- Email: [email protected]
Privatization's impact on private productivity: The case of Brazilian iron ore
Review of Economic Dynamics 11 (2008) 745–760 www.elsevier.com/locate/red
Privatization’s impact on private productivity: The case of Brazilian iron ore ✩ James A. Schmitz Jr. a,∗ , Arilton Teixeira b a Federal Reserve Bank of Minneapolis, 90 Hennepin Avenue, Minneapolis, MN, USA b Fundacao Capixaba de Pesquisa, Brazil
Received 29 March 2005; revised 13 December 2007 Available online 17 January 2008
Abstract A major motivation for the recent wave of privatizations of state-owned enterprises (SOEs) was a belief that privatization would increase SOE productivity. There are now many studies showing most privatizations achieved this goal. Our theme is that the productivity gains from privatization are much more general and widespread than has typically been recognized in this literature. In assessing the productivity gains from privatization, the literature has only examined the productivity gains accruing at the privatized SOEs. But privatization may have significant impact on the private producers that often exist side-by-side SOEs. In this paper we show that this was indeed the case when Brazil privatized its SOEs in the iron ore industry. That is, after their privatization, the iron ore SOEs dramatically increased their labor productivity, but so did the private iron ore companies in the industry. © 2008 Elsevier Inc. All rights reserved. JEL classification: L33; L70 Keywords: State-owned enterprises; Privatization; Productivity
1. Introduction A major motivation for the recent wave of privatization of state-owned enterprises (SOEs) was a belief that privatization would increase SOE productivity. There are now many studies showing most privatization achieved this goal. Our theme is that the productivity gains from privatization are much more general and widespread than has typically been recognized in this literature. In assessing the productivity gains from privatization, the literature has only examined the productivity gains accruing at the privatized SOEs. But privatization may have significant impact on the private producers that often exist side-by-side SOEs. In this paper we show that this was indeed the case when Brazil privatized its SOEs in the iron ore industry. That is, after their privatization, the iron ore SOEs dramatically increased their labor productivity, but so did the private iron ore companies in the industry. ✩
The views expressed herein are those of the authors and not necessarily those of the Federal Reserve Bank of Minneapolis or the Federal Reserve System. * Corresponding author. E-mail address: [email protected] (J.A. Schmitz). 1094-2025/$ – see front matter © 2008 Elsevier Inc. All rights reserved. doi:10.1016/j.red.2008.01.001
746
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
Why would the existence of state-owned enterprises lower private productivity (and, hence, privatization of SOEs raise private productivity)? In the next section, we discuss some general ideas, or theory, as to why SOEs can lower private productivity. We discuss two channels. First, with an SOE in an industry, it is easier for a government to deliver benefits to constituents in the industry, and this (as we argue below) can ultimately lead to lower private productivity. Second, with an SOE in an industry, there will be less competition, and this too will lower private productivity. Consider the first channel. Suppose workers in an industry-X are part of a government’s constituents. With an SOE in the industry, the government can more easily provide generous wage and work-rule benefits to workers in the SOE. But this will lead to SOE losses and the need to subsidize SOEs. The government has an incentive to hide the losses and subsidize “off-budget.” In practice, the methods used by governments to hide SOE losses often lead to lower private-firm productivity. The government can reduce SOE losses, and subsidies, if it can adversely impact the profitability of private firms in the industry. Methods used by governments to do so include (1) adversely impacting the technology of private firms and (2) raising the prices of private-firm inputs. Both approaches (we argue below) will reduce private productivity. Below, we present evidence that governments employ both approaches. For example, regarding approach (1), in the United States, the US Postal Service puts restrictions on private delivery services like FedEx (as when it forbids FedEx to use private mailboxes). Regarding approach (2), in some industries (like mining), the government argues that certain work rules are needed for safety. It then points to generous work rules in SOEs and argues that these are present to provide safety and that the private firms should use the same work rules. The second channel is that with an SOE in an industry, there will be less competition, and less pressure to reduce costs. We wait till the next section to discuss this channel. After discussing general ideas in Section 2, we turn to how the “theory” applies to Brazil in Section 3. In the Brazilian iron ore industry, the vast majority of public production was accounted for by CVRD (Companhia Vale do Rio Doce). It gave very generous wage and work rule packages to its workers. These work rules significantly lowered CVRD’s productivity. As the “theory” sketched above suggests, government had an incentive to see that these generous work rules were adopted at the private mines as well. One reason, mentioned above, is that it would reduce any SOE losses. But there is another reason: the workers at the private mines were constituents of the government as well. As we show, governments in Brazil, working with the union representing workers at both public and private mines, were able to use the generous contracts at CVRD as a way to pressure private firms to offer generous work rules to their workers.1 The second force we discussed was that SOEs limited competition. This was clearly the case in this industry. CVRD was a massive iron ore producer. Given that CVRD’s mines were under public control, and not private, meant that competition in the world iron ore industry was much less intense than it otherwise would have been. Hence, with the advent of privatization in Brazilian iron ore, there would be two forces that spurred private firms to increase productivity. First, CVRD began to change its labor contracts, and to roll back its generous work rules. Hence, local governments lost leverage in attempting to force private firms to provide generous work rules. Second, the transfer of control of CVRD’s mines to private investors clearly led to increases in the extent of competitive pressure in the world iron ore industry. This gave added incentive for firms to change their labor contracts. After the “theory” in Sections 2 and 3, we turn to the evidence in Sections 4 and 5. Section 4 gives institutional details. Prior to privatization, SOEs accounted for roughly 60% of Brazilian iron ore production, private firms about 40%. As mentioned, the vast majority of the SOE production was accounted for by CVRD. In Section 5, we present the productivity evidence. We show that there was little labor productivity gain in the Brazilian iron ore industry until the onset of the privatization process in 1990. Then, industry productivity dramatically increased, nearly doubling over the next decade. We then show that this pattern was found in both the public and private firms in the industry. That is, following the onset of the privatization process in 1990, both sets of firms roughly doubled their labor productivity over the 1990s. That changes in labor contracts of the type at the private mines could have significant impact on productivity has been shown in Schmitz (2005). Finally, a brief word on the literature. There is a very large literature examining the impact of SOE privatization on SOE productivity. This literature is reviewed, for example, in Megginson and Netter (2001). The literature typically 1 Governments, of course, often try to persuade private firms to provide greater wage and work rule benefits to workers. The key point is that the existence of an SOE in an industry may make the task simpler.
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
747
finds large productivity gains at SOEs following privatization, a leading example being La Porta and Lopez-de-Silanes (1999). There is also a literature that compares productivity in public and private firms, asking which are more efficient (see, e.g., the reviews by Borcherding et al., 1982 and Megginson and Netter, 2001). But, again, as far as we know, there is no literature that looks at the impact of SOE privatization on private productivity. This is an important issue, and we begin examining it here. 2. Existence of SOEs can lower private productivity We discuss two channels through which the existence of state-owned enterprises can lower private productivity. First, with an SOE in an industry, it is easier for a government to deliver benefits to constituents in the industry, and this (as we argue below) can ultimately lead to lower private productivity. Second, with an SOE in an industry, there will be less competition, and this too will lower private productivity. We will present some general ideas in this section, and discuss Brazil’s particular experience in two sections hence. 2.1. Having SOEs makes delivering benefits to constituents easier Governments want to deliver benefits to constituents. Suppose workers in an industry-X are part of the government’s constituents. By creating an SOE in the industry, the government finds it easier to deliver benefits to workers in the industry. For example, since the government appoints the managers of the SOEs, it can pressure the managers to provide generous wage and work rule packages to workers in the SOE. But this process of delivering benefits can ultimately lower private productivity. Continuing with the above example, since these wages and work rules benefits in the SOE are “generous,” they exceed those offered elsewhere in the industry, and these benefits will yield SOE losses. The government can provide direct subsidies to the SOE, but these will be observable. This can be politically costly, since the subsidies can be interpreted (in this case, correctly) as signals of nepotism, corruption, etc. Hence, the government has a strong incentive to eliminate these subsidies.2 There are much less transparent approaches to “subsidizing” SOE losses. These approaches will be “off-budget” and reduce private productivity. The government can reduce SOE losses if it can adversely impact the profitability of private firms in the industry. It can do this by either (1) adversely impacting the technology of private firms and/or (2) raising the prices of private-firm inputs. Both approaches will reduce private productivity. We discuss each in turn. 2.1.1. SOEs adversely impact private technology We give two examples where governments protect their SOEs by putting production restrictions on private firms. Australia had a policy of permitting only two domestic airlines, one private and one public. As Davies (1971) reports, the government endeavored to ensure that the private airline did not harm the SOE. For example, for each route in Australia, the government would determine biannually the expected capacity in ton-miles for the route over the next six months. The two airlines would then receive the right to carry exactly one-half that ton-mile capacity over the period. Neither airline could carry more than that capacity over the period. But the government soon realized that this was not enough to protect the SOE. By choosing a better mix of aircraft, the private airline was able to make profits (and cause the public to make losses). So, the government started requiring that the airlines have the same fleet of aircraft! In summarizing the consequences of such government intervention, Forsyth and Hocking (1980, p. 182) argue In such a situation, there is little scope for the private airline to be more efficient than the public one, and it is doubtful whether individuals associated with the private one are permitted to have any more incentives to improve efficiency. Any incentives in this system must finally be reduced to very low levels by the strong de facto controls on the profits of the private airline. 2 There is some theory to motivate why governments would make transfers to constituents in ways (like providing jobs with generous work rules) that did not involve simply monetary payments, that is, in seemingly inefficient ways (see, e.g., Coate and Morris, 1995).
