Escolar Documentos
Profissional Documentos
Cultura Documentos
POLICY PROPOSALS FOR MAXIMISING TAX REVENUE AND PROMOTING TELE-DENSITY GROWTH - A REPORT PREPARED FOR THE GSMA
August 2006
Draft
Conclusion..........................................................................................33 Statistical modelling techniques ........................................................................ 37 Regression results................................................................................................ 39 The relationship between the price of new handsets and the proportion of handset sales in taking place in the formal market .............................. 45
Contents
rpt-RL-taxes&mobiles-Bangladesh-FINAL-V2-STC-270406.doc
ii
Draft
iii
Draft
iv
Draft
List of Tables
Table 1: Estimated market elasticities ........................................................................ 16 Table 2: Estimated mobile elasticities at the operator level .................................... 17 Table 3: A comparison of the estimated elasticities for Bangladesh with elasticities reported elsewhere ............................................................................ 19 Table 4: Outline of the tax simulation scenarios ...................................................... 22 Table 5: Tax simulation scenario 1 reducing VAT on mobile services from 15% to 10%.................................................................................................................... 24 Table 6: Tax simulation scenario 2 reduce import tax on handsets from Taka 300 to zero............................................................................................................. 26 Table 7: Sensitivity analysis of tax simulation scenario 2 ........................................ 27 Table 8: Tax simulation scenario 3 removal of SIM tax (Taka 900) holding all other taxes unchanged......................................................................................... 28 Table 9: Tax simulation scenario 3 sensitivity analysis around the removal of SIM tax................................................................................................................... 30 Table A10: Estimated mobile elasticities ................................................................... 39 Table A11: GrameenPhone regression results post-paid..................................... 40 Table A12: GrameenPhone regression results pre-paid....................................... 41 Table A13: Aktel regression results pre- and post-paid together ...................... 42 Table A14: Banglalink regression results pre- and post-paid together............... 43 Table A15: Estimated market elasticities full regression results ......................... 45
Draft
Executive summary
The purpose of imposing taxes on any economic transaction is so that the government can collect tax revenues, which are then used to finance the various functions of government. The analysis presented in this report shows that a more streamlined tax regime for mobile customers in Bangladesh, in which mobile specific taxes are removed, could ultimately result in higher tax revenues for the government. The existing tax regime for mobiles in Bangladesh, and in particular the mobilespecific taxes levied on customers, are estimated to constrain penetration growth and mobile usage in the mobile sector. This is because mobile customers in Bangladesh are sensitive to the prices that are charged for mobile services. If certain taxes levied on mobile users were removed, such as the SIM tax or the import tax on mobile handsets, then prices for consumers would be expected to fall. Consumers will respond to this fall in prices by consuming more both increasing mobile penetration growth and also overall mobile usage. With respect to the removal of mobile-specific taxes, there are three important factors to be taken into account. 1. Removal of the Taka 900 SIM card tax for new connections
| We find that the introduction of the Taka 900 SIM tax for new connections
in mid-2005 could be expected to have a significant impact on the development of the Bangladeshi mobile market. We estimate that by 2009 there could be approximately two million fewer subscribers compared with the counter factual of no SIM tax. This translates to a 9% difference in the estimated subscriber base by 2009. We estimate that if the SIM tax was removed, tax revenues could actually be 2.5% higher by 2009 and in certain cases up to 4.4% higher in 2009, compared with the situation with the SIM tax in-place. The net increase in tax revenues is driven by the higher tax revenue from higher growth rate in new connections and higher usage due to the associated fall in mobile connection costs. This is estimated to counter-balance the fall in tax revenues from the removal of the SIM tax. Figure 1 illustrates the potential estimated impact of removing the SIM tax on tax revenues and subscriber growth in 2009.
| We also find that the removal of the tax could actually boost tax revenues.
Executive summary
Draft
Removing SIM tax may boost subscriber numbers & tax revenues
22.4m subscribers in 2009, keep SIM tax - tax revenues of BDT 30.3 bn 12.5m subscribers in 2006 24.3m subscribers in 2009, removing tax - tax revenues of BDT 31.1 bn
35 Projected tax revenues from mobile (BDT billions) 30 25 20 15 10 5 0 Tax revenues (2006)
Taxes from usage: BDT 12bn (62%)
Figure 1: The estimated effect on tax revenues and subscriber numbers from removal of SIM tax
Source: Frontier Economics
imported to the country at a flat rate of Taka 300. We find that the removal of this tax could also increase subscriber numbers, compared with the counterfactual of the import tax in place. We estimate that the subscriber base could be up to 1.8% to 2.9% larger by 2009, approximately 400 thousand to 600 thousand additional mobile subscribers, were this fixed import tax to be removed. This is because one of the key requirements for becoming a mobile phone subscriber is to actually buy a mobile phone. Therefore, if the import tax is removed, then this pushes down the price of buying a mobile phone and therefore the overall cost of becoming a mobile subscriber. phones in either the grey or second hand markets. If the handset tax was removed, we project that not only will more handsets be sold, but a greater proportion of these will be sold through legitimate means. In the mediumterm, by 2009, we estimate that the fall in taxes following the removal of the fixed import tax could be almost offset by higher taxes from more subscribers (variable taxes and the SIM tax), higher overall usage and more handset sales in the legitimate market (which are taxed according to general taxation, i.e. VAT at 15%). Figure 2 illustrates the impact of removing the handset tax on tax revenues and subscriber growth in 2009.
| The relatively high import tax prompts many consumers to buy mobile
Executive summary
Draft
Removing handst tax may boost subscriber numbers & tax revenues in the longer-term
35 30
Projected tax revenues from mobile (BDT billions)
22.4m subscribers in 2009, keep handset tax - tax revenues of BDT 30.3 bn 12.5m subscribers in 2006
Taxes from usage: BDT 22bn (74%)
Figure 2: The effect on tax revenues and subscriber numbers from removal of the handset import tax
Source: Frontier Economics
3. Removal of the Taka 300 import tax on new mobile handsets and the Taka 900 SIM card tax Removing both mobile specific taxes could result in significantly lower connection and usage costs for mobile customers. Figure 3 shows the difference in the projected subscriber base under the scenario where the mobile specific taxes remain in place, and the scenario where both taxes are removed1. The figure shows that even by 2009 the subscriber base is higher under the scenario with no mobile-specific taxes, by almost 11%.
In the main report we explain that the tax simulation model created for the analysis of tax changes is not a forecasting model. Therefore, the projections presented in the figure should not be viewed as Frontier Economics forecast of subscriber growth going forward. The figure only serves to illustrate the potential drag effect that mobile-specific taxes can have on the growth of mobile penetration in Bangladesh.
Executive summary
Draft
12% 10% 8%
The impact of removing both mobile specific taxes on tax revenues is shown in Figure 4. The figure shows that, according to the simulation, in the initial period following the removal of both taxes overall tax revenue falls by as much as 18%. Revenues recover quite quickly after this point driven by the higher variable taxes on increased penetration growth and higher overall levels of usage such that by 2009, the net effect on tax revenues is estimated to be zero. From 2009 onwards, we see that tax revenues are estimated to be higher, following the removal of the mobile-specific taxes.
