WPS4378 Policy ReseaRch WoRking PaPeR 4378 Do Regulation and Institutional Design Matter for Infrastructure Sector Performance? Luis Andres Jose Luis Guasch Stephane Straub The World Bank Latin America and the Caribbean Region Sustainable Development Department October 2007 Policy ReseaRch WoRking PaPeR 4378 Abstract This paper evaluates the impact of economic regulation shows that in three relevant economic aspects--aligning on infrastructure sector outcomes. It tests the impact costs and tariffs; dissuading renegotiations; and of regulation from three different angles: aligning costs improving productivity, quality of service, coverage, with tariffs and firm profitability; reducing opportunistic and tariffs--the structure, institutions, and procedures renegotiation; and measuring the effects on productivity, of regulation matter. Thus, significant efforts should quality of service, coverage, and prices. The analysis continue to be made to improve the structure, quality, uses an extensive data set of about 1,000 infrastructure and institutionality of regulation. Regulation matters for concessions granted in Latin America from the late 1980s protecting both consumers and investors, for aligning to the early 2000s. closely financial returns and the costs of capital, and for The analysis finds that as the theory indicates, capturing higher levels of benefits from the provision of regulation matters. The empirical work here reported infrastructure services by the private sector. This paper--a product of the Sustainable Development Department in the Latin America and the Caribbean Region--is part of a larger effort in the department to understand the determinants for performance in the infrastructure sectors. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at jguasch@worldbank.org and landres@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Do Regulation and Institutional Design Matter for Infrastructure Sector Performance?1 Luis Andres The World Bank Jose Luis Guasch The World Bank and University of California San Diego and Stephane Straub University of Edinburgh 1We are very grateful for the most helpful referees' comments and for those of the participants at the CUTS-CDRF Competition and Regulation Conference in New Delhi, India. Findings, interpretation and conclusions expressed herein do not necessarily reflect the views of the Board of Executive Directors of the World Bank or the governments they represent. Corresponding authors: jguasch@worldbank.org and landres@worldbank.org. 1. Introduction As part of structural reforms in infrastructure industries during the 1990s, more than US$ 750 billion was invested in 2,500 private infrastructure projects in developing economies. Nearly half went to the Latin American region, mainly through the divestiture of public assets in telecommunications and electricity sectors and transport concessions. Six countries ­ Argentina, Brazil, Chile, Colombia, Mexico and Peru ­ absorbed more than 90 percent of all private investments. Overall, the region was the largest beneficiary of the huge flow of private investments for infrastructure worldwide with private investment peaking at around US$ 130 billion in 1997. Since then, investors' appetites have waned, public support to privatization decreased and the role of public investments in the provision of infrastructure services has gained momentum again1. While the increase of public investments is welcomed, given the magnitude of infrastructure needs in the region ­ roughly 4 to 6 percent of GDP per year to catch up or keep up with countries that once trailed it, such as China and Korea ­ and the fiscal limitations of the public sector, private sector financing for infrastructure will always be important in Latin America. While in Latin American countries, state-owned enterprises continue to account for more than 10 percent of gross domestic product, 20 percent of investment, and about 5 percent of formal employment (Kikeri, 1999), the infrastructure sector has dramatically changed. Specifically, while at the beginning of the 1990s only 3, 3, and almost 0 percent of the subscribers of fixed telecommunications, electricity and water distribution, respectively, were in private hands, in 2003 these ratios were 86, 60, and 11 percent. The setting of regulatory frameworks has accompanied that increase of private sector participation in infrastructure. 2 There is strong evidence supporting the generally positive economic results of these policies. Some examples include Boardman and Vining (1989) and Megginson, Nash, and van Randenborgh (1994) (see Megginson and Netter, 2001, and Chong and Lopez-de-Silanes, 2003, for more recent reviews). Yet, public perceptions of the outcome are not very positive. Chong and Lopez-de-Silanes (2003) have, among others, summarized and addressed the most voiced criticism. In the case of Latin American countries (LACs) and for the infrastructure sector, beyond case studies, there is little empirical literature analyzing impact and determinants. Most of it has focused all sectors and on the performance of financial indicators (see Megginson, Nash, and van Randenborgh, 1994, and D'Sousa and Megginson, 1999). Recently Andres, Guasch, and Foster (2006) evaluate the impact of private sector participation on output, efficiency, labor productivity, quality, coverage and prices, using a large cross country data set for Latin America. Also, the impact of competition is analyzed in Andres, Guasch, and Foster, (2006), the issue of renegotiation of the concessions in Guasch, (2003), Guasch, Laffont, and Straub, (2003, 2004), the profitability of private infrastructure firms in Sirtaine, Pinglo, Guasch, and Foster, (2005). Yet, there is little work that has focused on the determinants of outcomes and particularly on the impact of regulation on those outcomes. While the theory tells that regulation matters, there is a shortage of empirical work analyzing that issue. Some exceptions are Wallsten (2001), Jamasb (2005), Cubbin and Stern (2005), Stern and Cubbin (2004), and Minogue and Carino (2006). The objective of this paper is to add to that scarce literature, testing the impact of regulation from three different angles: (a) on aligning costs with tariffs- firms profitability, (b) on 3 reducing/deterring opportunistic renegotiation, and (c) on its effects on productivity, quality of service, coverage and prices. That is done respectively in Sections 2, 3 and 4 respectively. The analysis uses an extensive data set of about 1000 concessions granted in Latin America from the late 1980s to the early 2000s compiled by Guasch (2004). 2. Testing the Impact of Regulation on Aligning Costs with Tariffs-firms Profitability Unlike normal competitive business sectors, the profitability of concessions is not simply a reflection of market conditions and managerial competence, but is to a considerable extent determinedor at least circumscribedby regulatory decisions. Infrastructure companies operate mostly under a monopoly regime and thus are subject to regulation of tariffs and other aspects of enterprise performance. Thus, the observed profitability of these concessions in part should reflect the quality of the regulatory framework and the performance of the regulators that oversee them. 2.1 Theoretical framework Regulation aims to protect consumers from abuse of monopoly power and investors from opportunistic behavior by the government, given the politically sensitive nature of infrastructure tariffs and the large sunk cost characteristics of the companies' investments. In consequence, regulatory decisions have a substantial impact on the profitability of companies. Ideally, the regulator's objective should be to maintain alignment between a company's rate of return and its cost of capital. This is because a rate of return in excess of the cost of capital inappropriately 4 penalizes consumers, while a rate of return beneath the cost of capital inappropriately discourages further investment. The closeness of that alignment will depend, among other things, on the quality of regulation. In theory, the closeness with which the rate of return tracks the cost of capital will also depend on the chosen regulatory regime. Under rate of return regulation, the regulator has the possibility of making frequent price adjustments to keep realigning the company's rate of return with its cost of capital. Under price cap regulation, on the other hand, the regulator sets tariffs so that expected returns match the cost of capital ex ante, but allows these returns to diverge ex post during the periods between regulatory reviews. However, in practice, in Latin America, the distinction between price cap and rate of return regulation is somewhat blurred due to frequent renegotiation of infrastructure contracts (Guasch, 2004; Guasch and Spiller, 1999; Gomez- Ibanez, 2003) , and to the fact that review methodologies sometimes take into account historic 2 divergences between the rate of return and the cost of capital in adjusting future prices, which goes against the forward looking principles of price cap regulation. Thus, the practice in the region would best be described as a hybrid regime. Therefore, instead of focusing on the dichotomy between price cap and rate of return regulation, the approach taken is to develop a measure of the overall quality of the regulator that oversees each of the companies in the sample. The purpose of this section, then, is to empirically evaluate the impact of the quality of regulation on the profitability of the firms. The hypothesis is that the better the quality of regulation, the closer is likely to be the correspondence between the firm's rate of return and the firm's cost of capital. 5 2.2 Measuring regulatory quality In order to test this hypothesis a quantitative measure of regulatory quality is needed. Good regulation is defined by clear, stable and predictable rules, a purely professional and technical interpretation of the law and contract, ability to withstand influences and pressures from the stakeholders such as government and operators, and the establishment of a predictable and adequate allocation of resources. In consequence, the index developed here considers three key dimensions of regulatory quality: legal solidity, financial strength, and decision-making autonomy. The construction of each of these indices and associated scoring method are detailed in Table 1 below. Legal solidity refers to the stability, and thus predictability, of the regulatory regime. The strongest legal foundation is when the regulatory framework is embedded into a law, as opposed to a less strong legal instrument-less difficult to change (such as a decree or a contract if the judiciary is not reliable). Financial strength refers to the resources the regulatory agency has to undertake its functions. This dimension has two aspects. The first aspect is financial independence, which is achieved when a regulatory entity has its own source of revenue (for example via a sectoral surcharge) that does not depend on the government budget. The second aspect is financial strength, which is a function of the size of the agency's budget. Decision-making autonomy measures the likelihood that regulatory decisions are based on technical as opposed to political criteria. This dimension has three aspects. The first aspect is independence of appointment, which measures the extent to which the appointment process 6 avoids a purely political appointee without adequate technical knowledge of the sector. The second aspect is duration of appointment, which indicates whether a regulator can be reappointed and hence might be less likely to act independently and issue professionally and technically based decisions. The third aspect is collegiality of decisions, which measures the relative difficulty of regulatory capture, thought to be lower when multiple regulators act jointly within a board structure. While each of these elements are individually relevant, it is also of interest to aggregate them into a single quality index that gives equal weight to each of the three dimensions that have been identified. For the sample of companies covered in this study, the average score on this index of overall regulatory quality is 0.51 as against a potential maximum of 1.0, suggesting that the quality of regulation is not overall very high. However, there is significant variation in quality across countries and sectors, with scores ranging widely between 0.12 and 0.85. The highest average score is obtained on legal solidity, 0.65, as against decision-making autonomy, 0.56, and financial strength, 0.34. Pair-wise correlations between each of the regulatory quality measures are typically low at around 0.20, and in no case greater than 0.57. In some cases, pair- wise correlations even take negative values, suggesting that high regulatory quality along one dimension is correlated with low regulatory quality along another dimension. This result illustrates that few countries have consistently applied all of the design principles needed to ensure good quality regulation. 