Enterprise Restructuring in Transition: A Quantitative...

54
Journal of Economic Literature Vol. XL (September 2002) pp. 739–792 Enterprise Restructuring in Transition: A Quantitative Survey SIMEON DJANKOV and PETER MURRELL 1 739 1. Introduction O VER THE LAST decade, more than 150,000 large enterprises in 27 transi- tion countries have encountered revolution- ary changes in every aspect of their political and economic environments. Some enter- prises have responded to the challenge, en- tering world markets with great dynamism and becoming indistinguishable from their competitors in mature market economies. Many others remain mired in their past, un- dergoing protracted deaths, delayed at times by their slippage into a world of barter and subsidies. Thus the revolutionary changes in transition countries have been matched by enormous variance in the degree to which enterprises have restructured their opera- tions and responded successfully to events. With changes in the institutional and policy environment much faster and more encom- passing than in virtually any other historical episode, this is as close to a policy laboratory as economics gets. This mammoth quasi-experiment offers lessons of profound importance for eco- nomic studies and economic policy. Since the pace at which firms restructure is a fun- damental determinant of economic growth, analysis of the determinants of restructuring in formerly socialist countries sheds light on the bases of economic progress. Such analy- sis addresses age-old questions and poses new ones. What are the relative productivi- ties of state and private enterprises? Does mass privatization work? What is the effi- ciency cost of diffuse share ownership rela- tive to block-holder ownership? Which pri- vate owners are most effective—managers, workers, banks, or investment funds? Does competition promote productivity change? Does it matter whether competitive pres- sure comes from foreign or domestic firms? To what degree do soft budgets dull enter- prise performance? Is a strengthening of managerial incentives sufficient to inspire turnaround, or is replacement of managers necessary for revitalization? 1 Djankov is senior economist at the World Bank. Murrell is professor of economics and chair of the Academic Council of the IRIS Center, University of Maryland. We thank Erik Berglof, Bernard Black, Olivier Blanchard, Morris Bornstein, Harry Broadman, David Brown, Wendy Carlin, Stijn Claessens, Alexander Dyck, William Evans, John Earle, Roman Frydman, Cheryl Gray, Irena Grosfeld, Laszlo Halpern, Judy Hellerstein, Janos Kornai, John McMillan, John Nellis, Tatiana Nenova, Guy Pfeffermann, Yingyi Qian, Mark Schaffer, Paul Seabright, Marcelo Selowsky, Andrei Shleifer, Jan Svejnar, and three anonymous referees for helpful ad- vice. This research was made possible through support provided by the World Bank and by the U.S. Agency for International Development under Cooperative Agreement DHR-0015-A-00–0031–00 to the Center for Institutional Reform and the Informal Sector (IRIS). The findings, interpretations, and conclusions expressed in this paper are entirely those of the au- thors. They do not necessarily represent the views of the IRIS Center, US AID, the World Bank, its execu- tive directors, or the countries they represent.

Transcript of Enterprise Restructuring in Transition: A Quantitative...

Page 1: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

Journal of Economic Literature Vol. XL (September 2002) pp. 739–792

Enterprise Restructuring in Transition:A Quantitative Survey

SIMEON DJANKOV and PETER MURRELL1

739

1. Introduction

OVER THE LAST decade, more than150,000 large enterprises in 27 transi-

tion countries have encountered revolution-ary changes in every aspect of their politicaland economic environments. Some enter-prises have responded to the challenge, en-tering world markets with great dynamismand becoming indistinguishable from theircompetitors in mature market economies.Many others remain mired in their past, un-dergoing protracted deaths, delayed at timesby their slippage into a world of barter andsubsidies. Thus the revolutionary changes in

transition countries have been matched byenormous variance in the degree to whichenterprises have restructured their opera-tions and responded successfully to events.With changes in the institutional and policyenvironment much faster and more encom-passing than in virtually any other historicalepisode, this is as close to a policy laboratoryas economics gets.

This mammoth quasi-experiment offerslessons of profound importance for eco-nomic studies and economic policy. Sincethe pace at which firms restructure is a fun-damental determinant of economic growth,analysis of the determinants of restructuringin formerly socialist countries sheds light onthe bases of economic progress. Such analy-sis addresses age-old questions and posesnew ones. What are the relative productivi-ties of state and private enterprises? Doesmass privatization work? What is the effi-ciency cost of diffuse share ownership rela-tive to block-holder ownership? Which pri-vate owners are most effective—managers,workers, banks, or investment funds? Doescompetition promote productivity change?Does it matter whether competitive pres-sure comes from foreign or domestic firms?To what degree do soft budgets dull enter-prise performance? Is a strengthening ofmanagerial incentives sufficient to inspireturnaround, or is replacement of managersnecessary for revitalization?

1 Djankov is senior economist at the World Bank.Murrell is professor of economics and chair of theAcademic Council of the IRIS Center, University ofMaryland. We thank Erik Berglof, Bernard Black,Olivier Blanchard, Morris Bornstein, Harry Broadman,David Brown, Wendy Carlin, Stijn Claessens,Alexander Dyck, William Evans, John Earle, RomanFrydman, Cheryl Gray, Irena Grosfeld, LaszloHalpern, Judy Hellerstein, Janos Kornai, JohnMcMillan, John Nellis, Tatiana Nenova, GuyPfeffermann, Yingyi Qian, Mark Schaffer, PaulSeabright, Marcelo Selowsky, Andrei Shleifer, JanSvejnar, and three anonymous referees for helpful ad-vice. This research was made possible through supportprovided by the World Bank and by the U.S. Agencyfor International Development under CooperativeAgreement DHR-0015-A-00–0031–00 to the Centerfor Institutional Reform and the Informal Sector(IRIS). The findings, interpretations, and conclusionsexpressed in this paper are entirely those of the au-thors. They do not necessarily represent the views ofthe IRIS Center, US AID, the World Bank, its execu-tive directors, or the countries they represent.

Article 1 8/12/02 9:13 AM Page 739

Page 2: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

Answers to these questions are obviouslyof vital significance for economic delibera-tions in general. But beyond this, the transi-tion process is important in and of itself be-cause of its geographical scope, the largechanges in levels of economic well-being inthe last decade, and the ramifications forthe world economy and polity. Analysis ofenterprise restructuring is central in any ef-fort to understand the effects of the mostimportant reform measures adopted intransition countries. With enterprise re-structuring apparently more successful insome countries than others, the naturalquestion that arises is whether such differ-ences relate systematically to policy. In thispaper, we address this question by examininghow the effects of policy have varied betweentransition countries.

Our objective is to survey the evidence onthe determinants of enterprise restructuringin transition.2 We provide such a synthesis,summarizing the composite conclusionsfrom more than one hundred empiricalstudies. Where possible, we compare the re-sults from the transition literature with thosefrom studies of market economies.

With such a large body of literature underreview, it is necessary to pay special atten-tion to the methodology of synthesis.Because there are so many results, verbaldescription alone would result in a hard-to-remember list. An interpretative summarypresents its own dangers. Bayesian priorsmight come to weigh too heavily in the syn-thesis, a danger that is all too great in thetransition arena, where the contentiousnessof the subject has encouraged forthrightstatements. Indeed, we have made such

statements, although the reader might be re-assured to note that our priors to some ex-tent cancel (Murrell 1992; Gerhard Pohl etal. 1997).

In view of these factors, we adopt moreroutinized methods of synthesizing the evi-dence, drawing on insights from meta-analysis, which has long been in use in otherdisciplines, particularly bio-medicine, psy-chology, and education (Morton Hunt1997).3 Apart from making our methods ofsynthesis transparent, application of meta-analysis has several other advantages. First,it provides tools to aggregate the results ofmany studies on a similar topic, combiningmany tests with weak power to produce asingle one with larger power. Second, thesemethods allow one to test hypotheses acrossgroups of studies. For example, we examinewhether the replacement of managers ismore effective than the addition of incen-tives, and we test whether privatization hasstronger results in Eastern Europe than inthe former Soviet Union. Third, the synthe-sis of results can address the thorny issue ofdifferences in the quality of studies, allowingone to gauge the extent to which the conclu-sions change when one gives greater weightto those studies that are methodologicallysounder.

We find that privatization is strongly asso-ciated with more enterprise restructuring.Economic effects are quite often very large,for example adding several percentagepoints to enterprise growth rates. The pri-vatization effect is, however, statisticallyinsignificant in the Commonwealth of In-dependent States (CIS). These results arerobust. They hold when we vary the empha-sis assigned to the results of different studiesby using weights that reflect the differing at-tention paid to selection bias or the overallquality of analysis.

740 Journal of Economic Literature, Vol. XL (September 2002)

2 The previous survey papers in this area, JosefBrada (1996) and Oleh Havrylyshyn and DonalMcGettigan (2000), used quite limited empirical evi-dence, which came almost exclusively from CentralEurope and China. Now studies of other countries (theformer Soviet Union, Mongolia, and Vietnam) are be-coming numerous, providing a much wider variety ofevidence. John Nellis (1999) covers the full range ofcountries; here we provide a more systematic summaryof the evidence and focus on a wider set of determi-nants of enterprise restructuring.

3 Examples of recent use of meta-analysis in eco-nomics are Kerry Smith and Ju Chin Huang (1995),Smith and Yoshiaki Kaoru (1990), David Neumark andWilliam Wascher (1998), Joseph Phillips and ErnestGoss (1995), and Thomas Stanley (1998).

Article 1 8/12/02 9:13 AM Page 740

Page 3: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

The survey also documents the effects ofdifferent types of owners on enterprise re-structuring. We find that state ownershipwithin traditional state firms is less effectivethan all other ownership types, except forworker-owners, who have a negative effect.Privatization to outsiders is associated with50 percent more restructuring than privati-zation to insiders (managers and workers).Investment funds, foreigners, and otherblock-holders produce more than ten timesas much restructuring as diffuse individualownership. State ownership within partiallyprivatized firms is surprisingly effective, pro-ducing more restructuring than enterpriseinsiders and non-block-holder outsiders.The effects of different owners vary be-tween regions. Workers are better owners inEastern Europe than in the CIS, whilebanks and concentrated individual owner-ship are significantly more effective in theCIS than elsewhere.

Product market competition has a signifi-cant effect in improving enterprise perfor-mance. The economic effects can be large,with a typical study indicating that enter-prises in highly competitive sectors are20–30 percent more productive than mo-nopolies. The sources of improvement differbetween regions, however. In EasternEurope, the beneficial effect comes mainlythrough import competition but is also evi-dent through domestic competition. In con-trast, in the CIS domestic competition issometimes statistically insignificant whileimport competition generally has a negativeeffect on enterprise restructuring.

We next explore the link between enter-prise restructuring and the hardening ofbudget constraints. The evidence is con-sistent with the view that hardened budgetconstraints have had a beneficial effecton enterprise restructuring in EasternEurope and the CIS. The effect in the CISis economically larger than that in EasternEurope.

We also examine whether managementturnover—or more broadly, bringing in

new human capital—is associated with im-proved enterprise performance. Statisticalanalyses show that this is the case in bothEastern European and CIS countries. Wefind evidence that the strengthening ofmanagerial incentives leads to a greateramount of restructuring, in economicterms.

Our review of existing research thenleads to suggestions on three priority areasfor future research. The design of thestudies we survey was not driven by at-tempts to address difficult methodologicalproblems like selection bias or simultane-ity, or by efforts to examine the effects ofless easily quantifiable aspects of policy,such as the quality of institutional con-struction, or by considerations that viewedtransition in the wider context of develop-ment. In the future, research should focuson these issues.

We conclude by comparing the strengthof the effects of the different policy reformsexamined in this paper. Privatization to out-siders is associated with the largest restruc-turing gains, while privatization to workershas no effect in Eastern Europe and is detri-mental in the CIS. Hardened budget con-straints are the second most importantdeterminant in the CIS, while the establish-ment of competition is second in EasternEurope.

The paper is organized as follows.Section 2 lays out the nature of the studiesunder review, in general terms. Section 3investigates the empirical evidence onwhether state-owned or privatized firmsundertake more economic restructuring.Section 4 studies the effects of differenttypes of owners. Section 5 links productmarket competition and enterprise restruc-turing efforts. Section 6 analyzes the role ofsoft-budget constraints in limiting produc-tivity enhancements. Section 7 documentsthe role of managers, focusing on manage-ment turnover and manager incentives.Section 8 develops the future researchagenda. Section 9 concludes.

Djankov and Murrell: Enterprise Restructuring in Transition 741

Article 1 8/12/02 9:13 AM Page 741

Page 4: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

2. The Nature of Reviewed Studies:Enterprise Restructuring and Its

Determinants

What is enterprise restructuring and whatchanges might induce it in transition coun-tries? The perspective taken here is found atthe core of every paper that we survey. Theenterprises initially functioned in socialisteconomies, and their behavior was a productof the institutions and policies of thoseeconomies. In the 1990s, those institutionsand policies changed radically. Thesechanges compelled enterprises to adapttheir behavior in order to survive, and per-haps to succeed, in a new, liberalized, mar-ket environment. The term enterprise re-structuring has come to denote the wholeprocess undertaken by enterprises as theyadapt for survival and success in a marketeconomy.4

The search for possible determinants ofenterprise restructuring therefore lies inthose institutions and policies that changedmost rapidly in the early years of transition.To identify these, it is useful to reflect on thecentral characteristics of the socialist econ-omy and its enterprises. These have beenwidely discussed in the literature and weonly need to reiterate a few central issues.5

The classical socialist enterprise was by def-inition state-owned and was oriented to an input-output plan rather than any market.Meeting the plan was of prime importanceand the plan was normally very ambitious.Therefore, production issues dominated en-trepreneurship, marketing, and cost minimiza-tion in managerial concerns. Consistently, thetypical manager was a production engineerand not a businessman. Managers faced a mixof monetary and career-based incentives,which were a function of plan fulfillment, en-terprise performance, and political loyalty.

Profits and efficiency were much less impor-tant than they are under capitalism.

The enterprise was organized along veryhierarchical lines. Workers had virtually norole in enterprise decision making, except inpersonnel policy, leading to firing rates thatwere extremely low by any standard (DavidGranick 1987). The enterprise was not onlya producer of goods, but also played an im-portant role in implementing the state’s so-cial welfare policies. For these reasons, effi-ciency considerations were often secondaryin determining the size of the workforce.

A labyrinthine bureaucracy replaced theinstitutions and markets of capitalism.Bureaucratic pressure substituted for com-petition. The bureaucracy found customersand determined prices, interceding betweenproducer and buyer. It generated contractsand enforced them. Its one-year plans guar-anteed short-term working capital, and itsinvestment projects automatically receivedlong-term credits. Given the ubiquitous roleof the state, much was decided by negotia-tion. One consequence of the frequency ofthese negotiations was the universal pres-ence of soft budgets, which further turnedattention away from profits and efficiency.

Pre-transition reforms did change this stan-dard picture in a few countries, notablyYugoslavia (which was centralized only for afew years), Hungary, and Poland (LeszekBalcerowicz 1995, and Kornai 1986). De-centralizing reforms reduced the scope of bureaucratic decision making. Markets andcompetition increased in importance.Enterprises came closer to ultimate con-sumers. Paradoxically, however, abandonmentof formal planning led to increased bargainingbetween bureaucracy and enterprise, furthersoftening budgets. Notably also, workersgained more power within enterprises, acquir-ing experience at being informal owners.

In sum, the pre-transition enterpriseswere state-owned, protected from competi-tion, shielded from failure by soft budgets,and managed by production engineers withincentives oriented toward the plan or politics.

742 Journal of Economic Literature, Vol. XL (September 2002)

4 See, for example, Phillippe Aghion and MarkSchankerman (1999), Roman Frydman et al. (2000),and Lubomir Lizal, Miroslav Singer, and Jan Svejnar(2001).

5 See Joseph Berliner (1976), Murrell (1990), andJanos Kornai (1990) for details.

Article 1 8/12/02 9:13 AM Page 742

Page 5: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

The enterprises were embedded in a set ofinstitutions, for contracting, for financing,for governance, that were far different fromthose of a market economy. Transition poli-cies aimed to change all of this. Therefore,the literature with which we are concernedemphasizes the following determinants ofrestructuring: ownership, competition, softbudgets, managerial incentives and charac-teristics, and broader institutional changes.

The widespread liberalization meant thatenterprises had to adapt in order to surviveand prosper in markets that were increas-ingly contestable, the process of change thatthe literature conventionally defines as en-terprise restructuring. In some studies, themeasurement of restructuring focuses di-rectly on enterprise decisions, for example,changes in the structure of corporate gover-nance and management (Saul Estrin andAdam Rosevear 1999a) or renovation of fac-tories (Djankov 1999a), or investment rates(Irena Grosfeld and Jean-Francois Nivet1999). But more usually, the degree of en-terprise restructuring has been captured byperformance, with performance measuredby variables that are objectives of companiesin market economies and that were less im-portant for socialist enterprises. Thus, pro-ductivity (e.g., Stephen Smith, Beom-CheolCin, and Milan Vodopivec 1997; and YoungLee 1999) or profits (e.g., Stijn Claessensand Simeon Djankov 1999a; and Estrin andRosevear 1999a) are often used. But sales orrevenue have also been used extensively(e.g., Frydman et al. 1999; and Derek Jones1998) under the premise that the ability tohold on to customers or to find new ones isan indicator of successful change within theenterprise, especially when accomplished inthe face of steep recessions (Frydman et al.1999).6 There seems to be no consensus in

the literature concerning which variables arethe best measures of restructuring, apartfrom a greater preference for measures ofperformance than for indexes of internaldecisions.7

The typical study relates these restructur-ing measures to their determinants, estimat-ing an equation of the form:

Y = α + X β + γ P + ε (1)

where the enterprise is the unit of observa-tion, Y is the restructuring measure, P cap-tures some aspect of the reforms to whichthe enterprise is subject, e.g., ownershipchange, degree of hardness of the budget,etc., X is a vector of other enterprise charac-teristics, and ε is an error term. γ is the pa-rameter of direct interest. The overwhelm-ing majority of papers justifies their exactchoice for (1) pragmatically, rather than pre-senting a structural derivation.

The studies we examine use data frommedium-large and large enterprises. Theseenterprises were the core of the socialisteconomy, and when liberalization came theyremained going concerns, leading to somecontinuity in operations and personnel. Thiscontinuity facilitated data collection. In con-trast, smaller enterprises were notoriouslyweak under socialism and soon wereswamped by new entrants. They often van-ished, with their assets resurfacing in newactivities, with new personnel. Data collec-tion for smaller enterprises, therefore, facedenormous difficulties, leading to few studiesexamining the progress of the small enter-prises established before transition.

The paragraphs above emphasize thecommon elements of the studies under re-view. We now turn to important ways inwhich studies differ. The cataloging of thesedifferences and the determination of theireffect on results are important featuresof our survey. We focus on three such vari-ations in the characteristics of studies:

Djankov and Murrell: Enterprise Restructuring in Transition 743

6 There is wide variation in whether levels or growthrates are used. This decision often depends on theavailability of data and also reflects whether lagged de-pendent variables are included as explanatory variablesin the equation. The differences between levels andgrowth specifications diminish greatly with the inclu-sion of lagged dependent variables.

7 The lack of consensus probably results from thedifficulties of obtaining data, with researchers using thebest measure available.

Article 1 8/12/02 9:13 AM Page 743

Page 6: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

measurement of Y, selection bias, and over-all quality.

Many varieties of Y appear in the studies,but one distinction, already suggestedabove, is worth emphasizing. One categoryof Y comprises quantitative indicators of en-terprise performance, which are based onaccounting information. Other Y’s are some-what softer, perhaps derived from surveyquestions on economic performance posedto managers, e.g., forecasts of sales in thesurveyed year as in John Earle (1998), or us-ing information collected about reorganiza-tion, e.g., the introduction of new productsas in Wendy Carlin et al. (2001), or reflect-ing operational factors only indirectly re-lated to current performance, e.g., the ex-tent of wage arrears, as in HartmutLehmann, Jonathan Wadsworth, andAlessandro Acquisti (1999). We will refer tothese two types of indicators as quantitativeand qualitative.8

The prevailing sentiment in the literatureis that the quantitative variables are to betrusted more (even with the misreportingand accounting difficulties that are rife intransition countries). They measure directlythe prime output of enterprise restructur-ing: economic performance. On the otherhand, there is also the view that quantitativeperformance might suffer when an enter-prise is investing in large-scale reorganiza-tion and that the results of this processmight be observed earliest in qualitativevariables. We focus primarily on the quanti-tative indicators in this paper, deeming themmore reliable. Where a sufficient number ofstudies is available, however, we examineboth types.

A second crucial variation between paperslies in the degree of attention paid to possi-ble biases in the estimates of γ due either to

selection effects or simultaneous causation.9Typical cases of selection bias occur when ei-ther the level of ownership or the use ofmanagerial incentives is systematically re-lated to some unobserved enterprise charac-teristic that also affects Y. Examples ofsimultaneous causation are when an enter-prise’s market share is being used to meas-ure competition or subsidies are being usedto measure soft budgets: in each case, themeasured determinant of restructuring re-flects enterprise performance. These prob-lems have been thoroughly recognized inthe literature, but solutions are not alwayseasy to obtain. Thus, for example, in thestudies examining private versus state own-ership in section 3, only 53 percent of the es-timates of γ employ methods that mightserve to counter selection bias.

The prevailing evidence suggests biasesdue to selection effects or simultaneouscausation are a real possibility. For exam-ple, Sweder van Wijnbergen and AntonMarcinin (1997) show that selection intoCzechoslovakia’s voucher program was non-random and that this must be taken into ac-count in ascertaining the effects of thevoucher program on outcomes. Moreover,when there are OLS estimates and estimatesemploying some technique to counter biasin the same paper, they quite often differconsiderably, suggesting the presence ofbias. (See, for example, David Brown andJohn Earle 2000; Rachel Glennerster 2000;James Anderson, Lee, and Murrell 2000;and Earle 1998.)

The fact that some studies have identifiednontrivial selection or simultaneity bias sug-gests that we must be sensitive to its pres-ence. Hence, in every section of this paper,we present conclusions that allow the readerto understand how the synthesis of resultsmight have been affected by these biases.We present sufficient information to alloweach reader to judge which of the broad

744 Journal of Economic Literature, Vol. XL (September 2002)

8 We do not use indicators for which there is sub-stantial disagreement in the literature on whether thesign of γ should be positive or negative. The most perti-nent example is employment, whose direction ofchange would depend very much on the extent of ex-cess labor under the old regime.

9 Our use of the term selection bias might not bewholly consistent with the precise terminology ofeconometrics. We simply follow the literature.

Article 1 8/12/02 9:13 AM Page 744

Page 7: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

conclusions emanating from this literaturefall under the shadow of selection bias. Howwe do this is best explained in a practicalcontext, in the next section. But it must beemphasized that we have directly con-fronted issues of selection or simultaneity bi-ases in reaching our conclusions, that wedraw back from stating strong conclusionswhere we feel the possibility of such biasleaves some residual doubt, and that never-theless there are many strong conclusionsthat we can reach.

A third variation between papers onwhich we focus is quality of analysis. One ofthe primary objections to the application ofmeta-analysis hinges on the fact that thequality of empirical work varies greatlyacross papers, meaning that a simple aggre-gation of results might give undue weight tolesser quality research. Therefore, somescholars prefer to focus reviews of empiricalliterature on the high points, ignoring pa-pers that are less convincing from a method-ological standpoint. However, it is also possi-ble to take a middle road, one that examineswhether the composite results change whenconsiderations of quality are taken into ac-count. This is our approach. How we do thisis, again, best left until our methods can beexplained in context in the next section.10

3. State versus Private Ownership

In early transition debates, there wasagreement on the goal of an economy domi-nated by private ownership, but conflictingviews on how best to attain this, through fastprivatization (Jeffrey Sachs 1992; Maxim

Boycko, Andrei Shleifer, and Robert Vishny1995) or through stimulation of a nascentprivate sector (Kornai 1990; Murrell 1992,1995). The relative emphasis on the differingstrategies waxed and waned with events.With deep crisis in Eastern Europe in theearly 1990’s, fast privatization gained empha-sis. However, after the recovery of Poland, arelatively slow privatizer, that emphasis de-clined (Brian Pinto, Marek Belka, and StefanKrajewski 1993; Phillippe Aghion, OlivierBlanchard, and Robin Burgess 1994; andBrada 1996). But Poland is only one of manytransition countries, an outlier at that. Nowthere are fast privatizers performing well(Estonia) and fast privatizers performingbadly (Russia), with similar variation acrossslow privatizers (Poland versus Romania),giving sustenance to a variety of opinionsabout the results of privatization (Pohl et al.1997; Havrylyshyn and McGettigan 2000;Joseph Stiglitz 1999; Bernard Black, ReinierKraakman, and Anna Tarassova 2000). Toreferee this debate, it is necessary to turn tothe microeconomic empirical literature.11

Studies examining whether privatized en-terprises perform better than state-owned en-terprises use equation (1), with P a measure ofthe degree of private ownership of an enter-prise of pre-1989 vintage. A large variety ofvariables takes the place of Y, X, or P with noconsensus on the nature of the equation to beestimated. For Y, a significant number of pa-pers use measures of output levels, captured,for example, by sales (e.g., Jozef Konings1997; and Earle 1998) or value added (e.g.,Anderson, Georges Korsun, and Murrell2000; and Smith, Cin, and Vodopivec 1997).Other papers use growth rates of the samevariables (e.g., Frydman et al. 1999; and Jones1998). Qualitative versions of Y are also exten-sively used, especially in studies of the formerSoviet Union, where there are greater difficul-ties in obtaining accurate accounting data.

