27
Innovation Diffusion in Developing Countries 1 Innovation Diffusion in Developing Countries: An Extension of Bass-Model in the Context of International Marketing by Dr. Sein Min Research Fellow UNIVERSITÄT PASSAU Lehrstuhl für Betriebswirtschaftslehre mit Schwerpunkt Absatzwirtschaft und Handel Prof. Dr. Dr. h.c. Helmut Schmalen August 2001

Working Paper

Embed Size (px)

DESCRIPTION

This paper investigates the diffusion process in developing countries, applying the well-known Bass' Diffusion Model and Shamlen's Model. The empirical data were gathered from a developing country, namely Burma (Myanmar).

Citation preview

Page 1: Working Paper

Innovation Diffusion in Developing Countries

1

I n n o v a t i o n D i f f u s i o n i n

D e v e l o p i n g C o u n t r i e s :

An Extension of Bass-Model in the Context of International Marketing

by

Dr. Sein Min Research Fellow

UNIVERSITÄT PASSAU Lehrstuhl für Betriebswirtschaftslehre

mit Schwerpunkt Absatzwirtschaft und Handel Prof. Dr. Dr. h.c. Helmut Schmalen

August 2001

Page 2: Working Paper

Innovation Diffusion in Developing Countries

2

Innovation Diffusion in Developing Countries:

An Extension of Bass-Model in the Context of International Marketing

Introduction

Innovation diffusion has been interested by many academic disciplines including

anthropology, (rural-, and medical -) sociology, education, economics, communication,

geography, etc.1 Martketing is also one of the disciplines being much interested in the nature

of innovation diffusion. as it has to deal with introducing new technological products into the

markets. Understanding it can help marketing managers to forecast sales of new products and

to formulate their marketing strategies. Todays, international marketing is executed by many

firms and hence markets of developing and least developed countries are more and more

interesting. For the task of internatioanl marketing which has a special interest in developing

markets, studies of innovation diffusion can help a lot.

The phenomena of innoation diffusion is studied at two levels: micro individual level

(adoption process) and macro aggregate level (diffusion process).2 Diffusion process is

observed with the help of models, i.e. modeling approach is used to capture the reality of

diffusion process. Diffusion models used in marketing area are particularly interested in how

diffusion occurs in the consumer domain.3 Thus, these models can or should relate to the

theoretical aspects of consumer behaviour in adopting innovations. Gatignon and Roberston

pointed out that „an integration of the behavioural and modeling literatures on diffusion could

be benieficial to both constttuenices“.4

In modeling the pattern of innovation diffusion, the Bass model was very successful

showing the high magnitude of fitness of the model.5 The Bass model reflects the individual

adoption behavior in two different ways: innovation and imitation. The model’s parameters p

and q represents the external influence to adopt (in other words, the propensity to innovate)

and the internal influence to adopt (in other words, the propensity to imitate) respectively.6

1 Mahajan, V. and M:E.F. Schoeman, „Generalized Model for the Time Pattern of the Diffusiom Process“, IEEE

Transactions on Enginerring Management, Feb. 1977, pp. 12-18., p. 12. See also Gatignon, H. and T.S.

Robertson, „ A propositional Inventory for New Diffusifon Research“, Journal of Consumer Research, Vol. 11,

March 1985, pp. 849-867, p. 849. 2 The lengthy discussion about this is found in: Rogers, M. „Diffusion of Innovations“, New York , The Free

Press, 1995. 3 Gatignon, H. and T.S. Robertson, Ibid, p. 849.

4 Ibid.

5 Bass, F. M. „A New Product Growth Model for Consumer Durables“, Management Science, 15(1969), pp.

215-2127. 6 Gatignon, H., J. Eliashberg and T.S. Robertson, „Modeling Multinational Diffusion Patterns: An Efficient

Methology“, Marketing Science, 8(1989, pp. 231-247, p. 232. The discussion on interpretation of the parameters

Page 3: Working Paper

Innovation Diffusion in Developing Countries

3

The Bass model was tested in the actual diffusions of products, mostly consumer durables.7 It

was also extended or modifed to overcome its assumptions by introducing such variables as

market potential, marketing-mix (advertising, price), effects of multiple products, etc.8

Diffusion researches via modeling which involve both Bass type and non-Bass type models

were increasing. The Bass model was earlier applied in a single market, mostly of U.S. Later

the application was extended to include European (industrialzed countries). There were also

investigations of the Bass-model in international setting. Nowadays, the diffusion research

covers almost the entire world and it has reached even the stage of global diffusion research.

For example, Dekimple et al conducted the research on diffusion of cellular phone comprising

184 countries, although their approach used a differnt modeling method by employing a

hazard function.9

Although the diffusion research becomes an international and global research, little

attention has been paid to the distinct nature of diffusion process in developing countries.

Dekimpe et al. criticized this point sharply as:

„The set of countries considered in most international diffusion research is not

only limited in scope, but also severely biased towards the study of industrialized

countries.10

Little is known, however, about the nature of the diffusion process in

developing countries.11

..more research is needed on the extent of an internatioanl

learning effect (cf. Infra) both among developing countries, and between developed and

developing countries.“12

Developing countries should be paid attention by intrnational and global marketers as

well as diffusion researchers, because the developing countries (and also least developed

countries) are the major part of world market at least in terms of population. Due to market

saturation in domestic markets and increased competition among suppliers of industrial

products the markets of developing countries play increasingly an important role. Thus, it is

p and q, see Schnmlen, H. and H. Xander, „Produkteinführung und Diffusion“, in S. Albers and A. Hermann

(ed.),“Handbuch Produktmanagement: Strategieenwicklung – Produktplanung – Organisation – Kontrolle“,

Gabler, 2000.The criticism on Bass Model was discussed in many literatures. For examples, see Schmalen, H.

„Das Bass-Modell zur Diffusionsforschung: Darstellung, Kritik und Modifikation“, ZfBf 41(3/1989), pp. 210 –

226, Tanny, S.M. and N. A. Derzko, „Innovators and Imitators in Innovation Diffusion Modelling“, Journal of

Forecasting, Vol. 7, 1988, pp.225-234. 7 Mahajan, V. E. Muller and F.M. Bass, „New Product Diffusion Models in Marketing: A Review and Directions

for Research“, Journal of Marketing, 54(1990), pp. 1-26, p.16 (Table 4) 8 See Ibid, p. 10-15.

9 Dekimpe, M.G., P.M. Parker and M. Sarvary, „Comparing Adoption Patterns: A Global Approach“, Working

Paper (96/37/MKT), INSEAD; 1996. 10

Dekimpe, M.G., P.M. Parker and M. Sarvary, „Multi-Market and Global Diffusion“, Working Paper,

INSEAD, 1998, p.3. 11

Ibid. 12

Ibid. p.4.

Page 4: Working Paper

Innovation Diffusion in Developing Countries

4

necessary to pay a close attention towards the diffusion process in developing countries. This

effort should include specific theoretical consideration about the developing countries and

diffusion there and then the development of appropriate models for them. Moreover, we need

to pay attention towards these countries because of the changing condition in the world such

as globalization.

