MIS Workshop Proceedings Format - SSRN papers

Salcedo & Kuhlmann

A model to estimate the broadband and Internet access demand for typical Mexican rural communities

A model to estimate the broadband and Internet access demand for typical Mexican rural communities Ante Salcedo ITAM [email protected]

Federico Kuhlmann ITAM [email protected]

BIOGRAPHIES

F. Kuhlmann and A. Salcedo are full time professors at the Digital Systems Department at ITAM. They conduct, among other things, research on telecommunications industry trends, infrastructure, economics, and public policy. ABSTRACT

Benefits of developing telecommunications infrastructure are beyond questioning; however, challenges stand at small rural communities, which in Mexico encompass ~15% of the population, living in townships with less than ~800 residents. Previous research indicates that technology concepts like the Internet of Things (IoT) can create economic value in local economies. Thus, this paper presents a model to estimate wireless access demand in such communities. Estimated projections of IoT product adoption are built for a representative scenario, considering publicly available statistical information (accounting for variables like the number of residents, households, cars, commercial establishments, or local economic indicators). A basket of highly probable products, like payment terminals, is constructed and characterized. Then the adoption process for each considered product estimated with an S-curve, to integrate the corresponding broadband/Internet access demand. Telecommunications industry is experiencing exciting times, yet careful analysis is crucial to support well informed strategic planning processes. Keywords

Mobile Broadband Access, Rural Community, ICT

INTRODUCTION

The objective of this paper is to present a model to describe small rural communities, as well as preliminary estimations of the Internet and broadband access demand that can exist in them, assuming that broadband/Internet availability will trigger the adoption of information and communication technology (ICT) products based on the Internet of Things (IoT), and other emerging technology concepts. Digitization has been defined as the social transformation triggered by the massive adoption of digital technologies to generate, process, share, and transact information. Recent studies, like The Digital Ecosystem and Economy in Latin America[1] show the importance of increasing the digitization index in countries like Mexico. The benefits of developing telecommunications infrastructure and increasing its penetration and utilization are beyond any doubt. However, there are important challenges to be resolved at small rural communities with less than ~800 residents that lack of broadband and/or Internet access, which in Mexico encompass approximately 15% of the total population. Figure 1 illustrates, as an example, the 3G/4G cellular coverage over the state of Chiapas, reported by the Open Signal WEB application. As it can be appreciated, cellular service supporting wireless access is only available at very few singular points, which overlap with the largest townships in the region. It was found that in the particular case of Chiapas townships with less than 3,000 residents are not connected yet.[2]

150

CPRLATAM Conference, Mexico, June 22-23rd, 2016 in conjunction with CLT2016, June 20-23rd, 2016

Electronic copy available at: https://ssrn.com/abstract=2863367

Salcedo & Kuhlmann

A model to estimate the broadband and Internet access demand for typical Mexican rural communities Figure 1. Mobile wireless broadband availability (from Open Signal WEB application).

In order to restrain the length and scope of the paper, the case of Chiapas has been taken as a case of study; not only because of its low wireless access penetration, but also because it was recently defined by the Mexican government as a “Special Economic Zone” that needs to be developed. While the work in progress that is presented in the paper only considers a particular scenario in the state of Chiapas, the presented model can be easily extrapolated to study other regions of the country. Previous studies[3] make the authors believe that small rural communities in Mexico may have sufficient economic activity to justify the deployment of the infrastructure that is necessary to provide wireless Internet and/or broadband access, and that there are viable schemes to capture economic and social value from them. The authors also believe that emerging technology concepts like the Internet of Things (IoT), or new business models like the Mobile Virtual Network Operators (MVNO), will impact a diversity of local economic activities (such as farming, logistics and commerce), in such way that the economic value that could be captured should have an eventual impact on the overall local Gross Domestic Product (LGDP). However, significant quantitative analysis and research work are still necessary in order to gain a deeper understanding of the problem, to dimension it, to evaluate alternatives to solve it, to develop strategic plans, and to take effective actions to capture the possible existing value. TARGET TOWNSHIP SELECTION

