Market Competition, Infrastructure Sharing, and Network Investment in China’s Mobile Telecommunications Industry

Abstract

The relationship between market competition and network investment in the mobile telecommunications industry has always been one of the focuses for scholars and regulatory agencies. The conclusions of previous studies on this topic remain ambiguous. Infrastructure sharing has become a noticeable trend in the global mobile telecommunications industry, but its impact on operators’ investment and innovation behaviors is controversial. This paper uses fixed effects and dynamic panel data models to empirically examine the relationship among the market competition, infrastructure sharing, and network investment in China’s mobile telecommunications industry. The results show that market competition has a significant positive impact on the total industry network investment, but the interaction of market competition and infrastructure sharing has undermined network investment, which indicates that the competitive strategy of rational operators will shift from facility-based competition to service-based competition when both deep infrastructure sharing and fierce market competition appear at the same time, and this is likely to cause insufficient incentives for investment in advanced technology. This paper suggests that China’s regulators should improve the market structure, enhance market competitiveness continuously, and support infrastructure sharing, but at the same time they should exercise caution when conducting in-depth infrastructure sharing. In addition, they should accelerate the development of 5G vertical industry applications to expand the market space for industry development.

1. Introduction

Over the past three decades, the regulation of the global telecommunications industry has undergone a fundamental transformation. The highly regulated monopoly business model has long been regarded as the industry’s most technically efficient business method. However, since the mid to late 1980s, some countries have initiated telecommunications reforms to eliminate the market entry barriers, privatize state-owned telecommunications enterprises, and promote market competition. Since the 1990s, this reform movement, called “liberalization” and “deregulation”, has swept the world [1]. Although controversial, the reform is mainly based on the premise that competition will stimulate investment and innovation. The marketization of China’s telecommunications industry synchronized with that of the world. Since 1994, by learning from other countries’ reform experiences and considering China’s own realities, China’s telecommunications have experienced a marketization reform directed at opening up the market and promoting competition, and this trend is still under way. Especially in June 2019, the Ministry of Industry and Information Technology (MIIT) officially issued 5G commercial licenses to four companies, namely China Mobile, China Unicom, China Telecom, and China Broadcast Network, indicating that China Broadcast Network has become the fourth mobile network operator in addition to the other three. The reform has also been the most striking event since the restructuring in China’s telecommunications industry in 2008. This top-down, gradual reform idea of introducing new entrants and deepening full-service competition has inherited the tradition in the past three decades of focusing competition reform in China’s telecommunications industry and made some innovations, such as encouraging four operators to share the 5G network (On 20 May 2020, the MIIT and the State-owned Assets Supervision and Administration Commission of the State Council (SASAC) issued the “Implementation Opinions on Promoting the Sharing of Telecommunications Infrastructure and Accelerating the Development of 5G Networks”. At present, the four operators have reached 5G network sharing agreements in pairs, namely China Telecom + China Unicom and China Mobile + China Broadcast Network). The new round of technological and industrial transformation represented by 5G is currently emerging and becoming both a significant technical support for the economic development of all countries around the world and a strategic high ground for global industrial competition. The Chinese government has proposed to speed up the construction of new infrastructure such as 5G networks and promote the deep integration of advanced information technology and the real economy. In this context, can the reform further stimulate technological innovation and network investment in the industry? Can infrastructure sharing reduce operators’ investment pressure effectively? How to find a balance between implementing national strategies and realizing a sustainable and healthy development for operators in the future? The answers to these questions are crucial for further deepening structural reform, promoting the high-quality development of the telecommunications industry, and stimulating the growth of China’s digital economy.

The development of China’s mobile telecommunications industry in the past ten years has provided a good case for our research purposes. On the one hand, the 2008 telecommunications restructuring introduced a new mobile operator (China Telecom, Beijing, China) besides the two incumbents (China Mobile (Beijing, China) and China Unicom (Beijing, China)). Moreover, the reform has achieved remarkable results after two new technology cycles, 3G and 4G. On the other hand, under the joint leadership and coordination of the state-owned Assets Supervision and Administration Commission of the State Council (SASAC) and MIIT, China Tower was established to promote the telecommunications infrastructure sharing in 2014. These two reform events coincided with the introduction of CBN in the 5G era, and the regulator has encouraged operators to share the 5G networks. Therefore, this paper plans to use industry-level data from 2010 to 2019 and adopt empirical analysis methods to examine the relationship among market competition, infrastructure sharing, and network investment. The paper will also evaluate the reform performance and forecast investment outlook and make policy recommendations. The main contributions include: (i) this article focuses on the recent development of the world’s largest mobile telecommunications market, with a unique institutional environment—China’s mobile telecommunications industry—to examine the relationship between market competition and network investment. The results of the research can provide unique and important empirical evidence for improving and developing the theory about the link between competition and investment. (ii) Distancing itself from the existing literature, this paper also includes and analyzes the impact of telecommunications infrastructure sharing, an essential supplement to the existing research in this field. (iii) Competition reform, regulation reform, and infrastructure sharing are all complicated issues in telecommunications reform. The results presented have important policy implications for deepening the structural reform of the telecommunications industry.

The remainder of this paper is organized as follows: Section 2 briefly reviews and summarizes the findings in the literature on the relationship between competition and investment, focusing on the research in the mobile telecommunications industry. Section 3 combines theory with practice to present research hypotheses. Section 4 introduces the data, variables, and models applied in this article. Section 5 introduces and analyzes the empirical results of the paper. Section 6 summarizes the conclusions and proposes and discusses policy recommendations.

