Do market pressures induce economic efficiency? The case of Slovenian manufacturing, 1994-2001.1. Introduction
A long-held but infrequently tested proposition in economics is that competitive pressure will force firms to be efficient. (1) The transition to market in formerly planned economies offers a natural laboratory to test the competitive hypothesis. As reviewed by Boeri (2000), the consensus expectation at the outset of transition is that competitive pressures from the emerging market system would force greater productive efficiency on enterprises that remained from the old system. Furthermore, converting state-owned enterprises into profit-maximizing firms was expected to create incentives to improve the efficiency of these often-under-performing sectors, either through profit motives or through the rigors of investor scrutiny (Brada 1996). This study examines whether market competition had the expected effect on productive efficiency in Slovenia.
The early evidence from transition economies does not demonstrate immediate evidence of efficiency gains. From an initial reduction in gross domestic product (GDP) averaging 25% in Central and Eastern Europe and 50% in the former Soviet states, production recovered unexpectedly slowly. Only 2 of 25 transition countries had matched their 1989 production levels 10 years later (Campos and Coricelli 2002). Some of the slow recovery may be that competitive pressures grew slowly. For example, most countries only gradually abandoned tax and transfer policies that effectively taxed the expanding sectors to subsidize those in decline, and it took time to establish legal institutions that supported property rights and limited corruption (Boeri 2000; Svejnar 2002). However, some have argued that increased competition itself may have contributed to reduced production because it disrupted the formerly well-organized trading systems linking Warsaw Pact countries (Blanchard and Kremer 1997).
As the transition to market has progressed, a consensus is emerging that efficiency gains are forthcoming from competitive market pressures; although, the magnitude of the effect is uncertain. The review by Djankov and Murrell (2002), summarizing 23 studies of the impact of increased competition on firm performance, suggests that competition raised efficiency in Central and Eastern Europe but not in the former Soviet Union. Similarly, their examination of 37 studies on the impacts of privatization finds that competition raised efficiency in Central and Eastern Europe but not in countries of the former Soviet Union. Even within regions, however, there is substantial variation in the magnitude, and even the sign of the productivity effects, so the average masks considerable variation across studies. (2)
Past studies of the impact of transition on firm efficiency have concentrated on potentially unrepresentative subsets of the firms in transition economies. As we demonstrate in this study, these subsamples can yield misleading inferences regarding efficiency gains in these economies. One common strategy is to concentrate on large, formerly state-owned enterprises that survived into the transition. The focus is natural because these are the types of firms that existed under socialism, but this approach misses the contributions to efficiency from the entry of new, more efficient firms and from the closing or exit of the least efficient firms. McMillan and Woodruff (2002, p. 154) argue that the "success or failure of a transition economy can be traced in large part to the performance of its entrepreneurs." They conclude that the most successful transition economies were those that fostered the entry and success of new firms and not necessarily those that most aggressively privatized former state enterprises. In addition, most studies use a single cross section of data or else a short time frame, but most countries adopted competitive policies only gradually over several years. If efficiency gains were only realized over time, short panels may understate the impact of market competition on efficiency.
This study contributes to the existing knowledge regarding the impact of market forces on production efficiency by utilizing a unique data set encompassing every manufacturing firm in Slovenia over a long time span. Our data include not only continuously existing firms but also newly formed firms and firms that cease production during the period of study. In contrast to many existing studies, we can therefore identify the impacts of firm entry and exit on productive efficiency in the transition to market. Second, we examine the progress of efficiency over a long period of time: from 1994 through 2001. The eight-year period is sufficiently long to determine whether measured efficiency gains or losses are permanent or a consequence of short-term economic shocks. And, third, the article deals with Slovenia, a country often considered a special case among transition countries, and thus it can provide interesting case findings from the comparative perspective.
Our results strongly confirm the importance of competitive pressures in raising firm total factor productivity (TFP). The efficiency gains were progressive, rising each year, and they are broad based, occurring in almost all industries examined. Although the largest gains were in private firms, consistent with findings in other transition economies, competitive pressures at the industry level appeared to increase TFP in firms under state and other (3) ownership types as well. Firm ownership type did matter in the aggregate. In industries with higher market shares attributable to private firms, foreign-owned firms, and imported goods, all firms were more productive, whether privately owned, state owned, or foreign owned. An important factor driving this increased productivity was competitive pressure, which drove out the least efficient firms, including inefficient state-owned enterprises. At the same time, newly entering firms were at least as efficient as surviving firms. This sorting effect is at least as large as the effect of competition on continuing firms in our preferred specification. These conclusions are not sensitive to alternate specifications or controls or firm-specific factors. As a result, the role of market forces in generating economic efficiency is strongly confirmed.
2. Policies That Affect the Market Competition in Slovenia
As part of former Yugoslavia, Slovenia's economy was characterized by government rather than private ownership of assets. Although nominally under a worker-managed system, there was extensive political interference in firm decisions regarding investment, employment, and wages. To meet mandated payrolls, a massive system of discretionary taxes and transfers taxed away net revenue from profitable enterprises to subsidize failing firms that could not meet their payrolls. Inefficient firms could lose money indefinitely; whereas, efficient firms could not build up reserves that could allow expansion. (4) Restrictions on capital mobility also restricted efficient resource allocations. Socially owned firms were not allowed to sell their assets, nor could workers obtain a return on capital if they invested in the firm by accepting wage concessions. Consequently, there was little incentive to invest in capital. Private firms were limited to no more than 10 workers and so also faced limits to growth.
[FIGURE 1 OMITTED]
Slovenia's transition to market began toward the end of 1988, but most reforms came into effect much later. Based on European Bank for Reconstruction and Development (EBRD) transition reports, summarized in Figure 1, Slovenia's structural reforms progressed steadily but unevenly across sectors. Liberalization of foreign trade and of prices was already well underway by 1991, as was privatization of small firms. Other reforms began later and made slower progress. The legal process for privatization of large state enterprises began in 1993 and started in earnest in 1994. About the same time, reforms of the banking system and of other financial institutions began, as did relaxation of rules governing the formation of new firms. (5)
Slovenia was slow to attract foreign direct investment, and foreign investors purchased less than 1% of the initially offered shares of Slovenian privatized firms. Consequently, the most important source of competition from foreign firms is through imports. Slovenia already had liberalized trade restrictions before the transition began, and the Custom and Tariff Acts of 1996 reduced average tariffs to 5.7%.
By the end of 1993, government subsidies to failing firms were discontinued. Roughly 27% of manufacturing firms in existence in 1994 had failed by 2001. (6) Even state-owned firms were allowed to fail. In contrast, many transition economies maintained formal and informal transfers that retarded both state and private sector firm failures (Estrin 2002). (7) In western economies, a large share of productivity growth in industrialized economies has been attributed to resource reallocation through new firm start-ups and closings (Olley and Pakes 1996; Bartelsman and Doms 2000).
Compared to other transition economies, Slovenia's pace of structural reforms was slower than average (Svejnar 2002). Nevertheless, by the end of the period, Slovenia's slow but steady growth in the overall EBRD transition index, shown in Figure 1, reached the average of other countries. (8) Our interest is in assessing whether there are coincident changes in measures of firm efficiency that correspond to cross-sectional or time series variation in measures of the degree of competition facing firms. Our analysis begins in 1994 when newly installed firm reporting procedures created a consistent set of accounting rules for all incorporated firms operating in Slovenia. By that point, the government transfer system was completely disabled, and so measures of firm revenues are not clouded by remaining direct political and economic interference in firm decisions regarding entry, exit, or resource allocation. Instead, efficiency should reflect the ongoing process of institutional reforms.
