate post- and the more conservative the selling sale relationship between the firm and government, the more likely it is that Megginson and Netter: From State to Market 341 privatization will occur through public and Megginson, Nash, Netter, and offerings. However, SIPs are less likely Poulsen (2000). in French civil law countries. Bortolotti, A related practical question about Fantini, and Carlo Scapa (2000) exam- privatization is whether governments ine factors that lead countries to sell should restructure SOEs (e.g., lay off shares in SOEs abroad. redundant workers) prior to selling or leave this to the new owners. This is re- lated to questions discussed in section 4.3 Restructuring SOEs, and 3.4: can governments reform SOEs (in- Sequencing and Staging of Sales cluding reform without privatization) Some of the most complex issues in- and should reform and privatization volve the interrelated questions of proceed quickly or slowly? Early advice when to privatize and at what pace, from the World Bank (John Nellis and what order to follow in privatizing Sunita Kikeri 1989) was that govern- (sequencing), whether to sell an SOE ments should restructure SOEs prior to all at once or in stages (staging), divestment, since governments are bet- whether to restructure an SOE prior ter able than private owners to cushion to sale, and the role of macroeconomic the financial blow to displaced workers reform in privatization. Since these are by using unemployment payments or complex issues that involve factors pensions. Government-led restructuring outside the scope of this article (espe- can thus provide a private buyer of the cially macroeconomic reform which we SOE with a “clean slate.” Preparing do not discuss) we do not spend much companies for privatization was stan- time on them. Further, their complex- dard practice in the United Kingdom ity has limited empirical work in this during the 1980s, in part to smooth the area. transition with the trade unions. How- Several authors have theoretically ever, by 1992, the same authors (Kikeri, modeled the sequencing and staging of Nellis, and Shirley 1992) had become SOE sales, including Barbara Katz and more nuanced in their interpretation of Joel Owen (1993, 1995), Boycko, the optimal strategy. They said (p. 54) Shleifer, and Vishny (1996b), Francesca that small and medium-sized SOEs Cornelli and David Li (1997), Enrico “should be sold ‘as is’ at the best price Perotti (1995), and Bruno Biais and possible, as quickly as possible.” They Perotti (2000). The models illustrate also noted that in all cases (p. 60) new the importance of sequencing and stag- investments “should be left to private ing to build reputational capital with owners once a decision has been made investors by the privatizing govern- to privatise the enterprise.” ment, building domestic support for Two empirical papers that examine the program, and identifying bidders SOE reform prior to privatization are that will maximize the efficiency of the López-de-Silanes (1997) and Dewenter firm. While the complexities of these and Malatesta (2000). López-de-Silanes interrelationships have limited empiri- examines whether prior government re- cists’ ability to identify factors in se- structuring of SOEs improves the net quencing and staging, several articles price received for the company, and that empirically examine them are finds evidence that it does not. He Perotti and Serhat Guney (1993), shows that prices would have increased Dewenter and Malatesta (1997), Jones, by 71 cents per dollar of assets if the Megginson, Nash, and Netter (1999), only restructuring step taken had been 342 Journal of Economic Literature, Vol. XXXIX (June 2001) to fire the CEO and if the assets had of Mexican privatizations empirically been divested an average of one year supports this theoretical conclusion that earlier. He argues that other restructur- maximizing the number of bidders in an ing steps slow down the process and open auction is usually the best way to consume too many resources to be maximize revenues. 19 He finds that worthwhile. The 71 cents per dollar in prices received are sensitive to the level added value would be a significant im- of competition in the auction process provement on the average 54 cents per but that the Mexican government fre- dollar of assets actually received. How- quently restricted participation (par- ever, this evidence is based on a small ticularly by foreigners) in spite of this sample of banks, which limits its useful- fact. Nonetheless, the amount of reve- ness. Dewenter and Malatesta (2001) nue generated was the main criteria in find some evidence that the improve- selecting the winning bidder for more ments brought about by privatization than 98 percent of the SOEs sold. occur before the SOE is privatized. Rondinelli and Iacono (1996) exam- ine auctions in central and eastern 4.4 Pricing and Allocation of Control Europe, where thousands of small busi- and Ownership nesses have been auctioned off, as well Although mass or voucher privatiza- as in Latin America and Russia, where tion programs have attracted a great larger SOEs have been sold. Many deal of academic interest, asset sales types of auctions have been used, in- and SIPs account for most of the value cluding English, Dutch, first price, sec- of assets that have been divested by ond price, double, and pro-rata sales. governments in the past two decades. 18 Auctions have been used to sell both Thus we focus on the latter two methods. lease rights and ownership rights. In other cases, governments have sold 4.4.1 Pricing Decisions in Asset Sales SOEs directly to groups of private in- Four papers study the revenue im- vestors or firms, setting prices and pact of SOE direct sale pricing deci- terms by negotiation. In some cases, the sions. At a theoretical level, Jeremy Bu- groups of investors consist of manage- low and Paul Klemperer (1996) ask ment or employees. In other cases, the whether it is more profitable to sell a government has liquidated the SOE company through an auction with no re- and sold physical assets to a group of serve price or by using an optimally investors. structured direct negotiation with one Archana Hingorani, Kenneth Lehn, less bidder. They show that under most and Anil Makhija (1997) examine an ac- conditions, a simple competitive auc- tual voucher privatization program, the tion with N + 1 bidders will yield more first round of the Czech Republic’s expected revenue than a seller could ex- mass privatization in 1991. Because the pect to earn by fully exploiting his or mechanics of how companies are di- her monopoly selling position against N vested by this government are actually bidders. López-de-Silanes’ (1997) study more similar to an asset sale than to any 18 However, it is also true that a much larger other method, we discuss their work number of companies were transferred to private here. Hingorani, Lehn, and Makhija ownership through mass privatization programs. It test whether the level of share demand, is also likely that more employees were from firms that were transferred in mass schemes than from 19 The Mexican program relied almost exclu- firms that were sold in SIPs. We thank John Nellis sively on direct sales, rather than SIPs, as its prin- for pointing this out to us. cipal divestment technique. Megginson and Netter: From State to Market 343 as measured by voucher redemptions by offer, a book-building exercise, or at a Czech citizens, effectively predicts the fixed price. If the latter, the govern- actual level of stock prices in the second- ment must decide whether the offering ary market. The authors confirm the price should be set immediately prior to predictive power of share demand, and the offer or many weeks in advance. also document that share demand is The share allocation decision requires positively related to the level of insider the government to choose whether to shareholdings and the extent of foreign favor one group of potential investors ownership in a company being sold. over another (i.e., domestic investors, They find that share demand is posi- SOE employees, or both, over foreign tively related to the level of past profit- and institutional investors). It also re- ability, which itself shows that even im- quires deciding whether to use the best perfect accounting statements convey available investment banker as lead un- useful information. Additionally, they find derwriter (regardless of nationality) or that share demand is inversely related to favor a national champion. to the firm’s market risk, which they Several papers empirically examine measure as the post-offering coefficient the choices governments make in de- of variation of stock prices. 20 signing SIP programs. Kojo Menyah and Krishna Paudyal (1996) and Men- 4.4.2 Pricing and Share and Control yah, Paudyal, and Charles Inyangete Allocation in SIPs (1995) investigate how the aims and ob- jectives of privatization influence the Any government that intends to pri- procedures and incentives used in the vatize SOEs using public share offer- sale of state-owned shares on the Lon- ings faces three sets of interrelated de- don Stock Exchange by the U.K. gov- cisions: (1) how to transfer control, (2) ernment. Jones, Megginson, Nash, and how to price the offer, and (3) how to Netter (1999), Qi Huang and Richard allocate shares. The control transfer de- Levich (1998), and Dewenter and cision includes whether to sell the SOE Malatesta (1997) present comprehen- all at once or through a series of partial sive studies of the pricing and share and sales. If the government chooses the control allocation decisions made by latter course, then it must determine governments disposing of SOEs through how large a fraction of the company’s public share offering. The results are shares to issue in the initial versus sub- broadly similar, so we concentrate on sequent offers. The government must the paper by Jones et al. (1999) since it also decide whether to insert any post- has the largest sample. privatization restrictions on corporate Jones et al. (1999), whose results are control. The pricing decision requires summarized in table 2, provide evi- that the government determine the dence on the way political factors im- amount of underpricing, and whether pact the offer pricing, share allocation, the offer price should be set by a tender and other terms in SIPs. They analyze a 20 Stijn Claessens (1997) examines the relation large sample of 630 SIPs from 59 coun- between ownership concentration and equity tries made over the period June 1977 to share prices from the voucher bidding rounds and the secondary market prices for the 1491 firms July 1997. 