748
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
As another example, consider the mail and package delivery industry. In the United States, the US Postal Service (USPS) is able to put restrictions on private delivery services. These restrictions include the regulation that private companies are prohibited from placing items in mailboxes. Amazingly, the Postal Service itself has the power to issue regulations regarding mail service (see US Congress, 2004). In his statement before the US Senate regarding the issue of postal reform, Fred Smith, the CEO of FedEx, stressed how the restriction on private firms from placing items in mailboxes led these companies to be less efficient. He urged dropping this restriction, arguing that “the end result of this reform will be to make the nation’s delivery services sector more efficient by extending the efficiencies of the national mailbox system to private delivery services.” These are two clear cases where the government gives generous SOE benefits, which lead to losses, which are then minimized by putting restrictions on private technology. 2.1.2. SOEs raise private-firm input prices Another approach to minimizing SOE losses is to raise the costs of private-firm inputs, like wages. Note well that if the government is able to raise the wages in private firms (i.e., the wages of its constituents), then it has killed two birds with one stone: it has lowered SOE losses and provided benefits to some of its constituents. If, for example, the government is able to raise wages at United Parcel Service (UPS), a private delivery service, it pleases UPS workers and it reduces USPS losses. How does the government raise the wages and improve the work rules of workers in the private firms? Governments, of course, often try to persuade private firms to provide greater wage and work rule benefits to workers. The key point is that the existence of an SOE in an industry may make the task simpler. For example, in some industries (like mining) the government can argue that certain work rules are needed for safety. It can then point to work rules in the SOE and argue that these are present to provide safety and that the private firms should use the same work rules. The government can use other laws (besides safety laws) as a way to coerce private firms to match SOE work rules and benefits. But the government does not have to be directly involved in using its safety laws, etc., to coerce private firms to provide similar work rules and wages as provided in SOEs. By simply providing generous benefit packages to SOE workers, it can enhance the bargaining position of private workers. This was the case, for example, in the Canadian railroad industry. Canada’s railroad industry historically comprised two major firms, a public firm, Canadian National (CNR), and a private firm, Canadian Pacific (CPR). Timur and Ponak (2002) discuss how in the late 1980s the public railroad (CNR) gave significant job protection to employees. In particular, shop craft employees with more than eight years of experience were guaranteed full pay if their jobs were eliminated through new technology and if no suitable substitute position was found. Timur and Ponak quote a senior private firm (CPR) executive as saying, “CNR [the public firm] first negotiated it [the job guarantees]. Then it became a tactical issue whether we would take a strike, lose business permanently perhaps, and the decision was taken not to strike.” In bargaining with two firms in an industry, unions will likely try to first reach a deal with a firm that is in the weakest bargaining position. It can then try to extend the deal to other firms. This happens in an industry with all private firms. The key point, again, is that the existence of an SOE in an industry may make the task of getting a generous industry-wide contract simpler. Finally, we need to tie this discussion to productivity. Generous work rules will lead to overstaffing and low labor productivity. They can also lead to low materials and capital productivity. Schmitz (2005) has shown this to be true. In particular, he showed it was true in the US and Canadian iron ore industries. 2.2. Having SOEs reduces industry competition We now argue that having an SOE in an industry (accounting for a significant share of output) reduces competition relative to the case with all private production. And then we will argue that this leads to lower productivity. There is no general theory that tells us more competition will lead to higher productivity. For example, there is no general result that tells us an industry-X with one private firm (i.e., a monopoly) will be less productive than the same industry with two firms. But we think there are convincing arguments for why having an SOE sector in an industry will mean lower productivity of all firms.
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
749
Imagine an industry that comprises two large private firms (say, with both having 50% of the market). Contrast this with an industry with a large SOE and a large private firm (again, say both have 50% of the market). There are two senses in which the latter industry will be less “competitive” and will have lower productivity than the former. First, SOE management likely has less incentive to reduce costs and capture market share than does a private manager. Governments typically can only offer limited compensation to managers, for political reasons. Second, a private manager likely has less incentive to reduce costs if the “other” half of the industry is an SOE rather than a private firm. That is because the government may be committed to keeping the SOE open, and it may be hard to capture market share. This then ties back to the previous discussion, that a private firm might be wary of reducing costs because the government may put technology restrictions on the firm, as in the Australian airline industry. In sum, there will be less pressure to reduce costs in an industry with an SOE and private firms (as opposed to all private firms). Productivity will be lower. 3. How does the “theory” apply to Brazil? In this section, we discuss how the “theory” discussed above applies to Brazil’s privatization experience. In particular, we discuss the mechanisms through which privatization increased private productivity. Before we turn to the productivity of private firms, we need to briefly address the typical question in this literature: Why would privatization of SOEs lead to SOE productivity increases? or Why was SOE productivity so low? As we argued, governments use SOEs to deliver generous wages and work rules to constituents. This leads to low productivity. This was the case here, with CVRD. Generous benefits were given to both management and workers. First, there was significant overstaffing of management. For many management tasks, there was a separate management team in the south and the north (as discussed below, CVRD had two major operating areas that were thousands of miles apart). But for many of these tasks, a single management team was needed. During the privatization process, in the early 1990s, much of this management redundancy was eliminated. This information was provided in interviews with CVRD management. Second, there were generous wage and work rule packages given to workers. As was common in the United States and Canada, work practices in Brazilian iron ore mines were such that employees engaged in only a small range of tasks. For example, there was typically a sharp separation of repair and production work. Those workers who ran machinery, for example, were not permitted to tend to their machines—this was reserved for repair staff. As another example, there were many repair job classifications. Repair staff with a certain classification were not permitted to conduct repairs assigned to other job classifications. In the US and Canada, these work practices changed in the early to middle 1980s, as these iron ore industries were faced with the possibility of closure (see Galdon-Sanchez and Schmitz, 2002 and Schmitz, 2005). But they persisted in Brazil until the onset of the privatization process. In the early 1990s, the management at CVRD began changing contracts so as to require workers to be responsible for more tasks. This process was called “polivalente,” the Portuguese term for jack-of-all-trades. We now turn to the main question: Why did SOE privatization lead to private-firm productivity gains? or Why was private productivity low before privatization? We discussed two general forces whereby SOE participation in an industry led to lower private productivity. First, because SOEs give generous benefits to their workers, this ultimately could impact private technology and/or private costs. In the Brazilian iron ore industry, the existence of CVRD enabled local governments to pressure private firms to provide their workers with work rules similar to those at CVRD. In particular, all iron ore mines in Brazil were required to receive annual certifications to operate. This certification was given at the local government level. There were zoning regulations, and other regulations, that each mine had to satisfy in annual reviews. Local government officials used these laws to help union leaders get generous benefits from private firms. In Brazil, there was a single union, Sindicato Metabase (Sindicato dos Trabalhadores da Industria de Metais de Base), that represented workers at both the public and private iron ore mines. This information regarding use of local regulatory procedures to pressure private mines to grant generous benefits was provided in interviews with union management. The workers at the private firms, then, benefited from the narrow job classifications at CVRD. The second force we discussed was that SOEs limited competition. This was clearly the case in this industry. CVRD was a massive iron ore producer. The transfer of its control to private investors clearly led to increases in the extent of competitive pressure in the industry.
750
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
Hence, with the advent of privatization in Brazilian iron ore, there would be two forces that spurred private firms to increase productivity. First, CVRD began to change its labor contracts, and to roll back its generous work rules. Hence, local governments lost leverage in attempting to force private firms to provide generous work rules. Second, the transfer of control of CVRD’s mines to private investors clearly led to increases in the extent of competitive pressure in the world iron ore industry. This gave added incentive for firms to change their labor contracts. We now turn to examine the data. In the next section, we present the institutional background. In the section following this, we present the productivity record of the industry. 4. Brazilian iron ore privatization: background In this section we briefly discuss the history of iron ore privatization in Brazil and also the sources of our data. 4.1. Iron ore privatization The Brazilian iron ore industry had two public companies that owned mines: CSN (Companhia Siderugica Nacional) and CVRD (Companhia Vale do Rio Doce). CSN was primarily a steel producer; in fact, it was Brazil’s largest (see Andrade et al., 2002). It owned a single iron ore mine. CVRD was primarily an iron ore producer, a massive one. In 1990, CSN produced about 6 million metric tons (mmt) of iron ore, CVRD 85 mmt, and Brazil 152 mmt. Brazil was the largest producer of iron ore (152 mmt), followed by Australia (110 mmt) and the United States (56 mmt).3 Hence, CVRD produced more iron ore in 1990 than any other country except Australia. Steel industry privatization in Brazil began earlier than iron ore privatization. The official plan to privatize CSN (and other steelmakers) was announced in 1988. The process of privatizing CSN’s steel operation, begun in 1988, was completed in 1993, when the CSN mine was also sold. The CSN mine was the first sold to private investors. CVRD did not appear on official privatization lists till much later than CSN, in fact, not until 1995. The government had decided to sell CVRD well before this date but knew it was going to face stiff resistance to selling it (see Kandell, 1997 for some of the difficulties faced in selling CVRD). Hence, it delayed a formal announcement. CVRD was the crown jewel of Brazilian state enterprises. The government had to placate many nationalist interests before the sale. One restriction on the sale was that Australian iron ore producers (who, again, were the largest producers after Brazil) could not be part of a bidding consortium. Also kept out of the bidding were Japanese steel producers. In the end, only two consortiums, both led by local entrepreneurs, bid on CVRD. It was not finally sold until 1997. While the CVRD privatization process was officially started in 1995 and completed in 1997, both CVRD management and union officials were unanimous in agreeing that it was evident much earlier that CVRD would be privatized (many give 1989 as the date it was clear). By the early 1990s, CVRD was undergoing major reforms and changes in its organization to prepare for privatization. As the preceding discussion suggests, there is no one clear date to label as the date of privatization. The process of privatization started before the actual sales to private investors, and even before the actual announcement that sales would be happening at some future date. Looking at the impact of privatization in this case, and likely all cases, is not like examining the impact of, for example, an unexpected tariff change at some date. To pick a round number, we pick the date that privatization began as 1990, though it could be argued that the process began even earlier, in 1989, or even 1988 as the decision to sell CSN was announced. 4.2. Data sources The basic source of data for mining industries in Brazil, including iron ore, are annual mine reports submitted to regional DNPMs (Departamento Nacional de Producao Mineral—National Mineral Production Department). These reports include the mine’s production (tons of iron ore by type of iron ore) and employment (as of the middle of December). These reports are then submitted to the Ministry of Mine and Energy for tabulation of aggregate production and aggregate employment levels for Brazil. These are published in Relatorio Anual de Lavras. From this publication, we obtain aggregate industry data from 1972 to 1999. 3 These statistics for Australia and the United States are from the United States Geological Society Minerals Yearbook. We are excluding China and the USSR from these rankings, since these countries reported production differently from the rest of the world.
