Changing connectivities and renewed priorities: Status and challenges facing Nepali Internet
First Monday

Changing connectivities and renewed priorities: Status and challenges facing Nepali Internet by Shailesh Pandey and Nischal Regmi



Abstract
Evidence available after the devastating April 2015 Nepal earthquake (Gorkha earthquake) illustrates uneven coverage and poor data consumption in Nepal in spite of impressive mobile Internet subscription numbers. Places with favourable terrain, higher population densities, and higher income have better connectivity. Online activity levels, on the other hand, do not always correspond with these factors. Overall, ownership of digital technologies and its use exhibit clear regional unevenness and a large urban-rural inequality. These geographical factors reflect differences in socio-demographic characteristics. Unfortunately, in Nepal, dominant discourses on the Internet brush aside these linkages. With deep structural inequalities, a resource-scarce economy, and a track record of poor governance, broadband connectivity will not reduce this development chasm. This paper calls for Nepali Internet discourses to be grounded in reality, detaching from a ‘self-evident’ development vision of connectivity.

Contents

1. Introduction
2. Unevenness of mobile Internet
3. Unevenness of ICT facilities
4. Discussion
5. Conclusion

 


 

1. Introduction

A magnitude (Mw) 7.8 (local magnitude [ML] 7.6) devastating earthquake struck central Nepal on 25 April 2015 with an epicentre 85 kilometres northwest of the capital in the Barpak village of Gorkha district. Two aftershocks larger than Mw 6 followed soon after. The most powerful aftershock struck on 12 May 2015 with Mw 7.3. Overall, there have been 483 aftershocks with local magnitudes larger than four since April 2015. [1] Luckily the physical ICT infrastructure managed to escape with minimal damage. Data beyond the basic subscription numbers were also revealed in infrastructure damage assessments. This information allowed us to examine online activity at a district/city level and investigate the state of connectivity in Nepal.

The disaster was a blunt reminder of the importance of ICT and mass media in post-disaster recovery. It demonstrated that ICTs cannot be effective when there is poor preparation and the state itself exhibited “amnesia” in retrieving data (Raj and Gautam, 2015). The ICT in operation was a synergy of mobile telecommunication, radio (FM) broadcast, and the Internet. Radio stations were crucial in informing the public and particularly important in dispelling rumours. Rumours can circulate at an astonishing rate and volume thanks to social media, generating panic. The government largely used Twitter and other social media to inform the public, effectively in parallel to radio broadcasts. Overall, the disaster reminded that a pragmatic synergy of ICT and mass media in a disaster has to be an immediate priority. It is a more beneficial investment than the pursuit of a muddled vision of a ‘knowledge society’. [2]

The disaster served as a stark reminder of the fragility of Nepal’s ICT infrastructure. Thanks to a collapse of the energy infrastructure, a lack of electricity disrupted telecoms and ISPs in the absence of power for towers and networks. Diesel generators were used in some cases as backup. Our previous research found that the energy consumed by Nepali telecommunications in the last 15 years is quickly catching up with that utilized by the transportation sector (Regmi and Pandey, 2015). Unfortunately, the energy implications of an Internet-connected Nepal are largely misunderstood.

Ensuring universal connectivity has been a focus of Nepal’s information technology (IT) vision since 2000. The private sector established a Nepali Internet in the mid-1990s. These private ISPs largely relied on the telecommunication infrastructure of Nepal Telecom, the incumbent operator during that period. Only a few ISPs built their own independent networks. By 2017, many private ISPs have developed their own physical infrastructures but are still small compared to telecoms. They all compete within a few profitable cities, mainly inside the three main cities of the Kathmandu valley: Kathmandu, Lalitpur, and Baktapur.

Taking the Internet into Nepali rural areas has been a challenge. A universal service fund, created to provide broad telephone access, is being used to build and extend fibre-based networks. This fund was inactive since its creation in 2000. In March 2017 US$19.5 million was awarded to build an optical fibre network in two of the seven provinces. Almost an equal amount in 2016 was awarded to place optical fibre along 2,376 kilometres of the Mid-Hill Highway that connects 32 districts.

These large-scale projects are on the table to build a Nepali information infrastructure. The expectation is that broadband connectivity will enable Nepal to leapfrog to higher levels of developments by transitioning into a “knowledge based society and economy.” Realizing the potential of the Internet, there has also been individual and community efforts to bring Internet access to rural and underserved parts of the country. The Nepal Wireless Networking Project, for example, is a well-known effort targeting remote villages. A local initiative brought wireless Internet access to Tangting in 2016. Tangting is a remote village northeast of Pokhara, one of the largest Nepali cities. However, there is a lack of synergy to extend connectivity.

Technology has always been at the centre of infrastructure development. Landline and dial-up Internet was the focus in the early late nineties. Mobile telephony and broadband now have taken their place. Subscription numbers published by the telecom regulator illustrate both the achievements and challenges to these technological visions. [3] Fixed telephone subscription is just above three percent of the population in 2017 (Nepal Telecommunications Authority, 2017a). Meanwhile, there are 124 mobile subscriptions per 100 inhabitants. Ninety-six percent of total Internet subscribers are mobile Internet subscribers; however data, other than subscriptions, are not readily available.

Mobile subscriptions are based on counts of active SIM cards, and are inherently problematic. It does not reveal whether a given individual ever uses the service. An individual can own multiple SIMs without owning a mobile phone, yielding an overestimate of the number of devices in use. Ownership of basic information and communication technologies (ICTs) can be obtained from a population census and a few surveys which regard them as essential household amenities. This paper investigates district-level diffusion of ICTs using these sorts of publicly available data.

The paper’s concern is therefore twofold. Since the mobile Internet is central to availing government services, the first concern is to understand connectivity and data consumption. There are no readily available data beside subscription figures. This paper maps online activity at regional, zonal, and district levels to tweet volume. In doing so, the paper contributes a wider insight into Nepal’s mobile Internet landscape beside mere subscription statistics.

