Privatization Methods and Economic Growth*
John Bennett, Brunel University†
Saul Estrin, London Business School
James Maw, University of Wales Swansea
Giovanni Urga, Cass Business School, London
17 February, 2004
In low-income countries privatization, if implemented appropriately, may play an
important role in generating growth. Using data recently available from Central and
Eastern Europe, we therefore investigate the impact of alternative methods of
privatization on economic growth. Our analysis suggests that the use of conventional
privatization methods to match owners with firms can be inefficient in economies
with underdeveloped capital markets, particularly if wealth is poorly correlated with
managerial and entrepreneurial ability. In these circumstances mass privatization,
with firms being given away or sold at a nominal price, may be the appropriate policy
JEL Classification Numbers: L33, O40, P27, P31
Keywords: Privatization, Economic Growth, Capital Market Development
* We thank participants at the ACES/ASSA 2003 meeting (Washington, 3-6 January 2003), in
particular Barry Ickes and Mario Nuti, and participants at the 59th IIPF Conference (Prague, 25-28
August 2003), in particular our discussant Ugurhan Berkok, for valuable discussions and comments.
We are grateful to Joe Brada for detailed comments and to seminar audiences at London Business
School, Sheffield University, Queens University Belfast, and a CEPR workshop at Heriot-Watt
University (29 August, 2003). We acknowledge helpful discussions with Anders Aslund, Simon
Commander, Stephen Fries, Mark Schaffer, and Jan Svejnar. We thank Maria Bytchkova, Michele
Meoli and Alberto Montagnoli for data collection and research assistance. Any errors are our own.
† Corresponding author: Department of Economics and Finance, Brunel University, Uxbridge,
Middlesex, UB8 3PH (United Kingdom). Tel., +/44/1895/816201; email, firstname.lastname@example.org
Privatization has become a world-wide phenomenon. During the 1990s, it
generated government revenues of more than $700b according to the OECD Pri-
vatization Database. Since 1989 more than 70,000 enterprises have been priva-
tized in Central and Eastern Europe, and privatization is currently a major item
on the policy agenda in China, India, and many other developing economies. Al-
though the impact of privatization on the performance of firms has been studied
extensively (see William L. Megginson and Jeffry M. Netter, 2001), literature ad-
dressing its effects on growth and other macroeconomic implications is sparse.1
Yet many of the countries with nascent privatization programs, such as those
in Sub-Saharan Africa, are suffering from economic stagnation, and the choice
of an appropriate method of privatization could play an important role in jump-
starting growth. For the testing of hypotheses about the consequences for growth
of different privatization strategies, the transition process in Central and East-
ern Europe provides a unique sample, for extensive privatization programs were
undertaken in that region over a relatively short period of time using a variety
of methods. In this paper we therefore use data from a sample of 23 transition
economies over the period 1990 to 2001 inclusive to estimate an aggregate growth
model in which we include methods of privatization, as well as capital market
and private sector growth, as exogenous variables.
Our framework allows us to test empirically a number of competing hypothe-
1Two exceptions are Nico Hansen (1997), who analyzes technology choices under different
privatization schemes, and Alfred Schipke (2001), who sketches general macroeconomic themes
related to privatization.
ses. First, if the positive effects of privatization on the financial performance
and productivity of firms that are predicted by microeconomic theory obtain,
we would expect these effects to have a macroeconomic analog, raising growth.
Second, as Simeon Djankov and Peter Murrell (2002) have stressed, the type of
owner post-privatization may matter and this is related to both the method of
privatization and the extent of capital market development. For example, some
ownership types will, for a given level of capital market development, lead to
more efficient matching of buyers with firms and better corporate governance.
Third, methods of privatization may affect aggregate demand. The greater is the
expenditure of private agents on the purchase of shares from the government,
the more tightly will the spending ability of the private sector be constrained,
leading to different levels of investment spending by privatized firms and con-
sumption spending by households. However, greater revenue for the government
may also raise its willingness to spend on infrastructure, with a potential positive
feedback on aggregate productivity (Philippe Aghion and Mark Schankerman,
1999). Finally, if economic growth is positively associated with capital market
development (Asli Demirguc-Kunt and Vojislav Maksimovic, 1998), and if differ-
ent privatization methods have different effects on capital market development,
this suggests another channel through which the privatization method may affect
Our results have important implications for current privatization policy in
developing economies. We find that the conventional sale of state assets (‘full
privatization’) has no significant impact on growth, whereas mass privatization,
with firms being given away or sold at nominal prices, has a significant positive
effect. Stock market development also enhances growth. Our analysis suggests
that the use of full privatization to match owners with firms can be inefficient in
economies with underdeveloped capital markets, particularly if wealth is poorly
correlated with managerial and entrepreneurial ability. In these circumstances
mass privatization may be the appropriate policy choice.
