*
As elsewhere, wireless markets play a crucial role in Latin American economic growth. Mobile telephone networks increasingly provide the communications infrastructure that has largely been lacking throughout the region. Yet, governments have generally made only modest allocations of bandwidth available to Latin American wireless operators, either absolutely (in terms of spectrum each country could allocate at modest opportunity cost) or relative to countries in North America, Asia and the European Union. Using an empirical model estimated on mobile phone data for international markets, we show that very large social benefits are available to countries that make more spectrum available for mobile phone markets. We conduct simulations using our calibrated model to provide lower bounds for country-by-country gains from larger allocations. We also discuss the impact of alternative regulatory regimes on the feasibility to achieve those social gains.
Keywords: spectrum allocation, auctions, revenue extraction, mobile telephone competition, telecommunications policy, Latin American economic development.
JEL classification: D6, H4, L5, L96.
* The authors would like to thank to three anonymous referees, the editors of this journal, and seminar participants at George Mason University and the Centro de Investigación y Docencia Económicas (CIDE,
Mexico) for valuable comments on an earlier version of this paper. The excellent research assistance of
Diego Avanzini is also acknowledged. Roberto Muñoz expresses gratitude to the Chilean National Fund for Scientific and Technological Development (FONDECYT) for partial support of this research under the project N° 11060033. Of course, all content (including errors) should be solely attributed to the authors.
Hazlett (corresponding author): thazlett@gmu.edu
, Tel. (703)993-4244, Fax: (703)993-8124; Muñoz: roberto.munoz@usm.cl.
1

1.
INTRODUCTION
Wireless services are growing rapidly. In Latin America, as in other developing markets, mobile phone networks are supplying valuable social overhead capital (SOC), stimulating economic growth. Waverman et al. (2005, p. 18), studying African economies, find that “[d]ifferences in the penetration and diffusion of mobile telephony certainly appear to explain some of the differences in growth rates between developing countries… there are also increasing returns to the endowment of telecoms capital (as measured by the telecoms penetration rate)… Our analysis suggests the need for regulatory policies that favour competition and encourage the speediest rollout of mobile telephony.” This conclusion builds on the widespread view that telecommunications networks are key components of SOC (Hardy, 1980; Leff, 1984; Norton, 1992;
Greenstein and Spiller, 1996), and recent studies showing that wireless technologies are strong drivers of developing country growth.
1
Yet, spectrum policies in Latin America are, on the whole, extremely conservative. This appraisal reflects both the structure of regulation, and the quantitative outcome. On average, Latin American countries allocate only about 100 MHz to mobile phone carriers’ licenses, compared with a mean of about 266 MHz in the European
Union.
2
Some of the differential is attributable to demand differences, but controlling for those factors we still find a statistically significant restriction on valuable bandwidth imposed across both sets of markets. Given substantial amounts of unutilized bandwidth, the opportunity costs of more generous cellular allocations are typically quite modest.
Two Latin American countries, Guatemala and El Salvador, liberalized their spectrum regimes via far-reaching legislative measures enacted in 1996 and 1997, respectively. Wireless markets in both countries appear relatively robust, exhibiting high degrees of competitiveness, as measured by industry concentration and retail prices.
Other Latin American countries, while not undertaking such ambitious reforms, have initiated policies that distinguish heir wireless markets. These include Argentina, Chile,
Nicaragua and Paraguay. The experience of these nations may inform policy makers as to what might be achieved, and offers data for public choice scholars wishing to explain the divergence in regulatory regimes – a pursuit we do not undertake herein.
The primary task of this paper is to evaluate the Latin American spectrum policies now in place, an analysis we divide into four parts (Sec. 2 – Sec. 5). First, in Sec. 2, we provide a diagnostic analysis revealing the primary spectrum under-allocation problem confronting Latin American mobile services markets. We also define alternative regulatory regimes, providing a taxonomy for reform measures that alter bandwidth allocations by traditional administrative actions versus policy liberalization granting wireless rights-holders greater scope for accessing and utilizing airwaves. In Sec. 3 we then present a model that tracks the relationship between spectrum policies and retail mobile market results (prices and outputs), empirically calibrating the model using a panel consisting of 28 countries for the period 1Q1999-2Q2003. These estimates are
1 See for example: Dasgupta, et al. (2001); Waverman et al. (2005); U.N. (2004); Annan (2005).
2 For the sources, see Table 1 below for Latin America and Appendix 1 for the European Union.
2

then, in Sec. 4, used in simulations that forecast the social value of enhanced spectrum allocations for mobile phone service in six of the largest Latin American nations. We find that, in response to an increase of 20 MHz, the average change in consumer surplus in our sample is approximately US$50 per capita. This magnitude is over 10 times the average marginal revenue in regional countries employing wireless license auctions (on a per capita per MHz basis), suggesting that the social gains are of very substantial magnitude, and accrue overwhelmingly to consumers’ surplus rather than to mobile operator profits.
In Sec. 5 we consider the prospects for achieving such welfare gains under alternative regulatory structures. The gains described above could be realized via policies instituted by fiat under the administrative allocation systems now in place. Alternatively, statutory reforms may eliminate regulatory discretion, forcing a regime shift. We also explain that the important link between a liberalization of licensee property rights
(granting greater degrees of freedom in utilizing allocated bandwidth) and the expansion of spectrum allocations (increasing the bandwidth allocated for use by mobile operators).
We then offer a brief conclusion.
2. WIRELESS MARKETS IN LATIN AMERICA: A DIAGNOSIS
Spectrum allocation policies in Latin America are usefully evaluated with respect to the market for mobile phone services.
3
First, this application constitutes the dominant spectrum-based service in terms of service revenues. Second, the next most economically important industry, radio and television broadcasting, is intensely political and is subject to its own idiosyncratic analysis.
4
Third, mobile telephony plays a vital role in economic development, providing basic telecommunications infrastructure for most residents and businesses. And fourth, this industry is broadly studied by investors, yielding data and opening the way for empirical estimates of the social value of policy reforms.
5
The average amount of spectrum allocated to cellular service in Latin American countries (equally weighted) is 102 MHz (see Table 1), well below the average in the
European Union of 266 MHz. We consider two possible reasons for this gap.
(1) A market hypothesis: In Latin America the demand for mobile services does not justify a higher allocation. The lower amount of spectrum would optimize resource use were marginal frequencies more productively deployed elsewhere.
3 This paper uses the terms “wireless telephone,” “mobile telephone,” and “cellular” interchangeably.
4 The politics of broadcast regulation have figured prominently in explaining the spectrum allocation regime in the U.S. and other countries, but the emerging dominance of cellular in the 1980s and 1990s fundamentally shifted this paradigm. See Hazlett (1998, 2001).
5 For example, Hazlett et al. (2006) study the transition to digital TV in Europe. Using estimates from the mobile phone market, they provide a lower bound for the social value of TV band spectrum.
3

(2) A regulatory hypothesis: Regulatory authorities in the region have inefficiently constrained spectrum access, over-conserving bandwidth.
This would suggest non-market failure.
Figure 1 illustrates the situation for a sampling of countries included in quarterly mobile phone market data published by Merrill Lynch (2003).
6 This global dataset includes six Latin American countries: Argentina, Chile, Brazil, Mexico, Venezuela and
Colombia. As can be seen, all six countries allocate substantially less spectrum to mobile telephony than what is predicted by their per capita national income.
7
This L.A. subsample features the largest economies in the region, and we note that most of the remaining countries have even more parsimonious spectrum allotments.
The quantity of spectrum allocated to wireless phone service is presumably a function of other factors in addition to GDP per capita, and countries throughout Latin
America should be included in the analysis if we are attempting to provide a regional perspective. For this broader statistical inquiry, we generate a new database, adding all non-Caribbean Latin American countries not appearing in Merrill Lynch (2003). Data for the additional Latin American countries
8
are only available annually, however. Hence the new database is an annual panel from 1999 to 2002, with data for 40 countries (29 in
Merrill Lynch, to which we add eleven more Latin American countries).
F IGURE 1: S PECTRUM VS .
GDP PER C APITA ( SECOND QUARTER 2003)
600,0
New Zealand
500,0
400,0
300,0
200,0
100,0
Czech Republic
Chile
Argentina
Brazil
Colombia
Hungary
Mexico
Venezuela
Hong Kong UK
Netherlands
Greece
Portugal
Spain
Germany
Italy
Singapore
Australia
France
Austria
Sweden
Finland
Belgium
Canada
USA
Denmark
Ireland
Norway
0,0
0 10.000
20.000
30.000
GDP per Cápita (US$)
40.000
50.000
Sources: Authors’ calculations, country regulatory authority websites, and Merrill Lynch (2003).
6 The Merrill Lynch database consists of a quarterly panel with 46 countries, first quarter 1999 through the second quarter 2003. A subsample of 29 countries was used given that these were markets for which spectrum allocation data were available.
7 We use a simple linear model in this diagnostic because we are focusing on the relationship between spectrum allocations for mobile telephony (in MHz) and per capita income.
8 The additional 11 countries are found in Table 1, along with the six L.A. nations found in the Merrill
Lynch database.
4

