Presentation on theme: "Econometric evidence on the impacts of privatization, new entry, and separate regulatory authority on the mobile network expansion Yan Li University of."— Presentation transcript:
Econometric evidence on the impacts of privatization, new entry, and separate regulatory authority on the mobile network expansion Yan Li University of East Anglia 11 th /June/2008
2 Structure of Presentation Theoretical Background & Existing Studies Research Questions & Hypotheses Methodology Econometric Models & Test Results Conclusion
3 Privatization Historical discussions on theory of privatization: Vickers and Yarrow, 1991; Laffont and Tirole, 1993; Levy and Spiller (1996); Shleifer, 1998; Sheshinski and López-Calva, 2003; Summaries: Vickers and Yarrow (1988); Galal et al. (1994); Adam et al. (1992); Empirical works : Meggingson, et al. (1994) Oniki et al. (1994), Ramamurti (1996), Ros (1999), Boylaud and nicoletti (2001), Li and Xu (2002, 2004); Wallsten (2001, 2004). Competition Theoretical discussions: Hayek (1945); Leibenstein (1966); Matta (2004); Empirical studies: Wallsten (2001), McNary (2001), Fink et al. (2001) and Li and Xu (2002, 2004); Ros (1999); Empirical works on telecoms diffusion: Burki and Aslam (2000); Gruber (2001); Gruber and Verboten (2001); Gebreabs (2002); Koski and Kretschmer (2005). Separate regulatory authority Political institutions (regulatory governance) for economic performance Levy and Spillers (1994); Fink et al. (2001); Wallsten (2001, 2002); Gutierrez and Berg (2000); Ros (2003), Guttierez (2003a, b); Gual and Trillas (2003); Edwards and Waverman (2006); Maiorano and Stern (2007). Theoretical Background
4 Only a limited number of studies conducted econometric analyses; moreover, they offered mixed results on the reform effects (see, Ros, 1999; Wallsten, 2001; Fink et al., 2001; McNary, 2001; and Li and Xu, 2002; 2004). Especially, there is little econometric evidence on the effect of a separate regulator, per se. (Wallsten (2001) treads a separate regulator as a proxy of countrys propensity to reform, but not tests the effect of itself). Obviously, the gap of empirical study in general, and econometric analysis in particular, exists mainly because that regulatory reforms in mobile network markets started relatively recently, which means that it is premature to observe the thorough effects of regulatory reforms with the data used in earlier empirical studies (Most of earlier empirical works used data from mid-1980s to late-1990s when some mobile markets just began their reform or even just show an ambiguous propensity to reform). Existing Studies
5 Research Questions RQ1. Is privatization a necessary and sufficient condition for sustainable and effective market dynamics? RQ1. Is privatization a necessary and sufficient condition for sustainable and effective market dynamics? RQ2. How does new entry into a market matter to the effects of competition on the market performance and efficiency? RQ2. How does new entry into a market matter to the effects of competition on the market performance and efficiency? RQ3. Is a separate regulatory authority* essential to effective market dynamics? RQ3. Is a separate regulatory authority* essential to effective market dynamics? * The separate regulatory authority, in my study, must be created backed by legislation rather than executive decree, and claims to be independent in its decisions making.
6 Research Hypotheses RQ1: Hypothesis 1: The mobile penetration will increase, if the mobile incumbents are privately owned. Hypothesis 2: The interaction of entry* and privatization will enhance the increase in mobile penetration; the faster growth in mobile penetration will be observed when private-owned mobile incumbents operate in competitive market environments than that observed under either one situation.RQ2: Hypothesis 3: The growth of mobile penetration will be enhanced by each new entry into a market.RQ3: Hypothesis 4: The mobile penetration will be higher in a situation of having a separate regulatory authority than that in the situation of without one. Hypothesis 5: The interaction of a separate regulator and privatization will positively affect mobile penetration; better growth in mobile penetration will be obtained if there is a separate regulator in privatized markets than if there is no one. Hypothesis 6: The interaction of a separate regulator and entry* will positively impact mobile penetration; faster growth in mobile penetration will be observed if a separate regulator is created to monitor the market competition than if there is no one. * Entry is measured by number-of-entry into a market. And, more specific and precise effects of each entry can be estimated by setting number-of-entry dummies.
