THE JOURNAL OF INDUSTRIAL ECONOMICS Volume XLIX March 2001

0022-1821 No. 1

AN ECONOMETRIC ANALYSIS OF TELECOM COMPETITION, PRIVATIZATION, AND REGULATION IN AFRICA AND LATIN AMERICA*
Scott J. Wallsten{
This paper explores the eects of privatization, competition, and regulation on telecommunications performance in 30 African and Latin American countries from 1984 through 1997. Fixed-eects regressions reveal that competitionmeasured by mobile operators not owned by the incumbentis correlated with increases in the per capita number of mainlines, payphones, and connection capacity, and with decreases in the price of local calls. Privatization combined with an independent regulator is positively correlated with telecom performance measures. Privatization alone, however, is associated with few benets, and is negatively correlated with connection capacity.

i.

introduction

Telecommunications markets in developing countries have been undergoing dramatic reforms since the 1980s. Spurred by changes in technology, the abysmal performance of incumbent providers, and prodding by The World Bank and other international organizations, developing countries are privatizing state-owned telecom providers, opening portions of their telecom markets to competition, and building regulatory institutions. We have a good theoretical understanding of the potential eects of privatization and competition and of the importance of regulation when privatizing a monopoly provider.1 Our empirical knowledge of their eects in developing countries is much less comprehensive. Indeed, the empirical work to date consists primarily of case studies and non-econometric comparisons of telecom performance before and after privatization. These studies have provided important insights into reform eorts, but we have little econometric evidence of the eects of com-

* I am grateful to David Genesove, Luke Haggarty, Roger Noll, David Sappington, Mary Shirley, and two anonymous referees for comments. I also thank Frew Gebreab for research assistance. I, of course, am solely responsible for all errors. { Author's aliation: The World Bank and Stanford Institute for Economic Policy Research, 579 Serra Mall at Galvez St., Stanford, CA 94305, USA. email: Wallsten@stanford.edu 1 See, for example, Laont et al. [1997], The Economic Report of the President [1996], Wellenius et al. [1992], Noll [1987].
Blackwell Publishers Ltd. 2001, 108 Cowley Road, Oxford OX4 1JF, UK, and 350 Main Street, Malden, MA 02148, USA.

scott j. wallsten

petition, privatization, or regulatory changes.2 This gap exists largely because reforms in developing countries have occurred relatively recently, meaning that only now is enough data available to begin econometric analysis. This paper undertakes an econometric analysis of the eects of telecommunications reforms in developing countries. Using an original panel dataset of 30 countries in Africa and Latin America from 1984 through 1997, the paper explores the eects on telecommunications performance of privatization, competition (as measured by the number of mobile operators in the country not owned by the incumbent), and regulation. I nd that competition is signicantly associated with increases in the per capita number of telephone mainlines, payphones, and connection capacity, and with decreases in the price of a local call. Privatization by itself is signicantly associated with decreased connection capacity, and positively correlated only with payphones. Privatization combined with the existence of a separate regulator, however, is signicantly associated with increases in connection capacity and labor eciency (as measured by employees per mainline), and mainlines per capita. Interpreted causally, these ndings are broadly consistent with conventional wisdom: competition is the most eective agent of change, and privatizing a monopoly without concurrent regulatory reforms may not necessarily improve service. One important policy implication of these results is that granting exclusivity periods to an incumbent may seriously delay the real benets that competition can bring.

ii.

telecommunication reforms

Vigorous competition existed in telecommunications markets around the world in the nineteenth century [Petrazzini, 1996]. Nonetheless, most countries soon embraced the notion that telecom service was a natural monopolythat it could be provided at the lowest cost by one rm. Most developing countries nationalized telecommunications services in the 1960s. By the late 1980s, however, these countries began to reform their telecom sectors in response to three interrelated factors: changes in technology making less tenable the argument that telecommunications are a natural monopoly, the abysmal performance of state-owned telecom providers, and pressure by The World Bank and other international organizations. The notion that telecom was a natural monopoly began to appear less tenable as early as 1959, when the U.S. Federal Communications Commission decided to allow large rms to use microwave transmission to
2 Ros [1999], discussed in more detail below, provides one of the rst econometric analyses of privatization, though he does not focus specically on developing countries.

Blackwell Publishers Ltd. 2001.