In Section 1 we outline our theoretical framework, and in Section 2 we discuss
the specification of the estimating equations and the data used. The results
are reported in Section 3, while in Section 4, which concludes, we interpret our
findings. The data sources are reported in the Appendix.
1 Theoretical Framework
In this section we present the simple theoretical framework for our empirical
contribution. We take the method of privatization to be an exogenous policy
choice. First we classify privatization methods and then we specify equations for
real aggregate demand and supply, in each of which the method of privatization
is an argument. Combining these equations, an expression is obtained in which
real GDP depends on the method of privatization and a variety of other factors.2
2The empirical literature on growth commonly adopts a Cobb-Douglas production function
and is based on the assumption that each economy is close to its steady-state growth path (for
example, Robert J. Barro, 1991). However, by definition, transition economies are significantly
out of steady state. Instead, we use a general formulation of the determinants of real GDP,
The impacts on growth of private sector and capital market development are also
We distinguish three alternative privatization methods.3‘Mass privatization’
is defined to occur when the dominant form of privatization in an economy is
that firms are sold at a zero (or nominal) price. ‘Full privatization’ occurs when
the dominant form of privatization is the sale of firms to outsiders for positive
prices. ‘Mixed privatization’ covers all cases that are not adequately represented
by either of the first two categories, and includes manager-employee buyouts
(MEBOs) and leased buyouts. In practice, the choice of privatization methods
appears to have been driven primarily by political and ideological factors, and
does not appear to correlate with economic performance pre-transition.
For a given country and time, we denote real aggregate demand by ydand
real aggregate supply by ys. Our formulation of real aggregate demand, where
the sign above a variable is that of the relevant partial derivative, is
Thus, in addition to being negatively related to the price level p, ydis assumed to
depend on the method of privatization M. Let M = 1 denote mass privatization,
and thus economic growth, incorporating a range of independent variables.
3In practice, each country has privatized in a variety of ways. Nonetheless, we have taken
care to determine where the dominant mode can be identified in each country, and that is the
‘privatization method’ specified in our analysis (see the working paper version of this paper,
available from the authors).
M = 2 full privatization, and M = 3 mixed privatization. With mass privatiza-
tion, the recipients of shares may feel richer, at least in the short term, and this
can be expected to a have a positive, though perhaps relatively small, real wealth
effect on the demand for goods.4The other two methods of privatization involve
the transfer of funds from private agents to the government and so the net effect
on aggregate demand depends on relative marginal propensities to spend. Given
imperfect capital markets, the expenditure associated with these methods may
leave buyers of firms short of liquidity, which may have a negative short-term
effect on real consumption and investment demand. However this effect may be
small for full privatization if the number of buyers is relatively small or if firms
are purchased by foreigners. Hence we expect that real aggregate demand will
be greatest for M = 1 and smallest for M = 3.
Real aggregate supply is specified by
where K is the private sector capital stock; L is employment; A is the human
capital stock; S is a measure of capital market development; P is the share of
the private sector in national income, and G is the public sector capital stock.
We assume without comment that ysis positively related to p, K and A. We
4Insofar as the other methods of privatization underprice shares, we may expect similar
effects on demand, though generally less than for mass privatization.
also assume that ysis increasing in L, though, because of labor hoarding, the
relationship may be weaker than in Western economies. We focus our discussion
on the variables M, P, S and G.
Different methods of privatization may lead to different majority ownership
structures, with differentiated impacts on firm performance, as represented by
M in equation (2). Assuming that full privatization is associated with relatively
efficient matching of owners to firms, it may be expected to lead to the most
effective corporate governance of our three types of privatization. In contrast,
MEBOs and leased buyouts may lead to managerial and worker entrenchment,
while mass privatization may lead to diffuse ownership structures and long agency
chains.5Assuming that more effective corporate governance raises real aggregate
supply, this suggests that the effect on ysmay be strongest for full privatization
(M = 2), but we cannot rank, a priori, the expected effects of mass and mixed
privatization (M = 1 and 3).
