Ex-post Analysis of Competition Policy Decision:
An Empirical Analysis of Common Ownership in the Indonesian Mobile Telecom Market
Master Thesis
Tsuraya Nurrahma Maulana (11375302) MSc Economics
Supervisor: Dr. Jo Seldeslachts
2 Statement of Originality
This document is written by Tsuraya Nurrahma Maulana who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3 Ex-post Analysis of Competition Policy Decision:
An Empirical Analysis of Common Ownership in the Indonesian Mobile Telecom Market
Abstract
Based on a particular case study in Indonesia, this thesis investigates the efficacy of competition policy decision concerning common ownership of industry rivals. In 2007, the Indonesian Competition Commission (KPPU) opened an antitrust case against Temasek’s common shareholdings of two leading operators in Indonesia –Telkomsel and Indosat. The common ownership led to Temasek’s dominant control over Indonesian mobile telecom market, which in turn resulted in a very high price of mobile services charged for consumers. As a remedy, the KPPU adopted an antitrust decision which ordered Temasek to divest its share ownership in one of the operators with a view of helping preserve fair competition among mobile operators in Indonesia. This thesis provides a quantitative analysis that highlights the market-wide impact of the enforcement of KPPU’s decision in terms of mobile telecom prices at market level. The empirical model in this study employs annual average revenue per user (ARPU) at the operator level as the proxy for mobile prices. By using the Difference in Differences (DiD) approach, it compares the evolution of prices in Indonesia and the selected control countries in some years before and after the policy enforcement (i.e. 2001-2015). The main finding demonstrates that the implementation of competition policy decision, in which implied the breakup of Temasek’s common ownership of two operators in Indonesia, contributes to a 22-36% price decline. It indicates that the decision adopted by the KPPU was appropriate to keep a level playing field in the mobile market, and the KPPU’s intervention was beneficial for consumers.
4 Table of Contents
1. Introduction ... 5
2. The Temasek Antitrust Case... 7
3. Literature Review ... 10
4. Methodology and Data ... 14
A. Methodology ... 14
B. Data and Summary Statistics ... 16
C. Empirical Specification ... 20
5. Findings and Interpretation ... 24
6. Conclusions ... 28
References ... 31
5 1. Introduction
Competition authorities recently have taken increased interest in common ownership situations in which the competing firms operating in the same industry share similar shareholders (Nigro et al., 2016; Dentons Europe LLP, 2016). They have concerns that common ownership potentially generates substantial competitive harm in a market, even if the shareholdings are a minority. There are theoretically and empirically sound reasons for competition authority’s concern about common ownership. Since the 1980s, some scholars have developed theoretical literature that common ownership could reduce the firms’ incentive to compete. It consequently brings potential anticompetitive effect in a market: less output and higher prices (Bresnahan and Salop, 1986; Reynolds and Snapp, 1990). Recent empirical studies of Azar et al. (2015) and Azar et al. (2016) also established consumer harm from common ownership in the market, particularly in a highly concentrated market.
Treatment of common ownership then becomes an important issue for antitrust and regulatory policy. The Indonesian Competition Commission (KPPU) is one of the competition authorities which have experienced in dealing with common ownership.1 In 2007, the KPPU opened an antitrust case on Temasek concerning its common shareholdings of two leading mobile telecom operators in Indonesia (i.e. Telkomsel and Indosat). Temasek Holdings (Temasek) is an Asian investment firm incorporated in Singapore which manages a widely diversified portfolio investment in telecommunications and media, financial services, real estate, transportation, and logistics. Through its subsidiaries, in 2002 Temasek owned 35% of Telkomsel shares and 40.81% of Indosat, respectively. At that time, Telkomsel was the biggest mobile telecom operator in Indonesia with 68% market share in terms of total subscribers, followed by Indosat with 21% market share.
Temasek’s common ownership of Telkomsel and Indosat led to the company’s control of more than 50% of the relevant market, which in turn had lessened competition and had resulted in an excessive price for mobile telecom services in Indonesia in 2003-2007. In the KPPU decision, the KPPU elaborated the exorbitant price through mobile telecom price comparison among Asian countries, where Indonesia’s price was the highest – approximately twice times higher – compared to prices in Malaysia, Brunei, Thailand, India, Singapore, and Vietnam. Furthermore, the KPPU stated that the mobile tariff charged to consumers was way
1 Another competition authority, for instance, is UK Competition Commission (CC). In 2008, the CC found that the BAA’s common ownership of seven UK airports restricts competition, and thus ordered the divestiture of two BAA’s London airports and one Scotland airport (The Telegraph, 2008).
6
above ‘the recommended tariff’ in the report of OVUM2
to the Indonesian telecom regulator, which was calculated based on the origination and interconnection fee in Indonesia. A very high EBITDA margin obtained by the Indonesian mobile telecom operators (around 55-70%) during the common-ownership period also served as an indication for KPPU to claim the excessive price in the mobile industry due to Temasek’s common ownership of the two largest operators in Indonesia (KPPU, 2007).
The KPPU made a decision concerning Temasek’s common ownership in Telkomsel and Indosat, in which the KPPU ordered Temasek to cease its common ownership in the Indonesian mobile telecom market by divesting its shareholdings in Telkomsel or Indosat. As per the KPPU’s decision, Temasek then sold its all Indosat shares to Qatar Telecoms Q.S.C. (Qtel) in June 2008 which can remark as the breakup of Temasek’s common ownership in Indonesian mobile telecom industry.
The enforcement of competition policy historically has created substantial change in the functioning of mobile telecom industry (Lear, DIW Berlin, and Analysys Mason, 2016). Competition authorities conduct different policies as a response to potential anticompetitive issues related to the merger, abuse of dominance, and cartel in the mobile telecom market. It necessitates a case-by-case approach to assess the policy’s effect on competition and market performance and to evaluate whether the policy enforcement has achieved its desired impact. This thesis investigates the effectiveness of the KPPU’s decision on Temasek case related to the Temasek’s common shareholding of two mobile operators in Indonesia. It aims to examine the extent to which the Indonesian mobile telecom market has changed as a result of the KPPU’s competition policy and to estimate the impact of the KPPU’s intervention in terms of mobile telecom price.
Current literature on the ex-post evaluation studies on the KPPU’s decision is quite limited in Indonesia, while none of them discussed the common ownership issue. The KPPU has only conducted ex-post evaluation once, namely the evaluation of the KPPU decision on short message service (SMS) cartel case in 2010. The study employed the compensating variation model to measure total consumer welfare after the KPPU’s policy enforcement (OECD, 2011). Using DiD model which has been widely used as a quantitative technique for ex-post evaluation studies, this thesis proposes a new approach to evaluate the changes of
2
Ovum is an independent analyst and consultancy firm specializing in global coverage of IT, and telecommunications industries.
7
market performance due to the KPPU’s intervention and antitrust enforcement activity, particularly in the mobile telecom market.
The key finding in this study is that the implementation of the KPPU’s decision related to Temasek’s common ownership in the mobile telecom industry has a positive effect on the market outcome. Applying Difference in Differences (DiD) method, this thesis discovers that the Temasek’s breakup of common ownership in the market as per the KPPU’s decision is associated with a 22-36% decline in mobile prices. Following to the results, it shows that the decision adopted by the KPPU was appropriate to preserve competition among the operators and also beneficial for consumers. In the context of Indonesia, the results in this thesis suggest the importance of evaluation studies on the enforcement of KPPU’s decision/policy to assess the efficacy of enforcement activity undertaken by the KPPU.
