We consider two models to analyze the impact of regulatory policy on the prices of mobile calls in Africa. The two models differ because the first aggregates the different regulation components into a single variable while the second disaggregates them. This approach is consistent with the work of Katz and Jung (2021).
Several tests were carried out within the different models. Hausman's specification test (1978) leads us to retain fixed effects models. The Harris and Tzavalis test (1999) allows us to check and correct the stationarity of the variables. Finally, we opted for the weighted Ridge approach to consider the variables' multicollinearity and the residuals' autocorrelation. Fisher's validity test globally validates the model. The results of the two models are summarized in Table 3.
4.1. Analysis of the Regulation-Price Relationship
The results from Model 1 reveal that regulation has a negative and significant impact on all call price variables, except for weekend calls (P4), for which the coefficient is insignificant. An increase of 1000 in the regulation score is therefore associated with a reduction of just over 1.4 in the price of on-net calls, a drop of 1 in the off-net price, and 1.5 for mobile to landline prices. Thus, implemented jointly, the regulation's components favor reducing the prices of various mobile calls. Therefore, they ensure an environment conducive to competition, innovation, and well-being. Furthermore, through its actions, the regulator forms a clear counterweight to anti-competitive behavior and ensures consumer welfare. This last point results in the establishment of measures to keep prices relatively low. This result confirms the analysis of several authors (Bakhkhat and Allaf, 2018; Katz and Yung, 2021), who see regulation as a favorable tool for establishing and implementing effective competition, accompanied by a reduction in mobile call prices.
In further analysis, model 2 reveals that the regulatory authority (RA) component has a significant and negative impact on the prices of off-net calls (P2) and weekend calls (P4), but this impact is not significant for the prices of on-net calls (P1) and mobile to landline (P4). Thus, an increase of 1000 in the regulator's score leads to a reduction of more than 1.8 in the price of off-net calls and 10.6 in the price of weekend calls. It implies that strengthening policies that give the regulator greater autonomy is favorable to lower prices for off-net and weekend calls. Indeed, independence enhances the regulator's Effectiveness in procedural matters and in facilitating financing maneuvers actions to achieve the desired social and economic objectives. Since a single body is responsible for regulating operators' actions, this allows for greater efficiency in planning and bringing convergent technologies and services to the market. Separating the regulator from government agencies is essential in making impartial, fair, and transparent decisions. Wallsten (2001) obtained similar results, reflecting the negative effect of the regulatory authority on call prices. These results show that an independent regulator's presence increases the telecommunications market's performance and promotes lower prices for local calls.
The Regulatory Mandate significantly and positively impacts prices for off-net and weekend calls. For example, an increase of 1000 in the regulatory mandate score is linked to a price increase of 3.47 for off-net calls and 3.6 for weekend calls. Recall that the role of the Regulatory Mandate pillar is to promote the implementation of licenses, interconnection rates, spectrum, universal service, broadcasting, Internet, and computing and to reduce consumption problems. However, our results suggest that such a policy is ineffective enough to lead to lower prices; on the contrary, abundant and restrictive regulation will ultimately result in higher prices. Moreover, according to the analysis of Knieps (2005) and Sidak and Spulber (1998), if the incumbent operator's equipment is heavily regulated, alternative operators (potential entrants) will have no incentive to access the market by bringing new technologies knowing that they can access the basic infrastructures of established operators. Consequently, they will not be encouraged to invest in new infrastructure, which hinders innovation and leads to higher prices.
The Regulatory Regime component significantly negatively impacts all call prices except for weekend calls. An increase of 1000 in the score of the pillar of the regulatory regime is thus associated with a drop in prices of almost 3.9 for on-net calls, 2.6 for off-net calls, and 3.9 for mobile-to-fixed calls. It means that implementing the regulatory regime is favorable to reducing call prices, among other things because it makes it possible to increase transparency in the market. Also, as operators are required to publish their activity, do new entrants have enough information about the network to enable objective decision-making, thus reducing the time to enter and providing a basis for negotiation? Furthermore, publishing a standard offer reduces the possibility for a dominant operator to discriminate against interconnection candidates. In addition, adopting a national plan, including broadband, reinforces the need for consensus and coordination to deploy infrastructure and regulate the services provided. A more coordinated and accountable environment, in turn, accelerates innovation, stimulates investment, and increases productivity, which, together with the universal access objective of the broadband plan, contributes to higher levels of penetration and competition. Along the same lines, network-sharing agreements can optimize the use of coverage for operators, generally reducing costs, which benefits both service providers and consumers. This analysis is consistent with Galal and al.'s (1995) and Gutierrez's (2003) analysis. The first analysis confirms the close relationship between implementing new regulatory reforms and price declines in the communications sector. The second reveals the positive impact of the correct regulatory measures on the deployment and efficiency of the network as well as on the sector's performance level in both static (price reduction) and dynamic (investment) specifications.