Executive summary
Draft
Percentage change in tax revenues from mobile following removal of mobile-specific taxes 15% Percentage change in tax revenues compared with the base case (i.e. mobile specific taxes in place) 10% 5% 0% -5% -10% -15% 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
However, in later years increased subscription and usage have a counterbalencing positive effect on tax revenues
The focus of the analysis in this report is on the effect of changes in the tax regime on subscription, usage and tax revenues in the mobile sector. It has also been recognised that changes in mobile penetration growth can potentially have strong positive knock-effects in the economy more generally. Box 1, below, summarises the key results from this literature.
Executive summary
Draft
BOX 1 - Impact of a streamlined tax policy for mobiles on the economy more widely This report focuses on the impact of changing the tax regime in the mobile sector. Research carried out by economists at the London Business School (LBS*), shows that telecommunications infrastructure, and the way in which this is spread across the population is a significant driver of economic growth. Empirical evidence presented in this research indicates that a 10% increase in the penetration rate of mobile phones could boost GDP per capita growth rates by around 0.59 per cent per year. Referring back to Figure 3, we project that the removal of the mobile-specific taxes will boost the average mobile penetration rate by more than 10% from 2008 onwards. The potential impact of this additional subscriber growth on annual GDP per capita growth (using the LBS results) would therefore expected to be significant. While it is true that increased penetration growth and a more streamlined tax regime could have significant benefits for the economy more widely, these issues are not addressed directly in this report.
(*)Fuss, M., M. Meschi and L. Waverman, The Impact of Telecoms on Economic Growth in Developing Countries, Vodafone Policy Paper Series, No. 3, March 2005.
Executive summary
Draft
1 Introduction
This report presents an economic assessment of the effect of the taxation regime for mobile telecoms in Bangladesh on the development of the mobile market going forward. The research has been commissioned by the GSM Association, and has been carried out with the cooperation of the three major mobile operators in Bangladesh: GrameenPhone, Aktel and Banglalink. Together, these three mobile operators account for approximately 95% of the mobile market in Bangladesh (Q3 2005). The overarching objective of the study is to establish the extent to which there is scope to optimise the current tax regime facing mobile phone customers in Bangladesh. There are two stages to the analysis:
| Estimate the sensitivity of the demand for mobile services to the price
charged for these services - the demand sensitivity measure we estimate is called the (price) elasticity of demand. estimates, we simulate the effect of a change in the taxes facing mobile consumers on the demand for mobile services. We present estimates of the effects on consumer demand (subscriber numbers and minutes of usage) as well as estimates of the effect on overall tax revenues from mobile.
In the analysis section below, we look at the demand sensitivity for three different aspects of the overall demand for mobile services:
| The elasticity of demand for mobile minutes i.e. how sensitive is the
average mobile users minutes of usage to the price per minute that they pay.
| The elasticity of demand for mobile handsets i.e. how responsive is the
average customers demand for mobile handsets to changes in the handset price.
| The overall elasticity of demand for mobile services i.e. what is the
relationship between mobile subscriber growth and the average cost of mobile services, i.e. the cost of new connections, subscription and usage. The elasticities themselves are estimated using the statistical technique of regression analysis. The price sensitivity equations use time series data on subscriber numbers, minutes of usage and revenues for each operator, allowing us to relate total mobile subscribers and mobile minutes of usage to the average price for each service over time. In simulating the effect of tax changes on the demand for mobile services, we consider two types of taxes that consumers face:
| General consumption taxes: This is VAT, which is levied at a rate of 15%.
VAT is levied on all usage charges, i.e. the price per mobile minute each customer pays. VAT is also levied on sales of handsets, monthly subscription charges and upfront mobile connection charges.
Introduction
Draft
| Mobile specific taxes: there are two mobile specific taxes levied on mobile
users in Bangladesh: a fixed handset-specific tax of Taka 300; and a tax on all new SIM cards of Taka 900.
market in Bangladesh in recent years. This section describes the evolution of mobile penetration over time, and shows how growth in mobile penetration in Bangladesh has tended to follow changes in GDP per capita over time. This section also outlines the existing tax regime for mobiles in Bangladesh.
| Section 3 presents results from the analysis. We begin with an outline of the
methodological framework and the data used in the analysis. We then present the estimated elasticities of demand. Finally we present the results from the tax simulation analysis.
| Section 4 presents our conclusions and recommendations.
Before proceeding to section 2, we note the following. The operators role has been to provide Frontier Economics (Frontier) with the necessary data to carry out the analysis. This includes data on minutes of usage, number of subscribers, tariff information and revenues. All of the analysis presented in this report, and the conclusions that are derived from it, are based on work carried out by Frontier Economics. The focus of this research is on the relationship between the taxes consumers pay and the likely development of the mobile market in Bangladesh going forward. We also recognise however, that there are other taxes levied in Bangladesh which could potentially impact on the prices customers pay and hence the demand for mobile services. For example, import taxes on equipment used for building mobile infrastructure, are also likely to affect the prices customers pay. However, for the purposes of the current research, we concentrate on the role of consumer taxes in the development of the Bangladesh mobile market. The GSMA may wish to carry out research at a later stage into the effect of other taxes on the development of the Bangladeshi mobile market.
Introduction
Draft
2.1
In 1996, the Bangladesh government issued three mobile licences to compete with the incumbent D-AMPS (a forerunner to GSM technology) operator Pacific Bangladesh Telecom Limited (PBTL, which has since been re-branded as CityCell). At the end of 2005, there were approximately 8.1 million mobile subscribers in Bangladesh, representing a mobile penetration rate of around 6%. By February 2006, the number of subscribers was reported to have grown to 10.1 million, a penetration rate of 7.1%2. Figure 5 shows the growth in subscriber numbers over the 2001 to 2005 period. In September 2005, the market share of each of the four Bangladeshi operators, in terms of number of subscribers, was as follows: GrameenPhone, 59%; Aktel, 29%; Banglalink, 8.4% and CityCell 4.4%3. Figure 6 shows that the vast majority of mobile phone subscribers in Bangladesh around 93% - are on pre-paid contracts. This is similar to the situation in most other South Asian countries, where pre-paid customers are the dominant type of subscriber.
A 20 February 2006 press release by Telegeography (http://www.telegeography.com/cu/article.php?article_id=11281), stated that the total number of mobile subscribers in Bangladesh had risen to 10.1 million. This data is provisional. Sourced from the GlobalComms database at http://www.telegeography.com.