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
751
In order to examine production and employment by mine, both private and public, we used the archives of yearly mine reports at DNPM (in Minas Gerais). We limited our search of these archives in the following way. First, there are a large number of very small iron ore firms in Brazil. Hence, to limit the search we gathered mine records for the public companies, CSN and CVRD, and for the three largest private companies, Ferteco (Ferteco Mineracao SA), MBR (Mineracoes Brasileiras Reunidas SA), and Samitri. Second, it was difficult to collect records prior to 1986. Not many mine reports were found prior to 1986. So, we gathered mine reports from 1986 to 1998.4 We also decided to examine the data at the firm level and not the mine. While the companies are supposed to report employment by mine, in practice the mines of a firm are often close to each other so that some staff (like repair staff) may work in more than one mine. Hence, it is up to the firm how to assign workers to mines in the reports. Moreover, there are some mines that report production but no employment. Because of this, we have chosen to look at production and productivity records at the firm level. We have made one exception to this strategy of looking at performance at the firm level. CVRD has operations in the south and the north of Brazil. Since these operations are separated by more than two thousand miles, the issues of repair staff working in different mines, etc., is not a major issue. Hence, we break CVRD into two “firms,” CVRDSouth and CVRD-North. What we have constructed, then, is the output and employment records of three public firms, CSN, CVRD-South, and CVRD-North, and the three largest private firms, Ferteco, MBR, and Samitri, from 1986 to 1998. In Table 1, we present the production and productivity of these six firms, together with the rest of the industry, for 1990. Production by public firms (90.8 mmt) amounted to about 60% of total production (152.2 mmt) in 1990. Production by the three largest private firms (41.5 mmt) amounted to about 27% of total production. The small private firms contributed about 13% of production. The production of the three largest private firms (again, 41.5 mmt) would have ranked them as the fourth largest national producer, behind Australia (110 mmt), the United States (56 mmt), and India (54 mmt), but ahead of the next largest producers, Canada (35 mmt), South Africa (30 mmt), and Venezuela (20 mmt). Productivity (defined as thousands of metric tons per employee per year) varies widely across firms. CVRD-North, which is the huge Carajas mine, has very high productivity, over twice the industry average. CVRD-North’s high productivity was due to its very high quality deposits. MBR and Ferteco had higher productivity than the other two public firms, CVRD-South and CSN. The third large private firm, Samitri, had low productivity. However, Samitri produced a type of iron ore, pellets, that requires more labor per ton than other ores. Hence, its low productivity is not surprising. Also not surprising is that the “rest of the industry” had very low productivity, less than half the industry average. The rest of the industry mostly comprised very small firms that worked marginal deposits.
Table 1 Brazilian iron ore production and labor productivity, 1990 (by public firms, large private firms, and rest of industry) Production
Firm production as percent of industry
Productivity
Firm productivity relative to industry
Public Firms CSN CVRD-South CVRD-North
5.83 52.12 32.85
3.83 34.24 21.58
7.40 9.66 24.17
0.81 1.06 2.66
Large Private Firms Ferteco MBR Samitri
10.93 19.97 10.61
7.18 13.11 6.97
10.58 14.92 5.63
1.17 1.64 0.62
Rest of Industry
19.93
13.09
4.02
0.44
152.24
100.00
9.08
1.00
Total Industry
Note: Production in millions of metric tons (mmt). Productivity in thousands of tons per employee per year. Source: See text.
4 We collected the data in the mine reports in 2000, and at that time only reports through 1998 were available.
752
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
5. Privatization’s impact on productivity: evidence In this section we show that there was little productivity gain in the Brazilian iron ore industry until the onset of privatization. Then, industry productivity dramatically increased. We also show that this pattern was found in both the public and private firms in the industry. 5.1. Productivity and production: aggregate industry The production and productivity records of the Brazilian iron ore industry over the period 1971–1999 are given in Fig. 1. In the figure we have drawn a vertical line at 1990 to indicate the year when the privatization process began. If we examine the decade of the 1980s, there was a modest increase in productivity. But this was mostly driven by the opening of CVRD-North, the Carajas mine, in 1986. As we saw, CVRD-North had significantly higher productivity than all the other Brazilian producers, private and public. Its opening pulled up the industry average.5 To see this, in Fig. 2 we plot the labor productivity of the Brazilian iron ore industry (the dashed line), and the industry not including CVRD-North (the solid line). CVRD-North only opened in 1986, so the two lines coincide before 1986. Note that the labor productivity of the industry not including CVRD-North did not increase much at all over the 1980s. It fluctuates between 8 and 10 thousand tons per employee per year, and is at 9 by 1990. In 1990, productivity was at early 1970 levels. The total industry, and the total industry not including CVRD-North, show significant productivity growth once the privatization process begins. There were significant gains prior to the culmination of the privatization process in 1997, that is, from 1990 to 1997, and then a gain after CVRD was sold.
Fig. 1. Production and labor productivity (thousands of tons per employee), Brazilian iron ore industry, 1971–1999.
5 The existence of the Carajas (CVRD-North) deposits had been known well before the middle 1980s. However, they were located in the interior of the Amazon, and massive infrastructure investment was needed before the iron ore could be extracted and shipped to the coast. A World Bank project supplied these funds in the middle 1980s.
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
753
Fig. 2. Labor productivity, Brazilian iron ore industry (thousands of tons per employee) with and without CVRD-North, 1972–1998.
In Fig. 3, we plot the productivity of the public firms (aggregated as a group) and the private firms (aggregated as a group). That is, the public sector is the aggregate of CSN, CVRD-South, and CVRD-North. The private sector is an aggregate of Ferteco, MBR, and Samitri. There is little productivity growth in either of the groups of firms until the privatization process starts. Then productivity growth is significant in both groups. We now turn to the productivity records of the individual companies. Again, we only have data beginning a few years prior to privatization. 5.2. Productivity: public firms The productivity of the public firms over the period 1986–1998 is given in Figs. 4, 5, and 6. In Fig. 4, we plot the productivity of CSN. We plot the (natural) logarithm of productivity (we draw the vertical line at 1988 in this figure, since that was the date at which CSN’s privatization process began). CSN had few productivity gains between 1987 and 1989 (for CSN, 1987 is our first year of data). Over 1989–1998, its productivity dramatically increased, by more than a factor of four. Recall that CSN’s productivity was lower than CVRD-South’s in 1990 (Table 1). Its spectacular productivity growth likely reflects that it was run much poorer than CVRD prior to privatization. With its dramatic productivity growth, it was essentially catching up to CVRD-South’s level. In Fig. 5, we show the productivity of CVRD-North (again, in logarithms). CVRD-North shows good productivity gains between 1986 and 1988 (about 33% growth). This was due to the fact that the Carajas mine had just opened in 1986. CVRD-North was ramping up production during 1986–1988 (its output more than doubled over 1986–1988). When iron ore mines open, their productivity grows rapidly for the first few years, until mine output reaches its rated capacity (see, e.g., Schmitz, 2005, where this pattern is seen in new mines opening in Minnesota). After 1988, when the mine had reached its capacity, CVRD-North’s productivity did not change for four years. Following the onset of privatization, CVRD-North’s productivity roughly doubled over 1992–1998. In Fig. 6, we plot CVRD-South’s productivity. Productivity increased little from 1986 to 1990. It increased by about 0.1 log points over this four-year period. Over the next four years, 1990–1994, its productivity increased by about 0.3 log points, and over 1994–1998, by nearly 0.6 log points. If we break the period in half, productivity grew
754
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
Fig. 3. Labor productivity (thousands of tons per employee) private (large firms) vs. public, Brazilian iron ore industry, 1986–1998.
Fig. 4. Labor productivity (thousands of tons per employee, in ln’s), CSN (public), 1987–1998.
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
755
Fig. 5. Labor productivity (thousands of tons per employee, in ln’s), CVRD-North (public), 1986–1998.
Fig. 6. Labor productivity (thousands of tons per employee, in ln’s), CVRD-South (public), 1986–1998 (solid line—actual productivity; two dashed lines—linear trends, 1986–1990, 1990–1998).