21 One result they document that emerged from the mass privatization voucher scheme in the Czech and Slovak Republics. He 21 Though Jones et al. rely primarily on Privati- finds that the prices are related to the resulting sation International for the data used in this ownership structure, with more concentrated own- study; one of the authors has also developed from ership associated with higher prices. secondary sources (primarily the Financial Times, 344 Journal of Economic Literature, Vol. XXXIX (June 2001) TABLE 2 PRICING, SHARE ALLOCATION, AND CONTROL ALLOCATION PATTERNS IN SIPS Sample of 630 share issue privatizations (SIPs) executed by 59 national governments during 1977–97. Measures are broken down for the 417 initial public offerings of SIP shares and the 213 seasoned SIP offerings. Initial SIPS Seasoned Offers Measure Mean Median Number Mean Median Number Pricing Variables Issue size (US$ million) 555.7 104.0 417 1,068.9 311.0 172 Initial return1 34.1 12.4 242 9.4 3.3 55 Percent of offer at fixed price2 85.0 100.0 273 61.0 100.0 77 Cost of sales as a percent of issue3 4.4 3.3 178 2.5 2.6 61 Share Allocation Variables Percent of offer allocated to employees 8.5 7.0 255 4.8 2.6 76 Fraction of offers with some allocation to employees 91.0 255 65.8 76 Percent of offer allocated to foreigners 28.4 11.5 348 35.9 32.5 142 Percent of offers with some allocation to foreigners 57.1 348 67.6 142 Control Allocation Variables Percent of capital sold in offer4 43.9 35.0 384 22.7 18.1 154 Percent of offers where 100% of capital sold 11.5 384 0 154 Percent of capital where 50% or more of 28.9 384 8.4 154 capital sold Source: Jones, Megginson, Nash, and Netter (1999). Notes: 1 also known as initial underpricing, the return an investor who bought shares at the offering price could earn by reselling those shares at end of the first day’s trading. 2 measures the fraction of an issue offered to investors at a predetermined, fixed price rather than an auction- determined price. 3 a measure of the sum of cash expenses and underwriter discount charged by the investment banking syndicate managing the issue. 4 measures the fraction of a firm’s total common equity (which is not necessarily synonymous with total voting rights) sold in an offering. is the sheer size of SIP offers—the SIPs is 34.1 percent (median 12.4 per- mean (median) size of initial SIPs is cent). Even seasoned SIP offers are un- $555.7 million ($104.0 million) and the derpriced by an average of 9.4 percent mean size of seasoned issues is $1.069 (median 3.3 percent). We return to this billion (median $311.0 million), much issue in section 6. larger than typical stock offerings. They The evidence of Jones et al. on allo- also find that SIPs are significantly un- cation of control in SIPs supports a po- derpriced by government sellers. The litical interpretation of divesting gov- mean level of underpricing for initial ernments’ motives. Jones et al. find that nearly all SIPs are essentially secondary but also publications such as Price Waterhouse offerings, in which only the government 1989b) an appendix that details similar infor- sells its shares and no money flows to mation for an additional 500 SIPs. This appen- dix can be obtained upon request by contacting the firm itself. Since the divesting gov- wmegginson@ou.edu. ernment sells an average (median) of Megginson and Netter: From State to Market 345 43.9 percent (35.0 percent) of the investigate what they call the “voucher SOE’s capital in initial offers and 22.7 portfolio problem.” This problem arises percent (18.1 percent) in seasoned is- whenever the proportion of ownership sues, the offers represent significant re- resulting from a given voucher bid is ductions in direct government stock unknown, but the post-privatization ownership. Although governments typi- performance of a divested company cally surrender day-to-day operating largely depends on the skills of the new control of the SOE to private owners in owners and their respective ownership the initial SIP, they retain effective veto stakes. Katz and Owen also provide a power through a variety of techniques. good discussion of the philosophical dif- The most common technique is govern- ferences between the Czech program, ment retention of a “golden share,” which relied heavily on vouchers and which gives it the power to veto certain prohibited post-sale trading of stock, actions, such as foreign takeovers. 22 and the Russian program, which priva- tized relatively small (29 percent on av- 4.5 Voucher Privatizations erage) stakes in most firms and allowed Voucher privatization has been the unrestricted trading of vouchers. most controversial method of divesting Although most countries’ actual expe- state-owned assets. Boycko, Shleifer, rience with vouchers has been poor, and Vishny (1994) show that the decision none has been quite as dismal as Rus- to pursue mass privatization, and even sia’s. Although a variety of factors have the specific program design, is largely played a role, Frydman, Katharina Pis- dictated by politics. The privatization tor, and Rapaczynski (1996) show that programs practiced in western Europe insider control of privatized firms has and elsewhere were politically difficult been by far the most important impedi- to execute in eastern Europe, although ment to effective reform. Initially, the Hungary, Estonia, and Poland used Russian government had high hopes case-by-case privatizations, which have that the “voucher privatization funds” been successful at a macro level. None- (VPFs) formed during the initial theless, voucher privatization schemes voucher distributions might be able to can be made attractive from an economic overcome the collective action problem perspective, since they maximize value, inherent in mass privatization pro- foster free and efficient markets, and grams. Such funds might use their con- promote effective corporate governance. centrated ownership in privatized firms Barbara Katz and Joel Owen (1997) to force managers to restructure. Though most funds attempted to exer- 22 Though golden shares have been widely cise their “voice” in corporate board- adopted, they are in fact almost never used to af- fect control contests (Patrick McCurry 2000). The rooms, insider dominance completely EU is trying to block new adoptions of golden blocked their efforts. The VPFs turned shares and roll back those already in place, charg- instead to their “exit” option and sold ing they are designed to discourage free cross- border competition for corporate control. At a re- shares on the secondary market. cent OECD conference, the director of Italy’s pri- Pistor and Andrew Spicer (1996) also vatization program, Vittorio Grilli, pointed out an examine the early promise and sub- additional political problem with exercising a golden share: When a government uses its share to sequent failure of privatization invest- veto a takeover bid, this is equivalent to publicly ment funds in Russia and the Czech Re- stating it does not approve of the bidder. Such a public. In both countries, citizens have statement is awkward at best, and could cause an international incident if the bidder is a foreign become owners of the worst perform- company. ing privatized assets, while the “crown 346 Journal of Economic Literature, Vol. XXXIX (June 2001) jewels” have all come under insider standards vary from country to country control. As the authors say, “. . . estab- as well as within countries over time. A lishing property rights is a longer and large literature in accounting has shown more complicated process than allocat- that management can manipulate U.S. ing title.” Olivier Blanchard and Phillipe accounting data, and this problem is Aghion (1996) also conclude that priva- probably greater for international firms. tization is proceeding slowly in eastern Furthermore, the possibility of sample Europe, largely because insiders, who selection bias can arise from several currently have control of firms but no sources, including governments’ desire property rights, oppose outsider privati- to make privatization “look good” by zation. Given this reality, Blanchard and privatizing the healthiest firms first. An- Aghion examine whether privatization other sample selection problem is that would proceed more rapidly if govern- data availability tends to be greater in ments were simply to allocate property the more developed countries (and per- rights to insiders (insider privatization). haps for the better performing firms However, they find there is a wedge be- within countries), so developed coun- tween the private value of the firm to tries (and better performing firms) are insiders and its value to an outsider, overrepresented in empirical analysis. and that this difference might well pre- For example, in cross-sectional regres- clude value-increasing exchanges. Given sions using fixed effects, estimation will the actual experience with insider probably rely mainly on data from dominance of most voucher privatiza- developed