Djankov and Murrell: Enterprise Restructuring in Transition 745

10 An earlier version of this paper (Djankov andMurrell 2000) examined other variations in themethodological characteristics of studies. It applied themethods that appear below (on quality and selectionbias) also to the length of time (for the pertinent re-form) that is embodied in the estimates and the num-ber and appropriateness of the variables used in X. Inthe present paper, these dimensions of quality are sim-ply absorbed into the overall measure of quality. Thereader can refer to that earlier paper for the more de-tailed results. The data compiled for that study, exceptfor our assessments of the overall quality of papers, areavailable from the authors on request.

11 William Megginson and Jeffry Netter (2001) un-dertook an extensive survey of the effects of privatiza-tion worldwide. Their results are primarily germane tocountries with long-established market economies.

Article 1 8/12/02 9:13 AM Page 745

Page 8: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

One common form of P is the percentageof shares owned privately (e.g., Brown andEarle 2000). Other papers use a dummyvariable capturing whether private owner-ship is above some threshold value, for ex-ample 0 percent in Estrin and Rosevear(1999a) and 50 percent in Derek Jones andNiels Mygind (2002). Frydman et al. (1999)and Claessens and Djankov (2000) formtheir dummy variables on the basis of corpo-rate laws, the former defining a privatizedenterprise as one in which private ownershipis above the 33.3-percent threshold neededto block strategic company decisions, whilethe latter also uses the 66.7-percent levelthat guarantees that a state minority ownercannot veto such decisions.

Variety is also present in the X’s. Sectoral,regional, and (in the case of panels) time andfirm dummies are common. Among the X’sused in various studies are measures of com-petition, levels of soft budgets, size of firms,access to finance, year of privatization, andmany more. Pre-privatization levels of de-pendent variables are sometimes includedunder the premise that these might dampenthe effects of selection bias. One commonformulation includes capital and labor in theX’s and uses output for Y, with the basic equa-tion then representing a production functionand the estimate of γ capturing the effect ofownership on total factor productivity.

The variety in the formulation of estimat-ing equations is a reflection of two factors.First, there is the absence of a single com-pelling theory that models the process ofchange within an enterprise.12 Without suchtheory, specifications for estimating equa-tions rely on ad hoc formulations. Second,the set of variables for which data are avail-able varies greatly, with every study havingdeficiencies in some respect. Given thesereasons for the variety of approaches, withnone obviously superior to all others, itseems judicious to include a wide range ofstudies in drawing general conclusions.

The similarities and differences betweentwo papers (Frydman et al. 1999; andAnderson, Lee, and Murrell 2000) exhibitthe methodological decisions to be madewhen conducting such studies and the varia-tions in results that can be obtained.Frydman et al. examine the performancefrom 1990 to 1993 of a panel of 218 pri-vatized and state firms from the CzechRepublic, Poland, and Hungary, whileAnderson et al. focus on 1995 data for 211privatized (including partially state-owned)Mongolian enterprises. Data collection, in-cluding sample design, was carried outspecifically for each of these studies, raisingthe quality, extensiveness, and appropriate-ness of the information collected but result-ing in small sample sizes. Each study exam-ines the effects of privatization as a wholeand the effects of different owners. We dis-cuss the latter in the next section.

Both studies wrestle with the choice ofthe dependent variable, how to measureownership, which X’s to include, and how tocounter selection bias. Their decisions differa great deal. Frydman et al. use four differ-ent dependent variables, rates of growth ofrevenues, employment, revenues per em-ployee, and costs per unit of revenue.13

Their ownership variable is a dummy. Ascontrols, they use initial levels (not growthrates) of the four performance measures, ac-companied by sectoral, country, and timedummies. Selection bias is parried in a num-ber of ways, primarily by methods devel-oped for the analysis of treatment effectsin panels. In separate analyses, the authorsuse a dummy variable (in X) capturing pre-privatization differences between state firmsand privatized firms; they employ a firmfixed-effects model, and they verify that per-formance of those firms slated for privat-ization, but not yet privatized, is closer tothat of state firms than privatized ones. The

746 Journal of Economic Literature, Vol. XL (September 2002)

13 A major aspect of Frydman et al. (1999) is the ef-fect of privatization on different dependent variables.For reasons of space, we cannot examine this interest-ing issue here.

12 Additionally, there is also no agreed theory on howto model the selection of firms for privatization.

Article 1 8/12/02 9:13 AM Page 746

Page 9: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

cumulative effect of these analyses is to con-vince the reader that the privatization ef-fects are real, rather than an artifact of selec-tion for privatization.

Enterprise record-keeping during Mon-golia’s early transition was so poor thatAnderson et al. could not obtain panel data.Hence, they focus on performance in oneyear, using three different dependent vari-ables, total factor productivity (within aCobb-Douglas production function), salesper employee, and value-added per em-ployee. The equations for the latter two vari-ables include a lagged dependent variable,thus nesting a specification in which growthis the dependent variable. Since the esti-mated coefficients on the lagged dependentvariables are significantly different fromone, this suggests that growth measures arenot suitable dependent variables. The own-ership variable is the percentage of enter-prise shares held by non-state owners.Controls include regional and sectoral dum-mies, levels of competition, and the pres-ence of soft budgets. Selection bias is coun-tered through the use of instrumentalvariables. Suitable instruments were avail-able because of idiosyncratic features of theprivatization program and due to the varyingincentives of different types of owners dur-ing privatization. Comparison of OLS andinstrumental variables results suggests thatOLS estimates of the effect of privatizationare upwardly biased.

Frydman et al. and Anderson et al. obtainstrikingly different results. In CentralEurope, privatization improves revenuegrowth by approximately 7 percent a year; inMongolia, wholly private firms are 30 per-cent to 70 percent less efficient than com-pletely state-owned firms. Although thereare many differences between the analysesof these two papers, it is quite unlikely thatmethodological differences can explain thedivergence in results, since both papers useconventional methods to solve the usualproblems and pay close attention to possiblebiases. An obvious candidate to explain the

differences in results is the countries stud-ied, the most advanced transition countriesversus one of the most backward. We exam-ine this issue later in this section. Anotherpossibility is the types of owners generatedduring privatization, which is the subject ofsection 4.

One respect in which the two papers aresimilar is in their use of quantitative depen-dent variables derived from accounting infor-mation. In this respect, Carlin et al. (2001)provide an interesting contrast. They usedata from a survey of over three thousandfirms in 25 countries. Because of the logisti-cal and compatibility issues involved in col-lecting accounting data from such a wide va-riety of countries, they rely for theirdependent variables on responses to ques-tionnaires. One set of questions focused onchanges in real sales and employment, lead-ing to a measure of productivity growth.Another set focused on internal restructur-ing, asking about upgrading of existing prod-ucts, introduction of new products, openingof new plants, and quality accreditation.Using principal components to combine theanswers on this latter set, the paper derives ameasure of new product restructuring.

Carlin et al. do not find a direct effect ofchange in ownership on productivity or salesgrowth, but they do find a significant effectof privatization on new product restructur-ing. Moreover, new product restructuringdirectly increases sales and productivitygrowth, implying an indirect effect of own-ership on enterprise growth. These resultsgive some sustenance to the argument thatqualitative variables provide useful informa-tion where accounting ones are not avail-able. The authors nevertheless caution thereader that they were not able to implementprocedures to counter selection bias on theownership variable.

The three papers discussed above gener-ate highly varying results, which is a charac-teristic of the literature as a whole. In table1, we focus on the economic size of effects.Such information is usually not emphasized

Djankov and Murrell: Enterprise Restructuring in Transition 747

Article 1 8/12/02 9:13 AM Page 747

Page 10: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

in the papers under review. Indeed, it is re-markable that in most cases we had to calcu-late the numbers in table 1 ourselves, some-times using a considerable degree ofinterpretation. Moreover, a significant num-ber of papers did not offer enough informa-tion for analogous estimates to be calcu-lated. This suggests that the surveyed papersdo not emphasize the economic interpreta-tion of their estimates sufficiently.

The estimates in table 1 are sampled toreflect a variety of contexts and results. Thefirst conclusion from the table is that privat-ization has economically significant effects.A second conclusion is that these effects arefar from uniform. Privatization done in theright way, or under the right circumstances,can have huge positive effects, but privatiza-tion can also be hugely detrimental.

This large variation in results suggests thatan overall judgment on the effects of privat-

ization can only be reached by employingmethods that aggregate the results of themany papers under consideration. Moreover,to understand the reasons for variations be-tween estimates, the aggregation methodsshould offer the possibility of assessing howmethodological factors influence estimatesand should allow for comparisons betweensets of studies. It is to this aggregate analysisthat we now turn.

3.1 Statistical Significance: Combining t-Statistics

We are interested in the size and statisticalsignificance of the estimate of γ (γ̂ ) in equa-tion (1). The γ̂ ’s of different studies are notdirectly comparable because Y and P aremeasured in many different ways. Therefore,we seek methods of combining the results ofstudies with different characteristics. We firstbegin by combining t-statistics.

748 Journal of Economic Literature, Vol. XL (September 2002)

TABLE 1GAINS OR LOSSES DUE TO OWNERSHIP CHANGE: A SAMPLING OF ESTIMATES

Estimates of the performance of a 100% private firm relative to a 100% state firm. Positive signs indicate superiorprivate performance. For the growth measures, relative performance is private-firm yearly growth rate minus state-firm yearly growth rate. For the levels measures, relative performance is private-firm productivity minus state-firmproductivity expressed as a % of state productivity.

Output or Sales Growth Productivity Growth Levels of Productivity

% private % private private minusminus % minus % state as %

state country state country of state country

8.7 Poland1 4.3 Central Europe2 −43 Russia11

7.3 Central Europe2 3.5 Russia3 62 Russia12

−9.7 Russia3 10.6 Kyrgyz6 14 Estonia13

10.9 Bulgaria4 3.6 Georgia/Moldova7 −30 Lithuania14

(MBOs) (diffuse owners)2.5 Czech Republic4 0.9 Georgia/Moldova7 49 Latvia14

(voucher privatized) (individual owners)2.1 Hungary4 16.0 Romania8 −65 Mongolia15

−4.5 Russia5 1.0 25 countries9 36 Romania8

18.0 Kyrgyz6 3.1 Eastern Europe10 140 Slovenia16

Notes: These estimates are based on the authors’ calculations, derived from information presented in the following pa-pers: 1 Grosfeld and Nivet (1999); 2 Frydman et al. (1999); 3 Jones (1998); 4 Pohl et al. (1997); 5 Perevalov, Gimadi, andDobrodey (2000); 6 Roberts, Gorkov, and Madigan (1999); 7 Djankov (1999a); 8 Earle and Telgedy (2001); 9 Carlin etal. (2001); 10 Claessens and Djankov (2000); 11 Brown and Earle (2000); 12 Earle (1998); 13 Jones and Mygind (2002);14 Jones and Mygind (2000); 15 Anderson, Lee, and Murrell (2000); 16 Smith, Cin, and Vodopivec (1997).

Article 1 8/12/02 9:13 AM Page 748

Page 11: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

In this section, we combine the results of37 distinct studies. Table 2 lists these stud-ies, the countries included in them, their Y’s,and their P’s. Within these studies, we haveidentified 93 distinct empirical analyses andextracted 93 γ̂ ’s together with their corre-sponding t-statistics. We use more than oneanalysis from a single paper only in cases inwhich the estimates are derived from con-ceptually distinct analyses, e.g., from com-pletely different forms of the dependentvariable or from different countries.

The theory justifying the aggregation oft-statistics is analogous to that used whenconducting tests on the mean of a sample.Collect several independent observationsfrom the same distribution, find their mean,and deduce the distribution of the mean.The variance of the sample mean will besmaller than that of individual observations,implying that statistical tests based on themean will be more powerful than tests on in-dividual observations.

The data comprise M t-statistics on γ̂ ’s,t1, . . . , tM. The following is normally dis-tributed:14

(2)

M, which is the number of analyses, playsthe role of sample size. It is readily apparentthat a set of analyses with small positivet-statistics could be significant in the aggre-gate even with nonsignificance in each indi-vidual analysis. As it happens, less than one-halfof the t-statistics examined in this section arestatistically significant, but collectively theyare highly significant.

Our synthesis of results relies on tests of (2),whereas the most common method of combin-ing results in literature reviews is the methodof vote-counting (John Hunter and FrankSchmidt 1990). Vote-counting concludes thatthere is nonsignificance in the aggregate whena set of studies has a median t-statistic that isinsignificant. This method produces mislead-

ing conclusions, since it combines probabilityinformation erroneously, a point that is best il-lustrated with a simple example.

Researcher A obtains a t-statistic of 2 in astudy, pronouncing significance for the effect.Researcher B, not favorable to A’s conclusions,conducts two separate studies of two separatecountries and obtains t-statistics of 1 in eachstudy. Researcher B triumphantly announcesthat A has been mistaken, for the vote is now 2studies to 1 for nonsignificance. But the com-bined statistic obtained by applying (2) to allthree studies is 4/√3 = 2.31. B has actuallystrengthened support for A’s conclusions.

Column (1) of table 3 contains the resultsobtained by applying equation (2). The rowsreflect two different ways of grouping esti-mates. First, there is the quantitative-qualita-tive division of dependent variables, dis-cussed above. Second, there are regionalgroupings. Corresponding to much of therest of the literature (e.g., EBRD 1999), thebasic split is between the non-Baltic formerSoviet Union (the CIS) and the rest.15 Thecountries outside the CIS comprise EasternEurope and the Baltics (with one study ofChina). For brevity, but at the cost of someinaccuracy, we will refer to the two groups asEastern Europe (EE) and the CIS in the restof the paper. Interestingly, once we seek acriterion that corresponds to our split ofcountries, we find that it is the length of timethat the countries labored under commu-nism, seventy years for each CIS country andless than fifty years in EE.16 The readertherefore might like to think of our regionalgroups as “two generations” and “three gen-erations,” indicating the length of time undercommunism.

The significant effects of privatizationshow clearly in all but one of the statistics

Djankov and Murrell: Enterprise Restructuring in Transition 749

14 Assuming sample sizes in all analyses are suffi-ciently large, which is the case here.

15 In the set of papers under consideration, thereare two studies of Mongolia. Since this countrylooks like a typical member of the CIS (Korsun andMurrell 1995), Mongolia is included in the CISgrouping.

16 Moldova is an exception to this rule. This countrycontributes only a small amount of data to this paper.

ΣΜ

k=1

tk √ M

Article 1 8/12/02 9:13 AM Page 749

Page 12: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

TABLE 2STUDIES OF PRIVATIZATION

Levels or Privatization Paper Countries Dependent variables Growth Measure

Anderson, Korsun, Murrell 2000 Mongolia softness of budgets levels % ownershipAnderson, Lee, Murrell 2000 Mongolia TFP, labor productivity levels % ownershipBrown and Earle 2000 Russia TFP levels % ownershipBrown and Brown 1999 Russia sectoral profitability levels % of sectoral output

in private firmsCarlin et al. 2001 25 countries productivity, creation growth, dummy if privatized

of new products levelsClaessens and Djankov 2000 7 EE combined TFP, labor productivity growth dummy if 33% privateDjankov 1999a Georgia, Moldova sales, renovations growth dummy if privatizedDjankov 1999c Georgia, Moldova, labor productivity,

Kazakstan, Kyrgyz, renovations growth, % ownershipRussia, Ukraine levels

Earle 1998 Russia mid-year projection levels % ownershipof labor productivity

Earle and Estrin 1997 Russia qualitative restructuring levels % ownershipEarle and Estrin 1998 Russia mid-year projection of levels % ownership

labor productivityEarle and Rose 1997 Russia new goods production levels dummy if privatizedEarle and Sabirainova 1999 Russia wage arrears levels dummy if privatizedEarle, Estrin, Leshchenko 1996 Russia links to government, sales levels dummy if privatizedEarle and Telgedy 2001 Romania TFP levels % ownershipEstrin and Rosevear 1999a Ukraine profits, sales levels dummy if privatizedEstrin and Rosevear 1999b Ukraine qualitative restructuring levels dummy if privatizedEvans-Klock and Samorodov 1998 Kyrgyz excess employment levels dummy if privatizedFrydman, Gray, Hessel, Hungary, Poland, revenues, labor growth dummy if 33%

Rapaczynski 1999 Czech Republic productivity privateFrydman, Gray, Hessel, Hungary, Poland, default on debts levels dummy if 33%

Rapaczynski 2000 Czech Republic privateGlennerster 2000 Macedonia profits per worker growth dummy if privatizedGrigorian 2000 Lithuania TFP, labor productivity levels % ownershipGrosfeld and Nivet 1999 Poland output growth dummy if privatizedHendley, Murrell, Ryterman 2001 Russia success in transactions levels % ownershipJones 1998 Russia revenue, labor productivity growth dummy if 50% privateJones and Mygind 2002 Estonia TFP levels dummy if 50% privateKonings 1997 Hungary, Romania revenues levels % ownership

SloveniaLehmann, Wadsworth, Acquisti 1999 Russia wage arrears levels % ownershipLinz and Krueger 1998 Russia labor productivity growth dummy if privatizedMajor 1999 Hungary capital productivity levels ownership scalePerevalov, Gimadi, Dobrodey 2000 Russia revenues, labor growth, dummy if privatized

productivity levelsPohl, Anderson, Claessens, Bulgaria, Czech TFP growth dummy if privatized

Djankov 1997 Republic, Hungary, Poland, Romania, Slovenia, Slovakia

Roberts, Gorkov, Madigan 1999 Kyrgyz Republic sales, labor productivity, growth, dummy if privatizedqualitative restructuring levels

Smith, Cin, Vodopivec 1997 Slovenia TFP levels % ownershipWarzynski 2001b Ukraine productivity growth dummy if privatizedXu and Wang 1999 China return on equity and assets levels % ownershipZemplinerova, Lastovicka, Czech Republic restructuring index levels dummy if state

Marcincin 1995 majority owner

Article 1 8/12/02 9:13 AM Page 750

Page 13: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

appearing in column (1). Thus, the first con-clusion from this table is that the aggregateeffects of privatization are positive. This alsoapplies when both quantitative and qualita-tive indicators are examined separately. Theone case where the effects of privatizationare not significantly positive is for quantita-tive indicators for the CIS.

How robust are these conclusions? Thepapers that contribute analyses to our datavary a great deal, in the quality of analysis ingeneral, and the amount of attention paid toselection bias in particular. Therefore, duecaution suggests that we examine whetherthe above conclusions could reflect method-ological deficiencies in some of the surveyedpapers, rather than real economic phenom-ena.

A straightforward way to undertake such anexamination is to weight the various t-statis-tics, using weights reflecting methodologicaldifferences. Suppose that there are weights,w1, . . . , wM, for each t-statistic. Then the fol-lowing statistic has a normal distribution:

(3)

This weighting procedure discounts studieswith smaller weights. For example, if thereare particular concerns about selection bias,the weights would reflect attention paid tothis issue.

We have constructed two sets of weights.First, we rated papers on a scale of one tothree, reflecting the amount of attentionpaid to the problem of selection bias. Ascore of one indicates no attempt at coun-tering such bias, as in Djankov (1999a),while two suggests a less-than-complete attempt, for example, using fixed effectswithout any strong justification that thisapproach addresses the problem of bias(David Grigorian 2000). A three involves aconvincing attack on the problem, as inBrown and Earle (2000), which uses bothfixed effects and instrumental variables.Obviously the choice of scale is ad hoc, butit does fulfill our basic purpose, which is tocheck whether conclusions change notice-ably when discounting results derived fromweaker methodologies. This weightingplays a role analogous to the procedure ofdeemphasizing some papers in the stan-dard literature review. But it approaches

Djankov and Murrell: Enterprise Restructuring in Transition 751

TABLE 3EFFECTS OF PRIVATE VERSUS STATE OWNERSHIP ON ENTERPRISE RESTRUCTURING

Composite statistics derived from 93 analyses appearing in 37 studies

Weighting Method Used

(1) (2) (3)

attempts to handle overall rating

none selection bias of quality

normally distributed test statistics

All countries both 93 13.48 13.34 11.19EE both 37 15.71 15.49 15.18CIS both 53 3.95 2.84 0.07

All countries quantitative 64 11.78 11.07 9.54EE quantitative 34 15.37 15.14 14.92CIS quantitative 28 0.37 −2.28 −3.73

All countries qualitative 29 6.64 7.58 5.99EE qualitative 3 3.71 3.71 3.49CIS qualitative 25 5.35 6.76 5.06

ΣΜ

k=1Σ

Μ

k=1

wk wk2tk

Regions Type of Dependent Number of AnalysesVariable in the Contributing

Analyses to the CompositeTest Statistic

Article 1 8/12/02 9:13 AM Page 751

Page 14: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

the issue in a somewhat more transparentmanner, since the reader can directly observethe consequences of the deemphasis.

We also constructed a rating of the over-all quality of the papers. This rating reflectssystematic information collected on spe-cific issues such as selection bias, numberof reform years reflected in the data, thenature of Y, and appropriateness of X, aswell as our own subjective assessment ofthe strength of the analysis.17 It also re-flects the relative standing of the journal inwhich the paper is published, if it has beenpublished. The rating is on a scale of one toten. Again, the use of this rating is to exam-ine whether conclusions change noticeablyif the results from weaker methodologiesare discounted.

Comparisons across the columns of table3 suggest robustness in the conclusions forqualitative variables in both regions and forquantitative variables in EE. For example,the evidence in table 3 suggests quite clearlythat the overall conclusions for EE are notvitiated by the possibility of selection bias.Table 3 serves to blunt the perception heldby some that selection bias presents an in-surmountable problem in assessing the ef-fects of privatization.

There is one case where weighting byquality does change the aggregate result: thestatistic for quantitative variables in the CISchanges from nonsignificance to signifi-cantly negative. Given this sensitivity to theemphasis placed on different results andgiven the differences in the results for quali-tative and quantitative measures for the CIS,it is difficult to draw an overall conclusion onthe effects of privatization in this region. Ifthe results for the quantitative variables aredeemed most reliable, then our analysis sug-gests that privatization has not had positiveeffects.

3.2 Strength of Effect: Comparisons across Regions

Although it is tempting to do so, one can-not immediately conclude from table 3 thatthe effect of privatization on the quantitativevariables in EE is greater than the effect inthe CIS. Table 3 provides information onlyon the statistical significance of an effect rel-ative to zero. It is quite possible that an ef-fect can be weaker in statistical terms butstronger in economic terms. To compare thesize of the two economic effects, it is firstnecessary to identify a statistic that summa-rizes aggregate effect size for a heteroge-neous set of studies. In order to focus onsubstantive results, the more formal discus-sion of that statistic is relegated to an appen-dix. We concentrate on intuition below.

The t-statistic is not suitable because itsmagnitude reflects sample size, which hasnothing to do with the economic effect ofprivatization. γ̂ is not suitable because it isdependent on the units of measurement of Yand P, which vary greatly between studies.On the positive side, the t-statistic is invari-ant to units of measurement, while γ̂ doesnot depend on sample size. Thus, we seek ameasure that combines the positive proper-ties of the t-statistic and of the estimated pa-rameter, while eliminating their negativecharacteristics.

The standard procedure in meta-analysisis to use the partial correlation coefficient(Robert Rosenthal 1984). Partial correlationcoefficients are not usually published, butthey are easily calculated from regressionstatistics that appear in all papers.18 Sincecorrelation coefficients are unit-free, theyare comparable across a heterogeneous setof studies and therefore can be aggregated

752 Journal of Economic Literature, Vol. XL (September 2002)

17 An earlier version of this paper presented resultsfor many sets of weights, reflecting each methodologi-cal variable on which we collected systematic informa-tion. The reader can examine these results in Djankovand Murrell (2000).