Draft
20 02
20 01
20 01
20 01
20 01
20 02
20 03
20 03
20 04
20 04
20 05
Q 1
Q 3
Q 3
Q 1
100% Percentage of mobile subscribers that are pre-paid customers 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%
Q 2
20 Q 01 3 20 Q 01 4 20 Q 01 1 20 Q 02 2 20 Q 02 3 20 Q 02 4 20 Q 02 1 20 Q 03 2 20 Q 03 3 20 Q 03 4 20 Q 03 1 20 Q 04 2 20 Q 04 3 20 Q 04 4 20 Q 04 1 20 Q 05 2 20 Q 05 3 20 Q 05 4 20 05
Q 1
Q 3
Q 1
Q 3
Q 1
Q 3
Q 1
Q 3
Q 1
20 05
20 06
Draft
Due to the presence of the GrameenPhone Village Phone programme where a single mobile phone can have several different users - the headline mobile penetration rate for Bangladesh may underestimate the true number of mobile users. In 2004 GrameenPhone stated that of its 4.2 million subscribers, around 110 thousand (or 2.6%) were part of the village phone programme. Since then, as GrameenPhones subscriber base has grown to five million, it is estimated that around 250 thousand of these subscribers are part of the village phone programme. This can have important implications for the analysis that follows, as it would imply that all other factors the same, average tax revenues from usage of mobile services would be expected to be relatively high in Bangladesh.
See for example, the 2005 report by the GSMA and Frontier (among others) Tax and the Digital Divide. The paper by R. Lydon and M. Williams on Communications Networks and Foreign Direct Investment in Developing Countries (published in: Communications Strategies: Vol. 58 2nd Quarter 2005, pages 43 61) also contains a detailed discussion on the determinants of growth in mobile penetration.
Draft
25%
20%
15%
10%
Pakistan
5%
India Bangladesh
0% 0 100 200 300 400 500 600 700 800 900 1,000 GDP per capita ($US, 2004)
Figure 7: The relationship between GDP per capita and mobile penetration amongst Lower Income countries, 2005
Source: Frontier Economics. Penetration data is for September 2005 and is taken from the GlobalComms data base (http://www.telegeography.com/). Data on GDP per capita is from the World Bank World Development Indicators (WDI, 2006) database.
Figure 8 shows that the recent growth in mobile penetration in Bangladesh closely follows growth in GDP per capita. Whilst they are both on different scales, the figure shows that changes in the growth of GDP per capita are closely correlated with changes in mobile penetration. Figure 9 shows that between 2001 and 2004, changes in mobile coverage were also positively correlated with changes in mobile penetration.
Draft
160% Annual % growth in number of subscribers 140% 120% 100% Annual percentage change in GDP per capita (real, right scale)
7.0% 6.0% 5.0% 4.0% Annual % growth in GDP per capita (real)
80% 3.0% 60% 40% 20% 0% 2001 2.0% Annual % growth in number of subscribers (left scale) 1.0% 0.0% 2005
2002
2003
2004
Figure 8: The relationship between growth in real GDP per capita and growth in penetration
Source: Frontier Economics. Data on growth in subscriber numbers is from the operators. Data on annual growth (%) in real GDP per capita is taken from the World Bank and the Bangladesh Ministry of Finance
90% Network coverage (% of the population) 80% 70% 60% 50% Network coverage % of the population
6%
4%
3% 40% 30% 20% 10% 0% 2001 0% 2005 Mobile penetration (% of the population) 2%
1%
2002
2003
2004
Figure 9: The relationship between growth in mobile coverage (% of the population) and mobile penetration
Source: Mobile penetration data provided by the operators
Draft
Finally in this section, we compare the growth in the level of mobile penetration in Bangladesh with that in other South Asian countries, as well as the group of Low Income countries5. The figure suggests that growth mobile penetration in Bangladesh has tended to lag that in other South Asian and low income countries. As shown in Figure 5, however, mobile penetration is estimated to have grown to 7.1% in the first quarter of 2006, which would move Bangladesh just above the low income group of countries.
18% 16% Mobile penetration over time 14% 12% Pakistan, 10.9% 10% 8% 6% 4% 2% 0% 1997 Low-income country group, 6.8% Bangladesh 6.1% (2005) India, 5.8% Sri Lanka, 15.5%
1998
1999
2000
2001
2002
2003
2004
2005
Figure 10: Changes in mobile penetration over time comparing Bangladesh with other South Asian countries
Source: Frontier Economics. Data on mobile penetration for the 1997 2002 period is taken from the ITU database. Data for the period 2003 2005 is from the Telegeography GlobalComms data base (http://www.telegeography.com/), and is for December of each year.
Low-income economies are those in which 2004 GNI per capita is $825 or less. (World Bank World Development Indicators, 2006).
Draft
2.2
This section begins with an outline of the tax regime facing mobile subscribers in Bangladesh. We follow this with a comparison of the taxes levied on mobile services in Bangladesh with the taxes levied on mobile services in other South Asian countries.
The information on the import taxation regime for mobile handsets has been provided by the three Bangladeshi mobile operators who have supplied data as part of this study.
Draft
35 Tax burden in cost of mobile services (%) 30 25 20 15 10 5 0 Bangladesh Pakistan Sri Lanka India
Based on discussions with the operators, we estimate the average price of new handsets, inclusive of all taxes, to be in the range of Taka 4,000 to Taka 5,000 (equivalent to US$57 US$72). Therefore in the tax simulation modelling, we assume a starting value for the average handset price inclusive of all taxes of Taka 4,000. We also test the sensitivity of the simulation results to this assumption.
Mobile service costs includes all costs a customer incurs on an annual basis, with the exception of handsets costs. That is, the measure takes account of average usage patterns, average cost per minute and additional costs associated with subscription and connection (spread over a three year period).
Draft
3.1
METHODOLOGICAL FRAMEWORK
The aim of this study is to analyse the implications of a change in taxes on the demand for mobile services in Bangladesh. The methodological framework comprises three steps:
| Step 1: for a given change in taxes, we assume that the tax change is passed
on to consumers. For example, if the Taka 900 SIM tax were removed, we assume that the cost of becoming a mobile subscriber (new connections) also falls by Taka 900. Therefore, if the average cost of connection in Bangladesh is Taka 2,500, then removing the SIM tax reduces the connection cost by 36% (900/2,500). effect on the demand for mobile services. We model the demand sensitivities across three areas: the change in the take-up rate, or penetration/subscription rate, as a result of the tax change;
| Step 2: for a given change in prices due to a change in taxes, estimate the
the change in average minutes of usage per customer as a result of the tax change; the change in the demand for handsets as a result of the tax change.
| Step 3: Having estimated the change in demand due to a change in prices,
we then estimate the change in total tax revenues as a result - that is, taking account of the fact that taxes have changed, but mobile demand has also changed.
There are two main channels through which taxes can affect the mobile market and tax revenues: by affecting overall usage; or, by affecting handset demand. For example, a reduction in the tax rate, through either a reduction in VAT or mobile specific taxes, could increase subscription growth, handset demand and usage This would lead to higher growth of the mobile market than would otherwise be the case, and hence a (second round) increase in tax revenues. The methodological framework requires three elasticities. The next section describes how each of these elasticities is estimated, including a brief description of the data that has been provided by each of the operators to estimate the elasticities.