756
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
by 0.2 log points over 1986–1992, and 0.7 over 1992–1996. In the figure, we plot two linear regressions (the dashed lines) that are fitted to the productivity data, one for the period 1986–1990, the other for 1990–1998. The regression for the second period is much steeper than the first. Over 1990–1998, CVRD-South’s productivity more than doubled. Both CVRD-South and CVRD-North showed significant productivity growth prior to their formal sale in 1997. This pattern of major productivity gains in public firms before their formal privatization date is found in many privatization studies. For example, La Porta and Lopez-de-Silanes (1999) find this.6 The major changes at CVRD that led to productivity growth, the reduction of redundancy in management, and the changes in work rules, both discussed above, were begun well before CVRD’s formal sale. They were instituted so as to make CVRD a more attractive firm to purchase, thus enabling the government to raise more at its sale. 5.3. Productivity: large private firms The productivity of the three largest private firms over the period 1986–1998 is given in Figs. 7, 8, and 9. Here, again, we have only a few years of data prior to the onset of privatization. In Fig. 7, we plot Samitri’s productivity. There was little growth in Samitri’s productivity before the privatization process started (in fact, it fell over the short period we have before privatization). Samitri’s productivity shows much larger year-to-year swings than the other five firms (that is, the three public and two private firms). This may have been due to problems at some of its mines (such as major breakdowns of processing facilities), as its output also shows much greater variation than the other firms. In Fig. 8, we plot Ferteco’s productivity. There was little growth in Ferteco’s productivity before the privatization process started (in fact, it fell over the short period we have before privatization). Once privatization began, productivity grew strongly. Over 1990–1998, its productivity nearly doubled. In Fig. 9, we plot MBR’s productivity. MBR had little productivity growth between 1986 and 1988. Its productivity growth started earlier than the other firms, and it experienced significant growth over 1988–1991. MBR’s productivity, then, seems to have begun growing before privatization began, over 1988–1989. However, recall that the steel industry, and CSN’s, privatization began in 1988, so it is possible that MBR’s gains were tied to these developments. MBR experienced another surge in productivity beginning in 1995. Over the course of 1990–1998, its productivity more than doubled. In summary, the three private firms all show strong productivity growth during the privatization period. Productivity growth tends to start earlier in the private firms than it does in the public firms. This seems reasonable. Once privatization was known to be coming, local governments lost their leverage with the private firms. Private firms then could start changing their labor contracts, and could be expected to change them much more quickly than could public companies. 5.4. Summary, caveats, and quantitative discussion The results, then, are generally consistent with the “theory.” There was little productivity growth in the (aggregate) Brazilian iron ore industry until the onset of privatization. Then there was strong productivity growth. This pattern found at the aggregate level is found at the level of the public and private firms as well. It is unfortunate that we have only a few years of firm-level data before privatization. But the firm-level data before 1986 do not exist. The concern, of course, is that perhaps private productivity was growing strongly before 1986 when our data starts and that the flat private-firm productivity from 1986 to 1990 (in the case of Ferteco and Samitri) and from 1986 to 1988 for MBR was a fluke. There is one observation we can make regarding productivity growth before 1986. It perhaps adds a tiny amount of support to the view that private productivity growth was not strong before 1986. Recall there was little productivity growth at the aggregate level before privatization. Hence, if there was strong 6 They examine the employment levels in their sample of privatized firms in the four years leading up to the formal sale. On average, employment fell significantly in each of these years. In their Fig. I, they show average employment falls from a bit over 1450 in the fourth year before privatization to a bit over 1050 one year before privatization (a three-year period). In the last year before privatization, employment falls from a bit over 1050 to a bit over 550. La Porta and Lopez-de-Silanes surmised that these major reductions in employment before privatization were part of the government’s effort to prepare the SOEs for sale. That is certainly what happened in this industry: the large productivity gains at CVRD and CSN prior to privatization were due to the government’s effort to prepare these companies for sale (see below).
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
Fig. 7. Labor productivity (thousands of tons per employee, in ln’s), Samitri (private), 1986–1998.
Fig. 8. Labor productivity (thousands of tons per employee, in ln’s), Ferteco (private), 1986–1998.
757
758
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
Fig. 9. Labor productivity (thousands of tons per employee, in ln’s), MBR (private), 1986–1998.
growth at the private firms Ferteco, MBR, and Samitri before 1986, then it would likely mean that there was significant negative growth at CVRD-South (since before 1986 these four firms accounted for a very large share of the industry output, that is, CVRD-North had not yet opened). It is certainly possible that CVRD-South had a negative productivity trend till 1986, but perhaps not likely. Note that if its productivity growth was negative, it would imply that CVRD-South had significantly higher productivity than Ferteco and MBR in the years before 1986. That is because CVRD-South had roughly the same productivity as Ferteco and MBR in 1986. (This can be seen by looking at Table 1 where Ferteco and CVRD-South have roughly the same productivity in 1990. MBR’s productivity was about 50% higher than CVRD-South’s, but that was due to MBR’s very strong productivity growth over 1988–1990.) So, while it is possible that CVRD-South had higher productivity than the other firms in the decade before 1986, perhaps it is not likely. Finally, let us briefly discuss the quantitative gains in productivity. Many of the firms had productivity gains over the privatization period that were very large—roughly doubling of productivity. Much of these gains, we argue, were due to changes in the labor contracts used at the public and private mines. While we have not done an in-depth analysis of the labor contracts, from interviews we know the changes were very similar to those that occurred in the 1980s in the United States and Canada.7 Schmitz (2005) has done an extensive analysis of the impact of labor contract changes in these industries and has shown that much of the productivity gains (again, roughly doubling) in these industries resulted from the contract changes. He also discussed changes in Australia that were similar and led to large gains as well. The argument is by analogy, but we think it is a good one. 6. Other possible explanations? In this section we briefly consider the possibility that there were changes in the economic environment in the 1990s (besides privatization) that may have influenced the productivity of the entire Brazilian iron ore industry, that is, both 7 It proved very difficult to find copies of labor contracts that were more than a few years old.
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
759
the public and private iron ore producers. If this were the case, then we would be attributing too much of the gains in private productivity to the privatization of the public firms. First, we consider other economic reforms in Brazil (besides privatization). In fact, there was a lowering of tariffs in Brazil during the 1990s (see, e.g., Ferreira and Rossi, 2003). Trade liberalization could have led to industry-wide productivity improvements in Brazilian iron ore if imports of iron ore now challenged local iron ore in Brazil. But Brazil did not have any tariffs on iron ore. Brazil and Australia were the dominant players and exporters in the industry. Imports of iron ore were never in the cards. But productivity would also increase if the industry could now import intermediate or capital goods that were prohibitive to import before. Perhaps the industry could now import machinery that was too expensive to import before. Ferreira and Rossi (2003) provide effective rates of protection in Brazil over the period 1987–1997 by manufacturing industry (see their Table 2). Effective rates of protection in machinery and in transportation equipment were among the lowest in Brazil in 1987, and they did not fall very much over the period 1987–1997 (though effective rates of protection in some manufacturing industries fell dramatically). Hence, we believe that the impact of trade liberalization on the iron ore industry was small. Second, we consider general developments in the world iron ore industry. During the 1990s, as a result of changes in labor laws in Australia, its iron ore industry began to change its labor contracts, and this led to large gains in productivity (see, e.g., Galdon-Sanchez and Schmitz, 2002). Australia was now a much more efficient iron ore producer. Given that Brazil and Australia were by far the two largest exporters of iron ore, these developments may have significantly influenced the Brazilian industry. But we believe the impact on productivity from this effect was small as well. Australia was more productive, but adding capacity is very costly in this industry, so that without a major expansion in capacity (or a large fall in world demand for iron ore), Australia was not likely to capture much of Brazil’s market. 7. Conclusion Our theme has been that the productivity gains from privatization are much more general and widespread than has typically been recognized in the literature. This is because privatization of SOEs in an industry will typically lead to increases in private firm productivity in that industry—an influence that has been ignored in the literature. We began by presenting reasons why SOE privatization would raise private firm productivity. We then turned to the privatization of SOEs in Brazil’s iron ore industry. We showed that there was little productivity gain in the industry until the onset of privatization. Then there was significant productivity growth at both the public and private firms in the industry. We argued that much of the private productivity gain was due to forces set in motion with the privatization of the public companies. Acknowledgments We thank Sanghoon Lee and Shi Qi for excellent research assistance. We also thank the Associate Editor and two anonymous referees for very useful comments that have improved the paper a lot. For their help we thank Armando Castelar Pinheiro; staff of the labor union, Paulo Soares, Paulo Chaves, Jose Alberto Miguel, Ausgustinho Eliseo Vieira de Souza, Julio Freitas, Jose Maria, Anivaldo Coelho, Sebastiao Alves Oliveira, Joao Trevisan e Joao Cesar de Freitas Pinheiro; directors of CVRD, Flavio Ferreira, Jose Luis Martins, Marconi Tarbes Vianna; and the ex-President of CSN, Roberto Procopio Lima Neto. We are in great debt to all employees of DNPM-MG (Departamento Nacional de Producao Mineral—National Mineral Production Department) and the Ministry of Mine and Energy-MG. Teixeira thanks the Federal Reserve Bank of Minneapolis for its hospitality. References Andrade, M., Cunha, L., Gandra, G., 2002. Restructuring the Brazilian metallurgical sector. Sectoral Studies Series, BNDES. Borcherding, T.E., Pommerehne, W.W., Schneider, F., 1982. Comparing the efficiency of private and public production: The evidence from five countries. Zeitschrift für Nationalökonomie (Journal of Economics) (Supplement 2), 127–156. Coate, S., Morris, S., 1995. On the form of transfers in special interests. Journal of Political Economy 103 (6), 1210–1235. Davies, D.G., 1971. The efficiency of public versus private firms, the case of Australia’s two airlines. Journal of Law and Economics 14 (1), 149–165. Ferreira, P.C., Rossi, J.L., 2003. New evidence from Brazil on trade liberalization and productivity growth. International Economic Review 44 (4), 1383–1405.
760
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
Forsyth, P.J., Hocking, R.D., 1980. Property rights and efficiency in a regulated environment: The case of Australian airlines. Economic Record 56 (153), 182–185. Galdon-Sanchez, J.E., Schmitz Jr., J.A., 2002. Competitive pressure and labor productivity: World iron ore markets in the 1980s. American Economic Review 92 (4), 1222–1235. Kandell, J., 1997. The CVRD Factor. Institutional Investor (October), 186–192. La Porta, R., Lopez-de-Silanes, F., 1999. The benefits of privatization: Evidence from Mexico. Quarterly Journal of Economics 114 (4), 1193–1242. Megginson, W.L., Netter, J.M., 2001. From state to market: A survey of empirical studies on privatization. Journal of Economic Literature 39 (2), 321–389. Schmitz Jr., J.A., 2005. What determines productivity? Lessons from the dramatic recovery of the US and Canadian iron ore industries following their early 1980s crisis. Journal of Political Economy 113 (3), 585–625. Timur, A.T., Ponak, A., 2002. Labor relations and technological change at Canadian Pacific Railway. Journal of Labor Research 23 (4), 535–557. US Congress. Senate. Committee on Governmental Affairs. 2004. Statement of Fred Smith, “Postal Reform: Sustaining the 9 Million Jobs in the $900 Billion Mailing Industry (Day 2),” March 11, 2004. http://murray.senate.gov/comm/governmental_affairs/general/ index.cfm?Fuseaction=Hearings.Testimony&HearingID=157&WitnessID=545.
