New Technologies Adoption and Diffusion Patterns in Developing Countries. An Empirical Study for the Period 2000-2011.

The paper, purely empirical in nature, reports on the pace of adoption of new Information and Communication Technologies in developing countries, and – additionally – investigates country-specific ICTs diffusion patterns. We expect to uncover the S-shape curve in the diffusion process in most of developing countries, as well as in the whole country sample. For the analysis purposes we apply all counties, which – according to the World Bank nomenclature – are classified as low-income and lower-middle-income economies. Our sample covers 46 countries (upper-middle-income and high-income economies are excluded from the study purposely), which are classified as developing economies. The time framework is set for the period of 2000-2011. All data necessary for the analysis are derived from World Telecommunication/ICT Indicators Database 2012 (16th edition).


Introduction
In recent years, a pervasive explosion of new Information and Communication Technologies (ICTs) is observed in a magnitude of world economies. The process of diffusion of ICTs is fully associated with the spread of new ideas, knowledge and all kinds of information. These technologies provide new ways of mass communication, process of storing and processing all kind of knowledge and information.
ICTs are widely recognized as an effective tool for promoting economic growth and overall socioeconomic development. By creating a bundle of possibilities on education, running a business or inclusion of socially and economically deprived groups, they provide a good basis for wealth creation.
Wireless networks have enabled to open up economic "enclaves" by elimination of barriers such as i.e. geographical distance or information asymmetry. At a time, due to broad ICTs adoption, society members gain an opportunity for wide and unlimited access to different kinds of knowledge, information, which enable them to improve their professional skills.
The process of technology (ICTs) diffusion and adoption has been studied intensively in recent years.
Following E.Rogers and his seminal works, diffusion includes "both planned and the spontaneous spread of new ideas" (Rogers, E. 1962, p.7). Massive empirical and theoretical literature has been developed on the issue, and some general conclusions can be derived from the studies. The most crucial conclusion is that, at a given point of time, substantial differences can be perceived among different countries when level of ICTs adoption and diffusion is taken into account. Despite many discussions on widening -or diminishing -digital divide, observed huge differences in the degree of diffusion of technology among countries remain unquestionable.

