While the literature spanning ICT and its effects on growth is now considerably large, this paper specifically focuses on one aspect of ICT – telecommunications and its relationship to growth. Worldwide, the telecom sector has been deregulated, and market structures in this sector have become highly competitive. Because of this, prices of telecom services have decreased everywhere throughout the developing world, creating something close to a revolution in growth of telecommunication services. Simultaneously, telecom technology has also leapfrogged into second and higher generation cellular mobile systems. But developing countries still continue to be quite poor, and the digital divide is much larger than the income divide between the developed and developing world, as pointed out by Wong (2002). The intriguing question that has cropped up is how technology can be used to decrease the cost of access of providing telecom services in rural and remote parts of the developing world. If yes, how can the provision of telecom services affect growth more generally.

There are empirical investigations [Röller and Waverman 2001] that study how telecommunications infrastructure affects economic growth in developed OECD economies, taking into account the two-way causation between them. These relationships have not been studied in the context of developing economies.

Our objective is to develop a causal model that analyses the effect of telecom penetration on economic growth in developing economies, taking into account the two-way causation that exists between them. Based on the research, the contribution of telecommunication services towards growth can be used as benchmark to gain insights into developing policies for diffusion of these services in developing countries.

The paper addresses the following questions to understand the dynamics of this causal connection, i e, is it telecommunication services that accelerates economic growth or overall economic growth that creates the demand for more telecommunication services for their growth to occur? In the context of developing economies, what are the factors that determine demand for and supply of telecom services? Finally, given the importance of telecom infrastructure in growth, what determines the growth in telecom penetration in these economies?

The following section summarises the literature on the subject. The section following the literature survey describes the causal model. We present select quantitative and qualitative evidence from a few projects in developing countries to understand the nature of the impact telecommunications has on their economy and society. The final section discusses data limitations in empirical modelling, summarises the policy implications, and concludes.

The literature investigates the feasibility of telecommunication as one of the determinants of growth, and attempts to entangle the reverse causality between growth and the demand for telecommunication services.

The past literature clearly highlights how most infrastructure investments, including telecommunications, can favourably influence the economy in several ways. First, they reduce the cost of production and increase revenues, for reinvestment by firms. Productivity increases, made possible with the use of telephones (e g, physically dispersed activity possible), increase the productivity of all industries. There is some recent literature that shows that the internet has changed the markets by allowing more efficient search. Similar to other infrastructure investments, investing in telecommunication will increase the demand for the goods and services used in their production and increase total national output. Such investments can increase employment through both direct and indirect effects [Alleman et al 2002].

When compared with other traditional infrastructure, however, there are grounds to expect that the effect of telecommunication services on growth will be somewhat more pronounced. Telecommunications is a little different because of the existence of network externalities, a phenomenon that increases the value of a service with increase in the number of users. Kim et al (1997) demonstrates this phenomenon in their analysis of online service competition.

The impact of telecommunications on growth was first found by Hardy (1980) based on data from 45 countries, with the largest effect of telecommunication investment on GDP found in the least developed economies and the smallest effect, in the mostdeveloped economies.

Garbade and Silber (1978) found that the telegraph and (Trans-Atlantic) cable led to efficient markets by narrowing inter market price differentials. Interesting research by Bayes et al (1999) found that half of all telephone calls involved economic purposes such as discussing employment opportunities, prices of the commodities, land transactions, remittances and other business items. Bayes et al also noted that, the average prices of agricultural commodities were higher in villages with phones than in villages without phones. Leff (1984) argues that firms can have more physically dispersed activity with increased telecom services (in today’s context, this could even mean employee telecommuting) and enjoy economy of scale and scope.3 Using the Peterson Index, Cronin et al (1993b) find a statistically significant causal relationship between productivity growth and the portion attributable to telecommunications.

Eggleston et al (2002) show how basic telecommunication infrastructure can create a “digital provide” by making markets efficient through information dissemination to isolated local residents and improve the living standards of the world’s poor, which in turn accelerates growth. As Eggleston et al (2002) point out, their analysis is based on references and examples, more careful analysis is needed in the context of developing countries.