2. Literature Review

The link between market competition and innovation (or investment) is one of the most important relationships examined in economics. Although the research on this issue has been going on for decades, there is no definitive conclusion yet [2,3,4,5,6]. Fundamentally, there are two conflicting views on this relationship. Some emphasized that large firms operating in a concentrated market are more capable of, and motivated to, innovate—this was labeled the “Schumpeterian Effect”—while others argued that more competition leads to greater innovation, namely that more competition induces firms to innovate in order to escape competition—this opinion is labeled “Escape–Competition Effect” [7]. A theoretical hypothesis that unifies these two opposing views is the inverted U-shaped relationship between competition and investment. When the competition intensity increases from low to moderate level, firms tend to strengthen innovation to eliminate the entanglement of competitors (the escape–competition effect dominates). When the competition intensity gets closer to a perfectly competitive market level, market competition significantly reduces monopoly profits, which reduces firms’ incentive for innovation (the Schumpeterian effect dominates) [3]. Schmutzler [8] also indicates that the relationship between competition and investment is not affected in an unambiguous way by the level of pre-existing competition. Therefore, due to theoretical uncertainty, each specific industry needs empirical evidence.

Similar to the above, the empirical research results about the relationship between market competition and investment in the mobile telecommunications industry appear to be ambiguous. There is not much research literature at the industry level. Kang et al. [9] found that a positive correlation exists between market concentration and network investment in China’s mobile telecommunications industry by using regional panel data models during the 2003–2009 period, and the imbalanced market structure increased the total industry investment. Elixmann et al. [10] investigated thirteen mobile telecommunication markets in developed countries without finding a clear relationship between market competition and network investment. There is much more literature at the company level. Kim et al. [11] found that market competition, measured by the entry of mobile virtual network operators (MVNOs), has a negative impact on the incumbents’ network investment. Lestage et al. [12] argued that the impact of market competition on incumbents’ investment depends on whether the incumbents are private (negative effect) or state-owned (positive effect). Houngbonon and Jeanjean [13] found an inverted U-shaped relationship between competition and network investment in the mobile telecommunications industry by using firm-level data. Their further research indicated that the operators’ investment decreased with the increase of the competition intensity (the number of operators increased) in symmetric markets, and this decrease effect was relatively noticeable for non-dominant operators. Although the total industry investment rises (caused by the investment of new entrants) with the increase of competition intensity in the short term, it eventually declines in the long run due to significant adjustment costs of investment in the mobile telecommunications industry [14]. Genakos et al. [15] found that a higher market concentration promotes mobile operators’ network investment. In a word, the impact of competition on investment depends both on precise competition-enhancing measures and the type of investment at stake [16].

The research mentioned above provides valuable clues for revealing the relationship between market competition and network investment in the mobile telecommunications industry. However, the research conclusions are quite divergent, and it is still necessary to further verify the research results and expand the research ideas. First of all, most of the research objectives are the developed countries’ mobile telecommunications markets, and research about developing countries remains nearly nonexistent, given the apparent differences between the two in terms of development stage and institutional environment. Since the 2008 telecommunications restructuring in China, the market structure, telecommunication technology, and structural reforms have all undergone significant changes. It is doubtful whether the research conclusions based on developed markets or China’s provincial data before 2008 are still applicable. Secondly, telecommunications infrastructure sharing significantly impacts investment and regulation policies, as an important global mobile network deployment trend. However, the previous studies rarely consider infrastructure sharing. Finally, due to the difficulty of industry-level data acquisition, the previous studies focus on network investment at the company level. However, the regulators usually pay more attention to the total industry investment rather than to the distribution of the total investment within the industry. In light of the above discussion, this paper considers the characteristics of China’s mobile telecommunications industry, taking into account the impact of infrastructure sharing, and examines the relationship between market competition and industry network investment by using empirical methods.

3. Theoretical Analysis

China’s telecommunications has a unique institutional environment [9]. The main operators are all state-owned enterprises (SOEs), and the government appoints their senior management. The examination of the relationship between market competition and network investment must consider this environment. Lestage et al. [12] pointed out that intensified market competition increased network investment by state-owned operators while reducing network investment by private operators, mainly because these two types of enterprises have different objective functions. The investment decision of the former is based on social welfare maximization, and the investment in telecommunications infrastructure has a significant spillover effect which will increase social welfare. The latter’s investment decision is based on profit maximization, while intensified competition reduces the equilibrium price and firm profit, which will reduce firms’ incentives for investment. However, the investment decisions of state-owned operators are not so simple. As market entities that follow market rules and participate in competition equally, although competition occurs within the same subject of state-owned property rights, the business decisions of China’s state-owned operators are still subject to many realistic market conditions. Therefore, the logical starting point for this paper’s theoretical analysis is to analyze market factors. We should first consider the impact of the telecommunications’ inherent characteristics, and then consider the impact of China’s unique institutional environment.