While the largest townships in Mexico may be already connected, many questions about the deployment of infrastructure in the smaller unconnected communities remain unanswered. This article proposes a model to estimate the Internet and broadband access demand that could exist at such communities, as a part of a larger effort meant to size the economic potential existing value and to determine viable alternatives to capture it. Along the paper estimated projections of ICT product adoption are built, for an illustrative scenario that is representative of a small unconnected Mexican rural community. The scenario that is presented considers typical township variables, like the number of residents, households, cars, schools, hospitals, commercial establishments, as well as an indicator of the main local economic activity. Before developing the township model, a careful selection of communities was made to determine the type of township to be studied and modeled (the target township), which basically encompasses the typical, largest, but still unconnected, communities. For instance, in the case of the state of Chiapas, the target townships have a population between 900 to 3,000 residents. Figure 2 shows a plot of the number of townships in Chiapas (listed by INEGI), versus the number of residents they have. As it can be appreciated, there are very few with a population larger than 3,000 residents, while there are hundreds (or even thousands) with a population below 100. Careful analysis of the mobile service coverage by means of the Open Signal Web application, shows that there is a threshold for connectivity at ~3,000 residents. This means that the larger townships in the state have wireless access availability, while the smaller ones do not. By this means, communities below the threshold would be natural candidates to be considered as target townships. A lower population limit was set at 900 residents per community, considering the territorial extension (as explained afterwards).

CPR LATAM Conference, Mexico, June 22-23rd, 2016 in conjunction with CLT2016, June 20-23rd, 2016


151

Salcedo & Kuhlmann


Figure 2. Largest non-connected communities in Chiapas (elaborated with data from INEGI).

Localidades por número de habitantes en Chiapas Number of townships 10,000

Disperse population 1,000

Major connected cities Target townships

100

10

1 1

10

100

1,000

250

3,000

10,000

100,000

1,000,000

Number of residents

900 While the target township population (900 to 3,000 residents) may be relatively small, there should be a substantial rural population, as well as a relevant economic activity, scattered in the neighborhood. The target township model that is presented along this paper considers not only the population and economic activity within a target township itself, but also the one in the close vicinity. The first step to build a target township model was to look at the 118 municipalities1 that Chiapas is subdivided into, and identify those that are primarily urban. As a result, ten municipalities with less than 30% rural population were filtered out of the analysis as they were considered to be urban. The remaining 108 municipalities, which were considered rural, were further analyzed to determine typical parameters, and select a small sample of municipalities with average population. The ten largest urban municipalities that were filtered out have a combined population of 1,134,000 residents (~24% of the total state population). On the other hand, an average rural municipality was found to have a population of 33,900 residents, which for the 108 rural municipalities integrates into 3,661,200 residents (~76% of the state population). In order to continue building the model for a target township, average rural municipalities were further analyzed, finding that on the average each of them contains 2 principal rural cities, 4 target townships, 12 satellite townships, and additional scattered population. A principal rural city was assumed to be a township with more than 3,000 residents, which already has wireless access availability. The target townships, as explained before, have a population between 900 and 3,000 residents without Internet, nor broadband availability. Satellite townships were identified as much smaller locations, with a population lying between 250 and 900 residents, which usually can still be found in commercial maps, near a principal or a target township. The rest of the communities were just assumed to be scattered population settlements. Figure 3 illustrates the way an average municipality was composed.

1

From the Spanish name “municipios”.

152



Salcedo & Kuhlmann

A model to estimate the broadband and Internet access demand for typical Mexican rural communities Figure 3. Composition of an average rural municipality.