18 William Greene (2000, pp. 233–36) proves this re-sult within the classical regression framework, to whicha majority of the studies reviewed conform. When ap-plying this result outside this framework, we simply ap-peal to analogy. This suggests that more effort shouldbe made in including information in published papersthat allows the estimates to be more readily comparedto estimates in other papers that examine the same eco-nomic phenomena.

Article 1 8/12/02 9:13 AM Page 752

Page 15: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

to form a summary statistic on effect size forthe whole set of studies.

The first two rows of column (1) of table 4show the simple unweighted means of theestimates of the partial correlations derivedfrom two groups of studies, those on EE andthose on the CIS. As expected, the EE esti-mate is larger. As the appendix shows, it isstraightforward to construct a normally dis-tributed test statistic that reflects the nullhypothesis that these two means are equal.This test statistic appears in the third row oftable 4, providing evidence that the privat-ization effect is significantly larger in EasternEurope than in the CIS.

One disadvantage of working with partialcorrelation coefficients is that their size isnot obviously interpretable in terms of realeconomic phenomena. However, a few factsmight make the numbers in table 4 moreeasily interpretable. For example, if a studyhad the same sample size and standard er-rors as the Frydman et al. result in table 1(second row, first column), then a partialcorrelation coefficient of 0.05 would corre-spond to a percentage growth of sales of pri-vatized enterprises that is 3.51 higher thanthat of state enterprises. Using the Brownand Earle (2000) result (table 1, first row,last column) in a similar fashion, a partialcorrelation coefficient of −0.05 would corre-spond to privatized enterprises being 38percent less efficient than state enterprises.

Weighted means of the partial correlationcoefficients can also be calculated and a nor-mally distributed test statistic can be derivedto examine the significance of the differencebetween the two weighted means. Columns(2) and (3) of table 4 present the results forthe aggregate partial correlation coefficientscalculated using the weights that were de-scribed in the context of table 3. Thus, dif-ferences between the columns of the tablereflect the sensitivity of the overall results tovarying judgments on the importance ofmethodological quality.

When examining the test statistics forqualitative and quantitative variables to-

gether or when the quantitative variables areexamined alone, the test statistics affirm (atthe 1-percent significance level) that the pri-vatization effect is stronger in EE than in theCIS. For the quantitative variables, the useof the weights adjusting for methodologicalquality actually strengthens this conclusion,suggesting that selection bias, or othermethodological deficiencies, do not play anyrole in leading us to this conclusion. The re-sults are similar, but weaker statistically, forthe qualitative variables, perhaps becausethere are so few studies for EE that fit intothis category. In sum, there is strong evi-dence that the effect of privatization inEastern Europe has been much greater inmagnitude than in the CIS.

4. The Effects of Different Types of Owners

One reason that changes of ownershipmight have had varying effects across re-gions is that different privatization processesled to different mixes of owners. In theabsence of quick re-trading of shares(Anderson, Korsun, and Murrell 1999;Joseph Blasi and Andrei Shleifer 1996), theowners created by the privatization processhave had more than a temporary effect onenterprise behavior. This is important, ofcourse, only if ownership makes a differencein restructuring. As it happens, transition ex-perience offers unusually good evidence onthis score.

The pre-transition literature offers a vari-ety of disparate empirical results. There hasbeen general support for Adolf Berle andGardiner Means’ (1933) contention that dif-fuse ownership yields power to managerswith less interest than shareholders in re-structuring (Bruce Johnson et al. 1985;Richard Roll 1986). Randall Morck, AndreiShleifer, and Robert Vishny (1988) found thatsome managerial ownership can amelioratesuch problems, while Shleifer and Vishny(1986) argue that block-holders have an in-centive to monitor management. However,block-holders also have opportunities to

Djankov and Murrell: Enterprise Restructuring in Transition 753

Article 1 8/12/02 9:13 AM Page 753

Page 16: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

expropriate value from minority sharehold-ers. Quite separately, the comparative sys-tems literature pursued the question ofwhether the pathologies predicted by theneoclassical model of the labor-managed firmoutweigh the motivational effects of workerownership. The evidence is mixed (AlanBlinder 1990; John Bonin, Derek Jones, andLouis Putterman 1993; Douglas Kruse andJoseph Blasi 1995).

The existing literature on marketeconomies does not provide a clear pictureof which type of owner is most advantageousfor restructuring. Ronald Coase (1988) andHarold Demsetz and Kenneth Lehn (1985)suggest a reason for the lack of clarity. Theyargue that the evidence from mature marketeconomies might be spurious. If the transac-tion costs of taking value-maximizing posi-tions in firms are low, each firm would havethe “right” ownership structure: there is norelationship between ownership type and re-structuring.

Therefore, evidence from transition coun-tries has a vital contribution to make. Intransition economies, ownership was largelydetermined through political and adminis-trative processes rather than endogenouslyin markets with low transaction costs. Inmany cases, ownership structure is exoge-nous. In others, it is relatively easy to obtainreliable instruments to counter endogeneitybias. The evidence from transition shouldoffer much better information on the effectsof different owners than has been generatedpreviously.19

The results for transition countries areunique in another respect. Many differenttypes of owners were created. At the time ofprivatization, government agencies andfirms knew the structure of ownership andcould impart this information to researchers.

754 Journal of Economic Literature, Vol. XL (September 2002)

TABLE 4COMPARING THE SIZE OF PRIVATIZATION EFFECTS ACROSS REGIONS

Partial correlation coefficients and test statistics derived from 90 analyses from 36 studies

Weighting Method Used

(1) (2) (3)

attempts to handle overall rating

none selection bias of quality

partial correlation coefficients and test statistics on their differences

EE both 37 0.083 0.073 0.075CIS both 53 0.038 0.041 0.016Test statistic for both 3.89 2.74 4.82difference

EE quantitative 34 0.078 0.070 0.073CIS quantitative 28 0.016 −0.001 −0.019Test statistic for quantitative 4.14 4.63 5.41difference

EE qualitative 3 0.146 0.146 0.145CIS qualitative 25 0.063 0.089 0.064Test statistic for qualitative 1.99 1.36 1.84difference

19 For example, Megginson and Netter’s (2001) com-prehensive survey of the effects of privatization in alltypes of economies does not reflect on the relative ef-fectiveness of different types of private owners.

Regions Type of Dependent Number of AnalysesVariable in the Contributing

Analyses to the CompositeStatistic

Article 1 8/12/02 9:13 AM Page 754

Page 17: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

The resultant ownership information was ata level of detail that could not be obtained inmature market economies. Hence, there is alarge set of studies examining the effects ofdifferent owners. By aggregating the results,we are able to compare the effects of a com-prehensive set of owners, something notpreviously accomplished in the literature.

We begin by examining the characteristicsof the pertinent papers. The methodologiesused by the papers are essentially the sameas those discussed in section 3. In fact, manypapers begin with an analysis of state versusprivate and then turn to a disaggregation ofprivate ownership, e.g., Brown and Earle(2000), Earle (1998), Frydman et. al. (1999),and Jones (1998). Such papers use the samemethodology and independent variables inboth parts of their analyses, simply making Pin (1) a vector that captures the ownershipheld by different entities.

A further focus on Frydman et al. (1999)and Anderson, Lee, and Murrell (2000) ex-hibits the methodological decisions to bemade when choosing the elements of P andthe variations in results that can be obtained.Frydman et al. examine outsiders, insiders,and the state in one analysis, and foreigners,domestic financial firms, domestic nonfinan-cial firms, domestic individuals, the state (in aprivatized firm), the state (in a nonprivatizedfirm), managers, and workers in anotheranalysis. This is an unusually comprehensivelist of owners to appear in one paper, and mostpapers use a much smaller set, e.g., Smith,Cin, and Vodopivec (1997) using foreigners,insiders, and the state, and Claessens andDjankov (1999a) using managers and out-siders. The ownership variables in Frydman etal. are dummies capturing whether the givenowner is the largest shareholder, a commonapproach, e.g., Jones and Mygind (2000,2002). Measurement of ownership in this wayreflects a somewhat unusual feature of theFrydman et al. data, the highly concentratedownership in all privatized firms.

In Mongolia, examined by Anderson, Lee,and Murrell (2000), the variety of owners

after privatization is narrower and the de-gree of ownership concentration less.Managerial and worker ownership are highlycorrelated due to the nature of the privatiza-tion scheme and there is no value in differ-entiating between them. Individuals withsmall stakes are dominant among outsiderowners. Hence, analysis of the effects of dif-ferent owners in Anderson et al. only con-trasts the state (in a privatized firm) versusoutsiders versus insiders. In contrast toFrydman et al.’s dummy variables, owner-ship is measured by the percentage of sharesheld by each type of owner, perhaps themost usual approach, e.g., Brown and Earle(2000), Claessens and Djankov (1999a,b),Earle and Estrin (1997), and Smith, Cin,and Vodopivec (1997).

Frydman et al. find that foreign and do-mestic financial owners produce the largestpositive effects, while outsider owners out-perform insiders, results echoed later in oursynthesis of many papers. Their results aremixed when comparing insider ownership tostate ownership in enterprises with 100-percent state ownership. Anderson et al.find that outsiders and insiders perform lesseffectively than the state in privatized enter-prises, while there is no significant differ-ence between insiders and outsiders.

What explains the differences in thesetwo sets of results? Selection bias or omit-ted variables seem unlikely candidates,given the attention paid by both studies tothese issues. Rather, the outsider ownersare very different in Central Europe, wherethey are block-holders, than in CentralAsia, where they are individuals, usuallywith tiny amounts of ownership. Andersonet al. examine whether diffusion of owner-ship can explain their results and do findbetter performance in those few enter-prises in which block-holders have ob-tained seats on the boards of directors.20

The positive relation between ownership

Djankov and Murrell: Enterprise Restructuring in Transition 755

20 However, because of lack of instruments, they arenot able to control for selection bias on this variable.

Article 1 8/12/02 9:13 AM Page 755

Page 18: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

concentration and performance is a directfocus of Claessens and Djankov’s (1999b)study on the Czech Republic.

In reconciling the findings of Frydman etal. and Anderson et al., the results on stateownership warrant particularly close exami-nation. Anderson et al. were able to sampleonly privatized enterprises (including oneswith large state ownership). Thus their studyincludes no firms that remained under tradi-tional state ownership. Frydman et al. sam-pled both privatized and nonprivatized firms,thus having two state ownership variables.The Anderson et al. results on state owner-ship are comparable only to those ofFrydman et al. on state ownership in privat-ized firms. When this is understood, the twostudies’ results appear much more consis-tent, since Frydman et al. find that stateownership in a privatized firm performs atleast as well as the median type of privateowner.21 However, the two studies providevery different interpretations of their results.Frydman et al. view the state as passive, withprivate owners dominating the decisions ofpartially state-owned privatized firms.Anderson et al. view the Mongolian state asactive and enormously pressured by eco-nomic necessity. The government was morewilling than insiders to push for efficiencyand more able to do so than outsiders.

In providing a reconciliation of the resultsof the two studies, the above paragraphsprovide an important warning to the harriedreader trying to absorb a large number of re-sults very quickly. Details matter crucially,for example which study used the privatiza-tion agency to obtain its sample or whichstudy’s data reflect concentrated ownership.Choice of estimation method can even influ-ence the interpretation of which owners arereflected in results, fixed effects estimatesreflecting only ownership that changes overthe sample period. These facts are often notemphasized in papers.

4.1 Quantifying the Differences between Owners

We now turn to the quantitative synthesisof results. The objective of this synthesis isto find the relative sizes of the effects of dif-ferent owners and to test which owners havesignificantly different effects from others.We follow the literature in identifying theownership categories that we analyze, iden-tifying those ownership types that recur con-sistently in a number of studies.22 This re-sulted in eleven categories, some of whichoverlap:23

1) traditional state ownership: enterprisesthat are 100-percent state-owned and thathave not been part of a privatization program;

2) the state in commercialized (or corpo-ratized) enterprises: state ownership in en-terprises that have been legally separatedfrom the state bureaucracy, that are treatedas private under the corporate governancelaws, and, usually, that have been part of aprivatization program;24

756 Journal of Economic Literature, Vol. XL (September 2002)

21 As is clearly shown when we aggregate results be-low, this result is echoed in many other studies.

22 Since there is no source that provides a definitionof a standard set of ownership categories, there are in-evitable variations across the empirical papers in howdifferent owners are defined. Where papers used own-ership types that differed too much from the types thatappeared in other papers, we did not use the results.This happened in only a small number of cases.

23 For example, workers and managers together areinsiders. One should not assume that an aggregate cat-egory necessarily produces the same result as the sumof its parts. For example, estimates of the effects ofmanagers and workers usually appear in different stud-ies than those for insiders as a whole. Which owners areincluded in a study, however, is not random. Re-searchers are more likely to collect data on a specificowner when the privatization process gives scope forthat owner to be active in privatization procedures.Hence, the results for insiders probably reflect some-what different circumstances than those for managersand workers on their own. We should not expect to seethe insider effect simply equal to the weighted sum ofthe worker and managerial effect.

24 In most countries commercialization (or corpora-tization) occurred as preparation for privatization.Therefore, our data on this ownership type is domi-nated by results for enterprises that have less than 100-percent state ownership. Where a study did not provideenough information for us to know which of the twocategories of state ownership applied or where thestudy mixed the two types, we assigned the results tothe traditional state ownership category.

Article 1 8/12/02 9:13 AM Page 756

Page 19: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

3) enterprise insiders: a composite group,including both workers and managers;

4) outsiders: a composite group consistingof all nonemployee, nonstate owners;

5) workers (nonmanagerial employees);6) managers (managerial employees);7) foreign owners of all types;8) banks, except those included in 7;9) investment funds, except if ownership

of the fund is such that 2, 7, or 8 applies;10) block-holders: outsider ownership

concentrated in the hands of large individualowners, except where the block-holdercould be classified in 2, 6, 7, 8, or 9;

11) diffuse outsider: the residual outsiderownership category, when outsider ownersare not identified as belonging to 7, 8, 9, or10; a category dominated by individual out-sider owners, each of whom owns a minis-cule portion of any specific enterprise.25

All papers that we review examine a sub-set of these eleven types of owners. For ex-ample, a paper might focus on the state ver-sus insiders and outsiders and estimate:

Y = α + Xβ + δO+ θI + ε (4)

O and I are measures of the amount ofownership held by outsiders and insiders,state ownership is the omitted variable, δand θ are the parameters of interest, and allother variables are as defined before. Theinformation of prime interest is the compar-ison of δ to θ (outsiders versus insiders) andeach of these to 0 (insiders or outsiders ver-sus the state). We examine both statisticalsignificance and the magnitude of the own-ership effect.

The methods used here are extensions ofthose already introduced in section 3.However, a number of complications arise indeveloping the results. First, important in-formation is often missing from papers (rele-vant to the statistical precision of the com-parison between δ and θ). Therefore, it isnecessary to investigate how to fill this gap.

Second, there is no longer a binary compari-son between state and private, but a multi-lateral comparison between many types ofowners. Therefore, a dummy-variable re-gression framework is required to combineresults, rather than simply using differencesbetween means. Third, to take advantage ofinformation on the precision of estimatesfrom different papers, it is necessary to usegeneralized least squares. Fourth, theweighting of results according to method-ological quality also requires generalizedleast squares. A full consideration of thesecomplications entails an extended discussionof methodology, in which the average readeris probably not interested. The discussion istherefore relegated to an appendix.

From 24 studies on twenty countries, wehave compiled all usable information on theeffects of different types of owners on quan-titative enterprise outcomes. The pertinentpapers, together with the countries andowners in each study, are listed in table 5.

From these studies, we derive the t-statis-tics and then the partial correlation coeffi-cients corresponding to ownership effects.The data are then combined in a generalizedleast squares framework to obtain a compos-ite estimate of the effect of each of the newowners (2–11 above) relative to traditionalstate ownership. These estimates of owner-ship effects are in the form of partial correla-tions, as in section 3.

The estimates of the partial correlationsappear in figure 1. These estimates reflect aprocess that weights more heavily thosestudies that have paid more attention to theissue of selection bias.26 We focus on thismethodological issue because readers ofprevious versions of this paper have indi-cated that this is of most concern.27 By com-paring the results that are obtained when all

Djankov and Murrell: Enterprise Restructuring in Transition 757

25 This category is used only when the study differ-entiates between various types of outsiders. When alloutsiders are treated as one, owner group 4 applies.

26 The weighting scale is the same as in the previoussection.

27 Suspicions about selection bias arise naturally.Perhaps the state kept the best enterprises during pri-vatization, or managers fought harder to retain controlwhen prospects were good, or foreigners were willingto pay for efficient enterprises only.

Article 1 8/12/02 9:13 AM Page 757

Page 20: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

papers are treated equally (see appendix) tothose in figure 1, one can examine whetherthe results seem particularly sensitive to se-lection bias. Three conclusions follow fromthis comparison. First, the relative positionof most owners remains the same whateverthe treatment of selection bias. Second, se-lection bias reduces the estimated magni-tudes of most ownership effects. Third, onlythe results for managers and workers show a

considerable degree of sensitivity to how se-lection bias is handled. Thus, our assessmentof the results is that a high degree of confi-dence can be placed in the relative rankingthat appears in figure 1, except for someresidual doubt about the positions of man-agers and workers, especially the former.

Figure 1 suggests that differences be-tween owners are of great economic impor-tance. Privatization to workers is detrimental;

758 Journal of Economic Literature, Vol. XL (September 2002)

TABLE 5STUDIES OF PRIVATIZATION BY OWNERSHIP TYPE

Paper Countries Ownership Types

Anderson, Lee, Murrell 2000 Mongolia 2, 3, and 4Brown and Earle 1999 Russia 2, 3, 4, 7, and 11Claessens and Djankov 1999a Czech Republic 4 and 6Claessens and Djankov 1999b Czech Republic 2 and 7–10Claessens, Djankov, Pohl 1997 Czech Republic 2 and 7–9Cull, Matesova, Shirley 2002 Czech Republic 1 and 7–9Djankov 1999b Georgia and Moldova 1, 6, and 11Djankov 1999c Six CIS states 2, 5–7, 10, and 11Earle 1998 Russia 1, 4–11Earle and Estrin 1997 Russia 1, 4–11Earle and Telgedy 2001 Romania 2, 3, 10, and 11Estrin and Rosevear 1999a Ukraine 1–6Frydman, Gray, Hessel, Rapaczynski 1999 Central Europe 1–10Grosfeld and Nivet 1999 Poland 1 and 2Jones 1998 Russia 1, 4–6, and 8Jones, Klinedinst, Rock 1998 Bulgaria 1 and 2Jones and Mygind 2000 Baltics 1, 5–7, and 10Jones and Mygind 2002 Estonia 2, 5–7, and 10Konings 1997 Slovenia, Hungary, Romania 1 and 5Lee 1999 China 1 and 2Lehmann, Wadsworth, Acquisti 1999 Russia 1 and 7Roberts, Gorkov, Madigan 1999 Kyrgyz Republic 2–4Smith, Cin, Vodopivec 1997 Slovenia 1, 3, and 7Weiss and Nikitin 1998 Czech Republic 2 and 8–10

Ownership types are coded as follows (see text for details):1) traditional state ownership2) the state in commercialized (or corporatized) enterprises3) enterprise insiders4) outsiders5) workers (nonmanagerial employees)6) managers (managerial employees)7) foreign owners of all types8) banks9) investment funds

10) block-holders11) diffuse outsider

Article 1 8/12/02 9:13 AM Page 758

Page 21: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

privatization to diffuse individual owners hasno effect, and privatization to investmentfunds or to foreigners has a large positive ef-fect. Loosely speaking, privatization to fundsis five times as productive as privatization toinsiders, while privatization to foreigners orblock-holders is three times as productive asprivatization to insiders. Foreigners were ex-pected to make productive changes, but it isnotable that investment funds are signifi-cantly better than foreigners, and that banksand block-holders are close to foreigners.Similarly, diffuse individual ownership wasnot expected to be very effective, but it issurprising that it is indistinguishable fromtraditional state ownership.

A notable result is the difference betweentraditional state ownership and state owner-ship in commercialized enterprises, a resultthat recurs from Central Europe to China(Claessens, Djankov, and Pohl 1997; Lee1999). Remember, of course, that this resultis not for economies in which ownership has

been developed organically for decades, butrather where ownership has been artificiallytransferred, sometimes to private ownerswho are creatures of the state. Then, if cor-porate governance laws are weak, share re-trading is sluggish, and the state is focusedon solving economic problems, state owner-ship could easily be superior to some typesof ownership (Anderson, Lee, and Murrell2000). The superiority of state ownership incommercialized enterprises over traditionalstate ownership might arise because thepart-owners who are private are playing animportant role in enterprise affairs(Frydman et al. 1999) or because the veryact of commercialization changes the incen-tives facing the state when it intervenes intoenterprise affairs (Shleifer and Vishny 1994).

A conclusion implicit in the results of fig-ure 1 is that concentrated shareholding pro-duces larger effects than diffuse sharehold-ing. This is seen most clearly in thedifference between the effects of individual

Djankov and Murrell: Enterprise Restructuring in Transition 759

0.100

Category of Owners

0.004

Part

ial C

orre

latio

n C

oeff

icie

nts

Figure 1. How Ownership Affects Firm Performance after Privatization.Estimated Effects of Changing from Traditional State Ownership to Different Private Owners

0.020

0.040

0.060

0.080

0.000

–0.020

–0.040

foreign

insiders

0.025

–0.018

0.043

banks

0.011

0.049 0.048

0.017

0.035

0.079

workers

diffuse individual

outsiders

commercialized state

managers

block-holders

investment funds

Article 1 8/12/02 9:13 AM Page 759

Page 22: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

owners and those of block-holders, but it isalso implicit in the effects of foreigners,funds, and banks since these entities usuallyconcentrate their shareholdings. Claessensand Djankov (1999b) focus directly on theeffects of concentration in their study of theCzech Republic. They show that a 10-percentincrease in the percentage of shares held bythe largest five shareholders will increase la-bor productivity by 5 percent. They also finddiminishing returns to concentration. Theseresults are echoed in studies of Russia, e.g.,Earle and Estrin (1997) and Brown andEarle (2000).

Table 6 examines statistical significance ofthe differences in figure 1, reporting the val-ues of t-statistics for standard tests of nullhypotheses that one owner has the same ef-fect as another, for all pair-wise combina-tions of owners.28 A first inspection of table

6 immediately reveals an unusually largenumber of significant t-statistics.29 This ispartially a consequence of employing amethodology that combines results and em-bodies the precision of estimates from manystudies. Just as in the previous section, wehave been able to generate an unusualamount of statistical power. However, it isnot the case that all owners have signifi-cantly different effects from all others. Forexample, workers, diffuse individual owners,and traditional state ownership cluster at thebottom, with effects that do not differ signif-icantly. The effects of banks are significantlydifferent from the effects of less than halfthe other owners, partially reflecting the factthat banks are included in few studies.

4.2 Comparing Owners across Regions

We have found that privatization hasstronger effects in Eastern Europe than in

760 Journal of Economic Literature, Vol. XL (September 2002)

28 These test statistics are easily calculated usingstandard generalized least squares procedures and im-plicitly use the information on the variances of esti-mates that is extracted from the original studies.

29 The appendix suggests using a conservativebenchmark for significance, but this statement holdseven under this standard.

TABLE 6TESTING DIFFERENCES BETWEEN THE EFFECTS OF A VARIETY OF OWNERS

Statistics derived from 341 pair-wise comparisons of owners taken from 24 studies

Traditional DiffuseWorkers State Individual Managers

t-statistic for {(effect of owner listed on row) minus (effect of owner listed on column)}

Traditional State 1.36Diffuse Individual 1.78 0.50Managers 1.93 0.90 0.63Insiders 2.77 2.61 3.12 0.45Outsiders 3.20 3.45 3.54 1.09Commercialized State 4.12 5.69 7.21 1.99Banks 3.53 3.42 3.04 1.93Foreign 4.82 7.82 7.67 2.98Block-holders 4.90 7.73 7.74 3.09Investment Funds 5.55 7.05 5.98 4.28

Notes: To compare owners A and B: If the cell located at the intersection of A’s row and B’s column is blank, thenB is more productive than A. If the cell corresponding to A’s column and B’s row is nonempty, then the number inthat cell is the t-statistic for a test of the null hypothesis that B’s effect minus A’s effect is equal to zero.