The thesis is divided into the following sections. Section 2 provides background information on the antitrust case investigated by the KPPU in relation to Temasek’s common ownership of two mobile operators in Indonesia. This section also includes a review of the relevant literature on common ownership and price measurement in the mobile telecommunications market. In Section 3, a description of the methodology, the dataset, and the estimation model is presented. Section 4 presents the empirical results. The conclusions of the study for the ex-post evaluation of the KPPU’s decision on Temasek case are drawn in Section 5. Furthermore, the shortcomings to the thesis will be discussed in this chapter to provide suggestive remarks that might be useful for future research and policy applications. The last section contains an appendix which presents robustness checks of the model specification in this thesis.
2. The Temasek Antitrust Case
The Temasek antitrust case is one of the prominent cases investigated by the Indonesian Competition Commission (KPPU) in the history of antitrust enforcement in Indonesia. It is the first case in which the KPPU has opened an investigation and made a ruling about common ownership in Indonesia. This case involved two leading mobile operators in Indonesia: PT Telekomunikasi Seluler (Telkomsel) and PT Indosat Tbk (Indosat). Telkomsel was established in 1995 as a subsidiary of PT Telekomunikasi Indonesia (Telkom), the largest telecom services company in Indonesia. Indosat had involved in telecommunication services business since 1967, and in 1995 it started providing mobile telecom services.
8
Temasek Holdings (Temasek) is an Asian investment firm incorporated in Singapore which manages a widely diversified portfolio investment in telecommunications and media, financial services, real estate, transportation, and logistics. At the end of 2001, Temasek’s subsidiary – i.e. Singapore Telecommunications Ltd (Singtel) – purchased 17.3% and 5% of Telkomsel shares from KPN Netherland and Sedtco Megacell Asia through Singapore Telecom Mobile Pte Ltd. In the following year, Singtel acquired an additional 12.72% from Telkom, bringing its total to the 35% of share ownership in Telkomsel in 2002. Another Temasek’s subsidiary – i.e. Singapore Technologies Telemedia Pte Ltd (STT) – acquired 41.9% of Indosat shares from the Indonesian government through a public bidding process in 2002. In sum, Temasek owned 55% of Singtel, which in turn had 35% of Telkomsel shares; and wholly owns STT which in total held 40.81%3 of Indosat. This shareholding structure brings Temasek to have common ownership of two Indonesian mobile operators, Telkomsel and Indosat, as illustrated in the following figure.
Figure 1. Temasek’s common ownership in the Indonesian mobile telecom industry
As commonly occurred in telecommunication markets (Lear, DIW Berlin, and Analysys Mason, 2016; Sung and Kwon, 2011), the Indonesian mobile telecom industry has witnessed to high barriers to entry (such as the scare nature of spectrum, technology
3
Some slight changes occurred in the structure of Indosat share ownership took place during 2004-2005 due to ‘Employment Stock Program’ and acquisition of minority shareholding.
9
development, high costs of building national coverage) which have limited the number of operators that could profitably enter the market. Furthermore, the Indonesian market has also been characterized by unsymmetrical competition, where three large players dominated the market (Hasan and Afifah, 2008). Besides Telkomsel and Indosat, another leading operator is PT XL Axiata Tbk (XL). XL was founded in 1996, and its presence in the market has been intensified competition among the mobile telecom operators in the market.
During the common ownership period (i.e. 2002 – mid-2008), Telkomsel was a market leader with an average market share in terms of subscribers of roughly 61%. Indosat and XL had 25% and 13% subscriber market share respectively. Notwithstanding that (1) Temasek has no direct shares in both operators and (2) Singtel’s stake in Telkomsel as well as STT’s stake in Indosat were not majority shares, KPPU regarded Temasek as a single economic entity with the two mobile operators, Telkomsel and Indosat (KPPU, 2007). Thus, Temasek’s common ownership of the two operators led to Temasek’s control over more than 50% market share in the Indonesian mobile telecom industry.
Following to the concern about the dominance control of Temasek in the mobile industry, at the beginning of 2007, the Indonesian Competition Commission (KPPU) opened an investigation regarding Temasek’s common ownership of the two leading mobile operators, Telkomsel and Indosat. The KPPU published its decision on 19 November 2007 and stated that the common ownership had led to Temasek’s dominant control over the market, which in turn had significantly increased market concentration and operators’ market power to charge a higher price to consumers.
The quantitative evidence in the KPPU’s decision confirmed that the Temasek’s common ownership in the mobile telecom market facilitated Telkomsel as a market leader to conduct excessive pricing. By analysis of price trend comparison among the operators, the KPPU also asserted that other mobile operators followed the excessive pricing of a market leader (Hasan and Afifah, 2008), and this resulted in harm to consumers (KPPU, 2007). As a remedy for anticompetitive impacts of the common ownership in the market – i.e. higher mobile telecom prices for consumers, the KPPU imposed penalties on Temasek to sell off its shares in one of the two leading operators to cease the Temasek’s common ownership in the mobile telecom market.
The Indonesian Competition Law has not mentioned any sanctions to the undertakings which do not adopt the KPPU decision. The Law only stipulates that the firms
10
are obliged to implement the decision within 30 days after they received a copy of signed decision. If they are willing to object the KPPU decision, they allow appealing the decision to the District Court, and the Supreme Court. In this regard, Temasek filed an appeal against the KPPU’s ruling before the District Court asking for the annulment of the KPPU’s decision. But then the Court upheld the KPPU’s decision and confirmed the order to Temasek to divest its shares in Telkomsel or Indosat with the aim of breaking-up the Temasek’s common ownership in the mobile telecom industry. To comply with the KPPU’s decision, on June 2008 Temasek sold its Indosat share to Qatar Telecoms. Under the deal, Qatar Telecoms paid USD 1.8 billion to acquire the 40.81% stake in Indosat held by Asia Mobile Holdings, a joint venture between Qatar Telecom and STT. The shareholding structure of Telkomsel and Indosat after the Temasek’s divestment of Indosat shares is shown in the below figure.
Figure 2. The post-decision enforcement shareholding structure of Telkomsel and Indosat
3. Literature Review
This thesis empirically identifies an effect on the mobile telecom market of competition policy decision adopted by the Indonesian Competition Commission (KPPU) in relation to common ownership. On that subject, it examines the impact of the breakup of common ownership – as instructed by the KPPU – on market prices. Hence, the analysis builds on theoretical literature on the competitive effects of common ownership.
11
Salop and O’Brien (2000) distinguished two aspects of partial ownership in their economic framework analysis: financial interest and management control. The first point is related to firms’ profit maximization, and the latter refers to the company’s control over competitive decision making including pricing and production level. Both aspects become links how the partial ownership may change the competitive incentives of firms which are supposed to be rivals.
