LIBRARIES MICHIGAN STATE UNIVERSITY EAST LANSING, MICH 48824-1048 This is to certify that the dissertation entitled MULTINATIONAL DIFFUSION THEORY: A MACRO LEVEL ANALYSIS presented by ELIF SONMEZ has been accepted towards fulfillment of the requirements for the PhD. degree in Business Administration 0 WjoTTDrofessor’s Signature May 9, 2005 Date MSU is an Affirmative Action/Equal Opportunity Institution PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE DUE DATE DUE DATE DUE AUG 2 0 2005 # 2/05 c:/ClRC/DateDuo.lndd—p.15 MULTTNATIONAL DIFFUSION THEORY: A MACRO LEVEL ANALYSIS By Elif Sonmez A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Marketing and Supply Chain Management 2005 VINE-U \ .0 titration! 3.1.. mafiadm. 17.1%: of Sini'C.\ 351$} concerts: mctlc. as fir 21c are it Ex’f'llilll is to c Fifi; ilrect intc ti inflow an "iii: “‘3ch the . 5—: Q ' I‘ ' ,Mb itilm‘N‘C x F. k ' wing-mo do 610 .. ”a 9 J Li $.th Rdlmtj K‘I‘CC‘ZC'EI. ltlc Mason ol' énnma «:35. let. tin it‘s. J: ’5” , \; :fi\l'-"' . ~"‘~ \.Ls\! I \ ‘ go It i=5»: wt It shall} trace 4' .. . 8 45d dC‘iC ABSTRACT MULTINATIONAL DIFFUSION THEORY: A MACRO LEVEL ANALYSIS By Elif Sonmez Multinational diffusion of innovations is an increasingly important topic from both theoretical and managerial perspectives. From the theoretical perspective, a limited number of studies have focused on multinational diffusion of innovations, and these mostly concentrated on developed rather than developing countries. From the managerial perspective, as firms become more involved in global business, understanding how innovations are accepted in new markets becomes imperative. The purpose in this dissertation is to examine the impact of macro level globalization drivers (trade volume, foreign direct investment and income) on the diffusion of consumer durable products in both developed and developing countries. The model is based on the Generalized Bass Model, where the covariates are the globalization drives. Data for four consumer durable products (compact disc player, home computer, mobile phone and video camera) in twenty-two developed and twenty-one developing countries are analyzed using the Augmented Kalman Filter with Continuous State and Discrete Observations methodology. The results suggest that the macro level globalization drivers effect the diffusion of innovations process in a similar manner in developed and developing countries. Yet, significantly different diffusion parameters in developed and developing countries suggest that the diffusion patterns in developed and developing countries are not similar. It should also be noted that diffusion of innovations happens faster in both developed and developing countries due to increasing coefficient of innovation over time. Copyright by ELIF SONMEZ 2005 1r ind .1...» teen 3 Ion .11} somnztzt '. “C i..- ‘ '11.?)nuiii: t :L‘r 1 I'm- .' '- bummer: &.A I ,-.;.‘:;'a ten str. 31-1"... Due to l 3::1: ll: hub 91; 1e lilOll'i an. in w .. - “*5. \Olislam q 75‘3“ ‘ “t he 3:: c1 'I‘y ..,L ~L, .5") ' ‘tfl‘m m} ‘63: P ‘ ' . .. .1 $0.651: J‘ l0 1..- a" 4!. J153}. m‘ "Q: n; - w. . 4. a... 31: 3m“: lg -. "K"..- «Q. Ckr‘] I a,” c —u‘; 1. .; 55“»; -."‘I\.‘ v ‘ . *Tlt; ht; E i. .2 I, ‘V wit} w} ' “Hist. 'ht'i " n A.§’_‘ ‘ A . \ A. w .0; IF, L _r. . ”I: \: .v L d1 ACKNOWLEDGEMENTS This has been a long and difficult road. I am indebted to many people. I would like to thank my committee members Roger Calantone, Tamer Cavusgil, Ram Narasimhan and Jon Bohlmann for their valuable guidance and support. In particular, Dr. Calantone’s and Dr. Cavusgil’s patience and kindness helped me through quite difficult times. My brother Murat, who is very dear to me, has always been an important part of my life. He has provided a very strong emotional support throughout the years. He is also my technical consultant. Due to his efforts, I completed this dissertation with minimum computer problems. My husband Michael joined the journey toward the end. I appreciate him giving me motivation at times I was slowing down. I also thank him for discussing with me the statistical issues in the dissertation and providing support with data analysis. My aunt Serefnaz and my mother Gulten, who are thousands of miles away in Turkiye, have been my constant supporters through their prayers. Zeynep Altinsel and Emin Civi stand out among the several good friends who shared the daily ups and downs of life with me throughout my years at MSU. I would like to thank them for their invaluable friendship. I am also indebted to Kathy F orman, Ioanna Kalogiros, Meltem Avsar, Destan Kandemir, Pinar Ozbay, my fellow classmates Katrina Savitskie, Sangphet Hanvanich and Rosanna Garcia, and the staff in the Marketing and SCM Office (particularly Laurie Fitch, Kathy Mullins, Cheryl Lundeen and Kathy Waldie) for their support. Korkut Uygun is another dear fi'iend, who helped me with the computer code when I was stuck badly. I would not be able to finish the data analysis without his contribution. I am very fortunate and gratefill for being surrounded with such wonderful people through this difficult journey. iv ".5: or 1.131-155 {ZJLQIER l 373:)D‘.‘Cilt.l\ D313 and 5‘ Erzmazior. Conti‘iill‘] P135 0.: the ELDTCR " mum: RE Ks} Stalin" AISLE-id: .l: II“- ¥ ‘. I j «r‘ . .§ § 1"‘ ;.&. 5 IL: (/3 (/1 :ih’lER 3 lfjiilllDOIMi Y The Oxcn it Models Mil” Med \11 xi Mud Mariel: 1111;: \lflillkli‘lil‘c" “...1 Stan Ci‘lllil Aiglli D321 and D1: 1 . I. a. o». TABLE OF CONTENTS LIST OF TABLES ........................................................................... vii CHAPTER 1 INTRODUCTION ............................................................................ 1 Data and Sources of Data ........................................................... 10 Estimation Methodology ............................................................ 12 Contributions .......................................................................... I 4 Plan of the Dissertation .............................................................. 16 CHAPTER 2 LITERATURE REVIEW .................................................................... 17 Key Studies in the Multinational Diffusion of New Products Literature. . 21 Methodological Overview of the Diffusion of New Products Literature. . ....42 Single-equation time-invariant estimation procedures .................. 43 Single-equation time-variant estimation procedures .................... 48 Simultaneous equation estimation ......................................... 57 CHAPTER 3 METHODOLOGY ........................................................................... 66 The Overview of the Bass Model .................................................. 69 Models with Marketing Mix Variables ............................................ 71 Models with Price Alone ................................................... 71 Models with Advertising Alone ........................................... 76 Models with Price and Advertising ....................................... 78 Models with Macro Variables as Covariates ..................................... 82 Methodology and Model ............................................................ 88 What is a Kalman Filter? ................................................... 88 Standard Kalman Filter vs. Augmented Kalman Filter with Continuous State and Discrete Observations ............................ 91 Algorithm ..................................................................... 96 Data and Data Sources ............................................................... 99 CHAPTER 4 DATA AND ANALYSIS .................................................................... 100 Data .................................................................................... 100 Model .................................................................................. 101 Algorithm .............................................................................. 101 Analyses ............................................................................... 107 “Ur-Tr- < 1124;» ~ ggflJSANL w , . (Ci-11:2?“ liniuziti srxarrs 1991:: APPI\l APPI\l APPEXI APPLY CHAPTER 5 RESULTS AND DISCUSSION ............................................................ 112 Contributions ......................................................................... 1 26 Limitations and Future Research ................................................... 129 APPENDICES APPENDIX A: Model Parameters for Each Product and Country ............. 131 APPENDIX B: p and q Over Time for Each Product and Country ............ 175 APPENDIX C: Diffusion Patterns for Each Product and Country ............. 219 APPENDIX D: p and q Over Time for Each Product in Developed and Developing Countries ........................ 263 APPENDIX E: p and q Averages for Each Product in Developed and Developing Countries ........................ 280 APPENDIX F: p and q Over Time in Developed and Developing Countries for Each Product .................................................... 289 APPENDIX G: Interaction Effects ................................................ 294 LIST OF REFERENCES .................................................................... 300 vi LIST OF TABLES Table 2.1 Summary of International Diffusion of Consumer Durables Research. . ...62 Table 4.1 Results with and without Kalman Filtering .................................... 111 Table 5.] ANOVA Results for p ............................................................ 113 Table 5.2 ANOVA Results for q ............................................................ 114 Table 5.3 ANOVA Results for a ............................................................ 115 Table 5.4 ANOVA Results for b ............................................................ 116 Table 5.5 ANOVA Results for c ............................................................ 117 Table 5.6 MANOVA Results ................................................................ 118 Table 5.7 MANOVA Hypotheses and Results ............................................ 120 Table 5.7 Tukey’s Studentized Range (HSD) Test (for effect of country type) ...... 120 Table 5.8 Tukey’s Studentized Range (HSD) Test (for effect of product type) ...... 120 Table 5.9 Marginal Cell Means (for interaction effects) ................................. 121 vii "t‘szar; 01 no: ya. 51 11:11:11: .‘I ‘oh 1 ‘ \ JI' ‘ ' ' r‘ fil‘d‘bi O ‘. iii 1* 136 term ~- 57 ’7‘ J r 14—...65 T111316 I saw-"1.1 .1. ‘V h “01.1%: econ“ “mm" DC“ I 4.. 3‘ \ 4115‘; .3 T . x, \ 5, 1 \ \au\l p. \"\.I?_~.- "u\\ , . ““551 -.' ”KW .-,'_ ‘.1“l‘ \ ’l. LJI,‘ ~.j ‘ 5‘?“ :Ln h ..l CHAPTER 1 INTRODUCTION Diffusion of new products research has advanced a great deal in the recent decades. Researchers have succeeded in shedding light on what shapes and influences within- country diffilsion patterns. Issues regarding multinational (cross-country) diffusion of new products, however, have received limited attention. Due to recent economic trends such as the removal of political and trade barriers, increased foreign competition, and saturated home markets, firms have increasingly been seeking to establish global presence and compete effectively in the global marketplace. With rapid globalization of the world economy, cross-country diffusion has become a central issue for firms launching new products in foreign countries. The crucial role of new products for a firm’s viability as well as its competitive performance is well documented in the literature (e.g., Day and Wensley 1988). Firms must create and sustain competitive advantage (Porter 1985). The ability to develop and launch new products successfully is a major determinant of a firm’s competitive advantage. Such firms are more likely to increase their market shares and profits. To ensure success in introducing their products in foreign markets, firms need a solid understanding of their target markets. Part of this understanding comes from the knowledge of how their products diffuse in different countries, and why there are similarities or differences in the diffusion process between countries. 9.111111. court 11:31.1: in; 1.....15 116311 ;-..11 931:1: in T1: Jain: tren— I v- m- a , 1“ 5. .AA.‘."‘ ‘, L 5"- ) I. . \u‘ 5 ‘3'::“.i \“ “‘14 the 1. A ‘l “1.11.: ’ n." - ‘ " ‘r . I “3'ch- I. gun‘s»: \ ‘v- _ N~~ 11‘1r-“‘ «C L‘. r.‘\1~‘ 1 “(I 17A, . 14.- ‘.‘ "_ “\.-. . u 1 - . v l (‘1’ ”- I- Although countries differ from each other significantly, there is still considerable/increasing convergence in the markets due to globalization in the recent decades (Levitt 1983, and Ohmae 1985). The world economy has been moving toward a global system in which national economies are increasingly becoming interdependent. The decline trends in trade and investment barriers are among the main drivers of this globalization process. After the experience of the Great Depression of the 19303, the post World War II era is characterized by the commitment of the advanced industrial nations to removing barriers to free flow of goods, services, and capital between nations. As a result of the rounds of negotiations under the umbrella of GATT (General Agreement on Tariffs and Trade) and the WTO (World Trade Organization), there have been progressive reductions in trade barriers such as tariffs. Average tariff rates on manufactured products in countries such as France, Germany, Italy, United Kingdom, and United States declined from 15-25 percent in 1950 to 5-6 percent in 1990 and further down to 3.9 percent in 2000. Many countries have also been progressively removing restrictions to foreign direct investment (F DI). For example, between 1991 and 1999, more than one hundred countries made 1,035 changes in legislation governing F DI, and ninety-four percent of these changes involved liberalizing foreign investment regulations to make it easier for foreign companies to enter their markets (United Nations, World Investment Report 2000). Declines in trade and investment barriers facilitate globalization of markets where the consumer characteristics such as tastes and preferences have been converging. One reflection of the globalization of markets can naturally be expected on the diffusion rates :fgu FTLUUCIS In jig-on rates has I 1:111 11115511111} 9 3333,13} for 12:. an: otconsurm 1111261119931: 23529.1“. 119921 .1: 1:12.211: in 5:: c: 1.. 0153.1: and Itgttnn dc: cl. 1p. 7116;401:1331 [59:93 OI CI ElmC, 11257505 iii; y vI\‘ ”fl 1 ._ _.\. v - ' ' . 14“. DJ“ 11‘ I v... 4&3?) . I" “31' I “‘3“. R‘Ok pr, ‘5, - “51L ‘1 19‘ Mi“ “.131. £011" A“: ‘~<.L;'.‘-. J '4 I (‘1' Nu. ‘ -1,” '5. " x '6. Ki“ V‘ .5 . of new products in different markets. The convergence of market trends in terms of diffusion rates has been examined by Ganesh (1998) in the context of the European Union. This study empirically examined EU convergence/divergence trends and their implications for international marketing through time-series analysis of the diffusion patterns of consumer durable goods in individual EU countries. In fact, the objective of the Ganesh (1998) study was to reconcile the conflicting results of the Craig, Douglas, and Grein (1992) and Leeflang and van Raaij (1995) studies that examined the macro level trends in several developed countries. Craig, Douglas and Grein (1992) examined similarities in the macro level characteristics of eighteen developed countries (sixteen European countries, United States and Japan) from 1960 to 1988 to assess the effect of these characteristics on the evolution of these countries over time. The macro level variables included in the analysis were infant mortality, men’s life expectancy, cost of living, real per capita income, inhabitants per physician, population density, electrical production, rail passengers per kilometer, passenger motor vehicles, aviation passengers per kilometer, telephones in use, student population, book production, daily newspaper circulation, and radios in use. They concluded that, contrary to their initial hypotheses, countries are becoming more divergent. Leeflang and van Raaij’s (1995) study is a meta-analysis of several studies published in International Journal of Research in Marketing on changing consumer behavior (consumption patterns, credit usage, time allocation, information processing etc.) in 1.11; the macro It 1: merry-10} rr 1'1." 1111 :m ircnr 1.1M 1111:1211: :51: 111114111 stud 33:12.11: El' 11. :1 ITKCTIL 233:1W813imcd .:g;';t:; macro lc: :1 21122:: 01:: inc‘ ..1". y. 3.1 131111 mOIWTR, Vi?“ lira. ‘ . .5 1V0. the QUIET a ; . . .r‘b ' . 1“de p13\c1 31:; «.. ' ...xailOl'l pure» Europe. The macro level variables examined in these studies were economic environment (income, unemployment, inflation etc.), social/demographic environment (population, age and sex distribution, household size and composition, degree of urbanization etc.), and cultural environment (education, liberalization, health consciousness, ecological awareness, materialism etc.). Leeflang and van Raaij (1995) compared the main findings of the individual studies and concluded that, despite interesting and often substantial divergence, the EU nations are converging toward a similar macromarketing environment. Ganesh (1998) aimed at resolving the contradictory findings of these two previous studies regarding macro level convergence in the EU by examining variations in the diffusion parameters over time. He used sales data for ten products. Five of these products (car radios, lawn mowers, color televisions, deep freezers, dishwashers) were introduced before 1970; the other five products (VCRs, microwave ovens, home computers, cellular phones, and CD players) were introduced after 1970. 1970 was chosen as the midpoint of the unification process in Europe, which expanded from May 1950 to January 1993. Ganesh (1998) hypothesized that “convergence of macroenvironmental variables (economic, sociocultural, and demographic) among EU member nations will reflect in a relatively similar pattern of innovation diffusion among those countries” and “will accelerate innovation diffusion rates in those countries” (p.37). The examination of various comparisons of diffusion parameters for pre- and post-1970 data showed support for the first hypothesis while no support was found for faster diffusion rates in the EU 131:: due to unit EL Ci‘iml’lCS “'86 31? .121: 1.11.11.11.13 11:11:11.1.«1111 1 1. .. 5 be eeonom } 3:2: 2021.1 he lie 2.9121101: such .1: 1‘ 11:: 1? ar. earlier 5‘ 1522011111 E12119, 2:121 manet 1:111 ...‘ I ' A F’ ' 33* “up. !!~)‘.“-- ~ w .2: :;2;:at;on 01111 22: 2121211111215 intre Eitth-Sthaier ‘5 . 21'3"” "“ . panelling: ‘12, L L. .1116 I‘m 6W1? 1, it ' «NAIF? ‘ . ' .1.. ptttexxb 1r. ..‘L ‘36: countries due to unification. This unexpected result regarding the speed of diffusion in EU countries was attributed to the increased perceived risk associated with buying a new product immediately afier introduction due to high inflation and high unemployment rates in Europe in the earlyl9805, and recession in the early 19905. Ganesh (1998) noted that as the economy improved and consumer confidence increased, the diffusion rates in Europe would be faster as already evidenced by the penetration of more recent innovations such as cellular phones and CD players. This argument is in line with the results of an earlier study by Mahajan and Muller (1994). They simulated the effect of unification in Europe on new product diffusion and concluded that an integrated European market would result in faster penetration of new ideas, products and technologies. The unification of Europe is certainly a component of the globalization trend that has been continuing increasingly in the recent decades. Mahajan and Muller (1994) and Ganesh (1998) have provided valuable insight for understanding how globalization trend may impact the diffusion of new products in a multinational setting. However, these studies have not explicitly examined the impact of specific drivers of globalization on the diffusion processes in various countries. The phenomenon of globalization drivers has been identified by Yip (1992). There are four groups of globalization drivers that cover all the critical industry conditions affecting the potential for globalization, and the need for competing with a global strategy. These are market globalization drivers, cost globalization drivers, government globalization drivers, and competitive globalization drivers. Market globalization drivers include common customer needs, global customers, fiiifflr\iiitln~‘- ‘ $333.1“? “pm a»: 1n COUT' L,‘- 5" 5.312212%: 22;: 1922:1172: 1' #:2231225. 2m er: 22:75 redial to ' .1.... ..,,‘., ." ‘ h ..‘kl.“‘.t 'E’iiifidi ...- Ls, ' ‘ I will. tlYl'.l.'iCili\ 2?. ,3 ‘ ~ :72: m per: ' ~v ASS Ff. .arcclcmzf h: _ W11 relarixc r: «I ‘v t i I Hi (1‘05:er if}? i '- rron. comm “€1,111“ 1 tilli'ks I .“ ' L‘.‘i’{-m.. L ‘ :Ji\ O{.' - " uric "i; s- u. global channels, transferable marketing strategies, and existence of lead countries in terms of innovativeness, which require participation in these markets for exposure to latest innovations. Cost globalization drivers include global economies of scale and scope, steep experience curve effect, sourcing efficiencies, favorable logistics, differences in country costs including exchange rates, high product development costs. and fast changing technology. Government globalization drivers include favorable trade policies including trade incentives, compatible technical standards, common marketing regulations, govemment-owned competitors and customers, and host government concerns related to tax issues, weakening of national decision centers etc. Finally, competitive globalization drivers include high exports and imports, competitors from different continents, interdependence of countries due to sharing of business activities, and globalized competitors. Some of the recent trends in the globalization drivers include convergence in per capita income among industrialized nations, convergence of lifestyle and tastes, accelerating technological innovations, increasing cost of product development relative to market life, reduction of tariff and non-tariff barriers, shifi to open market economies from closed communist systems in eastern Europe and the former Soviet Union, continuing increase in the level of world trade (Yip 2002) In this dissertation the main objective is to analyze the diffusion process of new products by incorporating certain macro level globalization drivers, specifically, foreign direct investment inflows, trade volume, and income (GDP) into the model. These most crucial components of the globalization phenomenon are easily observed macroeconomic phenomena that managers can monitor. The declines in trade and investment barriers are mix»; 2516 main incl 1;" Further. lilC.‘ ..1 - i ‘ ' 21:2 01m: :01 4 1 ' . , V .. 52235. macaw I. i'riiflflt‘lcm‘} \s: ' - 1 .3 ' l' ..' :12 .‘x garmuson :2: women: rnr. grasrgtnce or as vs L In 1 b 2231';th 1n thrx \ petition rm erx \ .x-...r:r1101 imirariv ‘H as cramcrcristics r»..- . Rue. ll lm d0. [h Extra ...s. For amp}: '7“; ‘3 \“Ja ... lirkl‘tsL ,— dlf ‘. ““5313: . - )U\LC\\I hi sales for 1hr r ..- L. )0 ‘h 5* 2 . . “41' ‘1‘ {ch 1i 12M“) ‘ ..'"- \fgwj.‘ - - In the l h I ”ll! among the main factors that contribute the integration of economies across the globe (Yip 1992). Further, these globalization drivers are closely related to the economic and social well being of the countries. The openness of an economy due to market deregulation, and thus, increased international trade, and foreign direct investments lead to greater market efficiency with larger and more variety of products to consumers. It is expected that the globalization phenomenon reflected as increases in income levels and the foreign direct investment inflows as well as the overall trade volume have resulted in convergence of as well as faster diffusion rates (higher coefficients of external influence and coefficients of internal influence) for new products in developed and developing countries. The approach in this dissertation will allow quantifying the impact of specific globalization drivers on the diffusion parameters (coefficient of innovation and coefficient of imitation). This will also enable assessing whether the countries that exhibit similar characteristics in terms of globalization drivers are also similar in diffusion Patterns. If they do, this may have significant impact on the global marketing strategies of the firms. For example, if any two countries have similar diffusion patterns for the same product (class), and if the firm has experience with the product in either country, the manager may successfully predict the market size as well as the time and magnitude of the Peak sales for the product in the other market. If, on the other hand, countries with Similar characteristics of globalization drivers do not have similar diffusion patterns, the Contribution in the knowledge base in terms of which globalization driver impacts the filo: wmcrcfi I . ) 9:251:14 and grain Fm 1'22 seek to :51. 2:31: mpg: in the rfilraig 1995). lniti 355722653. lnzcrr. 2.122313ch {slur 2:41:01": reaardm ' 1: I 1:12:10 lac} maria Normand ri.~ aiman 1077 K' I g. \ u ' V 2:21-1:21 marl A ’ Li). ‘ . ‘%.. ‘~ \5\‘ - 1 1:} enter mulri r;- 's. flier inlllalh “Pa“ 1 ._..:arl:L$ in a \c ‘ h diffusion parameters by how much is still valuable, since this knowledge can be used in forecasting and strategy making by the managers and by the policy makers alike. Firms that seek to establish or continue to have presence in the global business arena need to engage in the dynamic evolutionary process of global strategy making (Douglas and Craig 1995). Initial market entry and expansion are among the most critical aspects of this process. International entry decisions include the identification and selection of potential markets along with the timing and order of entry (Ayal and Zif 1979). Decisions regarding identification and selection of markets have traditionally been based on macro level market characteristics such as market size, growth rate, attractiveness based on perceived risks (e.g., Cavusgil 1985). Timing of entry studies (e.g., Davidson and Harrigan 1977, Kalish, Mahajan, and Muller 1995), on the other hand, suggest that firrns can adopt either the sprinkler strategy, or the waterfall strategy when entering international markets. In the sprinkler strategy, firms adopt a simultaneous approach in Which they enter multiple foreign markets at the same time. Firms that adopt the waterfall Strategy enter initially one or more lead markets, and subsequently expand to other f(>l‘eign markets in a sequential manner. Managers of multinational firms are faced with critical questions regarding the choice of the entry strategy. Should the firm adopt the sprinkler strategy or the waterfall strategy? If the firm adopts the waterfall strategy, which foreign markets should it enter first? What is the global market potential for a given new product? What should the order of I -.I . o y- r‘tfmi‘; :0 01: blrh‘"T figxfimsl 15:22:20112: :1: 7:313:223: "W” V. ‘l- 1'51): szél L A ‘trL a 5‘. 1‘ g. ‘ ' HIT-37:63:: ll‘. 373': “9:181 .21 I '. . _ _ A. 3}! id: :“i\: L: 1 “' Sir, expansion to other foreign markets be for the realization of the market potential? These strategic issues facing the international managers can be addressed from the perspective of multinational diffusion of new products. More specifically, to capitalize effectively on the opportunities presented by globalization, managers need to fully comprehend the impact of globalization on market trends such as the diffusion of new products. Convergence in the diffusion rates of new products due to the impact of globalization drivers would not only mean the possibility of using a standardized global approach through global brands and marketing strategies but also have implications for the timing and order of entry in the global markets. Similarities in diffusion rates of new products among different countries may suggest that firms adopt a sprinkler approach (Kalish, Mahajan, and Muller 1995), whereas dissimilarities in diffusion rates would suggest a more prudent waterfall strategy. The development and application of the quantitative diffusion model in this dissertation Will provide guidance in new product planning and decision-making. Quantitative diffusion models specify mathematical relationships between quantifiable variables and inelude parameters that allow the model to be customized for a specific application. The fOCus will be on aggregate diffusion models representing the market penetration of new Products. It is aimed that the model results generate actionable information by a decision maker/policy maker. In this respect, the models may be used to describe the rate of difl:USion and to provide a better understanding of the drivers of adoption, to predict the filtllre penetration trajectory so that growth may be planned for, and to control the future penetration trajectory to provide inputs for strategy making decisions. The foundation of it mid is the hm: a 5.2253 in detail a 1:: :i‘sion prom: [)m and Sources 0' filters: on m 2:22.31 up b} rt‘x‘ Said 1980. Gatigm: :53; and Darrin 1‘1 rtifilicn 1907. X1: c' Tillmhiar. 82.1.51: 7 7?: gcrrbrmancc u l’ migrriL-‘lm but also ilrnuml‘er ol'thc p 311:: More prudLr-t'ts 27:: or la :1 of dm 1 ”£an Oldaml N named a: . -\. 1. 4w '1‘“. , i“ ll . . ‘ Jr: SchL‘S. I I ‘ a it"; “. v.35 LL’r‘ Q" . p'\h'\n\l\ C the model is the basic framework of the Bass (1969) model (Bass model will be described in Chapter 2 in detail). The concentration will be on the dynamics of the macro (market) level diffusion process of new products in a multinational setting. Data and Sources of Data In the literature on multinational diffusion of new products, certain issues regarding data are brought up by researchers repeatedly. These issues (see for example, Heeler and Hustad 1980, Gatignon, Eliashberg, and Robertson 1989, Takada and Jain 1991, Helsen, Jedidi, and Desarbo 1993, Putsis, Balasubramanian, Kaplan, and Sen 1997, Van de Bulte and Lilien 1997, Xie et al. 1997, Putsis et al. 1997, Kumar, Ganesh, and Echarnbadi 1 998, Talukdar, Sudhir, and Ainslie 2002) can be summarized as follows: 1) The performance of an estimation procedure is determined not only by its formulation and algorithm but also by its data sources. 2) The number of the products and countries in the multinational diffusion studies are 1iInited. More products and countries (particularly countries with drastic differences in cultures or level of development) should be included in future analyses. 3) The number of data points, that is, the length and/or frequency of time series data ShOuld be increased and observations from the tail end of the diffusion process should be inCIUded in the series. 4) More comprehensive set of covariates should be incorporated in the diffusion models. The database to be used for this dissertation research aims at addressing these issues. Data for forty-three countries on annual sales or possession of goods, exports, 10 gm foreign 11er :25: 78563911 are a some computer. cell “mm Daub» this-22g Data and lama. ALBIT‘dlEa. as. France. Ge: tamer-1:1. Ital}. hour. Portugal. l Skier Suitzerland 7:3: States 011m: 33 W noting tu1 l 0:1) [our Con consistentl} .1 'l ‘ he'll} «in: 0 Ml are the dc the dilllslttn l lillhe 317.1th imports, foreign direct investment inflows, and GDP are collected. The products included in this research are all consumer durables (Video camera, compact disc (CD) player, home computer, cellular phone). The main data sources are the Global Market Information Database, European Marketing Data and Statistics, and International Marketing Data and Statistics by Euromonitor. The list of countries is as follows: Argentina, Australia, Austria, Belgium, Brazil, Canada, Chile, China, Denmark, Egypt, Finland, France, Germany, Greece, Hong Kong (China), Hungary, India, Indonesia, Ireland, Israel, Italy, Japan, Malaysia, Mexico, the Netherlands, New Zealand, Norway, Pakistan, Portugal, Poland, Russia, Singapore, South A fiica, South Korea, Spain, Sweden, Switzerland, Taiwan, Thailand, Tunisia, Turkey, the United Kingdom, the United States of America (the countries written in italic are the developing countries). It is worth noting two issues here: 1. Only four consumer durables are included because these are the products that are consistently available across the forty-three countries in the study. 2. Twenty-one out of the forty-three countries are developing countries while the rest are the developed countries. This will enable us to examine the differences in the diffusion parameters between developed and developing countries in addition to the analysis of the diffusion of new products phenomenon in each country. 11 LtrimitiOI .llethodolo Tr: :ehrdolog} to he [shirts State and l) £13997]. Although I arsed in Chapter 3 a £331".th atta) in the .‘.- ' r... marshes ot the t1-.. 7:: ..rt‘tiel will contain .1 32': is mttiel. lhe p. .‘ii'ite'isttcs of the .-\K 315.133 of the ptmmt Welt-he products and 11:21:11 others little 1 iii'xtiut can mile u 3:2“ . “433% ll?“ ddlii (up '21. ‘. ““‘C‘I'r'l .W‘ .1 ldt d t . M19, “‘Lfllltf‘ it»; .1 1 * J13.“ m Mme-m1». Estimation Methodology The methodology to be used in this dissertation is the Augmented Kalman Filter with Continuous State and Discrete Observations (AKF(C-D)) methodology developed by Xie et al. (1997). Although the details of this methodology and model development are to be addressed in Chapter 2 and Chapter 3, respectively, there are certain issues that can be settled right away in the light of the advantages of the AKF(C-D) procedure and the characteristics of the data set. The model will contain a differential diffusion model, theoretical underpinnings of which is the Bass model. The parameters will be estimated in a time-varying manner due to the characteristics of the AKF(C-D) methodology. The information regarding prior distribution of the parameters will be incorporated in the estimation procedure. Given the scope of the products and countries, it is expected that in some cases strong priors will be available, in others little information about priors will be known. Employing a Bayesian approach that can make use of already available information and also can update the estimates as new data comes in is also crucial for this reason. The advantages of the AKF(C-D) methodology developed by Xie at al.(1997) include not only its superior predictive performance compared to other procedures but also other desirable characteristics. First, it is a general estimation approach that is not restricted by the model structure or by the nature of the unknown parameters. It can be applied directly to a differential diffusion model without requiring the diffusion model to be replaced by a disel‘ete analog or requiring that the diffusion model have an analytical solution. It can be 12 s: to estimate par :iztgfsl. Also. AK 2322?: from the e. o: it: prior distrihtt times sup-rite!) at? :t ofthout 3 pr amt} ahtd; pg: 2st ‘t situation ;. -,: - 3-.>1...t.1ors “he: Limitation of tl T:_lI[-Ylll.01.'l used to estimate parameters changing over time (both deterministic and stochastic changes). Also, AKF(C-D) is a Bayesian estimation procedure. It can provide better forecasts from the early stages of the diffirsion process by incorporating any information on the prior distributions of the parameters in the estimation process and updating the estimates adaptively. Third, AKF(C-D) is capable of estimating time-varying parameters with or without a prior knowledge of how the parameters change over time. Further, uncertainty about parameter estimates can be built into the estimation. Hence, the model is useful in situations where strong priors on the parameter estimates are present as well as in situations where little information is known about the parameter estimates. The formulation of the AKF(C-D) model is as follows: dc T33) = fx[x(t),¢(1)afl,’]+ wx d 7? =mlfi1x--~- ‘ uvl)mtlixlilmnl\m " “fiat-MI i. J’Ij‘iul'r ,. .. . will”: “:1ch , it" ‘h- h ‘ t _ it i ‘fl ‘ l ‘4. ‘ 1418 Social 3 for this dissertation research. In the following subsections a summary/review of these key studies will be provided. Gatignon, Eliashberg, and Robertson (1 989) Gatignon, Eliashberg, and Robertson (1989) is one of the early studies that provide an application of the diffusion paradigm in a global market setting. A statistically efficient methodology for analyzing and predicting multinational diffusion patterns is presented in this study. It allows for the estimation of the parameters of the diffusion model even in cases where these parameters cannot be estimated for separate product/country models. The complete set of diffusion parameters can be estimated due to the larger degrees of freedom resulting from pooling all observations for one product across all countries, and the constraints imposed concerning differences in diffusion parameters across countries. Consequently, this methodology enables the prediction of the diffusion rate of a product in a country before the product is introduced in that country, and the estimation of the diffilsion parameters for countries where data are not readily available, which occurs Comnmnly in international marketing research. Using their proposed methodology, Chaignon, Eliashberg, and Robertson (1989) modeled the heterogeneity among different colHitl‘ies in terms of their propensity to innovate and imitate within the diffusion process The Country characteristics they used to explain the diffusion pattern across countries were cosmopolitanism, mobility, and sex roles. OSmOpolitanism refers to the degree that the individuals of a society are oriented beyond their iImmediate social system (Gouldner 1957). The relationship between 22 figtxllmlSm and lt’nt tags 1995 i. organi/at : enter mileting tRo'rk grossed “countries it ith :rrtate and a smaller temptation is a le} Ut Pretend lifi. Brow :jesonal commitment Pissed “mobilitt \\ :1] 'r mes Lhe cumin-(m it} 1"he died of m, Walt unrelated. T act .0 model the hetero % net}. lite) related the matron iramler acri Eire ~ percentage oi \\ h. “fit n" t. “L lOl timeeonsur: ,_‘_ huh. ““ “m the “Mk U ‘n- F ‘0 E 5.. ‘ 4:3. ,- t “ammonium: ‘krl-J‘. ‘. ‘et df'iclttpcd b k. err; - tic ‘ “a BKd m("d”’ \i Q‘? i 3:1,]? ~ ‘ l ‘ '-t l‘lit’ it CHL‘C? H cosmopolitanism and tendency to innovate has been established in rural psychology (Rogers 1995), organizational behavior (Kimberly and Evanisko 1981), and consumer behavior/marketing (Robertson 1971). Gatignon, Eliashberg, and Robertson (1989) proposed “countries with a higher degree of cosmopolitanism show a greater propensity to innovate and a smaller propensity to imitate” (p.234). The second variable, mobility of the population, is a key underlying dimension of any spatial theory of diffusion (e. g., Hagerstrand 1953, Brown 1981), since the lack of mobility constitutes a barrier to interpersonal communication. Therefore, Gatignon, Eliashberg, and Robertson (1989) proposed “mobility will be positively associated with propensity to imitate, since it increases the opportunity for social interaction” (p.234). They did not have a proposition regarding the effect of mobility on propensity to innovate, as they appeared to be theoretically unrelated. The third variable Gatignon, Eliashberg, and Robertson (1989) used to model the heterogeneity across countries is related to the role of women in the society. They related the sex roles to the transmission of influence in terms of heterophily (information transfer across dissimilar individuals) within a social system, and proposed that “the percentage of women in the labor force is negatively related to the propensity to innoVate for time-consuming innovations and positively related to the propensity to imi tate when the work context provides a level of heterophilus influence” (p.235). Data regarding cosmopolitanism, mobility, and sex roles were cross-sectional. The rllode] developed by Gatignon, Eliashberg, and Robertson (1989) extended the single time series based model of Bass (1969) to multiple time series with a simultaneous CS ’ . . . . . tlrnatlon of the effects of the deterrmnants of the diffusron parameters across countries. 