countries. tions, we conclude that this wedge is There are also many problems in in fact alive, well, and fully operational. measuring performance changes that arise from using accounting or stock 5. Has Privatization Improved data. We discuss the problems with Performance? stock return data in section 6; the prob- lems with accounting data are more im- Since privatization has been part of portant since many empirical studies government policy tool kits for almost employ primarily accounting informa- two decades now, academic researchers tion. These problems include determin- have had enough time to execute many ing the correct measure of operating empirical studies of the effect of divest- performance, selecting an appropriate ment on the performance of former benchmark with which to compare per- SOEs. However, there are difficult formance, and determining the appro- methodological problems with research priate statistical tests to use (Ahmed in this area. 23 An important problem is Galal, Leroy Jones, Pankaj Tandon, and data availability and consistency. The Ingo Vogelsang 1994; and Brad Barber amount of information that must be dis- and John Lyon 1996). The finance lit- closed is much less in most countries erature has not reached a consensus on than in the United States, and these the ways to deal with these problems 23 Many of the difficulties are similar to those for U.S. companies, much less priva- discussed in Jonathan Temple (1999), who surveys tized international firms. Barber and cross-country research on the determinants of growth. Temple discusses the substantial problems Lyon (1996) argue that test statistics that arise in estimating and interpreting cross- designed to determine whether there is country regressions. James Tybout (2000) also dis- abnormal performance using accounting cusses the difficulties with data in attempting to assess the performance of manufacturing firms in data are misspecified when the sample developing countries. firms have performed unusually well or Megginson and Netter: From State to Market 347 poorly. They suggest that sample firms ies that employ accounting and/or real must be matched to control firms with output data to examine the impact of similar pre-event performance, which is privatization on the operating effi- especially difficult in studies of privatized ciency, ownership structure, and/or fi- firms. nancial performance of former SOEs in Therefore, the results of each of the developed, developing, and transition studies we discuss must be kept in economies. Though all these studies are perspective. We also note that the stud- detailed in the accompanying tables, ies of post-performance rarely examine and most are discussed at least briefly the welfare effects on consumers. Most in the text, we also specify which stud- important, few studies control for the ies we think are the most important— possible use of market power by the and why we think this is so. To effec- privatized firms; that is, performance tively synthesize such a large number of improvements could be due to greater empirical studies, we first categorize exploitation of monopoly power, which papers according to whether they exam- has harmful effects on allocative effi- ine privatization in transition or non- ciency, rather than productive effi- transition economies. The latter studies ciency. Many of the studies on perfor- are evaluated in section 5.1, while the mance changes after privatization transition economies are examined in examine the effects of divestiture on section 5.2. This dichotomization is nec- groups such as workers, but few exam- essary, since both direct observation ine the effect of privatization on con- and published research suggest that re- sumers. On the other hand, one of the forming transition economies invariably principal reasons for launching privati- requires embracing a great many eco- zations, particularly of monopoly utili- nomic and political changes simultane- ties, is consumer dissatisfaction with a ously, whereas privatization (and atten- firm’s service. Furthermore, the studies dant regulatory changes) is often the cited here almost unanimously report sole major component of reform pro- increases in performance associated cesses in non-transition economies. A with privatization. 24 This consistency is further organizational step is to present, perhaps the most telling result we re- in tables 3 through 7, summary informa- port—privatization appears to improve tion for each of the studies we examine. performance measured in many differ- Presenting this information in tabular ent ways, in many different countries. 25 form saves us from having to sequen- With the above caveats in mind, this tially discuss each paper’s sample con- section evaluates the results of 38 stud- struction methodology, estimation pro- cedure, and empirical results in the 24 A cynic might say that all of the gains re- section’s text. Instead, we can identify key searchers have documented after privatization are findings that appear in many different due to selection bias. However, while there is studies, and can discuss methodological pros some evidence discussed elsewhere that the better firms are privatized first, at least in SIPs the evi- and cons for entire groups of studies, dence is still strong that performance improves af- rather than for each paper in turn. ter privatization. Further, the paper that does the best job of controlling for selection bias, Frydman, 5.1 Empirical Studies of Non-Transition Gray, Hessel, and Rapaczynski (1999), finds pri- Economies vatized firms perform better than SOEs. 25 Temple (1999) also notes the importance of We separate non-transition studies by both historical case studies and cross-sectional analysis in assessing recent developments in the empirical methodology, depending upon economic theory of growth. how the papers compare performance 348 Journal of Economic Literature, Vol. XXXIX (June 2001) changes resulting from privatization. privatization of the U.K.’s Central Elec- The first set of papers examines a single tricity Generating Board (CEGB), and industry, a single country, or one or a document significant post-privatization small number of individual firms. While performance improvements. However, these studies employ a variety of em- they find that the producers and their pirical techniques, most compare post- shareholders capture all of the financial privatization performance changes with rewards of this improvement and more, either a comparison group of non- whereas the government and consumers privatized firms or with a “counterfac- lose out. The authors conclude that tual” expectation of what would have CEGB’s restructuring and privatization occurred if the privatized firms had re- was in fact “worth it,” but could have mained state-owned. The second set of been implemented more efficiently and studies examines only firms divested with greater concern for the public’s through public share offerings, and welfare. 26 measures privatization-related perfor- Two of the studies described in table mance changes by comparing the three- 3 examine national privatization experi- year mean or median operating and fi- ences. Stephen Martin and David nancial performance of divested firms Parker (1995) find that, after adjusting to their own mean or median perfor- for business cycle effects, less than half mance during their last three years as the British firms they study perform state-owned firms. better after being privatized. The authors do, however, find evidence of a 5.1.1 Case, Single-Industry, and “shake-out” effect, where several firms Single-Country Studies improve performance prior to being pri- The studies we examine in this sec- vatized (but not afterward). The results tion are summarized in table 3. The of the second national study are far less first study listed merits detailed analysis ambiguous. La Porta and López-de- because it has proven so influential, Silanes (1999) find that the former both due to the rigor of its methodology Mexican SOEs they study rapidly close and because it was sponsored by the a large performance gap with industry- World Bank. Galal, Jones, Tandon, and matched private firms that had existed Vogelsang (1994) compare the actual prior to divestment. These firms go post-privatization performance of twelve from being highly unprofitable before large firms—mostly airlines and regu- privatization to being very profitable lated utilities—in Britain, Chile, Malay- thereafter. Output increases 54.3 per- sia, and Mexico to the predicted perfor- cent, in spite of a reduced level of in- mance of these firms had they not been vestment spending, and sales per em- divested. Using this counterfactual ap- ployee roughly double. The privatized proach, the authors document net wel- firms reduce (blue- and white-collar) fare gains in eleven of the twelve cases 26 The privatization and liberalization of the considered which equal, on average, 26 British electricity industry is also discussed at percent of the firms’ pre-divestiture length in David Newberry (1997) and Vickers and sales. They find no case where workers Yarrow (1991), while the regulatory regime adopted for earlier utility privatizations is de- are made significantly worse off, and scribed in M. E. Beesley and S. C. Littlechild three where workers significantly bene- (1989). None of these works showers the Thatcher fit. David Newberry and Michael Pollitt government with praise for its policy decisions, though Beesley and Littlechild do find the RPI-X (1997) perform a similar counterfactual price regulation system adopted in the U.K. is analysis of the 1990 restructuring and superior to the U.S. rate of return regulatory regime. Megginson and Netter: From State to Market 349 TABLE 3 CASE STUDIES, COUNTRY AND INDUSTRY-SPECIFIC EMPIRICAL STUDIES: