The main goal of this paper is to estimate the welfare effects of the Peruvian telecommunications privatization. The authors also use their findings to provide an explanation for why privatization is not popular in Peru (as in much of Latin America), despite the fact that it seems to have had a number of salutary effects. The paper improves on the previous literature insofar as the data it relies on offer advantages relative to those typically analyzed. The use of these data, unfortunately, is simultaneously a significant weakness, in that information collected during a single year is used to make inferences on a lengthy period that witnessed many changes in the sector, which introduces several potential biases. This comment closes with some thoughts on what we can learn from such results regarding the low approval rates accorded to privatization.
This issue of how privatization affects the level and distribution of welfare has been analyzed before for Latin America, notably in a series of papers summarized by McKenzie and Mookherjee in an earlier issue of this journal.1 Such work generally relies on household survey data, which have a key limitation because they typically describe only households' expenditure on broad aggregates like telecommunication services. From these, it is sometimes impossible to identify the specific quantities of services consumed at specific prices. This is particularly problematic in the case of telecommunications, which involves a complex basket of available services (such as fixed-line local calls, national and international long-distance calls, and cellular telephony). Estimations thus have to rely on a series of assumptions and approximations.
In contrast, Torero, Schroth, and Pasco-Font use a survey specifically geared to telephone services, one that also entailed the transcription of billing information. The survey has a substantial sample size and follows a panel of households over a period with price variation, so that changes in quantities and prices are accurately observed. In this regard, this paper [End Page 123] offers a solid starting point for its estimations. Collecting such data is never easy, and the authors should be commended for doing it.
Unfortunately, the use of this information also introduces a number of potential biases, which arise because the data were collected over a period of only ten months, mostly in 1997. One illustration of the type of problems this may cause is the correction for access to fixed-line phone services in the demand estimations, which is a key aspect of the estimation implemented. The postprivatization expansion of fixed-line connections in Peru essentially took place between 1994 and 1996. This means that the characteristics of households that did not have access in 1997 (at the time the survey was taken) are probably quite different from those that did not enjoy it in 1993 or 1994, and this can affect the validity of the correction substantially.
A related point concerns the introduction of cellular telephones. Given their relative lack of importance in 1997, the survey used did not collect information on cellular services. Cell phone coverage has expanded dramatically since then, however. This is very important because as the authors themselves explain, the main source of variation in welfare changes across socioeconomic groups originates in the relative importance of fixed costs. A driving force behind the increase in cellular coverage, however, is precisely the reduction in the fixed costs of access (certainly relative to fixed-line services), and it is likely that new subscribers have experienced large welfare gains from this service. This simply cannot be elicited using the 1997 data. The very low penetration rate that existed then may explain, for instance, why the authors find cell phones are complements to rather than substitutes for fixed-line services. This might be the case for high-income households, but it would be surprising among low-income customers that gained access more recently-and these are precisely the individuals that, in some sense, are the ultimate focus of this study.
The bottom line is that information from one year is being used to make inferences on six rather different years. This is a key source of concern over and above any other methodological issues one...