- Email: [email protected]
Informational rents in oil and gas concession auctions in Brazil
Energy Economics 46 (2014) 93–101
Contents lists available at ScienceDirect
Energy Economics journal homepage: www.elsevier.com/locate/eneco
Informational rents in oil and gas concession auctions in Brazil Eric Universo Rodrigues Brasil a, Fernando Antonio Slaibe Postali b,1,⁎ a b
Tendencias Consulting and University of São Paulo, Brazil Department of Economics, University of São Paulo, Brazil
a r t i c l e
i n f o
Article history: Received 22 July 2011 Received in revised form 28 February 2013 Accepted 4 September 2014 Available online 16 September 2014 Keywords: C14 D44 D82 L71 Q35 Q38
a b s t r a c t This article aims to estimate the informational rents earned by winning bidders in oil and gas exploration and production auctions in Brazil. We estimate the distributions of bids and bidders' private valuations using a nonparametric structural model and assuming independence and asymmetry between participants. Petrobras, the Brazilian state-owned petroleum giant and former oil sector monopolist, was considered a competitor that was distinct from other competitors. Thus, we investigate a database based on information from all auctions held between 1999 and 2008. The results suggest that Petrobras earned significantly higher information rents than other competitors. Such rents ranged from 15% to 63%, depending on the number and type of competitors. © 2014 Elsevier B.V. All rights reserved.
Keywords: Oil and gas industry Auctions Regulation Nonparametric estimation
1. Introduction The modern regulatory framework for the Brazilian oil industry was instituted with the passing of the Oil Act (Lei do Petróleo, 9.478/97) in 1997. The new law regulated the Constitutional Amendment (Emenda Constitucional, 9/95) that broke the monopoly that Petrobras had enjoyed on oil and gas exploration and production activities. One of the most significant changes was the creation of auctions – known as Bidding Rounds – that controlled the concession rights for exploration in inland and continental shelf areas. The auctions are first-price sealed-bid in which bidders submit concealed bids that are opened simultaneously with one another. The winner is the bid with the highest score on a set of criteria, including the bid offered, the exploration program and the purchase commitment with domestic suppliers. Since 1999, 10 Bidding Rounds have been conducted, and the results have been quite significant in terms of revenue and the entry of private companies into oil and gas activities in Brazil.
⁎ Corresponding author at: Av. Prof. Luciano Gualberto, 908, Cidade Universitária. CEP: 05508-900, São Paulo-SP, Brazil. Tel.: +55 11 30915915; fax: +55 11 30916013. E-mail addresses: [email protected] (E.U.R. Brasil), [email protected] (F.A.S. Postali). 1 Sponsored by CNPq — Brazilian National Council for Research and Development.
http://dx.doi.org/10.1016/j.eneco.2014.09.002 0140-9883/© 2014 Elsevier B.V. All rights reserved.
There are two significant issues related to the new regulatory framework and the role of these auctions in Brazil. The first concerns the involvement of Petrobras, the state-owned Brazilian oil giant. Because it operated as a monopolist for four decades and has amassed data on the potential of Brazilian oil fields, it naturally enjoys an information asymmetry in such auctions. The second issue is related to the efficiency of the auctions in extracting information rent from bidders, i.e., how closely the bid approaches to the private company's actual valuation of the auctioned area. The aim of this paper is to estimate the information rents earned by the winning bidders in ten rounds of bidding through the use of a nonparametric estimate of the distributions of auction bidders' valuations. Thus, it is possible to analyze how the government collects the bidders' surplus and how successful bidders are in winning the concession with the lowest possible value. Specifically, we are interested in assessing whether there are significant differences in rents earned by Petrobras compared to the other bidders. The results suggest that asymmetries that favor Petrobras generate higher information rents for the company and that, as expected, auctions with fewer participants generate a lower surplus extraction for the government. Because the Brazilian government plans resumed the auctions in 2013, we aim to contribute to the discussion about improving the concession policies of exploration areas following the pre-salt drilling program.
94
E.U.R. Brasil, F.A.S. Postali / Energy Economics 46 (2014) 93–101
Auctions are selling mechanisms that are often characterized by asymmetric information (McAfee and McMillan, 1987). Auctions are attractive to sellers because it may be possible to extract the whole consumer surplus, even though the auctioneer does not know the bidders' willingness to pay. Auctions would be meaningless as selling mechanisms if private valuations were public – if there were perfect information – because the bidder would merely set a price equal to the bidder's valuation of the object sold. Auctions are also useful when the value of the auctioned good is unknown for both sellers and bidders, as in the case of an oil and gas concession area (McAfee and McMillan, 1987). The relationship between the auctioned object and the auction's information structure has generated two main models (Laffont and Vuong, 1996). The first one is the private value auction, in which bidders perform their own valuation of the good, independently or not from the valuations of other bidders. The second model is known as the common value auction (also known as the mineral rights model), in which bidders receive a common signal on the value of the auctioned object that is unknown ex ante. Auctions are the most common form of allocating oil and gas exploration rights worldwide. The choice of the underlying model to analyze them is a complex task and can be controversial. The pure common value model was developed driven by the concession of oil and gas exploration areas (to such an extent that it is also called the mineral concession model) and includes Wilson (1969) and Ortega-Reichert (1967) among its pioneers. An important characteristic of mineral auctions is that companies attempt to estimate a common value for the auctioned goods. Until the mid-1990s, the common value model was dominant in hydrocarbon concessions because of the specific characteristics of the sector, which involve granting the rights to explore and produce an unknown amount of resources at uncertain costs. Technological constraints that prevailed in the oil industry until the 1950s contributed to the permanence of the mineral concession model. Under this paradigm, participating bidders receive a signal on the value of the auctioned object and thereby set their bids. Reece (1978) shows that increased uncertainty and a reduced number of bidders contribute to raising the rents of the winning bidder at the government's expense. Reece (1979) also simulated the rent sharing between the government and bidders under different contractual arrangements and concluded that expected governmental revenue from auctions is higher under a profit-sharing plan and lower under a single bonus plan. Under this structure, imperfect information may result in a ‘winner's curse’ (Kagel and Levin, 1986); i.e., the winning bidder's tendency is to overestimate the value of the object, assuming that the average estimate is non-biased. Thus, the winning bidder would have paid a value above the actual value of the object. In this context, there are several contributions to the literature on oil and gas auctions, including articles concerning the performance of bids when there are information asymmetries between bidders. Hendricks et al. (1994) examined how such information asymmetries affect the strategic behavior of bidders under the common value model. The analysis was performed using a bidder with more private information than all the other bidders (who only have access to public information). The authors conclude that the equilibrium behavior in this information structure often drives bidders with little information to offer less than informed bidders. Similarly, Porter (1995) builds a stylized model with asymmetric competitors (informed versus non-informed bidders). A significant result of this model is that informed bidders are more likely to win the auction and to show positive information rent when they win, whereas noninformed competitors report a null payoff. Thaler and Tucker (1995) report strong evidence of a winner's curse in auctions in the Gulf of Mexico between 1954 and 1969, based on a quantitative survey of unproductive and barely profitable wells. Although it initially focused on common value, the literature on oil auctions began to recognize the existence of a private component in
the bid launching process because of bidder heterogeneity (Cramton, 2007). Indeed, Li et al. (2000) showed that the private component is relevant to oil auctions in the US OCS data while estimating a conditionally independent private value model. This led to a reformulation of the analysis model regarding this form of bidding to include the private component of bids. Moreover, recent technological developments have enabled companies to gain cost advantages that contribute to affect their assessment of the good, which created a rationale for the use of private value that is based on structural models. Accordingly, the number of empirical analyses has substantially increased in recent years. The premise of such models is to estimate the unobserved private values (termed pseudo-values) under the hypothesis that optimal strategies are defined according to an underlying game.2 Currently, the literature features significant contributions to the structural estimation of auction models using the non-parametric approach, including Guerre et al. (2000), Li et al. (2002), Hendricks et al. (2003), Campo et al. (2003), Bajari and Hortaçsu (2005), Flambard and Perrigne (2006), among others. In private value auctions, information rent is defined as the winning bidder's surplus, i.e., the difference between the bidder's private value and the actual bid launched. Information rents are so named because the bidder's valuation of the auctioned object is private information that precludes the auctioneer from fully discriminating bidders (Krishna, 2002, chapter 5), and the winning bidder benefits from a rent for the partial disclosure of his private information. The lower the winning bid is relative to the private valuation, the higher the bidder's information rent for the auctioned object. The presence of information rents is inherent in the private value mode. Depending on the type of auction, the role of the equilibrium bid – the optimal bid as a function of private value – produces a lower offer than the actual private value. This occurs, for example, in firstprice sealed-bid auctions (Paarsch and Hong, 2006, chapter 2); the auctioneer will be more efficient the closer the winning bid is to the maximum private value among bidders involved in the auction. The estimation techniques of these private values became progressively more complex over time. Following Li et al. (2000), Campo et al. (2003) estimated information rents of US OCS auctions in the 1970s using an affiliated private value model with asymmetric bidders. The results show information rents of approximately 65%, which were higher for consortium bids. Using simulated private value bids, Li et al. (2002) estimated information rents of approximately 38% for twobidder auctions and 29% for three-bidder auctions. Flambard and Perrigne (2006) estimated information rents and evaluated the inefficiency of auctions in hiring snow removal services in Canada under the hypothesis of independent private value and asymmetry between bidders, which was defined by costs that depended on the location of the company. The asymmetry of costs is conclusively relevant and accounts for inefficient allocations. The current paper has an empirical nature and aims to apply the methodology of the articles described above to the evaluation of information rents in oil and gas bidding auctions in Brazil. If companies are appropriating large private surpluses, there may be a low level of competition in such auctions, which might indicate that policy measures should be undertaken to correct this. This article is structured in 5 sections, including this introduction. Section 2 describes and analyzes the database, including a discussion about the main determinants of bids, and Section 3 discusses the theoretical model underlying the estimation of information rent and empirical estimation procedures in general terms. Section 4 describes 2 As highlighted by Perrigne and Vuong (1999), there are two approaches for estimating and treating data. The first is parametric, which estimates the parameters through maximum likelihood under a known distribution of unobserved values. The second is nonparametric, which is based on indirect procedures linking private value to observed bids. The disadvantage of the nonparametric approach is that it requires a large database of bids, whereas the parametric one requires less data but utilizes the computation of a function defining the optimal strategy, which may not be feasible.