Common sense suggests that increases in purchasing power (contributed by increased telecom services) also increase demand for such services. Chatterjee et al (1998) point out that income patterns decide the disposable income levels, i e, purchasing power for telecommunication services, and in turn the growth of services.

The reverse causality between telecom services and growth has also been investigated by Cronin et al (1991). Roller and Waverman (2001), in their econometric model, used a simultaneous approach to validate the hypothesis of reverse causality in the context of data for OECD countries. Sridhar and Sridhar (2004) study the impact of mobile phones on growth in developing countries. Waverman, Meschi and Fuss (2005) study

Table 1: Network Readiness Index Ranking of DevelopingTable 1: Network Readiness Index Ranking of DevelopingTable 1: Network Readiness Index Ranking of DevelopingTable 1: Network Readiness Index Ranking of DevelopingTable 1: Network Readiness Index Ranking of Developing

EconomiesEconomiesEconomiesEconomiesEconomies

(1=high; 102=low)

Country Network Readiness Index

Angola 99 Bangladesh 93 Cameroon 83 Ethiopia 101 Ghana 74 Gambia 82 Haiti 100 Indonesia 73 India 45 Kenya 84 Madagascar 92 Mali 96 Mozambique 97 Malawi 88 Nigeria 79 Nicaragua 94 Pakistan 76 Senegal 81 Chad 102 Tanzania 71 Uganda 80 Ukraine 78 Zambia 85 Zimbabwe 95

Source: World Economic Forum, 2003.

the impact of mobile telephony in developed and developing economies.

In this paper, we examine the relationship between telecom infrastructure and economic growth in developing economies, as these countries can use diffusion of telephones for spreading growth more rapidly. In developing countries rural teledensity is very low. One of the reasons is the high cost of providing telecommunication services in rural areas and low purchasing power of rural population. While in developed countries, 90 per cent of the households can afford monthly expenditure of US$30 on telecommunication services, only 5-6 per cent of the households can afford such expenditure in developing countries such as India [Jhunjhunwala 2000].

One way to improve rural teledensity is to reduce the cost of access loop for providing telecom services using wireless technologies [Jain and Sridhar 2003].4 Jha and Majumdar (1999) also note that for developing countries, where penetration rates of telephones are extremely low, catching up with developed countries in terms of telecom infrastructure has meant investment in wireless and cellular mobile systems, bypassing investment in fixed landlines. Reduced per line cost, quick deployment and better available technology are reasons for such growth of cellular services in developing countries [Jain and Sridhar 2003; Jha and Majumdar 1999]. Further, it is easy to see why mobile phones could have a great impact in developing countries, they do not rely on a permanent source of electricity supply.

Thus, there are a number of issues that are relevant to be considered exclusively in the context of developing countries. This provides the motivation for us to more comprehensively model the growth of telecommunication services and investigate the strength of its causal relationship with the level of economic growth, and examine how telecommunications can be used as a tool to enable growth in developing countries.

Calculations from the data from developing countries show that in these economies as a whole, the compounded annual growth rate (CAGR) of cellphones over the period 1996-2001 was 78 per cent compared to a growth of mere 7 per cent for main telephone lines over 1990-2001 (Table 2). Further, we find simple Pearson’s correlation coefficients between GDP per capita and total, main and cellphone penetration to be respectively 0.59, 0.58 and 0.24 (all statistically significant). Based on this, one can expect to find substantial effects of telecom penetration on GDP.

Below we discuss our approach and the model that can be potentially used to empirically model such relationships.

We develop a structural model that traverses from the microlevel demand for and supply of telecom services to aggregate changes in telecom penetration and the macro production function in which GDP is determined by traditional inputs including capital (net of telecom), and labour stock, along with telephones. We do not attempt to provide a general explanation of the determinants of national output. This means we have not included measures of government deficits or of trade openness that, the literature shows, affect national output. Rather, we use a macro production function approach that relates inputs to output. We endogenise telecom investment. The causal model we develop demonstrates various relationships between telecom penetration and growth, factors that determine the demand for and supply of telecom services, and those that influence the change in telecom penetration.