Private sector development is the consequence of output growth by both pri-
vatized and de novo firms. We assume that an increase in the output of either,
5An extensive literature addresses how different privatization methods may have influenced
the structure of private ownership post-privatization (see Djankov and Murrell, 2002). Full
privatization is found typically to have led to outsider ownership, with, in Hungary and Estonia,
a high proportion of foreign participation. Mixed privatization has led to insider ownership,
often dominated by managers, and sometimes with a large retained state ownership share (for
example, Romania and Slovenia). The consequences of mass privatization for ownership have
been more complex. In Russia and Ukraine, widespread insider ownership resulted, while in the
Czech Republic and Poland, mass privatization was constructed to ensure primarily outsider
ownership. The impact of privatization method and ownership form also depends on capital
market development; for example, on whether insiders choose to sell their mass privatization
as a proportion of national income, will tend to raise ys, and we treat this effect
as a form of neutral technical progress. The sources of the positive effect of pri-
vatization on productivity at the level of the firm include the better definition of
corporate goals by private firms and some resolution of the incentive problems
associated with the softer budget constraints of state-owned enterprises. Also,
privatization may generate network externalities, with more extensive market
transactions creating a climate of trust, raising business confidence. A major
contribution to productivity growth is made by small and medium-sized de novo
firms through their ability to fill the gaps left under communism by biases towards
high capital intensity and against the provision of services. Private sector devel-
opment is not correlated in the data with the method of privatization, though it
is likely to be associated with other policy measures with respect to the private
sector, for example towards small-scale privatization.
The development of capital markets may be associated with more widespread
and cheaper corporate finance, reducing the need for firms to rely on internally-
generated funds for investment, and thus raising ys. More mature capital market
structures can also be an important element in improved corporate governance
(Megginson and Netter, 2002). Privatization itself may generate development
of the capital market, and the larger the proportion of output that comes from
the private sector, the greater is the scope for benefiting from capital market
development. Thus, we expect that
∂2ys/∂P∂S > 0. (3)
Additionally, there is a potential interaction between privatization method and
capital market development. For example, implementation of mass privatization
policies in Poland was explicitly associated with plans for capital market devel-
opment (Leszek Balcerovic,1995), whereas extensive use of MEBOs may restrict
the expansion of the capital market.
Privatization methods may have a direct impact on the macro-economy through
the effects on government revenue, and hence on its ability to invest in infrastruc-
ture, thereby enhancing growth. Full privatization would be expected to yield
the most revenue for the government, and mass privatization the least. With
mixed privatization, firms are usually sold at a positive, but preferential, price.
Insofar as privatization programs in transition economies are a non-distortionary
source of revenue, productive investment financed by this revenue will have a
positive impact on growth. However, a large proportion of public investment is
financed in other ways, particularly by distortionary taxation, and so the effect
of government expenditure on economic growth may be of either sign (Barro,
1991). Furthermore, the transition economies’ investment performance during
the communist era exhibited extreme inefficiency. Hence, we expect at best a
weak positive relationship between public sector investment and real aggregate
Let y denote real GDP. Setting yd= ys= y, we can solve (1) and (2) for p:
p = P (M,K,L,A,S,P,G) (4)
Substituting (4) into (2), we obtain
y = Ys[P (M,K,L,A,S,P,G),M,K,L,A,S,P,G]
≡ Y (
Real GDP is increasing in K,L, A, S, and P, and may be increasing in G. Putting
together our comments concerning the effects on ydand ysof each privatization
method suggests that, compared to the other methods of privatization, the direct
effect of mixed privatization will be a relatively low level of y, but the ranking of
the other two methods in this respect is unclear. However, given that with mass
privatization the positive effect on demand may be relatively small, we expect
that full privatization, because of its relatively efficient matching, will have the
greatest effect on real GDP, at least in economies with more developed capital
2 Specification and Data
We estimate a cross-country growth model along the lines of, for example, Barro
(1991) and Gregory Mankiw, Davis Romer and David Weil (1992). However,
equation (5) leads us to supplement the model with the method of privatiza-
tion, government investment, and indicators of private sector and capital market
development. We also explore potential complementarities between privatiza-
tion method, private sector development and capital market development. For
emerging markets, similar methodology has been applied to capital market de-
velopment by, for example, Gert Bekaert and Campbell R. Harvey (2000) and
Peter B. Henry (2000). The basic model is therefore
GDPit = a1+ a2INVit+ a3EMPit+ a4IHCit+ (6)
a5STOCKMit+ a6PRIVit∗ a7GISit+
a8FULLit+ a9MASSit+ a10MIXEDit+ εit
where all variables are re-labelled for a more immediate interpretation of the es-
timates. i denotes country and t time; GDP is the first difference of the log of
real gross domestic product y; INV is the first difference of the log change in
the real capital stock K; EMP is the first difference of the log of employment L;
IHC is the first difference of the log of investment in human capital A, which is
measured by gross enrolment in tertiary education; STOCKM is stock market
capitalization as a proportion of GDP, which is our measure of stock market de-
velopment S; PRIV is the share of private sector output in GDP, corresponding
to P in equation (2); and GIS is the share of government expenditure devoted
to investment, corresponding to the variable G.6The three methods of priva-
tization, M, are denoted by FULL, MIXED and MASS; and εitis an i.i.d.