7 Table 1: Regulatory quality index: components and construction Weight Scoring Legal solidity 0.33 1 if regulatory framework established by law, 0 otherwise. Financial capacity 0.34 Sum of scores on factors detailed below. · Financial independence 0.17 · 1 if funded from regulatory levy, 0 if funded from public budget · Financial strength 0.17 · Regulatory budget as % sectoral GDP normalized on [0,1] scale Decision-making autonomy 0.33 Sum of scores on factors detailed below. · Independence of 0.11 · 0 if appointed directly by Executive, 1 if screening by appointment legislature · Duration of appointment 0.11 · 1 for a single fixed term, 0 for indefinite appointment · Collegiality of decisions 0.11 · 1 if headed by regulatory commission, 0 if by individual regulator Note: Scores between 0 and 1 are given for intermediate cases. These indices of regulatory quality are used to try to explain differences in the divergence between rate of return and cost of capital across the different companies in the sample2. This is done by regressing the difference between the Project Internal Rate of Return and the Weighted Average Cost of Capital (IRR-WACC) against this set of explanatory variables. The hypothesis is that the greater the quality of regulation, as measured by the described index, the smaller the differential should be, suggesting that the regulatory quality sub-indexes would enter the regression with negative signs. Two separate measures of the IRR-WACC differential are considered. The first measure is the simple IRR-WACC differential. This captures the quality of regulation purely from a short-term consumer's perspective, since the smaller the IRR-WACC differential (including negative values), the lower the resulting tariffs for consumers. However, this constitutes a myopic view since a negative IRR-WACC undermines investment incentives and therefore 2From that universe of private contracts, we used a sample of 34 concessions built by Sirtaine, Pinglo, Guasch, and Foster (2005), using the following criteria: (i) to include most Latin American countries with meaningful privatization programs; (ii) to include companies from all main infrastructure sectors; (iii) to focus on companies with at least 5 years of operations (in order to have a time series of data of adequate duration for the analysis); and (iv) to focus on companies publishing good quality financial statements. 8 ultimately penalizes consumers through declining service quality, decelerating service expansion, and potential flight of investors. Therefore, the absolute IRR-WACC differential is taken as a second relevant measure. According to this indicator, what matters is minimizing the distance between IRR and WACC, with positive and negative differentials regarded as equally reflective of poor regulatory decisions. 2.3 Simple differential (myopic consumer protection) The results for the first set of regressions are reported in Table 2, using each of the four measures of IRR-WACC differential.3 Despite small sample sizes, three out of the four models show that the regulatory quality variables are significant in overall terms, and are on their own capable of explaining 20-25 percent of the IRR-WACC differential. Moreover, some of the regulatory quality variables are also individually significant. Thus, the financial strength variable is significant at the 5 percent level in most of the regressions with the expected negative sign, indicating that regulators with larger budgets tend to have greater success in minimizing the IRR- WACC differential. In addition, the collegiality of decision variable is also significant at the 5 percent level, but takes a positive sign. This suggests that, arguably contrary to expectations, regulatory entities headed by a single superintendent do a better job at reducing the IRR-WACC differential than do broader based regulatory commissions.4 9 Table 2: Summary of regression results Dependent variable Simple Simple Simple Simple differential 1 differential 2 differential 3 differential 4 Financial independence -0.340 -0.174 -0.151 -0.135 Financial strength -0.372 -0.332** -0.355** -0.370** Legal solidity -0.026 0.077 0.070 0.080 Independence of appointment -0.109 -0.068 -0.101 -0.109 Duration of appointment -0.125 -0.011 -0.038 -0.030 Collegiality of decisions 0.455** 0.256** 0.271** 0.267** Constant -0.341 -0.047 -0.022 0.002 P-value 0.156 0.072* 0.052** 0.045** Adjusted R-squared 0.124 0.208 0.237 0.248 No. of observations 32 30 30 30 Notes: Regressions based on 30 observations; , , * ** ***indicate significance at 10%, 5%, and 1% level respectively 2.4 Absolute differential (protecting both consumers and investors) The results of the second set of regressions are reported in Table 3. Given that taking the absolute value of the IRR-WACC differential reduces the spread across observations in an already small sample, a log-linear specification is used to ensure that there is adequate variation for the purposes of the regression. Overall, this second set of regressions does not perform as well as the first. Nevertheless, two of the models show overall significance at the 5-10 percent level and are able to explain around 20 percent of the variation in the IRR-WACC differential. As before, the financial strength variable proves to be significant in some specifications, though not always with the expected sign. On the other hand, the collegiality of decisions is no longer statistically significant. The lower level of significance and explanatory power associated with this second set of regressions may simply be reflecting the fact that regulatory efforts are more strongly motivated by short-term considerations of keeping prices as low as possible for current 10 consumers, than by long term considerations of keeping returns as close as possible to hurdle rates for investors. Table 3: Summary of regression results Dependent variable Absolute Absolute Absolute Absolute differential 1 differential 2 differential 3 differential 4 Financial independence 1.071 -0.653 -0.001 0.071 Financial strength 2.619** -2.478 -2.488** -2.140** Legal solidity -0.697 0.928 0.412 0.844** Independence of 1.147 0.974 0.577 -0.050 appointment Duration of appointment -0.478 1.412 1.053 0.767 Collegiality of decisions -1.771 -0.810 -0.456 -0.243 Constant -1.104 -2.618** -2.365** -2.487** P-value 0.094* 0.273 0.125 0.049** R-squared 0.171 0.069 0.156 0.242 No. of observations 32 30 30 30 Notes: Regressions based on 30 observations; , , * ** ***indicate significance at 10%, 5%, and 1% level respectively The conclusion of this analysis is that regulation matters in aligning cost of capital and rate of return, as variations in quality across regulatory regimes are significant and material in determining the size of the IRR-WACC differential. However, regulatory efforts seem to be more closely associated with minimizing the simple IRR-WACC differential (and thereby keeping tariffs as low as possible for current consumers), than with minimizing the absolute IRR-WACC differential (and thereby keeping profitability well aligned with hurdle rates of return). Another striking feature of the results is that regulatory quality variables seem to have overall significance, more than individual significance, in determining IRR-WACC differentials. This is in fact consistent with the fact that performance along different dimensions of regulatory 11 quality is not highly correlated, and that the benefits of high regulatory quality along one dimension can be completely offset by low regulatory quality along another dimension. Thus for regulation to be effective, one needs the whole package of regulatory characteristics. If some of the key ingredients are missing the effectiveness of regulation is highly diminished. 2.5 Summary We have analyzed the differences between returns and costs of capital and shown that the variation of net returns across concessions can be partially explained by the quality of regulation. We have shown that the better the quality of regulation the closer the alignment between financial returns and costs of capital as is desirable. Quality of regulation is found to be a significant determinant of the divergence between the overall profitability of the concession and its corresponding hurdle rate, explaining around 20 percent of the variation. Thus we have shown that regulation indeed matters. However, regulatory efforts seem to be more closely associated with keeping tariffs as low as possible for current consumers, than keeping profitability well aligned with hurdle rates of return. The policy implications are clear. Significant efforts should continue to be placed to improve the quality of regulation. 3. Testing the Impact of Regulation on Reducing/Deterring Opportunistic Renegotiation 12 3.1 Concessions contracts in Latin America In Latin America, a majority of the privatization cases took the form of concession contracts. This was mostly to avoid political, legal and sometimes constitutional impediments to the outright sale of state assets to private operators that were often foreign firms. A concession contract grants a private firm or consortium the right to operate a given infrastructure in exchange for the revenues generated by users' payments, and lasts for a limited period of time (in general between 15 and 30 years), after which the underlying assets are devolved to the state. However, concession contracts have suffered from a number of problems, the most serious of which has been renegotiation. Considering an exhaustive sample of more than 1,000 concessions in Latin America and the Caribbean during the period 1985-2000, and excluding telecommunications where most projects were real privatizations with transfer of assets, 41 percent of the total projects in the three remaining sectors were renegotiated at some point. In water and transport, renegotiations have affected 74 and 55 percent of the projects respectively, and have occurred 1.6 years and 3.1 years on average after the award (Guasch, 2004). Such renegotiations have had a negative impact on users, including the need for additional risk premium ex ante (Guasch and Spiller, 1999), and ex post service disruption, non- compliance with expansion targets and excessive prices due to cost pass-through charged to customers, among others. For example, the Mexican toll road program, which consisted of 52 highways built in the early 1990s, was finally bailed out by the government in 1997. The estimated cost was between 1 and 1.7 percent of GDP (Guasch, Laffont, and Straub, 2005). It therefore becomes important to understand the reasons for these failures and in particular the role that regulation has in determining those outcomes. That is the aim of this section. 13 3.2 Renegotiations of concession contracts and their determinants Renegotiations may be of two types: renegotiations initiated by operators (Guasch, Laffont and Straub, 2003) or those at the initiative of local or national governments (Guasch, Laffont and Straub, 2005). Firm-led renegotiations might be related to economic shocks such as a devaluation or a recession, or might be opportunistic, when a firm that was previously awarded a concession seeks a bilateral negotiation with the government or the regulatory agency to strike a better deal than the initially agreed one. This may significantly reduce the benefit of the competitive pressure introduced by the ex ante auction procedure, first simply because the agreed parameters (tariffs, transfers) are modified and second because firms that anticipate this may be tempted to strategically undercut rivals at the bidding stage. Government-led renegotiations may sometimes be of a Pareto improving nature (related to unforeseen contingencies), but most of them are opportunistic, with politicians during or after an election campaign reneging on previous contracts to please their constituencies. Recent cancellations of water concessions in 2005 in Bolivia and the ongoing renegotiations of most concessions in Argentina after the 2001 crisis, in which the government refuses any significant adjustment of the rates converted to devalued pesos despite contract clauses that contemplated indexation to the dollar and US inflation, and are examples in case. A look at the data in Table 4 shows that regional volatility seems to play an important role in the timing of these renegotiations. For example, a number of them occurred around the hyperinflation at the end of the 80s in Argentina, during and after the Tequila crisis in 1995 in Mexico and at the time of the Real devaluation in 1999 in Brazil. It is therefore interesting to 14 find out if economic shocks were the only determinants of renegotiations, or if there were other flaws, in contract or regulatory framework design, that were pivotal in explaining the high incidence of renegotiation. Table 4: Renegotiation by Type of Initiator and Year 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 Total All countries outstanding concessions 10 38 38 50 78 103 123 132 156 187 187 165 Number of renegotiations 0 13 3 9 12 14 23 15 15 11 27 20 162 Firm-led reneg. 