By focusing on horizontal joint ventures among competitors, Bresnahan and Salop (1986) concluded that partial ownership reduces the incentives of firms to expand output and thereby lower price toward the competitive level. In a market where horizontal joint ventures among competitors exist, competition among the rivals would be unprofitable, or at least less profitable, since their profit maximizations are structurally linked to each other. As a consequence, they compete less vigorously; and then result in less output and higher prices than otherwise. As the degree of partial ownership increases, the market becomes less competitive and then leads to a more profitable non-cooperative equilibrium (Reynolds and Snapp, 1990; Farrell and Saphiro, 1990). Furthermore, they also addressed that even if the shares ownership is relatively small, the competitive effects basically might be substantial. Shelegia and Spiegel (2012) used a Bertrand oligopoly model and acknowledged that partial common ownership creates a less competitive outcome. The equilibrium even could be as high as the monopoly price of the most efficient firm.
Common ownership can be seen as a form of partial integration between firms (Azar, 2011). Having a common shareholder is likely to influence firms’ objective functions, such that the companies account for shareholders’ partial interests in competing firms. Recent studies of Azar et al. (2015) and Azar et al. (2016) provided empirical evidence of adverse effects of common ownership on market performance. Both studies respectively investigated the US airline and banking industry in which diversified institutional investors create a network of common ownership. Their results are in line with the theoretical literature and established that the increased common ownership is correlated with higher prices for consumers, therefore, brings less competitive market outcome.
Taking into account a review of literature above, the enforcement of KPPU’s decision which implied the breakup of Temasek’s common ownership of two leading players in the market is expected to restore the competitive conditions which would have been in the
12
absence of the common ownership. Hence the decision implementation could decrease the price of mobile telecom services in Indonesia.
To measure the causal effects of the antitrust enforcement activity in the mobile telecom market, this thesis deals with a specific challenge in obtaining a precise measure of the price of mobile telecom services. Affeldt and Nische (2014), Aguzzoni et al. (2015), and McCloughan and Lyons (2016) acknowledged that the appropriate measure of price for cellular services is not trivial despite the fact that mobile operators typically announce their prices to end customers. Mobile telecom services relate to a bundling of various services – e.g. voice call, text messages and data services, and operators often offer one tariff for a bundling of services. In other cases, operators determine several tariff categories (e.g. a per-minute or per-text charge for services) and tariff variants at the same time, and their tariff portfolios are continually adapted. Customers who typically have the different usage patterns may have the different relevance of tariffs since the actual price paid (tariff charged) depends on the usage of each customer.
Mobile pricing involves complex nonlinear ‘tariff schedules’ thus it is impossible to account for all aspects of the complexity of measuring the price of mobile telecom services. Several past empirical studies had attempted to use simpler rules, and in sum, there are three approaches to estimate the price in mobile telecom.
a. The price of service baskets
This method combines information on a cellular tariff with a fixed usage consumption profile to determine monthly expenditure of a customer for mobile services and then uses the estimates as a proxy for the price of mobile telecom. Usage profiles are defined by several consumption baskets for services which combine minutes of calls, a number of text messages sent, and volume of data services. To account for possible heterogeneous effects due to different usage characteristics, Aguzzoni et al. (2015) and Lear, DIW Berlin, and Analysys Mason (2016) distinguished three consumer types (i.e. low, mid, and high usage) which then produce three mobile expenditure levels. A sample of tariffs for each user profile is identified by the lowest tariff in a country, taking into account that consumers typically self-select their preferred (cheapest) tariff plan offered by operators in the same market. Tariffs from the largest two network operators in a country each year
13
are usually picked as tariff sample to construct a monthly mobile expenditure (Affeldt and Nische, 2014).
Some studies about the effects of antitrust enforcement in mobile telecom market are in the view that the price basket-based is the most appropriate measure for mobile price (Aguzzoni et al. (2015); Lear, DIW Berlin, and Analysys Mason (2016); Genakos et al. (2015); Csorba and Pápai (2013)). The price basket-based measure allows us to estimate the market price for a known basket of services offered by operators. The problem with this approach is that one often may choose irrelevant tariffs as tariff sample or define usage profile which is not suited for a market. Moreover, by using fixed usage consumption profile, changes in consumer usage behavior intra- and inter-profile are not considered. This measure also requires careful tracking of mobile tariffs and their changes over the period of the study.
b. Average revenue
Another alternative approach is to track aggregated revenue figures as employed in McCloughan and Lyons (2016). This approach simply uses the average revenue per user (ARPU) as the indicator of the effective price. It is the unit revenue of mobile services on a monthly basis which includes all revenues generated at the retail and wholesale level, in other words, it considers wider services than price basket-based measures which generally measures only voice calls, text messages, and data consumption. ARPU is commonly used by operators and regulators to compare the performance of mobile telecommunications market. ARPU could be considered as a proxy of the affordability of a service as the decreasing ARPU shows that mobile services have become more affordable for consumers over time (Lear, DIW Berlin, and Analysys Mason, 2016). Higher ARPU may be linked with ‘significant market power’ owned by operators in a market thus some national regulatory authorities argued that higher ARPU is the result of lack of effective competition (McCloughan and Lyons, 2016).
Although ARPU is commonly used as the average revenue measure in the mobile telecom services, academicians relatively have not given much attention to this approach. McCloughan and Lyons (2016) and Genakos et al. (2015) provided the main drawbacks of ARPU which is that the underlying usage changes over time. Unlike the price basket-based measure, ARPU measure does not incorporate the
14
differences in usage of services thus the effect of changes in unit prices and usage cannot be distinguished separately. It might lead to an unreliable estimate for the price as high ARPU may be due to usage induces by low prices or high prices or a combination of both.
c. Average revenue taking usage into account
Considering the ignorance of differences in usage levels present in price measures may lead to a serious issue, some academic studies extended the ARPU approach by including information on usage. Affeldt and Nische (2014) referred to a number of subscribers as a proxy for usage. Another commonly used indicator is MoU (minutes of use) referring to a number of total billable minutes for a particular period. Some scholars combined information on ARPU and MoU to estimate average revenue per minute (RPM), which is a monthly voice-only average revenue per user by minutes of use (Sung and Kwon, 2011; Hazlet and Munoz, 2009; Hausman and Agustin, 2013). RPM is a more appropriate measure of mobile telecom price if a study focuses on the cost per minute or only relate to voice call services.
4. Methodology and Data A. Methodology
The methodology of this research is to develop an empirical model to measure the market-wide effect of competition policy enforcement. Specifically, the model shall be able to estimate the impact of antitrust decision enforcement on the evolution of mobile prices. Ex-post policy evaluation study involves comparison of the two outcomes in a market: the actual outcome after the policy enforcement and the hypothetical outcome which would have occurred in the absence of the policy. A simple comparison of ‘before and after’ conditions, however, may not be an appropriate approach in examining the effect of one policy on mobile prices. Mobile telecom market has witnessed to a long-term path of a price decrease. Thus it is likely to fail an underlying assumption of ‘before and after’ approach that pre-policy enforcement price would have been similar to but-for prices (Aguzzoni et al., 2015).
Treatment evaluation model was developed as an extension of simple ‘before and after’ approach and commonly applied to the ex-post assessment of policy evaluation. The
15
idea is to measure the effect of policy at a given period by comparing both outcomes, both before and after the policy enforcement, of the country in which the concerned policy took place with other countries in which the policy is absent over the same period. This approach allows us to estimate the change in market outcomes that would have been observed in the country of interest absent the policy. This model is also widely known as Difference in Differences (DiD) and has been used in the economic studies on program evaluation for the estimation of treatment effects (Ashenfelter, 1978; Ashenfelter, Card, 1985; Imbens and Wooldridge, 2009).