NON-TRANSITION ECONOMIES Study Sample description, study period, and methodology Summary of findings and conclusions Galal, Jones, Compares actual post-privatization performance of 12 Documents net welfare gains in 11 of the 12 cases Tandon, and large firms (mostly airlines and regulated utilities) in which equal, on average, 26% of the firms’ pre- Vogelsang UK, Chile, Malaysia, Mexico to predicted perfor- divestiture sales. Find no case where workers were 1994 mance if the firms remained SOEs. made worse off, and 3 where workers were made significantly better off. Martin and Using 2 measures (ROR on capital employed and an- Mixed results. Outright performance improvements Parker 1995 nual growth in value-added per employee-hour), exam- after privatization found in less than half of firm- ines whether 11 UK firms privatized in 1981–88 im- measures studied. Several firms improved prior to proved performance after divestment. Attempts to divestiture, indicating an initial “shake-out” effect control for business cycle effects. upon privatization announcement. Ramamurti Surveys studies of 4 telecom, 2 airline, and 1 toll-road Concludes privatization very positive for telecoms, 1996 privatization programs in Latin America during 1987– partly due to scope for technology, capital investment, 91. Discusses political economic issues, methods used and attractiveness of offer terms. Much less scope for to overcome bureaucratic/ideological opposition to productivity improvements for airlines and roads, and divestiture. little improvement observed. Boles de Estimates impact of 1987 deregulation and 1990 Documents significant declines in price of phone ser- Boer and privatization of Telecom New Zealand on price and vices, due mostly to productivity growth that cut costs Evans 1996 quality of telephone services. Examines whether in- at a 5.6% annual rate, and significant improvement vestors benefited. in service levels. Shareholders also benefited signifi- cantly. Petrazzini Using International Telecommunications Union (ITU) Deregulation and privatization both are associated with and Clark data through 1994, tests whether deregulation and significant improvements in level and growth in tele- 1996 privatization impact level and growth in teledensity (main density, but have no consistent impact on service quality. lines per 100 people), prices, service quality, and Deregulation associated with lower prices and increased employment by telecoms in 26 developing countries. employment; privatization has the opposite effect. Ramamurti Examines restructuring and privatization of Documents a 370% improvement in labor productivity 1997 Ferrocarilla Argentinos, the national railroad, in 1990. and a 78.7% decline in employment (from 92,000 to Tests whether productivity, employment, and need 19,682). Services were expanded and improved, and for operating subsidies (equal to 1% of GDP in 1990) delivered at lower cost to consumers. Need for op- change significantly after divestiture. erating subsidies largely eliminated. Eckel, Examines effect of British Airways’ privatization on Stock prices of US competitors decline on average 7% Eckel, and competitors’ stock prices. Tests whether fares on upon BA’s privatization, and fares on routes served Singal 1997 competitive routes decline after privatization. by BA and competitors fall by 14.3%. Compensation of BA executives increases and becomes more performance-contingent. Newberry Performs cost-benefit analysis of the 1990 restruc- Restructuring/privatization of CEGB resulted in per- and Pollitt turing and privatization of Central Electricity Generat- manent cost reduction of 5% per year. Producers and 1997 ing Board (CEGB). Compares actual performance of shareholders capture all this benefit and more. Con- privatized firms to a counterfactual assuming CEGB sumers and government lose. Shows that alternative remained state-owned. fuel purchases involve unnecessarily high costs and wealth flows out of country. Ros 1999 Uses ITU data and panel data regression methodology Countries with at least 50% private ownership of main to examine effects of privatization and competition on telecom firm have significantly higher teledensity network expansion and efficiency in 110 countries levels and growth rates. Both privatization and com- over 1986–95. petition increase efficiency, but only privatization is positively associated with network expansion. La Porta and Tests whether performance of 218 Mexican SOEs Output of privatized firms increased 54.3%; employ- López-de- privatized through June 1992 improves after di- ment declined by half (though wages for remaining Silanes 1999 vestment. Compares performance with industry- workers increased). Firms achieved a 24% point in- matched firms, and splits improvements documented crease in operating profitability, eliminating need for between industry and firm-specific influences. subsidies equal to 12.7% of GDP. Higher product prices explain 5% of improvement; transfers from laid- off workers, 31%, and incentive-related productivity gains account for remaining 64%. 350 Journal of Economic Literature, Vol. XXXIX (June 2001) TABLE 3 (Cont.) Study Sample description, study period, and methodology Summary of findings and conclusions Wallsten Performs econometric analysis of effects of tele- Competition is significantly associated with increases 2000a communications reforms in developing countries. in per capita access and decreases in cost. Privatization Using panel dataset of 30 African and Latin American is helpful only if coupled with effective, independent countries from 1984–97, explores effects of pri- regulation. Increasing competition is single best re- vatization, competition and regulation on telecom- form; competition in combination with privatization is munications performance. best. Privatizing a monopoly without regulatory re- forms should be avoided. Laurin and Compares productivity and profitability of 2 large Total factor productivity of CN much lower than that Bozec 2000 Canadian rail carriers, before and after 1995 pri- of privately owned Canadian Pacific (CP) during 1981– vatization of Canadian National (CN). Compares ac- 91, but became as efficient during pre-privatization counting ratios for 17-year period 1981–97 and 3 (1992–95), exceeded it after 1995. CN stock price sub-periods: the fully state-owned era (1981–91), pre- outperformed CP, the transportation industry, and the privatization (1992–95), and post-privatization. Com- Canadian market after 1995. Both firms shed workers pares stock returns from 1995–98. Creates 6-firm after 1992, but CN’s employment declined more (34% comparison group of Canadian privatizations and com- vs 18%) as average productivity almost doubled (97% putes accounting ratios and stock returns for these increase). CN’s capital spending increased signifi- firms. cantly, though CP’s increased more. Six-firm Canadian privatization comparison group experienced significant increases in investment spending and productivity and decline in employment. Boylaud and Uses factor analysis and database on market structure Prospective and actual competition both bring about Nicoletti and regulation to investigate effects of liberalization productivity and quality improvements and lower 2000 and privatization on productivity, prices and quality of prices in telecom services, but no clear effect was long-distance and cellular telephone services in 23 found for privatization. OECD countries over 1991–97. employment by half, but those workers BA would result from the divestiture. 27 who remain are paid significantly more. Claude Laurin and Yves Bozec (2000) The authors attribute most of the per- compare the productivity and profit- formance improvement to productivity gains ability of two large Canadian rail carri- resulting from better incentives, with at ers (one state-owned and one private- most one-third of the improvement being sector), both before and after the 1995 attributable to lower employment costs. privatization of Canadian National Three of the papers described in ta- (CN). They find that CN’s relatively ble 3 are essentially case studies of indi- poor performance during the “fully vidual privatized companies, though state-owned period” (1981–91) rapidly two of the articles benchmark perfor- converges on Canadian Pacific’s perfor- mance changes with respect to one or mance levels during the pre-privatization more private companies. Catherine Eckel, but post-announcement period (1992–95), Doug Eckel, and Vijay Singal (1997) ex- and then surpasses it thereafter. These amine the effect of British Airways’ (BA) 1987 privatization on competitors’ stock 27 Eckel, Eckel, and Singal also examine the prices and on fares charged in those two-stage privatization of Air Canada (from 100 routes where BA competes directly with percent state ownership to 57 percent, then to zero). Unlike BA, Air Canada does not compete foreign airlines. They find that the with U.S. carriers on many routes, so there is no stock prices of U.S. competitors fall, as significant competitor stock price effect resulting do airfares in markets served by BA; from its divestiture. Air Canada’s fares do not fall after the first, partial privatization, but fall a sig- both findings suggest that stock traders nificant 13.7 percent after the final, complete di- anticipated a much more competitive vestiture of state ownership. Megginson and Netter: From State to Market 351 findings suggest two separable impacts ered and methodology employed. Ben of privatization on firm performance: an Petrazini and Theodore Clark (1996), “anticipation” effect prior to divestiture Agustin Ros (1999), and Scott Wallsten and a “follow through” effect sub- (2000a) examine developing countries sequently. The final case study, Ravi exclusively or as separate subsamples, Ramamurti (1997), examines the 1990 while Ros (1999) and Olivier Boylaud restructuring and privatization of Ferro- and Giuseppe Nicoletti (2000) provide carilla Argentino, the Argentine na- similar coverage of OECD countries, tional freight and passenger railway sys- and David Boles de Boer and Lewis tem. The author documents a nearly Evans (1996) study the deregulation incredible 370 percent improvement in and privatization of Telecom New Zea- labor productivity and an equally strik- land. Though Ros, Wallsten, and Boy- ing (and not unrelated) 78.7 percent de- laud and Nicoletti all use some variant cline in employment—from 92,000 to of panel data methodology, they arrive at 18,682 workers. 