error term. The estimation period is the twelve years 1990-2001 and covers 23
We use panel-data analysis (within-groups estimators) to exploit both time-
series and cross-section variation in data, in particular in the relationship between
growth and privatization method. We test for time-specific as well as country-
specific fixed effects in each regression, and compare the performance of model (6)
with its dynamic counterpart. For the dynamic version of the model, with lagged
dependent variables, we use generalized-method-of-moments (GMM) estimation,
dealing with potential problems of endogeneity of the explanatory variables by
instrumenting on lagged values. GMM estimation also allows us to address the
6Equation (6) is estimated in first-difference form to take out country-specific fixed effects.
Since we are addressing the impact of privatization method, and capital market and private
sector development on growth, the variables are included in the estimating equations in levels
form. This is clearly appropriate for privatization methods, since it is constructed as a dummy
variable. However, since the others are continuous variables, we also estimated the equations
including these in first differences. The principal findings with respect to INV , EMP, IHC, G
and methods of privatization are not affected by the change in specification. However, neither
PRIV nor STOCKM, nor their interaction is significant in the OLS estimation. STOCKM
and its interaction with PRIV are significant in the GMM estimation, with the same signs as
in Table 4.
7Our data set covers all the transition countries listed by EBRD (2002), except for Bosnia
and Herzegovina, FR Yugoslavia, Tajikistan and Turkmenistan for which relevant data are not
available. We cover Albania, Armenia, Azerbaijan, Belarus, Bulgaria, Croatia, Czech Repub-
lic, Estonia, FYR Macedonia, Georgia, Hungary, Kazakhstan, Kyrgyzstan, Latvia, Lithuania,
Moldova, Poland, Romania, Russia, Slovakia, Slovenia, Ukraine and Uzbekistan.
correlation between the error term and lagged endogenous variables,.
The three privatization dummies have both a cross-section and a time-series
dimension. We identify the chosen method of privatization in each country and
then identify the date at which this privatization method was introduced.8In
each case the dummy variable is zero in the years before the relevant method of
privatization was introduced and unity thereafter. The classification of privati-
zation method by year and country is presented in Table 1. When the primary
method is identified by the European Bank for Reconstruction and Development
(EBRD) as voucher, we classify privatization as ‘mass.’ When the EBRD pri-
mary method is direct sales, we classify privatization as ‘full.’ In other cases we
classify privatization as ‘mixed.’9
[Table 1 about here]
8In preparing the paper, we explored the effects of using three different ways of classifying
privatization methods (see our working paper). The first was based on official reports available
on government websites. The second used external documentary sources. The third was based
on EBRD classifications. In this paper we report regressions based on the third approach be-
cause it derives from a single source and does not rely on our subjective judgements. However,
all the equations were estimated using all three approaches. When the approach was changed,
five countries shifted category, but none of the conclusions with respect to method of privati-
zation were affected. (The regressions are available from the authors on request.) Thus, our
findings are not dependent on how we categorize the few countries for which the dominant
privatization method is somewhat unclear, and which may shift between Mass and Mixed. Our
results are driven by the performance of countries about which there is little debate as to the
dominant method of privatization.