The competition framework has a significant impact on overall call prices. However, this impact harms the price of weekend calls and is positive for the other prices. So, an increase of 1000 in the score of the Competition Framework pillar leads to an increase in the price of nearly 3.8 for on-net calls, 1.9 for off-net calls, 3.7 for the price of mobile-to-landline calls, and a drop of 2.1 for weekend calls. The impact of this component on prices is therefore balanced but overall positive. It measures competitive intensity in local and long-distance, mobile, and broadband services. It also includes criteria for determining dominance or significant market power (SMP), as well as tolerance of foreign presence in the ICT sector. Although controversial, our results remain consistent with previous analyses. While in theory, competitive markets are known to increase consumer welfare by lowering prices, promoting innovation, improving consumer choice, and improving the quality of services (Ambrose and al., 1990), several analyses question the positive effect of this pillar on lowering prices. According to these analyses, improving the competition framework can create price spikes in all-out calls. Indeed, a competitive framework not accompanied by solid innovation can lead to degrading services and a rise in prices. Moreover, in the absence of economies of scale, each operator would be tempted to increase its prices to make more profits. The analyses ofKnieps (2005) and Sidak and Spulber (1998) support this notion. Moreover, as observed by Pietrunti(2008), the asymmetry of information between regulator and operator undermines the Effectiveness of regulatory policies. Operators can set higher prices by taking advantage of this regulator's weakness.
4.2. Impact of other variables on prices
In the rest of the analysis, the problem is choosing the model that best explains the relationship between the price and the control variables. In other words, which of Models 1 and 2 best minimizes information loss? As the Akaike Information Criterion (AIC) cannot allow us to choose, we use the Hannan-Quinn Information Criterion (HQC). The latter is an alternative to the Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC).
The analysis of the Hannan-Quinn information criterion (HQC) shows that the index of model 2 is globally lower than that of model 1. Therefore, model 2 is better at minimizing information errors (Table 3) and explaining the relationship.
Using information from model 2, we find that the coefficient associated with investments at the 5% level is not significant for all mobile phone call prices except for weekend call prices. In other words, the evolution of investments during our analysis directly affects only the prices of calls on weekends. A 1000% increase in investment leads to a 4.33% drop in call prices. This result highlights two essential aspects of investment. The first one designates the difficulty of understanding the latter's effects on a static level, and the second highlights the negative impact of the investment on prices. According to this second aspect, by promoting innovation, investment tacitly leads to economies of scale and consequently to reductions in the price of services.
In the same way, the mobile service subscription rate has a significant and positive impact on weekend call prices. For example, an increase of 1000 in the subscription rate is associated with an increase of 1.4 in the price of calls during the weekend. Indeed, in the absence of adequate infrastructure, the increase in the number of subscribers will require additional investments, which, in the end, will be paid for by the end consumer, hence the increase in prices.
The Internet user rate has a significantly negative impact on the total price of mobile calls. An increase of 1000 in the user rate is associated with a price reduction of almost 0.7 for on-net calls, 1.1 for off-net calls, 0.8 for calls to landlines, and 1.1 for calls during the weekend. Internet-related services are gradually becoming proper substitutes for traditional telephony services, like voice service for calls. This service-based competition leads operators to reduce call rates to remain competitive in the market.
The HHI significantly impacts the prices of various calls, except on-net calls. This impact is positive for the prices of mobile off-net and mobile-to-fixed calls but negative for weekend calls. An increase of 1000 in the HHI is thus associated with an increase of nearly 44.9 for the price of off-net calls, 68.5 for calls from mobiles to landlines, and a drop of 82.8 for weekend calls. Thus, an increase in market concentration leads to a fall in call prices during the weekend but an increase in other prices. The impact of greater competition within the mobile telephony market on prices depends on the type of calls. The literature review has widely documented this debate. For example, Hausman and Ros (2013) argue that the high concentration of the telecommunications market in Mexico has favored lower call prices and consumer welfare. In the same context, Friesenbichler (2007) and Aghion et al. (2005) found that there is an optimal level of competition intensity that promotes price performance (low prices) and outside of which competition will inevitably result in higher prices and welfare loss. Continuing the analysis, Jeanjean (2015) concludes that the role of competition remains ambiguous because it can lead to lower prices and maintain an inverted U-shaped relationship with investment and therefore slow down the fall in the unit price. Finally, Houngbonon and Jeanjean (2016) assume that the optimal number of operators that promotes sector performance (and therefore prices) must vary between 2 and 5, but this interval can vary from one market to another.