0 12 2 2 0 1 3 3 11 4 1 14 53 Govt-led reneg. 0 0 0 0 10 13 19 11 3 7 25 6 94 Joint-led reneg. 0 1 1 7 2 0 1 1 1 0 1 0 15 Source: Guasch, Laffont and Straub (2005) Guasch, Laffont and Straub (2003, 2005) results are based on a sample comprising 307 projects in the water and transport sectors, in five countries (Argentina, Brazil, Chile, Colombia and Mexico), across 12 years, for a total of 1287 observations (see Guasch, 2004). For each contract, there is information on the general characteristics of the projects (sector, year of award, duration), on the award process, the investment and financing conditions, the institutional and regulatory context and the type of price regulation in place (price cap versus rate of return), and other contract clauses (arbitration, income guarantees, take-over clauses, etc.). These are completed by macroeconomic data (growth rate, exchange rate evolution), dummies for national and local elections and a full set of institutional indicators (corruption, quality of the bureaucracy, rule of law). 15 The initial estimations are based on a random effect probit, which is a linearized version of the equations giving the probabilities of firm-led and government-led renegotiations in the respective theoretical models: yint = I [y*int = xi1 + 2 zint + Ent3 + eint < 0], For concession i, in country n, at time t, yint is the binary variable indicating whether there was a renegotiation by the firm (resp. by the government), x is a vector of time-invariant characteristics of the contract, z is the time elapsed since the award, and E is a vector of environmental characteristics, including economic shocks, elections and quality of institutions. Alternatively, Guasch, Laffont and Straub (2003) present a competing risk duration model, which allows for both type of renegotiations hazard simultaneously. The specific model used follows Han and Hausman's (1990) semi-parametric competing risk model, with a non- parametric baseline hazard consisting of a set of dummy variables for each period. This model is estimated using a bivariate probit with the complete set of period dummies. One major econometric issue is the fact that most contract clauses, such as the type of price regulation or specific guarantees included, must be considered to be endogenous. Indeed, we expect the contracting parties to choose them according to their observable and unobservable characteristics and those of the projects. For example, the type of tariff regulation chosen is likely to be affected by the potential efficiency of the concessionaire (more efficient firms would prefer price cap regulation, which is more risky but makes them residual claimant for their cost savings) and also by the fact that riskier projects would call for lower-powered (rate of return) regulation. Similarly, most types of guarantees have in general been included to convince private agents to take on more risky concessions, as in the case of toll road programs for which demand 16 proves very difficult to predict accurately. The challenge is thus to control for this ex ante self- selection effect in order to assess correctly the ex post specific incentive effect of the variables under study. To tackle this, we implement in the two models mentioned above a two-stage instrumental variable procedure using as instrument a number of exogenous characteristics of the environment such as, institutional quality, sectors of activity and the existence of a regulator. For the variables found to be endogenous according to the Rivers and Vuong (1988) test, we take the predicted values from the first stage estimations, insert them in the second stage model and adjust the standard errors with a bootstrapping procedure. Unsurprisingly, the variables for which exogeneity is rejected are price cap regulation, the investment and financing variables, and clauses such as minimum income guarantee and existence of an arbitration body. The results arising from both models are strongly consistent. Table 5 presents the results both types of renegotiations. It shows that contract characteristics, political and economic variables, and regulation all matter in explaining the frequency of renegotiations. 17 Table 5: Estimates of the determinants of renegotiations Government-led Firm-led Renegotiations Renegotiations -1.09*** -1.40*** Existence of regulatory body (0.22) (0.34) 0.68* -0.46* Price cap (0.38) (0.40) 0.96** -0.70*** Investment requirements (0.40) (0.24) 0.35 -1.23*** Private financing (0.28) (0.24) -0.35** -0.57*** Bureaucratic quality (0.15) (0.16) 0.31 0.21 Elections -1 (0.20) (0.19) -0.06*** -0.05** Growth -1 (0.02) (0.03) -0.14*** -0.08** Growth -2 (0.02) (0.03) 0.53 -0.38 Transport dummy (0.36) (0.36) Log likelihood -251.1 Number of observations 1132 Source: Guasch, Laffont and Straub (2003, 2005). Significance at the 1%, 5%, and 10% level is noted by ***, **,* respectively. 3.3 Regulation and renegotiations First, the existence of a regulator at the time the concession contract is signed appears to be crucial in avoiding failures during the early life of concession projects. This aspect has the strongest marginal effect of all variables found to be statistically significant. Comparing three specific contracts out of the initial sample, and using the probabilities predicted by the empirical model, Guasch, Laffont and Straub (2003) show that had a regulator been in place at the time of awarding the contract, the respective probabilities of renegotiation in the last year of existence of the contract would have been reduced from 29.7, 9.9, and 3.1 percent, to 5.3, 0.3, and 0.2 percent respectively. 18 Depending on the type of renegotiation that is considered, at least two complementary lines of explanation are relevant here. On the one hand, regulators seem to allow for better contracts from the start, which reduces the necessity of posterior adjustments for unforeseen contingencies (this is particularly relevant for firm-led renegotiations). In the Latin American context, characterized by frequent, and difficult to predict, economic shocks and by the imperfect enforcement of contracts, drafting complete contracts is bound to fail. Moreover, long and complex contracts are often inefficient, because they lack transparency and lend themselves to contradictory interpretations and therefore opportunistic revision claims. As a consequence, most contracts are short concession-specific documents that rely on complementary rules contained in the relevant jurisprudence. This approach makes previous regulatory experience in dealing with the design of concessions contracts pivotal in limiting the occurrence of later renegotiations. On the other hand, regulators are even more effective in weak governance environments and appear to constitute a barrier against opportunistic behavior by governments (Guasch, Laffont and Straub, 2005). This conclusion is supported by several significant interactions showing for example that the previous existence of a regulator has a stronger marginal effect in a context characterized by more corruption, or that a good quality bureaucracy is more effective in limiting the incidence of renegotiations after elections. Finally, Guasch, Laffont and Straub (2005) also show that the fact that the regulator does not belong to a ministry significantly reduces the probability of government-led renegotiation. In that regard, these firm-level results confirm some cross-country studies results that show the importance of experienced and independent regulators in the telecommunication and electricity sectors (Wallsten, 2001; Cubbin and Stern, 2005). 19 Second, the choice of price regulation, between a price cap and a rate of return scheme, is of utmost importance. Beside well-known concerns with price cap regulation, in particular regarding the impact on quality and the implied risk transfer from consumers to the firm, Guasch, Laffont and Straub (2003, 2005) show that the main consequence of choosing a price cap regulatory scheme is the increased probability of renegotiation. Looking again at the marginal effect, they show for example that had the three random sample contracts-used to make this simulation analysis- been under a rate of return scheme, the respective probabilities of firm- led renegotiation in the last year of existence of the contract would have been reduced from 29.7, 9.9, and 3.1 percent, to 13.8, 3.3, and 0.8 percent respectively. Given that in the sample under study, above 70 percent of the concessions are regulated by price cap, this is clearly a major concern. Moreover, price cap schemes increase the riskiness of projects, which is reflected in an increase of the cost of capital and implies that firms end up facing higher interest. In contexts where institutions are weak, inexperienced and often unable to resist political pressures, the consequence is that most regulated firms (or the government and interest groups related to the firms) appropriate the gains when the conjuncture is favorable, but are able to transfer the losses to consumers during bad times. As a consequence, there is a growing pragmatic tendency to advocate the abandonment of price cap regulation, a synonym for the higher risk of renegotiation and higher cost of capital, and the return to an hybrid type of regulation, including some elements of rate of return (see for example Estache, Guasch and Trujillo, 2003). Such a move would imply recognizing that the shift to a hybrid regulatory scheme is imposed de facto by ex post renegotiations, which carry high associated social costs, because they tend to endogenize the regulatory review lags. In this 20 situation, it could prove less costly to adapt regulatory rules from the start by adopting lower- powered price regulation schemes. 3.4 Summary In summary, two related dimensions of regulation matter when it comes to avoiding disruptive renegotiations. The first one is the regulatory environment, including the very existence of a regulator from the start, but also its independence from potential political pressures. The second one is the type of price regulation itself. It should be noted that these two aspects can hardly be separated. Indeed, price cap regulation has often been the salient choice of governments lacking previous experience with regulation, because it appeared to be less informationally demanding. The absence of a regulator when initiating transfers of infrastructure to the private sector and the choice of price cap therefore often went in tandem. The results mentioned above show that a better strategic approach would be for governments to consider a sequence including first the development of a correctly endowed and reasonably independent regulatory agency, which would subsequently be in charge of the definition of the contract and the appropriate price regulation. 