In this paper, we are going to present the development of a diffusion model which is

extended from the popular Bass model based on behavioural argument of adoption in these

countries. Before we present our model, there will be a review about the international difusion

research and cross-country diffusion models. After the extended model for developing

counties has been described, we will continue to test the model by using data of a developing

country. We have selected Myanmar for this purpose compiling data of computer, room air-

conditioner, TV-set and telephone for the period 1985-2000. As required in the model, we

will use Japan as a lead country. For this, we have collected the statistics of per capita TV

and telephone possession extracting out of the UN Statistical Yearbooks. In the empirical

study, we first plot the data and estimate the parameters using OLS method and nonlinear

least square (NLS) estimation method with the help of SPSS software. Finally, we present the

conclusion and implications from our study.

Review of International Diffusion Research

The very well.known Bass model has been used to study diffusion of innovations

including new ideas, new products and new technologies. These studies focused on each

individual country. It is interesting however to compare and conclude whether there are some

systematic djfferences in characteristics of diffusion process of countries under study. This

task was done by the estimation and analysis of parameters of Bass diffusion model. Some

scholars applied a form of econometric model to include some exogeneous variables. A new

phase of international diffusion research began with incorporating lead-leg effect into a basic

diffusion model.13

Heeler and Hustad studied in 1980 the Bass model with international data.14

They

believed that the use of Bass model in international setting cannot guarantee its success

because of environmental differences like government policy and trade restrictions. They

13

Summary of International Diffusion Studies is found in; Dekimpe, M.G., P.M. Parker and M. Sarvary, „Multi-

Market and Global Diffusion“, Working Paper (98/73/MKT), INSEAD, 1998. 14

Heller, R. M. and T.P. Hustad „Problems in predicting New Product growth for Consumer Durables“,

Management Science, 26 (1980), pp. 1007-1020. .

Page 5: Working Paper

Innovation Diffusion in Developing Countries

5

found that the results showed instability with limited data and systematic underreporting of

estimated time to attain peak level of first purchase sales. 15

To search cross-national differences in diffusion processes between the home market

and foreign markets, Takada and Jain conducted an investigation of parameters of Bass

diffusion model in the Pacific Rim Region comprising US, Japna, Korea and Taiwan.16

The

diffusion process starts usually in US and later the rest countries subsequently follow the

process. Thus, there exists a phenomena of lead-lag effect. They explained the cross-national

differences due to two types effects: country effect and time effect. Especially due to latter,

they expected that lag countries would have a faster rate of diffusion. Their empirical study

could confirm this proposition.

As an application of diffusion knowledge into international marketing, Helsen et al.

tried to segment the countries based on diffusion parameters.17

They investigaed whether

diffusion parameters of one country-cluster differ from those of others. Their findings

indicated that, for all practical purposes, little agreement exists between the traditional-

derived country segements and diffusion-based segements.18

However, they proposed to

actually segement the countries on the basis of how the diffusion process evolves within these

countries for various consumer durables. 19

Attempts have benn also made to explain the differences in parameters of diffusion

model among countries. Gatignon et al. developed an econometric model for the diffusion of

innovations at the individual country level.20

In their model four factors were considered:

cosmopolitanism, mobility and women in the labour force. Their studies included 14

European countries and adoptions of six consumer durables. The major finding was that

cosmopolitansim is related positively to the population propensity to innovate for the six

products studied. By determining the parameters‘ values based on other exogeneous factors it

makes possible to forecast sales of one country based on the experiences of other countries.21

There were also studies which extended the scope of the internatioanl diffusion

research by relating one country’s diffusion with others. The relationship among countries in

this respect can be classified into (1) reciprocal relationship (mutual effect) and (2) one way

15

More specifically, they found that stable and reasonably accurate predictions of the peak year of sales only

occur with at least ten years of input data, generally after the peak has already occured. 16

Takada, H. and D. Jain, „Cross National Analysis of Diffusion of Consumer Goods in Pacific Rim Countries“,

Journal of Marketing, 55 (1991), pp. 48-57. 17

Helsen, K. K. Jedidi, K. and W.S. DeSarbo, „A New Approach to Country Segmentation utilizing

Multinational Diffusion Patterns“, Journal of Marketing, 57(1993), pp. 60-71.. 18

Ibid. p. 69 19

Ibid. p. 62. 20

Gatignon, H., J. Eliasherrg and T.S. Robertson, Op. cit. 21

Ibid. p.245.

Page 6: Working Paper

Innovation Diffusion in Developing Countries

6

relationship (lead-lag effect). The reciprocal relationship among countries in the international

diffusion process was expressed by Mahajan et al. with special emphasis on unification of

European community.22

Their study however did not empirically base upon the model. They

produced the different parameter values among EU member countries which varied across

countries

The one-way effect in terms of led-lag relationship was explored by Ganesh and

Kumar23

and Ganesh et al.24

. They put forward the concept of „learning effect“ which reflects

that when a new product/technology is introduced in one country and with a time lag in

subsequent countries, there exists an opportunity for consumers in the lag countries to learn

from the experience of the lead country adopters.25

The „learning effect“ was discovered by

Ganesh and Kumar for industrial technological product (retail scanner). In this study the

existence of learning effect was confirmed but the size was not homogeneous. In similar way,

Ganesh et al. investigated empirically for consumer durables (Home Computers, Microwave

Ovens, Cellular Phones and VCRs). This study, conducted for consumer durables in selected

Euopean countries, showed that there existed a cross-country learning effect that was a

function of cultural similarity, economic similarity, time-lag, type of innovation and existence

of technical standard but not of geographic proximity.26

All these two studies used basically

the model proposed by Peterson and Mahajan27

in order to systematically capture the learning

effect. In their study of consumer durables, they linked the coefficient of lerning effect with

covariables mentioned abouve.

Dekimpe et al. had interestingly investigated the breadth of adoption or variabilityin

adoption timing across countries.28

They distinguished the global diffusion process into

breadtn and depth dimensions. The dynamic of diffuson within a country is termed as depth

process. Their model used a hazard function, which gives the adoption rate of a country

during a particular time interval. Subsequently they used a general relationship between a

distribution’s hazard and survivor function. The parameters were estimated through maximum

22

Mahajan, V. and E. Muller, „Innovation Diffusion in Borderless Global Market: Will the 1992 Unification of

the European Community Accelerate Diffusion of New Ideas, Products and Technologies“, in Technological

Forcasting and Social Change, 45 (1994), pp. 221-237. 23

Ganesh, J. and V. Kumar, „Capturing the Cross-national Learning Effect: An Analysis of an Industrial

Technology Diffusion“, Journal of the Academy of Marketing Research, 24 (1996), pp. 328-337. 24

Ganesh, J., V. Kumar and V. Subramaniam, „Learning Effect in Multinational Diffusion of Consumer

Durables: An Exploratory Investigation“, Journal of the Academy of Marketing Science, 25 (1997), pp. 214-228. 25

Ganesh, J. and V. Kumar, Op. cit. p.330. 26

The lack of support for this proposition does not mean that geographical distance is unimportant in

influencing the learning process but, rather, that the other factors probably play a greater role. Ibid. p. 223. 27

Peterson, R.A. and V. Mahajan, „Multi-Product Growth Models“, in Research in Marketing ed.. Jagdish

Sheth, Green wich, CT: JAI Press, 1978. 28

Dekimpe, M.G., P.M. Parker and M. Sarvary, „Globalization: Modeling technology adoption timing across

countries“, Working Paper (96/38/MKT), INSEAD, 1996.

Page 7: Working Paper

Innovation Diffusion in Developing Countries

7

likelihood function. The adoption rate was the function of endogenous and exogenous factors.

