Transcript Slide 1

Economic Models to Value
Spectrum
Rohit Prasad
Associate Professor,
MDI Gurgaon
International Workshop on Spectrum Management
National Institute of Communication Finance
April 29, 2013
Agenda
• Tragedy of the Commons
• Pricing Practices
• Difference between market and administered
prices
• Economic Models
• Reflections on the 2013 auction
Thermometer Test
Understanding
Low
High
Low
HELP!
I PREFER IPL
High
I PREFER KAMAL
LOVE IT!
Interest
Tragedy of the Commons
What is being valued?
• The right to exclusively transmit signals for a
specified
• Service
• Technology
• Frequency range
• Geographical area
With rules on power levels etc.
Why is it being valued?
• Why are exclusive rights given?
• Why are these rights valuable?
Tragedy of the commons
• If operators are given free access to a scarce
resource there will be congestion/overuse and
reduction of social welfare
– Toll is recommended
• In the absence of spectrum sensing
technology, even without scarcity, there could
be congestion (unlike roads)
Hence exclusive rights
Why rights valuable?
• Those given rights earn supernormal profits,
and these supernormal profits represent the
value of the rights
Example
Consider a common pasture of a village where
the five people can potentially graze their
cattle. Each head of cattle costs Rs. 100 which
can be raised at 13% interest. Thus Rs. 13 is
the marginal cost of putting another cattle to
pasture. As the number increases the value
reduces due to the lower availability of
pasture per cattle.
No. of
cattle
1
2
3
4
5
Price per
cattle at
the end of
1 year
126
119
114
111
109
Average
Revenue
per cattle
26
19
14
11
9
Total
revenue
from
grazing
26
38
42
44
45
Marginal
revenue
26
12
4
2
1
Exercise 1: How many cattle will graze?
Exercise 2: How many cattle should graze?
Excess over
interest
due to
bank
13
-1
-9
-11
-12
Solutions
• Toll
– Exercise 3: What should be the toll
• Privatization
– Exercise 4: What is the value of the pasture?
No. of
cattle
1
2
3
4
5
Price per
cattle at
the end of
1 year
126
119
114
111
109
Average
Revenue
per cattle
26
19
14
11
9
Total
revenue
from
grazing
26
38
42
44
45
Marginal
revenue
26
12
4
2
1
Excess over
interest
due to
bank
13
-1
-9
-11
-12
Thermometer Test
Understanding
Low
High
Low
HELP!
I PREFER IPL
High
I PREFER KAMAL
LOVE IT!
Interest
Mechanisms to discover value
Mechanisms to discover value
• Market mechanism
– Operators gather data and build models, based on
which they place bids in an auction
• Bureaucratic mechanism
– Bureaucrats gather data and build models, based
on which they decide on an administered price
– May use a discount
– Administered price may be different from value
Pricing Practice
• Captive users use radio frequencies for their
communication network
• not in the business of providing telecom service
• Commercial users in the business of providing
telecom services using radio frequencies
• Most spectrum in control of government agencies –
captive users
Pricing Practice
• Auctions
–
–
–
–
License (2G spectrum) in 1995
Migrated to revenue share in 1999
License (2G spectrum) in 2001, 2012-2013
3G, BWA spectrum in 2010
• 2001 to 2012: 2G ‘spectrum price’ benchmarked
to value discovered in 2001 auction
– administered price progressively diverges from market
value
Pricing Practice
Auction
Administered Price
Commercial
spectrum
1995, 2001, 2010,
2012-13
1999, 2001-2008
Captive spectrum
NA
Since 1980s
Sample Calculation for Captive
Spectrum
• A. Royalty charges for single carrier radio spectrum (Usually
below 1 GHz).
– Annual Royalty per Carrier (in Rs.) = MxW, where, M is distance factor and W
is bandwidth factor
– Values of M are:
Distance Category
Max. Distance
Royalty(in Rs.)