21 4. Testing the Impact of Regulation on Productivity, Quality of Service, Coverage and Prices 4.1 Overview This section uses the framework developed in Andres, Guasch, and Foster (2006). As we have already described, their analysis splits the data into three periods: "pre-privatization", transition, and post-privatization periods, where the transitional period commences after the concession announcement and lasts until one year after the concession award. The motivation for this segmentation is that some of the more important changes start simultaneously with the privatization announcement and lasts one year after the change in ownership. In addition, some of these indicators are driven by firm specific time trends and not privatization itself; therefore, the authors also control for this effect. Their main results are summarized as follows: (i) After controlling for a positive firm-specific time trend, data for service coverage suggests that privatization has an upward impact on telecommunications, but no effect on electricity and water and sanitation. (ii) Indicators for technical losses are positively affected by privatization. While most of the improvement for electricity happens during the transition period, those for telecommunications, water, and sanitation occur later. (iii) Prices also significantly increased for the sectors during and after the transition except in telecommunications as the average cost of installation of a residential line decreased in every period (the monthly charge for residential service, however, increased substantially). 22 (iv) Labor productivity significantly changed in all the three sectors, mainly during the transition period, and fundamentally caused an important reduction in labor redundancy: in the electricity and water and sanitation sectors, employment decreased on average 10 percent per year during the transition period. (v) The outcomes' results are significantly heterogeneous across firms. The current analysis is based on the last conclusion that shows the heterogeneity across firms. Our proposal attempts to better understand the determinants for this heterogeneity across utilities. The hypothesis is that some procedural and regulatory differences might explain some of these variances. Here we focus on four basic regulatory characteristics: (1) budget autonomy; (2) the legal autonomy of the regulatory body; (3) tariff regulation (price cap, rate of return, among others); and (4) duration of the regulatory board. Additionally, we will control for some additional features such as the award process (direct selection vs. auction process), the award criterion (highest price; lower tariff or investment plan), and the nationality of the concessionaire. The premise is that these divergences may significantly affect the incentives involved in the managerial decision process, which, in turn, affects firm performance on efficiency, quality, and price. 4.2 Procedure Ideally, to assess the impact of privatization, the performance of utilities under private operation should be evaluated against comparable public-operated firms from similar 23 environments, assuming these firms are the contra-factual of the privatized ones. In most cases, it is hard to identify an analogous firm; hence, most of the literature compares the evolution of selected indicators before and after the change in ownership. Most of the literature employs two different strategies to estimate the effect of the privatization. First, since Megginson, Nash, and van Randenborgh (1994), there have been several studies using means and medians of the periods before and after the event of privatization, as there has also testing on the significance of the change. Some research considers different samples of SOEs among countries and evaluates indicators. Another branch of literature assumes these policies to be treatments and follows the literature of program evaluation (see Heckman and Robb, 1985) by proposing a dummy for those periods where the SOE was privately owned, and checks its significance, as well as other interactions with characteristics specific to each paper (for example, Boardman and Vining, 1989; Ros, 1999). In this section we propose a modification of Andres, Guasch, and Foster (2006), by introducing interactions between the privatization dummies and the characteristics described previously. More specifically, we define a dummy for the transition and another for the after- transition period: ln yijt = DUMMY _TRANijt + DUMMY _ POSTijt + ijDij +ijt ( ) T P (1) ij where 1 DUMMY _TRANijt if - 2 sijt +1 0 otherwise and 24 DUMMY _ POSTijt 1 if sijt 2 0 otherwise where yijt are the variables of interest (outputs, inputs, labor productivity, efficiency, quality, coverage and prices). The main coefficients in this model are the dummies DUMMY _TRANijt and DUMMY _ POSTijt that are equal to one, if the firm i of country j were in a transitional or port-transitional year at time t . Given the fact that there are several variables not observable to the econometrician, fixed effects are included to capture the characteristics of the firm, such as, management, initial conditions, size, density of the network, as well as other aspects, which we assume to be constant for each firm across time. This fixed effect is captured by Dij . Additionally, sijt is a time trend that has a value equal to zero for the privatization award year. Thus, the first dummy identifies the average change in the dependent variable during the transition with respect to the average level previous to those years. The second dummy identifies the average change of the dependent variable after the transition with respect to the first period. Therefore, T and capture the effect on the outcome of interest, P during the transition and after that, given by the change in ownership. A second version of the equation (1) will also be estimated here with the introduction of a firm-specific time trend: ln yijt = DUMMY _TRANijt + DUMMY _ POSTijt + ijDij + ijtij +ijt ( ) T P (2) ij ij Equation (2) will use the same dependent variables as well as the dummies used in the static model. However, the fourth coefficient captures the time trend of the variable of interest. Several factors may affect this, like the initial conditions. Hence, it is important to control for the firm's specific value. 25 To identify the different characterization effects of the privatization process as well as the regulation, we test the variables with the two main dummies. More precisely: ln yijt = DUM _TRANijt * Xijt + DUM _ POSTijt * Xijt + ijDij +ijt (3) ( ) T P ij ln yijt = DUM _TRANijt * Xijt + DUM _ POSTijt * Xijt + ijDij + ijtij +ijt ( ) T P (4) ij ij Now T , which was used as a scalar number in our previous specifications, becomes a vector with the coefficients for each characteristic of the vector Xijt that is of the form , xijt ,...,xijt (1 1 N ) with N as the total number of characteristics evaluated. The first coefficient of the vector T will became the average effect of change in ownership during the transitional period on a given indicator for a firm without the characteristics evaluated in the other elements of the vector Xijt . Equivalently, the vector contains the coefficients for the different characteristics of P vector Xijt , but for the post-transitional years. Since we are using a semi-logarithmic functional form of these models for each of the indicators, when interpreting the coefficient estimates of the dummy, it should be remembered that the percentage impact in each indicator is given by e -1. Correcting for potential nonspherical errors requires a more adequate approach, such as, the Generalized Least Square (GLS); however, this estimation requires the knowledge of the unconditional variance matrix of ijt , , up to scale. Hence, we must be able to write = C , 2 where C is a known GxG positive definite matrix. As this matrix is unknown, we will follow a Feasible GLS (FGLS) approach that replaces the unidentified matrix with a consistent estimator. Hence, our models specify heteroskedastic error structure with no cross-sectional correlation. 26 4.3 Data For our research we use an official data set provided by public and private sectors, as well as a novel one built by the World Bank. First, by using the official data reported by the firms to their investors and statistical reports of the regulator agencies of each country, we build an unbalanced panel data set of key indicators on outputs, inputs, labor productivity, efficiency, quality, coverage, and prices. Furthermore, we requested information from each of the companies and international organizations like the ITU (International Telecommunication Union), the OLADE (Latin American Organization of Energy), as well as information provided by each regulatory office. We make a particular effort in corroborating the company data with several public sources and with data from the firms provided by different governmental offices. We are also particularly cautious about the consistency and comparability of the data across time and countries (see Andres, Guasch, and Foster, 2006). Secondly, the novel dataset built by the World Bank describes the characteristics of nearly 1,000 infrastructure projects awarded in Latin American and Caribbean countries from 1989 to 2002, in the sectors of telecommunications, energy, transportation and water. (See Guasch, 2003.) The analysis focuses on several indicators of outcomes, inputs, labor productivity, efficiency, quality, coverage and prices. Some of these variables are used by other authors with other samples, such as, Ros (1999), who employs equivalent indicators for coverage, labor productivity, quality and prices for the telecommunications sector. Ramamurti (1996) uses analogous indicators in output, coverage, and labor productivity for the four Latin American 27 telecommunications firms of his study. Saal and Parker (2001) use similar indicators for output, employment, quality, and prices for water and sewerage companies in England and Wales. Table 6 shows the summary statistics of these variables in each sector. Table 6: Summary statistics Variable N Mean Median SD Min Max Electricity Distribution Number of subscribers 98 497,776 225,230 681,698 2,700 3,884,579 Output [thousand of KWHs] 100 2,850 789.5 5,282 13.8 34,300 Number of employees 87 1,421 625 2,115 18 13,642 Subscribers per employee 84 558.81 506.67 244.20 210.45 1,523.27 Output per employee 84 2,343.48 2,116.46 1,298.60 663.86 7,323.09 Distributional losses 90 15.3% 13.6% 6.6% 2.0% 33.9% Duration of interruptions per subscriber 65 25.26 20.36 21.01 1.75 100.00 Frequency of interruptions per subscriber 67 22.63 16.03 21.24 1.07 100.00 Subscribers per 100 HHs 86 74.6% 81.3% 20.7% 7.0% 100.0% Avg price per KWH [in u$s] 92 88.70 85.34 35.43 7.47 323.61 Fixed Telecommunications Number of subscribers 16 2,423,040 824,594 3,150,005 28,048 9,642,200 Output (million of minutes) 13 20,500 6,200 28,800 774 83,100 Number of employees 16 12,268 9,732 12,097 966 47,949 Subscribers per employee 16 209.30 109.27 241.96 33.81 736.65 Output per employee 13 1,627.35 844.29 1,790.44 257.10 6,419.45 P% of digital lines 16 67.0% 70.3% 26.4% 14.6% 100.0% % of completed calls 12 67.0% 64.8% 20.4% 20.0% 98.8% Subscribers per 100 inhabitants 16 9.84 8.40 5.83 2.96 22.01 Price of 3-minute call [in u$s] 14 0.13 0.07 0.25 0.01 0.99 Monthly charge for a resid. Sv. [in u$s] 15 6.16 6.01 4.52 0.36 19.97 Price for the installation of a line [in u$s] 15 343.75 309.51 339.35 1.20 1,102.26 Water and Sewerage Total Subscribers for water 48 147,119 78,864 223,803 1,894 1,282,074 Total Subscribers for sewerage 43 107,286 42,991 173,795 435 799,994 Water Production 47 91,400 28,900 2,110 145.6 13,700,000 Number of employees 42 528 258 997 9 6,346 Water subscribers per employee 42 312.23 283.10 153.56 43.34 772.36 Water production per employee 33 39.1% 37.3% 12.7% 15.3% 62.8% Continuity [hours per day] 21 19.40 22.97 6.57 - 24.00 Potability [%] 29 88.5% 98.9% 26.1% 0.0% 100.0% Water subscribers per 100 HHs 44 74.83 88.29 34.30 0.01 100.00 Sewerage subscribers per 100 HHs 34 64.61 71.99 27.83 0.30 97.70 Avg price for water [u$s/m3] 27 0.48 0.44 0.16 0.17 0.84 Avg price for sewerage [u$s/m3] 12 0.40 0.39 0.22 0.07 0.97 Note: each observation is the average for the available information since 5 years before the change in ownership and 5 years after that. The countries analyzed include: Argentina, Bolivia, Brazil, Chile, Colombia, El Salvador, Guatemala, Guyana, Jamaica, Mexico, Nicaragua, Panama, Peru, Trinidad and Tobago, and Venezuela. The sample consists of unbalanced panel data that include 181 firms and 1,885 firm- 28 year observations. Each of the sample firms contain at least one year of pre-privatization data, while 150 of the 181 firms have information for at least the previous 3 years. We matched our previous data set with a novel dataset built by the World Bank that describes the characteristics of nearly 1,000 infrastructure projects awarded in Latin American and Caribbean countries from 1989 to 2002, in the sectors of telecommunications, energy, transportation and water. (See Guasch, 2004). This dataset contains information with respect to the privatization process we know how many bidders participated, the contract process5, the award criterion6, and the type of concession7. With respect to the regulatory framework, we know how the establishment of the legal framework8, the regulation of tariffs9, if there were a possibility of contractual renegotiation, and (if this was the case) who would initiate it10. The data also contain additional contractual clauses, such as, if it considered a termination clause, about the arbitration process, claim solving institution, obligation to provide universal service, duration of the contract, contract renewal, government's guaranties, if the government granted subsidies, frequency of the tariff review, and how the exchange and commercial risk were born. If the contract was renegotiated, we know when it was, the reason given for it, and its outcome. Some characteristics of the regulator include: an index of its autonomy, its budget source, the duration of the regulatory board member mandate, as well as the year of the regulatory board's inceptions. Among these variables we selected those with enough variation across firms that allow us to better identify the effect of the differences in each outcome. Hence, Table 7 indicates the variables that we were able to use in this analysis, while Table 8 shows the summary statistics of the characteristics across the sectors. 