Argentina
Bolivia
Brazil
Chile
Colombia
Costa Rica
Ecuador
El Salvador
Guatemala
Honduras
Mexico
Nicaragua
Panama
Paraguay
Peru
Uruguay
Venezuela
Country
Average LA
(a)
Average Price per Minute (APM)
US$
(October 2003)
T
ABLE
1: B
ASIC
S
TATISTICS FOR
M
OBILE
P
HONE
M
ARKETS IN
L
ATIN
A
MERICA
(b) (c) (d) (e) (f) (g) (h)
Spectrum
MHz
( October 2003)
HHI
(1-10000)
Penetration
(%)
(Dec. 2003) (Dec. 2003)
GDP
(US$ billions)
(Dec. 2003)
GDP adjusted by PPP
(US$ billions)
(Dec. 2003)
Population
(millions)
(Dec. 2003)
(i)
GDP GDP per capita per capita adjusted by PPP
(US$) (US$)
(Dec. 2003) (Dec. 2003)
0.097
0.126
0.174
0.144
0.101
0.085
0.244
0.155
0.111
0.311
0.267
0.273
0.307
0.201
0.215
0.260
0.298
120
88
110
140
100
93
80
137.87
140
65
120
84.84
49.56
176
80
90
57
3266
3942
4298
2790
5501
10000
5115
3297
3591
10000
6154
5009
5000
3566
4041
5919
3570
18
15.2
26.4
51.1
14.1
14
18.9
17.6
17
6.2
29.1
8.5
26.76
29.85
10.61
25
27.3
127.30
8.10
505.54
73.37
80.01
17.70
27.20
14.94
24.74
6.95
636.55
4.15
12.86
5.60
60.79
11.21
84.28
442.04
22.73
1373.22
161.90
296.44
39.50
49.31
28.38
51.06
18.12
943.89
14.20
19.80
26.38
143.10
27.50
125.81
38.32
7.80
172.99
16.65
45.85
4.15
13.86
7.54
16.56
7.02
105.98
8.71
3.04
5.90
28.23
3.42
27.63
3322
1038
2922
4408
1745
4263
1962
1981
1494
989
6006
477
4231
948
2154
3275
3050
8763
3574
7708
15575
7724
8169
5249
3458
4551
2500
9504
2454
8431
3776
5020
8307
8535
0.198
101.84
5003.47
20.92
100.07
222.55
30.22
2604 6665
Sources:
(a)
(b)
(c)
Authors' calculations based on advertised postpaid subscription prices using equally weighted averages for plans having 100-400 minutes of use per month
offered by the larger wireless carriers in each country. Information collected from web pages of these companies.
Authors' calculations based on database in Hazlett (2008b).
Authors' calculations based on Pyramid Report 2003. In countries divided in several zones, the HHI is obtained as a simple average of HHI per zone.
(d) AHCIET, Observatorio económico del sector de las telecomunicaciones.
(e), (f), (g), (h), (i) World Economic Outlook, IMF.
Note : Penetration is defined as phone subscribers per 100 population .

This expanded database permits estimation of a simple regression where the quantity of spectrum assigned to mobile telephony is the dependent variable. The vector of independent variables includes Population, Population Density, GDP per capita , a Not
Calling Party Pays dummy (=1 if mobile phone calls received incur connection charges), a Liberal dummy (=1 if country is Australia, New Zealand, Guatemala or El Salvador 9 ) and an Auction dummy (=1 if country assigns licenses by competitive bidding). Table 2 contains regression results excluding and including a Latin America dummy (=1 if country is in Latin America).
Here the coefficients of interest are those associated with the Latin America and the Liberal dummies. The first is estimated to be negative and of a statistically significant magnitude. These estimates are similar in both models, one a completely pooled regression, the other using random effects. Moreover, the Latin America dummy improves model specification. Controlling for other factors, cellular spectrum allocations appear substantially lower in Latin America —by an estimated 55 MHz -- than in other countries. This magnitude is over 50% of the Latin American mobile allocation mean.
Alternatively, the estimated coefficient for the Liberal dummy is positive, consistent with the goal of economic reform, although the result is not statistically significant.
10
The data reveal a reduced level of spectrum availability for mobile services in
Latin American countries versus other markets. Can this difference be explained by valuable alternative deployments? In fact, there is no evidence to suggest that any wireless services – in broadcasting, satellite, local area networking or elsewhere -- productively employs the frequency spaces withheld from mobile operators. This suggests that the unallocated spectrum has little or no social opportunity cost. Hence, the data are best interpreted consistent with Hypothesis 2: regulatory constraints limit access to bandwidth. We now turn to ways in which regulatory structure may impact spectrum allocations.
9 These four countries undertook far-reaching statutory spectrum reforms in the 1990s, moving substantially to “property rights” regimes. See Hazlett (2008b).
10 The data are limited, of course, yielding few degrees of freedom.

T ABLE 2: R EGRESSION R ESULTS .
D
EPENDENT VARIABLE
: B
ANDWIDTH
A
LLOCATED TO
M
OBILE
P
HONE
S
ERVICE
(MH
Z
).
T
OTALLY
P
OOLED
M ODEL W/O
LAD
( robust
M
R
E
ANDOM
FFECTS
ODEL W/O
LAD
T
P
OTALLY
OOLED
M ODEL
( robust estimation ) estimation )
Independent Variable
Constant
Population
GDP/capita
Pop Density
Estimated coefficient
(t-statistic)
78.57154*
(5.99)
-0.058682
(-0.60)
0.004129*
(6.46)
0.010969**
(2.04)
-77.92255*
(-4.11)
Estimated coefficient
(z-statistic)
75.15470*
(3.43)
-0.024181
(-0.12)
0.004227*
(4.68)
0.012346
(1.26)
-82.34330***
(-1.70)
Estimated coefficient
(t-statistic)
120.8150*
(4.82)
-0.012468
(-.012)
0.002713*
(2.74)
0.010462
(1.95)
-76.66508*
(-4.14)
Not Calling Party Pays
Dummy
Latin American Dummy
(LAD)
Liberal Dummy
Auction Dummy
No.Observations
R-Squared
37.49638
(0.69)
43.89497*
(2.80)
134
0.2579
36.72925
(1.00)
41.89476**
(1.92)
134
0.2572
-53.54065*
(-2.51)
34.75306
(0.65)
34.81981**
(2.20)
134
0.2862
*, **, *** refer to 99%, 95%, and 90% confidence levels, respectively.
R
E
ANDOM
FFECTS
M ODEL
Estimated coefficient
(z-statistic)
124.2632*
(3.89)
0.011336
(0.06)
0.002590**
(2.19)
0.011294
(1.19)
-78.96361***
(-1.69)
-58.50892**
(-2.05)
(1.00)
(1.49)
134
0.2858
35.52704
32.21637
Spectrum Regimes: Three Simple Benchmark Models
The goal of this subsection is to provide three benchmark models representing the most commonly used spectrum policy regimes in Latin America. Suppose that there exist just two firms and two services in the industry. We will identify the firms with super indexes and the services with sub indexes.
We assume that firm i solves:
Max q i
1
, q i
2

P
1
 q
1 i  q
1 j
 q
1 i 
P
2
 q i
2
 q
2 j
 q i
2
 c
 q
1 i
, q i
2
 
Analogously, firm j solves for q
1 j , q
2 j
. Distinct cost functions result from differing regulatory schemes. The following models are introduced as benchmarks to discuss
7

spectrum policies found in Latin America.
11
While simple, the models reflect an important economic reality. In their use of radio spectrum, wireless networks view the resource as an input. The value of this input can be increased either by expanding the amount of bandwidth available to carriers (MHz), or by improving the terms on which an existing frequency allotment is utilized. For instance, suppose a regime allocates each mobile operator’s license 50 MHz of radio spectrum, while mandating that analog technology be used. Regulators could then effectively increase the spectrum allocation either by adding, say, 25 MHz to each license, or by a reform that permitted operators to employ other technologies (including digital systems). In essence, a liberalization that effectively yields increasing output is analogous to an increase is allocated bandwidth.
Model I: Spectrum Assigned to Services
In this case, the regulatory authority assigns spectrum to each operator to provide a specific service, with licenses awarded by either beauty contest or competitive bidding.
The problem of minimization of costs
12
for firm i is then given by: c
 q
1 i
, q i
2


Min
K i
1
, K i
2
 r

K
1 i 
K
2 i


F ( S i
)

, s.t. f
2 f
1


K
K
1 i i
2
, S
1 i


 q
1 i i
, S i
2
 q
2
S
1 i 
S i
, S
2 i 
0 and S i 
S j 
S
The two first constraints are the production functions for services 1 and 2. For simplicity, we assume that there are just two factors, capital ( K ) and spectrum ( S ). The capital cost is given by r , while spectrum rights i
S is acquired by firm i at cost F ( S i
) .
13
The last constraint is determined by regulation. A quantity of i
S MHz is allocated to the license of firm i , which authorizes the firm to supply service 1. Service 2 is not authorized by the license, although the licensee, if unconstrained, could provide the service with some fraction of the capacity provided by i
S .
The constraints for firm j are a mirror image of those of firm i , and total allocated spectrum is divided between the two licenses:
S
2 j 
S j
, S
1 j 
0 , S i 
S j 
S .
11 These models are based on Muñoz (2004).
12 Cost minimization is implied by profit maximization, and is the component of the optimization process directly linked to spectrum policies. How spectrum inputs are used in the wireless operator’s production function, subject to different regulatory constraints, imply cost functions defining the relevant margins.
13 This cost represents the payment offered in the corresponding auction, or the implicit payment (in the form of rent-seeking and commitment to non-economic expenditures) for licenses in a beauty contest.
8

if not fully, extracted through F
Mirroring firm i , firm j is authorized to provide service 2 but not service 1. As a result, each firm enjoys monopoly power in its authorized output market, yet rents are partially,
( S i
) in the license award.
Model II: Spectrum Assigned to Firms
This case represents an intermediate position on a spectrum liberalization continuum.
14
The regulatory authority allocates spectrum to licenses, which are then distributed to firms (through auctions or beauty contests), and it does not constrain the services that firms may supply or the technologies employed in accessing this bandwidth.
Cost minimization for firm i is then given by: c
 q
1 i
, q i
2