Article 1 8/12/02 9:13 AM Page 760

Page 23: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

the CIS and that different types of ownershave different effects. This immediatelyraises the question of whether the lattercould explain the former. This questionwould be best addressed using data on thedistribution of ownership in different coun-tries, but there is no systematic collection ofsuch data. The available evidence does sug-gest that worker and diffuse individual own-ership are more prevalent in the CIS than inEE, while foreign, investment fund, concen-trated individual, and bank ownership areless prevalent. (See, for example, Anderson,Lee, and Murrell 2000; Frydman et al. 1999;Brown and Earle 2000; and Claessens andDjankov 1999b.) Thus, since the CIS owner-ship portfolio contains a greater share of lesseffective owners, structure of ownership is astrong candidate to explain regional differ-ences in the effects of privatization.

The effects of different types of ownerscould also vary between regions. To investi-gate this, the natural first step is to seewhether the vectors of estimated ownershipeffects differ significantly between regions.

The data reject the null hypothesis that own-ership effects in the CIS are all the same asthose in EE, at the 1-percent significancelevel. The next step then is to redo figure 1,for each region separately.

In all cases except one (workers), the own-ership effects for the CIS in figure 2 aregreater than for EE.30 How is this consistentwith section 3, which showed that privateownership had stronger effects in EE than inthe CIS? There are two obvious reasons. First,the effect of workers is strongly negative andthey own a much larger share of privatized en-terprises in the CIS than elsewhere. (InFrydman et al.’s 1999 Central European sam-ple, workers are the dominant owner in 9 per-cent of firms, while they are dominant in 50percent of Earle and Estrin’s 1997 Russiansample.) Second, commercialized state own-

Djankov and Murrell: Enterprise Restructuring in Transition 761

30 It is worth emphasizing that these results do notshow that the various owners are more productive inthe CIS than Eastern Europe in absolute terms.Rather, these results could just as easily be due to therelative unproductiveness of traditional state enter-prises in the CIS, which are the basis for comparison.

TABLE 6 (Cont.)

Statistics derived from 341 pair-wise comparisons of owners taken from 24 studies

CommercializedInsiders Outsiders State Banks Foreign Block-holders

t-statistic for {(effect of owner listed on row) minus (effect of owner listed on column)}

1.675.28 2.282.09 1.39 0.636.66 4.18 3.44 0.446.71 4.36 3.71 0.56 0.385.30 4.60 3.97 2.47 2.96 2.84

Article 1 8/12/02 9:13 AM Page 761

Page 24: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

ership is separated from traditional state own-ership in figure 1 but not in table 4.Commercialized state ownership has aboveaverage effects (relative to traditional stateownership) in the CIS and below average inEE. Thus, if one conceives of the privatizationprocess as two steps, first the move away fromtraditional state ownership and then the moveaway from state ownership of all forms, figure2 and table 1 together suggest that the firststep alone was a important contribution of theprivatization process in the CIS, while this wasnot so in Eastern Europe.31

While we have treated the two regions asindividually homogenous in deriving figure 2,it should be noted that there are also largedifferences between the estimates for differ-ent countries within the same region. For ex-ample, Romania seems to be an outlier inEastern Europe. John Earle and AlmosTelgedy (2001) find that diffuse individualand insider ownership are surprisingly pro-ductive in Romania. Indeed, the studies forthe CIS are somewhat more homogenousthan those outside the CIS, as evidenced bythe R-squares for the regressions that providethe data for figure 2: 0.73 in the CIS and 0.31outside that region. This is perhaps a reflec-tion of the greater diversity, in history and inreform programs, within EE. Hence, figure 2is hardly a last answer, but rather a beginning,in a search for the reasons why different own-ers are more effective in some settings thanothers. In section 8, we provide some discus-

762 Journal of Economic Literature, Vol. XL (September 2002)

0.200

Category of OwnersEE

0.028

0.063

Part

ial C

orre

latio

n C

oeff

icie

nts

Figure 2. Regional Variations in the Effects of Different Types of Owners.Comparing Ownership Effects in Eastern Europe to Those in the CIS

0.000

0.050

0.100

0.150

–0.050

–0.100

foreign

0.017

0.0790.087

0.000

–0.053

0.005

0.162

0.011

0.102

–0.003

0.069

CIS

0.028

0.120

0.068

0.126

0.040

0.115

insiders

banksworkers

diffuse individual

outsiders

commercialized state

managers

block-holders

investment funds

0.094

31 Lest some readers find this statement confusing, itshould be emphasized that it was sometimes possiblefor an enterprise to pass through a privatization pro-gram and still end up with 100-percent state owner-ship! Going through a privatization program and reduc-tion of state ownership were not synonymous, althoughhighly correlated.

Article 1 8/12/02 9:13 AM Page 762

Page 25: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

sion of this issue, examining an institutionalinterpretation of the patterns in figure 2.

5. Product Market Competition

There is a substantial theoretical literatureanalyzing the relationship between competi-tion and enterprise efficiency. The generalhypothesis is that increased competitionstimulates improvements in productivity.Two lines of argument support this hypothe-sis. The first is derived from the literature onX-inefficiency, while the second centers onindustry rationalization. The X-inefficiencyliterature argues that managerial effort is un-der-supplied in the absence of vigorous com-petition. Henrik Horn, Harold Lang, andStefan Lundgren (1995) show that greatercompetition induces an expansion of outputby incumbent firms through improved inter-nal technical efficiency without any realloca-tion of resources across firms. Earlier studies(Bengt Holmström 1982; Barry Nalebuff andJoseph Stiglitz 1983) argue that incentiveschemes for managers will generate betterresults the greater the number of players(firms) involved because of greater opportu-nities for performance comparison.

A second line of argument is that in-creased competition may lead to a rationali-zation of oligopolistic industries as firms areforced to compete for market share (KlausSchmidt 1997). Resource reallocation oc-curs across firms, within and between sec-tors. Although the shake-out may result in atransitional decline in productivity as firmswith increasing returns to scale lose domes-tic market share, over time this may be re-versed as market share expands due to exitof competitors and access to export markets.Much depends on the existence of scaleeconomies and the ease of entry and exit.

In transition economies, increased compe-tition may have negative effects on efficiency.Competition creates incentives for breakingcontracts when institutions are weak (OlivierBlanchard and Michael Kremer 1997). BarryIckes, Randi Ryterman, and Stoyan Tenev(1995) suggest that excessive competition,

especially from abroad, destroys networkcapital and harms enterprise performance inthe early transition period. The formerSoviet Union experienced significant lossesin markets with the disintegration of thetrade relations between Russia and the otherformer republics (Simeon Djankov andCaroline Freund 2002). As traditional mar-kets collapsed and as countries were openedup to foreign competition, many firms foundit difficult to restructure their product linesand at the same time reorient their sales to-wards new markets.

The empirical literature outside of transi-tion economies yields a mixed picture on therelationship between competition and enter-prise efficiency. Few empirical investigationsusing firm-level data have established astrong link between greater competition andsubsequent changes in enterprise perfor-mance.32 Two studies of British manufactur-ing firms (Stephen Nickell, Sushil Wadhwani,and Martin Wall 1992; Stephen Nickell 1996)use a panel to show that market concentra-tion has an adverse effect on total factor pro-ductivity. In contrast, David Blanchflowerand Stephen Machin (1996) find no effectof changes in domestic market structure onthe productivity of British plants. SilkeJanuszewski, Jens Koke, and Joakim Winter(2001) use a sample of publicly listed Germancompanies to show that neither import com-petition nor domestic market competition hasbeneficial effects on TFP growth. Studies ondeveloping countries also find ambiguous re-sults. Ann Harrison (1994) finds that a reduc-tion of tariffs in Cote d’Ivoire and the subse-quent increase in import penetration had aninsignificant effect on TFP growth. In con-trast, Sweder van Wijnbergen and AnthonyVenables (1993) find a strong positive effectof increased import penetration on labor pro-ductivity in Mexican manufacturing firms.

Djankov and Murrell: Enterprise Restructuring in Transition 763

32 On the macro level, Francisco Rodriguez andDani Rodrik (1999) survey the literature on open tradepolicies and economic growth and conclude that theevidence that openness is associated with growth isvery weak.

Article 1 8/12/02 9:13 AM Page 763

Page 26: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

The initial period of transition provides aunique opportunity to test the importance ofproduct market competition. A majority oftransition economies rapidly liberalizedtheir trade regimes. Many went on to demo-nopolize their industrial sectors by breakingup conglomerates, spinning off individualproduction units, and allowing entry of newprivate firms (Simeon Djankov and BernardHoekman 2000; Lizal, Singer, and Svejnar2001). The short period in which thesechanges took place allows the researcher toidentify the timing of the policy change andto control for other economy-wide and firm-specific factors that influence the change inproductivity.

We find 23 studies that explicitly investi-gate the effect of product market competi-tion on enterprise restructuring. Amongthose, thirteen focus on EE countries(Claessens and Djankov 2000; RumenDobrinsky et al. 2001; Djankov andHoekman 1998, 2000; Grigorian 2000; IrenaGrosfeld and Thierry Tressel 2001; PhilipHersch, David Kemme, and SharadBhandari 1994; Laszlo Halpern and GaborKörösi 1998; Könings 1997, 1998; NikolayMarkov et al. 2000; Mary Shirley and LixinColin Xu 2001; Frederic Warzynski 2001a)and ten use data for Russia (ManuellaAngelucci et al. 2001; Annette Brown andDavid Brown 1999; Earle and Estrin 1998;Brown and Earle 2000; Boris Kuznetsov etal. 2002; Yuri Perevalov et al. 2000), Georgia(Vladimir-Goran Kreacic 1998), Ukraine(Warzynski 2001b), Mongolia (Anderson,Lee, and Murrell 2000), or all transitioneconomies (Carlin et al. 2001). We are ableto distinguish 82 separate analyses, wherethe authors use either different measures ofenterprise restructuring (total factor produc-tivity levels or growth rate, labor productivitylevel or growth, sales growth, and qualitativevariables like renovation of facilities) or dif-ferent indicators of competitive pressures.Twenty-eight analyses use import competi-tion as the main explanatory variable, while52 focus on domestic market structure.

The analyses are fairly homogeneous. Inmost cases, the dependent variable is quanti-tative: 55 analyses use TFP as the indicator ofenterprise restructuring, nineteen use laborproductivity, and two use sales growth. Sixanalyses, all derived from Kreacic (1998), usequalitative indicators of restructuring (facili-ties renovation, establishment of a marketingdepartment, and computerization of the ac-counting function). Import competition isproxied by the import penetration ratio(twenty analyses), or the industry-level tariffrate (eight analyses). Domestic market com-petition is measured by either the Herfindahlindex (25 analyses), the percentage of salesrevenues of the top two, three, or four firmsin the respective industry (fifteen analyses),the number of local competitors that enter-prise managers perceive as rivals (twelveanalyses), or the change in the enterprise’sown market share (two analyses).

Some explanatory variables raise the issueof endogeneity. In proxying for the level ofdomestic competition, Hersch, Kemme, andBhandari (1994) asked enterprise managersto “categorize the number of other firms nowproducing in your main domestic market aseither none, 1 to 10, 11 to 50, 51 to 100, ormore than 100” (p. 358). Dobrinsky, Dochev,and Markov (2001) use lagged own marketshare as one of their proxies. Both variablesare problematic since the number of com-petitors or one’s market share may be en-dogenous to the performance of the enter-prise. Accordingly, when we aggregate theempirical analyses using weights, we assign aweight of 1 to both studies for “attempts tohandle endogeneity.” Studies using theHerfindahl index, e.g., Könings (1998), avoidthis problem and are given a rating of three.

The endogeneity problem is also presentin studies that use manager perceptions ofimport competition or the import penetra-tion ratio as a measure of competition fromabroad. The former suffers from obvious en-dogeneity, while the second depends on thevolume of imports, which is endogenous tothe performance of the producers in the

764 Journal of Economic Literature, Vol. XL (September 2002)

Article 1 8/12/02 9:13 AM Page 764

Page 27: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

domestic sector. Accordingly, we assign avalue of 1 for handling endogeneity to stud-ies that use perception of import competi-tion, e.g., Konings (1997); a 2 to all analysesthat use import penetration, e.g., Angelucciet al. (2001); and a 3 for all analyses that usechanges in the statutory tariff rate, e.g.,Djankov and Hoekman (2000). The endo-geneity ratings are also reflected in the over-all quality rating.

Table 7 illustrates the variety of analysesin terms of country samples, the choice ofdependent and explanatory variables, andthe estimated economic effects of competi-tion on enterprise efficiency. For example,Angelucci et al. (2001) study the effect ofcompetitive pressure on firm performancein Romania, using panel data from 1997 to1998.33 The study finds that a 10-percent-

age-points reduction in the Herfindahl indexresults in an increase in TFP growth of 19.1percent; a 10-percentage-points increase inimport penetration results in a TFP growthloss of 6 percent. The negative effect ofcompetition is particularly pronounced forstate-owned enterprises, and is insignificantfor privatized firms. The authors concludethat the technology gap between domesticfirms and importing rivals may have beentoo wide and thus acted to discourage enter-prise restructuring. In effect, domestic(state-owned) firms are priced out of theirmarkets.

Grosfeld and Tressel (2001), using a panelof 153 Polish firms, find that the reductionin own market share by 10 percentage pointsresults in 1.4-percentage-points increase inproductivity growth. Again, this effect isstronger for privatized firms. The results arerobust to the use of alternative estimationtechniques, and the inclusion of controls for

Djankov and Murrell: Enterprise Restructuring in Transition 765

TABLE 7GAINS OR LOSSES DUE TO PRODUCT MARKET COMPETITION: A SAMPLING OF ESTIMATES

Study Country Dependent Variable Explanatory Variable Economic Effect1

Angelucci et al. 2001 Romania Log TFP level Herfindahl level 19.1%Perevalov et al. 2000 Russia Log TFP level Number of competitors 2.4%Grosfeld and Tressel 2001 Poland TFP growth Own market share 1.4 ppDjankov and Hoekman 2000 Bulgaria TFP growth Herfindahl level 1.4 ppWarzynski 2001b Ukraine TFP growth Number of competitors −3.8 ppBrown and Earle 2000 Russia Log TFP level Herfindahl level 1.4%Konings 1998 Bulgaria TFP growth Herfindahl level 4.4 ppKonings 1998 Estonia TFP growth Herfindahl level −1.0 ppHalpern and Korosi 1998 Hungary Log TFP level Herfindahl level −2.6%Anderson, Lee, Murrell 2000 Mongolia Log TFP level Own market share 9.2%Claessens and Djankov 2000 Hungary Labor prod. growth 3-firm concentration 7.1 ppMarkov et al. 2000 Bulgaria Log TFP level Own market share −23.7%Markov et al. 2000 Bulgaria Log TFP level 3-firm concentration 5.9%Angelucci et al. 2001 Romania Log TFP level Import penetration −6.0%Markov et al. 2000 Bulgaria Log TFP level Import penetration −2.1%Brown and Earle 2000 Russia Log TFP level Import penetration 1.4%Djankov and Hoekman 2000 Bulgaria TFP growth Import penetration 1.1 ppHalpern and Korosi 1998 Hungary Log TFP level Import penetration 3.0%Dobrinsky et al. 2001 Bulgaria Log TFP level Import penetration −0.7%Kreacic 1998 Georgia Log TFP level Import penetration −5.8%Claessens and Djankov 2000 Czech Rep. Labor prod. growth Tariff levels 4.3 pp

1 The economic effect is calculated from the coefficient estimates assuming a 10-percentage-points change in theexplanatory variable. pp means percentage points.

33 The study also uses a large panel of Polish firms.As the data cover both small and large companies, theyare not used in our analysis.

Article 1 8/12/02 9:13 AM Page 765

Page 28: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

ownership concentration and ownershiptypes. The evidence supports the view thatcompetition-enhancing policies and privat-ization should be pursued simultaneously.

The analyses indicate that product mar-ket competition has been a major force be-hind improvements in enterprise productiv-ity in transition economies as a whole, asshown in table 8A. When we divide thesample into analyses based on import com-petition versus domestic market structure(for all countries together), we find thateach is generally significant in explainingenterprise performance. The effect of do-mestic market structure is always signifi-cant. Import competition is insignificant inexplaining enterprise restructuring whengiving less weight to studies that do not con-front endogeneity problems. Examining theresults by region, the effects are strong forEE countries, where in all cases the t-statis-tics are significant at conventional levels.For CIS countries, increased competitivepressures are not associated with enhancedrestructuring.

A further subdivision of the analysesshows an interesting pattern: import compe-tition in the CIS countries does not have asignificant effect on enterprise restructuringand has a negative sign. In contrast, importcompetition is significant in explaining en-terprise restructuring in the EE sample,with t-statistics varying between 3.37 and4.15. Since all the Russian studies use im-port penetration variables that are regionspecific, it would not be correct to concludethat this difference is due to the fact thatRussia has a larger internal market and theeffect of import competition may be mutedor imprecisely measured. What explains thisdifference then? The evidence is consistentwith the theoretical predictions in Blanchardand Kremer (1997) and Ickes, Ryterman,and Tenev (1995). In the CIS, the rapidopening to competition from abroad actedto deter domestic restructuring. Perhaps agradual approach to trade liberalization isthe preferred policy choice.

Changes in domestic market structure areimportant in explaining enterprise restruc-turing in both the CIS and EE samples.(The exception is in the CIS when we weightaccording to attention paid to endogeneitybias.) These results are upheld in a study ofover 3,300 enterprises in 25 transitioneconomies (Carlin et al. 2001) that showsstrong positive effects of the reduction ofmarket concentration on firm efficiency.The significant effect of changes in domesticmarket structure on enterprise restructuringin the CIS resonates with recent evidenceon high barriers to entry in transitioneconomies. Djankov et al. (2002a) documentthe number of procedures and the associ-ated time and cost for starting a new busi-ness in 85 countries around the world, in-cluding 22 transition economies, to explainpoor aggregate growth performance in theCIS. Establishing a new business in Russiaor Ukraine takes twice as much effort, time,and money than a start-up in EasternEurope, and three times longer than estab-lishing a new business in Latvia orLithuania. The authors argue that entry bar-riers serve to impede product market com-petition. Similarly, Andrei Shleifer andDaniel Treisman (2000) document the pres-ence of significant geographic segmentationin Russia and lackluster competition at theregional level.

Table 8B examines the statistical signifi-cance of the variation in the effect of prod-uct market competition across regions andacross types of competition. In the firstpanel, we show competition from local pro-ducers has a stronger effect than importcompetition, but the difference is not statis-tically significant. The second panel showsthat competition has a stronger effect in ex-plaining enterprise restructuring in EEcountries than in CIS countries and that thisdifference is statistically significant whenweighting for attempts to handle endogene-ity, and when controlling for the overallquality of the paper. The third panel com-pares the relative importance of foreign

766 Journal of Economic Literature, Vol. XL (September 2002)

Article 1 8/12/02 9:13 AM Page 766

Page 29: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

competition in the CIS and EE countries,suggesting that foreign competition has alarge positive effect in Eastern Europe andthe Baltics and a large negative effect in theCIS countries. The last panel shows thatthere are no discernible patterns in the wayin which the effects of domestic marketstructure differ between the CIS and EEcountries.

The findings on the effect of import com-petition deserve special attention. In the CIS,import competition has a large negative effectin economic terms, although this effect is sta-tistically not robust. In Eastern Europe, im-port competition has a positive effect in eco-nomic terms, but the results of the individualstudies are mixed, consistent with the litera-ture on other developing economies.

Djankov and Murrell: Enterprise Restructuring in Transition 767

TABLE 8IMPORT AND DOMESTIC COMPETITION AND ENTERPRISE RESTRUCTURING

Composite statistics derived from 82 analyses appearing in 23 studies

Weighting Method Used

(1) (2) (3)

attempts to handle

endogeneity overall ratingnone bias of quality

A. Testing the Effects of Competition normally distributed test statistics

All countries 82 5.13 4.14 5.01Import competition 28 2.86 1.79 3.30Domestic competition 54 4.27 3.78 3.79

EE, all competition 58 5.09 4.90 5.08CIS, all competition 24 1.58 0.68 1.22

EE, import competition 20 3.97 3.37 4.15CIS, import competition 8 −0.93 −1.12 −1.55

EE, domestic competition 38 3.41 3.65 3.16CIS, domestic competition 16 2.59 1.56 2.10

B. Comparing Types of Competition and Regions partial correlation coefficients and test statistics on their difference

Import competition 28 0.003 −0.013 0.002Domestic competition 54 0.025 0.015 0.021Test statistic for difference −1.43 −1.80 −1.53

EE countries 58 0.032 0.026 0.025CIS countries 24 −0.018 −0.030 −0.014Test statistic for difference 2.68 3.05 2.08

EE, import competition 20 0.036 0.029 0.028CIS, import competition 8 −0.080 −0.087 −0.096Test statistic for difference 3.41 3.52 3.59

EE, domestic competition 38 0.030 0.024 0.023CIS, domestic competition 16 0.014 −0.002 0.017Test statistic for difference 0.46 1.21 0.30

Category Number of AnalysesContributing to

the Composite TestStatistic

Article 1 8/12/02 9:13 AM Page 767

Page 30: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

6. Soft Budgets

The concept of soft budget constraintswas coined by Kornai (1979, 1980) to explainshortages in socialist economies, especiallyin the market socialist system. Kornai de-scribes the soft budget phenomenon as“firms are bailed out persistently by stateagencies when revenues do not cover costs”(Kornai 1998, p. 12) and defines soft bud-gets as “the expectation of the decision-maker as to whether the firm will receivehelp in time of trouble . . . ” (Kornai 1998,p. 14). Stiglitz further narrows the definitionto situations when “enterprises believe thatany losses they incur will be made good bythe government” (Stiglitz 1994, p. 184).

Two alternative theories explore the causesof soft budgets.34 Kornai (1979, 1980) relatesthe softness of budget constraints to the pa-ternalistic attitude of the government, whichresults in the accommodation of enterpriserequests for new finance. Firms are financedeven when the expected new project’s returnis below the real interest rate. The govern-ment’s goal is to provide economic securityfor enterprise employees and supply socialservices (kindergartens, schools, hospitals,recreation facilities) in the enterprise.

A different reason for the existence of softbudgets is advanced in Shleifer and Vishny

(1994) and Boycko, Shleifer, and Vishny(1996). These papers model bargaining be-tween politicians and managers, which leadsto equilibria with subsidies to firms and ex-cess employment. Politicians use govern-ment subsidies to induce firms to maintain ahigher-than-efficient level of employment inorder to enlarge their own political con-stituency. Since the model predicts that sub-sidies increase when there is a decline in en-terprise profits, these political decisionsconstitute a soft-budget policy regime.35

These two theories of the causes of softbudgets differ significantly. The first explainsaccommodating lending behavior deter-mined by concerns of social stability, whilethe second suggests that soft budgets arisefrom politicians’ narrower self-interest. Inboth, soft budgets compensate the enterprisefor keeping surplus employment. The pre-dicted effect on enterprise restructuringfrom soft budgets is lack of productivity im-provements and continuation of unprofitableproduction (and nonproduction) activities.

The theoretical literature suggests that anempirical measure of the incidence of softbudgets needs to satisfy two criteria. First, ithas to capture the expectations of enterprisemanagers of a future government bail-out,not the current policy of the government.Imprudent behavior of managers that re-sults in poor enterprise performance can oc-cur in enterprises that have not yet receivedgovernment subsidies but expect to do sowhen they are in trouble. Also, enterprisesthat have received subsidies in the past maynot expect those to continue in the future,perhaps as a result of new government poli-cies. Second, the expectation of governmentsupport is contingent on the enterprise be-ing in financial distress. Both Kornai’s theoryand the Shleifer-Vishny-Boycko model showthat without the fear of imminent job losses,

768 Journal of Economic Literature, Vol. XL (September 2002)

34 A third theory, relevant for both transition and de-veloped market economies, views soft budgets as thecontinued extension of credit even when the substan-dard performance of an already-financed investmentproject has been revealed (Michel Dewatripont andEric Maskin 1995; Maskin 1999). Because of asymmet-ric information, even poor projects are initially fi-nanced. By the time creditors can observe project qual-ity they will continue to lend, because refinancing maystill maximize the expected value of funds that can berecovered. A fourth theory, advanced in Stiglitz (1994),suggests that soft budgets arise when financing institu-tions have an incentive to make large gambles. In thismodel, an insolvent bank is willing to invest in a riskyproject with low or even negative expected payoff, be-cause the bank will become solvent if the gamble paysoff; and if the gamble doesn’t pay off, the bank will beno worse off than it was before it made the loan, i.e., itwill still be insolvent. We do not discuss these two the-ories in the context of soft budget constraints in transi-tion, because the source of soft budgets here is a (pri-vate) financial institution, not the government.