econometric analysis of telecom privatization

bypass the telephone network [Crandall and Waverman, 1995]. Continuous improvements in technology make it increasingly unlikely that telecom is a natural monopoly [Spiller and Cardilli, 1997; Noll, 2000]. Advances in wireless technology alone allow competing rms to roll out telecommunications services with relatively low sunk costsan attractive option in many developing countries. By the 1980s, it was clear that nationalized monopoly telecommunications rms in developing countries could not eectively provide telecom services. In 1981, Africa boasted an average of only 0.8 telephones per hundred people, and Latin America only 5.5, compared to 83.7 in the United States [Saunders et al.., 1983]. Potential customers, meanwhile, faced long waiting periods before being connected to the network, and many large rms attempted to bypass the monopoly provider by building their own facilities [Wellenius et al.., 1992]. This poor performance helped generate pressure on governments to reform their telecom sectors. Advisers typically recommend three components to telecommunication reforms: privatizing the state-owned monopoly provider, introducing competition and creating an independent regulatory agency. This description is a gross oversimplication, of course. Each component can be implemented in a variety of ways and to greater or lesser extents. For example, privatization is rarely complete. The government often retains partial ownership of the incumbent, at least initially. The government may also give the newly-privatized rm a temporary monopoly by prohibiting competition in order to entice investors. Finally, regulation can take many forms and its details can have large impacts on sector performance and the ability of the incumbent to exercise market power. In general, there is broad agreement that competition is likely to be the most eective method of promoting improvements in the telecom sector. A monopoly provider, whether state-owned or private, faces fewer incentives to improve service and lower prices than do rms operating in a competitive environment. Most agree that while privatization can bring about great improvements, it must be combined with eective regulation. As Ambrose et al.. [1990] note, `simply moving a monopoly from the public to the private sphere will not result in competitive behavior.' The existing literature on reforms in developing countries primarily contains two types of analyses: case studies and empirical work that compares average performance indicators across rms or countries before and after reforms took eect. Not surprisingly, given the region's relatively early start in reforms, most of this evidence is from Latin America. In general, these studies nd positive eects of reforms. On the eects of privatization, Wellenius et al.. [1992] concluded from their case studies that early results of privatization were encouraging, with rms in some countries reporting good nancial performance and increased investment from internal funds and international capital markets. Petrazzini and
Blackwell Publishers Ltd. 2001.

scott j. wallsten

Clark [1996] study the eects of competition in Latin America and Asia. Using the existence of cellular rms as evidence of competition, they compare the performance of competitive and noncompetitive markets. They nd that cellular and mainline penetration in competitive markets is higher than in noncompetitive markets.3 Wellenius et al.. [1992] sound a note of caution from their case studies, though. While many countries privatized their telecom rms quickly, they have built up regulatory capacity much more slowly. Galal et al.. [1995] adopt a hybrid approach to study the eects of regulatory reforms. They compared the performance of the telecom sector in several countries before and after regulatory reforms. They attempt to explore how well countries were able to balance regulatory objectives: commitment, information asymmetry and pricing issues. They nd that the country in their sample (Chile) that resolved all three issues achieved the greatest improvement, while the country (the Philippines) that did not, experienced the worst performance. Countries that resolved some issues experienced mixed success. The existing literature has given us important information on which aspects of telecom reforms seem to succeed and which are more problematic. But this work needs to be complemented by econometric studies that can more rigorously explore the eects of telecom reform and control for other factors that may aect performance. For example, Boubraki and Cosset [1998], in a study of 79 rms (in several industries) in 21 developing countries nd signicant productivity improvements after privatization. But they also note that privatization appears most successful in wealthier countries. This observation highlights the need for econometric analysisin this case to control for income. While it may be the case that privatization is most successful in wealthier countries, it may also be the case that rms in general in wealthier countries experienced improved productivity during the few years studied. The point here is that it is important to conduct econometric analyses to complement existing work and to begin to address such lingering issues. Enough time has elapsed from the start of reforms to make such analyses possible. One recent study undertakes such an analysis, although not specically in developing countries. Ros [1999] uses a xed-eects model to explore the eects of privatization and competition. He nds that countries that privatize their telecom providers have more telephone mainlines per capita than countries that do not privatize. Competition is not correlated with network expansion, but is correlated with eciency in terms of lines per
More research has been done on privatization in industrialized countries. Megginson et al.. [1994] compare pre- and post-privatization nancial and operating performance of 61 companies (in 32 industries, including telecommunications) from 18 countries. They nd increased sales, prots, investments, and employment following privatization.
Blackwell Publishers Ltd. 2001.
3

econometric analysis of telecom privatization

employee. I extend this analysis in several ways. First, the competition indicator in Ros [1999] is whether the country ocially permits competition in either local, long distance, or international telecom service. The problem with this measure is that a country may ocially allow competition while the incumbent provider manages to block actual competition. I deal with this issue by using a measure of actual competition. Second, that study does not deal with the question of regulation. I begin empirically to address the question of regulation by including a variable, measuring whether the country has established an independent regulator and interacting that variable with reform indicators. Finally, Ros [1999] does not focus specically on developing countries, which face dierent problems in reforming their telecom sectors.4 Most industrialized countries, for example, already have a substantial regulatory infrastructure when undertaking reforms, while developing countries often do not. In the remainder of the paper I explore empirically the eects of privatization, competition, and regulatory changes on telecommunications performance. In the following sections I describe the data, methodology, and econometric results.

iii.

data

The dataset contains information on 30 countries in Africa and Latin America from 1984^1997 (Table I lists the countries, which were chosen on the basis of the availability of telecom reform data). An observation, therefore, is a country-year. Not all data exist for all years for all countries, meaning that the sample size diers depending on the indicator used in a particular analysis. Nonetheless, the data allow us to begin exploring empirically the eects of liberalization. Specically, using this unbalanced panel, we can investigate the relationships between competition, privatization, having an independent regulatory agency and telecommunication performance measures. Primary telecommunication indicators include the number of mainlines per capita, the number of payphones per capita, network connection capacity per capita, telecom employees per main line, and price of a threeminute local call. These statistics come from the International Telecommunications Union (ITU). Table II shows trends in these indicators by region (calculated from the countries in the sample). Main line penetration (main lines per capita, or teledensity) is the most
4 In addition, the paper relies almost entirely on ITU data, which collects primarily telecom variables. As a result, the only arguably exogenous variables are GDP and country and year dummy variables. Other control variables include telecom investment, telephone connection prices, access charges, and the price of a three-minute local phone call. All of these variables are likely to be aected by telecom reforms and cannot easily be considered exogenous.