K i
1
,
Min
S i
1
, K i
2
, S i
2
 r

K
1 i 
K
2 i


F ( S i
)

, s.t.
S
1 i 
S
2 i  f
2 f
1


K
K
1 i i
2
, S
1 i


 q
1 i i
, S i  q
2 2
S i and S i 
S j 
S
15
The result is distinguished from that of the previous model due to the altered regulatory constraint. In this case, a total amount of S i
MHz is allotted firm i
’s license, with the firm being authorized to freely distribute bandwidth between the two services.
16
Under this scenario, both firms compete in both markets.
Increased spectrum flexibility has two effects. First, firm i is able to deploy frequency space where it produces the greatest incremental profit, increasing the potential value of the license. Second, as firm j also receives the same flexibility, monopoly rents are dissipated, decreasing the value of the license. As a result, the sign of license value windfalls associated with a shift from Model I to Model II is theoretically ambiguous.
14 Liberalization of spectrum policy, following Kwerel and Williams (2002), can be broken down into two component parts. The first encompasses the flexibility given a particular licensee to use the spectrum allocated to its license. More flexibility cedes additional property rights to wireless operators. The second encompasses the process whereby spectrum is allocated (or reallocated) from one category to another. This permits spectrum to be bid out of a given deployment and used in another without special regulatory action.
15 This constraint does not preclude “spectrum sharing” between the two services. The total capacity of the allocated spectrum, then, can be defined in multiple dimensions.
16 This distribution “between” services may (in fact, likely would) involve sharing frequency spaces. But splitting capacity between multiple services cannot be done without cost. Hence, this does not change the basic trade-off we model, for simplicity, as a choice between service-specific bands.
9

Model III: Spectrum Assigned by Markets
In this case, private property rights are assigned to the spectrum resource, wireless firms enjoying full flexibility in the use of assigned airwave space. Distinct from Model
II, spectrum rights can flow between firms.
17 The cost minimization problem for firm i becomes: c
 q
1 i
, q i
2


K
1 i
,
Min
S
1 i
, K i
2
, S i
2
S
1 i 
 r

K
1 i 
K
2 i
 
S
1 i 
S i
2

 f
2 f
1


K
K
1 i
2 i
, S
1 i


 q
1 i i
, S i 
2 q
2
S
2 i 
S
1 j 
S
2 j 
S
F S
  i

, s.t.
Distinct from the previous models, the objective function here includes a market price for spectrum (

), which results from the equilibrium between supply, S , and demand in the factor market. In this case spectrum flows not only to the more efficient use intra-firm (as in Model II), but also to the more efficient firms. Rent extraction, represented by F ( S i
) , continues within the assignment process, with a theoretically ambiguous sign for license value windfalls associated with a shift from Model II to
Model III. In addition, it is theoretically possible that a price increase obtains from
Model II to Model III, because in the latter the cost of Spectrum is included as a marginal cost. The sign of both ambiguous effects needs to be studied.
Hazlett (2008b) finds that, for countries implementing ambitious spectrum liberalizations, wireless license sales prices are observed to be about 61% lower than in other countries, adjusting for other factors. This implies that the effect of increasing competition in the transition from Model I through III can easily dominate the higher valuation for flexibility. This creates a dichotomous policy choice. Governments will tend to prefer Model I if the policy goal is to maximize auction receipts; Model II or III if the objective is to maximize social welfare.
Blanco (2005) studied the impact on Consumer Surplus (and then on prices) of the transitions from Model I through III. Assuming Cobb-Douglas production functions in each service, she showed in simple examples that consumer surplus increases significantly more in the transition from Model I to Model II than in the transition from II to III. She recognizes that this result underestimates the impact of property rights because her evaluation does not consider the gains from flexibility in the introduction of new services or new technologies. There are, however, other reasons why the benefits from the transition from Model II to III are underestimated in her study. The total amount of
17 This was the logic employed by Coase (1959), who suggested that private property rights in spectrum would not only substitute for government allocation, they would lead to more efficient levels of radio interference.
10

spectrum liberalized is assumed to be the same in Models II and III, which severely truncates potential effects of a liberalization that permits transactions to increase bandwidth. Nevertheless, her study suggests that regulatory inertia may not result from ideological opposition to property rights, per se, because Model II does not require explicit property rights and yet represents a significant social welfare improvement over
Model I.
Categorizing Latin American Spectrum Regimes
Models I, II and III represent benchmarks allowing us to chart the fundamental economic changes associated with spectrum liberalization. We now ask: How extensive are exclusive spectrum ownership rights in Latin America?
We review the spectrum regulatory regimes of each country in the region, 18 identifying key aspects. A summary is provided in Table 3, where in the last column we classify each country according to
Models I to III, based on the characteristics indicated by the variables displayed in other columns. Of course, our categorization reflects the judgments we make, and may be viewed somewhat differently by other methods. In the analysis just above and in Table 3 we attempt to be clear about the manner in which our judgment is exercised.
When a service license does not permit the introduction of new services by incumbent firms without permission (column (c)), then Model I is implied. The classification is confirmed by the fact that in those countries the license is revoked when the licensee provides an unauthorized service (column (h)). On the other hand, the existence of explicit property rights (column (f)) is a sufficient condition to classify the country as Model III. The information in the rest of the columns was not conclusive, but in aggregate permits a reasonable classification scheme.
For instance, when technological neutrality (column (d)) is not supported, then
Model III is excluded. A regime where the absence of a positive administrative determination fails to defer to a petitioner’s usage request (column (g)) indicates strong regulatory interference in the provision of new services. In combination with the absence of technological neutrality, this approach classifies Peru as a Model I country. Argentina and El Salvador present challenges in categorizing. Both present relatively flexible regimes, but column (e) suggests that in El Salvador the spectrum can flow more freely across firms than in Argentina, so Model III best applies to the former while Model II to the latter. We note that columns (a) and (b) are simply descriptive. Column (a) shows that the broadly used regulatory practice of separating service licenses from wireless licenses is not correlated with the flexibility of the spectrum regime. Finally, column (b) reflects the fact that wireless licenses that define permitted services are (definitionally) associated with inflexible rights.
18 Details of legal documents reviewed are given in Appendix 3.
11

Country
(a)
Is the service license distinct from the wireless
license?
(b)
Is wireless license operator restricted to a particular service?
T
ABLE
3.
L
EGAL
A
SPECTS OF
L
ICENSES IN
L
ATIN
A
MERICA
(c) (d) (e) (f)
Does the service license permit the introduction of new services?
Technological neutrality
Spectrum under administrative control?
(g) (h)
Explicit radio spectrum property rights?
Positive administrative silence?
Is the license revoked if the licensee provides an unauthorized service?
(i)
Model
Argentina
Bolivia
Brasil
Chile
Colombia
Costa Rica
Ecuador
El Salvador
Guatemala
Honduras
Mexico
Nicaragua
Panama
Paraguay
Peru
Uruguay
Venezuela yes yes yes yes no yes yes yes yes yes yes yes yes yes yes no no yes yes yes yes yes yes yes yes no yes yes no yes yes no no no no no no yes no no yes no no yes, but restricted ones. * no no yes no no yes no no no yes yes no no no no no no yes intermediate yes yes yes yes yes yes no no yes yes intermediate yes yes yes yes yes yes no no no no no no no no no no no no no no no no yes yes yes yes no no no yes yes yes yes yes yes no yes yes yes yes it depends on the contract it depends on the contract yes
Model II
Model I
Model I
Model I
Model I
Model I
Model I
Model III
Model III
Model I
Model I
Model II
Model I
Model I
Model I
Model I
Model I
* Those considered "services of added valued."
(a) In most of the Latin American countries a wireless operator needs two licenses, one to authorize the service (usually a ‘public service license’) and a second one granting spectrum access.
(b) The wireless license defines the service(s) to be provided.
(c) The cross constraint established in (b) is usually redundant because, as column (c) shows, the licensee usually needs new service licenses to provide services not explicitly authorized.
(d) In this paper we take technological neutrality in a broad sense. The regulatory authority in a country is technologically neutral if it does not impose direct constraints over the technology used by providers of a service. A “neutral” authority can, however, impose indirect constraints over technology such as quality of service requirements.
(e) Administrative allocation is the traditional regime where regulators, rather than market-based owners, determine spectrum use.
(f) In most countries the wireless license does not convey property rights over spectrum. The exception is the case of Guatemala, which grants title to spectrum use rights.
(g) When an existing service provider or an entrant applies for a new service license the authority usually has a deadline to answer the request. If the authority does not provide an answer before the deadline is met, then a positive administrative silence means the application is approved.
(h) The provision of an unauthorized service by a licensee is usually subject to penalty as severe as license revocation.
(i) Based on the information provided in the previous columns, we assigned each country to one of the reference models discussed in the previous section.