Draft
3.2
We first describe the data and statistical techniques we have used in the analysis, and then summarise the results from the elasticity analysis, which is covered in more detail in the Annexe.
S t = A + bPt + c Pt + D X t
s m
s
[1],
m
minute of usage and X t is a vector of other factors that affect subscription, such as coverage, GDP per capita and other demographic factors. The relationship described by equation [1] is estimated using the statistical technique of regression analysis. The results from the regression analysis provide us with estimates of the own- and cross-price elasticities, that is parameters b and c in equation [1]. Annexe 1 provides a more detailed description of the statistical estimation techniques. The Annexe also lists the full regression results with details of the variables used and the number of observations. As well as estimating the model given by equation [1], we estimated the following relationship for the average mobile minutes of usage per subscriber at time t, Mt :
m s M t = + Pt + Pt + X t
[2],
where, as in equation [1], and are the own- and cross-price elasticities respectively. Ideally, we would also want to estimate two further equations: one that describes the relationship between mobile subscription and the price of (new) handsets; and another that describes the relationship between new handset sales and the price of new handsets. However, in the absences of a long time series of data on
Draft
handset sales and handset prices we have been unable to estimate these two further relationships. We discuss the implications of this for the analysis below.
Elasticity of demand for new connections and elasticity of demand for usage at the market level
The tax simulation model, which considers tax revenues for the entire market, requires market elasticities. Given that GrameenPhone has a significant market share, then we might expect GrameenPhones elasticities to more closely reflect the market elasticities. However, we have also attempted to come up with our own estimates of the market elasticities. In order to do this, we estimate two regressions at the market level, which are variants on the regressions shown in equations [1] and [2]:
Quarterly change in subscribers=A +b * (quarterly change in total revenue per subscriber) [3]
Quarterly change average minutes per user=C +d * (quarterly change in average revenue per minute) [4]
All variables are measured as natural logarithms; therefore the coefficients b and d can be interpreted as elasticities. We estimate each regression for the market as a whole - that is, combining revenue, subscriber and usage data for all operators (but one). We are unable to separate the revenue data for each operator into revenues from connection and usage. Hence we are unable to calculate separate own-price and cross-price market elasticities8. The estimated elasticities are reported in Table 1.
Technically, the coefficients b and d are constrained estimates of the subscription and usage price elasticities given in equations [1] and [2]. That is, constraining the coefficient on the subscription price and the per minute usage price to be equal.
Draft
Dependent variable Quarterly change in subscribers Quarterly change in average minutes per user Table 1: Estimated market elasticities
Source: Frontier Economics
Independent variable and estimated elasticity Quarterly change in total revenue per subscriber (*), -0.30 Quarterly change in average revenue per minute(**), -0.66
Notes: (*) Quarterly change in total revenue per subscriber is defined as total revenues from mobile operations for all operators divided by the total number of subscribers. (**) Quarterly change in average revenue per minute is defined as total revenues from mobile operations for all operators divided by total minutes of usage across all operators.
for a 10% decrease in the average cost (or average mobile revenue) per subscriber, penetration growth increases by 3%; for a 10% decrease in the average cost (or average mobile revenue) per subscriber, average minutes of usage per subscriber increases by 6.6%
Ideally, we would require cross-elasticity estimates also, that is estimates of all of the parameters in equations [1] and [2]. However, it is not possible to do this at the market level, as we do not have market based estimates of revenues from subscription/connection and revenues from usage. Some of the operators have, however, provided us with this data, and it is therefore feasible to estimate the cross-elasticities at the operator level. The next section presents these results
Elasticity of demand for connection and elasticity of demand for usage at the operator level
We also estimated demand elasticities for each of the mobile operators, subject to availability of the required data. The connection and usage elasticities at the operator level are shown in Table 2. The elasticities in Table 2 are not market elasticities, as they only show the price sensitivities at the operator level. These elasticities are likely to overstate the market elasticity as the price sensitivities implicitly include switching between operators 9.
The technical Annexe reports the regression diagnostics for each of the operator elasticity estimates. Note that for reasons of data confidentiality, we do not report the regression coefficients for each operator separately. The tables in the annexe do, however, show the sample size used in the estimation, as well as the usual regression diagnostics, such as the goodness-of-fit and the statistical significance (t-statistics) of the estimated parameters.
Draft
Prices
-0.46 -0.22
-0.18 -0.73
A fall in the cost of connection has two effects. First, there is the direct effect on demand (own-price elasticity), that is, a 10% fall in the connection price leads to a 4.6% increase in the number of new subscribers/connections above the base case. Second, there is the indirect effect on demand (cross-price elasticity), that is, a 10% fall in the cost leads to a 2.2% increase in the average MOU per subscriber10. For mobile usage costs, there are also direct and indirect effects. The direct effect implies that a 10% fall in the average cost leads to a 7.3% increase in average MOU per user. The indirect effect implies that a 10% fall in the average cost per MOU also leads to a 1.8% increase in the average number of new connections above the base case.
Although the elasticities in Table 2 are not directly comparable with those reported in Table 1 because the variable definitions are different we would expect the elasticity of demand with respect to the price of mobile usage (-0.18 for subscription and -0.73 for MOU) to be fairly similar. This is because the vast majority of total revenues are revenues from usage at least for GrameenPhone and Banglalink, the operators who have provided the revenue data at such a disaggregated level. We find that this is indeed the case, with the estimate elasticity on average revenue per minute of usage coming out at -0.66. For the tax simulation model we propose to use the market based elasticity of demand of -0.66 for MOU. For the elasticity of subscription with respect to the price of usage, we propose to use average operator elasticity reported in Table 2, that is -0.18; and similarly for the elasticity of usage with respect to the price of connection/subscription (-0.22). These two elasticities, while based on the
10
This could be due to the externality that affects existing subscribers when new subscribers join the mobile network. That is, as more subscribers join the network, there are more individuals to call and hence minutes of usage of subscribers increases as a result.
Draft
operator elasticities, are the only cross-elasticities we have been able to estimate with the data provided. For the elasticity of demand for new connections with respect to the price of connection we have the estimate of -0.46 from the operator elasticities. However, for the reasons we have outlined above, this is likely to be biased upwards (in absolute value), i.e. because it includes switching between operators. The other estimate we have is -0.30 from the market-based elasticities. However, this is actually a weighted average of the market elasticity of demand for new connections with respect to the price of connection and with respect to the price of usage. In the absence of any information as to which elasticity is likely to be more or less biased (or more or less robust), we propose to use the mid-point of the two estimates in the tax simulation model, -0.3811.
The GSMA/Frontier report Tax and the Digital Divide (GSMA, 2005); and The review carried out on behalf of the New Zealand Commerce Commission into the price elasticity of demand for mobile telecommunications services, (NZCC, 2003)12.