Our strategy is to trace changes in individual firm efficiency over time, using a measure of TFP. (9) To derive our TFP measure empirically, we assume that the technology faced by the ith firm in the jth industry in year t is assumed to be approximated by the translog production function
ln [q.sub.ijt] = [[alpha].sub.0] + [n.summation over (k = 1)] [[alpha].sub.k] ln [x.sub.ijkt] + 1/2 [n.summation over (k = 1)] [n.summation over (l = 1)] [[beta].sub.kl] ln [x.sub.ijkt] ln [x.sub.ijlt] + [[epsilon].sub.ijt], (1)
where output is denoted by [q.sub.ijt]; the n inputs [x.sub.ijkt] include measures of labor, capital, and material inputs; [[alpha].sub.k] and [[beta].sub.kl] are, respectively, first- and second-order translog production parameters; and [[epsilon].sub.ijt] is an error term. The error term, a variant of the Solow residual, is our measure of TFP. (10)
The TFP has three components that we will explore: time-varying industry-specific factors, [[eta].sub.jt]; time-varying firm-specific factors, [[psi].sub.it], and time-invariant firm-specific factors, [[theta].sub.i]. In addition, we allow a purely random technology shock, [[xi].sub.ijt]. (11) The formulation for the error term in Equation 1 is written
[[epsilon].sub.ijt] = [[eta].sub.jt] + [[psi].sub.it] + [[theta].sub.i] + [[xi].sub.ijt]. (2)
Our strategy is to specify the elements of the error components in a manner that will allow us to identify factors that are tied to changes in TFP across firms and across time. The industry-specific component is specified as
[[eta].sub.jt] = [I.sub.jt][gamma] + [[iota].sub.jt], (3)
where [I.sub.jt], is a 1 x J vector of industry attributes, such as industry concentration or import penetration, [gamma] is a J x 1 parameter vector that translates industry attributes into measured TFP for firms in the industry, and [[iota].sub.jt] is a random error. Similarly, we can specify the time-varying firm-specific component as
[[psi].sub.it] = [f.sub.it][delta] + [[phi].sub.it], (4)
where [f.sub.it] is a 1 x M vector of firm attributes that change over time, such as ownership structure, [delta] is an M x 1 vector of parameters describing how these firm attributes affect TFP, and [[phi].sub.it] is a random error.
The time-invariant firm component is specified as
[[theta].sub.i] = [F.sub.i][mu] + [[upsilon].sub.i], (5)
where [F.sub.i] is an N x 1 vector of observable firm attributes that do not change over time, [mu] is a 1 x N vector of parameters, and [[upsilon].sub.i] is a vector of unobserved time-invariant firm productivity. (12)
Equation 5 summarizes the selection issues that could bias our estimates of [gamma] and [delta]. Suppose that [[theta].sub.i] represents a firm-specific technology component that is observable by potential investors. Then changes in firm ownership status to private ownership or stock ownership from state ownership will be correlated with [[theta].sub.i]. (13)
If [[upsilon].sub.i] = 0 for all firms, then selection into firm types is based on the observables, [F.sub.i]. Attractive candidates for inclusion in the vector [F.sub.i] are ultimate ownership status measures for the firms. In other words, [F.sub.i] will contain dummy variables indicating whether the firm ultimately became privately owned, foreign owned, a company issuing common stock, or some other ownership type. The coefficients on these measures, [mu], will reveal whether firms that ultimately attained ownership status [F.sub.i] had atypically high or low TFP before any changes in their ownership. The related estimate of [delta] will reveal whether there was a change in TFP associated with the change in ownership status.
When [[upsilon].sub.i] = 0 for all L we can estimate [gamma], [delta] and mu] by inserting Equations 2-5 into Equation 1 and applying ordinary least squares (OLS) to the resulting reduced form equation. (14) If [[upsilon].sub.i] in Equation 5 is not zero but is distributed N(0, [[sigma].sub.i]), then selection into ownership states on the basis of expected efficiency will still be driven by the observables, [F.sub.i]. All the parameters [gamma], [delta], and [mu] can be estimated with the appropriate substitutions of Equations 2-5 into Equation 1. However, additional efficiency can be obtained by applying a random effects estimator to accommodate the firm-specific error variance, [[sigma].sub.i].
If E([[upsilon].sub.i]) [not equal to] 0 for at least some i, then selection into ownership types will be based in part on the unobservable [[upsilon].sub.i]. The correlation between [F.sub.i] and [[upsilon].sub.i] will yield biased coefficients on the [gamma] and [delta]. With multiple years of data, we can use fixed-effects to estimate a separate [[theta].sub.i] for each firm. We will no longer be able to capture the [mu], but we can derive unbiased estimates of [gamma] and [delta].
Although individual elements of [I.sub.jt], [f.sub.it], and [F.sub.i] may be of interest, we are particularly interested in exploring the relative importance of industry and firm factors in affecting TFP growth. Competitive forces will have common effects across firms in the industry; whereas, ownership structure will affect TFP growth for specific firms in the industry. The aggregate impacts of industry, time-varying and time-invariant firm attributes on TFP growth over the sample period can be estimated respectively by ([[bar.I].sub.t] - [[bar.I].sub.0])[gamma], and ([[bar.f].sub.t] - [[bar.f].sub.0].) [delta], and ([[bar.F].sub.t] - [[bar.F].sub.o])[mu]. The terms of form ([[bar.X].sub.t] - [[vbar.X].sub.0]) are the change in the average values of X from the base period to period t. Notice that although the luted effects, [[bar.F].sub.t], do not change over time for an individual firm, the average values of these firm fixed effects will change as weak firms are sorted out of the market and strong firms are retained.
The data for this study are based on the universe of manufacturing firms existing in Slovenia between 1994 and 2001. The primary information on firms comes from four data sources. The official financial records of the firm, submitted annually under uniform accounting procedures to the government of Slovenia, provide information on the firm's capital stock, material inputs, and revenues from domestic and foreign sales. The Slovenian Business Registry includes information on the four-digit industries that describe each firm's product line(s), the year the firm initiated production, and the firm's type and ownership structure. (15) We supplement these categorizations of ownership type with additional information from the Bank of Slovenia on firms that are at least 10% owned by foreign individuals or entities. The work history data set tells us how many employees of each education level work for each firm. These three data sets can be integrated using a common firm identification number used in all three series. (16) The variable definitions and sample means are reported in Table 1. (17)
The employment information includes the number of two- or four-year college graduates, the number of high school graduates, and the number of primary educated workers in the firm. This employment information is in real terms by construction. However, the accounting data on firm output and capital and material inputs are reported in nominal terms. We convert the nominal data into real data, using industry input and output price deflators reported for all years 1994-2001. The material input price deflator is a weighted sum of sectoral prices where the weights are sectoral input shares generated from an input-output matrix of the Slovenian economy. Output price deflators are reported for each industry. There is a single capital price series that was applied to all firms. Using these input and output price series, we generate series for real output, capital, and material inputs for each firm and for each year. (18)
Numerous changes suggest an increase in the competitive pressure on Slovenian manufacturing firms from imports, foreign owners, more firms, more new firms, and more private firms that presumably will be trying to produce efficiently. There is also considerable evidence that firms fail over the period. The sample means show a dramatic increase in the number and the market share of private firms. Within individual four-digit industries, the Herfindahl index falls over time, indicating greater competition. The proportion of firms under foreign ownership changes only modestly over time, but their market share doubles. Import penetration, measured by the proportion of industry sales attributable to imports, rises by 79%. The market share of industry output attributable to new entrants (firms that initiated sales after 1993) rises from 4% to 21%. In 1994, firms that will cease operations by 2001 were responsible for 17% of industry sales. (19) All of these trends suggest rising competitive pressure on firms.
5. Total Factor Productivity Growth over Time and across Firms
Our first task is to document the extent of TFP growth in the Slovene economy. This section demonstrates that efficiency gains in Slovenia following the policy reforms were experienced in virtually all sectors of the economy. We also demonstrate that measured TFP growth is sensitive to the inclusion or exclusion of entrants, exiters, and small firms, and that their exclusion greatly understates the efficiency gains during the transition.
TFP Growth Is Not Sensitive to Measurement
As shown in Table 2, trend growth of productive efficiency in Slovenia manufacturing is robust to alternative assumptions about the error process. Three specifications of the translog formulation of Equation 1 were estimated: OLS, a fixed-effects variant that allows for a separate constant term for each firm, and a random effects variant that assumes a different variance for each firm. The three series are highly correlated and yield the same general inference: There has been a consistent increase in TFP in the 1994-2001 period. The increase in TFP per firm is substantial, varying from 0.222 to 0.244 log points, which implies a 24.9% to 27.6% increase in TFP. (20) In other words, the average manufacturing firm in Slovenia was producing about 25% more from the same level of inputs in 2001 as in 1994.
We also report measures of skewness and kurtosis for the least squares variant of the TFP. (21) TFP is skewed left in every year, and so there are more outliers at the bottom tail of the productivity distribution. However, the skewness and kurtosis measures are similar at the beginning and the ending of the period. Mean TFP is rising because the entire distribution of firm TFP is shifting to the right over time and not just because an atypical upper tail is driving up the mean. Lower tail firms in 2001 are more productive than lower tail firms in 1994. In 1994, three-quarters of firms had TFP measures in the negative range, but by 2001, only one-third had below average productivity.