The second concern is to understand Nepal’s household ownership of ICTs, specifically the fixed Internet. It focuses on district-level analysis, using data from the last (2011) population census, examining associations between ownership and socio-demographic factors. The third Nepal living standards survey (NLSS-III) was also completed in 2011. It provides some insights into technological adoptions at a household level.

It is necessary to understand discourses that have shaped policies determining the trajectory of the Nepali Internet. With some insight into technological use, we can understand who benefits and who is left out with alterations in digital connectivities. This paper ends with a discussion on larger discourses that continue to influence an emphasis on connectivity.

 

++++++++++

2. Unevenness of mobile Internet

According to published reports, in 2009 mobile Internet subscriptions nudged ahead of fixed connections (Nepal Telecommunications Authority, 2009). Private ISPs account for less than two percent of Internet subscriptions in 2017 (Nepal Telecommunications Authority, 2017a). With subscription numbers at 58 percent of the total population, new Internet users are primarily obtaining access by mobile phones.

Unsurprisingly, mobile phones and mobile Internet are at the forefront of online service delivery plans. Consider the Nepali government ICT policy in 2015 [4] as an example. It sets overarching agenda to ensure by 2020 there is access to 80 percent of citizens to online services as well as some form of Internet access for the entire population (Government of Nepal, 2015).

Amidst the rapid growth in mobile telephony subscriptions, local surveys have found that individuals are discouraged from using the Internet on their mobile devices. Smartphones make up 32 percent of mobile phones according to the 2014 survey conducted by Martin Chautari (Pandey and Raj, 2016). Internews (2014) similarly puts this number at 38 percent. Both of these surveys place respondent mobile phone ownership at around 72 percent. Official figures (Nepal Telecommunications Authority, 2014) show 87 percent of the population have subscribed to voice services in 2014. Much of the difference is explained by choice of measurement, ownership versus subscription.

We demonstrate unevenness associated with the mobile Internet in the following parts of this paper. We found locations associated with higher tweet volume corresponded to areas with better 3G site cell coverage. Similarly, places in a favourable terrain, higher population density, and higher income levels have better 3G coverage. However, not all of these locations yield a larger volume of tweets. Sluggish growth in fixed Internet subscriptions suggests new users primarily access the Internet through a mobile telecommunication network. We show a rise in mobile Internet subscription has not translated to impressive data consumption.

2.1. Uneven online activity

The GSM Association (2015), an organization representing mobile network operators, reported that 16.75 percent of 2G and 23.23 percent of 3G sites were down after the 2015 earthquake. Two days later, 17 percent of 2G and two percent of 3G sites were not available, reflecting almost normal operations (13 percent 2G and one percent 3G were reported on 20 April). Lack of energy (commercial power and back-up fuel) as well as damage to roads were major challenges (Khanal, 2015). Within five days most of the cell sites were available, although with some power problems. Overall the network managed to escape with minimal damage.

The Nepali government perceived Twitter as an effective way to inform the public. Nepal Police set up their Twitter account on 26 April. Similarly, the National Emergency Operation Centre (NEOC), under the Ministry of Home Affairs, established their Twitter account and issued official statistics immediately. Telecom operators recommended using SMS and the Internet due to high congestion in voice calls (Khanal, 2015). Minimal damage to the network meant social media was buzzing. Hashtags and new accounts were set up on Twitter to communicate local situations and consequently to coordinate relief efforts.

To understand the use of the Internet from mobile phones we examined tweet statistics from 24 April to May 11 in both 2015 and 2017. StatCounter report Twitter’s [5] popularity was three times higher during April-May 2015 compared to 2017. We choose 2015 because we expect those who used Twitter to be active during this period due to the severity and duration of the disaster. We choose 2017 as a baseline representing an average activity period for comparison.

 

Percentage of tweets originating from 14 zones
 
Figure 1: Percentage of tweets originating from 14 zones (Data source: Centre for Geographic Analysis, Harvard University [6]).

 

Most affected districts were Gorkha, Lamjung in Gandaki Zone and Kavre, Sindhupalchowk, Dhading, Nuwakot, Rasuwa, Dolakha, and Kathmandu valley in the Bagmati zone (Ministry of Home Affairs [Nepal], 2015). Figure 1 provides a graphic description of the volume of tweets at a zonal level. Note that 60 percent of total tweets originate from the Bagmati zone. Kathmandu alone commands 41 percent of tweets. [7] Second, and more interestingly, the proportion of tweets is almost the same in 2017 for the top two. [8]

The relative ordering of the zones was largely maintained in both years. It was visually apparent and reflected in a high correlation score of 0.98. A majority of active users were unevenly located in a few districts, many were in cities of the Kathmandu valley.

Mountainous sites in the Sagarmatha zone usually contribute noticeable Twitter activity from trekkers and mountain climbers, explaining the dip in 2015. When we look at the zones, we found 16 districts featured lower Twitter activity. Twelve of these were from the midwestern and far-western development regions, all noted for poor 2G and almost nonexistent 3G infrastructure. Some combination of lower income, lower population density, and difficult terrain characterizes these locations.

Another way to gauge online activity was to look at search queries in Google Search. Search volume for ‘Nepal earthquake’ peaked on 26 April (day after the main quake) and again on 12 May (day of a major aftershock); see Figure 2. Kathmandu (Bagmati), Biratnagar (Koshi) and Pokhara (Gandaki) were the origin of most searches. These locations were also the top three locations for tweets.

 

Search volume from Google Trends for 2015 Nepal Earthquake, 15 July 2017
 
Figure 2: Search volume from Google Trends [9] for ‘2015 Nepal earthquake’ (15 July 2017). A score of 100 is assigned to a location with the most popularity. Biratnagar score of 74 is then interpreted as three-forth as popular compared to Kathmandu. Locations with less than one percent popularity as the peak were assigned a zero value.