They considered the importance of endogenous demonstration effect exerted by earlier

adoptions in „similar“ countries. The exogenous factors were political disposition (communist

or not), socioeconomic characteristics (GNP per capita, crude death rate, population growth),

social system homogenity (number of ethniic groups) and population concentration (number

of major population centers). To illustrate the model, data from cellular telephone industry of

184 countries were collected. The findings supported the extant theories of cross-country

adoption. The demonstration effect of earlier adoptions was also accepted. Similar approach

was also applied by Dekimpe et al to explain the global diffusion of network innovations, i.e

Telecom.29

But they argued that gloabal adoption is comprised of two stages: implementation

and confirmation. The models of the same type were developed to investigate the diffusion

proceess in each country. Their approach was applied for the study of 160 countries.

From these and other studies, a few generalizations seem to emerge about the cross-

country diffusion Dekimpe et al. summarized them as follows:30

(1) The wealth of a country (mostly operationalized through GNP/ capita) has a postive effect

on diffusion process.

(2) There are cross-national learning effects. Countries whcih introduce the innovation a later

point in time seem to have faster with-in country diffusion patterns.

(3) The size of cross-region experience effect is not homogeneous. There was no uniform

finding about the relationship between geographic proximity and learning effect.

(4) The social system heterogeneity has a negative effect of diffusion for both timing and

speed of diffusion of a country.

Cross-country Diffusion Models

Earlier studies of diffusion models focused on a single country, mostly using the Bass

diffusion model. Marketing research on multinational diffusion process has been relatively

sparse in 1980s.31

In 1990s, however, the attention had been paid towards to cross-country

diffusion process, i.e. the diffusion process among countries. In markeing literature the

underlying phenomenon of international diffusion has been observed under the concept of

29

Dekimpe, M.G., P.M. Parker and M. Sarvary, „Global Diffusion of Network Technologies: A Double-Hazard

Approach“, Working Paper (97/50/MKT), INSEAD, 1997. 30

Dekimpe, M.G., P.M. Parker and M. Sarvary, „Multi-Market and Global Diffusion“, Op. cit, p. 8-9. 31

Helsen, K., Jeddi; K. and W.S. DeSarbo, „A new Approach to Country Segmentation Utilizing Multinational

Diffusion Patterns“, Journal of Marketing, 57, (1993), pp. 60-71.

Page 8: Working Paper

Innovation Diffusion in Developing Countries

8

international product cycle.32

The models were not developed for this purpose, while studies

on international diffusion were increasingly conducted by introducing relevant economic,

soical and cultural variables into the diffusion parameters of individual countries.

In 1978 Peterson and Mahajan constructed a model for diffusion of multi-products

which have interrelationships among them. 33

This model was not orginally intended to apply

in the multicountry markets, but the structure of model is applicable for cross-country

diffusion process, since it has based on the lead-lag phenemenon.34

The model appears as

follows:

dX/dt = (p1 + q1 X) (N1- X)

dY/dt = [p2 + (q2 + 2X) Y] (N2 –Y)

dZ/dt = [p3+ (q3 + 3Y) Z] (N3–Z)

These equations can be explicitly assigned to particular countries. For example, X is

for US, Y is for Japan and Z is for Korea. US is the lead country and Japan is firstly the lag

country, but later it becomes the lead country for Korea. The coefficients s are termed as

lead effect.35

Another diffusion modeling application in a multinational setting was provided by

Eliashber and Helsen. They expanded the Bass-model to reflect the lead-lag phenomena by

incorporating a term that captures the impact of the diffusion in the lead market.36

Mahajan,

Muller, and Kalish used a similar version of the model to analyze whether firms should

launch their product in all their target markets simultaneously („sprinkle djffusion strategy) or

sequentially („waterfall“ strategy).37

.

The cross-country diffusion process was considered for longer time in German

literature. Sartorious38

applied one kind of diffusion model in the combined markets where

every country has a reciprocal effect on diffusion of other countries. The diffusion model was

32

Vernon, R. „International Investment and International Trade in the Product Cycle“, in:Quarterly Journal of

Economics, Vol. 80, No.2, (May 1966), pp. 190-207. See also Wells, L.T. Jr. „International Trade: The Product

Life Cycle Approach“, in: Wells, L.T. Jr. (ed.) „The Product Life Cycle and International Trade“, Boston, 1972. 33

Peterson, R.A. and V. Mahajan, Op. cit. 34

Takada, H. and D. Jain, Op. cit. 35

Ibid. p.52. 36

Helsen, K. et al., Op. cit. p. 63. See also Eliashberg, J.and K. Helsen, „Cross-country Diffusion Processes and

Market Entry Timing“, Working Paper, University of Pennsylvania, 1987 and Eliashberg,, J. and K. Helsen,

„Modeling lead/lag Phenomena in Global Marketing: The case of VCRs“, Working Paper, University of

Pennsylvania, 1996. 37

Ibid. p.63, See also Mahajan, V., E. Muller and S. Kalish, „Waterfall and Sprinkler New-Product Strategies in

Competitive Global Markets“, Working Paper, University of Texas at Austin, 1990 or Kalish, S., V. Mahajan

and E. Muller, (the same title) in International Journal of Research in Marketing, 1995, pp. 105-120. 38

Sartorious, B. „Exportmarkeing ffür neuartige Gebrauchsgüter auf verbundenen Märkten“, Dissertation,

Universität Passau, 1983.

Page 9: Working Paper

Innovation Diffusion in Developing Countries

9

named as Schmalen model39

which has its root in Bass model despite it applies the

compartment system dividing the potential market into innovator segement and imitator

segement. Both Schmalen and Sartorious used their models for the normative purpose of

searching optimal strategies of markeing-mix rather than for the purpose of sales

forecasting.40

Min considered the diffusion process between developed and developing countries

with lead-lag phenomenon, viz. developed countries as lead countries and developing

countries as lag countries.41

The model for diffusion process was constructed accordingly to

include the lead effect of developed countries. Despite the same principle of lead-lag

relationship, his argument was based on cosumer behaviour. It was postulated that in

developing countries consumer categories with respect to adoption of an innovation could be

distisnguished into three types, namely innovators, imitators and international imitators.42

International imitators are imitators whose orientation to imitate is however towards outside

their society, specifically international spreadness and acceptance of the innovation. As

discussed by many authors, developing countries can be characterized by „dualism“

consisting of rich people on one side and poors on the other side. A portion of rich people

who are educated, cosmopolited and internationally outlooked, always try to copy the

lifestyles of people in other countries, especially in advanced countries.43

They are of course

modern elite of rich people and they can play the role of a change agent demonstrating the

innovation in their surroundings, thus the process of diffusion is accelerated. Hypothetically,

they may be pushing the adoption and diffusion process more effectively than innovators.