E.U.R. Brasil, F.A.S. Postali / Energy Economics 46 (2014) 93–101
95
Table 1 Summary of ten annual oil and gas bidding rounds in Brazil. Round
Blocks sold Total bids Auctions with Petrobras involvement (%) Petrobras bids Petrobras winning bids Petrobras success rate (%) Consortium bids Consortium winning bids Success rate of consortium bids (%)
1
2
3
4
5
6
7
8
9
10
Total
12 21 58 7 5 71 11 6 55
21 46 48 10 8 80 19 10 53
34 57 59 20 15 75 19 13 68
21 33 43 9 8 89 5 5 100
101 106 88 89 88 99 3 3 100
154 188 73 113 107 95 79 64 81
251 379 43 109 96 88 131 87 66
38 84 58 22 21 95 24 15 63
117 276 49 57 27 47 108 46 43
54 92 52 28 27 96 29 22 76
803 1282 58 464 402 87 428 271 63
Source: Prepared by the authors. The first round occurred in 1999 and the last occurred in 2008.
Table 2 Observations per number of bidders. n
Auctions
Bids
Petrobras
Others
1 2 3 4 5 6 7 8
528 164 60 26 14 7 2 2 803
528 328 180 104 70 42 14 16 1282
283 97 45 17 12 6 2 2 487
245 229 135 87 58 36 12 14 795
Source: Prepared by the authors.
the results and Section 5 offers concluding remarks, which include certain policy suggestions that might increase competition in these auctions and (therefore) the governmental appropriation of private rents. 2. Data The database was obtained from the National Petroleum Agency (Agência Nacional do Petróleo—ANP) and is composed of information on all auctions held over ten annual Bidding Rounds3 between 1999 and 2008. The data provide information on the auctioned block area, its basin and sector, the round in which it was sold, its location (land or offshore), the number of bidders in each auction, the bidding companies and/or bidding consortia, the operating companies and the bid value offered. Companies and/or pre-qualified consortia have 1 min to place their bids in sealed envelopes. All envelopes are opened simultaneously. Until 2002, the bid value represented 85% of the score defining the winning bidder; the remaining 15% of the score was related to a commitment to using domestic suppliers. From 2003 onward, there was an important shift because the exploration program offered by the investor began to be incorporated into the score, and the percentages assigned to each criterion (bid, local content and exploration program) began to vary from year to year. In approximately 8% of the auctions, the winning bidder failed to offer the highest bid.4 The database records 109 companies that took part in some way in ANP auctions over these ten years, in addition to Petrobras. Although some of these are large international corporations, such as Amerada Hess, Shell, Repsol YPF, Chevron and BP, none have the same level of knowledge that Petrobras has with respect to the quality of Brazilian deposits, which justifies the classification of bidders into two main 3 The term “Bidding Round” refers to a set of auctions held once a year. Each round is held over one or two days in which multiple blocks are sequentially auctioned. 4 The Brazilian auctions suffer from intrinsic inefficiencies arising from the possibility of selling the tract to a candidate that has not offered the highest bid. Therefore, we chose to evaluate only information rent because the efficiency (that is, the block allocation to the bidder with the highest willingness to pay) does not appear to be a priority for the Brazilian government in these auctions, particularly after 2003.
groups – Petrobras and Others – according to the assumption described below.5 Table 1 shows a quantitative summary of the involvement and bids of Petrobras in Bidding Rounds: 803 blocks were auctioned that received a total of 1282 bids, which represents an average of 1.53 bids per block sold; 311 blocks were located offshore and received 396 bids; 192 were located on land and received 886 bids. The success rate of Petrobras – that is, the number of winning bids in relation to total bids – was 87% (in 2003, it was 99%); Petrobras participated in 53% of the auctions held. Conversely, consortium bids show a 63% success rate. Table 2 shows the distribution of auctions based on the number of bidders. More than half of all blocks sold received only one bid (528 out of 803), which shows the low-level of competitiveness in the auctions. The reasons for this phenomenon may be related to the high degree of geological uncertainty at the auctioned areas, according to Reece (1978), or because of the perception of a regulatory risk that is reflected in the high standard deviation of expected returns (Reece, 1979); this scenario is consistent with high information rents for winning bidders. Table 3 shows a statistical summary of bids launched over the ten rounds; it indicates that the bids of Petrobras (per unit area) were, on average, much higher than those of other bidders.6 Another important feature is that bids made by Petrobras against non-informed bidders (Petrobras versus Others) are, on average, higher than the others' bids among themselves (Others versus Others); in addition, non-informed bidders tend to offer more aggressive bids when competing with Petrobras (Others versus Petrobras) than when competing against their peers (Others versus Others). This suggests that each bidder's strategy depends on the bidder's opponent, which may correlate with the presence of asymmetric information between Petrobras and the other bidders, as speculated by Campo et al. (2003). Finally, the last section in the table indicates that consortium bids are slightly higher than bids launched by single companies. This difference is somewhat more significant when the consortia contains only non-informed bidders (excluding Petrobras).7 Table 4 shows the results from the estimation of the probability of winning the auction given certain observable characteristics and using a probit model; this estimation aims to report a preliminary study on the determinants of bids. The dependent variable is a dummy assuming
5 There have also been large national companies that had no prior experience in oil and gas exploration but decided to enter into auctions after the Petrobras monopoly was ended, such as Vale do Rio Doce, Queiróz Galvão and Odebrecht. 6 This is stylized evidence from the findings of Porter (1995), in which the informed bidder has a higher probability of launching bids and a higher rate of success. 7 As Petrobras has a type-A operator status, it has the option of launching individual or consortium bids (with other companies). Although it may theoretically compete with itself, no records are found in the database of situations in which Petrobras launched an individual bids against one of its consortia. It is unlikely that this would occur in practice because Petrobras is almost always the operator company – the company responsible for conducting the operations – when participating in consortia, and it therefore has the prerogative to decide to bid.
96
E.U.R. Brasil, F.A.S. Postali / Energy Economics 46 (2014) 93–101
Table 3 Statistical summary of bids. Variable
Obs.
Average (R$/km2)
Standard deviation
Coefficient of variation
Minimum
Maximum
All bids Petrobras bids Others bids Petrobras vs. Others Others vs. Petrobras Others vs. Others Consortium bids (including Petrobras) Individual bids (including Petrobras) Consortium bids (excluding Petrobras) Loose bids (excluding Petrobras)
754 181 563 174 335 225 266 488 155 408
11,805.10 18,269.59 7416.29 15,988.71 7959.86 6082.61 12,773.92 11,277.01 9810.50 6506.72
30,720.35 30,408.55 17,597.70 24,043.56 18,124.16 15,783.04 27,106.81 32,536.00 20,413.60 16,337.91
2.60 1.66 2.37 1.50 2.28 2.59 2.12 2.89 2.08 2.51
1.58 2.99 1.58 2.99 1.58 1.64 2.99 1.58 3.70 1.58
440,494.00 208,789.90 206,146.00 118,089.00 206,146.00 118,788.70 208,789.90 440,494.00 118,788.70 206,146.00
Source: Prepared by the authors. Table 4 Determinants of the probability of winning the auction.
Table 5 Determinants of observed bids. Dependent variable: log(bid/km2)
Dependent variable: winning bid = 1 (1)
(2)
−1.046⁎⁎⁎ (0.0816) 0.181 (0.158) −1.172⁎⁎ (0.457) 0.619⁎⁎⁎ (0.0605) 0.00016 (0.00010) 3.664⁎⁎ (1.548) −1.564 (1.688) 2.852⁎⁎⁎
−1.087⁎⁎⁎ (0.0950) 0.247 (0.189) 0.127 (1.484) 0.459⁎⁎⁎ (0.0864) 0.00043⁎⁎
Year dummy Basin dummy Enterprise dummy Constant
(0.619) (0.756) – – Yes Yes Yes −5.751⁎⁎⁎
(0.610) (0.838) 0.00056⁎⁎⁎ (0.00011) Yes Yes Yes −11.46⁎
Observations Pseudo-R2
(1.429) 1339 0.506
(6.612) 1107 0.544
Number of bidders Consortium dummy Offshore block dummy Log(Bid/km2) Area (km2) Exploration (%)(a) Development (%)(b) Petrobras dummy
Proposed investment program(c)
(0.0002) 4.051 (3.628) 5.095 (9.587) 3.032⁎⁎⁎
Robust standard errors in parentheses. ⁎⁎⁎ 1%, ⁎⁎ 5%, and ⁎ 10% significance levels. (a) (b) (c)
Commitment to local suppliers in the exploratory phase. Commitment to local suppliers in the development phase. Introduced after 2003.
the value 1 when the bidder wins the auction.8 Specifications (1) and (2) differ due to the inclusion of the exploration program offered, which was introduced in 2003. The number of competitors reduces the bidder's probability of winning an auction, whereas consortium formation does not affect this probability.9 As might be expected, the bid launched (log(bid/km2)) and the minimum exploration program positively affect this probability, whereas the commitment to domestic suppliers in the exploration phase positively affects this probability only with specification (1)—when the exploration program offered is not considered. In addition, the offshore blocks are less likely to be auctioned in this case, which reflects the lower levels of competition for these areas. The signal obtained for the dummy variable identifying
8 It should be noted that this is not necessarily the highest bid but, rather, is the bid from the bidder who actually won the auction under the three scoring criteria (bid, purchase commitment with local suppliers and exploration program, if applicable). 9 This result is consistent with Iledare et al. (2004), who conclude that consortium bids are common strategies in Outer Continental Shelf (OCS) auctions in the Gulf of Mexico (USA) and do not significantly affect the bids made. Hoffman et al. (1991) found the same evidence using data from 1970.
Number of bidders Consortium dummy Offshore block dummy Exploration (%) Development (%) Petrobras dummy Proposed investment program Year dummy Basin dummy Enterprise dummy Constant Observations R-square
(1)
(2)
0.108⁎⁎⁎ (0.0239) −0.214⁎⁎
0.0389⁎⁎ (0.0183) −0.120 (0.0821) 2.748⁎⁎⁎ (0.605) 0.517 (2.066) 13.16⁎⁎⁎
(0.0980) 0.802⁎⁎ (0.311) 2.557⁎⁎ (1.115) −1.721 (1.178) −0.0604 (0.339) – – Yes Yes Yes 4.445⁎⁎⁎ (0.710) 1468 0.858
(4.640) −0.284⁎⁎ (0.126) 0.00068⁎⁎⁎ (4.11e−05) Yes Yes Yes −4.141 (2.700) 1226 0.932
Robust standard errors in parentheses. ⁎⁎⁎ 1%, ⁎⁎ 5%, and ⁎ 10% significant.