TFP growth in Slovenia was faster than rates reported for 13 Organisation for Economic Co-operation and Development (OECD) manufacturing sectors over the 1980-1988 period (Benjamin and Ferrantino 2001). It is also faster than the annual TFP growth rates reported for the overall business sectors of those 13 OECD countries over the 1981-1995 period, and faster than 12 of the 13 over the 1996-2000 period (Gust and Marquez 2004). (22)
TFP Grew across All Firm Ownership Types
In Table 3 we report TFP growth for different firm ownership structures. Because there was little substantive difference in the time paths of TFP growth using the various estimation methods, we used the TFP levels based on OLS. The first column repeats the estimates from Table 2 of the average TFP level across all firms to provide a frame of reference. The second column lists average TFP for privately owned firms, and the third column lists TFP for all other firms. Firm efficiency was initially significantly lower in private firms, but TFP grew faster in private firms. Some of the growth was due to relatively efficient firms moving from the state sector to the private group, but sorting cannot explain much of the rise in TFP among private firms. First, the initial gap in efficiency is less than 0.03 log points, so the rise in efficiency is much larger than can be explained by sorting alone. Second, TFP is rising in both groups, not just the private group. If migration across firm types were the only factor, we would see decreases in TFP among the firms remaining in the nonprivate group as the more efficient state firms switched to the private group. One conclusion from Table 3 is that privately owned firms have more rapid TFP growth. However, a second conclusion is that TFP grows in state-owned enterprises as well, albeit more slowly. Over the full period, efficiency in privately owned firms rose 28%; whereas, it rose 18% in nonprivate firms.
Foreign-owned firms were slightly more efficient (0.036 log points) than average in 1994. They retained that TFP advantage through the end of the period. Over the eight-year period, TFP grew almost the same in foreign-owned firms as in the average manufacturing firm at about 25% growth.
Firms that entered limited liability arrangements may be under private, state, or a mixture of private and state ownership. They began the period with below average efficiency, but gained efficiency somewhat more rapidly than average. By 2001, limited liability firms were significantly more efficient than other firms, having experienced a 27.5% gain in TFP versus 24.9% for firms on average.
The firms with other ownership modalities began the period with a small TFP advantage, but experienced slower efficiency gains. By 2001, their TFP advantage had disappeared. Stock-owned companies also started the period with a TFP advantage, but experienced slower TFP growth. By 2001, stock-owned companies had significantly lower TFP levels than did the average manufacturing firm.
Birth and Death of Firms and Small Firm Growth Are Important Elements of TFP Growth
Table 4 reports TFP levels by firm size and by entry or exit status. Initially, large firms had a significant TFP advantage, but the faster TFP growth in small firms erased the gap by 1998. The implied efficiency growth was 25.6% in small firms versus 21.3% in large firms, so ignoring small firms understates efficiency growth.
Firms that opened for business after 1993 maintained a 0.01 log point TFP advantage over the average firm throughout the period. The average TFP advantage of 0.03 log points for new entrants over the full period is even larger than the annual advantage of 0.01 log points. The reason is that even though TFP levels for new entrants were similar to TFP levels for older firms, there were many more new entrants by the end of the period when prevailing efficiency levels were higher. Hence the weight of the effect of new entrants is to raise efficiency.
Firms that exited business by 2001 were significantly less efficient than the average firm. The disadvantage for firms destined to close was quite large, with an average TFP gap of 17% over the eight years. Eliminating these inefficient firms and growth of their more efficient rivals had a substantial effect on overall Slovene TFP growth.
We can illustrate the impact of firm entry and exit by replicating our TFP measurement and including only firms that were continually in business for the entire period. As shown in column 7, focusing on a balanced panel results in markedly smaller estimated TFP growth. The implied productivity gain of 0.177 log points, or 19.4%, is three-fourths of the true growth of 24.9% reported in column 1. If we further exclude firms with fewer than 100 employees, the productivity gains are even smaller at 0.154 log points, or 16.6%. Clearly, ignoring the productivity contributions of entrants, exiters, and small firms significantly biases downward the estimated growth in firm efficiency during the transition.
TFP Growth Occurs in Almost Every Manufacturing Sector
Table 5 carries the investigation of the distribution of TFP growth to the three-digit industry level. The included industries represent about two-thirds of all manufacturing firms. Industries were chosen so that they would have a sufficient number of firms to allow us to estimate the production function with some degree of precision. We estimated the Cobb-Douglas variant of Equation 1 to conserve on degrees of freedom. (23) The results support the view that TFP growth was widespread in the Slovenian economy. Only in the bakery industry did TFP levels fall, and in only three others did TFP rise by less than 10% (footwear, books and periodicals, and printing). In all other sectors TFP grew rapidly.
The evidence in Tables 2-5 tells a convincing story that virtually all manufacturing firms in Slovenia became more efficient as the transition progressed, regardless of sector, firm size, ownership modality, or date of entry. Firms that did not become more efficient went out of business. Furthermore, ignoring the role of small firms, entering firms and exiting firms can understate TFP growth by as much as one-third of the true productivity growth.
6. Regression Analysis of the Factors Affecting Total Factor Productivity
This section reviews the extent to which measures of market competition can be tied to the heterogeneity in TFP growth across firms. By embedding Equation 2 into the translog specification of Equation 1, we can identify factors that are tied to atypically rapid or slow increases in TFP. Our results are reported in Table 6.
Least Squares Analysis
To set a basis of comparison, the first specification includes only current firm attributes, including whether the firm was a new entrant. The results suggest that private firms and firms with other ownership are more efficient. Firms that entered after the passage of the Amended Law on Enterprises in 1993 are also more efficient; although, the impact is small. Stock-owned companies have marginally lower efficiency, and foreign-owned firms have comparable efficiency to domestically owned firms. At the bottom of the table, we have aggregated the impacts of the various changes in current firm attributes from 1994 to 2001 into a single metric. The implied TFP growth due to changes in current firms attributes is just 0.046, or 21% of the total growth over the period.
These first column results do not control for selection into the various ownership modalities. If, for example, only the most efficient firms are privatized, then private firms may be more productive because of efficiencies that predate the private ownership. To control for this selection bias, we add the remaining constant firm attributes that include the ultimate ownership status for the firm. The coefficients on the future status variables will capture the average effect of all firms that eventually become private firms. The coefficient on the current firms attributes will then capture the change in efficiency associated with the move to the new ownership status.
The coefficients on future attributes suggest that firms that were targeted for foreign or stock ownership had average productivity. Firms that came under other ownership or limited liability arrangements were more productive than average. The impacts are small, suggesting that there is not a strong selection process driving the results. However, there is strong evidence that firms that will ultimately go out of business have significantly lower TFP. The coefficient on EXIT implies that firms that are destined to close have TFP that is 18% below firms destined to survive. (24)
The combined effect of fixed firm attributes is 0.054 log points. (25) Adding these fixed firm attributes lowers the impact of current firm characteristics to 0.034 log points. Together, these current and fixed firm effects account for only 40% of TFP growth.
In column 3, we add the measures associated with the extent of competitive pressure in the industry. In contrast to the firm attributes, industry attributes are extremely important in explaining variation in firm efficiency. The Herfindahl index is based on domestic market shares, with 0 indicating perfect competition and 1 indicating the firm is a monopolist. The coefficient on the Herfindahl index implies that a monopolist would be 23% less efficient than an otherwise equivalent perfectly competitive firm. Firms in industries with a higher share of foreign-owned firms were significantly more efficient. Note that the foreign-owned firms themselves were not more efficient, but their presence made all firms in the industry more efficient. Private firms were more efficient, but their presence in the industry made all other firms more efficient as well. Firms destined to exit are presumably weak competitors. Firms in industries in which exiting firms have a higher output share are less efficient, even those that do not ultimately exit. Firms in industries in which entrants have a greater market share are also more efficient. There is no apparent impact of greater import penetration on incumbent firm efficiency. Overall, increases in these competitive pressures at the industry level increased firm productivity by 0.172 log points, or over three-quarters of the total TFP growth over the period. Meanwhile, current firm attributes drop still further in importance, as do firm fixed effects.
The specification in column 3 presumes that selection into ownership types is based solely on observable attributes, so that [[upsilon].sub.i] = 0 in Equation 5. If [[upsilon].sub.i] # 0, but E([u.sub.i]) = 0 for all i, selection will still depend only on observables, but a random effects estimator will provide added efficiency. Results from that specification are reported in column 4. The test for nonzero variance of the [[upsilon].sub.i] favored the random effects estimator over the least squares estimate of column 3. Individual results are similar to those in column 3 with the exception that current firm attributes lose magnitude and significance, while firm constant attributes gain strength. On the other hand, all measures of competitive pressure are now significant and pushing incumbent firms to be more efficient. Overall, industry competitive pressures account for 0.12 log points of TFP growth, which is half of the total TFP growth. Changes in firm fixed attributes due to the sorting out of weak firms and sorting in of strong firms can account for an additional 0.061 log points of TFP growth, or one-quarter of the overall gain. Firm current attributes become irrelevant.