In industrial economics, there exists an extensive literature on empirical studies which employ DiD as a quantitative technique of the analysis concerning mobile telecommunications market. Most past studies using DiD have revolved around the ex-post evaluation of mergers to assess how the merger decisions (including the remedies) have affected prices in a market (Aguzzoni et al., 2015; Lear, DIW Berlin, and Analysys Mason, 2016; Csorba and Papai, 2015; Genakos et al., 2015; RTR-GmbH, 2016).
This thesis would employ the DiD approach as often conducted in ex-post evaluation of merger decision enforcement. The breakup of Temasek’s common ownership in the mobile industry which also implies the implementation of the KPPU’s decision would act as the ‘treatment or event’ in the model. The DiD approach allows us to isolate the pricing effect due the antitrust enforcement by comparing price fluctuations in the treated group (i.e. the country affected by the KPPU’s decision / Indonesian market) as against prices in other markets (i.e. the control group / outside Indonesian market) in which no similar policy enforcement has taken place over the same period. Price variation over time of the control group can be used to estimate the level of prices in the but-for/counterfactual scenario. Then the average effect of the competition policy enforcement then can be assessed through a price comparison of the treated and control group before and after the enforcement of the KPPU’s decision (i.e. the breakup of Temasek’s common ownership in the mobile industry).
The identification of appropriate control group is a critical key to the validity of the analysis in the DiD approach (Aguzzoni et al., 2015). Control group must satisfy two requirements: they are affected by similar shocks as the treated country, but no spillover effects occurred due to the analyzed event. The two underlying conditions then allow us to assume that price change over time observed control group approximates the price change that would have occurred in treated country absent the ‘event’. However, the prices of treated
16
and control groups may not share common trend due to structural differences across the groups that affect price level. Some explanatory control variables are therefore necessary to be included in the model to capture the differences to some extent. This thesis would perform the visual and formal test of common trend hypothesis to test whether both treated and control groups have a similar trend in prices once the observable control variables are included in the model.
B. Data and Summary Statistics
The dataset consists of an unbalanced panel of 13 mobile telecom operators from 5 countries with observations running from 2001 to 2015. Table 1 below provides a list of operators along with its corresponding country.
Table 1. List of operators and its origin countries in the dataset
No. Mobile operators Country
Treated group
1. Telkomsel Indonesia
2. Indosat Indonesia
3. XL Indonesia
4. Smartfren (Mobile-8/Fren) Indonesia Control group
5. AIS Thailand
6. DTAC Thailand
7. Smart Philippines
8. Global Telecom Philippines
9. China Mobile China
10. China Unicom China
11. NTT Docomo Japan
12. Softbank Japan
13. Au Japan
The data have been collected from several resources including each operator’s annual reports. A specific database was created which covers annual-basis information on types of mobile services offered (i.e. prepaid or postpaid), number of subscribers for each type service (i.e. prepaid or postpaid) and total subscribers, ARPU for each type service (i.e. prepaid or postpaid) as well as overall ARPU, and time of entry. Unfortunately, not all operators in the
17
dataset published their respective ARPU figures for prepaid and postpaid services. Mobile operators from China and Japan estimated only the total ARPU, while the operators from the other countries publish data on three types of ARPU (prepaid, postpaid, and total). Thus, this study focuses on ARPU total as the proxy for mobile telecom prices and does not further distinguish whether the operators’ revenue comes from prepaid or postpaid services.
Information on operators’ time of entry is used to generate the variable entry which indicates the order of operators’ entry to the corresponding mobile market: incumbent, second entrant, third entrant, etc. The categorization in variable entry also takes into account the entry of operators who are not in the dataset.
Furthermore, data at the country-level covering economic and demographic variables were supplemented from World Bank and OECD Database. They are real GDP per capita, total mobile subscriptions, mobile subscriptions per 100 people (i.e. mobile penetration), population density, rural to total population ratio, and exchange rates from the local currencies of the countries in the dataset to US Dollar and Indonesian Rupiah. In addition to that, there is a dummy variable MNP indicating whether the state has implemented the ‘mobile number portability’. ‘Mobile number portability’ allows users who are willing to change the mobile network operator to retain the same telephone number.
Based on the data on the operators’ number of subscribers and total mobile subscriptions in the corresponding country, the market share of each operator was firstly estimated. These figures were then used to estimate the market concentration ratio of 2 operators (CR2)4. The operators’ ARPUs are reported in the corresponding country’s local currency; thus they also were converted into US Dollar and Indonesian Rupiah.
Table 2. Descriptive statistics
Variable Mean Standard
deviation Minimum Maximum ARPU (USD) 17.785 20.960 1.289 72.925 ARPU (IDR) 178,681.3 209,543.2 11,600 781,917.5 𝑪𝑹𝟐 0.779 0.115 0.575 1.055 GDP percapita (USD) 12,421.4 17,143.88 1,691.207 47,150.37
4 It is likely that the CR2 is less precisely estimated given that the data on the operators’ number of subscribers and total mobile subscriptions are collected from two different sources.
18
Variable Mean Standard
deviation
Minimum Maximum
GDP percapita (IDR) 1.09e+08 1.56e+08 7,397,903 4.87e+08 Population density 203.189 97.693 118.377 351.338 Rural ratio 44.612 18.331 6.502 67.428 Mobile penetration 74.902 39.658 3.076 152.731 Mobile number portability 0.229 0.421 0 1 Entry 1.877 0.884 1 4 Operator’s subscribers
7.92e+07 1.44e+08 414,315 8.26e+08
Table 2 above presents the descriptive statistics for variables employed in this study. ARPU (in terms of US Dollar and Indonesian Rupiah) has large standard deviations across the countries, showing that there is substantial variability in the average revenue received by the operators during the studied period. In general, the CR2 indicates that the mobile telecommunication markets in each country are highly concentrated. Two largest operators have the average market share of 77 percent. The demographic variables, which are population density and rural to total population ratio, also have a notable variability. Large standard deviations relative to mean for mobile subscriptions per 100 people and each operator’s subscription indicates that the penetration of mobile telecommunication services quite varies across time and countries. Some operators in the sample are the incumbent, while other operators are second up to the fourth entrant in the market.
19
Figure 3. ARPU trend in Indonesia by operator in US Dollar and Indonesian Rupiah
Figure 3 provides a graphical comparison of ARPU trend among Indonesian operators over the years. It indicates clear downward price trend from 2001-2015. Telkomsel, as an incumbent, has charged the consumers with the highest price compared to other operators, while the latest entrant (Smartfren) has set the lowest price at least from 2006 onward. However, it is important to note that we should be cautious about this graph as ARPU has often declined in nominal terms over time as a mixed result of economies of scale effects, technical progress, and the benefits of competition (Lear, DIW Berlin and Analysys Mason, 2016). 0 5 10 15 20 a v e ra g e r e v e n u e p e r u s e r in U S D 2000 2005 2010 2015 year period Indosat Telkomsel XL Smartfren US Dollar 0 50000 100000 150000 200000 a v e ra g e r e v e n u e p e r u s e r in I D R 2000 2005 2010 2015 year period Indosat Telkomsel XL Smartfren Indonesian Rupiah
20
Figure 4. Graphical comparison of ARPU between Indonesia and control countries
By plotting the unweighted average ARPU from the ARPUs of each mobile operator in one country, Figure 4 shows the evolution of price for each country over the study period. Red line in the graph indicates the year of 2008 when Temasek complied with the KPPU’s decision to divest its shares in Indosat, which remarks as the breakup of Temasek’s common ownership in the Indonesian mobile telecom industry. Visual comparison of the ARPU series by countries also reflects a clear downward trend along with substantial fluctuation across time for each country for both before and after the enforcement of the KPPU decision. Furthermore, the ARPU in Indonesian Rupiah presents more and larger variations than which in US Dollar. Taking into account these circumstances, this study includes the ARPU (and GDP per capita) variable in US Dollar currency in the model specifications. Besides, Japan shares were a quite different trend in mobile price compared to Indonesia as well as other control countries.