Target township 1

Target township 3

Principal rural city 2

Principal rural city 1

Target township 2

Principal rural city

Target township 4

Target township

Satellite township

A closer look at figure 3 shows that the different communities considered in the municipality model have also been connected to each other with dotted lines, which represent roads and/or trails. The analysis of the different municipalities shows that the topology that was considered is indeed a good estimation to describe them. Figure 4 illustrates, as an example, the way the average municipality model fits to the Arriaga municipality in Chiapas. While the municipality does not match exactly the proposed model, for the purpose of analysis it was considered that the similarity between them is good enough to understand its general behavior. Figure 4. Average municipality model fit to the Arriaga municipality.


153

Salcedo & Kuhlmann


Assuming that an average municipality can be integrated as proposed, then the target township and its close vicinity were modeled, considering the corresponding fraction of the overall municipality. TARGET TOWNSHIP MODEL

As explained before, a target township is considered to be one of the constitutive components of an average rural municipality; same as the principal rural city, the satellite township, and the scattered population. For modeling purposes, it was considered that on the average a principal rural city has 5,850 residents; a target township 1,300; a satellite community 650; and all other townships with less than 250 residents, in a municipality, a total population of 9,200 altogether. Then, considering the availability of roads and trails, satellite townships were associated to principal cities or target townships, as if they were part of their immediate vicinity. In such fashion, one satellite community was assigned to each of the target townships, and four satellite communities to each of the principal rural cities. The disperse population was distributed as well, assigning 80% to principal cities, and 20% to target townships. Considering the new population distribution, a principal city and its close vicinity was assumed to have 12,130 residents, while a target township and surroundings 2,410. With the population distribution that was just explained, an average municipality would be integrated by two principal cities and four target townships (considering that the population in their vicinity is already accounted for). In such fashion, adding up the population in the 108 rural municipalities, and the 1,134,000 people living in urban municipalities, it is possible to obtain the population of Chiapas (4,795,200 residents). Considering 432 target townships (108x4), the total population in such communities is 1,041,120, which represent the ~21% of the total population in the state that do not have Internet or broadband service. Once the population in a target township (including its close vicinity) was obtained, then the territorial surface was also estimated to build the model. It was found that in average rural municipalities have a 648 Km2 surface, which was divided equally by 6 (two principal rural cities and four target townships) to obtain an average of 108 Km2. For the purpose of the model, it was assumed that the target township has a rectangular shape with 12Km by 9Km, as shown in figure 5. Assuming that such communities have extremely few commercial activity and government services, a distance of 12Km may be reasonable for people who don’t have access to a vehicle but have to commute frequently to a larger township, in timeframes shorter than a day. The 432 target townships in the stat add up for a total of 46,656 Km 2 (~64% of the total state territory). Figure 5 also shows how a target township may look like. Figure 5. Proposed model for a target township.

The territorial distribution was further analyzed in order to identify farming fields, urbanized land, and natural forests/reserves. Figure 6 illustrates how the territory was segmented and classified. A full municipality, such as Arriaga, was segmented into regular 2Km by 2Km squares, so that each square could be classified according to the type of use that was observed on the map. As a result, it was found that approximately 6% of the territory is urban, 23% is natural forest, and 71% 154


Salcedo & Kuhlmann


is farming fields. The estimated distribution was taken for the Arriaga municipality, and still needs to be validated for other municipalities. It was assumed, for the purpose of this paper, that farming is the only relevant economic activity. Some cattle rising was also considered later on, but no other economic activities (like tourism, energy, and/or manufacturing). Figure 6. Urban-farming surface distribution.