The Hausman test suggests that unobservable (to the econometrician) productive attributes were also important, so we turn to the fixed-effect estimates. When fixed effects are imposed, only one firm-level current measure retains significance and the joint test that the coefficients on current firm attributes were equal to zero could not be rejected at standard significance levels. The aggregated impact of the [delta] suggests that current firm attributes explain none of the growth in TFP over the sample period.
On the other hand, all industry-level measures still retain signs that are consistent with the implied impact of market competition on productive efficiency. The only imprecisely estimated effect is that of the Herfindahl index, which implies a monopolist is only 2% less efficient than a perfectly competitive firm. Recall that the Herfindahl index was defined only on domestic production, so even a monopolist could face competition from foreign producers. (26) All the other estimated industry effects support the role of competition in significantly enhancing firm efficiency. Firms in industries that have higher market shares (net of own firm production) controlled by private firms, foreign-owned firms, and new entrants all had rising TFP. We emphasize that private firms and foreign-owned firms were not themselves more efficient, but their presence in an industry made all the firms in that industry more efficient. Firms in industries with higher import penetration were also more efficient. Firms are less efficient in industries with weak competitors, as indicated by a high net market share going to eventual exiters. Aggregating these industry effects, we find that rising industry competition can explain 0.091 log points, or 41% of the growth in TFP over the period. These represent external benefits from market competition, independent of the impact of current firm-specific factors.
The fixed-effects estimation eliminates the firm fixed factors from the model. Nevertheless, we know that a portion of the remaining efficiency gain must be due to the entry and exit of firms. If we take only the implied impacts of these two factors from column 3, we have that firm exits contributed 0.04 log points and new entrants contributed 0.005 log points to TFP growth. Together, firm entry and exit contribute an additional 21% of the estimated TFP growth over the sample period. The role of firm entry and exit in explaining efficiency growth in Slovenian manufacturing corresponds closely to the proportion of efficiency growth attributable to entry and exit in western economies, as summarized by Bartelsman and Doms (2000).
In sum, we estimate that 62% of the growth in firm efficiency between 1994 and 2001 can be attributable to the combined influence of increased industry market competition, to the birth of relatively efficient firms, and to the death of relatively inefficient firms. The remaining 38% of TFP growth is due to common effects across all firms. These could be due to business cycle effects or to reversion to mean efficiency levels following the initial shock of the transition. However, if the competitive policy changes, summarized in Figure 1, have any impact, they would be responsible for at least some of the 38% of TFP growth not explained by entry, exit, or our measures of industry-level competition. Our estimate that competitive pressures and sorting are responsible for 62% of TFP growth is therefore a lower-bound estimate of their true effects.
The Konings, Van Cayseele, and Warzynski (2005) investigation of price-cost margins in Romanian and Bulgarian firms finds similar roles for competition in fostering efficiency. They report that firms in more competitive sectors had significantly lower price cost margins. To the extent that their results generalize to Slovenia, our finding that firms in more competitive markets also had atypically large TFP gains, and hence falling production costs, suggests a double benefit of competition to a transition economy. Not only does competition foster more efficient production, but the benefits to consumers in the form of lower prices are even greater than the cost saving to the firm.
The relative unimportance of firm ownership structure in our study contrasts with findings in earlier studies. For example, Konings, Van Cayseele, and Warzynski (2005) find that privatized firms and foreign-owned firms had higher price-cost margins, other things equal. It is possible that our finding that ownership structure did not matter for efficiency gains is a consequence of the fact that all Slovene enterprises, state or private, faced the possibility of bankruptcy. As a result, state firms behaved more like private firms in Slovenia. Second, the scope for improvements of corporate governance in Slovenia may have been smaller than in other transition countries to the extent that managers in the worker self-managed Yugoslav system had considerable autonomy, which may have led some firms to be relatively efficient even before transition. Consequently, the privatized state firms did not experience atypical growth in efficiency. However, the most interesting finding in our context is that although privately owned firms or foreign-owned firms were not atypically efficient, industries with higher market shares attributable to private firms, foreign-owned firms, newly entered firms, or foreign-sourced goods had faster efficiency growth. When we ignore these industry-level aggregates of firm attributes, it appears that private or foreign ownership behave more as in previous studies. Consequently, earlier productivity effects attributed to ownership structure may have masked the market competition effects that dominate our article.
In Table 7, we replicate the fixed-effect estimates under various scenarios. In column 2, we report a variant of the Olley-Pakes (1996) estimation strategy. Their concern was in deriving unbiased estimates of [alpha] and [beta] in Equation 1, which is tangential to our concern with evaluating factors affecting the time path of [[epsilon].sub.ijt]. Nevertheless, there may be a concern that unmeasured firm heterogeneity in production is correlated with our industry-level measures of market competition. Under the assumption that firm exit and investment decisions are predicated on firm expectations of future market structure and factor prices, and that firm profits are increasing in capital, the firm's idiosyncratic productivity in Equation 4, [[phi].sub.it], can be written as
[[phi].sub.it] = h([i.sub.t], [a.sub.t], [k.sub.t], [[member of].sub.it]), (6)
where it is current investment, [a.sub.t] is the age of the firm, [k.sub.t] is the firm's capital stock, and [[epsilon].sub.it] is an approximation error assumed to be purely random. Inserting Equation 6 into Equation 4 gives us
[[psi].sub.it] = [f.sub.it][delta] + h([i.sub.t], [a.sub.t], [k.sub.t], [[member of].sub.it]). (4')
Olley and Pakes used an explicit formulation for h(.) to derive unbiased structural estimates of the [alpha] and [beta] in Equation 1. We are not interested in those parameters, so instead we replace h(*) with its second-order Taylor approximation and then estimate Equation 1 imposing Equations 2, 3, and 4'. The results are reported in the second column of Table 7. The proportion of TFP growth that can be jointly attributed to market competition falls somewhat from 41% to 38%. Nevertheless, the coefficients are very consistent in both sign and magnitude. The null hypothesis that all the coefficients in the approximation of h(.) are jointly zero could not be rejected. Current firm attributes continue to have no effect.
In column 3, we repeat the fixed-effect estimation excluding entrants and exiters from the sample. Current firm attributes continue to have no effect. The market competition measures still retain sign and significance. Their joint effect actually rises to 0.107, or 60% of the TFP growth in those firms. Thus, although using a balanced sample biases downward the overall estimate of TFP growth, it increases the proportion attributable to market competition.
The last column repeats the exercise but excludes firms with fewer than 100 employees. The conclusion that current firm attributes have no effect still holds. However, several of the market competition variables switch signs and significance from the full sample. The share of foreign-owned firms and market entrants now lower efficiency. The impact of import penetration is reduced by half, and the joint effect of market competition on efficiency falls to 0.052, or just 34% of the total TFP growth in those firms. It is apparent that excluding small firms from the sample causes significant downward bias in the estimated impact of market pressures on firm efficiency.
[FIGURE 2 OMITTED]
The results from Table 7 demonstrate that the general finding that increased market competition leads to increased efficiency holds when alternative definitions and assumptions about the error process are imposed. They also show that excluding small firms, entrants, and exiters can have large effects on the estimated magnitude of the market competition effect.
8. Structural Reform and Efficiency
We cannot perform a rigorous test of the relationship between structural policy reforms and our measures of productive efficiency because the EBRD data presented in Figure 1 are common across all firms and industries, and we can identify only relative efficiency effects between firms. Nevertheless, the correlation between the two series is of interest. In Figure 2 we present the simple bivariate relationship between our aggregate TFP measure and the average of the structural reform indexes shown in Figure 1. The average of the EBRD indexes explains 93% of the variation in the TFP measure over time. The average of the various EBRD indexes strongly outperformed any single index, consistent with the presumption that it is the mixture of liberalization policies supporting competition that is important, as opposed to any single policy. Of course, with only seven degrees of freedom, this can be viewed only as a suggestive result.
One of the oldest propositions in economics is that competition spurs economic efficiency. The introduction of competition was expected to improve the efficiency of formerly planned economies, moderating the adverse consequences of transition for output. Our evaluation of the data from Slovenian manufacturing is strongly supportive of the role of market competition. TFP growth in Slovenia over the period averaged 2.8% per year, a growth rate that compares favorably to most OECD countries. The TFP growth is broad based across industries, across private and state firms, across small and large firms, and at both the top and bottom of the efficiency distribution.
Changes from one ownership type to another had virtually no impact on firm TFP growth. Beyond a firm-specific, time-invariant productivity level, firm-level variables do not alter TFP. However, changes in industry attributes, such as the extent of foreign competition, foreign ownership, private ownership, and the market share of new entrants and eventual exiters, can explain 41% of TFP growth. An additional 21% can be attributed to the entrance of relatively efficient firms, and, more importantly, the exit of relatively inefficient establishments. These changes coincide with the timing of policy liberalization in Slovenia, suggesting that policies fostering market competition contributed to the growth of market pressures.