While this model has been developed by Roller and Waverman (2001), we view the contribution of the model in this paper as being two-fold. First, in modelling these relationships, we

Table 2: Annual Growth of GDP Per Capita, Main TelephoneTable 2: Annual Growth of GDP Per Capita, Main TelephoneTable 2: Annual Growth of GDP Per Capita, Main TelephoneTable 2: Annual Growth of GDP Per Capita, Main TelephoneTable 2: Annual Growth of GDP Per Capita, Main Telephone

Lines and Cellphone Penetration in Developing EconomiesLines and Cellphone Penetration in Developing EconomiesLines and Cellphone Penetration in Developing EconomiesLines and Cellphone Penetration in Developing EconomiesLines and Cellphone Penetration in Developing Economies

Country CAGR, 1990-CAGR, 1990-2001, CAGR, 1996-2001, 2001, GDP Mainlines per Cellphones per Per Capita 100 Inhabitants 100 Inhabitants

Afghanistan NA -4.29 NA Angola -2.29 -2.09 66.54 Armenia -2.95 -0.96 101.53 Azerbaijan -2.33 2.10 81.94 Burundi -1.19 0.88 NA Benin 1.54 9.49 83.99 Burkina Faso 2.06 8.70 100.00 Bangladesh 2.75 5.74 NA Bhutan 3.03 17.41 NA Central African

Republic -0.65 2.92 45.95 Cote d’Ivoire -0.53 9.29 88.32 Cameroon -0.85 5.68 101.53 Republic of Congo -1.20 0.00 NA Comoros -45.66 4.14 NA Eritrea 3.67 7.71 NA Ethiopia 0.95 4.28 NA Georgia -6.78 4.84 130.89 Ghana 1.08 12.25 53.90 Guinea 1.17 4.97 62.26 Gambia 0.21 12.18 57.49 Guinea-Bissau -0.77 4.03 NA Haiti -45.26 2.88 NA Indonesia 2.48 15.85 49.45 India 3.34 16.66 66.10 Kenya -1.08 2.65 140.19 Kyrgyz Republic -3.93 0.72 NA Cambodia 1.30 19.33 40.05 Lao PDR 3.52 16.30 37.89 Liberia 0.66 -4.02 NA Lesotho 1.53 2.95 87.89 Moldova -7.93 2.65 152.07 Madagascar 0.23 3.90 90.30 Mali 1.76 11.50 87.89 Myanmar NA 11.80 6.99 Mongolia -1.11 4.15 142.43 Mauritania 1.48 10.49 NA Malawi 1.11 4.40 54.31 Niger -1.06 4.66 NA Nigeria 0.80 3.63 87.17 Nicaragua -0.39 7.32 70.62 Nepal 2.62 12.76 NA Pakistan 1.33 9.91 49.58 Rwanda -1.54 3.93 NA Sudan 3.05 15.57 NA Senegal 0.96 12.44 132.02 Solomon Islands -2.29 1.93 16.06 Sierra Leone -4.78 3.07 NA Somalia NA 6.20 NA Sao Tome Principe -44.01 5.50 NA Chad 0.04 5.95 NA Togo -1.14 11.23 NA Tajikistan -8.22 -1.92 NA Tanzania 0.61 3.84 86.69 Uganda 2.75 3.44 96.75 Ukraine -5.80 3.80 104.74 Uzbekistan -1.49 -0.25 35.72 Vietnam 5.07 31.55 60.53 Yemen 0.32 6.11 54.31 Democratic

Republic of Congo 17.06 18.78 NA Zambia -2.03 -0.69 83.62 Zimbabwe -0.68 4.93 NA Average, all

developing countries -2.03 6.61 77.99

Figure: Causal Model of Telecommunications and GrowthFigure: Causal Model of Telecommunications and GrowthFigure: Causal Model of Telecommunications and GrowthFigure: Causal Model of Telecommunications and GrowthFigure: Causal Model of Telecommunications and Growth

Economic Growth (GDP)

+++-++-++++ + + -+ + -+ + + Demand SideDemand SideDemand SideDemand SideDemand Side Supply SideSupply SideSupply SideSupply SideSupply Side Capital (Net of Telecom Investment) Labour Force Demand for Telecom Competition in the Market Place Price of Telecom Services Government Deficits/Debt Telecom Investment Land Area Telecom Penetration

identify variables that are of relevance, and operationalise them in the context of developing economies. Next, we highlight the use of different variables than what Roller and Waverman (2001) use in their paper, based on certain considerations.