28 Operating subsidies slightly different conclusions regarding declined almost to zero, and consumers the relative importance of deregulation/ benefited from expanded (and better liberalization and privatization in pro- quality) service and lower costs. Rama- moting expanded teledensity (number murti concludes that these performance of main lines per 100 population) and improvements could not have been operating efficiency of national telecom achieved without privatization. companies, and the quality and pricing No less than six of the studies de- of telecom services. On balance, these tailed in table 3 examine the telecom- studies generally indicate that deregula- munications industry, which has been tion and liberalization of telecom ser- transformed by the twin forces of tech- vices are associated with significant nological change and deregulation (in- growth in teledensity and operating ef- cluding privatization) since 1984—the ficiency, and significant improvements year when the AT&T monopoly was in the quality and price of telecom broken up in the United States and the services. The impact of privatization, Thatcher government began privatizing per se, is somewhat less clear-cut, but British Telecom. Five of these are em- most studies agree that the combina- pirical studies, while Ramamurti (1996) tion of privatization and deregulation/ provides a simple, though highly read- liberalization is associated with signifi- able, summary of empirical studies ex- cant telecommunications improvements. amining four telecom privatizations in This is certainly the result predicted by Latin America. Ramamurti concludes Noll (2000) in his analysis of the politi- that all were judged to be political and cal economy of telecom reform in de- economic success stories. Unfortu- veloping countries. The Juliet D’Souza nately, the empirical studies tell some- and Megginson (2000) study’s findings— what conflicting stories, probably due in described in the following section—also part to differences in the nations cov- support the idea that telecom privatization yields net benefits. 29 28 Ramamurti (1997) details the intense political maneuvering that accompanied the attempt to re- 29 Though they do not quite fit into our empiri- structure and slim down FA. The generous sever- cal classification scheme, six related studies de- ance payments awarded to displaced workers were serve mention here. Peter Smith and Björn Wel- instrumental in winning union acquiescence in the lenius (1999) and Wellenius (2000) present restructuring plan, while the presence of effective normative analyses of telecom regulation in devel- road transport competition for rail traffic reduced oping countries, while Walter Wasserfallen and the threat of a potentially crippling strike weapon. Stefan Müller (1998) discuss the privatization and 352 Journal of Economic Literature, Vol. XXXIX (June 2001) 5.1.2 Pre- vs Post-Privatization MNR methodology is its need to exam- Performance for SIPs ine only simple, universally available ac- counting variables (such as assets, sales, The studies summarized in table 4 all and net income) or physical units such examine how privatization affects firm as number of employees. Obviously, performance by comparing pre- and researchers must be careful when post-divestment data for companies pri- comparing accounting information vatized via public share offering. Since generated at different times in many the first study to be published using different countries. Most of the stud- this methodology is Megginson, Nash, ies cited here also ignore (or, at best, and van Randenborgh (1994), we will imperfectly account for) changes in refer to this as the MNR methodology. the macroeconomy or industry over This empirical procedure has several the seven-year event window during obvious economic and econometric which they compute pre- versus post- drawbacks. Of these, selection bias privatization performance changes. Fi- probably causes the greatest concern, nally, the studies cannot account for the since by definition a sample of SIPs will impact on privatized firms of any regu- be biased towards the very largest com- latory or market-opening initiatives that panies sold during any nation’s privati- often are launched simultaneously with zation program. Furthermore, since or immediately after major privatization governments have a natural tendency to programs. privatize the “easiest” firms first, those In spite of these drawbacks, studies SOEs sold via share offerings (particu- employing the MNR methodology have larly those sold early in the process) two key advantages. First, they are the may well be among the healthiest state- only studies that can examine and di- owned firms. 30 Another drawback of the rectly compare large samples of eco- nomically significant firms, from differ- deregulation of western Europe’s telecom indus- ent industries, privatized in different try. Michael Pollitt (1997) analyzes the impact of liberalization on the performance of the interna- countries, over different time periods. tional electric supply industry, and Bortolotti, Since each firm is compared to itself (a Fantini, and Siniscalso (1999b) document that ef- few years earlier) using simple, inflation- fective regulation is a crucial institutional variable in electric utility privatization. Establishing such a adjusted sales and income data (that regulatory regime allows governments to increase produce results in simple percentages), the pace of privatization, sell higher stakes, and this methodology allows one to effi- maximize offering proceeds. Finally, Wallsten (2000b) shows that exclusivity periods, which are ciently aggregate multinational, multi- usually granted to telecom monopolies as they industry results. This point is made are being privatized, are economically harmful to clear in table 5, which summarizes the consumers and do not achieve the efficiency ob- jectives assigned to them at the time of divest- results of three studies that use pre- ment. Exclusivity periods do, however, raise the cisely the same empirical proxies and price that investors are willing to pay for priva- test methodology—and can thus be ag- tized telecoms, which largely explains why they are employed. gregated and directly compared—yet 30 Megginson, Nash, Netter, and Poulsen (2000) examine non-overlapping samples. In find that governments selling SOEs tend to sell total, these three studies examine seven the more profitable SOEs in the public capital markets and the less profitable in the less trans- performance criteria for 204 companies parent private markets. Those sold in the public from 41 countries. Second, while focus- capital markets are the firms that appear in studies ing on SIPs yields a selection bias, it of performance. Dewenter and Malatesta (2001) also show performance improvements before pri- also yields samples that encompass the vatization in firms that are being privatized. largest and most politically influential Megginson and Netter: From State to Market 353 TABLE 4 EMPIRICAL STUDIES ON PERFORMANCE CHANGES FOR FIRMS PRIVATIZED VIA PUBLIC SHARE OFFERINGS: NON-TRANSITION ECONOMIES These studies each employ samples from more than one country and more than one industry. Study Sample description, study period, and methodology Summary of findings and conclusions Megginson, Compares 3-year average post-privatization per- Documents economically and statistically significant Nash, van ormance ratios to 3-year pre-privatization values for 61 post-privatization increases in output (real sales), Randenborgh firms from 18 countries and 32 industries from 1961– operating efficiency, profitability, capital investment 1994 89. Tests significance of median changes in post versus spending, and dividend payments; significant de- pre-privatization periods. Binomial tests for percent of creases in leverage; no evidence of employment de- firms changing as predicted. clines, but significant changes in firm directors. Macquieira Compares pre- versus post-privatization performance Unadjusted results virtually identical to MNR: sig- and Zurita of 22 Chilean firms privatized over 1984–89. Uses nificant increases in output, profitability, employment, 1996 Megginson, Nash and van Randenborgh (MNR) investment, dividend payments. After adjusting for methodology to analyze first without adjusting for market movements, changes in output, employment, overall market movements (as in MNR), then with and liquidity are no longer significant, and leverage adjustment for contemporaneous changes. increases significantly. Boubakri Compares 3-year average post-privatization per- Documents significant post-privatization increases in and Cosset formance ratios to 3-year pre-privatization values for output (real sales), operating efficiency, profitability, 1998 79 firms from 21 developing countries and 32 indus- capital investment spending, dividend payments, tries over 1980–92. Tests for significance of median employment; significant decreases in leverage. Per- changes in ratio values post- versus pre-privatization. formance improvements are generally larger than Binomial tests for percentage of firms changing