Many studies have attempted to measure the impact of transition by comparing the performance of state enterprises against that of private firms. For example, Frydman et al. (1999) find that private firms generate more sales than state enterprises, but have similar unit costs. Anderson, Lee, and Murrell (2000) find that Mongolian state enterprises had a TFP advantage over privately owned firms. Djankov and Murrell's (2002)review finds that privatization had a wide range of effects on productivity, most positive but some negative. In Slovenia, state firms are not protected from competition or risk of bankruptcy. Our results suggest that the distinction between firm ownership types is not as important as whether those firms face competitive pressures. However, firm ownership type did matter in that an increased industry share of private firms, foreign-owned firms, and imported goods raises competitive pressures on all firms in the industry, resulting in increased TFP growth for all surviving firms regardless of ownership structure.
Small firms, new market entrants, and exiting firms have a large impact on measured TFP growth in transition. Efficiency gains appear to occur over time and not at one instance. Our findings suggest that past studies that concentrated only on large firms, balanced panels, and short time frames may have understated the efficiency gains that resulted from the transition to market.
Received July 2006; accepted September 2008.
Anderson, James H., George Korsun, and Peter Murrell. 2000. Which enterprises (believe they) have soft budgets after mass privatization. Journal of Comparative Economics 28:219-46.
Anderson, James H., Young Lee, and Peter Murrell. 2000. Competition and privatization amidst weak institutions: Evidence from Mongolia. Economic Inquiry 38:527-49.
Bartelsman, Eric J., and Mark Doms. 2000. Understanding productivity: Lessons from longitudinal microdata. Journal of Economic Literature 38:569-94.
Benjamin, Nancy, and Michael J. Ferrantino. 2001. Trade policy and productivity growth in OECD manufacturing. International Economic Journal 15:95-115.
Blanchard, Olivier, and Michael Kremer. 1997. Disorganization. Quarterly Journal of Economics 112:1091-126.
Blanchflower, David, and Stephen Machin. 1996. Product market competition wages and productivity: International evidence from establishment-level data. Annales-d' Economie-et-de-Statistique 17:21933.
Boeri, Tito. 2000. Structural change, welfare systems and labour reallocation. Oxford: Oxford University Press.
Boeri, Tito, and Catherine Terrell. 2002. Institutional determinants of labor reallocation in transition. Journal of Economic Perspectives 16:51-76.
Bojnec, Stefan, and Aria Xavier. 2004. Entry and exit in transition economies: The Slovenian manufacturing sector. Post-Communist Economies 16:191-214.
Brada, Josef C. 1996. Privatization is transition--Or is it? Journal of Economic Perspectives 10:67-86.
Campos, Nauro F., and Fabrizio Coricelli. 2002. Growth in transition: What we know, what we don't, and what we should. Journal of Economic Literature 40:793-836.
Djankov, Simeon, and Peter Murrell. 2002. Enterprise restructuring in transition: A quantitative survey. Journal of Economic Literature 40:739-92.
Estrin, Saul. 2002. Competition and corporate governance in transition. Journal of Economic Perspectives 16:101-24. European Bank for Reconstruction and Development (ERBD). Transition Report. Various Issues. London, UK. Frydman, Roman, Cheryl Gray, Marek Hessel, and Andrzej Rapaczynski. 1999. When does pfivatization work? The impact of private ownership on corporate performance in the transition economies. Quarterly Journal of Economics 114:1153-91.
Gust, Christopher, and Jaime Marquez. 2004. International comparisons of productivity growth: The role of information technology and regulatory practices. Labour Economics 11(1): 33-58.
Haltiwanger, John, and Milan Vodopivec. 2003. Worker flows, job flows and firm wage policies: An analysis of Slovenia. Economics of Transition 11:253-90.
Institute of Macroeconomic Analysis and Development (IMAD). 2003. Development report 2003. Ljubljana, Slovenia: IMAD.
Joskow, Paul L., and Nancy L. Rose. 1989. The effects of economic regulation. In Handbook of industrial organization, vol. 2, edited by R. Schmalensee and R. Willig. Amsterdam: Elsevier Science Publishers, pp. 1449-1506.
Konings, Jozef, Patrick Van Cayseele, and Frederic Warzynski. 2005. The effects of privatization and competitive pressure on firms' price-cost margins: Micro evidence from emerging economies. Review of Economics and Statistics 87(1): 124-34.
Li, David D., and Minsong Liang. 1998. Causes of the soft budget constraint: Evidence on three explanations. Journal of Comparative Economics 26(1): 104-16.
McMillan, John, and Christopher Woodruff. 2002. The central role of entrepreneurs in transition economies. Journal of Economic Perspectives 16:153-70.
Megginson, William L., and Jeffrey M. Netter. 2001. From state to market: A survey of empirical studies on privatization. Journal of Economic Literature 39:321-89.
Nickell, Stephen. 1996. Competition and competitive performance. Journal of Political Economy 104:724-46.
Olley, G. Steven, and Ariel Pakes. 1996. The dynamics of productivity in the telecommunications equipment industry. Econometrica 64:1263-97.
Orazem, Peter F., and Milan Vodopivec. 2000. Male-female differences in labor market outcomes during the early transition to market: The cases of Estonia and Slovenia. Journal of Population Economics 13:289-303.
Riboud, Michelle, Carolina Sanchez-Paramo, and Carlos Silva-Jauregui. 2001. Does eurosclerosis matter? Institutional reform and labor market performance in central and Eastern European countries in the 1990s. Mimeo, the World Bank.
Simoneti, Marko, Andreja Bohm, Marko Rems, Matija Rojec, Joze P. Damijan, and Boris Majcen. 2001. Secondary privatization in Slovenia: Evolution of ownership structure and company performance following mass privatization. Warsaw: Center for Economic and Social Research (CASE) Reports No. 46
Svejnar, Jan. 2002. Transition economies: Performance and challenges. Journal of Economic Perspectives 16:3-28.
Vodopivec, Milan. 1993. Determination of earnings in Yugoslav firms: Can it be squared with labor management? Economic Development and Cultural Change 41:337-48.
Peter F. Orazem * and Milan Vodopivec ([dagger])
* Department of Economics, 267 Heady Hail, Iowa State University, Ames, IA 50011-1070, USA; E-mail firstname.lastname@example.org; corresponding author.
([dagger]) World Bank, 1818 H Street NW, Washington, DC 20433, USA; E-mail email@example.com.
The authors are grateful for support from a grant provided by the U.S. Department of State's Program for Study of Eastern Europe and the Independent States of the Former Soviet Union (Title VIII) and administered by the William Davidson Institute, and from World Bank research project RF-P064129-RESE-BBRSB. The article would not have been possible without tremendous help in obtaining the data from the Statistical Office of Slovenia, the Agency of Slovenia for Public Statistics and Services, and the Bank of Slovenia. We thank the Bank of Slovenia for conducting analyses using restricted data on foreign ownership. Helpful comments and advice were provided by Mico Mrkaic, Matija Rojec, Marko Simoneti, Andreja Vodopivec, and the referees. Tomaz Rejec and Jakob Tomse provided excellent assistance in setting up the data sets, and Donna Otto prepared the manuscript. The opinions, findings, and conclusions or recommendations expressed herein are those of the authors and do not necessarily reflect those of the World Bank, the U.S. Department of State, or the William Davidson Institute.
(1) The Joskow and Rose (1989) summary suggests that increased market competition from deregulation in the United States and privatization of public enterprise in Europe generally led to increases in labor productivity. Megginson and Netter (2001) present a more recent review including other areas of the world and derive similar conclusions. The few studies that examine the role of competition in fostering efficiency in unregulated environments yield mixed results. For example, in two studies of British firms, Nickell (1996) finds that competition enhances efficiency; whereas, Blanchflower and Machin (1996) find no effect.
(2) Of course, much of the variation reflects differences in methodology and measures of firm performance. However, even the most careful studies that control for selection problems can generate conflicting results. For example, Anderson, Lee, and Murrell (2000) find that state enterprises were more efficient than private firms; whereas, Frydman et al. (1999) find that privatization raises measures of firm performance. The differences may be in the measure of firm performance used. When Frydman et al. use a measure of efficiency, namely, unit cost, the differences between private and state enterprise disappear. Their other measures (revenue growth, employment growth, and revenue per employee) do not have an obvious connection to efficiency.
(3) For the purpose of this study, we define other ownership to include cooperative ownership and split state and private ownership, two categories used by the Statistical Office of Slovenia for firms that do not fit naturally into the other firm types.
(4) Vodopivec (1993) discusses this system in detail.
(5) A more detailed summary of the Slovenian reforms can be found in Orazem and Vodopivec (2000).
