3.1 Deployment Context
The CCN deployments, shown in Fig. 21.2, are located in the province of Aurora, in barangays scattered across three municipalities along the Pacific coastline and Sierra Madre mountain range. From Manila, it takes 8 to 10 hours of land travel to reach Baler, the provincial capital. From Baler, several hours’ travel by sea or unpaved roads is required to reach these remote and isolated coastal communities, a trip sometimes made hazardous or impossible by seasonal typhoons. Community members mainly depend on fishing and farming for income and sustenance. A few locals earn their living by reselling retail goods brought from the town centers to their respective communities. Owing to their geographic location, these sites do not have access to terrestrial radio and television broadcasts. Prior to the commencement of the CCN deployments, they were beyond the reach of cellular coverage; locals would need to travel several hours to use cellular services like calls, SMS, and data.
The project installed seven community cellular networks in Aurora from September 2017 to January 2019. These sites were randomly selected from a pool of 14 candidate sites as part of a randomized controlled trial (RCT) impact assessment study.
Prior to the networks’ installation, all candidate sites underwent a household-level baseline survey in December 2016. The baseline survey involved three parts: (1) a household survey, (2) a listing of all adults 15 years or older, and (3) a one-on-one adult survey. The household survey queried household demographic composition, asset ownership, and economic activity. The core adult survey asked about social networks and included a travel diary. Across the 7 CCN sites, 1131 households were interviewed, and we listed 3057 adults (Keleher et al., 2020).
The survey data was used as input to a pairwise matching procedure as described in Bruhn and McKenzie (2009). First, potential cell tower locations were sorted into pairs that were as similar as possible along observable characteristics, and then one location within each pair was randomly selected to receive a cell tower in the first wave (treatment) of deployments. The other would receive a cell tower in the second wave (control). Specifically, pairs were formed so as to minimize the Mahalanobis distance between the values of all selected characteristics within pairs, using an optimal greedy algorithm. The research design required the CCNs to be deployed in a specific order.Footnote 1 Table 21.2 shows the final list of pilot sites. Several months after a site’s network was launched, a household-level randomized experiment was performed on promotional pricing such as free credits and discounted rates.
Finally, from May and September 2019, surveyors returned to all 14 candidate sites to conduct an end line survey to measure changes after the network installations. Surveyors administered the same household and adult surveys, prioritizing interviews with the same individuals as in the baseline.
As part of the RCT study, SIM card distribution was tightly controlled as shown in Fig. 21.3. While SIM cards were given for free, only eligible individuals could receive them. Eligible individuals were defined to be 15 years old and above, residing in the community for at least 6 months at the time of site launch. The impact evaluation team had initially wanted to limit SIM cards only to residents of the barangay. However, upon request, exemptions were given to civil workers such as public school teachers and soldiers, as they were often assigned from other municipalities.
The network was dubbed “VBTS (Village Base Transceiver Station) Konekt Barangay” and branded separately to differentiate it from the mainstream Globe network. Pricing for the network services was already set by Globe Telecom and the national regulator. As the CCN was categorized as an experimental network, the per-minute or per-SMS service rates were lower than those of the mainstream network. Any price changes (such as the time-limited promotional pricing or “promos”) were to be approved by Globe first and then by the national regulator. Service rates are shown in Table 21.3.
3.2 Implementation Challenges
Aside from developing the technical intervention, the project required coordination and work with a wide variety of stakeholders. These included a national mobile network operator (Globe), the national regulatory agency (National Telecommunications Commission – NTC), local government units, research collaborators in the academe, local cooperatives, and network end users within the remote communities. A large amount of time and resources were invested to establish the necessary partnerships and agreements with these stakeholders.
3.3 Spectrum Negotiations
With the core technical pieces in place, the researchers quickly discovered that starting these community cellular networks was not as straightforward as they had originally envisioned. The Philippines does not have a dedicated spectrum policy for last-mile service delivery, and the current regulatory framework forces small operators to adapt to the model used for national telecoms and other large organizations. MNOs are given licenses that span the whole country, even in areas where they are not providing service. Furthermore, current regulations on radio equipment use, SIM card production, and interconnect limit these activities to only MNOs.
There was no path for community networks to apply for their own licenses. In response, the project team first attempted to acquire a license exemption from the national regulator, the NTC. However, the regulatory officers advised the researchers to instead reach out to any of the current license holders and ask if they would allow co-use of frequencies under their respective licenses. The NTC would allow the project, provided an official agreement was acquired with a current licensee. The researchers sent out proposals to the major Philippine MNOs, to which Globe Telecom responded.
The long negotiation process with Globe took more than 2 years to close.Footnote 2 The project worked its way internally through various departments, explaining how different components of the project would pan out upon deployment. The research team also had to reassure Globe that the project would shoulder the CCN’s capital and operational expenses. Eventually, the project was taken under Globe’s corporate social responsibility arm and was granted approval to use Globe’s spectrum for an initial 1-year pilot period. One of the risks the project took on due to this spectrum arrangement was that the CCNs’ operation was critically dependent on Globe’s continued support.