Technology diffusion -general conceptualization.
Conceptually, the discussion on the "technology diffusion" has started with Joseph Schumpeter (1934Schumpeter ( , 1947, and his pioneering work emphasising the crucial -for economic growth and developmentprocess of technology (innovation) diffusion. In his works, Schumpeter -inspired by origin works of Charles Darwin (Darwin 1859), argued that the process of technology diffusion is mostly linear.
However further studies on the phenomenon of the technology diffusion process accompanied by technological change, has shown that the process of adoption of invention is far from linear. In the fifties due to some theoretical discussion on technology diffusion, there emerged new concepts of epidemic models of diffusion 2 . The models were based on some analogies between process of technology diffusion and diffusion of contagious diseases. The epidemic models were claimed to be correct by Mansfield (1968), Metcalfe (1988), Karshenas and Stoneman (1995), Stoneman (2001Stoneman ( , 2005. In time, many started o claim that epidemic models of technology diffusion are "blind" for the dynamics of the process, and ignore some societal, educational, demographical or economic conditions which highly influence rate and pattern of technology diffusion. Now, theoretical and empirical analysis of technology diffusion, mostly refer to the seminal concept of Everett Rogers (see Rogers 1962Rogers , 1995, explaining diffusion process of innovation (Rogers presumes that the word "technology" is a synonymous to "innovation". As consequence Rogers uses these two alternatively, along with the diffusion rates and stages. His concepts and theoretical framework (model) of technology diffusion are widely applied in political sciences, economics, public health policies, history or education. For Rogers (1962Rogers ( , 1995 the "technology" stands for "design for instrumental action that reduces the uncertainty in the cause-effect relationships involved in achieving a desired outcomes" (Rogers, E. 1962, p.6). He (see Rogers, 1962, p.10) defines the diffusion "as a process by which an innovation (technology -author`s note) is communicated through certain channels over time among members of a social system". Following the previous, the process of technology diffusion covers 4 aspects (elements): innovation (defined as "an idea, practice, or project that is perceived as new (…) (Rogers, 1962, pp.12)), communication channels ("process in which participants create and share information", (Rogers, 1962, pp.5), time (time in behavioural studies is perceived as an element determining the strength of the process) and social system ("a set of interrelated units", (Rogers, 1962, pp.23). Also Gort and Klepper (1982) define technology diffusion as the "spread in the number of producers engaged in manufacturing a new product". The definition was slightly adjusted by Adres et al. They claim that the technology diffusion can also refer to the "number of consumers of Internet" (see Andres et al., 2007).
In accordance to the diffusion of innovation theory of Rogers E. (Rogers 1962), there emerged the concepts of theoretical diffusion patterns that countries shall follow, passing sequent steps (phases) in adoption process. The adoption patterns of technology, are assumed to follow the S-shaped curve. The group of S-shaped curves is originated from Pierre-Francois Verhulst, who in 1833 proposed to the public, first logistic equation, explaining natural population growth in time. The "natural growth" stands for particular ability of "species" (see Darwin, 1859) for multiplying in a specified period of time.
In economic sciences, the term" species" refer to any kind of variable which tends to grow in time. The S-shaped trajectories are mainly observed when innovation (technology) diffuses over large societies.
The curve shape explains adopter population dynamics in time, assuming the heterogeneity of the whole population. The theoretical trajectory shape shows that technology tends to spread at high pace in early stages of diffusion process, and afterwards the diffusion rates slow as the overall saturation (share of population using given technology) of the technology in a given society if growing. It is also assumed that, after passing the "inflection point", the rate of adoption shall slow down, as the higher percentage of population is having access to technology.
The general concept of the S-shaped curve explaining stages of technology diffusion is clarified in the Picture 1 (see below). Clearly, the process of technology diffusion is fully accounted for the nature of technology itself. Certain features of technology, absorptive capacities of society, its knowledge and skills, information asymmetries, channels of diffusion and further adoption, state policies -all these highly condition diffusion rates, creating friendly environment for technology diffusion or -reverselyposing barriers to the process.
Despite the above, the technology diffusion can be described as a long-run process, which results in broad spread of different kinds of innovations. It is widely agreed that the process enhances deep changes in economic structures, fosters economic growth and development, contributing to overall welfare of societies. Picture 1. Theoretical S-shaped curve. Source: own elaboration.
As stated before, in the following paper, by using the term "technology" we refer strictly to New Information and Communication Technologies, and the two terms -technology and ICTs -are used alternatively, as equivalent issues.