(6) Bojnec and Xavier's (2004) analysis of the number of firms by industry in Slovenia suggests exit rates from manufacturing of 5% per year, roughly consistent with our longitudinal data on Slovenian manufacturing firms.
(7) The likelihood that the government will prevent state firm failure may vary even within countries. Anderson, Korsun, and Murrell (2000) find that only 27% of Mongolian state enterprise thought the government would bail them out at least partially if the firm failed, with 73% stating the government would do nothing to help. In China, Li and Liang (1998) find that state enterprises with negative cash flow did not reduce their employment of redundant workers, apparently because all expected the government to make up the losses.
(8) The EBRD index did not include labor market reforms, but Slovenia also took a gradualist approach in adjusting labor market policies. Early on, the government imposed many provisions to protect jobs in traditional sectors. By 1991, restrictions on layoffs and mandated severance were reduced. Despite the liberalization, Riboud, Sanchez-Paramo, and Silva-Jauregui (2001) conclude that Slovenia's labor policies were the most restrictive of the formerly planned economies that were being targeted for accession to the European Union. Boeri and Terrell (2002) provide a comparative review of labor market policies in transition economies.
(9) Konings, Van Cayseele, and Warzynski (2005) derive a dual formulation of the traditional Solow residual framework that we employ. Their method assumes that firms are maximizing profits, an assumption that is inconsistent within our framework in which some firms may become more efficient over time due to competitive pressures, while other firms that are more insulated from competition are able to remain inefficient. The Konings, Van Cayseele, and Warzynski framework is aimed at estimating price-cost margins and not firm efficiency per se, and they ultimately difference away the Solow residual that is the focus of our analysis. Nevertheless, our results correspond with theirs in interesting ways, as will be discussed below.
(10) Note that by construction, [[epsilon].sub.ijt] is orthogonal to the inputs, so it is productivity attached to the firm's overall production but not to specific inputs.
(11) We could also specify a time varying error component that is common across all firms and industries. The most likely source of such common national shocks would be government tax and transfer policies and regulatory policies. However, these policies were stable over the sample period.
(12) Bartelsman and Doms (2000) conclude that there is considerable persistence in firm productivity so that highly productive firms in one year are likley to be highly productive in other years. This suggests that the fixed effect component [[upsilon].sub.i] is likely to be important.
(13) This is almost certainly true. Simoneti et al. (2001) find that insider investment is heaviest in firms that had higher profits in the years preceding privatization. It is not clear if the higher profitability is a permanent or transitory state. Our own results suggest the latter, in that firms that become stock-owned have slower TFP growth than other firms.
(14) Note that it is more efficient to estimate the system of equations in one step than to estimate Equation 1, derive estimates of [[epsilon].sub.ijt], and then to estimate Equation 2 with appropriate substitutions for q [[eta].sub.jt], [[psi].sub.it] and [[theta].sub.i].
(15) We distinguish private and state ownership, as well as ownership by domestic and foreign owners. We also have information about whether the firm is a publicly traded stock company or a limited liability company.
(16) For more information about the data sets, see Haltiwanger and Vodopivec (2003), who use the same sources.
(17) Note that to suit our empirical analysis, the variables ENTRY and EXIT are defined in a specific way. Namely, ENTRY is set equal to one in all years if the firm came into existence any year after 1993 (thus a firm that was founded in 1995 is considered an entrant also in 1996 and subsequent years of analysis). Similarly, EXIT is set equal to one in all years if the firm had no employees in 2001. Thus, a firm that began operations in 1996 and ceased operations in 2000 will have ENTRY = 1 and EXIT= 1 for all years of its existence.
(18) Most studies of market efficiency in western economies have concentrated on manufacturing because data on inputs and outputs are more readily available and comparable across firms. In our case, we did not have sufficient detail on input and output prices to allow us to perform the analysis on industries outside manufacturing. For example, we would not be able to assess how much of the revenue changes in the service sector was due to increases in service output versus increases in unit prices. However, our results for manufacturing may not carry over to all sectors. According to IMAD (2003), growth of market-oriented services in Slovenia has been so slow that the gap in the share of these services in GDP rose between Slovenia and the European Community from 1995 to 2001. The slow growth in these services has been attributed to their being shielded from market pressures, causing the service sector to lag behind other sectors in productivity growth.
(19) Employment shares are similar. New entrants were responsible for 16% of all employment in manufacturing in 2001, while firms that exit by 2001 employed 15% of all manufacturing employees in 1994.
(20) Computed, for example, as 100*(exp(0.222) - 1).
(21) These measures were similar across the various TFP measures, and so these are reported as a representative example.
(22) Finland had faster TFP growth over the 1996-2000 period.
(23) This constrains all the [[beta].sub.kl] = 0 in Equation 1.
(24) Bojnec and Xavier (2004) report that firms are more likely to exit in sectors with greater import penetration and lower (more competitive) Herfindahl indexes, consistent with our presumption that competition helps force exits of inefficient firms.
(25) Even though these are firm fixed effects, the composition of firms is changing over time to favor more efficient factors. Firms whose fixed factors are associated with lower efficiency are exiting, while entering firms tend to avoid attributes that lower efficiency.
(26) The simple correlation between the Herfindahl index and import share was 0.27, suggesting a modest increase in import penetration in more concentrated sectors.