7 Methodology The relationship between the regulatory reform and the mobile penetration Econometric approach to a panel dataset for 30 national mobile markets (including China and other 29 OECD countries) over period from 1991 to 2006 (total 480 observations, consisting of three balanced datasets, namely operating performance dataset, regulatory information dataset, and selected national macroeconomic indicator dataset, and involving 24 variables); An general econometric model developed based on the logistic model of technology diffusion; Given considerations of potential endogenous problems, a econometric 3-equation system is further developed, which is approached by panel fixed-effect instrumental estimates under 2SLS technique; In addition, the model is also approached by panel equation-by- equation estimates and panel dynamic estimates (Arellano-Bond (1991) dynamic panel estimation).
8 Data (1) Table 1: Summary of Statistics (mean) Mobile penetration Fixed-line penetration Mobile subscribers Mobile prices Number of mobile Privatized A separate Relevant Law privatization & year (in per 100 inhabitants) (in per 100 inhabitants) served per mobile (3-mins local call operators regulator and regulations separate regulator Freq. staff in USD$) created passed 19911.6639.214.901.36117%20%57%10% 30 19922.1140.406.741.44120%20%60%10% 30 19932.8841.609.831.38123%23%67%10% 30 19944.4942.8415.491.40227%27%70%10% 30 19957.1444.1024.791.42233%30%70%10% 30 199611.0345.5357.551.45237%30%77%13% 30 199715.9147.3876.751.31250%37%80%23% 30 199824.2248.27122.691.22260%60%90%37% 30 199937.5749.79171.781.04383%67%90%57% 30 200052.8750.83254.610.94383%70%97%60% 30 200163.8550.47287.940.77387%80%100%70% 30 200270.0950.09333.890.89387%83%100%73% 30 200376.1549.30381.960.99387%87%100%77% 30 200483.7248.81435.220.98487%87%100%77% 30 200590.3047.34471.061.00487%87%100%77% 30 200696.7946.87511.051.00487%87%100%77% 30 Total40.0546.43197.891.16360%56%85%43% 480
9 Data (2) P: the mobile market is characterized by private monopoly, without a separate regulator; C: the mobile market is characterized by public ownership and operates in a competitive environment, without a separate regulator; PS: the mobile market is characterized by private monopoly, with a separate regulator; CS: the mobile market is characterized by public ownership and operates in a competitive environment, with a separate regulator; CP: the mobile market is characterized by private ownership and operates in a competitive environment, without a separate regulator; CPS: the mobile market is characterized by private ownership and operates in a competitive environment, with a separate regulator. Graph C-1: Average mobile penetration by regulatory practicesGraph C-2: The relationship between the mobile penetration and the number-of-entry
10 Privatization dummy (prvt) equals one if at least 50% of assets are held by private sector; and equals zero, if the mobile incumbent is state-owned or, the share held of private assets do not reach 50%. It is recorded as one beginning the following year the privatization was announced by government. New entry is measured by the number of mobile incumbents in a market (NoF) that will be operated in two ways (separately) in the model. Any entries will be recorded one year after commencing their services in the market, which is the year effective competition was observed. First, the quadratic form of NoF is used to explore the non-linear relationship between number-of-entry and mobile penetration. Second, dummies of the number-of-entry are set to explore more precise effects of each entry on the mobile penetration (using monopoly as base dummy). For instance, dumNoF2 equals one if a second mobile operator enter a market, and equals zero, otherwise; dumNoF3 equals one if a third mobile operator enter a market, and equals zero, otherwise; and so on. Separate regulator dummy (sprtrg) equals one if the following two necessary conditions are held simultaneously – (1) the created separate regulatory agency is backed by legislation and (2) claims to be independent from the political power, or says, to make its decisions without being subject to any other government bodies; equals zero if any one of the two conditions fails. This variable, again, will follow the rule that being one starting one year after the eligible separate regulator was created. Regulatory Variables
11 Econometric Models (1) To start with Griliches (1957) logistic growth model: Considering the mobile diffusion in the country i at the time t follow the logistic growth curve, and then obtain: After a natural logarithmic transformation and re-arrangement, I obtain
12 Econometric Models (2) By expanding with more detailed specification, the general econometric model for my hypotheses tests is produced as follows: where, is a k × 1 vector of regulatory variables in the country i at the time t, involving three dimensions, namely privatization dummy, entry dummies and separate regulator dummy; is a j × 1 vector of exogenous control variables in the country i at the time t, which includes GDP per capita, fixed-line penetration, fixed-line price of 3-minute local call, urban population rate, total national population, total telecommunications investments rate and time trend; is an m × 1 vector of endogenous variables in the country i at the time t, including the mobile price and the mobile labour productivity; is countries individual specific effects; is country individual specific effect error, and is normal model error term with white-noise.