Privatization’s impact on private productivity: The case of Brazilian iron ore ✩ James A. Schmitz Jr. a,∗ , Arilton Teixeira b a Federal Reserve Bank of Minneapolis, 90 Hennepin Avenue, Minneapolis, MN, USA b Fundacao Capixaba de Pesquisa, Brazil
Received 29 March 2005; revised 13 December 2007 Available online 17 January 2008
Abstract A major motivation for the recent wave of privatizations of state-owned enterprises (SOEs) was a belief that privatization would increase SOE productivity. There are now many studies showing most privatizations achieved this goal. Our theme is that the productivity gains from privatization are much more general and widespread than has typically been recognized in this literature. In assessing the productivity gains from privatization, the literature has only examined the productivity gains accruing at the privatized SOEs. But privatization may have significant impact on the private producers that often exist side-by-side SOEs. In this paper we show that this was indeed the case when Brazil privatized its SOEs in the iron ore industry. That is, after their privatization, the iron ore SOEs dramatically increased their labor productivity, but so did the private iron ore companies in the industry. © 2008 Elsevier Inc. All rights reserved. JEL classification: L33; L70 Keywords: State-owned enterprises; Privatization; Productivity
1. Introduction A major motivation for the recent wave of privatization of state-owned enterprises (SOEs) was a belief that privatization would increase SOE productivity. There are now many studies showing most privatization achieved this goal. Our theme is that the productivity gains from privatization are much more general and widespread than has typically been recognized in this literature. In assessing the productivity gains from privatization, the literature has only examined the productivity gains accruing at the privatized SOEs. But privatization may have significant impact on the private producers that often exist side-by-side SOEs. In this paper we show that this was indeed the case when Brazil privatized its SOEs in the iron ore industry. That is, after their privatization, the iron ore SOEs dramatically increased their labor productivity, but so did the private iron ore companies in the industry. ✩
The views expressed herein are those of the authors and not necessarily those of the Federal Reserve Bank of Minneapolis or the Federal Reserve System. * Corresponding author. E-mail address: [email protected] (J.A. Schmitz). 1094-2025/$ – see front matter © 2008 Elsevier Inc. All rights reserved. doi:10.1016/j.red.2008.01.001
746
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
Why would the existence of state-owned enterprises lower private productivity (and, hence, privatization of SOEs raise private productivity)? In the next section, we discuss some general ideas, or theory, as to why SOEs can lower private productivity. We discuss two channels. First, with an SOE in an industry, it is easier for a government to deliver benefits to constituents in the industry, and this (as we argue below) can ultimately lead to lower private productivity. Second, with an SOE in an industry, there will be less competition, and this too will lower private productivity. Consider the first channel. Suppose workers in an industry-X are part of a government’s constituents. With an SOE in the industry, the government can more easily provide generous wage and work-rule benefits to workers in the SOE. But this will lead to SOE losses and the need to subsidize SOEs. The government has an incentive to hide the losses and subsidize “off-budget.” In practice, the methods used by governments to hide SOE losses often lead to lower private-firm productivity. The government can reduce SOE losses, and subsidies, if it can adversely impact the profitability of private firms in the industry. Methods used by governments to do so include (1) adversely impacting the technology of private firms and (2) raising the prices of private-firm inputs. Both approaches (we argue below) will reduce private productivity. Below, we present evidence that governments employ both approaches. For example, regarding approach (1), in the United States, the US Postal Service puts restrictions on private delivery services like FedEx (as when it forbids FedEx to use private mailboxes). Regarding approach (2), in some industries (like mining), the government argues that certain work rules are needed for safety. It then points to generous work rules in SOEs and argues that these are present to provide safety and that the private firms should use the same work rules. The second channel is that with an SOE in an industry, there will be less competition, and less pressure to reduce costs. We wait till the next section to discuss this channel. After discussing general ideas in Section 2, we turn to how the “theory” applies to Brazil in Section 3. In the Brazilian iron ore industry, the vast majority of public production was accounted for by CVRD (Companhia Vale do Rio Doce). It gave very generous wage and work rule packages to its workers. These work rules significantly lowered CVRD’s productivity. As the “theory” sketched above suggests, government had an incentive to see that these generous work rules were adopted at the private mines as well. One reason, mentioned above, is that it would reduce any SOE losses. But there is another reason: the workers at the private mines were constituents of the government as well. As we show, governments in Brazil, working with the union representing workers at both public and private mines, were able to use the generous contracts at CVRD as a way to pressure private firms to offer generous work rules to their workers.1 The second force we discussed was that SOEs limited competition. This was clearly the case in this industry. CVRD was a massive iron ore producer. Given that CVRD’s mines were under public control, and not private, meant that competition in the world iron ore industry was much less intense than it otherwise would have been. Hence, with the advent of privatization in Brazilian iron ore, there would be two forces that spurred private firms to increase productivity. First, CVRD began to change its labor contracts, and to roll back its generous work rules. Hence, local governments lost leverage in attempting to force private firms to provide generous work rules. Second, the transfer of control of CVRD’s mines to private investors clearly led to increases in the extent of competitive pressure in the world iron ore industry. This gave added incentive for firms to change their labor contracts. After the “theory” in Sections 2 and 3, we turn to the evidence in Sections 4 and 5. Section 4 gives institutional details. Prior to privatization, SOEs accounted for roughly 60% of Brazilian iron ore production, private firms about 40%. As mentioned, the vast majority of the SOE production was accounted for by CVRD. In Section 5, we present the productivity evidence. We show that there was little labor productivity gain in the Brazilian iron ore industry until the onset of the privatization process in 1990. Then, industry productivity dramatically increased, nearly doubling over the next decade. We then show that this pattern was found in both the public and private firms in the industry. That is, following the onset of the privatization process in 1990, both sets of firms roughly doubled their labor productivity over the 1990s. That changes in labor contracts of the type at the private mines could have significant impact on productivity has been shown in Schmitz (2005). Finally, a brief word on the literature. There is a very large literature examining the impact of SOE privatization on SOE productivity. This literature is reviewed, for example, in Megginson and Netter (2001). The literature typically 1 Governments, of course, often try to persuade private firms to provide greater wage and work rule benefits to workers. The key point is that the existence of an SOE in an industry may make the task simpler.
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
747
finds large productivity gains at SOEs following privatization, a leading example being La Porta and Lopez-de-Silanes (1999). There is also a literature that compares productivity in public and private firms, asking which are more efficient (see, e.g., the reviews by Borcherding et al., 1982 and Megginson and Netter, 2001). But, again, as far as we know, there is no literature that looks at the impact of SOE privatization on private productivity. This is an important issue, and we begin examining it here. 2. Existence of SOEs can lower private productivity We discuss two channels through which the existence of state-owned enterprises can lower private productivity. First, with an SOE in an industry, it is easier for a government to deliver benefits to constituents in the industry, and this (as we argue below) can ultimately lead to lower private productivity. Second, with an SOE in an industry, there will be less competition, and this too will lower private productivity. We will present some general ideas in this section, and discuss Brazil’s particular experience in two sections hence. 2.1. Having SOEs makes delivering benefits to constituents easier Governments want to deliver benefits to constituents. Suppose workers in an industry-X are part of the government’s constituents. By creating an SOE in the industry, the government finds it easier to deliver benefits to workers in the industry. For example, since the government appoints the managers of the SOEs, it can pressure the managers to provide generous wage and work rule packages to workers in the SOE. But this process of delivering benefits can ultimately lower private productivity. Continuing with the above example, since these wages and work rules benefits in the SOE are “generous,” they exceed those offered elsewhere in the industry, and these benefits will yield SOE losses. The government can provide direct subsidies to the SOE, but these will be observable. This can be politically costly, since the subsidies can be interpreted (in this case, correctly) as signals of nepotism, corruption, etc. Hence, the government has a strong incentive to eliminate these subsidies.2 There are much less transparent approaches to “subsidizing” SOE losses. These approaches will be “off-budget” and reduce private productivity. The government can reduce SOE losses if it can adversely impact the profitability of private firms in the industry. It can do this by either (1) adversely impacting the technology of private firms and/or (2) raising the prices of private-firm inputs. Both approaches will reduce private productivity. We discuss each in turn. 2.1.1. SOEs adversely impact private technology We give two examples where governments protect their SOEs by putting production restrictions on private firms. Australia had a policy of permitting only two domestic airlines, one private and one public. As Davies (1971) reports, the government endeavored to ensure that the private airline did not harm the SOE. For example, for each route in Australia, the government would determine biannually the expected capacity in ton-miles for the route over the next six months. The two airlines would then receive the right to carry exactly one-half that ton-mile capacity over the period. Neither airline could carry more than that capacity over the period. But the government soon realized that this was not enough to protect the SOE. By choosing a better mix of aircraft, the private airline was able to make profits (and cause the public to make losses). So, the government started requiring that the airlines have the same fleet of aircraft! In summarizing the consequences of such government intervention, Forsyth and Hocking (1980, p. 182) argue In such a situation, there is little scope for the private airline to be more efficient than the public one, and it is doubtful whether individuals associated with the private one are permitted to have any more incentives to improve efficiency. Any incentives in this system must finally be reduced to very low levels by the strong de facto controls on the profits of the private airline. 2 There is some theory to motivate why governments would make transfers to constituents in ways (like providing jobs with generous work rules) that did not involve simply monetary payments, that is, in seemingly inefficient ways (see, e.g., Coate and Morris, 1995).