(Km)
I
<= 2
1500
II
<= 5
3000
III
>5 <= 25
6000
IV
>25 <= 60
12000
V
>60 <= 120
22500
VI
>120 <= 500
37500
VII
>500
50000
Limitations
•
•
•
•
Band neutral
Demand neutral
Geography neutral
Technology neutral
Administered Price for UAS License
Subscriber Linked Criteria
• Additional spectrum beyond startup amount
allocated on the basis on subscriber
milestones and attracts additional usage
charges
Sample Subscriber Linked Criteria
Difference between market and
administered price
Difference between market and
administered price
Market Price
Administered Price
Based on bidding
Based on administrative decision
Based on market information
Based on administrative information
Does not take social impact into account
( externalities)
Takes social impact into account
( externalities)
Susceptible to market collusion
Susceptible to misuse of discretionary
power
Administrative pricing in India has tended to be
• focused on capabilities rather than demand
• unchanging over long periods (static) rather than responsive to
changing conditions ( dynamic)
• driven by engineers rather than economists
• Non-transparent
EXTERNALITIES AND MARKET
INEFFICIENCY
An externality refers to the uncompensated
impact of one person’s actions on the wellbeing of a bystander.
• Externalities cause markets to be inefficient,
and thus fail to maximize total surplus.
• Two types of externalities
1. Negative : Adverse Impact on the bystander.
2. Positive: Beneficial Impact on the bystander.
Externalities: Examples
Negative Externalities:
– Automobile exhaust
– Cigarette smoking
– Carbon emissions of telecom towers
Positive Externalities:
– Research into new technologies
– Diffusion of mobile telephony and data services
• Provision of basic services
• Productivity increase
EXTERNALITIES AND MARKET
INEFFICIENCY
• Negative externalities lead markets to produce
a larger quantity than is socially desirable at
lower price
– Tax needed
• Positive externalities lead markets to produce
a smaller quantity than is socially desirable at
higher price
– Subsidy needed
Economic Models of Valuation
Economic Models of Valuation
• “Nothing is more useful than water; but it
will purchase scarce anything. A diamond on
the contrary, has scarce any value in use, but
a great quantity of other goods may …be had
in exchange of it.” Adam Smith
Economic Models of Valuation
• Focuses on demand as well as supply
• Values something in terms of the additional
surplus earned by its possession ( opportunity
cost)
• Explicitly factors externalities
Economic Models
• Cash Flow
• Production Function
• Benchmarking with closest market value
Thermometer Test
Understanding
Low
High
Low
HELP!
I PREFER IPL
High
I PREFER KAMAL
LOVE IT!
Interest
Cash Flow Method
• Imagine you get access to a park
• Using some playground equipment I am able to
earn Rs. 100 of revenue
• My costs are Rs. 40
• What is my profit?
• If I had put the investment I made in the park in
the next best alternative I would have made Rs.
35
• What is my ‘supernormal profit’?
• Exercise 5: What is the value of the park to me?
Cash Flow Method
• 25/(1+r) + 25/(1+r)^2 + …
• Problem: the earning in the alternative
investment cannot be determined without
knowing the value of the park
• So we are going to have to solve an equation!