C. Empirical Specification
The DiD model constructed in this thesis aims to measure the causal effect of the enforcement of KPPU’s decision on Temasek case on mobile telecom price. The competition
0 20 40 60 80 A R P U i n U S D 2000 2002 2004 2006 2008 2010 2012 2014 year period Indonesia Thailand Philippines China Japan US Dollar 0 2 0 0 0 0 0 4 0 0 0 0 0 6 0 0 0 0 0 8 0 0 0 0 0 A R P U i n I D R 2000 2002 2004 2006 2008 2010 2012 2014 year period Indonesia Thailand Philippines China Japan Indonesian Rupiah
21
policy decision adopted by the KPPU implied the breakup of Temasek’s common ownership of the two leading operators in the mobile telecom industry.
The dependent variable in the model specification is the price for mobile telecommunication services, measured with the log of ARPU. As elaborated in Section 2, measuring the price of mobile services presents significant challenges. Price basket-based approach has its difficulties in implementation, chiefly in requiring formidable quantities of tariff data over the years which are time-consuming and expensive to gather. Furthermore, the other approach is also quite impossible due to a limitation of the availability of usage consumption data in both treated and control groups.
The key explanatory variable of interest is the variable did, which is the interaction terms of dummy variable event and variable treat. Dummy variable event takes the value of one for the years during the post-event period, i.e. from 2008 onward. Variable treat equals to 1 for Indonesia as the treated group, otherwise for the control countries. Some other variables are also included in the model as controls at the operator- and country-level. The primary econometric model in this study is:
log(𝐴𝑅𝑃𝑈)𝑐𝑖𝑡= 𝛼 + 𝛽1𝑒𝑣𝑒𝑛𝑡𝑡+ 𝛽2𝑡𝑟𝑒𝑎𝑡𝑐+ 𝛽3𝑑𝑖𝑑𝑐𝑡+ 𝛽4𝐶𝑅2𝑐𝑡+ 𝛽5log(𝐺𝐷𝑃𝑝𝑒𝑟𝑐𝑎𝑝𝑖𝑡𝑎)𝑐𝑡
+ 𝛽6log(𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛𝑑𝑒𝑛𝑠𝑖𝑡𝑦)𝑐𝑡+ 𝛽7log(𝑚𝑜𝑏𝑖𝑙𝑒𝑝𝑒𝑛𝑒𝑡𝑟𝑎𝑡𝑟𝑖𝑜𝑛)𝑐𝑡+ 𝛽8𝑀𝑁𝑃𝑐𝑡
+ 𝛽9𝑒𝑛𝑡𝑟𝑦𝑐𝑖𝑡+ 𝛽5log(𝑠𝑢𝑏𝑠𝑐𝑟𝑖𝑏𝑒𝑟𝑠)𝑐𝑖𝑡+ 𝜇𝑐𝑖+ 𝜏𝑡+ 𝜀𝑐𝑖𝑡
where c, i, t indicate country, operator, and year respectively; 𝜇𝑐𝑖 is country-operator fixed-effects, 𝜏𝑡 is time fixed-effects, and 𝜀𝑐𝑖𝑡 is error term.
The coefficient of variable did measures the effectiveness of the KPPU’s decision, which indicates how the competition decision enforcement affected prices of mobile telecom services in Indonesia. In other words, it means the additional price variation in the Indonesian market compared to the control groups due to the breakup of Temasek’s common ownership. According to economic literature in Section 2, common ownership likely brings higher prices for the consumer. Thus we expect the coefficient on the variable did to be statistically significant and negative. The negative value of this coefficient indicates that the breakup of Temasek’s common ownership of two mobile operators – as the competition authority’s intervention in the mobile market – was successful at decreasing in price at the market level.
To account for differences across operators and countries as well as over time, several controls were included into the model. The choice of control variables is critical to hold the
22
common trend hypothesis between Indonesia and the control groups in the pre-policy enforcement period. The control variables were drawn from the determinants of mobile price which come from four classes of structural factors for mobile telecommunication prices. They are service quality, market environment, regulation, and quantity service as studied in several previous studies, such as Shew (1994), Affeldt and Nische (2014), and McCloughan and Lyons (2016). Fixed-effects with respect to time and country-operator were also included to capture influences from any unobservable factors other than the control variables across countries and over time.
Table 3. List of control variables in the model specification
No. Variable Description
1. 𝐶𝑅2 It indicates the level of market concentration, estimated by the sum of market shares of the two largest operators in the market.
2. GDP per capita It is one of the push factors of demand for mobile telecommunication services; hence it determines ARPU.
3. Population density It is a proxy for user density. This variable is included to capture differences in cost due to economies of scale (economies of density).
4. Rural ratio It is one of the demographic variables influencing the supply of telecom service. Higher prices may be partly attributable to higher rural to total population ratio which requires more capital costs for coverage expansion.
5. Mobile penetration It uses mobile cellular subscriptions per 100 people as the proxy variable.
6. Mobile number portability (MNP)
It is a dummy variable for the categorisation of whether the country has implemented the MNP. The availability of MNP reduces switching costs of customers who want to change operators. In other words, it implies how flexible the users to switch operator for their mobile telecom services providers. 7. Entry It is the categorization of operators in terms of an
incumbent, second entrant, etc. 8. Operator’s
subscribers
This variable is included to capture quantity supplied in a particular period.
Fixed-effects
23
No. Variable Description
fixed-effects given operator but tend to be invariant over time (e.g. the quality level of mobile telecommunication services).
10. Time fixed-effects It is included to allow for time-varying unobserved factors common to all operators and countries (e.g. falling quality-adjusted costs of technology).
Table 4 below presents statistical evidence that there is a high correlation between the variables GDP percapita and rural ratio. They have a 0.95 correlation one to another. Following to the likely multicollinearity issue, this thesis employs two separate model specifications: the primary regression employs only the log of GDP percapita and excludes the log of rural ratio. In specification 2, the model uses the log of rural ratio instead of the log of GDP percapita.