Table 3 reveals that the dominant regulatory structure assigns spectrum use rights for specific services (Model I). Most countries define standards for operators, restricting technology choices and violating the commonly stated regulatory goal of technological neutrality.
Departing from Model I are the regimes found in Argentina and Nicaragua, which we categorize as Model II. Both countries feature relatively liberal rules with respect to the use of spectrum allocated to a given license, but does not enable firms to freely move spectrum between firms or to move unallocated spectrum into productive use. Since relatively little spectrum has been allocated to licenses, this is an important constraint. Guatemala explicitly defines property rights to the use of frequencies ( Titulo de Usufructo de Frecuencia) and extends the property regime by granting parties the right to petition for access to unoccupied bandwidth, imposing an obligation on the regulator to issue such rights by competitive bidding. El Salvador has enacted a functionally similar process, although property rights are not explicitly granted. Licenses are defined by international spectrum allocation templates, but licensees are granted the freedom to offer other applications within the bandwidth allocated to the license. Mechanisms are also specified for parties to obtain unassigned licenses from the regulator. While the regulatory authority is not mandated to issue to such licenses, as in Guatemala, we note the empirical similarity in the mobile markets. As of 2003, both countries featured four national networks, with carriers using 140 MHz in Guatemala, and 137.9 MHz in El Salvador. We classify these systems as approximated by Model III.
Prices and Market Concentration in Latin America
Thus far we have not discussed the impact of spectrum policy on retail prices. The regimes identified above, however, permit us to identify some correlations between regimes and market behavior in our test market -- mobile phone services. We are particularly interested in two indicators. First, how much spectrum has been made available for mobile phone markets in the different regimes.
19
Second, how do retail prices differ across regimes?
Consider first the spectrum available for wireless telephony in Guatemala and El
Salvador. Although in both countries the number is still below the trend line in Figure 1, the gaps
-- 16 MHz for Guatemala and 20 MHz for El Salvador – are well below the regional average of
55 MHz.
20
This suggests that liberal regimes facilitate the efficient flow of spectrum in Latin
America.
21
Figure 2 shows that higher spectrum allocations are directly related with lower average price per minute (APM). It is then not surprising that cellular telephone rates appear to be generally below average under the more liberal Latin American spectrum regimes, while competitiveness appears to be higher as it is shown in Figure 3 where we chart the APM and
HHI data found in Table 1.
19 In the following sections we show that, at the current margin, more spectrum allocated to mobile phone markets implies higher social welfare.
20 Of course, the spectrum gap for Latin America would be wider were Guatemala and El Salvador not relatively well endowed.
21 See Hazlett et al. (2007).

F IGURE 2: A VERAGE P RICE PER M INUTE (APM) AND S PECTRUM A LLOCATIONS
IN
L
ATIN
A
MERICA
200
180
160
140
120
100
80
60
40
20
0
0.000
Gu Ch
CR
Ar
Co Bo
ES
Br
Par
Pe
Me
Ur
Ni
Ec
Ve
Ho
Pan
0.050
0.100
0.150
APM
0.200
0.250
0.300
0.350
12,000
10,000
8,000
6,000
4,000
2,000
0
0.000
F IGURE 3: A VERAGE P RICE PER M INUTE (APM) AND M ARKET
C
ONCENTRATION
(HHI)
IN
L
ATIN
A
MERICA
0.050
CR
Ar
Co
Bo
Gu
Ch
ES
Br Pe
Par
0.100
0.150
APM
0.200
Ur
Ec
Me
Ni
0.250
Ho
Pan
Ve
0.300
0.350
14

While we have strong theoretical and empirical reasons for associating high concentration with high prices,
22
Figure 3 suggests the relationship is complicated. The lowest price is observed in Costa Rica where the supply of wireless telephony is concentrated in a state monopoly, unique in our sample. As Blanco (2005) points out, subsidies seem to explain these low prices. At the other extreme, Venezuela shows one of the highest values for APM, despite of the fact that the concentration index is one of the lowest in the region. Importantly, however, the countries with the most liberal regimes, El Salvador and Guatemala, feature both relatively low APMs and concentration ratios. Their situation is comparable to Chile and better (from a social point of view) than Brazil, despite differences in size and GDP per capita tending to favor the larger, wealthier markets. Our hypothesis is that spectrum liberalization drives these results.
3. MODELING THE IMPACT OF SPECTRUM SCARCITY
3.1. A SIMPLE THEORETICAL MODEL
Consider a market where N firms will be producing a homogeneous mobile telephone service, with output levels given by q i
where i identifies the firm. We assume there is no initial incumbent. Aggregate output is given by
 q i

Q . The market price associated with this output i is defined by the inverse demand function p(Q) .
The cost function of firm i is given by:
C i
( q i
)
 c ( K i
, S i
) q i
(1)
Where K i
denotes capital and S i
the spectrum allocated to the license awarded to firm i .
We assume a two-stage game, where in the first stage capital and spectrum investments are defined simultaneously and independently. In order to focus the analysis on spectrum allocation policies, we assume symmetric investments ( K i
= K for all i ). In the second stage we assume
Cournot competition. Second stage’s marginal cost c ( K , S i
) is decreasing in capital and spectrum, and these two inputs are substitutes (Reed, 1992; 11-12, 20-21).
In what follows we denote market share as s i
 q i
/ Q , while

( Q ) is the price elasticity of demand. In such a context, a Mark Up equation is defined by: p ( Q )


 1

HHI

( Q )



1 i
N 

1 s i c ( K , S i
)
In particular, when spectrum allotments are equal across competitive licenses, we obtain:
22 See Hazlett and Munoz (2008).
15

p ( Q )


 1

HHI

( Q )



1 c
S
K ,
N
(2) where S is the total amount of spectrum assigned to mobile wireless services.
We interpret the equilibrium Mark Up equation (2) as one where the supply depends on the elasticity of demand

( Q ), the level of investment ( K ), the total amount of allocated spectrum to the service ( S ), and the Herfindahl-Hirschman Index ( HHI ). We assume demand for wireless telephony to be a function of the price of wireless service ( p ), income level ( Y ), and the price of alternative telephone services ( F ). In principle, we can posit a constant elasticity of demand function for wireless telephony
23
such that:
Q
 
Y

F
 p

(3)
3.2. THE ECONOMETRIC MODEL
In Hazlett and Muñoz (2008), we sought to answer the question: Looking across countries, what factors influence the prices and outputs of mobile phone usage?
The Latin
American country simulations conducted in this paper are enabled by the empirical model developed therein.. The econometric model is a system formed by an empirical Mark Up equation motivated by the variables in Eq. (2), and an expanded log-log version of the demand function given by Eq. (3). Both include nonlinear terms.
Price is predicted by a system of simultaneous equations, one an empirical Mark Up equation (4) and the other an empirical Demand equation (5). A traditional fixed effects model would permit us to adjust for unobserved characteristics between markets (countries) in our panel dataset. However, our inclusion of important time-invariant explanatory variables led us to employ a pseudo-fixed effects model (Plümper and Troeger, 2007). The endogenous determination of RPM, TOTMIN and HHI is embedded in the estimated model, which uses a
3SLS approach in the first and the third stages of the Plümper and Troeger procedure. Included variables are described in Table 4. Identification is achieved in the estimation of the system because the number of excluded explanatory variables in each equation is greater than the number of endogenous variables.
Merrill Lynch data were available for 28 wireless phone markets, providing quarterly information on prices (proxied by mean revenue per minute of use) and output (minutes of use),
1999-I to 2003-II.
24 Results are displayed in Table 5 for different specifications. Model 6 was the preferred specification, and the one whose results are discussed here.
23 A more general log formulation would include nonlinear terms.
24 The database in Hazlett and Muñoz (2008) is detailed in Appendix 2. We emphasize that it differs from that used to generate Table 2 primarily due to the fact that quarterly price and output data are not available for most Latin
American countries.
16

Empirical Mark Up Equation (4): ln( RPM it
)
 
0
 
1 ln( Q it
)
 
2
[ln( Q it
)]
2  
3 ln( HHI it
)
 
4
[ln( HHI it
)]
2 
 
5 ln(
 
9
[ln(
Spectrum it
Spectrum it
)
)
 
6
* ln(
[ln( Spectrum it
Density it
)]

)]
2

10
 
7 ln(
Auction it
Density it
 
)
 
8
11
NotCPP it
[ln( Density it
  it
)]
2 
Empirical Demand Equation (5): ln( Q it
)

 
5
A ln(

0
 
1 ln( RPM it
)

Fixprice it
)
 
6
[ A ln(

2
[ln( RPM it
)]
2
Fixprice it
)]
2
 
3 ln( GDPpc it
)
 
4
[ln( GDPpc it
)]
2
 
7
NotCPP it
  it
T ABLE 4: D ESCRIPTION OF V ARIABLES IN E MPIRICAL M ODEL

RPM Revenue per minute in constant 2000 US$ for mobile voice services, a
proxy for price. 25
Q Output, measured as total minutes of use per month (totmin),
in millions.
HHI Herfindahl-Hirschman Index in the market (0 to 10,000), with market
shares based on subscribers.
Spectrum Aggregate bandwidth available for mobile phone service by all operators in the market. Measured in MHz.
Density A proxy for capital costs. Measured as mean inhabitants per
square kilometer.
GDPpc Gross Domestic Product per capita in constant 2000 US$.
Auction Dummy variable = 1 if wireless licenses awarded via auction; 0
elsewise.
NotCPP Dummy variable = 1 if the market not using calling party pays rule.
Fixprice Mean price of 3-minute call in constant 2000 US$ using fixed network
(peak period).
Dumfix Dummy variable = 1 if Fixprice is zero, and zero otherwise.
Aln ( Fixprice ) It is obtained as (1dumfix )* ln ( Fixprice )
25 Note that RPM, obtained as revenues over minutes of use, is not identical to APM, average price per minute posted by wireless operators in their plans. Either variable is a proxy for price.
17