• Assume park was being given for a year
• Let the rate of interest be 10%
• Denote value by v
• Earning in alternative investment = v.10%
• Then v = (60 – v. 10%)/(1+ 10%)
Profit
• Suppose park being given in perpetuity
• V = (60 – v. 10%)/(1+ 10%) + (60 – v. 10%)/(1+
10%)2 + (60 – v. 10%)/(1+ 10%)3 + …
• Using formula on sum of geometric series,
• V = (60 – v. 10%)/ 10%
• V = 60/(2. 10%)
Application to Spectrum
• Start with ‘representative firm’ with subscribers in
proportion to its spectrum holding, 6.2 MHz
• Cash Flow in a Year = Revenue – (License Fee
•
•
•
•
+Spectrum Charge + Network Cost +
Administration, Marketing, & Operational Cost)
Revenue = subscribers . ARPU
Network cost assessed by taking average number of
BTSs held by operator with 6.2 MHz of spectrum
Admin, marketing and operations 28% of revenue
Assume return on investment in alternative use is
20%
Extracting Value of 1800 MHz
• Let the percentage of 900 MHz spectrum be 60%
• Then
60%. Value of 900 MHz + 40%. Value of 1800 MHz =
value of blended spectrum
• Let ‘k’ represent the additional productivity of
900 MHz spectrum
– This is different in different markets
60%. K. Value of 1800 MHz + 40%. Value of 1800
MHz = value of blended spectrum
Solve equation for value of 1800 MHz
Table 7: Price of contracted spectrum 1800 MHz
Rs. Crore per MHz 2010 (20 year license)
Service Area
Price of contracted spectrum
% of 900 MHz spectrum
Metro
Delhi
Mumbai
Kolkata
Category A
Maharashtra
Gujarat
Andhra Pradesh
Karnataka
Tamil Nadu
Category B
Kerala
Punjab
Haryana
Uttar Pradesh (West)
Uttar Pradesh (East)
Rajasthan
Madhya Pradesh
West Bengal, Andaman &
Nicobar
Category C
Himachal Pradesh
Bihar
Orissa
Assam
North East
57.1
54.8
49.3
Price of 1800 Mhz
spectrum
60.78
71.56
66.67
60.78
57.41
60.19
55.86
109.81
74.71
37.25
0.00
82.53
107.41
110.12
98.14
136.26
0.00
55.03
53.64
9.46
43.40
116.45
79.77
66.24
57.39
32.31
58.49
57.41
66.67
66.67
78.57
6.97
38.07
18.15
7.76
7.91
57.9
63.1
57.9
57.9
56.6
Thermometer Test
Understanding
Low
High
Low
HELP!
I PREFER IPL
High
I PREFER KAMAL
LOVE IT!
Interest
Production Function Method
• Production function gives the total number of
subscribers that can be serviced, for any given
combination of spectrum and BTSs
– Derived by statistical estimation using data on
subscribers, spectrum and BTSs of operators in
relevant circles over relevant years
• Allows estimation of the number of BTSs saved by
possessing an additional unit of spectrum
• Cost of BTSs saved represents value of spectrum
Method of estimation
• We estimate a Cobb Douglas production
function
ln X  ln A   ln y   ln z
• Very general specification that encompasses
diminishing returns, economies of scale
• Used widely in the literature
Data
• We use a panel data collected over 7 years across
all the 23 circles for different GSM operators
providing services in their respective circles
• Estimate statistically significant values of A, β, γ
• Ln A = -6.83
• β = 1.23
• γ = 0.66
• Limitation: Data not available at sub-circle level
Model for Allocative Efficiency
First Order Condition
Price Ratio
Quantity Ratio
Feb 12, 2009
Seminar @ NUS
Productivity Ratio
49
Benchmarking with Auction Values
• Many methods
– Sector -specific
– General
• Sector-specific
– Value in 2001 . Percentage increase in revenue per
MHz/Percentage increase in BTSs
– Adaptive expectations
• Such methods are relevant at proximate times
Limitations of Economic Models
• Sensitive to assumptions
• Difficult to factor externalities
Auctions
• Need to guard against collusion ( reserve
price)
• Possibility of winner’s curse
Failure of 1800 MHz auction
• Divergence between value of 1800 MHz and
2100 MHz
– 1800 MHz 40-50 % in value
– More valuable in cat B and cat C circles
•
That is why benchmarking reserve price to
2100 MHz resulted in poor bidding in Metros
but OK results in cat B and cat C
Concluding Thoughts
• Economists and engineers both invaluable for
valuation
• Academics and practitioners need to provide
inputs
• Auctions cannot be discarded without
understanding reasons for failure
• New technological possibilities
Can NICF lead the way?
Thank you