Of the two comparator studies, the previous work carried out by Frontier for the GSMA study Tax and the Digital Divide provides probably the most relevant comparator as the sample of 50 developing countries is more similar (in terms of economic development) to Bangladesh. The estimated elasticities are similar in value to those for Bangladesh, with the elasticity of demand for usage with respect to the price of usage (U-U) being almost identical, -0.73 compared with -0.76 (at the top end of the range).
11
Note, in the tax simulation modelling, we also test the sensitivity of the model-projections to assumptions regarding the elasticity of the demand for new connections with respect to the overall cost of connection. Review of the price elasticity of demand for fixed line and mobile telecommunications services, New Zealand Commerce Commission. See also New Zealand Commerce Commission, Schedule 3 investigation into regulation of mobile termination, October 2004.
12
Draft
Estimated elasticities across the Bangladeshi operators, weighted average by market share Date of study =>
Quantity Price
2006
2005
2003
-0.50
-0.06 to -0.54
-0.09 to -0.80
Table 3: A comparison of the estimated elasticities for Bangladesh with elasticities reported elsewhere
Source: Frontier Economics (*) The Tax and the Digital Divide study did not include estimates of the elasticity of demand for usage (U) with respect to the price of subscription (S)
Draft
handsets and the number of new handset sales, this time series is not sufficiently long enough for us to back out a subscription-handset price elasticity. In the tax simulation model, therefore, we use the estimate of the subscription elasticity of demand with respect to the price of subscription as our estimate of the elasticity, i.e. -0.38.
The relationship between the price of new handsets and the number of handsets sold in the formal market
The final piece of information we need for the tax simulation model is an estimate of the relationship between the proportion of handset sales taking places in the formal (or legitimate) market (i.e. handsets that are not sold in either the informal market or the second-hand market). Data on relative prices and the sales of handsets in the formal and informal market in Bangladesh is difficult to come by. We therefore rely on evidence from the GSMA report Tax and the Digital Divide, which used cross-section data for 25 developing countries to analyse the relationship between the market share of grey (i.e. informal market and second hand handset sales) and the average retail price of legitimate handsets (as a percentage of GDP per capita in each country). The analysis in the Tax and the Digital Divide study, which is describe in more detail in the Annexe, finds that for 10% decrease in the price of new legitimate handsets, holding the price of handsets in the informal market constant, the share of new handset sales in total handset sales increases by 4.7%. To conclude this section, there are two channels through which an X% reduction in the handset price as a direct result of a change in the handset tax can affect tax revenues:
a first-round reduction in tax revenues as a result of the reduction in the tax; an increase in tax revenues from a second round increase the growth in mobile subscribers above the base level (by X%*elasticity of demand of subscription). In the tax simulation model, we use an estimate of the elasticity of demand of -0.38; and an increase in tax revenues from handset sales, because of an increase in the proportion of handsets sold through legitimate marketing channels relative to the base case (by X%*0.47)
Draft
3.3
The previous section outlined a number of different channels through which a reduction in taxes, and the subsequent reduction in each of the component prices of mobile services, can impact on tax revenues:
The initial first round effect is to reduce total tax revenues. For example, if the Bangladeshi government were to remove the Taka 900 SIM tax we estimate that tax revenues for mobile could fall by up to 20%, as illustrated in Figure 12, for the 2006 tax year. However, there are counter-balancing effects which can also act to increase tax revenue in the short- to medium term. This is because a reduction in taxes ultimately reduces the prices that customers have to pay. If prices fall, then demand for mobile services could also rise leading to higher penetration, usage and handset sales. The aim of this section is to try and estimate the size of these potentially counterbalancing effects.
Figure 12 shows the projected consumer tax revenues from mobile for 2006. Note that the projected revenues, in the order of Taka 20 billion, depend very much on the modelling assumptions we have made in the tax simulation model, in particular around:
subscriber growth in 2006; average usage patterns across subscribers in 2006; the number of new handsets bought in 2006 on which taxes are paid; the number of new connections on which taxes, including the SIM tax, are paid (in the base case, this is always assumed to be 100% of new connections).
Draft
(2) Removal of SIM card tax (3) Removal of handset import tax (4) Removal of both the SIM card tax and the handset tax
0 900
300
Annual growth in total mobile subscriber numbers: The base case for growth in subscriber numbers is taken from the GSMAs own forecasts for market growth presented as part of the 2005 study Tax and the Digital Divide. The base case for growth in subscriber numbers over the 2007 2009 period is an average annual growth rate of 22%. Pre-paid versus post-paid subscribers: Not all mobile charges apply to pre-pay and post pay subscribers. We therefore need to make an assumption about the pre-paid/post-paid market split in the base case. We use the pre-paid subscriber market share at the end of 2005, equal to 93%, as our estimate going forward. This split remains constant under each of the three scenarios.
Draft
Usage: Data on MOU provided by the three operators shows that the
average minutes of usage (average MOU) per subscriber has changed little in recent years. We therefore use the average annual MOU in 2005 for all subscribers with each of the three operators as our base case MOU assumption. This works out at 2,189 minutes per year in 2005. we use a value of Taka 4,000 ($59) as our estimate of the average price of handsets. Therefore, if the Taka 300 handset import tax were to be removed, the average price of a handset would fall by 7.5% (i.e. Taka 300/4,000)
Handset sales: Based on data on handset sales provided by one of the operators for 2005, we estimate the total number of new handset sales at around 1.85 million in 2005. Given that the market grew by 4.63 million subscribers in 2005, this implies approximately 40% (1.85/4.63) of all handset sales (assuming all new subscribers also acquire a new handset) were first-hand, or legitimate handset sales. Other estimates of the number of legitimate handset sales range from 60%14 to 30%15. In the absence of more robust information on the number of new mobile subscribers that actually purchase a new mobile phone through the legitimate market - and therefore pay taxes we propose to use an estimate of 50% in the base case. If the market share of legitimate handset sales in total handset sales were to rise about 50% (assuming total handset sales are constant or increasing), then tax revenues from handset sales would also increase. Handset upgrades for existing customers: The typical assumption in
modelling the cost of mobile usage is to assume a mobile phone lifespan of 3 years, i.e. a third of all mobile users upgrade their mobiles every year. For Bangladesh, we have been slightly more conservative and assumed an upgrade rate of 4 years. We also assume that of those existing subscribers who upgrade their mobile phones, only 50% do so in the legitimate market.
the SIM card tax from Taka 900 to zero. In order to simulate the impact on subscription and usage, we need to translate this change to a proportionate change in the connection price, and then multiply this proportionate price change by the relevant elasticity. Our base-case estimate of the costs associated with connecting for the first time is Taka 2,500. This figure is derived from two sources. First the Bangladesh government reportedly valued the cost of a new mobile connection in 2005, including all the relevant taxes, at Taka 2,17216. Second, Frontiers
14 15 16
Draft
own discussions with mobile operators indicate an average cost of connection of around Taka 3,000
Total consumer tax revenues from mobile 2009 (Taka, millions) 30,279 21,721 -28.3%
Table 5: Tax simulation scenario 1 reducing VAT on mobile services from 15% to 10%
Source: Frontier Economics
Draft
increased VAT on connection and subscription from subscriber boost relative to the base case and also increased revenues from the SIM tax; increased VAT on sales of additional new phones relative to the base case.
