Service quality in Brazilian mobile telephony: an efficiency frontier analysis

Abstract

The mobile telephony sector is characterized by the dynamic interplay of rapid changes in technology and an apparently growing competition, as indicated by the fierce nonprice competition, and is yet associated with the entry of new operating companies in some cases. In this context, a relevant and neglected issue is the assessment of how service quality responded to an increasingly competitive environment. This study utilizes data envelopment analysis to assess the quality efficiency of mobile telephony companies in Brazil during the 2000–2003 period. Window analysis was conducted for the entire period, taking as reference different quality indicators pertaining to different forms of complaints and calls completed and interrupted. The efficiency measurement was made feasible by interpreting the indicators reflecting a positive dimension of quality as outputs and those reflecting negative aspects of quality as inputs. Given potential heterogeneities across firms that relate to the frequency band and to the technology (time division multiple access technology and code division multiple access technology among others), this article considered adjusted efficiency scores. The Tobit model for censored data was estimated to control for the aforementioned aspects. Rescaled residuals from the econometric estimation produced efficiency scores for service quality. The evidence indicated an overall improvement in efficiency over time. Nonparametric tests were consistent with a positive evolution of quality over time even for shorter subperiods.

Notes

¹Although some strengthening of competition appears to have occurred in the sector there is evidence that multimarket contact and cross-ownership are important in explaining noncompetitive prices as indicated by Parker and Roller (1997).

²Overviews on quality in the context of regulation appear in Sappington (2003) and Façanha and Resende (2004).

³See Hausman (2002) and Gruber and Valletti (2003) for comprehensive overviews on mobile telephony.

⁴The exception was the state of Rio de Janeiro that used the B band but in 1997 was to migrate to band A.

⁵A fifth band (band C) was predicted but no interested operators presented bids. That band is defined in terms of the 1725–1740 MHz range for mobile terminal and 1820–1835 MHz for ERB.

⁶The protocol IS-41 guaranteed ‘roaming’ between the areas covered by different concessionaries in mobile telephony under TDMA and AMPS technologies. In Brazil, all operating companies in band A keep AMPS channels in the totality of its coverage area so as guarantee national ‘roaming’ for all subscribers.

⁷The program above is consistent with an output expansion perspective (output orientation) where one attempts to maximize a virtual output given the level of a virtual input. An alternative perspective would be that of input conservation (input orientation) where one seeks to minimize the level of the virtual input given the level of the virtual output under the CCR model with constant returns to scale the two approaches are equivalent.

⁸See Cooper et al. (2000) and Thanassoulis (2001) for introductions to DEA methods.

⁹There is a growing interest in the application of windows analysis. See for example Asmild et al. (2004).

¹⁰Tupper and Resende (2004) also considered these adjusted DEA efficiency scores to devise a yardstick scheme that rewarded units that were relatively more efficient in Brazilian water and sewarage utilities. In principle, one could conceive some yardstick scheme to reward quality performance in telecommunications. However, one has to bear in mind that mobile telephony is essentially deregulated.

¹¹The intention of the regulator was to secure roaming in the different regions.