as those documented by MNR. predicted. D’Souza and Documents offering terms, sale methods, and Documents significant post-privatization increases in Megginson ownership structure resulting from privatization of 78 output (real sales), operating efficiency, and profit- 1999 firms from 10 developing and 15 developed countries ability, and significant decreases in leverage. Capital over 1990–94. Compares 3-year average post- investment spending increases insignificantly, while privatization performance ratios to 3-year pre- employment declines significantly. More of the firms privatization values for subsample of 26 firms. Tests privitized in the 1990s are from telecoms and other for significance of median changes in ratio values post- regulated industries. vs pre-privatization. Binomial tests for percent of firms changing as predicted. Verbrugge, Study offering terms and share ownership results for Documents moderate performance improvements in Megginson, 65 banks fully or partially privatized from 1981 to OECD countries. Ratios proxying for profitability, fee Owens 2000 1996. Then compare pre- and post-privatization per- income (noninterest income as fraction of total), and formance changes for 32 banks in OECD countries capital adequacy increase significantly; leverage ratio and 5 in developing countries. declines significantly. Documents large, ongoing state ownership, and significantly positive initial returns to IPO investors. Boubakri Examine pre- versus post-privatization performance of Document significantly increased capital spending by and Cosset 16 African firms privatized through public share of- privatized firms, but find only insignificant changes in 1999 fering during 1989–96. Also summarize findings of profitability, efficiency, output and leverage. three other studies pertaining to privatization in developing countries. D’Souza and Examines pre- versus post-privatization performance Profitability, output, operating efficiency, capital Megginson changes for 17 national telecom companies privatized spending, number of access lines, and average salary 2000 through share offerings during 1981–94. per employee all increase significantly after pri- vatization. Leverage declines significantly; employ- ment declines insignificantly. Dewenter Compares pre- versus post-privatization performance Documents significant increases in profitability (using and of 63 large, high-information companies divested dur- net income) and significant decreases in leverage and Malatesta ing 1981–94 over both short-term [(+1 to +3) versus labor intensity (employees ÷ sales) over short- and 2001 (−3 to −1)] and long-term [(+1 to +5) versus (−10 to −1)] long-term horizons. Operating profits increase prior to horizons. Examines long-run stock return performance privatization, but not after. Significantly positive long- of privatized firms and compares relative performance term (1–5 years) abnormal stock returns, mostly in of a large sample (1,500 firm-years) of state and pri- Hungary, Poland, and UK. Results strongly indicate vately owned firms during 1975, 1985, and 1995. that private firms outperform SOEs. 354 Journal of Economic Literature, Vol. XXXIX (June 2001) TABLE 4 (Cont.) Study Sample description, study period, and methodology Summary of findings and conclusions Boardman, Compares 3-year average post-privatization per- Profitability, measured as return on sales or assets, Laurin, and formance ratios to 5-year pre-privatization values for 9 more than doubles after privatization; efficiency and Vining 2000 Canadian firms privatized during 1988–95. Computes sales increase significantly (though less drastically). long-run (up to 5 years) stock returns for divested Leverage and employment decline significantly; firms. capital spending increases significantly. Privatized firms significantly outperform Canadian stock market over all long-term holding periods. privatizations. As discussed in section 4, firms over the study period, but many SIPs account for more than two-thirds of these improvements cease to be sta- of the $1 trillion of total revenues tistically significant once such adjust- raised by governments since 1977. ments are made. Verbrugge et al. With these methodological caveats in (2000) document significant, though mind, we turn to a summary of the modest, increases in the profitability findings of studies using the MNR and capital adequacy of commercial technique. banks privatized in OECD countries, as All of these studies offer at least lim- well as significant declines in leverage, ited support for the proposition that but they also find substantial ongoing privatization is associated with signifi- state involvement in these banks’ af- cant improvements in the operating and fairs. Consistent with the result that financial performance of SOEs divested state connections matter in bank opera- via public share offering. Two of these tions, Philip Hersch, David Kemme, studies focus on specific industries: and Netter (1997) find that in Hungary banking (James Verbrugge, Wanda the banks made it much easier for firms Owens, and Megginson 2000) and tele- headed by former members of the no- communications (D’Souza and Meggin- menklatura to get loans than other son 2000); one examines data from a firms. single country, Chile (Carlos Mac- Finally, Dewenter and Malatesta quieira and Salvador Zurita 1996); and (2000) estimate the effects of govern- the other six employ multi-industry, ment ownership and privatization using multinational samples. Five of these a sample of large firms from three sepa- studies—MNR (1994), Narjess Boubakri rate time periods (1975, 1985, and and Jean-Claude Cosset (1998), D’Souza 1995) compiled by Fortune. They esti- and Megginson (1999, 2000), and mate regressions explaining profitability Boardman, Laurin and Vining (2000)— controlling for firm size, location, in- document economically and statistically dustry, and the business cycle. They significant post-privatization increases find that net income-based profitability in real sales (output), profitability, effi- measures increase significantly after ciency (sales per employee), and capital privatization, but operating income- spending, coupled with significant de- based measures do not. Instead, they clines in leverage. Macquieira and find that operating profits increase Zurita find similar results for Chilean prior to divestiture, once more support- firms using data that is not adjusted for ing the idea that privatization can have changes experienced by other Chilean a significant anticipation effect. Megginson and Netter: From State to Market 355 TABLE 5 PERFORMANCE OF NEWLY PRIVATIZED FIRMS Results of three empirical studies comparing three-year average operating and financial performance of a combined sample of 211 privatized firms with average performance of those firms during their last three years as SOEs. The studies employ the Wilcoxon rank sum test (with its z-statistic) to test for change in median value, and multiple proxies for most economic variables being measured. This table summarizes one proxy per topic, and emphasizes the one highlighted in the studies (usually the variable that uses either physical measures, such as number of employees, or financial ratios using current-dollar measures in the numerator or denominator, or both). Efficiency and output measures are index values, with the value during the year of privatization defined as 1.000; inflation-adjusted sales figures are used in efficiency and output measures. Mean Mean Z-Statistic % of Firms Z-Statistic Value Mean Change for with for Number of before Value after due to Difference Improved Significance Variables and Observa- Privatiza- Privatiza- Privatiza- in Perfor- Perfor- of % Studies Cited tions tion tion tion mance mance Change Profitability (%) net income ÷ sales Megginson, Nash, and van 55 0.0552 0.0799 0.249 3.15∗∗∗ 69.1 3.06∗∗∗ Randenborgh 1994 (0.0442) (0.0611) (0.0140) Boubakri and Cosset 78 0.0493 0.1098 0.0605 3.16∗∗∗ 62.8 2.29∗∗ 1998 (0.0460) (0.0799) (0.0181) D’Souza and Megginson 78 0.14 0.17 0.03 3.92∗∗∗ 71 4.17∗∗∗ 1999 (0.05) (0.08) (0.03) Weighted Average 218a 0.0862 0.1257 0.0396 67.6 Efficiency (real sales per employee) Megginson, Nash, and van 51 0.956 1.062 0.1064 3.66∗∗∗ 85.7 6.03∗∗∗ Randenborgh 1994 (0.942) (1.055) (0.1157) Boubakri and Cosset 56 0.9224 1.1703 0.2479 4.79∗∗∗ 80.4 4.60∗∗∗ 1998 (0.9056) (1.1265) (0.2414) D’Souza and Megginson 63 1.02 1.23 0.21 4.87∗∗∗ 79 5.76∗∗∗ 1999 (0.87) (1.16) (0.29) Weighted Average 170 0.9733 1.1599 0.1914 81.5 Investment (%) capital expenditures ÷ sales Megginson, Nash, and van 43 0.1169 0.1689 0.0521 2.35∗∗ 67.4 2.44∗∗ Randenborgh 1994 (0.0668) (0.1221) (0.0159) Boubakri and Cosset 48 0.1052 0.2375 0.1322 2.28∗∗ 62.5 1.74∗ 1998 (0.0649) (0.1043) (0.0137) D’Souza and Megginson 66 0.18 0.17 −0.01 0.80 55 0.81 1999 (0.11) (0.10) (−0.01) Weighted Average 154 0.1405 0.1900 0.0493 60.6 Output (real sales, adjusted by cpi) Megginson, Nash, and van 57 0.899 1.140 0.241 4.77∗∗∗ 75.4 4.46∗∗∗ Randenborgh 1994 (0.890) (1.105) (0.190) Boubakri and Cosset 78 0.9691 1.220 0.2530 5.19∗∗∗ 75.6 4.58∗∗∗ 1998 (0.9165) (1.123) (0.1892) D’Souza and Megginson 85 0.93 2.70 1.76 7.30∗∗∗ 88 10.94∗∗∗ 1999 (0.76) (1.86) (1.11) Weighted Average 209a 0.9358 1.7211 0.8321 80.3 Employment (total employees) Megginson, Nash, and van 39 40,850 43,200 2,346 0.96 64.1 1.84∗ Randenborgh 1994 (19,360) (23,720) (276) Boubakri and Cosset 57 10,672 10,811 139 1.48 57.9 1.19 1998 (3,388) (3,745) (104) D’Souza and Megginson 66 22,941 22,136 −805 −1.62 36 −2.14∗∗ 1999 (9,876) (9,106) (−770) Weighted Average 162 22,936 23,222 286 49.5 356 Journal of Economic Literature, Vol. XXXIX (June 2001) TABLE 5 (Cont.) Mean Mean Z-Statistic % of Firms Z-Statistic Value Mean Change for with for Number of before Value after due to Difference Improved