Mahajan and Muller proposed the model which takes into account the reciprocal

effects of innovation diffusion among member countries of an economic integration.44

The

framework was particularly taken for the unification of European Community in 1992. They

recommended that this unification effort would make faster the diffusion of ideas, products

and technologies, Their model has the following structure:45

39

Schmalen, H. „Marketing-Mix für neuartige Gebrauchsgüter“, Gabler, Wiesbaden, 1979. 40

See. also Schmalen, H. „Markteröffnungssstrategien für Neuheiten“, in ZfB, 41, March 1989, pp. 210-226. 41

Min, Sein „Exportmarketing im Wirtschaftsverkehr zwischen Industrie- und Entwicklungsländern: Eine

computer gestützte Simulationsanalyze zur Ermiitlung optimaler Strategien für die abzatzpolitischen

Instrumente“, Dissertation, Universität Passau, 1990. 42

Ibid. p. 62-63. 43

Ibid. p. 62. 44

Mahajan, V. and E. Muller, „Innovation Diffusion in a Borderless Global Market: Will the 1992 Unification of

the European Community Accelerate Difuusion of New Ideas, Products and Technologies“, Technological

Forecasting and Soical Change, 45, (1994), pp. 221-237.

Page 10: Working Paper

Innovation Diffusion in Developing Countries

10

dX/dt = ( a + b X + q Y ) (N – X)

M+N M+N

dY/dt = ( p + q Y + b X ) (N – Y)

M+N M+N

The extended elements in the adpotion rates of respective countries denote the

incremental contributions to the rates of diffusion. They explained: „By supporting a border-

free zone of market potential size (M+N), unification permits potential customers and

individual adopters in two countries to travel, trade, shop, and settle anywhere in the union.

Because of this freedom, it is reasonable to assume that adopters from each country are also

likely to influence the remaining potential adopters in the other country.“ 46

In the study of cross-country diffusion of industrial technology (retail scanners)

Ganesh and Kumar formulated a diffusion model based on the multi-product model of

Peterson and Mahajan.47

They called it as „learning model“. 48

d F1(t) = [p+q*F1(t)+c*F2(t)]*[1 _

F1(t)]

dt

where

F1(t) = N1(t)]/ m1 = the cumulative penetration ratio till time t for the lag country

F2(t) = N2(t)]/ m2 = the cumulative penetration ratio till time t for the lead country

p = the coefficient of innovation for the lag country

q = the coefficient of imitation for the lag country

c = the learning coefficient for the lag country

In the same study they formulated another version that excluded the innovation coefficient,

thus including only the coefficients of imitation and learning. They assumed that in the lag countries,

the diffusion of industrial technological innovation was seldom shaped by the innovativeness there.

They named their model as „pure learning model“.49

d F1(t) = [q*F1(t)+c*F2(t)]*[1 _

F1(t)]

dt

Ganesh et. al. captured the learning effect by studying diffusion of consumer durables, but the

model used was a little different from the (previous) work of Ganesh and Kumar. The model of

Ganesh et al. oriented more towards the model of Mahajan and Muller for „borderless markets“.50

The

45

Mahajan V. and E. Muller, Op. cit. p. 224. 46

Ibid. 47

Peterson, R.A. and V. Mahajan, Op. cit. 48

Ganesh, J and V. Kumar, „Capturing the Cross-National ................“, Op. cit., p. 332. 49

Peterson, R.A. and V. Mahajan, Op. cit. p. 333. 50

Mahajan, V. and E. Muller, Op. cit.

Page 11: Working Paper

Innovation Diffusion in Developing Countries

11

model formulation was different from their „learning model“ in the way that the learning effect varies

with not only the dynamic influence of the market saturation level in the lead country but also the

market potentials of lead and lag countries. This model can be expressed as follows:51

d F1(t) = [p+q*F1(t)+c*(m1/m2)*F2(t)]*[1 _

F1(t)]

dt

where

m1 = market potential of the lag country

m2 = market potential of the lead country

Modeling Cross-Country Diffusion Process between Developed Countries and

Developing Countries (Extension of the Bass Model)

Although the diffusion research has been extended towards the global context, there

has been not yet a particular research conducted for developing countries. Accordingly we

could not find distinct models which were formulated particularly for developing countries.

As Dekimpe et al. reviewed52

, little is known about the nature of the diffusion process in

developing countries. However, the lead-lag phenomenon as focused by international

diffusion studies can be directed towards the cross-country diffusion process between

developed countries and developing countries. Now we will focus on building an appropriate

model which will focus on developing countries in the context of international diffusion.

When we stage the diffusion process into developed and developing countries,

developed counties act usually as leader whilst the latter as follower. We can assume that

although the extent of diffusion in developed countries will cause an impact on that in

developing countries, the reverse does not occurm in general. As explained earlier, the impact

will increase the adoption rate of innovation through resulting in international imitation or in

ohter word „demonstration effect“,. Thus, we add a new component in the rate of adoption

that reflects incremental rate by international influence as the following:

dX/dt = [p + q (X/M) + r (X*/M*)}[M – X]

where p, q = innvation and imitation coefficient respectively

X = cummulative number of adoption in time t

M = market potential

X*/M* = Level of saturation in lead country, i.e. Industrialized country in time t

r = the coefficient of international influence or the coefficient of international imitation

51

Ganesh, J., V. Kumur and V. Subramaniam, Op. cit. p. 221 52

Dekimpe, M. G., P. M. Parker and M. Sarvary, „Multi_Market und Global Diffusion“, Op. cit. p. 3.

Page 12: Working Paper

Innovation Diffusion in Developing Countries

12

This model is of course an extension of the typical Bass model. This formulation is as

the same as the model used by Ganesh and Kumar („learning model“) in investigating the

industrial technology diffusion. The difference, however, is in theoretical point- of - view and

the interpretation of the new coefficient ( r ). We call the demonstration effect or international

imitation effect, while Ganesh and Kumar termed this phenomen as lerning effect. We believe

that although both are acceptable, the lerning effect is more appropriate for industrial

technology diffusion and the demontration effect is more suitable for consumer durables.

According to the model shown above, it can be expected that the greater „r“ the faster is the

speed of the diffusion process in developing countries. The increasing extent or level of

saturation in developed countries also raises that speed. As a variation of this basic model, we

can introduce a new variable () for time lag in peception of the innovation diffusion in

external environment (lead country) by potential adopters in developing countries.

dX/dt = [p + q (X/M) + r (X*/M*)t-}[M – X]

where = time lag; = 0, 1, 2, 3, ....

Empirical Study

In order to test the appropriateness of the proposed extended Bass model for the

developing countries, we will use data of innovation diffusion in a developing country.

Firstly we will estimate the parameters of Bass model. It is also necessary to have data of

diffusion in an industrialized country to investigate the existence of demonstration effect.

After getting the required data, we will estimate parameters of our extended model for the

same data. Then we will ask ourselves whether the Bass model fits in a setting of developing

country and whether the parameter values lie in normal ranges. For the extended model too,

we will investigate whether the extended model fits with actual data and whether there exists

actually the demonstration or learning effect between developed countries and developing

countries.

Data

We collected annual import volume and units of four consumer durables, namely

computer, room air-conditioner, television set and telephone which were imported into

Myanmar53

during the period from 1985-86 to 1999-2000. Because there was no standard

53

Myanmar is officially a least developed country (LDC). Myanmar can be however considered as a developing

country in general.