In the figure, we summarise factors that affect telecom penetration and GDP of the country through a causal model, highlighting the existence of all major cause-and-effect links, indicating the direction (cause → effect) of each relationship. The relationship is positive (or negative) if a change in the causal factor produces a change in the same (or opposite) direction in growth. A closed sequence of causal links represents a causal loop. The causal loop is positive if it has all positive links or even number of negative links. Otherwise, it is a negative loop.

The primary relationship in the figure is the explanation of growth, which depends on traditional inputs, apart from telephones. We expect all the inputs highlighted in the figure – capital (net of telecom), labour, and telecom penetration – to have a positive effect on total national output (GDP). In the figure, there is at least one positive loop indicating increased growth of both economic growth and telecom penetration.

Note that, based on the importance of mobile phones, and data available, it is possible to empirically estimate at least three specifications of the model shown in the figure, as it concerns telecom penetration – main landlines only, mobile phones only, and all telephone lines. Further, when one empirically estimates the impact of telecom penetration on growth, unobserved country-specific characteristics are to be taken into account, if one intends to do this cross-regionally.

Our next step in the causal model, demonstrated in the figure, is to examine demand for telecom service in the network of relationships that determine growth. Increased economic growth increases the disposable income of population and hence affects demand for telecom services positively. In most of the developing countries, initially, the government was providing telecommunications service and there was a huge waiting list for main telephones. For example, in India, even after private operators were allowed to provide competitive service, the waiting line for main telephones was around 1.649 million in 2001. Even after the introduction of cellular services in 1995, the waiting list continued to grow from 2.28 million to 2.89 million in 1996 and 2.71 million in 1997. As this example demonstrates, the waiting list for mainline telephones tends to be quite long in developing countries. Hence, to operationalise demand for telephones in the context of developing countries, one should define effective demand for telecommunications infrastructure as the sum of existing teledensity and waiting list for mainline telephones, as Roller and Waverman (2001) rightly point out.

Next, the supply of telecom investment depends on potential demand measured by the waiting list for main telephone lines. As with the price, the expected effect is positive.

Further, competition and liberalisation in the market in developing countries has encouraged private operators in addition to government operators, to invest in telecommunication services. The more competitive the market structure is, the greater would be the supply of telecom investment that would be forthcoming, other factors remaining the same. However, as the standard framework shows, as the competition increases, prices also decrease which could have a negative effect on telecom investment, all else remaining the same. Empirical work should throw light on the relative strengths of these effects.

Government deficit or central government debt, determine governments’ ability to invest in telecom infrastructure. While higher telecom investment by the government, could, in principle, cause higher deficits or government debt, the first-order effect is that lower government deficits or debts cause telecom investment to be higher. This is, of course, assuming that the government is a major player in telecom investment, as they are in developing countries.

The empirically oriented reader might note that the causeeffect relationships that are indicated in dotted lines in the figure are those that may or may not be necessarily captured in the empirical model, depending on the availability of appropriate data. See Sridhar and Sridhar (2004), for examples of the typical data limitations one would encounter in empirically validating the model in the context of developing countries.

Finally, consistent with Roller and Waverman (2001), we characterise the growth of telecom penetration as a function of telecommunications investment and the country’s geographic area. We expect that total telecom investment will have a positive impact on the penetration rate, whether it is landlines, cellphones, or total penetration. The land area is relevant for determining the extent of change in telecom penetration. Other things (such as investment) remaining the same, a country with a large land area would have a slower rate of change in its teledensity, when compared to one with a smaller land area.

In sum, the figure shows that telecom penetration, which positively impacts GDP, is determined by GDP per capita and the price of telecom services. Competition or market structure determines the price of telecom services and the magnitude of total investment in the sector. To close the loop, telecom investment affects penetration positively if other factors were to remain the same. Sure enough then, the model shows that competition is the only exogenous policy variable that affects both demand for and supply of telecom services through their effect on prices. This summary of the model also clearly demonstrates that single equation or OLS methods would be unsuitable for empirically estimating it, systems and simultaneous methods would be more appropriate.