Table 4. Matrix of correlations for the explanatory variables
treat event did 𝑪𝑹𝟐 GDP percap Pop. density Rural ratio Mobile penetration MNP Entry Subscribers Treat 1 Event -0.27 1 Did 0.68 0.39 1 𝑪𝑹𝟐 -0.30 -0.24 -0.49 1 GDP percap -0.40 0.14 -0.22 -0.41 1 Pop. density -0.52 0.14 -0.31 -0.07 0.66 1 Rural ratio 0.30 -0.19 0.17 0.41 -0.94 -0.74 1 Mobile penetration -0.27 0.72 0.28 -0.36 0.40 0.39 -0.35 1 MNP -0.37 0.32 -0.25 -0.33 0.70 0.44 -0.71 0.4 1 Entry 0.42 0.06 0.32 -0.24 -0.04 -0.11 -0.03 -0.01 -0.02 1 Subscribers -0.36 0.39 0.02 0.10 0.10 0.09 -0.07 0.44 0.02 -0.39 1
The period of the empirical analysis goes from 2001-2015: 7 years prior to the enforcement of the KPPU’s decision in 2008 – i.e. the breakup of Temasek’s common ownership – and eight years thereafter. Given that Temasek started its common shareholdings of the two operators in 2002, I also attempted to shorten the pre-event period to the years 2003-2015 with the aim of testing the robustness of the model. As an additional robustness check, I constructed other model specifications in which Japan data are excluded from the sample considering that the country’s price trend shows quite different to Indonesia and other control groups, as reflected in Figure 4. Table A in the Appendix provides a list of model specifications performed in this thesis. In all specifications, the standard errors are clustered
24
at the country level and are corrected for the problem of heteroscedasticity arising from non-constant variance across observations.
Control groups in this study are the mobile telecom operators from some Asian countries, e.g. Thailand, Philippines, China, and Japan. They are not subject to the enforcement of the KPPU’s decision on Temasek case (i.e. the breakup of Temasek’s common ownership). Furthermore, they are also geographically close to Indonesia thus both treated and control groups are likely to share similar unobservable characteristics. These circumstances support the underlying assumption of common trend hypothesis in the DiD method. Graphically and formally, I tested the common trend hypothesis in each model specification. As a graphical check, I conducted a visual comparison of the development of the price series in the pre-event period. Common price trend between Indonesia and the control groups before the decision enforcement implies that the selected countries and operators are suitable control groups for DiD method.
Furthermore, I also performed a formal test similarly conducted in Aguzzoni et al. (2015). The basic concept of this test is to estimate the deviation of the Indonesian price from the average price of the control countries in each year, and then to assess whether these deviations follow a different trend than the average price of control groups. To conduct this test, I first replaced the event, treat, and did dummy variables in the model equation with one new dummy variable for each year that equals to one only for Indonesia in the pre-event period. Then I performed this ‘new regression model’ only in the pre-policy enforcement period to estimate the slope of a linear trend of the coefficient of this new dummy and test whether the estimated slope is statistically different from zero5.
5. Findings and Interpretation
First, I commenced this section with the evaluation of common trend assumption, one of the main identifying assumptions of the DiD approach. It aims to check the suitability of control groups in estimating the evolution of prices in the treated country before the decision enforcement. The appropriate control groups should be able to graphically and formally present similar price trends to the prices in Indonesia in the pre-policy enforcement period.
5
25
Figure 5 presents the visual comparison between prices in Indonesian and in the control groups. The prices of the control groups were estimated by taking the unweighted average ARPU. Red line in the graph indicates the year of 2008 when Temasek complied with the KPPU’s decision to divest its shares in Indosat, which remarks as the breakup of Temasek’s common ownership in the Indonesian mobile telecom industry. The two series share similar and decreasing price trend. Focusing on the price trend during the pre-policy enforcement (i.e. before 2008), this figure graphically confirms the common trend assumption in Indonesia and the selected control groups.
Figure 5. Graphical comparison of average ARPU in Indonesia and control countries
As the additional check, I also performed a formal test of common trend hypothesis as explained in Section 3. To be appropriate control groups, price deviations in Indonesia shall follow a similar trend to prices in the control groups during the pre-decision enforcement period. In statistical terms, the slope of pre-decision enforcement price trend in Indonesia should not be statistically different from zero to prices in the selected control countries. The result of this test is presented at the bottom of Table 6 below. The null hypothesis of no trend differences between Indonesia and the control groups is rejected for Specification 3 and 4, meaning that the price evolution in Indonesia statistically has a similar trend to the prices in
0 5 10 15 20 25 30 35 40 A R P U i n U S D 2000 2002 2004 2006 2008 2010 2012 2014 year period Indonesia control US Dollar 0 100000 200000 300000 400000 A R P U i n I D R 2000 2002 2004 2006 2008 2010 2012 2014 year period Indonesia control Indonesian Rupiah
26
the control groups during the pre-decision enforcement period. In sum, the graphical and formal test results support the validity of the selected control groups.
Table 5. Estimation Results
Specification 1 (2001-2015) Specification 2 (2001-2015) Specification 3 (2003-2015) Specification 4 (2003-2015) Treat -2.343*** (1.131) -10.632* (2.190) -1.429 (1.011) -12.317* (3.309) Event -0.389* (0.076) 1.147* (0.309) -0.292* (0.075) 1.553* (0.515) did -0.364* (0.054) -0.254** (0.108) -0.357* (0.071) -0.207* (0.043) 𝑪𝑹𝟐 -0.204 (0.283) -0.063 (0.361) -0.430 (0.322) -0.580 (0.374) Log(GDP percapita) 0.740* (0.041) 0.962* (0.030) Log(Population density) -1.738 (1.203) -6.206* (2.148) -1.331 (1.071) -6.900** (2.854) Log(Rural ratio) 1.034* (0.218) 1.541* (0.439) Log(Mobile penetration) -0.335** (0.151) -0.406** (0.180) -0.474* (0.125) -0.650* (0.196) Mobile number portability 0.035 (0.022) -0.175 (0.111) 0.029 (0.033) -0.230*** (0.119) Entry -0.157* (0.009) -0.147* (0.009) -0.164 (0.018) -0.147* (0.008) Log (Subscribers) -0.042 (0.065) -0.025 (0.072) -0.025 (0.052) -0.017 (0.059) Constant 9.089 (7.198) 39.925* (12.959) 4.754 (6.312) 44.028** (16.357) 𝑹𝟐 0.986 0.9839 0.9868 0.9848 N 179 179 161 161
Cluster Country Country Country Country
Time fixed-effects Yes Yes Yes Yes
Operator fixed-effects Yes Yes Yes Yes
Deviation from the common trend
-6.577 -6.799 -2.474 -11.119
Common trend test Failed Failed Passed Passed
Notes: Robust standard error in parentheses, *** p<0.15 ** p<0.10 * p<0.05
Common trend test: ‘Failed’ means that we reject the null hypothesis of common trend at 5% level
Table 5 reports the estimation results for the first four model specifications which include all control groups in the dataset (i.e. Thailand, Philippines, China, Japan). To evaluate the efficacy of the KPPU’s decision on Temasek case, I examined how the enforcement of the decision affected prices in the mobile telecom industry. The evolution of the price in the control groups approximates what would have been the prices in Indonesia in the absence of the enforcement of KPPU’s decision (i.e. the breakup of Temasek’s common ownership).
27
The DiD approach investigates the causal effect of the antitrust enforcement on prices by comparing the price outcomes in Indonesia before and after the enforcement, with the prices in the chosen control countries over the same period.