Table 1. Sample Means and Standard Deviations for the Full Sample and Means for 1994 and 2001 1994-2001 Std. Variable Mean Dev. Endogenous TFP Total factor productivity from OLS 0.000 0.363 TFPfe Total factor productivity assuming 0.000 0.406 fixed effects TFPre Total factor productivity assuming 0.023 0.367 random effects lnrq Log of real output 6.01 2.03 Inputs lnrk Log of real capital stock 4.62 2.45 lnrm Log of real value of materials 5.46 2.08 lnuniv Log of two- or four-year university 0.60 1.11 educated employees lnhigh Log of high school educated 1.70 1.64 employees lnprim Log of employees with < high 1.02 1.59 school education lnmonth Log of months of operation 2.481 0.073 Current firm attributes state owned Firm is currently state owned 0.101 0.301 (reference) private Firm is private in current year 0.837 0.369 stockco Firm currently issues publicly 0.075 0.264 traded stock ltdliab Firm is currently a limited 0.858 0.349 liability firm other Firm is currently under other 0.061 0.239 ownership type forown Firm is currently at least 10% 0.098 0.297 foreign owned Constant firm attributes ENTRY Firm's birth year after 1993 0.254 0.435 EXIT Firm has no employees by 2001 0.111 0.314 PRIVATE Firm becomes private by 2001 0.884 0.321 STOCKCO Firm issues publicly traded 0.108 0.311 stock by 2001 LTDLIAB Firm becomes a limited liability 0.895 0.306 firm by 2001 OTHER Firm under other ownership type 0.089 0.285 by 2001 FOROWN Firm under at least 10% foreign 0.118 0.323 ownership by 2001 Four-digit industry attributes (a) HERF Herfindahl concentration index 0.137 0.151 PRIVSHR Industry market share of private 0.589 0.305 firms FORSHR Industry market share of firms 0.166 0.207 with forown = 1 ENTSHR Industry market share of 0.147 0.142 firms entering after 1993 EXITSHR Industry market share of firms 0.059 0.097 that will exit by 2001 IMPORTSHR Industry market share of imports 0.338 0.22 N 28047 1994 2001 Variable Mean Mean Endogenous TFP Total factor productivity from OLS -0.137 0.086 TFPfe Total factor productivity assuming -0.159 0.086 fixed effects TFPre Total factor productivity assuming -0.116 0.108 random effects lnrq Log of real output 6.188 6.088 Inputs lnrk Log of real capital stock 4.797 4.667 lnrm Log of real value of materials 5.746 5.433 lnuniv Log of two- or four-year university 0.783 0.565 educated employees lnhigh Log of high school educated 1.890 1.693 employees lnprim Log of employees with < high 1.303 0.948 school education lnmonth Log of months of operation 2.480 2.483 Current firm attributes state owned Firm is currently state owned 0.284 0.039 (reference) private Firm is private in current year 0.636 0.906 stockco Firm currently issues publicly 0.037 0.085 traded stock ltdliab Firm is currently a limited 0.823 0.862 liability firm other Firm is currently under other 0.08 0.054 ownership type forown Firm is currently at least 10% 0.076 0.102 foreign owned Constant firm attributes ENTRY Firm's birth year after 1993 0.074 0.324 EXIT Firm has no employees by 2001 0.265 0.000 PRIVATE Firm becomes private by 2001 0.788 0.916 STOCKCO Firm issues publicly traded 0.154 0.094 stock by 2001 LTDLIAB Firm becomes a limited liability 0.868 0.893 firm by 2001 OTHER Firm under other ownership type 0.123 0.074 by 2001 FOROWN Firm under at least 10% foreign 0.106 0.111 ownership by 2001 Four-digit industry attributes (a) HERF Herfindahl concentration index 0.14 0.041 PRIVSHR Industry market share of private 0.204 0.732 firms FORSHR Industry market share of firms 0.100 0.230 with forown = 1 ENTSHR Industry market share of 0.036 0.205 firms entering after 1993 EXITSHR Industry market share of firms 0.168 0.000 that will exit by 2001 IMPORTSHR Industry market share of imports 0.196 0.350 N 2904 4244 (a) Except for IMPORTSHR, these measures are net of own firm's output share. Table 2. Time Path of Alternative Estimates of Firm Total Factor Productivity in Slovenia Manufacturing, 1994-2001 All Firms, All Firms, All Firms, Year TFP (a) TFPfe (b) TFPre (c) 1994 -0.136 -0.158 -0.115 1995 -0.115 -0.119 -0.090 1996 -0.048 -0.046 -0.023 1997 0.010 0.014 0.034 1998 0.015 0.021 0.039 1999 0.032 0.036 0.055 2000 0.081 0.085 0.104 2001 0.086 0.086 0.108 1994-2001 0.222 0.244 0.223 Average 0.000 0.000 0.023 Correlation matrix of the alternative TFP estimates over 28,047 observations TFP TFPfe TFPre TFP 1.0 TFPfe 0.90 1.0 TFPre 0.99 0.941 1.0 Skewness and kurtosis of the TFP distribution Year Skewness Kurtosis 1994-1996 -1.88 44.13 1998-2001 -1.89 40.00 (a) TFP is total factor productivity measured as the error from OLS estimates of the translog production function (Eqn. 1). (b) TFPfe is total factor productivity measured as the error derived from a fixed-effects estimate of the translog production function. (c) TFPre is total factor productivity measured as the error derived from a random effects estimate of the translog production function. Table 3. Time Path of Firm Total Factor Productivity by Slovenian Manufacturing Firm Ownership Type, 1994-2001 (a) All Firms, Not Foreign Year TFP Private (b) Private (b) Owned (b) 1994 -0.136 -0.147 ** -0.119 ** -0.107 1995 -0.115 -0.116 * -0.143 * -0.113 1996 -0.048 -0.053 -0.079 -0.017 1997 0.010 0.015 ** -0.018 ** -0.005 1998 0.015 0.022 ** -0.032 ** 0.027 1999 0.032 0.036 ** 0.000 ** 0.054 2000 0.081 0.085 * 0.046 * 0.094 2001 0.086 0.090 0.050 0.120 * 1994-2001 0.222 0.247 0.169 0.227 Average 0.000 0.017 ** -0.053 ** 0.026 ** Limited Liability Other Stock Year Firm (b) Ownership (b) Company (b) 1994 -0.148 ** -0.107 -0.122 1995 -0.117 -0.105 -0.052 ** 1996 -0.048 0.009 ** -0.001 ** 1997 0.010 0.021 0.044 ** 1998 0.015 0.011 0.015 1999 0.032 0.054 0.032 2000 0.084 * 0.083 0.061 2001 0.095 ** 0.087 0.039 ** 1994-2001 0.243 0.194 0.161 Average 0.001 0.012 0.018 ** (a) Total factor productivity is measured as the error from OLS estimates of the translog production function (Eqn. 1). (b) t-tests of the null hypothesis that mean TFP are equal between the stated ownership type versus all other firms were conducted, allowing for different variances in the two groups. * Significant differences at the 0.10 confidence level. ** Significance at the 0.05 level. Table 4. Time Path of Firm Total Factor Productivity by Slovenian Manufacturing, Firm Size, Entry Cohort, and Mortality (a) All Firms, <100 100+ Year TFP Employees (b) Employees (b) Entry (b) 1994 -0.136 -0.142 ** -0.101 ** -0.129 1995 -0.115 -0.118 * -0.090 * -0.115 1996 -0.048 -0.050 ** -0.025 ** -0.058 1997 0.010 0.008 * 0.031 * 0.006 1998 0.015 0.014 0.024 0.022 1999 0.032 0.034 0.018 0.045 2000 0.081 0.081 0.084 0.092 2001 0.086 0.086 0.092 0.097 Difference between 2001 and 1994 0.222 0.228 0.193 0.226 Average 0.000 -0.0001 0.001 0.031 ** Balanced Balanced Large Panel, (c) Firm Panel, (d) Year Exit (b) TFP (a) TFP (a) 1994 -0.229 ** -0.112 -0.087 1995 -0.223 ** -0.097 -0.073 1996 -0.164 ** -0.032 -0.012 1997 -0.114 ** 0.016 0.035 1998 -0.126 ** 0.012 0.024 1999 -0.182 ** 0.027 0.009 2000 -0.205 ** 0.067 0.066 2001 0 (e) 0.065 0.067 Difference between 2001 and 1994 0.229 0.177 0.154 Average -0.181 ** 0.000 0.000 (a) TFP is measured as the error from OLS estimates of the translog production function (Eqn. 1). (b) t-tests of the null hypothesis that mean TFP are equal between the stated ownership type versus all other firms were conducted, allowing for different variances in the two groups. (c) TFP estimate over the subsample of firms in continuous production from 1994 through 2001. (d) TFP estimate over the subsample of firms with more than 100 employees in continuous production from 1994 through 2001 (e) By definition, TFP = 0 for firms no longer in business. * Significant differences at the 0.10 confidence level. ** Significance at the 0.05 level. Table 5. Total Factor Productivity Estimates by Detailed Manufacturing Sectors Industry SIC (b) Share (c) 1994 1995 Bakery 15.8 2.9% 0.055 0.008 Woven textiles 17.4, 17.5 1.6% -0.105 -0.087 Clothing 18.2 8.0% -0.125 -0.042 Footwear 19.2, 19.3 1.9% 0.02 -0.14 Lumber 20.1 2.3% -0.07 -0.095 Plywood 20.2 2.0% -0.134 -0.075 Wooden crates 20.4 1.3% -0.15 -0.103 Paper products 21.21-21.23 0.9% -0.135 -0.167 Book, 22.11-22.13 1.4% -0.021 -0.146 periodicals Printing 22.21, 22.22 2.6% -0.065 -0.118 Rubber 25.1 0.8% -0.097 -0.183 Plastics 25.2 5.3% -0.119 -0.179 Cement and 26.6, 26.7 1.2% -0.121 -0.149 stone products Metal castings 27.5 0.7% -0.055 -0.075 for plumbing, etc. Metal finishing 28.5 9.8% -0.108 -0.089 Cutlery, hand 28.6 2.4% -0.044 -0.113 tools Manufacturing 29.2 1.7% -0.13 -0.149 equipment Power hand 29.5 2.0% -0.117 -0.166 tools Electrical 31.6 3.5% -0.221 -0.107 machinery Radio, TV, 32 1.9% -0.185 -0.092 communication equip. Precision testing 33.2, 33.3 1.2% -0.286 -0.15 and control Furniture 36.1 