T ABLE 5: 3SLS R ESULTS IN H AZLETT AND M UÑOZ (2008)
First Stage: Final 3SLS results for variables in deviations
MODEL 1 MODEL 2 MODEL 3 MODEL 4 MODEL 5 MODEL 6
MARK UP EQUATION
DV: lrpm ltotmindev
Coef.
-10.967
ltotmin2dev lhhidev
0.920
SD
22.387
1.787
-199.885
363.629 lhhi2dev lspectrumdev lspectrum2dev ldensitydev ldensity2dev lspecldendev constant
Number of obs.
"R-squared"
12.707
22.884
2.626
4.899
0.024
0.283
-98.849
146.744
8.778
14.637
-0.458
0.819
0.114
452
0.171
-31.028
DEMAND EQUATION
DV: ltotmin Coef.
lrpmdev lrpm2dev lgdppcdev lgdppc2dev alfixpricedev alfixprice2dev constant
Number of obs.
"R-squared"
-0.665
0.147
11.497
-0.487
1.240
0.265
0.004
452
0.625
SD
0.604
0.172
3.757 *
0.240 **
1.364
0.292
0.009
Coef.
0.505
-19.189
1.297
1.014
-0.078
-25.319
1.451
-0.058
0.033
452
-0.094
SD
0.223 **
17.995
1.080
0.504 **
0.045 ***
4.484 *
0.383 *
0.030 **
0.013 *
Coef.
0.572
-21.932
1.471
0.218
-26.149
1.479
-0.058
0.036
452
-0.207
SD
0.236 **
Coef.
0.633
2.537
18.816
1.131
0.163
4.676 *
0.399 *
0.031 ***
0.014 *
0.058
-23.381
1.206
-0.033
0.037
452
-0.080
SD
0.218 *
0.748 *
0.101
3.936 *
0.320 *
0.022
0.013 *
SD
0.068 *
3.359 *
0.214 *
1.181
0.265
0.009
Coef.
-1.182
13.889
-0.646
1.548
0.320
0.005
452
0.635
SD
0.067 *
3.335 *
0.213 *
1.170
0.263
0.008
Coef.
-1.173
13.443
-0.620
1.788
0.388
0.004
452
0.638
SD
0.067 *
3.290 *
0.210 *
1.153
0.258
0.008
Coef.
0.627
2.511
-0.083
-21.141
0.923
0.036
452
-0.070
SD
0.218 *
0.751 *
Coef.
0.627
2.511
0.049 ***
3.577 *
0.263 *
0.013 *
-0.083
-21.141
0.923
0.036
452
-0.070
Coef.
-1.166
13.275
-0.610
1.810
0.395
0.004
452
0.641
SD
0.067 *
3.280 *
0.209 *
1.148
0.256
0.008
MODEL 1
DV: uimarkup notcpp auction
_cons
Number of obs.
R-squared
Adj R-squared
Coef.
-68.001
SD
27.674 **
8.862
20.735
1046.140
16.393 *
28
0.197
0.132
DV: uidemand notcpp
_cons
Number of obs.
R-squared
Adj R-squared
Coef.
-3.295
-53.942
28
0.154
0.121
SD
1.515 **
0.573 *
Second Stage: Regressions including Time-Invariant Variables
MODEL 2
Coef.
-0.173
7.162
135.122
28
0.024
-0.055
Coef.
-3.382
-66.055
28
0.168
0.136
SD
12.734
9.541
7.543 *
SD
1.478 **
0.559 *
MODEL 3
Coef.
0.225
7.494
150.037
28
0.024
-0.054
Coef.
-3.469
-62.857
28
0.171
0.139
SD
13.307
9.970
7.882 *
SD
1.500 **
0.567 *
MODEL 4
Coef.
SD
2.811
12.787
7.438
41.968
9.581
7.574 *
28
0.031
-0.046
Coef.
-3.633
-60.759
28
0.189
0.158
SD
1.476 **
0.558 *
MODEL 5
Coef.
SD
5.918
12.851
7.700
38.082
9.629
7.612 *
28
0.044
-0.033
Coef.
-3.639
-60.048
28
0.191
0.160
SD
1.470 **
0.556 *
Coef.
-1.205
14.762
-0.705
1.505
0.311
0.005
452
0.625
Coef.
-1.166
13.275
-0.610
1.810
0.395
0.004
452
0.641
SD
0.067 *
3.280 *
0.209 *
1.148
0.256
0.008
MODEL 6
Coef.
8.806
SD
12.247
4.629 * 42.895
28
0.020
-0.018
Coef.
-3.639
-60.048
28
0.191
0.160
SD
1.470 **
0.556 *
SD
0.218 *
0.751 *
0.049 ***
3.577 *
0.263 *
0.013 *
18

T
ABLE
5 (
CONT
.): 3SLS R
ESULTS IN
H
AZLETT AND
M UÑOZ (2008)
Third Stage: Final 3SLS results including time invariant explanatory variables
MODEL 3 MODEL 4 MODEL 5 MODEL 6 MODEL 1 MODEL 2
MARK UP EQUATION
DV: lrpm ltotmin ltotmin2 lhhi lhhi2 lspectrum lspectrum2 ldensity ldensity2 lspeclden
FEmarkup notcpp auction
Coef.
-2.448
0.286
-162.561
92.616 ***
10.346
-2.003
0.328
-62.793
30.140 **
5.518
-0.216
0.640
-42.063
20.067 **
5.917
SD
2.891
0.222
5.849 ***
1.780
0.206
2.651 **
0.131 ***
0.306 **
2.732 ** constant
Number of obs.
"R-squared"
795.872
452
-5.659
437.950 ***
Coef.
0.489
133.248
452
0.609
-18.886
1.279
0.228
-0.012
-24.303
1.387
-0.048
0.961
-0.087
6.891
SD
0.034 *
4.593 *
0.287 *
0.330
0.032
1.898 *
0.108 *
0.015 *
0.074 *
0.083
0.517 *
21.910 *
DEMAND EQUATION
DV: ltotmin lrpm lrpm2 lgdppc lgdppc2 alfixprice alfixprice2
FEdemand notcpp constant
Number of obs.
"R-squared"
Coef.
-0.200
0.295
11.039
-0.459
1.241
0.265
1.001
-3.340
-51.737
452
0.970
SD
0.439
0.140 **
0.360 *
0.022 *
0.051 *
0.014 *
0.011 *
0.102 *
1.594 *
Coef.
-1.154
14.226
-0.675
1.516
0.316
0.993
-3.340
-63.649
452
0.975
SD
0.047 *
0.317 *
0.019 *
0.044 *
0.012 *
0.009 *
0.066 *
1.347 *
Coef.
-1.151
13.395
-0.617
1.551
0.323
0.994
-3.437
-60.682
452
0.975
Coef.
0.563
-24.442
1.630
0.106
SD
0.043 *
5.400 *
0.338 *
0.074
-25.893
1.458
-0.046
0.991
0.283
7.436
159.740
452
0.536
2.219 *
0.124 *
0.016 *
0.083 *
0.104 *
0.614 *
25.986 *
SD
0.048 *
0.315 *
0.019 *
0.044 *
0.012 *
0.009 *
0.067 *
1.342 *
Coef.
0.618
2.555
-0.046
-22.494
1.153
-0.023
0.964
2.821
7.192
39.839
452
0.607
Coef.
-1.152
13.030
-0.596
1.789
0.390
0.995
-3.605
-58.938
452
0.975
SD
0.042 *
0.215 *
0.064
1.664 *
0.084 *
0.015
0.070 *
0.259 *
0.512 *
2.984 *
SD
0.050 *
0.312 *
0.019 *
0.044 *
0.012 *
0.009 *
0.068 *
1.330 *
Coef.
0.610
2.505
-0.147
-20.240
0.878
0.962
5.787
7.423
35.974
452
0.612
Coef.
-1.139
12.852
-0.586
1.809
0.397
0.995
-3.606
-58.170
452
0.975
SD
0.040 *
0.194 *
0.024 *
1.414 *
0.062 *
0.067 *
0.445 *
0.505 *
2.574 *
SD
0.050 *
0.310 *
0.019 *
0.044 *
0.012 *
0.009 *
0.068 *
1.322 *
Coef.
0.614
2.503
-0.149
-20.389
0.885
0.969
8.625
41.095
452
0.612
Coef.
-1.120
12.784
-0.582
1.804
0.396
0.993
-3.586
-57.841
452
0.976
SD
0.050 *
0.308 *
0.019 *
0.044 *
0.012 *
0.009 *
0.068 *
1.317 *
SD
0.039 *
0.195 *
0.024 *
1.373 *
0.060 *
0.065 *
0.620 *
2.741 *
Note: "DV" stands for Dependent Variable. Significance: * 1%, ** 5%, *** 10%. "R-squared" is a pseudo-R-
Squared that can be less than 0. Source: Hazlett and Muñoz (2008).
Regression results show, through the Mark Up equation, a positive relationship between equilibrium prices and HHI and a negative relationship with the amount of spectrum. The magnitude of either relationship is substantial, leading us to conclude that policy makers should focus on promoting lower retail cellular prices by allocating more radio spectrum (or allowing markets to allocate more radio spectrum), and/or by permitting more liberal use of spectrum already allocated. The increase in spectrum availability both increases efficiency directly, lowering marginal costs of wireless services, and indirectly, by lowering fixed costs. The latter facilitates entry by competitors, intensifying price rivalry and reducing HHI . It also enables the provision of innovative services such as wireless broadband access (Rosston, 2001), although these effects are beyond the scope of our empirical analysis. This implies, importantly, that the welfare effects produced by more liberal spectrum policies are very conservatively estimated in this study.
By obtaining estimates of consumer welfare changes associated with differences in spectrum allocations across countries, the calibrated model can then be deployed to estimate the change in consumer surplus and service revenues in a given market for a hypothetical increase in
19