Following to what has been discussed in Section 2, theoretical and empirical studies conclude that common ownership could bring higher prices for the consumer. Thus the breakup of Temasek’s common ownership as per the KPPU’s competition decision is expected to decrease the price for mobile telecom services in Indonesia. The estimation result in Table 5 presents empirical evidence that the decision adopted by the Indonesian Competition Commission (KPPU) – i.e. the breakup of Temasek’s common ownership – was associated with a significant relative price decline in the Indonesian market. Coefficients for variable did are always significant and systematically negative in the four specifications at the 10% significance level or higher. The KPPU’s intervention in the Temasek’s common shareholdings in the mobile market has caused a positive impact on prices. Given the differences across the four specifications, the enforcement of the KPPU’s decision on the Temasek case seems beneficial; it led to a substantial decrease in mobile telecom prices of between 20-36%.
A series of alternative specifications were also performed to test the robustness of the estimates. Considering that Japan’s price trend seems to be quite different from Indonesia and other control countries, I constructed additional model specifications in which Japan data were excluded from the dataset. Table B in the Appendix presents the results. It broadly shows the similar results: the coefficient for variable did is statistically significant and negative, and the magnitude of the estimated effects is 23-36%. The quantitative evidence over the different specifications collected points to the conclusion that the enforcement of KPPU’s decision –which implies the breakup of Temasek’s common ownership in mobile telecom industry– was successful at decreasing mobile telecom prices in Indonesia. The result of formal common trend test increases the reliability of the estimates in this ex-post study.
The estimated coefficients for the included control variables have the expected signs. GDP percapita has a significant and positive effect on ARPU, meaning that higher GDP percapita increases mobile telecom prices. The result is in line with McCloughan and Lyons (2016) and Affeldt and Nische (2014), established that GDP percapita as the proxy for
28
personal incomes determines the level of demand for mobile telecom services and hence ARPU; and this also applies to Indonesian market (Modjo, 2008).
Looking at the coefficients on a rural ratio, it indicates that demographic effects on mobile telecom prices are significant and positive. As a variable influencing supply of mobile service (McCloughan and Lyons, 2016), higher rural to population ratio is likely to attribute to higher prices since operator requires higher costs of capital to expand the mobile coverage in a market. The coefficients associated with economies of scale are always negative, suggesting as expected a negative relation between mobile telecom prices and these measures.
The inclusion of population density and mobile penetration captures the effects of differences in cost in the market (Sung and Kwon, 2011; Hazlett and Munoz, 2008) thus both variables indicate the level of user density in a market. A higher level of both variables accommodates mobile to increase their economies of scale in providing mobile telecom services to consumers in a market. The coefficient for the variable entry indicates that incumbent seems to have more market power than the operators who enter a given market afterward, which in turn allows the incumbent to charge consumers with a higher price. The effects of other price determinants, such as CR2, MNP, and a number of operator’s subscribers are insignificant on mobile telecom prices, yet they are necessary to be included in the model to take variation across the markets into account.
6. Conclusions
The econometric analysis using DiD approach in this thesis shows that the antitrust enforcement on Temasek case is associated with a decline in mobile telecom prices. The intervention of the Indonesian Competition Commission (KPPU) in the mobile industry, followed by the breakup of Temasek’s common ownership of the two largest operators in the market, was successful at lowering mobile telecom prices. The enforcement of KPPU’s decision, which ordered Temasek to relinquish its common shareholdings of the two operators, had limited Temasek’s control over the market, which in turn reduced Temasek’s significant market power in the Indonesian mobile telecom industry. In sum, the enforcement of competition policy has been beneficial on the functioning of the mobile telecom market. It
29
helped to preserve vigorous competition among the operators, which resulted in a 22-36% price decrease at the market level.
Proper ex-post evaluation study of policy enforcement using DiD approach relies on the underlying common trend assumption, meaning that trends of the outcome variables for the treated and control groups would share the similar pattern in the absence of the policy enforcement. This thesis had performed both graphical and formal test to examine the common trend assumption in the dataset. It found that Indonesia and the control groups have similar trend before the enforcement of KPPU’s decision in 2008. Thus the selected control groups seem to be reliable to approximate the prices in Indonesia but for the enforcement of KPPU policy decision.
However, it is important to be aware of the limitations of this study. There are some reasons for caution in the precision of the estimated price effects. First, this study employs ARPU as the proxy for mobile prices. Past studies show that ARPU fails to consider the usage changes over time. By using the ARPU as price measure, this thesis could not take into account the differences in consumers’ usage which typically influence the mobile prices. Nevertheless, I believe ARPU is an appropriate proxy for the mobile price since other available data on tariff does not cover the entire period of analysis. Second, this study uses data on an annual basis since the available ARPU data are not insufficiently detailed (i.e. they are not monthly or quarterly). In consequence of using annual data, I must include the relatively long period of analysis (i.e. 2001-2015) to meet the minimum statistical sample size. The more extended period of study may distort the estimates result in the event that the effects of the decision enforcement – i.e. the breakup of common ownership – had not lasted until the end of the analysis period.
Another limitation is that this thesis did not take into account the 2008/2009 global economic crisis in the Temasek’s decision to divest all its Indosat shares in 2008. The KPPU decision concerning Temasek’s common ownership of the two leading Indonesian operators was made and adopted in 2008 when Lehman Brothers collapsed, and it sparked massive sell-offs stock exchanges around the world. Even though Indonesia, like the other Southeast Asian countries, withstood the financial turbulence well; the crisis may induce Temasek’s decision to sell all Indosat shares to other investors. In this regard, the divestiture of Indosat by Temasek –which implies the breakup of common ownership in the Indonesian mobile telecom industry – might not be solely a result of the enforcement of the KPPU decision; it
30
can be influenced by the 2008/2009 global economic crisis. It could be occurred considering that the Indosat stock price kept decreasing from the Q4 of 2007 to Q1 of 2009 as shown in Figure C in the appendix.6 If such case happened, it basically would reduce the effectiveness of the antitrust enforcement in relation to Temasek’s common ownership in the Indonesian mobile telecom market.
To conclude, the key finding in this thesis is that the enforcement of KPPU decision on Temasek case (which led to the breakup of Temasek’s common ownership in the Indonesian mobile telecom industry) contributes to a 22-36% price decline. This might be related to a decrease in control and market power of Temasek in the market after the company divested the shareholdings in one of the leading operators. Hence, the antitrust enforcement related to Temasek’s common shareholding in the mobile telecom industry has been appropriate to preserve a level of playing field among the operators; and beneficial for the functioning of mobile telecom market as well as consumers. By using the DiD approach which is a common quantitative technique for ex-post policy evaluation study, this thesis provides a valuable addition to the existing literature how the competition decision adopted by the Indonesian Competition Commission (KPPU) has affected the market performance. This thesis hopefully shows that the proper ex-post evaluation of KPPU’s competition policy enforcement might be carried out despite the typical constraints in the availability of data, human resources, and funds in Indonesia (OECD, 2011).
6
31 References
Affeldt, Pauline and Nitsche, Rainer (2014) A Price Concentration Study on European Mobile Telecom Markets: Limitations and Insights. ESMT Working Paper.
Aguzzoni, Luca et al. (2015) Ex-post Analysis of Two Mobile Telecom Mergers: T-Mobile/tele.ring in Austria and T-Mobile/Orange in the Netherlands.