Blackwell Publishers Ltd. 2001.

scott j. wallsten
Table I Countries in Telecommunications Database Africa Botswana Cameroon Cote d'Ivoire Ghana Kenya Malawi Morocco Mauritius Mozambique Nigeria Senegal Tanzania Uganda South Africa Zambia Latin America Argentina Brazil Bolivia Chile Costa Rica Colombia Dominican Republic Ecuador Guatemala Jamaica Mexico Panama Paraguay Peru Uruguay Venezuela

common indicator of telecom performance as it is the variable most commonly reported to the ITU. A main line `is a telephone line connecting the subscriber's terminal equipment to the public switched network and which has a dedicated port in the telephone exchange equipment' [ITU, 1998c]. This variable provides an indication of the level of penetration of telephone service in the country. In some ways, the number of mainlines per capita overstates access to phones and in other ways, understates it. It may overstate access since some people have several linesone at home and two or more at work. It may understate access since some lines are for public phones or call centers where many people use one line. Nonetheless, it remains the most widely available indicator of telephone penetration. The number of payphones per capita provides a measure of the extent of universal service. While this number does not tell us what percentage of the poor or rural population actually has access to a phone, a payphone is, in principle, useful to anyone who lives close to it. Access to a payphone is probably a better indicator of universal service for the poorest countries than is the percentage of people with telephones in the home.5 Connection capacity, `corresponds to the maximum number of main lines which can be connected [and] includes . . . mainlines already connected and mainlines available for future connection' [ITU 1998c]. Capacity tends to grow at about the same rate as connected mainlines, with capacity remaining slightly ahead of connections. The number of

5 A good indicator of universal service may be the percentage of the population within walking distance of a public phone.

Blackwell Publishers Ltd. 2001.

 Blackwell Publishers Ltd. 2001.

Table II Telecom Statistics for Latin American and Africa (excluding South Africa) Mainlines per hundred population Year 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 Average annual percent change Latin America 4.56 4.95 5.22 5.54 5.86 6.16 6.66 7.14 7.77 8.50 9.19 9.91 10.86 11.67 7.51 Africa 0.74 0.82 0.88 0.94 1.00 1.08 1.17 1.30 1.48 1.72 1.95 2.15 2.45 2.94 11.25 Payphones per hundred population Latin America 0.06 0.06 0.07 0.08 0.09 0.09 0.09 0.09 0.10 0.11 0.12 0.14 0.15 0.18 9.40 Africa 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.02 0.03 0.05 19.57 Connection capacity per hundred population Latin America 5.83 5.77 5.92 6.85 7.61 7.16 8.23 8.54 9.17 10.07 10.78 12.47 13.49 15.34 7.93 Africa 1.16 1.11 1.46 1.48 1.29 1.58 1.55 1.86 1.97 2.26 2.40 2.66 3.36 4.14 11.01 Employees per line Latin America 0.03 0.03 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.01 0.01 0.01 0.01 7.97 Africa 0.08 0.07 0.07 0.07 0.07 0.06 0.06 0.05 0.05 0.05 0.04 0.04 0.04 0.03 7.31 Price of 3-min local call (USD) Latin America Africa

econometric analysis of telecom privatization

0.03 0.05 0.06 0.06 0.05 0.05 0.04 0.05 0.05 0.05 0.07 0.07 11.20

0.16 0.11 0.11 0.09 0.07 0.06 0.09 0.09 0.07 0.08 0.09 0.06 5.58

Source: International Telecommunications Union Database, 1999.