35 In Shleifer and Vishny (1994) there are severaltypes of equilibria, depending on whether bribes occurand on the nature of property rights. The effect of a de-cline in profits is either an increase or no change insubsidies, depending on which equilibrium pertains.

Article 1 8/12/02 9:13 AM Page 768

Page 31: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

i.e., if the enterprise is solvent, there is noreason for the government to extend new financing to the enterprise.

Few empirical papers meet these criteriafor measuring soft budgets. The study thatcomes closest to the prescribed measure isAnderson, Korsun, and Murrell (2000).They survey 249 enterprises that underwentMongolia’s mass-privatization program bymid-1995, asking enterprise managers whatproportion of lost revenues the state wouldmake up if losses threatened the enterprise’sability to maintain its employment level.36

The resulting measure of soft budgets is ex-pectational in nature, and the expectationsare limited to the situation where the enter-prise has lost revenues. Almost three-quartersof the surveyed enterprise managers, 73.1percent, thought that the government woulddo nothing to help out, 2.4 percent of man-agers thought that the government wouldbail them out completely, and 13.6 percentof managers thought that the governmentwould make up at least half of the decline inrevenues. The perceived degree of budgetsoftness was shown to be strongly correlatedwith the amount of central governmentownership. Almost half of the enterpriseswith majority central government ownershipexpected some financial assistance, while31.3 percent of the managers of these enter-prises expected that the government wouldmake up at least half the decline in rev-enues. The data on managers’ expectations

is used in Anderson, Lee, and Murrell(2000) to study the effect of soft budgets onenterprise performance.

While theoretically most appropriate, theexpectational measure used in Anderson etal. is not easily replicable in other studiessince it relies on a survey instrument de-signed for studying the soft budgets phe-nomenon. In the absence of such data, otherauthors use quantitative variables that proxyfor the expectations of enterprise managers,while different variables are used to controlfor the financial distress of the enterprise.The best example of this approach is DavidLi and Minsong Liang (1998), who constructindicators of soft budgets from survey dataon employment of nonproduction workers,investment with below-average rates of re-turn, and distribution of bonuses in excess ofthese regulated by the Chinese government.All three factors are shown to contribute tothe financial distress of the enterprise. Usinga sample of 681 enterprises for the period1980 to 1994, Li and Liang find that if all ofthe non-production workers, defined asworkers with zero marginal product (such ascleaners, political counselors, etc.), wereeliminated, 38 percent of financial losseswould be avoided. As another example, ex-cessive bonuses accounted for 39 percent oftotal enterprise losses. The empirical ques-tion that the authors ask is whether enter-prises respond to financial losses, defined asnegative cash flows, by reducing any of thefactors that cause these losses initially. Theanswer is No. All enterprises in financialdistress expected to be bailed out by thegovernment.

Several other papers illustrate the diffi-culties in constructing a soft budget proxy inthe absence of survey data. These studiesrely on data from the financial statements ofenterprises. Wafa Abdelati and StijnClaessens (1996) and Fabrizio Coricelli andSimeon Djankov (2002) use samples of overfour thousand Romanian companies and dis-tinguish among chronic loss-makers and ad-justed enterprises. Chronic loss-makers are

Djankov and Murrell: Enterprise Restructuring in Transition 769

36 The question was phrased as follows: “Suppose thatunfortunate market conditions resulted in a sudden dropin your enterprise’s revenues, so that you might have tolay off workers. How likely is it that the government (ei-ther national or local) would help your enterprise out, sothat it would not be forced by its financial situation to layoff workers? Please indicate your expectation of thelikely government reaction by choosing a point on thescale from 0 to 10—a “0” means that you think the gov-ernment would do absolutely nothing to help out and a“10” means that you think the government would com-pletely make up for the decline in revenues in some way,and a “5” means that the government would make uphalf the decline in revenues. Choose any number be-tween 0 and 10, indicating your expectation concerningthe extent to which the government would help out”Anderson, Korsun, and Murrell (2000, p. 222).

Article 1 8/12/02 9:13 AM Page 769

Page 32: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

defined as enterprises with negative cashflows during the 1991 to 1994 period.Adjusted enterprises are defined as enter-prises that had negative cash flows in 1991or 1992 or both, but then recorded positivecash flows in 1993 and 1994. The proxy forsoft budgets is the amount of new bank fi-nancing, net of accrued interest expenses.Since the banking system was fully in statehands in 1994, this proxy measure capturesthe major channel of government funding,as tax arrears and direct government subsi-dies are shown to be small relative to theflow of new credit. The authors seek to an-swer the question whether chronically loss-making enterprises received financial sup-port in order to keep employment as high asthat of profitable enterprises. The answer isYes. In fact, during the sample period, loss-making enterprises received about 50-per-cent more bank credit as a share of revenuesand shed less labor than did adjusted firms.Assuming that the government would con-tinue its policy of keeping loss-making en-terprises afloat, these studies provide empir-ical support for the presence of soft budgets.Stijn Claessens and Kyle Peters (1997) use asimilar approach to study the effect of softbudget constraints in Bulgaria. They usetrade arrears to other state enterprises, andthe share of external financing in total assets,as alternative proxies for soft budgets. Thedependent variable is the change in valueadded.

Earle and Estrin (1998) use data forabout three hundred Russian enterprisesduring 1990 to 1994 and construct a mea-sure of government assistance to enter-prises, which they interpret as a proxy forsoft budgets. The measure relies on flow-of-funds data and includes all possible chan-nels of government support, such as federalsubsidies and investments, tax benefits,preferential credits, extra-budgetary funds,tax exemptions and other benefits associ-ated with foreign trade, and local govern-ment subsidies and tax arrears. In additionto relying on a current policy proxy, the

analysis does not capture the financial dis-tress of the enterprise, making the resultsdifficult to compare with the theories of softbudgets. For example, product specific sub-sidies, often the result of price controls, canbe given to both profitable and loss-makingenterprises. Studies using such data docu-ment the effect of distortions due to pricecontrols as much as the incentives of softbudgets.

Earle and Estrin are in good company.Claessens and Djankov (1998), Djankov andHoekman (2000), Grigorian (2000), andCarlin et al. (2001) use similar variables asproxies for soft budgets. This pattern is inpart explained by the fact that all five stud-ies focus on the effects of ownershipchanges and use the soft budgets proxy onlyas a control variable. We include these stud-ies in the survey since they provide valuableinformation on the incidence of govern-ment financing of loss-making enterprisesin transition economies, and some of the ef-fects that this financing has. If anything, ourcriticism of their approach reflects thedearth of research focused on the soft bud-gets issue.

The studies discussed so far document nu-merous channels of government support forailing enterprises. While Abdelati andClaessens (1996) and Coricelli and Djankov(2002) identify new credit from the state-owned banking sector as the main conduit ofsoft budgets in Romania during the earlytransition period, and Claessens and Peters(1997) finds similar results for Bulgaria,Mark Schaffer (1998) shows that bank lend-ing was not a major source of soft budgets inHungary during the early transition period,even before the banking sector was privat-ized. Once the Hungarian banking systemwas largely privatized, by 1995, bank creditwas restricted to profitable projects(Claessens and Djankov 1998). Tax and socialsecurity arrears to the central governmentwere the main source of soft financing inPoland in 1993, while trade credit was not asource of soft budgets in either Hungary or

770 Journal of Economic Literature, Vol. XL (September 2002)

Article 1 8/12/02 9:13 AM Page 770

Page 33: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

Poland.37 These results are reflected also inClaessens and Djankov (1998) who find that,by the mid-1990s, financing of loss-makingenterprises by the still state-owned bankingsector was the primary channel of soft bud-gets in Bulgaria, the Czech Republic,Romania, Slovakia, and Slovenia, but not inHungary or Poland. Not coincidentally, bythe end of the 1990s all of these countrieshad experienced severe (Bulgaria andRomania) or at least significant (the CzechRepublic, Slovakia, and Slovenia) bankingcrises.

Shleifer and Treisman (2000) show thattax exemptions by the local government arethe main channel of soft financing in Russia,as do Gilles Alfandari, Qimiao Fan, and LevFreinkman (1996) and Brown and Earle(2000). Direct subsidies from the centralgovernment are an important source of softbudgets in Kazakhstan (Simeon Djankovand Tatiana Nenova 2000) and Lithuania(Grigorian 2000). Finally, Shumei Gao andMark Schaffer (1998) show that credit is al-ways available to loss-making state enter-prises in China.

What is the economic significance of hard-ened budgets on enterprise restructuring? Inaddition to the results already discussed, Liand Liang (1998) find that if all inefficient in-vestment projects were eliminated, the aver-age enterprise’s profit margin would changefrom –8.7 percent to 2.3 percent. Claessensand Peters (1997) find that the presence ofsoft budgets in Romanian enterprises resultsin a reduction of labor shedding by 4 percentannually during 1992 to 1994. Coricelli andDjankov (2002) find that labor shedding wasreduced by 4.6 percent during 1993–1995.Claessens and Djankov (1998) find a 2.7 per-cent unrealized TFP growth as a result ofcontinued soft financing in the Eastern

European countries. Djankov and Hoekman(2000) find an unrealized annual growth of 3percentage points in Bulgaria over the1992–95 period. Earle and Estrin (1998) finda 5.7 percent unrealized labor productivitygrowth. Alfandari, Fan, and Freinkman(1996) show that recipients of soft financingrecord labor productivity growth that is 6 per-cent less than that of non-recipients. Finally,Abdelati and Claessens (1996) find that aone-standard-deviation increase in the flow offinancing from state banks in Romania is as-sociated with a 14.2 percent unrealized laborproductivity growth per annum. This evi-dence suggests that soft budgets are a majordeterrent of enterprise restructuring.

We next turn to a discussion of the statisti-cal significance of the results on soft budgets.We discuss the results of ten papers that useregression analysis to answer these questions.The data come from Bulgaria (Claessens andPeters 1997; Djankov and Hoekman 2000),Kazakhstan (Djankov and Nenova 2000),Lithuania (Grigorian 2000), Romania(Abdelati and Claessens 1996; Coricelli andDjankov 2002), Mongolia (Anderson, Lee,and Murrell 2000), Russia (Earle and Estrin1998), and cross-country studies of the sevenCentral and East European countries(Claessens and Djankov 1998) and 25 transi-tion economies (Carlin et al. 2001). Theycover the period between 1992 and 1999, andcontain 31 separate analyses.

Before discussing the aggregate results,however, we acknowledge the simultaneityproblem that plagues most empirical workon soft budgets.38 Theory suggests thatpoorly performing firms get bailed out, whilegood firms do not receive government fi-nancing since they don’t need it. Researcherswant to examine whether soft budgets causepoor performance, yet the data may be over-whelmed by the relationship operating in theother direction. Researchers have addressedthis problem in various ways, but none isentirely satisfactory. Weighting papers by a

Djankov and Murrell: Enterprise Restructuring in Transition 771

37 Overdue trade credit between private enterprisesshould not be considered as evidence of soft budgets.Also, John McMillan and Christopher Woodruff (1999)show that private trade creditors in Vietnam stop financing enterprises once their payments are twomonths in arrears. 38 We owe this point to Mark Schaffer.

Article 1 8/12/02 9:13 AM Page 771

Page 34: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

subjective measure of how the authors havehandled the simultaneity bias and by theoverall quality of the empirical analysis givessome perspective on the importance of theproblem, but does not eliminate it.

Table 9A shows that the effect of hard-ened budgets on enterprise restructuring,defined as sales growth, TFP, or labor pro-ductivity, is very significant in EE countriesand generally significant in CIS countries.The positive signs imply that hardenedbudgets have a beneficial effect on restruc-turing. Table 9B compares the size of thehardened budget effect across the two re-gions. The studies on EE and CIS countriesshow effects of similar magnitude, with thet-statistics for differences between partialcorrelations insignificant.

How can soft budgets be limited? BothKornai and Shleifer-Vishny suggest, and theempirical literature shows, that privatizationof enterprises is the main policy choice.Anderson, Korsun, and Murrell (2000) showthat a 10-percent-larger share of central-government ownership increases by 9 percentthe probability of receiving soft financing.Their estimates suggest that privatization re-duced the percentage of enterprises with softbudgets from 78 percent to 23 percent.Similarly, Alfandari, Fan, and Freinkman(1996) show that the probability of receivingstate support is reduced in half if the enter-prise is private. Frydman, Hessel, andRapaczynski (2000) find evidence to suggestthat hardening budgets induces cost efficiencybut not revenue restructuring, and that the onlypolicy for the Czech, Hungarian, and Polishgovernments to commit to hard budgets wasthrough privatization of enterprises. Withoutprivatization, it is often rational for the gov-ernment to extend assistance to loss-makers.39

The literature surveyed here also showsthat larger state-owned enterprises are morelikely to be the recipients of soft financing.This implies that policies that reduce therole of industry giants (through de-monopo-lization, split-ups, or spin-offs) will also re-duce the presence of soft budgets (Lizal,Singer, and Svejnar 2001).

Other findings are less robust. Privatizingthe banks may bring about a reduction in softfinancing. A study of 92 economies, includ-ing eleven transition economies, shows thatgovernment ownership of banks is associatedwith lower TFP growth in the enterprise sec-tor, stemming from inefficient allocation ofresources across enterprises (Rafael LaPorta, Florencio Lopez-de-Silanes, andAndrei Shleifer 2002). Beware of the Stiglitz(1994) model, however. There are manyexamples of private banks extending softcredits to enterprises as part of their ownsurvival-through-gambling strategy. Finally,the decentralization of the governmentbudget and decision making seems to havelimited soft budgets in China (Yingyi Qianand Gérard Roland 1996) and Mongolia(Anderson, Korsun, and Murrell 2000) butserved to increase the incidence of support-ing failing enterprises in Russia (Blanchardand Shleifer 2000; Ekaterina Zhuravskaya2000). Further empirical research is neededto be more definitive on the effects of bankprivatization and fiscal decentralization.

7. The Role of Managers

Economic theorists have taken two broadapproaches in assessing the role of managersin improving corporate performance. Onefocuses on manager incentives. MichaelJensen and William Meckling (1976) holdthat managers want to pursue projects orconsume perquisites that benefit them per-sonally. This argument is based on the obser-vation that in most modern corporations, asshown in Berle and Means (1933) and morerecently in La Porta, Lopez-de-Silanes, andShleifer (1999), ownership is separated from

772 Journal of Economic Literature, Vol. XL (September 2002)

39 One example: A policy of offering generous sever-ance pay for workers was tried in Romania in 1992 to1994. Employees of large loss-making enterprises wereoffered twelve months of wages as severance. However,the government continued the flow of soft financing tolarge loss-making enterprises even after severancepackages were offered, thus dooming the initiative tofailure (Djankov 1999b).

Article 1 8/12/02 9:13 AM Page 772

Page 35: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

control and hence managers (who have thecontrol rights) do not bear the consequencesof their actions. A variation of the agencyproblem, developed in Holmström (1979),studies unobservable managerial effort.When owners cannot judge how hard man-agers work, there is an incentive for man-agers to slack off. The implication of boththeories is that when manager ownership ishigh, managers will act so as to increase en-terprise efficiency.40 Theoretical models onthe early transition, e.g., Aghion, Blanchard,and Burgess (1994) and Aghion andBlanchard (1996), follow the incentive liter-ature by showing how the incentives of man-agers to behave in a profit-maximizing man-ner improve once they hold a significantstake in the performance of their firms.41

A second theory focuses on managementturnover (Sherwin Rosen 1992). The theorystipulates that if the manager is not capableor knowledgeable, no amount of incentiveswill do. In such instances, it is better to bringin a new manager. Barberis et al. (1996) ap-ply this theory to data on Russian shops andfind that the hiring of new managers bringsabout significant improvements in enter-prise performance, while giving incentivesto old managers does not.

In this section, we survey the literaturethat has studied government policy affectingthe role of managers in improving enterpriseperformance in transition. Unlike the stud-ies on western firms where corporate boardsdesign manager incentives and fire man-agers for bad performance, the transition lit-erature on managers to-date has focused pri-marily on government policies. Many of thestudies discussed here use data on Chineseenterprises. There are few studies that testthe theories of manager incentives and man-ager turnover in Central and EasternEurope or the former Soviet Union. Thisdisparity is a reflection of the policies taken

Djankov and Murrell: Enterprise Restructuring in Transition 773

40 In both theories, the optimal level of managers’ownership is below 100-percent if managers are riskaverse (Morck, Shleifer, and Vishny 1988).

41 Furthermore, career concerns may lead managersof state-owned enterprises to restructure if privatizationin the economy is imminent or under way, since privat-ization creates a market for managers and hence in-creases managers’ incentives to outperform their peers(Roland 1994; Gérard Roland and Khalid Sekkat 2000).

TABLE 9HARD BUDGET CONSTRAINTS AND ENTERPRISE RESTRUCTURING

Composite statistics derived from 31 analyses appearing in 11 studies

Weighting Method Used

(1) (2) (3)

attempts to handle

simultaneity overall ratingnone bias of quality

A. Testing the Effects of Hardening Budgets composite normally distributed statistics

All countries 31 8.44 6.81 7.63EE 23 7.34 6.52 7.78CIS 8 4.17 2.26 2.13

B. Comparing the Effects partial correlation coefficients andin the Two Regions test statistics on their difference

EE 23 0.027 0.025 0.029CIS 8 0.065 0.045 0.045Test statistic for difference −1.53 −0.64 −0.50

Number of AnalysesRegionContributing to

the Composite TestStatistic

Article 1 8/12/02 9:13 AM Page 773

Page 36: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

by the Chinese government at the outset ofeconomic reforms. Privatization was notseen as an option and instead the govern-ment sought to improve the performance ofstate-owned enterprises by giving their man-agers more autonomy and better incentives(Barry Naughton 1994).42

The decision to provide more autonomyand better incentives in China started in1978 as an experiment in the agricultural sec-tor (John McMillan, John Whalley, andLijing Zhu 1989). It was quickly emulated inmanufacturing (Keun Lee 1990). The evi-dence shows that managers responded to thenew autonomy by strengthening the disci-pline imposed on workers, by increasing theproportion of workers’ income paid in theform of bonuses, and by raising the fractionof workers on fixed-term contracts(McMillan and Naughton 1992). Enterprisesalso spent more on productive investment(Theodore Groves et al. 1994).

The incentive system let state enterprisesretain profits, separated management frompolitics, and allowed managers to be treatedas individual contractors, with their pay andbonuses to be linked to the performance oftheir units. These changes in incentiveswere so fundamental, that the process issometimes called “privatization from below,”e.g., Naughton (1994, p. 266). The policywas designed to give managers controlrights, contingent on keeping governmentownership intact.

The move to autonomy in decision mak-ing at the enterprise level may account for alarge part of the estimated effect on enter-prise restructuring. The autonomy effectneeds to be controlled for, in order to esti-mate the effect of incentives per se.Fortunately, most of the studies that we sur-vey use separate variables for the autonomycomponent of the responsibility system, e.g.,Lee (1990), Groves et al. (1994), RogerGordon and Wei Li (1995). The simulta-

neous introduction of several policies withinthe responsibility system also makes it diffi-cult to differentiate the effect of manager in-centives from that of manager turnover.Indeed, “stronger incentives and bettermanagers are complementary changes. Theymight be so strongly complementary thatneither change would be effective by itself.Some managers might be so inadequate asto be unable to respond to new incentives,no matter how well designed. Good man-agers might not work well under badly struc-tured incentives. If so, restructuring is effec-tive only if both changes—new managersand new incentives—are introduced to-gether” (McMillan 1997, p. 7). We try togive the reader some sense of the relativeimportance of turnover versus incentives,but one should consider them two parts of asingle policy.

The measurement of incentives createssome problems. While data on equity own-ership and salaries of managers do exist andhave been used in a number of studies, dataon stock options and bonuses are less readilyavailable. Yet these types of incentives couldaccount for a large part of the manager’s de-cision to stay with or join a particular firmand work hard on improving its perfor-mance. While liquid capital markets are rarein transition economies (Claessens, Djankov,and Daniela Klingebiel 2000) and stock op-tions are not much used, bonuses or man-agement pay linked to enterprise perfor-mance are used in all transition economies(Derek Jones and Takao Kato 1996). InChina, bonuses or, in the case of enterpriseswith auctioned management positions,bonuses based on the bid price of the newmanager are used in 97 percent of enter-prises (Gordon and Li 1991). Outside ofChina, however, we could not find any em-pirical studies using such data.

The earliest study on management incen-tives and enterprise performance in China isLee (1990), which constructs dummies forwhether the enterprise has implemented a)all incentive reforms; or b) any one of these

774 Journal of Economic Literature, Vol. XL (September 2002)

42 The Polish government followed a similar policyin the first years of transition (Pinto, Belka, andKrajewski 1993).

Article 1 8/12/02 9:13 AM Page 774

Page 37: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

reforms. The results indicate that the adop-tion of all reforms yields about a 4 percenttotal factor productivity increase, while ac-counting for other firm characteristics suchas the initial productivity level, export orien-tation, the ratio of nonmanagerial to mana-gerial employees, and the ratio of temporaryworkers to total employment.

Groves et al. (1995) show that manage-ment incentives improved profitability inChinese enterprises by 7.3 percentagepoints and that managers benefited signifi-cantly from this improvement: the averagepay with new incentives was 37.4 percenthigher than the previous average salary.Using the same data set to construct a ten-year panel, Groves et al. (1994) show thatbonuses are positively associated with pro-ductivity in five manufacturing industries:textiles, chemicals, building materials, ma-chinery, and electronics. In all but one,chemicals, this association is statistically sig-nificant. Gordon and Li (1995) and Li(1997) also find manager incentives to be ef-fective in increasing total factor productivityin China. Both papers control for the timingof decentralization, as well as for non-linear-ities in the production function. Using datafor 1983 to 1987, Gordon and Li find thatfirms implementing manager incentives ex-perienced an additional 17 percent increasein productivity beyond that found for otherfirms. Li (1997) finds that an increase inbonuses by one percentage point raised theTFP growth rate by 0.089 percentage pointsbetween 1980 and 1984, and by 0.060 per-centage points between 1985 and 1989.Similarly, Chun Chang, Brian McCall, andYijiang Wang (2001) find manager incen-tives to be associated with a 4 percent in-crease in the return to equity, using a paneldata set of eighty agro-processing enter-prises from 1984 to 1993.

In contrast, Harry Broadman and GengXiao (1997) find a positive but generally in-significant effect of incentives on laborproductivity in the early reform period,1980 to 1985, and Shirley and Xu (2001)

document a negative but generally in-significant relation between manager per-formance contracts and TFP growth inChinese state enterprises. Claessens andDjankov (1999b) use data on stock owner-ship in publicly listed Czech companiesand fail to find evidence of the effect ofmanager incentives on enterprise perfor-mance. However, all three papers lackgood proxies for manager incentives. Forexample, Broadman and Xiao do not havedata on managerial wages and bonuses anduse the gross value of nonindustrial fixedcapital as a proxy for in kind compensation.Shirley and Xu use a dummy variable forthe presence of a performance contract asa proxy, but cannot differentiate amongcontracts that offer generous bonusesbased on performance and contracts thatare merely window dressing.

We turn to the effect of managerturnover next. The seminal study in thetransition literature linking managerturnover and enterprise performance isGroves et al. (1995). They use a sample ofChinese state-owned companies to investi-gate the link between manager turnoverand productivity growth. Poorly performingfirms are more likely to have a new managerselected by auction,43 and such managersare required to post a higher security de-posit and are subject to more frequent per-formance reviews. Managers are frequentlyfired for poor performance and in suchcases often lose part of their security de-posit. The incidence of turnover during thesample period of 1980 to 1989 is very high.Only 11 percent of managers in place in1980 remained managers by the end of theperiod, and almost half (44 percent) of man-agers were appointed after 1985. Thesechanges are primarily a reflection of thefunctioning of the market for managers:only 25 percent of turnover is the result ofretirement. The economic effect of replac-ing bad managers is high: new managers

Djankov and Murrell: Enterprise Restructuring in Transition 775

43 Auctions for managerial jobs took place in 14 per-cent of the sample enterprises.

Article 1 8/12/02 9:13 AM Page 775

Page 38: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

bring about a 16-percent increase in laborproductivity.44

The literature on the effects of changes inmanagers in Central and Eastern Europeand the former Soviet Union suffers from anacute problem: management turnover oftentakes place as an endogenous decision afterprivatization. New owners pick new man-agers, and the effect of management changeis confounded with ownership change, espe-cially when the new owners themselves arethe new managers. For example, the retailshops studied in Barberis et al. (1996) are inmany cases manager-owned. Even for mid-size and larger enterprises, managers may belarge owners (Kuznetsov, Mangiarotti, andSchaffer 2002) and hence lose their jobswhen there is ownership change.