 

Within Nepal the volume of search for the term ‘Nepal’ in the last five years peaked in April 2015. In the last 12 months (June 2016 to June 2017) search volume was disproportionately associated with Kathmandu valley. All of the first 10 cities are in the valley. Places outside the top 10 included Biratnagar (Koshi), Dharan (Koshi), Birgunj (Narayani), Nepalgunj (Bheri), Bharatpur (Narayani), Pokhara (Gandaki), and Butwal (Lumbini). The top locations matched with the relative ordering based on tweet proportions.

Poor 3G coverage provides one explanation for lower activity online. However, it was not sufficient for determining the level of activity. Terrain, population density, and income provide a fuller justification for decreased activity.

2.2. Uneven coverage

A few national level statistics surfaced during the 2015 disaster which were not publicly available. Figure 3 illustrates the distribution of 2G and 3G sites.

 

Distribution of 2G (brown) and 3G (pink) sites
 
Figure 3a: Distribution of 2G (brown) and 3G (pink) sites. Source: GSM Association (2015), modified to include district and development region boundaries.
 
2G sites on 20 April 2015
3G sites on 20 April 2015
Figure 3b: 2G sites (top) and 3G sites (bottom) as on 20 April 2015. Thirteen percent of 2G was down while only one percent of 3G was not available. Source: From the presentation by a telecom regulator (Khanal, 2015).

 

There are three straightforward observations. First, telecom services elude a large number of districts in the midwestern and far-western regions, based on an absence of Twitter activity. Secondly, available service in these areas is largely 2G. Finally, mountain districts in the north, obviously thanks to rugged terrain, have the worst service.

Three G cell coverage agreed to the proportion of tweets at the district level. A better picture of service availability can be seen taking population density and terrain into consideration. Three G sites (pink dots) are few even in a populous district like Kailali despite a favourable terrain. Though Kailali and Kaski have similar population densities, around 240 people per square kilometre, Kailali has a lower per capita income (KailaliGNI, ppp: 942 and KaskiGNI, ppp: 1561) [10] below the national average of 1,080.

More voice (2G) and less data (3G) characterizes Kailali which has a large population and a favourable terrain, but low income. Unsurprisingly, Kaski featured nine times more tweets than Kailali. Bhaktapur, a city in the Kathmandu valley, was ranked eighth in tweet volume. Two other districts in the valley occupied the top two positions. This was despite good 3G coverage, second highest population density, and higher than average GNI of 1,379. Parsa has higher than average GNI, population density, and favourable terrain, but was ranked thirtieth in tweets, suggesting a socio-demographic influence on mobile Internet usage.

2.3. Low data consumption

At the end of 2015, Nepal Telecom had a mobile subscriber base equal to 48 percent of the population (Nepal Telecommunications Authority, 2017b). Major competitor Ncell had 53 percent. Seventy percent of the voice subscribers could use data services. Table 1 provides a breakdown of subscriber numbers and earnings from telecom services.

 

Table 1: Consumption of telecommunication services in 2015.
Source: Nepal Telecommunications Authority (2017b).
   Earnings per subscriber per month
CompanyMobile subscribersData subscribersVoice
US$ (rupees)
Data
US$ (rupees)
Nepal Telecom12,799,1388,945,8652.31 (227)0.45 (45)
Ncell14,145,7589,887,0752.83 (279)0.69 (68)

 

On average, a Nepal Telecom data subscriber uses 45 megabytes (MB) of data per month; a Ncell subscriber uses 68 megabytes. In 2015, one megabyte of data cost the equivalent of one rupee or one US cent [11]. In terms of today’s prices, this would give a Nepal Telecom subscriber 84 megabytes and a Ncell subscriber 127 megabytes. These numbers are low compared to global averages.

Monthly data consumption per smartphone was 1.5 gigabytes in 2015 and is assumed to rise as high as 8.9 GB by 2021 (Ericsson, 2016). Increasing average data consumption in three years is ambitious given a variety of factors described in this paper.

Based on monthly data use, monthly household expenditures for mobile phone bills would be equivalent to 1,036 rupees (or about US$10). Mobile phone bills would claim, on average, 3.5 percent share of monthly household income in a family with two phone owners (for rural residents, four percent; urban, three percent). [12]

Mobile Internet consumption is uneven, thanks to supply-side developments that lead to uneven geographical coverage of services. Skewed distribution of tweet volume across regional, district, and city levels also demonstrates uneven distribution of users. Kathmandu centricity exists in mobile data consumption despite impressive 60 percent data subscription.

 

++++++++++

3. Unevenness of ICT facilities

We found high inequality in household ownership of fixed Internet and other ICT technologies. Districts in midwestern and far-western regions and mountains in the north are the worst off in all technologies and electricity use. These districts do not provide a return on investment or generate sufficient profit for Internet service providers. Districts with better connectivity provide better value in terms of income, higher education, and electricity use. It suggests geographical differences reflect unevenness in socio-demographic characteristics, based on data from the national population and housing census (Central Bureau of Statistics [Nepal], 2012b) and the second part from the NLSS-III (Central Bureau of Statistics [Nepal], 2012a).

3.1. District level variation

We performed a district level cluster analysis for six household ICTs — fixed Internet, computer (desktop or laptop), mobile phone, television, cable television, and landline telephone. [13] See Figure 4 for the resulting cluster assignment. [14]

 

ICT clusters for Nepal's 75 districts
 
Figure 4: ICT clusters for Nepal’s 75 districts.

 

Three clusters corresponding to high (red), medium (yellow) and low (blue) penetration have pronounced differences between cluster averages; see Table 2. Just four out of the 75 districts are in the red or high cluster — Kathmandu, Bhaktapur, Lalitpur, and Kaski, with the first three in the Kathmandu valley. The Kathmandu and Pokhara valleys are major hubs of urbanization (Government of Nepal, 2017), illustrating the significance of urban-biased supply-side driven development.