scott j. wallsten

employees per main line provides a measure of labor eciency. It is not clear how benecial reforms would aect this indicator. While privatization and competition should increase eciency, benecial reforms could have two, opposing, impacts. Reforms may induce the incumbent either to eliminate excess sta or to increase employment as it improves its network. Thus, the expected results of reforms on this indicator are indeterminate. The price indicator is the price `of a peak rate three-minute call within the same exchange area using the subscriber's own terminal (i.e., not from a public telephone)' [ITU 1998c]. Price data are the least commonly reported of the telecom indicators and the most problematic to analyze. Even if the reported prices are accurate, they may not be comparable across countries. The size of a local exchange area may dier dramatically. Prices are often subject to extensive cross-subsidies. Nonetheless, it is worthwhile to begin exploring the eects of reforms on prices, since price will be among the consumer's primary concerns once connected to the network. Again, it is unclear how benecial reforms would aect prices. On one hand, competition should reduce prices. On the other hand, residential prices are often subsidized. Reforms often include eorts to end cross-subsidies, which could increase consumer prices. The privatization, competition, and regulatory variables come from ITU publications, Economist Intelligence Unit (EIU) publications, and the U.S. Department of Commerce. Privatization is a dummy variable that equals one beginning the year the rm was privatized, if ever. This variable does not capture the extent of privatization; it simply indicates whether the government sold part of the rm. Competition is measured by the number of wireless operators in the country not owned by the incumbent. This is not a perfect measure of competition. For example, cellular service is eectively available only to the wealthy in very poor countries. The least expensive cellular plan oered in 1999 by Telcell (one of three mobile operators) in Ghana, for example, was $25 per month, plus $0.50 per minute after the rst ten minutes [Telcell 1999]. Even if the user pays only the base charge his bill is $300 per yearquite steep in a country with an average per capita income of less than $400. Nonetheless, mobile operators oer benchmark comparisons of the incumbent and are potential threats to the incumbent since they can increase penetration swiftly at relatively low cost per additional subscriber. In this capacity they could spur the incumbent to improve its services. The regulatory measure is a dummy variable indicating whether the country has a separate telecommunications regulatory agency not directly under the control of a ministry. This information comes from ITU publications. Having a separate regulator is generally associated with attempts at regulatory reform, and its eect in a regression analysis is probably better characterized as indicating a country's propensity to undertake regulatory reforms rather than the eect of a separate regulator, per se.
Blackwell Publishers Ltd. 2001.

econometric analysis of telecom privatization

Control variables include per capita income, population, percent of the population living in an urban area, a dummy variable indicating whether the country passed telecom reform legislation, a dummy variable indicating whether a World Bank telecommunications project was active in the country-year, net World Bank aid as a percent of GDP, exports as a share of GDP, and a variable measuring the risk of expropriation. The macroeconomic and demographic variables come from The World Bank's Statistical Information and Management Analysis (SIMA) database. The risk variable comes from the International Country Risk Guide (ICRG), and is a 10-point scale where 10 indicates the lowest risk of expropriation.6 Table III presents summary statistics for the independent telecom variables for Africa and Latin America. (Summary statistics for nontelecom independent variables are available on the Journal's editorial website). The table demonstrates that in general, countries in Latin America undertook reforms earlier than in Africa. Indeed, Chile passed reform legislation in 1982 while in Africa, Nigeria was the rst country to pass reform legislation, in 1992.7 On average, Latin American nations undertook privatizations earlier than African nations. Mobile competition, too, tends to be more intense in Latin America. Table IV shows correlations between the reform variables. The tables demonstrates, for example, that of the countries that passed reform legislation, 50 percent privatized the incumbent telecom provider, about 80 percent introduced competition and 85 percent created a separate regulator. On the contrary, of the countries that did not pass reform legislation, none privatized the incumbent rm, 20 percent introduced competition and 40 percent had a separate regulator. Among the countries that introduced competition, nearly half also privatized the incumbent, more than 90 percent had passed reform legislation and the same share had a separate regulator. Among the countries that did not see competition, only 25 percent privatized the incumbent, 63 percent had passed reform legislation and 38 percent had a separate regulator.

6 The ICRG variable goes only through 1995. Because this variable exhibits little variation over time, I assume that it remains constant from 1995 through 1997. This assumption adds error and may bias coecients on this variable in a regression. 7 In Chile, Ley General de Telecomunicaciones (Ley no. 18.168) `Establishes inter alia: responsibilities with respect to telecommunication services; compulsory interconnection between public service licensees; and machinery for setting telecommunication service taris where existing market conditions are not sucient to guarantee a free tari system' [ITU 1998]. Decree 75 of 1992 in Nigeria liberalized the telecommunications industry and established the Nigerian Communications Commission [ITU, 1998].

Blackwell Publishers Ltd. 2001.

 Blackwell Publishers Ltd. 2001.

10

Table III Summary Statistics for Telecom Independent Variables Latin America Year Year reform independent legislation regulator passed established 1990 1993 1982 ^ 1991 1994 1994 1989 1991 ^ 1990 1995 1991 1995 1987 1996 1995 ^ 1996 1994 1995 1996 1996 1994 ^ 1991 1997 1995 1996 ^ Year incumbent privatized 1990 1995 1987 ^ ^ ^ ^ 1990 1994 ^ 1991 ^ ^ 1997 ^ Number mobile competitors by 1997 3 0 7 0 6 2 1 7 1 1 1 0 1 1 2 Africa Year Year reform independent legislation regulator passed established ^ ^ 1994 1994 1994 ^ 1996 ^ 1992 1996 1994 1994 1994 1993 ^ ^ 1996 1997 ^ ^ 1997 1992 1992 ^ 1994 ^ 1994 1997 Year incumbent privatized ^ ^ 1997 1997 ^ ^ ^ ^ ^ ^ ^ ^ ^ 1997 Number mobile competitors by 1997

Country Argentina Bolivia Chile Costa Rica Colombia Ecuador Guatemala Mexico Peru Uruguay Venezuela Brazil Paraguay Panama Dominican Republic