Petrobras as participating bidder – which is positive and significant in both specifications – indicates asymmetry, such that Petrobras has an above-average probability to win a block.10 Table 5 shows the results of the regression of the bid value per unit area (in log). As the sign of the coefficient regarding the number of bidders is positive and significant, competition drives bidders to launch higher bids,11 and higher bids are launched for offshore blocks. Interestingly, consortia launch lower bids when the exploration program is not considered (specification 1); in this case, the national commitment to the exploration phase also positively affects the bid. The effect of the investment program offer, though small, is positive and statistically significant at 1%. Petrobras bids, controlled for all other characteristics, are not significantly different from the other companies' bids in specification (1); this is notable because the exploration program (specification 2) indicates that the Brazilian state company launches bids below the average bids of other companies. Because Petrobras is more likely to win the auction (Table 4), launching bids that are not higher than the mean may be further evidence of asymmetry, which indicates that Petrobras has more accurate information on the value of the blocks. But this result merits further investigation, which are beyond the goal of this study.
10
These results are consistent with predictions by Porter (1995). In a detailed investigation of the determinants of North-American auction bids (OCS) in the Gulf of Mexico between 1983 and 1999, Iledare et al. (2004) note that competition increased the mean value of bids and that large companies launched bids above the mean. 11
E.U.R. Brasil, F.A.S. Postali / Energy Economics 46 (2014) 93–101
3. Information rents and asymmetries in Brazilian auctions Assuming the independent private value model, the aim of this study is to estimate the information rent (IR) earned by the winning bidders of oil and gas exploration auctions in Brazil, calculated as follows: IRi ¼
vi −bi vi
ð1Þ
where vi is the bidder's private value, and bi is his/her actual bid. Each bid is known, but the bidders' respective private values are not. We use the method proposed by Guerre et al. (2000) to estimate them, complemented by Flambard and Perrigne (2006) for asymmetric auctions. This is a non-parametric two-stage procedure, computationally feasible and applied to models of first-price sealed-bid auctions under independent private value. The method has the advantage of avoiding a functional form for the Bayesian Nash equilibrium strategy (that is, bidding as a function of private value bi(vi)). We assume that each bidder is risk neutral with private value vi and ignores the private value of others. Furthermore, all private values independently originate from the same distribution F(.) with density f(.). As shown in Guerre et al. (2000)12 and further explained by Flambard and Perrigne (2006),13 the distribution function of private values can be identified from the distribution of bids, G(.) for n ≥ 2, where n is the number of bidders. We assume that there is no minimum bid, unlike Flambard and Perrigne (2006). This is a simplifying assumption, but it fits reasonably well in the Brazilian case because the reserve bids fixed by the government were low and varied very little from area to area; it is virtually impossible that they would have caused the exclusion of any potential candidate. 14 Because the bidder of type i launches a bid bi, its probability of winning the auction is G(bi)n − 1, given the assumption of independent private value. Therefore, the bidder's expected payoff from launching a bid bi is given by: n−1
Πi ¼ ðvi −bi ÞGðbi Þ
ð2Þ
Maximizing the above expression regarding bi results in: 1 Gðbi Þ vi ¼ bi þ n−1 g ðbi Þ
12
bidder is a function of the bidder's own private signal and a common unknown value arising from a signal received by all the participants. However, the estimation of this model requires the selection of auctions with exactly two participants, which would entail excluding a substantial portion of our database. Thus, we opted to utilize the independent private value model to maximize the use of data. In addition to independence, we also assumed additive values, i.e., the value of each block is independent of the value of previously auctioned blocks. A consequence of this hypothesis is that the revenue earned by a set of blocks auctioned in a single package is equivalent to the sum of revenues of blocks auctioned separately. As Cramton (2007) argues, additive values are good approximations for oil and gas auctions because the amount of hydrocarbon resources in a set of blocks is the sum of the resources of each block at a given price. As a result of the state monopoly in oil exploration that Petrobras enjoyed between 1953 and 1997 in Brazil, it acquired extensive knowledge about the geological characteristics of Brazilian sedimentary basins. Now, it is recognized as one of the world's most productive energy companies. Therefore, oil exploration auctions in Brazil, according to the data analysis from the previous section, suggest a strong asymmetry favoring Petrobras. As emphasized by Maskin and Riley (2000), asymmetric auctions are extremely complex and have distinct characteristics, including the possibility of inefficient allocation (because the object may be sold to a bidder who does not have the highest private value among bidders) and non-validity of the Revenue Equivalence Theorem (Myerson, 1981). In symmetric auctions, bidders use the same equilibrium strategy as a function of their types—that is, the function bi(vi) is the same for all bidders; in asymmetric auctions, bidders may have the same information on the object but might differ with respect to opportunity costs and budget constraints, which may lead them to yield different valuations (Maskin and Riley, 2000). Thus, in the Brazilian case, we assume independent private value with asymmetric competitors – Petrobras versus Others – to balance the simplifying assumptions and model the asymmetry. Type i = 1 represents Petrobras and type i = 2 represents Others. In this case, Eq. (3) may be applied to the two groups of bidders. Given that G1(.) and G2(.) are the probability distributions of the bids of Petrobras and Others, respectively, the private values may be estimated by: v1 ¼ b1 þ
1 G2 ðb1 Þ n−1 g 2 ðb1 Þ
ð4Þ
v2 ¼ b2 þ
G1 ðb2 ÞG2 ðb2 Þ g 1 ðb2 ÞG2 ðb2 Þ þ ðn−2Þg2 ðb2 ÞG1 ðb2 Þ
ð5Þ
ð3Þ
The model adopted in the present study assumes the independence of the bidders' private values. As described above, the first studies on oil auctions focused on the common value model. However, as the methodology developed, the literature on private value gained momentum. In the case of Brazil, each company's peculiarities, including different technological capabilities in oil exploration, suggest little likelihood that the auctioned block has the same value ex post, irrespective of the winner; this is true for Petrobras in particular because it has developed specific technologies over time as the sole operator in the Brazilian oil sector. Conversely, it is reasonable to assume that auctions for oil exploration areas lie between the extremes of private value and common value, in what Milgrom and Weber (1982) termed an “Affiliated Private Value Auction”. In this class of models, the utility of each Theorem 1, p. 553. 13 Proposition 1, p. 1023. 14 In fact, until 2002, the areas were ranked into three categories, each with a uniform reserve price. In subsequent years, reserve bids were individualized per area, but their values were close to those set in the early rounds. The surprisingly high premiums – some reaching 50,000% – suggest that fixed reserve bids were more symbolic than effective; it is virtually impossible that they might have caused the exclusion of a competitor previously eligible to participate in the auction. Thus, for the purpose of simplification, we chose to assume that the auctioneer's reserve bid is irrelevant for the decision to enter the auction because this also represents the reality.
97
where n is the total number of bidders, including Petrobras. After dividing the observations into two groups according to the pattern of asymmetry described above (Petrobras versus Others), the results were estimated based on a three-step procedure, following Guerre et al. (2000) and Campo et al. (2003). The first step is to non-parametrically estimate the distribution of observed bids using a Kernel estimator. According to Li and Racine (2007, chapter 1), the type of Kernel does not have a significant effect in practical terms; therefore, we use a Triweight function because it satisfies all hypotheses required, as argued by Guerre et al. (2000). Thus, the probability density function at a point x is given by: f ðxÞ ¼
m 1X x−xð jÞ K m j¼1 h
ð6Þ
where m is the number of observations (that is, the number of auctions corresponding to each type), K(.) is the Kernel function, and h is the bandwidth such that ∫ K(t)dt = 1. The Triweight Kernel is given by: K ðuÞ ¼
35 2 3 Iðjuj≤1Þ 1−u 32
ð7Þ
98
E.U.R. Brasil, F.A.S. Postali / Energy Economics 46 (2014) 93–101
Distribution function
Cumulative distribution function
0.2
1 Petrobras Others
0.16
0.8
0.14
0.7
0.12
0.6
0.1 0.08
0.5 0.4
0.06
0.3
0.04
0.2
0.02
0.1
0
0
2
4
6
8
10
12
Petrobras Others
0.9
Probability
Probability
0.18
0
14
0
2
4
6
log(1+b) Distribution function
12
14
Cumulative distribution function Petrobras Others
0.18 0.16
0.8
0.14
0.7
0.12
0.6
0.1 0.08
0.5 0.4
0.06
0.3
0.04
0.2
0.02
0.1
2
4
6
8
10
12
14
Petrobras Others
0.9
Probability
Probability
10
1
0.2
0
8
log(1+b)
16
0
2
4
6
8
10
12
14
16
log v
log v Fig. 1. Auctions with n = 2 bidders.
The bandwidth parameter is a sequence of constants converging toward zero as the number of observations tends to infinity. It defines the neighborhood surrounding a point x in which the Kernel performs the estimation. Low h values decrease the bias, but increase the variance of the nonparametric estimator ^f ð:Þ (Paarsch and Hong, 2006, chapter 4). The bandwidth regulates the smoothness of distribution; high h values mean smoother distributions with low variances but with a higher bias. According to Li and Racine (2007, chapter 1) and Guerre et al. (2000), we use the optimal bandwidth given by: ^ bn h ¼ cσ
−1=5
ð8Þ
^ b is the standard deviation of observed bids, n is the number of where σ bids and c is a constant. This is the bandwidth used in Li et al. (2002) and Campo et al. (2003) and is based on a “rule of thumb” developed by Silverman (1986) that minimizes the integrated mean squared error, which is composed of the sum of two components, i.e., a measure of the bias and a function of variance (Li and Racine, 2007, chapter 1). For a Triweight Kernel, we use c = 2.96. The second step is to estimate the pseudo-values using (4) and (5) and the information rent using (1). In the third and final step, the pseudo-values calculated are used to non-parametrically estimate the
distributions of private values with the procedure applied in the first step.
4. Results Figs. 1 and 2 show the estimated distributions of observed bids and private values (in log15) for 2- and 3-bidder auctions, respectively. The difference in the densities estimated for the groups is notable. In both cases, we observe the stochastic dominance of the distribution of bids and the private values of Petrobras; i.e., the Brazilian state company shows, on average, higher private values than the other bidders, which is further evidence of the asymmetries discussed above. The optimal bids – b(v)– that are part of the Bayesian Nash equilibrium strategies may be plotted for illustration purposes after estimating the pseudo-values based on the observed bids (Fig. 3). The observations for the group “Others” noticeably lie, on average, slightly above observations of the “Petrobras” group, showing that the information rent of Petrobras exceeds the information rent obtained by Others. 15 Eqs. (4) and (5) must be adjusted following Campo et al. (2003) to calculate the pseudo-values, because the estimation was performed using the logarithm of bids. This is why we use log(1 + bi) instead of log(bi). For high values of bi, log(bi) ≈ log(1 + bi).