The data -preliminary analysis.
In the study we concentrate exclusively on low and lower-middle-income economies. We base on the broadly accepted country classification introduced by World Bank 3 . As results, our country sample covers 46 economies, where 15 countries are classified as low-income and 31 are classified as lower-3 World Bank classifies countries according to their per capita Gross National Income using Atlas Method as tool for international comparisons. The low-income countries are those were average annual GNI per capita does not exceed 1025 US dollars. The lower-middle-income countries are those were average annual GNI per capita varies from 1026 US dollars per capita to 4035 US dollars per capita. middle-income ones. Full list of countries is available in the statistical appendix (see statistical Appendix 1). The applied time span is 2000-2011.
For the analysis purposes, we have selected 3 basic ICTs indicators, treated as proxy variables of information and communication technologies. They are thought to reveal the average level of usage and application of ICTs among society. In the study we use the following variables: share of total population using (having access) to fixed telephone lines (FTL -abbrev.); share of population using mobile cellular phones (we use mobile cellular subscribers ratio; MCS -abbrev.); and share of population (individuals) using Internet (Internet users ratio, IU -abbrev.). All statistical data referring to the usage of ICTs were derived from World Telecommunication/ICT Indicators Database 2012 (16 th edition) developed by ITU 4 . Internet Users ration data were derived directly from the ITU statistics. In case of FTL and MCS data, we have drawn absolute numbers of given ICTs users in each country and -in the next step -applying data on the absolute population number 5 , we have calculated the shares of population using (having access) given ICTs tool 6 . Full, balanced panel dataset is put in the statistical Appendix 1.
To get a general idea about the variables values in low and lower-middle-income countries, we run a preliminary descriptive statistical analysis. For this purpose we use raw data for FTL, MCS and IU in the period 2000 and 2011.
In the following Table 1 (see below), we present basic descriptive statistics of the ICTs variables. 4 International Telecommunication Union. 5 All data on population number were derived from World Development Indicators 2012 database. 6 In case of break in series, we have filled the missing values by estimating average annual exponential growth rate for given time-series. In Table 1, each variable stands for share of total population using given telecommunication tool. As can be easily concluded, changes in FTL adoption are slight, and do not present any crucial increases.
But, if we turn to the following 2 variables: MCS and IU, changes in their level of basis adoption and usage are enormous. In 2000, the average share of population having adopted MCS in low and lowermiddle-income countries was at 2,3%, while -after 11 years of dynamic growth 7 , the level is at 76,9%.
In 2000, the best performing country was Paraguay -with 15,3% of population using MCS, and in 2011 it was Vietnam -144,9% of population using mobile phones 8 .
Similar conclusion can be drawn in case of IU. In 2000, the average level of population share defined as Internet Users was at 0,7%, while in 2011 -17,5%. In 2000 the best performer was Belize with 5,95% of population (individuals) using Internet, and in 2011 -Morocco with the 51% of individuals using Internet.
The lower/upper quartile analysis also reports on huge changes on the field. For MCS (in 2000) in the 25% of countries (lower quartile) the share of population using MCS was below 0,2%, while in 2011, the analogous value reached the level of 55,8%. This, again, proofs highly dynamic diffusion and increase in usage of basic ICTs tools in low and lower-middle-income countries.
Additionally, we report on Gini coefficients for each variable, separately in 2000 and 2011 (see Table 1).
For FTL variable, changes in value of Gini coefficients are not reported, and the degree of inequality was not affected by the process of FTL diffusion in given years. Just the opposite situation is reported 7 For detailed growth rates see Appendix 1. 8 The number higher than 100% proof that one person subscribes more than one mobile phone.
in case of MCS and IU. For both variables, the process of ICT diffusion has affected significantly the magnitude of concentration. After 11-year period of dynamic diffusion of ICT -especially for MCSthe inequality in access to ICT has diminished, and the usage of basic ICT tools has become more common for all society members (for additional evidence on changes in inequalities, see Lorenz curves for FTL, MCS and IU in the Appendix 3).

Identifying ICTs diffusion patterns in low and lower-middle income countries.
The main aim of the study is to identify diffusion patterns of ICTs in low and lower-middle-income countries. As stated by Jovanovic and Lach (1989), there exists wide empirical evidence that technology diffusion patterns follow the logistic trajectories graphically approximated by S-shaped curve. Following the above, we expect to uncover the "classical" S-shaped diffusion curve in most of countries for ICTs diffusion process. The empirical part encompasses 3 sections. In section 1, we aim to detect changes in Finally, in the third section, we estimate the ICTs diffusion patterns of FTL, MCS and IU in country sub-samples: C L (low-income countries) and C LM (lower-middle-income countries), for the analogous time span.

Distribution changes of ICTs in low and lower-middle-income countries.
In the first stage we seek for changes in the distribution of ICTs in the group of low and lower-middleincome countries. To capture these we apply densities functions approach, we analyze distributions of Analogous conclusion could be derived from descriptive statistics in Table 1, where the only slight growth in usage of fixed phones is detected.
A completely different picture presents Chart 2 (see below). In 2000, the average adoption/usage of mobile phones was at relatively low level, and the level of adoption/usage in the countries in the sample was highly uneven. In 2000, high diversification is evident 10 . After the 11 years (in 2011), the distribution line (dash line) has "moved" to the right and is highly peaked. It proofs much higher average level of MCS usage in the whole sample, and -on the other hand -decreases in inequalities. In terms of MCS usage, in the analyzed period, the group of low and lower-middle-income countries has become much more homogenous.
10 Again for MCS.