9For countries, such as Ukraine, in which privatization has been relatively slow, the year
specified in Table 1 may be contentious. However, our main findings are found not to be
sensitive to adjustment of the assumed privatization dates.
We first estimate versions of equation (6), before addressing issues of dynamics
using GMM methods and undertaking sensitivity tests. In all equations we use
White’s correction for robust standard errors. In Table 2, we report four ver-
sions of equation (6). Column (1) represents the simplest formulation, with no
interactions, while in column (2) we include a term for the interaction between
capital market development and private sector development. We have also cal-
culated regressions interacting each method of privatization with stock market
development and with private sector development. None of these interactions
with the method of privatization are significant, but, as a ‘representative’ ex-
ample, in column (3) we report a regression that includes an interaction term
between stock market development and mass privatization. Column (4) includes
both interaction terms.
[Table 2 about here]
All four formulations in Table 2 yield good fits, with R2> 0.6 and joint
Wald tests ranging from 130.7 in column (1) to 179 in column (3). The Wald
tests for country (dummy) and time confirm the strong significance of fixed-
and time-specific effects in the growth process, while the AR tests indicate that
autocorrelation is not present.Country fixed effects represent a particularly
important element in the explanation, with the value of the χ2in the Wald tests
ranging from 2703 in column (1) to 7108 in column (4), all with probability [0.00].
The coefficients on factor inputs are stable and significant across the four for-
mulations. The findings are consistent with the type of growth process identified
by Barro (1991) and many others, with the coefficient on investment estimated
to be around .08. The coefficient on the change in employment is also highly
significant, but lower than typically obtains in the West, perhaps because of la-
bor hoarding in the immediate post-transition period. Additionally, we identify a
relatively small, but significant, impact of labor-quality change on GDP growth.
Private sector and stock market development are not found to be indepen-
dently significant (column (1)). However, once their interaction term is included
(column (2)), we identify a significant positive impact of stock market develop-
ment on GDP growth, together with a small negative interaction effect. This
suggests that the growth-enhancing effects of capital market development relied
on private sector development, but tailed off as an economy approached a West-
ern ownership and capital market structure. As noted above, the interaction of
stock market development with mass privatization (column (3)) is not significant.
Turning to the impact of privatization methods on growth, the findings are
consistent across the four specifications. Neither full nor mixed privatization is
found to exert a significant independent influence on GDP growth, but the co-
efficient on mass privatization is always positive and at least weakly significant.
Moreover, the coefficient on government investment expenditure is insignificant
in each specification, and so we conclude that full privatization influences growth
neither directly, through productivity enhancement, nor indirectly, through the
potential macro-economic externality that could derive from spending the in-
creased government revenues on infrastructure.10
To check the robustness of
the finding with respect to government investment expenditure, we re-estimated
columns (1)-(4) of Table 2 replacing GIS by the EBRD (1996, 2003) index of
infrastructure reform. The coefficients and standard errors on the other indepen-
dent variables in Table 2 were hardly affected by the change and the index was
not significant in any regression (see our working paper).
Since the results could be sensitive to the absence of dynamics in the spec-
ification, we re-estimated the OLS regressions with the inclusion of a lagged
endogenous variable. The results, which are reported in Table 3, indicate that
a dynamic specification is appropriate: the lagged endogenous variable is signifi-
cant in all three columns, and its inclusion widens the standard error around the
contemporaneous change in employment variable, which is no longer significant.
However, our conclusions about the impact of capital market development and
privatization method on growth are the same as in our original specification. The
coefficients on stock market development and its interaction term with private
sector development are significant in columns (2) and (4), and the coefficient on
mass privatization is positive and significant in each column. The coefficients on
mixed privatization, full privatization and government investment expenditure
10The data suggest that, even though the choice of full privatization must have relaxed the
government’s budget constraint, the authorities did not choose to use this incremental revenue
on capital expenditure. The correlation coefficient between full privatization and government
investment expenditure is only .05.
are not significant in any column.