Country Botswana Cameroon Cote d'Ivoire Ghana Kenya Malawi Morocco Mozambique Nigeria Senegal Tanzania Uganda Zambia South Africa

scott j. wallsten

0 0 3 1 0 0 0 0 0 0 1 0 1 2

Sources: ITU (1998a,b,c); Pyramid Research (1997); U.S. Department of Commerce (1996).

econometric analysis of telecom privatization

Table IV Reform correlations: Share of countries with particular reforms that also adopt other reforms Countries with that also have Reform Legislation ^ 91.7 100.0 91.3 ^ 57.1 72.2 62.5 Separate Regulator 84.6 ^ 84.6 91.3 40.0 ^ 72.2 37.5

11

Privatized Incumbent Competition 50.0 45.8 ^ 47.8 0.0 30.0 ^ 25.0 80.8 87.5 84.6 ^ 20.0 30.0 66.7 ^

Reform Legislation Separate Regulator Privatized incumbent Competition No reform legislation No separate regulator Not privatized No competition

iv.

methodology

Because countries dier in so many ways, I employ a xed-eects model to control for unobserved country-specic factors. I also include year dummies to control for time trends.8 I rst estimate equation (1) to get a preliminary look at the eects of competition, privatization, and regulation. 1 yit ai gt b1 cellit b2 privateit dRegit yXit eit

I estimate this equation ve times, using as the dependent variable the dierent telecom indicators discussed above (mainlines per capita, payphones per capita, network connection capacity per capita, employees per mainline and the cost of a three-minute call). Cellit is the number of mobile operators not owned by the incumbent and privateit is a dummy variable indicating whether the incumbent was privatized. Regit is the dummy variable indicating the existence of a separate regulator. Xit is a vector of control variables described above. As discussed above, the regulatory dummy variable cannot capture the many aspects of regulation that aect telecom performance. This simple variable, however, can provide us with information beyond that derived from estimating equation (1). Theory suggests that simply privatizing a monopoly may not generate telecom improvements. Careful regulation is required to encourage a monopoly to improve its performance. To explore further the eects of regulation, I interact the regulation dummy with the privatization dummy and with the number of competitors, and estimate equation (2).
8 I correct the standard errors for serial autocorrelation using the Newey-West [1987] method as described in Greene [1993, pp. 422^423].

Blackwell Publishers Ltd. 2001.

12 2

scott j. wallsten yit ai gt b1 cellit b2 privateit b3 cellit regit b4 privateit regit dRegit yXit eit

Equation (2) allows us to explore separately the eects of competition, privatization, regulation and how they interact. That is, how do competition and regulation together, and privatization and regulation together, correlate with telecommunication sector performance? This is an especially important question given that theory suggests, and most policy makers and advisers agree, that reforms must give careful thought to regulation when privatizing the monopoly provider. While these data are the best currently available for these developing countries, the data and the methods suer from some shortcomings. The biggest potential problem is that competition, privatization and regulation may be endogenous to reforms. That is, reforms aect telecom performance, but performance may also aect reforms. Likewise, the same unobserved factors aecting reform may also aect performance. The analysis deals with some of these issues by including country and year xed eects and a variable indicating whether the country passed reform legislation. Country xed eects control for a country-specic propensity to reform, while year xed eects control for general trends of changes in telecom service. The reform dummy variable, too, helps control for a propensity to reform, which could be correlated with performance changes. Another problem is that the dummy variable for an independent regulator is an oversimplication of telecom regulation. Ideally, we would like much more detailed data on regulation. Such details would include the type of regulation (e.g., price caps or cost-of-service) and regulatory institutions (e.g., annual budget, employees, employee training and experience, enforcement powers, and the method of appointing regulators and their tenure). Unfortunately, acquiring such informationespecially for developing countriesis a Herculean task.9 Despite these shortcomings, this crude indicator provides a rst empirical look at the importance of regulatory reforms in the context of telecom reforms. v. results

Table V presents the results of estimating equations (1) and (2) above for each of the ve dependent variables. To preview, the results are, in general, consistent with conventional wisdom. Competition is associated with increased mainline penetration, payphones, connection capacity, and lower prices for local calls. Privatization by itself, meanwhile, is associated with few benets. Privatization combined with an independent regulator,
9

Such an eort is underway in the Research Group at the World Bank.

Blackwell Publishers Ltd. 2001.

econometric analysis of telecom privatization

13

however, is associated with increased payphone penetration, connection capacity and increased labor eciency as measured by employees per main line. Regulation interacted with competition had no signicant impact. Below I rst discuss in detail the results when using mainlines as the dependent variable, then provide a more cursory discussion of the results when using other indicators, as they follow the same general pattern. V(i). Number of mainlines