ICTs diffusion patterns in low and lower-middle-income countries.
The process of technology diffusion is widely discussed on the literature. Empirical works mainly concentrate on Internet diffusion patterns, explaining rate of diffusion and the extent to which Internet is adopted by potential users. Such analysis can be find in the works of Chong and Micco (2003), Klobas and Clyde (1998), Quibria (2002) or i.e. Estache et al. (2002). There are also several authors which intended to identify the S-shaped curve in technology adoption process in countries. The empirical evidence on the existence of S-shaped diffusion path can be found in works of Mansfield Following the logic, we assume that the technology diffusion -in each country, shall follow heterogeneous path, resulting in S-shaped curved as graphical illustration of the process. In early stages of diffusion the rate of diffusion is relatively low. Later on, the adoption rates are growing exponentially, finally achieving the inflection point. After passing the inflection point, the diffusion process is slowing down again, aspiring to stabilization phase when the saturation 11 (share of population using the given technology/innovation) is high (reaching 100% or above).
Let us assume that technology diffusion rate is a function of time, where φ j (t), for j-country in year t.
Following the previous, the absolute technology saturation change for j-country in the period 12 t, for given T 13 -technology, can be formally expressed as:  14 Diffusion. 15 FJI -international country code. 16 SLV -international country code. 17 PRY -international country code. 18 UKR -international country code. 19 All classified as lower-medium-income countries.
To clarify the picture and achieve more comparability among selected economies, in Charts 8-10 (see below), we apply mutually four diffusion curves strictly for one ICTs tool to see differences in spread and dynamic of the process in selected economies.
The most important finding is that in each of them, the diffusion patterns of MCS (dash-lines in Charts 4-7), they form nearly the classical S-shaped diffusion curve. Analogous conclusion we draw analyzing We also need to note, that the case of Ukraine is very particular. In the period 2000-2011, the MCS diffusion process has passed all "traditional" phases explained by S-shaped curve. In the initial period, In Table 2, (see below) we present summary estimates of the dynamics of ICT diffusion process in the analyzed countries. Again, we see that in 2011, the saturation levels with MCS result to be much higher, than in case of FTL and IU. For all 4 countries, the ϑ j,MCS(2011) reaches the level of almost 100% or even higher. This implies that almost 100% of country`s population enjoys access to given ICT tool. Also, solely in case of MCS, developed diffusion trajectories seem to be described by S-shaped pattern. 21 In Paraguay it is only another 0,38% of population.  (Φ j,T,2000-2011 ), average absolute technology saturation growth (ϕ j,T,2000-2011 ), and final saturation level with ICT in 2011ICT in (ϑ j,T,2011  Country Φ j, T(2000-2011) [%p] ϕ j, T(2000-2011) [annually, %p] ϑ j, T(2011)  Source: own calculations.

ICTs diffusion patterns in country sub-samples.
In the final section, we investigate the ICT diffusion curves in 2 country sub-samples (sub-groups) -C L (low-income) and C LM (lower-middle-income countries). Again we seek for S-shaped diffusion trajectories revealed in time, but this applying average values for FTL, MCS and IU as variables. Chart 11 plots the evidence of the ICT diffusion paths in C L and C LM separately. Finally, we conclude on the historical evolution of the ICTs diffusion in the 2 heterogeneous country groups. Despite the fact, that growth rates of ICTs diffusion differ significantly, regarding differences among variables and among countries, in both cases the ICT diffusion trajectories result to be quite similar in shape. In low-income countries (15 countries in the group), the highest saturation level is reported for MCS (for variablespecific diffusion patterns see Chart 12(b)) -in 2011, the average level of ϑ Chart 12 (c) Source: own elaboration.
In case of IU, the diffusion trajectories stand for the initial "parts" of the theoretical S-shaped curve, indicating that the IU diffusion, both on C L and C LM , refer to the early stages of diffusion process. For the given variable, in low-income countries the ϑ !" The FTL diffusion process is hardly visible in both country groups, revealing little higher dynamic in lower-middle-income countries than in low-income ones. The average saturation level in 2011, for C L and C LM , results as ϑ ./ , 2,06% and ϑ ./ , 9,82%. At a time, no logistic curve is displayed. This again confirms the hypothesis on growing substitution of "old" ICTs by "new" ones. In low-income and lower-middle-income economies the process of substitution can be noticed by no difficulties. This is mainly due to special features of ICTs. Easiness for adoption and usage, minimum financial and infrastructural requirements constitute them an attractive development tool especially in low and lower-middle-income countries. Our analysis was not the long-term one (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011), and was limited strictly to the analysis of low and lower-middle-income countries. In a general view, it can be concluded that despite diversity in ICTs diffusion paths in countries, the process is characterized by high dynamics and wide spread of new technologies is observed. The finding is seminal, for further economic growth and development perspectives in backward economies.