748
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
As another example, consider the mail and package delivery industry. In the United States, the US Postal Service (USPS) is able to put restrictions on private delivery services. These restrictions include the regulation that private companies are prohibited from placing items in mailboxes. Amazingly, the Postal Service itself has the power to issue regulations regarding mail service (see US Congress, 2004). In his statement before the US Senate regarding the issue of postal reform, Fred Smith, the CEO of FedEx, stressed how the restriction on private firms from placing items in mailboxes led these companies to be less efficient. He urged dropping this restriction, arguing that “the end result of this reform will be to make the nation’s delivery services sector more efficient by extending the efficiencies of the national mailbox system to private delivery services.” These are two clear cases where the government gives generous SOE benefits, which lead to losses, which are then minimized by putting restrictions on private technology. 2.1.2. SOEs raise private-firm input prices Another approach to minimizing SOE losses is to raise the costs of private-firm inputs, like wages. Note well that if the government is able to raise the wages in private firms (i.e., the wages of its constituents), then it has killed two birds with one stone: it has lowered SOE losses and provided benefits to some of its constituents. If, for example, the government is able to raise wages at United Parcel Service (UPS), a private delivery service, it pleases UPS workers and it reduces USPS losses. How does the government raise the wages and improve the work rules of workers in the private firms? Governments, of course, often try to persuade private firms to provide greater wage and work rule benefits to workers. The key point is that the existence of an SOE in an industry may make the task simpler. For example, in some industries (like mining) the government can argue that certain work rules are needed for safety. It can then point to work rules in the SOE and argue that these are present to provide safety and that the private firms should use the same work rules. The government can use other laws (besides safety laws) as a way to coerce private firms to match SOE work rules and benefits. But the government does not have to be directly involved in using its safety laws, etc., to coerce private firms to provide similar work rules and wages as provided in SOEs. By simply providing generous benefit packages to SOE workers, it can enhance the bargaining position of private workers. This was the case, for example, in the Canadian railroad industry. Canada’s railroad industry historically comprised two major firms, a public firm, Canadian National (CNR), and a private firm, Canadian Pacific (CPR). Timur and Ponak (2002) discuss how in the late 1980s the public railroad (CNR) gave significant job protection to employees. In particular, shop craft employees with more than eight years of experience were guaranteed full pay if their jobs were eliminated through new technology and if no suitable substitute position was found. Timur and Ponak quote a senior private firm (CPR) executive as saying, “CNR [the public firm] first negotiated it [the job guarantees]. Then it became a tactical issue whether we would take a strike, lose business permanently perhaps, and the decision was taken not to strike.” In bargaining with two firms in an industry, unions will likely try to first reach a deal with a firm that is in the weakest bargaining position. It can then try to extend the deal to other firms. This happens in an industry with all private firms. The key point, again, is that the existence of an SOE in an industry may make the task of getting a generous industry-wide contract simpler. Finally, we need to tie this discussion to productivity. Generous work rules will lead to overstaffing and low labor productivity. They can also lead to low materials and capital productivity. Schmitz (2005) has shown this to be true. In particular, he showed it was true in the US and Canadian iron ore industries. 2.2. Having SOEs reduces industry competition We now argue that having an SOE in an industry (accounting for a significant share of output) reduces competition relative to the case with all private production. And then we will argue that this leads to lower productivity. There is no general theory that tells us more competition will lead to higher productivity. For example, there is no general result that tells us an industry-X with one private firm (i.e., a monopoly) will be less productive than the same industry with two firms. But we think there are convincing arguments for why having an SOE sector in an industry will mean lower productivity of all firms.
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
749
Imagine an industry that comprises two large private firms (say, with both having 50% of the market). Contrast this with an industry with a large SOE and a large private firm (again, say both have 50% of the market). There are two senses in which the latter industry will be less “competitive” and will have lower productivity than the former. First, SOE management likely has less incentive to reduce costs and capture market share than does a private manager. Governments typically can only offer limited compensation to managers, for political reasons. Second, a private manager likely has less incentive to reduce costs if the “other” half of the industry is an SOE rather than a private firm. That is because the government may be committed to keeping the SOE open, and it may be hard to capture market share. This then ties back to the previous discussion, that a private firm might be wary of reducing costs because the government may put technology restrictions on the firm, as in the Australian airline industry. In sum, there will be less pressure to reduce costs in an industry with an SOE and private firms (as opposed to all private firms). Productivity will be lower. 3. How does the “theory” apply to Brazil? In this section, we discuss how the “theory” discussed above applies to Brazil’s privatization experience. In particular, we discuss the mechanisms through which privatization increased private productivity. Before we turn to the productivity of private firms, we need to briefly address the typical question in this literature: Why would privatization of SOEs lead to SOE productivity increases? or Why was SOE productivity so low? As we argued, governments use SOEs to deliver generous wages and work rules to constituents. This leads to low productivity. This was the case here, with CVRD. Generous benefits were given to both management and workers. First, there was significant overstaffing of management. For many management tasks, there was a separate management team in the south and the north (as discussed below, CVRD had two major operating areas that were thousands of miles apart). But for many of these tasks, a single management team was needed. During the privatization process, in the early 1990s, much of this management redundancy was eliminated. This information was provided in interviews with CVRD management. Second, there were generous wage and work rule packages given to workers. As was common in the United States and Canada, work practices in Brazilian iron ore mines were such that employees engaged in only a small range of tasks. For example, there was typically a sharp separation of repair and production work. Those workers who ran machinery, for example, were not permitted to tend to their machines—this was reserved for repair staff. As another example, there were many repair job classifications. Repair staff with a certain classification were not permitted to conduct repairs assigned to other job classifications. In the US and Canada, these work practices changed in the early to middle 1980s, as these iron ore industries were faced with the possibility of closure (see Galdon-Sanchez and Schmitz, 2002 and Schmitz, 2005). But they persisted in Brazil until the onset of the privatization process. In the early 1990s, the management at CVRD began changing contracts so as to require workers to be responsible for more tasks. This process was called “polivalente,” the Portuguese term for jack-of-all-trades. We now turn to the main question: Why did SOE privatization lead to private-firm productivity gains? or Why was private productivity low before privatization? We discussed two general forces whereby SOE participation in an industry led to lower private productivity. First, because SOEs give generous benefits to their workers, this ultimately could impact private technology and/or private costs. In the Brazilian iron ore industry, the existence of CVRD enabled local governments to pressure private firms to provide their workers with work rules similar to those at CVRD. In particular, all iron ore mines in Brazil were required to receive annual certifications to operate. This certification was given at the local government level. There were zoning regulations, and other regulations, that each mine had to satisfy in annual reviews. Local government officials used these laws to help union leaders get generous benefits from private firms. In Brazil, there was a single union, Sindicato Metabase (Sindicato dos Trabalhadores da Industria de Metais de Base), that represented workers at both the public and private iron ore mines. This information regarding use of local regulatory procedures to pressure private mines to grant generous benefits was provided in interviews with union management. The workers at the private firms, then, benefited from the narrow job classifications at CVRD. The second force we discussed was that SOEs limited competition. This was clearly the case in this industry. CVRD was a massive iron ore producer. The transfer of its control to private investors clearly led to increases in the extent of competitive pressure in the industry.
750
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
Hence, with the advent of privatization in Brazilian iron ore, there would be two forces that spurred private firms to increase productivity. First, CVRD began to change its labor contracts, and to roll back its generous work rules. Hence, local governments lost leverage in attempting to force private firms to provide generous work rules. Second, the transfer of control of CVRD’s mines to private investors clearly led to increases in the extent of competitive pressure in the world iron ore industry. This gave added incentive for firms to change their labor contracts. We now turn to examine the data. In the next section, we present the institutional background. In the section following this, we present the productivity record of the industry. 4. Brazilian iron ore privatization: background In this section we briefly discuss the history of iron ore privatization in Brazil and also the sources of our data. 4.1. Iron ore privatization The Brazilian iron ore industry had two public companies that owned mines: CSN (Companhia Siderugica Nacional) and CVRD (Companhia Vale do Rio Doce). CSN was primarily a steel producer; in fact, it was Brazil’s largest (see Andrade et al., 2002). It owned a single iron ore mine. CVRD was primarily an iron ore producer, a massive one. In 1990, CSN produced about 6 million metric tons (mmt) of iron ore, CVRD 85 mmt, and Brazil 152 mmt. Brazil was the largest producer of iron ore (152 mmt), followed by Australia (110 mmt) and the United States (56 mmt).3 Hence, CVRD produced more iron ore in 1990 than any other country except Australia. Steel industry privatization in Brazil began earlier than iron ore privatization. The official plan to privatize CSN (and other steelmakers) was announced in 1988. The process of privatizing CSN’s steel operation, begun in 1988, was completed in 1993, when the CSN mine was also sold. The CSN mine was the first sold to private investors. CVRD did not appear on official privatization lists till much later than CSN, in fact, not until 1995. The government had decided to sell CVRD well before this date but knew it was going to face stiff resistance to selling it (see Kandell, 1997 for some of the difficulties faced in selling CVRD). Hence, it delayed a formal announcement. CVRD was the crown jewel of Brazilian state enterprises. The government had to placate many nationalist interests before the sale. One restriction on the sale was that Australian iron ore producers (who, again, were the largest producers after Brazil) could not be part of a bidding consortium. Also kept out of the bidding were Japanese steel producers. In the end, only two consortiums, both led by local entrepreneurs, bid on CVRD. It was not finally sold until 1997. While the CVRD privatization process was officially started in 1995 and completed in 1997, both CVRD management and union officials were unanimous in agreeing that it was evident much earlier that CVRD would be privatized (many give 1989 as the date it was clear). By the early 1990s, CVRD was undergoing major reforms and changes in its organization to prepare for privatization. As the preceding discussion suggests, there is no one clear date to label as the date of privatization. The process of privatization started before the actual sales to private investors, and even before the actual announcement that sales would be happening at some future date. Looking at the impact of privatization in this case, and likely all cases, is not like examining the impact of, for example, an unexpected tariff change at some date. To pick a round number, we pick the date that privatization began as 1990, though it could be argued that the process began even earlier, in 1989, or even 1988 as the decision to sell CSN was announced. 4.2. Data sources The basic source of data for mining industries in Brazil, including iron ore, are annual mine reports submitted to regional DNPMs (Departamento Nacional de Producao Mineral—National Mineral Production Department). These reports include the mine’s production (tons of iron ore by type of iron ore) and employment (as of the middle of December). These reports are then submitted to the Ministry of Mine and Energy for tabulation of aggregate production and aggregate employment levels for Brazil. These are published in Relatorio Anual de Lavras. From this publication, we obtain aggregate industry data from 1972 to 1999. 3 These statistics for Australia and the United States are from the United States Geological Society Minerals Yearbook. We are excluding China and the USSR from these rankings, since these countries reported production differently from the rest of the world.