29 Table 7: Description of the characteristics used in the analysis Variable Description Regulatory Board AUTON_YES Dummy with value 1 if the Regulatory Board was fully autonomous. AUTON_PART Dummy with value 1 if the Regulatory Board was partially autonomous. DURATION Dummy with value 1 if the duration of the Regulatory Board was 5 or more years Tariff Regulation TARIFF_RR Dummy with value 1 if the tariffs were regulated according to the Rate of Return TARIFF_PC Dummy with value 1 if the tariffs were regulated according to Price Cap. Table 8: Summary statistics of the characteristics used in the analysis Fixed Telecommunic. Electricity Distribution Water and Sanitation Variable # firms Mean # firms Mean # firms Mean Regulatory Board AUTON_YES 11 36.4% 84 39.3% 33 0.0% AUTON_PART 11 9.1% 84 38.1% 33 27.3% DURATION 4 75.0% 56 41.1% 9 100.0% Tariff Regulation TARIFF_RR 8 25.0% 106 20.8% 38 23.7% TARIFF_PC 8 62.5% 106 91.5% 38 89.5% 4.4 Main Results Tables 9 through 11 present the results of the regression analysis according to the four different specifications for each indicator. For some indicators, the interpretation of the results will be more relevant when considering firm-specific time trends. This applies when analyzing output, labor productivity, and coverage indicators; therefore, for these variables we include firm-specific time trends. The table clarifies when these trends were included. The results in this section suggest that most of the design characteristics of the private sector participation significantly affect the outcomes of each of the indicators; however, while some characteristics have positive effects on certain indicators, the same characteristics have 30 negative outcomes in other instances. The set of available choices is important to consider and analyze when focusing on specific targets. If the target is the expansion of the network, the strategy will focus on certain characteristics; however, if the target is an efficiency increase, another set of characteristics may be analyzed. In addition, when evaluating these same cases, we have found that not all the sectors react evenly to an identical set of characteristics. The main objective of this section is not to advocate a certain type of regulatory or contract design but rather to emphasize that privatization is not simply a straightforward yes-no decision. Indeed, there are many privatization design variables that can influence performance outcomes. The results in the following sections show that depending on the priorities of a country, certain privatization contracts or regulatory characteristics might be more important than others. For example, if reducing prices is of central importance to a country, then either a partially or fully autonomous regulatory body would be preferred over a non-autonomous one. The findings of the chapter can be summarized in three main points. First, regulatory and contract characteristics matter: the manner in which privatizations are undertaken can generate significant performance differences. Second, each regulatory and contract characteristic affects each performance variable differently. In other words, a certain contract characteristic could have a positive influence on one performance variable while having a negative or insignificant impact on another. Third, some regulatory and contract variables have bigger impacts that others. For instance, in some cases the changes attributed to having a fully autonomous regulatory body are much larger than changes attributed to other regulatory variables. The following section describes and expands on the more robust results across the different specifications. 31 32 )61( * 1* 14 )660 8 01 )040 1 04 )160 *2 *** ** 0* 5 *9 *** **6 *** 4 .6 eeyolp 0. 0.( 0. 0.( 0. 0.( .070- )140 02 )360 )440 )570 05 )060 s s 0.( .20 )170 0.( 32.0- )560 0.( 00.0- )760 0.( .070- )640 092 0.( -0. 0.( 040.0 0.( .170- 0.( .30 0.( ***855. )380 0.( Ye Ye 885 285 mer ) 15( 4* ) ) ) ) *** ) ) *** ) ) ) ) ) ) ) 19 12 s s pe 0.- .1 (0 960.0 12 8 7 8 9 563 .1 01 19 .2 .1 .1 .1 .1 .0 .1 .0 09 79 33 11 32 (0 0.- (0 **482.0- 11 71 51 29 68 16 95 11 83 (0 .40 12 (0 0.- (0 .20 )101.0( 551.0 )690.0( 03 0. (0 710.0 (0 -0 (0 0.- .0 (0 671.0- .1 (0 470.0 .1 (0 *4**686. .2 (0 Ye Ye 283 5.823 onticen ) 14( 7 04 con)nl( 0.- )140.0( ***131.0 )040.0( ** 520.0- 3) 05 0.( 460.0- 7) ***4 9* 04 16 05 02 s .7 66 0.( 19.0 )450.0( 2* 0.- )240.0( * 82 .00 )540.0( 0 0.- )040.0( *4* 630.0 )440.0( 6 0.- )730.0( 32 5. )001.0( s Ye Ye 106 425 ) * 13( 3* ) ) ) ) *** ) ) ** ) ) ) ) ** ) 24 01 9 01 99 9 16 01 53 2 54 7* 52 73 0.- .1 45 s s (0 0.- .0 (0 180.0- )080.0( 711.0 )080.0( 6 25 06 .2 0. (0 ***103.0- 11 09 7 07 1 .1 (0 .30 .1 (0 0.- .1 (0 362.0 93 91 .0 10 .0 (0 0. (0 0.- .0 07 .0 98 .1 (0 0. (0 6. (0 Ye Ye 503 7.563 )21( see **261.0 4) * 9) 3) 9) 3) 3) *** * 6) 5) 9* 1) *** ** *** 1 07 70 0.( .00- 03 0.( ***731.0- ***2 05 63 53 13 07 04 02 08 07 06 04 0.( .00 03 08 0.( -0. 0.( 790.0 07 75 08 61 05 47 2) 4) 4* 2) 1 6) 7* 6) 0) 0.( .30- 0.( .00- 0.( -0. 0.( .10- 0.( .00 29 05 08 0.( 0.- 0.( -0. 0.( 0.- 0.( 202.6 13 0.( seY oN 4 56 1.581 oypl )11( em of ***054.0 )311 *** ** ** ** 48 51 9* 04 5* 6) 43 1) 71 1* 4 3* 1) ** 40 4* 4) *** 7 56 (0. .10- )790 (0. .60- )911 (0. 430.0 )860 76 )841 12 5 1. (0. 0.- (0. .10- )921 29 10 (0. 0.- 0.( .10- 09 0.( ***713.0- )801 (0. .00- )190 15 )280 (0. 0.- (0. .00 )170 (0. 320.0- )121 06 )011 92 13 (0. 0.- (0. 0.- 35 (0. 0. (0. 423.7 )111 (0. seY oN 33 13 erb m nu)nl( )01( **7 )430 ** ** * * 8* 6* 6) 5* 4) **0 **5 *** 8 2* ) ) ** 7 8* 33 .060 0.( 17.0- )130 0.( 19.0- 05.0( 09.0- 04.0( .100- )050 153 0.( .080 )830 )550 0.( -0. 0.( .050- )440 0.( 12.0- )350 0.( .040 )140 270 610 280 510 0.( 0.- 0.( 0.- 0.( 61.6 )932 0.( seY oN 6 58 4.651 ) (9 ***4 ) ** 051 6 ) 900 ***2 )3 ***2 )7 9* ) 321 *4 ) *** ) 850 965 101 1 ) *** ) 940 653 980 0* ) ) ) ** ) 830 5 660 4 ) *** ) ** 0* 2 763 .390 (0. 04.0- (0. .220- .070( .180- .060( 44.0- s (0. .150 221 572 7 7. (0. 0.- (0. .040- (0. -0. (0. 15.0- (0. 00.0 (0. 00.0- 570 209 311 093 (0. -0. (0. 0. (0. 04.7 (0. Ye oN 35 12 productivity 8)( *** * ** 6) 2) *** 3) 9) 4* 7) 7) 3) 3) 8 3) 1) 2) 0 5) 99 8 2) 6 6) 1* 6) s s 2. 66 labor .00 03.0( 30 0.0 03.0( 61 61 .00- 02.0( 00 0.0 01.0( 05.0- 02.0( *64 0.0- 02.0( ***51 0.1- 03.0( 0.0- 03.0( 03.0- 03.0( 81 0.0 03.0( 44 0.0 03.0( 01.0- 02.0( 02.0 03.0( 00.0 02.0( 21. 16 02.0( Ye Ye 176 489 sit *** and (7) 16 un .20 )450. 2* 40 ***68 * ** 36 7 83 3* 9) 8) 92 45 5 19 34 8 8 8) 8 4) 24 (0 .00- )160. 93 .55 s s (0 0.0- )230. (0 .00- )920. (0 08.0- )550. (0 0.0- )260. 12. (0 -0 04.0( 0.0 05.0( 0.0- )150. (0 0.0- )730. (0 01.0- )340. (0 .00 )830. (0 .00- )650. (0 03.0 )830. (0 02.0 07. (0 01.0 08. )850. (0 13 (0 Ye Ye 084 8.685 owlf)nl( 7) 7) *** ** 6)( *** 3) 8) 4 1) 9) 9) 6) 78 19 6 6) 8 2) 1* 9) 8 ployees .00 01.0( .00 01.0( 16 )62 10 )22 70 s s 9. .00- 0.0( .00- 0.0( .00- 02.0( 01 .00- 01.0( 00.0- 02.0( **54 0.0- 01.0( **54 0.0- 01.0( 62 0.0- 01.0( 03.0 02.0( 01.0 02.0( 85. 19 03.0( Ye Ye 396 979 em of *** (5) 80 .10 )930.0( * ** ** 06 0 4. 0.0- )350.0( ) ) ** * * 2* 12 38 10 37 76 5 464 .00- .00( .00 .00( .00- )830.0( 29 .00- )330.0( 0* 71 5 1) 7 8) 09.0- )040.0( 310.0 )650.0( 2* 10.0- )530.0( .00 )050.0( 46 .00- )620.0( 41 .00- )520.0( 04 0. 08.0( 02 0. 08.0( 59 2.1 )880.0( seY seY ber 43 64 num )4( * 2) 8) 20 9 1) 9) 8) 8) 2* 0) 8) 2) 0 8) 05 4* 1) 9* 0) 0* 3) .5 27 .00 01 0.( 400.0 8) 00 0.( ***530.0- * 2) ** * ** 01 05 10 03 02 26 0.( .00- 00 0.( 00 01 01 01 08 01 3 06 0. 0.( .00 00 0.( .00- 00 0.( .00- 00 01 0.( 0.- 0.( 800.0 00 0.( .00- 01 00 00 0.( -0. 0.( 02 0. 0.( 01 0. 0.( 0.1 0.( seY seY 69 16 1% output,­ s ) oni (3 42 )510. *** 2* 29 72 0 01 01. 01 4 1) 12 6) *81 ** 6 8) 7) 34 03 04 11 0 4* 4 s s ect 0.0 (0 .00 )810. (0 .00- )020. (0 .00 )310. )510. .65 (0 -0 (0 0.0- )910. (0 00.0- 01.0( 0.0- 01.0( 0.0- )010. (0 .00 )110. (0 01.0- )110. (0 .00 )110. (0 .00- )710. 01. )310. (0 -0 (0 03.0 01. (0 02.0 01. )220. (0 12 (0 Ye Ye 284 9.9701 nnoc) attnaicfingis***; ** 0* 2) 9 7) 1* 1 * 7 ) ) *** (ln 2)( 4* analysis 01 0. )500.0( ***410.0 )500.0( ** ** 06 01 00 00 03 00 55 s s 0.- 0.( 0.- 0.( 0.- )010.0( 00 0.- )700.0( ** *** ** 47 5% 30 .00 )210.0( 0 01 0. )800.0( 7* 01 0. )700.0( 0.- )600.0( .00 140.0( 200.0 100.0( 191. 10 )410.0( Ye Ye 157 0.5761 ses (1) **4 ) ) * he 6 9 ) 3 ) ) 0 ) ) 8* ) ) 2 6 ) *** ) ) *** ) .020 110 ***0 ***0 )3 )2 48 9. 54 nt attnaicfingis (0. .040 150 (0. .070- .020( 00.0 .020( 03.0- 270 (0. 01.0- 220 350 240 (0. 0. (0. .030- 260 090 100 0 36 (0. 0.- (0. 02.0- 140 080 (0. .010 (0. 00.0- 080 610 240 140 230 (0. 0. (0. 0. (0. 1.41 290 (0. seY seY 45 11 reap ** Regression t 10 9: r os iv_trp_ iv_prp_ trap_n t ar se se at f b b cr rc _p t erro _y _y r r _p nsio id id uto ontu at ont ontu du du f_no b_ ati n_iota n_iot n_iota _pdraw ard rret nt tas EF dnertficepS %; s ni doohileki mrifforeb rs rd da Table mud mud _btr _btp _atr _atp au_rt _atp b_r_rt _rtp _ntr _ntp na_rt _ntp _atr aw_ pt ar_t_rt _rrerat_ pt onC mriF mriF servbO g-l Lo muN antS ntacifingis* 33 * 32)( 5 0.033*** .012)0( 0.001 .008)0( .023**0- .011)0( .0050- .009)0( .028**0- .007)0( .0030- .008)0( .035***0- .008)0( .0040- .008)0( .036***0- .011)0( 0.008 .009)0( .0010- .011)0( 0.017** .008)0( 1.385*** .036)0( esY seY 666 1634.2 * * ) * ager 31( 0** 87 .0150 .020)0( .0290 .021)0( 022*.0- .013)0( 010.0 .013)0( 036**.0- .008)0( 002.0- .017)0( .0090- 011).0( .0200- 015).0( .034**0- .012)0( 020*.0 .012)0( .022*0- .012)0( 011.0 .012)0( 023.0 .018)0( .0180- .017)0( 14.4 .032)0( esY seY 224 131.51 cove n)l( 30)( 91 13) 80 12) 31** 14) 02 12) 21 10) 00 09) 80 * 74 06) 06) 07) 02 06) 08 18) 5.0 0.0 0.0( 0.0 0.0( .00- 13* 06 0.0( .00 0.0( 0.0- 0.0( .00 0.0( 0.0- 0.0( 0.0 0.0( .00 0.0( .00- 0.0( .00- 0.0( seY esY 688 169 29)( **7 6) 1* 8) 8 )8 0 )8 7 6) 8 6) *7 4) 9 7) **6 )1 **8 )3 9 9) 0 9) ** 4* 6) .0 075 .030 0.01( 0.03 0.01( .020- .010( 0.01 .010( 0.01- 0.01( 0.00 0.01( .020- 0.01( .010- 0.01( .020- .010( .020- .010( 0.00 0.00( .000- 0.00( 1.37 0.03( esY seY 444 1172 28)( 3 .159**0- 0.065)( .069*0 0.036)( .197**0- 0.098)( 0.077*- 0.045)( .358***1- 0.228)( 0.007 0.095)( 0.014 0.063)( 0.028 0.033)( .266***0 0.054)( 0.132*** 0.033)( .559***1 0.234)( 0.034- 0.092)( .918***0- 0.233)( esY oN 346 208.4 3 3 3 3) 3 index 7)2( 64 266.0 ,469.902( 213**.0 .089)0( 500.0- ,469.902( .249**0- .114)0( 063.0 ,469.902( 150.0- .115)0( .4690- 469.90,2( 000.0- 063).0( .0630- .068)0( .0210- .042)0( 258**.0 .116)0( 022.0 .071)0( .293**0- .121)0( 014.0 .081)0( 047.0 ,469.902( esY No 192 14.51 coverage quality)nl( ) * * * * * * 26( 32 018 019) 17 016) 31** 073) 06** and 040) 155** 819** 239) 0.- 0.( .00 0.( .20 0.( .20 0.( 275**.0- 110) 053) 0.( 0.- 0.( 014.0- 068) 081** 033) 0.( 0.- 0.( 265**.0- 067) 0.( 102**.0- 028) 0.( 0.- 0.( esY 6 4.0 No 34 20 25)( ***8 4) 4 1) 0 )3 ***4 )4 ***4 3) ** 7* 6) ***3 3) ** 0* )3 3 )8 6 1) 0 1) ***1 4) 642 quality .071- 0.21( 0.00 0.04( 0.04 .120( .660 .110( .700- 0.11( 1.08 0.24( .580- 0.11( 1.06 .210( 0.01 .030( .080- 0.07( .030- 0.04( .251- 0.21( seY .6 No 219 200 losses, 24)( 6 0.080- 0.052)( 0.080*** 0.015)( 0.010 0.040)( .106***0- 0.036)( 084* 0. 