8.3% -0.148 -0.053 Industry 1996 1997 1998 1999 Bakery -0.054 0.039 0.018 -0.008 Woven textiles -0.004 0.048 -0.001 -0.009 Clothing 0.014 0.032 0.037 -0.024 Footwear -0.03 0.007 0.02 0.023 Lumber -0.064 -0.017 0.025 0.041 Plywood -0.046 -0.015 -0.004 0.009 Wooden crates -0.084 0.053 0.029 0.024 Paper products -0.004 0.051 0.03 0.051 Book, 0.002 0.021 0.032 0.065 periodicals Printing -0.031 0.06 0.053 0.093 Rubber -0.054 0.037 0.05 -0.149 Plastics -0.06 -0.02 0.026 0.038 Cement and -0.096 0 0.029 0.082 stone products Metal castings 0 -0.074 0.028 0.053 for plumbing, etc. Metal finishing -0.026 -0.077 -0.035 0.033 Cutlery, hand -0.026 -0.052 0.006 0.072 tools Manufacturing -0.079 -0.078 -0.04 -0.042 equipment Power hand -0.045 0.014 -0.011 0.019 tools Electrical -0.077 -0.037 0.036 0.078 machinery Radio, TV, -0.093 0.045 0.067 0.131 communication equip. Precision testing -0.112 -0.019 0.019 0.018 and control Furniture -0.019 0.028 0.011 0.016 Cumulative Industry 2000 2001 1994-2001 Bakery 0.028 -0.067 -0.122 Woven textiles 0.04 0.042 0.147 Clothing 0.036 0.076 0.201 Footwear 0.02 0.03 0.01 Lumber 0.018 0.106 0.176 Plywood 0.117 0.049 0.183 Wooden crates 0.072 0.124 0.274 Paper products 0.022 0.063 0.198 Book, 0.022 0.003 0.024 periodicals Printing 0.025 0.003 0.068 Rubber 0.113 0.117 0.214 Plastics 0.114 0.09 0.209 Cement and 0.064 0.059 0.18 stone products Metal castings 0.068 0.051 0.106 for plumbing, etc. Metal finishing 0.069 0.029 0.137 Cutlery, hand 0.086 0.073 0.117 tools Manufacturing 0.092 0.152 0.282 equipment Power hand 0.192 0.221 0.338 tools Electrical 0.121 0.226 0.447 machinery Radio, TV, 0.287 0.288 0.473 communication equip. Precision testing 0.11 0.153 0.439 and control Furniture 0.065 0.084 0.232 (a) Total Factor Productivity measured by residuals from OLS estimation of the Cobb-Douglas form of Equation 1, restricting all second-order coefficients to zero. (b) Industrial classification numbers used for the Slovenian National Income and Product Accounts. (c) Industry's share of total manufacturing output in Slovenia. These sectors represent approximately two-thirds of Slovenian manufacturing output over the period. Table 6. Estimation of Impacts of Firm and Industry Variables on Total Factor Productivity in Slovenian Manufacturing Firms, 1994-2001 1: OLS: 2: OLS: 3: OLS: Variable Variable and Firm and Firm Fixed Firm Industry Attributes Attributes Attributes Current firm attributes, [delta] private 0.159 ** 0.117 ** 0.066 ** (16.70) (8.65) (4.73) stockco -0.036 ** -0.025 -0.048 ** (2.71) (1.41) (2.68) ltdliab -0.014 -0.038 ** -0.035 ** (1.47) (2.67) (2.46) other 0.123 ** 0.069 ** 0.039 ** (10.20) (4.01) (2.23) forown -0.008 0.003 -0.014 (1.04) (0.16) (0.80) Constant firm attributes, [mu] ENTRY 0.035 ** 0.028 ** 0.020 ** (6.83) (5.50) (3.69) EXIT -0.201 ** -0.154 ** (28.6) (21.2) PRIVATE 0.008 0.018 (0.60) (1.35) STOCKCO -0.001 0.02 (0.04) (1.31) LTDLIAB 0.053 ** 0.058 ** (4.02) (4.37) OTHER 0.032 ** 0.037 ** (2.30) (2.68) FOROWN -0.001 0.021 (0.05) (1.34) Industry attributes, [gamma] HERF -0.258 ** (17.1) PRIVSHR 0.152 ** (17.9) FORSHR 0.117 ** (10.4) ENTSHR 0.043 ** (2.45) EXITSHR -0.271 ** (11.10) IMPORTSHR -0.014 (1.32) N 27,958 27,949 25,735 [R.sup.2] 0.97 0.97 0.97 Estimated growth components from 1994 to 2001 ([[bar.f].sub.01] - 0.046 0.034 0.018 [[bar.f].sub.94]) [delta] ([[bar.F].sub.01] - NA 0.054 0.042 [[bar.F].sub.94]) [gamma] ([[bar.I].sub.01] - NA NA 0.172 [[bar.I].sub.94]) [gamma] Total TFP growth 0.222 0.222 0.222 4: Random 5: Fixed Effects: Firm Effects: Firm and Industry and Industry Attributes Attributes Current firm attributes, [delta] private 0.035 ** 0.019 (2.51) (1.23) stockco -0.040 ** -0.021 (2.33) (1.10) ltdliab -0.047 ** -0.038 ** (3.06) (2.16) other 0.025 0.01 (1.47) (0.53) forown -0.024 -0.027 (1.47) (1.57) Constant firm attributes, [mu] ENTRY 0.030 ** (Dropped) (3.08) EXIT -0.173 ** (Dropped) (15.0) PRIVATE 0.100 ** (Dropped) (5.08) STOCKCO -0.050 ** (Dropped) (2.11) LTDLIAB 0.055 ** (Dropped) (2.59) OTHER 0.050 ** (Dropped) (2.30) FOROWN -0.01 (Dropped) (0.52) Industry attributes, [gamma] HERF -0.122 ** -0.021 (7.27) (1.08) PRIVSHR 0.098 ** 0.050 ** (11.00) (4.91) FORSHR 0.069 ** 0.062 ** (5.48) (4.25) ENTSHR 0.073 ** 0.081 ** (3.58) (3.35) EXITSHR -0.186 ** -0.149 ** -8.05 -5.86 IMPORTSHR 0.026 * 0.099 ** (1.74) (4.99) N 25,735 25,735 [R.sup.2] 0.97 0.97 Estimated growth components from 1994 to 2001 ([[bar.f].sub.01] - 0.012 0.002 [[bar.f].sub.94]) [delta] ([[bar.F].sub.01] - 0.061 NA [[bar.F].sub.94]) [gamma] ([[bar.I].sub.01] - 0.120 0.091 [[bar.I].sub.94]) [gamma] Total TFP growth 0.244 0.223 Coefficients are taken from translog production function estimation of Equation 1 augmented with the variables that make up Equation 2. The coefficients on the translog specification, including all first- and second-order terms in the logs of real capital, materials, numbers of university, high school, and primary school trained workers, are withheld to conserve space. Coefficients on the log of months of firm operation, dummy variables indicating no employees and education group, and the constant are also suppressed. t-statistics are reported in parentheses. * Significance at the 0.10 level. ** Significance at the 0.05 level. Table 7. Alternative Fixed-Effect Estimation of Impacts of Firm and Industry Variables on Total Factor Productivity in Slovenian Manufacturing Firms, 1994-2001 1 2 Current firm Column 5 of attributes, [delta] Table 6 Olley-Pakes private 0.018 0.006 (1.19) (0.37) stockco -0.021 0.002 (1.13) (0.12) ltdliab -0.038 ** -0.038 ** (2.15) (1.97) other 0.010 -0.003 (0.54) (0.16) forown -0.010 -0.007 (0.61) (0.40) Industry attributes, [gamma] HERF -0.023 -0.019 (1.16) (0.93) PRIVSHR 0.054 ** 0.048 ** (5.29) (4.27) FORSHR 0.107 ** 0.094 ** (5.23) (4.66) ENTSHR 0.070 ** 0.077 ** (2.88) (3.02) EXITSHR -0.149 ** -0.117 ** (5.90) (4.20) IMPORTSHR 0.097 ** 0.082 ** (4.90) (3.66) N 25,726 22,447 [R.sup.2] 0.97 0.96 Estimated growth components from 1994 to 2001 ([[bar.f].sub.01] - 0.002 0.0001 [[bar.f].sub.94]) [delta] ([[bar.I].sub.01] - 0.091 0.084 [[bar.I].sub.94]) [gamma] Total TFP growth 0.223 0.223 3 4 Current firm Balanced Balanced Large attributes, [delta] Panel Firm Panel private 0.022 0.034 (1.36) (1.44) stockco -0.030 -0.024 (1.58) (0.96) ltdliab -0.043 ** -0.041 (2.44) (1.57) other 0.020 0.026 (1.06) (1.06) forown -0.015 -0.057 (0.78) (2.37) Industry attributes, [gamma] HERF -0.013 -0.021 (0.56) (0.34) PRIVSHR 0.061 ** 0.060 ** (5.44) (2.13) FORSHR 0.066 ** -0.010 (4.13) (0.32) ENTSHR 0.132 ** -0.149 ** (4.86) (2.21) EXITSHR -0.156 ** -0.217 ** (5.66) (2.89) IMPORTSHR 0.104 ** 0.051 (4.83) (0.99) N 16,914 2,258 [R.sup.2] 0.97 0.93 Estimated growth components from 1994 to 2001 ([[bar.f].sub.01] - 0.002 0.004 [[bar.f].sub.94]) [delta] ([[bar.I].sub.01] - 0.107 0.052 [[bar.I].sub.94]) [gamma] Total TFP growth 0.177 0.154 Column I is taken from the last column in Table 6. Column 2 is a variant of the Olley-Pakes (1996) specification. In this application, we supplement the specification in column I with i, [i.sup.*]k, [a.sup.*]k, [a.sup.*]i, and [i.sup.2] where i is the logarithm of real investment, k is the logarithm of the capital stock, and a is the logarithm of the firm's age. Linear and quadratic terms in firm age are controlled by the faced effect. The null hypothesis that the five terms can be excluded could not be rejected at standard significance levels (F(5,16,893)) = 1.81. Column 3 replicates Column 1 but excludes firms that enter or exit the sample. Column 4 replicates Column 3 but excludes firms with fewer than 100 employees. Other notes are the same as in Table 6.