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
751
In order to examine production and employment by mine, both private and public, we used the archives of yearly mine reports at DNPM (in Minas Gerais). We limited our search of these archives in the following way. First, there are a large number of very small iron ore firms in Brazil. Hence, to limit the search we gathered mine records for the public companies, CSN and CVRD, and for the three largest private companies, Ferteco (Ferteco Mineracao SA), MBR (Mineracoes Brasileiras Reunidas SA), and Samitri. Second, it was difficult to collect records prior to 1986. Not many mine reports were found prior to 1986. So, we gathered mine reports from 1986 to 1998.4 We also decided to examine the data at the firm level and not the mine. While the companies are supposed to report employment by mine, in practice the mines of a firm are often close to each other so that some staff (like repair staff) may work in more than one mine. Hence, it is up to the firm how to assign workers to mines in the reports. Moreover, there are some mines that report production but no employment. Because of this, we have chosen to look at production and productivity records at the firm level. We have made one exception to this strategy of looking at performance at the firm level. CVRD has operations in the south and the north of Brazil. Since these operations are separated by more than two thousand miles, the issues of repair staff working in different mines, etc., is not a major issue. Hence, we break CVRD into two “firms,” CVRDSouth and CVRD-North. What we have constructed, then, is the output and employment records of three public firms, CSN, CVRD-South, and CVRD-North, and the three largest private firms, Ferteco, MBR, and Samitri, from 1986 to 1998. In Table 1, we present the production and productivity of these six firms, together with the rest of the industry, for 1990. Production by public firms (90.8 mmt) amounted to about 60% of total production (152.2 mmt) in 1990. Production by the three largest private firms (41.5 mmt) amounted to about 27% of total production. The small private firms contributed about 13% of production. The production of the three largest private firms (again, 41.5 mmt) would have ranked them as the fourth largest national producer, behind Australia (110 mmt), the United States (56 mmt), and India (54 mmt), but ahead of the next largest producers, Canada (35 mmt), South Africa (30 mmt), and Venezuela (20 mmt). Productivity (defined as thousands of metric tons per employee per year) varies widely across firms. CVRD-North, which is the huge Carajas mine, has very high productivity, over twice the industry average. CVRD-North’s high productivity was due to its very high quality deposits. MBR and Ferteco had higher productivity than the other two public firms, CVRD-South and CSN. The third large private firm, Samitri, had low productivity. However, Samitri produced a type of iron ore, pellets, that requires more labor per ton than other ores. Hence, its low productivity is not surprising. Also not surprising is that the “rest of the industry” had very low productivity, less than half the industry average. The rest of the industry mostly comprised very small firms that worked marginal deposits.
Table 1 Brazilian iron ore production and labor productivity, 1990 (by public firms, large private firms, and rest of industry) Production
Firm production as percent of industry
Productivity
Firm productivity relative to industry
Public Firms CSN CVRD-South CVRD-North
5.83 52.12 32.85
3.83 34.24 21.58
7.40 9.66 24.17
0.81 1.06 2.66
Large Private Firms Ferteco MBR Samitri
10.93 19.97 10.61
7.18 13.11 6.97
10.58 14.92 5.63
1.17 1.64 0.62
Rest of Industry
19.93
13.09
4.02
0.44
152.24
100.00
9.08
1.00
Total Industry
Note: Production in millions of metric tons (mmt). Productivity in thousands of tons per employee per year. Source: See text.
4 We collected the data in the mine reports in 2000, and at that time only reports through 1998 were available.
752
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
5. Privatization’s impact on productivity: evidence In this section we show that there was little productivity gain in the Brazilian iron ore industry until the onset of privatization. Then, industry productivity dramatically increased. We also show that this pattern was found in both the public and private firms in the industry. 5.1. Productivity and production: aggregate industry The production and productivity records of the Brazilian iron ore industry over the period 1971–1999 are given in Fig. 1. In the figure we have drawn a vertical line at 1990 to indicate the year when the privatization process began. If we examine the decade of the 1980s, there was a modest increase in productivity. But this was mostly driven by the opening of CVRD-North, the Carajas mine, in 1986. As we saw, CVRD-North had significantly higher productivity than all the other Brazilian producers, private and public. Its opening pulled up the industry average.5 To see this, in Fig. 2 we plot the labor productivity of the Brazilian iron ore industry (the dashed line), and the industry not including CVRD-North (the solid line). CVRD-North only opened in 1986, so the two lines coincide before 1986. Note that the labor productivity of the industry not including CVRD-North did not increase much at all over the 1980s. It fluctuates between 8 and 10 thousand tons per employee per year, and is at 9 by 1990. In 1990, productivity was at early 1970 levels. The total industry, and the total industry not including CVRD-North, show significant productivity growth once the privatization process begins. There were significant gains prior to the culmination of the privatization process in 1997, that is, from 1990 to 1997, and then a gain after CVRD was sold.
Fig. 1. Production and labor productivity (thousands of tons per employee), Brazilian iron ore industry, 1971–1999.
5 The existence of the Carajas (CVRD-North) deposits had been known well before the middle 1980s. However, they were located in the interior of the Amazon, and massive infrastructure investment was needed before the iron ore could be extracted and shipped to the coast. A World Bank project supplied these funds in the middle 1980s.
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
753
Fig. 2. Labor productivity, Brazilian iron ore industry (thousands of tons per employee) with and without CVRD-North, 1972–1998.
In Fig. 3, we plot the productivity of the public firms (aggregated as a group) and the private firms (aggregated as a group). That is, the public sector is the aggregate of CSN, CVRD-South, and CVRD-North. The private sector is an aggregate of Ferteco, MBR, and Samitri. There is little productivity growth in either of the groups of firms until the privatization process starts. Then productivity growth is significant in both groups. We now turn to the productivity records of the individual companies. Again, we only have data beginning a few years prior to privatization. 5.2. Productivity: public firms The productivity of the public firms over the period 1986–1998 is given in Figs. 4, 5, and 6. In Fig. 4, we plot the productivity of CSN. We plot the (natural) logarithm of productivity (we draw the vertical line at 1988 in this figure, since that was the date at which CSN’s privatization process began). CSN had few productivity gains between 1987 and 1989 (for CSN, 1987 is our first year of data). Over 1989–1998, its productivity dramatically increased, by more than a factor of four. Recall that CSN’s productivity was lower than CVRD-South’s in 1990 (Table 1). Its spectacular productivity growth likely reflects that it was run much poorer than CVRD prior to privatization. With its dramatic productivity growth, it was essentially catching up to CVRD-South’s level. In Fig. 5, we show the productivity of CVRD-North (again, in logarithms). CVRD-North shows good productivity gains between 1986 and 1988 (about 33% growth). This was due to the fact that the Carajas mine had just opened in 1986. CVRD-North was ramping up production during 1986–1988 (its output more than doubled over 1986–1988). When iron ore mines open, their productivity grows rapidly for the first few years, until mine output reaches its rated capacity (see, e.g., Schmitz, 2005, where this pattern is seen in new mines opening in Minnesota). After 1988, when the mine had reached its capacity, CVRD-North’s productivity did not change for four years. Following the onset of privatization, CVRD-North’s productivity roughly doubled over 1992–1998. In Fig. 6, we plot CVRD-South’s productivity. Productivity increased little from 1986 to 1990. It increased by about 0.1 log points over this four-year period. Over the next four years, 1990–1994, its productivity increased by about 0.3 log points, and over 1994–1998, by nearly 0.6 log points. If we break the period in half, productivity grew
754
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
Fig. 3. Labor productivity (thousands of tons per employee) private (large firms) vs. public, Brazilian iron ore industry, 1986–1998.
Fig. 4. Labor productivity (thousands of tons per employee, in ln’s), CSN (public), 1987–1998.
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
755
Fig. 5. Labor productivity (thousands of tons per employee, in ln’s), CVRD-North (public), 1986–1998.
Fig. 6. Labor productivity (thousands of tons per employee, in ln’s), CVRD-South (public), 1986–1998 (solid line—actual productivity; two dashed lines—linear trends, 1986–1990, 1990–1998).