0.044)( .253***0- 0.048)( .0070- 0.047)( 0.328***- 0.028)( 0.043- 0.048)( .234***0- 0.024)( 0.008 0.048)( 0.139*** 0.044)( .782***2- 0.200)( esY oN 582 358.6 es * * * * * ** loss 3)2( 76 0.218**- 0.087)( 488**.0 0.090)( .0580- 0.054)( 053.0 0.060)( 160**.0 0.056)( .470**0- 0.104)( 110*.0 .057)0( .483**0- .083)0( .0320- 0.066)( .127**0- 0.062)( 0.021- 0.069)( .165**0- 0.062)( 062.0 0.087)( .0340- 0.090)( .821*1- 0.134)( esY No 313 32.92 utionalbi * ) * * distributional strdi 22( 83 032) 029) 049) 142** 038) 042) 032) 032) 036 031) 400) 4 1.2 n)l( 069**.0- 0.( 071**.0- 0.( 076.0- 050).0( 102**.0- 040).0( 059.0 0.( 0.- 0.( 152**.0 0.( 042.0- 0.( 018 0. 0.( 0.- 0.( 027**.2- 0.( esY No 60 33 ity, ** ** ** * * ** ** ** 21)( 34* 55) 8* 88) 16 13) 67* 16) 5 08) 5* 15) 6* 06) 50* 35) 5* 43) 67* 82) 63 42) 9 42) 57* 75) 963 .20- 0.0( 0.48 0.0( .00- .10( 0.3- .10( 0.01 0.1( 0.29 0.1( 0.25 0.1( .20- 0.1( 0.14 .00( .50- .00( .00- 0.0( 0.04 0.0( 1.9- 0.3( seY .5 No 353 241 productiv *** 20)( 0 labor­ ee 0.054 0.083)( .0180- 0.040)( .0270- 0.070)( .162***0- 0.052)( .433***0 0.095)( .210***0- 0.073)( 0.133 0.095)( 0.025 0.053)( 0.074- 0.078)( 0.088* 0.045)( .318***0- 0.090)( 0.088 0.070)( 13.130 0.126)( esY seY 1% at 530 447.5 oylp antci em 19)( 46 per 0.009 0.143)( 0.159 0.132)( .1410- 0.191)( .542***0- 0.135)( 0.381** 0.191)( 0.145- 0.151)( .0680 .133)0( .1240 .114)0( 0.007 0.143)( .0650- 0.113)( 0.022- 0.102)( .0920- 0.098)( 0.185- 0.163)( 0.073- 0.133)( 8.520*** 0.190)( esY seY gnifis 313 267.4 *** * analysis units ) * * * * owlf 18( 23 5%;ta .0000- n)l( .052)0( 157**.0 .052)0( .0440- .064)0( .140**0- .058)0( .176**0 .057)0( .197**0- .048)0( .142**0 .049)0( 103**.0- .042)0( 003.0- .049)0( 019.0 .041)0( 256**.4 .104)0( esY seY 525 64.34 es * ** anticifn gressioneR 17)( 24 33) .00- 56 18) 07 98) 97) 60 98) 00** 23) 40) 53** 24) 22) 06) 55** 68) 37 66) 79) .1 663 nthese sig 0.1( .00 0.1( .00 .00( 12 .10 .00( 49 0.1- 0.1( 0.5- 0.1( 0.1 0.1( .20- 47 0.1( 0.0 .10( 55 0.0 .10( .10- 0.0( .00- 31* 0.0( 8.5 0.1( seY seY ** 335 297 par t dn s t t ins 10: se es c c orr 10%;ta v_trir v_posir d_pr d_pr ter er d _p _p d ton_par ton_par ton_y ton_y dur b_ _durb tion_f tion_f tion_b tion_b terr_ _rr ant EF treficepS ationsv elihoodkil mrifforeb anticif Table dum dum _bidtr pt_bi _autr pt_au _autr pt_au _rtr pt_r _natr pt_na _natr pt_na _awartr pt_awar _tartr pt_ta onstC mriF rm Fi bserO Log- muN Standar sign* Table 11: Regression analysis ­ prices (ln) average prices in dollars (ln) average prices in real currency (33) (34) (35) (36) (37) (38) (39) (40) dum_priv_tr 0.213*** 0.140*** 0.389** 0.565*** 0.165*** 0.111*** 0.300*** 0.250*** (0.035) (0.021) (0.152) (0.098) (0.029) (0.019) (0.083) (0.048) dum_priv_post -0.189*** 0.104*** -0.235*** -0.111* 0.198*** 0.130*** 0.259*** 0.169*** (0.032) (0.019) (0.056) (0.059) (0.025) (0.017) (0.036) (0.033) tr_bid 1.086*** 0.191*** 0.511*** 0.144*** (0.046) (0.065) (0.086) (0.055) pt_bid -0.340*** -0.060** -0.210*** -0.120*** (0.038) (0.029) (0.038) (0.026) tr_auton_part -0.601*** -0.738*** -0.224 -0.723*** -0.456*** 2.009*** -0.338*** (0.066) (0.079) (0.195) (0.081) (0.069) (0.129) (0.063) pt_auton_part 0.316*** -0.108*** 0.239*** -0.023 0.266*** -0.108*** 0.183** -0.223*** (0.072) (0.039) (0.070) (0.046) (0.083) (0.037) (0.079) (0.043) tr_auton_yes -1.142*** -0.301*** -0.137 -0.153** -0.382*** -0.079 0.193*** -0.010 (0.049) (0.072) (0.090) (0.063) (0.081) (0.056) (0.053) (0.038) pt_auton_yes 0.243*** -0.161*** -0.090** -0.253*** 0.133*** 0.098*** 0.019 0.063** (0.036) (0.030) (0.041) (0.037) (0.043) (0.025) (0.031) (0.029) tr_rb_dur -0.083** 0.019 -0.096*** -0.182*** (0.034) (0.105) (0.028) (0.066) pt_rb_dur 0.306*** 0.126*** -0.112*** -0.175*** (0.029) (0.027) (0.022) (0.023) tr_nation_f -0.007 0.073** -0.112 -0.182* -0.108*** -0.059** -0.020 -0.189*** (0.035) (0.036) (0.117) (0.093) (0.026) (0.025) (0.052) (0.038) pt_nation_f 0.227*** 0.024 0.156** 0.085 0.070*** 0.046** 0.070** 0.021 (0.030) (0.027) (0.064) (0.061) (0.025) (0.021) (0.031) (0.030) tr_nation_b -0.255** -0.403*** (0.109) (0.096) pt_nation_b 0.176*** 0.197*** (0.058) (0.057) tr_award_prc -0.060 -0.184** -0.052 -0.157*** (0.198) (0.078) (0.043) (0.043) pt_award_prc 0.034 0.094** -0.010 -0.076*** (0.047) (0.038) (0.027) (0.029) tr_tar_rret -0.188* 0.061 (0.113) (0.052) pt_tar_rret -0.126*** -0.060* (0.040) (0.033) Constant 3.839*** -1.227*** 4.435*** -1.183*** 4.193*** 6.614*** 4.593*** 6.715*** (0.027) (0.093) (0.095) (0.088) (0.033) (0.085) (0.056) (0.082) Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Firm Specif trend No No No No No No No No Observations 372 550 350 528 370 548 348 526 Log-likelihood 316.7 281.2 288.7 280.6 381.3 400.5 373.5 388.7 Number of firms 44 65 42 63 44 65 42 63 Standard errors in parentheses * significant at 10%; ** significant at 5%; *** significant at 1% 4.4.1 Sale Method Privatizations and concessions that were sold via an auction process experienced a reduction of 5.8 - 6.8 percent in connection numbers (depending on the regression specification) below the firm specific time trend during the transition. In contrast, no significant changes were encountered during the post-transition period. Whether or not privatizations or concessions were auctioned did not have a significant effect on output. When privatizations/concessions were 34 auctioned, there was a relative decrease in coverage of about 3 percent during the transition yet no significant changes after the transition. Privatizations and concessions that followed an auction process experienced a drop in employment of about 18-20 percent during the transition. After the transition, an additional drop of 9-17 percent occurred, depending on the econometric specification. In order to measure labor productivity, both output per employee and connections per employee were analyzed. In both cases, whether or not a utility was auctioned made little difference. Some evidence was found of a decrease in labor productivity after the transition, but this decrease was accompanied by several non-significant results, depending on the regression specification. Regardless of whether privatization processes took place through an auction, there was no effect on distributional losses during the transition period. However, after the transition, cases that were auctioned displayed distributional loss reductions between 10 and 31 percent. Finally, when companies were privatized via auctions, average prices in dollars rose during the transition and then fell after the transition. Prices in real local currency followed a similar trajectory. 4.4.2 Autonomy of Regulatory Body When the regulatory body was partially autonomous, the number of connections decreased between 3.1 and 6.9 percent during the transition yet no significant changes occurred after the transition. When the regulatory body was fully autonomous, there were no significant changes during either period. With respect to output, when the regulatory body was partially autonomous, output decreased between 5.9 and 8.2 percent during the transition; however, there were no significant changes after the transition. When the regulatory body was fully 35 autonomous, output fell by 5.3- 8.6 percent during the transition and by about 4.5 percent after the transition. Coverage appeared to decrease slightly during the transition, when the regulatory body was partially autonomous, while no significant changes were encountered after the transition. When the regulatory body was fully autonomous, output fell by between 1.6-3.5 percent during the transition; no substantial changes occurred after the transition. Privatizations that had a partially autonomous regulatory body experienced employee reductions during the transition that were between 10 and 48 percent greater than reductions experienced without an autonomous body. The analysis found some evidence of relative increases in employee numbers after the transition when the regulator was partially autonomous, but the results were not always significant. When the regulatory body was fully autonomous, employee reductions were greater than those observed under partial autonomy. During the transition, privatization processes with a fully autonomous regulator experienced employment reductions that were between 27 and 54 percent greater than cases where the regulator was not autonomous. Changes in employment after the transition were not significant. When the regulatory body was partially autonomous, connections per employee increased between 14 and 21 percent during the transition and then fell by 15 and 26 percent after the transition. Labor productivity defined as output per employee followed a similar pattern although the drop experienced after the transition was even greater, resulting between 14 and 42 percent. When the regulatory body was fully autonomous, large increases in labor productivity were experienced during the transition: connections per employee increased between 27 and 60 percent and output per employee increased between 15 and 48 percent. After the transition, the full autonomy cases experienced decreases of 10 to 35 percent in labor productivity. 36 No significant effects on distributional losses were detected during the transition when the regulatory body was partially autonomous. Mixed results were found during the post- transition period, but the majority of the specifications pointed toward reductions in distributional losses when regulators were partially autonomous. Results were quite different when the regulatory body had total autonomy: distributional losses increased between 11 and 29 percent during the transition, followed by significant losses (between 22 and 38 percent) after the transition. When regulatory bodies were partially autonomous, prices in dollars fell during the transition by 45-52 percent. After the transition, the analysis exhibits mixed results for prices in dollars. Diverse results were also found for partial autonomy in real local currency in both periods. In cases where the regulatory body was completely autonomous, significant price reductions in dollars--i.e. 26 to 68 percent--were visible during the transition. Similar to the partial autonomy cases, various results (in dollars) were found after the transition for full autonomy, although most of the regression specifications indicated a drop in prices. In real local currency, mixed results were found for the full autonomy cases during the transition and price increases of about 10 to 14 percent were found after the transition. In cases where the regulatory body was partially autonomous, the quality index fell by approximately 24 percent during the transition period and an additional 14 to 50 percent during the post-transition period. When the regulator was fully autonomous, most results pointed to large increases in quality during the transition. After the transition, the results for full autonomy were varied. 37 4.4.3 Duration of Regulatory Body Appointments With respect to the number of connections, no significant changes were observed during either period. However, privatizations/concessions that were regulated by bodies where regulators were appointed for terms of 5 or more years experienced decreases in output between 9.7 and 11.6 percent during the transition. However, there were no significant changes after the transition. Privatizations/concessions that were regulated by similar bodies appeared to experience small drops in coverage both during and after the transition. However, these results are dependent on the regression specification and some specifications for no significant results. When regulatory board appointments lasted 5 or more years, the number of employees decreased by roughly 25-30 percent during the transition. After the transition, the results were less robust, but employment appeared to fall by an additional 14 percent. Additionally, connections per employee increased between 27 and 31 percent during the transition. Results after the transition were less clear-cut, but connections per employee again seem to have increased. Changes in output per employee were not very robust, but there was some evidence of a decrease during the transition followed by an increase after the transition. These results became insignificant when additional controls were added to the regression specifications. When regulatory board appointments were of a longer duration, the results suggest that there were increases in losses between 12 and 16 percent during the transition. In the post- transition period, relatively large reductions in losses--roughly 40 percent--were observed, more than offsetting the increases seen during the transition period. When considering the impact of longer duration of regulatory body appointments on dollar prices during the transition, mixed results were found. Dollar prices rose during the post- 38 transition period while prices in real local currency fell between 9 and 16 percent during both the transition and post-transition periods. There was some evidence that quality increased when appointments to the regulatory body were of a longer duration, although not all results were significant. Changes after the transition were not significant. 