Page 13: Working Paper

Innovation Diffusion in Developing Countries

13

publication of data we required, the statistics of import items produced by the Custom Office

were reviewed and collected by the Central Statistical Organization (CSO) in Yangon. Since

Myanamr did not produce these products domestically or if so as in the case of television, it is

negligible, we can assume that the units imported represent the number of adoption for each

consumer durables. The imported volume is expressed in thousands of Ks which is Myanmar

currency. The rate of convrsion between the CIF value in US dollar to Myanmar currency was

not changed significantly over the time, we have already excluded the effect of inflation of the

Myanamr currency.54

Assuming that the prudcts‘ prices denominated in US$ were stable

during the observed time, we think that it is possible to use these data as an alternative to

units. The following table (Tab. 1) shows the import of selected consumer durables for the

period 1985 – 2000. Unfortunately data in units of room air-conditioner and telephone for the

period from 1985- 86 to 1991-1992 are not available. But we estimate them by average value

per unit of the period from 1992-93 to 1999-2000 for which complete data are available. The

estimated data are shown in the brackets. Among the data, it is noticable that the imports of

computers and telephone in the begining of 90s were dropped substantially. It might be due to

the political situation in Myanmar, particularly the turmoil in 1988 and consequently the

siezure of the State’s power by the present military government. .

When we plot the data as shown in Fig. 1, we get different pattterns of diffusion; we

cannot see typical pattern with peak point in almost all products except air-conditioner in Ks.

value. Moreover, very irregular amounts have been seen in telephone both Ks. value and

units.

Regarding with the diffusion of innovatioan in industrialized countries, we took the

data of Japan (lead country), because there are many cultural, economic and trade

relationships between Japan and Myanmar.The consumer durables under study were imported

mainly from Singapore, Japan.and Korea. Moreover, Japan is more or less playing a leading

role to the Far-east and South-east Asian countries. We collected thus the data of innovation

diffusion in Japan. From the UN Statistical Yearbook it is possible to compile two cultural

and communication indicators: possession of TV per 1000 inhibitants and Telephone per 100

inhibitants. We will use these data as time-varying variable that reflects increasing pressure of

international product acceptance inducing potential adopters in developing countries to adopt

the innovation, i.e. international demonstration effect. 55

The two diffusion patterns of Japan

for TV and telephone are shown in the following Tab. 2 and Fig. 2.

54

The official exchange rate was 1 US$ = 6 Ks. 55

The statistics thus show the level of market saturation denoted as (X*/M*)t.

Page 14: Working Paper

Innovation Diffusion in Developing Countries

14

Tab. 1: Diifusion of Innovative Consumer Durables in Burma (from1985-86 to 1999-2000)*

Computer Air-conditioner TV Telephone

Ks Unit Ks. Unit Ks. Unit Ks Unit

1985-86 0 0 6480 n.a.(1536) 37042 20685 1017 n.a.(216)

1986-87 142 15 11814 n.a.(2600) 31116 7270 2291 n.a.(487)

1987-88 948 33 10744 n.a.(2546) 33142 11048 726 n.a.(154)

1988-89 651 37 2267 n.a.(537) 28238 11910 2153 n.a.(458)

1989-90 1138 152 9417 n.a.(2232) 23092 7477 1 n.a.(0)

1990-91 2342 74 10714 n.a.(2539) 97670 52740 0 n.a.(0)

1991-92 211 24 13450 n.a.(3187) 90525 55207 67 n.a.(14)

1992-93 5236 2863 17012 2079 86688 65234 331 738

1993-94 13569 6939 23782 8928 110213 69935 2526 10269

1994-95 13908 6467 35475 4612 147546 65122 3575 414

1995-96 29068 16540 43212 26024 111418 107462 281 33

1996-97 31603 54221 69013 26943 43274 33619 6537 1280

1997-98 32678 54830 97081 18190 2297 13235 12171 2061

1998-99 41895 41840 139680 40746 124228 86959 16203 2647

1999-2000 51837 71051 83295 36979 107952 101003 1264 476

*Ks in thousands

.n.a. = not available and (....) = estimated units

Source: compiled by the Central Statistical Organization (CSO), Yangon, Myanmar

Page 15: Working Paper

Innovation Diffusion in Developing Countries

15

Fig. 1: Diffusion Patterens of Selecdted Consumer Durables in Myanmar

Computer (Units)

0

20000

40000

60000

80000

86-

87

88-

89

90-

91

92-

93

94-

95

96-

97

98-

99

Year

Un

its

Computer (Ks.)

0

10000

20000

30000

40000

50000

60000

86-

87

88-

89

90-

91

92-

93

94-

95

96-

97

98-

99

Year

Ks.

(,000

)

Air-conditioner (Ks.)

0

50000

100000

150000

85

-86

87

-88

89

-90

91

-92

93

-94

95

-96

97

-98

99

-00

Year

Ks

. (,

00

0)

Air-conditioner (Units)

0

10000

20000

30000

40000

50000

85-

86

87-

88

89-

90

91-

92

93-

94

95-

96

97-

98

99-

00

YearU

nit

s

Page 16: Working Paper

Innovation Diffusion in Developing Countries

16

TV-set (Ks.)

0

50000

100000

150000

20000085-

86

87-

88

89-

90

91-

92

93-

94

95-

96

97-

98

99-

00

Year

Ks.

(,000

)

TV-set (Units)

020000

4000060000

80000100000

120000

85

-86

87

-88

89

-90

91

-92

93

-94

95

-96

97

-98

99

-00

Year

Ks

. (,

00

0)

Telephone (Ks.)

0

5000

10000

15000

20000

85-

86

87-

88

89-

90

91-

92

93-

94

95-

96

97-

98

99-

00

Year

Ks.

(,000

)

Telephone (Units)

02000400060008000

1000012000

85-8

6

87-8

8

89-9

0

91-9

2

93-9

4

95-9

6

97-9

8

99-0

0

Year

Un

its

Page 17: Working Paper

Innovation Diffusion in Developing Countries

17

Tab. 2: Innovation Diffusion in Japan

(TV and Telephone)

Year

TV Telephone

(per 1000 inhibitants)

(per 100 inhibitants)

1985 579 55.5

1986 n.a. n.a.

1987 n.a. 39.7

1988 n.a. 41.1

1989 610 42.6

1990 611 44.1

1991 613 45.4

1992 614 46.4

1993 618 47.1

1994 681 47.9

1995 684 48.7

1996 683 48.9

1997 686 47.9

Source: Statistical Yearbooks, UN

Parameter Estimation

Bass Model

We estimate firstly the parameters of Bass-model, namely p (the coefficient of

external coefficient), q (the coefficient of internal coefficient) and m (the market potential).

The market potential can be also determined from other exogenous sources of information

rather than from the diffusion model and as a result better results of estimation for p and q can

Fig. 2: Innovation Diffusion of TV and

Telephone in Japan

0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

19

85

19

87

19

89

19

91

19

93

19

95

19

97

Ma

rke

t S

atu

rati

on

(%

)

TV

Telephone

Page 18: Working Paper

Innovation Diffusion in Developing Countries

18

be obtained.56

For the estmation of all three parameters in Bass model, it is possible to use

different estimation methods. They are:57

1. The ordinary least squares (OLS) procedure (Bass 1969)58

2. The maximum likelihood estimation (MLE) procedure (Schmittlein and Mahajan 1982)59

3. The nonlinear least squares (NLS) procedure (Srinivasan and Mason 1986)60

4. The algebraic estimation (AE) procedure (Mahajan and Sharma1985)61

The fourth method, the algebraic estimation method, is to be used when no prior data is

available. Mahajan and Sharma developed the estimation method by which the parameters

p and q of the Bass model are estimated. It requires however information from managers

about three items: (1) the market size (m), (2) the time of the peak of the noncummulative

adoptioin curve, and (3) the adoption level at the peak time (n*).62

We should also note

here the criticism by F.M. Bass: „one of the key outputs of the diffusion model is the

prediction of the timing and magnitude of the peak. Therefore, if one can guess these

items, there is no need to estimate model parameters.“ 63

Since we have already had a series of historical data about diffusion of consumer

durables under study, we can use the first three methods of estimation described above. We

beign with OLS method. The estimation results are summarized in Tab.3.