Significance Variables and Observa- Privatiza- Privatiza- Privatiza- in Perfor- Perfor- of % Studies Cited tions tion tion tion mance mance Change Leverage (%) total debt ÷ total assets Megginson, Nash and van 53 0.6622 0.6379 −0.0243 −2.41∗∗ 71.7 3.51∗∗∗ Randenborgh 1994 (0.7039) (0.6618) (−0.0234) Boubakri and Cosset 65 0.5495 0.4986 −0.508 −2.48∗∗ 73.1 2.11∗∗ 1998 (0.5575) (0.4789) (−0.0162 D’Souza and Megginson 72 0.29 0.23 −0.06 −3.08∗∗∗ 67 3.05∗∗∗ 1999 (0.26) (0.18) (−0.08) Weighted Average 188 0.4826 0.4357 −0.0469 67.0 Dividends (%) cash dividends ÷ sales Megginson, Nash and van 39 0.0128 0.0300 0.0172 4.63∗∗∗ 89.7 8.18∗∗∗ Randenborgh 1994 (0.0054) (0.0223) (0.0121) Boubakri and Cosset 67 0.0284 0.0528 0.0244 4.37∗∗∗ 76.1 4.28∗∗∗ 1998 (0.0089) (0.0305) (0.0130) D’Souza and Megginson 51 0.015 0.04 0.025 4.98∗∗∗ 79 5.24∗∗∗ 1999 (0.00) (0.02) (0.02) Weighted Average 106 0.0202 0.0655 0.0228 80.4 a Number exceeds 211 because of overlapping firms in different samples. ∗∗∗ Indicates significance at the 1 percent level ∗∗ Indicates significance at the 5 percent level ∗ Indicates significance at the 10 percent level 5.1.3 Summary and Analysis privatized firms. All governments fear that privatization will cause former These 22 studies from non-transition SOEs to shed workers, and the key economies offer at least limited support question in virtually every case is for the proposition that privatization is whether the divested firm’s sales will associated with improvements in the increase enough after privatization to operating and financial performance of offset the dramatically higher levels of divested firms. Several of the studies of- per-worker productivity. Three studies fer strong support for this proposition, document significant increases in em- and only Martin and Parker (1995) ployment (Galal et al. 1994; Megginson, document outright performance de- Nash, and van Randenborgh 1994; and clines (for six of eleven British firms) Boubakri and Cosset 1998), two find in- after privatization. Almost all studies significant changes (Macquieira and that examine post-privatization changes Zurita 1996; and D’Souza and Meggin- in output, efficiency, profitability, capital son 2000) while the remaining five investment spending, and leverage docu- document significant—sometimes mas- ment significant increases in the first four sive—employment declines (Ramamurti and significant declines in leverage. 1997; La Porta and López-de-Silanes 1999; The studies examined here are far Laurin and Bozec 2000; D’Souza and Meg- less unanimous regarding the impact of ginson 1999; and Boardman, Laurin, and privatization on employment levels in Vining 2000). These conflicting results Megginson and Netter: From State to Market 357 could be due to differences in method- tion, a natural follow-on question is to ology, sample size and make-up, or ask why performance improves. As we omitted factors. However, it is more will discuss in the next section, a key likely that the studies reflect real differ- determinant of performance improve- ences in post-privatization employment ment in transition economies is bring- changes between countries and be- ing in new managers after privatization. tween industries. In other words, there No study explicitly documents system- is no “standard” outcome. Perhaps the atic evidence of this occurring in non- safest conclusion we can assert is that transition economies, but Catherine privatization does not automatically Wolfram (1998) and Michael Cragg and mean employment reductions in divested I. J. Alexander Dyck (1999a,b) show that firms—though this will likely occur the compensation and pay-performance unless sales can increase fast enough sensitivity of managers of privatized after divestiture to offset very large U.K. firms increases significantly after di- productivity gains. vestment. The only study that explicitly In our opinion, the Galal et al. addresses the sources of post-privatization (1994), La Porta and López-de-Silanes performance improvement using data (1999), Dewenter and Malatesta (2001), from multiple non-transition econo- and the three articles summarized in mies, D’Souza et al. (2000), finds D’Souza and Megginson (1999) are the stronger efficiency gains for firms in most persuasive studies examined in developing countries, in regulated in- this section. As mentioned, the main dustries, in firms that restructure op- strength of Galal et al. is its construc- erations after privatization, and in coun- tion and use of a clear “counterfactual” tries providing greater amounts of that (virtually uniquely) allows both the shareholder protection. financial and welfare gains from privati- We now turn to an examination of re- zation to be measured. La Porta and search findings about privatization’s im- López-de-Silanes execute what we con- pact in transition economies. Privatiza- sider the best single-country study, tion is both more difficult and more since it examines almost the entire all-encompassing in these countries population of Mexican privatizations than it is in either industrialized or and compares performance changes to non-transition developing countries. industry-matched private firms. Dewen- This is because in transition economies, ter and Malatesta both contrast the privatization is only part of the massive performance of private-sector and changes in the economy as countries move state-owned firms over three non- from communism to more market ori- overlapping periods and study how the ented methods of allocating resources performance of privatized firms changes and organizing production. over an extended time period. Finally, 5.2 Privatization in Transition D’Souza and Megginson’s summary and Economies comparison of three studies that use the same methodology—but non-overlapping We categorize the empirical studies samples—provides compelling evidence that examine privatization in transition that the operating and financial gains to economies into more manageable privatization are pervasive. groups. Both direct observation and the Since the empirical studies discussed findings of these studies suggest that a in this section generally document per- logical classification scheme is to evalu- formance improvements after privatiza- ate separately studies that examine 358 Journal of Economic Literature, Vol. XXXIX (June 2001) firms privatized in central and eastern eastern Europe are summarized in table Europe and those which study the pri- 6. These countries employed varying vatization programs of Russia and the methods of privatizing SOEs, including other republics of the former Soviet asset sales (Hungary and eastern Ger- Union. These categories are evaluated many), voucher privatizations (the in sections 5.2.1 and 5.2.2, respectively. Czech Republic and early Polish dives- We then conclude section 5 with a brief titures), “spontaneous privatizations” overview of China’s liberalization and (Slovenia), share offerings (later Polish privatization program. sales), or a combination of techniques. Note that testing for the effects of The studies also cover differing event privatization on firm performance is periods during the 1990s, employ dif- even more difficult in transition econo- fering empirical methodologies, and ask mies than in non-transition economies. somewhat different questions—though As mentioned above, privatization in all directly or indirectly ask how privati- these countries occurs at the same time as, zation impacts firm-level operating per- and is part of, other massive economy- formance. Additionally, all of these wide changes. Thus, isolating the ef- studies must contend with the fact that fects of privatization itself is problem- output typically fell dramatically in atic. Further, as discussed by Djankov every central and eastern European and Murrell (2000b, p. 9) “mis-reporting country during the period immediately and accounting difficulties are rife in after the collapse of socialism in 1989– transition economies.” In general, the 91, though in most cases output later data from transition economies is much snapped back smartly. 31 These studies worse and much more limited than must therefore examine whether, for from non-transition economies. Finally, example, the output of privatized firms the transition economies are under- contracted less than did the output of going many other major changes in their firms that remained state-owned. These political and economic environments. and other econometric challenges that The number of firms privatized in some must be faced in disentangling the ef- way in transition is much greater than fects of privatization, ownership struc- in non-transition economies (Djankov ture changes, and other influences on and Murrell 2000a report over 150,000 the post-divestment performance of pri- large firms in 27 transition economies vatized firms in transition settings are faced the revolutionary changes of tran- discussed at length in Andrew Weiss sition). However, we do not have good and Georgiy Nikitin (1998) and Frydman, data or even any data on many of these Gray, Hessel, and Rapaczynski—hereafter firms. The data that do exist often come FGHR—(1999). from surveys rather than mandated dis- In spite of all the caveats spelled out closure. Thus, the studies of privatiza- above, the studies summarized in table tion in transition economies has greater 6 yield surprisingly consistent results problems with significant selection bias, as well as omitted variables, than in the 31 This “U-shape” pattern of aggregate output in studies of non-transition economies. 