E.U.R. Brasil, F.A.S. Postali / Energy Economics 46 (2014) 93–101
99
Cumulative distribution function
Distribution function
1
0.25 Petrobras Others
Petrobras Others
0.9 0.8
0.2
0.15
Probability
Probability
0.7
0.1
0.6 0.5 0.4 0.3 0.2
0.05
0.1 0
0
2
4
6
8
10
0
12
0
2
4
6
8
10
12
log(1+b)
log(1+b)
Cumulative distribution function
Distribution function
1
0.2 Petrobras Others
0.18
Petrobras Others
0.9 0.8
0.16
0.7
Probability
Probability
0.14 0.12 0.1 0.08
0.6 0.5 0.4 0.3
0.06
0.2
0.04
0.1
0.02
2
4
6
8
10
12
0
14
2
4
6
log v
8
10
12
14
log v Fig. 2. Auctions with n = 3 bidders.
n=2 Optimal bids
n=3 Optimal bids
14
12 Petrobras Others
12
Petrobras Others 10
10
8
ln b(v)
ln b(v)
8
6
6 4 4 2
2
0 2
4
6
8
10
12
14
16
0
2
4
ln v
6
8
ln v Fig. 3. Optimum bids, as a function of estimated pseudo-values.
10
12
14
100
E.U.R. Brasil, F.A.S. Postali / Energy Economics 46 (2014) 93–101
Table 6 Estimated information rents and additional statistics. Bidder
n
Mean (%)
Standard deviation (%)
Coefficient of variation
Min (%)
Max (%)
Petrobras Petrobras Other Other
2 3 2 3
62.96 44.04 33.80 15.60
37.82 37.04 39.98 28.70
0.60 0.84 1.29 1.83
0.00 0.00 0.00 0.00
88.95 79.72 99.83 79.23
Source: Prepared by the authors.
Thus, Others offer larger portions of their private values, especially in three-bidder auctions. Table 6 exhibits the information rents calculated based on the procedure described above. Petrobras, as an informed competitor, is able to earn higher information rents than the other competitors. Furthermore, as the number of competitors grows, the information rent earned by Petrobras drops, which increases the surplus extracted by the government. In two-bidder auctions won by Petrobras, the estimated information rent is approximately 63%; i.e., 37% of the true willingness to pay is offered to the government, on average. The dispersion of information rent is higher for Others insofar as the coefficient of variation (i.e., the ratio standard deviation to the average) is greater for them and increases with the number of bidders. Such evidence is consistent with the conclusions of Reece (1978, 1979), according to which the industry tend to earn greater proportions of the surplus when uncertainty is high and competition is low. Therefore, policy recommendations must act to remove barriers to the entry of bidders in auctions and to provide the maximum possible information on the areas auctioned. Currently, consortia enrolled in auctions acquire a database from the regulatory agency regarding the geological characteristics of the areas to be auctioned. The improvement of access to such information – or even its wider dissemination – may help to reduce uncertainty and thereby increase the governmental share of revenues generated by the activity. Another method of reducing uncertainty is to improve regulatory stability. Brazil has changed certain rules governing the relationship between government and the oil industry based on the pre-salt discoveries. The profit sharing contracts that have been adopted for the new areas are more suitable for the division of risks between the government and investors, which might favor revenue gains by society.
and based on a database designed with information from all Bidding Rounds held in Brazil until 2008. The findings suggest strong asymmetries between Petrobras and the other bidders in these auctions in light of the high information rents obtained by the state-owned enterprise compared to those earned by other winning bidders. Those rents ranged from 63% (when Petrobras was the winner of two-bidder auctions) to 15% (when other companies won three-bidder auctions). Furthermore, as expected, information rents decrease as the number of bidders increases, which suggests implementing policies to facilitate the entry of investors into these auctions. The regulatory apparatus improvement that was triggered by the new exploratory frontier of pre-salt discoveries should be guided by the dissemination of information regarding auctioned areas and the design of stable, long-term rules to reduce investor risk. Furthermore, the investment in exploration efforts – including funds invested in the geological mapping of Brazilian sedimentary basins – that aim to reduce uncertainty may help raise the governmental share of the resource rent. Another measure that may be studied to increase competition is to change the criteria for the accreditation of potential bidders (minimum net worth required and previous experience, among others) with the aim of increasing the number of interested parties. To our knowledge, this is the first exploratory study on the results of oil and gas auctions in Brazil. A simplifying assumption was adopted herein – independent private value auction with a null reserve price – that aimed to maximize the number of database observations, but several extensions of this study are possible. The natural path to be followed is to estimate a structural model that simultaneously acknowledges asymmetry and affiliated private value, as developed and applied by Campo et al. (2003). The dataset available to date (1999–2008) is still insufficient for implementing this model in Brazil, but this type of estimation may be performed in the near future because the government is planning new auctions. Acknowledgment The authors are grateful to the anonymous referees for the comments and suggestions. References
5. Concluding remarks and policy implications From 1999 to 2008, the Brazilian government has held ten bidding rounds to award oil and gas exploration rights to private investors. Approximately 800 blocks were auctioned and 1,282 bids received from the more than 100 companies that were involved in the auctions. Currently, dozens of private companies, in addition to Petrobras, have exploration rights for hydrocarbon resources in Brazilian territory, both inland and offshore. Those auctions generated a considerable body of data that enabled this study to test whether the government has been successful in collecting revenues from the oil industry. This study aimed to estimate the information rents earned by the winning bidders in such auctions and to evaluate the success of the government and private companies involved in the process. From the government's standpoint, the study aimed to quantify whether the public power is appropriating resource rents, as stipulated by precepts in the Brazilian Constitution. Following pre-salt discoveries in 2007, Brazil has drawn the attention of international investors in the energy sector and our study may offer insights for the design of more effective bidding strategies. Moreover, the pre-salt discoveries have significant implications in terms of risk sharing and block quality, which may require more complex auctions than the first-price sealed-bid auctions conducted thus far. We estimated the distribution of private pseudo-values nonparametrically, under the assumptions of asymmetry and independence
Bajari, P., Hortaçsu, A., 2005. Are structural estimates of auction models reasonable? Evidence from experimental data. J. Polit. Econ. 113 (4), 703–741. Campo, S., Perrigne, I., Vuong, Q., 2003. Asymmetry in first-price auctions with affiliated private values. J. Appl. Econ. 18, 179–207. Cramton, P., 2007. How best to auction oil rights. In: Humphreys, M., Sachs, J.D., Stiglitz, J.E. (Eds.), Escaping the Resource Curse. Columbia University Press. Flambard, V., Perrigne, I., 2006. Asymmetry in procurement auctions: evidence from snow removal contracts. Econ. J. 116, 1014–1036. Guerre, E., Perrigne, I., Vuong, Q., 2000. Optimal nonparametric estimation of first-price auctions. Econometrica 68, 525–574. Hendricks, K., Porter, R.H., Wilson, C.A., 1994. Auctions for oil and gas leases with an informed bidder and a random reservation price. Econometrica 62 (n°.6), 1415–1444. Hendricks, K., Pinkse, J., Porter, R.H., 2003. Empirical implications of equilibrium bidding in first-price, symmetric, common value auctions. Rev. Econ. Stud. 70 (1), 115–145. Hoffman, E., Marsden, J.R., Saidi, R., 1991. Are joint bidding and competitive common value auction markets compatible? Some evidence from offshore oil auctions. J. Environ. Econ. Manag. 20, 99–112. Iledare, O., Pulsipher, A., Olatubi, W., Mesyanzhinov, D., 2004. An empirical analysis of the determinants of high bonus bids for petroleum leases in the US Outer Continental Shelf (OCS). Energy Econ. 26, 239–259. Kagel, J., Levin, D., 1986. The winner's curse and public information in common value auctions. Am. Econ. Rev. 76, 894–920. Krishna, V., 2002. Auction Theory. Academic Press, Pennsylvania. Laffont, J.J., Vuong, Q., 1996. Structural analysis of auction data. Am. Econ. Rev. 86, 414–420. Li, Q., Racine, J.S., 2007. Nonparametric Econometrics: Theory and Practice. Princeton University Press, Princeton, New Jersey. Li, T., Perrigne, I., Vuong, Q., 2000. Conditionally independent private information in OCS wildcat auctions. Journal of Econometrics 98. Elsevier, pp. 129–161. Li, T., Perrigne, I., Vuong, Q., 2002. Structural estimation of the affiliated private value auction model. RAND J. Econ. 33 (2), 171–193. Maskin, E., Riley, J., 2000. Asymmetric auctions. Rev. Econ. Stud. 67 (3), 413–438. McAfee, R.P., McMillan, J., 1987. Auctions and bidding. J. Econ. Lit. XXV, 699–738.
E.U.R. Brasil, F.A.S. Postali / Energy Economics 46 (2014) 93–101 Milgrom, P.R., Weber, R.J., 1982. A theory of auctions and competitive bidding. Econometrica 50, 1089–1122. Myerson, R., 1981. Optimal auction design. Math. Oper. Res. 6, 58–73. Ortega-Reichert, A., 1967. Models of competitive bidding under uncertainty. Ph.D. thesis Stanford University. Paarsch, H.J., Hong, H., 2006. An Introduction to the Structural Econometrics of Auction Data. The MIT Press, Cambridge, Massachusetts. Perrigne, I., Vuong, Q., 1999. Structural econometrics of first-price auctions: a survey of methods. Can. J. Agric. Econ. 47, 203–223. Porter, R.H., 1995. The role of information in U.S. Oil and Gas lease auction. Econometrica 63, 1–27.
101
Reece, D., 1978. Competitive bidding for offshore petroleum leases. Bell J. Econ. 9 (2), 369–384. Reece, D., 1979. An analysis of alternative bidding systems for leasing offshore oil. Bell J. Econ. 10 (2), 659–669. Silverman, B., 1986. Density Estimation for Statistics and Data Analysis. Chapman and Hall, London. Thaler, R.H., Tucker, S.M., 1995. The Winner's Curse: Paradoxes and Anomalies of Economic Life. J. Consum. Aff. 29 (1), 295–296. Wilson, R.B., 1969. Competitive bidding with disparate information. Manag. Sci. 446–448.