23 liads‘rete-time econ intro ofthc 3455 "W manor. in multiple er mm characteristit smite? emplming a g em simulttneouxl; ltitrgouers. pocket cal. times Austria Bel g5 leftists. Norma}. Pt 3‘55 and l‘lllU were use. 9 y “al.“ .13th 2 Lenerall) Prl‘ ttzmcatton factors it 1:32;?) leiel. and SC\ 0 fr. consumer durables i fits stud} uith cautior liltd Further. the Ct ...o- 0; dt\ Cltlpmcnl l an i Will); thcmx ~-Ql.‘ ‘ Minn oi coun .“ if: - «e: lo M ‘ sling {he [hCI It is a discrete-time econometric data-generating model that consists of the discrete version of the Bass model (with notation that allows for simultaneous diffusion of an innovation in multiple countries), and the equations that link the diffusion parameters and the country characteristics (cosmopolitanism, mobility, and sex roles). The model is estimated employing a generalized least squares (GLS) estimation procedure on all three equations simultaneously. Data for consumer durables (dishwashers, deep freezers, lawnmowers, pocket calculators, car radios, and color televisions) for fomteen European countries (Austria, Belgium, Denmark, Finland, France, West Germany, Italy, the Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, United Kingdom) between 1965 and 1980 were used. The results generally provided support for the importance of incorporating internal social communication factors in the diffusion model. Hence, a country’s cosmopolitanism, mobility level, and sex roles account for systematic pattern differences in the diffusion of new consumer durables in an international context. One should view the general results Of this study with caution though due to the limited number of products and countries analyzed. Further, the country characteristics used are not fully representative of the Variety of development levels and cultures in the world. Gatignon, Eliashberg, and Robertson (1989) themselves called for possible modification of predictor variables as we“ as inclusion of countries with drastic differences in level of development or in cultures to refine the theory of multinational diffusion of new products. 24 “may: ("/9911 round lain t 1991 i immunities in to from not represent .1 it} deteioped h} polite mm. The} did not e mi segment in the nmr 3:} finned tRogers 1% LL11. fittmimig gt‘ttg‘ irectl} atlect the dt :‘t‘u :2 Ill. 1987. and Rt. fit-“Pitta in countries 1 mm lSUch it It: for in imitation Ct E’flirjed b} l0“ L‘U.’ who; ~. - when) added i' (p. < or: «some ' ' t- has in e" l ff; . .1: is intrt‘tduced t' new “ al.-on coefficient 5.33 i w, ‘3'; ' “w filer than for fizzw',‘ \ l ' . .. :0llh1"' bdllL’ If" i . WJ new - imoutes of 9 ' Takada and Jain (I 991) Takada and Jain (1991) applied the Bass model to analyze the diffusion process of consumer durables in four Pacific Rim countries (United States, Japan, South Korea and Taiwan) that represent a variety in terms of socioeconomic and cultural characteristics. They developed hypotheses regarding country effect and time effect on the coefficient of imitation. They did not consider the coefficient of innovation since innovators are a very small segment in the market and their role in diffusing the innovation to other segments is very limited (Rogers 1995). The country effect refers to the uncontrollable factors (e. g., cultural, economic, geographic, legal, and political environments) that directly or indirectly affect the diffusion processes of new products in different countries. Based on Hall 1976, 1987) and Rogers (1995), Takada and Jain (1991) hypothesized that “the rate of adoption in countries characterized by high context culture and homophilous communication (such as Japan, South Korea, Taiwan, parenthesis added) is faster (higher value for the imitation coefficient, i.e., word-of-mouth effect) than that in countries Characterized by low context culture and heterophilous communication (such as US, Parenthesis added)” (p.50). The time eflect refers to the lead and lag relationship of diffusion processes in different countries. According to Takada and Jain (1991) “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 value for the country in which the product is introduced later than for the country in which the product is first introduced” (p.50). The fratneVvork for this argument is provided by Rogers (1995), which suggests that four per‘ceived attributes of innovations accelerate their rate of adoption. These attributes are 25 terslatitc’ 3d‘ ”my 12:513. and obsen: .11.. and lain t l W a" tale: potential I is” t its]; formulatit The islet isions. elec slant refrigerators. . 1133 lot 115.. laps: Wilt . ‘l “ ‘ . wk.) almtdl‘lf. £1le :11: unit sales to ur al.-1.189: found SUD erasure influence 1'11 INCA-ICE lll ll'lt‘ lit t.:e< '1. Jru'ltjl (Hid Dr tier. ledidi. and D» . smentation 5 ,. '.-‘ ' 4 a. “V - .h . 1C event 10 1 5:1. .. ‘W.‘I;"t“.-‘ . U ' ' mum‘lan div: - l l }i,‘ . "5311991 lST'IUL'l" the relative advantage of the product, compatibility with the needs of potential adopters, triability, and observability, and they are all positively related to time. Takada and Jain (1991) estimated the coefficient of innovation, coefficient of imitation, and market potential by applying nonlinear least squares (N LS) estimation to the Jain and Rao (1990) formulation of the Bass model. Data consisted of sales data for black and white televisions, electric washing machines, room air conditioners, passenger cars, electric refrigerators, calculators, vacuum cleaners, and radios during different time periods for US, Japan, South Korea and Taiwan. In cases where sales data were not readily available, data were derived by subtracting the export unit sales and adding the import unit sales to unit production, which gives apparent consumption data. Takada and Jain (1989) found support for both of their hypotheses. Their empirical results suggested the positive influence of cultural/communication system, and time on the diffusion of new products in the four Pacific Rim countries. Helsen, Jedidi, and DeSarbo (1993) Helsen, Jedidi, and DeSarbo (1993) study is an attempt to understand the merits of country segmentation schemes based on multinational diffusion parameters. They a‘nalyzed the extent to which countries belonging to the same (different) grouping reveal Similar (dissimilar) diffusion patterns. The research questions Helsen, Jedidi, and l)eSarbo (1993) sought answer for were stated as follows (p.62): 26 1.10 that extent the mcttrreswnd to St glitters? lilovi hell do \aria ’11.:5 xterm it her Flies sable are diti Litrent innox ations lithehxlolog} ll e acres on the ha\i~' 2:;dt‘rlog} for re gr sanitation of the n‘ ‘zlapfi . “W 30ml adt an 5E5 ‘1' - - . ““3 “35 Pooled i'j‘pu: ~ “ *mWSlnt‘ am fem and parame z. ‘ a on statzstical cr g. V 4:11“ i» “’05 dllCde C' ‘ 1. To what extent the country segments derived from traditional analyses of macro-level data correspond to segments derived from multinational, product-class specific diffusion parameters? 2. How well do variables that are typically used in macro-level country segmentation studies perform when used to profile diffusion-based country groups? 3. How stable are diffusion-based country segmentation schemes estimated across different innovations? The methodology Helsen, Jedidi, and DeSarbo (1993) used to segment the set of countries on the basis of observed diffusion patterns is a latent class structure methodology for regression models (for applicability to the Bass model). It is a modification of the methodology developed by DeSarbo et al. (1992). This technique Offered several advantages. First, short time series data were not a problem because the Sales data was pooled across countries. Second, the country segments were derived Without imposing any a priori segmentation scheme. Another benefit was that the country Seglnents and parameter estimates were determined simultaneously. Also, the technique relied on statistical criteria to evaluate appropriate number of country segments. Finally, the method allowed each country to belong to fractionally more than one grouping. Data consisted of annual sales data for color television sets, VCRs, and CD players for twelve countries (AUStria, Belgium, Denmark, France, Finland, Japan, the Netherlands, NorWay, Sweden, Switzerland, UK. and US). Twenty-three macro level variables gr Ouped into six constructs (mobility, health, trade, lifestyle, cosmopolitanism, and 27 astiianeousl “ch teed segmentation. errant compositior t: iiiision paramett felon-based seem moral countr} se riot cultural critert react introductione ‘Ci SK; 'cl‘iC‘Sc mau- tutti-ill} dtti'eren: “Pl'ikiucts Sl‘lalc .\ NC?» ‘ .4 .tsl‘utl} (all lltl’t‘t - l :- a": ', ’l Rama] and} gated K' ‘ umar, and 5 S’had.‘ v - -‘\“~-111't lean-“m: C 9. 121;. miscellaneous) were also used to segment the same countries in addition to diffusion- based segmentation. The results of the study showed little agreement between the segment composition based on macro level variables and the segment composition based on diffusion parameters. Further, segmentation comparisons showed no stability of diffusion-based segments across new product introductions. These results suggest that traditional country segmentation schemes based on macro level socioeconomic, political, and/or cultural criteria may provide little guidance as to the success of specific new product introductions. Also, cross-country diffusion process differences are not explained well using these macro level characteristics. Finally, the same country may exhibit substantially different diffusion patterns for different new product introductions even if these products share similar characteristics just like in the Helsen, Jedidi, and DeSarbo (1993) study (all three products were consumer entertainment electronics). Ganesh, Kumar, and Subramaniam (1 99 7) Ganesh, Kumar, and Subramaniam (1997) study investigates the existence of a systematic learning effect between pairs of lead and lag countries in the case of consumer durables. They proposed a theoretical framework that identifies the factors that influence the learning process, and empirically examine the relationship between these factors and the learning effect. The learning effect is defined as the phenomenon contributing to an accelerated diffusion of a product in the lag countries due to the success of the product in the lead country (Ganesh and Kumar 1996). It is critical to note that the learning effect examined in the Ganesh, Kumar, and Subramaniam (1997) study is not organizational learning, or changes to the marketing mix decisions based on prior experience in the lead 28 11 ct. RJlllCI. it re :2 niets 10113111 the n: l .A copiers in the lead n 33.1ng from p851 TC 131m". Kumar. and i it ..1'ning process 1 permit}. cultural si: ..1 standard. (It :1o:of'he lag count ifgtite index of the firestorm lettteen t Etifiede (1981i) idem Elm?) alOidfince ésirt ditlerencee j] 353...“ . .....ent) ot (jlll 5‘. ".511" ‘ ‘ . ....eso>nditte leat riff. market. Rather, it refers to the behavioral response of potential adopters in the lag markets toward the new product based on their observation of the experiences of the adopters in the lead market. Drawing from past research in international marketing, and multinational diffusion, Ganesh, Kumar, and Subramaniam (1997) identify six factors that potentially influence the learning process between a pair of lead and lag markets. These are geographical proximity, cultural similarity, economic similarity, time lag, type of innovation, and technical standard. Geographical proximity is measured as the distance between capital cities of the lag countries and the lead country. Cultural similarity is measured as a negative index of the sum of absolute differences in each of the four Hofstede (1980) dimensions between the corresponding lead and lag countries. The cultural dimensions Hofstede (1980) identifies are individualism, power distance, masculinity/femininity, and uncertainty avoidance. Economic similarity is measured as a negative index of the sum of absolute differences in the standardized values (due to differences in the unit of measurement) of GDP per capita, level of urbanization, and unemployment rate between the corresponding lead and lag countries. Time lag is measured as the difference in the years of introduction of the product between the lead and the lag countries. Type of innovation refers to whether the innovation is a continuous or a discontinuous one. And finally, technical standard refers to whether there were conflicting technologies when the product was first introduced in the marketplace (for example, VCRs) or not. Ganesh, Kumar, and Subramaniam (1997) hypothesized that the learning effect will be stronger if the lead and the lag countries are more similar geographically, culturally, and 29 orimicall}. the time tiisontinuoin one. a 1 data used in this s triers. and cellula Eastern countries t.-\ it}. lreiand. lhe Net}: thing-110ml. lhe : 1:15 oldie countries. snegorized as. the lea. mailman and W rigo capture the dirt} "a; Lie non-linear lea litntlQSoiThQ ‘1- it‘e’ttsin‘lte \ead ct 111.1113 similar to th ll; “itch allow .. includes the c {1‘1}?! ; “"15 L), i 1 f.-. .1, gr: ::.', 4- ‘ .% . . ““th 311 1a.. a. ‘s r , "mid / 7 economically, the time lag is longer, the innovation is a continuous innovation rather than a discontinuous one, and a technical standard for the product already exists. The data used in this study includes annual sales of VCRs, microwave ovens, home computers, and cellular phones for sixteen (eleven in the case of cellular phones) European countries (Austria, Belgium, Demnark, Finland, France, Germany, Greece, Italy, Ireland, The Netherlands, Norway, Portugal, Spain, Sweden, Switzerland and the United Kingdom). The starting point of data for each product is its first introduction year in each of the countries. The country in which a product was first introduced in the region is categorized as the lead country for that product category. Ganesh, Kumar, and Subramaniam (1997) estimated the classical Bass (1969) model in order to capture the diffusion process of each of the products in the individual countries using the non-linear least squares (N LS) procedure recommended by Srinivasan and Mason (1986). They also estimated a learning model that captures the influence of the adopters in the lead country on the potential adopters in the lag country. This learning model is similar to the independent product model developed by Peterson and Mahajan ( 1978), which allows for a one-way interaction between a pair of products. The learning model includes the coefficient of learning for the lag country, which is modeled as a function of the four (geographical proximity, cultural similarity, economic similarity, time lag) of the six factors described above. The other two factors (type of innovation, and technical standard) are tested with a dummy variable regression model where the 30 tines ofthe leamrn; re of innoi ation. at 11c empirical tindin, 1.: he learning.v etlec 11:37.1; .odel explai 5": to 97%). Further. £1st}. time lag. t}; torn} related to the 1 mg etlect rather th 11 snares of the cue 3313?, N“ “W are Pl. first. Kumar. and ..... .‘.‘ 1'13 values of the learning coefficient are regressed with the dummy variables representing the type of innovation, and the presence of technical standard. The empirical findings of the Ganesh, Kumar, and Subramaniam (1997) study suggest that the learning effect exists systematically across all product categories, and that the learning model explains most of the variations in the sales data (adjusted R2 ranging from 87% to 97%). Further, five of the six factors examined (cultural similarity, economic similarity, time lag, type of innovation, and the existence of a technical standard) are strongly related to the learning process. Although the main focus of the study is on the learning effect rather than the parameters of the basic Bass model, it is worth noting that the estimates of the coefficient of innovation, and coefficient of imitation as well as the market potential are plausible. Also the basic diffusion model explains 80% to 99% of the variance in annual sales. Ganesh, Kumar, and Subramaniam (1997) evaluated the forecasting performance of the learning model in comparison to the basic Bass model as well. Data from several countries were used as holdout samples. Based on the mean absolute deviations (MAD) and the mean squared errors (MSE) of the two models for all of the countries and each product category, they found that the forecasting performance of the learning model is better than that of the basic Bass model in 12 of the 17 cases. This result underscores the robustness of the learning model and the role of the learning coefficient in explaining the diffilsion process in the lag markets. 31 P315 8.11.1101th Paisetalt 1997 or tour} al‘t‘ee effect of nitrite t if: no product cc:::::um urth s 1.. 1;- . ‘ terreuate 1017‘. 1:; consisted o 3" @9215erst'1034 312.115 Austria I \citcrlmds. Spai option cross-cc Strain as a on 1:5”: as a proxy 1 irnce: the para limitation in th it»: i 4 he Prtldttc Putsis, Balasubramanian, Kaplan, and Sen (1 99 7) Putsis et al.(1997) study addresses the extent to which prior adoption of a new product in one country affects adoption in other countries. They investigated the importance and effect of mixing (the pattern of communication within and across countries) in the context of a new product diffusion model. In their model, mixing is viewed as occurring across a continuum with segregation (no mixing) at one end, and random mixing at the other. Intermediate forms of mixing that lie along this continuum are called Bernoulli mixing. Data consisted of annual sales data for VCRs (1977-1990), microwave ovens (1975-90), CD players (1984-1993), and home computers (1981-1991) for ten European Community nations (Austria, Belgium, Denmark, France, Italy, Germany, Great Britain, the Netherlands, Spain, and Sweden). A diffusion model that incorporates cross-country prior adoption, cross-country mixing patterns, and individual country covariates (television set ownership as a proxy for non-word-of-mouth effects, and gross domestic product per capita as a proxy for information-seeking behavior and susceptibility to word-of-mouth influence; the parameters of these covariates correspond the coefficients of innovation and imitation in the Bass model) was estimated simultaneously across all countries for each of the products. The results provide preliminary evidence of the importance of considering cross-country interactions when estimating diffusion models. The mean absolute percentage error (MAPE) declined when the random mixing assumption was relaxed and Bernoulli mixing was allowed for all of the products, and increased dramatically as segregation which is 32 he most common 17 literature leg- Gati retouchoi lhe pa nicornpact disc p' Luring parameter ti eons} emanates I it 'talue of their p; 9 tepnman contrii rheological. Th matched on exte landing to Deitir 't: - ' 61m. that requrre 7'} .‘ .1.; * ‘s. iear accept union or the Po 3}»? . : «5111.011 Ofm.\ Wired D .l g I , c t 7 11m, Yr- .rs in 184 CW“ 1‘... \C‘fh $58. and 15 fr tire ., . magmm 11‘ \. \f'jki" v the most common form of cross-country diffusion pattern previously addressed in the literature (e.g., Gatignon, Eliashberg, and Robinson 1989, Takada and Jain 1991) was approached. The patterns MAPE followed for videocassette recorders, microwave ovens and compact disc players was similar with mixing parameters around 0.5 while the mixing parameter for home computers was around 0.7. Further, results showed that country covariates matter in determining the diffusion patterns in different countries, and the value of their parameters differ across countries and products. Dekimpe, Parker, and Sarvary (1998) The primary contribution of the Dekimpe, Parker, and Sarvary (1998) study is methodological. Their estimation procedure is based on the premise that samples should be matched on external criteria before valid comparisons can be made among them. According to Dekimpe, Parker, and Sarvary (1998) the four components of a diffusion pattern that require matching are the social system size, the long-run penetration ceiling, the first year acceptance level (the intercept of the penetration curve), and the speed of diffusion or the growth rate between the intercept and the ceiling. The secondary contribution of this study is with respect to the scope of cross-cultural variation considered. Dekimpe, Parker, and Sarvary (1998) used data on the adoption of cellular phones in 184 countries (55 from Africa, 37 from Asia, 32 from Europe, 15 from the Americas, and 15 from other regions, mostly island countries) between 1979 and 1992. Using the matching criteria and incorporating several macro level covariates in the model, Dekimpe, Parker, and Sarvary (1998) estimated their model using a staged 33 Smart procedure \\ the social system 51/ the intercept term \sl astral estimation of e; user the ceiling. and :1." he estimation proct omit“ section. The c 911533011 gromh rate. It anther of com pet 11's and number of . 7' . '1 1“ ‘ .1 than 91 the “chill criminated h\ crud; 'fii" it“ a I .l‘...l&l penetrattor ...1: is negate e\.\ to: ~ ~ \Q\ {121112 3! 9mm T318 0ft 15‘.“ “ll," -- ' Mt'firr- _ dre‘d. .ng w- J L, t 4.7," F -._/ tea/11.3 pf“, estimation procedure with the following sequence: 1) external estimation and validation of the social system sizes, and long run adoption ceilings across countries; 2) calculation of the intercept term which is exogenous to the subsequent growth process; and 3) internal estimation of each countries’ grth parameter which is endogenous to the social system, the ceiling, and the time-origin (intercept) concept. More details about the model and the estimation procedure used in this study will be given in the methodological overview section. The covariates included in the model are such factbrs as average annual population growth rate, number of major population centers, GNP per capita, crude death rate, number of competitors, political environment (communism or not), number of ethnic groups, and number of countries that adopted cellular phones previously. The results of the Dekimpe, Parker, and Sarvary (1998) study suggest that poverty (approximated by crude death rates), and ethnic heterogeneity are negatively related to both initial penetration level and the speed of diffusion. The number of major population centers is negatively related to the initial penetration level (because it is difficult to provide coverage service for cellular phones in all areas at once), and positively related to the growth rate of diffusion. Factors that are positively related to the initial penetration level are population growth rate and number of competitors. All other factors such as GNP per capita, state control of the economy, and number of previous countries that adopted cellular phone service previously do not have any impact on either the penetration levels or the growth rates. 34 he .417, (um \ litI" ,hfianes Liaiigton. ll .14 are ‘ DeSa'hot throw the O\ C tilb‘L’ia. B A, eiei Portal : .Spair ahe' toleot‘ e 31.": _ meters b --.|. - “R. '1 \‘u " 1.4:... ~ 1 ‘~‘\.\. 'tc ‘l ere“ ..tat‘? on V Kumar, Ganesh, and Echambadi (I 998) Kumar, Ganesh, and Echambadi (1998) study replicated and extended the findings of Gatignon, Eliashberg, and Robertson (1989), Takada and Jain (1991), and Helsen, Jedidi, and DeSarbo (1993) studies using a common set of product categories (VCRs, microwave ovens, cellular phones, home computers, and CD players) and countries (Austria, Belgium, Denmark, Finland, France, Germany, Italy, the Netherlands, Norway. Portugal, Spain, Sweden, Switzerland, and UK). Its objectives were to empirically verify (a) the role of country-specific effects in explaining the differences in diffusion parameters, (b) the presence of a lead-lag effect, (c) the use of cultural variables to explain systematically the diffusion patterns across countries, and (d) the merit of country segmentation schemes based on diffusion parameters. The general result that came out of the Kumar, Ganesh, and Echambadi (1998) study is that diffusion parameters across countries are influenced by certain country-specific characteristics related to social communication (cosmopolitanism, mobility, sex roles) and time lag effects. This is in accordance with Gatignon, Eliashberg, and Robertson (1989) and Takada and Jain (1991). Another result of the Kumar, Ganesh, and Echambadi (1998) study is that although the relative influences of country-specific and time lag variables differ across product categories, different products, on the average, have similar diffusion parameters, and hence, the average of the coefficient of these variables across existing products can be used to generate forecasts for a new product. 35 (my, Ganes lehdi. and l): utilities. lltc gametes for m ofthe ti: gnagraphieil p real: prox ides This et al. t 1‘ lk \rn' t Lee—tr. Ganesl mined Nth itch. are cm» to mt‘dtl R; J“ 0n the 3. 1': “‘~ Pen Kumar, Ganesh, and Echambadi (1998) did not find corresponding results to Helsen, Jedidi, and DeSarbo’s (1993) country segmentation composition based on macro level variables. The country segments they formed based on the similarities of diffusion model parameters for each product category suggested that countries seem to group together in terms of the time of introduction. Second, given similar time periods of introduction, geographical proximity appeared to influence the formation of segments (note that this result provides support for the argument/result regarding the importance of mixing in Putsis et al. (1997)). Further, cultural and economic similarity also seemed to influence the segmentation scheme. Kumar, Ganesh, and Echambadi (1998) did also an extension study where the model contained both country-specific characteristics (cosmopolitanism, mobility, sex roles), which are cross-sectional data as well as the time lag variable. This cross-sectional time series model was estimated using GLS, and the results supported the influence of these variables on the cross-country diffusion parameters. Further, the comparison of the forecasting performance of this extended model to the forecasting performance of Gatignon, Eliashberg, and Robertson (1989) model, which is the cross-sectional model, suggested that the cross-sectional time series model has superior forecasting accuracy. T alukdar, Sudhir, and Ainslie (2002) Talukdar, Sudhir, and Ainslie (2002) is the most recent study that recognizes the increasing importance of global marketing, and multinational diffusion of new products. They contribute to the multinational diffusion research by addressing three important 36 3 Whine lite 7? ; t‘ulOLlS stud}: deteioped cou nicentiates a 313:. the} UM rjcmuon an ' I 333M 3?: me ‘ 5 “$13.13 581 T‘ gaps in the literature. The first one is related to the scope of the countries. Unlike previous studies, Talukdar, Sudhir, and Ainslie (2002) included developing as well as developed countries in their analysis. Second, they investigated the impact of a larger set of covariates on the diffusion process than any other single multinational diffusion study. Third, they used Hierarchical Bayesian estimation procedure that allows combining information about past diffusion patterns across products and countries with the aim of improving the predictive power of the Bass (1969) model. The data set Talukdar, Sudhir, and Ainslie (2002) used consists of 6 consumer durable products (camcorders, cellular phones, CD players, fax machines, microwave ovens, and VCRs) for 31 countries from Europe, Asia, North America, and South America. The base model of the study is the discrete time version of the Bass (1969) diffusion model. They used per capita sales in order to correct for the influence of scale in countries with respect to varying populations, and to account for population growth over the time period of analysis. They modified the basic Bass model by incorporating the error term on the demand in a multiplicative fashion, as done in Van den Bulte and Lilien (1997), in order to reduce the effects of heteroscedasticity, and to prevent the possibility of support for negative demand. They also allowed for autocorrelated errors in order to account for any serial correlation in errors between successive periods. Afier estimating the modified (nonlinear) Bass model using non-linear least squares estimation (Srinivasan and Mason 1986), Talukdar, Sudhir, and Ainslie (2002) transformed the parameters of the model (market penetration potential, coefficient of 37 extend lllllUt - - '3" 3733.033th ll mfomed \‘ lecnrttpisitior i'JCilll’c‘ of th h: is cornmu run all prod :e’ncglar prtx 33.1333 Cit”: , Q . .ejigtm ’J're L I . 4 c ' ‘ . 3'4. «‘2 41- ' . “‘13: \ i I n ‘ -. \Qrfrz h ‘ 5' 5 ;.'?;~«'. . "i'ut‘t’a‘ . h. .333“ t 1th like} -. 3. "- 1“ “it'lii‘ “31'! «n ' ‘ L‘Jr“\1 . :x. '~£C\ ~ -. p. ., “dear: external influence, and coefficient of internal influence) using an exponential, non-linear transformation to restrict the actual parameters to be positive, while allowing their transformed values to lie along the full real line to allow estimation of the variance decomposition using the normal distribution. Next, they modeled the heterogeneity structure of the transformed parameters by separating each parameter into a component that is common across all countries for a particular product, a component that is common across all products for a particular country, and a component that is unique to the particular product-country combination. It is worth noting that Talukdar, Sudhir, and Ainslie (2002) assumed the parameters of the diffusion model are time-invariant. Therefore, the country specific covariates are included in the model as time-invariant. The product specific component is modeled as a random effects model (i.e., without any hierarchical regressors). More details will be given about the model Talukdar, Sudhir, and Ainslie (2002) used in the methodological overview section. The heterogeneity structure of the diffusion parameters across products and countries are analyzed using Hierarchical Bayes estimation methodology. In doing so, Talukdar, Sudhir, and Ainslie (2002) draw strongly on two previous diffusion studies, Gatignon, Eliashberg, and Robertson (1989), and Lenk and Rao (1990), although there are considerable differences that are worth noting. Unlike the Gatignon, Eliashberg, and Robertson (1989) study, Talukdar, Sudhir, and Ainslie (2002) focused on explaining not only the coefficients of external and internal influence but also the penetration potential. They also considered heterogeneity of the diffusion parameters across products and countries, whereas Gatignon, Eliashberg, and Robertson (1989) pooled the data only 38 ins countries. a or.) linear models heel. With respe ifi'lli present a n‘. aliiermliieal 8.2} ainpier error strt 5th products and .mcar. Sudhir. 2 4.,» J - .. nonel. These 5 tennis: 1) The Ira} lpiIICllLbln; firlzi ' _..,...-..llOll Ht. ml 3:? ‘All . ' with! pnoncx .rprenes for fax 1 it? h‘-. _ until 53 Ol CXprtj 13mm? ' -~ "“4 mllUka e53; ‘ Ul {El-771$ 0f; 1' {“55 10 ”1135 align tr lltillcrah IE» - tin" across countries, and estimate the model one product at a time. Lastly, Gatignon, Eliashberg, and Robertson (1989) used an estimation methodology that can be applied to only linear models; Talukdar, Sudhir, and Ainslie (2002) estimated a nonlinear diffusion model. With respect to the Lenk and Rao (1990) study, Talukdar, Sudhir, and Ainslie (2002) present a more through analysis as well. Although Lenk and Rao (1990) too used a Hierarchical Bayesian approach, their analysis contained no hierarchical regressors and a simpler error structure, and they pooled the data only across products rather than across both products and countries as Talukdar, Sudhir, and Ainslie (2002) did. Talukdar, Sudhir, and Ainslie (2002) included an extensive collection of covariates in their model. These covariates and their expected impacts on the diffusion parameters are as follows: 1) The factors that affect the penetration potential (m) are consumers’ ability to pay (purchasing power parity (+), Gini index (-), proportion of dependents in the population (-)), willingness to pay (percentage of customers waitlisted for terrestrial lines for cellular phones (+), TV penetration level for VCRs (+), per capita installed base of telephones for fax machines (+)), and access to the product (urbanization (+), and percentage of exports and imports in GDP (+)). 2) The factors that affect the coefficient of external influence (p) are consumers’ access to product related information (external contact in terms of number of minutes of incoming and outgoing international phone calls (+), access to mass media in terms of TV and newspaper penetration levels (+), and years of lag in product introduction (+)), inclination and ability to process non-word-of-mouth information (illiteracy level in a country (-)). 3) The factors that affect the coefficient of internal influence (q) are population homogeneity (Gini index (-), number of distinct 39 me groups 1 men; Monte T‘ 4;... 0: it" blundtlon Ci 735 Ct‘lifidli - w J ; 1.de. 0‘ Ln: mud. . I. artisan \t; the mount t frieze in he mpg. Yr“. | ' W m. _ ' rm... 5‘“) L11” he7:13: to 13* “milon ah . s 'L. 5“ “ii-'3 if e tir ’t‘wnfl ; L-kl.A; APR?“ JCF .r.f; A \. ”I“ ”at 3m: 9} ir 5.; ‘5’»? (1}:sz _ . \ “‘\‘.C 3.5133. ’Il ‘ 2‘. \ h r_ MIC u‘“'"‘.1 «I. .1.?2. ethnic groups (-), and percentage of women in the labor force (+)), and persuasiveness of existing adopters (years of lag in product introduction (+)). The estimation results for the diffusion model support almost all of the expected impacts of the covariates on the diffusion parameters. One surprising result was the negative impact of time lag on the coefficient of external influence. The hypothesis regarding this relationship was that as the number of years of lag increased, there would be an increase in the amount of information about the product externally which in turn would lead to an increase in the coefficient of external influence. Talukdar, Sudhir, and Ainslie (2002) explain this unexpected result by reversing the direction of causality given that timing of entry in a market is in control of the managers unlike other factor such as illiteracy rate. According to this explanation, the introductory lag might really be a proxy for firm information about the relative take-off times for different products in different countries, and there are unobservable characteristics of a country that reduce the coefficient of external influence and slow the take-off time which in turn delay the entry to that country. Other interesting, although not unexpected, results of the Talukdar, Sudhir, and Ainslie (2002) study include that developing countries have lower coefficients of innovation and higher coefficient of imitation than developed countries. With respect to the variance decomposition of the diffusion parameters, Talukdar, Sudhir, and Ainslie (2002) found that while country effects (past experiences of other products in a country) are relatively more useful to explain the penetration level or cumulative sales, product effects (past 40 anenences ii I ' ‘h to ermin the l :l): reel Old horler to ex; tamed it 1.; minim prot econ-l one mt 'W' 'J. The f: e’“ v.4 L" ‘ I \ - .LLl: ..ier ”fir I “r-O‘mi‘nl} t». \ “\‘ll‘i 'n We £21.... experiences in other countries where the product was introduced earlier) are more useful to explain the coefficient of external influence and the coefficient of internal influence, or the speed of diffusion. In order to evaluate the performance of their model, Talukdar, Sudhir, and Ainslie (2002) compared it against two simpler benchmarks. The first one modeled all countries for a particular product (in line with Gatignon, Eliashberg, and Robertson (1989)), and the second one modeled all products for a particular country (in line with Lenk and Rao (1990)). The percentage improvements in the forecasts (measured by improvements in mean square errors) when the full model (the model that accounts for both country and product effects) is used ranged from 17% to 36%. The comparisons were based on improvements in one period and two period ahead forecasts of the three types of models (full model, country model, and product model). An overall summary of the multinational diffusion studies reviewed in this section is given in Table 1.1. With respect to the individual results one needs to be aware of the idiosyncrasies of each study in terms of the data characteristics, model specifications and the methodology used to analyze the diffusion process. Still, collectively, these studies provide the knowledge base in the multinational diffusion literature. They point to the importance of several macro environmental variables along with the interdependence of the diffusion process across countries. Further each and every one of them emphasize the crucial impact of the globalization phenomenon on the diffusion of new products. 