Zero-Rating: Walled Garden or Gateway Drug? An Empirical Framework for Analyzing the Effects of Pricing on Mobile Data Usage

Abstract

This paper performs an empirical study on mobile usage behavioral patterns in the context of pricing practices such as zero-rating. Zero-rating – or the practice of offering free cellular data access to certain mobile services or applications – has been widely adopted by network carriers in developing countries such as South Africa and Kenya. First, we perform a longitudinal exploratory study of mobile usage patterns in the United States versus South Africa over a period of three years (2013–2015). Using data collected from the MySpeedTest application, we analyze differences in usage behavior for the top 5 applications in terms of total data usage in each country, comparing usage on different connection types (Wi-Fi vs. cellular) as well as for devices on different data plans (unlimited vs. limited monthly data-cap vs. prepaid). Our findings show that US users slightly prefer cellular connections to Wi-Fi connections for most of the US top 5 most used applications, while South African users generally prefer Wi-Fi connections (with the notable exception of Facebook). Further, US users on unlimited and limited plans display much higher average monthly usage than those on prepaid plans, while South African users on prepaid plans generally display much higher usage than those on unlimited and limited plans. Next, we attempt a deeper analysis into the possible behavioral effects of zero-rating and prepaid data bundle pricing. We find that in one instance, zero-rating increases overall usage of the application, regardless of connection type. In another case, we observe increased mobile data usage of a zero-rated application during and immediately after zero-rating, but not in the long term. However, a lack of sufficient data renders many of these results inconclusive. In addition, data insufficiencies prevent a conclusive study on the behavioral effects of changing data prices for users on prepaid plans. This lack of data motivates the employment of a pay-per-install recruitment method to guarantee user installs for the final portion of the thesis. In order to better understand the user motivations behind our empirical observations, this part of our research shifts to a more focused short-term comparative study on zero-rating behavioral effects in South Africa and Kenya. In addition to the ongoing collection of usage data from these installs, we administer a survey in these two countries to gain more qualitative insights into zero-rating and usage practices across our user base. We find zero-rating penetration is actually quite low among respondents, and that zero-rating seems to serve more as a popular data conservation option and possible gateway to full Internet usage than as a walled garden discouraging respondents from venturing beyond zero-rated applications