Whether in the era of 2G voice or in the era of 3/4/5G data traffic, the development history of the mobile telecommunications industry shows that since it provides a basic communication service, network quality is one of the key factors affecting customer satisfaction. Therefore, an important way for operators to gain a competitive advantage is to build high-quality mobile telecommunication networks (wider coverage, more advanced technologies and larger capacity), which can help them take the lead in contending for new customers and retaining existing customers. Moreover, due to the significant scale economies of the mobile telecommunications industry, the high investment of network construction in the early stage will be compensated by the vast customer base brought about by the advantage of mobile network quality at a later stage (since the marginal cost of adding per user is almost negligible), which can form a positive feedback of “fine network quality → successful user development → nice profit based on scale economies → further strengthened investment capacity”. In the 2G era, it is the best example that China Mobile relied on its network quality advantages to crush China Unicom in acquiring customers and establishing its leading position. Valletti and Cambini [17] also pointed out that, due to the scale economies of telecommunications networks, operators’ network quality has a positive impact on their customer market share and a negative impacts on competitors’ customer market share, which is labeled the substitution effect of the network. Therefore, operators should adopt follow-up strategies, as much as possible, to deal with their opponents’ investment behaviors. In addition, unlike building differentiated advantages in operations or services, which requires more innovative efforts, differentiated advantages based on network quality are relatively easier to achieve. To a large extent, operators only need to have a large amount of funds and invest heavily in network construction. Compared with operators in developed countries, preferring service-based competition, China’s operators in the factor/investment-driven growth phase prefer facility-based competition, and their assets seem to be “heavier” (According to a Deloitte report, the number of base stations per 10,000 people in China is 14.1, which is significantly higher than the 8.7 in Europe and 4.7 in the United States. https://www2.deloitte.com/content/dam/Deloitte/us/Documents/technology-media-telecommunications/us-tmt-5g-deployment-imperative.pdf (accessed on 15 November 2021)). Therefore, under the combined influences of the characteristics of mobile telecommunications and the development stage of China’s telecommunications industry, market competition will prompt operators to increase investment. Dominant operators must maintain or strengthen their competitive advantages, while non-dominant operators, must follow the competitors’ strategies and avoid falling too far behind the opponents in the competition. In addition, the previous literature about the negative correlation between competition intensity and network investment in the mobile telecommunications industry usually implicitly assumes that the market is almost wholly covered, which means that the market demand remains basically unchanged [14]. Vives [18] argued that an increase in competition intensity (increase in the number of firms) would produce two effects. One refers to the decline of each firm’s market share (new entrants cut the “cake” of existing firms), which is labeled “demand effect”. The other refers to the expansion of the market size (increased competition lowers prices and increases the market demand, and the “cake” becomes bigger), labeled “price-pressure effect”. When the market is fully covered, demand is insensitive to changes in either prices or market structure. Meanwhile, the price-pressure effect is weak, and the demand effect is dominant. Therefore, intensified market competition has a negative impact on output and investment. The competition intensity in China’s mobile telecommunications market is far below that of other major markets in the world (the statistics from GSMA (Global Association for Mobile Communications Systems) show that at the end of 2020, the Herfindahl–Hirschman Index (calculated by mobile users) in China’s mobile telecommunications market was 0.4320, and China ranks first in the world’s major mobile markets in terms of this index, compared with 0.3522 in Japan, 0.3522 in the United States, 0.2974 in Indonesia, 0.2847 in India, and 0.2738 in the United Kingdom during the same period), and it is still far from the inflection point in the inverted U-shape. In terms of market demand, China’s mobile phone penetration in 2010 reached 64.4 phones per 100 people (2010 is the starting point for the sample data selected by this paper). Compared with the current mobile phone penetration in developed countries, there is still much room for improvement (http://www.gov.cn/govweb/gzdt/2011-01/26/content_1793136.htm (accessed on 20 November 2021)). Therefore, it is unreasonable to give up competing for the market while choosing inactivity for a rational operator, in this case.

The objective function of the SOEs is a linear combination of its profit and the government objective [19]. On the one hand, SOEs must achieve the economic goal of maximizing profit and shareholder value. On the other hand, they also need to act as a public interest defender to maximize social welfare and fulfill their political and social responsibility [20]. Under the requirements of the latter, China’s operators have undertaken some significant policy burdens, including accelerating the deployment of advanced infrastructure, supporting independent technological innovation, upgrading internet speed, and reducing service tariffs. These policy burdens have changed and even aggravated as China’s internal and external development environment continually changes. For example, in the 3G era, the Chinese government appointed China Mobile, the largest telecom operator in China, to build a 3G network with China’s homegrown and immature TD-SCDMA technology. In the 4G era, the Chinese government required the three major operators to upgrade the internet speed while reducing the service tariff, and the state-owned Assets Supervision and Administration Commission of the State Council (SASAC) has put the operators’ implementation of “upgrading internet speed and reducing service tariff” and improving customer service into the performance evaluation of SOEs (to which the SASAC replied: “What have China’s state-owned telecommunication companies done for upgrading internet speed and reducing service fees?” http://www.sasac.gov.cn/n2588040/n2590387/n9854147/c18391392/content.html (accessed on 22 November 2021)). Currently, telecommunications has become the fundamental and leading industry that comprehensively supports economic and social development in China. China hopes that state-owned operators will play a more significant role in promoting industrial upgrading and achieving high-quality development. Since the senior managers of China’s operators are all appointed by the government, the investment and operation decisions of these operators will inevitably be affected by the will of the state, which is mainly reflected in the fact that China’s operators are facing increasing investment pressure. Therefore, state-owned operators tend to increase investment under the combined effects of market competition and country needs. As operators undertake policy burden, the government usually compensates them in other areas or formulates favorable policies to alleviate their operating pressure. For example, the MIIT directly allocates valuable radio frequencies to operators with low fees, which is significantly different from the fact that operators in other countries need to spend vast amounts of money purchasing frequencies. Moreover, the government encourages operators to share telecommunications infrastructure. In the 3G era, the MIIT issued the “Emergency Notice on Promoting Telecommunication Infrastructure sharing” in October 2008 (the document requires operators to share existing towers, and operators must jointly build new towers. It is forbidden to sign an exclusive agreement when renting third-party facilities). In the 4G era, due to the ineffectiveness of the policies mentioned above, China Tower, a state-owned telecommunication company providing telecommunication tower construction, tower maintenance, ancillary facilities management, and other services, was established in July 2014. In the 5G era, with the issuance of four 5G commercial licenses, the MIIT encourages the four operators to share the 5G network. Implementing the policy is to solve the over-investment, improve resource utilization, and ease operators’ investment pressure.

In the light of the above analysis, three hypotheses are proposed here.

Hypothesis 1 (H1).

Market competition can significantly increase the total industry network investment.

Hypothesis 2 (H2).

Infrastructure sharing can significantly reduce the total industry network investment.

Hypothesis 3 (H3).

When market competition and infrastructure sharing work together, the effect of market competition on network investment is relatively smaller than that without infrastructure sharing. (The regression coefficient is positive but less than the regression coefficient value of Hypothesis 1, or the regression coefficient is negative).