Contents lists available at ScienceDirect
Energy Economics journal homepage: www.elsevier.com/locate/eneco
Informational rents in oil and gas concession auctions in Brazil Eric Universo Rodrigues Brasil a, Fernando Antonio Slaibe Postali b,1,⁎ a b
Tendencias Consulting and University of São Paulo, Brazil Department of Economics, University of São Paulo, Brazil
a r t i c l e
i n f o
Article history: Received 22 July 2011 Received in revised form 28 February 2013 Accepted 4 September 2014 Available online 16 September 2014 Keywords: C14 D44 D82 L71 Q35 Q38
a b s t r a c t This article aims to estimate the informational rents earned by winning bidders in oil and gas exploration and production auctions in Brazil. We estimate the distributions of bids and bidders' private valuations using a nonparametric structural model and assuming independence and asymmetry between participants. Petrobras, the Brazilian state-owned petroleum giant and former oil sector monopolist, was considered a competitor that was distinct from other competitors. Thus, we investigate a database based on information from all auctions held between 1999 and 2008. The results suggest that Petrobras earned significantly higher information rents than other competitors. Such rents ranged from 15% to 63%, depending on the number and type of competitors. © 2014 Elsevier B.V. All rights reserved.
Keywords: Oil and gas industry Auctions Regulation Nonparametric estimation
1. Introduction The modern regulatory framework for the Brazilian oil industry was instituted with the passing of the Oil Act (Lei do Petróleo, 9.478/97) in 1997. The new law regulated the Constitutional Amendment (Emenda Constitucional, 9/95) that broke the monopoly that Petrobras had enjoyed on oil and gas exploration and production activities. One of the most significant changes was the creation of auctions – known as Bidding Rounds – that controlled the concession rights for exploration in inland and continental shelf areas. The auctions are first-price sealed-bid in which bidders submit concealed bids that are opened simultaneously with one another. The winner is the bid with the highest score on a set of criteria, including the bid offered, the exploration program and the purchase commitment with domestic suppliers. Since 1999, 10 Bidding Rounds have been conducted, and the results have been quite significant in terms of revenue and the entry of private companies into oil and gas activities in Brazil.
⁎ Corresponding author at: Av. Prof. Luciano Gualberto, 908, Cidade Universitária. CEP: 05508-900, São Paulo-SP, Brazil. Tel.: +55 11 30915915; fax: +55 11 30916013. E-mail addresses: [email protected] (E.U.R. Brasil), [email protected] (F.A.S. Postali). 1 Sponsored by CNPq — Brazilian National Council for Research and Development.
http://dx.doi.org/10.1016/j.eneco.2014.09.002 0140-9883/© 2014 Elsevier B.V. All rights reserved.
There are two significant issues related to the new regulatory framework and the role of these auctions in Brazil. The first concerns the involvement of Petrobras, the state-owned Brazilian oil giant. Because it operated as a monopolist for four decades and has amassed data on the potential of Brazilian oil fields, it naturally enjoys an information asymmetry in such auctions. The second issue is related to the efficiency of the auctions in extracting information rent from bidders, i.e., how closely the bid approaches to the private company's actual valuation of the auctioned area. The aim of this paper is to estimate the information rents earned by the winning bidders in ten rounds of bidding through the use of a nonparametric estimate of the distributions of auction bidders' valuations. Thus, it is possible to analyze how the government collects the bidders' surplus and how successful bidders are in winning the concession with the lowest possible value. Specifically, we are interested in assessing whether there are significant differences in rents earned by Petrobras compared to the other bidders. The results suggest that asymmetries that favor Petrobras generate higher information rents for the company and that, as expected, auctions with fewer participants generate a lower surplus extraction for the government. Because the Brazilian government plans resumed the auctions in 2013, we aim to contribute to the discussion about improving the concession policies of exploration areas following the pre-salt drilling program.
94
E.U.R. Brasil, F.A.S. Postali / Energy Economics 46 (2014) 93–101
Auctions are selling mechanisms that are often characterized by asymmetric information (McAfee and McMillan, 1987). Auctions are attractive to sellers because it may be possible to extract the whole consumer surplus, even though the auctioneer does not know the bidders' willingness to pay. Auctions would be meaningless as selling mechanisms if private valuations were public – if there were perfect information – because the bidder would merely set a price equal to the bidder's valuation of the object sold. Auctions are also useful when the value of the auctioned good is unknown for both sellers and bidders, as in the case of an oil and gas concession area (McAfee and McMillan, 1987). The relationship between the auctioned object and the auction's information structure has generated two main models (Laffont and Vuong, 1996). The first one is the private value auction, in which bidders perform their own valuation of the good, independently or not from the valuations of other bidders. The second model is known as the common value auction (also known as the mineral rights model), in which bidders receive a common signal on the value of the auctioned object that is unknown ex ante. Auctions are the most common form of allocating oil and gas exploration rights worldwide. The choice of the underlying model to analyze them is a complex task and can be controversial. The pure common value model was developed driven by the concession of oil and gas exploration areas (to such an extent that it is also called the mineral concession model) and includes Wilson (1969) and Ortega-Reichert (1967) among its pioneers. An important characteristic of mineral auctions is that companies attempt to estimate a common value for the auctioned goods. Until the mid-1990s, the common value model was dominant in hydrocarbon concessions because of the specific characteristics of the sector, which involve granting the rights to explore and produce an unknown amount of resources at uncertain costs. Technological constraints that prevailed in the oil industry until the 1950s contributed to the permanence of the mineral concession model. Under this paradigm, participating bidders receive a signal on the value of the auctioned object and thereby set their bids. Reece (1978) shows that increased uncertainty and a reduced number of bidders contribute to raising the rents of the winning bidder at the government's expense. Reece (1979) also simulated the rent sharing between the government and bidders under different contractual arrangements and concluded that expected governmental revenue from auctions is higher under a profit-sharing plan and lower under a single bonus plan. Under this structure, imperfect information may result in a ‘winner's curse’ (Kagel and Levin, 1986); i.e., the winning bidder's tendency is to overestimate the value of the object, assuming that the average estimate is non-biased. Thus, the winning bidder would have paid a value above the actual value of the object. In this context, there are several contributions to the literature on oil and gas auctions, including articles concerning the performance of bids when there are information asymmetries between bidders. Hendricks et al. (1994) examined how such information asymmetries affect the strategic behavior of bidders under the common value model. The analysis was performed using a bidder with more private information than all the other bidders (who only have access to public information). The authors conclude that the equilibrium behavior in this information structure often drives bidders with little information to offer less than informed bidders. Similarly, Porter (1995) builds a stylized model with asymmetric competitors (informed versus non-informed bidders). A significant result of this model is that informed bidders are more likely to win the auction and to show positive information rent when they win, whereas noninformed competitors report a null payoff. Thaler and Tucker (1995) report strong evidence of a winner's curse in auctions in the Gulf of Mexico between 1954 and 1969, based on a quantitative survey of unproductive and barely profitable wells. Although it initially focused on common value, the literature on oil auctions began to recognize the existence of a private component in
the bid launching process because of bidder heterogeneity (Cramton, 2007). Indeed, Li et al. (2000) showed that the private component is relevant to oil auctions in the US OCS data while estimating a conditionally independent private value model. This led to a reformulation of the analysis model regarding this form of bidding to include the private component of bids. Moreover, recent technological developments have enabled companies to gain cost advantages that contribute to affect their assessment of the good, which created a rationale for the use of private value that is based on structural models. Accordingly, the number of empirical analyses has substantially increased in recent years. The premise of such models is to estimate the unobserved private values (termed pseudo-values) under the hypothesis that optimal strategies are defined according to an underlying game.2 Currently, the literature features significant contributions to the structural estimation of auction models using the non-parametric approach, including Guerre et al. (2000), Li et al. (2002), Hendricks et al. (2003), Campo et al. (2003), Bajari and Hortaçsu (2005), Flambard and Perrigne (2006), among others. In private value auctions, information rent is defined as the winning bidder's surplus, i.e., the difference between the bidder's private value and the actual bid launched. Information rents are so named because the bidder's valuation of the auctioned object is private information that precludes the auctioneer from fully discriminating bidders (Krishna, 2002, chapter 5), and the winning bidder benefits from a rent for the partial disclosure of his private information. The lower the winning bid is relative to the private valuation, the higher the bidder's information rent for the auctioned object. The presence of information rents is inherent in the private value mode. Depending on the type of auction, the role of the equilibrium bid – the optimal bid as a function of private value – produces a lower offer than the actual private value. This occurs, for example, in firstprice sealed-bid auctions (Paarsch and Hong, 2006, chapter 2); the auctioneer will be more efficient the closer the winning bid is to the maximum private value among bidders involved in the auction. The estimation techniques of these private values became progressively more complex over time. Following Li et al. (2000), Campo et al. (2003) estimated information rents of US OCS auctions in the 1970s using an affiliated private value model with asymmetric bidders. The results show information rents of approximately 65%, which were higher for consortium bids. Using simulated private value bids, Li et al. (2002) estimated information rents of approximately 38% for twobidder auctions and 29% for three-bidder auctions. Flambard and Perrigne (2006) estimated information rents and evaluated the inefficiency of auctions in hiring snow removal services in Canada under the hypothesis of independent private value and asymmetry between bidders, which was defined by costs that depended on the location of the company. The asymmetry of costs is conclusively relevant and accounts for inefficient allocations. The current paper has an empirical nature and aims to apply the methodology of the articles described above to the evaluation of information rents in oil and gas bidding auctions in Brazil. If companies are appropriating large private surpluses, there may be a low level of competition in such auctions, which might indicate that policy measures should be undertaken to correct this. This article is structured in 5 sections, including this introduction. Section 2 describes and analyzes the database, including a discussion about the main determinants of bids, and Section 3 discusses the theoretical model underlying the estimation of information rent and empirical estimation procedures in general terms. Section 4 describes 2 As highlighted by Perrigne and Vuong (1999), there are two approaches for estimating and treating data. The first is parametric, which estimates the parameters through maximum likelihood under a known distribution of unobserved values. The second is nonparametric, which is based on indirect procedures linking private value to observed bids. The disadvantage of the nonparametric approach is that it requires a large database of bids, whereas the parametric one requires less data but utilizes the computation of a function defining the optimal strategy, which may not be feasible.