[Table 3 about here]
The results in Table 3 may be biased by correlation between the error term
and the lagged endogenous variable, so we also undertook GMM estimation. The
GMM estimates for the four specifications, with factor inputs (investment and
the changes in employment and labor quality) and government investment expen-
diture all instrumented on lagged values, are reported in Table 4. This replicates
the main results from the other two tables. The lagged endogenous variable is
always significant, as are the coefficients on investment and the change in employ-
ment. However the measure of the change in human capital is not quite significant
in any specification using GMM estimation methods. As in the previous tables,
government investment expenditure is found to be insignificant in all the four
specifications, as are the coefficients on full and mixed privatization. The con-
clusions from Table 2 concerning private sector and capital market development
are confirmed by Table 4, as are the findings concerning mass privatization.
[Table 4 about here]
We sought to understand these results further by breaking the data set into
sub-samples geographically (Conference of Independent States (CIS) versus non-
CIS) and over time (1990-1995 and 1996-2001), but the decline in degrees of
freedom made the estimated coefficients much less precise. Since the equations are
less reliable, we exclude them here (they are reported in full in our working paper).
What these experiments suggest is that our main results with respect to stock
market development, private sector development and method of privatization
hold in particular in the CIS, and in the period after 1995. This is unsurprizing
given the data Table 1, which indicate the concentration of mass privatization in
CIS countries and in the post-1995 period.
However, these findings suggest that the mass privatization dummy might be
proxying for improved demand and cost conditions in the CIS countries between
1996-2001. Sources of these improved conditions might include oil prices, which
were increasing in this period to the benefit of several oil-supplying CIS countries,
and exchange rate depreciations. We therefore re-estimated our equations to
control for this issue. A sample of the results is reported in Table 5, which is
based on column (2) of Table 2, but includes first the exchange rate (column (1))
and then the exchange rate and the oil price (column (2)). It can be seen that
the new coefficients are significant in both columns (1), suggesting that oil (and
perhaps other commodity) price increases and exchange rate depreciation both
had a positive effect on growth. However, as in the other tables, the coefficient
on mass privatization in Table 5 is positive and significant. Thus, although mass
privatization was the dominant method in CIS countries, the variable MASS is
not proxying for demand or costs factors in this region. This suggests that the
correct interpretation of our previous findings is that the countries in the CIS that
introduced mass privatization were in a better position to exploit the improved
market conditions in the late 1990’s than those countries that employed other
[Table 5 about here]
Our empirical analysis shows that the method of privatization plays an important
role in economic growth. A finding that full privatization had a significant growth
effect would have verified the hypothesis that efficient corporate governance and
the matching of buyers with firms are critical. An advantage of full privatiza-
tion is that it leads to concentrated ownership, whereas mass privatization has
the converse effects. However, we find that it is mass privatization that has
the positive effect, suggesting that the matching argument is relatively weak for
transition economies. This may because in any economy the ability to purchase
a firm, or at least a substantial ownership share, is imperfectly correlated with
the skills required to run the firm efficiently. In an economy with an extremely
underdeveloped capital market, ‘wrong’ owners will tend to persist for longer.11
The positive effect of mass privatization has been justified on political economy
11This argument is strengthened insofar as the income distribution inherited from the com-
munist era was misaligned with the ability to run firms. In contrast, in the Czech Republic,
the distribution of shares at nominal cost to the general public led to shares being placed in
the hands of privatization funds, which may have exerted pressure on managers to be relatively
efficient. The argument does not apply to full privatization to foreign investors, but, as we
have already noted, the amount of such privatization has been relatively small across all tran-
sition economies, amounting to less than 2% of world foreign direct investment in 1999 (United
Nations Conference on Tariffs and Development, 2003).
grounds (Maxim Boycko, Andrei Shleifer and Robert W. Vishny, 1995), but it is
also consistent with the hypothesis that the method of privatization may foster
growth through the effects on the demand for consumption or investment goods
and by separating the management of firms from state ownership (see Djankov
and Murrell, 2002).
We hypothesized that government capital investment might have played a
significant role in generating growth through the provision of public goods, in
which case full privatization could have yielded benefits through the greater rev-
enue it generates for the government. However, we are unable to identify such
an effect in our empirical work, presumably because transition governments did
not use incremental revenue in this way. Thus, although full privatization raises
more (immediate) government revenue than mass privatization does, the extent
to which this is translated into faster economic growth may be disappointing.
Our analysis has significant implications for developing countries that have
still to undertake large-scale privatization programs, for example China, India
and Vietnam. It suggests that the method of privatization is an important policy
choice, and that, despite the great criticism it has received in recent years, mass
privatization may be the appropriate choice in situations where capital markets
are highly imperfect and the distribution of wealth is not well correlated with the
distribution of managerial ability.