56

Mahajan, V., E. Muller and F.M.Bass, Op. cit., p. 6. 57

Mahajan, V. and Y. Wind „Innovation Diffusion Models of New Product Acceptance: A Reexamination“, in

Mahajan, V. and Y. Wind (eds.) „Innovation Diffusion Modes of New Product Acceptance“, N.Y., Ballinger,

1986, pp.3-25, p. 11. 58

Bass, F.M. Op. cit. 59

Schmittlein, D.C. and V. Mahajan, „Maximum Likelihood Estimation for An Innovation Diffusion Model of

New Product Acceptance“, Marketing Science, Vol. !, No. 1, Winter 1982, pp. 57-78. 60

Srinivasan, V. and C:H. Mason, „Nonlinear Least Squares Estimation of New Product Diffusion Models“,

Marketing Science, Vol. 5, No. 2, Spring,(1986), pp. 169-178. 61

Mahajan, V. and S. Sherma, „Simple Algebric Estimation Procedure for Innovation Diffusion Models of New

Product Acceptance“, Techonological Forecasting and Social Change, 30 (1986), pp. 331-46. 62

Mahajan, V., E. Muller and F.M.Bass, Op.cit., p. 8.

Page 19: Working Paper

Innovation Diffusion in Developing Countries

19

Tab. 3: Estimated Parameters' Values and R2 Values in Bass Model (OLS)

Product Name

Basis p q m R2

(Significant Level)

Computer Ks. 0.0064 0.5884 316894 0.916 (0.00)

Computer Units 0.0086 0.8646 289367 0.834 (0.00)

Air-conditioner

Ks. -2.45578E-08 0.4039 1,09783E+11 0.861 (0.00)

Air-conditioner

Units -0.0019 0.5030 301219 0.831 (0.00)

TV Ks. 0.0142 -0.0036 4138681 0.253 (0.174)

TV Units 1.5169E-07 0.2090 1,02956E+11 0.431 (0.034)

Telephone Ks. -0.0497 0.8662 53144 0.552 (0.008)

Telephone Units 0.0490 0.3060 19870 0.019 (0.889)

When we look at the results, it has been found that the Bass model fits well with two

products, namely computer and air-conditioner (both volume and units), while it cannot

explain the diffusion process of the rest products, namely TV and telephone. This may be due

to the distinctive nature of original data as we have noticed in the diagrams. Even though the

model fits with the data of air-conditioner, the parameter values of p are given in minus sign.

Thus it must be said that the Bass model generates unplausible results in some products

diffused in developing country (Myanmar). Now again we can compare the parameter values

of our study with those in international study. Sultan et al. reported in their meta analysis of

15 diffusion studies, that the coefficient of innovation (p) averages 0.03 and the coefficient of

imitation (q) averages 0.38 but values vary very considerably.64

Compared with this, we can

find that the parameters of our study show different results beyond the internatioal range.

Particularly, the innovation coefficient p is much smaller than the international average of

0.03, while the coefficient of q is much greater than the international average of 0.38 (at least

in the case of computer).Hence, we can interpret that in developing countries the diffusion

process is mainly accelerated by the imitators rather than innovators in relation to developed

countries. As expected and agreed with theoretical point of view, we propose thereby that

developing countries are less innovative, however, they are more imitative than developed

countries. Moreover, our results support to the following hypothesis of Takada and Jain:

63

Mahajan, V., E. Muller and F.M.Bass, Op.cit., p. 8. 64

Sultan, F., J.U. Farley und D.R. Lehmann, „A Meta-Analysis of Applications of Diffusion Models“, Journal of

Marketing Research, 27b (1990), pp. 70-77, p. 71.

Page 20: Working Paper

Innovation Diffusion in Developing Countries

20

„The later a product is introduced in a market, the faster will be the rate of adoption.

Consequently, the imitation coefficient will have a larger for a country in which th product is

introduced later than for the country in which the product is first introduced.“65

In addition to the OLS estimation, we have also applied the nonlinear least square

(NLS) estimation method proposed by Srinivasan and Mason. In this method, the sales X(i) in

the ith time interval (ti-1, ti) is given by:66

X(i) = m[F(ti) – F(ti-1)] + ui

1 – e – (p+q) ti

1 – e –(p +q ) ti-1

X(i) = m [ ________________ _ __________________ ] + ui for i=1, 2, ....., T,

1 – (q/p) e- (p+q)q) ti

1 – (q/p) e –(p +q ) ti-1

where m = the number of eventual adopters

In OLS method the parameters p, q and m are not directly estimated and the regression

analog equation of the Bass model has been instead applied, we cannot obtain the signifiance

level and standard error of each parameter. Moreover, there can be some problems with

multicolinearity among input data, since the cummulative adoption at a time has been used

twice but in different orders.67

Using the nonlinear least squares (NLS) estimation, the

parameters are directly estimated and standard errors are also provided by using asymptotic

approximations.68

65

Takada, H. and D. Jain, Op. cit., p. 50. 66

Srinavasan V. and C. H. Mason, Op. cit., p. 170. 67

Mahajan, V. E. Muller and F.M. Bass; Op. cit. p. 9. See also Schmittlein, D.C. and V. Mahajan, Op. cit,

p. 59-60. 68

Srinivasan, V. and C.H. Mason, Op. cit. p.171

Page 21: Working Paper

Innovation Diffusion in Developing Countries

21

The following is the results estimated by NLS.

Tab. 4: Estimated Parameters' Values and R

2 Values in Bass Model (NLS)

by Srinivasan and Mason(1986)

Product Name Basis p q m R2

Computer Ks. 0.00114 0.41182 490309 0.96755

Computer Units 0.00025 0.58613 431500 0.89295

Air-conditioner Ks. 0.00006 0.69123 668256 0.92233

Air-conditioner Units 0.00058 0.44385 345763 0.87755

TV Ks. 0.15832 0.27926 1313059 0.32518

TV Units 0.10913 0.22211 1200694 0.44496

Telephone Ks*. 0.00004 0.75234 62534 0.61748

Telephone Units** 0.00018 1 17493 0.2681

* Constraint: p greater than or equal to 1

** Constraints: both p and q greater than or equal to 1

When we compare the results between the OLS and NLS estimation, it has been found

that NLS method produces higher levels of goodness of fit (R2) than OLS. Moreover, NLS

can produce plausible results (with proper direction of signs), because it is possible to set the

constraints like p > 0, q > 0. Nevertheless, the interpretation on parameters are not different

from that of OLS.

The Extended Bass Model

We will now estimate the parameters of the extended Bass model which is built to

apply in the setting of developing countries. The model as described earlier requires to

estimate four parameters: p, q, r and m. Since it is impossible to estimate by means of OLS,

we will use the nonlinear least square (NLS) estimation method available in the SPSS

software. We estimate, however, only three parameters, p, q and r; m is set according to the

value obtained in earlier estimation works. Here we take the values from NLS estimation by

Srinavasan and Mason. As explained in data, data for lead country (Japan) are available for

only two products, TV and telephone, hence we estimate the parameters of the extended Bass

model for these two products. Tab.5 shows the results of our estimation.