41 r ’1 ‘ Tbg‘lpprOuL over :zrrie. a: ninnltentent 135'» {WONG \ ‘3';- «. ~.' mankdl‘ lOS‘I 5 I Wail-.1. ' '1'?“"‘ in U]: '1 r— ‘i . ' - “- ~33 ('1? L .‘ier -j,_. a," 0 v ’ ’— l ‘. ',r Fir“ .‘ L s The approach we adopt in this dissertation follows directly from this perspective. The main argument we make is that the drivers of globalization have a direct impact on the diffusion patterns of new products across countries. We expect that the speed of diffirsion (reflected in the coefficient of innovation and the coefficient of imitation) has increased over time, and that there are identifiable diffusion patterns across countries based on their involvement in globalization at the macro level as well as their level of development (developed versus developing). In order to investigate these issues we will use the Augmented Kalman Filter with Continuous State and Discrete Observations (AKF (C-D)) methodology developed by Xie et al. (1997). This is a methodology that has not been applied in the multinational diffusion literature before, and has very appealing advantages over the previously used methodologies in this area. These advantages are mainly due to the Bayesian nature of the AKF (C-D) procedure, which also yields time-variant parameter estimates. In the next section, we will present a methodological overview of the diffusion of new products research. The AKF (C-D) methodology will be explained in more detail along with the other methodologies commonly used in previous studies. The specific model and the algorithm to be used in this dissertation, however, will be explained in Chapter 3. Methodological Overview of the Diffusion of New Products Literature Accurate prediction of the diffusion of new products requires the specification of the diffusion model and the estimation of the model parameters. A variety of methods for estimating diffusion models have been proposed (for extensive reviews of the estimation 42 resiziques see 1 matador: proc .» VI ‘ . ‘2 1 5";6'61. Lad; .' \.\ n -——.——h—-— 3651 O u C K .; J , t ' ' }:i;£6'c&.l'.l.1[lklt \\5.V ‘ C C ' as 1 Veal t“. lel {Lil L minim like gs; ‘4‘.dl\.‘_-ln m‘ C fi)..~‘~’ techniques see Mahajan, Muller, and Bass 1990, Putsis and Srinivasan 2000). These estimation procedures can be grouped under three subheadings. S_ir;gle-egugtion time-invagant estimation procedures Most of the work on estimation of diffusion models has focused on the estimation of single-equation time-invariant models. Time-invariant estimation procedures include the conventional estimation methods such as ordinary least squares (OLS) (Bass 1969), maximum likelihood estimation (MLE) (Schmittlein and Mahajan 1982), and nonlinear least squares (N LS) (Srinivasan and Mason 1986). These estimation procedures suffer some common limitations. First, to obtain stable and robust parameter estimates, time- invariant procedures often require data to include peak sales (Mahajan, Muller and Bass 1990). Time-invariant procedures are not helpful in forecasting a new product diffusion process because by the time sufficient data have been collected, it is too late to use the estimates for forecasting or planning marketing strategies. Second, though diffusion models often are expressed by a continuous differential equation, the time—invariant procedures can be applied only to the discrete form of the diffusion model or to the solution of the diffusion model. It becomes necessary to estimate a continuous model using data across discrete time intervals because sales, and cumulative sales are not observed continuously. The discrete form used to estimate diffusion models ofien results in biased and high variance estimates. Requiring a diffusion model to be analytically solvable limits the applicability of the estimation procedures (Putsis and Srinivasan 2000). 43 are t l W ‘-~r.J'J :3“ 12315-5 ex: used a 'l ‘r “We 5; l5 Bass (1969) originally suggested the OLS estimation of the parameters m, p, and q using the discrete analog of Equation (2.1). The discrete analog of Equation (2.1) can be expressed as: _ 2 st—a+bxt_1+cxt_1+et (2.11) Where s, is sales over the t‘h time interval, and x,-1 is cumulative sales at the end of period —b—(b2 -4ac)y2 a t-1.Further,m= ,p=——,andq=—mc. 2c m Using OLS to estimate the Bass (1969) model attracts many inherent problems (e.g., Schrnittlein and Mahajan 1982, Srinivasan and Mason 1986, Putsis 1996, Van de Bulte and Lilien 1997). First, parameter estimates can be extremely unstable when there are only a few data points. Second, the standard errors of the parameter estimates of p, q, and m are not readily available since they are nonlinear firnctions of a, b, and c. Third, there is a time interval bias due to estimating Equation (2.1) using discrete time-series data. This bias results from attempting to estimate a continuous time model using 3, = x(t) - x(t — 1) on the left-hand side of Equation (2.1), where, to be consistent with the right-hand side, it should represent the derivative of x at H. As a result, the application of OLS will overestimate sales when cumulative sales are growing quickly (such as before peak sales), and underestimate when cumulative sales are growing slowly (such as afier peak sales). In theory, the shorter the data interval used the smaller the time interval bias under OLS (Putsis 1996). 44 XIII and XLS 5mm; sing? subject to the ti: SLED-53d ClTOfS hllE appma; or. . fW‘n 9'”. L-.C:.3Lli.£ um anion «2.3ic iztfn°ql. Al: 755:. then the en SztittIein and i. which detcm TL)?" 5 , MLE and NLS are the other two estimation approaches that have been used frequently in estimating single-equation time-invariant diffusion models. These two approaches are not subject to the time interval bias. Further, they have the added benefit of providing standard errors for the estimated parameters. Schmittlein and Mahajan (1982) proposed a MLE approach that provided significant improvement over OLS by appropriately aggregating the continuous time model over the time intervals in the data. Note that Equation (2.3) can be rewritten as F(t) = (1- e- 7t)/(1+ lie— 7' ) where it = q/p and y = (p + q). Also note that in a sample of size M, if the eventual probability of adopting is p. , then the expected number of eventual adopters is M and E[x(t)] = uM F (t). Schmittlein and Mahajan (1982) first generate maximum likelihood estimates of ch , and u , which determine the estimates of p, q, and m. Formulae for approximate standard errors of the estimates for p, q, and m are also derived. MLE is consistent, asymptotically normal, and asymptotically efficient. In the NLS approach proposed by Srinivasan and Mason (1986), an expression for the right-hand side of Equation (2.1) is derived such that the right-hand side equals the same difference on the left-hand side. That is, there is a continuous form expression for the difference x(t) — x(t — 1) and the parameters p, q, and m can be estimated directly via this difference. Specifically, p, q, and m can be estimated via NLS using s, = m[F(t) — F(t -1)]+ u, 01' 45 there u, is an a. The formulation 2‘ t‘qufiion (ll l _l SI-[m-rtr -1.] a: 3 ’lf‘tll- F' - ll‘Ftr :Em up llll per m’ub grim" lur- 315$th Clt‘n Em39150111 data. I— — 1_e—(p+q)t 1_e-(p+q)(t-1) s, =mt — +u, (2.12) 1+£1_e-(p+q)t 1,18-(p+q)(1-1) p p — — where u, is an additive error term. The formulation suggested by Jain and Rao (1990) is slightly different. Their alternative to Equation (2.11) is _ _ _ [P(t)-F(t-l)l s,_[m x(t 1){ [1—F(t-l)] +5, (213) where [[130 ;:(t 131”] is the probability that an individual who has not purchased the product up till period t-l will purchase in the (1” time interval. NLS estimators obtained by both Srinivasan and Mason (1986) and Jain and Rao (1990) are consistent but not unbiased (Van den Bulte and Lilien 1997). Consequently, NLS parameter estimates obtained from data sets with few and noisy observations should be viewed with caution. Although both the MLE and the NLS procedures eliminate the time interval bias, there are some important differences between the two estimation procedures. For example, since the MLE approach of Schmittlein and Mahajan (1982) focuses on the sampling errors and ignores all other sources of errors (such as omitted variables), it seriously underestimates the standard errors of the estimated parameters (Srinivasan and Mason 1986). On the other hand, the error term under NLS includes both sampling errors, and errors due to omitted variables and functional form misspecification. Hence, NLS may 46 . r ‘ SCI-£13353“ -01 jetttth'd M3 13;, uere bit“ 35¢ [hf hill? gjlieal‘le to d me Further. l :a'zrneters the got sense ot‘t $2335. Farie}. r for sin. uitane :fision resea: racemes are 5 Extreme. have an advantage in instances where nonsampling errors are substantial (Putsis and Srinivasan 2000). Still, it is useful to note Srinivasan and Mason (1986) report that the downward biases in the MLE standard error estimates are negligible when the adoption data were based on sample sizes of 200 or fewer customers. Both the MLE and the NLS approaches suffer similar limitations. They are only directly applicable to diffusion models for which F (t) can be expressed as an explicit function of time. Further, both procedures require specifying starting values for each of the parameters that are used in the estimation algorithm. Although we have a somewhat good sense of what the parameters in applications of the Bass model are likely to be (e. g., Sultan, Farley, and Lehmann 1990), this may not be true for more complex specifications or for simultaneous equations models that are particularly suitable for multinational diffusion research. Since the estimated parameters in both the MLE and the NLS procedures are sensitive to the starting values, this issue may become of major importance. Overall, the main advantages of MLE and NLS techniques are that they overcome the time interval bias and that they provide direct estimated of the diffusion parameters and their standard errors. In the context of a single equation Bass model applied to consumer durables, the noncumulative form of NLS (Equation (2.11)) will do well in most settings and may be preferred to MLE (Putsis and Srinivasan 2000), especially since there might be serious downward bias in standard error estimates when MLE is used (Srinivasan and Mason 1986). 47 3': rtrn‘. kl lsdbhil ‘i'.:§\ ‘.~. LU ’g Single-equation time-variant estimation procedures Conceptually there are numerous reasons why it is important to address parameter variation in diffusion models over time. First, a diffusion model’s parameters may vary over time due to competitive activity, changes in the advertising quality, or changes in the product itself (Eliashberg and Chatterjee 1986) as well as due to perfection in technology and associated increases in product quality, increased benefits that result from an expanding installed base, and changing consumer expectations (Horsky 1990). Second, different segments of the market are likely to adopt at different points in time. This can cause variation in the estimated diffusion parameters over time. For example, early adopters may have different coefficients of external and internal influence than late adopters, high-income households are likely to have different parameter values than low- income households (e.g., Horsky 1990) and so on. Third, specification and measurement errors can result in parameters that vary over time. For example, “contamination” of first purchase data with repeat purchases can lead to the appearance of varying parameters (Putsis 1998). Finally, aggregation, use of proxy variables, nonlinearity and omitted variables can also result in varying estimators (Sarris l973, Judge et al. 1985). Understanding parameter variation in diffusion models over time is important for various reasons. First, the form of the variation can provide insight into the nature of the diffusion process. Second, it provides information on the appropriate estimation technique or theoretical model to use to estimate the relevant parameters that vary over time. Finally, it allows to relatively easily improve the within sample fit and forecasting ability of 48 hfhuion mt-‘JC e hmntiall) l0 TUBE-Vii!) is g " itS}stematic t: hzhese mixiei: rzr: period to 1 techastic \ am him. The es! ' -it 3:234:10“ theor fir‘a ~*~ 31¢ Spec if . ‘ @8st Dex; :65 a l. 3’32; ~. diffusion models (Putsis 1998). In fact, time-varying parameter specifications can substantially lower forecast errors (Putsis 1998, Xie et al. 1997) Time-varying parameter models can be grouped into three groups: 1) Systematic (nonstochastic) variation models: In these models, the form of transition is prespecified by the researcher, and transition from period to period is not allowed to vary stochastically. Although more restrictive than stochastic variation models, these models offer added flexibility over time-invariant NLS method. The estimated parameters can change over the time span covered by the data set and allow theory to guide the transition. Further, the rigidity in the Bass model is reduced since the specification in systematic variation models allow for nonsymmetric diffusion curves and flexible inflection points. These models are not free of limitations however. The requirement to specify the form of transition a priori is often difficult, especially in the case of weak priors. The application of these models is problematic also when F (t) is not known or when solution to the differential equation does not exist. In these cases stochastic or Bayesian specifications may be preferred. Stochastic parameter models allow the parameters in a diffusion model to vary stochastically over time according to either a stationary or a nonstationary process. Although stochastic parameter models have received limited attention to date, evidence shows that models allowing for flexible forms or parameter variation can provide a very good fit to diffusion data and have better predictive validity (e.g., Bretschneider and Mahajan 1980, Easingwood 1987, 1988, Putsis 1988). 49 :r Stethastic : feedback tiiIC meter esti :rorided 5) th example. Bret: rcc'elas 51H : o:- the time-\ .1: .lé-Ierrtt is call »‘ rs lt can 32 limine the n {ref to be adj Li; i‘ectl}, 1‘. d. ‘31:: lili’ -' : Lv'Jk.an “"57 Ol‘lhc Bki I [u ‘L' . . can also h‘ l“ ‘5; (I ghf‘i.’ “SQllQT‘, 2) Stochastic stationary processes Feedback filters and Bayesian approaches have been used in a variety of ways to update parameter estimates. The basic approach of adaptive models is to use the information provided by the forecast error e(t) to update the estimated coefficients 01,- (t) . For example, Bretschneider and Mahajan (1980) express the discrete analog of the Bass model as s(t) = a1(t) + a2 (t)x(t -1) + a3 (t)x2 (t — 1), and sales s(t) are predicted based on the time-varying coefficients, which are defined by ai (t) = ai (t — 1) + Ai (e(t)) . Ai (e(t)) is called the feedback filter, which is a ftmction of the one-step ahead forecast errors. It can also be specified as any weighted average of past errors. The objective is to determine the magnitude and direction by which the previously estimated coefficients need to be adjusted. This can be prespecified and imposed on the system, or estimated directly. Although the application by Bretschneider and Mahajan (1980) is based on the discrete analog of the Bass model, this need not be the case. Another application of feedback filters can also be found in adaptive estimation procedure introduced by Carbone and Longini (1977). An alternative stochastic stationary process is the Kalman filter (Kalman 1960, Kalman and Bucy 1961) estimation technique designed to estimate state variables in a dynamic system in an optimal way. It is based on a probabilistic treatment of process and measurement noises. It can also be used, often in the form of a feedback filter, to update the parameter estimates when new observations become available. The basic form of the 50 fixer-etc Kalma dcgibe the ex grbsen‘ations a! Swim 611113110 tit-ere '10 ~ if, lleastrement e Briers 1‘- “lll. )1} rs the state ‘ arise processes ‘- .‘x .~ pdm Slices, and U 13-7't‘Ctitelt. discrete Kalman filter consists of two sets of equations. System (or transition) equations describe the evolution of the state variable yk. Measurement equations describe how the observations are related to the state of the system. System equations are yk +1: fk lyk ,fl,uk ,tk J+ Gk wk (2.14) where yo ~ linsPoka ~ (0,Q). Measurement equations are 2 k = H k yk + Vk (2.15) where vk ~ (0, R). y k is the state vector, 2 k is the observation vector. Wk and v k are stationary white noise processes uncorrelated with y k and with each other. f k is a vector function of state ( y k ), parameter vector ([3 ), control vector u k and time t k . G k and H k are known matrices, and Q and R are covariance matrices of the process and measurement noises, respectively. Together the set of system and measurement equations represents a state space model. Once a state space model is specified, the Kalman filter can be used to estimate and trace out or smooth the parameter estimates across time periods. Harvey and Phillips (1982), Judge et al. (1985) and Putsis (1998) are other studies that provide applications of the Kalman filter estimation technique. Xie, Song, Sirbu, and Wang (1997) proposed a new approach to diffusion model estimation: the Augmented Kalman Filter with Continuous State and Discrete 51 [tigenalions l which uses dis gametes. an continuous K3 .jtsoi and Ste inadirfrsion s l 1 ll} 3 513551.th t discrete or con irierential eq'. iiScTete nor the Scranton. 1' than the dg ‘E-‘Jd Kill m; B ifSchneider a g he Standm '3‘; p.- - -..4‘ A g; I..- -- Observations (AKF(C-D)). This technique is built based on the extended Kalman filter which uses discrete observations to estimate the state of a continuous system with known parameters, and augmented Kalman filter which estimates unknown parameters in a continuous Kalman filter model. The reviews of these methods can be found in Lewis (1986) and Stengel (1986)). The limitations of the standard Kalman filter method applied in a diffusion setting are addressed by the AKF(C-D) approach. These limitations are: 1) In a standard Kalman filter, both the system and the measurement equations are either discrete or continuous. Because difiusion models often are expressed by a continuous differential equation whereas sales data are obtained in discrete time intervals, neither the discrete nor the continuous Kalman filter is directly applicable to diffusion model estimation. 2) When the discrete version of the Bass model is used as the system equation, the standard Kalman filter method is subject to the same time interval bias as is OLS. (e.g., Bretschneider and Mahajan 1980). 3) The standard Kalman filter can only be applied in situations where the differential equation has an analytic solution (e.g., Lenk and Rao 1990, Sultan, Farley, and Lehmann 1990). It is desirable for a diffusion model to facilitate forecasts early in the product cycle, when there are only a few observations are available. In order to do so, the estimation method should provide a systematic way of incorporating prior information about the likely values of model parameters and an updating formula to upgrade the initial estimates as additional data become available. Further, the estimation method should be directly 52 appiicahle to dit apparent cortsurr solution to the e equation be reu has An analflic Britten as an an; aiiresses these 1 Filter Jet clk‘l‘s‘d 54:31. Fdflm a) 3513' R510 1 l 991.) 1 ‘1 iferential cqw chitin. Secnmi mimetic 0, 323330,] Pmt‘t‘s 71*." "Eran =1 - ““‘Idldllun () 35'; \: [ “fill I l”) o «3 \: f, M T:Xk ‘1 applicable to difiusion models expressed as a differential equation for cumulative sales/ apparent consumption. It should require neither a discrete analog nor an analytical solution to the equation. A discrete analog would require that a continuous differential equation be rewritten as a discrete time equation in a way that introduces a time interval bias. An analytic solution would require that cumulative sales/apparent consumption be written as an analytic function oft. AKF(C-D) method Xie et al. (1997) proposed addresses these issues. It is superior to the previous estimation procedures (e.g., Adaptive Filter developed by Bretschneider and Mahajan (1980), the meta-analysis conducted by Sultan, Farley, and Lehmann (1990), and the Hierarchical Bayesian introduced by Lenk and Rao (1990)) because it does not introduce a time interval bias due to discretizing the differential equation, and does not require the differential equation to have an analytical solution. Second, it can be used to estimate parameters that change over time (deterministic or stochastic). Third, it explicitly incorporates observation error in the estimation process. Another advantage of AKF(C-D) is that its algorithm is straightforward and easy to implement. Furthermore, a parallel AKF(C-D) procedure can be used to overcome the uncertainty in choosing diffusion model structure and/or prior distributions of unknown parameters. The formulation of the AKF(C-D) model is as follows: £92 _ d“) — f, [x(t).¢(t).fl,r]+ w, 1:; =ffilfl,x(’)atl+wfl (2.16) Zk = xk + vk 53 it here : V ‘"W‘ ,. 33.. l. HRH}! ‘-,‘ y? '1‘. ar- “I‘Q‘n-y. ['1 . .g\ 4“ I‘- k‘ . . N '2‘. “\1'1‘ . ac T D. (r ‘ I where x(t) is the cumulative number of adopters, ¢(t) is a vector of covariates (for example, marketing mix variables), B is the unknown parameter vector, wx and wfl are the process noise, x), and Z]: are the true and observed cumulative number of adopters at time tk, and vk is the observation noise. It is assumed that x(0) ~ (x0 ,O’xo) and fl(0) ~ (,60,Pfl0) , {wx,wfl } and {vk } are white noises. {Wx,Wfl }~ (0,Q) and {vk }~ (0, R), and {Wx , wfl } and { v k } are not correlated to one another. The process noise, w includes model specification errors (due to omitted variables and/or x 9 misspecification), and sampling errors which can occur when using the model to describe the difiusion process of a sampled group instead of the entire population. The measurement error, v k , includes random errors in the data collection. The AKF(C-D) algorithm is essentially a Bayesian updating procedure. Using prior experience and knowledge, one gives the initial estimates for the unknown parameters. AKF(C-D) updates the parameter estimates of the diffusion model as new data become available. It estimates parameters and updates the state variables through a time updating process and a measurement updating process. The details of the AKF(C-D) algorithm will be presented in Chapter 3. Xie et al. (1997), using diffusion data for seven products (room air conditioner, color TV, clothes dryer, ultrasound, mammography, foreign language education, and accelerated educational program) compared four time-invariant procedures, namely OLS (Bass 1969), MLE (Schmittlein and Mahajan 1982), NLS (Srinivasan and Mason 1986), and 54 llfxl o gercenta \‘Ft‘f‘w; - “\iu \ algebraic estimation (AE) (Mahajan and Shanna 1986), one time-varying model, namely Adaptive Filter (AF) (Bretschneider and Mahajan 1980), and their AKF(C-D) model. Based on three criteria (mean absolute deviation, mean squared error, and mean absolute percentage deviation (Mahajan, Mason, and Srinivasan 1986)) for comparing one-step- ahead forecasts of the different methods, Xie et al. (1997) found that AKF(C-D) performs better than each of the estimation procedures considered. Further, they extended the AKF(C-D) estimation to the situation in which there is uncertainty in choosing model structure and prior distributions of unknown parameters. There are several advantages of the AKF(C-D) estimation approach that should be considered along with its superior predictive performance. First, it is a general estimation approach that is not restricted by the model structure or by the nature of the unknown parameters. It can be applied directly to a differential diffusion model without requiring the diffusion model to be replaced by a discrete analog or requiring that the diffusion model have an analytical solution. It can be used to estimate both constant parameters and parameters changing over time (both deterministic and stochastic changes). Second, AKF(C-D) is a Bayesian estimation procedure. It can provide better forecasts from the early stages of the diffusion process by incorporating any information on the prior distributions of the parameters in the estimation process and updating the estimates adaptively. Third, AKF(C-D) is a better estimator compared to other methods because it avoids the time-interval bias problem, it is capable of estimating time-varying parameters with or without a prior knowledge of how the parameters change over time, and it accounts explicitly for possible noise during the data collection process. Fourth, the 55 .tKFtC-Dt al.—‘1 rolel structure can be employe there strong p where little info in the latter case AKF(C-D) algorithm is straightforward and easy to implement. Fifth, when multiple model structures or prior distributions are considered, the parallel AKF(C-D) procedure can be employed to deal with the uncertainty. Thus, the model is useful in situations where strong parameters on the parameter estimates are present as well as in situations where little information is known about the parameter estimates. It should be noted that in the latter case nonstationary stochastic processes may have some advantages in developing the priors when prior adoption data are available (Putsis and Srinivasan 2000). 3) Stochastic nonstationary processes Putsis (1998) applied the nonstationary stochastic Cooley and Prescott (1973, 1976) model, which divides each parameter vector in to a permanent component (which persists from one period to the next) and a transitory component (which dissipates at the end of each period). Examples of the changes that persist from one period to the next, and hence would be incorporated into the permanent component of the parameter vector include changes in price elasticity due to improvements in product quality, the impact of increased product familiarity over time, increased benefits due to a larger installed base, and the effect of replacement sales (Eliashberg and Chatterjee 1986, Kamakura and Balasubrarnanian 1987, Horsky 1990). The effect of changes in transitory income, sale price changes, short-term advertising expenditures are likely to be included in the transitory component of the parameter vector. 56 lhe deans of 3 19761. Jud:26 6‘ Simultaneous et There are situati Liar- and Head Deiixrpe. Parke P35110115 section 5512.235 mi [101 liefihood (F N] 125.31 are neces The Dekimpe, P; ...uiar phones 1 13"? .., Gatignon. 1 ‘21'1.. stricter estima 5:54 ‘Wfirted at t‘. \ ._ «11' .(1 r - mil sct‘ ‘— The details of the Cooley-Prescott procedure can be found in Cooley and Prescott (1973, 1976), Judge et al. (1985), and Hanssens et al. (1990). Simultaneous eguation estimation There are situations, for example successive—generation models (Norton and Bass 1987, Islam and Meade 1997) and multinational diffusion of new products (Putsis et al. 1997, Dekimpe, Parker, Sarvary 1998; please note that these studies are also reviewed in the previous section), where estimation techniques designed to address single-equation settings may not be appropriate to use, but methods like full information maximum likelihood (F IML), seemingly unrelated regression (SUR) or three-stage least squares (3SLS) are necessary. The Dekimpe, Parker, and Sarvary (1998) study provides an illustration for adoption of (cellular phones) based on previous research on multinational diffusion of new products (e.g., Gatignon, Eliashberg, and Robertson 1989, Takada and Jain 1991). They reported parameter estimates for the Bass model using NLS applied separately to data on cellular phones service between 1979 and 1992 for 184 countries. However, there are significant issues that resulted in implausible parameter estimates in almost all of the cases. Since service started at different dates in each country, the available degrees of freedom ranged from 1 to 13. Further, the diffusion processes across all 184 countries are certainly not independent. Putsis et al. (1997) point out significant interaction across countries in a multinational setting. Also, network extemalities (Dekimpe, Parker, and Sarvary 2000) as well as supply-side relationships such as production economies (Jain, Mahajan, and 57 llulier 1991 ) m Finally. it is alst cross countries considerations r: neEl. flhlL. 351 DEWEY Parker $333515 15 [Cm ”minim coun t 51111 ‘ ti = '\ 51.1 ‘5 Sal. .aek‘l‘lltmgj threw aafimdk‘lthe' it < < I ‘ llcaPU \ RnleaI lilr 215.3. . .‘i‘i‘er ..P. 8115 dc te— Muller 1991) make it highly unlikely that diffusion in multiple countries are independent. Finally, it is also likely that omitted variables such as price and income are correlated across countries, implying that, at a minimum, SUR estimation is required. Given the considerations regarding the system’s specification and parameter identification issues as well, FIML, 3SLS, or nonlinear 3SLS are more than likely choices for estimation. Dekimpe, Parker, and Sarvary (1998) proposed a three-stage estimation process that they suggested is temporally consistent with the evolutionary nature of the diffusion process. For a given country i, they defined the following adoption function over time: - Si(1) Biin—l) . . _ . _ sh, .. llCiSZi ] +[_—_C,szi ]][C,SZ, x,(t 1)] (2.17) where Si t is sales in country iover the t’h time interval, and xi (t — 1) is cumulative sales (adoptions) through the end of period H. The term 51' (1) represents the penetration level at the end of the first period, SZi denotes the social system size (population), and C ,- C-SZ s- 1 (0 S C,- s 1) captures the long-run adoption ceiling. A“ ={ '( )J denotes the intercept z r of the penetration curve, which is specified as endogenous to the social system. The parameter Bi is defined as the growth-rate parameter that captures the growth that occurs between the intercept and the adoption ceiling. Country-specific covariates are —l . . . — d X . incorporated using longth transformations Ait =[1+ e 1 1 :l and —-1 B,- =[1 + e- dei ] where X ,- denotes a country-specific set of covariates. 58 Tne first stage of (19381qu cons adoption ceilings he intercepts for that stage. interr erdogenous to th. esumation proce or. extemal criter the diffusion ; ”$11 and the g1 19‘ hese criteria Bierszc-od betre “stole. Social i Sap-PS can be U SI:431116015].1 . The first stage of the three-stage estimation procedure Dekimpe, Parker, and Sarvary (1998) used consists of external estimation of the social system sizes and long-run adoption ceilings across countries. These are specified as exogenous. In the second stage, the intercepts for each country, which are exogenous to the next stage are calculated. In third stage, internal estimation of each country’s growth parameter, specified as endogenous to the social system is undertaken using data pooled across countries. The estimation procedure requires sample matching, which entails matching countries based on external criteria. Dekimpe, Parker, and Sarvary (1998) suggested that four components of the diffusion process (the social system size, the adoption ceiling, the time intercept origin, and the growth rate) require matching. Once the countries are matched according to these criteria, the diffusion process based on similar sets of countries can be understood better. Since the diffusion process is temporally sequential in nature (for example, social infrastructure is established before the product launch), the matched samples can be used to estimate diffusion parameters sequentially, rather than simultaneously. The advantages of the staged estimation procedure include sufficient flexibility to explain cross-country variation using either endogenous or exogenous covariates as well as provision of a reasonable basis for testing hypotheses when only the earliest adoption figures are available. This procedure also allows for estimation when products are launched at different times across countries. However, there are also limitations. For example, estimated parameters for the intercept, the long-run adoption ceiling, and the 59 social system size for each country are fixed in the third stage while the ceiling and the population size may change over time. Sequential rather than simultaneous estimation may make parameter estimates inconsistent (Anderson 1970). Accordingly, the estimation procedure employed needs to consider that diffusion of the product may be related across countries after launch in a simultaneous manner (see e. g., Putsis et al.1997). Putsis et al. (1997) developed a model that empirically addresses cross-country relationships. Each country’s adoption equation was specified not only as a function of its own prior adoption but also as a function of prior adoption in the other countries. A mixing parameter and country covariates were also incorporated in the model. Ten individual-country adoption equations were estimated simultaneously with the aim of measuring the individual-country diffusion parameters and the mixing parameter. It should be noted that the technique Putsis et al.(1997) resulted in plausible estimates because the product launches for the products studied had occurred at approximately the same time. Otherwise, Bayesian or unbalanced panel methods would have been more appropriate techniques. There are serious methodological issues that make use of nonlinear methods such as nonlinear SUR, nonlinear 3SLS, or F IML problematic. Convergence issues and concerns about parameter instability are such examples as the models get complex and nonlinear. Also, in 3SLS and FIML, misspecification in a single equation can seriously affect parameter estimates in all of the equations of the system. Further, the solutions of 60 complex nonlinear systems are highly sensitive to the starting values of the parameters and great care should be given to ensure that a global solution is obtained (Putsis et al. 1997). It should also be considered that convergence and identification problems may arise once covariates are incorporated in the model, but not necessarily before. Given the pros and cons of using the different methodologies that are reviewed in this section, we have decided that the most appropriate method to use for the purposes of this dissertation is the Augmented Kalman Filter with Continuous State and Discrete Observations (Xie et al. 1997). The main reasons for this decision are the ability of the AKF (C-D) methodology to do Bayesian updating of the parameter estimates as new data becomes available, and the time-varying nature of the parameter estimates that will enable us to examine the impact of our covariates on the diffusion process over time. 61 gain. :DU'ai - ?.d-A-:--:A- L.v=-=?==. ii: (equalling-I .11 - .N b..—.-.-. .3 can 5 Sage do use 95 £298 2 .33.: 22: 3038—8 2n acute 825.5 28 ._u>o_ cargocoa 05 5293 8 58: 2o:— b03228 93 acute 9550 6 :0 Soto 03:89 a 32 fiancee wfifixo we mmocozmusfioa 52382— floaova weaves acne—30m mo mmocozmgflua .3anan i so “coho 03:8“. corn—anon ”a ..om a 96: cones—Sufi 5:oE¢o.Eo3-=o= accesses 332a 9 bags fiaoE¢o683é8 EB 532:2: $805 9 bane doze—E9: @822 28 nouns—05 828.5 9 mmooo< downer—£5 noes—o.— Szvoa 30> autos—4... .5 so Soho 8 338 “Q .5.”— .m=o>o 053825 5.5m 98 £2.55 328a e 32 .9550 some .mofinomfi 5.82 £2 .oqosm 8:35 05 3 388 6.69:— 05 9 $88 a 5 :38:ng a £2.23 5 85.58 x8 8 whims—=3 5a 8 3035:; 832a 33:3 .3933 mafia—05v 88$ tho—v.54. 23 Sen 3 b=B< mozam 302883: Sam 2 mam—Ba ”E ..om coca who» a EE no ..coEooEuU 98 cone—gov .m 555 mags—uh 4839.4...” mus—a...— 3?2 .5502 8:526 8.3.. EE. .2. :— 8535 gm £933.3— 83959 383.80 he new—EB Quota—=8:— ue Pan—Esm —.N 033—. 62 S U)...- .I.:......Ji.....4_...£ . , 32...: 3:: 5.3).... p............;. _ 331...! 25...... ...1....a>< \ ..23. 3‘3. \ 5.22.... 1.1...- n~n< .fihfiv NOFQGV 5 ”03500 £~ «mam C Siam Ea metal 6&5..on .bLCOU ~ .N U~n~mnh 855.8 den-£6 $88 889? go 3on 05 2: mama—“82%“ 35> 2.3: so Soto 358a a 5:950 .255 053828 sum was .5296. 96: 838 ..ua mzo 05 .«o cotafiumo 838 ..oq n=40 33. £8388 85:38 .§_§Efln=mu_am can $5555 >H 3853383 28 dEEoEso >H 339852 83 25: 6.8.33 no 5285”.“ 2 £35 EEUSE .uo_:au._woow 8: :5 «dong—go daugm 32538 Soto 9:58. no 3:055 93 05 ..8 8:33 Emu—5.9. 63835: .8 090 “muo— BEEocow 335.08 uo 353:8 Q: «E: 55:83 gong—saga £2835: .3 25 28288 55:8? .638 .wfl 2:: 55:83 82 mMU> 85:58 cannon—m .8an mo 85:3 835 man 033 £80.88 .thER 8 8.605 :80 he £8383 08o: Aflofio =a a m_ Soto wEEmB 05 8.“ 33:3 3.52:0 .bfizxoa 89A :38:ng .803 05320? 8.: m: 8 $2.23 €35 Sam—3:593 3:23.380 unno— 535202 3023380 858m 8on .mocozn 33:00 33:8 8.: S can .383— £855 3:5: Emma 05 mo 3.38.. 05 vow—0&3 $82.00 .880583 535$“. 5:83 022 amp—3a 05 8 .3885 8%: Samoa EEobG .motowouao Ba 42an don—o: 8:85 328 at? can .283 52. £93:— 8335, mm. oEu v5 nus—E. .83: 32 no can £8358 Ea 058%-?58 8:35 3:38 05 gauged Ba 8 8355 some .8.“ 2.8: .8523 .«o 883:5 «>522 5 no 8322852: .wbnnmazm .coamzao an» couozvobfi 33:3 .m=o>o 85:38 83: 63852.“ 05 .3352 2.3m .3 >25. 8:833. 8283 52m 053838 55> 52.83— 3 v5 £850 .383. P280 flu 2an .2: z: 5.5:... A1:_:_:::..:_ :7.:3._:;::.£3 :. :2...:::::: 92.21:». 9):: 13:21 b.....u...._r ..u. 3...“... .35: 3:33.: ’2. v) ...r..v»..:L ..I..: 3:: 33.25.... «533‘ ~ 2.1.}.2‘ £1.74; 1:: 5:293 {.39 :71: _ ..2.:.:. 25221;). ..S :5}. .2: 2:1.— ‘ A F v .325 2:... w 22.2312; r; :2: ..::\..<. .2....\.05: €283sz mo 520508 05 so 088 Spa. 95 .flonoob 83: cognac”. Soto 0358 a 8: 03:3 5 5:83 5:39: noon .868 395m EB @3535 Emfiafioaoemoo 68— cos—80:00 .Em_=§_onoEmoU 32-32 :3 a: 8.00 E 85.58 3 .couwumo .88. Eight 525 95503 8:33 Baa 022 83. E flea—:28 .mao how—888 “nova 35 83.500 . ..8 3263— ”.md .85..an Soto 2:: 05 ..8 S anfiowEB wfizmss 058.0 eggs: :o_mmo._w2 295m .8 32-52 km .Eoufiowtuo. be 820508 35.8 3.520 A523. .8 Soto 9632. =£§co cofiémv :08 5 838a $23.. 