The data used in this paper describe 23 of the 27 transition economies covered
in the EBRD Transition Reports (various years), which provide complete infor-
mation for 1990-2001 on macroeconomic variables including GDP, employment
and gross fixed capital formation, and the indicators of institutional investment
used in the paper. Bosnia and Herzegovina, FR Yugoslavia, Tajikistan and Turk-
menistan were excluded from the analysis because complete data on these coun-
tries were unavailable.
Gross Domestic Product. The base year for the GDP series was sourced
from the World Bank’s Historically Planned Economies: A Guide to the Data,
taking the 1988 figure, measured in constant 1987 market prices. Figures were
converted into US dollars using the 1987 exchange rate.12
For each country
that later disintegrated (Czechoslovakia, Yugoslavia and the USSR), we broke
the total GDP into constituent parts using information provided by UN, World
Bank and national sources on the constituent countries’ share in total GDP.
The total USSR figure was divided into Armenia, Azerbaijan, Belarus, Estonia,
Georgia, Kazakhstan, Kyrgyzstan, Latvia, Lithuania, Moldova, Russia, Ukraine
and Uzbekistan. The total figure for Yugoslavia was divided into separate data
for Slovenia, Macedonia and Croatia. The total figure for Czechoslovakia was
divided to obtain separate data on Slovakia and the Czech Republic. To extend
12In the case of Albania, 1988 GDP is provided in constant 1986 market prices, and was
converted into US dollars using the 1986 exchange rate.
the series from 1988 real GDP growth data provided by the EBRD were used.
Fixed Capital Investment. Fixed capital investment figures were obtained
from the EBRD (1999, 2000, 2001, 2002) by taking the real gross fixed investment
rate, measured in annual percentage change. For the few cases in which such in-
formation was unavailable, alternative measures were used. The main alternative
source was data on investment share in GDP provided by the IMF and EBRD.
This ratio was applied to our GDP levels data to obtain fixed capital investment
levels figures. An annual percentage change in fixed capital investment was cal-
culated from the levels figures. We also used GDP level figures to calculate fixed
capital investment growth in the early 1990s in the few cases when information on
annual percentage change in investment was not available. We calculated fixed
capital investment figures by applying fixed capital investment to GDP ratios,
provided by IMF and National Statistics sources, to our GDP levels figures.
Employment. Information on employment growth was obtained from EBRD
employment time series, measured in annual percentage change, for 1989-2001.
Investment in Human Capital. The measure chosen for investment in
human capital was gross enrolment in tertiary education, defined as the total
number of students who had attained a certain level of education as a percentage
of the total population in the age group. The data were obtained using the
TransMonee Database, produced by UNICEF, by taking 5-year period averages.
These series were preferred to UNESCO data, which are inconsistent with the
World Bank source and show unconvincingly high growth of enrolment rates for
Government Investment Share in GDP. Data were derived from the
measurement of government investment expenditure provided by IMF Country
Reports. For CIS countries, information on the early years of transition was
unavailable from this source, so we used the CIS national databases. For the
Baltic countries, the Baltic International Centre for Economic Policy Studies
provided additional data for Estonia (1991 and 1995); for Latvia (1994 and 1995);
and for Lithuania (1991, 1993, 1994 and 1995.)
Private Sector Share in GDP. Data were taken from EBRD (1999, 2002).
Stock Market Capitalization as a Proportion of GDP. Data were taken
from EBRD (2002) and the Emerging Stock Market Facts Book. Since in many
transition countries the stock market did not exist in the early 1990s, a zero value
was assigned for those years.
Indices of Reform. The EBRD Infrastructure Reform Index was sourced
from EBRD (2002).
Privatization Data and Mode. Information on privatization mode was
sourced from the EBRD (1995, 2002), which classifies privatization methods into
voucher, direct sale, and MEBO, and identifies the first year in which the primary
type of privatization was implemented.
 Aghion, Philippe, and Mark Schankerman, “Competition, Entry and the So-
cial Return to Infrastructure in Transition Economies,” Economics of Tran-
sition, 7 (1999), 79-102.
 Balcerovic, Leszek, Socialism, Capitalism, Transformation (Budapest: Cen-
tral European University Press, 1995).