The data set in Claessens and Djankov(1999b) is well suited for empirical testing ofthe importance of management turnover.First, the privatization process in the CzechRepublic prevented incumbent managersfrom obtaining significant ownership. As a re-sult, management changes were separatedfrom ownership change. The data display highturnover: 35.6 percent of managers in pri-vately owned enterprises and 42.1 percent inthose under state ownership. The authorsconstruct dummy variables for enterpriseswithout management changes, for enterpriseswhere managers were replaced while understate ownership, and for privatized enterpriseswhere the new owners hired new managers.These dummies are used alongside variablesfor the initial labor productivity level, the en-terprise size, and manager’s equity ownershipto explain differences in annual labor produc-tivity growth rates. Labor productivity growthincreases by 4.2 percent with the appointmentof a new manager in privatized enterprises.The appointment of new managers in state-owned enterprises yields a 3.5-percent in-crease in labor productivity growth.

In a similar study, Frydman, Hessel, andRapaczynski (2000) use a sample of state-owned and newly privatized enterprises inthe Czech Republic, Hungary, and Poland tostudy management turnover and its effecton the annual rate of revenue growth duringthe period 1991 to 1993. The turnover ratewas extremely high: nearly two-thirds (64percent) of managers were dismissed ormoved voluntarily. Managers changed in 75percent of state-owned enterprises, in 72percent of enterprises with a dominant out-side owner, and in 55 percent of insider-owned enterprises. In state-owned enter-prises, management turnover is associatedwith a 3.1-percent increase in revenuegrowth, but this effect is statistically insignif-icant. In privatized enterprises, the effect ismuch larger, at 18.5 percent, and statisticallysignificant. Using the same methodology in astudy of large Ukrainian firms, Warzynski(2001b) finds that management turnoverdoes not affect the change in productivity instate-owned enterprises, but has a small pos-itive effect in privatized enterprises.45

The studies reviewed in this section showthat manager incentives work and manage-ment turnover is almost always effective inimproving enterprise performance. To un-derstand the composite implications of thesestudies, we combine the regression resultsfrom twelve studies with 33 separate analy-ses, testing the importance of both manage-rial turnover and managerial incentives inrestructuring (table 10).46 We find that in allcases manager incentives and turnover, con-sidered together, are an important determi-nant of restructuring. The pertinent t-statis-tics have values between 8.19 and 9.46.Separately, manager turnover and managerincentives also have significant effects on re-structuring. Panel B directly compares the

776 Journal of Economic Literature, Vol. XL (September 2002)

44 Shirley and Xu (2001), using the same data set for1986 to 1989, also find the economic effect of managerturnover to be quite high: a new manager brings about4 percentage points higher TFP.

45 The turnover was higher in state-owned enter-prises, where 60.7 percent of management positionschanged hands in 1993–97, than in privatized enter-prises, with 47.5 percent of management turnover.

46 The construction of panel A employs the methodused in table 3, while the construction of panel B usesthe method of table 4.

Article 1 8/12/02 9:13 AM Page 776

Page 39: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

size of the two economic effects. There is nostatistically significant difference, indicatingthat management turnover can be as effec-tive as manager incentives for enterprise re-structuring. Note, however, that the evi-dence on incentives comes from Chinaalone, and that this result seems to be muchweaker in countries where incentive policiesare not used as a substitute for ownershipchange (Qian 1999). In such countries, theevidence shows that privatization has a largeand significant effect on performance, whilemanager-related variables usually do nothave statistical significance (Warzynski2001b).

Before concluding this section, we sug-gest one avenue for further research onmanager incentives in transition. It is strik-ing that the literature has not studied thepresence of penalties (disincentives) formanagers in the event of poor performance,asset-stripping, or spending enterprise re-sources for personal enrichment. While wehave not been able to find any empirical lit-erature on this topic, a multitude of press

reports on China suggests that the govern-ment has imposed harsh penalties, includingthe death penalty, for enterprise managerswho abuse their power, e.g., China Daily(2000). In contrast, the literature on Russiasuggests that the central government wasunable to impose any such disciplines(Shleifer and Treisman 2000). The presenceof such “sticks” may be an important deter-minant of the apparent positive effect ofmanager incentives in China (Chen Qingtai2001). In an environment where negligentor fraudulent behavior by managers is se-verely punished when uncovered, managershave the choice of working hard and gettingbonuses or slacking off and living off theirsalary alone. In contrast, where bad behaviorgoes unpunished, managers have the choiceof stripping enterprise assets and getting ahuge windfall now, as opposed to workinghard through the years and receiving bettercompensation through bonuses. The evi-dence in Black, Kraakman, and Tarassova(2000) shows that many Russian managerspursue the former path.

Djankov and Murrell: Enterprise Restructuring in Transition 777

TABLE 10THE ROLE OF MANAGERS IN ENTERPRISE RESTRUCTURING

Composite statistics derived from 33 analyses appearing in 12 studies

Weighting Method Used

(1) (2) (3)

attempts to handle

simultaneity overall ratingnone bias of quality

A. Do Managers Matter? composite normally distributed statistics

Both turnover and incentives 33 8.19 9.18 9.46Management turnover 13 7.63 7.55 8.34Management incentives 20 4.38 5.67 5.40

B. Comparison of Turnover partial correlation coefficients and testand Incentives statistics on their difference

Management turnover 13 0.034 0.032 0.038Management incentives 20 0.039 0.047 0.040Test statistic for difference −0.39 −1.01 −0.21

Number of Analyses

Region Contributing tothe Composite Test

Statistic

Article 1 8/12/02 9:13 AM Page 777

Page 40: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

8. The Future Research Agenda

In aggregating the results of the many pa-pers that examine the determinants of enter-prise restructuring, we have arrived at strongconclusions. Given the number of analysesbeing aggregated, almost six hundred in total,it is unlikely that any new study would modifythese conclusions significantly unless itswhole approach constituted a new departure.

Many of the studies under review werestimulated in part by the easy availability ofnew types of data, which opened up wholenew avenues of enquiry. The design of thestudies was not driven by attempts to ad-dress difficult methodological problems, byefforts to examine the effects of less easilyquantifiable aspects of policy, such as the na-ture of institutions,47 or by considerationsthat viewed transition in the wider context ofdevelopment. In the future, the most pro-ductive research will focus directly on thesethree issues. In the ensuing subsections, wediscuss each of the three.48

8.1 Selection, Simultaneity, andComplementarities

In designing a new research project thatexamines those determinants of restructur-ing that we have already reviewed, whatshould the central focus be? Our reading ofthe concerns of fellow researchers is that themost important source of skepticism aboutexisting results centers on the possible biasesin estimates that might arise either from si-multaneous causation or from the selectionof enterprises to be subjected to some policy.We have focused on this issue in the previoussections and we have argued that our conclu-sions are robust to such bias. However, wesense enough residual concern over this

issue to suggest that there is a large pay-off from any study that goes beyond themethods employed in the existing literature.

A study aimed at quelling all possibledoubts about bias in estimates would focus,especially in the collection of data, on ways todiminish selection or simultaneity bias.Existing studies usually confronted theseproblems after data collection, then beingforced to employ ad hoc methods. In con-trast, it is with data collection that the mostprofitable route would begin. In the case ofprivatization and the effects of different own-ers, this would entail detailed, micro-institu-tional study of the process that generatedprivate ownership. The hope would be thatsuch study would uncover instrumental vari-ables whose properties were so transparentthat they left no concerns. In the case ofcompetition or soft budgets, the research de-sign would entail the construction of mea-sures of competition or of budget hardnessthat were independent of enterprise per-formance. Again, detailed study of institu-tional processes would be highly beneficial,in leading the researcher to the types of datathat have the desired exogeneity propertiesand in convincing the reader of this fact.

An examination of complementarities inpolicies provides a second area where new re-search might offer a significant possibility ofmodifying the conclusions reached above. Inpolicy debates at the beginning of transition,there was great emphasis on complementari-ties. However, neither theory nor existingempirical studies provide strong guidance onwhere such complementarities exist or how tomodel or measure them.49 Existing empiricalwork on complementarities relies mainly ontesting the effects of simple interactions be-tween policy variables, suggesting an ad hocapproach formulated after data collection. Amore convincing analysis would focus on

778 Journal of Economic Literature, Vol. XL (September 2002)

49 Angelucci et al. (2001), Grosfeld and Tressel(2001), Earle and Estrin (1998), Perevalov, Gimadi,and Dobrodey (2000), and Djankov and Hoekman(2000) all contain results on complementarities.However, the patterns in the results defy an easy sum-mary.

47 This is a general characterization to which thereare notable exceptions. For example, Joel Hellman,Geraint Jones, and Daniel Kaufmann (2000) and Carlinet al. (2001) sought measures of nonstandard enter-prise determinants.

48 When focusing on deficiencies in the literature,we should make it clear that our own papers are not ex-empted from such criticism.

Article 1 8/12/02 9:13 AM Page 778

Page 41: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

complementarities at the beginning of re-search, designing data collection on the basisof both theory and detailed country-specificstudy. Using a structural approach to the esti-mation of complementarities, rather than try-ing different ad hoc interaction effects, wouldprovide a more convincing analysis and wouldoffer the possibility of policy implications ofprofound importance.

This is an appropriate juncture to make amore prosaic suggestion on future research.As we have become painfully aware whilewriting this paper, the existing literature hasa number of deficiencies in the presentationof results and in the amount of informationoffered to the reader. All too often, papersobscure the type of quantitative informationon economic effects that we presented in ta-bles 1 and 7. Often, this information canonly be extracted with great effort; indeed,in a significant number of papers, we couldnot deduce the pertinent magnitudes.However, the size of economic effectsshould arguably be the most significant re-sults in papers that examine policy.50

Deficiencies in presentation of informa-tion become most significant when trying tocompare or combine the results of papers onidentical topics. For example, we wouldhave been able to present more incisive con-clusions in the previous sections, if authorsof the surveyed papers had devoted moreconsideration to presenting results so thatthey could be accurately compared with therest of the literature. In this respect, it seemsthat economics lags some other empiricalfields (Hunt 1997).

8.2 The Elusive Determinant: Institutions

The beginning of transition coincidedwith the publication of North’s (1990) influ-ential book, with its central message that in-stitutions were crucial to the success of mar-ket capitalism. This message was not at the

forefront of policy discussions during theearly years of transition.51 Gradually the fo-cus has changed, spurred by a recognitionthat the relatively poor performance of theCIS was not explained by differences instandard reforms. Institutions are now invogue (Johnson, Kaufmann, and Shleifer1997; Blanchard and Kremer 1997; Stiglitz1999). It is no coincidence that the sameemphasis is now present in the more generalliterature on growth and development(Robert Hall and Charles Jones 1999; DaronAcemoglu, Simon Johnson, and JamesRobinson 2001).

As Oliver Williamson (2000, p. 595)states, however, “we are still very ignorantabout institutions.” This is no more so thanwhen relying on the evidence that is the focus here, the determinants of large-enterprise restructuring in transition. Thereis a dearth of studies including institutionalvariables. One reason for this is that researchhas followed policy, focusing on privatiza-tion, competition, and soft budgets. A sec-ond is that the institutional framework is of-ten the same for all enterprises, leading todifficulties in designing tests. This has led tothe use of indirect estimates to judge institu-tional effects.

One indirect approach uses variationsacross enterprises in the need for institu-tions. This has been employed to examineBlanchard and Kremer’s (1997) theory thatthe erosion of contract enforcement mecha-nisms was a primary factor causing the largeoutput fall in the CIS. They hypothesize thatthe need for contract enforcement is morecritical for enterprises with more complexinput requirements. Such enterprises wouldperform relatively worse in the period afterthe abandonment of contract enforcementthrough planning but before the creation ofany effective alternative. This prediction

Djankov and Murrell: Enterprise Restructuring in Transition 779

50 Other information is also frequently omitted. Forexample, the methodological appendix discusses howcomparisons between different owners are made diffi-cult because papers omit crucial statistical information.

51 For a fuller history of the ebb and flow of policiesin the first decade of transition, see Cynthia Clementand Peter Murrell (2001). The essays in ChristopherClague and Gordon Rausser (1992) constitute an ex-ception to the early lack of focus on institutions.

Article 1 8/12/02 9:13 AM Page 779

Page 42: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

also follows from the observation that thesupply of information and the coordinationof decisions was a central task of the now de-funct planning apparatus (Murrell 1992).

Tests of this prediction relate measures ofthe complexity of enterprise relationships toeconomic performance, with complexitycapturing the institutional effect indirectlyby proxying the need for institutions.Konings (1998) measures complexity by thenumber of firms in the enterprise’s sectorand Jozef Konings and Patrick Paul Walsh(1999) use number of products produced byan enterprise, obtaining evidence that isgenerally supportive of the prediction inBulgaria, Estonia, and Ukraine. FrancescaRecanatini and Randi Ryterman (2000) usethe measure suggested by Blanchard andKremer (1997), the diversity of sources ofinputs, with negative results. However, theydo find that growth was lower in those enter-prises that formerly received the strongestinstitutional support from central planning,one interpretation being that this variable isa proxy for some institutional need.

Johnson, McMillan, and Woodruff (2002)also examine contract enforcement, using adifferent, indirect approach. They exploitdifferences in managerial beliefs about theeffectiveness of institutions, finding thatmanagers who believe that courts can en-force contracts extend more trade creditthan those who do not. These results suggestthat effective courts can significantly en-hance economic efficiency. In a similarstudy, Timothy Frye (2001) finds that thevariations in the quality of police services, asmeasured in opinions, are important deter-minants of investment. Kathryn Hendley,Murrell, and Ryterman (2001) use theamount of legal human capital in an enter-prise to assess the effects of law on Russianenterprises. They find that such capital is animportant determinant of success in transac-tions, concluding that legal institutions af-fect enterprise performance even in suppos-edly lawless Russia. Although their measureof human capital is based on facts rather

than opinions, the conclusion that institu-tions affect performance is still only an indirect implication.

Corporate governance is another impor-tant element of the institutional infrastruc-ture pertinent for large enterprises. But theevidence on this too is sparse and indirect.Using largely anecdotal evidence, Black,Kraakman, and Tarassova (2000) for Russia,and Stiglitz (1999) more generally, claimthat the failure of corporate governance in-stitutions has been of great importance.Anderson, Korsun, and Murrell (1999) douse systematic survey data to show that cor-porate governance laws work poorly inMongolia, but they present no evidence onwhether there is a cost in terms of foregonerestructuring.

Black (2001) focuses on corporate gover-nance practices in Russia enterprises. Heestimates that a 600-fold increase in stock-market valuation would follow from imple-menting best-practice, rather than worst-practice, corporate governance. Althoughthe differences in practices are not due todifferent institutions and although valuationis not the same as efficiency, these estimatesare suggestive of the improvements thatmight be stimulated by more effective insti-tutions.

Our analysis in section 4 also offers indi-rect evidence on the importance of corpo-rate governance institutions. The ownersmost dependent on institutional support arediffuse individual owners, outsiders wherethere are a number of different block-hold-ers, and perhaps even workers. Hence, thepattern of ownership effects in figure 2 isbroadly consistent with the argument thatcorporate governance institutions func-tioned less well in the CIS than elsewhere(John Coffee 1999). This interpretation isreinforced when one examines the amountof variation of ownership effects in the tworegions. There is much more variation inownership effects in the CIS, suggesting thatinstitutions have been more effective inEastern Europe in providing the support

780 Journal of Economic Literature, Vol. XL (September 2002)

Article 1 8/12/02 9:13 AM Page 780

Page 43: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

that is essential to some types of owners butnot needed by others.52

Competition policy was a strong focus ofearly institutional reforms (Paul Joskow,Richard Schmalensee, and NataliaTsukanova 1994). Our results in section 5certainly show that competition is important,but existing research does not allow one totake the extra step to conclude that competi-tion policy itself was a critical factor. MarkDutz and Maria Vagliasindi (2000) aggregatedata on individual firms to the country leveland show that competition policy led to anincrease in enterprise mobility. Brown andEarle (2001) show that competition-enhanc-ing policies, such as price decontrol and in-frastructure investment, increase the effectof competition on enterprise productivity inRussia. Again missing from this area of studyis research that directly estimates the link be-tween institutions and restructuring at thelevel of the enterprise.

This literature survey provides ample tes-timony to the fact that the evidence on theconnection between institutions and enter-prise restructuring is both scant and indi-rect.53 Evidently, if institutions are to de-serve the prominence in policy deliberationson enterprise restructuring that theypresently have, empirical work at the enter-prise level is a matter of great urgency.

8.3 Transition as Development

Over the last decade, the empirical re-search on enterprise restructuring in transi-tion has developed as a separate field, withonly sporadic cross-pollination with themore general literature on economic effi-

ciency, growth, and development. Two fac-tors explain this somewhat isolated stance.First, the research on transition partiallyemerged from the older comparative eco-nomics literature, which focused most of allon comparisons of socialism and capitalism.With the collapse of socialism and with capi-talism constituting the only viable target, the subsequent rapid transformation in theformer socialist economies led to a one-dimensional focus centering on the pathfrom plan to market, when analyzing policyand the associated changes at the enterpriselevel. Second, the very specific characteris-tics of socialist enterprises caused re-searchers to view restructuring in transitionas a singular task, rather separate from the is-sues of attaining corporate efficiency in mar-ket-capitalist economies. In section 2 we de-scribed a prototypical socialist enterpriseand the many dimensions of the enormouschanges that it faced in the 1990s. Many ofthese changes, e.g., selling shares for vouch-ers or the first contacts with foreign busi-nesses, were unparalleled in other econ-omies, both in character and in size.

In the years of transition, the former so-cialist countries have developed capitalisteconomies, albeit functioning quite differ-ently from each other and differently fromcapitalist economies that traversed otherhistorical paths. There is the odd countrystill clinging to a socialist-type economy, e.g.,Belarus, and Uzbekistan, but even in suchplaces there is privatization, entry of newprivate firms, import competition, and theestablishment of some institutions charac-teristic of capitalism. With the most basicfeatures of capitalism now established, thepolicy issues that China, Eastern Europe,and the CIS now encounter in their restruc-turing efforts are commonplace across theworld’s economies. For example, govern-ments face the task of producing more effec-tive corporate governance institutions andthe need to resist the political pressuresfrom industrial dinosaurs, while enterprisesconfront the challenges of entering new

Djankov and Murrell: Enterprise Restructuring in Transition 781

52 Grigorian (2000) suggests an interesting institu-tional reason for the different results for banks in theCIS and Eastern Europe. Commercial banks in severalEastern European countries were reluctant to takelarge equity positions in privatized enterprises, in orderto make their prudential practices consistent with thosein the European Union. This led to less ability to im-pose restructuring.

53 We should emphasize that we are restricting dis-cussion to the focus of this paper, empirical work ex-plaining restructuring in large enterprises.

Article 1 8/12/02 9:13 AM Page 781

Page 44: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

markets and shaping corporate structures toattract finance. In short, the process of tran-sition and the research on restructuring intransition have both matured to a pointwhere future research efforts can integrateeasily into a broader literature that studiesdifferences between capitalist economiesand between varieties of those institutionsthat support a vibrant corporate sector.

Such integration would greatly enrich thestudy of enterprise restructuring in transition,as it would bring to bear the findings of a vo-luminous literature on growth and develop-ment in capitalist economies. For example,this wider literature recognizes the fact thatdifferent capitalist economies evidence manydifferent ways of solving the institutionalproblem of promoting enterprise efficiency.But the institutional solutions are not always,nor perhaps often, the result of direct effortsto improve efficiency. Rather, they are theproduct of the efforts of powerful interestgroups that aim to enrich themselves. Hence,it is incorrect to consider a country’s policy re-forms and new institutions as efficient adap-tations to a country’s environment (Djankovet al. 2002b), as is quite often implicitly as-sumed in the literature that we have re-viewed. For example, the difference betweenthe effects of privatization and increasedcompetition in Eastern Europe and the CIS,documented in this paper, would be prof-itably studied from the perspective of the po-litical economy of the creation of institutions.

Scholars outside the field of transition willalso benefit from the integration of researchon restructuring in transition into the moregeneral study of institutions and economicdevelopment. As evidenced in this paper, thetransition experience provides multiple casestudies of reform, exhibiting great successesand spectacular failures. Scholars of policy re-form have much to learn from the study ofthe former socialist economies. The distinc-tive history of the transition economies,which still affects present structures, providesthe extra exogenous variation in independentvariables that is so important in the genera-

tion of incisive empirical results. But this vari-ation is only useful if it is one of degree ratherthan kind, meaning that lessons can be drawnfrom country comparisons. Now with transi-tion and transition research into its seconddecade, we conclude that this is the case.

9. Concluding Comments

We have surveyed the research on enter-prise restructuring in transition to synthesizeits main findings and to convey the evolvingstructure of this field of study. We find plenti-ful work examining the effects on restructur-ing of privatization, different types of newowners, competition, hardened budget con-straints, and policies on managers. Using amethodology that enables us to combine themany results on each of these determinants,we analyze the effects of each separately, con-trast the size of the various effects in EasternEurope and in the CIS, and compare the re-sults on different determinants to each other.

A capsule summary of the results appearsin the Introduction and no reiteration isneeded here. But we do leave the readerwith one last perspective, in figure 3. The fig-ure pulls together the comparable quantita-tive evidence on the effects of the main pol-icy reforms pursued by transition economiesin their efforts to restructure the enterprisesector. Although it is a somewhat hazardousexercise to compare partial correlation coef-ficients reflecting the effects of such dis-parate policies, such comparisons constitutethe only concrete evidence that is available tojudge which policies are most effective. Inour view, this figure aptly summarizes thefindings of the preceding sections.54

Privatization to outsiders is found to havethe largest positive effects on enterprise re-structuring, both in Eastern Europe and inthe CIS. Hardened budgets are also eco-nomically significant in explaining restruc-turing. Increased competition is associatedwith positive results in Eastern Europe but

782 Journal of Economic Literature, Vol. XL (September 2002)

54 Indeed, it is based on the partial correlation coef-ficients reported in figure 2, and tables 8b and 9b.

Article 1 8/12/02 9:13 AM Page 782

Page 45: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

not in the CIS. In fact, in the initial transitionperiod, import competition is detrimental toenterprise restructuring in Russia, Ukraine,Kazakhstan, and the other CIS countries.Finally, privatization to workers has not en-hanced restructuring in Eastern Europe andhas had negative effects in the CIS.

In section 8, we suggested directions forfuture research. Large rewards await the re-searchers who venture into those areas, espe-cially those who are able to isolate the effectsof institutions on enterprise restructuring,and who do so in a manner facilitating pre-cise comparisons between the effects of insti-tutional reform and the effects of the policiesexamined above.

Methodological AppendixThis appendix develops the methods introduced in

sections 3.2 and 4 and employed in sections 3–7. Thisappendix should be read in conjunction with sections3 and 4.

(a) Comparing Sets of Studies

For expositional purposes, begin with the simplestlinear model:

Y = α + γP + ε (A.1)

where all variables and parameters are as defined as inequation (1) of the text. Variances of the pertinent vari-ables (and their estimates) are denoted by σ2

Y, σ2P, and

σ2ε where σ2

Y = γ2σ2P + σ2

ε . The t-statistic correspondingto the kth study’s γ̂ for equation (A.1) is then:

(A.2)

where nk is degrees of freedom in the kth study. We as-sume that sample size is large relative to the number ofparameters estimated, so that it approximates degreesof freedom. This is a pragmatic assumption, sincemany studies do not indicate precisely how many pa-rameters are estimated, leaving degrees of freedomunknown.55

On inspection of equation (A.2), it is apparentthat the presence of sample size in the t-statisticrenders it inappropriate for cross-study comparisons

Djankov and Murrell: Enterprise Restructuring in Transition 783

0.10

Policy ReformEE

0.087

Part

ial C

orre

latio

n C

oeff

icie

nts

Figure 3. How Policy Reform Affects Enterprise Performance.Comparing Economic Effects in Eastern Europe to Those in the CIS

0.000

0.06

0.04

0.02

0.08

–0.02

–0.04

–0.06

–0.08

–0.10CIS

0.025

0.045

0.024

–0.002

0.029

–0.087

0.000

outsider privatization

hard budgets

domestic competition

import competition

privatization

to workers

0.094

–0.053

tk k k= γ̂ n 1/2 σ σPk εk

55 For example, studies often include industry dum-mies without stating the number of sectors.

Article 1 8/12/02 9:13 AM Page 783

Page 46: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

of economic size of effects (of ownership, competi-tion, etc.). But (A.2) shows that t-statistics have animportant property: they are invariant to changes inunits of Y or P. When comparing estimates across aheterogeneous collection of studies, we seek a sta-tistic that does not reflect sample size but is invari-ant to units.