592323 d23— finom 5.25. a was 88?? 308 05 .5.“ 8.329. genomes—on :03 Ba 80:3 .m>.~ 38m ..m.a SE 058m :3: 05 mo genome—co: 58— 305782 829? Ecom 2:: 32:3 £22328 :< 05 E 85.58 v £2 28 £3qu 4588.9“ 5 8: 2a 5:85—58 woman saga 98 5:85.53 >258 .32 282 .=o_§_E_ 23 539,85 mo £56508 3% 2: so "gecko 3289308 Em_=§_oqumOo S noun—6.55 358a 32 £58m 832:5 6:08.:— 635 638:— 35> .9533 .m.D can 85:58 38: 0885 532 2a £33 820 223 .532 5:33; “a a: 05 so; no at 3.8 533m N. Ea .53 .830: 6.280 fin 039—. 65 ‘ marsh; :t I4 “$.- “l CHAPTER 3 METHODOLOGY The early models of diffusion of new products such as the Bass model (1969) have the major limitation that they do not explicitly incorporate variables that managers and policy makers use for decision/policy making. Therefore, these models are not useful for strategic planning with respect to, for example, pricing and advertising policies or entry decisions of the firms. The importance of the role of decision variables or covariates in understanding the dynamics of the diffusion process has been emphasized by various researchers (e.g., Mahajan and Muller 1979, Kalish and Sen 1986, Mahajan and Wind 1986, Mahajan, Muller, and Bass 1990). In fact, the earliest study to modify the Bass model to include covariates was by Robinson and Lakhani (1975). Their study was followed by others that propose numerous modifications and extensions of the Bass model to examine the effects of covariates on the diffilsion process of new products. Some of these studies incorporated the effect of price (e.g., Bass 1980, Kalish 1985, Kamakura and Balasubramanian 1988, Jain and Rao 1990, Horsky 1990) into the Bass model while others examined the impact of advertising (e. g., Horsky and Simon 1983, Simon an Sebastian 1987). Few other studies (e.g., Bass, Krishnan, and Jain 1994, Bass, Jain, and Krishnan 2000) included both price and advertising effects on the diffusion of new products. Marketing mix variables are not the only covariates researchers included in the diffusion models. For example, Dekimpe, Parker, and Sarvary (1998) include demographic factors such as annual average population grth rate and number of major population centers, economic factors such as GNP per capita and crude death rate, and 66 n ',-l i’lteg Q 4‘. —- w“ . . .- 5.“, “‘4”. 'h-Jm d1_Mv ‘ .‘ fl; :5 ' social system factors such as number of ethnic groups in a country in their model for investigating the diffusion of cellular phones in 184 countries. Similarly, Talukdar, Sudhir, and Ainslie (2002) include factors such as purchasing power parity adjusted per capita income, Gini index, urbanization, international trade volume, illiteracy rate, number of ethnicities in their Hierarchical Bayesian analysis of diffusion of six consumer durable products for 31 countries. For the purposes of this dissertation, these studies are interesting from a methodological point of view. The manner in which covariates are incorporated in the Bass model is our main concern. Hence, the remaining of the discussion will include an overview of the Bass model, and examples of its modifications where covariates are introduced to the model. Then the model to be used in the dissertation will be described along with the AKF(C-D) estimation methodology, and the data characteristics. Before getting into the details of the methodological issues, however, it is necessary to lay out the desirable properties of a diffusion model with covariates. This will enable better evaluation of the usefulness of the diffusion model developed in this dissertation. The guidelines regarding the desirable properties of diffusion models with covariates are mainly based on the prescriptions by Bass, Krishnan, and Jain (1994) and Bass, Jain and Krishnan (2000): o It is desirable for the diffusion models with covariates to describe the empirical diffusion process. Since Bass model is an empirical generalization that describes the diffusion of new products, diffusion models with covariates should reduce to 67 flan-.51.»...- A or be approximately equivalent to the Bass model curve. Empirical generalization can be described as “a pattern or regularity that repeats over different circumstances and that can be described simply by mathematical, graphic, or symbolic methods” (Bass 1995, p. G7). The diffusion patterns for new products usually take the form of an S-shaped curve. The central proposition of the Bass model describes this phenomenon as the conditional probability that an adoption will be made at time t given that an adoption has not yet been made is a linear function of the number of previous adopters. It is appropriate to call the Bass model an empirical generalization because although the indicated diffusion pattern does not always prevail, it usually does. The behavioral rationale of the Bass model suggests that the timing of adoption of those who have not yet adopted at any time t is influenced by the number of previous adopters due to learning or imitation. This implies that the greater the number of adopters today the greater will be the influence on the remaining potential adopters to adopt at each future time period. In the context of diffusion models with covariates, this “carry-through” property means that any change in the covariate variables will affect the diffusion of the new product not only today but also in all future time periods. Diffusion models with covariates should be capable of empirical estimation, and yield plausible parameter estimates. Such empirical support will enhance the credibility of the model, particularly when the number of data sets for which empirical estimates of parameters is large. 68 Goodness of fit alone is not an adequate criterion for evaluating the validity of the diffusion models with covariates. When a diffusion model has a closed-form solution, this makes it easier to understand and interpret the relationships among the variables and the time pattern of adoption. Hence, it is desirable, though not always sufficient for validity, that the diffusion models with covariates have closed-form solutions. Diffusion models with covariates should be constructed in such a way that the managers/policy makers can interpret the parameters of the model either by intuition or with reference to comparison to other products. Diffusion models with covariates should also be implementable for new products without data. The Overview of the Bass Model The framework originally proposed by Bass (1969) constitutes a single-equation model of first purchase with parameters that remain constant over time. It postulates the likelihood that an individual would adopt a new product at time t, given that he/she has not yet adopted. The model can be stated as follows: s(t)=—(—)= [p+{q —')}][mx(— x(t)]= pm+[(q- p)x(t)]— [(3)402] (3.1) where s(t) is the sales rate at time t, x(t) is the cumulative number of adopters at time t (with x(0) = 0), and m is the maximum number of potential adopters. The parameter p is the coefficient of innovation or the coefficient of external influence. It reflects the tendency to adopt at time 0. Note that s(O) = pm. The parameter q is the coefficient of 69 imitation or [ht conununicalior Definingfin 3: Fm as the cum atfime! is giu flu 14111-!) filming R n i‘g’ {1+9 L P Fin: in“ in b\ : imitation or the coefficient of internal influence. It reflects word—of—mouth communication. Defining fit) as the density fimction describing the time of adoption for a population, and F (t) as the cumulative density, the hazard function describing the probability of adoption at time t is given by: f0) ]_ F(t) = p + qF(t) (3.2) Assuming F (0) = 0, 1_e-(p+q)t F(t) = (3.3) 1+ ie_(p +q)’ _ p Assuming that each adopter only buys one unit, the sales rate, s(t) = mf(t), where fit) is given by: (PW)2 e-(p+q)t p f (I) = (3-4) 2 [1+ie-(p+q)t] u p .1 The point of inflection in F (t) occurs at time t*, which corresponds to the maximum penetration rate. F(t*) = g 7’;— (3.5) “‘ifi‘ai'ffl 7O I"! ,l. : -. i - 411v" f(t*)=1+-§+£— (3.7) Models with Marketing Mix Variables In the next two sections an overview of the modifications to the original Bass (1969) model will be given. We will start with examples of models that include marketing mix variables as covariates. Then we will give examples of studies where macro level covariates are included in the Bass model. For our purposes, these models are interesting in terms of the manner in which the covariates are incorporated in the model. Models with Price Alone Robinson and Lakhani (I 975) and Others Robinson and Lakhani (1975) were the first to introduce decision variables into the diffusion models. The model they used to examine the profit differences due to optimal pricing sequence over the planning horizon versus a myopically optimal pricing policy (i.e., marginal revenue equals marginal cost in each period) is: 3(1) = (m— Y)(p+qY)e—k Pr“) (3.8) where Y is the cumulative sales, Pr(t) is price at time t, and k is a price coefficient. The main problem with the Robertson-Lakhani model is that it is not viable for empirical estimation unless k is near 0 or price is constant (then it reduces to the Bass model). Other models similar to the Robertson-Lakhani model are used in Horsky and Simon (1983) 71 and Dockner ar problematic as Bass I} W); lerass ( 108 Q“: .fltlrll than: Qm is 2'56 Bass mod pm 613.51le 305" and Dockner and Jorgenson (1988). Estimation with these models is generally problematic as well (Bass, Jain, and Krishnan 2000). Bass (1 980) The Bass (1980) model starts with a constant elasticity demand function: Q(t)=f(t)c[Pr(t)]'7 (3.9) where Q(t) is the quantity demanded at time t, Pr(t) is the price of product at time t,f(t) is the Bass model specification given in Equation 3.2, c is the cost function, and I] is the price elasticity parameter. Under the assumptions firms are choosing prices to maximize profits myopically, and marginal costs follow the experience curve with A as the learning rate parameter the closed form solution for demand at time t is: Q(t)=[(1_':n)]f(t)F(t)M/(1_M) (3.10) This equation is capable of empirical estimation of p, q, m, and 77 once xi is estimated from the historic relationship of price and experience curve, and then substituted in the equation. Although closed form solution and capability of empirical estimation are desirable properties, the above equation does not reduce to the Bass (1969) model when prices are changing. It also lacks the carry-through property; it is a current-effects model. Hence its managerial usefulness and implementation are limited. Kalish (I 985) In Kalish (1985) study the main argument is that it is the information about the new product, not the new product per se, that diffuses in the social system through advertising 72 pert: .":.. Lilli. (036 318. 4'1 541 113', . 1".) . “35? ’V23 in: his: and word-of-mouth. Hence the model is built to capture how the new product is perceived in the market with respect to the uncertainty surrounding its performance, its utility to the potential adopter, and its price. The two simultaneous differential equations (one for the awareness process, the other for the adoption process) representing the model are: fl = [l — l][f(A(t)) + b] + UPI—(tin dt m 4!. Pro) _ dt -[g[u(Y(t)/m)]1 ”0]" where Y(t) is the cumulative sales up to time t, Pr(t) is the price at time t, A(t) is the (3.11) advertising at time t, I is the information or awareness level, m is the initial market - potential, g, j: and u are function operators, and b, b', and k are parameters. Assuming I to be 1, g has an exponential and u has a quadratic specification at the empirical stage, the model reduces to the following empirically testifiable form where advertising is no longer included: S(t)= mex — ,BPr(t)(l+;a) —Y(t)k (3.12) 7a+(Y(t)/m)2 where 5 and y are new parameters. This model has the desirable carry-through property since price is included in the model in such a way that an impact on sales at time I created by a price change at time t is carried to the future periods through the diffusion effect. However, this model has no solution expressed as a function of time, price and advertising alone (a closed form 73 «s ‘0 51L )‘V «I. I. rd“ \ solution), and contains constructs difficult to measure in a real market, which in turn limits its managerial usefulness. Kamakura and Balasubramanian (I 988) The main purpose of the Kamakura and Balasubramanian (1988) model was to use an extended data set that goes well beyond the usually used a few years of data in a modified Bass model to test the impact of price on diffusion speed or market potential. The model where the price effect is incorporated explicitly is: 5(1) = [p + qY(t)]Pr(t)'6] [6M(t) Pr(t)’62 — no] (3.13) where [31 represents the impact of price on the adoption speed, ,6; represents the impact of price on the market potential, M(t) is the number of households with electricity at time t, and 6 is the ultimate penetration level. Kamakura and Balasubramanian (1988) estimated this model using a discrete time formulation while the diffusion equation is a continuous one. This may have introduced a time-interval bias in the parameter estimates (Schmittlein and Mahajan 1982). Although this model has the carry-through property, it does not reduce to the Bass model unless price is constant, and it does not have a closed form solution. Jain and Rao (1 990) Jain and Rao (1990) used the continuous time formulation of the Bass (1969) model (to overcome the methodological shortcomings of the Kamakura and Balasubramanian (1988) model, and proposed two model specifications that include price: 74 vi run. F(t)— F(t —1) l-Fo—n F(t)— F(t —1) 1— F(t —1) 5(1) = (mPr(t)- '7 — Y(1—1)) (3.14) so) = (m — Y(t —1))Pr(:)“ '1 where F (.) is the Bass (1969) model given by Equation 3.3. In the first model, price affects the market potential whereas in the second model it affects the remaining sales potential. The estimation results yielded that the second model is better in terms of fit. It should be noted that price is modeled to affect the sales function (Equation 3.3) rather than the basic diffusion process (Equation 3.1). This sales fimction was derived independently by solving the differential Equation 3.1. Also, the Jain and Rao (1990) model does not have the carry-through property, and has a closed form solution for part of the model that does not include the price variable. Horsky (I 990) The model Horsky (1990) built is based on an economic principle that a new product is adopted because it maximizes the net utility for the household. The consumer evaluates the benefits of the new product (time saving, and utility enhancement) using reservation price (the maximum price a consumer is willing to pay given the benefits of the new product), his wage rate, and the price of the new product. The final form of the diffusion model Horsky (1990) used is: 6M0) S = — . (I) { K+w(t)_kpr(t)} Y(t) [p+qY(t)] (315) 1+exp - L 59) . 75 where the pr tine-i "IMU “lulu-i ‘y-ei \ u.“ \ ‘h- where w(t) is the average wage rate of the population, Pr(t) is the average market price of the product, 6 represents the dispersion of both of these distributions, K and k are the time-saving and utility enhancing attributes of the new product, respectively, Y(t) is the cumulative sales up to time t, M(t) is the total number of households in the US. population, 6 is the fraction who are potential buyers, and p and q are the diffusion parameters. Note that both the wage distribution across the population and the price distribution change over time. The first part of the model is the income-price-market saturation component, and the second part is the diffusion process. The Horsky (1990) model has the desirable carry-through property. It does not have a closed form solution though. Further, the empirical results of the study are open to criticism that Horsky (1990) used data sets with long time intervals introducing bias to the parameter estimates due to replacement purchases. Although it was assumed that 70 percent of the demand in later years was replacement purchases, there is no empirical support for this assumption on which empirical results heavily depend. Still, it should be noted that Horsky (1990) attempts to explain why rather than when individuals buy a new product, an approach that should be commanded. Models with Advertising Alone Horsky and Simon (1 983) The underlying proposition of the Horsky and Simon (1983) model is that advertising affects the innovative characteristics of the adoption population. Their model is: 3(1) = [a + mil/1(1) +qY(t —1)Im— Y(t —1)] (3.16) 76 fne p19) .1, £41er H a- 11" “mm where A(t) is the advertising at time t, and fl is the effectiveness of advertising. The empirical estimation of the model (using data for telephone banking in five cities) provided good fit and significant estimates for advertising effectiveness. Simon and Sebastian (I 98 7) The argument of the Simon and Sebastian (1987) study is that advertising could affect either innovation or imitation component of the diffusion process. Hence, they empirically test the following two models using telephone adoption data in Germany. 80) = [p + 121040)) + qm —1)Im — Ya — 1)] 3.17 so) = lp+ {q+qf(A(t»}Y(t-1)Im—Y(t—1)] ( ) where A(t) is the advertising in time t. The function f had one of the three alternative forms: ln(A(t)), 2; = Oln(/4(1)) , or ln( WA (1)) where WA (t) is the weighted average of the advertising from time 0 to time t. The empirical estimation of the (six) models yielded the conclusion that advertising affects the imitation component of the diffusion process. Given that both of the two studies that incorporated advertising in the diffusion model used data for only one product and OLS as the estimation procedure, their empirical results need to be accepted with caution. Also notice that, although both Horsky and Simon (1983) and Simon and Sebastian (1987) models possess the carry-through property, neither have a closed form solution. 77 Sicie ———— (It: (16' there Ziil: Models with Price and Advertising Generalized Bass Model (Bass, Krishnan, and Jain 1994) Bass, Krishnan, and Jain (1994) have proposed a generalized form of the Bass model that incorporates the effects of marketing mix variables on the likelihood of adoption of the new product. They added a multiplicative factor Z(t) to the original Bass (1969) model: s(r)- - —Q- —[p+{q "( —)mH[m— x(t)]Zm (3.18) where _. P(t) 4(1) Z(t)—1+a[-P—()]+,B[ A(t)] (3.19) P(t) and A(t) are the price and advertising at time t. P'(t) and A'(t) are the rate of change in price and advertising respectively at time t. The generalized Bass model is solvable to yield sales as an explicit function of time, price, and advertising in the following form: e— (p + q)(t + ,6] ln(Pr) + )62 ln(A) 2 . . S(,)=m(p+q) [1+3] Pr(t)”, A (a) P 2 Pr(t) Am [I + i e‘ (p + q)(t + 131 ln(Pr) + 192 ln(A) ]2 P (3.20) The generalized Bass model has all of the desirable properties described at the beginning of this chapter. It reduces to the Bass model when prices and other decision variables are changing by a constant percentage; it has a closed form solution; and it has the carry- through property. The model is managerially useful and implementable as well. The generalized Bass model has strong theoretical and empirical support. It provides an 78 explanation of the existence of the empirical generalization of the Bass diffusion curve, which is almost always observed in diffusion phenomena. One practical concern, however, arises because the covariates in the generalized Bass model act through percentage changes in the variables rather than the levels of the variables. This problem can partially be handled by an assumption that the initial price acts as a reference point, and subsequent changes in price may be used to convert to price levels (Krishnan, Bass, Jain 1999). Proportional Hazard Models (Bass, Jain and Krishnan 2000) Bass, Jain and Krishnan (2000) present a proportional-hazard model of diffusion using the proportional-hazard fiamework introduced by Cox (1972). Although these models have the limitation of being current-effects models, they are worth comparing with the generalized Bass model, particularly because both changes and levels of the covariates can be used in these diffusion models. The hazard function is given by A(t, Z) where t, time to adoption, is a random variable with the distribution function F and the density function f. A(t, Z) represents the instantaneous probability of adoption at time t given that the individual did not adopt before time t. Hence, it can be expressed as: _ f (AZ ) 30,2) _ —————-[1 _ F(t,” (3.21) Integrating Equation (3.21) and using F (t, Z) = O as the initial condition, the distribution function showing the one-to-one correspondence between the distribution function and the hazard function of the random variable Z is: 79 l F(t, Z) = l — exp[- Il(u,Z)du] (3.22) 0 Following Cox (1972), the conditional hazard rate A(t, Z) can be expressed as 20,2) = 20(t)0{2(1)}. Using the specification of the Bass (1969) model (Equation 3.2) as the base-line hazard function (7.0), and modeling function of the marketing mix variables, 6{Z(t)}, that act multiplicatively on the baseline hazard rate as an exponential function (i.e., exp{Z(t),6} where ,8 is the set of unknown coefficients associated with the marketing mix variables), the hazard function and the distribution function can be expressed as: 20,2) = p + ‘1 exp{Z(t)} [I + %exp(( p + q)t)] (3.23) l F(t, Z) = 1 —- exp - j p + ‘1 exp{Z(u),6}du (3.24) 0 1+ %exp((p + q)u b -1 Equation 3.24 takes the following form after using the properties of definite integrals and evaluating it: i r q - + exr)((p + q)r) F(t,Z)=l—exp 22:1‘CXP{Z(TW}1"8 p t (3.25) %+exp«p+qxr—l) \ 4 ..J Equation 3.25 represents the distribution function of adoption time T in the presence of marketing mix variables. 80 Since, in practice, the number of individuals who have adopted the product in each time period rather than the individual adoption times are known, Bass, Jain and Krishnan (2000) worked on the discrete proportional-hazard model for grouped data (Kalbfleisch and Prentice 1980) in order to estimate the diffusion parameters. i r -- 1 . F(t,-,Z) - F(t,-=12) + ex13((1) + (1)1, ) 1- F 1 z :1. “P ‘exl’iZUiWi'Og (Fl, ) %+CXP((P+Q)(11_1) P ' -'expiZ(t,-)fl} 1+—lq;+exp(—(p+q)ri_l) =1" 6"? ’(P+‘1)(’i"i—1) (3.26) 1+ 1 + exp<

— x(t)]g(t) —m— m + b_D_.(_t) + 01:92 T(t) 13(1) 1(1) g0)=1+a where T(t) is trade volume at time t, T ‘(t) is the rate of change in trade volume at time t, D(t) is foreign direct investment inflows at time t, D ‘(t) is the rate of change in foreign direct investment inflows at time t, 1(t) is GDP at time t, I ‘(t) is the rate of change in GDP at time t. The ratios of growth rate to level for the covariates create indexes that allow comparability across countries. The specification of our model is inspired by the Generalized Bass Model as mentioned earlier. 2 k is observed sales. Algorithm The formulation of the AKF (C-D) procedure is: £12 - (10) — fx[x(t)9¢(t)9fl9t]+ Wx w_ 711— - fflLa,x(1),1]+wfl (3.38) %=%+% 96 where x(t) is the cumulative number of adopters, ¢(t) is a vector of covariates (for example, marketing mix variables, or globalization drivers), [3 is the unknown parameter vector, wx and Wfl are the process noise, x k and z k are the true and observed cumulative number of adopters at time t k , and vk is the observation noise. It is assumed that x(0) ~ (x0,0'x0)and ,6(0) ~ (flOJ’flO), {wx, wfl} and {vk } are white noises. {wx , wfl }~ (0,Q) and {vk }~ (0, r), and {Wx , wfl } and {vk } are not correlated to one another. Note also that the augmented version is: 9% =fy(y(1).¢(1),t)+wy f, = F Model 7 0.09569633 0.01367090 15.30 < 0.0001 Error 164 0.14655195 0.00089361 Total 171 0.24224828 Source DF Sum of Squares Mean Square F-value Pr > F Country type 1 0.04163342 0.04163342 46.59 < 0.0001 Product 3 0.03070763 0.01023588 1 1.45 < 0.0001 Country it Product 3 0.02217083 0.00739028 8.27 < 0.0001 Tukey’s Studentized Range QISD) Test for p Alpha 0.05 Error Degrees of Freedom 164 Error Mean Square 0.000894 Critical Value of Studentized Rang; 3.67070 Minimum Sigpificant Difference 0.0167 Means with the same letter are not significantly different. Tukey Grouping Mean N Product A 0.048650 43 CD Player B 0.028355 43 Home Computer C B 0.023623 43 Mobile Phone C 0.010793 43 Video Camera 113 Table 5.2 ANOVA Results for q Cell Means for q CD Player H. Computer M. Phone Video Cam Develomtd Countries 0.2821 0.2385 0.2425 0.1973 Less Developed Co. 0.1979 0.2002 0.2401 0.1790 ANOVA Dependent Variable: 1 Source DF Sum of Squares Mean Square F-value Pr>F Model 7 0.17589373 0.02512768 9.96 < 0.0001 Error 164 0.41378102 0.00252306 Total 171 0.58967475 Source DF Sum of Squares Mean Square F-value Pr > F Country type 1 0.05504926 0.05504926 21.82 < 0.0001 Product 3 0.07947266 0.02649089 10.50 < 0. 0001 Country 11 Product 3 0.04049389 0.01349796 5.35 0.0015 Tukey’s Studentized Range LHSD) Test for q Alpha 0.05 Error Deg'ees of Freedom 164 Error Mean Square 0.002523 Critical Value of Studentized Range 3.67070 Minimum Significant Difference 0.0281 Means with the same letter are not significantly different. Tukey Grouping Mean N Product A 0.24129 43 Mobile Phone A 0.24095 43 CD Player A 0.21980 43 Home Computer B 0.18833 43 Video Camera 114 Table 5.3 AN OVA Results for a Cell Means for a CD Player H. Computer M. Phone Video Cam Developed Countries 0.2985 0.3044 0.2249 0.2943 Less Developed Co. 0.2816 0.7801 0.33157 0.3031 ANOVA Dependent Variable: a Source DF Sum of Squares Mean Square F-value Pr>F Model 7 4.57054236 0.65293462 1.26 0.2740 Error 164 85.07702515 0.51876235 Total 171 89.64756751 Source DF Sum of Squares Mean Square F-value Pr > F Country type 1 0.83786927 0.83786927 1.62 0.2056 Product 3 2.12861 139 0.70953713 1.37 0.2545 Country it Product 3 1.68659298 0.56219766 1.08 0.3576 Tukey’s Studentized Range (HSD) Test for a Alpha 0.05 Error Degas of Freedom 164 Error Mean Square 0.518762 Critical Value of Studentized Range 3.67070 Minimum Sigpificant Difference 0.4032 Means with the same letter are not significantly different. Tukey Grouping Mean N Product A 0.5367 43 Home Computer A 0.2986 43 Video Camera A 0.2902 43 CD Player A 0.2692 43 Mobile Phone 115 Table 5.4 ANOVA Results for b Cell Means for b CD Player H. Computer M. Phone Video Cam Developed Countries 0.1617 0.1742 0.1658 0.1296 _rlsess Developed Co. 0.1219 0.5512 0.1668 0.1400 ANOVA RESULTS Dependent Variable: 0 Source DF Sum of Squares Mean Square F-value Pr>F Model 7 2.99897480 0.42842497 2.77 0.0095 Error 164 25.35825131 0.15462348 Total 171 28.35722612 _Source DF Sum of Squares Mean Square F-value Pr > F @1111! type 1 0.32635737 0.32635737 2.11 0.1482 £30111.“ 3 1 .51439940 0.50479980 3 .26 0. 0229 \(fimtry x Product 3 1.21920468 0.40640156 2.63 0.0521 ukey’s Studentized Rapge (HSD) Test for b $111111 0.05 wr Degpees of Freedom 164 agror Mean Square 0.154623 Critical Value of Studentized Rage 3.67070 Minimum Significant Difference 0.2201 I\tleans with the same letter are not significantly different. Tukey Grouping Mean N Product A 0.35832 43 Home Computer B A 0.16630 43 Mobile Phone B A 0.14223 43 CD Player B 0.13470 43 Video Camera 116 Table 5.5 ANOVA Results for c -.-'1‘~. a" ”we“: .. Cell Means for c CD Player H. Computer M. Phone Video Cam Developed Countries 0.3545 0.2299 0.2247 0.2008 Less Develqped Co. 0.2972 0.6488 0.2624 0.3105 ANOVA RESULTS Dependent Variable: c Source DF Sum of Squares Mean Square F—value Pr>F Model 7 3 .06473 834 0.43 78 1 976 1.49 0.1746 Error 164 48.23314227 0.29410453 Total 171 51.29788061 Source DF Sum of Squares Mean Sirare F-value Pr > F Country type 1 0.69625159 0.69625159 2.37 0.1258 Product 3 1.03990055 0.34663352 1.18 0.3196 Country it Product 3 1.36898336 0.45632779 1.55 0.2032 Tukey’s Studentized Rang; (HSD) Test for c Alpha 0.05 Error Degrees of Freedom 164 Error Mean Square 0.294105 Critical Value of Studentized Range 3.67070 Minimum Sigpificant Difference 0.3036 Means with the same letter are not sigpificantly different. Tukey GroupinL Mean N Product A 0.4345 43 Home Computer A 0.3265 43 CD player A 0.2544 43 Video Camera A 0.2431 43 Mobile Phone 117 Following the individual ANOVA analyses, a MANOVA analysis is run as well in order to ensure there are no differences in the results when the model parameters (p, q, a , b and c) are used as dependent variables simultaneously. The MANOVA analysis allows assessing the effect of country type (developed and developing), product type (CD player, home computer, mobile phone and video camera) and the interaction of country type and product type on all of the model parameters (p, q, a , b and c) as a set of dependent variables. The results are given in the following tables. Table 5.6 MANOVA Results For p Source DF Sum of Squares Mean Square F—value Pr > F Model 7 0.09569633 0.01367090 15.30 < 0.0001 Error 164 0.14655195 0.0008936] Total 171 0.24224828 Source DF Sum of Squares Mean Square F—value Pr > F Country type 1 0.04163342 0.04163342 46.59 < 0.0001 Product 3 0.03070763 0.01023588 1 1.45 < 0.0001 Country 11 Product 3 0.02217083 0.00739028 8.27 < 0.0001 For q Source DF Sum of Squares Mean Square F-value Pr>F Model 7 0.17589373 0.02512768 9.96 < 0.0001 Error 164 0.41378102 0.00252306 Total 171 0.58967475 Source DF Sum of Sgpares Mean Square F-value Pr > F Country type 1 0.05504926 0.05504926 21.82 < 0.0001 Product 3 0.07947266 0.02649089 10.50 < 0. 0001 Country x Product 3 0.04049389 0.01349796 5.35 0.0015 118 Table 5.6 Cont’d. For a Source DF Sum of Squares Mean Square F -value Pr>F Model 7 4.57054236 0.65293462 1.26 0.2740 Error 164 85.07702515 0.51876235 Total 171 89.64756751 Source DF Sum of Squares Mean Square F—value Pr > F Country type 1 0.83786927 0.83786927 1.62 0.2056 Product 3 2.12861139 0.70953713 1.37 0.2545 Country x Product 3 1.68659298 0.56219766 1.08 0.3576 Forb Source DF Sum of Squares Mean Square F-value Pr>F Model 7 2.99897480 0.42842497 2.77 0.0095 Error 164 25.35825131 0.15462348 Total 171 28.35722612 Source DF Sum of Squares Mean Square F-value Pr > F Country type 1 0.32635737 0.32635737 2.11 0.1482 Product 3 1 .51439940 0.50479980 3.26 0.0229 Country it Product 3 1.21920468 0.40640156 2.63 0.0521 Forc Source DF Sum of Squares Mean Square F-value Pr>F Model 7 3 .06473 834 0.43 78 1 976 1.49 0.1746 Error 164 48.23314227 0.29410453 Total 171 51 .29788061 Source DF Sum of Squares Mean Square F-value Pr > F Country type 1 0.69625159 0.69625159 2.37 0.1258 Product 3 1.03990055 0.34663352 1.18 0.3196 Country x Product 3 1.36898336 0.45632779 1.55 0.2032 119 -.»a.- . 2 L. .. . Table 5.7 MANOVA Hypotheses and Results " “ ' 4.1. 3‘. u- “1&2... ..J- ._:x 1 ' 5'- " Ho: No overall effect of country type Wilks’ Lambda F-value lkgrees of Freedom Pr >F 0.73298807 11.66 5 < 0.0001 Ho: No overall effect of product type Wilks’ Lambda F-value Defies of Freedom Pr >F 0.68959859 4.25 15 < 0.000] H0: No overall interaction effect of country type and product type Wilks’ Lambda F-value Degees of Freedom Pr >F 0.78908264 2.64 15 0.0008 All three hypotheses are rejected. Significant overall effects of country type, product type and interaction of country type and product type are found. Table 5.7 Tukey’s Studentized Range (HSD) Test (for effect of country type) Means with the same letter are not significantly different. Country Mean Mean Mean Mean Mean type for p for q for a for b for c Developed 0.043 A 0.240 A 0.281 A 0.158 A 0.253 A Developing 0.012 B 0.204 B 0.420 A 0.245 A 0.380 A Table 5.8 Tukey’s Studentized Range (HSD) Test (for effect of product type) Means with the same letter are not significantly different. For p Tukey Grouping Mean Product A 0.048650 CD Player B 0.028355 Home Computer C B 0.023623 Mobile Phone C 0.010793 Video Camera For q Tukey Groupipg Mean Product A 0.24129 Mobile Phone A 0.24095 CD Player A 0.21980 Home Computer B 0.18833 Video Camera 120 Table 5.8 Cont’d. For a Tukey Grouping Mean Product A 0.5367 Home Computer A 0.2986 Video Camera A 0.2902 CD Player A 0.2692 Mobile Phone For b Tukey GroupinL Mean Product A 0.35832 Home Computer B A 0.16630 Mobile Phone B A 0.14223 CD Player B 0.13470 Video Camera Forc Tukey Grouping Mean Product A 0.4345 Home Computer A 0.3265 CD player A 0.2544 Video Camera A 0.2431 Mobile Phone Table 5.9 Marginal Cell Means (for interaction effect of country and product type). p q a b c Developed CD Player 0.082369 0.282063 0.298467 0.161685 0.354450 Countries Home Comp. 0.042124 0.238490 0.304352 0.174184 0.229882 Mobile Phone 0.031052 0.242469 0.224870 0.165816 0.224653 Video Camera 0.016679 0.197259 0.294337 0.129641 0.200835 Developing CD Player 0.013325 0.197878 0.281579 0.121859 0.297204 Countries Home Comp. 0.013930 0.200213 0.780106 0.551221 0.648788 Mobile Phone 0.015841 0.240061 0.315738 0.166816 0.262429 Video Camera 0.004628 0.178970 0.3031 13 0.140001 0.310526 The graphical representations of the interaction effects for p, q, a, b, c are given in Appendix G. 121 These results indicate that: 0 There is a significant main effect of country type for the coefficient of external influence (p). 0 There is a significant main effect of product type for the coefficient of external influence (p). o For coefficient of external influence (p): CD player is significantly different than home computer, mobile phone and video camera; video camera is significantly different than CD player and home computer; mobile phone and video camera are not significantly different from each other. 0 There is a significant interaction effect of country type and product type for the coefficient of external influence (p). 0 There is a significant main effect of country type for the coefficient of internal influence (q). 0 There is a significant main effect of product type for the coefficient of internal influence (q). 0 For coefficient of internal influence (q): Video camera is significantly different than CD player, home computer and mobile phone; CD player, home computer and mobile phone are not significantly different from each other. 0 There is a significant interaction effect of country type and product type for the coefficient of internal influence (q). 0 There is not a significant main effect of country type for the coefficient that shows the increase in diffusion speed resulting fi'om a one percent increase in per capita trade volume (a). 122 There is not a significant main effect of product type for the coefficient that shows the increase in diffusion speed resulting from a one percent increase in per capita trade volume (a). There is not a significant interaction effect of country type and product type for the coefficient that shows the increase in diffusion speed resulting from a one percent increase in per capita trade volume (a). There is not a significant main effect of country type for the coefficient that shows the increase in diffusion speed resulting fi'om a one percent increase in per capita FDI inflows (b). There is a significant main effect of product type for the coefficient that shows the increase in diffusion speed resulting from a one percent increase in per capita FDI inflows (b). For the coefficient that shows the increase in diffusion speed resulting from a one percent increase in per capita F DI inflows (b): Video camera is significantly different than CD player, home computer and mobile phone; home computer is significantly different than CD player, mobile phone and video camera; CD player and mobile phone are not significantly different from each other. There is not a significant interaction effect of country type and product type for coefficient that shows the increase in diffusion speed resulting from a one percent increase in per capita FDI inflows (b). There is not a significant main effect of country type for the coefficient that shows the increase in diffusion speed resulting from a one percent increase in per capita GDP (6). 123 There is not a significant main effect of product type for the coefficient that shows the increase in diffusion speed resulting from a one percent increase in per capita GDP (c). There is not a significant interaction effect of country type and product type for the coefficient that shows the increase in diffusion speed resulting from a one percent increase in per capita GDP (c). Based on these, we can conclude that: The coefficients of external influence (p) for the developed countries and the developing countries have not converged and are still different from each other. The coefficients of internal influence (q) for the developed countries and the developing countries have not converged and are still different from each other. Globalization drivers (per capita trade volume, per capita FDI inflows and per capita GDP) impact the difiusion process in a similar way for the developed countries and the developing countries. One might find this result surprising, particularly with respect to the impact of per capita GDP. It should be noted that all of the products used in the data analysis were technological products and we used the percentage of urban population as the market potential. Urban population in both developed countries and developing countries are probably more technologically savvy and less price sensitive with respect to the purchase of products such as CD player, home computer, mobile phone and video camera compared to the general population. Hence, no difference regarding the impact of 124 per capita income on the diffusion process in the developed and the developing countries are detected. The values of the model parameters p, q, a , b and c for each of the forty-three countries and each of the four products are given in Appendix A. Appendix B contains the graphical representation of the coefficient of external influence (p) and coefficient of internal influence (q) over time for each country-product combination. Appendix C contains the graphical presentation of the diffusion patterns of each product in each country. In order to facilitate the visual examination of the results, same information is presented in different forms. In Appendix D, the graphs show the coefficients of external influence (p) and coefficients of internal influence (q) of each developed country and each developing country over time plotted as a group for each of the four products. Appendix E contains the graphs where the averages of p and q across developed countries and developing countries for each of the ten time periods are plotted for each of the four products. Appendix F presents the same information in a third way through graphs of the averages of p and q across developed countries and developing countries for all four products shown for each country type. Finally, Appendix G contains the graphical representations of interaction effects for p, q, a , b and c. Based on the examination of the graphs given in the Appendices, one can conclude that: o The coefficient of external influence (p) increases over time for both developed and developing countries contributing to faster diffusion of innovations. 