 Barro, Robert J., “Economic Growth in a Cross Section of Countries,” Quar-
terly Journal of Economics, 106 (1991), 407-443.
 Bekaert, Gert, and Campbell. R. Harvey, “Foreign Speculators and Emerging
Equity Markets,” Journal of Finance, 55 (2000), 565-614.
 Boycko, Maxim, Andrei Shleifer, and Robert W. Vishny, Privatizing Russia
(Cambridge, Mass., and London: MIT Press,1995).
 Demirguc-Kunt, Asli, and Vojislav Maksimovic, “Law, Finance, and Firm
Growth,” Journal of Finance, 53 (1998), 2107-2138.
 Djankov, Simeon, and Peter Murrell, “Enterprise Restructuring in Transi-
tion: a Quantitative Survey,” Journal of Economic Literature, 40 (2002),
 European Bank for Reconstruction and Development (EBRD), Transition
Report (London: EBRD, various years).
 Hansen, Nico, “Privatization, Technology Choice and Aggregate Outcomes,”
Journal of Public Economics, 64 (1997), 425-42.
 Henry, Peter B., “Stock Market Liberalization, Economic Reform, and
Emerging Equity Prices,” Journal of Finance, 55 (2000), 529-564.
 Mankiw, Gregory, David Romer, and David Weil, “A Contribution to the
Empirics of Economic Growth,” Quarterly Journal of Economics, 107 (1992),
 Megginson, William L., and Jeffry M. Netter, “From State to Market: a Sur-
vey of Empirical Studies of Privatization,” Journal of Economic Literature,
39 (2001), 321-89.
 Schipke, Alfred, Why Do Governments Divest? The Macroeconomics of Pri-
vatization (Heidelberg: Springer, 2001)
 United Nations Conference on Tariffs and Development (UNCTAD), World
Investment Report (New York and Geneva: United Nations, 2003).
Table 1: Country Privatization Table
Country Classification of
Czech Republic Mass
Table 2: Growth Equations, 1991-2001, Interacting Private Sector Share and
Mass Privatisation with Stock Market Capitalization
PRIV ∗ STOCKM
MASS ∗ STOCKM
T × N
Notes: (a) Significance levels: ***: 1% or less; **: less than 5%; *: less than 10%; (b) Time
Dum and Group Dum=time and group dummies, respectively; Σ= equation standard error, R2=
determination coefficient; total T × N=number of observations, N=number of countries and
k=number of parameters; (c) W(joint) = Wald tests the significance on all regressors except the
dummies;W(dummy) = Wald tests the significance of all dummies; and W(time) = Wald tests
the significance of the time dummies and the constant. All these statistics are asymptotically
distributed as χ2
(n)under the null of no relationship, where n represents the degree of freedom;
(d) AR(1) and AR(2) statistics test for the first- and second-order serial correlation respectively
in the residuals. The statistics are asymptotically distributed as standard normal under the null
of no serial correlation.
Table 3: Growth Equations, 1991-2001, OLS Dynamic Models, Interacting Pri-
vate Sector Share and Mass Privatization with Stock Market Capitalization
PRIV ∗ STOCKM
MASS ∗ STOCKM
T × N
Notes: For explanation see notes to Table 2.
Table 4: Growth Equation, 1991-2001,GMM Dynamic Models, Interacting Pri-
vate Sector Share and Mass Privatization with Stock Market Capitalization
PRIV ∗ STOCKM
MASS ∗ STOCKM
T × N
Notes: For explanation see notes to Table 2.
ing GDP, INV , EMP, IHC and GIS. (b) Transformation used: first differences. (c)
Level instruments: dummies, GMM (GDP,1,2), GMM (INV ,1,2), GMM (EMP,1,2), GMM
(IHC,1,2), GMM (GIS,1,2). (d) The Sargan statistic is a test for the over-identifying restric-
tions (k), asymptotically distributed as χ2(k) under the null of instruments validity..
In addition: (a) GMM model instrument-
Table 5: Growth Equations, 1991-2001, Including Ex- Download full-text
change Rates and Oil Price.
PRIV ∗ STOCKM
T × N
Notes: For explanation see notes to Table 2. In addition: in
columns (1): model with exchange rates and time dummies;
in column (2): model with exchange rates and oil prices (no