13 Econometric Models (3) Without endogenous explanatory variables: With endogenous explanatory variables – structural model: * new_entry is measured by number-of-entry that will be operated in two ways – i.e. quadratic form and dummies, separately; number-of-entry dummies are used in the structural model to test more specific effects of each entry.
14 Test Results (1) Eq. (1): Pure effects Eq. (2): With other exogenous control variables Eq. (3) With endogenous variables VariableModel 1Model 2Model 2' Model 3Model 3' Model 4Model 5 lnGDPpc1.9434***1.8041***1.6787***0.03780.5975***1.4049***0.5405 4.073.8919.380.064.552.631.06 NoF0.7977*** 2.67 sqNoF-0.0790* -1.29 prvt-0.12810.5466***0.1738*0.18060.1944**0.09010.0213 -0.304.561.541.251.800.700.23 sprtrg0.7834**0.3283***0.3973***0.2278*0.3400***0.09260.0679 1.812.773.991.552.900.710.70 itcSP0.4351**0.4005**0.11720.3548**0.4241*** 2.002.240.872.233.50 itcNoFS-0.4316* -1.33 itcsqNoFS0.0559 1.01 itcNoFP0.4479* 1.35 itcsqNoFP-0.0985** -1.68 dumNoF20.4765***0.3718***0.5962***0.5977***0.3591***0.1919** 3.872.765.075.633.321.83 dumNoF31.0231***0.8150***1.1265***1.0068***0.7066***0.5768*** 5.934.617.017.084.773.84 dumNoF40.6041***0.3650**0.8153***0.9571***0.5937***0.3370** 2.801.903.885.913.162.24 dumNoF50.7483***0.4601**1.0582***1.1085***0.9172***0.7780*** 3.071.954.595.624.484.60 dumNoF6-0.4383-1.1995***0.24650.36310.18810.1518 -1.11-3.350.671.130.580.47 dumNoF7-0.5623*-1.2056***0.10080.1611-0.0309-0.1630 -1.30-4.340.250.51-0.09-0.57 lnFxpn0.8637***0.8328***-0.3912**-0.1211 4.027.66-1.73-0.32 lnFxp_ee7.0406***1.3573**5.8863***3.2176*** 4.432.054.112.47 lnUbpr1.0492-0.9704***2.8483**2.3552* 0.55-4.721.671.45 lnPop-9.3651***-0.2147***-7.4927***-4.7131*** -5.44-7.99-4.87-3.58 lnMbp_ee-0.2506*-0.4475** -1.62-2.01 lnMSps0.5011***0.6651*** 10.104.25 time trend0.3732***0.3770***0.4153***0.4675***0.3938***0.2609***0.1977*** 12.7913.5030.5515.1835.787.752.96 n = 447n = 447n = 447 n = 447n = 447 n = 447n = 383 R-sq = 0.9324R-sq = 0.9338Wald Chi2(10) = 4177.68R-sq = 0.9468Wald Chi2(15) = 7412.59 R-sq = 0.9583 R-sq = 0.9759 Log likelihood = -465.385 Het-corrected Std.Err.No Fixed-effectNo Fixed-effectYes No Fixed-effectYes GLS No Fixed-effectNo Fixed-effect Dependent variable: y = ln(Mbpn/(100-Mbpn)) Table 2: estimation results by panel equation-by-equation and panel instrumental approaches
15 Test Results (2) Table 3: regulatory effects on mobile price and labour productivity Variable Eq. (4): Mobile price Eq. (5): Mobile labour productivity lnGDPpc0.2240** 0.9035* 2.20 1.58 lnL1Mbp_ee0.5298*** 12.86 y-0.0393*** -3.79 lnCpu0.0343** 1.76 NoF-0.0290*** -2.42 prvt-0.0152 0.0206 -0.54 0.17 sprtrg0.0128 0.3493*** 0.44 2.94 dumNoF2 0.5319*** 4.48 dumNoF3 0.4902*** 2.84 dumNoF4 0.1115 0.52 dumNoF5 -0.1174 -0.45 dumNoF6 0.0415 0.08 dumNoF7 -0.7782** -1.81 lnUbpr 4.7258*** 2.52 lnPop -5.3109*** -2.74 lnTIr 0.4311*** 4.89 time trend 0.3785*** 12.20 n = 383 n = 386 R-sq = 0.5159 (within) R-sq = 0.8971 (within) Fixed-effect Fixed-effect T-statistics / z-statistics are reported below each coefficient in italic type. ***significant at 1%; **significant at 5%; *significant at 10%.