The standard procedure in the meta-evaluation liter-ature is to use a statistic that is intermediate betweenthe t-statistic and γ̂ (Rosenthal 1984). This is the corre-lation coefficient, which is scale free and does not de-pend on sample size:

(A.3)

It is now simple to make an analogy to the case inwhich other variables (X) are present. We simply usepartial correlation coefficients, where the σYk and σPkthat appear in equation (A.3) are now the standard devi-ations of the errors in a regression of X on Y and thestandard deviation of the errors in a regression of X on P.Loosely speaking, the variables that are correlated arethose that capture variations in P and Y after P and Yhave been purged of any variations due to X: P and Ycontrolling for X. There is a similar adaptation in thet-statistic: equation (A.2) should also use the standarddeviations of P and Y controlling for X. Since partial cor-relation coefficients are usually not published, it is fortu-nate that (A.2) and (A.3) imply that there is a simple re-lation between t-statistics and the correspondingcorrelations:

(A.4)

where the formula applies equally to partial correla-tions (Greene 2000).56

When testing correlation coefficients, it is usual touse a transformation, “Fisher’s Zr”(ln[(1 + r)/(1 −r)]/2), which is approximately normally distributed. Itsvariance is [1/(n − 3)], where n is the sample size usedto calculate the correlation coefficient (WilliamShadish and Keith Haddock 1994). Tests on the meansof a set of correlations involve transforming each corre-lation and using the means of the transformations, alsonormally distributed. Moreover, weighting procedurescan be employed to find weighted-mean partial corre-lations and to conduct tests on the weighted means ofthe Zr transformations, so that one can investigate, justas in table 4, how sensitive are the results to such fac-tors as study quality.

(b) Combining the Results of Studies on the Effects of Different Types of Owners

Two new problems arise when combining the resultsof studies using (4), instead of (1). First, important in-formation is missing. Second, bilateral comparisons ofownership effects must be combined to produce a mul-tilateral comparison. We address each problem in turn.

(b.1) Missing Information on the Variance ofOwnership Effects

Here, there is no loss in generality in examining acase where the typical study estimates:

Y = α + Xβ + δO + θI + ε (A.5)

O and I are the amounts of ownership held by two dif-ferent owners, e.g., insiders and outsiders. (The gener-alization to more owners appears below.) Whatever isnot owned by these two is owned by a third entity, thestate, the omitted ownership share. δ and θ are the pa-rameters of interest and all other variables are as de-fined before. For estimates of (A.5) to be usable in thepresent context, it is necessary that O and I be mea-sured on the same scale within a single study (but notnecessarily across studies), so that δ and θ are compa-rable.

In papers estimating (A.5), the information of primeinterest is the comparison of δ to θ (outsiders versus in-siders) and each of these to 0 (insiders or outsiders versusthe state). Papers present the information to make thelatter comparison, in the usual listing of coefficients andt-statistics, which can be converted into partial correla-tion coefficients showing the effect of changing owner-ship from state to outsider or state to insider. But obtain-ing all information pertinent to the comparison of δ and θpresents some difficulties.

We require a t-statistic for a test of the null hypothe-sis that δ = θ. This statistic is a function of δ̂ and θ̂ andestimates of the variances of δ̂ and θ̂, all of which usu-ally appear in papers, plus an estimate of the covari-ance of δ̂ and θ̂, which is invariably omitted, since:

Variance (δ̂ − θ̂) = Variance (δ̂)

+ Variance (θ̂) − 2Covariance (δ̂ , θ̂) (A.6)

We sought a pragmatic method of estimating thevariance of (δ̂ − θ̂). O and I will almost always be nega-tively correlated because they are shares of ownership.Hence, the covariance of the estimates of the parame-ters attached to O and I will almost certainly be posi-tive. Therefore assuming the covariance equals zerowill lead to an overestimate of the variance of (δ̂ − θ̂).

We have verified this point in three ways. First, takea simple theoretical case. Assume that outsiders have100-percent ownership in a third of enterprises, insid-ers 100 percent of another third, and the state com-pletely owns the last third. Then, the covariance of δ̂and θ̂ is equal to one-half of the variance of either δ̂ orθ̂ . Second, we have used simulated ownership datawith five ownership types and have consistently foundin these simulations that the estimated covariance of δ̂and θ̂ is positive and at least half the size of the smallerof the variances of the two estimated parameters.57

Third, we have investigated the size of the variance of(δ̂ − θ̂) in two data sets, those used in Anderson, Lee,

784 Journal of Economic Literature, Vol. XL (September 2002)

57 Note that the comparisons between the relativesizes of variances and covariances does not depend oneither the size of errors or the data on Y. As in the the-oretical result, we assume that the equation does notcontain X.

56 The sign of the rk is obtained from the sign of theestimated γ.

rk k= =γ γ^kγ̂2 2σ σPk σPk σYkPk

2σ εk( )+

r k nk=2 t k2 t k

2( )+

Article 1 8/12/02 9:13 AM Page 784

Page 47: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

and Murrell (2000) and Claessens and Djankov(1999b). For those data sets (with various configura-tions of X’s and ownership types), we found that thestandard error of (δ̂ − θ̂) varied between 75 percentand 122 percent of the standard errors of the δ̂’s andthe θ̂’s.

Hence, the variance of (δ̂ − θ̂) will almost certainlylie in the interval between the sum of the variances of δ̂and θ̂ and the mean of the variances of δ̂ and θ̂. Thissuggests that we can take the t-statistics of δ̂ and θ̂, cal-culate the corresponding standard errors (S.E.’s) orvariances (var.’s), and then form a crude estimate of theS.E. of (δ̂ − θ̂):

If the first term within the braces on the right-handside of (A.7) applies, the estimated variance will lie be-tween the mean estimated variance of the individualparameters and the sum of the estimated variances ofthe individual parameters. The 1.25 factor is a conserva-tive adjustment, corresponding to the assumption thatthe covariance will usually be somewhat less than one-half of the mean of the two variances. (Alternatively, ifthe covariance were zero, which is unlikely, this factorwould need to be 21/2). The second term within thebraces is again a conservative adjustment, used whenthe variances of δ̂ and θ̂ are of quite different sizes(which is the case in only 25 percent of the analysesused in this paper). That element of (A.7) ensures that avery small variance does not influence the results tooheavily.

Obviously, the procedure that we apply is not first best,but first best is not possible given the available informa-tion. Without a pragmatic second-best procedure it wouldnot be possible to exploit the vast amount of informationin the literature on the effects of different types of own-ers. In formulating (A.7), we have used conservative as-sumptions to ensure that we do not overestimate differ-ences between owners. However, below, we do testwhether these assumptions are too conservative.

(b.2) Combining Many Bilateral Ownership Comparisons

The present subsection examines how to aggregateinformation on the effects of different types of ownerstaken from many studies. Since each study usually con-tains several types of owners, each contributes severaldifferent pair-wise comparisons of owners (i.e., observa-tions). The central variables of interest in each observa-tion are the t-statistics and sample sizes, the latter de-noted nijk. The t-statistics are obtained either directlyfrom the studies or by using formula (A.7). Applicationof (A.4) gives partial correlation coefficients, rijk, wherei,j =1, . . . ,11 (the number of categories of owners) andk = 1, . . . ,Kij, with Kij the number of studies that con-tribute information on the i,j-th ownership comparison.rijk estimates the effect of privatizing to ownership type jrather than to ownership type i.

For some comparisons, e.g., insiders versus managers,Kij is zero because studies using enterprise data in-variably make the sensible methodological decision notto use overlapping ownership categories within a singleregression. For many other ownership comparisons, Kijis quite small, since data on many types of owners (e.g.,banks, funds, etc) are not available for many countries.Therefore applying the methods of section 3 in astraightforward manner does not get us very far. Weseek a method that combines information from all datapoints when obtaining estimates of the effect of eachtype of owner.58 This is accomplished using a dummy-variable regression framework:

(A.8)

As we have information on only the relative perform-ance of owners, it is not possible to estimate all elevenpartial correlation coefficients (λm). Thus, we compareall owners to traditional state ownership and focus onλ̂

m − λ̂ 1 (m = 2,. . .11), the effect of switching from

traditional state ownership to owner m.The rijk reflect many observations and it is impor-

tant to use this fact in estimating (A.8). If the underly-ing data are normally distributed, then the variance ofrijk is approximately:

var (rijk| ijk) = (1 − r2ijk)2 / nijk (A.9)

Given that var(εijk) = var(rijk|ijk), generalized leastsquares can be employed to use this variance in-formation in estimating the λ̂

m − λ̂ 1 and their stan-

dard errors.59,60

For 30 percent of the observations, it is not neces-sary to apply (A.7) since the pertinent t-statistic ap-pears in the paper. We can therefore ask whether thereare any systematic differences between the estimatedownership effects found from these 30 percent of ob-servations and those found with the remainder of the

Djankov and Murrell: Enterprise Restructuring in Transition 785

,S.E. S.E. ( )̂δmax θ( )̂

=^ ^( () )+S.E. δ δ̂θ θ− max [ ] /2,1.25 Var. ( )̂Var.

(A.7)

r ijk ijkijk=

= −

λ ε+Σ11

m=1m

m

ijkm

D

D1 if m = i1 if m = j0 otherwise

58 For example, if there is information on owners Aand B versus owner C and information on A and B versusD, then the effect of C versus D can be estimated with-out having access to any study that matches C versus D.

59 Since we have estimates of the actual error vari-ances, our GLS procedure uses the assumption that theerror variances in the regression are equal to these esti-mates. The numerical results are easily obtained usingany standard GLS routine. However, the standard out-put must be reinterpreted to take into account the as-sumptions on error variances. See Larry Hedges(1994).

60 Since each data point is a bilateral comparison andsince the number of bilateral comparisons taken from apaper is greater than the number of owners in the pa-per, some underestimate of standard errors might oc-cur leading to some overestimate of the t-statistics intable 6. This argues for a small amount of discountingof the t-statistics in table 6.

(A.7)

Article 1 8/12/02 9:13 AM Page 785

Page 48: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

observations, which use (A.7). This question is ad-dressed by estimating the following equation usinggeneralized non-linear least squares:

(A.10)

D is a dummy variable equal to 1 when the observationon rijk was derived using (A.7). The estimate of η pro-vides the information on whether we have been tooconservative in formulating (A.7). In fact, we did findthis to be the case, η̂ implying that application of (A.7)leads to rijk that are 32 percent smaller than appropri-ate.61 When comparing the λ̂m − λ̂1 derived from (A.8)and those from (A.10), the latter leads to slightly largerdifferences in the effects of different owners thanthose found when estimating (A.8).62 This is to be ex-pected given that η̂ implies that (A.7) creates rijk thatare smaller than appropriate and that estimation of(A.10) compensates for this. However, the ordering ofthe effectiveness of the different owners is unchanged.Thus, we conclude that our pragmatic approach to fill-

ing in missing information is warranted and does notdistort the results. Given the information generated inestimating (A.10), we use the results from this formula-tion in the paper and in the remaining parts of thisappendix.

As in section 3, we classified papers into threegroups and generated corresponding weights reflectingthe papers’ attempts to counter the problem of selec-tion bias. Then in estimating (A.10), we applied theseweights to the observations. Thus, in this appendix wepresent two sets of estimates, ones that treat all studiesequally and ones that more than proportionately reflectthe results obtained in papers that pay greater atten-tion to solving the problem of selection bias.

The two sets of estimates appear in figure A.1. Thereis a large amount of consistency between them, sug-gesting that the overall interpretation of the results isnot vitiated by the possibility of selection bias. The cor-relation between the two sets of estimates is 0.70.Nevertheless, the results for some owners, especiallymanagers and workers, do change substantially withthe selection bias correction. With the estimates forworkers and managers removed, the correlation be-tween the two sets of estimates is 0.96. These facts pro-vide the basis for using the selection-bias corrected es-timates in section 4 and for general confidence in theseestimates, except for some residual doubt about man-agers and workers.

REFERENCES

Abdelati, Wafa and Stijn Claessens. 1996. “Losses,Financing, and Adjustment of State Enterprises inRomania,” mimeo, World Bank.

786 Journal of Economic Literature, Vol. XL (September 2002)

61 Formula (A.7) contains two elements, to be ap-plied in different situations. We also tested whetherthese two elements led to systematically different bi-ases in the estimated partial correlation coefficients, us-ing a method analogous to that in (A.10). The testsstrongly endorse the hypothesis that the two biases areequal.

62 The relevant details are in Djankov and Murrell(2000).

0.100

Category of Owners

Basic Results

–0.002

0.004

Part

ial C

orre

latio

n C

oeff

icie

nts

Figure A.1. Judging the Effects of Selection Bias on the Results

0.000

0.040

0.020

0.060

0.080

–0.020

–0.040

0.006

0.025

0.012

–0.018

0.043

0.0350.040

0.011

Results corrected for selection biases

0.040

0.049

0.079

0.0420.048

0.008

foreign

insiders

banksworkers

diffuse individual

outsiders

commercialized

state

managers

block-holders

investment funds

0.017 0.017 0.017

0.057

r ijk ijkijk= λ ε+Σ η11

m=1m

mD ijkλ +Σ11

m=1m

mDD

Article 1 8/12/02 9:13 AM Page 786

Page 49: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

Acemoglu, Daron; Simon Johnson and James Rob-inson. 2001. “The Colonial Origins of ComparativeDevelopment: An Empirical Investigation,” Amer.Econ. Rev. 91:5, pp. 1369–401.

Aghion, Philippe and Olivier Blanchard. 1996. “OnInsider Privatization,” Eur. Econ. Rev. 40:5, pp.759–66.

Aghion, Philippe; Olivier Blanchard and RobinBurgess. 1994. “The Behavior of State Firms inEastern Europe: Pre-Privatization,” Eur. Econ. Rev.38:2, pp. 1327–49.

Aghion, Phillipe and Mark Schankerman. 1999.“Competition, Entry, and the Social Returns toInfrastructure in Transition Economies,” Econ.Tran. 7:1, pp. 79–101.

Alfandari, Gilles; Qimiao Fan and Lev Freinkman.1996. “Government Financial Transfers to IndustrialEnterprises and Restructuring,” in EnterpriseRestructuring and Economic Policy in Russia. SimonCommander, Qimiao Fan and Mark Schaffer, eds.Economic Development Institute (EDI) De-velopment Studies, Washington, DC: World Bank,pp. 166–202.

Anderson, James H.; Georges Korsun and PeterMurrell. 1999. “Ownership, Exit, and Voice afterMass Privatization: Evidence from Mongolia,” Econ.Tran. 7:1, pp. 215–43.

———. 2000. “Which Enterprises (Believe They)Have Soft Budgets after Mass Privatization?Evidence from Mongolia,” J. Compar. Econ. 28:2,pp. 219–46.

Anderson, James H., Young Lee and Peter Murrell.2000. “Competition and Privatization amidst WeakInstitutions: Evidence from Mongolia,” Econ. Inq.38:4, pp. 527–49.

Angelucci, Manuella; Saul Estrin, Jozef Konings andZbigniew Zolkiewski. 2001. “The Effect ofOwnership and Competitive Pressure on FirmPerformance in Transition Countries: MicroEvidence from Bulgaria, Romania, and Poland,” dis-cus. paper 2985, CEPR, London.

Balcerowicz, Leszek. 1995. Socialism, Capitalism,Transformation. Budapest: Central European U.Press.

Barberis, Nicholas; Maxim Boycko, Andrei Shleiferand Natalia Tsukanova. 1996. “How DoesPrivatization Work? Evidence from the RussianShops,” J. Pol. Econ. 104:4, pp. 764–90.

Berle, Adolf and Gardiner Means. 1933. The ModernCorporation and Private Property. NY: MacMillan.

Berliner, Joseph. 1976. The Innovation Decision inSoviet Industry. Cambridge, MA: MIT Press.

Black, Bernard. 2001. “Does Corporate GovernanceMatter? A Crude Test Using Russian Data,” U.Penn. Law Rev. 149:2, pp. 2131–50.

Black, Bernard; Reinier Kraakman and AnnaTarassova. 2000. “Russian Privatization andCorporate Governance: What Went Wrong?” Stan.Law Rev. 52:7, pp. 1731–808.

Blanchard, Olivier and Michael Kremer. 1997.“Disorganization,” Quart. J. Econ. 112:4, pp. 1091–126.

Blanchard, Olivier and Andrei Shleifer. 2000. “Goodand Bad Federalism: China versus Russia,” work.paper, econ. dept., Harvard U.

Blanchflower, David and Stephen Machin. 1996.“Product Market Competition, Wages, and Produc-tivity,” Ann. Econ. Statist. 17:41, pp. 219–53.

Blasi, Joseph and Andrei Shleifer. 1996. “CorporateGovernance in Russia: An Initial Look,” in CorporateGovernance in Central Europe and Russia. Vol. 2.Roman Frydman, Cheryl Gray and Andrej Rapaczynski,eds. Budapest: Central European U. Press.

Blinder, Alan S. ed. 1990. Paying for Productivity: ALook at the Evidence. Center Econ. ProgressEmployment series, Washington, DC: BrookingsInst., pp. 15–88.

Bonin, John P.; Derek Jones and Louis Putterman.1993. “Theoretical and Empirical Studies ofProducer Cooperatives: Will Ever the Twain Meet?”J. Econ. Lit. 31:3, pp. 1290–320.

Boycko, Maxim, Andrei Shleifer and Robert Vishny.1995. Privatizing Russia. Cambridge, MA: MIT Press.

———. 1996. “A Theory of Privatisation,” Econ. J.106:435, pp. 309–19.

Brada, Josef. 1996. “Privatization Is Transition: Or IsIt?” J. Econ. Perspect. 10:2, pp. 67–86.

Broadman, Harry and Geng Xiao. 1997. “TheCoincidence of Material Incentives and MoralHazard in Chinese Enterprises,” work. paper 128,HIID, Harvard U.

Brown, Annette and David Brown. 1999. “DoesMarket Structure Matter? New Evidence fromRussia,” discus. paper 1946, CEPR, London.

Brown, David and John Earle. 2000. “Privatization andEnterprise Restructuring in Russia: New Evidencefrom Panel Data on Industrial Enterprises,” work.paper 1, RECEP, Moscow.

———. 2001. “Competition Enhancing Policies andInfrastructure: Evidence from Russia,” discus. paper3022, CEPR, London.

Carlin, Wendy; Steven Fries, Mark Schaffer and PaulSeabright. 2001. “Competition and EnterprisePerformance in Transition Economies: Evidencefrom a Cross-Country Survey,” discus. paper 2840,CEPR, London.

Chang, Chun; Brian McCall and Yijiang Wang. 2001.“Incentive Contracting vs Ownership Reforms:Evidence from China’s Township and VillageEnterprises,” work. paper 14, Carlson SchoolManagement, U. Minnesota.

China Daily. 2001. “Two Officials Sentenced to Deathfor Corruption amid China’s Resolve to EliminateFraud,” Oct. 11, A6.

Claessens, Stijn and Simeon Djankov. 1998.“Politicians and Firms in Seven Central and EasternEuropean Countries,” Policy Research work. paper1954, World Bank.

———. 1999a. “Ownership Concentration andCorporate Performance in the Czech Republic,” J.Compar. Econ. 27:3, pp. 498–513.

———. 1999b. “Enterprise Performance andManagement Turnover in the Czech Republic,” Eur.Econ. Rev. 43: 4–6, pp. 1115–24.

———. 2000. “Government Regulation and Pri-vatization Benefits in Eastern Europe,” work. paper,World Bank.

———. 2002. “Privatization Benefits in EasternEurope,” J. Pub. Econ. 83:3, pp. 307–24.

Djankov and Murrell: Enterprise Restructuring in Transition 787

Article 1 8/12/02 9:13 AM Page 787

Page 50: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

Claessens, Stijn; Simeon Djankov and DanielaKlingebiel. 2000. “Stock Markets in TransitionEconomies,” SSRN work. paper, http://papers.ssrn.com/sol3/papers.cfm?abstractsub id=240703.

Claessens, Stijn; Simeon Djankov and Gerhard Pohl.1997. “Ownership and Corporate Governance:Evidence from the Czech Republic,” work. paper,World Bank.

Claessens, Stijn and Kyle Peters. 1997. “State EnterprisePerformance and Soft Budget Constraints: The Caseof Bulgaria” Econ. Tran. 5:2, pp. 305–22.

Clague, Christopher and Gordon C. Rausser, eds.1992. Emergence of Market Economies in EasternEurope. London and NY: Blackwell Publishers.

Clement, Cynthia and Peter Murrell. 2001. “Assessingthe Value of Law in the Economic Transition fromSocialism: An Introduction,” in Assessing the Valueof Law in Transition Economies. Peter Murrell, ed.Ann Arbor: U. Michigan Press.

Coase, Ronald. 1988. “The Theory of the Firm,” in TheFirm, The Market, and the Law. Ronald Coase, ed.Chicago: U. Chicago Press.

Coffee, John, Jr. 1999. “Privatization and CorporateGovernance: The Lessons from Securities MarketFailure,” J. Corp. Law 25:2, pp. 1–39.

Coricelli, Fabrizio and Simeon Djankov. 2001.“Hardened Budgets and Enterprise Restructuring:Theory and an Application to Romania,” J. Compar.Econ. 29:4, pp. 749–63.

Cull, Robert; Jana Matesova and Mary Shirley. 2002.“Ownership and the Temptation to Loot: Evidencefrom Privatized Firms in the Czech Republic,” J.Compar. Econ. 30:1, pp. 1–25.

Demsetz, Harold and Kenneth Lehn. 1985. “TheStructure of Ownership: Causes and Con-sequences,” J. Polit. Econ. 93:6, pp. 1155–77.

Dewatripont, Michel and Eric Maskin. 1995. “Creditand Efficiency in Centralized and DecentralizedEconomies,” Rev. Econ. Stud. 62:4, pp. 541–55.

Djankov, Simeon. 1999a. “Restructuring of Insider-Dominated Firms,” Econ. Tran. 7:2, pp. 467–79.

———. 1999b. “The Enterprise Isolation Program inRomania,” J. Compar. Econ. 27:2, pp. 281–93.

———. 1999c. “Ownership Structure and EnterpriseRestructuring in Six Newly Independent States,”Compar. Econ. Stud. 41:1, pp. 75–95.

Djankov, Simeon and Caroline Freund. 2002. “TradeFlows in the Former Soviet Union, 1987–96,” J.Compar. Econ, 30:1, pp. 76–90.

Djankov, Simeon and Bernard Hoekman. 1998.“Conditions of Competition and MultilateralSurveillance,” World Econ. 21:8, pp. 1109–28.

———. 2000. “Market Discipline and CorporateEfficiency: Evidence from Bulgaria,” Can. J. Econ.33:1, pp. 190–212.

Djankov, Simeon and Peter Murrell. 2000. “EnterpriseRestructuring in Transition: A Quantitative Survey,”work. paper, U. Maryland College Park. http://www.bsos.umd.edu/econ/murrell.

Djankov, Simeon and Tatiana Nenova. 2000. “Why DidPrivatization Fail in Kazakhstan?” work. paper,World Bank.

Djankov, Simeon; Rafael La Porta, Florencio Lopez-de-Silanes and Andrei Shleifer. 2002a. “The

Regulation of Entry,” Quart. J. Econ. 117:1, pp.1–37.

———. 2002b. “The New Comparative Economics,”work. paper, econ. dept. Harvard U.

Dobrinsky, Rumen; Nasko Dochev and NikolayMarkov. 2001. “Productivity and the Sources of FirmLevel Efficiency in Bulgarian Manufacturing,” workpaper, Center Econ. Strategic Research, Sofia.

Dutz, Mark and Maria Vagliasindi. 2000. “Competitionpolicy implementation in transition economies: AnEmpirical Assessment” Eur. Econ. Rev. 44:3, pp.762–72.

Earle, John. 1998. “Post-privatization Ownership andProductivity in Russian Industrial Enterprises,”Working Paper 127, Stockholm Institute forTransition Economies, U. Stockholm.

Earle, John and Almos Telgedy. 2001. “Productivityand Ownership Structure in Romania: DoesPrivatization Matter?” Working paper, CentralEuropean U., Budapest.