125 o The coefficient of internal influence (q) tends to slightly decrease over time for both developed and developing countries. Hence, it does not contribute to faster diffusion of innovations. In summary: Hypothesis 1: The diffusion rates for the developed and the developing countries have converged and became similar. Not supported. liYDOthesis 2a: The coefficient of external influence (p) for the developed and the developing countries have increased over time contributing to faster diffusion rates. Supported. Hypothesis 2b: The coefficient of internal influence (q) for the developed and the developing countries have increased over time contributing to faster diffusion rates. Not supported. Contributions The contributions of this dissertation from academic and managerial perspectives are discussed below. The approach in this dissertation allowed quantifying the impact of macro level globalization drivers on the diffusion of innovations process in forty-three developed and developing countries. The model parameters a , b and c reflect the effectiveness of per capita trade volume, per capita FDI inflows and per capita income over the simple time-based diffusion. It was expected that, once the impact of the globalization phenomenon was accounted for by the globalization drivers in the model, we would find convergence of the coefficient of external influence and coefficient of 126 internal influence for developed and developing countries. This would have significant managerial implications as well. Convergence in the diffusion rates of new products due to the impact of globalization drivers would not only mean the possibility of using a standardized global approach through global brands and marketing strategies but also have implications for the timing and order of entry in the global markets. Similarities in diffusion rates of new products among different countries could suggest that firms adopt a sprinkler approach (Kalish, Mahajan, and Muller 1995) instead of a more prudent waterfall strategy, which in turn could translate into first-mover advantages such as consumer loyalty and high market share. Although the hypothesis of convergence of the diffusion rates for developed and developing countries was not supported, the results still have significance in terms of increasing the knowledge base regarding the values of diffusion parameters in a wide range of countries for moderately to highly technological consumer durable products. This knowledge not only provides more informed priors regarding diffusion parameters for researchers but also can be used in forecasting and strategy making by the managers and by the policy makers alike. It is also important to note that globalization impacts the developed and developing countries in a similar manner since no statistically significant differences regarding the coefficients of the globalization drivers between developed and developing countries were found. The methodology used in this dissertation, Augmented Kalman Filter with Continuous State and Discrete Observations (AKF (C-D)) was introduced in the literature by Xie et al. (1997). The advantages of the AKF (C-D) methodology include not only its superior predictive performance compared to other procedures such as Ordinary Least Squares 127 (OLS) (Bass 1969), Maximum Likelihood Estimation (MLE) (Schmittlein and Mahajan 1982), Nonlinear Least Squares (N LS) (Srinivasan and Mason 1986), Algebraic Estimation (AB) (Mahajan and Sharma 1986) and Adaptive Filter (AF) (Bretschneider and Mahajan 1980) but also other desirable characteristics. First, it is a general estimation approach that is not restricted by the model structure or by the nature of the unknown parameters. It can be applied directly to a differential diffusion model without requiring the diffusion model to be replaced by a discrete analog or requiring that the diffusion model have an analytical solution. It can be used to estimate parameters changing over time (both deterministic and stochastic changes). Also, AKF (C-D) is a Bayesian estimation procedure. It can provide better forecasts from the early stages of the diffirsion process by incorporating any information on the prior distributions of the parameters in the estimation process and updating the estimates adaptively. Third, AKF (C-D) is capable of estimating time-varying parameters with or without a prior knowledge of how the parameters change over time. Further, uncertainty about parameter estimates can be built into the estimation. Hence, the model is useful in situations where strong parameters on the parameter estimates are present as well as in situations where little information is known about the parameter estimates. We have taken advantage of all of these characteristics of the AKF (C-D) methodology and applied it for the first time in the literature to a Generalized Bass Model that incorporates the impact of three macro level globalization drivers using data for a wide range of developed and developing countries for four moderately to highly technological consumer durable products. As a result, we obtained five model parameter estimates (p, q, a , b, c) in a highly accurate manner where two of them, the coefficient of external influence (p) and the coefficient of internal 128 influence of the dis oxathn both (1e) influene mmda qmau Change incoqx: lelta Luniuy endeax influence (q), were estimated over time. This enabled evaluation of the second hypothesis of the dissertation, which suggested faster diffusion rates (reflected in increasing p and q over time) due to globalization. Evidence for increasing p over time for all products in both developed and developing countries were found. The coefficient of internal influence, on the other hand, shows a slight decline trend over time for all products in both developed and developing countries. One important implication of estimating p and q in a time-varying manner is that now we have a better idea of how these parameters change over time. This will enable us to model the behavior of these parameters and incorporate in the model and the analysis explicitly in future research. Limitations and Future Research Limitations of the research in this dissertation provide the opportunity for future research endeavors related to the multinational diffusion of innovations. These are: 0 Although this dissertation addresses an important gap in the multinational diffusion of innovations research by including a wide range of countries (both developed and developing countries that have economic significance in their respective regions and the world), only four products were included in the analyses. There is a definite need to expand the scope of products in the future research of multinational diffusion of innovations. 0 Similarly, there is need to expand the scope of the covariates included in the model. In this dissertation only three macro level globalization drivers, which are also the major indicators of a country’s economic health, were included. In future research, other covariates such as social and cultural factors as well as factors 129 reflecting the level of development of the country (technological infrastructure, education infrastructure, etc.) should be included in the model. In this dissertation, only two of the five model parameters are estimated in a time- varying manner. Another possible improvement via future research would be estimation of all model parameters over time. This would allow better understanding of the dynamics of the diffusion process in different countries. It would also provide the knowledge base to explicitly model the over time behavior of model parameters, which can easily be incorporated to the AKF (C-D) algorithm in the subsequent research projects. 130 APPENDIX A x: Estimated per capita sales 2: Observed per capita sales p: Coefficient of external influence q: Coefficient of internal influence a: Coefficient that shows the increase in diffusion speed resulting from a one percent increase in per capita trade volume b: Coefficient that shows the increase in diffusion speed resulting from a one percent increase in per capita FDI inflows c: Coefficient that shows the increase in diffusion speed resulting from a one percent increase in per capita GDP SSE: Sum of squared errors 131 em: Sooood H mwm noowwooooood H mmm 336 n ”L Knooood u n _ VNmo; n .u mflmmd u o _ vawd H a _ evoNd u u wVNNvod oommvood Smovod 833d 3 wNNNNd mooEd wmvfld ”32.0 S mNthod mmcomood NoN_vo.o NON_vo.o o 3:: .o we? 36 scammed 338.0 a aooofio mvwmcood memmmod memmmod m we mNd own—ed 8386 833.0 m oVNoaod Nooovood choNod eNeoNod \1 _amVNd Smemood manNod manNod 5 v3.35 wNSNood 8886 888.0 0 SONd 55:86 2386 2356 o :2 .o NNomood 3336 wowiod m SmNNd 9 MNSod Nvomood NNvomood m 3 Eve _ommmood vomoSd vomoSd v NONd 333090 20890 wNSNood v mwoONd wvwoNood 85023 30886 m m w 386 ES Sod ma Sod $3 Sod m N336 33386 2386 3380.0 N 2:26 comm Seed emvoood @3386 N :No.m.o mmoNSod own Sod News Sod _ Row M .o _Ovaooood wNoooood wmmNeooood L a m N x u a m N a . $.3— 23» neg—93.95:: 93:30 82> 3:— Ba?» gag—55:5 «:25 23:2 Nogvoooocod u mmm NNNoeoooooood n mmm $985 1 e N $28 1 e _ 3.38 n .e 83 u e A 8883 n e _ 8886 u e mvefd 338.0 Said vooid 2 >926 3586 292.0 292.0 3 omNNNd memNSd aged 32:6 a mNmN_.o vaNod N326 waQd o $526 ~60de M2336 $38.0 w Shame £5an 90de memoro w wNELd mMNNSd NNEvod NNfivod \1 waEd wmnmfiod mmwvnod mmwvnod \t 022d 3308.0 nmaoNod smocNod o _vENd .5356 8836 @0886 o cenomd Semood mom 5o mode m NMZNd 022 So NS mod NS 8.0 m N926 meEood @2306 $2806 v enmeo SVNwood NMKSd NmKSd v ovSNd 08:86 3386 3386 m gem mo Goomood M3306 838.0 m woos _ .o _mmo _ cod 3m ~ cod Nag Sod N :36 vane _ cod 3386 50806 N 33 H .o mwamNoood N. _Noood non _ Noood _ NNmNNd SQ Sod wvo Sod wvoSod _ a a a a d a N x a Cha— Caom gees—.235 not: :80 9:8: 9&3 £3» seen—595:5 ..o a: 3.5 gonna—e0 «Zr—zany: N 132 n2: 22003000000 M 01200 0008000000 u mmw 8.2.01: 2:031: .821... Even: 30931: $0221.. 002 20 00000.0 002020 00200 2 .0 02 00000.0 00N~N0 N202 0.0 N00200 02 000000 00080 .2122 .0 2.2.2 .0 0 2 2N000 032.0 00.00N0 00>0N0 0 0002 20 000200 022 20 022 2.0 0 2.02000 000N000 2NN.2.0 2NN20 0 002VN2.0 NO0N200 NN02000 2N02000 \1 0N02N0 0000N00 20020 2.0020 \1 22.00 2 .0 202.0200 020N0000 0VN000.0 0 30000 Ken 200 00Nv000 00Nv000 0 2.02.0000 002N200 0200000 300200 0 R2N.0 0000200 2N00000 2N00000 0 220000 0000800 NN00.0 20NN000 v 00020 022200 0000200 0000200 2. 200000 02000000 2000200 0000200 0 000020 00000000 0N00000 0 2 N00000 0 0N02v0 03.0000 NV0000 00220000 N NNNVNO 00202000 02800 002N000 N 32.00 03800 0000000 00000000 2 0202000 0028090 02VN0000 02.008000 2 2. .2 N a a 0 .2 u 2. a 2002 £30 5225:0335 anon—flue; 2.002 v.30 528509.322 2.22.2 022.2032 0002000000 M 01200 NNVN000000 n 0200 $8.30 1 e _ N88: 1 .2 fl 8203 u 8 £085 1 e m 03.3 n e 2 39.3 n e 2N02 20 0000000 0 200N0 0 200N0 02 00N2N0 080000 N300 N300 02 N000000 00 0000.0 0.2.NNO 000NNO 0 0N00 2 .0 N02v0o0 000VNO 0000N0 0 20020 0002200 20020 0020 0 002020 0000000 00002.0 00002.0 0 200020 02.0000 0202.0 2.2020 0 200000 05080 0200000 0200000 0 020wN.0 0.30200 000N000 000N000 0 2.0.\.N0 0N00200 0000000 0000000 0 0000N0 00N0200 0VN00.0 002N000 0 002.20 002N200 v0>0N00 002.0N00 0 2 20000 00000000 0000000 0000000 2. 0NN020 2. 2200000 200200 200 200 v 02000 2 23.0000 00080 20080 0 000 2N0 80200.0 20000.0 N0200000 0 000020 $002000 N0 0000.0 02 0 0000.0 N 22230 N000 2 00.0 000N000 0000N000 N 000NNO 000000000 02.1000 0222000 2 02.0200 000000000 20800 02 20000.0 2 0 .2 N 2. 2 0 .2 u 2. a $002 in”; 52.93.3322 08:. :50 2.5: €002 "usemetoaeebuc 20.9.2.2 8202 3.22500 <22<¢Hm22< 2 133 00002000000 H 0200 N00 2 2000000 u mmm 880 n o _ 822 n a _ 88$ u a $38.0 u o 0 N380 n a _ 0035 u .6 200020 0000000 0000N0 0000N0 02 000020 0000000 00002.0 00002.0 02 0 20020 N00000 20NNNO 20NNNO 0 N000N0 N0 0N0.0 0NON2 .0 0NON 2 .0 0 0000 2 .0 00N000 0000 2 .0 0000 2 .0 0 0000 2 .0 0008.0 000 200.0 000 200.0 0 0000N0 000 08.0 00 2 .0 00 2 .0 0 0000 2 .0 0008.0 00000.0 00000.0 0 2002 20 0N00 20.0 2 2002.0 2 2002.0 0 000NNO 00N20000 0002000 0002000 0 0000N0 0000N00 0000000 0000000 0 0NON20 0002 20.0 0020N00 0203.0 0 0000 2 .0 0 2 0080 0000000 0000000 0 N0 2 0N0 00 0000.0 0000200 0000200 0 N000 2 .0 0020200 000080 0000N00 0 20 2 0N0 00 0N0000 0000200 0000200 0 0 2002.0 0002 20.0 00020.0 000200 N 00000.0 NONON000 00000.0 00000000 N 0N00 2 .0 00000000 N000000 0 2000000 2 000 2 0.0 000 0N000 020N000 002 0 N000 2 22 .2 n u u 2. .2 n n a A 0002 ".390 gag—22.8320 225.0 83> A 0002 2.30 532225.550 28.2.2 9.2.2.222 00 00N0000000 u mmm 0 0000000000 n 000 0302 u o _ 0023 n a _ Ego u w. 0808.0 u o _ 0823 u i 383 u w 0 2000.0 0N02N0 00000.0 00000.0 02 000NNO NONN00 N00000 N00000 02 00NON0 00002.0 00000.0 00000.0 0 20N00 02N020 N020.0 N020.0 0 000NNO 0202 2.0 NO0N0 NO0N0 0 000N000 0202000 N002N.0 N002N0 0 0000 20 00000.0 000NNO 000NNO 0 N02020 0000000 20002.0 20002.0 0 0000N0 0020000 0000 2 .0 0000 2 .0 0 2000 2 .0 0002000 0 2000.0 0 2000.0 0 00NON0 0000000 N00020 N00020 0 000020 0000N00 0000000 0000000 0 0000N0 0000N00 0002000 0002000 0 000020 2000200 0000000 0000000 0 000020 0000200 N020000 N020000 0 0002N0 00N00000 2000200 2000200 0 000 2 0.0 00N0 200 0000200 0000200 N 0000N0 00N 2 000.0 N0020000 00020000 N 002000 0.0000000 00N00000 00N00000 2 NNON0 N2NON000 00002000 00002000 2 22 .2 n x 2 2. .2 n u 2 20002 9:30 528.22.2322 58:. :30 2.822 20002 2.30 nozoseoh2aquoza2m 8222 3228.5 «0:22.030. 134 2000N000000 n H00 02 0800000 u H00 8033 u a $586 u i 8035 u 0. 0002.0 n a V 2886 u A_ F $02.0 u 3 200020 0000000 00000.0 00000.0 02 000020 0000200 00N0000 00N0000 02 00202 .0 00N0000 0NO0N0 0NO0N0 0 2 200N0 00000000 0002000 0002000 0 0NN2 20 0000000 0 00NNO 0 00NNO 0 0 0002.0 0 2000000 0NO0N00 0NO0N00 0 000020 020N000 00002.0 00002.0 0 0N02 2.0 00000000 2002N00 00380 0 20002 .0 0000N00 000N2 .0 000N20 0 0N00000 00000000 0000200 0000200 0 200N0 0000200 0000000 0000000 0 N000N0 NON02000 0000200 0000200 0 0000N0 NON0200 N200000 0200000 0 00002.0 00280.0 0020000 N0020000 0 00000.0 2N000000 0000000 0000000 0 02 2000 00002000 00000.0 00000000 0 0N0000 02 0000.0 0200200 0200200 N 000020 000000000 0222000 20222000 N 0 2 0 2 N0 00000000 20N0000 0 2 0N0000 2 2NO0N0 0000 0000.0 0000000 000000000 2 2. .2 N N a 2. .2 N a 2 @0002 £30 case—22.05250 FEE—“U 82> 20002 230 acrosuchflm 2.22.2 2302 0 2002000000 u H00 00002000000 n H00 $0020 u o _ 82 S u a _ 823 u .0 800.8 N o _ 3385 u a _ $8.0 u 0 0000.0 20N000 020N00 020N00 02 0020.0 20N200 N00000 N00000 02 N02000 002N0 00000.0 00000.0 0 N000N0 02002.0 02000.0 02000.0 0 0N0000 0000 2 .0 00020.0 000 2 0.0 0 00000.0 0000000 00000.0 00000.0 0 0020N0 000N000 00000.0 00000.0 0 NONON0 NN02000 000N00 000N00 0 0NO0N0 0000000 00000.0 00000.0 0 0002N0 0000000 020NNO 020NNO 0 000020 0N00000 0000N0 0000N0 0 000NNO 0002000 02N02.0 02N02.0 0 NONON0 0N00000 0N00 2 .0 0N002 .0 0 0000N0 2 200N00 0002000 0002000 0 000020 0202000 0020000 N020000 0 200000 0N00200 0000000 0000000 0 002 20.0 200NN00 N000000 N000000 N 00000.0 00020000 00N0200 00N0 20.0 N 002000 000N200 0000200 0000200 2 2 20000 00000000 N020000 NN020000 2 0 .2 N x a 2. .2 N a a 20002 9:60 accuauctuauouamyou 2.302 00002 :50 509509.252 Lama—.2 8:2 2222500 SD—UHHH 135 N00000000000 H H00 000000000000 n H0m $5.22 M o _ 0.2.08.2 u 3 283.22 u 0 0005.2 u o _ 85.3 n fl2 _ 02000.2 u H. mm: 2... 0002.286 2.288... 028232 2 0.33 88822 $82.22 N08; .22 22003.; 02.022822 38232 00% 2.2.0 0 2.803 0080.222 2 20.2 2 .o 2 20.2 2 .o a 32 222 888.222 £82.; £82.; 0 00022 885.2 0808.2 0308.2 0 0 2.20 2.22 382.232 88.2.2 883 0 882.22 220288.22 $38.0 $38.2 0 8.022 0022.88.22 00008.22 02.2.8.3 0 £02 22 22.02.0820 02.2.0 3.2.2.22 0 RS222 8828.0 02822 2308.2 0 2.8222 0258.222 N883 022208.22 0 $82.22 88 28.0 808.222 828.2 2. 220022 2.02.00.28.22 $028.22 800285 0 82.22 0828.0 088.222 2.88.2.0 m 2.: :2 $388.20 $88.22 £8820 m 08023 Swag: $88.2 $88.22 N 20222 8.2-0880 8.2-80.0 89%:00 N SE0... 082886 228.2822 82885 2 020:2 8.2-0880 8.2-00 8.2-0008... 2 22 .2 N 2. 2 22 22 N 2. 2 200024.290 2202293202252 «.2 2.520 .2022; 200021.243» 22.222922229222222 22.2.2.2 9.2.2.222 2.888886 n 000 03888.0 n 2200 2.0022 n o 2 0022.222 u ,2 0 08022 n .2 N88... n o 2 8202.23 u n2 2 80022 n 0 2.200222 8250225 223222 0222 .22 802.2 835.22 882.2 882.22 .22 a 222.2 0308.2 532.0 2282.22 0 8002.2 0082.3 2308... 230822 a $22.2 $0222.22 882.22 882.2 0 02.02.25 0002.08.22 02003.2 02008.2 0 23.22 $002822 $82.; .833 0 $002 .22 02.2 :3 as 02.22 82 0032 0 232.22 9.: 2.2.2 2023.2 2.28.22 0 808.0 858.0 2.58.22 00 386 o 0:03 82.28.22 $83.22 $825.2 0 288.22 8082222 220232 20$ 2.2.2 0 $00222 8028.222 208.2 2.0822 2. 08.2 £088.22 885.0 88232 2. 0.82 2208.22 $02.3 30.522 0 388 0208.3 00008.2 E00822 m 2.8.20.0 28008.22 2.28.2.2 888.3 N 30222 0028822 2028.222 020283 N 2.20:2 80.2823 8 2.2.; 8028.2 2 082.2 28.2882 2.08822 $288.2 2 22 .2 N 2. 2 22 .2 N 2. 2 20002 ".2230 22022950822222 .52... 8.20 222.222 20002 2.39.2 522222292252 203.2 3.222 22:25.20 122322 136 «‘1 r1.— Ill} .-.-.--‘q ". ‘ It. .II.~A.~ pi..- ..: \.:2N.= «~.V \ III! I . v M N00000000000 H H00 002N200000.0 u H00 N00020 n o 2 0NNN000 u .2 2 N00020 n .2 00202.0 .1. o 2 0000N0 u .2 _ 00N0000 u 0 0NNO 2 .0 0N00 00.0 NNONNO NNONN.0 02 0000N0 20N0000 0NN2 0.0 0NN 2 0 .0 02 22002.0 0000000 00002.0 00002.0 0 000N.0 0N00000 000NNO 000NNO 0 000020 0808.0 0002.0 0002.0 0 00N020 00008.0 00002.0 00002.0 0 0002N0 0000200 00002.0 00002.0 0 0NON00 0000200 0NN2 2.0 0NN2 2.0 0 0000N0 0000200 0000000 0000000 0 00N2 20 00008.0 0N00000 0N00000 0 0200N0 00N0200 N002000 N002000 0 00002.0 00008.0 0000000 0000000 0 0NO0N0 0002 20.0 020N000 020N000 0 00N0.0 N002 000.0 0000N00 0000N00 0 N200N0 20000000 0000200 0000200 0 002NNO 00000000 0000200 0000200 0 N000N0 N 2 N00000 0 2 0000.0 0N 2 0000.0 N 000N00 02 N00000 2000000 2000000 N 000020 20002000 N02 2000 0N02 200.0 2 0NO0N.0 00002000 00NN00.0 00NN00.0 2 22 H N u 2 2.2 .2 N 2. 2 20002 2.2.2.2 22022322022225 2.28.20 82> 20002 ”.592 actuatoflv 22.2.2.2 22.2.22 NON0 2000000 n H00 2NON2000000 n H00 E022 n 22 2088.2 u a 2 a: ...2 u 2 2.2.20.2 u 2 2 888.2 M .2 2 883 u 2 000NNO 00000.0 02000.0 02000.0 02 0N002.0 0002000 0000N0 0000N0 02 202 2N0 0000000 0000N0 0000N0 0 0000N0 0000000 00002.0 00002.0 0 000NNO N002000 002 2N0 002 2N0 0 020020 0200000 00002.0 00002.0 0 0 2 00 2 .0 N020000 0002.0 000 2 .0 0 000 2 N0 0000N00 00000.0 00000.0 0 0N00 2 .0 0000N0.0 0002 2.0 0002 2.0 0 002NNO 0000200 0020000 0020000 0 02N0 0000200 00N200.0 00N200.0 0 0N000.0 0200200 00N0200 00N0200 0 000N.0 0200200 0002000 0002000 0 200200 000N2000 0000000 00000000 0 00N0 .0 N0000000 N000N00 N000N00 0 2 2N0 00000000 000 2000 00002000 0 00000.0 00000000 000N200 000N200 N 0N002.0 0088000 0200000 200200000 N 0002N0 N0 20N000 NON0000 0NON0000 2 00N020 0000N0000 0000000 000000000 2 22 .2 N 2. 2 22 .2 N 2. 2 2N002 9:30 5222.22.25.25 2.22:. 222.20 222.222 2N002 ”2.3.2 =0222=22022=2mowmm 3.202 2222.22.20 0.0—0.20.0 137 000NN000000 n 0200 020N2000000 u H00 $02.3”; 0228.722 2.2.2.2 nu 02.203”: 2.22.2”: .2202“... 0000000 000N000 N000N00 0000N00 02 N200N0 0000000 00002.0 00002.0 02 0000000 000N0000 N00 0N00 200 0N00 0 0 2 00 2 .0 0008.0 0 00000.0 0000000 0 00N0000 0 2N0 000.0 0808.0 0N00 20.0 0 0N20N0 0000200 0000000 0000000 0 2000000 0020N000 N000200 2000200 0 000020 0000200 0000000 0000000 0 000NNO 000N000 20N2 20.0 NON2 200 0 000NNO 0N000000 000 0N00 00 00N0.0 0 000 2 0.0 02 0080.0 0000000 00000000 0 00002 .0 00 2 0000.0 0NNO 200 0NNO 2 0.0 0 N020N0 0N0000000 0 2 0000.0 N0200000 0 00000.0 00 0080.0 0000000 00000000 0 000020 000000000 0020000 00020000 0 00000.0 0N002000 020000 00020000 0 002000 2 20000000 N00 2000 2 2002000 N 02NON0 000200000 0002000 00002000 N 00020 000000000 0000000 000000000 2 00N0.0 000000000 0000000 00N000000 2 22 .2 N 2. 2 22 .2 N 2. 2 20002 :30 :022295002§_==.2U|?.2222> 20002 ".2230 5222329222222 22.2.2.2 22.2.22 2002000000 u H00 00N200000000 n H00 020N0n22 0020N0u.22 N02000n0 0002.0”: 0000N.0n.2 fl 00002.0u22 002020 0000000 002NNO 002NNO 02 0N00000 0000200 02000.0 02000.0 02 00002 .0 20N0200 N200 2 .0 N200 2 .0 0 0002N0 00N00000 NN00000 NN0000.0 0 0000000 200000 NON020 NON020 0 0000N0 00020000 0000.0 0000.0 0 000020 0000200 200020 200020 0 0200N0 0N00200 0000000 0000000 0 00002.0 00020.0 N00000.0 N000000 0 000200 00000000 0000N00 0000N00 0 0N202.0 0000200 2000000 2000000 0 000N.0 00000000 0002 20.0 N002 20.0 0 000020 02000000 0200000 N200000 0 0000N0 N0202000 0020000 0020000 0 000NNO 00 00000.0 2000200 2000200 0 00NON0 000000000 00N000 50080.0 0 00020.0 000800 0200000 00200000 N 000020 000200000 0202000 00202000 N 2000000 000N000 2 20N00.0 20N000 2 0000000 000000000 0N000.0 N000N0000 2 22 .2 N 2. 2 22 .2 N 2. 2 20002 230 52292220222222 .52... 222.20 2.2.22.2 20002 "222.201.3223220552 5.32.2 8222 22.2.2500 E 138 N000N0000000 H H00 00002000000 n 000 0.2003 n o 2 2.03.2.2 u .20 2.2.22.2 u u 3.52.2.2 n o 2 32.3 n 2 0 $0.22.; n 2 200020 0200N0000 02 2000 02 2000 02 NON00 0000200 0000200 0000200 02 00NN2.0 000020000 0000000 000000000 0 0200N.0 00000000 0000000 00000000 0 0000000 000000000 0000000 00N000000 0 N00 2 .0 20000000 000N000 0000N000 0 00000.0 202000000 0200000 N20200000 0 000N20 00002000 0000000 000000000 0 NN02 .0 0000N0000 0000000 000000000 0 0000000 000000000 2NN0000 00 2NN0000 0 0 0002.0 000020000 0000000 00N000000 0 N000000 000N20000 000-0000 000-002 00.0 0 02 2 0N0 000-00 2 N0.N 0N000.0 0000N0000 0 0000000 000-00 200. 0 000-0N0.N 000-00000.N 0 0000000 0002N0000 N020000 2 2N020000 0 2002.0 000-00020.N 000-o0N.2 000-0020N.2 0 000200 000-00000.N 0020000 000020000 N 000200 000-o0N20.0 000-0200 000-0000N.0 N 2000N0 220020000 000-0000 000800200 2 0002 20 000-0002 0.0 000-0020 000-00000 2 22 .2 N 2. 2 22 .2 N 2. 2 20002 v.80 2202223202225 «8.2.2.0 8222> 20002 "28.2 =o22o=22.2..2§.2.2.2.2 9220.222 00022000000 n 0200 0N0000000000 n 0200 2000200 n o 2 0088.0 n .2 2 0000200 U .22 0000000 n o f 0000000 U .22 N00020 u .u 00N02.0 00002.0 00N02.0 00N02.0 02 000020 00000000 0NNN20.0 0NNN20.0 02 0002 2.0 002N20 0002 2.0 0002 2.0 0 002020 0NN20000 0NN00000 0NN00000 0 000NN.0 0000000 N000000 N000000 0 0 002.0 00N0000 0N000000 0N000000 0 0NO0N0 N000000 002 00.0 002000 0 00N2 2.0 00000000 00000000 00000000 0 0000000 0000000 0000000 0000000 0 N000 2 .0 0000N000 0N020000 0N020000 0 0000000 0000000 00N020.0 0N0200 0 020020 00N20000 N000N000 0000N00.0 0 2002N0 000N2000 000000000 200000000 0 00000.0 000000000 00002000 00002000 0 2000 2 .0 N00000000 220000000 000 00000.0 0 0 000N.0 00002000 00000000 200000000 0 N000 2 .0 000000000 0088000 005880 N 00000.0 000 2 0000.0 000N0000 0000N0000 N 00N000 000020000 000000000 000000000 2 020000 200NN000.0 000000000 N20000000 2 22 .2 N 2. 2 22 .2 N 2. 2 22002 2.2.3.2 82222229222222 .82—2.25.20 222.222 @002 9.80 22.228.225.552 8.22.2.2 3.222 23.28.20 «0722220 139 0N002000000 u H00 00002000000 n H00 503 n o 2 0.88.2 u .2 2 8.23.2 u 2 08:..2 u o 2 0.522 n .2 2 23.0.2 n 2 2 2002.0 00000.0 00000.0 00000.0 02 0000N0 0000000 0N000.0 0N000.0 02 00002.0 00008.0 N02 200 N02 200 0 000020 0000000 0NNON0 0NNON0 0 N00020 0000000 N000N.0 N000N.0 0 000N20 0000000 02002.0 02002.0 0 0000 2 .0 N000000 000020 00002 .0 0 0000N0 0000200 0NN02 .0 0NN02 .0 0 020020 8080 02002.0 02002.0 0 00N02.0 0200200 00002.0 00002.0 0 0000N0 0000200 0N2020 0N2020 0 0000N0 20N00000 0NN0000 0NN0000 0 000020 0000200 00000.0 00000.0 0 0N002.0 000200 0000000 00000.0 0 0000N0 N2020000 0000000 0000000 0 0002N0 00000000 N0000 N0000 0 000N00 20000000 0000200 0000200 N 00000.0 00000000 0000N00 0000N00 N 0000 2 .0 0N000000 0000000 00000000 2 002 2N0 NN 2 0000.0 0000000 00000000 2 22 .2 N 2. 2 22 .2 N 2. 2 20002 .280 82232202252 2.22.50 8222> 0002 28.2 822222202252 22.2.2.2 22.2.22 00002000000 n 0200 N2002000000 u 000 .0003 u o 2 0.0286 n .2 2 20.03 H 2 20023 n o 2 2.00.2.2 u .2 2 22 .02 n .2 002020 N00020 2N2000 NN2000 02 2 20000 000200 00200 00200 02 0202 0.0 0000000 0200N.0 0200N.0 0 2000.0 2020N0 00N000 00N000 0 N000N.0 000 2 00.0 000N.0 000N.0 0 00000.0 0000 20 00000.0 00000.0 0 00N02.0 20NON00 00002.0 00002.0 0 0020N0 200020 0000N.0 0000N0 0 200NN.0 0000200 00002.0 00002.0 0 002 20.0 00N000 00002 .0 000020 0 000020 2020200 000N000 000N000 0 00N02.0 0000000 020N000 N20N000 0 0N20N0 0000200 0200000 0200000 0 N00020 00000000 0000000 0000000 0 0200N.0 00000000 0NO0N00 0NO0N00 0 00000.0 00000000 02N020.0 02N020.0 0 000020 00000000 N002 20.0 N002 20.0 N 000N.0 00000000 00000.0 00000000 N NON2 0 .0 2000N000 000N000 08080.0 2 0000N0 0002000 000N000 00080.0 2 22 .2 N 2. 2 22 .2 N 2. 2 20002 :80 8222228252 .82... 2.2.20 222.222 20002 “.822 22.228.200.252 8.2.22.2 3.222 22:28.5 v2~2<2>272H22 140 0 200N000000.0 n H00 0000N00000000 n H00 0800.702 0.6.7.22 .2203”... 08.8.2”: :82 "22 Eamon. 0000000 000N200 0000000 0000000 02 0000000 000020000 0000000 0000000 02 00000.0 00080.0 0200000 0200000 0 0020000 200020000 N000000 N000000 0 0002000 0N000000 20NON00 20NON00 0 000020 200 2 0000.0 0000000 00N000000 0 0202 2.0 0N0 2N000 00NON00 00NON00 0 0000000 000-002 20.0 00000.0 00000.0 0 0000 20 0000200 0000200 0000200 0 0002.0 002N20000 0000000 000000000 0 00000.0 00NN0000 00N2 20.0 00N2 20.0 0 2002000 000-0N0N.0 N000000 002000000 0 0000N0 00000000 0020000 00020000 0 0NNO 000000000 0088.0 0088.0 0 20NON0 00 6800 0000000 20000000 0 00 2 000.0 00NON0000 0020000 0020000 0 020.0 00002000 0002000 00002000 N 00000.0 000-00000.0 0220000 02 20000 N 0000N0 0N0 0 2 00.0 0 2 0000.0 000 2 0000.0 2 0000N00 000-0000 2 .0 000-0000 000000000 2 22 .2 N 2. 2 2. .2 N 2. 2 20002 "28.2 822328.250 0.. 2.5.0 8222> 20002 "00001822220922.222 0.5.2.2 0.2.2.252 00008000000 u 000 000NN00000000 n 0200 822-02 0.2.22.2": 20.0.2“... 0:222”: 0.3.2.2”: S... n... 0000 20 20000000 000NN00 000NN00 02 0000000 2NN2 000.0 0 N00200 0N00200 02 0002000 000NN000 02 20200 0220200 0 020N000 00000000 0020200 0020200 0 000N2.0 00020000 N000200 N000200 0 2000000 00020000 NO0N200 NO0N200 0 02 2N2.0 N0002000 0002 20.0 0002 20.0 0 00000.0 0002N000 2000000 2000000 0 02N00 00000000 0200000 0200000 0 00N02.0 00000000 0000000 0000000 0 0000N0 20000000 2 200000 2 200000 0 002020 00000000 0000000 0000000 0 0000 2 .0 0000800 2000000 00000000 0 00N000 0 00000.0 NO0N000 NO0N000 0 0N00 2 .0 NNON 200.0 2NON000 N 2 NON000 0 0000N.0 0000800 000 200.0 000 200.0 0 02 202.0 00N02000 0N2 200.0 0N2 200.0 N 0000N0 00002000 0000000 0000000 N 200000 NO0N00000 00N0000 0000N0000 2 0020000 000N20000 0020000 0020000 2 22 .2 N 2. 2 22 .2 N 2. 2 20002 2.200 8220322252 ..02... :50 0.822 20002 0.000 822058.252 Sam—.2 00.222 200.2800 0.2.000— 141 200N00000000 n mmm 000200000000 H mmm @003 u o _ 0002.; u p _ 8003 u a 8.080 n o _ 80080 n a 2 020$ n .0 000020 2020000 020020 020020 02 200020 0000N00 0NON20 0NON20 02 0000000 0000000 002 2 20 002 2 20 0 00020 0000200 02 0000.0 0200000 0 0000 2 .0 0 2 NN00 0020000 0020000 0 N000000 000 2 0.0 080000 0N00000 0 0N0 2 .0 0200 200 0000000 0000000 0 0000000 0000000 000 200.0 000 2 00.0 0 0000 2 .0 0000200 N000000 2000000 0 00N0 2 .0 0000200 0022000 0022000 0 000020 2NON200 0N2 2000 0N2 200.0 0 000020 20200 0000000 0000000 0 NON000 58000 0000200 0000200 0 N000N.0 200N0000 0000N00 0000N00 0 0000N0 00000000 0000000 0000000 0 000 2 .0 N0000000 2000 20.0 2000 20.0 0 00 2 0 2 .0 00000000 0000000 00000000 N 0000N0 00NN0000 0000200 0000200 N 0000N0 200000000 0000000 000000000 2 00000.0 000N0000 2N00000 N2N00000 2 22 .2 N 2. 2 22 .2 N 2. 2 @002 ".23.. 52222220353800 32> A0002 ".2430 5222.25.22.25 22.2.2.2 922.2022 N0002000000 n 0200 02 2000000000 n 0200 .320": 3000.5”: 08.270 :82": 0.031: Egaouw 0020N0 N000000 00000.0 00000.0 02 N000N.0 2N0020 00N000 00N000 02 0NONNO 0000000 N000N.0 N000N.0 0 02NNO 000020 0000N0 0000N0 0 N002 20 0000000 0000N0 0000N0 0 000NN.0 0NON000 N00020 N00020 0 000020 0000000 02 2 020 02 2020 0 00000.0 0000000 N02020 N02020 0 0N2N0 00N0000 200020 200020 0 0000N0 0000N00 0000000 0000000 0 0000 2.0 0N00000 00000.0 00000.0 0 N000N.0 200NN00 0000000 0000000 0 0000 2 .0 00 2 N00 00N0000 NON0000 0 0N000.0 N202 200 00N0000 00N0000 0 00000.0 0000200 0200N00 0200N00 0 00020 0020200 2000200 2000200 0 020000 00000000 N000 200 N000 200 N 0000N0 N0000000 20N0000 20N0000 N NON20 00000000 00N0000 000N0000 2 00N000 00N000000 0000000 000000000 2 22 .2 N 2. 2 22 .2 N 2. 2 20002 ".230 3229:2233: .32... 8.20 222.222 20002 ".890 =c229=220~2u2v Loam—.2 902: 2922.25.20 0272012722..— 142 000000000000 n 0200 000000000000 n 0200 838.0 M o _ 368.0 N .2 _ 03.3 n 0 0.2030 n 00 22.0 n .2 _ 00003 n .0 000020 0000200 00000.0 00000.0 02 N00020 0N02 200 0200000 0 2 0000.0 02 NN000 0000200 2N00000 2N00000 0 000020 000N0000 0000 00.0 0 000000 0 0000 2 .0 00000000 2N0 2000 2N02000 0 000NN.0 00000000 N000N00 N000N00 0 200020 00000000 02N0000 02N0000 0 0N2020 00000000 0000200 0000200 0 002020 00020000 0020000 0020000 0 0NO0N0 N0000000 0000200 0000200 0 000NN.0 N2N00000 000NN00 000NN00 0 020N00 00002000 2000000 00000000 0 0002N0 00000.0 0000200 0000200 0 N00020 00202000 000N000 00080.0 0 2000N0 02N00000 08000.0 08000.0 0 NON000 000000000 0000000 000000000 0 000020 0000N000 200N000 2 200N000 N 0000N0 000020000 0 0N000.0 2NO0N0000 N 0000N0 N2N000000 20N0000 0000N0000 2 2000N0 000600000 000-20N0 000-0000N.0 2 22 .2 N 2. 2 22 .2 N 2. 2 300a: 2202293292252 «22.2.20 0.2—22> 20002 2.243» 22.222922229222222 22.2.2.2 22.2.22 0N00 2000000 n 0200 00002000000 n 0200 .823”: 0808.0”: 20.200”. 0052": ezgui 0028". N20000 0000000 02000.0 02000.0 02 000N2.0 0N002.0 20N000 20N000 02 0000N0 00002.0 00000.0 00000.0 0 00N02.0 N00020 00000.0 00000.0 0 000NN.0 200N000 NONON0 NONON0 0 0000N0 0020N00 N000N.0 N000N.0 0 000020 0000000 NO0NNO NO0NNO 0 0000N0 2200000 002 2N0 002 2N0 0 0000N0 0000N00 N20020 N20020 0 0002N0 200N000 000020 000020 0 0002N0 2000000 002 2.0 002 2.0 0 00N02.0 0020000 000020 000020 0 000020 0000N00 0000000 0000000 0 N002 20 00008.0 0200000 0200000 0 N000 2 .0 N000N00 2 0000.0 2 0000.0 0 00000.0 NON00000 000 0 00.0 N00 000.0 0 000N00 200N200 00008.0 000 0N00 N NN0000 00000000 N020.0 N0 200 N 20 000.0 0N 2 0000.0 0 00000.0 0N000000 2 0000N0 002 2N000 N02000 00 202000 2 22 .2 N 2. 2 22 .2 N 2. 2 20002 :30 522922292252 .52... :50 222.222 20002 ”.2230 522932222222 Log 8222 2223 02207203222 143 02 202000000 n 0200 22N200000000 n 0200 2:03 u i :82; u 02 000mg u w 0023 u o _ 0.88... n 2 000.8 n 2 0000000 0NO0N00 00002.0 00002.0 02 000200 200NN00 NO0N20 NO0N2 .0 02 0002 20 0020N00 000020 000020 0 0000N0 0200200 0022000 0022000 0 NO0N20 0NO0N00 NO0N20 NO0N20 0 200N20 00N200 0000000 0000000 0 000020 200NN00 0000000 0000000 0 20NON0 0N000000 00N0.0 00N0.0 0 020000 0002 200 N000000 N000000 0 0 20NNO 0N00000 2000200 2000200 0 000020 N200000 0000000 0000000 0 0000N0 0000N000 2N00000 2 2N00000 0 000020 00N20000 0000N00 N000N00 0 000N00 000200000 0000000 N0000000 0 002000 00000000 000N200 000N200 0 00 20 2 .0 02022000 000N000 000N000 0 000 2 00 NNO0N000 N000000 2N000000 N 000 2N0 N00 2 0000.0 002 200.0 0N02 2000 N 0NO0N0 00002000 000 2000 0000 2 00.0 2 N00 2 .0 000000000 0000000 0 00000000 2 22 .2 N 2. 2 22 .2 N 2. 2 20002 ".230 526320222222 «.8800 .2922; 20002 2.22292 9.222252292222I22 22.2.2.2 0.2.2.222 0000 2000000 n 0200 000000000000 n 0200 00020": 00N200.0”.2_ 0002N0n0 000000”: 000220”: 220000”.2 0N002.0 0N002.0 2N0000 2N0000 02 0NO0N0 000N000 00000.0 00000.0 0 002N0 0NN2000 0000N0 0000N0 0 00N000 00200.0 00000.0 00000.0 0 20N020 0020000 02 2N0 02 2N0 0 0N2NNO 02002 .0 0N02 0.0 0N0200 0 0000N0 000N00 0N002.0 0N002.0 0 N02020 00020 00000.0 00000.0 0 2 2002.0 000 2 00.0 2000000 2000000 0 002020 0202000 0020N0 0020N0 0 0000N0 00N0.0 200000 2200000 0 002000 0000000 200020 20002 .0 0 200000 0020 200 0NO0N00 0NO0N00 0 00000.0 0000200 0000000 0000000 0 N000N.0 00N00000 0000000 00000000 N N000N.0 2000N00 N200N00 N200N00 N 2N0 00.0 000N0000 00N0000 2 000N0000 2 00000.0 000000000 2000000 000000000 2 22 .2 N 2. 2 22 02 N u 2 22002 ".230 5222.28.22.25 .52... 8.20 222.222 20002 2.2.3.2 3229522222222 29202.2 3.222 222mg. >72<2>2~2m20 200000000000 n 0200 0000800000 n 0200 02.22 H o 2 00023 n .2 _ 08022 n 2 0.82 n o _ 00030 n n _ 0.022 n 2 0000000 0NO0N00 00000.0 00000.0 02 N00020 0020N0 00000.0 00000.0 02 000020 N000200 02N0000 02N0000 0 0N000.0 20N0000 00202.0 00202.0 0 0000000 000N200 0N000.0 0N000.0 0 0000N0 0000000 00N02.0 00N02.0 0 0000000 2200200 0000000 0000000 0 0200 2 .0 0000000 0002000 0002000 0 00NNNO 00020000 02 2N000 02 2N000 0 20002 .0 00060002 0200N00 N200N00 0 0202N0 0002 200 002NN00 00 2NN00 0 000020 NO0N0000 0200 200 0200200 0 0NONNO 00000000 0N00 200 0N00 200 0 000020 0000N000 0000000 20000000 0 2N02 0.0 0000N000 NON0000 NON0000 0 000N.0 00202000 00000.0 00000.0 0 0000N0 00020000 0800 08080.0 N N00000 000N00000 0002000 00002000 N 2000N0 000000000 00 0000.0 000 00000.0 2 0 20000 000000000 0000000 00000000 2 22 .2 N 2. 2 22 .2 N 2. 2 20002 ”..Iawh 5222;202:222 «8.5.0 82> 20002 ".28.. 52222328252 2.0.2.2 922.2052 00000000000 n 0200 000000000000 n 0200 820.5“: .2086": 000N2.0". $0.23“: 80022“: 0.0022”... 000020 0000000 0002 2.0 0002 20 02 00008.0 000-0200N0 00002.0 00002.0 02 000080 000-000N00 2000000 0000.0 0 N000 2 .0 0000000 N000 2 .0 N00020 0 000NN.0 0N20200 0000000 0000000 0 000N00 N000N00 0N02 2.0 0N02 20 0 0000N0 000N200 N000000 N000000 0 2000N0 0000N00 00N0000 00N0000 0 0000N0 22000000 000 2 00.0 000 2 00.0 0 N020N0 0000200 0000000 0000000 0 2020N0 00000000 0000200 0000200 0 0020N0 0002200 02NON00 02NON00 0 000N.0 N0000000 0N02 200 0N02 20.0 0 0000N0 0000200 0000200 0000200 0 N000N.0 N0000000 20N0000 000N0000 0 020N0 00N00000 N000000 02000000 0 00N0 0N20N000 00N0000 020N0000 N 00000.0 00 202000 000N000 2000N000 N 00020 00002000 0000000 002000000 2 00N2N0 0N002000 0000000 N00000000 2 22 .2 N 2. 2 22 .2 N 2. 2 20002 ”.28.. 8222222822222 82:. 8.20 2.2.22.2 20002 ".28.. 82225228252 82222.2 82.2 28.23. HUHHMO 145 00002000000 n mmm 0200800000 H mwm 020000 N o _ 0NO0N0 u .2 _ 002020 n a 0000N0 u o f 000020 N .2 _ 0000000 U 0 0N002.0 0000000 0000N0 0000N0 02 0000N0 0000000 00000.0 00000.0 02 0002 20 8008.0 0002N0 0002N0 0 002 0N0 2000000 0000N0 0000N0 0 02 2020 00008.0 00002.0 00002.0 0 . N02N0 NN20N00 200020 200020 0 000NN.0 NN0200 020020 020020 0 020020 00NON00 000N2.0 000N2.0 0 002 2.0 0N000000 2000000 2000000 0 200020 2000200 00N0000 00N0000 0 0N2000 20000000 0000000 0000000 0 0 2N020 000200 0000000 0000000 0 N000N.0 0000200 00000.0 00000.0 0 00NON0 00200000 0002 00.0 0002 00.0 0 0000N0 00000000 00008.0 00008.0 0 NNO0N0 00000000 0000200 0000200 0 0000N0 00000000 00NN200 00NN200 N 00000.0 0000N000 0000000 N0000000 N 00000.0 00200000 0000000 N0000000 2 000020 00022000 00 2000 00002000 2 .2 N 2. 2 22 .2 N 2. 2 22002 ".59. 52222222250 2.2.520 82> 20002 :8 82222228250 2.22.2 32.2.22 0N002000000 n 0200 02002000000 n 0200 0. 20.0 n o _ 00200.5 u p 2 9.88.0 u a 0.2023 n o _ 0:832 n .2 _ $9.3 n u. 20N020 2020000 200N0 200N0 02 0N00000 0N2000 00000.0 00000.0 02 000020 0000200 02N0 02N0 0 000020 0000000 0000N0 0000N0 0 000020 0000200 00202.0 00202.0 0 002N0 0000N00 0NONNO 0NONNO 0 0002 2.0 0N00200 00N02.0 00N02.0 0 00N20 0N00000 00002.0 00002.0 0 0000000 00080 0200000 0200000 0 0NON00 0000200 0NON20 0NON20 0 N20NNO 002 2 20.0 0N00000 0N00000 0 NONON0 000 200 00N0000 00800.0 0 N 2 00 20 N2 20200 0000000 0000000 0 00N2N0 020N200 0N00000 0N00000 0 2 00000 00000000 2NO0N00 2NO0N00 0 0N002 .0 2000000 00008.0 00008.0 0 00NNO 0NN00000 0000200 0000200 N 0000N0 000N0000 0000000 00000000 N 00000.0 002 2N000 0000000 00000000 2 000 2 20 N0000000 2000000 2000000 2 22 .2 N 2. 2 22 .2 N 2. 2 20002 28.. 52220328252 .82... 5.20 2.822 20002 9.80 52222228252 8.2.22.2 8202 222.23 @202 0720: 146 20000000000 H 0.200 000000000000 n Mmm 00000.0“: 00:20.0”.27 20N020nu. 00000.0": 00N02.0“: 00202.0n0 202 00.0 00000000 0000000 0000000 02 0N000.0 000N2.0 0000N0 0000N0 02 N200000 000N0000 0000000 0000000 0 200N0 000N000 N00020 N00020 0 0000000 20000000 0000N00 0000N00 0 00N200 0020000 0N00000 0N00000 0 0000000 NO0N0000 000NN00 000NN00 0 0000N0 02 0N000 N200000 N200000 0 00000.0 00000000 0000200 0000200 0 0000N0 20000000 0002N00 0002N00 0 002 2 20 00000000 00020.0 00020.0 0 0200N.0 20000000 N000000 0N000000 0 N000N.0 0000N000 N000200 2000200 0 000000 00002000 2000000 2000000 0 0000N0 00000000 0000000 0N000000 0 0002 2.0 00002000 0022000 00022000 0 2NO0N0 0002000 0020000 02020000 N 00000.0 0NON20000 0000000 00N000000 N 000N000 00002000 0N2 2000 00N2 200.0 2 0020N00 000020000 0220000 000220000 2 22 .2 N 2. 2 22 .2 N 2. 2 00002 228.2 5222328252 «8.2.2.0 82> 00002 ".28.. ...2222328250 94822.2 0.2.2.222 N20200000000 n 0200 0022 2000000 U 0200 00000.0".24 0002N00u.2_ 200200u22 00000.0uo_ 000N200n.2_ 0200N.0u22 000N2.0 0000200 00002.0 00002.0 02 000020 000N000 02002.0 02002.0 02 0000000 00NON00 0NON20 0NON2 .0 0 NNN020 0000000 NN2 2.0 NN2 20 0 N020000 00080 000020 000020 0 00000.0 NN0200 NN00000 NN00000 0 0020N0 2000200 002000 002000 0 0000N0 0N00200 000N000 000N000 0 000020 00080 0000000 0000000 0 N00000 200N200 0020000 0020000 0 0000N0 N200200 N000000 N000000 0 0020N0 N20N0000 0000200 0000200 0 0200N.0 00000000 0200200 0200200 0 020NNO 0N000000 0000000 00000000 0 000N00 00000000 0020200 0020200 0 0000N0 000N000 0000000 00000000 0 N20N00 00N2N000 00N0000 000N0000 N 00NON0 0N002000 00 2000 00 200.0 N N000N.0 00002000 0002000 00002000 2 00N000 000000000 0N00000 000N00000 2 22 .2 N 2. 2 22 .H N 2. 2 20002 ".28.. 8229328250 .82... 8.20 2.2.222 20002 ".28.. 8222—2228250 8.0.22.2 8222 22.05.20 >~2 20002 2.2.8.2 522022228252 22.2.2.2 022.2052 0000N00000000 n 0200 2000N0000000 u mmm 50080 n o 2 02.... n a _ 20008.2 u 2 2.3 u o _ 0.0.8.0 n .2 _ 0N00-:2 u 2 020000 0000N000 00200000 00200000 02 02 0000.0 00000000 00000000 00000000 0 00020 00002000 00200000 00200000 0 00 2 0 2 .0 2NO0N000 00000000 00000000 0 00002 .0 N000N000 000N0000 000N0000 0 0000 2.0 200N000 00200000 00200000 0 000N.0 000800 02 2N0000 02 2N0000 0 0N20N0 00N02000 00000000 00000.0 0 2 2 2000 00002000 0000N000 00080.0 0 200N00 00002000 020NN000 02 0NN000 0 2202 20 00000000 00002000 00002000 0 0000N0 00002000 22002000 2002000 0 N000N.0 000000000 N000 2 00.0 N000 2 00.0 0 00 2 NNO 00002000 000N00000 000N00000 0 002 020 000000000 200000000 N00000000 N 0 20020 000000000 N000N0000 2000N0000 N 0000N0 00 0 0000.0 000020000 000020000 2 00N0 2 .0 000 2 N0000 000-00 200.0 000-0000 0 .0 2 22 .2 N 2. 