Page 22: Working Paper

Innovation Diffusion in Developing Countries

22

Tab. 5: Estimated Parameter Values and R2 Values in Extended Bass Model

by NLS*

Product Name Basis p q r R2

TV Ks. 0 0.2585 0.0366 0.3484

TV Units 0 0.2095 0.0257 0.4102

Telephone Ks. 0 0.6832 -0.0579 0.5011

Telephone Units 0 0.4029 0.1211 0.0126

* Constraint: p greater than or equal to 1

When we look at the results, we find firstly that the parameter p’s values are changed

into 0: this is also due to imposing of the constraint p = 0 or p > 0. Without this constraint, p

values result in negative sign.69

In place of innovative coefficients, the parameter r

(coefficient of international imitation) enters into existence. But an unplausible minus sign

appears for telephone (Ks.). As a result of the existence of the influence of international

imitation (r), the coefficient q’s values become lower. When we look at the degree of

goodness of fit, the extended model can improve a little bit only in one case (TV –Ks.) and in

the other three cases the model’s degree of fitness has decreased.

At the next stage, we examine the lag effect of international diffusion i.e. the

saturation level in industrialized countries, here Japan. The results shown above are obtained

for the condition in which the international diffusion is perceived simultaneously by the

potential adopters in developing countries, here Myanmar. It means that there is no time lag (

= 0). We have tested the effect of varying time lag ( = 1, 2 and 5). on the parameter r and the

degree of goodness of fit (R2). These results are shown in the Tab. 6.

When we analyze the effects of time-lag in the perception of international diffusion by

potential adopters in developing countries, we find that the best fit comes from simultaneous

perception instead of time-lag. It can be also seen that the longer the time lag, the lesser is the

degree of goodness of fit as well as the value of parameter „r“. As there is a long time-lag and

consequently the coefficient of international imitation, paramter „r“ declines and the

coefficientof innovation or parameter p comes into existence again.

69

This condition is comparable with the estimation results of Ganesh and Kumar, Op. cit. p. 334-335. In their

study of diffusion of retail scanner, the coefficient p values were negative when the learning effect model was

applied.

Page 23: Working Paper

Innovation Diffusion in Developing Countries

23

Tab. 6: The Effect of Time-lag in perception of international Diffusion on Parameters and

Goodness of Fit

Product Type Base Time Lag ()

= 0 = 1 = 2 = 5

TV Ks. p 0 0.0212 0.0212 0.0212

q 0.2585 0.2651 0.2651 0.2651

r 0.0366 0 0 0

R2 0.3484 0.3467 0.3467 0.3467

TV Unit p 0 0 0.0151 0.0151

q 0.2095 0.2096 0.216 0.216

r 0.0257 0.026 0 0

R2 0.4102 0.4098 0.4092 0,4092

Telephone Ks. p 0 0 0 0

q 0.6832 0.5564 0.5564 0.5564

r -0.0579 0 0 0

R2 0.5011 0.4611 0.4611 0.4611

Television Unit p 0 0 0 0.0503

q 0.4029 0.4121 0.4197 0.4964

r 0.211 0.116 0.112 0

R2 0.0126 0.0102 0.006 0.0048

Conclusion

Innovation diffusion is much interested by marketers, because this phenomenon can

explain how a new innovative product spreads in a market. As traditional domestic marketing

is expanded towards international marketing and global marketing, the application of innvatin

diffusion and its models go also beyond a single country, and the multiple countries become

the focus of the study. The international aspect of innovation diffusion has been widely

considered in 1990s. The international study of innovation diffusion comprises two kinds:

breadth and depth. Some studies attempted to link between countries applying the lead-lag

concept, while the others continued to study individual markets without such relationship. It

has been also found that the models used in the cross-nation studies include Bass-type models

and non-Bass type models.

Despite the increased interest in multinational aspects of innovation diffusion, no

emphasis on developing countries has been given. Some authors like Dekimpe et al. stress the

need to focus on developing countries with respect to innovation diffusion. It can be expected

that there are some differences between developed countries and developing countries. For

Page 24: Working Paper

Innovation Diffusion in Developing Countries

24

instance, the innovation diffusion in developing countries may be slower due to economic,

social and cultural factors, whereas it may be speedy due to time effect (together with

demonstration effect or learning effect generated by developed countries) and greater word-

of-mouth communication. Todays, as there are more and more developments in

communication and globalization emerges, however, we need to investigate thoroughly the

actual diffusion processes in developing countries.

We proposed a cross-nation diffusion model extending basic Bass-model. In the

literature, we found that our proposed model is very similar to that of Ganesh and Kumar who

studied the industrail technology diffusion. In our model, we add a new type of coefficient

which reflects the behaviour of international imitation into the adoption rate of a period. Thjs

model is appropriate for developing countries, because the consumers there orient towards the

consumption in developed countries: this kind of behaviour is obviously seen in a particular

segment of elite consumers. We termed the resulting effect as international demonstration

effect rather than learning effect. By this way, our proposed extended model is in agreement

with theorectical behavioural premises. We tested the soundness of our model with empircal

data.

We used the data of innovation diffusion for some consumer durables from two

countries: Myanmar and Japan. Myanmar plays the role of developing countries and Japan for

developed countries. So Japan is a lead country and Myanmar is a lag country. We firstly

tested the appropriateness of Bass model by estimating parameters. Out of four products

(computer, air-conditioner, TV set and telephone), only the first two products showed a high

degree of goodness of fit. By this point, we like to conclude that Bass model cannot be always

applicable when innovatin diffusion in developing countries is investigated. The reason

behind this is that in developing countries, many factors can influence in the diffusion

process. For example, government policy and restrictions, political situation, trade

relationships, infrastructure, etc. may be imposing the constraints upon free process of

diffusion. Compared to this, developed countries may have negligible restrictions and

constraints. In our Myanamr data, we can obviously see the irregularities in yearly adoption

amounts in telephone. Under such conditions, the Bass model will not function well.

When we studied the parameter values, the results were as we expected. The parameter

of innovation or external influence (p) is very small and the parameter of imitation or internal

influence (q) is considerably large. It can be even said that the adoption behaviour in

developing countries is shaped solely by imitators. In the parameter estimation, we applied

two methods, ordinary least square (OLS) and nonlinear least square (NLS). NLS is clearly

Page 25: Working Paper

Innovation Diffusion in Developing Countries

25

better than OLS in terms of degree of goodness of fit (R2) values. We found also that using

Bass model estimated by OLS had produced in some cases unplausible results, especially the

wrong sign.

When we tested the soundness of the extended Bass model which is particularly built

for developing countries, the results were not decisive. Due to unavailablity of data of the

lead-country (Japan) in some consumer durables, namely computer and air-conditioner, we

could only test in two products which data originally do not show typical diffusion patterns.

Out of our estimatin work, we could find however the existence of influence of internatinal

imitation. The parameter ( r ) was determined in all cases, except that there was a case with

negative sign. Another finding is that the innovation coefficient (p) was no more in existence

as the new parameter ( r ) entered. Thus, we like to propose that the diffusion process in

developing countries is merely the imitation process (comprising conventional imitation and

international imitation) without any effect of innovative behaviour. This finding is very

plausible in the context of developing countries. We cannot, however, conclude firmly this

pattern because we found that the extended model could not improve the goodness of fit

except in one case and the extent of improvement is not significant. Actually we need to study

further and collect findings in order to make sound conclusion about this peculiar pattern.