16 Test Results (3) Table 4: panel dynamic estimation results Dependent variable: D.y Short-run effects zLong-run effects LD.y0.7507***12.45---- D.lnGDPpc-0.1550-0.22-0.6219 D.dumNoF20.2598***3.561.0420*** D.dumNoF30.3716***3.601.4905*** D.dumNoF40.2237**1.660.8973** D.dumNoF50.17520.890.7028 D.dumNoF60.02600.100.1042 D.dumNoF7-0.0573-0.16-0.2298 D.prvt-0.2320*-1.96-0.9307* D.sprtrg-0.0038-0.04-0.0154 D.itcSP0.1464*1.550.5872* D.itcNoFP0.0638*1.500.2558* D.itcNoFS0.04010.960.1608 D.lnFxpn0.22590.730.9063 D.lnFxp_ee-0.7374-0.57-2.9580 D.lnUbpr-0.6570-0.24-2.6353 D.lnTIr0.0727*1.440.2918* D.lnPop-2.3380-1.01-9.3783 D.year0.1243***2.440.4985*** n = 321 Wald chi2(19) = 29877.25 Robust Std. Err.Yes Arellano-Bond test that average autocovariance in residuals of order 1 is 0: H0: no autocorrelation z = -2.48 Pr > z = 0.0132 Arellano-Bond test that average autocovariance in residuals of order 2 is 0: H0: no autocorrelation z = -2.33 Pr > z = 0.0197 T-statistics / z-statistics are reported below each coefficient in italic type. ***significant at 1%; **significant at 5%; *significant at 10%. The long-run effects are obtained by removing subscript t from the model, and then, divide each coefficient by (1-0.7507).
17 First, the effect of privatization on network expansion is less important than that of new entry and of a separate regulator in most cases. Especially, dynamic model estimates imply that the change in the ownership of mobile incumbents from public to private does not benefit the mobile network expansion, in both the short-run and the long-run, unless there has been (is) a separate regulator in the mobile network sector and/or the market operates in a competitive environment. Conclusion (1)
18 Second, as long as there are sufficient number of entries into a mobile market, sustainable and more rapid expansion can be achieved, even without privatization taking place. More specifically, the effect of new entry on mobile penetration follows an inversed U-shape as the number of entry increases, and a three-to-five-firm oligopolistic market structure is associated with the highest mobile penetration and expansion. Conclusion (2)
19 Conclusion (3) Third, a separate regulator is essential, in particular, in a market characterized by private ownership in mobile incumbents. And, its effects (e.g. pro-competition) on promoting mobile network expansion are particularly outstanding when interacting with private entries. Finally, the mobile services price and labour productivity partially mediate the effects of regulatory reforms, in particular, of new entry and a separate regulator, suggesting that the impacts of those reforms on mobile network expansion are partially through their effects on price reduction and labour productivity improvements.
20 Policy implications Building up a formal independent regulatory authority for individual sectors is increasingly necessary, as the industrial privatization has been becoming popular around the world. It should be noticed that privatization alone can no longer bring benefits to vigorous market dynamics, since its comparative advantages of timely financial decisions and efficient resources allocations motivated by profit-maximization have been more than offset by the benefits from the sufficient and effective market competitions (that should be the eternal power for vigorous market dynamics). Therefore, it is imperative for those countries that have not fully privatized their markets to take urgent action to foster an effective competitive environment by encouraging new entries into the market rather than rush into privatization alone. More importantly, only an appropriate competitive level (neither inadequate nor over-competitive) is beneficial to economic performance and efficiency. Therefore, we strongly recommend that an apposite range of the number of entries into a market should be careful regulated by individual competition policy across countries, given their different and specific economic and political conditions. Conclusion (4)