The rst two columns of Table V show the results of estimations using the number of mainlines as the dependent variable. The rst column shows the results of equation (1), and the second column shows the results of equation (2). The number of competitors is positively correlated with main lines. Indeed, this regression suggests that, ceteris paribus, each mobile competitor is associated with an increase of almost 0.2 additional main lines per hundred population. Privatization alone turns out to be correlated with a decrease in main lines per hundred population, though it is not statistically signicant (t-statistic 1.32). Interacting the regulation dummy variable with the privatization variable (the second column) changes some results. The eect of competition remains approximately the same (in magnitude and statistical signicance), while there does not appear to be any interaction eect. Privatization interacted with the separate regulator, however, is positively correlated with mainline penetration. The regulation dummy by itself is insignicant. The variable indicating the passage of reform legislation is negative and signicant, probably because poor performance helps stimulate reform eorts. Legislation, of course, is necessary before competition and privatization can be pursued. As the EIU [1997] notes, telecom reform `often requires changes in telecommunications laws or national constitutions, and is subject to legislative and political delays. Once the formal regulatory structures have been implemented, additional time and experience is needed to develop policymakers who understand how new technologies and services impact their markets.' It is therefore not surprising that passage of reform legislation would be negatively correlated with performance. As expected, income per capita is positively correlated with main lines per capita. Population and the percentage of population that lives in an urban area are negatively correlated with main lines per capita. Aidboth as a percentage of GDP and the presence of a World Bank telecom projectare insignicant. Exports as a share of GDP is positive, longterm debt as a share of GDP is negative, and lower expropriation risk is positively correlated with mainlines per capita. I have also estimated similar regressions to test the eects of any mobile competition (in addition to the number of competitors) and a time trend.
Blackwell Publishers Ltd. 2001.

 Blackwell Publishers Ltd. 2001.

14

Table V Regression Results (absolute t-statistics in parentheses)

Dependent Variable Mean of dependent variable Number observations # mobile operators Incumbent privatized? Separate regulator Private * separate regulator # mobile operators * separate regulator Reform legislation passed Population (millions) Urban population (as share of total) 1.09 (5.13) 0.11 (3.96) 0.06 (1.85) 0.17 (2.19) 0.46 (1.32) 0.15 (0.60) Mainlines per hundred population 4.48 401 0.23 (2.43) 1.16 (2.58) 0.05 (0.18) 1.21 (2.20) 0.07 (0.61) 1.01 (4.75) 0.1 (3.88) 0.05 (1.74) 0.77 (3.17) 0.12 (3.56) 0.06 (1.72) 0.26 (3.08) 0.62 (1.58) 0.5 (1.70) Connection capacity per hundred population 5.32 343 0.24 (2.24) 1.27 (2.54) 0.18 (0.52) 1.59 (2.48) 0.26 (1.60) 0.65 (2.71) 0.1 (3.03) 0.05 (1.53) 124.68 (2.34) 58.35 (7.35) 3.9 (0.44) 72.24 (4.21) 226.84 (2.65) 29.28 (0.45) Payphones per million population 811.6 277 118.05 (5.91) 212.51 (2.19) 168.95 (2.43) 845.78 (7.20) 48.46 (1.90) 44.4 (0.91) 59.15 (8.48) 8.12 (1.04) 0.56 (1.96) 0.03 (0.91) 0.08 (2.06) 0.14 (1.36) 0.20 (0.44) 0.13 (0.37) Employees per hundred lines 3.69 386 0.11 (0.86) 0.20 (0.32) 0.04 (0.12) 0.04 (0.06) 0.06 (0.33) 0.55 (1.87) 0.03 (0.90) 0.08 (2.06) 0.01 (1.49) 0.003 (1.46) 0.004 (2.12) 0.01 (3.67) 0.01 (0.49) 0.01 (1.35) Price of 3 minute local call (in USD) 0.073 163 0.02 (3.36) 0.01 (0.41) 0.02 (1.70) 0.04 (1.54) 0.004 (0.76) 0.01 (1.15) 0.003 (1.42) 0.004 (2.37)

scott j. wallsten

 Blackwell Publishers Ltd. 2001.

Income per capita ($ thousands) World bank telecom project Net World Bank aid ($ hundred millions) Exports / GDP Long-term debt / GDP Short-term debt / GDP Risk of expropriation (0-10; 10 lowest) R-squared Net eect of privatization and regulation F-test of joint signicance

1.15 (9.54) 0.18 (0.98) 0.02 (1.10) 2.56 (2.39) 1.06 (3.87) 1.35 (1.17) 0.14 (1.66) 0.97

1.18 (10.00) 0.19 (1.05) 0.01 (0.87) 2.8 (2.64) 1.06 (3.96) 1.22 (1.07) 0.12 (1.48) 0.97 0.05 3.56

1.46 (10.42) 0.06 (0.28) 0.01 (0.48) 1.43 (1.19) 0.76 (2.50) 2.16 (1.55) 0.03 (0.30) 0.96

1.53 (11.19) 0.02 (0.11) 0.002 (0.09) 2.12 (1.80) 0.73 (2.47) 1.51 (1.11) 0.01 (0.14) 0.97 0.32 3.88

96.96 (3.34) 65.63 (1.20) 6.33 (1.45) 293.57 (1.00) 197.37 (1.73) 339.68 (0.88) 12.42 (0.51) 0.93

110.85 (4.35) 55.56 (1.13) 8.40 (2.13) 410.74 (1.58) 91.48 (0.90) 518.27 (1.52) 17.11 (0.79) 0.95 464.32 30.45