E.U.R. Brasil, F.A.S. Postali / Energy Economics 46 (2014) 93–101
95
Table 1 Summary of ten annual oil and gas bidding rounds in Brazil. Round
Blocks sold Total bids Auctions with Petrobras involvement (%) Petrobras bids Petrobras winning bids Petrobras success rate (%) Consortium bids Consortium winning bids Success rate of consortium bids (%)
1
2
3
4
5
6
7
8
9
10
Total
12 21 58 7 5 71 11 6 55
21 46 48 10 8 80 19 10 53
34 57 59 20 15 75 19 13 68
21 33 43 9 8 89 5 5 100
101 106 88 89 88 99 3 3 100
154 188 73 113 107 95 79 64 81
251 379 43 109 96 88 131 87 66
38 84 58 22 21 95 24 15 63
117 276 49 57 27 47 108 46 43
54 92 52 28 27 96 29 22 76
803 1282 58 464 402 87 428 271 63
Source: Prepared by the authors. The first round occurred in 1999 and the last occurred in 2008.
Table 2 Observations per number of bidders. n
Auctions
Bids
Petrobras
Others
1 2 3 4 5 6 7 8
528 164 60 26 14 7 2 2 803
528 328 180 104 70 42 14 16 1282
283 97 45 17 12 6 2 2 487
245 229 135 87 58 36 12 14 795
Source: Prepared by the authors.
the results and Section 5 offers concluding remarks, which include certain policy suggestions that might increase competition in these auctions and (therefore) the governmental appropriation of private rents. 2. Data The database was obtained from the National Petroleum Agency (Agência Nacional do Petróleo—ANP) and is composed of information on all auctions held over ten annual Bidding Rounds3 between 1999 and 2008. The data provide information on the auctioned block area, its basin and sector, the round in which it was sold, its location (land or offshore), the number of bidders in each auction, the bidding companies and/or bidding consortia, the operating companies and the bid value offered. Companies and/or pre-qualified consortia have 1 min to place their bids in sealed envelopes. All envelopes are opened simultaneously. Until 2002, the bid value represented 85% of the score defining the winning bidder; the remaining 15% of the score was related to a commitment to using domestic suppliers. From 2003 onward, there was an important shift because the exploration program offered by the investor began to be incorporated into the score, and the percentages assigned to each criterion (bid, local content and exploration program) began to vary from year to year. In approximately 8% of the auctions, the winning bidder failed to offer the highest bid.4 The database records 109 companies that took part in some way in ANP auctions over these ten years, in addition to Petrobras. Although some of these are large international corporations, such as Amerada Hess, Shell, Repsol YPF, Chevron and BP, none have the same level of knowledge that Petrobras has with respect to the quality of Brazilian deposits, which justifies the classification of bidders into two main 3 The term “Bidding Round” refers to a set of auctions held once a year. Each round is held over one or two days in which multiple blocks are sequentially auctioned. 4 The Brazilian auctions suffer from intrinsic inefficiencies arising from the possibility of selling the tract to a candidate that has not offered the highest bid. Therefore, we chose to evaluate only information rent because the efficiency (that is, the block allocation to the bidder with the highest willingness to pay) does not appear to be a priority for the Brazilian government in these auctions, particularly after 2003.
groups – Petrobras and Others – according to the assumption described below.5 Table 1 shows a quantitative summary of the involvement and bids of Petrobras in Bidding Rounds: 803 blocks were auctioned that received a total of 1282 bids, which represents an average of 1.53 bids per block sold; 311 blocks were located offshore and received 396 bids; 192 were located on land and received 886 bids. The success rate of Petrobras – that is, the number of winning bids in relation to total bids – was 87% (in 2003, it was 99%); Petrobras participated in 53% of the auctions held. Conversely, consortium bids show a 63% success rate. Table 2 shows the distribution of auctions based on the number of bidders. More than half of all blocks sold received only one bid (528 out of 803), which shows the low-level of competitiveness in the auctions. The reasons for this phenomenon may be related to the high degree of geological uncertainty at the auctioned areas, according to Reece (1978), or because of the perception of a regulatory risk that is reflected in the high standard deviation of expected returns (Reece, 1979); this scenario is consistent with high information rents for winning bidders. Table 3 shows a statistical summary of bids launched over the ten rounds; it indicates that the bids of Petrobras (per unit area) were, on average, much higher than those of other bidders.6 Another important feature is that bids made by Petrobras against non-informed bidders (Petrobras versus Others) are, on average, higher than the others' bids among themselves (Others versus Others); in addition, non-informed bidders tend to offer more aggressive bids when competing with Petrobras (Others versus Petrobras) than when competing against their peers (Others versus Others). This suggests that each bidder's strategy depends on the bidder's opponent, which may correlate with the presence of asymmetric information between Petrobras and the other bidders, as speculated by Campo et al. (2003). Finally, the last section in the table indicates that consortium bids are slightly higher than bids launched by single companies. This difference is somewhat more significant when the consortia contains only non-informed bidders (excluding Petrobras).7 Table 4 shows the results from the estimation of the probability of winning the auction given certain observable characteristics and using a probit model; this estimation aims to report a preliminary study on the determinants of bids. The dependent variable is a dummy assuming
5 There have also been large national companies that had no prior experience in oil and gas exploration but decided to enter into auctions after the Petrobras monopoly was ended, such as Vale do Rio Doce, Queiróz Galvão and Odebrecht. 6 This is stylized evidence from the findings of Porter (1995), in which the informed bidder has a higher probability of launching bids and a higher rate of success. 7 As Petrobras has a type-A operator status, it has the option of launching individual or consortium bids (with other companies). Although it may theoretically compete with itself, no records are found in the database of situations in which Petrobras launched an individual bids against one of its consortia. It is unlikely that this would occur in practice because Petrobras is almost always the operator company – the company responsible for conducting the operations – when participating in consortia, and it therefore has the prerogative to decide to bid.
96
E.U.R. Brasil, F.A.S. Postali / Energy Economics 46 (2014) 93–101
Table 3 Statistical summary of bids. Variable
Obs.
Average (R$/km2)
Standard deviation
Coefficient of variation
Minimum
Maximum
All bids Petrobras bids Others bids Petrobras vs. Others Others vs. Petrobras Others vs. Others Consortium bids (including Petrobras) Individual bids (including Petrobras) Consortium bids (excluding Petrobras) Loose bids (excluding Petrobras)
754 181 563 174 335 225 266 488 155 408
11,805.10 18,269.59 7416.29 15,988.71 7959.86 6082.61 12,773.92 11,277.01 9810.50 6506.72
30,720.35 30,408.55 17,597.70 24,043.56 18,124.16 15,783.04 27,106.81 32,536.00 20,413.60 16,337.91
2.60 1.66 2.37 1.50 2.28 2.59 2.12 2.89 2.08 2.51
1.58 2.99 1.58 2.99 1.58 1.64 2.99 1.58 3.70 1.58
440,494.00 208,789.90 206,146.00 118,089.00 206,146.00 118,788.70 208,789.90 440,494.00 118,788.70 206,146.00
Source: Prepared by the authors. Table 4 Determinants of the probability of winning the auction.
Table 5 Determinants of observed bids. Dependent variable: log(bid/km2)
Dependent variable: winning bid = 1 (1)
(2)
−1.046⁎⁎⁎ (0.0816) 0.181 (0.158) −1.172⁎⁎ (0.457) 0.619⁎⁎⁎ (0.0605) 0.00016 (0.00010) 3.664⁎⁎ (1.548) −1.564 (1.688) 2.852⁎⁎⁎
−1.087⁎⁎⁎ (0.0950) 0.247 (0.189) 0.127 (1.484) 0.459⁎⁎⁎ (0.0864) 0.00043⁎⁎
Year dummy Basin dummy Enterprise dummy Constant
(0.619) (0.756) – – Yes Yes Yes −5.751⁎⁎⁎
(0.610) (0.838) 0.00056⁎⁎⁎ (0.00011) Yes Yes Yes −11.46⁎
Observations Pseudo-R2
(1.429) 1339 0.506
(6.612) 1107 0.544
Number of bidders Consortium dummy Offshore block dummy Log(Bid/km2) Area (km2) Exploration (%)(a) Development (%)(b) Petrobras dummy
Proposed investment program(c)
(0.0002) 4.051 (3.628) 5.095 (9.587) 3.032⁎⁎⁎
Robust standard errors in parentheses. ⁎⁎⁎ 1%, ⁎⁎ 5%, and ⁎ 10% significance levels. (a) (b) (c)
Commitment to local suppliers in the exploratory phase. Commitment to local suppliers in the development phase. Introduced after 2003.
the value 1 when the bidder wins the auction.8 Specifications (1) and (2) differ due to the inclusion of the exploration program offered, which was introduced in 2003. The number of competitors reduces the bidder's probability of winning an auction, whereas consortium formation does not affect this probability.9 As might be expected, the bid launched (log(bid/km2)) and the minimum exploration program positively affect this probability, whereas the commitment to domestic suppliers in the exploration phase positively affects this probability only with specification (1)—when the exploration program offered is not considered. In addition, the offshore blocks are less likely to be auctioned in this case, which reflects the lower levels of competition for these areas. The signal obtained for the dummy variable identifying
8 It should be noted that this is not necessarily the highest bid but, rather, is the bid from the bidder who actually won the auction under the three scoring criteria (bid, purchase commitment with local suppliers and exploration program, if applicable). 9 This result is consistent with Iledare et al. (2004), who conclude that consortium bids are common strategies in Outer Continental Shelf (OCS) auctions in the Gulf of Mexico (USA) and do not significantly affect the bids made. Hoffman et al. (1991) found the same evidence using data from 1970.
Number of bidders Consortium dummy Offshore block dummy Exploration (%) Development (%) Petrobras dummy Proposed investment program Year dummy Basin dummy Enterprise dummy Constant Observations R-square
(1)
(2)
0.108⁎⁎⁎ (0.0239) −0.214⁎⁎
0.0389⁎⁎ (0.0183) −0.120 (0.0821) 2.748⁎⁎⁎ (0.605) 0.517 (2.066) 13.16⁎⁎⁎
(0.0980) 0.802⁎⁎ (0.311) 2.557⁎⁎ (1.115) −1.721 (1.178) −0.0604 (0.339) – – Yes Yes Yes 4.445⁎⁎⁎ (0.710) 1468 0.858
(4.640) −0.284⁎⁎ (0.126) 0.00068⁎⁎⁎ (4.11e−05) Yes Yes Yes −4.141 (2.700) 1226 0.932
Robust standard errors in parentheses. ⁎⁎⁎ 1%, ⁎⁎ 5%, and ⁎ 10% significant.