2 22 .2 N 2. 2 20002 "28.2 8220328252 202—2225.20 0.22.222 20002 ".280 22220520222222 8.2.22.2 82: 2022.25.20 «2.202722 148 220000000000 n 0200 N002000000000 n mmm 0000000”: 00002 n: 0000N0u.2 0002.0n0_ 2000200u.2 _ 0020200n22 NN02000 N002N0000 000N00000 000N00000 02 0202N0 00002000 000N000 000N000 02 0000000 00020000 00000000 00000000 0 0000 2 .0 200000000 000 2 00.0 000 2 00.0 0 0000 2 .0 0N00 2 000.0 000000000 000000000 0 N00000 2 N0000000 0000000 0000000 0 00002 .0 000-02 200 220000000 000000000 0 0000 2 .0 0N0000000 0200000 000200000 0 0000000 000-0N02 .0 00 2 00000.0 00 2000000 0 000NN.0 200020000 0N00000 000N00000 0 0002 .0 000-00000.0 0NO0N0000 0NO0N0000 0 N 200 2 .0 000020000 0N000.0 5008000 0 000020 000000000 002020000 202020000 0 002020 00020000 020000 020000 0 0000.0 N00020000 200N20000 0NON20000 0 0202 0.0 000000020 0220000 000220000 0 0002 2.0 000-0000N0 000-000N00 000-000N00 N N00200 000-000000 000-0000 000000000 N 0000N0 000020000 000-00000.N 000-00000.N 2 N0 2 0N0 000000000 000-000.N 000-00000.N 2 22 .2 N 2. 2 22 .2 N 2. 2 20002 "Ha- ...2220202282:I22‘a.20800 82> €002 0.30.2 5220522824222 0.20.2.2 0.2.2.222 00002000 n mmm N002 2000000 n 0200 80.0.. n o 2 82.2 n .0 2 0002.0 u 2 0.02830 u o 0 2285.0 u .0 _ 2.02028 n 2 00N0000 0000N000 0002200 0002200 02 000-02 000NN00 0000000 00000000 0 0200000 00200000 00000000 00000000 0 000-02 000000000 2020000 00020000 0 NNO 2 2.0 0000N000 00000000 00000000 0 000-02 000-02 N000000 20000000 0 N000000 000N0000 00000000 00000000 0 000-02 000-02 N000000 20020000 0 00N0 2 .0 0002N000 02200000 02200000 0 000-02 0000200 N000000 00000000 0 0000 2.0 0000N000 00020000 00020000 0 N00020 000-02 002N000 0002N000 0 0000000 000N000 000N000 0000N000 0 0000000 000-02 00N200.0 000N2000 0 0000N0 00002000 N20N2000 020N2000 0 00000.0 0 00-00000.N 0000000 N0 N00000 N 00000.0 00002000 NN0000000 000000000 N 00000.0 0N0020000 0020000 000020000 2 200020 N00000000 00N020000 002020000 2 22 .2 N 2. 2 22 .2 N n 2 2N002 .08.. 5220—2228252 002:. 8.20 05.222 20002 28.2 5222—2282:: 20002.2 00222 2922.28.20 0200272002722 149 0NON00000000 .I- mmm 000000000000 H 0200 22228.0 n o 2 02008.00 u .02 20000.0 u 2 20.2.8.0 u o 2 2088.0 u .02 80.8.0 u 2 200020 0022200 2000000 2000000 02 2 2002.0 0000N00 0002000 0002000 02 00N02.0 020N200 00000.0 00000.0 0 N000N.0 N00200 00000.0 00000.0 0 2NN020 2202200 0NN0000 0NN0000 0 000N.0 00000000 N000000 N000000 0 2NN20 0002200 2000000 2000000 0 00N02.0 0N000000 0NNON00 0NNON00 0 0000000 0000200 00N000 00N000 0 000020 00N0000 02020.0 02020.0 0 000020 00000000 000N00 000N00 0 020020 02000000 N202200 N202200 0 00 2 0N0 00000000 00N0 200 00N0200 0 0N00000 0020N000 0220000 00220000 0 0000N0 0 2000000 0000000 00000000 0 00N0 2 .0 00000000 0 2 0000.0 00 2 0000.0 0 0002 0.0 20000000 0000000 00000000 N N000N.0 20020000 NN00000 0 2N00000 N 000020 300800 0002000 00002000 2 0N002.0 000000000 000-000.N 000-00000.N 2 22 .2 N u 2 22 .2 N 2. 2 20002 9.82 .202205228252 8.28.20 82> 20002 28.2 52.20.222.2MHW 0.8.2.2 022.202 002000000000 n 0200 000000000000 n 0200 N000000 n 0 2 0000N00 n .2 2 200N0 n a. NO0N000 n o _ 0200000 n .2 2 N020000 u 0 000N2.0 0000000 2N000 2N000 02 0000N0 0002000 0002.0 0002.0 02 000020 0000000 000N.0 000N.0 0 0000N0 0000N00 0000000 0000000 0 000020 0000000 0000N0 0000N0 0 00NNNO 2000200 0000000 0000000 0 2000N0 0000N00 000020 000020 0 000020 2020200 0000000 0000000 0 2000N0 2N00200 02N20 02N20 0 00NON0 200N200 200NN00 200NN00 0 0 2 0 2N0 0002 00.0 0000000 0000000 0 N000N.0 00000000 2002200 2002200 0 00000.0 N020200 0000000 0000000 0 N000N.0 0 2000000 02000000 2N000000 0 000N.0 02 20200 2000000 0000.0 0 0000N0 00000000 0000N000 0000N000 0 0NO0N0 2020200 0NN0200 0NN0200 N 0002N0 000020000 000020000 000020000 N 0000N0 00000000 00000000 00000000 2 0000000 000020000 000002000 000-0N00.0 2 22 .2 N 2. 2 22 .2 N x 2 20002 :82 82220228252 .02... 8.20 08.222 20002 0282 82222228252 8.2.22.2 00222 200. 8.20 027221222222 150 000N00000000 H 0200 200N2000000 n mmm 0200N.0-o2 20.2.0702 00.22020”2 8208.702 00000.2”.02 22.882 0000 2 .0 N0200000 0000000 0000000 02 0N00000 0000200 0020000 0020000 0 0000N0 20000.0 000200 000200 0 0020 0020000 0000000 0000000 0 00000.0 02002.0 00NN00 0NN00 0 020020 N02 20000 0020000 0020000 0 002N00 N020000 000020 000020 0 0N020 0000N000 00N0.0 00N0.0 0 2000N0 2000200 0N00000 0N00000 0 NNO 2 .0 000N0000 000 200 000 2 0.0 0 0 20000 0N000.0 0000N00 0000N00 0 0 2 2NNO 0000N000 002 20.0 002 200 0 N00020 00200000 00N0200 00N0200 0 00NON0 00200000 00N0000 000N0000 0 00000.0 0N002000 0000200 0000200 0 0002 0.0 00002000 200N000 N200N000 N 2000 2 .0 0000N000 0000000 00000000 N 000020 220000000 0000000 00N000000 2 00000.0 00002000 0002000 00002000 2 22 .2 N 2. 2 22 .2 N u 2 22002 0082 80205228252 80.5.9 82> 20002 ".82-82.202.220.225 0.5.2.2 0.2.2.252 00002000000 n 0200 002000000000 n 0200 0NN020 n 0 _ 00000.0 n .2 2 020020 n .0 0000N0 u 0 2 0000000 u .20 002000 N .0 0000000 NN00000 00000.0 00000.0 02 0000000 NON2000 0000N0 0000N0 02 2000N0 0000000 00N0.0 00N0.0 0 200020 2000000 000020 000020 0 000020 0000000 0N20N0 0N20N0 0 2020 0000200 0N002.0 0N002.0 0 0200N.0 002000 20002.0 20002.0 0 000020 N000200 000N2.0 000N2.0 0 0N002.0 NO0N00 000020 000020 0 0002N 0 0N02 200 2000000 2000000 0 0000N0 00020.0 000020 000020 0 2000N0 0000200 0002000 0002000 0 0000 2 .0 N000N00 0000000 0000000 0 2 20 0.0 00000000 0200000 0200000 0 00000.0 0000200 0N00000 0N00000 0 200000 00200000 2000N00 2000N00 0 2N2 200 00000000 0220200 0220200 N 00000.0 00N00000 02N0000 N02N0000 N 000N00 000N0000 0000000 0000000 2 20N000 00N200.0 0002000 20002000 2 22 .2 N 2. 2 22 .2 N 2. 2 22002 .282 822020228252 82:. 8.20 05.222 20002 :82 :022osuothhohfl 151 00002000000 n mmm 00N2 2000000 H mmm 80.3 n o 2 0.38.0 u .0 2 0.02.... n 2 0 00:0 n o 2 2000.00 n .0 2 0023 u 2 000020 0000200 020020 020020 02 20N200 000200 0022000 0022000 02 00002 .0 0002200 00 2 0 2 .0 00 2 020 0 000NN.0 0000200 0002000 N002000 0 20 2 N 2 .0 0000200 2002 20 2002 20 0 00NNO 0000200 0000000 0000000 0 02 20 2 .0 020N200 0N00000 0N00000 0 00000.0 00000000 0000200 0000200 0 0000 2 .0 2008.0 00 0000.0 0000000 0 002 2 0.0 00N00000 0000000 00000000 0 0 200 2 .0 0002 200 2002000 200 2 00.0 0 000 2 0 .0 N2 2000000 00NN00.0 00NN00.0 0 000NN.0 20NN200 0000 00.0 0000000 0 0N2020 000N00000 2002000 02002000 0 020N0 00020.0 0002N00 0002N00 0 00NON0 00NON0000 0000000 000000000 0 N000N.0 00000000 NON0200 NON0200 N 000N.0 0002 20000 0020000 00020000 N 000020 00000000 00N0000 000N0000 2 0000N0 000-0000N.0 000-0000 000000000 2 m 22 N 2. 2 22 .2 N 2. 2 20002 2282 82222228252 20800 82> 20002 2282 92205228252 0.8.2.2 0.2.2.2: N0002000000 n 0200 00022000000 n 0200 N000N.0 u 0 2 0000000 u .2 2 0000N0 n 0. 0200000 n 02 0200000 u .2 2 00N000 u 2 000020 NN00000 00000.0 N00000 02 000N2.0 2200000 00000.0 00000.0 02 0000N0 0000000 0000N0 0000N0 0 000 2 .0 0000000 0000N0 0000N0 0 002020 02N2000 2020N0 2020N0 0 020020 N020000 002N0 0002N0 0 2N00 2 .0 0200000 0000 2 .0 0000 2 .0 0 0 00000.0 0000000 00 2 0 2 .0 00 20 2 .0 0 N002N0 0002000 0000 2.0 000 02.0 0 00NNO 20N0200 000N2.0 000N2.0 0 2000N0 N000N00 0NON000 0NON000 0 002 20 0000000 N000000 2000000 0 0002 2.0 20200000 0 00000.0 0000000 0 000NN.0 00 2 0N00 0000000 0000000 0 000NN.0 00008.0 00 2 N000 00 2 N000 0 00N2 2 .0 00000000 N00 2 00.0 N00 2 00.0 0 02N000 00N200 0000200 0000200 N 00NN00 020N200 2N00200 2N00 20.0 N 0 2000.0 N0000000 NON000 00 2 0N000 2 00N0.0 2000000 000N000 0000N000 2 22 .2 N 2. 2 2-2 .2 N x 2 20002 28.2 822222282022 .02... 3.20 08.222 20002 028.2- 82200228250202022 00222 2002.25.20 >122: 152 0N0000000000 n 0200 200000000000 n mmm 0083” 00 8330122 02822”.u 020020”: 2002...”: 20002.0”Nu N00020 N020000 20N200 20N200 02 00002.0 N02N200 200020 200020 02 NN0020 N000000 0 200N0 0200N.0 0 0000N0 0000N00 N002000 N002000 0 200020 N002000 000020 000020 0 002000 00000000 0000000 0000000 0 0N00000 0000000 00002 .0 00002 .0 0 000020 000N200 0N00000 0N00000 0 0000000 020N200 0N020 0N020 0 00N02.0 0NON0000 0008.0 0008.0 0 0N002.0 00008.0 00000.0 00000.0 0 00000.0 0N200000 0000200 0000200 0 02000 .0 0000200 0000000 0000000 0 N000N.0 2 000 2000 0000000 N0000000 0 N020N0 0000200 N000N00 N000N00 0 N000 2 .0 0NO0N000 0200000 00200000 0 00 0N00 00020000 N000000 02000000 N 0000 2 .0 0N00 200.0 08800 008800 N N0002 .0 200NN000.0 2 2N0000 N00 2 N0000 2 00000.0 00000000 0000000 00000000 2 a .2 N a 2 2. .2 N x 2 00002 ".230 2202222252250 «.5800 32> A0002 ".2230 2202222222250 22.2.2.2 022.202 00002000000 n 0200 0NON2000000 n 0200 02 2 2% n o 2 $2.80 N .2 _ ENS u 0 2.086 n o 2 .3080 u .2 _ 20002.0 u .0 200020 0002000 000020 000020 02 00002.0 0000N00 N0002 .0 N00020 0 00020.0 00000.0 00000.0 00000.0 0 0N002.0 202NN00 000020 000020 0 020000 20002.0 00000.0 00000.0 0 000NN.0 00 2 0 20.0 N00000 N00000 0 0000N0 00N0 2 .0 00000.0 00000.0 0 000 2 N0 0000000 00000.0 00000.0 0 2000 .0 00000.0 0000N0 0000N0 0 000N2.0 0000200 N000000 2000000 0 20N000 0002000 00020 00020 0 000N2.0 02002000 002NN00 002NN00 0 NO0N00 00NON00 00002.0 00002.0 0 02 20N0 0N00000 00N200 00N200 0 0N2020 000N200 0000000 0000000 0 000200 000N0000 2N00000 2 2N00000 N 0000N0 0000200 00N0200 00N0200 N 0N0 2 0.0 000 2000 0002000 00002000 2 02 2N00 20000000 000 000.0 20000000 2 0 .2 N 2. 2 2. .2 N a 2 22002 28.2 23222222822220 .32... 8.5 22822 2N002 ".2920 25222225222220 3002.2 8202 22.3 72320.2. 153 N00200000000 n 0200 N000000000000 H 0200 E020 u o _ 523 n .2 2 08.2.9 u .3 00002.20 u o 2 0002822 u .2 2 00008.2 u u 0N20N0 02N020.0 2Nm020 2Nm020 02 0002.0 N002000 200N20 200N20 02 00002.0 0200N00 000N2.0 000N2.0 0 0N002.0 0000000 00000.0 00000.0 0 0020000 000000 N000000 N000000 0 002020 00NON00 0000000 0000000 0 000020 0002N00 00000.0 00000.0 0 0N02N0 N002 20.0 00NON00 00NON00 0 000220 0200200 0020000 0020000 0 00002.0 0002200 0200200 0200200 0 N20NNO 0000200 0000000 0000000 0 000020 00000000 N200200 N200200 0 00002.0 0000200 0000N00 0000N00 0 200m 2 .0 000 20000 0000000 0000000 0 00000.0 002N0000 2200200 2200200 0 00N20 0200N000 00N200.0 00N0000 m 0000N0 00N00000 N00000.0 02000000 N NO0Nm0 000N500 0002000 0002000 N 0.2 2N2.0 00000000 000N000 0000N000 2 00002.0 02N02000 N000000 00N000000 2 22 W N 2. 2 22 .2 N 2. 2 202202 also» 22.206329222121958220 0.2—22> 00002 2301:0229—2229:2220 ya 0 2 0 2 00000000 H Wmm 20NN00000000 H mam 000080 N 20 000020 n .2 _ 00002.2 u 2 00820 n o 2 882.2 u .2 0 00020.20 u .u 000NN.0 0000200 20 202.0 20202 .0 02 2N00000 00000.0 0N002 .0 0N002 .0 02 000NN.0 2000200 00N220 00N2 2.0 0 00202.0 0NO0N00 02N20 02N20 0 2002.010 000200 2000000 2000000 0 N00020 0000N00 0000000 0000000 0 002 2 2.0 000NN00 0000000 0000000 0 0002 2.0 2000N00 0N00000 0N00000 0 000020 0000200 0000000 0000000 0 0000N0 0200000 0N00000 0N00000 0 M002N0 00000000 000N000 000N000 0 00022.0 0000200 0000000 0000000 0 2000000 2000200 000N00 000N00 0 0002N0 0NON200 200N00 200N00 0 220000 20000000 0000200 0000200 0 2000N0 0200200 2000200 2000200 0 0000N0 00000000 0200000 20200000 N 0000 2 .0 00000000 0000000 00000000 N 0000 2 .0 00 20200.0 0000000 000000000 2 0000N0 0200N000 NMN2000 0NNN2000 2 22 .2 N 2. 2 22 .2 N 2. 2 20002 ".230 22.222922229222220 2.22:. 8.20 222.222 20002 ".300 5222.22.25.33 020252.202 154 0NON2000000 n mmm 000000000000 H mwm 00002.0”.22 0000000“: 0222000“.2 00002.0“.22 00002.0".22 0N0N20u.2 N000 00002000 2000200 2000200 02 200020 200200 2 2N0000 2 2N0000 02 200N20 020N2000 2N00200 N00200 0 000020 00000000 0000000 0000000 0 0000000 000N000 0000200 0000200 0 000020 0000000 NO0N00 NO0N00 0 000020 00002000 00N2 200 00N2 200 0 000NN.0 0000000 20 20 200 2020 200 0 0000 2 .0 00N0 2000 0N00000 NON00000 0 0000 2 .0 0000N000 0000200 0000200 0 0000N0 0N02 2000 0000000 00000000 0 00000.0 0000N000 0N00000 20N00000 0 0002 00.0 2 202N000 02N0000 00 2 N0000 0 0000N0 00002000 000N000 0NO0N000 0 02 200 0N00000 000N000 0000N000 0 000N.0 020000000 0000000 00000000 0 00020 2 2N000000 00N200.0 000N2000 N 000020 000N20000 0020000 N20020000 N 0000N0 000000000 0000000 000000000 2 N002N0 000-0N0N0.N 000-0000 000600000 2 22 .2 N 2. r 2 22 .2 N 2. 2 20002 Blue» ...22._o=22.2..2|:2fl.2|....28a0 0.2—22> 20002 9:30 =022u=2202fl2l20 22.2.2.2 22.2.22 0 20000000000 n 0200 200000000000 n 0200 20N000Ho2 N000000n.2_ 00202.0u0 00022.0".22 2N0000u.: 20022.0“.2 00 2 0 2 .0 00000000 0000000 0000000 02 2000 2 .0 20000000 0000000 0000000 02 00002.0 000N0000 0000000 0000000 0 000020 00000000 0200000 0200000 0 0000 2 .0 00 00000.0 0000N00 00008.0 0 00N0 2 .0 20N00000 00N0000 00N0000 0 0000 20 0020000 0N00200 0N00200 0 00020 00000000 02 $80 02 080 0 0000N0 00020000 00NN200 00NN200 0 0000N0 0N00000 0000200 0000200 0 N000N.0 0000N000 2 200000 2200000 0 00000.0 N200N000 2000200 2000200 0 0NO0N0 00002000 00000.0 20000000 0 2N0000 000N000 0000000 0000000 0 200 2N0 02022000 2002000 2 2002000 0 000NN.0 0200N000 0NON000 0NON000 0 000020 000000000 0000000 N00000000 N 0000N0 000000000 0000000 020000000 N 0000 20 00060002 .0 000-0000 08.00000 2 0000 2 .0 00020000 0020000 020020000 2 22 .2 N 2. 2 22 .2 N 2. 2 20002 "2.2.2.0 ...2229328250 2.22:. :50 9.22.222 20002 v.30 5220.22.32.20 2.20.2.2 3.222 20225.20 2202220252 155 N0002000000 H Mmm 000000000000 u mmm 022.6": 6.6.26": 230.26% 222.26“: 23.22661: 0220.256”d .8. 26 0606.66 3.26 $.26 62 66.26 .6. 266 8.2666 8.22066 62 .2.2 .6 2.6.66 0.2.26 02.26 a 20.626 E20666 .5266 .5066 a 26.226 ....26.6 028.66 020.866 . .826 023.666 82.86 82.86 . 2.2226 2.3666 ...2066 ...2066 . 83.6 5.866 ..2866 ..2NN66 0 .6226 8.0.666 $0.86 $0.86 0 .6686 2.20.666 $0266 $0266 0 .2326 222.666 $3.66 62.0.66 . 324.6 .8886 $2866 2.02.666 . 0.8.6 $06866 .8086 .8086 2 2.0.6 200.2666 .586 00.866 2 0.0.26 .0...66.6 .6386 2602266 . 0....6 $86666 .2866 .32 .666 . $.26 .2286 .6866 26.6866 N 8826 33.6666 2.202666 22.202666 N 2.32.6 8.8666 ...866 0....666 2 6666.6 023.6666 £06666 0602206666 2 22 .2 N 2. 2 m .2 N 2. 2 00002 9:290 nag—2.22.2552 22.22520 .2022; A0002 9:30 52229522322222 22.2.2.2 0.2.2.252 8.2666666 H mm. 2826666666 H .5. 0.8.6 u o 0 .6826 u i 0.006 n 22 5.6 u o _ $02.6 u 3 200.6 n a 22.686 .8066 .8306 ...36 62 8.026 26.0.66 .20....2 $00.6 .. . 56.66 866.66 2.00.6 2.00.6 . 6.8.6 0.8.6 $8.6 .0806 . 0.06.6 2.0.66 $8.6 $8.6 . 0.0.6 2....6 .8606 .8606 0 23:6 2.66.66 0.6.26 .826 o 33.6 .0826 .206 .306 o .326 2.60.66 8.26 8.626 m ...26 2.26666 8.6.6 8.6.6 . .386 .06. 266 86.066 30.066 0 206026 82366 2826 0.826 2 262.06 8.3666 .586 25.66 . 2036.6 26.266 .662 26 .662 26 . 222.36 286.666 54266 5366 m 300.6 822.266 3.6.66 2.6.66 N 232206 56.666 26.666 2006.666 2 0.0.26 8.0.266 602.266 602.266 2 22 .2 N 2. 2 22 .2 N 2. 2 220.2 :8. 52.2.2283: 222.085 2.522 20002 ".39. 22.226.225.250 Loan—.2 8202 23.22.30 0027201222th? 0222.0 156 N0002000000 n mmm 200000000000 n mmm 22.0.26": 626626”: 00.36% 263.6": 03226”: 06.626”.Nu 000020 0020000 0020N0 0020N0 02 000200 2020 000200 000200 02 000020 0000N00 N020N0 N020N0 0 000NN.0 002020 0020N0 0020N0 0 0N002.0 20080 000020 000020 0 N00020 002N000 0N020 0N020 0 000020 000NN00 0002 20 0002 20 0 0000N0 000080 0002 2.0 0002 2.0 0 0N 2 NNO 00 2 0 20.0 20N0000 20N0000 0 00NON0 NN00 200 00000.0 00000.0 0 02 0N0 2000200 N000000 80000.0 0 NO0N20 0000200 0000000 0000000 0 000020 2NN0200 0000000 0000000 0 202000 00200 $0080 0000N00 0 0000N0 20000000 0200200 0 200200 0 0000N0 00000000 002N200 002N200 0 000020 0N000000 0000000 00000000 N 000000 0NO0N000 00000.0 00000.0 N 00N20 00002000 2202000 002 0 200.0 2 0000000 00000000 200000 200000 2 22 .2 N 2. 2 22 .2 N 2. 2 2002 9:30 5222.25.22.26 a. 2.5.0 8225 00002 ".39. 5.230.222.2120 22.2.2.2 9.2.2.222 00002000000 n 0200 0022000000 n 0200 000020": 00000”: 000000n0 000000”: 0N002.0”: 00000.0”.2 000020 0000000 0000N0 0000N0 02 0000N0 0N000.0 00000.0 00000.0 02 0002.0 N202000 0020N0 0020N0 0 0000N0 000020 200000 200000 0 0000 2 .0 0NN0200 N000 2 .0 0000 2 .0 0 0000N0 000N2 .0 020000 0 20000 0 00NON0 0200N00 00N02.0 00N02.0 0 0200N.0 NN00000 000.0 000.0 0 N000N.0 000 2 20.0 000020 000020 0 0200N.0 0N00000 00NON0 00NON0 0 0002.0 0308.0 00N000 000N000 0 000020 N000000 200020 200020 0 000020 N000200 0000000 0000000 0 0000N0 000N00 00NN2.0 00NN2.0 0 00N02.0 0202200 00008.0 00008.0 0 2NNON0 0000N00 0N00000 0N00000 0 000NN.0 N0000000 000N200 000N200 N 000200 0000200 0000000 0000000 N N2N000 0N00000 0000000 0N000000 2 000N.0 00000000 000N200 000N200 2 22 .2 N 2. 2 22 .2 N 2. 2 20002 :30 3229222250 .52... .250 2.2.222 H0002 ".39. 52295229222220 5002.2 3202 2922.25.20 GZ<2 00002 ".230 :o_,2o=22.2w2u..22v 22.2.2.2 9.2.2.200 00022000000 n 0200 000000000000 n 0200 8.2 .6 u o _ .926 u 220 .0086 u .0 83.6 n u _ 0.3206 n 2 2 .226 u 02 000020 00N02.0 00000.0 00000.0 02 N00020 0000N00 2000N0 2000N0 0 N00000 00000.0 2 2 20.0 2 2 200 0 0002 20 0000000 2 200N0 2 200N0 0 NO0N00 002000 200000 0000.0 0 N00000 00080 000020 000020 0 N02000 200000 02000.0 02000.0 0 000020 0000N00 2002 2.0 2002 2.0 0 0200N.0 N02 2 2.0 000N.0 000N.0 0 000N2.0 0000200 000N000 000N000 0 0000000 0002N00 00002.0 00002.0 0 000NN.0 0000200 00000.0 2000000 0 0NO0N0 NON0000 000020 000020 0 0002N0 2000 200 0000000 0000000 0 00000.0 2000000 0000000 0000000 0 200000 00000000 2000200 2000200 N N002N0 00000000 2000200 2000200 N 0000N0 0000000 00N0000 000N0000 2 0000.0 N2 00 2000 0000000 2 20000000 2 22 .2 N 2. 2 22 .2 N 2. 2 00002 9:30 52222229222222 .2222... 8.20 222.222 00002 222.201.238.222222220 2902.2 32.2 2922.28.20 ><>>~2072 158 0000N0000000 H 0200 000000000000 n M00 2222.86 u o 0 05.66 H .20 2622.066 u .3 0222.26 u o _ $2.66 u 22 0 0.0.06 u 22 0000.0 N00020000 N00000000 000000000 02 0002000 0N000.0 0200000 0200000 02 0000000 000NN0000 000000000 000000000 0 0NO0N0 00002000 NO0NN000 NO0NN000 0 00 0000.0 020N20000 000000000 000000000 0 NN0020 00022000 00 202000 00 2 0200.0 0 00N0 2 .0 0000N0000 000000000 000000000 0 00002 .0 00000000 0NN200000 002200000 0 20N0000 2NO0N0000 2N2 0000.0 002200000 0 0000 2 .0 000N0000 N00000000 000000000 0 0N00000 0 2 0NN0000 0000N0000 0000N0000 0 2000000 000N00000 000000000 0000 0000.0 0 20NN000 000020000 000020000 200020000 0 0002 20 200200000 00 2NN0000 00 2 NN0000 0 0000000 0N0020000 N202 20000 000220000 0 0000N0 0NN020000 N00020000 N00020000 0 20N0000 0N0020000 000-00000 000600000 N 0000000 000600000 000-0N2020 000-0N02.0 N 0 2 02 2.0 000-00N000 00060000. 2 08.000002 2 000N 2 .0 000-00N000 000-0000N. 2 000-200 2 0. 2 2 0 .2 N 2. 2 22 .2 N 2. 2 00002 ".5220 3222222325 «2222.20 0.2—22> 00002 2.2430 52223229222240 22.2.2.2 9.2.2.222 NO0N0000000 n 0200 000220000000 n 0200 .326”: 22.6.2”: 2222.2”. 2262.26": 2.226": .0363 020020 00080 0002200 0002 200 02 000N000 00N00000 0000N00 0000N00 02 2000000 20202000 0200000 20200000 0 000N2.0 00020000 000NN00 000NN00 0 00000.0 0N020 00000000 0N000000 0 00N2 .0 000N0000 2000200 2000 20.0 0 00NNNO 00020000 00000000 N0000000 0 NON0 20 NN20000 0000200 0000200 0 000020 N002 200.0 02000000 0N000000 0 200020 020N0000 2 200 2 200 0 0NON000 000000000 0 2000000 N0000000 0 0N02 2.0 00020000 0000000 20000000 0 0 2 N0 2 .0 000 2000 00NON000 00880.0 0 0000 2 .0 00000000 00NN0000 00NN0000 0 00000.0 000000000 00022000 20022000 0 020N00 0000N000 00020000 00020000 0 000020 00NN00000 000000000 020000000 N 200000 0002N000 00002000 00002000 N 0N020 220000000 0N2020000 002020000 2 002020 02002000 000000000 000000000 2 22 .2 N 2. 2 22 .2 N 2. 2 00002 2.2220 5222222222220 2222220800 2.2.22.2 00002 2.2.501 522322292252 2902.2 3.222 2223 Z<.2.02V2<.2 159 160 200000000000 n 0200 000000000000 n 0200 0000N0 u o _ 0000200 n .2 _ 2000000 u .2 02NON00 u o 2 00000000 n .2 _ 0000000 u .2 020020 NN000000 0000000 0000000 02 0020N0 000020 00N02.0 00N02.0 02 002 2N0 00020000 00080 2000N00 0 002 200 00000.0 0200000 0200000 0 202N0 N0020000 2000N00 2000N00 0 00000.0 02 00N0.0 0020000 0020000 0 0000N0 00000000 000200 000200 0 0N000 20000000 0N20000 00N20000 0 000N00 00N2N000 20N00000 00N00000 0 2 2NN00 00002000 2200000 22200000 0 200NN.0 00002000 2000000 N2000000 0 N002N0 002N2000 0002000 0002000 0 000020 00N02000 0020N000 220N000 0 00000.0 220000000 0000000 000000000 0 0NO0N0 00022000 0002000 00002000 0 0000N0 00060000 0020000 200020000 0 N20000 0 2 0000000 N00 00000.0 00000000 N 0002 0.0 0006N000 .0 0006000 000600000 N 000N2.0 00062N00.0 000602 20.0 000600000 2 2000000 000600000 000600.N 0006000.N 2 22 .2 N 2. 2 22 .2 N 2. 2 00002 2.230 3222.22.23.25 9:28.20 .2622; 00002 Elna» nets—2.2.2.222 6.5.2.2 6.2.2.222 00 2000000000 n 0200 000N2000000 n 0200 82.6 n o _ $662.66 u .2 2 0.8.6 n 22 ...26 n o 2 8.0.66 n .2 _ 6.6626 n 22 00N0000 2002000 NONON0 NONON0 02 000020 0N00000 0N02N0 0N02N0 02 000020 0000000 02 2N0 02 2N0 0 0NN0000 020 2 000 00002.0 00002.0 0 000020 0NN0000 00202.0 00202.0 0 2NON000 0000000 000020 000020 0 N002 2.0 000N000 000020 000020 0 2NNN20 00NON00 000020 000020 0 2002N0 0020N00 00000.0 00000.0 0 000NN.0 0N00 20.0 0N02 00.0 0N02000 0 00N2N0 200NN00 0000000 0000000 0 00N02.0 0000200 0N20000 0N20000 0 202NNO 0000200 0000000 0000000 0 0002 0.0 0002200 0 2N0000 0 2N0000 0 NN0000 000N200 0000N00 0000N00 0 0000N0 00000000 0000200 0000200 0 02 2N0 000N200 0000200 0000200 N 0000N0 0000000 20020000 N0020000 N 00000.0 00000.0 000N000 N000N000 2 00000.0 2 200N000 02002000 0200 2000 2 22 .2 N 2. 2 fl .2 N 2. 2 00002 "2.2.2.0 522622202252 .32... 3.20 68.222 00002 2.80 52226323250 30.2-22.2 8202 26:3 027241203 000000000000 H 0200 NON000000000 n mmm .6226"; 32.26”: 92.0.6".N2 020.666“: 662626”: 8.0.6.on 000000 000N000 200000 200000 02 00NON0 0000000 000020 0000 20 0 00000.0 00000.0 2000N0 2000N0 0 20200022 0000000 N00000 N00000 0 000N00 NO0N00 220000 220000 0 02N02.0 N20000 0000000 0000000 0 020000 0002000 0020000 0020000 0 0000 20 0000000 0N000 0N000 0 0000N0 0000000 N000000 N000000 0 0000N0 0NO0N00 0000000 0000000 0 0 200 2 .0 20 $80 NON0 20.0 NON0200 0 0000N0 00002022 N000N022 N000N022 0 0N0000 20000022 00000022 000000022 0 0000N0 00N0 200 0000200 0000200 0 N00000 0NONN000 2 2NN000 2 2 2NN000 0 000000 00NN:00 00000002 200000022 N 000000 00020002 00000022 N000000022 N 0000000 00000000 000N000 N000N000 2 020000 000000000 0N000.0 0808000 2 0 a N a 2 0 a N x 2 @002 2.560 =02226.%2Ea0 .2622; 00002 2.36 32263292252 22.2.2.2 a 000000000000 n 0200 0 20000000000 n 0200 020000”: N200N0u.2_ 0020N0u22 020020": N0002.0n.2_ 00002.0u..2 0002N0 0000000 0002N0 0002N0 02 N02000 0000N0 0N20N0 0N20N0 02 000000 2000090 002000 002000 0 N000N0 N00000 0N02NA2 0N02N0 0 0000N22 00020Q0 220000 220000 0 0000N0 00000 000000 000000 0 0N000 N020NQO 000000 000000 0 000N00 0002000 002000 002000 0 000000 00000022 00200022 00200022 0 0000N22 0000022 02NN0022 02NN0022 0 000000 20000022 20000022 20000022 0 000000 0000N022 00000022 00000022 0 0000N0 0020000 00N200.0 00N200.0 0 0N2020 0N00200 000200 000200 0 0N000.0 002 0 200 0000200 0000200 0 000020 00000000 0000000 00000000 0 0NON22 00022022 00000022 200000022 N 020N22 000200022 N00000A2 020000022 N 00N0 2 .0 00 2 0N000.0 0006000 000600000 2 200 200 0000N000 0002000 00002000 2 0 a N x 2 0 0 N n 2 00002 "28015222522822: 2625.28.20 6.22.2: 00002 ".360 gazes—2.2.2.222 .50....2 6.2202 26.23 iUDHflOm 161 000000000000 u mmm N00000000000 H 0.200 20N200 n o _ 00N0000 u .2 _ 00000.0 u .u 0N00.2 u o _ 00000.0 n 3 0200.2 n .0 0000000 0002200 0 2 0000.0 0 2 0000.0 02 000 2 N0 2002N00 000N000 000N000 02 00 2000 N0 00000.0 0020000 0020000 0 NONN 2 .0 00000000 0NNON00 0NNON00 0 00N0 2 .0 0000000 0 2NN000 0 2 NN000 0 00000.0 00N00000 2000 2 0.0 2000 20.0 0 0N020 0N000000 0002000 0002000 0 00000.0 000N0000 00000.0 20000000 0 000000 00000000 0000N00 0000N00 0 0000.0 N20000000 000N000 0000N000 0 000 0N0 N0 2 N0000 0000200 0000200 0 02 0000.0 00 2 00000.0 0000000 N00000000 0 N000 2 .0 0000N000 0000000 0000000 0 2000 2 .0 0NO0N0000 00N0000 0000N0000 0 0000N0 00000000 0000000 00000000 0 00 20N0 200000000 2 2000.0 N00020000 0 0000 2 .0 00002000 200N000 N 200N000 N 0002 20 000-000 2 .0 000-00N0 000-0020N0 N NNO 2 0.0 000000000 0000000 000000000 2 0N0 0 2 .0 000-0NN0. 2 000-0N0.N 08.00000. 2 2 22 .2 N x 2 22 .2 N 0 2 0 2002 am 52222252252 Eoflao 83> 02002 v.30 5229322250 25.2.2 22.2.2: 0 20000000000 n 0200 00NN00000000 n 0200 022 u up 3903 u a _ So u a. 022 u o 0 8230 u i 2923 n 0 0NN0000 0000200 0000000 0000000 02 00 2 0N0 0 2 00 200 20000.0 20000.0 02 2000000 0000200 00N0000 00N0000 0 000000 N2NN200 0000000 0000000 0 002000 00200000 00N000 00N000 0 200020 00000000 2220000 2220000 0 200020 2 2N00000 N202 00.0 N202000 0 200020 00000000 N000000 N000000 0 200000 0000200 00NON00 00NON00 0 0NNON0 0000200 0000N00 0000N00 0 000020 00N00000 N000200 N000200 0 020020 00000000 2000200 2000200 0 000020 0NO0N000 0000000 20000000 0 000020 2020N000 0000000 00000000 0 0 20020 00000000 0000000 00000000 0 00N02.0 20000000 0000000 00000000 0 2N002.0 NNO0N000 202000 N0202000 N 00N000 20002000 0202000 00202000 N 00000.0 200020000 N020000 N020000 2 0N20 00 000000000 0000000 N00000000 2 2. .2 N u 2 0 .2 N x 2 00002 is?» 3229:0235 .32... :80 2.82.2 00002 ".390 :022952292252 302.2 8202 23.2.50 ZO2H<~2020202~2 72200052 162 0NON00000000 u M00 00202000000 n 0200 163 wmeénfi ”23".; 00:;de Nomad": 2886“: Egouu 0N00000 00000000 NN00000 NN00000 02 N2000 00N00 .0 2 20000 2 2000.0 02 0000 2.0 00000000 0000000 0000000 0 0000N0 20002.0 00000.0 00000.0 0 000000 0000000 0000000 0000000 0 0000N0 000 200.0 2 2000.0 2 20000 0 N000000 00000000 0000000 0000000 0 00NON.0 0000000 0000N0 0000N0 0 0NO0N0 0N000000 000N00 000N00 0 0N00 2 .0 0000000 0000 2 .0 00002 .0 0 0N0000 N0000000 0000200 0000200 0 N20N0 0000N00 00002.0 00002.0 0 000020 0000890 0000200 0000200 0 200000 000N200 0000000 0000000 0 0000N0 0N002000 0000000 0000000 0 N00000 N020200 0000000 0000000 0 0N000.0 00002000 00N000 800800 N 00000.0 N0200000 00N0200 00N0200 N 2N0000 000000000 0200000 000200000 2 00N000 02N00000 0NN0000 00NN0000 2 0 .2 N a 2 0 .2 N a 00002 v.30 522935.252 «.5500 82> $002 Baum—329:2.E2fl22 25.2.2 022.2022 0002000000 H 0200 02002000000 n 0200 £83 u o 0 28086 n a _ 525 u .3 $32 n o 2 8.3 a _ 232 u w 0202 20 0000000 002 00.0 002000 02 000020 0002000 00000.0 00000.0 02 0 2 002.0 002 2000 0000N0 0000N0 0 000NN.0 0N00000 00000.0 00000.0 0 000N2.0 0200N00 000NN.0 000NN.0 0 000020 0000000 N00000 N00000 0 000020 0002000 00002.0 00002.0 0 0000N0 0000N00 0000N0 0000N0 0 0NN000 0080 020020 020020 0 0000N0 0200000 00002.0 00002.0 0 0000 00 002NN00 2000000 2000000 0 0200N.0 0000N00 NONN20 NONN20 0 2 2000.0 000N200 000N000 NO0N000 0 0NO0N0 00N2N00 00000.0 00000.0 0 N000N.0 0002200 20000.0 20000.0 0 N2000.0 000N200 0000N00 0000N00 0 0000 2 .0 00000000 0 2 0020.0 0 2 0020.0 N 000 2 N0 0N000000 0N00000 00N00000 N 200000 0N000000 0N000.0 00NN0000 2 200 2N0 0NO0N000 000N000 N0 0080.0 2 # .2 N u 2 0 .2 N x 2 00002 ".390 52292252250 32:00:00 2.52.2 00002 "LEN 5222—03250 3.02.2 3202 23.25.20 §0m<07220 _1. I. 0- 1 \li’..- 4‘ I‘I.‘ I1‘ .-luululiIlIV I41.Ilqin~ nIIWA~ \aiflunhnnhqi. \ \ 2.53: .0 tr? :NJQ 00NON000000 H mwm 000N00000000 H 0200 Sm; u o _ 30035 n a _ 38.2 n a a .93 u o _ .00de n .2 _ 883 n .3 002 200.0 00000000 0202000 0 202 00.0 02 0020N0 2000000 00002 .0 000000 02 2000000 0N000000 2000N00 2000N0.0 0 0000N.0 020000 0200000 0200000 0 0N2NNO 20000000 0002N0.0 0002N0.0 0 0N000 N000200 0000000 0000000 0 2N02 .0 00000000 0000200 000200 0 200000 0000200 NNON00 NNON00 0 00N2N0 00000000 NO0N 20.0 200N200 0 200000 0002 200 0000200 0000200 0 0000 2 .0 0000N000 0000000 00000000 0 0002 0.0 000000000 000 200.0 000 2 00.0 0 0N000.0 000000000 000 000.0 0000 000.0 0 002N0 000000000 0000000 N00000000 0 0000N0 0000800 0020000 N002 000.0 0 000N2 .0 000000000 000000 000000000 0 0000.0 00002000 200 2000 02002000 N 000000 000N20000 00N0000 NO0N0000 N 00000 N00 00000.0 N000000 002000000 2 2000N0 38800.0 02 2000.0 000220000 2 0 .2 N n 2 0 .2 N a 2 00002 ".890 522952.324: 95:30 82%; 00002 9:30 5222.28.22.25 2: .2.2 922.2052 0NO0N0000000 n 0200 0 2NNN0000000 n 0200 0880 u o _ £830 u p _ 08.0 n u 223 u o _ 8880 n .2 _ 8203 u .3 00202 .0 0002 200 0000000 0000000 02 N000000 000N0000 0000000 0000000 02 00002.0 000N0000 20000.0 20000.0 0 000N2.0 200N0000 0000000 0000000 0 0000 00 0000200 0200000 0200000 0 0000 00 NNON000.0 N000000 N000000 0 0200000 0000000 00008.0 00008.0 0 0202 00 0000000 0000N00 0000N00 0 0000 2 .0 00000000 N000N00 N000N00 0 0N002.0 0NN00000 002 2N00 002 2N00 0 0000N0 N2000000 N000200 N000200 0 0000N0 00200000 0200200 0200200 0 N0000.0 00000000 0000000 0000000 0 2N002.0 20000000 0000000 0000000 0 00200.0 008800 0000000 00000000 0 00NON0 00000000 2 200000 00200000 0 02000.0 00280.0 0002000 0002000 N 00000.0 00NON000 N002000 N002000 N NO0N2.0 000000000 0020000 000020000 2 200020 0NO0N0000 0020000 000020000 2 0 .2 N a 2 0 .2 N 2. 2 00002 v.30 522309.252 .52... 5.5 2:022 00002 :80 5222—2282:: 5.0.42.2 8202 23% 0.02.5240 22.2.0200 164 T . 1 0.0.23 :.: :flg 0N000000000 n 0000 000N00000000 n 0200 508.0 u o _ @285 u a _ 03080 n 0 32035 n fl 8280 u a _ £23 u .u 00202.0 000N000 0000 2.0 0000 2 .0 02 00 2 N0 0N2 00.0 200N000 2 0 0N00.0 02 2N002.0 80080 00002 .0 00002 .0 0 0000N.0 000N200 0000000 0000000 0 0N00 2 .0 00NON00 0N2 2.0 0N2 2.0 0 2000N0 02000000 0NO0N0.0 0NO0N00 0 000N.0 0002 20.0 000N000 000N000 0 00000.0 N0000000 000 20.0 00020.0 0 200200 00020000 N000000 N000000 0 020000 00N000 0000000 0000000 0 0N20N.0 000N200 0200000 0200000 0 00000.0 00222000 0000000 00000000 0 0NO0N.0 0000200 000NN00 000NN00 0 00NNO 20000000 002000 00N0 200.0 0 N000N.0 0 2 00000.0 N002 200 N00 2 20.0 0 NONNO 000000000 0000000 00000000 0 000000 NO0N0000 00000.0 00000000 N 00NON0 0088000 200000 0N020000.0 N 00 0000.0 20N2 2000 00000.0 000000000 2 0NO0N0 000020000 02000.0 000020000 2 0 .2 N a 2 0 .2 N n 2 0N002 9:30 5222.25.22.25 9.25.0 82> 00002 ".230 5222.23.22.26 a... .2.2 22.2.22 000N2000000 n 0200 002N00000000 n 0200 8280 u 02 8&2 u a 2 288.0 M 0 0.83 n o _ 553 u n _ 382... n N. 0200000 000080 NNO0N0 0NO0N0 02 0000 2 .0 002 00.0 20N0 2 .0 20N0 2 .0 02 0000 2 .0 NN2 2N00 0000N0 0000N0 0 0002N0 0NON000 NN02 00 NNO 2 2.0 0 000020 N000000 00202.0 00202.0 0 2002N0 00N200 N000000 N000000 0 0000N0 2N0200.0 20NN2.0 20NN2.0 0 00000.0 0200N00 0N000.0 0N000.0 0 002 2N0 0002 20.0 00N000.0 00N000.0 0 00200.0 00000.0 0002 00.0 0002000 0 00NNO 0000N00 0200000 0200000 0 0000 2 .0 00000000 N000200 N000200 0 002N0 2000 200 0000000 0000000 0 00NON0 0000N000 NON0000 2NON0000 0 00000.0 00 2 0000.0 0000N0.0 0000N0.0 0 0000 2 .0 220000000 08800 08800 0 00000.0 00000000 0000000 N0000000 N 00200.0 0020N000.0 00000.0 000000000 N 2 0 00N0 00002000 0002000 N0002000 2 N0 2 000.0 00 2 0N000.0 0 2 N0000 20N 2 N0000 2 0 .2 N a 2 2. .2 N a 2 00002 ".390 5223222252 .52... E00 2.2.222 00002 ".280 5222.25.20.52 Loam—.2 3.202 23.2.5.0 00002 243.0 5220:2552”: 25.2.2 022.2022 000000000000 n mmm 200000000000 n mmm 082... u u _ 3.08... n p _ 0.53... n a. Ems... u o _ 882. n a _ 3.08... n a 0202 00 2000000 000NN.0 000NN.0 02 020020 0000000 0000N.0 0000N0 02 20202.0 NNO0N00 00002.0 00002.0 0 200000 0000000 0000N.0 0000N.0 0 0N0000 0000N00 NON020 NON020 0 00N0 0000000 00202.0 00202.0 0 2020N0 N020200 N20020 22002.0 0 00000.0 0200200 200N20 200N20 0 N000 2 .0 000 2 N00 0200000 0200000 0 N0 002 .0 2N0 2 00.0 000N000 000N000 0 000NN.0 0200 200 0000000 0000000 0 2000N0 0000200 0N02000 0N02000 0 00000.0 00200000 0200000 0200000 0 200020 20NN200 2000N00 0008.0 0 002 0N0 0200200 0000N00 0000N00 0 0NON0.0 00020000 0020200 0020200 0 0000N0 00000000 0 20000 20020000 N 000 2 0.0 N 2 00N000 0NN0000 00NN0000 N 0 2 000.0 00NON000 NO0N000 NO0N000 2 000NN.0 00N0 200.0 20 2000 00002000 2 2. .2 N u 2 2. .2 N u 2 00002 2.30. 522252252 .52... 5.5 25.22 00002 aghast—59.252 Lows—.2 3.202 23.2.5.0 7224.20 166 l‘ l‘l|‘ . \ 0 Z....2...N. . . 14 00002000000 H mmm 200200000000 .I. mmm 0200000 U o 2 0NN000 n .2 _ 00N2 2.0 u .u NNO0N0 n o _ 0000000 n .2 _ 020000 n 0 020020 2002000 000020 000000 02 2N002.0 0000N0 0000.0 0000.0 02 0N2020 000N200 200220 200200 0 00002.0 00002.0 00N000 0080.0 0 N000 2 .0 2000 200 0000000 0000000 0 00N20000 0 0N02000 N00 2 0.0 N00 2 0.0 0 0000 2 .0 0002200 000N000 00N000 0 0020 2 .0 2000000 0000N.0 0000N.0 0 0000 2 .0 0NN00000 0000000 0000000 0 0N02N.0 0000N00 20202.0 20202.0 0 0020N0 N0020000 0000000 0000000 0 0NONN.0 0N00 0 002 2.0 0002 2.0 0 00020 20000000 0000200 0000200 0 N00020 002N00 0020000 0020000 0 2000N0 00000000 0000000 00000000 0 000N.0 0000200 0N000.0 0N000.0 0 000 2 0.0 200N000 0N00000 NON00000 N 0200 2 .0 0000200 0000N00 0000N00 N 200000 000000000 0000000 0N0000000 2 00000.0 N0 0000.0 00 2 0 20.0 002 0 2 0.0 2 2. .2 N u 2 2. .2 N x 2 0N002 v.30 0022.52.32.20 «20:30 82> 00002 9:290 5223252252 2.22.2 022.2022 00N000000000 n 0200 0000N0000000 n 0200 223%... u o 2 .338... u .2 2 23.2... n 2. 3.8... u o 2 £28... u .2 2 8.8... u 0 000020 000000 0NON00 0NON00 02 0NNN000 0000000 0NO0N0 0NO0N0 0 000000 0000.0 00N02.0 00N02.0 0 0N000.0 0N0200 00200.0 00200.0 0 NN00000 00000.0 000000 000020 0 02 20.0 002NNO 00000.0 00000.0 0 0NN020 0000N00 000000 000000 0 00800 00002.0 20000.0 20000.0 0 NNO0N0 0000200 0000000 0000000 0 000000 000020 0000N.0 0000N0 0 00 0 .0 2 2000000 N000000 N000000 0 002000 000000 20002 .0 20002 .0 0 2000N0 000N200 0000N00 0000N00 0 00000.0 0020000 0000000 0000000 0 0N000.0 00000000 0200200 0200200 N 200200 0000N00 00N000.0 00N000.0 N 2N0000 00020000 0020000 N0020000 2 0000 2 .0 0 2 00N000 00NN00.0 200NN000 2 2. .2 N x 2 2. .2 N .. 2 00002 "23.. 522368252 .52... :80 2.322 00002 as...» 5222—2522:: 2922.2 8202 23.0.80 7220020230 167 11...] 1.11 I I‘ll.| lll'llll... 1'11. .II l1lil§1l - :72 0. .2 2....N0 0... ...IN 00002000000 n mwm 000N00000000 u mmm .208... u o _ 28.8... n .2 2 8.2.... u a 0.3.... u u 0 32.3... n .2 2 300.... u u 000N000 0000200 N020N0 N020N0 02 000N.0 020N20 0N000.0 0N000.0 02 0NON00 00020000 000020 000020 0 0NO0N0 N000000 00 20N0 00 20N0 0 0000000 0000N00 000020 00002 .0 0 20N2N.0 2N0000 00002.0 00002.0 0 000000 0000N00 000020 000020 0 0000000 00020.0 200020 200020 0 000020 00N000.0 000N2.0 000N2.0 0 00NON0 NONON00 00N00.0 00N00.0 0 NO0N0 0000N0.0 0000000 0000000 0 2002 0.0 0NO0N00 0N 200.0 0N 2 00.0 0 NN0000 0NN0200 N02N000 N02N000 0 0NON00 00N0200 0200000 0200000 0 N2000.0 0000200 2N00000 2N00000 0 N000N.0 0000200 0000200 0000200 0 0 2N200 000N200 00N0200 00N0200 N 00N00.0 N000000 2 200000 2200000 N 0 2 N000 2 2N00000 N00 000.0 02000000 2 2 2N000 0N0 20000.0 00000.0 N0000000 2 2. .2 N x 2 2. .2 N 2. 2 00002 .2430 002.232.922.222 «.5800 82> 00002 "2.3.0 3229.22.32.22 25.2.2 022.2022 000000000000 M 0200 000200000000 n 0200 .80.... u o 2 E... u .2 2 .0020... u 3. R2... n 02 3.20.... n .2 2 88.... n .. 000N000 0220000 00200 00200 02 02000.0 2000N0 00N00.0 00N00.0 0 0000N00 0000N00 020000 020000 0 N00000 00002.0 NON000 NON000 0 00000.0 00000.0 00000.0 00000.0 0 2N000 0000000 N02NN.0 202NNO 0 20000.0 0000N0 0200N.0 0200N.0 0 000000 2000000 0N2020 0N2020 0 00000.0 0002.0 0000 2 .0 0000 2 .0 0 000N2.0 0000N00 0200000 0200000 0 000020 0NO0N00 00000.0 00000.0 0 00000.0 N000N0.0 02N0000 0 2N0000 0 000N2.0 000N200 N002000 N002000 0 000N00 00008.0 00008.0 0000N00 0 2020.0 0020000 0000200 0000200 N 00000.0 00NN0000 0000200 0000200 N 000000 00020000 NO0N000 NNO0N000 2 NNO0N0 0002000 0000000 000000000 2 0 .2 N .. 2 0 .2 N x 2 00002 ".592 522325.552 .52... :50 22...: 00002 “.592 522935.252 $51.22 2202 23.2500 QZ<222HNP2>>0 168 :- 1...... \.uqu:-h-\. \\ , i «12...... . {fl 0 20000000000 n ”.200 0 20000000000 n mmm $.28... n .2 80...... n .2 _ 0.258... n a. 8.200... n u 2 3.0.... u .2 2 28...... u 0 N00 000.0 00 2 0 2 0.0 N000000 2000000 02 20N0000 0000000 20N000 20N000 02 00002.0 00N00000 000N000 000N000 0 200000 00000 0000N0 0000N0 0 000000 00000000 0N0000.0 08000.0 0 0N20N.0 N002000 N00020 N00020 0 0020N0 N000000.0 0000000 0000000 0 0000N.0 000NN00 000N2.0 000N2.0 0 000NN.0 000N000.0 0000N00 000 0N00 0 0000N.0 0000200 02N0000 02N0000 0 000NN.0 00000000 0020 200 0020200 0 00N02.0 2000200 00000.0 0000000 0 2000N0 0000000 0000000 00000000 0 00000.0 0NON200 0000000 0000000 0 0000N0 0NON000 0200000 N0 20000.0 0 N002 0.0 00N0 20.0 000 20.0 00020.0 0 000N00 00002000 NO0N000 0NO0N000 N 020NNO 00000000 0000.0 N0000000 N 0002N0 2 2N000000 0N00000 000N00000 2 0000N0 0NO0N000 2002000 2002000 2 2. .2 N .. 2 2. .2 N .. 2 00002 ”aflnc2m_o=2.e..2u22 «Sn—«D 82> 00002 ".392 022222.022”: 0.3.2.2 922.2022 02002000000 n 0200 00002000000 n mmm .88... u .0 2.8.... n .2 2 23.2... n .3 $82... n o _ .228... u .2 2 .82... u .. 0002N0 0000000 2002N.0 2002N.0 02 0002N0 00N02.0 00000.0 00000.0 02 00002.0 2N00000 00002.0 00002.0 0 2NO0N0 0002.0 00000.0 00000.0 0 00202.0 00NON00 000N2.0 000N00 0 2 200N0 2000000 000N.0 000N.0 0 00002.0 00008.0 00N000.0 00N000.0 0 000NN.0 0000000 0NO0N0 0NO0N.0 0 0000N0 00 0NN00 0020000 0020000 0 NO0N0 0000000 0000 2 .0 0000 2 .0 0 0N20N.0 0000200 0200000 0200000 0 0000N0 0002000 N020000 N020000 0 0NNON.0 N002 20.0 000NN00 000NN00 0 0000N0 0000N00 0N00000 0N00000 0 000 0N0 0N000000 0 0NN200 00NN200 0 0000 2 .0 00NON00 0000N00 0000N00 0 0N20N.0 2N200000 00 0000.0 00000000 N NN0000 20N0000.0 0000000 00000000 N 20200 .0 NON000000 002 2000 20022000 2 0000 2 .0 0NO0N000 2NNN000 00NNN000 2 2. .2 N .. 2 2. .2 N a 2 00002 2...... 522322.252 .32... 5.5 2.5: 00002 v.80 3223002252 29002.2 3.202 2222.250 72552.2. 169 - 1: ..1-:2: £3200 0.22:...2. 0W - - 2:1: 22...... «fishy I]! 000020000000 n 0200 N000N0000000 n 0200 0000000 n o _ NO0N000 u .2 2 0000000 u .u 0N0000.0 u o _ 02000.0 n .2 _ 2N00000 u u 202 2 20 0000200 00NON00 00NON00 02 00020.0 00N200.0 0000000 0000000 02 0020000 0N00200 0000200 0000200 0 0000N0 N00000 2000N00 2000N00 0 000N2.0 00N2 200 0000200 0000200 0 2NN000 N000N0.0 0020200 0020200 0 000 2 .0 00000000 202 2 20.0 202 2 20.0 0 000200 002N0000 0000000 0000000 0 0000N0 0 2000000 00N000.0 00N000.0 0 000000 00000000 0N00000 00N00000 0 00N02.0 02N00000 N000000 02000000 0 0000N.0 00N00000 0NON000 0NON000 0 0 2002 .0 00200000 0000000 00000000 0 020000 0000N000 0022000 20022000 0 00 0000.0 N0 0000.0 00NN00.0 000NN000 0 00000.0 N00000000 0000000 N 2 0000000 0 00000.0 NONON000 0022000 20022000 N 200 2 .0 200000000 N0 2000.0 0 2 N0 20000 N 0000 2 .0 000 2 N000 0N00000 000N00000 2 000N.0 000600000 000-0N0. 2 00060020. 2 2 2. .2 N 2. 2 2. .2 N 2. 2 00002 v.30 neg—93.9.2345? 82> @002 ".390 52295252252 0.5.2.2 922.202 0N0020000000 n 0200 0000N0000000 n 0200 5.22... n o 2 22%.... u .2 _ 22.2.... n 0 0.28... n o 2 0.8.... u .2 2 2.2.50.2... u a N2 2N2.0 0000000 02N000.0 02N000.0 02 0002000 0002000 NON000 NON00.0 02 N020N0 02000.0 0000000 0000000 0 000020 000080 0200000 0200000 0 200020 0 2 0000.0 02NN00.0 02NN00.0 0 0000000 0NN2 00.0 02 2000.0 02 2000.0 0 000N2.0 0002000 2008.0 0008.0 0 0000N0 02020.0 0000N00 00008.0 0 200NN.0 0000N00 0N00200 0N00200 0 200000 0000200 0008.0 0008.0 0 000N.0 0000200 0002 20.0 0002 20.0 0 000000 0000200 N020200 N020200 0 0 2 00 2 .0 0000200 0000000 20000000 0 000 2 .0 NO0N200 0000000 00000000 0 00 2 0N0 0NON0000 000 200.0 2 000 2 00.0 0 0000N0 00000000 0N00000 00N00000 0 0000N00 0020N000 0200000 000200000 N 020020 00000000 0202000 0202000 N N000000 20022000 NNN0000 002NN0000 2 0N2 2.0 N000 200.0 00N000.0 N000N0000 2 2. .2 N .. 