Once the population and territorial characteristics were settled, the rest of the parameters for the average target township model were established. Table 1 summarizes the assumed parameters, including the number of residents, households, cars, schools, hospitals, commercial establishments, and the local economic activity (farming), among other things. The parameters were estimated from public information, considering pessimistic circumstances. Table 1. Target township model considered parameters. Total population

2,410

People

Child

362

15%

Youth

241

10%

Working

1,398

58%

Elder

410

17%

Professional

41

2%

Surface area

108

Km2

Farming

77

71%

Roads

30

Km

State roads

10

33%

Municipal roads

20

67%

Places

946

Constructions

Homes

532

95%

Retail stores

19

2%

Restaurant, bar, hotel

16

Schools

13

2%

Churches

4

0%

Medical centers

2

0%

Municipal modules

3

0%

Libraries

2

0%


155

Salcedo & Kuhlmann

A model to estimate the broadband and Internet access demand for typical Mexican rural communities Federal buildings

1

0%

Vehicles

329

Vehicles

Personal cars

127

39%

Trucks and tractors

100

37%

Motorcycles

70

21%

Transportation buses

10

3%

The parameters shown in table 1 reflect how small are the target townships, and how very little economic activity they have. Yet, a significant farming area is also observed, which indeed generates sufficient economic value to attract all the people who lives there. Communications availability may be an economic booster, not only by increasing the efficiency of the local economy, but also by attracting new people, businesses, and resources, from other places. In particular, wireless Internet or broadband availability will not only support personal communications devices, but also, any other technology that can help to create local economic value. In this sense, the Internet of Things can bring a variety of connected devices that can help boost economy, and create local value and attractiveness around the main economic activities, which in the case of the present case of the analyzed municipalities are farming and cattle rising. DEVICES TO BE CONNECTED WITHIN THE TARGET TOWNSHIP BLOCK MODEL

As explained, this article considers the imminent adoption of IoT and other technology concept solutions, triggered by availability of wireless access services. Therefore, a heuristic exercise was made to determine which connected devices could be adopted within a target township (and its neighborhood) if wireless Internet/broadband was available everywhere, and the corresponding business models were developed to provide all possible products and services.[3] A selection of highly probable ICT products was integrated and characterized. Keeping in mind a very conservative scenario, the considered devices are: water and electricity utilization measurement and control devices, IP cameras, personal devices like smart phones or tablets, GPS tracking devices, alarms and panic button systems, 3G/LTE-WiFi access points, smart connected TVs, IP car stereos, car insurance trackers, inventory management systems, fingerprint access and assistance control, payment terminals and modules, automatic speed ticketing devices, and ATMs. The amount of products that could be adopted by each person, place, road, piece of land, and vehicle considered in Table 1, was heuristically estimated. As a result, approximately 9,000 devices were found to be probably adopted. While figure 7a summarizes the estimated percentage for each type of device, figure 7b shows the estimated percentage per possible user. As it can be appreciated, the number of adopted devices widely surpasses the number of residents (by ~373%), as the most significant reason for adoption should be related to the local economic activity (to create economic value), and not necessarily to personal communications and entertainment. Figure 7a. Distribution of connected devices that can be adopted within a target township.

BAM 3G/LTE Smart IP TV

1% 2% Alarm & panic button 4% GPS tracking

Figure 7b. Possible users of the adopted connected devices.

Total of ~9,000 devices Vehicles

2% Other

Cattle 5%

Other places

6%

30%

6%

Water measurement and/or control devices

People

IP camera 7%

16%

Other include: • • • •

Personal device 16%

•

30% Electricity measurement and/or control device

156

• •

IP car stereo Inventory control Insurance tracking Fingerprint/access assistance control Pay terminal or pay module Speed tickets ATM

4%

58% Farm fields

Homes

11%


Salcedo & Kuhlmann


As far as this paper goes, it is assumed that each adopted device requires one single wireless Internet or broadband access; that is, an average target township may eventually require ~9,000 of such wireless access accounts. While some products may be broadband demanding, like an IP camera, other products may not require broadband capacity at all, like an IP alarm button; yet, both of them have enormous potential to create economic value. At this point, the quality of the access is not accounted for, nor its potential economic value. Such analysis is left for future work. Instead, once the potential access demand was estimated, then a time projection for possible future adoption was generated and analyzed. S-CURVE PARAMETERS FOR FUTURE ADOPTION PROJECTIONS

The adoption process for each of the considered products was projected in time with a Gompertz S-shaped curve.[4][5] Then the overall Internet/broadband access adoption in time was integrated by adding up all the individual product adoption time growth estimations. The Gompertz S-shaped curve is defined as: −γj t

Sj (t) = αj e−βje

(1)

.