 

Table 2: Average penetration of ICT technologies.
ICTTelevisionCable TVComputerInternetTelephoneMobile phone
Red725529142587
Yellow331741567
Blue11410243

 

Lack of appropriate infrastructure, such as electricity, characterizes the blue ICT cluster. Figure 5 provides a graphic illustration of clustering due to electricity penetration. Districts belonging to the lowest spectrum in electricity and ICT — blue clusters — overlap to a large extent. All of the lower 20 districts in electricity access belong to the blue cluster. Districts in the second lowest (green) electricity cluster also correlate to the blue cluster.

 

Electricity cluster assignment
 
Figure 5: Electricity cluster assignment. Districts are coloured red (87 percent), yellow (69 percent), green (46 percent), and blue (21 percent), from highest to lowest average values.

 

Districts in the midwestern and far-western regions and mountains in the north are not well off in technologies and electricity. Districts in the blue ICT cluster have the lowest values for higher education, income, population density, electricity access, adult literacy, and youth population (Table 3). For a more detailed description, see Table A1.

 

Table 3: Mean values in different ICT clusters.
ICT clusterHigher education (percent)Income, GNI PPP (US$)Population densityElectricity for lighting (percent)Adult literacy (percent)Youth population (percent)
Blue3.1848.7163.032.850.931.2
Yellow4.61185.4250.272.960.633.6
Red18.61899.52109.097.080.142.4

 

Blue areas do not provide a return on investment for Internet service providers (ISPs) and telecoms. Combination of low population density, lower income levels, and difficult terrain characterizes districts in the yellow cluster, which in turn does not translate into better subscription numbers.

Table 4 illustrates the percentage of households with ICT technologies organized into quintiles. The topmost quintile Internet penetration is double the national average. About nine out of 10 households with the Internet are in the topmost quintile (Figure 6).

A similar skew is seen in distribution of computers, with mobile phones an exception. However, the evenness is distorted when we consider types of handsets (smartphone v. feature phone) and mobile data consumption.

 

Table 4: Average household penetration of technologies across quintiles of districts.
TechnologyQuintilesNational average
12345
Internet0.20.40.61.27.83.3
Mobile35.149.760.067.781.764.6
Telephone1.32.23.34.815.67.4
TV4.312.924.633.656.536.4
Computer0.61.31.93.215.47.3

 

 

Quintile households with technology expressed as percentage of total households
 
Figure 6: Quintile households with technology expressed as percentage of total households.

 

Next, we constructed a decision tree for each technology to identify patterns that characterize clusters. First, districts with close penetration values were clustered together by the k-means algorithm as in ICT and electricity maps. The decision tree was built to label districts based on distinctive features.

Decisions at the nodes are on one of the 11 infrastructures and socio-demographic features listed in Table A1. The recursive partitioning (rpart) method then builds decision trees (Therneau and Atkinson, 2017). Rpart works by either splitting or not splitting each node into two nodes by asking a sequence of Boolean questions on socio-demographic features. Figure B1 in the Appendix includes the resulting decision trees.

Three districts make up the topmost red Internet cluster. Three cities in the Kathmandu valley account for most Internet households. At the opposite end, 50 districts in the blue cluster have one-sixteenth its penetration value. These districts also have the lowest level of electricity, with a low threshold of 68 percent separating them.

The threshold is lower still for ICT, narrowing down to the district with lowest values for socio-demographic variables. The telephone has a similar disproportionate cluster size as the Internet. Unlike the Internet landscape, higher education separates the bottom cluster from the top, with the threshold close to the national average.

Education and electricity were able to differentiate districts in the lower clusters and the top across all technologies, with the mobile phone no exception. Though the grading is less substantial, electricity is still the most important discriminant separating the lowest three clusters from the top two. Subscription statistics masks this tendency. Districts in the top clusters have a far higher share of subscription. Except for one district, the lowest three clusters for mobile phones maps to the blue ICT cluster.

3.2. Household level variation

Data analysed in this section is taken from NLSS-III (Central Bureau of Statistics [Nepal], 2012a). We limit our analysis because NLSS does not explicitly provide important indicators, like income and adult literacy; many other variables are simply not available.

 

Table 5: Association of household Internet with electricity and higher education [15].
Table 5a: Electricity access and fixed Internet connection.
Electricity as the main source of lightingInternet connection
YesNoYes (percent)
Yes2164,1874.90
No31,5850.19
 
Table 5b: Electricity bills quintiles (for households using electricity as the main lighting source).
Annual electricity bill (rupees)Internet connection
YesNoYes (percent)
≤60079110.76
601 to 1,00068960.67
1,001 to 1,800208862.21
1,801 to 3,860327644.02
>3,86015173017.14
 
Table 5c: Higher education and the Internet connection.
Higher educationInternet connection
YesNoYes (percent)
Yes1671,26411.67
No302,8331.05
 
Table 5d: Electricity bill, higher education, and the Internet.
HouseholdsElectricity bill (rupees)Higher education (percent)
With Internet10,428.9684.77
Without Internet2,479.1030.85

 

Ninety-six percent of households do not have a fixed Internet connection (Table 5a). Households that have Internet access almost exclusively use electricity as a source for lighting. A large proportion of households with connectivity are in the top quintile. The chance of a household with an Internet connection increases non-linearly with the amount of electricity it consumes (Table 5b). Connectivity further increases with higher educational attainment (Table 5c).

Affordability is a legitimate concern. The least expensive Internet connection (landline-based) costs 660 rupees (or US$6.60) per month in 2011. [16] Households without the Internet would have to spend three times their electricity bill to pay for the service.