We project that tax revenues would be approximately 2% lower in 2009, compared with the base case. However, tax revenues will slowly recover, as sales of handsets, growth in subscriber numbers, and increases in usage all increase steadily over time, compared with the base case.
17
The implicit assumption we are making is that a reduction in the handset price is equivalent to a decrease in the overall cost of becoming a mobile subscriber, and that the rate of amortisation of both of these connection costs is the roughly the same.
Draft
Total consumer tax revenues from mobile 2009 (Taka, millions) 30,279 29,676 -2.0%
Table 6: Tax simulation scenario 2 reduce import tax on handsets from Taka 300 to zero
Source: Frontier Economics
The handset tax simulation requires information on the average price of handsets. As we have argued above, this is one of the areas of analysis where the empirical evidence for Bangladesh is limited, and we have therefore relied on a number of assumptions to obtain the results in the simulation. It is therefore worth checking the sensitivity of the results to certain key assumptions. Specifically, we alter the parameters of the simulation in two ways:
increase the elasticity of subscription with respect to the handset price from -0.38 to -0.60. Recall from section 3.2 that the data was not amenable to estimating a subscription-handset price elasticity, and we therefore relied on the subscription own-price elasticity of demand; increase the number of handsets sold in the grey market in the base case from 50% to 70%. Clearly, if fewer handsets are sold in the legitimate market, then taxes from such sales will account for a smaller proportion of total tax revenues therefore, removing these taxes will have a smaller impact on the total tax take.
Table 7 shows the results of the sensitivity analysis. The results show that the simulated tax revenues in later years are sensitive to assumptions about the key parameters. Under both sensitivity scenarios, the fall in tax revenues following the removal of the handset import tax (the first round effect) is almost completely offset by the increase in tax revenues from higher subscription growth, usage and legitimate handset sales.
Draft
Sensitivity 1: assume subscription-handset price elasticity of -0.60 Base case Tax simulation, % difference from base case 15.1 15.4 2.1% 22.4 23.0 2.9% 21,140 20,864 -1.3% 30,279 30,235 -0.1%
Sensitivity 2: assume subscription-handset price elasticity of -0.38, and share of legitimate handset sales in total handset sales of 30% in the base case Base case Tax simulation, % difference from base case 15.1 15.3 1.3% 22.4 22.8 1.8% 20,102 19,923 -0.9% 28,882 28,806 -0.3%
Draft
3.3.5 Scenario 3: removal of Taka 900 SIM tax for new connections
The removal of the Taka 900 SIM tax, has the effect of reducing the cost of becoming a mobile subscriber by 36%. That is Taka 900/Taka 2,500 (the estimated total cost of connecting to a mobile network for the first time). The subsequent effect on subscription and tax revenues is large and significant (Table 8):
growth in new subscribers increases by almost 9% relative to the base case scenario by 2009, equivalent to almost 2 million more subscribers relative to the base case (i.e. 24.3 million minus 22.4 million). in accordance with the elasticities we have estimated above, these new subscribers also have higher MOU relative to the base case, because the overall cost of mobile usage has fallen.
The additional subscribers and the higher minutes of usage - both relative to the base case result in higher tax revenues for the government. By 2009, tax revenues are 2.5% higher, a difference which continues to grow significantly over time.
Total consumer tax revenues from mobile 2009 (Taka, millions) 30,279 31,032 2.5%
Table 8: Tax simulation scenario 3 removal of SIM tax (Taka 900) holding all other taxes unchanged
Source: Frontier Economics
As with all of the tax simulation scenarios, the output of the simulation is dependent on certain key assumptions. In the case of the SIM tax simulation, the key assumption is what proportion of current connection costs are currently accounted for by the SIM tax. As we outlined in setting out the base case, precise estimates of the average cost of connection are not available for Bangladesh. However, based on discussions we have had with the operators, combined with the available evidence on average connection costs, we proposed a range of Taka 2,172 Taka 3,000 for the average connection cost.
Draft
The simulation shown in Table 8 assumes a connection cost (inclusive of the SIM tax) of Taka 2,500, approximately the mid-point of the proposed range. However, given the level of uncertainty around the current value of the average connection charge, it is useful to illustrate the sensitivity of the simulation results to assumptions around the value of the connection charge. Table 9 compares the base case tax revenues and subscriber growth projections with two tax simulations, each of which makes different assumptions about the value of connection charges. The first sensitivity test assumes the connection charge is Taka 2,172, the lower end of the proposed range. In this case, the removal of the SIM tax reduces the overall connection charge by 41% (i.e. Taka 900/2,172). The second sensitivity test, sets the connection charge to Taka 3,000, meaning that the removal of the SIM tax lowers the overall costs of connection by 30% (i.e. Taka 900/3,000). The results from the sensitivity analysis show that the effect of the removal of the SIM tax on subscriber growth is quite robust to assumptions around the value of the average cost of connection. For example, in the scenario where the SIM tax is removed and the average cost of connection is Taka 2,172, we project that by 2009 the subscriber base could be 10% higher, compared with the base case. The same figure for the scenario where the connection cost is Taka 3,000 is 7.2%. The change in tax revenues from the removal of the SIM tax is slightly more sensitive to assumptions about the average cost of connection however it should be noted that this delays the counterbalancing tax revenue gains, rather than remove them completely. For example, in 2009 tax revenues are 4.4% higher than the base case, assuming an average connection cost of Taka 2,172, and 0.5% higher, assuming an average connection cost of Taka 3,000. In the latter case by 2011, tax revenues are 3.0% higher.
Draft
Total consumer tax revenues from mobile 2009 (Taka, millions) 30,279
Base case Tax simulation (connection cost of Taka 2,172) % difference Tax simulation (connection cost of Taka 3,000) % difference
15.1
22.4
21,140
16.2 7.3%
24.6 10.0%
20,748 -1.9%
31,611 4.4%
15.9 5.3%
24.0 7.2%
20,148 -4.7%
30,437 0.5%
Table 9: Tax simulation scenario 3 sensitivity analysis around the removal of SIM tax
Source: Frontier Economics
3.3.6 Scenario 4: removal of both the Taka 900 SIM tax and the Taka 300 handset
Removing both mobile specific taxes could result in significantly lower connection a costs for mobile customers. Figure 3 shows the difference in the projected subscriber base under the scenario where the mobile specific taxes remain in place, and the scenario where both taxes are removed18. The figure shows that by 2009 the subscriber base is estimated to be almost higher under the scenario with no mobile-specific taxes.