4. Model and Data

4.1. Data and Variables

The dataset is comprised of China’s mobile telecommunications market in 31 provinces over the period 2010–2019, giving a total of 310 observations. This period was chosen because of two considerations: The restructuring of China’s telecommunications industry frequently occurred before 2009, and it remained stable after 2009. The three major operators have achieved a nationwide coverage of mobile networks. Besides, the popularization of new technologies represented by 3G and 4G has driven a new round of rapid development of China’s telecommunications industry during this period. The network investment and the number of base stations have increased significantly compared with the previous period (http://www.stats.gov.cn/ztjc/ztfx/ggkf40n/201809/t20180911_1622071.html (accessed on 10 December 2021)), and the infrastructure sharing policies have been continually introduced, which also provides a good case for the research purposes of this paper. The dataset includes variables of mobile network investment, infrastructure sharing, and market competition and variables reflecting the market environment. The data come from the Annual Report of China’s Communication Industry Statistics, China’s Telecommunications Statistical Yearbook, China’s Statistical Yearbook, and the statistical forms of the three major operators.

The main variables are defined as follows:

• Dependent variable: The total industry mobile network investment is denoted by inv. It refers to the investment dedicated to constructing the mobile telecommunication network, including purchasing and constructing tower resources (this construction task will be undertaken by China Tower in 2015 and later. In order to keep the relevant data content consistent, the total industry mobile network investment in 2015 and later used in this paper already includes the investment for tower undertaken by China Tower), base station equipment, antennas, and necessary mobile supporting equipment. These investments form fixed assets.
• Independent variable: The intensity of competition is denoted by comp. The Herfindahl–Hirschman Index (HHI) and Market Concentration Rate (CRn) are the commonly used indicators for measuring market competition. Referring to the relevant literature [9,11,15], and making the empirical results more robust as well, this paper selects the HHI calculated by the mobile service revenue and the CR1 denoted by the market share of the largest operator’s mobile service revenue as the basis for measuring the intensity of competition in China’s mobile telecommunications market. To facilitate understanding, the paper applies 1-hhi and 1-cr1 to measure the intensity of competition, denoting by comp1 and comp2, respectively.
• Control variable: Theoretically, a few exogenous variables, such as market demand and deployment costs, can affect the company’s optimal investment. The mobile network investment of the operators depends on the market demand for mobile services, and the demand in a region is usually related to its population and economic development. In addition, investment is also affected by the cost of network deployment. It is generally believed that the radio frequency resources (given the physical characteristics of radio frequencies, low-frequency bands provide wide coverage, while high-frequency bands provide large capacity. Operators with a large number of continuous frequency resources in different frequency bands can obtain a more efficient combination of frequency bands, bring customers a better experience, and effectively reduce the network deployment cost at the same time. On the contrary, having only low-frequency or high-frequency resources or highly fragmented frequency resources means that providing the same customer experience requires more network investment. Because the allocation and use of radio frequency resources in China are managed uniformly by the MIIT, there is no difference in different provinces. In terms of time, it changes synchronously with the introduction of new technologies (3G or 4G), so this factor is not considered in designing the empirical model of this paper), the geographical distribution of the population, and the infrastructure sharing will affect the cost of network deployment. The selection of control variables in this paper follows the above principles.

The permanent resident population in a region is usually used as an appropriate proxy to measure its overall market demand for mobile services, denoted by pop. In order to reflect the difference in demand for mobile services at different levels of economic development, this paper also selects per capita GDP as a proxy to accurately measure market demand, denoted by gdpc. As mentioned above, the cost of network deployment is another significant factor affecting the investment of the mobile telecommunication network. Some studies choose population density as a proxy for network deployment cost [13,21], because the more dispersed the population geographical distribution (low population density), the more mobile base stations a region needs for covering the same population (the greater the investment), and vice versa. Nevertheless, we believe that population density is not a good proxy for network deployment cost, because, in unpopulated remote areas such as most areas in western China, operators will not build networks in these areas. Mobile connections per square kilometer of network area would be a good proxy for the cost of network deployment—for which, however, data are difficult to obtain. Therefore, drawing on the research of Elixmann et al. [10], this paper uses the share of the rural population as a potentially better proxy for deployment costs than population density. Infrastructure sharing is another critical factor affecting the cost of network deployment. Both theory and practice have proved that infrastructure sharing can effectively reduce network deployment costs [22]. The higher the level of infrastructure sharing, the more network investment can be saved. China’s mobile telecommunications industry began to truly introduce infrastructure sharing since the establishment of China Tower in July 2014. Given that the running of China Tower is a gradual process that first acquires the existing tower resources from the three major operators and then puts these resources into operation (China Tower was established on 18 July 2014, and it completed the acquisition of the tower resources of the three major operators on 14 October 2015). In order to accurately capture the different effects of infrastructure sharing changes in this process, this paper introduces a comprehensive dummy variable shar varying with time to measure the grades of infrastructure sharing. Assign shar to 0, 0.5, and 1, representing 2010–2014, before the establishment of China Tower, 2015, during the establishment of China Tower, and 2016–2019, after the establishment of China Tower, respectively. Finally, in order to explore the combined impact of market competition and infrastructure sharing on network investment, this paper introduces the interaction variable compshar.

Figure 1 shows the theoretical logic of the variables selection in this paper.

Table 1. Variable Definition and Descriptive Statistics.

Variable	Definition	Mean	SD	Min	Max
inv	The total mobile industry network investment, RMB(billion)	5.0024	3.2350	0.5241	20.1590
comp1	1-HHI	0.4554	0.0757	0.2483	0.6409
comp2	1-CR1	0.3021	0.0718	0.1388	0.5436
pop	Number of permanent residents in a region, Million	44.38	28.64	3.00	124.89
gdpc	Local GDP per capita in a region, RMB	49523	26355	12882	164563
rural	Share of the resident population in rural areas in a region	0.4391	0.1340	0.1021	0.7733
shar	The level of infrastructure sharing (0, 0.5, and 1)	0.4500	0.4725	0.0000	1.0000

shows the definition and descriptive statistics of the variables.