Consumption of electricity or having someone with a higher education does not fully explain non-adoption (Table 5d). The majority of the top quintile households do not have an Internet connection. These include households that can afford an Internet connection and having someone with a higher education. This figure has not improved significantly in 2017. Less than eight percent of total households have a fixed Internet connection. [17]

There is also a disparity between rural and urban household Internet adoption. Less than one percent of households that rely solely on agriculture for their income have an Internet connection. This increases to nearly seven percent in households involved in enterprises. Household income matters, but so does experience with older ICT technologies. Around four percent of households have an Internet connection, jumping to 22 percent for households with both cable TV connections and telephones.

This observation supports recent moves by ISPs to bundle cable TV and Internet. Telephone-based Internet users and homes with cable TV but without the Internet in urban areas are the target of this strategy. Ownership of multiple technologies is strongly associated with overall household electricity bills and characteristic of households in the red ICT cluster.

There are several reasons to believe that the picture of household connectivity and use in Nepal has not changed in 2017. First, fixed Internet in March 2017 accounts for less than a three percent share of total Internet subscriptions (Nepal Telecommunications Authority, 2017a). The number has leveled off since 2011 when it was at 3.7 percent. Second, rural/urban disparities in device ownership and service adoption at the household level were very pronounced for ICTs in 2014 (Pandey and Raj, 2016; Internews, 2014). Lower income level and slow economic growth has become a fundamental bottleneck for Internet policies and current universal broadband ambitions (Regmi, 2017).

 

++++++++++

4. Discussion

4.1. Limited understanding of access

There is a need to clarify how ‘access’ to the Internet is being characterized. The large disparity in mobile Internet and fixed Internet subscriptions is one indication that new users primarily access the Internet from their mobile phones. But low mobile data consumption demonstrates the plausibility that disparity can run in the opposite direction as well.

A large volume of Internet traffic reaching mobile phones could be travelling through Wi-Fi at home, school, or office and therefore not registering on mobile networks. For those with regular access to the fixed Internet, it provides a way around expensive mobile data.

There could be a large consumption gap between Internet users with personal computer based Internet experiences and those with access based on mobile phones. Impressive subscription numbers hide inequality in bandwidth availability and consumption. Unlike subscriptions, bandwidth is a persistent dynamic target affected by the arrival of new technologies and their diffusion (Hilbert, 2014).

4.2. Centrality of electricity

Electricity should occupy a central position in discussions about Internet adoption. Less than one-third of the households in the blue ICT cluster in Nepal use electricity for lighting.

Household adoption of electricity can be used to identify districts in lower clusters with high accuracy. The ownership figures, however, do not improve much for some clusters despite greater use of electricity for lighting. The living standards survey further reveals on average households currently without the Internet would have to spend three times their electricity bill in order to secure the lease expensive Internet connections.

Three policy considerations emerge from these observations. First, a large number of rural households are not using electricity. Second, a significant proportion of households primarily use electricity for lighting. Finally, affordable Internet policies have to combine Internet tariffs and ever-increasing costs for electricity. Distinct geographies appear when we compare broadband Internet costs with household income. It is unlikely to disappear by merely having districts connected to a fibre-based backbone. However, the most critical factor that affects Nepal’s digital ambitions is a reliable and abundant supply of electricity.

In the fiscal year 2016/17, peak demand for electricity was 1,444.1 MW with Nepal’s total installed capacity of 967.85 MW (Nepal Electricity Authority, 2017). The production is almost halved in dry seasons (Subedi, 2017b). To reduce deficits, in addition to scheduled power cuts popularly known as ‘load shedding’, Nepal has been importing electricity from India. These imports have doubled in the past three years to about 250 MW in 2016/17 (Nepal Electricity Authority, 2017). In the past two years, there has been an apparent elimination of load shedding inside the Kathmandu valley by imposing more power cuts to industries (Himalayan Times, 2017; Subedi, 2017a) and increasing electrical imports from India.

India itself is facing chronic energy shortages. Nepal imports electricity from power plants located in the Indian state of Bihar (Nepal Electricity Authority, 2017). A survey suggests only 20 percent rural households in Bihar use electricity as their primary source for lighting. A lack of infrastructure prevents 43 percent of households in Bihar to be without electricity (Jain, et al., 2015). It is significant to note that the population of rural Bihar is three times larger than the entire population of Nepal. Expansion of an electrical infrastructure coupled with an increase in demand from households will mean less surplus available for Nepal from Bihar. A reliance on electricity imported from India is a temporary fix at best.

4.3. Dominant visions and discourses

Claims in some Internet policies are considered ‘self-evident’ despite mixed and inconclusive evidence. It is common to view technological fixes to problems of development as being geographically independent. That is why, “arguments about inevitable changes can be made without ever pinning down those arguments to specific places and contexts” (Graham, 2015).

These policies develop from imaginaries of the Internet that render issues of geography and physical distance irrelevant. Dominant discourses on connectivity call for immediate investments in broadband infrastructure or “lose the opportunity to reap the economic and social benefits that broadband brings” (ITU/UNESCO Broadband Commission for Sustainable Development, 2011). Such international discourses were crucial in the development of Nepali broadband policies.

These notions were stimulated with information from organisations like the World Bank that provided a quantitative basis to these discourses. For example, Qiang, et al. (2009) claimed 1.38 percentage points of economic growth with an increase of 10 percent penetration in broadband. The World Bank (2016) later admitted that the contribution of ICT for development is unobservable in countries that lack so-called “analog complements”: — favourable business environment, strong human capital, and good governance.

Scholarly research on the impact of ICT appears inconclusive as well (Friederici, et al., 2017). It has even been argued ICT can widen existing income inequality (Tyson and Spence, 2017). Jensen’s (2007) work on the use of mobile phones by the fishermen of Kerala has been strongly criticized for over-generalizations based on unrealistic premises and even suspect sources (Steyn, 2016). Another line of criticism found market dynamics of the Internet far more complicated to be captured by neoclassical economic models favoured by policy-makers (Alleman, et al., 2009).