18
In the main report we explain that the tax simulation model created for the analysis of tax changes is not a forecasting model. Therefore, the projections presented in the figure should not be viewed as Frontier Economics forecast of subscriber growth going forward. The figure only serves to illustrate the potential drag effect that mobile-specific taxes can have on the growth of mobile penetration in Bangladesh.
Draft
12% 10% 8%
Figure 13: Comparative growth in subscriber base following removal of mobile-specific taxes
Source: Frontier Economics
The impact of removing both mobile specific taxes on tax revenues is shown in Figure 4. The figure shows that, according to the simulation, in the initial period following the removal of both taxes overall tax revenue falls by as much as 18%. Revenues recover quite quickly after this point driven by the higher variable taxes on increased penetration growth and higher overall levels of usage such that by 2009, the net effect on tax revenues is estimated to be zero. From 2009 onwards, we see that tax revenues are estimated to be higher, following the removal of the mobile-specific taxes. It is important to point out that in simulating the impact of removing both of the mobile-specific taxes considered here, we have assumed that the average cost of a new handset is Taka 4,000, and the average cost of a new connection is Taka 2,500. These are the central case scenarios set out scenarios 2 and 3 above.
Draft
Percentage change in tax revenues from mobile following removal of mobile-specific taxes 15% Percentage change in tax revenues compared with the base case (i.e. mobile specific taxes in place) 10% 5% 0% -5% -10% -15% 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
However, in later years increased subscription and usage have a counterbalencing positive effect on tax revenues
Figure 14: Percentage change in tax revenues following removal of mobile-specific taxes
Source: Frontier Economics
Draft
4 Conclusion
In concluding the research presented in this report it is important to re-iterate once more that the projections presented in this report are the output from a tax simulation model, and are subject to the assumptions that have gone into the construction of that model. The projections presented here do not represent economic forecasts, and should not be viewed as such. The counter-balancing tax revenue effects that we obtain under the tax simulation scenarios that remove the mobile-specific taxes are based on plausible assumptions that we have backed up with the available empirical evidence, both from the operators and on the market more generally. If these assumptions were to be changed, as we have illustrated in the report, then this could impact significantly on the output from the tax simulations. The main findings are as follows: Taka 900 SIM card tax for new connections
| We find that the introduction of the Taka 900 SIM tax for new connections
in mid-2005 could be expected to have a significant impact on the development of the Bangladeshi mobile market. We estimate that by 2009 there could be approximately two million fewer subscribers compared with the counter factual of no SIM tax. This translates to a 9% difference in the estimated subscriber base by 2009. We estimate that if the SIM tax was removed, tax revenues would actually be 2.5% higher by 2009, and in certain cases up to 4.4% higher in 2009, compared with the situation with the SIM tax in-place. The net increase in tax revenues is driven by the higher tax revenue from higher growth rate in new connections and higher usage patterns due to the fall in mobile connection costs associated. This counter-balances the fall in tax revenues from the removal of the SIM tax. Figure 15 illustrates the impact of removing the SIM tax on tax revenues and subscriber growth in 2009.
| We also find that the removal of the tax could actually boost tax revenues.
Conclusion
Draft
Removing SIM tax may boost subscriber numbers & tax revenues
22.4m subscribers in 2009, keep SIM tax - tax revenues of BDT 30.3 bn 12.5m subscribers in 2006 24.3m subscribers in 2009, removing tax - tax revenues of BDT 31.1 bn
35 Projected tax revenues from mobile (BDT billions) 30 25 20 15 10 5 0 Tax revenues (2006)
Taxes from usage: BDT 12bn (62%)
Figure 15: The estimated effect on tax revenues and subscriber numbers from removal of SIM tax
Source: Frontier Economics
imported to the country at a flat rate of Taka 300. We find that the removal of this tax could also significantly boost subscriber numbers, compared with the counterfactual of the import tax in place. We estimate that the subscriber base could be up to 1.8% to 2.9% larger by 2009, approximately 400 thousand to 600 thousand additional mobile subscribers, were this fixed import tax to be removed. This is because one of the key requirements for becoming a mobile phone subscriber is to actually buy a mobile phone. Therefore, if the import tax is removed, then this pushes down the price of buying a mobile phone and therefore the overall cost of becoming a mobile subscriber.
| The relatively high import tax prompts many consumers to buy mobile
phones in either the grey or second hand markets. If the handset tax was removed, we project that not only will more handsets be sold, but a greater proportion of these will be sold through legitimate means. In the mediumterm, by 2009, we estimate that the fall in taxes following the removal of the fixed import tax is could be almost completely offset by higher taxes from more subscribers (variable taxes and the SIM tax), higher overall usage and more handset sales in the legitimate market (which are taxed according to general taxation, i.e. VAT at 15%). Figure 16 illustrates the impact of removing the handset tax on tax revenues and subscriber growth in 2009.
Conclusion
Draft
Removing handst tax may boost subscriber numbers & tax revenues in the longer-term
35 30
Projected tax revenues from mobile (BDT billions)
22.4m subscribers in 2009, keep handset tax - tax revenues of BDT 30.3 bn 12.5m subscribers in 2006
Taxes from usage: BDT 22bn (74%)
Figure 16: The estimated effect on tax revenues and subscriber numbers from removal of the handset import tax
Source: Frontier Economics
Removal of SIM tax and handset import tax We have also considered the potential impact on subscription growth and tax revenues from the removal of both mobile specific taxes considered in this report the SIM tax and the handset import tax. The main findings from the tax simulation are as follows:
removing the two major mobile specific taxes facing consumers may provide a strong boost for growth in the mobile subscriber base in Bangladesh. We estimate that with the space of approximately two- to three-years, the number of mobile subscribers in Bangladesh could be over 11% higher without the mobile-specific taxes, compared with the counterfactual of those taxes remaining in place. as we might expect, the immediate effect of removing the mobile specific taxes is to reduce the tax take from mobile by approximately 18%. However, the counter-balancing effect from added subscriber growth, and the associated overall higher usage, quickly acts to offset this decrease. From 2010 onwards, tax revenues are projected to be above the level they would be with the taxes in place.
Conclusion
Draft
Subscriber numbers quarterly/monthly growth in subscriber numbers back to 2000, where applicable. Minutes of usage data quarterly/monthly MOU data, and in some cases split by pre-pay and post-pay; Revenue data quarterly revenue data, split by pre-/post-pay and revenues from connection/subscription and usage (for some operators).