the amount of spectrum allocated to mobile telephony. In particular, we perform this exercise in each of the six Latin American countries included in the Merrill Lynch dataset.
26
4. SIMULATIONS
The model’s empirical estimates suggest that the amount of allocated spectrum is critical, but also important is the level of industrial concentration, which is heavily influenced by the regulator determining the number of licenses and setting other rules impacting operator size, entry barriers, and competitive rivalry. In this section, we consider choices a regulatory authority makes with respect to the total quantity of spectrum allocated to the mobile telephone sector.
4.1. DIRECT EFFECTS OF CHANGES IN SPECTRUM ON RETAIL PRICE
Figure 4 illustrates the direct effect of an increase in spectrum on the average revenue per minute ( RPM ) in mobile telephony (including 95% confidence intervals). The function is obtained with empirical results displayed in Table 5, fixing all the other (non-Spectrum) exogenous variables at their mean sample values and then varying the quantity of spectrum (in
MHz) allotted to the mobile telephony sector. As is seen, price is decreasing in the amount of allocated spectrum, with the rate of decrease declining. Retail prices are reduced because marginal costs are lower with more abundant inputs and because additional spectrum generally accommodates more intense competition, reducing HHI.
To simulate the direct effect of a spectrum allocation shift in a particular country, the
Spectrum-Price relationship in Figure 4 changes due to particular characteristics of that country
(which differ from the sample mean). We are able to make these adjustments for each of the six
Latin American countries in the quarterly database, and present them as six panels of Figure 5.
We observe that, while the opportunity for consumer welfare gains is available across all countries, the magnitude of the gains (via retail price reductions) varies.
26 Out of sample predictions are difficult to obtain due to the use of a pseudo-fixed effects model.
20

F IGURE 4: R ETAIL P RICE AS A F UNCTION OF S PECTRUM A LLOCATION :
T HE G ENERAL C ASE
4.2. COUNTRY PRICING SIMULATIONS
The country simulations ask the question: If regulators were to make more spectrum available to mobile telephone networks, where would retail prices (RPM) be?
The results suggest that, starting from the current situation, Venezuela has the potential to reach one of the lowest retail prices (for countries in the sample), with revenue per minute falling around
US$0.085, while Argentina is (lower) bounded at about US$0.15. The sources of these differences are in the country specific mean values for the other explanatory variables, and differences in institutional aspects captured in the fixed effects constants. The result is not surprising because the most constrained the current state, the higher the marginal benefits from reallocations.
21

0.5
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
F IGURE 5: P RICE AS A F UNCTION OF S PECTRUM A LLOCATION :
S IX L ATIN A MERICAN M ARKETS
A
RGENTINA
B
RAZIL
C
100
M
HILE
100
200 300
Spectrum (MHz)
400
200 300
Spectrum (MHz)
400
EXICO
500
500
600
600
0.16
0.14
0.12
0.1
0.24
0.22
0.2
0.18
0.08
0.06
0.04
0
0.3
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0 100
C
100
V
200 300
Spectrum (MHz)
400
OLOMBIA
200 300
Spectrum (MHz)
400
ENEZUELA
500
500
100 200 300
Spectrum (MHz)
400 500 600
0.25
0.2
0.15
0.1
0.05
0 100 200 300
Spectrum (MHz)
400 500
600
600
600
22

4.3. COUNTRY WELFARE ANALYSIS
To evaluate the impact of spectrum-policy induced price changes in economic welfare terms, we used the model reported in Table 5. Simulations were performed in 1000 groups of
1000 cases each, with shocks to the system of equations from a Bivariate Normal distribution with mean zero for each dimension. The variance-covariance matrix was estimated from the residuals of the Stage 3 regressions contained in Appendix 2 of Hazlett and Muñoz (2008), which provides detailed explanation of the procedure. These experiments were conducted at least a hundred of times each to achieve convergence.
Results for the six Latin American countries are displayed in Figure 6. Potential gains partly depend on the size of the market, so it is unsurprising that the most populous countries, Brazil and Mexico, are predicted to have the highest incremental consumer surplus. Other results, however, are more interesting. For instance, additional bandwidth generates approximately as high an increase in consumer surplus in Venezuela as in Argentina, despite the Venezuelan economy’s smaller size (see Table 1). This outcome appears to be driven by the extremely parsimonious spectrum allocation for mobile telephony in Venezuela (57 MHz).
It is noteworthy that, across countries, revenues increase slowly when more spectrum is made available to operators. Given that the change in producer surplus is even lower than the change in revenues, more spectrum can reduce the aggregate sum of operators’ benefits. However, even when total benefits increase, it is possible that more spectrum leads to more players, so incumbents’ benefits decline. This may help explain why governments are so conservative.
Concerns regarding lower revenue extraction (in license auction regimes) or in protecting profits of wireless incumbents would tend to bias regulation away from welfare maximization.
Finally, welfare changes follow a linear path rather than one of diminishing marginal returns due to a condition of the selected model, wherein elasticity of demand is assumed constant.
Alternatively, in Model 1 of Table 5 the elasticity of demand declines (in absolute value) as more spectrum is made available to the market, generating a nonlinear impact on welfare. The intuition is explained by the fact that, beyond some level of output, the demand elasticity falls to below unity, creating a situation where further increases in MHz would reduce market revenues.
23

FIGURE 6: CHANGE IN CONSUMER SURPLUS AND REVENUES: THE EFFECT OF
ADDITIONAL SPECTRUM ALLOCATIONS
Argentina Chile
6,000
5,000
4,000
3,000
2,000
1,000
0
3,000.00
2,500.00
2,000.00
1,500.00
1,000.00
500.00
0.00
20 40 60 delta Spectrum
80 100 20 40 60 delta Spectrum
80
Brazil Colombia
30,000
25,000
20,000
15,000
10,000
5,000
-
3,000
2,500
2,000
1,500
1,000
500
-
20 40 60 delta Spectrum
80 100 20 40 60 delta spectrum
80
Mexico Venezuela
18,000
16,000
14,000
12,000
10,000
8,000
6,000
4,000
2,000
-
20 40 60 delta Spectrum
80 100
4,500
4,000
3,500
3,000
2,500
2,000
1,500
1,000
500
-
20 40 60 delta Spectrum
80
100
100
100

5. POLICY IMPLICATIONS
As shown, many Latin American countries could capture substantial social gains by expanding the bandwidth available to supply mobile phone services. Such enhancements would improve voice services, broadband services, and broadly boost economic development (see, e.g., Jensen 2007). Policies to achieve such changes can be implemented only by overturning existing political equilibria. What strategies may affect such movement is not the subject of discussion here. Rather, we tackle the more general topic of how more generous spectrum allocations may be structurally accommodated.
5.1. SPECTRUM ALLOCATION REFORM
There are two primary sets of institutional reforms that effectively make additional bandwidth available to mobile markets. The first entails a broadening of the spectrum use rights granted in the traditional license. In the limit, this produces a private property right to allocated spectrum, what has been called a Liberal License (Hazlett,
2008a) or EAFUS, exclusively assigned, flexible use spectrum (Hazlett and Spitzer,
2006). This approach leaves in place the underlying spectrum allocation structure, meaning that EAFUS rights are granted case-by-case. Such policy changes are sometimes associated with rules for “secondary markets,” where regulators may permit licensees to not only trade rights explicitly granted but to enjoy flexible use of the spectrum allocated to the license.
This liberalization of license terms facilitates the market allocation of frequencies anticipated by Coase (1959). Licensees become de facto owners of spectrum, and then seek to deploy bandwidth to maximize value. License rigidities, constituting both competitive and technological barriers to entry, are removed. While bandwidth remains limited to the frequency space allocated to liberal licenses, greater flexibility allows for enhanced efficiencies (optimizing input combinations with greater degrees of freedom).
This drives productive expansion of wireless services.
The alternative reform is for government to simply allocate more radio spectrum to mobile network licenses. Under any given set of rules, such increases lower the opportunity cost of spectrum inputs and increasing output, ceteris paribus.
Of course, liberalizing both licenses and spectrum allocations can be undertaken in tandem, and many reform efforts have taken this approach. One particularly articulate proposal in this direction is the FCC staff Working Paper by economist Evan Kwerel and engineer John Williams (2002). In their “Big Bang” proposal they advocate increasing the spectrum (under 3 GHz) allocated to liberal licenses from about 180 MHz to 438
MHz, being careful to note that the licenses would be most efficiently utilized under
“flexible use” rules.
25