We have studied also the effect of time-lag in the perception of internatinal diffusion.

We entered the degree of international diffusion differently with time-lag 1, 2 and 5, we found

that the best fit occurs when there is a simultaneous perception without delay. As there is a

time-lag or delay, the model’s abiltiy to fit with actual data declines. The parameter of

international diffusion becomes also smaller and smaller till it reaches zero. Again the

innovation coefficient (p) appears and exits, as the parameter of international imitation

disappears due to time-lag.

By looking these results, we like to conclude that the extended model has a good

potential to be used in the setting of developing countries, although the model cannot

guarantee its soundness. It is probable that the model’s usefulness depends on the particular

case. Thus, we need to contiume to test the model with many atual data. Here we can propose

to study a theoretical aspect of innovation diffusion in developing countries, whether there

exists innovation inflence or international imitation influence. So we raise a question about

the dilemma of innovator vs. international imitator in developing countries. The

comprehensive scale of study in developing countries about the diffusion process to be

conducted in the future will answer this.

Page 26: Working Paper

Innovation Diffusion in Developing Countries

26

References

Bass, F. M. „A New Product Growth Model for Consumer Durables“, Management Science, 15(1969), pp. 215-2127.

Dekimpe, M.G., P.M. Parker and M. Sarvary, „Comparing Adoption Patterns: A Global Approach“, Working Paper (96/37/MKT), INSEAD; 1996. --------------------------------------------------------------„Globalization: Modeling technology adoption timing across countries“, Working Paper (96/38/MKT), INSEAD, 1996. ---------------------------------------------------------------„Global Diffusion of Network Technologies: A Double-Hazard

Approach“, Working Paper (97/50/MKT), INSEAD, 1997.

-------------------------------------------------------------- „Multi-Market and Global Diffusion“, Working Paper, INSEAD, 1998.

Eliashberg, J.and K. Helsen, „Cross-country Diffusion Processes and Market Entry Timing“, Working Paper, University of Pennsylvania, 1987. Eliashberg,, J. and K. Helsen, „Modeling lead/lag Phenomena in Global Marketing: The case of VCRs“, Working Paper, University of Pennsylvania, 1996

Ganesh, J. and V. Kumar, „Capturing the Cross-national Learning Effect: An Analysis of an Industrial Technology Diffusion“, Journal of the Academy of Marketing Research, 24 (1996), pp. 328-337.

Ganesh, J., V. Kumar and V. Subramaniam, „Learning Effect in Multinational Diffusion of Consumer Durables: An Exploratory Investigation“, Journal of the Academy of Marketing Science, 25 (1997), pp. 214-228.

Gatignon, H. and T.S. Robertson, „ A propositional Inventory for New Diffusifon Research“, Journal of Consumer Research, Vol. 11, March 1985, pp. 849-867.

Gatignon, H., J. Eliasherrg and T.S. Robertson, „Modeling Multinational Difusion Patterns: An Efficient Methodology“, Journal of Management Science, Vol. 8, No. 3, Summer 1989, pp. 231- 246.

Heller, R. M. and T.P. Hustad „Problems in predicting New Product growth for Consumer Durables“, Management Science, 26 (1980), pp. 1007-1020. .

Helsen, K. K. Jedidi, K. and W.S. DeSarbo, „A New Approach to Country Segmentation utilizing Multinational Diffusion Patterns“, Journal of Marketing, 57(1993), pp. 60-71..

Mahajan, V., E. Muller and S. Kalish, „Waterfall and Sprinkler New-Product Strategies in Competitive Global Markets“, Working Paper, University of Texas at Austin, 1990

Mahajan, V. and E. Muller, „Innovation Diffusion in a Borderless Global Market: Will the 1992 Unification of the European Community Accelerate Difuusion of New Ideas, Products and Technologies“, Technological Forecasting and Soical Change, 45, (1994), pp. 221-237.

Mahajan, V. and S. Sherma, „Simple Algebric Estimation Procedure for Innovation Diffusion Models of New Product Acceptance“, Techonological Forecasting and Social Change, 30 (1986), pp. 331-46.

Mahajan, V. and Y. Wind „Innovation Diffusion Models of New Product Acceptance: A Reexamination“, in Mahajan, V. and Y. Wind (eds.) „Innovation Diffusion Modes of New Product Acceptance“, N.Y., Ballinger, 1986, pp.3-25.

Mahajan, V. E. Muller and F.M. Bass, „New Product Diffusion Models in Marketing: A Review and Directions for Research“, Journal of Marketing, 54(1990), pp. 1-26.

Mahajan, V. and M:E.F. Schoeman, „Generalized Model for the Time Pattern of the Diffusiom Process“, IEEE Transactions on Enginerring Management, Feb. 1977, pp. 12-18.

Min, Sein „Exportmarketing im Wirtschaftsverkehr zwischen Industrie- und Entwicklungsländern: Eine computer gestützte Simulationsanalyze zur Ermiitlung optimaler Strategien für die abzatzpolitischen Instrumente“, Dissertation, Universität Passau, 1990. Peterson, R. A. and V. Mahajan, „Multi-Product Growth Models“, in „Research in Marketing“ ed. Jagdish Sheth, Greenwich, CT: JAI Press, 1978.

Rogers, M. „Diffusion of Innovations“, New York , The Free Press, 1995. Sartorious, B. „Exportmarkeing ffür neuartige Gebrauchsgüter auf verbundenen Märkten“, Dissertation, Universität Passau, 1983.

Schmalen, H. „Marketing-Mix für neuartige Gebrauchsgüter“, Gabler, Wiesbaden, 1979.

Page 27: Working Paper

Innovation Diffusion in Developing Countries

27

Schmalen, H. „Das Bass-Modell zur Diffusionsforschung: Darstellung, Kritik und Modifikation“, ZfbF, 1989, pp.1191-1209. Schnmlen, H. and H. Xander, „Produkteinführung und Diffusion“, in S. Albers and A. Hermann (ed.), “Handbuch Produktmanagement: Strategieenwicklung – Produktplanung – Organisation – Kontrolle“, Gabler, 2000. Schmittlein, D.C. and V. Mahajan, „Maximum Likelihood Estimation for An Innovation Diffusion Model of New Product Acceptance“, Marketing Science, Vol. !, No. 1, Winter 1982, pp. 57-78.

Srinivasan, V. and C:H. Mason, „Nonlinear Least Squares Estimation of New Product Diffusion Models“, Marketing Science, Vol. 5, No. 2, Spring,(1986), pp. 169-178. Takada, H. and D. Jain, „Cross National Analysis of Diffusion of Consumer Durable Goods in Pacific Rim Countries“, Journal of Marketing, 55 (1991), pp. 48-54. Tanny, S.M. and N. A. Derzko, „Innovators and Imitators in Innovation Diffusion Modelling“, Journal of Forecasting, Vol. 7, 1988, pp. 225-234. United Nation, „Statistical Yearbook“, various issues. Vernon, R. „International Investment and International Trade in the Product Cycle“, in: Quarterly Journal of Economics, Vol. 80, No. 2, (May 1966), pp. 190-207. Wells, L.T.Jr. „International Trade: The Product Life Cycle Approach“, in: Wells, L.T.Jr. (ed.) „the Product Life Cycle and International Trade“, Boston, 1972.