4.2. Empirical Methodology

Some previous literature applied panel data models to explore the relationship between market competition and network investment in the telecommunications industry [9,12,13,14,15,23,24]. However, there is no literature introducing infrastructure sharing into the econometric model. To explore the relationship among competition intensity, infrastructure sharing, and network investment in China’s mobile telecommunications industry, the following econometric models of investment are estimated.

$$in{v}_{it}=\alpha +{\beta }_{1}comp{1}_{it}+{\beta }_{2}po{p}_{it}+{\beta }_{3}gdp{c}_{it}+{\beta }_{4}rura{l}_{it}+{\beta }_{5}sha{r}_{it}+u_i+{\lambda }_t+v_{it}$$

(1)

$$in{v}_{it}=\alpha +{\beta }_{1}comp{2}_{it}+{\beta }_{2}po{p}_{it}+{\beta }_{3}gdp{c}_{it}+{\beta }_{4}rura{l}_{it}+{\beta }_{5}sha{r}_{it}+u_i+{\lambda }_t+v_{it}$$

(2)

The dependent variable inv_it denotes the total industry mobile network investment in province i at period t. Both comp1_it and comp2_it are the vectors of market competition variables (see Table 1). pop_it is the permanent resident population. gdpc_it denotes the GDP per capita. rural_it is the share of rural resident population. shar_it denotes the level of telecommunications infrastructure sharing. β_i is the parameter to be estimated. u_i and λ_i are individual-specific and time-specific parameters that represent unobserved explanatory variables, respectively, and v_it is a random disturbance term with zero mean and constant variance that satisfies the classical assumptions of homoscedasticity and has no serial correlation.

In order to examine the interaction of market competition and infrastructure sharing on industry network investment, the following model is used, where both comp1shar and comp2shar denote the interaction of market competition (comp1 and comp2) and infrastructure sharing.

$$in{v}_{it}=\alpha +{\beta }_{1}comp{1}_{it}+{\beta }_{2}po{p}_{it}+{\beta }_{3}gdp{c}_{it}+{\beta }_{4}rura{l}_{it}+{\beta }_{5}sha{r}_{it}+{\beta }_{6}comp1sha{r}_{it}+u_i+{\lambda }_t+v_{it}$$

(3)

$$in{v}_{it}=\alpha +{\beta }_{1}comp{2}_{it}+{\beta }_{2}po{p}_{it}+{\beta }_{3}gdp{c}_{it}+{\beta }_{4}rura{l}_{it}+{\beta }_{5}sha{r}_{it}+{\beta }_{6}comp2sha{r}_{it}+u_i+{\lambda }_t+v_{it}$$

(4)

The historical level of a firm’s investment is likely to influence its investment decision in the next year [23]. Therefore, the model assumes that the amount of mobile investment in year t is likely to be influenced by the amount of investment in year t−1 by using the lagged dependent variable inv_i(t−1) term to represent the total industry mobile investment in province i at period t−1. The following dynamic panel data model is used:

$$in{v}_{it}=\alpha +{\beta }_{0}in{v}_{i(t}{}_{-1)}+{\beta }_{1}comp{1}_{it}+{\beta }_{2}po{p}_{it}+{\beta }_{3}gdp{c}_{it}+{\beta }_{4}rura{l}_{it}+{\beta }_{5}sha{r}_{it}+u_i+{\lambda }_t+v_{it}$$

(5)

$$in{v}_{it}=\alpha +{\beta }_{0}in{v}_{i(t}{}_{-1)}+{\beta }_{1}comp{2}_{it}+{\beta }_{2}po{p}_{it}+{\beta }_{3}gdp{c}_{it}+{\beta }_{4}rura{l}_{it}+{\beta }_{5}sha{r}_{it}+u_i+{\lambda }_t+v_{it}$$

(6)

$$in{v}_{it}=\alpha +{\beta }_{0}in{v}_{i(t}{}_{-1)}+{\beta }_{1}comp{1}_{it}+{\beta }_{2}po{p}_{it}+{\beta }_{3}gdp{c}_{it}+{\beta }_{4}rura{l}_{it}+{\beta }_{5}sha{r}_{it}+{\beta }_{6}comp1sha{r}_{it}+u_i+{\lambda }_t+v_{it}$$

(7)

$$in{v}_{it}=\alpha +{\beta }_{0}in{v}_{i(t}{}_{-1)}+{\beta }_{1}comp{2}_{it}+{\beta }_{2}po{p}_{it}+{\beta }_{3}gdp{c}_{it}+{\beta }_{4}rura{l}_{it}+{\beta }_{5}sha{r}_{it}+{\beta }_{6}comp2sha{r}_{it}+u_i+{\lambda }_t+v_{it}$$

(8)

Estimating the dynamic panel data model using OLS is problematic due to the inclusion of the lagged dependent variable as a regressor. The result of the OLS estimate will be biased and inconsistent because of the individual-specific unobserved effects, u_i. Although the first difference method can be used to remove the individual effect u_i, the latent heterogeneity estimation of the equation is still biased because of a correlation between the new error term and the differenced lagged-dependent variable [25]. Therefore, a generalized method of moments (GMM) estimator developed by Blundell and Bond [26] following the work of Arellano and Bover [27] is used to deal with this problem. This method is also known as System GMM. The method applies a different number of instruments in each time period and uses a large set of moment conditions to provide efficient and consistent parameter estimates in a wide variety of settings. This estimator is designed for the T < N short dynamic panel dataset (just like the dataset used in this paper). Since this method uses many instruments, the Sargan test for over-identifying restrictions is used [25].

5. Empirical Results

5.1. Estimation Results

In view of the powerful data processing and analysis functions of Stata, all estimates in this paper were calculated using Stata 14.

There are two types of approaches for static panel data models variation, namely fixed effects (FE) and random effects (RE). The difference between the two lies mainly in the assumptions made about the individual specific effect. The FE model assumes that differences between individuals can be accommodated from different intercepts. In contrast, the RE model assumes that the difference between intercepts is accommodated by the error terms of each company. Referring to the previous research, we use the fixed effect (FE) regression models to examine the relationship among market competition, infrastructure sharing, and network investment. We also use Hausman’s specification test to determine whether a fixed or random effects model is most appropriate in this case. The main regression results are listed in

Table 2. Effects of market competition variables (comp1 and comp2) on network investment (inv).