5.2. REFORM WITHIN EXISTING REGULATORY STRUCTURES
Kwerel and Williams (2002) focus on regulatory agency policy making in the
U.S. rather than appealing to Congress for statutory reform. This approach has produced some results, as the U.S. Federal Communications Commission broadened the spectrum use rights conveyed in mobile network licenses through the cellular (1980s) and PCS
(1990s) assignments. While relatively little bandwidth was made available prior to 2006
(Bazelon, 2006; Faulhaber, 2006),
27
despite the argument made in Kwerel and Williams
(2002) and the high social values associated with incremental allocations (Hazlett and
Muñoz, 2008), the movement towards a spectrum market was facilitated via agency liberalization.
A more ambitious reform program was announced by Ofcom, the wireless regulatory authority in the United Kingdom, in 2004. By 2010, the Ofcom plan aims to remove restrictions on license use for 72% of the bandwidth in the most valuable frequencies, between 300 MHz and 3 GHz.
28 This strategy relies on loosening constraints embedded in licenses, and to expand the frequency space allocated to assigned licenses. It follows the strategy of exhaustive licensing, which we deem a
“Model II” approach, where the only spectrum not assigned to liberal licenses is withheld by specific justification. This is the mirror image of the standard approach, which issues
(rather than retains) license rights only when specific justifications are found to be in the
“public interest.”
5.3. REFORMS REQUIRING STRUCTURAL CHANGES
The alternative path to enabling market allocation of radio spectrum entails structural reform of the spectrum allocation process. While the barriers to such institutional change are considerable, the interesting fact is that two Latin American economies succeeded in spectrum liberalization via national statutes just over a decade ago. In 1996, Guatemala created explicit private property rights to radio spectrum, titulo de usufructo de frecuencia (TUFs), a “Model III” approach. Such devices define ownership by specifying: a. the band or frequency ranges; b. hours of operation; c. geographical coverage area; d. maximum effective radiated power by the TUF holder; e. maximum field strength or signal strength on the border of the coverage area; f. order and title number; g. issue date and expiration; h. name of title holder;
27 The U.S. cellular market could access less than 190 MHz of radio spectrum, about 100 MHz less than other countries of similar income levels. See Figure 1.
28 Ofcom, The Spectrum Review Framework: A Consultation on How Radio Spectrum Should be
Managed (Nov. 23, 2004), www.
ofcom .org.uk/consult/condocs/sfr/sfr2/presentation.ppt
.
26

i. blank spaces for endorsement of reassignment to another party.
29
Any person or entity is entitled to request a frequency, triggering the TUF assignment process.
30
The independent regulator is constrained to issue requested, nonconflicting rights and to keep an up to date record of TUF holders. Petitions are subject to opposition on the grounds of radio interference with existing services, but strict time limits for adjudication, as well as binding arbitration mandates, are designed to block excessive administrative barriers to entry. Rights to contested TUFs (with multiple claimants) are required to be assigned via auction. TUFs carry 15-year terms, with an additional 15 years at the option of the TUF holder. According to the International
Telecommunications Union, Guatemala has “probably the world’s most liberal radio spectrum regulatory model.” 31
El Salvador arrived at a similar policy outcome via less radical means in a 1997 statute.
32
Concessions for the use of frequencies extend for a 20-year period, and can be transferred or subdivided in frequency, geographic, and time dimensions without regulatory approval.
33
Rights holders are free to choose technologies. While a National
Table of Frequency Allocation (TFA) describes the type of services frequencies are allocated for,
34
rights holders may deviate from TFA specifications without penalty. This results in generic license flexibility.
35
Additionally, the regulator is directed, as in
29 Ley General de Telecomunicaciones. Legislative Decree No. 94-96 (Oct. 17, 1996, Article 57). A copy of a TUF is featured in Hazlett (2001, p. 447).
30 Ley General de Telecomunicaciones. Legislative Decree No. 94-96 (Oct. 17, 1996, Article 61).
31 International Telecommunications Union, Radiocommunications : SPU newslog on radiocommunication issues (Dec. 19, 2003),
http://www.itu.int/osg/spu/newslog/categories/radiocommunications/2003/12/19.html.
32 Ley General de Telecomunicaciones. Legislative Decree No. 142 (Nov. 6, 1997).
33 Ley General de Telecomunicaciones. Legislative Decree No. 142 (Nov. 6, 1997, Articles 15-16). The transfer of concession rights is treated as a private contract and must be entered in the telecommunications registry of the regulator ( Reglamento de la Ley de Creacion de la Superintendencia General de
Electricidad y Telecommunicaciones. Executive Decree No. 56 . May 13, 1998, Articles 19, 27).
Concession holders are liable for violations, including out of band emissions. Ley General de
Telecomunicaciones. Legislative Decree No. 142 (Nov. 6, 1997, Article 15).
34 Ley General de Telecomunicaciones. Legislative Decree No. 142 (Nov. 6, 1997: Articles 10). See also
Regulation of the Law of Telecommunications. Executive Decree No. 64 (May 15, 1988: Article 52). At the time a concession is awarded, the regulator issues a document called a “Resolution” in which the characteristics of the concession are specified. This includes: “(a) a reference to the fulfillment of the dispositions of the CNAF [national table of frequency allocation] that are applicable, and, (b) the technical background of the system in terms of the service to offer; central frequency and bandwidth of the transmitting stations; geographical locations of the fixed transmitting stations; coverage area or link direction; operation timetable; nominal power of the transmitters; effective maximum radiated power; maximum intensity of the electrical field in the surrounding of the covered area; modulation type; type, gain and pattern of the radiation of the antennas of the transmitter stations; type, gain and pattern of reception of the antennas of the receiving stations, whenever they have to be protected; altitude and location of the antennas above the terrain level and above sea level; and a spectrum diagram of the signals emitted by the transmitters after the filtering state, as it corresponds.” Regulation of the Law of
Telecommunications. Executive Decree No. 64 (May 15, 1988: Article 55).
35 The classification of services, while non-binding, may provide a coordinating function. In any event, service categories are defined by International Telecommunications Union allocations (non-binding agreements between countries) and by international markets for telecommunications equipment. A small
27

Guatemala, to issue requested licenses, using auctions for contested applications. To the extent these rules are administratively or legally enforced, they can be said to enable a market in radio spectrum (Hazlett et al., 2007). These idiosyncratic spectrum policies illuminate possible paths to liberalization, an important normative exercise left for later research.
6. CONCLUSION
Most Latin American countries appear to inefficiently constrain access to radio spectrum. In a model calibrated via cross-country differentials in frequency allocations and retail mobile phone rates, we find substantial social gains available from removing such limits. Policy reform, which entails an expansion of administrative allocations or more general liberalization measures, can be pursued either by independent regulatory actions or via statute.
Regulators in Chile and Paraguay have administratively made relatively generous bandwidth available for mobile phone use. Guatemala and El Salvador, enacting statutory reforms in the 1990s, more fundamentally liberalized their regimes. Networks in both countries utilize relatively abundant bandwidth under rules allowing discretion in the deployment of services and technologies. Such policies accommodate productive use of radio spectrum, lowering mobile phone rates and increasing consumer welfare. Yet, our simulations for six of the largest Latin American markets suggest that the social value of increasing mobile spectrum allocations is still high -- on average, US$53.70 per capita for an increment of 20 MHz. Since there are no competing economic employments for such bands, which effectively lie fallow, additional allocations would produce the proverbial
“free lunch.” Why public policy makers rarely exploit such opportunities is an excellent question for further analysis.
More generally, the relationship between spectrum policy measures and retail market outcomes has been little studied. Scholars and regulators would benefit from greater visibility over the linkages between policy inputs and economic welfare outputs.
We hope that this research paper might stimulate more interest in this endeavor. country’s spectrum, even in the most open regulatory environment, will be likely to largely conform to world markets to capture economies of scale in manufacturing transmission and receiving equipment.
28

7. REFERENCES
Annan, K., 2005. ICTs Can Support Faster Economic Growth and Thus Strengthen the
Material Basis for Development. Foreword, UNCTAD Information Economy Report.
Bazelon, C., 2006. Licensed vs. Unlicensed Spectrum: A New Economic Model for
Determining the Trade-offs. Paper Presented to the 34 th
Annual Telecommunications
Policy Research Conference (Sept. 30).
Blanco, V., 2005. Impactos de la Flexibilización en la Asignación del Espectro
Radioeléctrico. Unpublished thesis, CIDE, Mexico.
Coase, R. H., 1959. The Federal Communications Commission. Journal of Law &
Economics 2, 1-40.
Dasgupta, S., Lall, S., Wheeler, D., 2001. Policy Reform, Economic Growth, and the
Digital Divide: An Econometric Analysis. Development Research Group, Infrastructure and Environment, World Bank (March).
Faulhaber, G. R., 2006. The Future of Wireless Telecommunications: Spectrum as a
Critical Resource. Information Economics and Policy 18, 256-271.
Greenstein, S., Spiller, P., 1996. Estimating the Welfare Effects of Digital Infrastructure.
National Bureau of Economic Research, Working Paper No. 5770.
Hardy, A., 1980. The Role of the Telephone in Economic Development.
Telecommunications Policy 4, 278-86.
Hazlett, T. W., 1998. Assigning Property Rights to Radio Spectrum Users: Why Did
FCC License Auctions Take 67 Years? Journal of Law and Economics 41, 529-576.
Hazlett, T. W., 2001. The Wireless Craze, the Unlimited Bandwidth Myth, the Spectrum
Auction Faux Pas, and the Punchline to Ronald Coase’s ‘Big Joke’: An Essay on
Airwave Allocation Policy. Harvard Journal of Law & Technology 14 (Spring), 335-469.
_____, 2008a. Optimal Abolition of FCC Spectrum Allocation. Journal of Economic
Perspectives 22 (Winter), 103-128.
_____, 2008b. Property Rights and Wireless License Values. Journal of Law &
Economics 51, 563-598.
_____, Ibarguen, G., Leighton, W., 2007. Property Rights to Radio Spectrum in
Guatemala and El Salvador: An Experiment in Liberalization. Review of Law and
Economics 3, 437-484.
29