Table 3
Estimation of models 1 and 2
Variables
|
Model 1
|
Model 2
|
|
dP1
|
dP2
|
dP3
|
P4
|
dP1
|
dP2
|
dP3
|
P4
|
RA
|
|
|
|
|
.0000864
( 0.945 )
|
− .00186*
(0.069)
|
.000052
(0.967)
|
− .01068***
(0.000)
|
RM
|
|
|
|
|
− .0002661
( 0.838 )
|
.00347***
(0.001)
|
.0004
(0.757)
|
.0036***
(0.015)
|
dRR
|
|
|
|
|
− .0039**
(0.017)
|
− .0026**
(0.046)
|
− .0039**
(0.014)
|
.00285
(0.130)
|
CF
|
|
|
|
|
.0034***
(0.000)
|
.0019***
(0.000)
|
.0037***
(0.000)
|
− .0021***
(0.001)
|
ROIT
|
− .0014***
(0.055)
|
− .0010***
(0.070)
|
− .00151**
(0.043)
|
− .00090
(0.308)
|
|
|
|
|
lnINV
|
− .00002
(0.987)
|
− .00017
(0.893)
|
.000053
(0.973)
|
− .004457*
(0.017)
|
.00028
(0.864)
|
.00015 (0.908)
|
.00034
(0.833)
|
− .00433*
(0.024)
|
dCOUV
|
− .0006
(0.180)
|
− .00031***
(0.391)
|
− .0005
(0.264)
|
− .000235
(0.659)
|
− .00045
(0.318)
|
− .0002
(0.580)
|
− .00033
(0.452)
|
− .00047
(0.368)
|
dABON
|
− .00028
(0.422)
|
− .00012
(0.657)
|
.00008
(0.817)
|
.00212***
(0.000)
|
− .000101
(0.774)
|
9.28e-06
(0.974)
|
.00014
(0.683)
|
.0014***
(0.000)
|
dUTIL
|
− .00006
(0.938)
|
− .00033
(0.624)
|
− .00025
(0.767)
|
.00179***
(0.079)
|
− .00027
(0.747)
|
− .00018
(0.785)
|
.00010
(0.899)
|
.0012
(0.222)
|
IHH
|
.086008**
(0.01)
|
.06404**
(0.017)
|
− .10892***
(0.000)
|
− .13617***
(0.004)
|
.04927
(0.144)
|
.04499*
(0.098)
|
.0685**
(0.042)
|
− .0828*
(0.067)
|
RN
|
1.36e-06
(0.479)
|
6.73e-07
(0.662)
|
0.0003*
(0.100)
|
-0.0008***
(0.000)
|
− .00001
(0.518)
|
-8.11e-07
(0.604)
|
2.63e-07 (0.892)
|
-0.000303
(0.274)
|
dCC
|
.03584
(0.256)
|
.01494
(0.556)
|
.04381
(0.166)
|
.05252
(0.166)
|
.03172
(0.310)
|
.013485
(0.592)
|
.038998
(0.892)
|
.0368
(0.313)
|
dPS
|
.016694
(0.249)
|
.00376
(0.746)
|
.01920
(0.186)
|
.01704
(0.325)
|
.014601
(0.310)
|
.0033867
(0.770)
|
.01745
(0.224)
|
.00989
(0.555)
|
_cons
|
2.5026***
(0.000)
|
3.432***
(0.000)
|
2.31856***
(0.000)
|
-19.809***
(0.000)
|
-2.053***
(0.000)
|
1.2978***
(0.000)
|
-1.808***
(0.000)
|
10.817***
(0.000)
|
Number of observations
|
416
|
416
|
416
|
416
|
416
|
416
|
416
|
416
|
Hausman test
chi2(13)
|
90.89***
(0.000)
|
84.86***
(0.000)
|
54.90***
(0.000)
|
33.75***
(0.000)
|
37.42***
(0.000)
|
41.37***
(0.000)
|
49.92***
(0.000)
|
26.81***
(0.008)
|
Hannan-Quinn information criterion (HQC)
|
0.0019
|
0.0012
|
0.0019
|
0.0026
|
0.0019
|
0.0012
|
0.0019
|
0.0025
|
Note: (.) we have the P-values, * significance at 10%, ** significance at 5%, *** significance at 1% |
Source: Author, from STATA 16 |