Further, in modelling telecom services and their effects on growth, it would be difficult to empirically examine threshold effects, which network externalities in telecom services suggest could be important. Both diffusion of innovation theory [Rogers 1983] and economics of network externalities [Rai et al 1998] suggest that the cumulative number of subscribers of communication services over time is expected to follow a non-decreasing S-shaped distribution.7 After a “critical mass” is attained, the services take-off following an exponential growth pattern. Finally, given a finite potential population, the growth levels off representing the saturation level.

There are two reasons why one cannot empirically examine saturation levels or threshold effects in modelling telecom services in the context of developing countries:

Note that the relationships shown in the figure can be empirically estimated separately for mainline and mobile phones. In Table 3, we provide a list of the variables that can be used in these models of telecom services, if they were to be empirically estimated, as in Sridhar and Sridhar (2004).

Roller and Waverman (2001) find a 1 per cent increase in the number of main telephone lines per capita, increases economic growth by 0.15 per cent in the OECD countries, not controlling for country-specific fixed effects, and a much lower, 0.045 per cent when controlled for fixed effects.

The previous section highlighted the effect of telecom penetration on growth, in the network of several exogenous and endogenous relationships. In this section of the paper, we focus distinctively on the applications of telecom, and how they have contributed to growth and social development in many developing countries.

A number of projects initiated by voluntary non-profit organisations and entrepreneurs in developing countries have demonstrated the positive impacts of telecom penetration on economic and social development in these countries. This section summarises representative projects as qualitative evidence.

Richardson et al [2000] summarises rural Bangladesh’s Grameen Telecom’s Village Phone Programme. Bangladesh’s Grameen Bank, with expertise in microcredit, collaborated with Grameen Telecom, a wireless telecommunications service provider, enabling women members of the Grameen Bank’s revolving credit system, to retail cellular phone services in rural areas. This pilot project involved the setting up of 950 village phones providing telephone access to more than 65,000 people. When fully completed, the access was envisaged to increase to about 40,000 village phone operators, having begun from a modest 950. When completed, the village phone operators would be employed for a combined net income of $24 million per annum.

This study estimates that the consumer surplus for a single phone call from a village to Dhaka, a call that replaces a physical trip to the city, ranged from 2.64 per cent to 9.8 per cent of mean monthly household income. This study also found that 42 per cent of calls made using the village phones, were regarding secure remittance of savings into Bangladesh from other countries. The income village phone operators derived, was about 24 per cent of the household income on average, and in some cases it was as high as 40 per cent of the household income.

These case studies, while being primarily pilot projects in nature, demonstrate qualitatively, but clearly the many applications of telephones, especially of mobile phones, and how they can be used to further the social and economic growth of remote areas in developing countries.

Tables 4 and 5 give descriptive details of various parameters of telecom services and growth, for all developing countries defined as low income by the World Bank. As Table 5 shows, the observations for cellphones are lower than those for main lines, as most of the developing countries started experiencing rapid cellphone penetration only after 1995. The average change in penetration for landlines and cellphones are greater than 1 and 2 respectively (Tables 4 and 5) suggesting continual increase in landline penetration, and more so for cellphones. Interestingly, the average and maximum waiting lists for landlines (per 100 population) in countries with cellphones (the smaller sample, Table 5), are much smaller than those in the full sample. This shows that countries with rapid cellphone penetration did not have waiting lists to the same extent as those without. The average telecom investment for the smaller sample of countries that experienced rapid cellphone penetration is also higher than that for the full sample.

On average, the total telephone penetration in the developing countries we examined, is much lower (being 2.6 per 100 inhabitants) than that observed in the OECD countries by Roller and Waverman (2001) (30 per 100 inhabitants). Among developing countries, the maximum total tele-density is itself 20, observed for Ukraine in 1999. The landline penetration is even less. On average the GDP per capita for these countries is much lower than that observed for the OECD group of countries, reasonable to believe.