4.4.4 Investor Nationality When only foreign investors were considered, the analysis yielded mixed results during the transition and non-significant results after the transition. When there was a mix of foreign and local investors, the number of connections fell by 1.0 and 2.2 percent during the transition. After the transition, the basic regression specification showed an increase of 1.4 percent in the number of connections. However, after other controls were added to the regression, the results were no longer significant, suggesting that covariance exists with some of the other variables. Additionally, output decreased between 4.4 and 10.9 percent during the transition and between 2.1 and 4.0 percent after the transition. When both foreign and local investors were involved, slight increases in output were observed before controlling for other regulatory characteristics. However, after adding other controls to the regression specification, the changes became insignificant. Coverage decreased by roughly 3 to 4 percent during the transition (relative to coverage levels with only local investors), when only foreign investors were present, while no significant changes were observed after the transition. Similar results were found were there was a mix of foreign and local investors. 39 Employee reductions were about 12 to 31 percent higher during the transition for companies with foreign investors vis-à-vis companies with no foreign investors. After the transition, no additional changes were observed. Companies with both foreign and local investors had smaller changes during the transition than those with only foreign investors but, after the transition, they experienced additional employee reductions. Furthermore, when contracts were awarded to firms with only foreign investors, the changes in labor productivity were either not clear-cut or insignificant. When both local and foreign investors were involved, labor productivity appeared to decrease during the transition, followed by an increase after the transition. Changes in both directions were generally between 6 and 10 percent. For firms with only foreign ownership, no significant changes were observed during the transition. In the post-transition period, distributional losses fell by roughly 12 to 26 percent. A similar pattern emerged for firms with both foreign and domestic ownership: no significant changes during the transition and a reduction in losses between 15 and 19 percent after the transition. When there were only foreign investors, mixed results were found in dollars during the transition, while prices increased by between 14 and 26 percent after the transition. In the case of only foreign investors and real local currency, prices fell during the transition and increased after the transition. When both foreign and local investors were involved, prices in dollars fell substantially (by roughly 23 percent) during the transition, only to recover after that. In contrast, prices in real local currency did not experience significant changes during the transition; however after the transition, there is some evidence that prices fell, but several of the regression specifications produced insignificant results. 40 Results for firms with foreign investors were not very robust, but some specifications exhibit decreases in quality both during and after the transition. For firms with both foreign and local ownership, quality increased by about 29 percent during the transition. Some specifications also found increases after the transition. 4.4.5 Award Criteria The impact of two types of award criteria was analyzed: highest price and best investment plan. Concessions awarded based on the highest price criterion experienced drops between 1.3 and 2.6 percent during the transition. There were no significant changes after the transition. On the contrary, concessions that were awarded to the bidder with the best investment plan experienced a small increase (roughly 2.5 percent) in the number of connections during the transition. Similar to the concessions awarded based on the highest price criterion, there were no significant results found after the transition for concession with the best investment plan. When privatizations/concessions were awarded based on a highest price criterion, no significant results were found. (Repetitive) When the award criterion was based on the best investment plan, no significant results were found during the transition, yet an increase in output of about 2 percent was noticed after the transition. It is worth noting that these results are those reflected after controlling for time trends. When privatizations/concessions were awarded based on a highest price criterion, coverage appeared to decrease slightly during the transition, but then increase faintly after the transition. When the award criterion was the best investment plan, the opposite occurred: coverage increase slightly during the transition but then decreased mildly after the transition. It is important to note that these are the obtained results are after controlling for time trends. 41 Privatizations/concessions awarded according to the highest price reported some reduction during the transition, but after controlling for other factors, these changes were not significant. Employment reductions were also encountered during the post-transition period. Privatizations/ concessions awarded according to the best investment plan had some relative increases in employee numbers both during and after the transition, but after controlling for other factors, these changes were less significant. Moreover, when contracts were awarded according to the best investment plan, no significant changes in labor productivity were observed. When contracts were awarded based on the highest price offer, output per employee appears to have fallen by roughly 20 percent during the transition, followed by an increase of about the same amount after the transition. Connections per employee for highest price cases seem to have increased by about 50 percent after the transition. Whether or not a contract was awarded according to the highest bid did not significantly affect performance during the transition. After the transition, distributional losses seem to have increased slightly in highest bid cases. In cases where the winner was determined by the best investment plan, there were no significant results. When contracts were awarded according to the highest price criteria, prices in dollars appeared to drop during the transition but then increase after the transition. However, different regression specifications produced somewhat mixed results. In real local currency, various results were found during the transition, followed by a drop after the transition. When contracts were awarded based on the best investment plan, no significant results were found in dollars during the transition, whereas an increase of 14 percent was observed after the transition. When the case of the best investment plan criteria is considered in real local currency, prices appear to 42 have fallen during the transition. However, no clear results in local currency were found after the transition. For the case in which contracts were awarded based on a highest price criterion, various results (some positive, some negative) on quality were found during the transition. After the transition, some results pointed to a decrease in quality, with others were not significant. No significant results in quality were exposed after the transition. When contracts were awarded based on a best investment plan criterion, quality seems to have decreased by about 18 percent during the transition, while no significant changes were found for the post-transition period. 4.4.6 Tariff Regulation In order to identify the effect of the type of tariff regulation on network expansion, both "rate of return" and "price cap" regulation were analyzed. Concessions regulated according to rate of return experienced an increase in the number of connections between 2.4 and 6.3 percent during the transition. After the transition, the number of connections increased an additional 2 percent. Concessions subject to price cap tariff regulation do not appear to have experienced significant changes, although a parallel analysis of changes in growth rates indicated a decrease in the number of connections during the transition. When analyzing output, no significant changes were found when tariffs where regulated according to rate of return. Price cap regulation yielded an increase in output of about 5 percent during the transition yet no significant changes after the transition. While, small increases in coverage were seen under rate-of-return tariff regulation during the transition, no significant changes were observed for price cap tariff regulation schemes. 43 Utilities subject to rate of return regulation showed large relative employee reductions-- roughly 60 percent--during the transition in some (but not all) of the econometric specifications. Relative employment increases were identified in some specifications after the transition, but one of the specifications showed a slight decrease. Firms regulated under price cap systems experienced some employee reductions during the transition; however these changes were not significant after controlling for other factors. When tariffs for privatizations/concessions were regulated according to a rate-of-return system, there is some evidence that labor productivity increased during the transition. Nevertheless, no significant results were encountered in labor productivity after the transition. While, under price cap regulation, labor productivity appears to have increased during the transition, no significant changes were observed after the transition. When tariffs were regulated according to rate of return, there were no significant changes during the transition, and distributional losses fell somewhat following the transition. However, under price cap tariff regulation, distributional losses seem to have increased after the transition. When tariffs were regulated according to a price cap methodology, no significant results were found in dollars. However, in real local currency, prices increased during both periods. When rate-of-return tariff regulation was implemented, prices in dollars first increased during the transition but then decreased after the transition. Prices in real local currency showed mixed results. When tariffs were regulated according to a rate of return method, mixed results were found during the transition. After the transition, one regression specification showed quality reductions, but these became insignificant once more controls were added to the regression. 44 4.5 Conclusions and policy recommendations We have provided a brief overview of infrastructure reforms in Latin America during the 1990s, and analyzed the determinants of performance, with a focus of the regulatory factor. A number of main messages emerge. First, private participation generated important improvements, but they were mostly concentrated in the transition period (around the privatization event). Second, only a share of those benefits was transferred to consumers. And third, significant performance heterogeneity within and among sectors may be explained by intrinsic characteristics of the reform process, such as the privatization mechanism, the level of regulatory development, and the concession design. In particular the analysis shows that: (i) Generally autonomous regulatory bodies seem to be correlated with greater reductions in the number of employees, while older-through experience and capacity- institutions result in lower price increases. (ii) When pricing is regulated according to the rate of return, companies have higher network expansion than in the case of price-cap regulation. Consistently, those firms under price-cap regulation experience higher reductions of their labor force, but lower increases in labor productivity. Additionally, the latter firms present less improvement in both distributional losses and quality, while also exhibiting higher price increases when compared to those under the rate-of-return regulation. These results suggest one main policy implication: change in ownership has significant effects in improving efficiency and quality. However, regulatory structure, framework and quality are important determinant of sector performance. 