Petrobras as participating bidder – which is positive and significant in both specifications – indicates asymmetry, such that Petrobras has an above-average probability to win a block.10 Table 5 shows the results of the regression of the bid value per unit area (in log). As the sign of the coefficient regarding the number of bidders is positive and significant, competition drives bidders to launch higher bids,11 and higher bids are launched for offshore blocks. Interestingly, consortia launch lower bids when the exploration program is not considered (specification 1); in this case, the national commitment to the exploration phase also positively affects the bid. The effect of the investment program offer, though small, is positive and statistically significant at 1%. Petrobras bids, controlled for all other characteristics, are not significantly different from the other companies' bids in specification (1); this is notable because the exploration program (specification 2) indicates that the Brazilian state company launches bids below the average bids of other companies. Because Petrobras is more likely to win the auction (Table 4), launching bids that are not higher than the mean may be further evidence of asymmetry, which indicates that Petrobras has more accurate information on the value of the blocks. But this result merits further investigation, which are beyond the goal of this study.
10
These results are consistent with predictions by Porter (1995). In a detailed investigation of the determinants of North-American auction bids (OCS) in the Gulf of Mexico between 1983 and 1999, Iledare et al. (2004) note that competition increased the mean value of bids and that large companies launched bids above the mean. 11
E.U.R. Brasil, F.A.S. Postali / Energy Economics 46 (2014) 93–101
3. Information rents and asymmetries in Brazilian auctions Assuming the independent private value model, the aim of this study is to estimate the information rent (IR) earned by the winning bidders of oil and gas exploration auctions in Brazil, calculated as follows: IRi ¼
vi −bi vi
ð1Þ
where vi is the bidder's private value, and bi is his/her actual bid. Each bid is known, but the bidders' respective private values are not. We use the method proposed by Guerre et al. (2000) to estimate them, complemented by Flambard and Perrigne (2006) for asymmetric auctions. This is a non-parametric two-stage procedure, computationally feasible and applied to models of first-price sealed-bid auctions under independent private value. The method has the advantage of avoiding a functional form for the Bayesian Nash equilibrium strategy (that is, bidding as a function of private value bi(vi)). We assume that each bidder is risk neutral with private value vi and ignores the private value of others. Furthermore, all private values independently originate from the same distribution F(.) with density f(.). As shown in Guerre et al. (2000)12 and further explained by Flambard and Perrigne (2006),13 the distribution function of private values can be identified from the distribution of bids, G(.) for n ≥ 2, where n is the number of bidders. We assume that there is no minimum bid, unlike Flambard and Perrigne (2006). This is a simplifying assumption, but it fits reasonably well in the Brazilian case because the reserve bids fixed by the government were low and varied very little from area to area; it is virtually impossible that they would have caused the exclusion of any potential candidate. 14 Because the bidder of type i launches a bid bi, its probability of winning the auction is G(bi)n − 1, given the assumption of independent private value. Therefore, the bidder's expected payoff from launching a bid bi is given by: n−1
Πi ¼ ðvi −bi ÞGðbi Þ
ð2Þ
Maximizing the above expression regarding bi results in: 1 Gðbi Þ vi ¼ bi þ n−1 g ðbi Þ
12
bidder is a function of the bidder's own private signal and a common unknown value arising from a signal received by all the participants. However, the estimation of this model requires the selection of auctions with exactly two participants, which would entail excluding a substantial portion of our database. Thus, we opted to utilize the independent private value model to maximize the use of data. In addition to independence, we also assumed additive values, i.e., the value of each block is independent of the value of previously auctioned blocks. A consequence of this hypothesis is that the revenue earned by a set of blocks auctioned in a single package is equivalent to the sum of revenues of blocks auctioned separately. As Cramton (2007) argues, additive values are good approximations for oil and gas auctions because the amount of hydrocarbon resources in a set of blocks is the sum of the resources of each block at a given price. As a result of the state monopoly in oil exploration that Petrobras enjoyed between 1953 and 1997 in Brazil, it acquired extensive knowledge about the geological characteristics of Brazilian sedimentary basins. Now, it is recognized as one of the world's most productive energy companies. Therefore, oil exploration auctions in Brazil, according to the data analysis from the previous section, suggest a strong asymmetry favoring Petrobras. As emphasized by Maskin and Riley (2000), asymmetric auctions are extremely complex and have distinct characteristics, including the possibility of inefficient allocation (because the object may be sold to a bidder who does not have the highest private value among bidders) and non-validity of the Revenue Equivalence Theorem (Myerson, 1981). In symmetric auctions, bidders use the same equilibrium strategy as a function of their types—that is, the function bi(vi) is the same for all bidders; in asymmetric auctions, bidders may have the same information on the object but might differ with respect to opportunity costs and budget constraints, which may lead them to yield different valuations (Maskin and Riley, 2000). Thus, in the Brazilian case, we assume independent private value with asymmetric competitors – Petrobras versus Others – to balance the simplifying assumptions and model the asymmetry. Type i = 1 represents Petrobras and type i = 2 represents Others. In this case, Eq. (3) may be applied to the two groups of bidders. Given that G1(.) and G2(.) are the probability distributions of the bids of Petrobras and Others, respectively, the private values may be estimated by: v1 ¼ b1 þ
1 G2 ðb1 Þ n−1 g 2 ðb1 Þ
ð4Þ
v2 ¼ b2 þ
G1 ðb2 ÞG2 ðb2 Þ g 1 ðb2 ÞG2 ðb2 Þ þ ðn−2Þg2 ðb2 ÞG1 ðb2 Þ
ð5Þ
ð3Þ
The model adopted in the present study assumes the independence of the bidders' private values. As described above, the first studies on oil auctions focused on the common value model. However, as the methodology developed, the literature on private value gained momentum. In the case of Brazil, each company's peculiarities, including different technological capabilities in oil exploration, suggest little likelihood that the auctioned block has the same value ex post, irrespective of the winner; this is true for Petrobras in particular because it has developed specific technologies over time as the sole operator in the Brazilian oil sector. Conversely, it is reasonable to assume that auctions for oil exploration areas lie between the extremes of private value and common value, in what Milgrom and Weber (1982) termed an “Affiliated Private Value Auction”. In this class of models, the utility of each Theorem 1, p. 553. 13 Proposition 1, p. 1023. 14 In fact, until 2002, the areas were ranked into three categories, each with a uniform reserve price. In subsequent years, reserve bids were individualized per area, but their values were close to those set in the early rounds. The surprisingly high premiums – some reaching 50,000% – suggest that fixed reserve bids were more symbolic than effective; it is virtually impossible that they might have caused the exclusion of a competitor previously eligible to participate in the auction. Thus, for the purpose of simplification, we chose to assume that the auctioneer's reserve bid is irrelevant for the decision to enter the auction because this also represents the reality.
97
where n is the total number of bidders, including Petrobras. After dividing the observations into two groups according to the pattern of asymmetry described above (Petrobras versus Others), the results were estimated based on a three-step procedure, following Guerre et al. (2000) and Campo et al. (2003). The first step is to non-parametrically estimate the distribution of observed bids using a Kernel estimator. According to Li and Racine (2007, chapter 1), the type of Kernel does not have a significant effect in practical terms; therefore, we use a Triweight function because it satisfies all hypotheses required, as argued by Guerre et al. (2000). Thus, the probability density function at a point x is given by: f ðxÞ ¼
m 1X x−xð jÞ K m j¼1 h
ð6Þ
where m is the number of observations (that is, the number of auctions corresponding to each type), K(.) is the Kernel function, and h is the bandwidth such that ∫ K(t)dt = 1. The Triweight Kernel is given by: K ðuÞ ¼
35 2 3 Iðjuj≤1Þ 1−u 32
ð7Þ
98
E.U.R. Brasil, F.A.S. Postali / Energy Economics 46 (2014) 93–101
Distribution function
Cumulative distribution function
0.2
1 Petrobras Others
0.16
0.8
0.14
0.7
0.12
0.6
0.1 0.08
0.5 0.4
0.06
0.3
0.04
0.2
0.02
0.1
0
0
2
4
6
8
10
12
Petrobras Others
0.9
Probability
Probability
0.18
0
14
0
2
4
6
log(1+b) Distribution function
12
14
Cumulative distribution function Petrobras Others
0.18 0.16
0.8
0.14
0.7
0.12
0.6
0.1 0.08
0.5 0.4
0.06
0.3
0.04
0.2
0.02
0.1
2
4
6
8
10
12
14
Petrobras Others
0.9
Probability
Probability
10
1
0.2
0
8
log(1+b)
16
0
2
4
6
8
10
12
14
16
log v
log v Fig. 1. Auctions with n = 2 bidders.
The bandwidth parameter is a sequence of constants converging toward zero as the number of observations tends to infinity. It defines the neighborhood surrounding a point x in which the Kernel performs the estimation. Low h values decrease the bias, but increase the variance of the nonparametric estimator ^f ð:Þ (Paarsch and Hong, 2006, chapter 4). The bandwidth regulates the smoothness of distribution; high h values mean smoother distributions with low variances but with a higher bias. According to Li and Racine (2007, chapter 1) and Guerre et al. (2000), we use the optimal bandwidth given by: ^ bn h ¼ cσ
−1=5
ð8Þ
^ b is the standard deviation of observed bids, n is the number of where σ bids and c is a constant. This is the bandwidth used in Li et al. (2002) and Campo et al. (2003) and is based on a “rule of thumb” developed by Silverman (1986) that minimizes the integrated mean squared error, which is composed of the sum of two components, i.e., a measure of the bias and a function of variance (Li and Racine, 2007, chapter 1). For a Triweight Kernel, we use c = 2.96. The second step is to estimate the pseudo-values using (4) and (5) and the information rent using (1). In the third and final step, the pseudo-values calculated are used to non-parametrically estimate the
distributions of private values with the procedure applied in the first step.
4. Results Figs. 1 and 2 show the estimated distributions of observed bids and private values (in log15) for 2- and 3-bidder auctions, respectively. The difference in the densities estimated for the groups is notable. In both cases, we observe the stochastic dominance of the distribution of bids and the private values of Petrobras; i.e., the Brazilian state company shows, on average, higher private values than the other bidders, which is further evidence of the asymmetries discussed above. The optimal bids – b(v)– that are part of the Bayesian Nash equilibrium strategies may be plotted for illustration purposes after estimating the pseudo-values based on the observed bids (Fig. 3). The observations for the group “Others” noticeably lie, on average, slightly above observations of the “Petrobras” group, showing that the information rent of Petrobras exceeds the information rent obtained by Others. 15 Eqs. (4) and (5) must be adjusted following Campo et al. (2003) to calculate the pseudo-values, because the estimation was performed using the logarithm of bids. This is why we use log(1 + bi) instead of log(bi). For high values of bi, log(bi) ≈ log(1 + bi).