In developing countries, the average mainline price (monthly subscription) is quite low when compared to that for cellphone subscription. Here it may be relevant to note that in developing economies, for basic mainlines, the tariffs are always kept low by the regulator to make the service more affordable. In contrast, the revenue per user is higher for the total telephone lines and landlines than they are for cellphones. This is due to the larger subscriber base and usage of landlines.

Cellphones

Note that developing countries with low penetration rates for main telephone lines, find the infrastructure for cellphones to be relatively inexpensive and less time-consuming to install. Cellphone penetration in developing countries started increasing rapidly during the second half of the 1990s due to changes in telecom regulation, and move to competitive market structures.

Table 4: Description of Relevant Data for Full (of Landlines and Cellphones) Sample (N=256)Table 4: Description of Relevant Data for Full (of Landlines and Cellphones) Sample (N=256)Table 4: Description of Relevant Data for Full (of Landlines and Cellphones) Sample (N=256)Table 4: Description of Relevant Data for Full (of Landlines and Cellphones) Sample (N=256)Table 4: Description of Relevant Data for Full (of Landlines and Cellphones) Sample (N=256)

Table 5: Description of Relevant Data for Cellphones (N=65)Table 5: Description of Relevant Data for Cellphones (N=65)Table 5: Description of Relevant Data for Cellphones (N=65)Table 5: Description of Relevant Data for Cellphones (N=65)Table 5: Description of Relevant Data for Cellphones (N=65)

Most of the developing countries leap-frogged into secondgeneration mobile cellular systems, deploying them at a much greater rate compared to landline installations.

Waverman, Meschi and Fuss (2005) argue that for economies without fixed lines, or where mobile phones supplement the low roll-out of fixed lines, there should be no inherent difference in the growth dividend of a phone, whether it is mobile or fixed. However, there are some reasons why we may expect cellphones to contribute positively to national output, which are not always available with service providers of landlines. Cellphone penetration reduces transaction costs, including, but not limited to, decisions relating to production of goods and services. For instance, value added services such as stock quotes and commodity prices provided by cellular service providers at affordable prices using the latest digital cellular technologies, may be expected to produce tangible economic outcomes. The Manobi case also demonstrates this further and shows the wide-ranging social applications of mobile phones in rural communities, which may not be available with landline phones.

Based on the data, we find that a number of countries that experienced rapid increases in cellphone penetration experienced political turmoil during the 1990s. These countries – Cote d’Ivoire, Moldova, Armenia, Ukraine, Kenya, Kyrgyz Republic, Togo, Zimbabwe – experienced rapid decreases in their national output during the 1990s (Table 2), specifically Armenia, Moldova, Kyrgyz Republic and Ukraine, due to the splitting of the USSR. While cellphones may be expected to contribute positively to national output, Sridhar and Sridhar (2004) find that the effect of cellphones may not be strong enough to offset the negative impacts of the country-specific factors that decrease national output significantly. This is not to say that cellphones decrease output, but a disclaimer to remember that their positive effect may not be strong enough to offset the effect of overarching exogenous factors such as political turmoil that caused to decrease national output in these countries quite significantly.

Further, we believe that traditional economic factors – price and income – that determine the demand for other telephone services (including main landlines) should explain demand for cellphones as well. Moreover, the income elasticity of demand must be higher for cellphones than for landlines, since cellphones are still viewed as a luxury and mobile handsets continue to be expensive in developing countries. Thus, we expect micro, household decisions relating to cellphone services to be dependent on economic factors, as with other phone services.

Finally, when one empirically examines the effect of investment on the rate of change in tele-density, we note that in most of the developing countries, cellular services are greenfield projects, requiring huge capital investment to commence. Initial investment in interconnection facilities to connect to other networks may not immediately translate in to increase in subscriber base. It is possible that there could be some time lag before greenfield projects translate to increased penetration in the case of cellular phone services.

There are data limitations that could limit the value of the model presented in the figure, if it were to be empirically estimated for main landlines and mobile phones. The cellphone sample could be quite small since data on cellphone related information are reliable and available for all developing countries only post 1996, as is clear from Table 5.

l1r

Email: kala@nipfp.org.in; sridhar@mdi.ac.in

NotesNotesNotesNotesNotes

ReferencesReferencesReferencesReferencesReferences