4.4. Summary

Overall, there is a lack of clarity in framing Internet ‘access’ and empirical understandings of the Internet (Martin Chautari, 2014). There is a strong impetus for increasing connectivity, little on understanding ownership, adoption, and usage. The emphasis on infrastructure development is following the suit of countries who desire to leapfrog to higher levels of development by transitioning into a ‘knowledge-based society and economy.’ What is worrying is the dominant discourses on connectivity have adopted a ‘build and they will come’ approach without consideration of alternative outcomes. As Friederici, et al. (2017) noted, the worrying aspect of a ‘self-evident’ vision of connectivity is that it offers “a powerful, aspatial and ahistorical teleology.” This vision allows policy-makers to “point to new technological fixes instead of focusing on how the political economy of any given context works to allocate power and wealth.” It is therefore not surprising that very little is said on who benefits and who is left out from a reconfiguration of connectivities.

 

++++++++++

5. Conclusion

There is a geographic pattern to ICT ownership at the district level in Nepal. District clusters are most uneven in electricity and educational attainment. Electricity alone, and equally well by education, can distinguish the lowest cluster. Analysis of NLSS-III data further shows households without Internet access would have to spend three times for electricity in order to pay for that access. Electricity costs and fixed Internet tariffs are immediate deterrents in Nepal. Growth in mobile telephony renews interest and necessary enthusiasm for ambitious Internet policies. These policies lack clarity in visualizing ‘access’.

We demonstrated in this paper that a rise in mobile Internet subscriptions did not translate into data consumption. Locations with the best terrain, higher population density, and higher income have better 3G site coverage, similar to associations identified for other ICTs. Locations with high tweet volume corresponded to areas with better 3G sites. However, not all such locations produced a larger volume of tweets, indicating that there are socio-demographic factors influencing mobile Internet usage.

We call for discourses surrounding broadband Internet to detach from a simplistic ‘self-evident’ development vision of broadband connectivity. A natural disaster like the 2015 earthquake served to show that ICTs were effective. It is crucial to calculate who benefits and by how much to justify the use of limited resources in a relatively impoverished country. End of article

 

About the authors

Shailesh Pandey is a researcher at Martin Chautari in Kathmandu, Nepal. In his research, he is interested in the access and use of the Internet and related technologies.
E-mail: pandey [dot] shailesh [at] gmail [dot] com

Nischal Regmi is a researcher at Martin Chautari in Kathmandu, Nepal. In his research, he is interested in the access and use of the Internet and related technologies.
E-mail: nischalregmi108 [at] gmail [dot] com

 

Notes

1. The total is calculated up to 28 June 2017. See http://www.seismonepal.gov.np/index.php?listId=161, accessed 11 July 2017.

2. For a description, see, for example, Towards knowledge societies, Paris: UNESCO, at http://unesdoc.unesco.org/images/0014/001418/141843e.pdf, accessed 8 December 2017.

3. The MIS reports are available from http://www.nta.gov.np/en/mis-reports-en.

4. Information and communication technology policy, Kathmandu: Nepal Ministry of Information and Communication, at https://www.moic.gov.np/upload/documents/ict_policy_2072.pdf, accessed 22 April 2017.

5. http://gs.statcounter.com/social-media-stats/all/nepal/#monthly-201503-201504-bar, accessed 16 July 2017.

6. The 2015 map and data can be found at https://worldmap.harvard.edu/maps/nepalquake. The 2017 tweet data was obtained from the Billion Object Platform (BOP), located at http://bop.worldmap.harvard.edu/bop/, accessed 14 July 2017.

7. We could only secure zone level data from WorldMap data for 2017. Locations for the 2015 geocoded tweets were obtained using OpenStreetMap&rsquoi;s Nominatim (http://nominatim.openstreetmap.org. We can, however, zoom to visually locate the origin of tweets at district and city scales. The map is available from https://worldmap.harvard.edu/maps/nepalquake, accessed 15 July 2017.

8. Social Aves (https://socialaves.com) estimates total Twitter accounts in Nepal as 3.2 million in 2017, equating to 13.5 percent of the population. See https://socialaves.com/social-media-landscape-nepal/, accessed 12 July 2017.

9. https://trends.google.com/trends/, accessed 17 July 2017.

10. Gross national income (GNI) data from National Planning Commission (Nepal) and United Nations Development Programme, 2014. “Nepal Human Development Report 2014,” at http://www.npc.gov.np/images/category/NHDR_Report_20141.pdf, accessed 17 July 2017.

11. In 2015, a 300 MB data pack offered by Ncell was valid for a month and cost one rupee per MB. NTC also charged one rupee per MB for its 2G and 3G Internet. The tariff was obtained from the company’s Web site as described in January 2015.

12. Household income and expenditures based on the annual household budget survey prepared by the central bank of Nepal (Nepal Rastra Bank, 2015). Martin Chautari’s survey reports two mobile phones are in use on an average (Pandey and Raj, 2016).

13. The number of clusters has been selected using the d-index criteria for the k-means clustering method.

14. Red, orange, yellow, green, and blue colours indicate levels from high to low.

15. The total numbers of households in these tables are not exactly equal because of missing values for different variables.

16. The old tariff is from 21 February 2011, obtained with the Wayback Machine (https://archive.org/web/) and from the telecom Web site at https://www.ntc.net.np.

17. We divided fixed Internet subscription by the total number of households to get this estimate. The 2011 census reported there were 5,427,302 households in Nepal.

 

References

James Alleman, Jonathan Liebenau, and Paul Rappoport, 2009. “The new economics of ICT: The regulatory implications of post-neoclassical economics for the ICT sector,” Proceedings of the Third ACORN-REDECOM Conference (Mexico City), pp. 215–228; proceedings available at http://www.acorn-redecom.org/papers/ProceedingsAcornRedecom2009.pdf, accessed 11 December 2017.

Central Bureau of Statistics (Nepal), 2012a. “Nepal — Nepal Living Standards Survey 2010–2011, NLSS Third,” at http://cbs.gov.np/nada/index.php/catalog/37, accessed 6 March 2017.