Ashenfelter, Orley (1978) Estimating the Effect of Training Programs on Earnings. The Review of Economics and Statistics Vol. 60 No. 1 1978 pp. 47-57.
Ashenfelter, Orley and Card, David (1985) Using the Longitudinal Structure of Earnings to Estimate the Effect of Training Programs. The Review of Economics and Statistics Vol. 67 No. 4 1985 pp. 648-660.
Azar, Jose (2011) A New Look at Oligopoly: Implicit Collusion through Portfolio Diversification. PhD Dissertation, Princeton University.
Azar, J., Schmalz, Martin C., and Tecu, Isabel. (2015) Anti-competitive Effects of Common Ownership. Ross School of Business Paper No. 1235.
Azar, J., Raina, S., and Schmalz, M. (2016) Ultimate Ownership and Bank Competition. University of Michigan Working Paper.
Bresnahan, Timothy F. and Salop, Steven C. (1986) Quantifying the Competitive Effects of Production Joint Ventures. International Journal of Industrial Organisation 4, 155-175. Csorba, Gergely and Papai, Zoltan (2015) Does One More or One Less Mobile Operator Affect Prices? A Comprehensive Ex-post Evaluation of Entries and Mergers in European Mobile Telecommunication Market. Econstor, ZBW Leibniz-Informationszentrum Wirtschaft Leibniz Information Centre for Economics.
Dentons Europe LLP (2016) “Common Ownership”: New Competition Law Issues with respect to Minority Shareholdings? University of Oxford, Faculty of Law. Available at:
https://www.law.ox.ac.uk/business-law-blog/blog/2016/11/common-ownership%E2%80%9D-new-competition-law-issues-respect-minority [Accessed 19 April 2017]
32
Farrell, Joseph and Shapiro, Carl (1990) Asset Ownership and Market Structure in Oligopoly. The RAND Journal of Economics Vol. 21, No. 2, 275-292.
Genakos et al. (2015) Evaluating Market Consolidation in Mobile Communications. Centre on Regulation in Europe (CERRE) Report.
Hasan, M. Fadhil & Afifah, Evi N. (2008) Kepemilikan Silang, Pola Tarif dan Persaingan Usaha pada Industri Telepon Seluler di Indonesia. Bisnis & Ekonomi Politik (Quarterly Review of the Indonesian Economy) Vol. 9 No. 1 January 2008 The Institute for Development of Economics and Finance (Indef).
Haussman, Jerry A. and Ros, Agustin J. (2013) An Econometric Assessment of
Telecommunications Prices and Consumer Surplus in Mexico Using Panel Data. Journal of Regulatory Economics Vol. 43 No.3 2013 pp. 284-304.
Hazlett, Thomas W. and Munoz, Roberto E. (2009) A Welfare Analysis of Spectrum Allocation Policies. RAND Journal of Economics Vol. 40 No. 3 2009 pp. 424-454.
Imbens, Guido M. and Wooldridge, Jeffrey M. (2009) Recent Developments in the Econometrics of Program Evaluation. NBER Working Paper Series No. 14251 2008 JEL No. C01.
KPPU (2007) Putusan Perkara Nomor 07/KPPU-L/2007 [The Decision of KPPU, Case Number: 07/KPPU-L/2007]
Lear, DIW Berlin & Analysys Mason (2016) Economic Impact of Competition Policy Enforcement on the Functioning of Telecoms Markets in the EU.
McCloughan, Patrick and Lyons, Sean (2016) Accounting for ARPU: New Evidence from International Panel Data. Telecommunications Policy 30, 521-532.
Modjo, M. Ikhsan (2008) Aspek Ekonomi dan Persaingan pada Industri Telekomunikasi Seluler. Bisnis & Ekonomi Politik (Quarterly Review of the Indonesian Economy) Vol. 9 No. 1 January 2008 The Institute for Development of Economics and Finance (Indef).
Nigro, Barry A. et al. (2016) Cross-Ownership by Institutional Investors. Harvard Law School Forum on Corporate Governance and Financial Regulation. Available at:
https://corpgov.law.harvard.edu/2016/03/31/cross-ownership-by-institutional-investors/
33
OECD (2011) Policy roundtables: Impact Evaluation of Merger Decisions.
Reynolds, Robert J. and Snapp, Bruce R. (1986). The Competitive Effects of Partial Equity Interests and Joint Ventures. International Journal of Industrial Organisation 4, 141-153. RTR-GmbH (2016) Ex-post Analysis of the Merger between H3G Austria and Orange Austria.
Salop, Steven C. and O’Brien, Daniel P. (2000) Competitive Effects of Partial Ownership: Financial Interest and Corporate Control. Georgetown Law Faculty Publications, 67 Antitrust L.J., 559-614.
Shelegia, Sandro and Spiegel, Yossi (2012) Betrand Competition When Firms Hold Passive Ownership Stakes in One Another. Economic Letters 114, 136-138.
Sung, Nakil and Kwon, Mi-ae (2011) An Empirical Analysis of the State of Competition in OECD Mobile Wireless Markets. Conference Paper for 22nd European Regional Confernce of the International Telecommunication Society.
The Telegraph (2008) BAA Airports: Competition Commission Findings. Available at:
http://www.telegraph.co.uk/finance/newsbysector/transport/2795038/BAA-airports-Competition-Commission-findings.html [Accesses 14 June 2017]
34 Appendix
Table A. List of specifications for empirical estimations
Model specification Control variable (GDP percapita or rural ratio) Sample period
Control countries included in data sample Specification 1 GDP percapita 2001-2015
4 countries: Thailand, Philippines, China, Japan Specification 2 Rural ratio 2001-2015
Specification 3 GDP percapita 2003-2015 Specification 4 Rural ratio 2003-2015 Specification 5 GDP percapita 2001-2015
3 countries excluding Japan: Thailand, Philippines,
China Specification 6 Rural ratio 2001-2015
Specification 7 GDP percapita 2003-2015 Specification 8 Rural ratio 2003-2015
Table B. Estimation Results – 3 control countries, excl. Japan 3 control countries: excluding Japan Specification 5 (2001-2015) Specification 6 (2001-2015) Specification 7 (2003-2015) Specification 8 (2003-2015) Treat -0.477* (0.055) -0.538*** (0.239) -0.422* (0.061) -0.472 (0.252) Event -0.331 (0.515) -0.115 (0.582) 0.202 (0.262) 0.344 (0.452) did -0.362* (0.108) -0.364* (0.114) -0.234* (0.069) -0.233** (0.074) 𝑪𝑹𝟐 0.095 (0.256) 0.345 (0.201) -0.169 (0.217) 0.072 (0.129) Log(GDP percapita) 0.604* (0.187) 0.729* (0.068) Log(Population density) -4.122** (1.507) -2.871 (4.027) -3.648* (0.963) -1.508 (3.814) Log(Rural ratio) -1.813 (1.360) -2.345 (1.332) Log(Mobile penetration) -0.147 (0.159) -0.266 (0.194) -0.484** (0.173) -0.670* (0.175) Mobile number portability -0.031 (0.054) -0.190*** (0.088) -0.043 (0.036) -0.222** (0.085) Entry 0.591* (0.076) 0.584* (0.073) 0.553* (0.064) 0.548* (0.064) Log -0.085 -0.077 -0.045 -0.040