2 0 .2 N .. 2 00002 2.30 322235.552 .52... :80 2.8: 00002 "29.0 Egg—2.9.2.222 3002.2 3202 22.0.2.0 02720. 1220.222. 170 «.2707... :2. wmomNoooooood H Mmm ochSooooood H mmm Kmoadno— Svtodunfl mNOONoduw 3856"”; $306”: mNcoSdnw Nwomwod homomood NYmoSd NVmoSd 3 >536 NfiNoWoood oiNood oCNood S @326 memNood Noomfiod Noode a «$36 @8886 KmSod 55286 a £386 mmw H mood chNSd NmoNSd w E :Ld mmONoood 9NSod ENSOd w mwNnvod cVNomood 3386 3386 n mmmawod 3. ~ 38.0 2823 _ aged 5 smoovod _vonood NONSod NONSod c voncod chNmoood ovooood Sovooood o aRNNd $0223 2386 238.0 m Swfld 83886 $386 wSVoood m Rammed mSNood umomood umomood v mmmmd moobcvfluc :Noood oooRoood v NmNVNd oVNN Sod owe fl cod ace _ cod m n SNNd moo-ooom.m cm 886 cm 58.0 m N33 M .o 2 Gooood oomocod oomoocd N N g veto moobw _ WV moobmws moobNomwN N moNEd @8886 N586 SoNSood _ mamde mcobNNVNé mooéva mooémogN _ w a N x a a a N a « Amwa use» nous—5.5:: 9.08:0 32> A53 tau» 5.8—695:: 2.25 2532 2530886 u mmm 23588.0 n mmm 82: u 0 # N326 u 9 fl 382 u a 833 u o _ ”8086 n n _ @323 n a amid onoood mfimvvod 233.0 2 $036 Ewmmood $286 inmod S 9626 $8306 5986 5986 a 388.0 moonmood 3386 3385 o wmwm _ .o moommood $0086 $336 w mvom fl .o NNM mood o _ ONNod o _ CNNod w wmfi _.o NVNfimood $33.0 $325 w ooVNfio 32806 30206 32.56 N. nmNMNd \(NooNood Nwmiod Nwmiod o nmmfld flogged aNcde oNode o 3%: .o mwvoSod 3856 58:: m NNflNd ammoNood 3386 22.386 m MNmoNd VOVNNood mvcoood ochoood v mnmm _ .o 88890 3886 233686 v QOONNd wwcoNood ooomood Goomood m _moEd ancNood $386 33256 m mmmNfio Smtood 8385 $8256 N NoNSd wonSod $050.9 $386 N omcoNd omvomoood wmmoocd 52386 H GmCd mvoonoood vvvoood $3386 _ v a N u a a q u u a 033 :3» E53683: .32: :50 0:5: Amwa £3» accuscchuu—v 3?..— 35 «23:30 «Sm—75h 171 .a ..stS Pfiw 2356656666 H mmm 335665666 u mmm 069.56 M u _ N5m566 u n— _ 3556 u u wNN656 u o _ 559666 N n _ 36666 H m 5636 Nm56Noo6 $556 3856 6. Nmmmm6 w_NNm666 w #5366 35366 A: m _ N5 H 6 oNv _ 666 3356 66:56 6 353.6 566-3 535666 535666 a 033 6 566566 36656 36656 6 3856 68.2 553666 5563666 6 3936 653566 vmm56 $366 5 N536 3556666 35566 $5566 5 NN5N6 5on566 N52 56 N52 56 o w _ m5N6 56366666 6566566 36566 o 3:26 N5N5 566 £5666 3566 m m5N~ #6 5556666 mNS 566 mNE 566 m mOmNm6 ©3566 5386 5386 v $686 066-3 6555666 NN65m6666 v 5:56 23566 Nm _ 566 Nm 6 566 m cmeo6 666-2 535666 33556 N 6N6m56 65656666 33666 53666 N 3&6 55066 5N55666 355666 N 5Nmmm6 6356666 35666 535666 6 3 ~5m6 666-2 666-0mN66.m 566-3 _ 66.m _ u n N x a u a n a . Anna Each nag—9:“.aazfiau 32> Gwyn» guns—5.5:: 0:25 2362 36656666666 6" umm 56m 366666666 n mwm 3%: u o _ 283 u n _ $83 u a 588.0 u o 6 332 u a _ 332 n a mmVVN6 6.886666 25656 25656 2 5mm_ #6 836666 2866 _5666 S owmoN6 NE5N666 mmN66 mmN66 6 Nm6m666 5_ 56 V5mm566 V5mm566 o 55mN666 w 3566 N3556 N35 56 w mNE #6 35 5666 am: 566 am: 566 m 3me6 556566 52 56 52 56 5 @336 55_N56 5856 5556 5 SNwm6 65566 6565666 635666 0 ©3656 mN: 56 55556 55556 0 meN6 555566 @6566 633666 m 6536 6 365666 5cN5No6 5cN5N66 m 5836 335666 665N666 665N666 v VNm5 _ 6 069.566 «665 56 6665 56 v 5956 25N566 53566 53566 m mvvom6 636566 636666 656666 m 0Nwa6 mooéoNcoN 556666 N656666 N 3836 56566 NNv566 NNv566 N 536 56-0563” $5666 $5666 _ 5236 5665566 56566 35566 H a n s x a u a u x . A563 tau» 52269.25 58.: :80 9:8: 353 is?» 52269.26 Baa—m 35 «23:50 >EDP 172 . .. .I‘l.\ llfi~|II~A~ o‘i...‘ ‘n‘.‘--~‘~. ‘\ -ECCCZfia. «35:; ~ w 3566666666 n Mmm 3366666666 n umm $38 u o _ 35:56 u a _ @280 u a :83 u 06 283 u a 6 $83 u .3 566666 vm5Nm666 6566666 6566666 66 6 6N6 66 86366 566666 55666 66 666N66 55566.6 666 666.6 666 6m66 6 6Nco6 .6 N56666 N666N6 N666N6 6 56666 663666 666656 6666666 6 mV6N66 665666 68666 $6666 6 66N666 N6N66V666 665666 6666666 5 66666 666266 m566 656.6 5 666Nm .6 6NmVN66 566NN66 566NN66 6 6565 6 .6 6 66N56 6566566 656%566 o 6voNN6 66666666 66556 666266 6 65¢5N6 656N666 66656 $656 6 666666 66666666 N666666 66666666 v 6656N6 6666666 NVN¢666 NVNv666 v 666 666 6665 56.6 NN6m666 NNNmm666 6 566 6 .6 66665666 66N556 66N5 56 6 6669.6 N66m 566 65NN666 665NN666 N 656NN6 66N66666 6665666 66665666 N 636626 66566666 N65666 66866666 6 «656 365566 66NN666 66NN666 6 6. .6 N x 6 6. 66 N x 6 A6566 ".635 5669:696456 «hum—flu 82> €666 Elan» 656335.556 2.26.6 066.6662 c66m 566666 N mmm 6636666666 u mmm 86836 n o _ 68623 n a _ 66385 n .u 833 u 06 833° n a _ 223 u .u mmmmN6 666N66 No66m6 N666m6 66 66N66 6 65N66 665566 665566 66 6566N6 Nv5mo6 m 656666 m 65666 6 6.666VN6 6666666 666666 666666 6 mmmm66.6 66N6 6.6 666666 36666 6 mmN666 6666566 66656 66mvN6 6 666666 5666666 56566.6 56566.6 5 6 5666 6666666 5mNo66 5mNo66 5 v 6cm 6 .6 6N6mmo6 66N¢N.6 66N¢N6 o 666 6 N6 NVN6 666 63666 $666 6 63666 $356 @6566 636566 m 6663.6 v6m6666 6656666 665566 6 6652.6 6N66N66 6656 66 6656 66 v cmmvflo 65666666 36666 6666666 v 5666~N6 666 6N66 6665666 6665066 6 666566 6656366 9666666 26656 6 665666 666556 v66N666 666N666 N 656666 666N666 5656666 565666 N Nm66m6 666656 366666 5866666 6 66N66 65666666 6666666 6N6666666 6 6 66 N a 6 6. .6 N x 6 $666 "63>. 6566:6563: .36: :50 0:866 @666 ".596 5.6835556 6962.6 25 62:66:60 EDA—U766! £9575 173 ‘1'..i|l1lll J It'll. III 2.3:. :15: Ennis-62:62: 5.5.55 92...: s. Illlhflllfllllurl :25: t . 6 .fl - I - I VFW C5 @136 «x ,3: 76.6.51)". 6. .1. «tsN‘Nka N 6 6556 566666 N mmw 666566666666 H mmm 665666 n o _ 666666 n 6 6 mN6N66 n u. 66Nv66 n o _ 665N6 u 6 6 66366 n .u 666 6N6 656666 56636 66636 66 v6 66N6 666666 6666666 66666666 66 NmoNN6 6665566 6Nm666 6Nm666 6 5666N6 66666.6 6 6 6N6 6 6 6N6 6 6N6666 56N5666 mmN6N6 mmN6N6 6 666N6 966666 666666 666666 6 666666 6665666 6566N6 6566N6 5 ~65366 6666N66 5mN66 5mN66 5 6566 66 666656 566666 566666 6 66666 V5N6.6 666N666 66N666 6 666666 6mN6666 6N6666 6N6666 m meNm6 666556 666N666 N66N666 m 665N6 665m 6 6.6 VN66666 6N66666 v 666666 66 556 6656666 6656666 v 65636 N66656 636666 636666 6 665666 5656566 656656 656656 6 6666N6 665366 666656 666656 N 6666N6 6366666 6665666 66665666 N 6V5VN6 66565666 665666 66665666 6 566566 6666N666 666N666 66m6N666 6 66 66 N a 6 66 a N a 6 A6666 ".396 neg—9:656:66 6:68.60 82> A6666 tau» 686695656656 25.6.6 066.6656 N6N666666666 u Mmm 6 66666666666 n mwm $22 u i 2866 u 3 $8.66 N a SM u a Emma u i 28.6 n .3 Nm6VN6 666666 666666 666666 66 66 6666N6 6665666 3636 3636 6 666566 6N5m66 665656 665656 6 N6Nv66 6666666 666666 666666 6 NN6666 Nv65v6 666666 666666 6 666666 6666666 Nm5m66 Nm5m66 5 mV66N6 6666N6 666666 66666666 5 666566 .6 6N66 666 56N5666 56N5666 6 6566N6 N666666 m5Nm6 m5Nm6 6 666666 6v6N56 6666666 6666666 6 555666 N666666 656N6 656N6 6 665666 666656 6Nm6666 6Nm6666 v 66666 .6 665N666 6666 66 6666 66 66 5666 6 .6 666N566 6666N66 6666N66 6 N6 65N6 66N6N66 666666 666666 6 N663 .6 66666666 6566666 66566666 N 665666 6665666 N66656 N66656 N 666N666 N6656666 6566666 66566666 6 6666666 6666566 666666 66666666 6 6 m N x 6 66 .6 N x 6 A6666 £35 636665668656 .56.: :80 2.566 66666 r596 noun—666666.666 6936.6 3.5 62:66:60 «GEM—>2 m0 £6.5de CHE—Z66 174 APPENDL‘ p and q Ow APPENDIX B p and q Over Time for Each Country and Bach Product 175 pf0r( 0.8 0.6 P 0.4 01 0.8 0.6 p 0.4 0‘: p for CD Player in Argentina 0.8 r P 0.4 « 0.2 d O-W 12345678910 time __ _- __ ~J p for Home Computer in Argentim 1 0.8 4 0.6 A P 0.4 4 0.2 1 0 .. 12345678910 time p for Mobile Phone in Argentina 0.8 « 0.6 w P 0.4 q 0.2 J 12345678910 time p for Video Camera in Argentina 1 MN 0.64 90.44 0.2“ 0 v.¢y¢,¢r¢T¢I¢'¢r¢yv 1 2 3 4 5 6 7 8 910 time 176 T -_ q for CD Player in Argentina 0.8 4 0.6 A q 0.4 ~ 0.2 « 12345678910 q for Home Computer in Argentina 1 0.8 + 0.6 4 q 0.4 4 0.2 4 W Orrrr 12345678910 time —_ __ __666 ...66 q for Mobile Phone in Argentina 0.8 ~ 0.6 ~ ‘1 0.4 « 0.24M 12345678910 time q for Video Camera in Argentina 12345678910 time p for CD Player in Australia 1 0.8 4 0.6 4 P 0.4 ~ 0.2 4 O~M 12345678910 time p for Home Computer in Australia 1 0.8 4 0.6 4 P 0.4 - 0.2 4 0 - 12345678910 time p for Mobile Phone in Australia I 0.8 4 0.6 4 P 0.4 « 0.24 : 04 . I2345673910 time p for Video Camera in Australia 1 0.8 4 0.6 4 P 0.4 « 0.2 4 0 .. 12345678910 177 ‘1 for CD Player in Australia 0.8 4 0.6 4 q 0.4 « 0.2 1 7 I l 12345678910 """° 1 q for Home Computer in Australia 0.8 4 0.6 4 ‘1 0.4 « 0.2 ~ 12345678910 time __ . CJ q for Mobile Phone in Australia 12345678910 time q for Video Camera in Australia "—1 0.8 . 0.6 - ‘1 0.4 « p for 0.8 0.6 P 0.4 0.2 0.8 0.6 P 0.4 0.2 p for CD Player in Austria l 0.84 0.64 P0.4~ 0.24 0‘ r 1 u r 1 2 3 4 5 6 7 8 9 10 time pforl-lome Computerin Austria I 0.84 0.64 Po.4« 0.24 o ‘ fi T f r 1 2 3 4 5 6 7 8 9 10 time pforMobilePhonein Austria I 0.84 0.64 p044 0.24 0.. 12345678910 time p for Video Camera in Austria 1 0.8 4 0.6 4 P 0.4 . 0.2 4 0 .. 12345678910 time 178 q for CD Player in Austria 0.8 w 0.6 ~ ‘1 0.4 « O-Z‘W 12345678910 time q for Home Computer in Austria 0.8 ~ 0.6 ~ q 0.4 « 0.24W 01....fi,,r 12345678910 time “_J 1 q for Mobile Phone in Austria 0.8 « 0. ~ ‘1 0.4 ~ 0.2 1 12345678910 time q for Video Camera in 1 Austria 0.84 0.64 q0.44 0.249% 12 3 4 5 6 7 8 910 time pfou 0.8 0.6 p 0.4 0.2 f____ _. p forCD Playerin Belgium 1 0.84 0.64 P044 0.24 0‘ V I T f 1 2 3 4 5 6 7 8 9 10 time p for Home Computer in Belgium 1 0.8 4 0.6 4 PM ~ 0.2 4 0 d 1 l u 1 l T 12345678910 T 0.8 4 q 0.4 ~ 0.2 - q for CD Player in Belgium 12345678910 time 0.8 4 0.6 4 q 0'4 d W 0.2 4 0 q for Home Computer in Belgium Tfifttfi'ji 12345678910 time 0.8 4 ‘1 0.4 « 0.2 . q for Mobile Phone in Belgium 12345678910 time time pforMobilePhonein Belgium 1 0.84 0.64 P0.4~ 0.24 0 ¢.:.¢.:,¢.:.¢,¢,¢# l 2 3 4 5 6 7 8 9 10 time p for Video Camera in Belgium 1 0.84 0.64 Po.4« 0.24 0.. 12345678910 time *-—_fi .2 q for Video Camera in Belgium 0.8 % 0.6 ‘l 0.4 - 0.2 ~ 12345678910 time P fu 0.8 (15 [1 0‘1 p for CD Player in Bran'l 0.84 0.64 P0.44 0.24 o ¢.:.¢j:,:,$.¢,¢.:.¢ 1 2 3 4 5 6 7 8 9 10 time p for Home Computer in Brad] 1 0.84 0.64 Po.44 0.2j 0 l 2 3 4 5 6 7 8 9 10 time p for Mobile Phone in Braa'l 1 0.84 0.64 Po.44 0.24 04% 12345678910 time pforVideo CamerainBran'l I 0.84 0.64 Po.44 0.24 0 ¢rcu¢ucu¢u¢u:1¢l¢u¢ 12345678910 time 180 q for CD Player in Brau'l “P (1 0.8 4 0.6 4 q 0.4 4 0.2 4 123456789 time 10 q for Home Computer in Bran'l 0.8 4 0.6 4 q 0.4 4 0.2 4 time 12345678910 q for Mobile Phone in Braa'l 12345678910 0.24 0 I r time _.J q for Video Camera in Brazil 1 0.8 4 0.6 4 ‘1 0.4 4 12345678910 time 0.3 0.6 p 0.4 0.: 0.8 4 0.6 4 P 0.4 4 0.2 4 p for CD Player in Canada 1 04% 12345678910 -—u 0.8 4 0.6 4 P 0.4 4 p for Home Computer in Canada 12345678910 time 0.6 P 0.4 0.2 0.8 4 p for Mobile Phone in Canada 1 1W 12345678910 time p for Video Camera in Canada 1 0.8 4 0.6 4 P 0.4 4 0.2 4 0 - 12345678910 time 181 q for CD Player in Canada 0.8 4 0.6 4 0.4 4 0.2 W 0 4 r q 12345678910 time I..— F'— _ q for Home Computer in Canada 0.8 4 0.6 4 q 0.4 4 0.2 12345678910 time q for Mobile Phone in Canada 0.8 4 0.6 4 q 0.4 4 0.2 4 12345678910 time q for Video Camera in Canada 0.84 0.64 q0.44 o-Z‘W 0 4 . , , 12345678910 l I time _J pfo 0.8 0.6 ‘ p 0.4 0.2 0.8 0.6 p 0.4 0.2 p for CD Player in Chile 0.84 0.64 P0.44 0.24 0 V1¢T¢r¢u¢r:r¢1:1:4 1 2 3 4 5 6 7 8 9 10 time p for Home Compurter in Chile 1 0.84 0.64 Po.44 0.24 04W 12345678910 time pfor Mobile Phone in Chile 1 0.84 0.64 Po.44 0.24 0.. 1 2 3 4 5 6 7 8 9 10 time __J pforVideo CamerainChile 1 0.84 0.64 Po.44 0.24 o er:4:.¢.:.¢.:.¢.¢.¢ 12345678910 time 182 q for CD Player in Chile 0.8 4 0.6 4 9 0.4 4 0.2 4 12345678910 time q for Home Computer in Chile 0.8 4 0.6 4 q 0.4 4 0.2 4 12345678910 time q for Mobile Phone in Chile 0.6 4 q 0.4 4 12345678910 time q for Video Camera in Chile 0.8 4 0.6 4 ‘1 0.4 4 0.2 4 time p for CD Player in China -—-— *1 l 0.84 0.6 Po.44 0.24 O ¢r¢1¢.¢,¢.¢T¢I¢,¢I¢ l 2 3 4 5 6 7 8 9 10 time p for Home Computer in China 1 0.84 0.64 Po.44 0.24 ...4 0‘ T I I 12345678910 time ____ _J p for Mobile Phone in China 1 0.84 0.64 Po.44 0.24 o :.:.¢.:.:.¢.:.¢,:,¢ 1 2 3 4 5 6 7 8 910 time pforVideo CamerainChina l 0.84 0.64 Po.44 0.24 o :.:.¢.¢.:.¢.:.c.¢,: 1 2 3 4 5 6 7 8 910 m _I 183 q for CD Player in China 0.8 4 0.6 4 ‘l 04 - 0.2 12345678910 q for Home Computer in Chrna I 0.8 4 0.6 4 q 0.4 4 0.2 0 T I I I I I I I 12345678910 time q for Mobile Phone inChina 12345678910 time qforVideo CamerainChina 1 0.84 0.64 90.44 0.24 0 I I I I I I I I I I 2 3 4 5 6 7 8 910 time p101 0.8 0.6 p 0.4 0.2 [__WA _H... -4. .. "'7“ 1 p for CD Player in Denmark T 1 12345678910 time F” __,_ _,___._ p for Home Computer in Denmark I 0.8 4 0.6 4 P 0.4 4 0.2 4 04M 12345678910 time p for Mobile Phone in Denmark 1 0.8 4 0.6 4 PM 4 0.2 4 0 4W 12345678910 time p for Video Camera in Dennsrk I 0.8 4 0.6 4 P 0.4 4 0.2 j 0 12345678910 184 q for CD Player in Denmark 1 0.84 0.64 q0.44 0.24 0 I I I I I a I I I l 2 3 4 5 6 7 8 9 10 time q for Home Computer in Denmark 1 0.84 0.64 (”PPM 0.24 0 I I 7 T I I r I I— 1 2 3 4 5 6 7 8 9 10 time ___ __J qfor Mobile Phone in Dennnurk 1 0.84 0.64 q0.44 024W 0 . 4 4 T I . . r 1 2 3 4 5 6 7 8 9 10 time q for Video Camera in Denmark 1 0.84 0.64 q0.44 024W 0 I I I I I I I I 7 1 2 3 4 5 6 7 8 9 10 time p for CD Player in Egypt 0.8 4 0.6 4 P 0.4 4 0.2 4 04W 12345678910 time p for Home Computer in Egypt 1 0.84 0.64 Po.44 0.24 0 ¢.¢.:,¢ t #1:; :,+ 1 2 3 4 5 6 7 8 910 time pforMobile PhoneinEgypt 1 0.84 0.64 Po.44 0.2 0 :':'¢I:T:T:T¢I$I:I: 1 2 3 4 5 6 7 8 910 time pforVideo CamerainEgypt 1 0.84 0.64 Po.44 0.24 0 :I:I:I¢I:I{7:Tfi::¢ 1 2 3 4 5 6 7 8 910 time ‘m‘—‘— __J 185 qforCD Playerin Egypt I 0.84 0.64 ‘10.44 0.24 0 1 fi' I f I 1 2 3 4 5 6 7 8 9 10 time qforHome Computerin Egypt 1 0.84 0.64 (10.44 0.24 0 I I 7 ' 1 I I I I 1 2 3 4 5 6 7 8 9 10 time qforMobile PhoneinEgypt 12345678910 02‘“ 0 time I q for Video Camera in Egypt I 0.8 4 0.6 4 q 0.4 4 12345678910 time pfu 0.8 0.6 P 0.4 0.: p for CD Player in Finland 1 0.8 4 0.6 4 PM 4 0.2 4 o M 12345678910 time L———-—..—-—_—— —. ___ _J F— a p for Home Computer in Finland 0.8 4 0.6 4 p04 4 0.2 4 o 4% 12345678910 time _ _.-__. ___.__.J p for Mobile Phone in Finland 0.8 4 1 0.6 4 P 0.4 4 12345678910 time p for Video Camera in F'mland l 0.8 4 6 0. 4 PM 4 0.2 4 o .. 12345678910 186 q for CD Player in Finland I 0.8 4 0.6 4 ‘1 0.4 4 0.2 4 0 fi T fi 1 l 1 12345678910 q for Home Computer in Finland 0.8 4 0.6 4 (10.4 A W 0.2 0 4444 12345678910 time l "‘7 q for Mobile Phone in Finland 0.8 4 0.6 4 ‘l 0.4 4 0.2 4 1 r T 12345678910 time q for Video Camera in Finland 0.6 4 ‘1 0.4 4 0.2 4 0 T I I I I I T 12345678910 time ___ ____ ___, ,fifi,,_ J MI 0.8 0.6 P 0.4 0.: p1 0.3 0.6 P 0.4 0.2 p for CD Player in France 0:2 4 0 W 12345678910 time p for Home Computer in France 0.8 4 0.6 4 P 0.4 4 0.2 4 0 444M 12345678910 time pforMobile Phone inFuance I 0.84 0.64 p044 0.24 0 ¢I¢I313I¢I¢I¢I¢I¢IT¢ l 2 3 4 5 6 7 8 910 time p for Video Camera in France 1 0.84 0.64 P04. 0.24 o :.¢.¢.:.¢.:.¢,¢.¢,¢ 1 2 3 4 5 6 7 8 910 time 187 I” __ q for CD Player in France 1 0.8 4 0.6 4 q 0.4 4 0.2 4 0 7 I I l ‘1 I l 7 l 12345678910 time q for Home 'Computer in France I 0.8 4 0.6 4 q 0'4 J W 0.2 4 0 I r I r 1 fi T f 7 12345678910 time [ q for Mobile Phone in France 0.84 0.64 q0.44 0.24 O I l 1 l l l T 1 1 1 2 3 4 5 6 7 8 9 10 time “ __;: qforVideo Camerain France I 0.84 0.64 90.44 024% 0 l T T l V l I I I 1 2 3 4 5 6 7 8 9 10 time 0.3 0.6 p 0.4 0.2 0.8 (If) p 0.4 0.4 s _ ‘1 p for CD Player in Germany 1 0.8 4 0.6 4 Po.4 4 0.2 4 0 ' I T I 1 T I 2 3 4 5 6 7 8 9 10 time fi_ “ __iJ p for Home Compurterin Germany 1 0.8 4 0.64 Po.4 4 0.24 1 0 2 12345678910 time E: — M: p for Mobile Phone in Germany 1 0.8 4 0.6 4 Po.4 4 0.2 4 0 a 1 2 3 4 5 6 7 8 9 10 time p for Video Camera in Germany 1 0.8 4 0.6 4 Po.4 4 0.2 4 0 4 12345678910 time ___._ __.._._,_f___._ ...___ -3”. 188 q for CD Player in Genmny 0.8 4 0.6 4 q 0'4 4 M 0.2 4 12345678910 time q for Home Computer in Germany I 0.8 4 0.6 4 q 0.4 4 0.2 4 0 ‘ I 12345678910 q for Mobile Phone in Germny 0.8 4 0.6 4 q 0.4 O'ZTW 12345678910 time q for Video Camera in Germany 0.8 4 0.6 4 9 0.4 4 0.2 4 12345678910 time P for CD Player in Greece 0.8 4 0.6 4 P 0.4 4 0.2 4 04W 12345678910 I' p for Home Computer in Greece 1 0.8 4 0.6 4 P 0.4 4 0.2 4 0 - 12345678910 time __"__-_ “ 44444 p for Mobile Phone in Greece 1 0.8 4 0.6 4 P 0.4 4 0% 4 : I l 2 3 4 5 6 7 8 9 10 time p for Video Camera in Greece 0.8 4 0.6 4 P 0.4 4 41 o o o o 4» No mo #0 (It 0‘ \l co ND 5 E‘ 189 q for CD Player in Greece l2345678910 0.24% 0 I I , , 12345678910 time time q for Home Computer in Greece I 0.8 4 0.6 4 9 0.4 4 .__1 q for Mobile Phone in Greece 0.8 4 0.6 4 q 0.4 4 0.2 12345678910 time [wfi - .-_. q for Video Camera in Greece 0.8 4 0.6 4 ‘1 0.4 4 0.24% 0 12345678910 "T~i».—.i._.__-__-_r ._—_ .—. __———‘ p for CD Player in Hong Kong 1 0.84 0.64 Po.44 0.24 044W 1 2 3 4 5 6 7 8 9 10 time p for Home Computer in HongKong 1 0.84 0.64 Po.44 0.24 02 1 2 3 4 5 6 7 8 9 10 time pforMobile Phonein HongKong l 0.84 0.64 Po.44 0.24 04W 1 2 3 4 5 6 7 8 9 10 time __ -___._J p for Video Camera in HongKong 1 0.84 0.64 Po.44 0.24 0.. 1 2 3 4 5 6 7 8 9 10 time 190 q for CD Player in Hong Kong 1 0.8 4 0.6 4 q 0.4 4 02 4 W. 0 T T T r 1 l l l T 1 2 3 4 5 6 7 8 9 10 time q for Home Compurter in Hong Kong I 0.8 4 0.6 4 ‘l 0.4 4 0.2 ‘ W 0 T fl T T *r l l l l 12345678910 time q for Mobile Phone in Hong Kong 1 0.8 4 0.6 4 qOA ‘ W 0.2 4 0 f l l r j l 1 Y 1 I 2 3 4 5 6 7 8 9 10 time q for Video Camera in Hong Kong I 0.8 4 0.6 4 (10.4 4 0.2 " M 0 I T I I T f T T T 12345678910 time pfu 0.8 (1.6 P 0.4 0.: o 0. 0. 0- p for CD Player in Hungary 1 0.8 4 0.6 4 P 0.4 4 0.2 4 0 4W 12345678910 p for Home Computer in Hungary 1 0.8 4 0.6 4 P 0.4 4 0.2 4 0 4% 12345678910 time p for Mobile Phone in Hungary 0.8 4 P 0.4 4 0.2 4 2 0 3 12345678910 time pforVideo Camerain Hungary I 0.84 0.64 Po.44 0.24 o :.:.:¢.c.¢.:.¢.c.¢ 12345678910 time 191 -_-_ 1 q for CD Player in Hungary 1 0.8 4 0.64 q0.44 0.24 0 I l l T 1T Ifi f 12 3 4 5 6 7 8 910 time q for Home Computer in Hungary 1 0.8 4 O 6 4 ‘1 0:4 4 0.2 4 0 I I I I I I I I I 12345678910 time q for Mobile Phone in Humgary 12345678910 time __I q for Video Camera in Hungary I 0.8 4 ‘18.: 1 0.24% 0 r I 12345678910 time p for CD Player in India 0.8 4 P 0.4 4 0.2 ~4 p for Home Computer in Indra l 0.8 4 0.6 4 P 0.4 4 0.2 4 O 6 r I p for Mobile Phone in India 0.8 4 0.6 4 P 0.4 4 0.2 4 AA A A LA A IVYVWYVI'IVTTV 345678910 A p for Video Camera in India 0.6 4 P 0.4 4 192 ~_H q for CD Player in India 0.8 4 0.6 4 q0.4 4 02 4 4% 12345678910 time 44 fl :_ q for Home Computer in India 1 0.8 4 0.6 4 ‘1 0.4 4 0.2 M 0 I I 1 ' ' I I I T 1 2 3 4 5 6 7 8 9 10 time q for Mobile Phone in India I _ 0.8 0.6 q 0.4 0.2 0 12345678910 time q for Video Camera in India I 0.8 4 0.6 4 q 0.4 4 0.2 4 0 I . . 12345678910 r w __; 4 ___._ m... _‘ p for CD Player in Indonesia I 0.8- 0.64 PM4 0.24 0 ¢,¢,¢.¢T¢I¢.%I¢I¢r¢ l 2 3 4 5 6 7 8 9 IO time L p for Home Computer in Indonesia 1 0.84 0.64 p044 0.24 04% l 2 3 4 5 6 7 8 9 l0 time pforMobilePhonein Indonesia 1 0.84 0.64 Po.44 0.24 0 Aivlvlvl¢I;I:F;T¢I$ l 2 3 4 5 6 7 8 9 l0 time p for Video Camera in Indonesia I 0.84 0.64 Po.44 0.24 o ¢,¢,¢.:.¢.:.:,:I:,~ l 2 3 4 5 6 7 8 9 l0 L l' 193 q for CD Player in Indonesia I 0.8 4 0.6 4 9M4 0.24% 0 123456789l0 time q for Home Computer in Indonesia 1 0.8 4 0 6 4 90:44 0.24 0 I I I WI¢I¢I¢I¢I l23456789l0 time q for Mobile Phone in Indonesia 0.8 ~ 0.6 4 90.44 0.24 12345678910 time q for Video Camera in Indonesia .3: 9999 archaeo— 44L; 123456789l0 9.9 A ‘ ’u A ,.___...___ p for CD Player in Ireland I 0.3 J 0.6 J Po.4 4 0.2 4 o --o—.+I—o-.+.-o-.4-I+I-o-w4=#— l23456789lO time p for Home Computer in Ireland 1 0.8 4 0.6 4 P 0.4 4 0.2 4 o 4% l23456789lO time pforMobilePhonein Ireland I 0.84 0.64 p041 0.24 0- l 2 3 4 5 6 7 8 9 10 time p for Video Camera in Ireland I 0.84 0.64 Po.44 0.24 o ¢,:rh91949164¢'¢.¢ 12345678910 time q for CD Player in Ireland 0.8 4 0.6 4 q 0.4 4 0.2 1 W 12345678910 time q for Home Computer in Ireland I 0.8 4 0.6 4 90.4 4 0.2 _ m 0 I I I I I r T I I l 2 3 4 5 6 7 8 9 l0 time q for Mobile Phone in Ireland I 0.8 4 0.6 4 ‘1 0.4 4 0.2 I W 0 I I I I I I I I I 12345678910 time q for Video Camera in Ireland 0.8 4 0.6 4 90.44 0-2‘W 12 3 456 7 8 910 time 194 p for CD Player in Israel 084 0.6 4 P 0.4 4 024 Ouow+w¢¥fiwfiwfi¥+¥fi¥‘¥‘4 l 2 3 4 5 6 7 8 9l0 time p for Home Computer in Israel 1 081 064 P 0.4 4 024 0“*q‘*‘*fi*fiy.$fi;t;t:1j l 2 3 4 5 6 7 8 9 m time p for Mobile Phone in Israel I 084 Q64 FOAI Q24 0- l 2 3 4 5 6 7 8 910 time pforVideo Camerain Israel I 084 064 Po.44 024 o ¢4¢,¢,:r:.¢.:4¢r¢r: I 2 3 4 5 6 7 8 9 I0 time 195 [__V-_ q for CD Player in Israel I 084 064 ‘1 0.4 4 024 0 I I I I j I I I I l 2 3 4 5 6 7 8 9l0 time __J q for Home Computer in Israel I 084 064 q044 024 0 I I 7 I I I If If I l 2 3 4 5 6 7 8 910 "m. e J ‘1 0.4 0.2 084 ._.*.—‘ q for Mobile Phone in Israel l23456789l0 time q for Video Camera in Israel l 2 3 4 S 6 7 8 9 m p for CD Player in Italy 0.8 4 0.6 4 p 0.4 4 0.2 4 12345678910 time 0 4W4 7—___ __v-_.. p for Home Computer in Italy I 0.8 4 0.6 4 Po.4 4 0.2 I 04 44, I23456789Io time p for Mobile Phone in Italy 1 0.84 0.64 Po.44 0.24 0 :1:I¢I¢,¢,¢,:,:T:,¢ l 2 3 4 5 6 7 8 910 time p for Video Camera in Italy I 0.84 0.64 Po.44 0.2 o ¢I:IVLI:I¢I:I$I:T:1¢ l 2 3 4 5 6 7 8 9 IO time 196 qforCDPlayerinItaly l 0.8 4 0.64 (10.44 0.24 0 T f I I I I I I 1 l 2 3 4 5 6 7 8 9 l0 time q for Home Computer in Italy I 084 0.64 90.44 0.24 0 I I I I I I I T I I 2 3 4 5 6 7 8 9 10 time _._ _-_.-- __ ___J qforMobile Phonein Italy I 0.84 0.64 90.44W 0.24 0 I I I I I I I T r l 2 3 4 5 6 7 8 9 10 time qforVideo Camerain Italy I 0.84 0.64 90.44 O-Z‘W 0 I I I I I I I I I 12345678910 time p for CD Player in Japan 0.8 4 Po.4 4 0.2 4 ' ./ O ‘ I T I I I I23456789l0 time p for Home Computer in Japan I 0.8 4 0.6 4 P 0.4 4 0.2 4 O - l23456789l0 time p for Mobile Phone in Japan I 0.8 4 0.6 4 P 0.4 4 0.2 4 O 4% l23456789lO time p for Video Camera in Japan I 0.8 4 0.6 4 P 0.4 4 0.2 4 0 A 123456789l0 time .__._,—___ ._ _ .-__ ____ 197 q for CD Player in Japan I 0.8 4 0.6 4 q 0'4 A W 0.2 4 0 I I I I I I I I r I 2 3 4 5 6 7 8 9 10 time _ __I q for Home Computer in Japan I 0.8 4 0.6 q0.4 4 0.2 4 W l2345678910 time q for Mobile Phone in Japan I 0.8 4 l23456789l0 time __ _._I q for Video Camera in Japan I 0.8 4 0.6 4 q 0.4 4 O‘ZW—H 0 I 12345678910 time p for CD Player in Malaysia I 08 4 0.6 4 P 0.4 4 0.2 4 04W 12345678910 time p for Home Computer in Malaysia I 0.8 4 0.6 4 P 0.4 4 04W l23456789lO time _V- _._—_._- p for Mobile Phone in P Malaysia 04W l2345678910 time p for Video Camera in Malaysia 0.8 4 0.6 4 P 0.4 4 04W l23456789l0 time 198 q for CD Player in Malaysia I 0.8 4 0.6 4 ‘1 0.4 4 O-Z‘W 0 I I I I I I I rfi' l23456789l0 time q for Home Computer in Malaysia I 0.8 4 0.6 4 q 0.4 4 024ka 0 I I I I I I I j‘r 12345678910 time q for Mobile Phone in Malaysia 0.8 4 0.6 4 q 0.4 4 0.2 0 I T I I I I I I I l23456789l0 time q for Video Camera in Malaysia 0.8 4 0.6 4 ‘1 0.4 4 0.2 4 o r I I I T f r r I23456789I0 time - -,_____2__ _ _WJ r—— __ ____ ‘_ _._-.441 p for CD Player in Mexico I 0.84 0.64 Po.44 0.24 0 9494:,eyh949491:4¢ I 2 3 4 5 6 7 8 9 10 L I. p for Home Computer in Mexico l.OOE+00 8008-01 4 6.003014 p4.00B-Ol 2.005-01 4 0.005400 clattewefia: —-- M V) b O‘ time L.__-_.____-_.__. _—, F” _____ pforMobilePhonein Mexico I 0.84 0.64 Po.44 0.24 0 ‘4:I:,¢,¢,¢,¢,:j¢,: 1 2 3 4 5 6 7 8 9 10 time p for Video Camera in Mexico I 0.84 0.6 Po.44 0.24 0 $131: ¢I¢I¢I¢r#r61¢ I 2 3 4 5 6 7 8 9 IO time 199 q for CD Player in Mexico 064 q0:44 0 I I I I I I I I r 12345678910 q for Home Computer in Mexico I 0.8 4 0.6 4 q 0.4 4 02‘W 0 _I T T I— _I— l I I I 12345678910 time q for Mobile Phone in Mexico 0.8 4 O6 4 9044 0.24% 12345 678910 time q for Video Camera in Mexico I 0.8 4 0.6 4 90.44 0.24% o 1 , 12345678910 time 0.8 0.6 - P 0.4 0.2 0.8 0.6 P 0.4 0.2 p for CD Player in The Netherlands p for Home Computer in The Netherlanrb 0.2 4 o 4% 12345678910 time F pforMobilePhonein TheNetherlands I 0.84 0.64 Po.44 0.24 0 :I:T:I¢T:T:I:I:I:I: 1 2 3 4 5 6 7 8 9 10 time p for Video Camera in TheNetherlands 1 0.84 0.64 Po.44 0.24 o ¢,¢4:,:.¢I¢,¢.¢.¢r¢ 12345678910 200 q for CD Player in The Netherlands 0.8 4 0.6 4 q 0.4 4 0.2 4 fir 7 I r 12345678910 time pun—— f.-- ______ q for Home Computer in The Netherlands 0.8 4 0.6 4 q 0.4 4 0.24 4 0 I I I 12345678910 time q for Mobile Phone in The Netherlands 0.8 4 0.6 4 0.4 4 0.2 4 q 12345678910 time q for Video Camera in The Netherlands 12345678910 p for CD Player in New Zealand 1 0.8 4 0.6 4 Po.4 4 0'2 1‘ _M/ 0 4W 4 ‘ , ¢ ¢ v 4 4 12345678910 12345678910 time L224 2..-.“ time p for Home Computer in New Zealand 1 0.8 4 0.6 4 P 0.4 4 0.2 4 O .. _-_—__k_ p for Mobile Phone in New Zealand I 08 4 0.6 4 P 0.4 4 0.2 4 04M 12345678910 time p for Video Camera in New Zealand I 0.8 4 0.6 4 P 0.4 4 0.2 4 o .. 12345678910 time 201 q for CD Player in New Zealand I 0.8 4 0.6 4 q 0.4 0.2 0 I I if I 1W 12345678910 0.2 4 time q for Home Computer in New Zealand I 0.8 4 0.6 4 q 0.4 4 0 I I I fir f 12345678910 I T “I time q for Mobile Phone in New Zealand 1 0.8 0.6 ‘1 0.4 0.2 04.,, I 12345678910 time q for Video Camera in New Zealand 0.8 4 0.6 4 ‘1 0.4 4 0.2 4 I I I I T I r 12345678910 time mm .____fi__.* p for CD Player in Norway F q for CD Playerin Norway 1 1 0.8 4 0.8 4 0.6 4 0.6 4 Po.4 4 q0.4 4 0.2 4 0.2 4 0 7 I I I I T O I T I I I I T‘ I I 12345678910 12345678910 time time P —————— O _._—Q r—— O .._ —. p for Home Computerm q for Home Computer rn Norway Norway 1 I 0.8 4 0.8 4 0.6 4 0.6 s Po.4 4 90.4 4 0.2 _4 0.2 -4 W o 4W o , 4 fl , , j , , , 12345678910 12345678910 time time _-_4- _-_ ._ _._ __J p for Mobile Phone in q for Mobile Phone in Norway Norway 1 1 0.8 4 0.8 _ 0.6 0.6 4 Po.4 4 90.4 4 0.2 4 ! 1: 0.2 4 W O a ‘ 0 I T T I I I I I I 12345678910 12345678910 time time p for Video Camera in q for Video Camera in Norway Norway 1 1 0.8 4 0.8 I 0.6 4 0.6 I Po_4 4 q04 4 0.2 4 02 I 0 7 o I I I I I I I I I 12345678910 12345678910 time time 202 p for CD Player in Pakistan 1 0.84 0.64 P044 0.24 0 f4¢r¢I¢I¢I¢I¢I¢I¢T¢ l 2 3 4 5 6 7 8 9 10 tinre :‘:f,_ 'm‘”” "i::’ p for Home Computer in Pakistan 1 0.84 0.64 Po.44 0.24 0.. 1 2 3 4 5 6 7 8 9 10 time pforMobilePhonein Pakistan 1 0.84 0.64 Po.44 0.24 0 VI¢I¢I:I:I;I;I:IVI 1 2 3 4 5 6 7 8 9 10 time p for Video Camera in Pakistan 1 0.84 0.6 Po.44 0.2 0 ¢,:I:I:.:4¢4:4:4:.: l 2 3 4 5 6 7 8 9 10 time 203 q for CD Player in Pakistan 1 0.84 0.64 q0.44 (121% 0 I I I I I W I r f 1 2 3 4 5 6 7 8 9 10 time F q for Home Computer in Pakistan 1 0.84 0.64 ‘10.44 0.24 0 I I 77 I I I I I I 2 3 4 5 6 7 8 9 10 time ~J qforMobile Phonein Pakistan 1 0.84 0.64 90.44 0.24 0 I 2 3 4 5 6 7 8 9 10 time qforVideo Camerain Pakistan 1 0.84 0.64 90.44 0.24 0- l 2 3 4 5 6 7 8 9 10 time _-_n _________ _m____4 p for CD Player in Poland time p for Home Computer in Poland 0% 12345678910 0.8 4 0.6 4 Po.4 4 0.2 4 04W 12345678910 time p for Mobile Phone in Poland 1 0.8 4 0.6 4 P 0.4 4 0.2 4 o 4M 12345678910 time L______._________ _ _ p for Video Camera in Poland I 0.8 4 0.6 4 P 0.4 4 0.2 4 0 GI¢I¢4¢4343434¢4¢4¢ 204 0.8 4 0.6 4 q 0.4 4 0.2 4 O I I I— I r I I r I 2 3 4 5 6 7 8 9 10 time q for Home Computer in Poland I 0.8 4 0.6 4 q 0.4 4 0.24% 0 I I 12345678910 time F. q for CD Player in Poland ,J __ _._- _I q for Mobile Phone in Pohnd 0.8 4 90:44 024 0 II I 12345678910 time q for Video Camera in Poland 12345678910 p for CD Player in Portugal 0.8 4 0.6 4 p0.4 4 0.2 M 0 ¢,:T:,:j , T . . I l 2 3 4 5 6 7 8 9 10 time p for Home Computerin Portugal l 0.8 4 0.6 4 p0.4 4 0.2 4 0 4W4 123456789l0 time p for Mobile Phone in Portugal I 0.8 4 0.6 4 P 0.4 4 0.2 4 0 ‘ 12345678910 time I——__. __—.. ____ p for Video Camera in Portugal l 00 8 '4 0.64 1’0.44 0.2] 0W '2 3 4 5 6 7 8 910 time 205 q for CD Player in Portugal I 0.8 4 0.6 4 (IO-4 .4 W 0.2 4 O a . , , , I I 1 l23456789|o time q for Home Computer in Portugal I 0.84 0.64 q0.44 0.24M 0 I I ‘ T fi V T r 7 l 2 3 4 5 6 7 8 9 l0 time qforMobile Phonein Portugal I 0.84 0.64 (10.44 0.24 O I r 1 ' r r I f I l 2 3 4 5 6 7 8 9 IO time q for Video Camera in Portugal I 0.84 0.64 q0.44 (”MM 0 I . . , , , , . . l 2 3 4 5 6 7 8 9 IO time p for CD Player in RussianFederation I 0.84 0.64 Po.44 0.24 o ¢.¢,:,¢,:,¢,¢I¢,¢I¢ I 2 3 4 5 6 7 8 9 IO ""‘°' J p for Home Computer in Russian Federation I 0.84 0.64 l’o.44 0.24 o :,:.:,:,:.+,:I¢,¢I¢ I 2 3 4 5 6 7 8 9 IO ,___I._ p for Mobile Phone in Rrrssian Federation I 0.8 4 0.6 4 p 0.4 4 0.2 0 V1¢I¢I¢I:I;I;I¢—T:IJV I23456789IO time p for Video Camera in Russian Federation 0.8 4 0.6 4 P 0.4 J 0 0 0 N 0 w I} A 0 VI 0 Ox 0 \r 0 oo o 8 E‘ 206 q forCD Player in Russian Federation I 0.8 4 0.6 4 q 0.4 4 0 I I I r I I I I f I 2 3 4 5 6 7 8 9 IO "m. e I q for Home Computer in Russian Federation I 0.8 4 0.6 4 q 0.4 4 0.2 4 0 " I 1 I I T I I 123456789|0 time __ _____--,_I qforMobile Phonein Russian Federation I 0.84 0.64 ‘10.44 0.24 0 r I I I T I I f T I 2 3 4 5 6 7 8 9 IO time qforVideo Camerain Russian Federation I 0.84 0.64 90.44 O'Z“W 0 I I I I I T I I F I 2 3 4 5 6 7 8 9 IO time _ J pfo 0.8 0.6 p 0.4 0.2 p for CD Player in Singapore 0.8 4 0.6 4 p0.4 4 0.2 I O M I 2 3 4 5 6 7 8 9 I0 time p for Home Computer in Singapore I 0.8 4 0.6 4 0.4 0.2 4 o 4W I23456789IO time p for Mobile Phone in Singapore 0.6 4 P 0.4 4 0.2 4 0 n 123456789IO time pforVideo Camerain Singapore I 0.84 0.64 Po.44 0.24 0 :I:I¢I¢I¢¢,¢ Qt: : l23456789IO time 207 q for CD Player in Singapore I 0.8 4 0.6 4 q 0.4 4 0,2 4 M OTI III;I I23456789IO f time _._—.— q for Home Computer in Singapore 0.8 4 0.6 4 ‘1 0.4 4 0.24% 0:71, I23456789IO time I I f q for Mobile Phone in Singapore 0.8 4 O6 4 I'M-MW 0.24 o ,T,,,,,FT I23456789IO time q for Video Camera in Singapore 0.8 4 0.6 4 ‘1 0.4 4 MIN OII 1 I I23456789|O time p for CD Player in SouthAfriea I 0.84 0.64 p0.44 0.24 0 ¢I¢I¢I:r+.:,:,:,efi: I 2 3 4 5 6 7 8 9 IO time p for Home Computer in SouthAfrica I 0.84 0.64 p0.44 0.24 0 ¢I¢I3I¢I¢ycy¢1¢,¢,A I 2 3 4 5 6 7 8 910 time p for Mobile Phone in SouthAfrica I 0.84 0.64 Po.44 0.g_¢,¢.¢1¢4:r¢fi,¢f°7* I 2 3 4 5 6 7 8 9 IO time p for Video Camera in SouthAfriea I 0.84 0.64 Po.44 0.24 0 ¢I¢I¢I¢I¢T¢I¢T¢$: I 2 3 4 5 6 7 8 9 I0 time _fiwwfl___ j q for CD Player in South Africa I 0.8 4 0.6 4 q0.4 4 0-2 I W 0 I I I f T fi T f r I 2 3 4 5 6 7 8 9 I0 time __..*_I q for Home Computer in South Africa I 0.8 4 0.6 4 q 04 0de I23456789IO time q for Mobile Phone in South Africa .11 0.8 4 O 6 4 q o 4 0:24W 12345678910 time 208 q for Video Camera in South Africa 0.3 0.6 4 ‘1 0.4 4 0.2 4 123456789|O time pforCDPlayerin SouthKorea I 0.84 0.64 p0.44 0.24 0 ¢¢I¢:¢,:¢:A,A I 2 3 4 5 6 7 8 9 IO time p for Home Computer in SouthKorea I 0.84 0.64 p0.44 0.24 0- I 2 3 4 5 6 7 8 9 IO time _.-.___ I ...._-,._._. p for Mobile Phone in South Korea I 0.8 4 0.6 4 P 0.4 4 0.2 4 0 .. I 2 3 4 5 6 7 8 9 IO time p for Video Camera in South Korea I 0.8 4 0.6 4 P 0.4 4 0.2 4 o - I23456789IO 209 q q for CD Player in South Korea I 0.8 4 0.6 4 0.4 4 0.2 4 I23456789IO time 0.8 4 0.6 4 q 0.4 4 0.2 4 q for Home Computer in South Korea I23456789IO time 0.8 4 q 024 4 0.2 4 _I.__ _ q for Mobile Phone in South Korea 123456789IO time _-_”I 0.8 0.6 ‘1 0.4 0.2 q for Video Camera in South Korea r T I23456789IO p for CD Playerin Spain I 0.84 0.6 4 p0.4 4 0.2 4 0 Ar: refer;,+4 44-4—9 fifiT I l23456789l0 time p for Home Computer in Spain I 0.8 4 0.6 4 P 0.4 4 0.2 4 0 _ I23456789IO time I- pfor Mobile Phone inSpain I 0.84 0.64 Po.44 0.24 0 ¢I¢I¢#,¢,:,:I¢I¢T¢ I 2 3 4 S 6 7 8 9 IO time p for Video Camera in Spain I 0.84 0.64 Po.44 0.24 0 :,¢T¢,:,:,¢,¢,¢,¢I¢ I 2 3 4 5 6 7 8 9 IO time 210 V 44~4 q for CD Player in Spain 0.8 4 0.6 4 q 0.4 4 0.2 4 IZ3456789IO time q for Home Computer in Spam 0.8 4 0.6 4 ‘1 0.4 4 0.24m 123456789IO time r— q for Mobile Phone in Spain I I 0.84 0.6- q0.44 0.24 o , , r T r I23456789IO time __I q for Video Camera in Spain I 0.8 4 0.6 4 ‘I 0.4 4 0.2 4 0 I I I I I r I I2'3456789Io “m“ 4 p for CD Player in Sweden q for CD Player in Sweden I ~ I 0.8 4 0.8 4 0.6 4 0.6 r p04 4 ‘10.4 4 NO“. 0.2 4 0.2 4 O 44 T T 0 T I r I j I I I I I23456789IO I23456789IO time time _._..J p for Home Computer in q for Home Computer in Sweden Sweden I I 0.8 4 0.8 4 0.6 4 0.6 I 1’0.4 4 90.4 4 0.2 4 02 4 o 4W 0 , , , fl , , , , 123456789IO l23456789l0 time time __; p for Mobile Phone in q for Mobile Phone in Sweden Sweden I I 0.8 4 0.8 4 0.6 4 0.6 _ Po.4 4 904 I 0.2 4 !_,/_. 0,2 I 0 d 0 I I I T I I Iv I I I23456789IO I23456789IO time time “_-_,” V__.-_I _I _ W T‘ __._ p for Video Camera in q for Video Camera in Sweden Sweden I I 0.8 4 0.8 4 0.6 4 0.6 q Po.4 4 90.4 4 0.2 4 0.2 I 0 d O I I I I I I I I I I23456789IO 123456789IO time time 211 p for CD Player in Switzerland I 0.8 4 0.6 4 p 0.4 4 0.2 4 :51 0 A 123456789IO “m J p for Home Computer in Switzefland I 0.8 4 0.6 4 l’04 4 0.2 4 Duo-147344, 4 4 44. r I 2 3 4 5 6 7 8 9 IO time p for Mobile Phone in Switzerland I 0.8 4 0.6 4 Po.4 4 0.2 4 O 4W I23456789IO time __— 7-... p for Video Camera in Switzerland I 0.8 4 0.6 4 Po.4 4 0.2 4 0 4% l23456789l0 time h—M*4a_—__i. ___I7__._. 212 q for CD Player in Switzerland .OIIIIIIIfI I23456789IO time q for Home Computer in Switzerland 0.8 4 0.6 4 ‘I 0.4 4 0.2 4 o , T I23456789IO time q for Mobile Phone in Switzerland I23456789IO time q for Video Camera in Switzerland 0.8 4 0.64 90.44 0.2% 0 TIIIIII 12 3 4 56 7 8 9IO time p for CD Player in Taiwan 0.8 4 0.6 4 p044 0:2 4 g: 0 _‘ I I23456789IO "m“. I I_.__-______ _ pforHome Computerin Taiwan I 0.84 0.64 Po.44 0.24 0- I 2 3 4 5 6 7 8 9 IO time rm .- «2 pforMobile Phonein Taiwan I 0.84 0.64 Po.44 0.24 04W I 2 3 4 S 6 7 8 9 IO time p for Video Camera in Taiwan I 0.84 0.64 Po.44 0.2 o :,:,:,:,.L,¢I¢,:,¢I¢ I 2 3 4 5 6 7 8 9 10 time q for CD Player in Taiwan 0.8 4 0.6 4 q 0.4 4 osz’N‘O’N—Q 12345678910 time q for Home Computer in Taiwan 0.8 4 0.6 4 ‘I 0.4 4 0.2 4 0 T—T *Ii I I r I I T 12345678910 time q for Mobile Phone in Taiwan 0 T I I I I I T I 12345678910 time 213 q for Video Camera in Taiwan 0.84 06 4 90:44 024% 12 3 456 7 8910 time pft 0.8 P 0.6 I 0.4 0.2 p for CD Player in Thailand I 0.8 4 0.6 4 P 0.4 4 0.2 I O time p for Home Computer in Thailand I W 12345678910 __; —___ 0.8 4 0.6 4 P 0.4 4 0.2 4 04W 123456789I0 time p for Mobile Phone in Thailand I 0.8 4 0.6 4 P 0.4 4 0.2 4 0 4W 12345678910 time F.“ pfor Video Camera in Thailand I 0.8 4 0.6 Po.44 0.2 0 ° 4°4¢r¢r¢4:4:,:4¢r: I23456789IO time 214 q for CD Player in Thailand I 0.8 4 0.6 4 II 0.4 4 0'2 W 0 I fl I I I I T I23456789IO time q for Home Computer in Thailand 0.8 4 0.6 4 q 0.4 4 0.2 o W I I 123456789IO time q for Mobile Phone in Tlmiland I 0.8 4 0.6 q o 4 123456789Io time q for Video Camera in Tlnriland I 0.8 4 0.6 4 ‘I 0.4 4 0.2 0 W 12345678910 time p for CD Player in Tunisia I 0.84 0.64 p0.44 0.24 o Arcveefcrameare I 2 3 4 5 6 7 8 9 IO time — "a p for Home Computer in Tunisia I 0.84 0.64 Po.44 0.2 O ;r¢1¢rcy¢r¢y¢r¢,¢r¢ I 2 3 4 5 6 7 8 9 IO time pforMobilePhonein Tunisia I 0.84 0.64 Po.44 0.24 o :,:.:T:,:I¢.c,:,:,¢ I 2 3 4 5 6 7 8 9 IO time p for Video Camera in Tunisia I 0.84 0.64 Po.44 0.24 0 ‘I¢.¢I¢I%I%I¢I¢rfrv I 2 3 4 5 6 7 8 9 IO time 215 q for CD Player in Tunisia 0.8 4 0.6 4 q 0.4 4 12345678910 time L____ q for Home Computer in Tunrsia I23456789IO time q for Mobile Phone in Tunisia 123456789IO time q for Video Camera in Tunisia I 0.8 4 0.6 4 ‘I 0.4 4 0.2 4 0 I I I I I 12345678910 time p for CD Player in Turkey 0.8 4 0.6 4 P 0.4 4 0.2 4 0 AAAAAAAA v A VIVIVIVIVI'IfIVT v I 12345678910 time I pfor Home Computerin Turkey I 0.84 0.64 Po.44 0.24 o e.ere.¢.¢.:.¢.: .¢.v 12345678910 time (_-_—_-_- -4 pforMobilePhonein Turkey I 0.84 0.64 Po.44 0.24 0 fI IVTVIVI:I¢I:T:TA I 2 3 4 5 6 7 8 9 IO time T pforVideo Camerain Turkey I 0.84 0.64 P04- 0.24 o :.¢.e.:.¢,¢.:.¢re# I 2 3 4 5 6 7 8 9 IO time 216 q for CD Player in Turkey 2 W I 12345678910 time ____I q for Home Computer in Turkey I 0.8 4 0.6 4 ‘I 0.4 4 0.2 4 0 I I I I I I I I I 12345678910 time q for Mobile Phone in Turkey 0.8 4 0.6 4 q 0.4 4 0.2 4 23456789 time q for Video Camera in Turkey I 0.8 4 0.64 90.44 0.24 0 I I I I I I I I I 12345678910 time 12345678910 time time 4——_~— A p for Mobile Phone in UK q for Mobile Phone in UK I I 0.84 0.84 0.64 0.6- p0.44 qo,44 0.24 0.24W 0‘ o I I I I I I I I T I23456789IO 12345678910 time time p for Video Camera in UK q for Video Camera in UK I I 0.84 0.34 0.64 0.64 Po.44 90.44 0.24 02. 04W 0 , , r , . , , T , I23456789IO 12345678910 time time 217 p for CD Player in UK q for CD Player in UK I I 0.8 ‘ 0.8 —I 0.6 4 0.6 4 P 0.4 4 ‘I 0.4 4 0.2 4 0.2 4 W 0 I I I I 0 I I I I I I I I a I23456789IO I23456789IO time time _.n __ _._; "“7 p for Home Computer in UK q for Home Computer in UK I 0.84 06 4 90:44 0.24W 0 Ifi I I I *F I 12345678910 p for CD Playerin USA q forCD Playerin USA I m 0.8 4 0.6 4 q0.4 4 0.2 4 I I I I I I O T T j T j I I I I I23456789IO 12345678910 time time I____ _ __ -_____ WM“ _-- p for Home Computer in q for Home Computer in USA USA I I 0.8 4 0.8 4 0.6 4 0.6 - Po.4 4 ‘104 4 0.2 1 0.2 4 W 0 MM 0 , , , , . , , , , I23456789IO 123456789IO time time _I _ _-__-__ - ..k _._ _-_._____1 _._- ...,_,_._ j p for Mobile Phone in USA q for Mobile Phone in USA I I 0.8 4 0.8 4 0.6 4 0.6 4 Po.4 4 90.4 4 0.2 4 I ! i 0.2 W O A O a I I I I T I I T 123456789IO I23456789IO I' I' J III—I ___ -_——_—_ ..... "—fl p for Video Camera in USA q for Video Camera in USA I I 0.8 4 0,3 4 0.6 4 0.6 4 p04 I qOA I W 0.2 4 ! i 0.2 0 I 0 I I I fi fi I I I I23456789IO 123456789IO time time ______ _ fi_ 2 w _-J _ _4 218 APPENDIX C Difiusion Patterns Over Time for Each Country and Bach Product 219 Estimated Per Capita Sales of CD Players in Argentina l 0.8 4 0.6 4 x 0.4 — 0.2 M 0 t r 3 r t I A T 1 t j I I l 2 3 4 5 , 6 7 8 9 l0 time Estimated Per Capita Sales of Home Computers in Argentina 1 0.8 4 0.6 A x 0.4 - 0.2 4 0 e T 4 r e r t 1 5 M I 2 3 4 5 , 6 7 8 9 l0 time Estimated Per Capita Sales of Mobile Phones in Argentina | 0.8 A 0.6 d x 0.4 ~ 0.2 e o e , e , c I e r e M I 2 3 4 5 , 6 7 8 9 IO time Estimated Per Capita Sales of Video Cameras in Argentina 1 0.8 4 0.6 d x 0.4 4 0.2 4 0:,A :,:jfl_:§fikyi§:1¢14v l 2 4 5 , 6 8 9 l0 time 220 Estimated Per Capita Sales of CD Players in Australia l 0.8 0.6 X 0.4 02 1 _K/ 0 6 Y ; I : I ; T I I I f I l 2 3 4 , 7 8 9 IO trme Estimated Per Capita Sales of Home Computers in Australia 0.8 4 0.6 T x 0.4 4 0.2 ~ M 0 c I : I 7 I T r I y t l 2 3 4 5 , 6 7 8 9 l0 trme Estimated Per Capita Sales of Mobile Phones in Australia I 0.8 d 0.6 s x 0.4 « 0.2 ~ 0 6 T 3 I c I A r T I I l l l 2 3 4 , 7 8 9 l0 trme Estimated Per Capita Sales of Video Cameras in Australia I 0.8 ~ 0.6 e x 0.4 4 0.2 4 W 0 ¢ 1 : l——.!F’T : r : l c 7 l l l l 2 3 , 7 8 9 l0 trme 221 Estimated Per Capita Sales of CD Players in Austria l 0.8 4 0.6 4 x 0.4 4 0.2 4 0 c c l A f r l l l W7 l 2 3 4 S , 6 7 8 9 l0 trme Estimated Per Capita Sales of Home Computers in Austria I 0.8 4 0.6 4 0.4 4 0.2 4 0 : I A I ' I j I I I I l 2 3 4 5 , 6 7 8 9 l0 trme Estimated Per Capita Sales of Mobile Phones in Austria I 0.8 I 0.6 4 x 0.4 - 02 4 AM 0 ¢ I t I : I W I f I I l 2 3 4 5 , 6 7 8 9 l0 trme Estimated Per Capita Sales of Video Cameras in Austria I 0.8 0.6 4 0.4 4 02 M 0 : I A ; I : I I I r I I 2 3 4 5 , 6 7 8 9 l0 trme 222 Estimated Per Capita Sales of CD Players in Belgium l 0.8 4 0.6 4 x 0.4 0.2 0 ¢ 7— l l T l l I 2 3 4 5 , 6 7 8 9 l0 trme Estimated Per Capita Sales of Home Computers in Belgium l 0.8 4 0.6 4 x 0.4 J 0.2 4 0 v I I I I I I f I r l 2 3 4 5 , 6 7 8 9 l0 trme (_._ _____ _ Estimated Per Capita Sales of Mobile Phones in Belgium l 0.8 4 0.6 0.4 i 0.2 4 A A _._... 