26 transition economies is documented and exam- ined econometrically in Andrew Berg, Eduardo 5.2.1 Privatization in Central and Borensztein, Ratna Sahay and Jeromin Zettel- meyer (1999). They find that structural reforms— Eastern Europe including privatization—are critically important in promoting rapid recovery from the initial eco- The empirical studies that examine nomic decline. Taken as a whole, their results privatization programs in central and strongly support a “radical” approach to reforms. Megginson and Netter: From State to Market 359 TABLE 6 SUMMARY OF EMPIRICAL STUDIES OF PRIVATIZATION IN TRANSITION ECONOMIES: CENTRAL AND EASTERN EUROPE Study Sample description, study period, and methodology Summary of empirical findings and conclusions Claessens, Examines determinants of performance improvements Privatized firms do prosper, primarily because of Djankov, and for 706 Czech firms privatized during 1992–95. Using resulting concentrated ownership structure. The more Pohl 1997 Tobins-Q, tests whether concentrated ownership concentrated the post-privatization ownership structure or outside monitor (bank or investment structure the higher the firm’s profitability and market fund) improves Q more than dispersed ownership. valuation. Large stakes owned by bank-sponsored funds and strategic investors are particularly value- enhancing. Pohl, Compares extent of restructuring of over 6,300 private Privatization dramatically increases restructuring Anderson, and state-owned firms in 7 east European countries likelihood and success. Firms privatized for 4 years Claessens, during 1992–95. Uses 6 measures to examine which will increase productivity 3–5 times more than similar and Djankov strategies improve performance the most. SOEs. Little difference in performance based on 1997 method of privatization, but ownership and financing effects impact restructuring. Smith, Cin Using a sample with 22,735 firm-years of data drawn Percentage point increase in foreign ownership is and from period of “spontaneous privatization” in Slovenia associated with a 3.9% increase in value added, and Vodopivec (1989–92), examines impact of foreign and employee for employee ownership with a 1.4% increase. Firms 1997 ownership on firms. with higher revenues, profits, and exports are more likely to exhibit foreign and employee ownership. Dyck 1997 Develops and tests an adverse selection model to ex- Privatized east German firms were more likely to put plain Treuhand’s role in restructuring and privatizing western (usually German) managers in key positions east Germany’s SOEs. In less than 5 years, Treuhand than were companies that remained state-owned. privatized more than 13,800 firms and parts of firms Treuhand emphasized sales open to all buyers rather and, uniquely, had resources to pay for restructuring than favoring east Germans. Principal message: itself—but almost never chose to do so. Instead, it privatization programs must carefully consider when emphasized speed and sales to existing western firms and how to affect managerial replacement in firms. over giveaways and sales to capital funds. Paper ra- Plans open to western buyers and which allow man- tionalizes Treuhand’s approach. agement change are most likely to improve firm per- formance. Frydman, Compares performance of privatized and state-owned Privatization “works,” but only when firm is controlled Gray, Hessel firms in central European transition economies, and by outside owners (other than managers or em- and asks “when does privatization work?” Examines in- ployees). Privatization adds over 18 percentage points Rapaczynski fluence of ownership structure on performance using to annual growth rate of firm sold to domestic finan- 1999 a sample of 90 state-owned and 128 privatized com- cial firm, and 12 percentage points when sold to a panies in Czech Republic, Hungary, and Poland. foreign buyer. Privatization to an outside owner also Employs panel data regression methods to isolate adds about 9 percentage points to productivity growth. ownership effects. Gain does not come at expense of higher unemploy- ment; insider controlled firms are less likely to restruc- ture, but outsider-controlled firms grow faster. Shows the importance of entrepreneurship in reviving sales growth. Weiss and Analyzes effects of ownership by investment funds on Ownership concentration and composition jointly Nikitin 1998 performance of 125 privatized Czech firms during affect performance of privatized firms. Concentration 1993–95. Assesses these effects by measuring rela- in the hands of a large shareholder, other than an in- tionship between changes in performance and in vestment fund or company, is associated with signifi- composition of ownership at the start of privatization. cant improvements for all measures of performance. Uses robust estimation techniques, in addition to Concentrated ownership by funds did not improve OLS, since data strongly reject normality. performance. Preliminary post-1996 data suggests changes in investment fund legislation may improve their performance. Claessens Studies effect of management turnover on changes Finds that the appointment of new managers is asso- and Djankov in financial and operating performance of 706 pri- ciated with significant improvements in profit margins 1999a vatized Czech firms over the period 1993–97. Exam- and labor productivity, particularly if the managers are ines changes in profitability and labor productivity. selected by private owners. New managers appointed by the National Property Fund also improve per- formance, though not by as much. 360 Journal of Economic Literature, Vol. XXXIX (June 2001) TABLE 6 (Cont.) Study Sample description, study period, and methodology Summary of empirical findings and conclusions Claessens Examines the relationship between ownership con- Finds that concentrated ownership is associated with and Djankov centration and corporate performance for 706 pri- higher profitability and labor productivity. Also finds 1999b vatized Czech firms during the period 1992–97. Use that foreign strategic owners and non-bank-sponsored profitability and labor productivity as indicators of investment funds improve performance more than corporate performance. bank-sponsored funds. Frydman, Examines whether the imposition of hard budget Finds privatization alone added nearly 10 percentage Gray, Hessel constraints is alone sufficient to improve corporate points to the revenue growth of a firm sold to outside and performance in the Czech Republic, Hungary, and owners. Most importantly, finds that the threat of hard Rapaczynski Poland. Employs a sample of 216 firms, split between budget constraints for poorly performing SOEs falters, 2000 state-owned (31%), privatized (43%), and private since governments are unwilling to allow these firms (26%) firms. to fail. The brunt of SOEs’ lower creditworthiness falls on state creditors. Frydman, Examines whether privatized central European firms Documents that all state and privatized firms engage Hessel and controlled by outside investors are more entrepre- in similar types of restructuring, but that product Rapaczynski neurial—in terms of ability to increase revenues— restructuring by firms owned by outside investors is 2000 than firms controlled by insiders or the state. Study significantly more effiective, in terms of revenue gen- employs survey data from a sample of 506 manufactur- eration, than by firms with other types of ownership. ing firms in the Czech Republic, Hungary, and Poland. Concludes the more entrepreneurial behavior of outsider-owned firms is due to incentive effects, rather than human capital effects, of privatization— specifically greater readiness to take risks. Harper 2000 Examines the effects of privatization on the financial Finds that the first wave of privatization yielded and operating performance of 174 firms privatized in disappointing results. Real sales, profitability, ef- the first—and 380 firms divested in the second—wave ficiency and employment all declined dramatically of the Czech Republic’s voucher privatizations of 1992 (and significantly). However, second wave firms ex- and 1994. Compares results for privatized firms to perienced significant increases in efficiency and those which remain state-owned. Employs Megginson, profitability and the decline in employment—though Nash and van Randenborgh methodology and varia- still significant—was much less drastic than after first bles to measure changes. wave (−17% vs −41%). Lizal, Singer, Examines the performance effects of the wave of There was an immediate (in 1991) positive effect on and Svejnar break-ups of Czechoslovak SOEs on the subsequent the efficiency and profitability of small and medium 2000 performance of the master firm and the spin-offs. The size firms (both master and spin-offs) and negative for regressions use data for 373 firms in 1991 and 262 the larger firms. The results for 1992 are similar but firms in 1992. not statistically significant. regarding the impact of privatization on related to post-privatization perfor- the performance of divested central and mance, and these studies document eastern European firms. This is espe- consistent and significant relationships. cially true of the five studies—Dyck Other things equal: (1997), Weiss and Nikitin (1998), Claes- sens and Djankov (1999b), Lubomir Li- i) Private ownership is associated with zal, Miroslav Singer, Jan Svejnar (2000), better firm-level performance than and Frydman, Hessel, and Rapaczynski is continued state ownership. Con- (2000)—we consider the most persua- centrated private ownership is asso- sive due to sample size, period of cover- ciated with greater improvement age and/or methodological rigor. All but than is diffuse ownership. one (Joel Harper 2000) of the studies ii) Foreign ownership, where allowed, detailed in table 6 explicitly test is associated with greater post- whether the type of ownership struc- privatization performance improve- ture that emerges from the process is ment than is purely domestic
Enter the password to open this PDF file:
-
-
-
-
-
-
-
-
-
-
-
-