	Model 1		Model 2
	Coeff.	s.e.	Coeff.	s.e.
comp1	3.406118 **	1.663192
comp2			0.8951967 **	1.531084
pop	0.181086 ***	0.049695	0.174723 ***	0.042369
gdpc	−0.0000051	0.0000113	0.0000001	0.0000108
rural	2.770647	3.560028	0.07595019	3.322116
shar	0.6145001 ***	0.1657857	0.5855355 ***	0.1579159
_cons.	−5.857775 *	3.417628	−3.28757	2.818567
R2	0.1986		0.1931
F	14.23		17.48
Hausman Test Prob.	0.0060		0.0000

Note: *** Implies significant at 1%, ** Implies significant at 5%, and * Implies significant at 10%.

,

Table 3. Effects of market competition and infrastructure sharing’s interactions (comp1shar and comp2shar) on network investment (inv).

	Model 3		Model 4
	Coeff.	s.e.	Coeff.	s.e.
comp1	6.709538 ***	1.691606
comp2			7.954982 ***	20.34562
pop	0.148551 ***	0.048317	0.1453014 ***	0.0046156
gdpc	0.0000011	0.0000105	0.00000118	0.0000102
rural	5.385133 *	3.130149	5.126497 *	3.017521
shar	4.639881 ***	1.442006	3.712727 ***	0.9430873
comp1shar	−8.527496 ***	2.848072
comp2shar			−9.92853 ***	2.654481
_cons.	−7.263797 **	3.166152	−6.317964 **	2.869408
R2	0.2422		0.2455
F	13.04		13.5
Hausman Test Prob.	0.0000		0.000

Note: *** Implies significant at 1%, ** Implies significant at 5%, and * Implies significant at 10%.

,

Table 4. System GMM dynamic panel estimation of market competition (comp1 and comp2) on network investment (inv).

	Model 5		Model 6
	Coeff.	The Windmeijer Corrected Standard Error	Coeff.	The Windmeijer Corrected Standard Error
invi(t−1)	0.0570776	0.1284947	0.0505953	0.1292665
comp1	4.691227 *	2.827366
comp2			2.974024 *	2.822676
pop	0.1154193 ***	0.0244542	0.115887 ***	0.0241804
gdpc	−0.00000514	0.00001662	−0.0000236	0.0000155
rural	−8.559383 *	5.14787	−7.633909 *	4.429021
shar	0.5844562 *	0.333991	0.5829143 *	0.3291368
_cons.	5.527128	3.556668	4.359806	2.846509
Wald test	Chi2(6) = 497.64 prob. = 0.0000		Chi2(6) = 511.57 prob. = 0.0000
Sargan test	Chi2(22) = 26.06636 prob. = 0.2488		Chi2(22) = 26.03619, prob. = 0.2501

Note: *** Implies significant at 1%, * Implies significant at 10%.

and

Table 5. System GMM dynamic panel estimation of market competition and infrastructure sharing’s interactions (comp1shar and comp2shar) on network investment (inv).

	Model 7		Model 8
	Coeff.	The Windmeijer Corrected Standard Error	Coeff.	The Windmeijer Corrected Standard Error
invi(t−1)	−0.1186118	0.1454147	−0.0814001	0.1295155
comp1	1.653766	3.387184
comp2			2.980198	3.422514
pop	0.1335804 ***	0.026341	0.1290192 ***	0.025263
gdpc	0.0000129	0.000015	9.08E-06	0.0000156
rural	−3.012117	4.888187	−3.092439	4.837704
shar	8.594917 **	3.760856	5.164272 **	2.101923
comp1shar	−16.95718 **	7.341892
comp2shar			−14.78144 **	5.96281
_cons.	−0.3669619	3.086185	−0.2407717	2.815169
Wald test	Chi2(7) = 279.44 prob. = 0.0000		Chi2(7) = 348.64 prob. = 0.0000
Sargan test	Chi2(22) = 24.02996 prob. = 0.3457		Chi2(22) = 26.03619 prob. = 0.2501

Note: *** Implies significant at 1%, ** Implies significant at 5%.

. All the four static models have passed the Hausman test at a 1% significance level, proving that our selections of FE models are appropriate. All the Sargan tests for dynamic models are insignificant at any significance level, and the null hypothesis is accepted. Over-identifying restrictions are valid.