Earle, John and Saul Estrin. 1997. “After VoucherPrivatization: The Structure of Corporate Own-ership in Russian Manufacturing Industry,” Workingpaper, London Business School.

———. 1998. “Privatization, Competition and BudgetConstraints: Disciplining Enterprises in Russia,”Discus. Pap. 128, Stockholm Institute for TransitionEconomies, U. Stockholm.

Earle, John; Saul Estrin and Larisa Leshchenko. 1996.“Ownership Structures, Patterns of Control, andEnterprise Behavior in Russia” in EnterpriseRestructuring and Economic Policy in Russia,Commander, Simon, Qimiao Fan and Mark Schaffer,eds., EDI Development Studies, World Bank,Washington, DC.

Earle, John and Richard Rose. 1997. “OwnershipTransformation, Economic Behavior, and PoliticalAttitudes in Russia,” Discus. Pap. 83, StockholmInstitute for Transition Economies, U. Stockholm.

Earle, John and Klara Sabirianova. 1999. “Under-standing Wage Arrears in Russia,” discus. paper173, Stockholm Institute for Transition Economies,U. Stockholm.

EBRD. 1999. Ten Years of Transition: TransitionReport. London: EBRD.

Estrin, Saul and Adam Rosevear. 1999a. “EnterprisePerformance and Ownership: The Case of Ukraine,”Eur. Econ. Rev. 43:4, pp. 1125–36.

———. 1999b. “Enterprise Performance andCorporate Governance in Ukraine,” J. Compar.Econ. 27:3, pp. 442–58.

Evans-Klock, Christine and Alexander Samorodov.1998. “The Employment Impact of Privatization andEnterprise Restructuring in Selected TransitionEconomies,” IPPRED-16, Enterprise and Co-operative Development, ILO, Geneva.

Frydman, Roman; Cheryl Gray, Marek Hessel andAndrej Rapaczynski. 1999. “When Does PrivatizationWork? The Impact of Private Ownership onCorporate Performance in the Transition Eco-nomies,” Quart. J. Econ. 114:4, pp. 1153–91.

———. 2000. “The Limits of Discipline: Ownershipand Hard Budget Constraints in the TransitionEconomies,” Econ. Tran. 8:3, pp. 577–601.

788 Journal of Economic Literature, Vol. XL (September 2002)

Article 1 8/12/02 9:13 AM Page 788

Page 51: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

Frydman, Roman; Marek Hessel and AndrejRapaczynski. 2000. “Why Ownership Matters?Entrepreneurship and the Restructuring ofEnterprises in Central Europe,” work. paper, econ.dept. NYU.

Frye, Timothy. 2001. “Keeping Shop: The Value of theRule of Law in Warsaw and Moscow,” in Assessingthe Value of Law in Transition Economies. PeterMurrell ed. Ann Arbor: U. Michigan Press.

Gao, Shumei and Mark Schaffer. 1998. “FinancialDiscipline in the Enterprise Sector in TransitionCountries: How Does China Compare?” discus. pa-per 1, Center Econ. Research Transition, Harriot-Watt U., Edinburgh.

Glennerster, Rachel. 2000. “Evaluating Privatization inthe Former Yugoslav Republic of Macedonia,” work.paper, Kennedy School, Harvard U.

Gordon, Roger and Wei Li. 1991. “Chinese EnterpriseBehavior under the Reforms,” Amer. Econ. Rev.81:2, pp. 202–206.

———. 1995. “The Change in Productivity of ChineseState Enterprises, 1983–1987,” J. Prod. Anal. 6:1,pp. 5–26.

Granick, David. 1987. Job Rights in the Soviet Union:Their Consequences. NY: Columbia U. Press.

Greene, William. 2000. Econometric Analysis. UpperSaddle River, NJ: Prentice Hall.

Grigorian, David. 2000. “Ownership and Performanceof Lithuanian Enterprises,” Policy Research work.paper 2343, World Bank.

Grosfeld, Irena and Jean Francois Nivet. 1999.“Insider Power and Wage Setting in Transition:Evidence from a Panel of Large Polish Firms,” Eur.Econ. Rev. 43:4, pp. 1137–47.

Grosfeld, Irena and Thierry Tressel. 2001. “Com-petition and Corporate Governance: Substitutes orComplements? Evidence from the Warsaw StockExchange,” work. paper, Delta Center, Paris.

Groves, Theodore; Yongmiao Hong, John McMillanand Barry Naughton. 1994. “Autonomy andIncentives in Chinese State Enterprises,” Quart. J.Econ. 109:1, pp. 183–209.

———. 1995. “China’s Evolving Managerial LaborMarket,” J. Polit. Econ. 103:4, pp. 873–92.

Hall, Robert and Charles Jones. 1999. “Why Do SomeCountries Produce So Much More Output per Workerthan Others?” Quart. J. Econ. 114:1, pp. 83–116.

Halpern, Laszlo and Gabor Korosi. 1998. “CorporateStructure and Performance in Hungary,” DavidsonInstitute Working Paper 187, U. Michigan BusinessSchool, Ann Arbor, Michigan.

Harrison, Ann. 1994. “Productivity, ImperfectCompetition and Trade Reform,” J. Int. Econ.36:1–2, pp. 53–73.

Havrylyshyn Oleh and Donal McGettigan. 2000.“Privatization in Transition Countries,” Post-Sov.Affairs 16:3, pp. 257–86.

Hedges, Larry. 1994. “Fixed Effects Models,” in TheHandbook of Research Synthesis. Harris Cooper andLarry V. Hedges, eds. NY: Russell Sage, pp. 285–300.

Hellman, Joel; Geraint Jones and Daniel Kaufmann.2000. “Seize the State, Seize the Day: State Capture,Corruption, and Influence in Transition,” PolicyResearch work. paper 2444, World Bank.

Hendley, Kathryn; Peter Murrell and Randi Ryterman.2001. “Law Works in Russia: The Role of Law inInterenterprise Transactions,” in Assessing the Value ofLaw in Transition Economies. Peter Murrell, ed. AnnArbor: U. Michigan Press.

Hersch, Philip; David Kemme and Sharad Bhandari.1994. “Competition in Transition Economies:Determinants of Price-Cost Margins in PrivateSector Manufacturing in Eastern Europe,” South.Econ. J. 61:2, pp. 356–66.

Holmstrom, Bengt. 1979. “Moral Hazard andObservability,” Bell J. Econ. 10:1, pp. 74–91.

———. 1982. “Moral Hazard in Teams,” Bell J. Econ.13:2, pp. 324–40.

Horn, Henrik; Harold Lang and Stefan Lundgren. 1995.“Managerial Effort Incentives, X-Inefficiency andInternational Trade,” Eur. Econ. Rev. 39:1, pp. 117–38.

Huang, Yiping and K. Kalirajan. 1998. “EnterpriseReform and Technical Efficiency of China’s State-Owned Enterprises,” Appl. Econ. 30:5, pp. 585–92.

Hunt, Morton. 1997. How Science Takes Stock. NY:Russell Sage.

Hunter, John and Frank Schmidt. 1990. Methods ofMeta-analysis: Correcting Error and Bias inResearch Findings, Newbury Park, CA.; London andNew Delhi: Sage.

Ickes, Barry; Randi Ryterman and Stoyan Tenev. 1995.“On Your Marx, Get Set, Go: The Role ofCompetition in Enterprise Adjustment,” work. paper, World Bank.

Januszewski, Silke; Jens Koke and Joakim Winter.2001. “Product Market Competition, CorporateGovernance, and Firm Performance: An EmpiricalAnalysis of Germany, “ work. paper, Center Europ.Econ. Research, Mannheim.

Jensen, Michael and William Meckling. 1976. “Theoryof the Firm: Managerial Behavior, Agency Costs,and Ownership Structure,” J. Finan. Econ. 3:4, pp.305–60.

Johnson, Bruce; Robert Magee, Nandu Nagarajan andHarry Newman. 1985. “An Analysis of the StockPrice Reaction to Sudden Executive Deaths:Implications for the Managerial Labor Market,” J.Acc. Econ. 7:2, pp. 151–74.

Johnson, Simon; Daniel Kaufmann and Andrei Shleifer.1997. “The Unofficial Economy in Transition,”Brookings Pap. Econ. Act. 0:2, pp. 159–221.

Johnson, Simon; John McMillan and ChristopherWoodruff. 2002. “Courts and Relational Contracts,”J. Law, Econ., Organ. 18:1, pp. 221–77.

Jones, Derek. 1998. “The Economic Effects ofPrivatization: Evidence from a Russian Panel,”Compar. Econ. Stud. 40:2, pp. 75–102.

Jones, Derek and Takao Kato. 1998. “Chief ExecutiveCompensation During Early Transition: FurtherEvidence from Bulgaria,” Davidson Institute work.paper 146, U. Michigan Business School, Ann Arbor.

Jones, Derek and Niels Mygind. 2000. “The Effects ofPrivatization upon Productive Efficiency: Evidencefrom the Baltic Republics,” Ann. Public Coop. Econ.71:3, pp. 415–39.

———. 2002. “Ownership and Productive Efficiency:Evidence from Estonian Panel Data,” Rev. Devel.Econ. 6:2, pp. 284–301.

Djankov and Murrell: Enterprise Restructuring in Transition 789

Article 1 8/12/02 9:13 AM Page 789

Page 52: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

Jones, Derek; Mark Klinedinst and Charles Rock.1998. “Productive Efficiency During Transition:Evidence from Bulgarian Panel Data,” J. Compar.Econ. 26:3, pp. 446–64.

Joskow, Paul; Richard Schmalensee and NataliaTsukanova. 1994. “Competition Policy in RussiaDuring and After Privatization,” Brookings Pap.Econ. Act. 24:1, pp. 301–81.

Konings, Jozef. 1997. “Competition and Firm Performancein Transition Economies: Evidence from Firm-LevelSurveys in Slovenia, Hungary, and Romania” Discus.Pap. 1170, Center Econ. Research Transition, Harriot-Watt U., Edinburgh.

———. 1998. “Firm Performance in Bulgaria andEstonia: The Effects of Competitive Pressure,Financial Pressure and Disorganization,” DavisonInstitute work. paper 185, U. Michigan BusinessSchool, Ann Arbor.

Konings, Jozef and Patrick Paul Walsh. 1998.“Disorganization in the Transition Process: Firm-LevelEvidence from Ukraine,” Discus. Pap. 1928, CEPR,London.

Kornai, Janos. 1979. “Resource-Constrained vs Demand-Constrained Systems,” Econometrica 4:4, pp. 128–64.

———. 1980. “‘Hard’ and ‘Soft’ Budget Constraints,” ActaOecon. 25:3–4, pp. 231–45.

———. 1986. “The Hungarian Reform Process: Visions,Hopes, and Reality,” J. Econ. Lit. 24:4, pp. 1687–737.

———. 1990. The Road to a Free Economy. NY: Norton.———. 1998. “The Place of the Soft Budget Constraint

Syndrome in Economic Theory,” J. Compar. Econ. 26:1,pp. 11–17.

Korsun, Georges and Peter Murrell. 1995. “The Politics andEconomics of Mongolia’s Privatization Program,” AsianSurvey 35:5, pp. 472–86.

Kreacic, Vladimir-Goran. 1998. “Restructuring ofManufacturing Firms in Georgia: Four Case Studies and aSurvey,” work. paper, World Bank.

Kruse, Douglas and Joseph Blasi. 1995. “EmployeeOwnership, Employee Attitudes, and Firm Performance,”work. paper 5277, NBER, Cambridge, MA.

Kuznetsov, Boris; Giovanni Mangiarotti and Mark Schaffer.2002. “Insider Ownership and Economic Performance ofRussian Firms,” work. paper, Center Econ. ResearchTransition, Harriot-Watt U., Edinburgh.

La Porta, Rafael; Florencio Lopez-de-Silanes and AndreiShleifer. 1999. “Corporate Ownership around the World,”J. Fin. 54:2, pp. 471–517.

———. 2002. “Government Ownership of Banks,” J. Fin.57:1 pp. 265–301.

Lee, Keun. 1990. “The Chinese Model of theSocialist Enterprise: An Assessment of ItsOrganization and Performance,” J. Compar. Econ.14:3, pp. 384–400.

Lee, Young. 1999. “Wages and Employment in China’sSOEs, 1980–1994: Corporatization, Market Dev-elopment and Insider Forces,” J. Compar. Econ.27:4, pp. 702–29.

Lehmann, Hartmut; Jonathan Wadsworth andAlessandro Acquisti. 1999. “Grime and Pun-ishment: Job Insecurity and Wage Arrears in theRussian Federation,” J. Compar. Econ. 27:4, pp.595–617.

Li, David and Minsong Liang. 1998. “Causes of theSoft Budget Constraint: Evidence on ThreeExplanations,” J. Compar. Econ. 26:1, pp. 104–16.

Li, Huagang. 1999. “State Factories in Transition—Openness, Competition, and Productivity,” J. Devel.Econ. 58:2, pp. 429–62.

Li, Wei. 1997. “The Impact of Economic Reform onthe Performance of Chinese State Enterprises,1980–1989,” J. Polit. Econ. 105:5, pp. 1080–106.

Linz, Susan and Gary Krueger. 1998. “EnterpriseRestructuring in Russia’s Transition Economy:Formal and Informal Mechanisms,” Compar. Econ.Stud. 40:2, pp. 5–52.

Lizal, Lubomir; Miroslav Singer and Jan Svejnar. 2001.“Enterprise Breakups and Performance During theTransition from Plan to Market,” Rev. Econ. Statist.83:1, pp. 92–99.

Major, Ivan. 1999. “The Transforming Enterprise:Company Performance after Privatization inHungary between 1988–1997,” Compar. Econ. Stud.41:2–3, pp. 61–110.

Markov, Nikolay; Boyko Nikolov and RumenDobrinsky. 2000. “Competitive Pressure, Efficiency,and Mark-up Pricing in Bulgarian Manufacturing,”work. paper 21, Center Econ. Strategic Research,Sofia.

Maskin, Eric. 1999. “Recent Theoretical Work on theSoft-Budget Constraint,” Amer. Econ. Rev. 89:2, pp.421–25.

McMillan, John. 1997. “Markets in Transition,” inAdvances in Economics and Econometrics: Theoryand Applications, David Kreps and Kenneth Wallis,eds. Cambridge, UK: Cambridge U. Press.

McMillan, John and Barry Naughton. 1992. “How toReform a Planned Economy: Lessons from China,”Oxford Rev. Econ. Pol. 8:1, pp. 130–43.

McMillan, John; John Whalley and Lijing Zhu. 1989.“The Impact of China’s Economic Reforms onAgricultural Productivity Growth,” J. Polit. Econ.97:4, pp. 781–807.

McMillan, John and Christopher Woodruff. 1999.“Interfirm Relationships and Informal Credit inVietnam,” Quart. J. Econ. 114:6, pp. 1285–320.

Megginson, William and Jeffry Netter. 2001. “FromState to Market: A Survey of Empirical Studies onPrivatization,” J. Econ. Lit. 39:2, pp. 321–89.

Morck, Randall; Andrei Shleifer and Robert Vishny.1988. “Management Ownership and MarketValuation: An Empirical Analysis,” J. Finan. Econ.20:1–2, pp. 293–315.

Murrell, Peter. 1990. The Nature of SocialistEconomies: Lessons from Eastern European ForeignTrade. Princeton: Princeton U. Press.

———. 1992. “Evolution in Economics and in theEconomic Reform of the Centrally PlannedEconomies,” in Emergence of Market Economiesin Eastern Europe. Christopher Clague andGordon Rausser, eds. London and NY: BlackwellPublishers.

———. 1995. “The Transition According toCambridge, Mass.,” J. Econ. Lit. 33:2, pp. 164–78.

Nalebuff, Barry and Joseph Stiglitz. 1983. “Prizes andIncentives: Towards a General Theory of Com-

790 Journal of Economic Literature, Vol. XL (September 2002)

Article 1 8/12/02 9:13 AM Page 790

Page 53: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

pensation and Competition,” Bell J. Econ. 14:1, pp.21–43.

Naughton, Barry. 1994. “Chinese InstitutionalInnovation and Privatization from Below,” Amer.Econ. Rev. 84:2, pp. 266–70.

Nellis, John. 1999. “Time to Rethink Privatization?”IFC discus. paper 38, World Bank.

Neumark, David and William Wascher. 1998. “Is theTime-Series Evidence on Minimum Wage EffectsContaminated by Publication Bias?” Econ. Inq. 36:3,pp. 458–70.

Nickell, Stephen. 1996. “Competition and CorporatePerformance,” J. Pol. Econ. 104:4, pp. 724–96.

Nickell, Stephen; Sushil Wadhwani and Martin Wall.1992. “Productivity Growth in UK Companies,1975–86,” Eur. Econ. Rev. 36:5, pp. 1066–85.

North, Douglass. 1990. Institutions, InstitutionalChange, and Economic Performance, Cambridge U.Press, Cambridge, MA.

Perevalov, Yuri; Ilya Gimadi and Vladimir Dobrodey.2000. “Analysis of the Impact of Privatization on thePerformance of Medium- and Large-size IndustrialEnterprises in Russia,” work. paper 2K, Econ.Education and Research Consortium, Moscow.

Phillips, Joseph and Ernest Goss. 1995. “The Effect ofState and Local Taxes on Economic Development: AMeta-Analysis,” South. Econ. J. 62:2, pp. 320–33.

Pinto, Brian; Marek Belka and Stefan Krajewski. 1993.“Transforming State Enterprises in Poland:Evidence on Adjustment by Manufacturing Firms”Brook. Pap. Econ. Act. 23:1, pp. 213–61.

Pohl, Gerhard; Robert Anderson, Stijn Claessens andSimeon Djankov. 1997. “Privatization andRestructuring in Central and Eastern Europe:Evidence and Policy Options,” technical paper 368,World Bank.

Qian, Yingyi. 2000. “The Process of China’s MarketTransition (1978–1999): The Evolutionary, His-torical, and Comparative Perspectives,” J. Inst. The-oretical Econ. 156:1, pp. 151–71.

Qian, Yingyi and Gérard Roland. 1996. “The SoftBudget Constraint in China,” Japan World Econ.8:2, pp. 207–23.

Qingtai, Chen. 2001. “Improved Corporate Gov-ernance in China,” speech at Corporate GovernanceConference, Beijing, Sept.

Recanatini, Francesca and Randi Ryterman. 2000.“Transition and Output Collapse: Measuring theImpact of Institutional Change,” work. paper, WorldBank.

Roberts, Bryan; Yevgeny Gorkov and Jay Madigan.1999. “Is Privatization a Free Lunch? New Evidenceon Ownership Status and Firm Performance,” work.paper, U. Miami.

Rodriguez, Francisco, and Dani Rodrik. 1999. “TradePolicy and Economic Growth: A Skeptic’s Guide tothe Cross-National Evidence,” work. paper 7081,NBER, Cambridge, MA.

Roe, Robert A.; Irina Zinovieva, Elizabeth Dienes andLaurens ten Horn. 1998. “Firm Ownership andWork Motivation in Bulgaria and Hungary: AnEmpirical Study of the Transition in the Mid-1990s,”Davidson Institute work. paper 205, U. MichiganBusiness School, Ann Arbor.

Roland, Gerard. 1994. “On the Speed and Sequencingof Privatization and Restructuring,” Econ. J. 104:426, pp. 1158–68.

Roland, Gerard and Khalid Sekkat. 2000. “ManagerialCareer Concerns, Privatization and Restructuring inTransition Economies,” Eur. Econ. Rev. 44:10, pp.1857–72.

Roll, Richard. 1986. “The Hubris Hypothesis ofCorporate Takeovers,” J. Bus. 59:2, pp. 197–216.

Rosen, Sherwin. 1992. “Contracts and the Market forExecutives,” in Contract Economics, Lars Werin andHans Wijkander, eds., Oxford U. Press, Oxford, UK.

Rosenthal, Robert. 1984. Meta-Analytic Procedures forSocial Research, Sage: Beverly Hills.

Sachs, Jeffrey. 1992. “Privatization in Russia: SomeLessons from Eastern Europe,” Amer. Econ. Rev.Pap. Proc. 82:2, pp. 43–48.

Schaffer, Mark. 1998. “Do Firms in TransitionEconomies Have Soft Budget Constraints? AReconsideration of Concepts and Evidence,” J.Compar. Econ. 26:1, pp. 80–103.

Schmidt, Klaus. 1997. “Managerial Incentives andProduct Market Competition,” Rev. Econ. Stud.64:2, pp. 191–213.

Shadish, William and Keith Haddock. 1994.“Combining Estimates of Effect Size,” in TheHandbook of Social Researc. Harris Cooper andLarry V. Hedges eds. NY: Russell Sage, pp.261–84.

Shirley, Mary and Lixin Colin Xu. 2001. “EmpiricalEvidence of Performance Contracts: Evidence fromChina,” J. Law, Econ., Organ. 17:1, pp. 168–200.

Shleifer, Andrei and Daniel Treisman. 2000. Without aMap: Political Tactics and Economic Reform inRussia. Cambridge, MA: MIT Press.

Shleifer, Andrei and Robert Vishny. 1986. “LargeShareholders and Corporate Control,” J. Polit. Econ.94:3, pp. 461–88.

———. 1994. “Politicians and Firms,” Quart. J. Econ.109:4, pp. 995–1025.

Smith, Stephen; Beom-Cheol Cin and Milan Vodopivec.1997. “Privatization Incidence, Ownership Forms, andFirm Performance: Evidence from Slovenia,” J.Compar. Econ. 25:2, pp. 158–79.

Smith, Kerry and Ju Chin Huang. 1995. “Can MarketsValue Air Quality? A Meta-analysis of HedonicProperty Value Models,” J. Polit. Econ. 103:1, pp.209–27.

Smith, Kerry and Yoshiaki Kaoru. 1990. “What HaveWe Learned since Hotelling’s Letter? A Meta-analy-sis,” Econ. Let. 32:3, pp. 267–72.

Stanley, Thomas. 1998. “New Wine in Old Bottles: AMeta-analysis of Ricardian Equivalence,” South.Econ. J. 64:3, pp. 713–27.

Stiglitz, Joseph. 1994. “Whither Socialism?” WicksellLectures, Cambridge and London: MIT Press.

———. 1999. “Whither Reform? Ten Years of theTransition,” paper presented at Annual BankConference on Devel. Econ., World Bank,Washington DC.

Van Wijnbergen, Sweder and Anton Marcincin. 1997.“The Impact of Czech Privatization Methods onEnterprise Performance Incorporating Initial SelectionBias Correction,” Econ. Tran. 5:2, pp. 289–304.

Djankov and Murrell: Enterprise Restructuring in Transition 791

Article 1 8/12/02 9:13 AM Page 791

Page 54: Enterprise Restructuring in Transition: A Quantitative Surveyecon-server.umd.edu/~murrell/articles/Enterprise Restructuring in Transition.pdfEnterprise Restructuring in Transition:

Van Wijnbergen, Sweder and Anthony Venables. 1993.“Trade Liberalization, Productivity, and Com-petition: The Mexican Experience,” discus. paper345, London School Econ.

Warzynski, Frederic. 2001a. “Do SchumpeterianWaves of Creative Destruction Lead to Higher Pro-ductivity? Evidence from Poland,” work. paper, LI-COS Center Transition Economies, Catholic U.Leuven, Belgium.

———. 2001b. “The Complementarity of Reforms:Evidence from a Survey of Ukrainian Firms,” work.paper, LICOS Centre for Transition Economics,Catholic U. Leuven, Belgium.

Weiss, Andrew and Georgiy Nikitin. 1998.“Performance of Czech Companies by OwnershipStructure,” Davidson Institute work. paper 186, U.Michigan Business School, Ann Arbor.

Williamson, Oliver. 2000. “The New InstitutionalEconomics: Taking Stock, Looking Ahead,” J. Econ.Lit. 38:3, pp. 595–613.

Xu, Xiaonian and Yan Wang. 1999. “OwnershipStructure and Corporate Governance in ChineseStock Companies,” China Econ. Rev. 10:1, pp.75–98.

Zemplinerova, Alena; Radek Lastovicka and AntonMarcincin. 1995. “Restructuring of CzechManufacturing Enterprises: An Empirical Study,”work. paper, U. Prague.

Zhuravskaya, Ekaterina. 2000. “Incentives to ProvideLocal Public Goods: Fiscal Federalism, RussianStyle,” J. Pub. Econ. 76:3, pp. 337–68.

792 Journal of Economic Literature, Vol. XL (September 2002)

Article 1 8/12/02 9:13 AM Page 792