45 Additionally, there is a need to complete the reforms, particularly the so-called "second generation regulatory reforms." Without these reforms ­ that include the completion and improvement of the regulatory framework, setting up safeguards and procedures to dissuade excessive-frivoulos-contract renegotiations, and increasing competition when feasible ­ post- privatization improvements are limited and probably unsustainable. Likewise without a transparent and predictable regulatory framework, needed private financing will be difficult to secure. Obviously, the importance of competition, regulation, and contract design is closely related to technological characteristics within an industry. For example, technological developments in the telecommunications sector, facilitates the emergence and feasibility of effective competition and substitution, providing alternatives to the service, by means other than fixed telephony. But that requires using regulation and antitrust, according to jurisdiction, as a tool to control for abuse of dominance by incumbents. In water and sanitation, remaining natural monopolies make the move towards increased competition a more difficult task. This implies relying more on well-designed concession contracts with regulation as a tool to guarantee the appropriate compliance and efficient (albeit second best) performance. In either case, regulation is a key instrument, especially if one needs to reduce regulatory risks, attract private investments to support the Latin American needs in infrastructure and capture a larger share of the efficiency gains generated by the private participation. 5. Final Remarks We have tested the impact of regulation of private infrastructure operators on sector performance, from three separate angles. We have found that: 46 · Quality of regulation is a significant determinant of the divergence between the overall profitability of the concession and its corresponding hurdle rate, explaining around 20 percent of the variation. However, regulatory efforts seem to be more closely associated with keeping tariffs as low as possible for current consumers, than keeping profitability well aligned with hurdle rates of return. · Price caps lead a significant increase of the probability of renegotiation. This phenomenon brings the convergence of rate of return regulation and price cap regulation even closer, because the outcomes of the renegotiation process often include increasing the number of cost components with an automatic pass-through to tariffs, toward a hybrid system. · Existence of a regulator at the signing of contract reduces renegotiations. Comparing three specific contracts out of the initial sample, and using the probabilities predicted by the empirical model, we show that had a regulator been in place at the time of awarding the contract, the respective probabilities of renegotiation in the last year of existence of the contract would have been reduced from 29.7, 9.9, and 3.1 percent, to 5.3, 0.3, and 0.2 percent respectively. Also autonomy/independence of regulator matters. When the regulator does not belong to a ministry, there is a significantly lower probability of government-led renegotiation. In that regard, these firm-level results confirm some cross-country studies results that show the importance of experienced and independent regulators in the telecommunication and electricity sectors · The regulators filter and dissuade opportunistic private operator led renegotiation and in the case of government-led renegotiation, the regulator acts as barrier against political opportunism. Regulation attempts to protect investors and ultimately consumers from the opportunistic behavior of the government. Ideally, the regulator's objective should be to maintain 47 alignment between a company's rate of return and its cost of capital. However, the closeness of such an alignment depends on the structure and institutionality of the regulatory framework and of course on quality of the regulator. · Impact of the regulator is stronger in weak governance environments. In those contexts, the regulator-protected by its autonomous characteristics- can play a key role in informing civil society and with actions aimed at dissuading opportunistic behavior, by governments or operators, championing transparency and rallying stakeholders for compliance with regulatory framework and to raise the costs of opportunistic behavior. Also, as regulators often have a role in the design of contracts, they can supply part of the missing governance, assisting in the design of better contracts from the start, which reduces the necessity of posterior adjustments for unforeseen contingencies or opportunistic behavior. · Differences in the outcomes of infrastructure services provided by private sector participation are explained to some extent by differences in the institutionality, characteristics and quality of the regulatory framework, such as autonomy, type of price regulation, and structure of the regulatory board and so on. In summary, we have shown that as the theory indicates, regulation matters. The empirical work here reported has shown that on three relevant economic aspects: aligning costs and tariffs, dissuading renegotiations and improving productivity, quality of service, coverage and tariffs--the structure, institutions and procedures of regulation matters. Thus significant efforts should continue to be made to improve the structure, quality and institutionality of regulation. Regulation matters in protecting both consumers and investors and in aligning closely 48 financial returns and the costs of capital and in capturing higher levels of benefits from the provision of infrastructure services by the private sector. 49 References Andres, L., J. L. Guasch, and V. Foster (2006), The Impact of Privatization on Firms in the Infrastructure Sector in Latin America Countries, Washington, D.C.: The World Bank. Angrist, J. (1991), `Instrumental Variables Estimation of Average Treatment Effects in Econometrics and Epidemiology', National Bureau of Economic Research, Technical Working Paper Number 115. Boardman, A. and A. R. Vining (1989), `Ownership and Performance in Competitive Environments: A Comparison of the Performance of Private, Mixed, and State- Owned Enterprises', Journal of Law Economics, 32, pp. 1-33. Chong, A. and F. López-de-Silanes (2003), `The Truth about Privatization in Latin America', Inter-American Development Bank, Latin American Research Network, Research Network Working Paper #R-486, October 2003. Cubbin, J. and J. Stern (2005), `Regulatory Effectiveness: The Impact of Regulation and Regulatory Governance Arrangements on Electricity Industry Outcomes', World Bank Policy Research Working Paper 3536, Washington D.C. Estache A., J. L. Guasch, and L. Trujillo (2003), `Price Caps, Efficiency Payoffs, and Infrastructure Contract Renegotiation in Latin America', World Bank Policy Research Working Paper 3129, Washington D.C. Gomez-Ibanez, J. A. (2003), Regulatory Infrastructure: Monopoly, Contracts and Discretion, The Harvard University Press, Cambridge, Mass. 50 Guasch, J. L. (2004), Granting and Renegotiating Infrastructure Concessions ­ Doing it Right, Washington, D.C.: The World Bank. Guasch, J.L., J.J. Laffont, and S. Straub (2003), `Renegotiation of Concession Contracts in Latin America', World Bank Policy Research Working Paper 3011, Washington D.C. Guasch, J.L., J.J. Laffont, and S. Straub (2005), `Concessions of Infrastructure in Latin America: Government-led Renegotiations', World Bank Policy Research Working Paper 3749, Washington D.C. Guasch, J.L. and P. Spiller (1999), Managing the Regulatory Process: Design, Concepts, Issues, and the Latin America and Caribbean Story, World Bank, Latin American and Caribbean Studies, The World Bank. Han, A.K. and J. Hausman, (1990), `Flexible Parametric Estimation of Duration and Competing Risk Models', Journal of Applied Econometrics, 5, 1-28. Heckman, J. and R. Robb (1985), `Alternative Methods of Evaluating the Impact of Interventions', In J. Heckman and B. Singer (eds), Longitudinal Analysis of Labmy Market Data. New York: Cambridge University Press. 156-245. Jamasb, T., Mota, R., Newberry, D., and Pollitt, M. (2005), `Electricity Reform in Developing Countries: A Survey of Empirical Evidence on Determinants and Performance', World Bank Policy Research Working Paper No. 3549. Kikeri, S. (1999), `Privatization and Labor: What Happens to Workers When Governments Divest?' World Bank Technical Paper 396. Washington, DC, United States: World Bank. 51 Megginson, W., R. Nash, and M. van Randenborgh (1994), `The Financial and Operating Performance of Newly Privatized Firms: An International Empirical Analysis', Journal of Finance XLIX: 403-452. Megginson, W. and J. Netter (2001), `From State to Market: A Survey of Empirical Studies on Privatization', Journal of Economic Literature 39: 321-389. Minogue, Martin and Ledivina Carino (eds.) (2006), Regulatory Governance in Developing Countries, Edward Elgar Press, Cheltenham UK-Northhampton Ma, USA Rivers, D. and Q. Vuong (1988), `Limited Information Estimators and Exogeneity Tests for Simultaneous Probit Models', Journal of Econometrics, 39, 347-366. Sirtaine, S., M. E. Pinglo, J. L. Guasch, and V. Foster (2005), `How Profitable Are Private Infrastructure Concessions in Latin America? Empirical Evidence and Regulatory Implications', The Quarterly Review of Economics and Finance 45:380-402. Stern, J. and Cubbin J.S. (2004), `Regulatory Effectiveness: The Impact of Regulation and Regulatory Governance Arrangements on Electricity Outcomes ­ A Review Paper', World Bank Policy Research Working Paper No. 3535. Wallsten, S. (2001), `An Econometric Analysis of Telecom Competition, Privatization, and Regulation in Africa and Latin America', The Journal of Industrial Economics, 49(1), 1-19. 52 Notes 1In Brazil, for example, dissatisfaction with privatization has increased from 40 to 60 percent of the population during 1998-2004 while in smaller countries, such as Guatemala and Panama, this index reaches more than 80 percent of the population. Even in Chile, commonly seen as the champion of structural reforms, dissatisfaction is predominant (see Latinobarómetro surveys for 1998 and 2004). Indeed, public authorities and multilateral institutions, such as the IMF and the World Bank, once sponsors of privatization, are now discussing ways of increasing public investments in infrastructure without jeopardizing sound fiscal management. The policy-making pendulum is, then, back to public investments as either if infrastructure reforms and privatization had never been implemented or, even worse, if reforms were fully completed, all lessons had been taken, and adjustments had been made. 2Guasch (2004) shows that the incidence of renegotiation is about 42 percent of all concessions and about 55 and 75 percent for concessions in the transport and water sectors. And the incidence is even much higher for concessions regulated under a price-cap regime. Even more striking is how fast those renegotiations take place. The time interval between the granting of the concessions and renegotiation is about 2.1 years, and for water concessions is even quicker, about 1.6 years. 3 Simple differential 1 excludes terminal value, Simple differential 2 includes terminal value, Simple differential 3 includes terminal value and adjustment for management fee, Simple Differential 4 includes terminal value and adjustments for management fee and transfer pricing. 4One weakness of regulatory commissions, perhaps captured here in these estimates, is the higher political intervention, since often each relevant political party gets to designate its own commissioner. 53 5Bid, Direct adjudication, invitation, petition or request. 6Highest cannon, highest price, tariff, lowest government subsidy, investment plan, shorter duration of the concession or multiple criteria. 7Operation, BOT, BOO, privatization, etc. 8Law, decree, contract or license. 9Revenue cap, price cap, rate of return or no regulation. 10The government, the concessionaire, both or nobody. 54