Central Bureau of Statistics (Nepal), 2012b. “National Population and Housing Census 2011,” version at https://unstats.un.org/unsd/demographic/sources/census/wphc/Nepal/Nepal-Census-2011-Vol1.pdf, accessed 11 December 2017.

Ericsson, 2016. “Ericsson mobility report: On the pulse of the networked society,” at https://www.ericsson.com/res/docs/2016/ericsson-mobility-report-2016.pdf, accessed 11 December 2017.

Nicolas Friederici, Sanna Ojanperä, and Mark Graham, 2017. “The impact of connectivity in Africa: Grand visions and the mirage of inclusive digital development,” Electronic Journal of Information Systems in Developing Countries, volume 79, number 1, pp. 1–20.
doi: https://doi.org/10.1002/j.1681-4835.2017.tb00578.x, accessed 11 December 2017.

Government of Nepal, 2017. “National urban development strategy 2017,” at http://moud.gov.np/wp-content/uploads/2017/02/NUDS_cabinet-document_approved-on-2073-10-9.pdf, accessed 11 December 2017.

Government of Nepal, 2015. “National information and communication technology policy, 2015,” at http://www.youthmetro.org/uploads/4/7/6/5/47654969/ict_policy_nepal.pdf, accessed 26 April 2017.

Mark Graham, 2015. “Contradictory connectivity: Spatial imaginaries and technomediated positionalities in Kenya’s outsourcing sector,” Environment and Planning A, volume 47, number 4, pp. 867–883.
doi: https://doi.org/10.1068/a140275p, accessed 11 December 2017.

GSM Association, 2015. “Disaster response — Nepal earthquake response and recovery overview” (3 December), at https://www.gsma.com/mobilefordevelopment/programme/disaster-response/disaster-response-nepal-earthquake-response-and-recovery-overview, accessed 11 December 2017.

Martin Hilbert, 2014. “How much of the global information and communication explosion is driven by more, and how much by better technology?” Journal of the Association for Information Science and Technology, volume 65, number 4, pp. 856–861.
doi: https://doi.org/10.1002/asi.23031, accessed 11 December 2017.

Himalayan Times, 2017. “Load shedding hits Udayapur cement industry hard” (17 July), at https://thehimalayantimes.com/nepal/load-shedding-hits-udayapur-cement-industry-hard/, accessed 25 November 2017.

International Telecommunication Union (ITU)/UNESCO Broadband Commission for Sustainable Development, 2011. “Broadband: A platform for progress,” at http://www.broadbandcommission.org/Documents/publications/Report_2.pdf, accessed 11 December 2017.

Internews, 2014. “Nepal: Media survey findings,” at https://www.internews.org/resource/nepal-media-survey-findings, accessed 30 June 2017.

Abhishek Jain, Sudatta Ray, Karthik Ganesan, Michaël Aklin, Chao-yo Cheng, and Johannes Urpelainen, 2015. “Access to clean cooking energy and electricity: Survey of states,” at http://ceew.in/pdf/CEEW-ACCESS-Report-29Sep15.pdf, accessed 11 December 2017.

Robert Jensen, 2007. “The digital provide: Information (technology), market performance, and welfare in the South Indian fisheries sector,” Quarterly Journal of Economics, volume 122, number 3, pp. 879–924.
doi: https://doi.org/10.1162/qjec.122.3.879, accessed 11 December 2017.

Ananda Raj Khanal, 2015. “Nepal’s experience in responding to a disaster: A telecommunication/ICT sector perspective,” at https://www.itu.int/en/ITU-D/Regional-Presence/AsiaPacific/Documents/Events/2015/August-RDF2015/Session-4/S4_Ananda_Raj_Khanal.pptx, accessed 12 July 2017.

Martin Chautari, 2014. “Universal connectivity in Nepal: A policy review,” Research Brief, number 12, at http://www.martinchautari.org.np/files/ResearchBrief12-UniversalConnectivityInNepal_APolicyReview.pdf, accessed 12 July 2017.

Ministry of Home Affairs (Nepal), 2015. “Nepal earthquake 2072: Situation update as of 11th May,” at http://drrportal.gov.np/uploads/document/14.pdf, accessed 13 July 2017.

National Planning Commission (Nepal) and United Nations Development Programme, 2014. “Nepal Human Development Report 2014,” at http://www.npc.gov.np/images/category/NHDR_Report_20141.pdf, accessed 6 March 2017.

Nepal Electricity Authority, 2017. “Nepal Electricity Authority: A year in review,” at http://www.nea.org.np/admin/assets/uploads/supportive_docs/87757284.pdf, accessed 20 November 2017.

Nepal Rastra Bank, 2015. “Fifth household budget survey, 2014/15,” at https://www.nrb.org.np/red/publications/study_reports/Study_Reports--Fifth_Household_Budget_Survey_2014-2015.pdf, accessed 22 April 2017.

Nepal Telecommunications Authority, 2017a. “MIS Report: Falgun, 2073 (12, February 2017 — 13, March 2017),” at http://www.nta.gov.np/en/2012-06-01-11-33-01/mis-archives/mis-reports/nta-mis-120/download, accessed 29 June 2017.

Nepal Telecommunications Authority, 2017b. “Nepal Telecommunications Authority — Annual report,” at http://nta.gov.np/en/public-notice-en/563-annual-report-of-nta-year-2072-73, accessed 22 April 2017.

Nepal Telecommunications Authority, 2014. “MIS Report: Mangshir, 2071 (17 November–15 December 2014),” at http://www.nta.gov.np/en/2012-06-01-11-33-01/mis-archives/mis-reports/nta-mis-93/download, accessed 10 July 2017.

Nepal Telecommunications Authority, 2009. “Management Information System, Baisakha, 2066 (May, 2009),” at http://www.nta.gov.np/en/2012-06-01-11-33-01/mis-archives/mis-reports/nta-mis-25/download, accessed 10 July 2017.