We can see from Table 2 and Table 4 that in both static and dynamic models the competition intensity variable comp1 is positively related to the total industry network investment at a 10% significance level. Specifically, the regression coefficients are 3.406118 and 4.691227, respectively—namely, an increase in market competition intensity by one unit induces a rise in total industry network investment of 340% and 469%. The other competition intensity variable comp2 is also positively related to network investment at a 10% significance level, and the regression coefficients are 0.8951967 and 2.974024, respectively. Therefore, Hypothesis 1 proposed in this paper is proved. The more balanced the mobile telecommunications market structure, the higher the total industry network investment (there is a possibility of an inverted U-shaped relationship between market competition and network investment in the mobile telecommunication industry. Therefore, we added the squared terms of the independent variables comp1 and comp2 into the regression equations (Equations (1) and (2)). The estimation results show that the coefficient is negative, but failed to pass the significance test. To simplify the presentation of the results, they are not listed in the table). The mobile telecommunications industry has the typical characteristics of scale economies and network effects [28]. Only when operators have a particular market share can they afford essential network deployment and expansion and benefit from economies of scale. The highly concentrated market structure is not conducive to increasing industry network investment. Under a highly unbalanced market structure, underinvestment is more likely to occur. In addition, contrary to our expectations, in almost all models, the regression coefficient sign of the infrastructure sharing level variable shar is positive and significant (at a 5% level), which means that telecommunication infrastructure sharing promotes the total industry network investment. There is an explanation for this abnormal result. Telecommunication towers are precious resources, especially towers with large numbers and good geographical locations. Operators have been facing the deteriorating problem of “difficulty in obtaining mobile towers” in China. For non-dominant operators, this problem is even more serious, which has seriously affected operators’ network construction (in June 2014, China’s CaiJing Magazine published the article “Telecom Reform Starts Again”, which revealed that constructing mobile towers in core urban areas will not only entail the payment of high fees to the property management company, but may also be strongly resisted by residents. In 2013, China Mobile planned to build 200,000 4G base stations, of which only 60,000 have been completed. The lack of site (tower) is one of the main reasons. http://finance.sina.com.cn/chanjing/sdbd/20140603/133119299533.shtml (accessed on 25 December 2021)). The establishment of China Tower is equivalent to a one-time enlargement of the operators’ tower resources, permanently enhancing the market position of tower operators to upstream property management companies (because China Tower has the exclusive monopoly right to construct and operate towers, it has formed a complete monopoly in the relevant market). At the same time, it also narrows the gap in tower resources among operators and strengthens effective competition among operators [29], which stimulates operators’ mobile network investment. It can be seen from Table 3 and Table 5 that in the static model, the interaction terms of the competition intensity and the infrastructure sharing level, comp1shar and comp2shar, are both negatively related with network investment at a 1% significance level. The dynamic model is negatively related to network investment at a 5% significance level, which proves the Hypothesis 3 proposed in this paper—when market competition and infrastructure sharing work together, the positive effect of market competition on network investment is relatively smaller than that of the case without infrastructure sharing. Given that infrastructure sharing promoted industry network investment, this paper believes that a possible explanation is as follows: increasing fierce market competition will not stimulate network investment once the infrastructure sharing level exceeds a certain threshold. This is mainly because the network quality gap between operators has become smaller after the introduction of infrastructure sharing. Operators’ strategic intentions, building a competitive advantage through constructing high-quality mobile networks, face the challenge of opening network resources under infrastructure sharing requirements, which means operators will encounter the “free rider” problem. In this case, operators are more inclined to shift their competition patterns from facility-based competition to service-based competition, causing the industry to face insufficient incentives for network investment. In addition, it can be seen from the empirical results of the dynamic model in Table 4 and Table 5 that there is no correlation between the lagged network investment invi_(t−1) and the current network investment, which means that the network investment last year had no impact on network investment this year, and there is no inertial effect in the mobile industry’s investment.

Table 2, Table 3, Table 4 and Table 5 also report that the environmental control variable pop is positively related to network investment at a 1% significance level (although the coefficient is minimal), while gdpc does not have a statistically significant impact on network investment. It is consistent with the finding of Elixmann et al. [10]: although the GDP per capita ranks relatively low, South Korea’s mobile network investment level is significantly higher than that of the United States, Germany, Japan, and other wealthy countries. A high GDP per capita is not necessary for the high level of mobile industry investment. The estimate for the environmental control variable rural is differentiated. The static models (3) and (4) show that rural is positively related to network investment at a 1% significance level, while other models have no apparent correlation. The estimates for these control variables are not consistent with the usual expectations. A significant reason is that the mobile network deployment model and investment model are becoming complicated. For example, in the early days of mobile telecommunications (such as in the 2G era), operators mainly deployed mobile networks with low frequencies, which provided a wide area coverage with fewer mobile base stations. Since the geographical distribution of population in rural areas is more dispersed than that in urban ones, to cover the same population by mobile base stations, the expenditure for building mobile networks in rural areas is higher than that in urban ones if all other things are equal. However, as the demand for mobile services increases continually, market competition in critical areas (densely populated areas) becomes increasingly fierce, mobile telecommunication technology evolves continually (an important feature is that high frequencies are widely used), and the number of mobile base stations (network investment) in urban areas rise continuously. Therefore, the impact of the density economy on network investment in the mobile telecommunications industry is becoming blurred and complicated.

5.2. Robustness Tests

To verify the reliability of the conclusions, this paper conducts robustness tests by the following three ways: (i) replacing the explanatory variable with another variable. We use CR1 to replace HHI in the regression equation (see Section 4.1), and the estimation results do not change our conclusions; (ii) For the problem of omitted variables, the fixed-effects model can be a good solution, so we use the fixed-effects model to estimate Equations (1)–(4); (iii) For the potential endogeneity of our econometric models, namely, there may be a two-way causality between market competition and network investment. First, the development history of China’s telecommunications industry tells us that the top-down structural reforms implemented by China’s regulators are the most important factor affecting market concentration changes [30]. To a great extent, it can be considered that the changes of competition intensity in China’s mobile telecommunication market are exogenous, and the impact of the total industry network investment on market competition can be ignored. Second, system GMM estimations are often used to address the issue of endogeneity in the absence of appropriate external instruments. This paper applies the system GMM method to estimate Equations (5)–(8). the Sargan tests do not reject the null hypothesis that the instrumental variables are valid.

6. Discussions and Policy Recommendations

Narrowing the digital divide and fostering the digital economy are generally considered to be the keys to increasing productivity and promoting economic growth in all countries. In order to achieve these goals, investing in telecommunications is essential [31]. The outline of China’s Fourth Five-year Plan proposed to “Systematically deploy new infrastructure, accelerate the construction of 5G network, industrial Internet, and big data center. Develop the digital economy, promote industrial digitization and digital industrialization, promote the deep integration of the digital economy and the real economy, and create an internationally competitive digital industry cluster” (“Proposal of the Central Committee of the Communist Party of China on Formulating the Fourteenth Five-Year Plan for National Economic and Social Development and the Long-term Goals for 2035”, http://www.gov.cn/zhengce/2020-11/03/content_5556991.htm (accessed on 30 December 2021)).