______, Mueller, J., Muñoz, R., 2006. The Social Value of TV Band Spectrum in
European Countries. INFO, The Journal of Policy, Regulation and Strategy for
Telecommunications 2, 62-73.
______, Muñoz, R., 2008. A Welfare Analysis of Spectrum Allocation Policies. George
Mason Law & Economics Research Paper No. 06-28 (Jan. 18). Available at SSRN: http://ssrn.com/abstract=908717.
_____, Spitzer, M., 2006. Advanced Wireless Technologies and Public Policy. Southern
California Law Review 79, 595-665.
Jensen, R., 2007. The Digital Provide: Information (Technology), Market Performance, and Welfare in the South Indian Fisheries Industry. Quarterly Journal of Economics 72,,
879-924.
Kwerel, E., Williams, J., 2002. A Proposal for A Rapid Transition to Market Allocation of Spectrum. Federal Communications Commission OPP Working Paper No. 38 (Nov.
15).
Leff, N. H., 1984. Externalities, Information Costs, and Social Benefit-Cost Analysis for
Economic Development: An Example from Telecommunications. Economic
Development and Cultural Change 32, 155-176.
Merrill Lynch, 2003. Global Wireless Matrix 2Q03. Quarterly Update on Global
Wireless Industry Metrics. Merrill Lynch Global Securities Research and Economics
Group.
Muñoz, R., 2004. Modeling Liberalization in Spectrum Regimes. Mimeo CIDE.
Noll, R., Peck, M., McGowan, J., 1973. Economic Aspects of Television Regulation.
Brookings Institution Press.
Norton, S., 1992. Transaction Costs, Telecommunications, and the Microeconomics of
Macroeconomic Growth. Economic Development and Cultural Change 41, 175-196.
Plümper, T., Troeger, V., 2007. Efficient Estimation of Time-Invariant and Rarely
Changing Variables in Finite Sample Panel Analyses with Unit Fixed Effects. Political
Analysis 15, 124-139.
Reed, D., 1992. Putting it all Together: The Cost Structure of Personal Communications
Services. Federal Communications Commission, OPP Working Paper No. 28
(November).
Rosston, G., 2001. Statement of 37 Concerned Economists. Comment submitted to the
Federal Communications Commission, Promoting Efficient Use of Spectrum Through
30

Elimination of Barriers to the Development of Secondary Markets (Feb. 7); http://www.aei-brookings.org/publications/abstract.php?pid=118 .
United Nations [U.N.], 2004. The Wireless Internet Opportunity for Developing
Countries. United Nations Publications (Nov.).
Waverman, L., Meschi, M., Fuss, M., 2005. The Impact of Telecoms on Economic
Growth in Developing Countries. Vodafone Policy Paper Series (March).
31

A PPENDIX 1: S PECTRUM A LLOCATED TO W IRELESS T ELEPHONY
IN THE E UROPEAN U NION (2003)
36
Countries with Auctions MHz
Austria
Belgium
Denmark
Germany
Greece
Italy
Netherlands
United Kingdom
Average Allocation with Auctions
304.2
199.0
375.0
305.0
265.0
286.8
355.8
341.0
304.0
Countries with Beauty Contests
France
Finland
Ireland
Luxembourg
Portugal
Spain
Sweden
Average Allocation without Auctions
207.0
238.8
251.0
85.6
223.6
297.2
255.1
222.6
Average Allocation 286.3
Source : Authors’ elaboration based on Hazlett (2008) database.
36 The list of EU countries was taken from http://userpage.chemie.fu-berlin.de/adressen/eu.html
. Only the members as of December 2003 were included.
32

A PPENDIX 2: M OBILE V OICE M ARKET DATABASE
Our main source of information was: “Global Wireless Matrix 2Q03: Quarterly Update on Global Wireless Industry Metrics,” Merrill Lynch Global Securities Research &
Economics Group, Global Fundamental Equity Research Department. This includes quarterly data for the wireless market in 46 countries, first quarter 1999 through second quarter 2003. All data were obtained from this source except the following:
Spectrum, Auction :
The main source is each country’s telecommunications regulator and
Communications Ministry. The Economist Intelligence Unit ViewsWire database, the
European Commission and the European Radio Communications Office are secondary sources.
GDPPC
: The World Bank’s World Development Indicators (WDI) database, 2008.
Values expressed in constant 2000 US$.
GDP Deflator: base year 2000. The World Bank’s World Development Indicators (WDI) database, 2008. All monetary variables have been re-expressed in constant 2000 US$ using this deflator.
Density : It was constructed as population/area, where population is from Merrill Lynch and area is from the World Bank’s World Development Indicators 2003.
Fixprice : It was taken from the International Telecommunications Union’s World
Telecommunications Indicators 2002 database and then expressed in 2000 constant US$.
Our sample is comprised of all observations in the Merrill Lynch database for which we have data for all the relevant variables from the first quarter in 1999 through the second quarter in 2003. (While Merrill Lynch data technically begin in fourth quarter 1998, the data for that quarter are very incomplete.) Our sample includes these 28 countries:
Argentina
Australia
Austria
Belgium
Colombia
Czech Rep
Denmark
Finland
Hong Kong
Hungary
Ireland
Italy
Norway
Portugal
Singapore
Spain
Brazil
Canada
Chile
France
Germany
Greece
Mexico
Netherlands
New Zealand
United Kingdom
United States
Venezuela
Of the 46 countries in the Merrill Lynch database, many could not be used due to missing data (for variables not included in the ML database). The most difficult data to identify included Spectrum and Fixprice . To enable the inclusion of additional country data,
Fixprice was adjusted in Canada: The reported values are zero from 1991 to 1994; thereafter it is not reported. We used an assumed value of “0” after 1994.
33

The case of Sweden was special because Spectrum data was fully available and so it was included in the analysis for Figure 1 and Table 2. However, Fixprice was unavailable and then it was excluded in the analysis of Hazlett and Muñoz (2008).
34

A PPENDIX 3: S PECTRUM L AW IN L ATIN A MERICA
Argentina
Bolivia
Brasil
Chile
Colombia
Costa Rica
Ecuador
El Salvador
Guatemala
Honduras
Mexico
Nicaragua
Panama
Paraguay
Peru
Uruguay
Venezuela
Reglamento de Licencias para Servicios de Telecomunicaciones. Decreto PEN 764/2000 con fecha 3 de Septiembre del 2000, Anexo I.
Reglamento sobre Administración, Gestión y Control del Espectro Radioeléctrico. Decreto PEN 764/2000 con fecha 3 de Septiembre del 2000, Anexo IV.
Ley de Telecomunicaciones: Ley 1632 aprobada el 5 de Julio de 1995 y modificada por las leyes 2328 y 2342 del 2002.
Fuente: Reglamento a la Ley de Telecomunicaciones. Decreto Supremo 24132 del 27 de Septiembre de 1995 (incluye modificaciones hasta Febrero del 2001).
Agencia Nacional de Telecomunicaciones, ATO N. 50.312 del 13 de Mayo del 2005.
Ley N° 18.168, General de Telecomunicaciones y modificaciones posteriores.*
LEY 555 DE 2000 por la cual se regula la prestación de los Servicios de Comunicación Personal, PCS y se dictan otras disposiciones.
Reglamento a la Ley Reguladora de los Servicios Públicos, N° 7593 de 9 de Agosto de 1996
Reglamento General de Servicios de Telecomunicaciones (RGST) Publicado en La Gaceta Nº 27 del 7 de febrero de 2002.
Ley especial de Telecomunicaciones Reformada, Ley No. 184 Registro Oficial No. 996 (incluye modificaciones hasta Marzo 2000).
REGLAMENTO GENERAL A LA LEY ESPECIAL DE TELECOMUNICACIONES REFORMADA, DECRETO EJECUTIVO No. 1790, REGISTRO OFICIAL No. 404, 4-SEP-2001
Reglamento para el Servicio de Telefonía Móvil Celular
Thomas W. Hazlett, "Wireless License Values", AEI-Brookings papers, 2004.
Thomas W. Hazlett, "Wireless License Values", AEI-Brookings papers, 2004.
Ley Marco del Sector de Telecomunicaciones, Decreto 185-95 del 5 de Diciembre de 1995 y Actualización de la Ley Marco del Sector de Telecomunicaciones,
Decreto 118-97 del 25 de Octubre de 1997. Reglamento General de la Ley Marco de Telecomunicaciones.
Ley Federal de Telecomunicaciones, LEY PUBLICADA EN EL DIARIO OFICIAL DE LA FEDERACION EL 7 DE JUNIO DE 1995.
Reglamento De Uso Del Espectro Radioeléctrico Y De Los Servicios De Radiocomunicaciones
Reglamento de Títulos Habilitantes. Acuerdo Administrativo N° 006 de Enero 7 del 2005.
LEY No. 31 (De 8 de febrero de 1996) "Por la cual se dictan normas para la regulación de las telecomunicaciones en la República Panamá"
Decreto Ejecutivo No. 21 (De 12 de enero de 1996) "Por el cual se dicta el Reglamento sobre la Operación del Servicio de Telefonía Móvil Celular"
LEY Nº 642 DE TELECOMUNICACIONES
Decreto Nº 14135, POR EL CUAL SE APRUEBA LAS NORMAS REGLAMENTARIAS, DE LA LEY Nº 642/95 “DE TELECOMUNICACIONES”, Asunción, 15 de julio de 1996.
D.S. No. 013-93-TCC, Aprueba el Texto Unico Ordenado de la Ley de Telecomunicaciones. Promulgada: 28 de abril de 1993.
Texto Único Ordenado del Reglamento General de la Ley de Telecomunicaciones, DECRETO SUPREMO Nº 027-2004-MTC
DECRETO SUPREMO Nº 040-2004-MTC, 21 Diciembre del 2004.
REGLAMENTO SOBRE EL ESPECTRO RADIOELÉCTRICO 25/03/03
REGLAMENTO DE LICENCIAS DE TELECOMUNICACIONES, 25/03/03
LEY ORGÁNICA DE TELECOMUNICACIONES, 1 Junio del 2000.
35