In the statistical model, we estimate the following relationship for the number of mobile subscribers at time t, S t :
St = A + bPr + cPt + DX t
s m s
[A1],
m
minute of usage and X t is a matrix of other factors that affect subscription, such as coverage, GDP per capita and other demographic factors. If the regression is specified in logs, then the parameters b and c can be interpreted as elasticity estimates. We also estimate, via regression analysis, the relationship between (the log of) mobile minutes of usage and the (log of the) price of mobile services:
M t = + Pr + Pt + X t
m s
[A2],
where, as in equation [A1], and are the own- and cross-price elasticities respectively. Both the regressions given by [A1] and [A2] are time series regressions - in other words, we have repeated observations of the same variables over time. A common statistical issue in time series regression is non-stationarity, which effectively means that there is an underlying long-run time trend in the dependent variable which means it is changing over time. Changes in the long run underlying time trend common to both the dependent ( S t and M t ) and independent variables ( Pt , Pt and M t ) may be common, which might induce a spurious positive correlation if we were to estimate the regressions as given by equations [A1] and [A2]. The solution to the problem is to transform the nonstationary variable to stationary variables by estimating the regression in first differences. That is, we analyse the relationship between the change in the dependent and the independent variables. The parameters ( b , c , and ) have the same interpretation as in the un-differenced equation, i.e. they are the estimated elasticity of demand.
s m
Draft
Finally, it is important to point out that in using quarterly, or in the case of one of the mobile operators, monthly, data for the regression analysis, it is difficult to obtain monthly data on quarterly changes in GDP per capita over time. Furthermore, other exogenous factors, such as demographic changes or coverage are unlikely to change on a quarterly basis. Therefore, in practice we do not include any exogenous factors, the X t factors in the first difference regression. While this may induce a small bias in the estimated elasticities the direction of which is difficult to establish we do not believe it will have a significant impact on the results. We include time-dummies in the difference equations in an attempt to try and pick up some of these effects.
Draft
REGRESSION RESULTS
Prices
(average subscription/connecti on revenue per user) GrameenPhone (post-paid) Subscription Mobile minutes of usage GrameenPhone (pre-paid) Subscription Mobile minutes of usage Aktel Subscription Mobile minutes of usage Banglalink Subscription Mobile minutes of usage Weighted average across the three operators Subscription Mobile minutes of usage -0.46 -0.22 No data No data N/A N/A
0.36
[] []
[] []
3.86
[] []
1.62
[] []
0.924
[] []
[] []
-0.18 -0.73
Draft
Dependent variable
Independent variables (t-statistic) Quarterly change average subscription price Quarterly change in average usage price [] (0.75) 22 0.25
[] (2.05)
Quarterly change average subscription price Quarterly change in average minutes of usage per subscriber Observations R-squared [] (1.17) 22 0.96
Draft
Dependent variable
Independent variables (t-statistic) Quarterly change average subscription price Quarterly change in average usage price per minute [] (0.75) 22 0.03
NA
Quarterly change average subscription price Quarterly change in average minutes of usage per subscriber Observations (quarterly) R-squared Table A12: GrameenPhone regression results pre-paid
Source: Frontier Economics
NA
0.52
Notes: All variables are measured in logs. Therefore the coefficients can be interpreted as elasticity estimates For reasons of data confidentiality, we do not report the regression coefficients for each operator separately. Parameters marked with an [] symbol have been redacted from the document.
Draft
Dependent variable
Independent variables (t-statistic) Quarterly change average subscription price Quarterly change in average usage price per minute [] (2.35) 23 0.17
NA
Quarterly change average subscription price Quarterly change in average minutes of usage per subscriber Observations (quarterly) R-squared NA
0.86
Draft
Dependent variable
Independent variables (t-statistic) Monthly change average subscription price Monthly change in average usage price [] (0.68) 12 0.20
[] (2.01)
Monthly change average subscription price Monthly change in average minutes of usage per subscriber Observations (monthly) R-squared [] (1.22) 12
0.21
Draft
Coefficient (elasticity) (t-statistic) Regression 1 - dependent variable: Quarterly change in subscribers Independent variable: Quarterly change in total revenue per subscriber -0.30 (2.37)
Observations R-squared
23 0.17
Regression 2 - dependent variable: Quarterly change in subscribers Independent variable: Quarterly change in total revenue per subscriber -0.65 (7.91)
Observations R-squared Table A15: Estimated market elasticities full regression results
Source: Frontier Economics
23 0.73
THE RELATIONSHIP BETWEEN THE PRICE OF NEW HANDSETS AND THE PROPORTION OF HANDSET SALES IN TAKING PLACE IN THE FORMAL MARKET
As outlined in the main text, we rely on evidence presented in the GSMA report Tax and the Digital Divide (GSMA, 2005), which shows a strong positive correlation between the market share of legitimate or formal handset sales (as a percentage of total handset sales) and the price of new mobile handsets. Specifically, the previous analysis for the GSMA estimates the following regression: log (informal handset sales market share, %) = A + b * log (legitimate handset price/GDP per capita) The dependent and independent variables are in logs. Therefore, the regression coefficient b can be interpreted as the estimate of the elasticity of demand for
Draft
informal mobile sales with respect to the price of legitimate mobile phones (normalised by GDP per capita)19. The regression line, which is plotted in Figure 17 below, shows that the estimated slope coefficient (i.e., b) is equal to 0.47. This implies that for a 10% reduction in the price of legitimate handsets we would expect to observe a 4.7% increase in the proportion of handsets sold in the legitimate or first-hand market20 - this has the effect of raising the tax revenues from new handsets purchased.
0.0 -0.2
-0.4 -0.6 -0.8 -1.0 -1.2 -1.4 -1.6 -1.8 -2.0 -5.0
-4.5
-4.0
-3.5
-3.0
-2.5
-2.0
-1.5
-1.0
-0.5
0.0
Figure 17: Estimating the relationship between the log of informal handset sales (black market and second had) and the log of the legitimate (first-hand retail) handset price (as a proportion of GDP per capita).
Source: Frontier Economics Notes: The data used to plot the figure was take from the work Frontier has carried out for the GSMA report Tax and the Digital Divide. Data on handset sales and handset prices are for 2004/05 in each country and are estimates provided by the major handset manufacturers and retailers in each country. The countries included in the figure are Chile, Mexico, Argentina, Venezuela, Brazil, Peru, Colombia, Ecuador, Guatemala, Russia, Turkey, Malaysia, Thailand, Sri Lanka, Cameroon, Philippines, Bolivia, India, Cote d'Ivoire, Ukraine, Indonesia, China, Bangladesh, Cambodia and Vietnam.
19
The regression coefficient b is effectively a cross price elasticity of demand for informal mobile handsets with respect to the price of legitimate mobile handsets. Note that this is a proportionate change. Therefore, in the above example, if the proportion of handsets sold in the legitimate market before the 1% price drop is 50%, the simulated proportion after the price drop is equal to 50%*(1+0.47%)=50.24%.
20
Frontier Economics Limited in Europe is a member of the Frontier Economics network, which consists of separate companies based in Europe (London & Cologne) and Australia (Melbourne & Sydney). The companies are independently owned, and legal commitments entered into by any one company do not impose any obligations on other companies in the network. All views expressed in this document are the views of Frontier Economics Limited.
Frontier Economics Ltd 71 High Holborn London WC1V 6DA Tel. +44 (0)20 7031 7000 Fax. +44 (0)20 7031 7001 www.frontier-economics.com