Given the importance of network investment in practice and the ambiguity of existing conclusions in theoretical research, this paper uses an econometric model based on the panel data of 31 provinces in China in the period 2010–2019 to empirically analyze the relationship among the competition intensity, the level of infrastructure sharing, and the network investment in China’s mobile telecommunications. The estimation results show that both market competition and infrastructure sharing individually have a significant positive impact on the industry’s network investment, while their interaction has a significant negative impact on network investment. One possible explanation is that when the higher level of infrastructure sharing and more fierce market competition appear at the same time, the competitive strategy of rational operators will change, which is about to shift from Facility-based competition to Service-based competition. The results of this study also indicate that to a certain extent, despite facing policy burdens and other issues, as long as certain internal and external conditions are met, such as no service price control, a relatively complete corporate governance structure, and competitive market environment, the investment behavior of China’ telecommunications operators will still be rational. The return on investment determines, to a large extent, the investment behavior of operators.

The conclusions of this paper are different from that of Kang et al. [9], based on the panel data of China’s mobile telecommunications from 2003 to 2009. The different development stages of the industry might explain this. In the early stages of development (2008 and before), the market structure of China’s mobile telecommunications was highly concentrated, and the investment from dominant operators accounted for a large proportion of industry investment, making the changes of industry investment consistent with those of dominant operators. In other words, the market concentration positively correlated with network investment in mobile telecommunication (see

Table 6. Comparison of the mobile telecommunication market share of China’ three major operators in 2007 and 2019.

	2007		2019
	Mobile Revenue Share	Network Investment Share	Mobile Revenue Share	Network Investment Share
China Mobile	75%	86%	60%	56%
China Unicom	25%	14%	20%	20%
China Telecom	0	0	20%	24%

). However, when the industry entered the mature stage, primarily when competition occurred among market players with relatively balanced strength, market competition stimulated operators’ network investment. The market concentration (or competition intensity) is negatively (or positively) correlated with network investment. During the coverage period of this paper, the market structure of China’s mobile telecommunications improved continuously, and the HHI decreased continuously (see Scheme 1). Intensified market competition reduced the price of telecommunication services, improved customer service, and improved network quality (mainly by increasing investment).

The conclusions of this paper have important policy implications for the telecommunications industry in deepening structural reforms and promoting high-quality development. First of all, we urge policy makers to adhere to the reform ideas of improving market structure and enhancing market contestability. The leapfrog development of China’s mobile telecommunication in the past three decades has fully proved the correctness of the competition-focused reform policy. It should be noted that the market structure of China’s mobile telecommunications still has room for improvement compared with other developed countries or its status quo. The introduction of CBN in the 5G era can accelerate the maturity of the 5G industrial chain, effectively respond to international technology competition, and further promote market competition and innovation, continuing to promote the sustainable and healthy development of China’s mobile telecommunications along with the logic of “competition enhancement → price reduction → innovation increase → market expansion”.

Secondly, we maintain that policy makers continue to bolster infrastructure sharing, but exercise caution when conducting in-depth infrastructure sharing (both degree and scope). By the degree of sharing, the network sharing of mobile telecommunication infrastructure can be divided into site/tower sharing (passive sharing), spectrum/access network/core network sharing (active sharing), and roaming sharing. The degree of sharing is gradually deepening. The sharing scope can be divided into two operators sharing one network, three operators sharing one network, and all operators of the industry sharing a network, which is the biggest scope. Although in-depth infrastructure sharing can further save network investment and improve network quality, it will also damage market competition and service differentiation, reduce operators’ willingness to invest in the advanced telecommunications network and develop new products and services. In extreme cases, there is only one network of one operator in a country, and this is the highest level of infrastructure sharing (both degree and scope), which can save investment to the greatest extent. However, there will be substantial efficiency losses because of monopoly in this case. The history of telecommunications has proven that the loss of efficiency is greater than the benefit of saving investment. Therefore, there is an optimal level of infrastructure sharing in theory. In this case, considerable investment savings and network quality improvement can be obtained, and the weakness of infrastructure sharing that weakens investment and innovation incentives can be controlled to a low level. In the 5G technology cycle, the Chinese government encourages four state-owned operators to share the 5G network. Compared with the infrastructure sharing of site or tower in the 4G technology cycle, the degree of infrastructure sharing is deepening, which effectively reduces the operators’ investment pressure. However, China’s regulators should take a cautious stance toward major operators’ requests for the expansion of the degree and scope of network sharing (for instance, instead of two operators sharing one mobile network, they may want to have three operators or even the entire industry share one mobile network; or in lieu of sharing mobile access network, they may ask for a sharing mobile core network and even sharing service-level resources).

Thirdly, we also recommend the government to accelerate the development of the 5G vertical industry applications and expand the market space for industry development. China’s mobile telecommunications has entered a stage of maturity, and there is no longer a high growth rate based on a “demographic dividend” and a “reform dividend” (according to the “Telecommunications Industry Statistics Bulletin 2020” issued by the MIIT, by the end of 2020, the number of mobile phone users in China had reached 1.594 billion, and the mobile phone penetration rate was 113.9 phones per 100 people. http://www.gov.cn/xinwen/2021-01/26/content_5582523.htm (accessed on 30 December 2021)). The development of mobile telecommunications in developed countries in recent years shows that when the market demand growth is slow, the return on investment in new technologies is uncertain, and when the government strictly regulates the telecommunications industry, operators generally lack investment motivation (https://www.mckinsey.com/~/media/mckinsey/dotcom/client_service/telecoms/pdfs/05_a%20new%20deal_driving_investment_in_europe_telecoms_infrastructure.ashx (accessed on 30 December 2021)). At this time, market competition cannot stimulate or even inhibit the operators’ investment motivation. Under the state-owned property arrangement system, it is also easy to induce soft budget constraints. The broad market space brings tremendous confidence for operators to increase network investment and innovate continuously. Therefore, the government, enterprises, universities, and research institutes should cooperate closely to promote the comprehensive and coordinated development of 5G jointly, thoroughly carry forward the 5G empowerment of countless businesses, spur the formation of a high-level development model in which “demand pulls supply, and supply creates demand” and strengthens China’s digital economy.