E.U.R. Brasil, F.A.S. Postali / Energy Economics 46 (2014) 93–101
99
Cumulative distribution function
Distribution function
1
0.25 Petrobras Others
Petrobras Others
0.9 0.8
0.2
0.15
Probability
Probability
0.7
0.1
0.6 0.5 0.4 0.3 0.2
0.05
0.1 0
0
2
4
6
8
10
0
12
0
2
4
6
8
10
12
log(1+b)
log(1+b)
Cumulative distribution function
Distribution function
1
0.2 Petrobras Others
0.18
Petrobras Others
0.9 0.8
0.16
0.7
Probability
Probability
0.14 0.12 0.1 0.08
0.6 0.5 0.4 0.3
0.06
0.2
0.04
0.1
0.02
2
4
6
8
10
12
0
14
2
4
6
log v
8
10
12
14
log v Fig. 2. Auctions with n = 3 bidders.
n=2 Optimal bids
n=3 Optimal bids
14
12 Petrobras Others
12
Petrobras Others 10
10
8
ln b(v)
ln b(v)
8
6
6 4 4 2
2
0 2
4
6
8
10
12
14
16
0
2
4
ln v
6
8
ln v Fig. 3. Optimum bids, as a function of estimated pseudo-values.
10
12
14
100
E.U.R. Brasil, F.A.S. Postali / Energy Economics 46 (2014) 93–101
Table 6 Estimated information rents and additional statistics. Bidder
n
Mean (%)
Standard deviation (%)
Coefficient of variation
Min (%)
Max (%)
Petrobras Petrobras Other Other
2 3 2 3
62.96 44.04 33.80 15.60
37.82 37.04 39.98 28.70
0.60 0.84 1.29 1.83
0.00 0.00 0.00 0.00
88.95 79.72 99.83 79.23
Source: Prepared by the authors.
Thus, Others offer larger portions of their private values, especially in three-bidder auctions. Table 6 exhibits the information rents calculated based on the procedure described above. Petrobras, as an informed competitor, is able to earn higher information rents than the other competitors. Furthermore, as the number of competitors grows, the information rent earned by Petrobras drops, which increases the surplus extracted by the government. In two-bidder auctions won by Petrobras, the estimated information rent is approximately 63%; i.e., 37% of the true willingness to pay is offered to the government, on average. The dispersion of information rent is higher for Others insofar as the coefficient of variation (i.e., the ratio standard deviation to the average) is greater for them and increases with the number of bidders. Such evidence is consistent with the conclusions of Reece (1978, 1979), according to which the industry tend to earn greater proportions of the surplus when uncertainty is high and competition is low. Therefore, policy recommendations must act to remove barriers to the entry of bidders in auctions and to provide the maximum possible information on the areas auctioned. Currently, consortia enrolled in auctions acquire a database from the regulatory agency regarding the geological characteristics of the areas to be auctioned. The improvement of access to such information – or even its wider dissemination – may help to reduce uncertainty and thereby increase the governmental share of revenues generated by the activity. Another method of reducing uncertainty is to improve regulatory stability. Brazil has changed certain rules governing the relationship between government and the oil industry based on the pre-salt discoveries. The profit sharing contracts that have been adopted for the new areas are more suitable for the division of risks between the government and investors, which might favor revenue gains by society.
and based on a database designed with information from all Bidding Rounds held in Brazil until 2008. The findings suggest strong asymmetries between Petrobras and the other bidders in these auctions in light of the high information rents obtained by the state-owned enterprise compared to those earned by other winning bidders. Those rents ranged from 63% (when Petrobras was the winner of two-bidder auctions) to 15% (when other companies won three-bidder auctions). Furthermore, as expected, information rents decrease as the number of bidders increases, which suggests implementing policies to facilitate the entry of investors into these auctions. The regulatory apparatus improvement that was triggered by the new exploratory frontier of pre-salt discoveries should be guided by the dissemination of information regarding auctioned areas and the design of stable, long-term rules to reduce investor risk. Furthermore, the investment in exploration efforts – including funds invested in the geological mapping of Brazilian sedimentary basins – that aim to reduce uncertainty may help raise the governmental share of the resource rent. Another measure that may be studied to increase competition is to change the criteria for the accreditation of potential bidders (minimum net worth required and previous experience, among others) with the aim of increasing the number of interested parties. To our knowledge, this is the first exploratory study on the results of oil and gas auctions in Brazil. A simplifying assumption was adopted herein – independent private value auction with a null reserve price – that aimed to maximize the number of database observations, but several extensions of this study are possible. The natural path to be followed is to estimate a structural model that simultaneously acknowledges asymmetry and affiliated private value, as developed and applied by Campo et al. (2003). The dataset available to date (1999–2008) is still insufficient for implementing this model in Brazil, but this type of estimation may be performed in the near future because the government is planning new auctions. Acknowledgment The authors are grateful to the anonymous referees for the comments and suggestions. References
5. Concluding remarks and policy implications From 1999 to 2008, the Brazilian government has held ten bidding rounds to award oil and gas exploration rights to private investors. Approximately 800 blocks were auctioned and 1,282 bids received from the more than 100 companies that were involved in the auctions. Currently, dozens of private companies, in addition to Petrobras, have exploration rights for hydrocarbon resources in Brazilian territory, both inland and offshore. Those auctions generated a considerable body of data that enabled this study to test whether the government has been successful in collecting revenues from the oil industry. This study aimed to estimate the information rents earned by the winning bidders in such auctions and to evaluate the success of the government and private companies involved in the process. From the government's standpoint, the study aimed to quantify whether the public power is appropriating resource rents, as stipulated by precepts in the Brazilian Constitution. Following pre-salt discoveries in 2007, Brazil has drawn the attention of international investors in the energy sector and our study may offer insights for the design of more effective bidding strategies. Moreover, the pre-salt discoveries have significant implications in terms of risk sharing and block quality, which may require more complex auctions than the first-price sealed-bid auctions conducted thus far. We estimated the distribution of private pseudo-values nonparametrically, under the assumptions of asymmetry and independence
Bajari, P., Hortaçsu, A., 2005. Are structural estimates of auction models reasonable? Evidence from experimental data. J. Polit. Econ. 113 (4), 703–741. Campo, S., Perrigne, I., Vuong, Q., 2003. Asymmetry in first-price auctions with affiliated private values. J. Appl. Econ. 18, 179–207. Cramton, P., 2007. How best to auction oil rights. In: Humphreys, M., Sachs, J.D., Stiglitz, J.E. (Eds.), Escaping the Resource Curse. Columbia University Press. Flambard, V., Perrigne, I., 2006. Asymmetry in procurement auctions: evidence from snow removal contracts. Econ. J. 116, 1014–1036. Guerre, E., Perrigne, I., Vuong, Q., 2000. Optimal nonparametric estimation of first-price auctions. Econometrica 68, 525–574. Hendricks, K., Porter, R.H., Wilson, C.A., 1994. Auctions for oil and gas leases with an informed bidder and a random reservation price. Econometrica 62 (n°.6), 1415–1444. Hendricks, K., Pinkse, J., Porter, R.H., 2003. Empirical implications of equilibrium bidding in first-price, symmetric, common value auctions. Rev. Econ. Stud. 70 (1), 115–145. Hoffman, E., Marsden, J.R., Saidi, R., 1991. Are joint bidding and competitive common value auction markets compatible? Some evidence from offshore oil auctions. J. Environ. Econ. Manag. 20, 99–112. Iledare, O., Pulsipher, A., Olatubi, W., Mesyanzhinov, D., 2004. An empirical analysis of the determinants of high bonus bids for petroleum leases in the US Outer Continental Shelf (OCS). Energy Econ. 26, 239–259. Kagel, J., Levin, D., 1986. The winner's curse and public information in common value auctions. Am. Econ. Rev. 76, 894–920. Krishna, V., 2002. Auction Theory. Academic Press, Pennsylvania. Laffont, J.J., Vuong, Q., 1996. Structural analysis of auction data. Am. Econ. Rev. 86, 414–420. Li, Q., Racine, J.S., 2007. Nonparametric Econometrics: Theory and Practice. Princeton University Press, Princeton, New Jersey. Li, T., Perrigne, I., Vuong, Q., 2000. Conditionally independent private information in OCS wildcat auctions. Journal of Econometrics 98. Elsevier, pp. 129–161. Li, T., Perrigne, I., Vuong, Q., 2002. Structural estimation of the affiliated private value auction model. RAND J. Econ. 33 (2), 171–193. Maskin, E., Riley, J., 2000. Asymmetric auctions. Rev. Econ. Stud. 67 (3), 413–438. McAfee, R.P., McMillan, J., 1987. Auctions and bidding. J. Econ. Lit. XXV, 699–738.
E.U.R. Brasil, F.A.S. Postali / Energy Economics 46 (2014) 93–101 Milgrom, P.R., Weber, R.J., 1982. A theory of auctions and competitive bidding. Econometrica 50, 1089–1122. Myerson, R., 1981. Optimal auction design. Math. Oper. Res. 6, 58–73. Ortega-Reichert, A., 1967. Models of competitive bidding under uncertainty. Ph.D. thesis Stanford University. Paarsch, H.J., Hong, H., 2006. An Introduction to the Structural Econometrics of Auction Data. The MIT Press, Cambridge, Massachusetts. Perrigne, I., Vuong, Q., 1999. Structural econometrics of first-price auctions: a survey of methods. Can. J. Agric. Econ. 47, 203–223. Porter, R.H., 1995. The role of information in U.S. Oil and Gas lease auction. Econometrica 63, 1–27.
101
Reece, D., 1978. Competitive bidding for offshore petroleum leases. Bell J. Econ. 9 (2), 369–384. Reece, D., 1979. An analysis of alternative bidding systems for leasing offshore oil. Bell J. Econ. 10 (2), 659–669. Silverman, B., 1986. Density Estimation for Statistics and Data Analysis. Chapman and Hall, London. Thaler, R.H., Tucker, S.M., 1995. The Winner's Curse: Paradoxes and Anomalies of Economic Life. J. Consum. Aff. 29 (1), 295–296. Wilson, R.B., 1969. Competitive bidding with disparate information. Manag. Sci. 446–448.