0.001 (0.01) 0.13 (0.54) 0.04 (1.85) 0.22 (0.15) 0.11 (0.30) 1.95 (1.24) 0.25 (2.31) 0.88

0.005 (0.03) 0.12 (0.47) 0.04 (1.82) 0.21 (0.15) 0.11 (0.31) 1.99 (1.26) 0.25 (2.29) 0.88 n/a n/a

0.01 (1.05) 0.04 (3.18) 0.003 (2.49) 0.17 (2.12) 0.02 (0.73) 0.05 (0.90) 0.01 (1.33) 0.78

0.01 (1.47) 0.04 (3.17) 0.003 (2.83) 0.18 (2.24) 0.01 (0.54) 0.04 (0.67) 0.01 (1.63) 0.78 n/a n/a

econometric analysis of telecom privatization

Notes: Net eect calculated from the statistically signicant privatization, regulation, and interaction coecients. F-test is joint signicance of the privatization and regulation variables that are statistically signicant alone. Autocorrelation-robust standard errors, calculated using the Newey-West (1987) method.

15

16

scott j. wallsten

Those results, available on the Journal's editorial web site, are consistent with the results discussed above. In addition, they suggest that not only is competition benecial, more competition is better (a positive and signicant coecient on the number of mobile competitors even when controlling for the presence of any competition). V(ii). Connection capacity, payphones, employment, and prices

The other available telecom indicators behave comparably to mainline penetration. Regression results for connection capacity per capita, the second set of columns in Table V, is almost identical to mainline penetration. Competition is signicantly positively associated with connection capacity, privatization alone is (weakly) negative, and privatization combined with an independent regulator is positive. The eect on payphones is similar. However, in this case even privatization by itself is positively correlated with penetration, although combined with an independent regulator the magnitude is much greater (about 220 additional payphones per million population with privatization alone, versus a net eect of almost 460 payphones per million with privatization and an independent regulator). Competition and privatization have no identiable impact on employment per line. The lack of a signicant nding may reect the opposing impacts on this measure discussed above: privatization and competition induce rms to increase eciency by reducing employment, but may also induce rms to increase employment to improve service. The reform dummy variable, meanwhile, is signicantly negatively correlated with this employment eciency measure. This may indicate that rms begin to increase labor eciency once reform legislation is passed, to prepare for upcoming market changes. For example, a state-owned monopoly may wish to increase eciency prior to privatization or competition to increase the sale price or become more competitive, respectively. Prices are negatively correlated with competition, but not correlated with privatization. The presence of a World Bank telecommunications project is positively associated with prices. A Bank project was associated with an increase of about four cents per three-minute phone call. There are many possible explanations for this result. These projects may have been a part of large investments, which were nanced in part through increased taris. Alternatively, the projects may have coincided with eorts to reduce crosssubsidization, thus increasing the price of residential phone service. vi. discussion and future issues

Interpreted causally, the empirical results presented here are largely consistent with conventional wisdom. Competition appears to have tangible
Blackwell Publishers Ltd. 2001.

econometric analysis of telecom privatization

17

benets across the boardon mainline penetration, payphones, connection capacity, and prices. Privatization by itself does not appear to generate many benets and is negatively correlated with main line penetration. Privatization combined a separate regulator, however, is correlated with increased connection capacity and payphones per capita. Moreover, this interaction mitigates the negative eects of privatization on mainline penetration. These results suggest that reformers are correct to emphasize regulatory reforms along with privatization, since privatization without attention to regulation may be costly to consumers. Because competition appears to be the most successful agent of change, reformers should give careful thought to the notion of granting exclusivity periods to their incumbent telecom providers. While temporary monopoly rights undoubtedly raise the value of the incumbent to potential investors, it may delay the arrival of improved services to consumers. These results are interesting and encouraging, since they suggest that reform eorts are, in general, on the right track. But the results are far from conclusive. A large number of issues remain to be addressed. A rst order of business is to gather more detailed regulatory data. Noting that a country is not ignoring the regulation issueindicated by establishing a separate regulatoris a start, but does not begin to uncover the possible permutations of regulation. As mentioned above, it would be useful to know not only what type of regulation the country adopts (price caps versus cost-of-service, for example), but also details on the regulatory agency itself. What is its annual budget? How many employees does it have? Where do the regulators come from? What sort of training and experience do they have? What enforcement powers does the regulatory authority have? Second, we should conduct rm-level analyses similar to the ones presented here. Country-level data, while interesting, simply loses too much detail in the aggregation. Moreover, country-level analysis relies on telecom indicators from the ITU. While these are the best country-level data available, they do not always capture all telecom activity in the country. Often they contain information only on the incumbent. Firmlevel data would give us a much more detailed and accurate view of the eects of privatization, competition and regulation, and the overall telecom market in the country. Despite its shortcomings, this paper suggests that reform eorts tend to be concentrated in the right areas: encouraging competition and emphasizing building regulatory capacity when privatizing an incumbent telecom provider. The benets associated with competition over privatization, however, should cause policy makers to think carefully when granting exclusivity periods to privatized incumbents and, at the very least, should pay careful attention to the regulatory authority. This paper represents a rst attempt at econometric analysis of telecom reform eorts
Blackwell Publishers Ltd. 2001.