756
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
by 0.2 log points over 1986–1992, and 0.7 over 1992–1996. In the figure, we plot two linear regressions (the dashed lines) that are fitted to the productivity data, one for the period 1986–1990, the other for 1990–1998. The regression for the second period is much steeper than the first. Over 1990–1998, CVRD-South’s productivity more than doubled. Both CVRD-South and CVRD-North showed significant productivity growth prior to their formal sale in 1997. This pattern of major productivity gains in public firms before their formal privatization date is found in many privatization studies. For example, La Porta and Lopez-de-Silanes (1999) find this.6 The major changes at CVRD that led to productivity growth, the reduction of redundancy in management, and the changes in work rules, both discussed above, were begun well before CVRD’s formal sale. They were instituted so as to make CVRD a more attractive firm to purchase, thus enabling the government to raise more at its sale. 5.3. Productivity: large private firms The productivity of the three largest private firms over the period 1986–1998 is given in Figs. 7, 8, and 9. Here, again, we have only a few years of data prior to the onset of privatization. In Fig. 7, we plot Samitri’s productivity. There was little growth in Samitri’s productivity before the privatization process started (in fact, it fell over the short period we have before privatization). Samitri’s productivity shows much larger year-to-year swings than the other five firms (that is, the three public and two private firms). This may have been due to problems at some of its mines (such as major breakdowns of processing facilities), as its output also shows much greater variation than the other firms. In Fig. 8, we plot Ferteco’s productivity. There was little growth in Ferteco’s productivity before the privatization process started (in fact, it fell over the short period we have before privatization). Once privatization began, productivity grew strongly. Over 1990–1998, its productivity nearly doubled. In Fig. 9, we plot MBR’s productivity. MBR had little productivity growth between 1986 and 1988. Its productivity growth started earlier than the other firms, and it experienced significant growth over 1988–1991. MBR’s productivity, then, seems to have begun growing before privatization began, over 1988–1989. However, recall that the steel industry, and CSN’s, privatization began in 1988, so it is possible that MBR’s gains were tied to these developments. MBR experienced another surge in productivity beginning in 1995. Over the course of 1990–1998, its productivity more than doubled. In summary, the three private firms all show strong productivity growth during the privatization period. Productivity growth tends to start earlier in the private firms than it does in the public firms. This seems reasonable. Once privatization was known to be coming, local governments lost their leverage with the private firms. Private firms then could start changing their labor contracts, and could be expected to change them much more quickly than could public companies. 5.4. Summary, caveats, and quantitative discussion The results, then, are generally consistent with the “theory.” There was little productivity growth in the (aggregate) Brazilian iron ore industry until the onset of privatization. Then there was strong productivity growth. This pattern found at the aggregate level is found at the level of the public and private firms as well. It is unfortunate that we have only a few years of firm-level data before privatization. But the firm-level data before 1986 do not exist. The concern, of course, is that perhaps private productivity was growing strongly before 1986 when our data starts and that the flat private-firm productivity from 1986 to 1990 (in the case of Ferteco and Samitri) and from 1986 to 1988 for MBR was a fluke. There is one observation we can make regarding productivity growth before 1986. It perhaps adds a tiny amount of support to the view that private productivity growth was not strong before 1986. Recall there was little productivity growth at the aggregate level before privatization. Hence, if there was strong 6 They examine the employment levels in their sample of privatized firms in the four years leading up to the formal sale. On average, employment fell significantly in each of these years. In their Fig. I, they show average employment falls from a bit over 1450 in the fourth year before privatization to a bit over 1050 one year before privatization (a three-year period). In the last year before privatization, employment falls from a bit over 1050 to a bit over 550. La Porta and Lopez-de-Silanes surmised that these major reductions in employment before privatization were part of the government’s effort to prepare the SOEs for sale. That is certainly what happened in this industry: the large productivity gains at CVRD and CSN prior to privatization were due to the government’s effort to prepare these companies for sale (see below).
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
Fig. 7. Labor productivity (thousands of tons per employee, in ln’s), Samitri (private), 1986–1998.
Fig. 8. Labor productivity (thousands of tons per employee, in ln’s), Ferteco (private), 1986–1998.
757
758
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
Fig. 9. Labor productivity (thousands of tons per employee, in ln’s), MBR (private), 1986–1998.
growth at the private firms Ferteco, MBR, and Samitri before 1986, then it would likely mean that there was significant negative growth at CVRD-South (since before 1986 these four firms accounted for a very large share of the industry output, that is, CVRD-North had not yet opened). It is certainly possible that CVRD-South had a negative productivity trend till 1986, but perhaps not likely. Note that if its productivity growth was negative, it would imply that CVRD-South had significantly higher productivity than Ferteco and MBR in the years before 1986. That is because CVRD-South had roughly the same productivity as Ferteco and MBR in 1986. (This can be seen by looking at Table 1 where Ferteco and CVRD-South have roughly the same productivity in 1990. MBR’s productivity was about 50% higher than CVRD-South’s, but that was due to MBR’s very strong productivity growth over 1988–1990.) So, while it is possible that CVRD-South had higher productivity than the other firms in the decade before 1986, perhaps it is not likely. Finally, let us briefly discuss the quantitative gains in productivity. Many of the firms had productivity gains over the privatization period that were very large—roughly doubling of productivity. Much of these gains, we argue, were due to changes in the labor contracts used at the public and private mines. While we have not done an in-depth analysis of the labor contracts, from interviews we know the changes were very similar to those that occurred in the 1980s in the United States and Canada.7 Schmitz (2005) has done an extensive analysis of the impact of labor contract changes in these industries and has shown that much of the productivity gains (again, roughly doubling) in these industries resulted from the contract changes. He also discussed changes in Australia that were similar and led to large gains as well. The argument is by analogy, but we think it is a good one. 6. Other possible explanations? In this section we briefly consider the possibility that there were changes in the economic environment in the 1990s (besides privatization) that may have influenced the productivity of the entire Brazilian iron ore industry, that is, both 7 It proved very difficult to find copies of labor contracts that were more than a few years old.
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
759
the public and private iron ore producers. If this were the case, then we would be attributing too much of the gains in private productivity to the privatization of the public firms. First, we consider other economic reforms in Brazil (besides privatization). In fact, there was a lowering of tariffs in Brazil during the 1990s (see, e.g., Ferreira and Rossi, 2003). Trade liberalization could have led to industry-wide productivity improvements in Brazilian iron ore if imports of iron ore now challenged local iron ore in Brazil. But Brazil did not have any tariffs on iron ore. Brazil and Australia were the dominant players and exporters in the industry. Imports of iron ore were never in the cards. But productivity would also increase if the industry could now import intermediate or capital goods that were prohibitive to import before. Perhaps the industry could now import machinery that was too expensive to import before. Ferreira and Rossi (2003) provide effective rates of protection in Brazil over the period 1987–1997 by manufacturing industry (see their Table 2). Effective rates of protection in machinery and in transportation equipment were among the lowest in Brazil in 1987, and they did not fall very much over the period 1987–1997 (though effective rates of protection in some manufacturing industries fell dramatically). Hence, we believe that the impact of trade liberalization on the iron ore industry was small. Second, we consider general developments in the world iron ore industry. During the 1990s, as a result of changes in labor laws in Australia, its iron ore industry began to change its labor contracts, and this led to large gains in productivity (see, e.g., Galdon-Sanchez and Schmitz, 2002). Australia was now a much more efficient iron ore producer. Given that Brazil and Australia were by far the two largest exporters of iron ore, these developments may have significantly influenced the Brazilian industry. But we believe the impact on productivity from this effect was small as well. Australia was more productive, but adding capacity is very costly in this industry, so that without a major expansion in capacity (or a large fall in world demand for iron ore), Australia was not likely to capture much of Brazil’s market. 7. Conclusion Our theme has been that the productivity gains from privatization are much more general and widespread than has typically been recognized in the literature. This is because privatization of SOEs in an industry will typically lead to increases in private firm productivity in that industry—an influence that has been ignored in the literature. We began by presenting reasons why SOE privatization would raise private firm productivity. We then turned to the privatization of SOEs in Brazil’s iron ore industry. We showed that there was little productivity gain in the industry until the onset of privatization. Then there was significant productivity growth at both the public and private firms in the industry. We argued that much of the private productivity gain was due to forces set in motion with the privatization of the public companies. Acknowledgments We thank Sanghoon Lee and Shi Qi for excellent research assistance. We also thank the Associate Editor and two anonymous referees for very useful comments that have improved the paper a lot. For their help we thank Armando Castelar Pinheiro; staff of the labor union, Paulo Soares, Paulo Chaves, Jose Alberto Miguel, Ausgustinho Eliseo Vieira de Souza, Julio Freitas, Jose Maria, Anivaldo Coelho, Sebastiao Alves Oliveira, Joao Trevisan e Joao Cesar de Freitas Pinheiro; directors of CVRD, Flavio Ferreira, Jose Luis Martins, Marconi Tarbes Vianna; and the ex-President of CSN, Roberto Procopio Lima Neto. We are in great debt to all employees of DNPM-MG (Departamento Nacional de Producao Mineral—National Mineral Production Department) and the Ministry of Mine and Energy-MG. Teixeira thanks the Federal Reserve Bank of Minneapolis for its hospitality. References Andrade, M., Cunha, L., Gandra, G., 2002. Restructuring the Brazilian metallurgical sector. Sectoral Studies Series, BNDES. Borcherding, T.E., Pommerehne, W.W., Schneider, F., 1982. Comparing the efficiency of private and public production: The evidence from five countries. Zeitschrift für Nationalökonomie (Journal of Economics) (Supplement 2), 127–156. Coate, S., Morris, S., 1995. On the form of transfers in special interests. Journal of Political Economy 103 (6), 1210–1235. Davies, D.G., 1971. The efficiency of public versus private firms, the case of Australia’s two airlines. Journal of Law and Economics 14 (1), 149–165. Ferreira, P.C., Rossi, J.L., 2003. New evidence from Brazil on trade liberalization and productivity growth. International Economic Review 44 (4), 1383–1405.
760
J.A. Schmitz Jr., A. Teixeira / Review of Economic Dynamics 11 (2008) 745–760
Forsyth, P.J., Hocking, R.D., 1980. Property rights and efficiency in a regulated environment: The case of Australian airlines. Economic Record 56 (153), 182–185. Galdon-Sanchez, J.E., Schmitz Jr., J.A., 2002. Competitive pressure and labor productivity: World iron ore markets in the 1980s. American Economic Review 92 (4), 1222–1235. Kandell, J., 1997. The CVRD Factor. Institutional Investor (October), 186–192. La Porta, R., Lopez-de-Silanes, F., 1999. The benefits of privatization: Evidence from Mexico. Quarterly Journal of Economics 114 (4), 1193–1242. Megginson, W.L., Netter, J.M., 2001. From state to market: A survey of empirical studies on privatization. Journal of Economic Literature 39 (2), 321–389. Schmitz Jr., J.A., 2005. What determines productivity? Lessons from the dramatic recovery of the US and Canadian iron ore industries following their early 1980s crisis. Journal of Political Economy 113 (3), 585–625. Timur, A.T., Ponak, A., 2002. Labor relations and technological change at Canadian Pacific Railway. Journal of Labor Research 23 (4), 535–557. US Congress. Senate. Committee on Governmental Affairs. 2004. Statement of Fred Smith, “Postal Reform: Sustaining the 9 Million Jobs in the $900 Billion Mailing Industry (Day 2),” March 11, 2004. http://murray.senate.gov/comm/governmental_affairs/general/ index.cfm?Fuseaction=Hearings.Testimony&HearingID=157&WitnessID=545.