In equation (1), the parameters αj , βj and γj , are used to describe each product Pj , and adjust its growth projection in time. In particular, parameter

αj allows determining the final number of adopted products, and γj the rate of adoption. The larger

that γj is, the faster the product is adopted. For each of the products that were assumed to be adopted within a target township, the adoption parameters were defined heuristically assuming, again, a pessimistic scenario. Table 2 shows the selected parameters for each of the considered products. Table 2. S-shaped time projection curve parameters for the adoption of selected IoT products. Connected things

Estimated quantity

Start adoption

Adoption rate

Water measurement/control

2,631

Late

Fast

Electricity measurement/control

2,631

Late

Fast

Personal device

1,369

Early

Fast

IP camera

640

Intermediate

Slow

GPS tracking

511

Early

Steady

Alarm system and panic button

330

Intermediate

Fast

Other

215

Early

Steady

BAM 3D/LTE

204

Early

Fast

Smart connected TV

105

Early

Slow

TOTAL:

8,636

From the parameters shown in table 2, the S-curve projection for each product was calculated. Figure 8 shows, for instance, the projection corresponding to the adoption of personal devices within the target township. Figure 9, on the other hand, shows all the 9 considered curves, for the corresponding products shown in Table 2.


157

Salcedo & Kuhlmann


Figure 8. Estimated projection in time for the adoption of personal devices in a target township.

Number of adopted personal devices 1400 1200

1,369 adopted devices

1000 800 600

Start adoption

400

200 0

0

1

2

3 4 5 Years to adoption

Figure 9. Estimated projection in time for the adoption of different connected devices in a target township.

Number of adopted IoT devices

3000

2,631 water measurement & control devices

2000 1,369 personal devices

1000

0 0

1

2


649 IP cameras 511 GPS trackers 330 Alarm systems Other 3G/LTE BAMs Smart TVs

From the different estimated time projections of IoT device adoption, the overall broadband/Internet access demand was estimated assuming that each adopted product requires only one access. Therefore, the access demand in time is given by: M

NBA (a) = ∑ Nj (a) ;

(2)

j=1

where NBA (a) is the number of broadband/Internet access demanded at the end of the year a. Direct integration gives the growth estimation shown in figure 10. While figure 10a shows the stacking of all the adopted products, figure 10b shows the overall estimated wireless access demand as a function of time.

158


Salcedo & Kuhlmann


Figure 10a. Stack of independent product Internet/broadband access demand.

Internet/broadband access demanded by connected things 9,000

8,636 adopted things

Figure 10b. Estimated time projection of overall Internet/broadband access demand.