18

scott j. wallsten

in developing countries. This work is merely a rst cut at empirical tests, however. As more data become available for conducting econometric tests, we should gain greater insights. Future work should concentrate on improving data on regulation, gathering rm-level data, and endogenizing reform eorts such as competition, privatization, and regulation.

references
Ambrose, W., Hennemeyer, P. and Chapon, J., 1990, `Privatizing Telecommunications Systems: Business Opportunities in Developing Countries', IFC Discussion Paper Number 10. Armstrong, M., 1997, `Competition in Telecommunications', Oxford Review of Economic Policy, 13, Spring, pp. 64^82. Boubraki, N. and Cosset, J., 1998, `The Financial and Operating Performance of Newly Privatized Firms: Evidence from Developing Countries', The Journal of Finance, 53, No. 3, June, pp.1081^1110. Braga, C. A. P., Forestier, E. and Stern, P. A., 1999, `Developing Countries and Accounting Rate ReformA Technological and Regulatory El Nio?' Public Policy for the Private Sector, Note No. 173, January, The World Bank. Crandall, R. and Waverman, L., 1995, Talk is Cheap: The Promise of Regulatory Reform in North American Telecommunications (The Brooking Institution: Washington, DC). Petrazzini, B., 1996, Global Telecom Talks: A Trillion Dollar Deal (Institute for International Economics: Washington, DC). Galal, A. and Nauriyal, B., 1995, Regulating Telecommunications in Developing Countries. World Bank Policy Research Working Paper Number 1520. Greene, W., 1993, Econometric Analysis, Second Edition (MacMillan Publishing Company, New York). International Telecommunications Union, 1998a, General Trends in Telecommunications Reform, Volume II: Africa (ITU, Geneva). International Telecommunications Union, 1998b, General Trends in Telecommunications Reform, Volume III: Americas (ITU, Geneva). International Telecommunications Union, 1998c, Telecommunications Indicators Handbook (ITU, Geneva). Klein, M. and Gray, P., 1997, `Competition in Network IndustriesWhere and How to Introduce It', Public Policy for the Private Sector, Note No. 104, January, The World Bank. Laont, J., Rey, P. and Tirole, J., 1997, `Competition Between Telecommunications Operators', European Economic Review, 41, pp.701^711. Megginson, W., Nash, R. and van Randenborgh, M., 1994, `The Financial and Operating Performance of Newly Privatized Firms', The Journal of Finance, 49(2), June, pp. 403^452. Melody, W. H., 1995, `Privatization and Developing Countries.' In Mody, B., Bauer, J. and Straubhaar, J. (eds), Telecommunications Politics: Ownership and Control of the Information Highway in Developing Countries, (Lawrence Earlbaum, New Jersey). Noll, R., 2000, Telecommunications Reforms in Developing Countries, Stanford Institute for Economic Policy Research (SIEPR) Discussion Paper No. 99-31. Noll, R., 1987, `The Twisted Pair: Regulation and Competition in Telecommunications.' Regulation, 3/4, pp. 15^22.
Blackwell Publishers Ltd. 2001.

econometric analysis of telecom privatization

19

Petrazzini, B., 1996, Global Telecom Talks: A Trillion Dollar Deal. (Institute for International Economics, Washington, DC). Petrazzini, B. and Clark, T., 1996, `Costs and Benets of Telecommunications Liberalization in Developing Countries.' Mimeo. Pyramid Research, 1997, Privatizing Telecoms Markets. The Economist Intelligence Unit, LTD, May. Ros, A., 1999, `Does Ownership or Competition Matter? The Eects of Telecommunications Reform on Network Expansion and Eciency.' Journal of Regulatory Economics, 15(1), January, pp. 65^92. Saunders, R., Warford, J. and Wellenius, B., 1993, Telecommunications and Economic Development, (The World Bank, Washington, DC). Smith, P., 1997, `What the Transformation of Telecom Markets Means for Regulation.' Public Policy for the Private Sector, Note No. 121, July, The World Bank. Spiller, P. and Cardilli, C., 1997, `The Frontier of Telecommunications Deregulation: Small Countries Leading the Pack', Journal of Economic Perspectives, 11(4), pp. 127^138. U.S. Council Of Economic Advisers, 1996, `Promoting Competition in Traditionally Regulated Industries', Economic Report of the President (U.S. Government Printing Oce, Washington, DC). U.S. Department of Commerce, 1996, A Guide to Telecommunications Markets in Latin America and the Caribbean (U.S. Government Printing Oce, Washington, DC). Wellenius, B., 1997, `Telecommunications ReformHow to Succeed', Public Policy for the Private Sector, Note No. 130, October, The World Bank. Wellenius, B. and Stern, P., (eds), 1994, Implementing Reforms in the Telecommunications Sector: Lessons From Experience (The World Bank, Washington, DC). Wellenius et al.., 1992, Telecommunications: World Bank Experience and Strategy, World Bank Discussion Paper 192.

Blackwell Publishers Ltd. 2001.