Shailesh Pandey and Yogesh Raj, 2016. “Free float Internet policies of Nepal,” Studies in Nepali History and Society, volume 21, number 1, pp. 1–60, and at http://www.martinchautari.org.np/files/SINHAS-Articles/SINHAS-Vol.21-No.1_Shailesh-Pandey-and-Yogesh-Raj.pdf, accessed 11 December 2017.

Christine Zhen-Wei Qiang and Carlo M. Rossotto, with Kaoru Kimura, 2009. “Economic impacts of broadband,” In: Information and communications for development 2009: Extending reach and increasing impact. Washington, D.C.: World Bank, pp. 35–50, and at http://siteresources.worldbank.org/EXTIC4D/Resources/IC4D_Broadband_35_50.pdf, accessed 11 December 2017.

Yogesh Raj and Bhaskar Gautam, 2015. Courage in chaos: Early rescue and relief after the April Earthquake. Kathmandu: Martin Chautari.

Nischal Regmi, 2017. “Expectation versus reality: A case of Internet in Nepal,” Electronic Journal of Information System in Developing Countries, volume 82, number 7, pp. 1–20, and at http://www.is.cityu.edu.hk/staff/isrobert/ejisdc/82-7.pdf, accessed 11 December 2017.

Nischal Regmi and Shailesh Pandey, 2015. “A regression analysis into Nepali ICT’s energy consumption and its implications,” Proceedings of the 2015 Ninth International Conference on Software, Knowledge, Information Management and Applications (SKIMA).
doi: https://doi.org/10.1109/SKIMA.2015.7400034, accessed 11 December 2017.

Jacques Steyn, 2016. “A critique of the claims about mobile phones and Kerala fisherman. The importance of the context of complex social systems,” Electronic Journal of Information Systems in Developing Countries, volume 74, at http://144.214.55.140/ojs2/index.php/ejisdc/article/view/1687, accessed 19 March 2017.

Bibek Subedi, 2017a. “NEA slashes industrial loadshedding by an hour,” Kathmandu Post (16 July), at http://kathmandupost.ekantipur.com/news/2017-07-16/nea-slashes-industrial-loadshedding-by-an-hour.html, accessed 25 November 2017.

Bibek Subedi, 2017b. “Power output falls 60pc as river flows shrink,” Kathmandu Post (24 February), at http://kathmandupost.ekantipur.com/news/2017-02-24/power-output-falls-60pc-as-river-flows-shrink.html, accessed 30 June 2017.

Terry M. Therneau and Elizabeth J. Atkinson, 2017. “An introduction to recursive partitioning using the RPART routines” (12 March), at https://cran.r-project.org/web/packages/rpart/vignettes/longintro.pdf, accessed 11 December 2017.

Laura Tyson and Michael Spence, 2017. “Exploring the effects of technology on income and wealth inequality,” In: Heather Boushey, J. Bradford DeLong and Marshall Steinbaum (editors). After Piketty: The agenda for economics and inequality. Cambridge, Mass.: Harvard University Press.

World Bank, 2016. “World development report 2016: Digital dividends,” at http://documents.worldbank.org/curated/en/896971468194972881/pdf/102725-PUB-Replacement-PUBLIC.pdf, accessed 5 December 2016.

 

Appendix A: List of infrastructure and socio-demographic features.

 

Table A1: Features used to construct decision trees.
VariableDescriptionData source
ICT penetration variables
Internet, Computer, Mobile, Telephone, Television, and Cable televisionPercentage of households in a district having the corresponding facilitiesCBS, 2011
Demographic variables
Higher educationPercentage of population with education level at least School Leaving Certificate (tenth grade)CBS, 2011
Electricity accessPercentage of household in a district that use electricity as main source of lightingCBS, 2011
IncomeGNI (ppp) per capita at the district levelUNDP, 2014
Richest neighbourGNI (ppp) per capita of the richest neighbouring district [calculations of authors]Based on GNI data provided in UNDP, 2014
Population densityPopulation density of a districtCBSN, 2011
Adult literacyAdult literacy percentageUNDP 2011
Best ICTBest overall ICT score in connected districts. Score is an average of TV, CABLE, COMPUTER, INTERNET, TELEPHONE and MOBILE percentagesCBSN, 2011
Youth populationPercentage of population aged from 15 to 34CBSN, 2011
Language entropyIndicates the homogeneity of mother tongues in a district [calculations of authors]Based on CBSN, 2011; Yadava, 2003
Ecological beltEcological belt for a given district; a variable equivalent to ‘tarai’ (plain), ‘hill’, or ‘mountain’CBSN, 2011
Development regionOne of five administrative groups for districts (eastern, central, western, mid-western, or far-western)CBSN, 2011

 

Appendix B: Decision trees for four ICTs.

ICT
(a) ICT (Accuracy: 92 percent)
 Internet
(b) Internet (Accuracy: 96 percent)
Mobile
(c) Mobile (Accuracy: 70.6 percent)
 Cable TV
(d) Cable TV (Accuracy: 85.33 percent)
Telephone
(e) Telephone (98.6 percent)
  

 

Mean and median values.
VariableMeanMedian
Electricity57.164.2
Higher education4.73.5
Adult literacy57.558.2
GNI1079.81007.0

 

 


Editorial history

Received 25 August 2017; revised 27 November 2017; accepted 8 December 2017.


Creative Commons License
This paper is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

Changing connectivities and renewed priorities: Status and challenges facing Nepali Internet
by Shailesh Pandey and Nischal Regmi.
First Monday, Volume 23, Number 1 - 1 January 2018
https://journals.uic.edu/ojs/index.php/fm/article/view/8071/6613
doi: http://dx.doi.org/10.5210/fm.v23i1.8071





A Great Cities Initiative of the University of Illinois at Chicago University Library.

© First Monday, 1995-2018. ISSN 1396-0466.