Overall estimated Internet/broadband access demanded 9,000

8,636 adopted things

6,000

6,000

3,000

3,000

0 0

1

2


0 0

1

2


Figure 10b constitutes the final outcome of the proposed model, which corresponds to the expected evolution in time of the broadband/Internet access demand, for the modeled rural target township. When all the target townships within a state are accounted for, the overall wireless access demand that is expected from the unconnected townships in such state can be estimated. The assumed parameters may very well change, of course, for the different townships, municipalities, or states; and even for the particular perspective that each stakeholder interested in providing or adopting broadband/Internet services in Mexico may have. While every stakeholder may use a model like the one proposed, considering particular circumstances and analysis goals, the obtained results show that:  The expected demand should be much larger (probably by a factor of four) than the number of personal devices.  The most relevant adopted devices should help improve all local activities, in order to create economic value.  There could be a significant value to be captured in a time frame of three to four years.  The overall impact within a state (considering hundreds of target communities) may get to have an eventual impact on the overall local Gross Domestic Product. There are still many questions that have not been answered, as they are out of the scope of this work, such as: What is the total Internet/broadband capacity required in a target township? What is the actual economic value that can be captured from such communities? And, what is the investment cost that is necessary to get a target township connected? The research team continues searching for answers; yet the presented model represents a good step towards finding them, as it allows breaking the complex problem of understanding a complete country, state, or even municipality, into the much more simple problem of understanding the typical basic component where the difficulties are. The authors are currently visiting municipalities and townships to validate some of the assumed parameters, are generating the characteristic parameters for the different states and municipalities in Mexico, are running extreme scenarios to analyze the sensitivity of the solution, and are in the process of extending the model to estimate broadband capacity and economic impact. The consequent results will be submitted for publication in the near future, with the final purpose of providing insightful thoughts to gain understanding of the problem, dimension it, evaluate alternatives to solve it, develop strategic plans, find viable schemes to capture economic and social value, and to take effective action for their implementation. CONCLUSION

While the benefits of developing telecommunications infrastructure and increasing its penetration and utilization are beyond any doubt, important deployment challenges rest at thousands of small rural townships with less than ~800 residents, which in Mexico encompass approximately 15% of the total population. Difficult research questions need to be addressed before


159

Salcedo & Kuhlmann


viable and justifiable actions can be taken to provide broadband/Internet access service in such communities. One difficult research question to answer refers to the potential wireless access demand that may exist. While personal communications may certainly be an important source of demand, the Internet of things and other emerging technology concepts may very well boost such demand, providing interesting applications that can help to improve the local economic activity. Along this paper, a model to estimate the possible Internet/broadband access demand in Mexican rural communities with less than 3,000 residents was presented. Also, estimated projections of product adoption were built for a scenario that corresponds to a representative Mexican rural target township in the state of Chiapas; considering publicly available statistical information. The presented model accounts for township indicators of the number of residents, households, cars, schools, hospitals, commercial establishments, and the local economic activity, among other things. A basket of very probable ICT products like connected cars or payment terminals was defined and characterized for the selected scenario. The adoption process for each considered product was calculated by means of a Gompetz S-shaped curve, and then the corresponding overall wireless access adoption process was integrated from the individual product growth estimations. The obtained projections of broadband/Internet access demand for an individual rural community, under the assumed constraints, was presented and discussed. While the obtained results may rise many questions to reflect on, they also have shown that the possibility to impact over local economic activity may increase the overall local access demand to a number that surpasses three to four times the possible demand for personal communications devices. The proposed model has proven to be useful, as well as versatile to be extended into wider scope analysis. Information and communications technology industry is going through exciting times, and attractive business opportunities are around the corner, yet careful analysis is crucial to support accurate and well informed strategic planning processes. This work contributes with insightful academic thoughts and discussions on different matters, such as how to increase broadband penetration and, as a consequence, the Mexican digitization index. AKNOWLEDGEMETS

This research work has been partially supported by the Asociación Mexicana de Cultura A.C. REFERENCES

Katz R. (2015) The Digital Ecosystem and Economy in Latin America. Telefonica Foundation. Roman L., Mora M., Salcedo A. (2016). Local layered algorithmic model for topological design of rural telecommunications networks. International Conference on OR for Development (ICORD), Mexico City. 2016. Salcedo A., Ávila A., y Kuhlmann F. (2015). Searching for economic value in Mexican rural communities. In Proceedings of the 9th Communication Policy Research (CPR) LATAM Conference, pp. 169 -177. Franses P.H. (1994). Fitting a Gompertz Curve. The Journal of the Operational Research Society, Vol. 45, No. 1, pp. 109113. Orendain J. (2007). Diagnóstico y prospectivas del sector de telecomunicaciones en México, en un entorno de convergencia. Tesis. Instituto Tecnológico Autónomo de México.

160