3.4 Stakeholder Needs and Interests
Over the course of discussions and consultations between the researchers, Globe, and LGUs, we determined that local community partners would be needed to handle the day-to-day operations, management, and first-level maintenance for the CCN installations. The researchers initially wanted to recruit only local organizations based in the same communities as the installations, but were not always able to find organizations that fit these criteria. We enlisted the help of the LGU to nominate cooperatives and organizations and ended up with cooperatives not based in any of the target communities, but local to the municipality. Table 21.4 describes the partnership setup per municipality.
Moreover, the researchers wanted to maximize community involvement to create an operational model for the network that was as close as possible to the ideals of community networking yet that would also satisfy the other partners’ requirements. While the team had past deployment experience on which to base the operational model, Globe’s stipulations had to be accommodated. Globe preferred to have a single point of contact, rather than dealing with many independent communities. Its existing trade and distribution processes also relied heavily on partners having easy access to financial institutions such as banks and means of electronic communications such as email.
The cooperatives were primarily worried about the capital and potential financial liability that this venture could bring should the project not become sustainable. They also recognized that they would need to visit the retailers at the sites often, which might become inconvenient or infeasible due to the danger of travel in bad weather. They were skeptical about the business viability, as some of the treatment sites were very small in population. These were all valid concerns, especially as these deployments would be a test of the CCN model. However, they were convinced to participate primarily by the 80% revenue share they would eventually receive from the network’s gross revenues. As L2 actors, the cooperatives would also receive a discount on the wholesale e-load purchases from Globe as well as their 80% share. Finally, the project team assured the cooperative that all initial investments for the infrastructure were funded by the project.
3.5 Unexpected Changes in the Field
While the spectrum negotiations were taking place, the deployment team from UPD had been visiting and surveying potential sites where little to no cellular coverage existed. The team had set criteria for selecting pilot sites for CCN deployments. A primary consideration was for a village to be outside existing cellular coverage. Initially, five isolated coastal barangays in San Luis, Aurora, were selected as deployment sites. However, the introduction of the RCT study required additional sites to meet the minimum requirement for the matched pairs design (at least 14 sites). Hence, the search was extended to include nearby coastal communities in the town of Dingalan.
The initial site listing was vetted by Globe to ensure that they were indeed excluded on their existing coverage map. At the time, the Dingalan sites were on the fringes of existing cellular coverage. This meant that coverage was not present in the village, but residents had identified spots several kilometers outside the village where they could acquire a signal by walking.
The data gathered from the site surveys were then passed to the impact assessment study team for evaluation. The researchers then generated a final listing of seven treatment sites, with the research design necessitating that the CCNs had to be installed and deployed in a specific order. As a result of this process, the sites were primarily selected according to RCT requirements and not for business viability. Some of the target sites were very small in population and had been identified from the onset as having potential difficulty with generating enough revenue to cover recurring costs without external support.
Over the course of site preparations (mid-2017), two of the treatment sites in Dingalan needed to be relocated due to unforeseen security threats from military insurgent groups in the area. This forced the research team to abort some of the initially selected sites, and the deployment team had to scout other nearby areas for candidate locations, prompting a re-evaluation of the control-treatment pairs. The project team then had to forge new partnerships with local stakeholders in these locations.
Once the treatment sites were identified, the next step was to secure a small lot (7 m × 7 m) in each site where the CCN tower and equipment shelter would be built. If possible, the project team preferred to have the CCNs erected on government land, as the project already had partnerships with the LGUs that had assisted in expediting site clearances and other permits.
However, the project team had a hard time acquiring land for the sites in Dingalan, as a private corporation, Green Square Properties, claimed ownership over almost the whole municipality. The team initially attempted to seek permission with this private entity, but the company wanted lease payments in exchange. As we could not guarantee that income would be generated, the negotiations with them failed. The team tried by all possible means, including leveraging the land-use agreement that the Dingalan LGU had with Green Square. The negotiations for land use took a considerable amount of time and was further complicated when sites had to be re-randomized. In the end, the team ended up making informal arrangements with private homeowners in the area.
In subsequent visits to Dingalan, the researchers observed that network coverage from mainstream networks had improved since these locations were first assessed. By mid-2018, mainstream network coverage had expanded so that residents could now utilize them in the comfort of their homes. With this change, residents in Dingalan preferred to use the mainstream networks over the community network since residents perceived them as more reliable and affordable, and they offered additional services such as mobile data and promos for “unlimited” usage. The community network could not compete with the incumbent MNOs. While the coverage expansion was detrimental to the survival of the community network, we acknowledge that it ultimately brought benefits to the community.