0::Y¢I¢,¢.d—$=‘WI'T'. 2 3 5 , 6 7 8 9 I0 trme J Estimated Per Capita Sales of Video Cameras in Belgium l 0.8 4 0.6 4 223 (A. _ _ _ Estimated Per Capita Sales of CD Players in Brazil I 0.8 0.6 I 0.4 4 0.2 4 A A -A.———--—’ ocrertrctdE—r—fT'I'T 2 3 4 5 , 6 7 8 9 l0 trme J Estimated Per Capita Sales of Home Computers in Brazil I 0.8 4 0.6 4 x 0.4 4 02 M 0 ¢ I t I I I I 1 I 2 3 4 5 , 6 7 8 9 IO trme _ .3 Estimated Per Capita Sales of Mobile Phones in Brazil l.OOE+OO 8005-0] 4 6005-0] 4 x 4.00E-Ol 4 2.0030] 7 ‘V/ 0.005+00¢IVLI¢1¢T¢15I.. 2 3 5 , 8 9 IO trme F_A_V__ Estimated Per Capita Sales of Video Cameras in Brazil I 0.8 4 0.6 4 x 0.4 4 0.2 0 ¢ I c I 4 I : T i I t I : I f I A .‘—J I 2 3 4 5 ti 6 7 8 9 IO __ _ _ me __ _J 224 P_A..‘_-__...AA—. .4_._._._..___h. — Estimated Per Capita Sales of CD Players in Canada 0.8 4 0.6 4 x 0.4 4 0.2 ‘ M 0 4v a f r ¢ 1 e r I fl I T T I 2 3 4 5 , 6 7 8 9 IO trme Estimated Per Capita Sales of Home Computers in Canada I 0.8 4 0.6 4 x 0.4 4 02 4 A// O ; j A I c T v I W I I I r I 2 3 4 5 , 6 7 8 9 IO trme MAW __ _._- _____ _AJ Estimated Per Capita Sales of Mobile Phones in Canada I 0.8 I 0.6 4 x 0.4 4 0.2 I #y/ 0 : I 5 fl I I fii fi ‘Ii I I 2 3 4 5 , 6 7 8 9 IO trme __ __ _J Estimated Per Capita Sales of Video Cameras in Canada I 0.8 4 0.6 4 x 0.4 4 02 M 0 : r ¢ T . I : I c T I I ' I 225 Estimated Per Capita Sales of CD Players in Chile 0.8 4 0.6 4 X 0.4 4 0.2 4 A A————'4_———. O c I % I : r t I £ , .3 1 v f v f I 2 3 4 5 , 6 7 8 9 l0 trme Estimated Per Capita Sales of Home Computers in Chile I 0.8 4 0.6 4 x 0.4 4 0.2 M 0 ; ' : ' T I I ‘l_ I I 3 4 , 7 8 9 IO trme Estimated Per Capita Sales of Mobile Phones in Chile I 0.8 0.6 4 x 0.4 4 0.2 ~ _._—M OeIertItr’fia—dfi°.:1 I 2 3 4 5 , 6 7 8 9 IO trme Estimated Per Capita Sales of Video Cameras in Chile I 0.8 4 0.6 I x 0.4 4 0.2 4 0 ¢ I ¢ 3 I t I t 1 : 1w 226 Estimated Per Capita Sales of CD Players in China I 0.8 4 0.6 4 x 0.4 4 0.2 4 o - e , A e e A . e , t t e , fl—a I 4 , 6 7 8 9 IO trme Estimated Per Capita Sales of Home Computers in China I 0.8 4 0.6 4 x 0.4 4 0.2 MM 0 e e , c e +, ° 4 ' T 3 4 5 , 6 7 8 9 IO trme J Estimated Per Capita Sales of Mobile Phones in China I 0.8 4 0.6 4 x 0.4 4 0.2 4 0 ¢ r ¢ r ¢ e c r 4. I e e e , : l 2 3 4 , 6 7 8 9 IO trme Estimated Per Capita Sales of Video Cameras in China I 0.8 4 0.6 4 x 0.4 4 0.2 4 0+T¢T¢ ¢¢,AT¢%¢,¢ I 2 3 5 , 8 9 IO rme 227 Estimated Per Capita Sales of CD Players in Denmark 0.6 4 0.4 4 0.2 4 m Estimated Per Capita Sales of Home Computers in Denmark I 0.8 4 0.6 4 0.4 4 0.2 4 0 X 1P 0 _._—1 .-. r. _--__. Estimated Per Capita Sales of Mobile Phones in Denmark ‘mj I 0.8 4 Estimated Per Capita Sales of Video Cameras in Denmark 0.8 4 0.6 4 0.4 4 228 Estimated Per Capita Sales of CD Players in Egypt l 0.8 0.6 0.4 0.2 4 o¢,¢:¢I:,:fi;,A,t——I——o— 2 3 , 7 8 9 IO trme Estimated Per Capita Sales of Home Computers in Egypt I 0.8 4 0.6 4 x 0.4 4 0.2 0¢.¢:¢,:,:fi;l—jIefi—j__ 3 4 _ 6 7 8 9 IO trme (_-_w -_ Estimated Per Capita Sales of Mobile Phones in Egypt I 0.8 4 0.6 4 0.4 4 0.2 O t I 4 I 3 I 3 6 I t I t I é I 3 I é I 2 3 4 , 7 8 IO trme Estimated Per Capita Sales of Video Cameras in Egypt I 0.8 4 0.6 0.4 4 0.2 OeIcIcIfItIflIh:I% I 3 , 6 7 8 9 IO trme 229 FT — Estimated Per Capita Sales of CD Players in Finland 0.8 4 0.6 - x 0.4 4 0.2 4 0 c I t I I I I I I I I 3 4 5 , 7 8 9 IO trme Estimated Per Capita Sales of Home Computers in Finland I 0.8 0.6 4 x 0.4 4 0 3 f t I c I : I I I I I I I 2 3 4 5 , 6 7 8 9 IO trme Estimated Per Capita Sales of Mobile Phones in Finland I 0.8 4 0.6 4 x 0.4 4 0.2 4 A A A e —4.——---*“ 0 : I Afi I’fii v I v I V I I I I 2 3 4 5 , 6 7 8 9 IO trme Estimated Per Capita Sales of Video Cameras in Finland I 0.8 4 0.6 - x 0.4 4 0.2 4 A AM 0¢Icr¢If°I'I'I I I I 2 3 4 , 7 8 9 IO trme 230 _ A! Estimated Per Capita Sales of CD Players in France 0.8 4 0.6 4 x 0.4 ~ 0.2 4 j I I 5 . trme __...J Estimated Per Capita Sales of Home Computers in France -22..“ -‘I I 2 3 4 5 . 6 7 8 9 IO trme Estimated Per Capita Sales of Mobile Phones in France I . 0.8 0.6 0.4 0.2 4 0¢I%I¢I¢I¢,:Tfifi—T—+—;—‘ I 2 3 4 , 6 7 8 9 IO trme 4 Estimated Per Capita Sales of Video Cameras in France I 0.8 0.6 4 x 0.4 0.2 4 0 c e I : Ifi—‘t” : I v I v r 2 3 4 5 , 6 7 8 9 IO trme 23] Estimated Per Capita Sales of CD Players in Germany I 2 3 4 5 6 7 8 9 IO 7’ p” 4 Estimated Per Capita Sales of Home Computers in Germany I 2 3 4 5 , 6 7 8 9 IO trme L__.~ __ -_ _.___._ [_.__A.-__~_ __ Estimated Per Capita Sales of Mobile Phones in Germany I 0.8 4 0.6 x 0.4 4 0.2 0 : I A I : I t I : I I I T 2 3 4 , 6 7 8 9 IO trme 4 Estimated Per Capita Sales of Video Cameras in Germany I 0.8 0.6 4 x 0.4 4 0.2 4 0 ; I : I I I I 3 4 5 , 6 7 8 9 IO trme L..I___-A___.___-._---_ _ -.-_-I ._________3_.__..-_._...3-__3_.- 232 *4 Estimated Per Capita Sales of CD Players in Greece 0.8 4 0.6 4 x 0.4 4 0.2 4 M o e I e t I * I I a I I I I 2 3 4 5 , 6 7 8 9 IO trme Estimated Per Capita Sales of Home Computers in Greece I 0.8 0.6 4 x 0.4 4 0.2 4 A 3' 4% 0:I:I:I¢.4t==tf'I f. I 2 3 4 5 , 6 7 8 9 IO trme Estimated Per Capita Sales of Mobile Phones in Greece I 0.8 4 0.6 4 x 0.4 4 0.2 0 3 I ¢ I L I : r : I I I j “I I 2 3 4 5 , 6 7 8 9 IO trme J Estimated Per Capita Sales of Video Cameras in Greece I 0.8 0.6 4 x 0.4 4 0.2 4 0 ¢ I ¢ r 3 ,fifl—1 3 ‘ t I v I v I 2 3 4 5 , 6 7 8 9 IO trme 233 'W "_' — 1 ”Estimated Per Capita Sales of CD Players in Hong Kong I 0.8 4 0.6 4 x 0.4 4 0 c I : ' V I I I I I I I 3 4 5 , 6 7 8 9 IO trme I Estimated Per Capita Sales of Home Computers in Hong Kong 0.8 0.6 x 0.4 4 0.2 0 " f I I I I f I 2 3 4 5 , 6 7 8 9 IO trme Estimated Per Capita Sales of Mobile Phones in Hong Kong I 0.8 0.6 4 x 0.4 4 0.2 0 : f ; I I I I I I I I I 2 3 4 5 , 6 7 8 9 IO trme Estimated Per Capita Sales of Video Cameras in Hong Kong I 0.8 4 0.6 4 x 0.4 4 0.2 4 O " j I I I I I I I 2 3 4 5 , 6 7 8 9 IO trme 234 Estimated Per Capita Sales of CD Players in Hungary I 0.8 4 0.6 4 x 0.4 4 0.2 0 ¢ I 3 t I Af I ' ' I T I 2 3 4 5 , 6 7 8 9 IO L trme Estimated Per Capita Sales of Home Computers in Hungary I 0.8 0.6 4 x 0.4 4 0.2 0 c I c I t I M I 4 5 , 6 7 8 9 IO trme J Estimated Per Capita Sales of Mobile Phones in Hungary I 0.8 4 0.6 4 0.4 4 0 2 1 4+’/ 0 ¢ I 6 I ¢ I : I c I I I I I I 2 4 , 6 7 8 9 IO trme Estimated Per Capita Sales of Video Cameras in Hungary I 0.8 0.6 - 0.4 4 0.2 4 0 3 I ¢ I 3 I C I 4=F==I=r=fir I 4H] ‘t T LJ I 2 4 5 , 6 7 8 9 IO trme J 235 V i _I _ Estimated Per Capita Sales of CD Players in India I 0.8 4 0.6 ‘I x 0.4 ~ 0.2 « 0 ——r 4 k I 2 3 4 5 6 7 8 9 10 time Estimated Per Capita Sales of Home Computers in India I 0.8 0.6 ' X 0.4 0.2 O : I é I 6 I A I #fir : I : I : I I 2 3 4 5 , 6 7 8 9 IO trme Estimated Per Capita Sales of Mobile Phones in India I 0.8 4 0.6 I x 0.4 4 0.2 I 0 6 ¢ I c I c I c I 3 I ¢ I ¢ I 4v I 4: I 2 3 4 5 , 6 7 8 9 IO trme r__..__ __ Estimated Per Capita Sales of Video Cameras in India I 0.8 4 0.6 4 x 0.4 4 0.2 4 0 3 I 3 r ¢ I 3 I L I ¢ I VL I ¢ I ¢ I 6 2 3 4 , 6 7 8 IO trme 236 I” - h-" .47 Estimated Per Capita Sales OfCD Players in IndoneSia 0.8 I 0.6 4 X 0.4 4 0.2 4 o e I 2 3 4 5 , 6 7 8 9 I0 trme J 0 0 0 Estimated Per Capita Sales of Home Computers in Indonesia 0.4 I A A A A I v I v I v I O {I (I -I w t} .m— __ trme ______J _Estimated Per Capita Sales ofMobile Phones in Indonesia 0.8 4 0.6 4 X 0.4 4 time b—————_ ___ time 237 238 F Estimated Per Capita Sales of CD Players in Ireland I 0.8 - 0.6 4 x 0.4 ~ 0.2 M 0¢r¢4:,¢Tcr—r—T¢fi', I 2 3 4 , 6 7 8 9 IO trme Estimated Per Capita Sales of Home Computers in Ireland I 0.8 0.6 4 x 0.4 0 ¢ 1 A I ; I I I I I I I 2 3 4 5 , 6 7 8 9 IO trme Estimated Per Capita Sales of Mobile Phones in Ireland I 0.8 0.6 4 x 0.4 4 0.2 4 OctI¢IiTtIflf°I° ' I 2 4 5 , 6 7 8 9 IO trme I Estimated Per Capita Sales of Video Cameras in Ireland I 0.8 4 0.6 4 x 0.4 4 0.2 4 0 3 I Afi r C l—‘&==-,==Er_$ I ¢ ‘ : I v v I 2 3 4 5 , 6 8 9 IO trme J Estimated Per Capita Sales of CD Players in Israel I 0.8 4 0.6 x 0.4 4 0.2 ‘ M 0 : I : T I I I W I I f I 2 3 4 5 , 6 7 8 9 IO trme [ _ Estimated Per Capita Sales of Home Computers in Israel I 0.8 0.6 I x 0.4 0.2 I 0 c I t I f I I I fi T I 2 3 4 5 , 6 7 8 9 IO trme I 0.8 0.6 I x 0.4 0.2 4 o r I 2 3 4 5 , 6 7 8 9 IO trme F Estimated Per Capita Sales of Video Cameras in Israel I 0.8 4 0.6 4 x 0.4 4 0.2 4 0:1:fitrtrflfi¢fi°.°I' I 2 4 5 , 6 7 8 9 I0 trme 239 Estimated Per Capita Sales of CD Players in Italy ,—.__- time Estimated Per Capita Sales of Video Cameras in Italy 0 0 0 240 _._-.. .___ _-J Estimated Per Capita Sales of CD Players in Japan I 0.8 I 0.6 I x 0.4 0.2 O : T fir I I I r I 2 3 4 5 , 6 7 8 9 IO trme Estimated Per Capita Sales of Home Computers in Japan I 0.8 0.6 x 0.4 I 02 q M 0 c I— : I I I I T I I I 2 3 4 5 6 7 8 9 IO trme j Estimated Per Capita Sales of Mobile Phones in Japan I 0.8 I 0.6 4 x 0.4 4 0.2 I _A/ 0 ; I ¢ I : I : I—fl—q : I v I 7 l 2 3 4 5 , 6 7 8 9 I0 trme Estimated Per Capita Sales of Video Cameras in Japan I 0.8 0.6 4 x 0.4 4 O 2 d 4’4/0/ 0 ¥ I c T I I T I I I I I 2 3 4 5 , 6 7 8 9 IO trme 241 Estimated Per Capita Sales of CD Players in Malaysia 0.8 4 0.6 4 x 0.4 4 0.2 4 A -w 0 c I c I A :k'I ¢ I v Ti 7 I r I I 3 4 5 , 6 7 8 9 IO trme Estimated Per Capita Sales ofHome Computers in Malaysia I 0.8 0.6 4 x 0.4 0.2 I - - w 0 c I c c I I T v I v I I I I 2 3 4 5 , 6 7 8 9 IO time Estimated Per Capita Sales of Mobile Phones in Malaysia I 0.8 0.6 x 0.4 0.2 0 ¢ I : I 3 f t I t 1 I I I I 3 4 , 6 7 8 9 IO trme Estimated Per Capita Sales of Video Cameras in Malaysia I 0.8 0.6 4 x 0.4 4 0.2 ~ w 0 : fl : I—gi : : I v I T I I 2 3 4 5 , 6 7 8 9 IO trme 242 i‘Estimated Per Capita Sales ofCD Players in Mexico Estimated Per Capita Sales of Home Computers in Mexico Estimated Per Capita Sales of Mobile Phones in Mexico Estimated Per Capita Sales of Video Cameras in Mexico 243 ,—_I‘ ._ _._._. __ ____ _._ Estimated Per Capita Sales of CD Players in The Netherlands 0.8 0.6 0.4 4 0.2 4 0 ' I I T T I I 7* I r I 2 3 4 5 , 6 7 8 9 IO trme Estimated Per Capita Sales of Home Computers in The Netherlands I Estimated Per Capita Sales of Mobile Phones in I The Netherlands 0.8 0.6 I 0.4 4 0.2 4 O t I é : I t I t T—‘g—‘T—ii : I 4' 3 4 , 6 7 8 9 IO trme Estimated Per Capita Sales of Video Cameras in I The Netherlands 0.8 4 0.6 4 x 0.4 4 0.2 . M Ocr‘I—‘gicIcfi'I'I I I 2 3 4 5 , 6 7 8 9 IO trme 244 Estimated Per Capita Sales of CD Players in 245 New Zealand 0.8 0.6 4 x 0.4 I 0.2 I 0 C T T I T I T I 2 3 4 5 , 6 7 8 IO trme Estimated Per Capita Sales of Home Computers in New Zealand 0.8 4 0.6 4 x 0.4 4 0.2 I 0 A r I 1 j T I 2 3 4 5 , 6 7 8 IO trme Estimated Per Capita Sales of Mobile Phones in New Zealand 0.8 0.6 4 x 0.4 4 0.2 0 4v 3 I 2 Estimated Per Capita Sales of Video Cameras in I New Zealand 0.8 4 0.6 4 x 0.4 4 0.2 4 O : T t I T I I I I 2 3 4 5 ti 6 7 8 IO me ~ “_J I. Estimated Per Capita Sales of CD Players in Norway l 0.8 4 0.6 4 x 0.4 4 0.2 0 a r I 2 3 4 5 , 6 7 8 9 IO trme Estimated Per Capita Sales of Home Computers in Norway I 0.8 4 0.6 4 x 0.4 4 0.2 « M/ 0 : fl I *I I fl I I f I l 2 3 4 5 , 6 7 8 9 IO trme Estimated Per Capita Sales of Mobile Phones in Norway I 0.8 4 0.6 4 x 0.4 4 0.2 4 0 : r r r r r I I I 2 3 4 5 , 6 7 8 9 IO trme I Estimated Per Capita Sales of Video Cameras in Norway 1 0.8 4 0.6 4 x 0.4 4 0.2 7 M 0 c r c ,H’T ° 1 ' , fi , . I 2 3 4 5 , 6 7 8 9 IO trme 246 Estimated Per Capita Sales of CD Players in Pakistan I 0.8 4 0.6 - x 0.4 4 0.2 0 ¢ A 3 f ¢ , 3 r t 7 m I 2 3 4 , 6 7 8 9 IO trme Estimated Per Capita Sales of Home Computers in Pakistan I 0.8 0.6 4 x 0.4 0.2 4 o: ¢,¢TL:1;r:.:T+,: , 7 8 IO trme J Estimated Per Capita Sales of Mobile Phones in Pakistan 1 , 0.8 4 0.6 4 x 0.4 4 0.2 4 0¢T¢1¢4¢r¢TATvLT¢r¢41A I 4 5 , 7 8 9 IO trme _.__ I Estimated Per Capita Sales of Video Cameras in Pakistan I 0.8 - 0.6 4 x 0.4 4 0.2 4 0 ¢ c I Af T # I c I t I c I ¢ f ; if V 2 3 4 6 7 8 9 IO trme J 247 Estimated Per Capita Sales of CD Players in Poland I 0.8 4 0.6 4 x 0.4 4 0.2 4 MM 0 ¢ r WA T t r r r r f r r I 2 3 4 5 , 6 7 8 9 l0 trme Estimated Per Capita Sales of Home Computers in Poland I 0.8 4 0.6 4 x 0.4 4 0.2 ~ M 0 : "I c I I I fi f I I I I 2 3 4 5 , 6 7 8 9 IO trme Estimated Per Capita Sales of Mobile Phones in Poland I 0.8 4 0.6 4 x 0.4 02 7 4.4/0 0 ¢ fl A I ¢ I c I : I A I : r I 7 l 2 3 4 , 6 7 8 9 IO I trme Estimated Per Capita Sales of Video Cameras in Poland I 0.8 0.6 4 x 0.4 4 0.2 4 o :TAT¢,AI:T:7H==I=*—.—*—I" I 3 , 6 7 8 9 IO trme 248 r—-——————— -—- ——— _—-— m Estimated Per Capita Sales of CD Players in Portugal I 0.84 0.6 4 x 0.4 4 02 *M/ 0 ¢ ¢ 1 c . r , , Y j , l 2 3 4 5 , 6 7 8 9 IO trme 0.8 4 0.6 x 0.4 4 02 M 0 : f : T I r r I r r j I 2 3 4 5 , 6 7 8 9 IO trme Estimated Per Capita Sales of Mobile Phones in Portugal I 0.8 0.6 4 0.4 4 02 4 g/ 0 ¢ 1 A 1 : 1 : r j r —I_ T I 2 3 4 5 , 6 7 8 9 IO trme h‘ I 0 Estimated Per Caprta Sales of Video Cameras rn Portugal I 0.8 4 0.6 4 x 0.4 4 0 ¢ 3 rfl: ° F ' f I r , T I 2 3 4 5 , 6 7 8 9 I0 trme 249 I Estimated Per Capita Sales of CD Players in Russian Federation 0.8 0.6 x 0.4 0.2 4 AM 0 3 G 1 c T t I t j—‘th : I— v T T 2 3 4 5 , 6 7 8 9 IO trme -- __2__ _-.II___ _. _._.“ -J I‘“ _‘_ _‘ —' "'"" '— I Estimated Per Capita Sales of Home Computers in Russian Federation 0.8 0.6 4 x 0.4 0 2 M O 4 I : T r I r I I I I I 2 3 4 5 , 6 7 8 9 IO trme I Estimated Per Capita Sales of Mobile Phones in Russian Federation 0.8 0.6 4 x 0.4 4 o 2 4*{0/ 0 4 A 1 ¢ I : I f I r 7 I 2 3 4 5 , 6 7 8 9 IO trme f- ____ Estimated Per Capita Sales of Video Cameras in Russian Federation 0.8 4 0.6 4 x 0.4 4 0.2 4 M 0 ¢ 7 t I+T ; I : I : I v T I I I 2 3 4 5 , 6 7 8 9 IO trme 250 Estimated Per Capita Sales of CD Players in Singapore I 0.8 0.6 4 x 0.4 4 0.2 4 0 : T A I I I T I I I 3 4 5 , 6 7 8 9 IO trme Estimated Per Capita Sales of Home Computers in Singapore I 0.8 4 0.6 4 X 0.4 4 02 ‘ T—Q/KKO/ 0 t I I I I T I I I? I I 2 3 4 5 , 6 7 8 9 IO trme ___H__ J Estimated Per Capita Sales of Mobile Phones in Singapore I 0.8 0.6 4 x 0.4 4 0.2 0 c I— f I fl I I I 2 3 4 5 , 6 7 8 9 IO trme Estimated Per Capita Sales of Video Cameras in Singapore I 0.8 0.6 4 x 0.4 4 0.2 4 0:.‘r¢fi¢.‘fi—¢t?—I_t’r°t:,v I 2 4 5 , 6 7 8 9 IO trme 251 F_-__fi__ _._2 Estimated Per Capita Sales of CD Players in South Africa 0.8 4 0.6 4 x 0.4 4 0.2 4 O Af T ¢ 3 r : I fi’r ; T ¢ If ¢ I #6 I 2 3 4 5 , 6 7 8 9 IO trme I Estimated Per Capita Sales of Home Computers in South Africa 0.8 4 0.6 4 X 0.4 4 0.2 O Af I c I k I : Ifiii—I c I 3 T —¢ I 2 4 5 , 6 7 8 9 IO trme 1 Estimated Per Capita Sales of Mobile Phones in South Africa 0.8 4 0.6 4 x 0.4 4 02 I M O c I ¢ I L I A I : I : I I I I I 2 3 4 , 6 7 8 9 IO trme 1 Estimated Per Capita Sales of Video Cameras in South Africa 0.8 4 0.6 4 x 0.4 4 0.2 4 0¢,¢,:,:r4,—tfié+,+—,L l 3 4 5 , 6 7 8 9 IO trme 252 [__ ___ _ __ I Estimated Per Capita Sales of CD Players in South Korea 0.8 4 0.6 4 x 0.4 4 0.2 0 4 c I A .L f : 4 ‘ M I 3 4 5 , 6 7 8 9 IO trme J Estimated Per Capita Sales of Home Computers in South Korea 0.8 0.6 4 X 0.4 4 0.2 M 0 : I : I I I I I f I I I 2 3 4 5 , 6 '7 8 9 IO trme [ I Estimated Per Capita Sales of Mobile Phones in South Korea 0.8 4 0.6 4 x 0.4 0.2 4 O 6 1 ‘ I ‘ , c r ‘ I C I W I 2 3 , 6 7 8 9 IO trme [ I Estimated Per Capita Sales of Video Cameras in South Korea 0.8 0.6 x 0.4 0.2 4 4M 0 ¢ T— A I : fl I c fi— : I I I I- I 3 4 5 , 6 7 8 9 IO trme 253 Estimated Per Capita Sales of CD Players in Spain I 0.8 4 0.6 4 x 0.4 4 0.2 7 M/ 0 : f 4 fl I T I I T I I 2 3 4 5 , 6 7 8 9 I0 trme Estimated Per Capita Sales of Home Computers in Spain I 0.8 4 0.6 4 X 0.4 4 0.2 4 M 0 : I : I I T I I I I I I 2 3 4 , 7 8 9 IO trme F Estimated Per Capita Sales of Mobile Phones in Spain I 0.8 4 0.6 4 x 0.4 4 0.2 4 0 c I A I : I A I 3 I ¢ 1 i I 3 1m 2 3 4 , 6 7 9 IO trme 7— Estimated Per Capita Sales of Video Cameras in Spain I 0.8 4 0.6 4 x 0.4 4 0.2 4 041%,:fi‘rfl=—r==fi==—1=ti°r°44' I 3 4 5 , 6 7 8 9 IO trme 254 F Estimated Per Capita Sales of CD Players in Sweden 0.8 4 0.6 4 0.4 4 0.2 4 X time Estimated Per Capita Sales of Mobile Phones in Sweden 7. Estimated Per Capita Sales of Video Cameras in Sweden I 0.8 4 0.6 4 x 0.4 4 0.2 W 0:,eTcrfl4—1—1’frcyv.+r I 2 4 5 _ 6 7 8 9 IO trme 255 Estimated Per Capita Sales of CD Players in Switzerland I 0.8 4 0.6 - x 0.4 4 0.2 4 0 4 : T I I I T T I 2 3 4 5 , 6 7 8 9 IO trme Estimated Per Capita Sales of Home Computers in Switzerland 0.8 0.6 4 X 0.4 0.2 4 O : I T I I T T T I I I I 2 3 4 , 7 8 9 IO trme Estimated Per Capita Sales of Mobile Phones in Switzerland l 0.8 4 0.6 4 x 0.4 4 02 4 MT 0 ¢ I : T A T I I I I I I 2 3 4 5 6 7 8 9 IO time I” . . . . _ Estrmated Per Caprta Sales of Video Cameras rn Swrtze rland I 0.8 0.6 0.4 4 0.2 A : : a 0 e W I I I I 2 3 4 5 , 6 7 8 9 I0 trme 256 Estimated Per Capita Sales of CD Players in Taiwan 0.8 4 0.6 4 X 0.4 4 0.2 4 Estimated Per Capita Sales of Home Computers in Taiwan I I I I T I I 10 time Estimated Per Capita Sales of Mobile Phones in Taiwan Estimated Per Capita Sales of Video Cameras in Taiwan 0.8 4 0.6 4 0.4 4 0.2 4 257 Estimated Per Capita Sales of CD Players in Thailand T '7 0.8 4 0.6 4 x 0.4 4 0.2 4 O: eIeILIfiIFI°I°4T I 3 4 5 , 6 7 8 9 IO trme r______ , Estimated Per Capita Sales of Home Computers in Thailand I 0.8 4 0.6 4 x 0.4 4 0.2 4 0 ¢ I A I : T t l : IT—I : I : TT ' I 2 3 4 5 6 7 8 9 IO time F Estimated Per Capita Sales of Mobile Phones in Thailand I 0.8 4 0.6 4 x 0.4 4 0.2 4 0:IAI¢I%I:I5I¢I—t—’I°T° I 2 4 , 6 7 8 9 IO trme Estimated Per Capita Sales of Video Cameras in Thailand I 0.8 4 0.6 4 0.4 4 0.2 0¢I4I¢I¢I¢ tItIH—I—fi—Ififi—4 time 258 Estimated Per Capita Sales of CD Players in Tunisia 0.8 4 0.6 4 x 0.4 4 0.2 4 0 3 I ¢ I 5 f 3 I C T—gfig—tfi—‘fT—Q I 2 3 4 , 6 7 8 9 IO trme Estimated Per Capita Sales of Home Computers in Tunisia I 0.8 4 0.6 4 X 0.4 4 0.2 4 0 ¢ , e 1 ¢ I 3 ¢ I—M. I 3 , 6 7 8 9 IO trme Estimated Per Capita Sales of Mobile Phones in Tunisia I 0.8 4 0.6 4 x 0.4 4 0.2 4 o¢¢I¢I¢I¢Icfi¢I¢Ieje l 3 , 7 8 9 IO trme Estimated Per Capita Sales of Video Cameras in Tunisia I 0.8 4 0.6 4 0.4 4 0.2 0 c I ¢ 1 ; 1 ¢ I C 1 ; r 3 1 w I 2 3 4 6 7 8 9 IO 259 Estimated Per Capita Sales of CD Players in Turkey 0.8 4 0.6 4 0.4 0.2 O 3 I C : T!fi=tft VT fl T T T I T v I 3 4 5 , 6 7 8 9 IO trme Estimated Per Capita Sales ofHome Computers in Turkey I 0.8 4 0.6 4 X 0.4 4 0.2 4 O c t ‘f c f t 1 t , : ,#t=,—_—-_=fl__ I 3 4 _ 7 8 9 IO trme 4 Estimated Per Capita Sales of Mobile Phones in Turkey I 0.8 4 0.6 4 x 0.4 4 0.2 4 0 ¢ I 3 3 ¢ I A I 6 1T : I T I : I—T—J 4 , 6 7 9 IO trme Estimated Per Capita Sales of Video Cameras in Turkey l 0.8 4 0.6 4 x 0.4 4 0.2 4 0 6 I 3 ¢ 3 I a. 1 : 1W— 1 3 4 5 6 7 8 9 I0 260 Estimated Per Capita Sales of CD Players in UK 0.8 4 0.6 4 X 0.4 4 0 IO Estimated Per Capita Sales of Home Computers in UK I 2 3 4 , 7 8 9 IO trme 7 Estimated Per Capita Sales ofMobile Phones in UK I I 0.8 4 0.6 4 x 0.4 4 02 4 X/ 0 : T i I g T : r v I T I I I 2 3 4 5 , 6 7 8 9 IO trme Estimated Per Capita Sales of Video Cameras in UK I 0.8 4 0.6 4 x 0.4 4 0.2 4 0c ¢I>I:Ifl=I==fi—I°I°fi°fi' 2 3 4 5 , 6 7 8 9 IO trme 261 Estimated Per Capita Sales of CD Players in USA 0.8 4 0.6 4 X 0.4 4 0.2 4 IO Estimated Per Capita Sales ofHome Computers in USA 0.8 4 0.6 4 0.4 4 0.2 4 time Estimated Per Capita Sales of Mobile Phones in USA Estimated Per Capita Sales of Video Cameras in USA 262 APPENDIX D p and q Over Time in Developed Countries and Developing Countries for Each Product 263 EIIIIIJ : 813. «55:. 98:2... 533,—. 352528 ..I 81.? 55m oeoamwfim Emma“ cam—om Saws—am 8:82 Ewan—mi 38:26:— I- . 32: I Paws...— I wcocho: I EQMI 25 0:5 :85 II- «55%? I h——_ o— 2:: 2 I I f I T I I I I I I I mow—2:50 Esme—96: a. 25.5 3332 5.. .— 3o 3 2d 2 m2 2 mg to as 3 .— 25 ed 35 8 who 2 3o 3 3o 273 x 3 ucdtugam 533m _._QO Ema—om \EEoZ EESNBQZ $510502 SEE. bu: 38m— cam—o: 88.5 PaEBO I 8:8,.— 13:. 93.5... I fusion— II €980 €335 mE§< 2:: o_ a w R. o w v m N .— T 86 T 56 I who 1 wd . mad I ad v 36 fi £7223 I mutt—.50 coma—gon— E 0:25 23:: ..8 a 7 7 7 7 but: «725; 7 95:2: 7 :wEmH W 7 7 , 3.3258 .. . not“? :58 Bonamfim 3337 9878 533.3 8:82 £9832 38:85 3?: I Dawes: .. .. @338: I Saw I «.ch 220 fig I- 2:: «Eu—Sm; I 7 meta—.26 Mina—969 E 2.2—m 233% he a 275 7 ..8 a 276 n-, 7 meta..- 7 «Es; van—Hf. 52m... «0.3—£30m Ii- _ mom: :58 2239—5 333— van—om 53.9.33 cows—2 «mm-332 38:85 «:2: I human: I wco-cho: I 83m I «£6 250 _.nEm I-.-- Sficowfix I o— ] mod I _.o I 26 . Nd r mNd I md I mmd I To 1 mvd I md a r mmd 1 ed 1 8.0 r Nd I mud , ad 1 mad 1 ad 1 mad 35:ch Ema—05n— 5 £2.30 32> ..8 a 277 958 Z 7 238N262 , muss—8502 :83. 3.2 78.5.. “Ego: 8020- baFEuOI 85c... - - .- vEEE I fugue-I «350 629$ 353‘ I 272632 I 2:: numb—=50 toga—95¢ E .2256 82> ..8 a 278 A81: ..- aESF 0:5 28:9: , o. o w R. o w v m N _ 7 52a... 7 _ _ _ 7 7 7 _ . . 1 mod 7 «83358 I..- 7 I _.c 8:2 58m 7 - 26 Boguwfim H MNOO £33. - md 7 van—om 1 mm.0 7 3%; . H who 8:82 . . m6 3 33222 - mmd 33:85.-.-1. H mono MEI - 3 _ haw—SII W Whoa 7 wcovflcoII . mad “SmI7 - 27 7 22o 7 a mm o 7 2:6 7 __Nfim I»! 7 “5.8sz mete—=50 ”5:232..— E 98:50 32> ..8 a 279 APPENDIX E p and q Averages for Each Product in Developed Countries and Developing Countries 280 on: I DO III 3 25. - 3o - S T 25 - 3 T mg a - no - m2 - to I 36 59?: GD ..8 mow§o>< a md 281 o— 2:: I I I I I I I 59?..— 90 ..8 mousse-3‘ a 85 _.o m3 2. mg a no mmd 282 703+ 7 UQIII A: 2:: I mod - m3 - Nd .. m3 ._ - 8 - m2 - to 1 36 53:95.5 25:.— ..8 mom§o>< .— md 283 2 2:: OB LT OD III - no.0 fio T 26 I Nd - mg a I md T mmd T Yo I mvd 58:95.0 0:8: ..8 mow2o>< a md _7_________74 284 2:: DOA LT 09 III mod fio I 2.0 I S - mg a - no - m8 - to I 3.0 m.o «:25 23:2 ..8 mow-295‘ .— 285 o— 2:: Oman? 7 UQIII - 3o - _.o - m3 - S - mg a - no - moo - to I med 2.25 2332 .8.— mowaao>< 3 m6 286 2:: on; LT 0Q III - 3o - 3 - m3 - mo - m2... - mo - m8 . we I 3.0 Eon—a0 82> ..8 mango-3‘ .— m6 287 I'M—”I o— 2:: - 3o - 3 - m; - S - mg a - no - m2 - so - med md Ens—EU 82> ..2 mamas-3. a 288 APPENDIX F p and q Averages in Developed Countries and Developing Countries for Each Product 289 0E: 8280 82> le 28.5 2522 Lil 83980 080: III E“: 8 I 1 mod .. 2d 1 Nd mNd 82.5:ch 88.32— ..£ mow§o>< a [7 “77—— ___- ._ 290 22. S o m N. o w v 7 8280 82> le 0:95 2502 I4! ()/ .oSanU 28: III .883 no I - 3o . _.o - m3 - do - mg a - no - m8 - to - mvd 83:80 88.98: ..8 8883‘ a 29] 7 7 8280 82> lxl 0:25 0.502 Icl .3:qu0 35: Ill 3:: 8 I 7 r mod 83:80 ”28.025 ..2 8w8~>< 3 III mmd - “— ..__._ _— 292 o— 2.... 7- EQEmU ooE> IXI 2.25 25c: LT 83980 062.. III 833 no I yarw--- __ . _ - 3o - 3 - m3 - 3 - mg a - mo - m2 - to - mvd 822.80 ”Ego—gan— 2 82225. a md h‘q-vx \ __.- _,,___7_.__- 293 APPENDIX C Interaction Effect of Country Type and Product Type on p, q, a, b and c 294 «2:3...— Eosmo 82> 2.95 2302 .8:an0 080: .983 Do + E a ..8 2:32 T T 1 1 I i 56 Nod 86 vod med 532 cod nod mod cod fio 295 on: on J |-—-—‘ ‘30.--“ -_‘ H‘ «.5980 OOE> 0:23 2502 «0:69..— .2:an0 0803 .6wa GU a ..8 2.82 T I T I I 'r 36 No mNd Eco—Z m6 26 Yo 36 We 296 w 80:30 82> 2.2: 2.82 nun—5...— hxsnfiou 0:83 .5me Q0 a ..8 2:32 I I T T l T T md v.0 md 53—2 9o 5.0 ad ad 297 Una III on :6! 8250 82> 2.2... 2.82 .269..— .oSquU 2.83 a ..8 832 .933 DU 5 I I I I I I I I I fio Nd md v6 md :32 0.0 fie wd ad 298 835.5 82.80 82> 28.3 2522 889.50 2.82 .985 Do on: nil DD IOI o ..8 3.82 I I I I I I I fid Nd md vd m d :32 ed 5d ad ad 299 REFERENCES Anderson, E. B. (1970, “Asymptotic Properties of Conditional Maximum-Likelihood Estimators,” Journal of the Royal Statistical Society, 32 (2), 283-301. Ayal, Igal, and .Iehiel Zif (1979), “Market Expansion Strategies in Multinational Marketing,” Journal of Marketing, 43 (Spring), 84-94. Bass, Frank M. (1969), “A New Product Growth Model for Consumer Durables,” Management Science, 18 (January), 215-227. ---- (1980), “The Relationship Between Diffusion Rates, Experience Curves, and Demand Elasticities for Consumer Durables Technical Innovations,” Journal of Business, 53 (July), 51-67. ---- (1995), “Empirical Generalizations and Marketing Science: A Personal View,” Marketing Science, 14 (3), part 2, G6-G19. ----, Trichy V. Krishnan, and Dipak Jain (1994), “Why the Bass Model Fits Without Decision Variables,” Marketing Science, 13 (Summer), 204-23. ----, Dipak Jain, and Trichy V. Krishnan (2000), “Modeling the Marketing-Mix Influence in New-Product Diffusion,” “Estimation Techniques for Macro Diffusion Models,” in New-Product Diffusion Models, Vijay Mahajan, Eitan Muller, and Yoram Wind (eds), 263-94, Norwell, MA: Kluwer. Bretschneider, Stuart 1. And Vijay Mahajan (1980), “Adaptive Technological Substitution Models,” Technological Forecasting and Social Change, 18, 129-39. Brown, Lawrence A. (1981), Innovation Difl’usion: A New Perspective, New York: Method. Carbone, Robert and Richard L. Longini (1977), “A Feedback Model for Automated Real Estate Assessment,” Management Science, 24 (3), 241-48. Cavusgil, S. Tamer (1985), “Guidelines for Export Market Research,” Business Horizons, 28 (November-December), 27-33. Cooley, Thomas, and E. C. Prescott (1973), “Systematic (Non-Random) Variation Models and Varying Parameter Regression: A Theory and Some Applications,” Annals of Economic and Social Measurement, 2 (4), 463-73. ----, and ---- (1976), “Estimation in the Presence of Stochastic Parameter Variation,” Econometrica, 44 (1), 167-84. 300 Cox, DR. (1972), “Regression Models with Life Tables,” Journal of the Royal Statistical Society, 13 (34), 187-200. Craig, Samule C., Susan P. Douglas, and Andreas Grein (1992), “Patterns of Convergence and Divergence Among Industrialized Nations: 1960-1988,” Journal of International Business Studies, 23 (4), 773-787. Davidson, William H. and Richard Harrigan (1977), “Key Decisions in International Marketing: Introducing New Products Abroad,” Columbia Journal of World Business, Winter, 15-23. Day, George S., and Robin Wensley (1988), “Assessing Advantage: A framework for Diagnosing Competitive Superiority,” Journal of Marketing, 52 (April), 1-20. Dekimpe, Marnik G., Philip M. Parker, and Miklos Sarvary (1998), “Staged Estimation of International Diffusion Models: An Application to Global Cellular Telephone Adoption,” Technological Forecasting and Social Change, 57, 105-32. ----, ----, and ---- (2000), “Global Diffusion of Technological Innovations: A Coupled Hazard Apprach,” Journal of Marketing Research, 37, 47-59. DeSarbo, Wayne S.,M. Wedel, M. Vriens, and V. Ramaswamy (1992), “Latent Class Metric Conjoint Analysis,” Marketing Letters, 3, 273-88. Dockner, Englebert, and Steffen Jorgensen (1988), “Optimal Advertising Policies for Diffusion Models of New Product Innovations in Monopolistic Situations,” Management Science, 34 (March), 119-130. Douglas, Susan and Samuel C. Craig (1995), Global Marketing Strategy, New York: McGraw-Hill. Easingwood, Christopher (1987), “Early Product Life-Cycle Forms for Infrequently Purchased Major Products,” International Journal of Research in Marketing, 4, 3-9. ---- (1988), “Product Life-Cycle Patterns for New Industrial Products,” R&D Management, 18 (1), 23-32. Eliashberg, Jehoshua, and Rabikar Chattetjee (1986), “Stochastic Issues in Innovation Diffusion Models,” in Innovation Difiitsion Models for New Product Acceptance, Vij ay Mahajan and Yoram Wind (eds), 151-203, Cambridge, MA: Ballinger. Ganesh, Jainshankar (1998), “Converging Trends within the European Union: Insights from an Analysis of Diffusion Patterns,” Journal of International Marketing, 6 (4), 32- 48. 301 ----, and V. Kumar (1996), “Capturing the Cross-National Learning Effect: An Analysis of an Industrial Technology Diffusion,” Journal of the Academy of Marketing Science, 24 (Fall), 328-337. -------- , and Velavan Subramaniam (1997), “Learning Effect in Multinational Diffusion of Consumer Durables: An Exploratory Investigation,” Journal of the Academy of Marketing Science, 25 (Summer), 214-28. Gatignon, Hubert, Jehoshua Eliashberg, and Thomas S. Robertson (1989), “Modeling Multinational Diffusion Patterns: An Efficient Methodology,” Marketing Science, 8 (Summer), 231-247. Gouldner, Alvin W. (1957), “Cosmopolitans and Locals: Toward an Analysis of Latent Social Roles. 1”, Administrative Science Quarterly, 2, 281-306. Haagerstrand, T. (1953), Innovation of Spatial Process, Translated by A. Pred, Chicago: University of Chicago Press. Hall, Edward T. (1976), Beyond Culture, New York: Doubleday. ---- (1987), Hidden Differences, New York: Doubleday. Hanssens, Dominique, Leonard J. Parsons, Randall Schultz (1990), Market Response Models: Econometric and Time Series Analysis, Norwell, MA: Kluwer. Harvey, A. C., and G. D. A. Phillips (1982), “The Estimation of Regression Models with Time-Varying Parameters,” in Games, Economic Dynamics, Time Series Analysis, Diestler M., E. F urst, and G. Schwudiauer, eds., Wein-WurzburgzPysica-Verlag. Heeler, Roger M. and Thomas P. Hustad (1980), “Problems in Predicting New Product Growth for Consumer Durables,” Management Science, 10 (October), 1007-1020. Helsen, Kristian, Kamel Jedidi, and Wayne S. Desarbo (1993), “A New Approach to Country Segmentation Utilizing Multinational Diffusion Patterns,” Journal of Marketing, 57 (October), 60-71. Hofstede, Geert (1980), “Culture and Organizations,” International Studies of Management and Organizations, 10 (4), 15-41. Horsky, Dan (1990), “A Diffusion Model Incorporating Product Benefits, Price, Income and Information,” Marketing Science, 9 (Fall), 342-65. ----, and Leonard S. Simon(l983), “Advertising and the Diffusion of New Products,” Marketing Science, 2, (Winter), 1-17. 302 Islam, Towhidul and Nigel Meade (1997), “The Diffusion of Successive Generations of a Technology: A More General Model,” Technological Forecasting and Social Change, 56. 49-60. Jain, Dipak, Vijay Mahajan, and Eitan Muller (1991), “Innovation Diffusion in the Presence of Supply Restrictions,” Marketing Science, 10 (1), 83-90. ----, and Ram C. Rao (1990), “Effect of Price on the Demand on Durables,” Journal of Business and Economic Statistics, 8 (April), 163-70. Judge, G. G., W. E. Griffiths, R. C. Hill, and T. C. Lee (1985), The Theory and Practice of Econometrics, 2"d edition, New York: John Wiley. Kalish, Shlomo (1985), “A New Product Adoption Model with Price, Advertising, and Uncertainty,” Management Science, 31 (December), 1569-1585. ----, and Subrata Sen (1986), “Diffusion Models and the Marketing Mix for Single Products,” in Innovation Diffusion Models for New Product Acceptance, Vijay Mahajan and Yoram Wind (eds.), Cambridge, MA: Ballinger. ----, Vijay Mahajan, and Eitan Muller (1995), “Waterfall and Sprinkler New-Product Strategies in Competitive Global Markets,” International Journal of Research in Marketing, 12 (2), 105-19. Kalbfleisch, J. D. and R. L. Prentice (1980), The Statistical Analysis of Failure Time Data, New York: Wiley. Kalman, R. E. (1960), “A New Approach to Linear Filtering and Prediction Problems,” Journal of Basic Engineering, March, 35-45. ----, and R. S. Bucy (1961), “New Results in Linear Filtering and Prediction Theory,” Journal of Basic Engineering, March, 95-108. Kamakura, W. A. and S. K. Balasubramanian (1987), “Long-Term Forecasting with Innovation Diffusion Models: The Impact of Replacement Purchases,” Journal of Forecasting, 6 (1), 1-20. ----, and ---- (1988), “Long-Term View of the Diffusion of Durables: A Study of the Role of Price and Adoption Influence Processes via Tests of Nested Models,” International Journal of Research in Marketing, 5, 1-13. Kimberly, John R. and M. J. Evanisko (1981), “Organizational Innovation: The Influence of Individual, Organizational, and Contextual Factors on Hospital Adoption of Technological and Administrative Innovations,” Academy of Management Journal, 24, 689-712. 303 Krishnan, Trichy V., Frank M. Bass, and Dipak C. Jain (1999), “Optimal Pricing Strategy for New products,” Management Science, 45 (December). Kumar, V., Jainshankar Ganesh, and Raj Echambadi (1998), “Cross-National Diffusion Research: What Do We Know and How Certain Are We?,” Journal of Product Innovation Management, 15, 255-68. Leeflang, Peter C. H. , and W. Fred van Raaij (1995), “The Changing Consumer in the European Union: A Meta-Analysis,” International Journal of Research in Marketing, 12 (5), 373-387. Lenk, Peter J., and Ambar G. Rao (1990), “New Models from Old: Forecasting Product Adoption by Hierarchical Bayes Procedures,” Marketing Science, 9(1), 42-57. Levitt, Theodore (1983), “Globalization of Markets,” Harvard Business Review, 61 (May/June), 92-102. Lewis, Frank L. (1986), Optimal Estimation with an Introduction to Stochastic Control Theory, New York: John Wiley & Sons. Lilien, Gary L.,Arvind Rangaswamy, Christophe Van den Bulte (2000), “Diffusion Models: Managerial Applications and Software,” in New-Product Diffusion Models, Vijay Mahajan, Eitan Muller, and Yoram Wind (eds.), 295-311, Norwell, MA: Kluwer. Lindberg, Bertil C. (1982), “International Comparison of Growth in Demand for a New Durable Consumer Product,” Journal of Marketing Research, 19 (August), 364-71. Mahajan, Vijay, Chatlotte H. Mason, and V. Srinivasan (1986), “An Evaluation of Estimation Procedures for New-Product Diffitsion Models,” in Innovation Diflusion Models for New Product Acceptance, Vijay Mahajan and Yoram Wind (eds.), Cambridge, MA: Ballinger. ----, and ---- (1979), “Innovation Diffusion and New-Product Growth Models in Marketing,” Journal of Marketing, 43 (Fall), 55-68. ---- and Eitan Muller (1994), “Innovation Diffusion in a Borderless Global Market: Will the 1992 Unification of the European Community Accelerate Diffusion of New Ideas, Products, and Technologies?,” Technological Forecasting and Social Change, 45 (September), 221-235. ----, Eitan Muller, and Frank M. Bass (1990), “New Product Diffusion Models in Marketing: A Review and Directions for Research,” Journal of Marketing, 54 (January), 1-26. 304 ----, and Subash Sharma (1986), “Simple Algebraic Estimation Procedure for Innovation Diffusion Models of New Product Acceptance,” Technological Forecasting and Social Change, 30, 331-46. ----, and Yoram Wind (1986), Innovation Diffusion Models of New Product Acceptance, Cambridge, MA: Ballinger. McGee L.A., and S. F. Schmidt (1985), Discovery of the Kalman Filter as a Practical Tool for Aerospace and Industry, National Aeronautics and Space Administration, Technical Memorandum 86847, November. Norton, John A., and Frank M. Bass (1987), “A Diffusion Theory Model of Adoption and Substitution for Successive Generations of High-Technology Products,” Management Science, 33 (September), 1069-86. Ohmae, Kenichi (1985), The Triad Power, New York: Free Press. Peterson, Robert A., and Vijay Mahajan (1978), “Innovation Diffusion in a Dynamic Potential Adopter Population,” Management Science, 24 (November), 1589-1603. Porter, Michael E. (1985), Competitive Advantage: Creating and Sustaining Superior Performance, New York, NY: Free Press. Putsis, Jr., William P. (1996), “Temporal Aggregation in Diffusion Models of First-Time Purchase: Does Choice of Frequency Matter?,” Technological Forecasting and Social Change, 51, 265-79. ---- (1998), “Parameter Variation and New Product Diffusion,” Journal of Forecasting, 17 (3/4), 231-57. ----, Sridhar Balasubramanian, Edward H. Kaplan, and Subrata K. Sen (1997), “Mixing Behavior in Cross-Country Diffusion,” Marketing Science, 16 (4), 354-69. ----, and V. Srinivasan (2000), “Estimation Techniques for Macro Diffusion Models,” in New-Product Diffusion Models, Vijay Mahajan, Eitan Muller, and Yoram Wind (eds.), 263-94, Norwell, MA: Kluwer Robertson, Thomas S. (1971), Innovative Behavior and Communication, New York: Holt, Rinehart and Winston. Robinson, Bruce, and Chet Lakhani (1975), “Dynamic Price Models for New Product Planning,” Management Science, 21 (June), 1113-1122. Rogers, Everett (1995), Diflusion of Innovations, 4th edition, New York: Free Press. 305 Sarris, Alexander H. (1973), “A Bayesian Approach to Estimation of Time-Varying Regression Coefficients,” Annals of Economic and Social Measurement, 2 (4), 501-23. Schmittlein David C., and Vijay Mahajan (1982), “Maximum Likelihood Estimation for an Innovation Diffusion Model of New-Product Acceptance,” Marketing Science, 1 (1), 57-78. Simon, Herman, and Karl-Heinz Sebastian (1987), “Diffusion and Advertising: German Telephone Campaign,” Management Science, 33 (April), 451-466. Srinivasan, V., and Charlotte H. Mason (1986), “Nonlinear Least Squares Estimation of New-Product Diffusion Models,” Marketing Science, 5 (2), 169-78. Stengel, Robert F. (1986), Stochastic Optimal Control: Theory and Application, New York: John Wiley & Sons. Sultan, Fareena, John U. Farley, and Donald R. Lehmann (1990), “A Meta-Analysis of Applications of Diffusion Models,” Journal of Marketing Research, 27 (February), 70- 77. Takada, Hirokazu and Dipak Jain (1991), “Cross-National Analysis of Diffusion of Consumer Durable Goods in Pacific Rim Countries,” Journal of Marketing, 55 (April), 48-54. Talukdar, Debabrata, K. Sudhir, and Andrew Ainslie (2002), “Investigating New Product Diffusion Across Products and Countries,” Marketing Science, 21 (Winter), 97-114. Van den Bulte, Christophe, and Gary L. Lilien (1997), “Bias and Systematic Change in the Parameter Estimates of Macro-Level Diffusion Models,” Marketing Science, 16 (4), 338-54. Yip, George S. (1992), Total Global Strategy: Managing for Worldwide Competitive Advantage, New Jersey: Prentice Hall. Xie, J inhong, Michael Song, Marvin Sirbu, and Qiong Wang (1997), “Kalman Filter Estimation of New-Product Diffusion Models,” Journal of Marketing Research, 34 (August), 378-93. 306 “Twirliiiiiiitiii